1 Solidification Processing Merton C. Flemings Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA, U.S.A.
List of Symbols and Abbreviations 1.1 Solidification Mode 1.2 Plane Front Solidification and Crystal Growing 1.3 Heat Flow in Solidification of Castings and Ingots 1.4 Alloy Solidification - Traditional and Rapid Solidification Processes 1.5 Equiaxed Structures 1.6 Heat Flow and Mechanical Properties 1.7 Microsegregation in Novel Near-Rapid Solidification Processes 1.8 Alloy Solidification - Columnar Growth 1.9 Alloy Solidification - Heat Flow into the Bulk Liquid 1.10 Mixed Cases of Rapid Solidification 1.11 Macrosegregation 1.12 Deformation of Semi-Solid Dendritic Structures 1.13 Grain Refinement 1.14 Semi-Solid Slurries 1.15 Flow Characteristics of Semi-Solid Slurries 1.16 Semi-Solid Composite Slurries 1.17 Processing Non-Dendritic Semi-Solid Slurries 1.18 References
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
2 4 4 8 12 16 20 24 26 29 31 33 38 42 44 46 50 52 54
2
1 Solidification Processing
List of Symbols and Abbreviations A a0 C C* CL, C o cm9 cs d D, D s , DL
Km9 Ks L l0 mL n n P R, RT S T t, tf To TL, T{ TM V v, vx z
surface area length scale related to interatomic spacing constant composition of solid at the interface composition of a liquid, composition of the initial liquid specific heat of the mold, specific heat of the solidified metal dendrite arm spacing solute diffusion coefficient, solute diffusion coefficient of a solid, solute diffusion coefficient of a liquid fraction liquid, fraction solid volume fraction liquid gravitational acceleration temperature gradient in a solid, temperature gradient in a liquid heat of fusion casting surface heat transfer coefficient permeability consistency partition ratio, effective partition ratio, equilibrium partition ratio thermal conductivity, thermal conductivity of a solid, thermal conductivity of a liquid thermal conductivity of a mold, thermal conductivity of a solid length one-half of the dendrite arm spacing slope of equilibrium liquidus line power law index rate of nucleus formation in growth direction pressure growth rate, isotherm velocity thickness solidified temperature time, total solidification time of casting mold temperature equilibrium liquidus temperature, initial temperature melting point of the metal being cast casting volume velocity, velocity perpendicular to isotherms distance in growth direction
a, as P y y 5
thermal diffusivity, thermal diffusivity of solid metal solidification contraction constant shear rate boundary thickness
/ L , fs gL gr Gs, GL H h K K k, k\ k0 fc, /cs, kL
List of Symbols and Abbreviations
S g ji £s> QL> Qm T
distance between nucleation events cooling rate viscosity density of a solid, density of a liquid, density of a mold shear stress
KGT LKT M.I.T SIMA U.T.S.
Kurz-Giovanola-Trivedi model Lipton-Kurz-Trivedi model Massachusetts Institute of Technology strain-induced melt activation ultimate tensile strength
1 Solidification Processing
1.1 Solidification Mode One way to categorize solidification processes is by their "solidification mode." For the primary phase of usual, nonfaceting binary alloys which solidify over a temperature range, there are six of these (Fig. 1-1): columnar dendritic, cellular, equiaxed dendritic, equiaxed non-dendritic, single-phase plane front, and two-phase plane front. For alloys which freeze over a temperature range, and in which thermal gradient is sufficiently steep and convection low, the Equiaxed dendritic
Equiaxed non-dendritic
usual structure is columnar dendritic. For steeper thermal gradient and slower growth rate, the structure may become cellular. Lower thermal gradient, agitation, or grain refinement alters the structure from columnar dendritic to equiaxed dendritic. Vigorous agitation or more effective grain refinement results in equiaxed non-dendritic solidification. Sufficiently steep thermal gradient at slow growth rate results in plane front solidified alloys that are either singlephase or two-phase, depending on the number of phases present at the equilibrium solidus of the alloy. The three structures on the left of Fig. 1-1 are the most common in usual casting and crystal growing processes. The structures on the right belong to processes of emerging or potential commercial importance.
1.2 Plane Front Solidification and Crystal Growing Cellular
Plane Front (Single-Phase)
Plane Front (Two-Phase)
Figure 1-1. Types of solidification structure that can be obtained in directionally solidified binary alloys (of the non-faceting type).
Solidification with a plane front is obtained in one of three ways: (1) with an ideally pure metal, (2) with an alloy at sufficiently high temperature gradient, G, and low growth rate, R; or (3) at sufficiently high R. The first way is of little interest (except as a physical approximation for treating heat flow in solidification or casting of ingots which solidify in a narrow freezing range). (This topic is treated in Vol. 5, Chap. 10, Sees. 10.2.2 and 10.2.3.) The third is of emerging interest in "rapid solidification processing." The second is the basic technology of crystal growing. The basic heat flow objectives of all crystal growing techniques are to (1) maintain a positive thermal gradient across a liquidsolid interface, and (2) independently control this gradient so that the liquid-solid interface moves at a controlled rate. A heat balance at a planar liquid-solid interface in
1.2 Plane Front Solidification and Crystal Growing ooooooooo
Solid
I
Liquid
To inert gas source or vacuum
o o o o oQoooco o
Heating coils Crystal (a)
To inert gas source or vacuum Crystal withdrawal and rotation Crystal "Floating" liquid zone
To inert gas source or vacuum
Heating coils o
Heating coils
(b)
(c)
crystal growth from the melt is written (Flemings, 1974)
where ks is thermal conductivity of the solid metal, kL is thermal conductivity of the liquid metal, Gs is temperature gradient in the solid at the liquid-solid interface, GL is temperature gradient in the liquid at the liquid-solid interface, QS is density of the solid metal, and H is heat of fusion. Note from Eq. (1-1) that growth velocity R is dependent, not on absolute thermal gradient, but on the difference between ksGs and /cLGL. Hence, thermal gradients can be controlled independently of growth velocity. This is an important attribute of single-crystal-growing furnaces since growing good crystals of alloys requires that the temperature gradients be high and growth rate be low.
Figure 1-2. Examples of crystal-growing methods, (a) Boat method; (b) crystal pulling; (c) floating zone (Flemings, 1974, courtesy of McGraw-Hill Book Company).
The basic feature of all crystal growing furnaces is, therefore, the ability to obtain a controlled flux of heat across the liquidsolid interface. Fig. 1-2 shows schematically several types of furnaces to accomplish this (Flemings, 1974). Principles of crystal growing will be illustrated in the following sections with reference to the type of furnace illustrated in Fig. 1-2 a, in which metal in a crucible of some length, L, is fully melted and then solidified by "normal solidification" from one end to the other. In the limiting ideal case of "complete diffusion in the liquid" (or very vigorous convection), no solid diffusion, and equilibrium interface kinetics, solute redistribution in crystal growth occurs as shown schematically in Fig. 1-3. The liquid composition during solidification is given by the well-known non-equilibrium lever rule
1 Solidification Processing LIQUID
LIQUID
2 O
2 c0
Co
cs kC 0 0
L DISTANCE • (b)
DISTANCE—*(a)
Figure 1-3. Solute redistribution in solidification with no solid diffusion and complete diffusion in the liquid. (a) At start of solidification; (b) at temperature T*; (c) after solidification; (d) phase diagram (Flemings, 1974, courtesy of McGrawHill Book Company). kC( DISTANCE -
COMPOSITION • (d)
(c)
which, for constant partition ratio, fc, is written —c f^
(1-2)
where CL and fL are liquid composition and fraction liquid, respectively, and C o is initial liquid composition. This equation, also termed the Scheil equation, may be written in terms of the solid composition at the interface, Cs*, and fraction solid, / s : Cg* =
fcC0(l-/,)<*""
(1-3)
This equation also describes the final solid composition along the length of the crystal, since it is assumed that no diffusion occurs in the solid during or after solidification.
In real crystal growing processes, diffusion in the liquid is not complete, and there is some buildup of solute in a boundary layer in front of the liquid-solid interface. When some convection is present, the thickness of this boundary layer is taken to be some value 8, usually small compared with the size of the liquid pool. Now, even though equilibrium is still maintained at the interface, the higher solute content in the liquid at the interface leads to a higher solute content in the solid. For convenience, we now define an "effective partition ratio," k\ which is equal to the solid composition forming, Cs*, divided by the bulk liquid composition (the
1.2 Plane Front Solidification and Crystal Growing
liquid outside the boundary layer). The relation between k! and k is (Burton et al., 1953): k k
(1 4 )
where DL is the solute diffusion coefficient in the liquid. This expression is of considerable engineering usefulness because it relates the composition of the solid forming in crystal growth to alloy composition and growth conditions. It can be used to describe solute redistribution in crucibles of finite extent, provided only that the thickness 8 of the boundary layer is small compared with the length of the crucible. When this is true, a dynamic equilibrium is attained between the bulk melt and growing solid and equations identical to Eqs. (1-2) and (1-3) are readily derived, except that the equilibrium partition ratio k is replaced by the effective partition ratio k': s*
=
fc'C0(l-/s)<*'-1)
(1-5 a)
Here CL is the bulk liquid composition and k' = C*/CL. Eqs. (1-5a) and (l-5b) constitute a modified "normal segregation equation." Fig. 1-4 shows some calculated distributions of solute for the alloy of the preceding examples, taking k! equal to fc, unity, and an arbitrary value between the minimum (k) and the maximum (unity). As seen from Eq. (1-4), the minimum value occurs when Rc5/DL<^1, that is, at slow growth rate, high liquid diffusivity, and maximum stirring, and so 5 is a minimum. At this limit, solute distribution is described by the special case given earlier where infinite diffusivity in the liquid was assumed. The maximum value of k! (equal to unity) is obtained for RS/DL> 1. The major single problem in crystal growing of alloys is maintaining a flat,
o
oC s M o
k'= 1.0 c
o
DISTANCE, x
Figure 1-4. Final solute distributions for solidification with limited liquid diffusion and different amounts of convection (hence different "effective partition ratios," k!) (Flemings, 1974, courtesy of McGraw-Hill Book Company).
smooth interface which moves at constant velocity. Because of the buildup of solute in front of the crystal, the "melting point" (liquidus temperature of the alloy) increases with distance from the interface as shown schematically in Fig. 1-5. The actual temperature gradient must be maintained as steep or steeper than this to avoid supercooling ("constitutional supercooling") in front of the interface. The basic condition for this, the "constitutional supercooling criterion," may be written (Tiller et al., 1953): GL>_mLCs*(l-fc) kDL R
(1-6)
where mL is the slope of the equilibrium liquidus line in the phase diagram. As written, this equation is valid regardless of degree of stirring. For no stirring Cs* equals C o , and for very vigorous stirring it is the partition ratio times the bulk liquid composition. Eq. (1-6) illustrates directly
1 Solidification Processing
SOLUTE ENRICHED LAYER IN FRONT OF LIQUID-SOLID INTERFACE LIQUID
CO
o o DISTANCE, (b)
(a)
t
Ld
a:
UJ
D
or ZD
\<
CONSTITUTIONALLY
or u
c
SUPERCOOLED
CL
REGION
TEM
QL
DISTANCE, x1
Figure 1-5. Constitutional supercooling in alloy solidification, (a) Phase diagram; (b) solute-enriched layer in front of liquid-solid interface; (c) stable interface; (d) unstable interface (Flemings, 1974, courtesy of McGraw-Hill Book Company).
DISTANCE, x ' -
(c)
the reason for the requirement in crystal growing of controlling GL and R independently - it is the ratio of these two that determines whether a plane front is maintained. The constitutional supercooling criterion, Eq. (1-6), remains today an engineering tool of great value, although we understand that the conditions for breakdown of the plane front are more exactly specified by the stability analysis originally developed by Mullins and Sekerka, and that significant deviations may occur from this equation for faceting alloys or at very high rates of solidification (Mullins and Sekerka, 1963; 1964; Sekerka, 1965). (See also Vol. 15, Chap. 10, Sec. 10.2.4.)
1.3 Heat Flow in Solidification of Castings and Ingots Solidification processing of castings and ingots is basically different from that of
(d)
single crystals in that heat is rarely added to the liquid during solidification, and solidification rates are usually much higher. Fig. 1-6 shows the general heat flow problem for solidification of a pure metal in a mold, with temperature drops shown across the various resistances to heat flow. Fig. 1-7 is a similar plot, except for an alloy solidifying in columnar dendritic manner. (In this latter plot, mold-metal interface resistance is assumed to be negligible, and the mold is assumed to be very thick.) The following description of casting processes is given for the case of a pure metal being cast; differences for the treatment of alloys are discussed later. A number of important casting processes employ mold materials that are relatively insulating compared with the metal being cast. In "sand casting" the mold is made of sand, usually bonded with clay or with a resin. The pattern to form the mold cavity is usually reusable, made of wood,
1.3 Heat Flow in Solidification of Castings and Ingots
AIR
SOLID
LIQUID SOLID
^
LIQUID
*S
_
X
L
LIQUIDUS SOLIDUS
5 AT, LJ
AT,
iyAETAL-MOLD INTERFACE
MOLD-AIR INTERFACE
To DISTANCE, x
DISTANCE
Figure 1-6. Temperature profile in solidification of a pure metal (Flemings, 1974, courtesy of McGraw-Hill Book Company).
Figure 1-7. Unidirectional solidification of an alloy against a flat mold wall (Flemings, 1974, courtesy of McGraw-Hill Book Company).
COMPLETED CASTINGBROKEN OPEN TO REVEAL INTERIOR
Figure 1-8. Sketch of sandcasting processes as used in manufacture of a household radiator (Taylor et al., 1959, courtesy of J. Wiley and Sons, Inc.). MOLD SECTION B
metal, or plastic (Fig. 1-8). In "investment casting" the pattern is not reusable; it is made of wax or plastic. The pattern (or group of patterns) is surrounded or "invested" with a slurry of ceramic which is allowed to "set." In one method a relatively
thin mold is made by dipping the wax pattern alternatively into the ceramic slurry and then into a fluidized bed of dry ceramic. After the mold is made, it is fired to bond the ceramic and to melt and burn out the pattern (Taylor et al., 1959).
10
1 Solidification Processing
The heat flow problem for casting in insulating molds is simplified by the fact that heat flow is usually limited primarily by the heat diffusion in the mold, as illustrated schematically in Fig. 1-9. When this is the case, and when mold and metal variables are constant, a simple expression relates thickness solidified, S, to time t after casting: S=
TM-T0
t (1-7)
where TM is melting point of the metal being cast, To is mold temperature, and Km, gm, and cm are thermal conductivity, density, and specific heat, respectively, of the mold. This equation predicts a square-root relation of thickness solidified with time that is proportional to certain metal variables and to the square root of the "heat diffusivity,"Km£mcm. Heat flow into a concave mold surface is divergent and therefore somewhat faster than into a plane surface, and heat flow with a convex mold surface is slower. However, taking as a first approximation that a
;| SAND $
SOLID
LIQUID
given mold surface area has a fixed ability to absorb heat, it follows that in castings of simple shape the volume of metal solidified divided by the casting surface area may be substituted for S in Eq. (1-7). The result may be written as "Chvorinov's rule": (1-8) where V and A are casting volume and surface area, respectively, C is a constant, and t{ is total solidification time of the casting. In a large number of casting processes, heat flow is controlled to a significant extent by heat transfer resistance at the casting surface. These processes include permanent mold casting, die casting, liquid metal atomization, and thin strip casting. The die casting process is illustrated in Fig. 1-10, and several rapid solidification processes in which heat flow is largely "/z-controiled" are illustrated in Fig. 1-11. When the surface heat transfer coefficient is overriding, the temperature distribution in the metal and mold is as in Fig. 1-12, and thickness solidified in a given time t is: TM-T0 (1-9) S=h where h is casting surface (usually metalmold) heat transfer coefficient. For castings of simple shape, Eq. (1-9) can be generalized to yield the total solidification time, tf, of the casting:
i
QsH
LU
a.
h(TM-T0) A DISTANCE ,
x
Figure 1-9. Approximate temperature profile in solidification of a pure metal poured at its melting point against a flat, smooth mold wall (Flemings, 1974, courtesy of McGraw-Hill Book Company).
(1-10)
In the limiting case of ingot casting, when sufficient solid metal has formed, the mold-metal resistance to heat transfer becomes negligible, and the temperature profile becomes as in Fig. 1-13 a for a watercooled mold, and as in Fig. 1-13 b for
1.3 Heat Flow in Solidification of Castings and Ingots Cover die Ejector die Stationary platen \ Cores J Sliding platen Ejector pins \
1
\
YA PC A/A
11
-s~ + erf y =
ye'
I
Chamber
(TM-T0) ___ Ejector Rack and pinion Sleeve
Parting line Die Cavity Gate (a) Metal ladled into chamber
Ejector block
(b) Metal forced into die cavity
(c) Die opened, cores withdrawn
where ks, QS, and cs are thermal conductivity, density, and specific heat, respectively, of the solidified metal, and erf signifies the error function. Fig. 1-14 shows a plot of the solution to Eqs. (1-11) and (1-12) for pure iron cast in a cast iron mold. Also shown in this figure are two other curves showing how solidification is affected by varying degrees of mold-metal interface resistance. Analytic solutions such as those described above are available for a limited number of additional cases. Today, however, computer modeling is being used to an increasing extent to analyze solidification behavior of alloys, of complex shapes, and of simple shapes with complex boundary conditions (as continuous casting). The basic equation used in most analyses is, simply: dT
(d) Casting and excess metal ejected
Figure 1-10. Die casting, cold chamber machine (Taylor et al., 1959, courtesy of J. Wiley and Sons, Inc.).
an uncooled thick mold. Thickness of metal solidified is again proportional to the square root of time:
(1-n) where as is the thermal diffusivity of the solid metal and 7 is a constant given by:
(1-12)
HJK
= ocV2T
(1-13)
where T is temperature and a is thermal diffusivity. This equation is typically used in the semi-solid region as well as in the fully solid region and in the mold material. In the semi-solid region, the thermal diffusivity is adjusted to account for the heat of fusion released at a given temperature, and to refer to the thermal diffusivity of the liquid-solid mixture (i.e., for alloy solidification as shown schematically in Fig. 1-1). In continuous casting, the equation is sometimes applied at steady state, in fixed coordinates, for a casting withdrawal rate of R: = 0 = aV2T-R-^
(1-14)
12
1 Solidification Processing
Metal
Water
Non-rotating tungsten electrode Vacuum
Inert gas
- Collection port (b) Rotating electrode process
(a) V- or cone-jet atomization
Liquid metal
Molten droplets Gas T V - . . . \..l.-:r".[ / atomizer ^r' ' •"• •""vlV.V '
Gas fired furnace
Closed type atomizer ^oo\ed drum
Nitrogen
Tractor
Sray chamber Deposited strip
Flake particles
d? cP 0
**
\
\
\
\
\
\
\
\
\
\
\
\
\
s \
\
\
Exhaust system (c) Production of splat particulate by impact of atomized droplets on a rotating cooled drum
(d) Multilayer spray deposition on to a translating flat chill surface
Figure 1-11 a. Rapid Solidification - Spray Processes (Jones, 1982).
where z is distance in the growth direction. Construction of a mathematical model for alloy solidification requires a detailed model for solidification mode, and this is discussed in subsequent sections.
1.4 Alloy Solidification Traditional and Rapid Solidification Processes Solidification processes involving dendritic solidification are carried out commercially over a range of cooling rates of more than 12 orders of magnitude - from
10 4 Ks x for large ingots to nearly 109 K s " 1 for surface treatments. Isotherm velocities range from as little as 10" 5 ms" l to more than 10ms" 1 . Table 1-1 lists examples of the various solidification processes and their respective regimes of cooling rates. A region of growing industrial significance is that of "near-rapid" solidification, the range of about 1 to 103 K s " 1 cooling rate, where dendrite arm spacings in the range of 5 to 50 microns are obtained. Here, casting structure is usually fine-grained equiaxed, and segregate spacings are sufficiently fine that a high degree of homogeneity can be obtained through
13
1.4 Alloy Solidification -Traditional and Rapid Solidification Processes
—
Pressure
Vacuum outlet RF coil
Melt - Coolant
-Mold cavity
- Molten metal
Photocell
Pistons
Orifice
Liquid stream Heat removal
i
Injection tube
N
Quenchant Liquid-solid interface - Fibre (wire)
Aluminum foil
(a) Injection-chill mould
(b) Two-piston method
(c) Free-jet melt-spinning
Gas pressure
Ribbon %—
Tension
Heating source
Heating
Melt
Winding
-d Rotating drum
Fibre + sheath
(d) Taylor wire process
(e) Chill-block melt-spinning
Ftokes
(f) Twin-roll method
Melt stock
Product Molten metal — Fibre Extraction disc
(g) Crucible melt-extraction
(h) Pendant-drop melt-extraction
(i) Melt-drag process
Figure 1-11 b. Rapid Solidification - Chill Processes (Jones, 1982).
subsequent thermal processing (or thermomechanical processing in the case of ingots and continuous castings). Processes carried out in this region of cooling rate include strip casting, die casting, coarse powder atomization, "premium quality" cast-
ing, spray casting, semi-solid forming, and metal matrix composite casting. At more rapid rates, processes of fine powder atomization, thin strip casting, laser surface melting, and others are becoming of increasing importance. (Rapid solidification
1 Solidification Processing
14
TIME , MINUTES 16 36
SOLID
100
LIQUID
TM
(a) NO INTERFACE RESISTANCE (b)CARBON BLACK CHILL WASH
< o: LJ Q.
DISTANCE,
x
Figure 1-12. Temperature profile during solidification against a large flat mold wall with mold-metal interface resistance controlling (Flemings, 1974, courtesy of McGraw-Hill Book Company). 2
', MOLD / SOLID
LIQUID
'MOLD ' SOLID
LIQUID
4 6 SQUARE ROOT OF TIME, MINUTES172
Figure 1-14. Unidirectional solidification of pure iron against a cast-iron chill, (a) No interface resistance; (b) carbon-black chill wash [h assumed to be K *]; (c) thick insulating wash [h 0.04 cal cm' (Flemings, 1974, courtesy of McGraw-Hill Book Company).
''*-~VA
1
i
1 TM
_._
~'~'A / 1
LJ DC Z>
/ //
LJ •L
TEN
/ /
i 1 1
1i 1 1 i
/ To
DISTANCE
(a)
DISTANCE
(b)
Figure 1-13. Temperature profiles for solidification against a flat mold wall when (a) resistance of solidifying metal is controlling and when (b) combined resistance of metal and mold are controlling (Flemings, 1974, courtesy of McGraw-Hill Book Company).
techniques are discussed in detail in Vol.15, Chaps. 2 and 3.) In considering alloy solidification of castings and ingots, we need first of all to distinguish between columnar and equiaxed solidification, and secondly to distinguish between cases in which heat flow direction during solidification is into the bulk liquid, and those in which it is outwards through the growing semi-solid. Figs. 1-15 a and 1-15 b illustrate schematically the growth of equiaxed versus columnar grains, with heat flow through the growing dendrites. From the point of view of solute redistribution, the major dif-
1.4 Alloy Solidification -Traditional and Rapid Solidification Processes
15
Table 1-1. Range of cooling rates in solidification processes. Range of cooling rate
Production processes
Dendrite arm dendrite
Designation l ( T 4 t o 1(T2 1(T2 to 1 1 to 103
Slow Medium Near-rapid
103 to 109
Rapid
(urn) Large castings and ingots Small sand castings and ingots; billet and bar continuous castings Strip casting; die casting; coarse powder atomization; "Premium Quality" casting; spray casting; semi-solid forming; metal matrix composite casting Melt spinning; fine powder atomization; thin strip casting; electron beam or laser surface melting
50 to 5
5 to 0.05
Equiaxed dendritic
Columnar dendritic
Liquid
Liquid
(a)
(b)
ference between these two cases is that, for the case of equiaxed grains, the velocity of the isotherm where solidification begins, RT, can be much higher than the velocity of any given dendrite tip. It is: RT = Sn
5000 to 200 200 to 50
(1-15)
where 8 is linear distance in the growth direction between nucleation events, and n is rate of nucleus formation in the growth direction. Hence, even at high rates of movement of the solidification front, RT, dendrite tip temperature can be quite close to the liquidus temperature.
Figure 1-15. Solidification of an alloy against a cold chill wall, (a) Columnar solidification; (b) equiaxed solidification.
The same is not true in columnar or cellular growth. Here, isotherm velocity and dendrite tip velocity must be very nearly equal, and, at the more rapid rates of solidification, we expect increasing deviation of the dendrite tip temperature from the liquidus temperature. Solidification of columnar or equiaxed structures when heat flow is into the bulk melt present a difficult set of problems for the solidification scientist and engineer, and these are discussed in a later section.
16
1 Solidification Processing
1.5 Equiaxed Structures We first consider solidification where heterogeneous nucleation is sufficient so that undercooling before nucleation is low, and also that, regardless of the velocity of solidification isotherms, dendrite tips grow at temperatures near the equilibrium liquidus. Fig. 1-16 shows the model schematically. Liquid composition at any temperature within the liquid-solid zone, Fig. 1-16 c, is given by the equilibrium liquidus line, Fig. 1-16 b. Fraction liquid (or solid) at any point within the liquid-solid zone, Fig. l-16d, is calculated by the "Scheil equation" or one of its derivative analyses to be described below. In addition to the assumptions of interfacial equilibrium and negligible undercooling implicit in Fig. 1-16, the Scheil equation assumes no macroscopic mass transport and no effect Equiaxed dendritic
of ripening or other remelting process on growth. In differential form it is written (Scheil, 1942): 8CL
-
n
(1-16)
The solution to this equation has been given earlier as Eq. (1-2). This result can be rewritten in terms of solid composition as Eq. (1-3), which describes the microsegregation within fully solidified castings or ingots. It is written for a binary alloy with constant partition ratio, k. Calculated results for a simple binary alloy are shown schematically in Fig. 1-17. It has long been understood that diffusion in the solid during dendritic solidification alters microsegregation in most real systems. This was first quantitatively analyzed by Brody and Flemings (1966), who derived a simple approximate analytical
Solid
Liquid
Solid
(a)
Distance (c)
Distance (b)
Figure 1-16. Model of equiaxed solidification (a) dendrite structure; (b) temperature profile; (c) composition at the liquid versus distance; (d) fraction solid versus distance (Flemings, 1974, courtesy of McGrawHill Book Company).
1.5 Equiaxed Structures
17
kCQ
(b)
Figure 1-17. Formation of microsegregation. (a) Isoconcentrates as seen in a cross-section of a primary columnar dendrite arm after solidification, (b) Phase diagram showing "solidification path." (c) Microsegregation, measured along a radial path from the center of the dendrite.
kC
expression (modified Scheil analysis) to account for solid diffusion: k-1
(1-17) and a = •
4Dstf
(1-18)
where Ds is the diffusion coefficient of solute in the solid, tf is solidification time, and d is dendrite arm spacing. This analysis is
quite approximate, valid only for small values of a. Calculated results from the same work using a more exact numerical solution are shown in Fig. 1-18 for Al-4.5% Cu alloy. Since publication of the above work and its companion experimental study (Brody and Flemings, 1966; Bower et al., 1966), many refinements of the theory have been developed. Ohnaka (1986) derived a simple expression for an improved estimate of liquid composition during solidification.
18
1 Solidification Processing
Room temperature Just below eutecticJust above eutectic1.38 7.1.3,
"0.61 0.2 0M 0.6 0.8 Fractional distance along dendrite
1.0
Figure 1-18. Calculated composition across a dendrite arm at different times during solidification, showing effect of diffusion in the solid. Each curve drawn extends from the center of the dendrite arm to the position of the liquid-solid interface at the given time (Brody and Flemings, 1966).
2.5 a = 0.2713 k = 0.77
Eq.(1-23) ^ Scheil
/
/ /
(3=2 Ea( * 2.0 c o Matsumiya-
rody %&•
'
Equilibrium
1.5
0.2
0M 0.6 Fraction solid
0.8
1.0
Figure 1-19. Comparison of calculated microsegregation for manganese in an Fe 1.52% Mn alloy (Ohnaka, 1986).
Clyne and Kurz (1981) presented an analytical expression that is also approximate but has the correct limit at high values of a. Kobayashi (1988) has recently developed an improved analytic solution, although it is more difficult to use. Yeum, Laxmanan,
and Poirier presented a computationally efficient finite element method to calculate the microsegregation (Yeum et al., 1989). Kirkwood (1984) and Ogilvy and Kirkwood (1987) first incorporated the phenomenon of dendrite arm coarsening into the microsegregation model, and Mortensen later derived an analytical expression for this effect (Mortensen, 1989). Allen and Hunt (1976, 1979) first showed that dendrite arms tend to "climb" up the temperature gradient as a result of concentrationdriven solute diffusion. Others, most recently Riedl and Fischmeister (1990), have amply demonstrated the effect of dendrite arm climb and its influence on microsegregation. No attempt has yet been made, however, to incorporate this effect into the more general microsegregation models described above. In equiaxed structures with heterogeneous nucleation at low undercoolings, the main factor causing microsegregation to deviate from that predicted by the Scheil equation appears to be diffusion in the solid during solidification. Fig. 1-19 compares calculated results on segregation of manganese in an Fe-1.52 wt.% Mn alloy, obtained using several of the different analyses referred to above. The analyses differ only in the analytic or computational method used to account for solid diffusion (including assumed dendrite geometry). The extent of diffusion in the solid occurring during solidification depends on solidification time and inversely on the square of the diffusion distance (and hence, interdendrite arm spacing). Since dendrite arm spacing, as discussed below, depends on the cube root of solidification time, the result is a weak dependence of final microsegregation on solidification time (i.e., for a given alloy or cooling rate). Fig. 1-20 is an example from work of Michael and Bever (1954) in which severity of microsegrega-
1.5 Equiaxed Structures
tion is measured by amount of non-equilibrium eutectic in aluminum-copper alloys. For the alloys studied, the amount of interdendritic eutectic increased by a factor of two or more over the range of cooling rates from 10" 2 K s " 1 to 102 Ks" 1 . At the highest cooling rate employed, about 102 Ks" 1 , the amount of non-equilibrium eutectic measured is about that predicted by the Scheil equation. More recently, these and similar experimental results have been interpreted quantitatively on the ba-
o
D
8
C 0)
c D
sis of the solidification model discussed above, considering the effect of solidification and coarsening (Brody and Flemings, 1966; Bower et al., 1966; Basaran, 1981; Roosz et al., 1984; Sarreal and Abbaschian, 1986; Haider et al., 1987; Jones, 1984). While the degree of final segregation is not much influenced by cooling rate in these equiaxed structures, even at the maximum cooling rates studied, the segregate spacing (i.e., the dendrite arm spacing) is very much affected. Fig. 1-21 is a summary of data from a wide range of investigations for the binary alloy Al-4.5% Cu (Jones, 1984). The line through the points is:
4.79% Cu
o 10
.
^
^
-is—
3.97% Cu —
a — DK)8
Cu ^
u>
2
—-—-r-
U-iyj
where d is dendrite arm spacing (in microns) and s is cooling rate (Ks" 1 ). The factor determining dendrite arm spacing is now well understood to be ripening, which occurs during solidification (Mortensen, 1989; Kattamis et al., 1967). It is seen that this relation, and hence the importance of ripening, extends to well above the range of cooling rates of near-rapid solidification.
2.92%-
dJ , Q. U
19
2.05%. Cu
n
0.01 0.05 0.1 0.5 1.0 5 10 50 100 Solidification index-degrees centigrade per second
Figure 1-20. Weight percent eutectic vs. "solidification index" (cooling rate) for aluminum-copper alloys (Michael and Bever, 1954).
Figure 1-21. Dendrite arm spacing versus cooling rate for alloys of approximately Al-4.5 wt.% Cu. Experimental data are from a number of different investigations (Jones, 1984). 10b
10
z K/s
1011
20
1 Solidification Processing
Dendrite arm spacing is of engineering importance primarily because it determines the distance over which solute elements must diffuse to obtain fully homogeneous material. For dendrite arm spacings above those obtained in near-rapid solidification (above about 50 jum), times required for full homogenization in most commercial alloys become prohibitively long. Moreover, experience indicates that even extensive hot working of coarse dendritic structures does not fully eliminate the compositional heterogeneities. For example, Fig. 1-22 shows solution treatment calculations for Al-4.5% Cu alloy (Flemings, 1974). Solution temperature, time, and segregate spacing (dendrite arm spacing) are the important variables determining the effectiveness of a solution treatment. A moderate size sand casting, with dendrite arm spacing of 200 jim and given a standard commercial solution treatment (10 h at 515 °C), will not have the volume fraction of eutectic present reduced very much by the heat treatment. According to Fig. 1-22, approximately 40 h would be required to eliminate the second phase. This agrees with observations in practice which show that considerable second phase
is left in sand castings after usual solution treatments. On the other hand, modern "Premium-Quality" aluminum foundries apply substantial chilling to their castings to ensure that dendrite arm spacing stays below about 50 jim. Furthermore, they employ high-purity materials and carefully controlled equipment so that they can solutionize within 10 to 20 °C of the melting point (eutectic temperature) of the alloy being solution treated. A 2.5-h solution treatment at these temperatures is now more than ample to dissolve all the second phase. It should be added that, in the solidification processes that occur at rates up to those of near-rapid rate, solidification behavior (including microsegregation and secondary dendrite arm spacing) is not influenced by grain size or by whether the grain structure is equiaxed or columnar.
1.6 Heat Flow and Mechanical Properties Modern mathematical models of heat flow in solidification of alloys are generally based on physical models similar to that of
1.0
0.8
Figure 1-22. Calculated rate of solution of Al-4.5% Cu alloy. The symbols g and g0 are final and initial volume fractions of the second phase, respectively, /0 is one-half of the dendrite arm spacing, and t is time (Flemings, 1974, courtesy of McGraw-Hill Book Company).
0.6
^-SOLVUS (515°C)
0.4
0.2 n
~\ \ v
A\
\EUTECTIC\ \(548°C) 0.5
1.0 t/,2 x 10~9 sec/cm2
1.5
2.0
1.6 Heat Flow and Mechanical Properties
/TimT
21
Al-4.5% Cu alloy 1
- 8x10 5
8 x 10"
LU
& = 0.6
\
O O
k
- \ o
Cooling rate
x
L.
- 2x10
2x10' 0.2
-
a
3
0.A
0.6
0.8
5
1.0
2.0x10"'
a. 1.5x10"*
E o o
o 1.0x10"'
"5 0.5 xUT 4
0.5
0.2
0.4
0.6
0.8
o
Figure 1-23. Solidification of Al-4.5% Cu alloy against a water-cooled chill. Biot Number is 0.6. Dimensionless distance is measured from surface relative to half-thickness of metal (Flemings and Shiohara, 1985).
1.0
Dimensionless distance, X/L
Fig. 1-16 (Clyne, 1984; Rappaz, 1989). Fig. 1-23 shows some calculated results from work of Campagna (Campagna, 1970; Flemings and Shiohara, 1985) in which his physical model was exactly that described by Fig. 1-16 and for which the model alloy chosen was Al-4.5% Cu. The results plotted are for solidification against a watercooled chill, with a Biot number of 0.6 (where the Biot Number is defined as h L/k9 where h is the heat transfer coefficient at the mold-metal interface, L is the halfthickness of the metal, and k is the thermal
conductivity of the metal). Movement of the dendrite tips and the liquidus isotherm versus the square root of time are shown at the upper left. Average cooling rate during solidification is shown in the center plot, and dendrite tip velocity at the bottom. We may use this and Fig. 1-21 (or Eq. (1-19)) to illustrate some engineering aspects of several important commercial processes. Consider first a 100 mm thick plate of Al-4.5% Cu alloy, chilled on both sides. Mold-metal interface resistance is typical of that encountered in casting processes
22
1 Solidification Processing
(about 0.04 cal cm " 2 s " x K x) so that the Biot Number is about 0.6. From Fig. 1-23, cooling rate varies from about 3.0 K s " 1 at the surface to l.OKs" 1 at the centerline. Corresponding dendrite arm spacings vary from about 35 jim at the surface to 50 }im at the center - a suitable range for good mechanical properties, as seen by the experimental data in Fig. 1-24 for the alloy after solution treatment, quench, and aging treatment. Fig. 1-25 plots this expected variation in dendrite arm spacing and mechanical properties across the plate thickness. A casting of the same thickness cast in a sand mold would solidify, of course, at a much slower rate. An aluminum alloy casting this size would solidify throughout at a cooling rate of about 0.02 Ks" 1 , or 50 times more slowly than the center of the above chilled plate. The dendrite arm spacing throughout this sand casting would be about 200 jim. Full homogenization would be impossible (Flemings, 1974), and, as can be inferred from Fig. 1-24, very poor properties would result. The foregoing relationships are used by foundrymen in producing "Premium Quality" castings (Flemings, 1974), in which high mechanical properties such as those of Fig. 1-24 are reliably produced throughout castings. In sand castings, this is done by placing "metal chills" in the mold in appropriate locations and at appropriately spaced intervals so that the sand casting solidifies at rates comparable to those of the chilled casting described above - and final dendrite arm spacings are under about 50 |im. Dendrite arm spacing measurements from actual castings are sometimes requested by customers as a measure of process control. We may also consider solidification behavior of a die casting. Die castings typically solidify with much higher mold-metal
• - unchilled plates o - chilled plates 30
25
ELONGATION
510
o
30 40 50 60 70 DENDRITE ARM SPACING (urn)
Figure 1-24. Mechanical properties of Al-4.5% Cu alloy versus dendrite arm spacing. Alloy was solution treated, quenched, and aged (Flemings and Shiohara, 1985).
heat transfer coefficients than do chilled sand castings, and we estimate that a die casting of about 3 mm thickness would solidify with a Biot Number of about 0.6, so that we may use Fig. 1-23 to estimate its cooling rate during solidification. Cooling rate at the surface would be over 3 x 10 3 Ks~ 1 and cooling rate at the center about a third of that. Dendrite arm spacing would be about a tenth of that of the chilled sand casting described above, about 3 |im at the surface and 5 jim at the center. Such high cooling rates would normally be expected to result in excellent properties, but, for a number of reasons, high mechanical properties are not obtained in most of today's die castings. Reasons for this include entrapment of air and other gases during mold filling, and difficulties in
23
1.6 Heat Flow and Mechanical Properties 50
T
45
40 •
35
30
Al-4.5% Cu alloy \ 10 cm thick plate \ Biot Number = 0.6 •
^ - 3 - 2 - 1 0 1 2 3 4
Position Relative to Centerline (cm)
bU •
50" Ssv \U.T.S.
40"
Yield
30"
I
20-
10"
""^^^ Elongation \ ^ _
-
5
-
4
-
3
-
2
-
1
0
1
2
3
4
5
Position Relative to Centerline (cm) Figure 1-25. Calculated dendrite arm spacings and mechanical properties across the thickness of a 100 mm chilled plate, Al-4.5% Cu alloy.
fully feeding shrinkage. In addition, alloys used for most die castings are not amenable to development of high properties. The alloys used are chosen for low cost and ease of casting. It is clear, however, that the very high cooling rates obtainable in die casting could result in castings of high mechanical properties and reliability. Perhaps new processes such as semi-solid forming ("Rheocasting") will permit die casting to reach its potential for high-performance products. As an example of solidification at the very high rates characteristic of rapid rate processes, we may consider the example of a "splat cooled" 100 jim thick sample of Al-4.5% Cu solidifying from both sides. The high heat transfer coefficient here will again result in a Biot Number of 0.6, so we can again use Fig. 1-23 to calculate cooling rate. Now cooling rate is about 3 x 106 Ks"* at the surface and a third of that at the center. Dendrite arm spacing, from Fig. 1-21, varies from 0.35 ^im at the surface to 0.5 jim at the center. It should be added that, in solidification processes above that of the "near rapid" range, it becomes increasingly unlikely that nucleation frequency will be sufficiently high so that dendrite tip growth rate can remain a low value. In such cases, the Scheil equation becomes invalid and we
Table 1-2. Cooling rates and dendrite arm spacings for some practical examples. Al-4.5% Cu Alloy. Equiaxed solidification. Example
Cooling rate (Ks- 1 ) 100 mm thick sand casting 100 mm thick chilled casting 3 mm thick die casting 100 urn thick splat-cooled droplet
Center
Surface
0.02 3.2 3.6 xlO 3 3.2 x 106
DAS
(urn) 184 35 3.3 0.35
Cooling rate (Ks- 1 ) 0.02 1.2 1.3 xlO 3 1.2 xlO 6
DAS (um) 184 47 5 0.5
24
1 Solidification Processing
must turn to the analysis of a later section of this paper to describe the dendrite growth process. Table 1-2 summarizes the cooling rate and dendrite arm spacing calculations outlined in the previous paragraphs.
1.7 Microsegregation in Novel Near-Rapid Solidification Processes It is of interest to consider several processes which are often practiced in the near-rapid range of cooling rates but in which cooling behavior is greatly different from that in the more conventional processing described above. One of these is "spray casting," shown schematically in Fig. 1-26, in which metal is spray atomized, partially solidified in flight, and then deposited in plate form or other shape. The first part of this solidification may involve some undercooling, and certainly involves rapid rate; the last part of solidification is slower, depending on heat extraction to the substrate or surroundings. Final segregate spacing is found to be somewhat less than MOLTEN METAL
SPRAY CHAMBER
would be expected from a direct linear correlation of dendrite arm spacing with the logarithm of solidification time (Mathur etal., 1989). In semi-solid forming (Rheocasting) (Flemings, 1974; Flemings and Young, 1978), Fig. 1-27, solid particles initially present in the metal may be relatively large (50 jim or more), but solidification in the last stages (i.e., in a metal mold), may be very rapid. Microsegregation, and homogenization of that microsegregation, are only now being studied in these alloys (Molenar and Kool, 1989). A third process in this category is casting of metal matrix composites. In one type of such casting, metal is infiltrated into a ceramic preform. If the preform is cooler than the liquidus temperature of the alloy, initial solidification of the advancing front is very rapid; subsequent solidification is then determined by heat extraction to the mold (Mortensen et al., 1989; Masur et al., 1989). The amount of dendrite coarsening that can occur in these composites is, however, limited, since (at least in the alloys studied to date) coarsening ceases when
TUNDISH / CRUCIBLE
Figure 1-26. Schematic of the Osprey™ process (Mathur et al., 1989).
1.7 Microsegregation in Novel Near-Rapid Solidification Processes
25
CONTINUOUS ALLOY FEED
INGOT SECTIONED INTO "CHARGES"
^ -
MOLD
[7771^—^ WJ RHEOCAST , INGOT
p ^ SLURRY IN SHOT CHAMBER
SLURRY INJECTED INTO DIE
Figure 1-27. Rheocast Process (Flemings and Young, 1978).
A
V
dendrite size becomes about that of the interstices of the preform. Final microsegregation in these cases can therefore be much less than in conventional solidification (Mortensen et al., 1988). Fig. 1-28 shows the coarsening process in a fibrous composite schematically. Fig. 1-29 illustrates that, for local solidification times above a critical value (which depends on size of the interstices in the composite) during directional solidification, the dendritic structure has disappeared as a result of solid state diffusion into arms which are limited in their ability to coarsen by the ceramic fibers. Note that the time for disappearance of the second phase in this case can be very short compared with the time required to eliminate it by conventional solution treatment. For example, the time \
I
w
p
8!
7
(b)
Figure 1-28. Schematic rendition of coarsening in (a) a usual casting or ingot; (b) the interstices between fibers in a metal matrix composite (Mortensen et al., 1988).
26
1 Solidification Processing
•
CT c o
LIQUIDUS ISOTHERM
Average for the unreinforced alloy \U\ tc
1
m
•
V)
B
mmm 1
E 102-
_
o
H
•
•
• - A
i
•
L i i
TJ c 0) Q
•
1
10°
tf
1
1
o Q.
iDUS ISO!"HI
3 E 10 -,
1
1o
~^ o Q. 0) ~
1
[ SOLID
3
J
(a)
n>
CHILL
(b) LASER
CAST
MELTING-, WELDING
o
] i | 10 2
10
Local solidification time Ms
Figure 1-29. Dendrite arm spacing versus solidification time in a directionally solidified, aligned fiber metal matrix composite. Dendrite structure (and microsegregation) are eliminated when dendrite arms coarsen to the maximum size permitted by the interstices between fibers (Mortensen et al., 1988).
in Fig. 1 -29 required to eliminate the interdendritic eutectic is less than 100 s whereas many hours of conventional solution treatment would be required for material with a comparable dendrite arm spacing.
1.8 Alloy Solidification Columnar Growth In "constrained" columnar growth (growth with heat flow through the semisolid region), dendrite tip temperature can fall significantly below the liquidus isotherm, as illustrated in the practical examples of Fig. 1-30. Calculation of the extent of tip undercooling has occupied the attention of solidification scientists for a number of years. It has long been understood that dendrite tip undercooling can arise from three sources: the effect of interface curvature on equilibrium melting point; diffusion of heat and/or solute from the growing tip; and interface kinetics. A recent model, that of Kurz, Giovanola, and Trivedi (the "KGT model") incorporates all three effects using
SPINNING
WHEEL
( c ) MELT
SPINNING
Figure 1-30. Examples of rapid solidification processing with columnar dendritic growth; heat flow through the growing solid.
920
Equilibrium Liquidus = 918 K
910 900
Al-4.5% Cu G=o
Equilibrium Solidus = 844 K 840
10 "4
io "^
io"^
io -1
io°
io x
Mr
R(m/s)
Figure 1-31. Dendrite tip temperature versus tip velocity. Columnar growth of Al-4.5% Cu calculated using the KGT model (Kurz et al., 1986) with k = k0.
the Ivantsov solution for the diffusion field, the shortest marginally unstable wavelength for determining the tip radius, and a growth rate dependent partition ratio, k (Kurzetal, 1986). Fig. 1-31 shows results of calculations of dendrite tip temperature versus growth velocity for Al-4.5% Cu alloy using the KGT model with k = k0. The calculations assume growth into an isothermal melt at the tip temperature, so are drawn for (G/R) = 0.
1.8 Alloy Solidification - Columnar Growth
27
However, the curve is essentially unchanged even if G is much increased. For example, solidification of Al-4.5% Cu with a Biot Number of 0.6 (the example of Fig. 1-23) occurs, regardless of solidification rate, with a value of (G/R) = 1.5 x 10 6 Ksm~ 2 . Fig. 1-31 is not visibly changed by using this value of G/R. The curve of Fig. 1-31 assumes equilibrium interface kinetics. Note that significant tip undercooling is present at tip velocities as low as 10~ 3 ms~ 1 . The undercooling increases with increasing tip velocity, reaching a maximum at about 6.3 ms" 1 , the "limit of absolute stability" first described by Mullins and Sekerka (Mullins and Sekerka, 1964). At this velocity, solidification is with a plane front of uniform solid composition C o . From a simple kinetic argument we expect that, at sufficiently high growth velocities, the partition ratio will differ significantly from the equilibrium ratio as a result of "solute trapping." A functional relationship initially proposed by Aziz (Aziz, 1982) and incorporated in the KGT model for dendritic growth is:
solute trapping can be important. For Al-4.5% Cu alloy, it is possible, as indicated above, that some significant trapping can occur at velocities as low as about lms"1. Fig. 1-32 a shows the variation produced in tip composition for the cases outlined above, for equilibrium k. Tip composition varies from k0 C o at the lowest velocities to C o when absolute stability is reached. Fig. l-32b shows the calculated final microsegregation, using the recent empirical formulation of Giovanola and Kurz (1990) to "patch" the KGT solution at the tip to the Schell relation behind the tip. Table 1-3 summarizes some calculations of tip undercooling for several examples of this type of solidification for Al-4.5% Cu. It should be understood, however, that these calculations are probably valid only up to a velocity of a few tenths of a meter per second. At higher velocities, we expect k no longer to equal k0 due to "solute trapping." The predicted deviation of k from k0 at the dendrite tip is apparently substantiated by experimental work on rapidly solidified Ag-15wt.% Cu alloy (Boettinger et al,
(1-20)
Table 1-3. Calculated tip undercoolings for some practical examples. Columnar growth of Al-4.5% Cu alloy.
k =
1+
D where k and k0 are the kinetic and the equilibrium partition ratios, respectively, D is the solute diffusion coefficient at the interface, and a0 is a length scale related to the interatomic spacing. Eq. (1-20) predicts a rising k from k0 to unity over a range of velocities. For the Al-4.5% Cu alloy used as example here, it is over the range of about 1 to 100 ms" 1 , assuming ao = 0.5 nm. For a sufficiently dilute alloy, we expect the limit of absolute stability to be at a lower velocity than that where solute trapping occurs, but for higher solute contents
Example*
100 mm thick chilled casting 3 mm thick die casting 100 jam thick splat-cooled droplet
Tip velocity (ms' 1 ) 2.5x10"
Cooling rate 1.5
Tip undercooling (K) 7
0.09
2xlO 3
25
2.5
1.5 xlO 6
67
* Calculation in each case is for a location halfway between the surface and center of the casting; columnar growth assumed with Biot Number equal to 0.6.
28
1 Solidification Processing i
i
i
i
i
i
0 4
3
J
Al-4.5% Cu G=o
/ /
ity Limit
5 C
/
1
2
J3
10 "6
10' 5
10 "4
10 "3
10 "2
10 *
10°
10x
R(m/s)
Figure 1-32. Microsegregation in Al-4.5% Cu, calculated using the KGT model (Kurz et al., 1986) with k = k0. (a) Dendrite tip composition as a function of tip velocity, (b) Solute redistribution as measured outward from the center of a dendrite arm, calculated using the empirical patching relation of Giovanola and Kurz (1990).
0.0
0.2
0.4
0.6
0.8
1.0
25
'I / / / /
Ag-15wt% Cu 20 -"
V = 12cm/s
/
/ /
/
-
15
GK^
% O
_
-•••,'•••;.
y
y
10 --
1 /
Range of measured values
V.QC
-
0 0.01
1 V[cm/s]
10
100
(a)
Figure 1-33. Microsegregation in rapidly solidified Ag-15 wt.% Cu. (a) Dendrite tip composition. Data from Boettinger et al. (1987). Solid curve, KGT model (Kurz et al., 1986). Dashed curve, calculation using a
0.2
0.4
0.6
0.8
1.0
(b)
similar model with k = k0. (b) Solute redistribution. Solid curve calculated from KGT model using experimental patching (Giovanola and Kurz, 1990). Dashed curve, Brody and Flemings (1966).
29
1.9 Alloy Solidification - Heat Flow into the Bulk Liquid
1987). The KGT analysis with k = k0 predicted a tip composition substantially less than that observed (Fig. 1-33 a), while assuming a velocity-dependent k resulted in close agreement of the KGT theory with experiment (Kurz and Giovanola, 1988). The resulting solute distribution across the dendrite arm is shown in Fig. 1-33 b.
DISTANCE (a) SOLIDIFYING
DROPLET
—*>
(b) TEMPERATURE ACROSS DROPLET DIAMETER
1.9 Alloy Solidification Heat Flow into the Bulk Liquid When heat flows from the bulk liquid directly to the surroundings before or during solidification, conditions prevail that after solidification behavior in an important way. Heat then flows from the growing dendrite tips into the liquid, and recalescence occurs (at least locally). We sometimes see a small amount of this recalescence at the beginning of solidification in conventional castings and ingots. We can obtain much larger undercoolings in small droplet or bulk specimens, or in continuous processes, by employing clean molten metal and solidifying it without contact with materials which catalyze crystallization. In Fig. 1-34, for example, a liquid droplet is originally undercooled to a temperature T{ below the equilibrium liquidus, TL. Nucleation occurs at the left of the droplet and dendrites grow from left to right. Temperature across the droplet at this time is as shown in Fig. l-34b with heat flowing predominantly into the liquid. Movement of the dendrite front across the specimen is quite rapid at high undercoolings, and so the rate of local recalescence can be very rapid indeed. The growth velocity of dendrites into an undercooled melt is treated in a manner similar to the constrained growth discussed in the previous section. Moreover,
TIME —+(c) AVERAGE DROPLET TEMPERATURE VERSUS TIME
Figure 1-34. Solidification of an undercooled alloy droplet. 7j is initial temperature.
numerical results of dendrite tip velocity versus undercooling are very close to those calculated earlier, as shown in Fig. 1-35 for Al-4.5% Cu alloy using the LKT model (Lipton et al., 1987). Superimposed on the model is the curve of Fig. 1-31 (KGT model) which is for growth into a non-undercooled melt (constrained growth). Many workers are now studying dendrite growth in undercooled alloys, and comparing experimental results with these and other analyses. Results of one such study by Wu et al. (1987) are summarized in Fig. 1-36. Solidification of this "unconstrained" type is obtained in continuous processes when heat flow from the liquid is rapid and nucleation is hindered. Fig. 1-37, for example, shows equipment for direct casting of steel wire developed at Michelin several decades ago (Massoubre and Pflieger, 1978). The molten steel is ejected from a small orifice at speeds up to about 15ms" 1 and solidifies before Rayleigh breakup oc-
30
1 Solidification Processing I i lilMj
: :
I
Al-4 5%Cu o
KGT
•
LKT
Ni-25% Sn LKT & BC model
/
1 • &
Experimental • High-speed cinematography.-?' o Thermal measurement • Thermal measurement ° o 10° = [9 mm dia. droplet]
o o
o
8 9
~ 10"
8
8 9 o
10" II
o
10-
8
s
io " 4 r
10"
o9
l
I I i iml
o c
10
l
i
i i mill
100 Bulk undercooling, K
i i
1000
o
AT (K) Figure 1-35. Comparison of predictions of growth velocity versus undercooling for the models of Kurz et al. (1986), KGT, and Upton et al. (1987), LKT, for constrained and free dendritic growth, respectively, assuming k = k0.
Figure 1-36. Dendrite tip velocity vs. undercooling in Ni-25 wt.% Sn alloy. Experimental results and comparison with predictions of three dendrite growth models (Wu et al., 1987).
Melt Gas inlet-
curs. The wire is only about 200 |im in diameter, so radial temperature differences, even at these high velocities, are small. Solidification, in the usual case, occurs by axial dendritic growth, Fig. 1-38, and so, at steady state, dendrite growth velocity is just equal to the ejection speed. Fig. 1-39 shows results of actual experimental temperature measurements in wire spinning for a range of ejection velocities. Dendrite tip undercooling is seen to increase with increasing tip velocity and to be significant at these tip velocities of some meters per second. The undercoolings, in fact, are quite close to those of the Ni-25% Sn alloy of Fig. 1-36.
Induction coil
Crucible — Pressurised vessel (Maxi pressure: 30bars) ~0rifica
Gas inlet —*•
>
Cooling vessel
.Take-up device
Figure 1-37. Apparatus for spinning steel wire from the melt (Massoubre and Pflieger, 1978).
1.10 Mixed Cases of Rapid Solidification
Dendritic growth is treated in detail in Vol. 5, Chap. 10, Sees. 10.4 and 10.5.4.
1.10 Mixed Cases of Rapid Solidification The classes of rapid solidification cited above are not necessarily mutually exclusive with respect to any given casting oper-
LIQUIDUS
AT
TEMPERATURE
\
I DISTANCE
Figure 1-38. Solidification of spun cast wire. At steady state, wire moves to right with velocity R and dendrites grow at constant velocity R into liquid metal undercooled an amount AT.
T°C
31
ation. For example, consider the wire spinning operation shown schematically in Fig. 1-38 at steady state, and suppose that at a time, t, a large number of effective heterogeneous nuclei are continuously introduced into the molten stream. The "upstream front" then moves quickly up to the liquidus and subsequent solidification is by heat flow through the liquid-solid zone, comparable to the schematic example of Fig. 1-30. In the unsteady state, there remain two additional solidification fronts which grow together as shown schematically in Fig. 1-40. This phenomenon is observed in melt spinning (Massoubre and Pflieger, 1978). Mixed types of solidification can also occur when a thermal boundary layer reaches the surface of a casting, as can occur in solidification of atomized powders or in melt spinning. Fig. 1-41 illustrates this for melt spinning. Growth is assumed to occur only by columnar growth of existing grains, with no nucleation at the growing interface. Hence, substantial undercooling must occur in the melt to propagate growth upstream, and growth is rapid. But if the strip thickness is small compared with the thermal boundary layer, then this type of solid-
C =0.4% Si = 3.5%
j
'LIQUIDUS 1500
\ 1400
AT= 190°
AT= 2301°
V= II.I m s - 1 ^ V=l3.8ms-' ' 0~* V = l 6 . 4 m s " ' :=r ^ ^ - ^ _ 4 - * - V = 19ms"
'
265<
1300
V=l2.4ms-1 ^ = 15.1 ms-'
Figure 1-39. Measured temperature distributions along spun wire during steady state solidification at different velocities (Massoubre and Pflieger, 1978).
V = 17.7 ms'
1200
I 100 300
LENGTH
FROM O R I F I C E mm
1 Solidification Processing
32
L I Q U I D ;". |,
•I*
m
LIQUID
mm
LIQUIDUS
-++
TEMPERATURE
DISTANCE
Figure 1-40. Schematic illustration of transient solidification resulting when nuclei are introduced to the melt stream in wire casting. Three solidification fronts result.
ification can occur only in the bottom portion of the strip. In later stages of solidification, the undercooling in the bulk liquid will be depleted, tip temperature will rise, and tip velocity will slow to that which results from external heat extraction. The result is a strip of two distinct regions - an outer fine structure and an inner, much coarser structure, as observed by a number
of workers and described and explained by Chu etal. (Chu et al, 1988; Chu and Granger, 1990). Figs. 1-42 a and l-42b are from that work, and employ the theory of Lipton etal. (1987) to interpret the two distinct regions observed in Al-Fe alloy melt-spun strip of 100 jiim thickness. Fig. 1-42 a shows calculated growth velocity, assuming 75 K undercooling at the start of solidification, and Fig. l-42b shows calculated dendrite (or cell) tip radius. Note that this radius increases by nearly an order of magnitude over the first 20 |im of growth (as undercooling is dissipated) and stays relatively constant as further growth is limited by heat transfer to the surroundings. The points in Fig. l-42b indicate experimental estimates of tip radius and agree quite well with calculations. It should be added that in this case the very narrow freezing range of the Al-Fe alloy, combined with a Biot number that is not too low, apparently permitted this fine structure to be retained. In other instances, if ripening takes place more uniformly throughout a specimen, it may not be possible to distinguish as readily between the two regions of solidification.
LIQUIDUS
|
LIQUIDUS TEMPERATURE
DISTANCE
Figure 1-41. Mixed solidification in melt spinning. Two distinct regions form in the microstructure during solidification. Top: Solidifying structure. Bottom: Temperature in the metal near mold surface, assuming /i-controlled heat transfer to the mold. Temperature distribution at the top surface would be similar in form, but displaced slightly to the right.
1.11 Macrosegregation
33
Al-2 wt % Fe AT = 75K = 105w/m2-K
uu
Figure 1-43. Sketch of an ingot solidifying in a metal mold (shaded) with a refractory "hot top." The center of the ingot (dotted) is fully liquid; the outer portion (white) is fully solid. A semi-solid region exists between the two.
80 AI-2wt%Fe
60
h = 105w/m2-K L = 100 pm A = observation
2 20
"
-
/
n
i
0.2
0.4
i
0.6 X/L (b)
i
0.8
1.0
Figure 1-42. Dendrite tip velocity and tip radius versus dimensionless distance from the wheel surface for melt-spun Al-2 wt.% Fe alloy. Curves are calculated from Lipton et al. (1987). Triangular points on Fig. l-42b are experimental (Chu et al., 1988).
1.11 Macrosegregation The cause of macrosegregation in castings and ingots is now understood to be physical movement of liquid or solid phases during solidification. Diffusional transport, in the times available, can be significant
only over very small distances. One way in which the physical displacement can occur is by floating or settling of precipitated phases early in solidification (e.g., kishing of graphite in cast iron, or settling of fine grains early in ingot solidification). Most macrosegregation, however, is caused by a different mechanism - the flow of liquid through the interdendritic channels in the liquid-solid zone. Fig. 1-43 shows schematically solidification of an ingot. The character of the liquid-solid zone is described quantitatively in three ways using appropriate thermal relations combined with the Scheil equation (Eq. (1-2)) or one of its various modifications described earlier (Flemings, 1974). The liquid-solid region is permeable to liquid flow, as illustrated by the schematic "volume element" shown in Fig. 1-44. The interdendritic liquid flows through the ele-
34
1 Solidification Processing
tion equation" (Flemings and Nereo, 1967; Flemings et al., 1968; Mehrabian et al., 1970 a) 0L
acL
(1-22)
where ft is solidification contraction and 8 is rate of temperature change. The physical significance of Eq. (1-22) can be seen by considering steady state solidification with planar isotherms moving with velocity R in the x direction. Eq. (1-22) then becomes:
Figure 1-44. Fluid flow through a solidifying "volume element" (Flemings and Nereo, 1967).
ment with a velocity v given by: K v= —
(1-21)
where K is permeability; /i, gL, and QL are viscosity, volume fraction, and density, respectively, of the interdendritic liquid; P is pressure; and gY is gravitational acceleration. The flow may be driven by gravity acting on a fluid of varying density within the liquid-solid zone, by solidification shrinkage, by convection in the bulk liquid, which can influence flow within the liquid-solid zone, especially near the dendrite tips, or by electromagnetic forces or centrifugal acceleration. Tt may also be influenced by solid movement, as in bulging in continuous casting. Because the liquid composition varies spatially within the liquid-solid zone, any interdendritic flow other than that parallel to an isotherm must alter the composition of the volume element. Conservation of solute requires modification of the Scheil equation to a new "local solute redistribu-
8CL where vx is isotherm velocity perpendicular to isotherms. Eq. (1-23) reduces to the Scheil equation when vx and /? both equal zero. It also reduces to the Scheil equation (written in terms of volume fraction) when
P
(1-24) vv = — l - p R This flow is just that required to feed solidification shrinkage, and in steady state solidification the result is no macrosegregation. Flow with a speed greater than (down the temperature gradient from) that of Eq. (1-24) results in negative segregation; a speed that is less, or in the opposite direction, results in positive macrosegregation. Fig. 1-45 shows results, compared with theory, for an early experiment conducted to verify Eq. (1-22). Solidification of Al-4.5 wt.% Cu was carried out horizontally with approximately planar isotherms. Gravity and solidification shrinkage cause the interdendritic flow shown in Fig. l-45b. The resulting calculated and experimental macrosegregation are shown in Fig. l-45c. A schematic illustration of application of these principles to continuous casting is shown in Fig. 1-46. No segregation results if interdendritic flow lines are all vertical.
35
1.11 Macrosegregation
Miyazawa and Schwerdtfeger (1981) were the first to demonstrate the foregoing by quantitative calculation of segregation during bulging in continuous casting. Fig. 1-47 a illustrates their model, and Fig. l-47b several calculated flow fields within the liquid-solid zone. Resulting carbon and manganese macrosegregation is shown in Fig. 1-48. Another limiting condition of Eq. (1-23) occurs when
(a) I.U
x = 10cm L=10cm Co= 4.35% copper tG/s=-3x10"*cm/s
A
_, 0.8
\ \\ \\ \\
en
f >1
i
°V
0.6 —
-
V
~o c o "o
0.4 _
\
o
\\ 0.2 0 -1.0
theoretical experimental
\\ \\
\\° \\ \\
\\ I
-0.5
0 AC S ,% copper (c)
0.5
1.0
(1-25)
that is, when flow is in the same direction as, and is greater than, isotherm velocity. In this case, as temperature continues to fall (and so local liquid composition increases), local melting occurs, rather than solidification. It is this local melting that results in the channel segregates known in their various manifestations as "freckles," "A" segregates, and "V" segregates. The basic relation of Eq. (1-25) was first understood qualitatively on the basis of laboratory experiments (McDonald and Hunt, 1970; Copley et al., 1970) and later interpreted analytically with the aid of Eq. (1-22) (Mehrabian et al, 1970 b).
Figure 1-45. Macrosegregation in a horizontal unidirectionally solidified ingot, (a) Sketch of mushy zone, (b) calculated flow lines in mushy zone, (c) macrosegregation (Flemings, 1974, courtesy of McGraw-Hill Book Company).
Negative segregation at mid-radius would result from flow lines such as those of Fig. l-46b, and positive segregation at the centerline would result from flow lines such as those of Fig. l-46c. It can easily be visualized that "bulging" will result in greater downward flow, and in particular, greater downward flow toward the centerline (as illustrated in Fig. l-46c), thus enhancing centerline segregation.
(a)
(b)
(c)
Figure 1-46. Interdendritic fluid flow in a continuous casting, (a) No segregation, (b) negative segregation at mid-radius, (c) positive segregation at centerline (Flemings, 1974, courtesy of McGraw-Hill Book Company).
1 Solidification Processing
36
solid shell
sotidus
line
mushy zone
In the early analytical work of Flemings and coworkers cited above, steady state solidification was assumed with the thermal field given a priori, either from an uncoupled calculation, or from measurements. In addition, Eqs. (1-21) and (1-22) were uncoupled for ease of solution, and the bulk liquid in front of the growing dendrites was assumed to be quiescent. A simple relation between permeability, dendrite arm spacing, and fraction solid was assumed, with permeability being isotropic. Refinements in these and other respects have been incorporated in subsequent papers on the subject (Miyazawa and Schwerdtfeger, 1981; Kou et al, 1978; Dilawari and Szekely, 1977; Asai and Muchi, 1978; Fujii et al, 1979; Mori et al, 1979; Ridder et al, 1981; Tacke et al, 1981; Ohnaka and Fukusoko, 1981; Petrakis
. •
U center
surface
(a)
center
center
'Ml ' ' M M
heterogeneous 1
0
2
I
1 CO
Distance from center x (cm) (b) Figure 1-47. Effect of bulging on interdendritic flow in vertically continuous cast carbon steel (Miyazawa and Schwerdtfeger, 1981). (a) Schematic model of half-thickness of continuous casting, (b) Left is flow assuming only bulging; right is flow including effects of both bulging and solidification shrinkage.
1.11 Macrosegregation Center
Center
0.72
0.72
Manganese
Carbon
5
0.68-
0.68-
max
(mm) 1 2 3
c 0.64-
0.60
0.56
0.64-
0.60—
37
Fig. 1-49 a shows flow vectors during horizontal solidification of an ingot of NH 4 Cl-70wt.% H 2 O. This material was chosen because of the many laboratory studies which have been conducted using it. Solutally driven upflows are seen within the mushy region; these penetrate the liquidus front in the upper one-third of the cavity. Note the formation of discrete flow "channels" within the mushy region outlined by the dashed lines. The higher upward velocities within these outlined zones result from the local remelting and therefore higher permeability in the channel region.
0.56 1 0 2 1 Distance from center x(cm)
Figure 1-48. Macrosegregation of carbon (left) and of manganese (right) for three different degrees of bulging in carbon steel (Miyazawa and Schwerdtfeger, 1981).
etal, 1981; Ramachandran et al., 1981; Chang and Brown, 1983; Yao etal., 1984; Nomura etal., 1981; Maples and Poirier, 1984; Ohnaka and Kobayashi, 1986; Yeum et al., 1988; Bennon and Incropera, 1987 a, 1987 b, 1987 c; Beckermann and Viskanta, 1988; Nandapurkar et al, 1989; Mori et al, 1989; Heinrich etal, 1989). Fig. 1-49 is an example of results obtainable from the current advanced methods of calculation. The figure is from work of Bennon and Incropera (1987 b), who solved the solute continuity equation (including both advection and diffusion mechanisms) simultaneously with equations for energy and momentum. In this way they were able to show by calculation an irregular dendrite front and to account for a mushy region substructure (e.g., channel formation).
— 0.55
mm
/s 0.47-0.56 0.56-0.64 0.64 - 0.72 0.72 - 0.80
(a)
(b)
Figure 1-49. (a) Flow vectors during solidification, and (b) macrosegregation after complete solidification in NH 4 Cl-70 wt.% H 2 O ingot (Bennon and Incropera, 1987 c).
38
1 Solidification Processing
Fig. 1-49 b shows the resulting macrosegregation schematically. At the given time during solidification, the entire remaining liquid is enriched above its initial 70 wt.% H 2 O. The liquid channels within the mushy zone are clearly visible, and zones of negative segregation, especially toward the lower left, are present to make the solute balance. These results appear to be the first quantitative prediction of the formation of channel segregates. At the time of the initial quantitative work on macrosegregation in castings and ingots (Flemings and Nereo, 1967; Flemings et al., 1968; Mehrabian et al., 1970 a), the data available for permeability in liquid-solid zones were limited to those of Piwonka and Flemings (1966). Since that time, many experimental studies have been conducted, as summarized and used by Poirier (1987) to develop correlations for macrosegregation calculations. Fig. 1-50 shows examples of results of permeability calculations for flow through columnar structures. The macrosegregation theory outlined above, including the improved analyses now available, serves as the foundation today for design of laboratory experiments on macrosegregation. It serves also as the starting point for control of macrosegregation in practice. A few examples are given below. The theory predicts that macrosegregation will occur at abrupt changes in crosssection in directional solidification. Fig. 1-51 shows experimental confirmation of this for carbon steel (Nomura et al., 1981) and for a Mar M200 alloy (Sellamuthu etal., 1986). In agreement with calculations, elimination of bulging, plus a slight amount of additional reduction ("soft reduction") can reduce and even eliminate centerline segregation in continuous casting. Fig. 1-52 is an example (Izutani et al., 1988).
\
y^~~~~ U
-II
\
.1 1
-
I/I
,/r^B
d| = 300/i.m
// I
-14 0.0
0.2
0.4
0.6
0.8
1.0
(a) 1 / / /"
-10 -
y
/
/
/
y
y y*
__
\ \
C—^
V -*
_
'/i
i/i
d, = IOO/i.m
ii i i -1
d 2 = 40^.m
i
I I i '; -16 0.0
' '
i
0.2
0.4
i
i
0.6
i
i
0.8
i
1.0
(b) Figure 1-50. Calculations for permeability through columnar dendrites versus volume fraction liquid for flow normal to primary dendrite arms, (a) Primary and secondary dendrite arm spacings are 300 and 125 microns, respectively, (b) Spacings are 100 and 40 microns, respectively. A, B, and C represent three different correlations which give comparable results in their range of validity. Extrapolation C is the most realistic since this has the correct physical result at high fraction liquid. dt and d2 are primary and secondary dendrite arm spacings, respectively (Poirier, 1987).
1.12 Deformation of Semi-Solid Dendritic Structures
39
Calcd. 0.2 6
o
u_
°\ \
o 024
/ "D /
O O
O
o
-- 0.9
•/•Si = 0.2 5
0.20 -15
1.0 u
7.O0.25
0.22
°
o - 0.8
i
t
-10
-5
0
5
10
15
Distance! mm) Bottom area -•—1-»- Neck area (a)
Mar M2OO + 2..O%Hf 13.1
-
13.0
1
12.9 12.8
/I
12.7 12.6 - 12.5
itio
c
1 \
12.4
o 12.3 Q.
/
E o 12.2 u 12.1
\
z
I
In addition to work done relating theoretical understanding of experiments on channel formation in transparent alloys, some work has been done also in metal systems (Mehrabian et al., 1970 b; Shaw et al., 1986). Fig. 1-53 shows graphically that the severity of channel segregates in a laboratory test increases with increasing change in liquid density during the solidification process (in agreement with theory). The figure also shows that the direction of the channel segregate depends on the sign of the density change (also in agreement with theory).
-
12.0 II.9 11.8
if
11.7
•
Change In 1 Croit-ffldlon •
0 1
1.12 Deformation of Semi-Solid Dendritic Structures
i
i
2
3
i
4
i
5
i
6
t
7
i
8
t
9
t
10
Distance , cm (b) Figure 1-51. Macrosegregation at abrupt section changes in directional solidification, (a) Carbon steel (Nomura et al, 1981); (b) Mar M200 alloy (Sellamuthuetal., 1986).
When a casting of equiaxed grains solidifies in "mushy" fashion, the top surface of the casting can be seen to settle more or less uniformly in the early stages of solidification. Early in solidification, the grains are free to move (to "settle"). At some critical local solid fraction, the grains form a network and this stage of "mass feeding" ceases. The fraction solid at which the den-
40
1 Solidification Processing Reduction zone
o o o
Xn100°/o
"5 1.1 Middle carbon steel C = 0.15-0.16%
O
High carbon steel C= 0 . 5 0 - 0 . 5 5 %
to
100
50
x = k. x 100 %
X%
Figure 1-52. Relation between macrosegregation in continuous cast steel and "soft reduction." (Reduction is that calculated to have occurred near the centerline within the range of fraction solid from 0.4 to 0.7 as shown on right) (Izutani et al., 1988).
I
I
I
I
I
40 3CS15 DCS16 (inverted channel) 20 D Cooling rate 0.015 Ks" 1 A Cooling rate 0.044Ks" 1
0 -
-
CS18A
a
CS14
CS136 OCS17 -20 _
-
"0 0.10.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Fraction solid, fs • CS9 _
-40 a CS10 a CS12
CS19 en -DU
Figure 1-54. Isothermal shear strength of semi-solid dendritic aluminum alloys. Shear stress plotted is that after a small amount of shear (6.25 mm shear at a strain rate of 0.049 mm s"1). Insert at upper right shows test specimen (Metz and Flemings, 1970).
• CS7 CS8 •
an
I
30
30
I 60
I
I
90 120 Area of channels, mrrr
i
150
170
Figure 1-53. Relation between liquid density change and area of channel segregate formation in some carbon steels, in a laboratory experiment (Shaw et al., 1986).
drites form a cohesive network, and at which the network begins to develop some strength, must depend on dendrite size and morphology, but a number of studies in different alloys show it to be in the range of about 0.1 to 0.2 and occasionally higher.
1.12 Deformation of Semi-Solid Dendritic Structures
41
Rotating outer cylinder
Figure 1-55. Isothermal shear strength of semi-solid dendritic S n - 1 5 wt.% Pb alloy. Shear stress is the maximum measured (at a shear rate of 0.16 s" 1 ). Insert at upper right shows test arrangement (Spencer et al., 1972).
100 -
o o 0
0.2
0.4 Fraction solid, fs
As one example, Metz and Flemings isothermally sheared small blocks of aluminum alloys and found negligible strength below about 0.2 fraction solid (Metz and Flemings, 1969, 1970). Above 0.2 fraction solid, shear strength increased with increasing fraction solid, Fig. 1-54. Shear strength was found also to increase somewhat with increasing strain rate and with increasing grain size. In well-grain-refined alloys, strength did not begin to develop until 0.4 fraction solid. Spencer et al. (1972) carried out similar tests on Sn-15% Pb alloy. His test apparatus consisted of two grooved counterrotating cylinders, as shown at the upper right of Fig. 1-55. Maximum stresses obtained in the tests were qualitatively similar to the results of Metz on aluminum
alloys. Measurable strength began to develop at a fraction solid of about 0.2. Typical isothermal stress-strain curves for Spencer's semi-solid alloy are shown in Fig. 1-56. At a given strain rate, stress increases with displacement to a maximum, after which it falls to a low value. Maximum strength increases with increasing fraction solid. Deformation at fractions solid up to about 0.9 is primarily by grain boundary sliding with some dendrite distortion. The increase of stress with initial strain is probably because strain increases the number of contacts (the "welds") between particles. At sufficiently high strain, continuous fissures open, and so stress falls to a low level. The fissures become filled with liquid except at very high fractions solid.
1 Solidification Processing
42
400 -
80
At sufficiently high solid fractions (above about / s = 0.9) liquid can no longer flow to compensate for thermal or other strains in the solidifying metal. Then, if stresses are large enough to overcome the strength of the partially solid material, internal or open "hot tears" result. Fig. 1-58 shows results from a typical type of "hot tear" test casting employed by foundrymen. When a casting is sufficiently long and the mold is sufficiently rigid, the casting tears apart during solidification. Hot tear susceptibility is strongly alloy-sensitive. Alloys most prone to hot tearing are those which solidify over a wide temperature range, but with a relatively small amount of residual liquid at the eutectic temperature (Rosenberg et al, 1960; Clyne and Davies, 1979). As shown in Fig. 1-58, greatest hot tear susceptibility in Al-Cu alloys is at about 5 wt.% Cu.
Angular displacement in degrees Figure 1-56. Isothermal shear test results, semi-solid Sn-15% Pb alloy (Spencer et al, 1972).
1.13 Grain Refinement
Strains such as the foregoing, with resulting liquid flow, result in localized regions of macrosegregation in actual castings and ingots. A common source of the strain is thermal contraction of the solidifying metal, hindered by mold constraint (Rosenberg et al, 1960; Clyne and Davies, 1979). Centerline segregation in continuous casting can be visualized as having a similar root cause, resulting from thermal contraction of the solid, accentuated by "bulging" of the casting due to metallostatic head (Miyazawa and Schwerdtfeger, 1981). Detailed analysis of compositional variations across a casting cross-section such as this requires analysis of the full thermal and interdendritic flow fields during solidification (Flemings, 1974; Flemings and Nereo, 1967). Fig. 1-57 is an example.
Columnar grains in castings and ingots may be many centimeters long. Their diameter may range from less than 0.1 mm to 1 cm or more. Generally, the diameter of columnar grains increases progressively as solidification proceeds, since those grains which have a preferred growth direction oriented near the heat flow direction tend to "crowd out" less favorably oriented grains. Equiaxed grains typically also range in size from less than 0.1 mm to over 1 cm. Effective grain refiners (heterogeneous nucleating agents) are available for most nonferrous alloys. Through the use of these it is possible in practice (especially with aluminum and magnesium alloys) to obtain grain sizes consistently under about 0.1 mm. Grain refiners have been developed for ferrous and nickel-base alloys which are effec-
1.13 Grain Refinement
tive as mold coatings for thin-section castings, but their effective life is not sufficiently long for refinement of sections more than a few millimeters thick (Flemings, 1974). To achieve smaller grain sizes than can be achieved with available grain refiners, and especially to refine the grain size of metals for which grain refiners are not available, we need to turn to other methods. Vibration (ultrasonic and sonic) has been shown to influence formation of new grains, and there has been much speculation as to the mechanism. An interpretation favored by many some decades ago was that the vibration promoted heterogeneous or even homogeneous nucleation (Richards and Rostoker, 1956). Others (Southin, 1966) were inclined to the earlier explanation of Schmid and Roll (1939) that the grain refinement derived from fragmentation of primary crystallites, providing an artificial source of more nuclei. This second one is the prevalent view today. Foundrymen and ingot casters have long understood that equiaxed grains are
43
favored over columnar grains by low pouring temperature (low "superheat"). As pouring temperature is lowered further, the equiaxed grains become finer. We now understand that the major cause of the foregoing is convection at low pouring temperature: the convection associated with the mold filling itself remains strong as solidification commences. With high pouring temperatures, on the other hand, the forced convection dissipates before solidification begins, so that solidification proceeds in a relatively quiescent melt. Columnar grains (or large equiaxed grains) are favored at a given pouring temperature when convection is minimized by application of a magnetic field (Uhlmann et al., 1966), or by casting in a thin section within which convection is quickly dissipated (Bower and Flemings, 1967). More importantly from a practical standpoint, introducing convection by mechanical or electromagnetic means during the early stages of solidification favors formation of fine, equiaxed grains. There are many references in the older cast metals center center
mushy ! zone '
I » 1 » 1
• 1
t
I
1
<
I
•
»
t
~*
'TT'I \ \ :X \ \;^^^
•
*
/
/
'
'
1
/ / 1
*
i
>
i
Qi
1 11 1 1 l
heterogeneous'
surface
o
\ x-x
\ 2
<
JOI
1 \ solid\ \
1 11 t 1 I on
solid shell
1
' 1 0
Distance from center in cm
0.56 Distance from center in cm
Figure 1-57. Centerline segregation due to "bulging" in continuous casting, (a) Schematic, (b) interdendritic flow lines, (c) resulting macrosegregation (Miyazawa and Schwerdtfeger, 1981).
44
1 Solidification Processing 60
1 933 K
50
•
—
L
-»_
.
"
I
-—-». •
—
-
^ •
\
—
•
—
X
/
X
821 K
\
K
/
/
1
X
X
X
/
/
z O
-
O
/
0 30
0 \x
20
-
~
/
o
X XX
X
/
XX
X
/X
O
O
o
/i
vxxX X X X
\
10
/
1
r
XX
5
/
/
\
\ ^ X XX / O\\ / >
f
—
/
O
O
o x -- torn casting O -- no visible tears |
|
5
10
15
Wt. % copper
Figure 1-58. Hot tear testing of Al-Cu alloy. Plot shows maximum length test casting that can be made without visible tears (Rosenberg et al., 1960).
literature as to the effect of such convection on grain structure. Mold oscillation, for example, was sometimes employed to achieve fine grains in sand castings, and stirring with a cold rod was used to refine the structure of ingots. Today, electromagnetic stirring is widely practiced in continuous casting to achieve fine grain size. There is general agreement today that vibration, low pouring temperature, and externally induced convection all promote grain refinement primarily by a dendrite fragmentation mechanism, although there is not yet agreement on what this basic
mechanism is. (It must be added also that we cannot rule out the possibility that in some cases the convection may also enhance the effectiveness of heterogeneous nucleation agents.) Some possible dendrite fragmentation mechanisms are: (a) Dendrite arm fracture. Arms shear off as a result of the force on the arm from the fluid flow (Garabedian and StricklandConstable, 1972, 1974). (b) Remelting of the arm at its root as a result of normal coarsening (Kattamis et al., 1967). The function of the fluid flow in this case is simply to carry the dendrite arm away from its "mother grain" to where it can grow as a new grain. (c) Remelting as above, enhanced by thermal perturbations which occur with turbulent convection in a liquid which is not at uniform temperature. (d) Remelting as above, but where the melting at the root is accelerated by the stress introduced at the dendrite root as a result of the force of the fluid flow. (e) As in (c) above, but where the melting at the root is further enhanced by a high solute content in the solid at the dendrite root (Uhlmann et al., 1966). (f) Recrystallization as a result of the stress introduced by the force of the fluid flow, with rapid liquid penetration along the new grain boundaries (Vogel, 1978; Apaydinetal, 1980).
1.14 Semi-Solid Slurries During the course of his doctoral thesis research in early 1971, Spencer was conducting the hot tearing tests on Sn-15% Pb alloy discussed earlier (Figs. 1-55 and 1-56) (Spencer, 1971). In the course of those experiments, he decided to use the same apparatus to conduct a quite different type of test. Instead of partially solidifying the
1.14 Semi-Solid Slurries
45
Solid Liquid Rotating outer cylinder
0.8
0.6
0.4
0.2
fi-3'S 06 0.2
0.4
0.6
Figure 1-59. The Spencer experiment. Experimentally determined viscosity and shear stress versus fraction solid for Sn-15 wt.% Pb alloy cooled at 0.006 Ks" 1 with a shear rate of 200 s" 1 . Inset at upper right is a schematic illustration of the test specimen (Spencer et al., 1972).
0.8
Fraction solid, fs
alloy before beginning shear, he began the shear above the liquidus and then slowly cooled his alloy into the solidification range while it was being sheared. The results were surprising. When shear rate was relatively high, stress increased only very slowly as temperature was decreased below the liquidus. The shear stress measured at a given temperature below the liquidus was orders of magnitude less than when the samples were cooled to the given temperature before shear. Comparison of Figs. 1-55 and 1-59 provides an example of this remarkable reduction in shear stress. In both of these figures, frac-
tion solid, / s , is calculated from the Scheil equation based on actual temperature measurements. The grain structure obtained in these early experiments was nondendritic, as shown schematically at the upper right of Fig. 1-59, and it was evident that shearing in the material was taking place more or less uniformly throughout the sample. The material was behaving as a liquid-like slurry, to which an apparent viscosity could be assigned, as has been done in Fig. 1-59 (Spencer, 1971; Spencer et al., 1972). We came, in our early work at M.I.T., to call the process of obtaining these new structures "Rheocasting" to sig-
46
1 Solidification Processing
nify the distinctive rheological behavior of the material. Vigorous agitation as solidification begins results in formation of new grains by one or another of the mechanisms described in the previous sections, presumably by dendrite fragmentation. The early growth of each dendrite fragment then apparently continues dendritically, as shown schematically in Fig. 1-60 a and b. With continuing shear and time during solidification, the dendrite morphology becomes that of a "rosette" (Fig. l-60c), as a result of ripening, shear, and abrasion with other grains. Ripening proceeds during further cooling, Fig. l-60d. With sufficiently slow cooling and high shear, the particles become spheroidal (or in some cases ellipsoidal), usually with a small amount of entrapped liquid, as shown in Fig. l-60e. The (a)
Structure evolution in rheocasting
(c)
Increasing shear rate Increasing time Decreasing cooling rate
Id)
(e)
Figure 1-60. Schematic illustration of evolution of structure during solidification with vigorous agitation, (a) Initial dendritic fragment, (b) dendritic growth, (c) rosette, (d) ripened rosette, (e) spheroid.
extent to which the morphology evolves along the spectrum from that of Fig. 1-60 a to that of Fig. 1-60 e increases with increasing shear rate and amount of solidification, and with decreasing cooling rate. The size of the individual grains (dendrites or rosettes) appears to depend only moderately on shear rate above some given minimum value, but depends strongly on cooling rate - at least cooling rate during the initial stages of solidification.
1.15 Flow Characteristics of Semi-Solid Slurries When viscosity of such a slurry is measured during continuous cooling, it is found to be a strong function of shear rate, decreasing with increasing shear rate, Fig. 1-61. One effect of the shear is to produce denser, more rounded particles which move more easily past one another; i.e., to accelerate the irreversible structural evolution illustrated in Fig. 1-60. The major effect of cooling rate on viscosity can be understood in a similar way. Viscosity increases with increasing cooling rate, as illustrated in Fig. 1-62, primarily because the higher cooling rate results in less dense, less spheroidal particles (Flemings, 1991). Shear rate affects structure in an important way in addition to that sketched in Fig. 1-60. A larger scale "structure" can build in these slurries by collision and coalescence of favorably oriented particles, as sketched in Fig. 1-63. The extent of this larger structure will depend on a balance between the rate of structure buildup and its breakdown from shear (Rames et al., 1989). We expect this structure buildup and breakdown to be more or less reversible with changing shear rate. In rheological terms, these semi-solid slurries are said to exhibit "pseudoplastic-
1.15 Flow Characteristics of Semi-Solid Slurries
(a)
47
(b)
Figure 1-61. Effect of shear rate on apparent viscosity of semi-solid alloys, (a) Sn-15wt.% Pb (Joly and Mehrabian, 1976); (b) Al-4.5% Cu-1.5 wt.% Mg (Kattamis and Piccone, 1990).
ity." Empirical equations developed to describe the full range of viscosity of pseudoplastic materials require at least four parameters. However, a simple and widely used relation that is often useful over wide ranges of shear rate is the well-known "power-law" model: fi = Kyn~i
(1-26)
where \i is viscosity, y is shear rate, n is called the "power law index," and K the "consistency." The smaller the power law index, the greater the pseudoplasticity. We may use this simple model to help understand the effects of process variables on viscosity of semi-solid alloys.
Fig. 1-64 plots results from recent thesis work at M.I.T. by Moon (1990) on Al6.5 wt.% Si alloy. Viscosity is shown at 0.4 fraction solid as a function of shear rate for a number of different initial conditions. Note that viscosity at a given shear rate can vary by over an order of magnitude. Data for the top curve were obtained by cooling to fs = 0.4 and recording the viscosity obtained immediately on reaching this fraction solid; cooling time from the liquidus to this fraction solid was approximately 5 min. The middle curve shows measurements obtained after continuing the shear at / s = 0.4 for an extended period (90 min). For longer holding times, change
48
1 Solidification Processing 10 7
I
I
I
i
I
i
Al-4.5%Cu-1.5%Mg Sn-15%Pb
y0=330s"1
_
6 -
8 0.33 Ks"1
5 -
continuously cooled
-
OJ Q_
•E
c
6 -
>^ 4 ~ <7> o i
i-t \A
c
3 -
c=0.42Ks"1-^ AH
-
i
cm
a <
0.03 Ks"' i
£- 0.017 Ks"1
2 -
2 / c= 0.0055
1 -
/ VKs"1
w I
0.2
0.4
0.6
c= 0.008 Ks"1
I
i
0.8
Fraction solid, f%
(a)
0.2
0.4
0.6
i
0.8
Fraction solid, fs
(b)
Figure 1-62. Effect of cooling rate on apparent viscosity of semi-solid alloys, (a) Sn-15wt.% Pb (Joly and Mehrabian, 1976); (b) Al-4.5 wt.% Cu-1.5% Mg (Kattamis and Piccone, 1990).
Figure 1-63. "Structure" buildup from collision and coalescence. Schematic (a) rosettes, (b) spheroids.
in viscosity with time was very slow, and so we expect this as the "steady state" curve. The large drop in viscosity resulting as the "steady state" is approached during isothermal shearing is due to the irreversible morphological evolution occurring with time, as sketched in Fig. 1-60. Note the difference between the apparent shear thinning behavior in these two cases. The power-law exponent is substantially higher for the steady state case (indicating less shear thinning). Approximately reversible pseudoplastic behavior appears to be typical for "steady state" curve experiments such as the two in Fig. 1-64. That is, "structure" of the type
49
1.15 Flow Characteristics of Semi-Solid Slurries
10 A Continuously cooled, 0.075 Ks"1 A Continuously cooled, 0.0083Ks"1 •
Steady state
O Viscosity immediately after shear rate abruptly changed from 900 s"1 to value shown
0.1
Fraction solid = 0.4
0.01
100
200
500
1000
steady state value. The difference between this curve of "instantaneous viscosity" and the steady state curve is a measure of the thixotropy of the slurry. This structural buildup is further enhanced by solidifying and then partially remelting the semi-solid material in the process we have come to call "Thixocasting." Measurements on viscosities of "thixocast" Sn-15% Pb alloy were made by Laxmanan (1980) using a modified parallel plate plastometer (essentially a compression test) at low shear rates. Viscosities in the range of 107 Pa • s were obtained for material at about 50% solid, as shown in Fig. 1-65. These viscosities are several orders of magnitude greater than those obtained for the same alloy with the same initial spheroidal structure after vigorous agitation, and enable small samples to be lifted and handled like solids (Kenney etal., 1988).
Shear rate in s 1
Figure 1-64. Viscosity versus shear rate for Al6.5 wt.% Si alloys (Moon, 1990).
illustrated in Fig. 1-63 builds and is destroyed depending on shear rate. There is, however, a time dependency (thixotropy), so that when shear rate is abruptly changed, the new steady state viscosity is attained only after some time at that shear rate. This effect is illustrated by the bottom curve of Fig. 1-64, in which instantaneous viscosity is given after dropping the shear rate from 900 s~* to the shear rate plotted. This instantaneous viscosity is less at any given shear rate than the "steady state" value, because the "structure" has not had time to adjust to that of the new shear rate. With time, agglomerates build and the viscosity at a given shear rate approaches the
"-5
- 4 - 3 - 2 - 1
0
1 1
f, shear rate in s"
Figure 1-65. Viscosity versus shear rate for Sn15 wt.% Pb alloys. Solid lines at the upper left are for "Thixocast" materials (Laxmanan and Flemings, 1980).
50
1 Solidification Processing
Results qualitatively similar to those described above for Sn-15% Pb alloy and two aluminum alloys have been described for a large number of other aluminum and low melting point alloys. Similar results have also been obtained for a wide range of other metal alloys including copper-base alloys, cast iron, steels, and superalloys. During an extended program at M.I.T. in the 1970's, alloys studied included hypoeutectic cast irons, several copper-base alloys, a cobalt-base superalloy, a nickelbase superalloy, several stainless steels, and a low alloy steel. References to these works are given in a recent summary paper (Flemings, 1991). We may gain more familiarity with the properties of semi-solid metal alloys by comparing them with some well-known materials. Vigorously agitated metals at 40 to 50% solid have viscosities typically in the range of 0.1 to 10 Pa • s. This is two to four orders of magnitude higher than the viscosities of water or fully liquid metal, and in the range of glycerol, liquid honey, and machine oil (Table 1-4). Yogurt exhibits the same range of viscosities; from about 1 to 10 Pa • s, depending on shear rate. At higher viscosities, molten polymers and molten silicate glass are in the range of 103 to 104 Pa • s. At this viscosity, the materials still flow readily under gravity or moderate pressure. A material behaves as a "solid" in our vocabulary when it undergoes negligible deformation during a time that is of interest to us. We can begin to handle these semi-solid alloys as if they were solids when their viscosity rises much above 106 Pa • s, as is seen by the following order of magnitude calculation. A 50 mm cube of semi-solid aluminum is imagined to be held between two parallel plates. Gravity produces a shear force. Suppose now that we arbitrarily wish no more displacement
Table 1-4. Viscosities of semi-solid alloys and of some familiar materials. Approx. viscosity Pas Silicate glass in its temperature range for forming Semi-solid alloys, 50% solid, shear rate 1 0 " 3 s - 1 Molten silicate glass Molten polymers Liquid honey Yogurt, shear rate 10 s" 1 Semi-solid alloys, 50% solid, shear rate 200 s" 1 Glycerol Yogurt, shear rate 200 s ~1 Water, liquid metals
10 4 -10 8 107 103 104 103 101 101 lO0-^1 10° lO" 1
io~ 3 -io- 2
than 1 mm in a ten second period. Required viscosity can be calculated from the simple equation defining viscosity: x = iiy
(1-27)
where T is shear stress. The result is a viscosity of approximately 106 Pa • s. Semisolid thixocast alloys at 40 to 50% or more solid easily reach this viscosity, permitting them to be handled as if they were solids.
1.16 Semi-Solid Composite Slurries It was recognized early in the Rheocasting research at M.I.T. that the high and controllable viscosity of semi-solid slurries made them excellent starting materials for processing metal matrix composites. It was found that a variety of ceramic particulates could be added to the semi-solid slurry and kept in suspension without floating or settling (Mehrabian et al., 1974; Flemings et al., 1976 b). Other researchers have shown that particles can be similarly added to fully liquid metal (Rohatgi et al, 1979; 1986), but
1.16 Semi-Solid Composite Slurries
51
10 Matrix: Al-6.5wt.%Si Steady state at 973K
30 vol.% SiC
20voL%SiC
0.1 continuously cooled SiC/Al-6.5wt.%Si Cooling rate: 0.075 Ks'1 Shear rate: 180 s'1
10vol.%SiC
0.01
100
500
200
1000
1
Shear rate in s"
Total volume fraction of (primary solid + SiC)
(a)
(b)
Figure 1-66. (a) Viscosity versus shear rate for Al-6.5 wt.% Si alloy with SiC. (b) Viscosity versus total volume fraction solid during cooling through the liquid-solid zone of Al-6.5 wt.% Si alloy. Contains 0, 0.1, and 0.2 volume fraction SiC (Moon, 1990).
that a semi-solid metal slurry provides better wetting and dispersion (McCoy et al., 1988). The ease with which particles are entrained in the melt is a sensitive function of surface chemistry (e.g., magnesium in an aluminum alloy greatly aids incorporation and retention of SiC particles; whereas chlorine degassing causes expulsion of the particles). Rheological properties of these slurry composites have been studied by a number of investigators (Girot, 1987; Loue and Kool, 1989; Mada and Ajersch, 1990; Moon, 1990). Most find that the composites (with or without some solidified metal) exhibit thixotropic behavior qualitatively similar to that of partially solidified fully metallic slurries. Fig. 1-66 is an example from recent thesis work of Moon (1990). Fig. 1-66 a shows viscosity versus
shear rate for Al-6.5 wt.% Si with 0.1, 0.2, and 0.3 volume fraction SiC. Fig. l-66b shows viscosity during cooling in the liquid-solid range with 0, 0.1, and 0.2 volume fraction SiC. It is of interest that the viscosity for a given total volume fraction of solids is less when a portion of that volume fraction is ceramic particles. This is possibly because the ceramic particles prevent the metal particles from joining and coalescing as readily as in the fully metal slurry. This conclusion is supported by the observation that both the pseudoplasticity and thixotropy also appear to be reduced in aluminum slurries by the addition of the particulate SiC. It may also be that the presence of the ceramic accelerates the morphological evolution of the solidified metal particles toward more perfect spheroids.
52
1 Solidification Processing
1.17 Processing Non-Dendritic Semi-Solid Slurries The main processes used to date for achieving non-dendritic structures are illustrated schematically in Fig. 1-67. Fig. 1-67 a is a simple "batch rheocaster," in which a crucible of molten liquid is mechanically mixed while being cooled. This type of stirrer was used in early M.I.T. work on casting of metal and composite components and in exploratory studies by others. It has been incorporated in vacuum or inert atmosphere chambers for use with high melting point metals or to reduce air entrapment. A second type of process which has been employed is the "continuous rheocaster" of Fig. 1-67 b, in which (1) higher shear is readily achievable, (2) the stirring is well below the surface of the metal, thus minimizing air entrapment, and (3) cooling rate can be high so as to achieve a fine structure. This type of rheocaster has been used in the course of much of the practical process development at M.I.T. for low melting point metals, copper-base alloys, and steels (Flemings and Mehrabian, 1973; Flemings
(a)
(b)
et al., 1976 a; Riek et al., 1975; Young et al., 1976; Flemings and Young, 1978). The third type of process, Fig. l-67c, is vigorous electromagnetic stirring of continuous castings (to produce billets for subsequent thixocasting) (Kenney et al., 1988; Young et al., 1984). This is a process of considerable technological importance today since it permits production of large tonnages using a variant of a well-established technology, and it is applicable to high temperature metals such as steel. There are other routes as well to production of non-dendritic material. In one, the shear is obtained by rapid electromagnetic pulse discharge (Nakada et al., 1990). In another, the slurry structure is obtained by flow of cooling metal within a tortuous channel to achieve the shear needed for dendritic breakup (Brook, 1982; Antona and Moschini, 1986). Another relies solely on long thermal treatment of fine grain structures in the semi-solid temperature region; one notable starting material in this process is spray-deposited metal (Ogilvy, 1990). Nature is kind in a very few alloys, in which the grain refiners employed in usual casting processes are so powerful
(0
Figure 1-67. Schematic diagrams of methods of producing non-dendritic structures, (a) Batch, (b) continuous, (c) electromagnetic stirring with continuous casting.
1.17 Processing Non-Dendritic Semi-Solid Slurries
1
o 1 1 o o o nrTT-n o o o o o o o 0 o o o [I JU o
m
CZ2
oooooc
that the original grain structure is nondendritic. One such case is zirconiumrefined magnesium-zinc alloy (Flemings, 1974). A completely different approach has been termed SIMA (Strain-Induced Melt Activation). In this case, an alloy billet or bar (usually of relatively small cross-section) is cold worked a critical amount so that, on reheating to the liquid-solid zone, the desired spheroidal structure is obtained (Kenney et al., 1988; Young et al., 1983). Critical steps in this process appear to be obtaining a fine grain structure in the billet or bar and then achieving grain boundary melting on reheating. The step of solidifying the original slurry in either a continuous casting or shaped casting has been termed "Rheocasting." When a rheocast billet is reheated for subsequent shaping, the process is termed "Thixocasting" or, at higher fraction solid, "Thixoforging." Today, industrial interest focuses primarily on these "Thixoforming" process routes. Small billets cut from continuous castings with the rheocast structure are reheated and formed, usually by a process which resembles either die casting (Fig. 1-68 a) or closed die forging (Fig. l-68b). Other types of processes have also been shown to be suitable for shaping the semi-solid material, including extrusion and rolling. Present manufacturing processes, such as that shown schematically in Fig. 1-68, are now highly automated, with billets being progressively heated under computer control and then transported by robot arms to the forming operation for automatic shaping and subsequent removal from the die. Current applications for the process are primarily to produce parts of higher integrity than die castings and lower cost than other competing methods.
53
o o _
1 1 L_ FT
ur
Figure 1-68. Parts manufacture by semi-solid forming: (a) by equipment similar to die casting; (b) by equipment similar to closed die forging.
54
1 Solidification Processing
1.18 References Allen, D.J., Hunt, J.D. (1976), Metall. Trans. 7 A, 161-110. Allen, D.J., Hunt, J.D. (1979), Solidification and Casting of Metals. London: The Metals Society, Book 192, pp. 39-43. Antona, R. L., Moschini, R. (1986), Metall. Sci. Technol. 4(2), 49-59. Apaydin, N., Prabhakar, K.V., Doherty, R.D. (1980), Mater. Sci. Eng. 46, 145-150. Asai, A., Muchi, I. (1978), Trans. ISIJ 18, 90-98. Aziz, M.J, (1982), J. Appl. Phys. 53, 1158-1168. Basaran, M. (1981), Metall. Trans. 12A, 1235-1243. Beckermann, C , Viskanta, R. (1988), Int. J. Heat Mass Transfer 31, 35-46. Bennon, W.D., Incropera, F.P. (1987a), Int. J. Heat Mass Transfer 30, 2161-2170. Bennon, W.D., Incropera, F.P. (1987b), Int. J. Heat Mass Transfer 30, 2171-2187. Bennon, W.D., Incropera, F.P. (1987c), Metall Trans. 18B, 611-616. Boettlnger, W.J., Bendersky, L.A., Coriell, S.R., Schaefer, R.J., Biancaniello, F.S. (1987), J. Crys. Growth 80, 17-25. Boettinger, W.J., Coriell, S.R., Trivedi, R. (1988), Rapid Solidification Processing: Principles and Technologies IV: Mehrabian, R., Parrish, P. A. (Eds.). Baton Rouge, LA: Claitor's Publishing Division, pp. 13-25. Bower, T.F., Brody, H.D., Flemings, M.C. (1966), Trans. AIME 236, 624-634. Bower, T.F., Flemings, M.C. (1967), Trans. AIME 239, 216-219. Brody, H.D., Flemings, M.C. (1966), Trans. AIME 236, 615-624. Brook, G.B. (1982), Mater. Design 3, 558-565. Burton, J.A., Prim, R.C., Slichter, W.P. (1953), J. Chem. Phys. 21, 1987-1991. Campagna, A. J. (1970), Ph.D. Thesis, Department of Metallurgy, Massachusetts Institute of Technology. Chang, C.J., Brown, R.A. (1983), J. Cryst. Growth 63, 343-364. Chu, M. G., Granger, D. A., Ludwiczak, E. A. (1988), Solidification Processing 1987. London: Institute of Metals, pp. 271-274. Chu, M.G., Granger, D.A. (1990), Metall. Trans. 21 A, 205-212. Clyne, T.W. (1984), Mater. Sci. Eng. 65, 111-124. Clyne, T.W, Davies, G.J. (1979), Solidification and Casting of Metals. London: The Metals Society, pp. 275-278. Clyne, T.W, Kurz, W. (1981), Metall. Trans. 12A, 965-971. Copley, S.M., Giamei, A. R, Johnson, S.M., Hornbecker, I F . (1970), Metall. Trans. 1, 2193-2204. Dilawari, A. H., Szekely, J. (1977), Metall. Trans. 8B, 227-236.
Flemings, M.C. (1974), Solidification Processing. New York: McGraw-Hill. Flemings, M.C. (1991), Edward Campbell Memorial Lecture, ASM International Fall Meeting, Detroit, MI. Metall. Trans. B (in press). Flemings, M . C , Nereo, G.E. (1967), Trans. AIME 239, 1449-1461. Flemings, M . C , Mehrabian, R., Nereo, G.E. (1968), Trans. AIME 242, 41-49. Flemings, M . C , Mehrabian, R. (1973), Trans. AFS. 81, 81-88. Flemings, M . C , Riek, R.G., Young, K.P. (1976a), Mater. Sci. Eng. 25, 103-117. Flemings, M . C , Riek, R.G., Young, K.P. (1976b), AFS. Int. Cat. Met. J. 3, 11-22. Flemings, M . C , Young, K.P. (1978), Yearbook of Science and Technology. New York: McGraw-Hill, pp. 49-58. Flemings, M . C , Shiohara, Y. (1985), Tetsu-toHagane 71, A204-A208. Fujii, T, Poirier, D.R., Flemings, M.C. (1979), Metall. Trans. 10 B, 331-339. Garabedian, H., Strickland-Constable, R.F. (1972), /. Cryst. Growth 13/14, 506-509. Garabedian, H., Strickland-Constable, R.F. (1974), J. Cryst. Growth 22, 188-192. Giovanola, B., Kurz, W. (1990), Metall. Trans. 21 A, 260-263. Girot, G. (1987), Ph.D. Thesis, L'Universite de Bordeaux I. Haider, E., Roosz, A., Exner, H.E., Fischmeister, H. F. (1987), Materials Science Forum 13/14, 547-558. Heinrich, J.C, Felicelli, S., Nandapurkar, P., Poirier, D.R. (1989), Metall. Trans. 20B, 883-891. Izutani, M., Soejima, T, Saito, T, Kobayashi, X, Ayata, K. (1988), 4th International Conference on Continuous Casting, Preprints; 115-127. Joly, P. A. (1974), Ph.D. Thesis, Department of Materials Science and Engineering, Massachusetts Institute of Technology. Joly, P. A., Mehrabian, R. (1976), /. Mater. Sci. 11, 1393-1418. Jones, H. (1982), Rapid Solidification of Metals and Alloys. London: Institute of Metals. Jones, H. (1984), /. Mat. Sci. 19, 1043-1076. Kattamis, T.Z., Coughlin, J.C, Flemings, M.C. (1967), Trans. AIME 239, 1504-1511. Kattamis, T.Z., Piccone, T.J. (1991), Mat. Sci. Eng. A 131, 265-272. Kenney, M.P., Courtois, J. A., Evans, R.D., Farrier, G.M., Kyonka, C.P., Koch, A.A., Young, K.P. (1988), Metals Handbook, 9th ed., Vol. 15. Metals Park, OH: ASM International, pp. 327-338. Kirkwood, D.H. (1984), Mat. Sci. Eng. 65, 101-109. Kobayashi, S. (1988), /. Cryst. Growth 88, 87-96. Kou, S., Poirier, D.R., Flemings, M.C. (1978), Metall. Trans. 9B, 711-719. Kurz, W, Giovanola, B. (1988), J. Cryst. Growth 91, 123-125.
1.18 References
Kurz, W, Giovanola, B., Trivedi, R. (1986), Acta Metall. 34, 823-830. Laxmanan, V., Flemings, M. C. (1980), Metall. Trans. 11 A, 1927-1937. Lipton, X, Kurz, W, Trivedi, R. (1987), Acta Metall. 35, 957-964. Loue, W.R., Kool, W.H. (1989), Extended Abstract, Conference, Institute of Metals, London. Loue, W.R., Nava-Vazquez, E., Kool, W.H. (1990), First International Conference on Semi-Solid Processing of Alloys and Composites, Sophia-Antipolis, France. Mada, M., Ajersch, F. (1990), Abstract, TMS Annual Meeting, Anaheim, CA, Feb. 1990. Maples, A.L., Poirier, D.R. (1984), Metall. Trans. 15B, 163-172. Massoubre, J.M., Pflieger, B.F. (1978), AIChE Symposium Series 74, 48-57. Masur, L.J., Mortensen, A., Cornie, J. A., Flemings, M.C. (1989), Metall. Trans. 20A, 2549-2557. Mathur, P., Apelian, D., Lawley, A. (1989), Acta Metall. 37, 429-443. McCoy, J.W., Jones, C , Wawner, F.E. (1988), SAMPE Quart. 19(2), 37-50. McDonald, R.J., Hunt, I D . (1970), Metall. Trans. 1, 1787-1788. Mehrabian, R., Keane, M., Flemings, M.C. (1970a), Metall. Trans. 1, 1209-1220. Mehrabian, R., Keane, M., Flemings, M.C. (1970b), Metall. Trans. 1, 3238-3241. Mehrabian, R., Riek, R.G., Flemings, M.C. (1974), Met. Trans. A 5 A, 1899-1905. Metz, S.A., Flemings, M.C. (1969), Trans. AFS. 77, 329-334. Metz, S.A., Flemings, M.C. (1970), Trans. AFS. 78, 453-460. Michael, A.B., Bever, M.B. (1954), Trans. AIME 200, 47-56. Miyazawa, K., Schwerdtfeger, K. (1981), Arch. Eisenh. 52, 415-422. Molenar, J.M.M., Kool, W.H. (1989), /. Mater. Sci. 24, 1782-1794. Moon, H.K. (1990), Ph.D. Thesis, Department of Materials Science and Engineering, Massachusetts Institute of Technology. Mori, N., Ogi, K., Matsuda, K. (1979), /. Jpn. Inst. Metals 43, 858-865. Mori, N., Okamoto, A., Ogi, K. (1989), Proceedings, Japan-U.S. Cooperative Science Program, Oiso, Japan. Mortensen, A. (1989), Metall. Trans. 20 A, 247-253. Mortensen, A., Cornie, J. A., Flemings, M.C. (1988), Metall. Trans. 19 A, 709-721. Mortensen, A., Masur, L.J., Cornie, J.A., Flemings, M.C. (1989), Metall. Trans. 20A, 2535-2547. Mullins, W.W., Sekerka, R.F. (1963), J. Appl. Phys. 34, 323-329. Mullins, W.W., Sekerka, R.F. (1964), J. Appl. Phys. 55,444-451.
55
Nakada, M., Shiohara, Y, Flemings, M.C. (1990), ISIJ International 30, 27-33. Nandapurkar, P., Poirier, D.R., Heinrich, J.C., Felicelli, S. (1989), Metall. Trans. 20B, 711-721. Nomura, J., Tarutani, Y, Mori, K. (1981), Tetsu-toHagane 67, 80-87. Ogilvy, A.J.W. (1990), Osprey Metals Ltd, Glamorgan, U. K. (private communication). Ogilvy, A.J.W, Kirkwood, D.H. (1987), Appl. Sci. Res. 44, 43-49. Ohnaka, I. (1986), Trans. ISIJ 26, 1045-1051. Ohnaka, I., Fukusoko, T. (1981), Trans. ISIJ21, 485494. Ohnaka, I., Kobayashi, K. (1986), Trans. ISIJ 26, 781-789. Ohnaka, I., Matsumoto, M. (1988), Solidification Processing 1987. London: Institute of Metals, pp. 98-99. Petrakis, D., Flemings, M . C , Poirier, D.R. (1981), Modeling of Casting and Welding Processes: Brody, H.D., Apelian, D. (Eds.). Warrendale, PA: TMSAIME, pp. 285-312. Piwonka, T.W, Flemings, M.C. (1966), Trans. AIME 236, 1157-1165. Poirier, D.R. (1987), Metall. Trans. 18B, 245-255. Ramachandran, N., Gupta, G.R., Jaluria, Y (1981), Num. Heat Transfer 4, 469=484. Rames, H.A., Hutton, J.F., Walters, K. (1989), An Introduction to Rheology. New York: Elsevier Science Publishers. Rappaz, M. (1989), Int. Mater. Rev. 34, 93-123. Richards, R.S., Rostoker, W. (1956), Trans. ASM 48, 884-903. Ridder, S. D., Kou, A., Mehrabian, R. (1981), Metall. Trans. 12 B, 435-447. Riedl, R., Fischmeister, H.F. (1990), Metall. Trans. 21 A, 264-266. Riek, R.G., Young, K.P., Matsumoto, N., Mehrabian, R., Flemings, M.C. (1975), SDCE 1975 Transactions, 8th International Die Casting Exposition and Congress, Detroit, MI, Paper No. G-T75, p. 153. Rohatgi, P.K., Pai, B.C., Panda, S.C. (1979), /. Mater. Sci. 14, 2277-2283. Rohatgi, P.K., Asthana, R., Das, S. (1986), Int. Metals. Rev. 31, 115-139. Roosz, A., Gacsi, Z., Fuchs, E.G. (1984), Acta Metall. 32, 1745-1754. Rosenberg, R.A., Flemings, M.C, Taylor, H.F. (1960), Trans. AFS. 28, 518-528. Sarreal, J. A., Abbaschian, G. J. (1986), Metall. Trans. 17A, 2063-2073. Schmid, G., Roll, A. (1939), Z. Elektrochem. 45, 769775. Scheil, E. (1942), Z. Metallic. 34, 70-72. Sekerka, R.F. (1965), /. Appl. Phys. 36, 264-268. Sellamuthu, R., Brody, H.D., Giamei, A.F. (1986), Metall. Trans. 17B, 347-356.
56
1 Solidification Processing
Shaw, L.H., Beech, J., Hickley, R.H. (1986), Ironmaking and Steelrnaking 13, 154-160. Southin, R.T. (1966), J. Inst. Metals 94, 401-407. Spencer, D.B. (1971), Ph.D. Thesis, Department of Metallurgy and Materials Science, Massachusetts Institute of Technology. Spencer, D.B., Mehrabian, R., Flemings, M.C. (1972), Metall. Trans. 3, 1925-1932. Tacke, K.H., Grill, A., Miyazawa, K., Schwerdtfeger, K. (1981), Arch. Eisenh. 52, 15-20. Taylor, H.F., Flemings, M . C , Wulff, J. (1959), Foundry Engineering. New York: J. Wiley. Tiller, W. A., Jackson, K. A., Rutter, J.W., Chalmers, B. (1953), Ada Metall. 1, 428-437. Uhlmann, D.R., Seward, T.P. Ill, Chalmers, B. (1966), Trans. AIME 236, 527-531. Vogel, A. (1978), Metal Sci. 12, 576-578. Wu, Y., Piccone, T.J., Shiohara, Y, Flemings, M.C. (1987), Metall. Trans. 18 A, 915-924. Yao, M., Nanba, A., Noguchi, H., Nakanishi, K., Shinjo, Y, Kinoshita, K. (1984), Trans. ISIJ 24, B178. Yeum, K.S., Poirier, D.R. (1988), Cast Metals 1, 161-170. Yeum, K.S., Laxmanan, V., Poirier, D.R. (1989), Metall. Trans. 20A, 2847-2856. Young, K.P., Riek, R.G., Boylan, J.F., Bye, R.L., Bond, B.E., Flemings, M.C. (1976), Trans. AFS.
84, 169-174; Die Casting Engineer, March-April, 45-52. Young, K.P., Kyonka, C.P., Courtois, J.A. (1983), U.S. Patent 4415374. Young, K.P., Tyler, D.E., Cheskis, H.P., Watson, W.G. (1984), U.S. Patent 4482012.
General Reading Flemings, M.C. (1974), Solidification Processing. New York: McGraw-Hill. Jones, H. (1982), Rapid Solidification of Metals and Alloys. London: Institution of Metallurgists (now: Institute of Metals). Mehrabian, R., Parrish, P. A. (Eds.) (1988), Rapid Solidification Processing: Principles and Technologies IV. Baton Rouge, LA: Claitor's Publishing Division. Solidification and Casting of Metals (1979), London: The Metals Society (now: Institute of Metals). Solidification Processing 1987 (1988), London: Institute of Metals. Kurz, W (1989), Fundamentals of Solidification, third revised edition, Fisher, D. J. (Ed.). Aedermannsdorf, Switzerland: Trans Tech Publications.
2 Rapid Solidification C. Suryanarayana Institute for Materials and Advanced Processes, University of Idaho, Moscow, ID, U.S.A.
List of Symbols and Abbreviations 2.1 Introduction 2.2 Rapid Solidification Techniques 2.3 Spray and Droplet (Atomization) Methods 2.3.1 Gas Atomization 2.3.2 Water Atomization 2.3.3 Ultrasonic Gas Atomization 2.3.4 Rotating Atomization Processes 2.3.4.1 Centrifugal Atomization 2.3.4.2 Laser Spin Atomization 2.3.4.3 Electron Beam Rotating Disk Process 2.3.4.4 Rotating Electrode Process (REP) 2.3.4.5 Rapid Spinning Cup (RSC) Process 2.3.4.6 Perforated Rotating Cup (PRC) Process 2.3.5 Soluble Gas Atomization 2.3.6 Electrohydrodynamic Atomization (EHDA) 2.3.7 The Drop Tube Method 2.3.8 The Spark Erosion Technique 2.3.9 Twin Roll Atomization 2.3.10 Vibrating Electrode Atomization 2.3.11 The Duwez "Gun" Technique 2.3.12 Spray Deposition Methods 2.3.12.1 Spray Rolling 2.3.12.2 Spray Forging 2.3.12.3 Centrifugal Spray Deposition 2.3.12.4 The ALCOA Flake Process 2.3.12.5 Plasma Spray Deposition 2.4 Chill Methods 2.4.1 The Die Method 2.4.2 The Piston-and-Anvil Technique 2.4.3 Twin-Roller Quenching 2.4.4 Melt Spinning Process 2.4.4.1 Free-Flight Melt Spinning 2.4.4.2 Chill-Block Melt Spinning (CBMS) 2.4.4.3 Centrifuge Melt Spinning Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
59 60 62 63 64 65 66 67 67 68 68 69 69 70 70 71 71 71 71 72 72 74 74 74 75 75 75 76 76 76 78 80 80 82 86
58
2.4.4.4 2.4.5 2.4.6 2.4.7 2.4.7.1 2.4.7.2 2.4.8 2.4.9 2.4.10 2.5 2.6 2.7 2.7.1 2.7.2 2.8 2.9
2 Rapid Solidification
Planar Flow Casting Process In-Rotating Water Spinning (INROWASP) Taylor Wire Process Melt Extraction Process Crucible Melt Extraction (CME) Pendant Drop Melt Extraction (PDME) Melt Drag Process Melt Overflow Process Comparison of Rapid Solidification Techniques Laser Surface Treatment Cooling Rates in Rapid Solidification Consolidation Methods Shock Wave (Dynamic) Compaction Hot Compaction Concluding Remarks References
86 88 88 89 89 91 92 93 93 93 96 99 100 101 102 104
List of Symbols and Abbreviations
List of Symbols and Abbreviations A a D d h K L n, n2 t Tm V z
constant constant diameter eutectic lamellar spacing heat transfer coefficient material constant length material constants solidification rate melting temperature wheel velocity section thickness
X
secondary dendrite arm spacing
ALCOA CBMS CME CSC CW EREP EBSQ EDM EHDA FFMS HIP INROWASP LREP PREP PDME PDRSC PFC PRC PSV r.f. REP ROC RS RSC RSR VHP
Aluminum Company of America chill-block melt spinning crucible melt extraction centrifugal shot casting continuous wave electron beam rotating electrode process electron beam splat quenching electric discharge machining electrohydrodynamic atomization free-flight melt spinning hot isostatic pressing in-rotating water spinning laser rotating electrode process plasma arc rotating electrode process pendant drop melt extraction pendant drop rapid spinning cup process planar flow casting perforated rotating cup process pulverisation sous vide radio frequency rotating electrode process rapid omnidirectional compaction rapid solidification rapid spinning cup process rapid solidification rate vacuum hot pressing
59
60
2 Rapid Solidification
2.1 Introduction Scientific investigations by metallurgists and materials scientists have been continuously directed towards improved performance of materials in ever increasing hostile environments. Thus, stronger, tougher, hotter and corrosion-resistant materials are required. These properties for materials have been traditionally achieved through chemistry modifications and conventional thermal, mechanical and thermo-mechanical processing treatments. In fact, materials processing occupies a central place in the interplay of internal structure, external properties and performance in service. Quenching of steels from the austenitic phase field to produce the strong and hard martensitic phase and subsequent tempering to control the microstructure and properties is a classic example of heat treatment. Enhancing the mechanical behavior through precipitation hardening by producing metastable transition phases through aging a supersaturated solid solution is another example. Independently of the details, all these methods to improve the performance of materials in general, and metals and alloys in particular, involve transforming the materials into a metastable condition either by quenching from a high temperature (but still in the solid state), deformation or irradiation. A major limitation in all these methods is the extent of metastability (departure from equilibrium) achieved; it is limited essentially by the low cooling (solidification) rates, reaching only about 103 K s ~ l , obtained during solid-state quenching. These low solidification rates arise through the inherent limited heat transfer efficiency of the convection and radiation modes. The advent of the "gun" technique by Duwez and his associates (1960) to rapidly
solidify metallic melts has heralded a new era in materials science by vastly expanding the processing window for a variety of existing materials and enabling processing of entirely novel compositions. This simple technique involved rapid heat removal by conduction from a thin liquid layer shot onto a good heat conducting substrate such as copper. Good thermal contact with the substrate and small thickness of the melt resulted in solidification rates approaching 10 6 Ks~ 1 and achievement of relatively large (hundreds of degrees) undercooling of the melt before significant amounts of a solid phase could form. Rapid solidification processing has now come to be recognized to result in characteristic constitutional and microstructural changes. The constitutional changes include (a) extension of solid solubility limits, (b) formation of new non-equilibrium crystalline or quasicrystalline intermediate phases, and (c) production of metallic glasses. Interestingly, all these effects were first observed by Duwez and his colleagues during the first year of the use of the "gun" technique (Duwez et al., 1960; Klement et al., 1960). The microstructural effects include changes in the morphology and refinement of dimensions of microstructural features (the size, shape and location of grains and phases present). The change is in the direction of a more uniform and finer microstructure with a large reduction in scale of solute segregation effects. Additionally, all these effects lead to increased chemical homogeneity of the material. In comparison with conventional ingot materials, rapidly solidified materials exhibit greatly reduced segregation levels. Thus, the grain size, dendrite arm spacing, and the size of constituent particles, precipitates and dispersoids are refined leading to significant reduction in diffusion times for homogenization. As an example, it has
2.1 Introduction
been shown (Cohen et al., 1980) that at 1600 K, the time required to reach 99% homogenization of tungsten in nickel in conventional solidification is 16 h when the segregate spacing is 100 jim. On the other hand, in rapidly solidified alloys when the segregate spacing is only about 1 jim, the corresponding time is only 6 s. Hence, chemical homogeneity can be achieved much more easily in rapidly solidified materials. The metastable effects mentioned above confer beneficial properties on the rapidly solidified materials. A vast amount of literature, amounting to over 20000 publications, is available today on various aspects of rapidly solidified materials. A major source of this literature is the proceedings of the international conferences on Rapidly Quenched Metals held once in three years regularly since 1975 (the seventh in the series was held in Stockholm, Sweden, during August 13-17, 1990). In addition, several books, bibliographies, reviews and other conference proceedings give state-ofthe-art situation on structure, properties and applications of rapidly solidified materials (Jones, 1982; Anantharaman and Suryanarayana, 1987; Jones and Suryanarayana, 1973; Suryanarayana, 1980; Guntherodt and Beck, 1981; Beck and Giintherodt, 1983; Luborsky, 1983; Anantharaman, 1984). In comparison to conventional materials, rapid solidification processing has produced materials with greatly superior strength properties, improved corrosion resistance and a highly desirable combination of magnetic properties. In addition to a general improvement in the properties, rapid solidification processing allows production of materials with compositions otherwise impossible. While the 1960's saw the emergence of the rapid solidification principles and observation of metastable
61
effects in a variety of alloy systems, the 1970's witnessed enhanced activity in the field of metallic glasses. This increased activity on glassy alloys was due to the development of techniques to produce long and continuous tapes with uniform cross section. The unusual soft magnetic properties (reasonably high saturation magnetization, very low coercivity, zero magnetostriction and high electrical resistivity) of metallic glasses prompted extensive research into this field and also development of applications for these novel magnetic materials, especially as transformer core materials (Egami, 1984). However, metallic glasses crystallize during annealing at high temperatures and lose all the beneficial effects. Thus, during the 1980's, there was a resurgence of activity in the area of rapidly solidified microcrystalline alloys - especially those based on light metals - for possible application in the aerospace industry. The discovery of quasicrystalline phases (Shechtman et al., 1984; Suryanarayana and Jones, 1988) has given a short-lived added impetus to this field. The metastable effects of rapid solidification can be achieved in other ways also. The starting material in these non-RSP techniques can be in the form of either a solid, liquid, or vapor. By their very nature, the quenching rates obtained in the solid state are not high and the metastable effects therefore are limited. The quantity of the materials that can be produced per run is low when the starting material is in the vapor phase. (Recently, however, it has been shown (Bickerdike et al., 1986; Gardiner and McConnell, 1987) that large sheets can be produced by consolidating the product after physical vapor deposition or vacuum evaporation methods.) None of these techniques can easily compete with rapid solidification processing in rate of production. Therefore, in the present chap-
62
2 Rapid Solidification
ter, techniques such as thermal vapor deposition (Chopra, 1969), sputtering (Dahlgren, 1978), chemical vapor deposition (Bryant, 1977), electrodeposition (Brenner, 1963), electroless deposition (Bonetti et al, 1981), ion or electron irradiation (Rechtin etal, 1978; Fujita and Mori, 1988), rapid pressure application (Reddy etal., 1986; Ponyatovsky, 1988), ion implantation (Borders, 1979), ion beam mixing (Liu, 1988), solid-state diffusion reactions (Yeh et al., 1983; Johnson, 1986), etc. will not be considered, even though these may have specific advantages in particular cases. Some of these alternative techniques are treated by Follstaedt in Chap. 6 of this volume. Ever since the advent of the "gun" technique, a great variety of techniques have been developed to achieve rapid solidification of metals and alloys. Most of these techniques are capable of producing grain refinement and inducing one or more of the metastable effects referred to earlier. The present chapter is devoted to a brief description of the important techniques available today for rapid solidification of metallic melts, resulting in the formation of powders, fibers, tapes, flakes, wire, etc. Variables that affect the solidification rates and product parameters are also discussed. The special features of laser treatment of solid surfaces are covered in Chap. 3 of this volume. A more general treatment of solidification will be found in Chap. 1 of this volume, while the underlying physical theory of solidification is treated in Chap. 10 of Volume 5.
2.2 Rapid Solidification Techniques Jones (1981) has presented a historical survey of the early techniques developed to rapidly solidify metallic melts. Although these methods were not designed primarily
to achieve striking non-equilibrium effects, these can be considered the precursors of the present-day rapid solidification techniques. The techniques of rapid quenching from the melt have been reviewed periodically in recent years with emphasis on different aspects (Duwez, 1968; Anantharaman and Suryanarayana, 1971; Chen et al., 1980; Jones, 1982; Liebermann, 1983; Savage and Froes, 1984; Fleetwood, 1987; Anantharaman and Suryanarayana, 1987). High enough cooling rates during solidification can be achieved when two important requirements are satisfied. The solidification rate, T, during ideal cooling is related to the section thickness, z (in mm) through the relation t = 10 4 z~ 2
(2-1)
suggesting that the solidification rate increases by two orders of magnitude for a decrease in section thickness by one order of magnitude. Thus, firstly, the molten metal must be delivered in a stream which is thin enough in at least one dimension and have a high surface area to volume ratio to allow rapid heat removal. Secondly, rapid heat removal can be achieved by maximizing the contact area between the melt and the cooling medium by rapidly increasing the liquid alloy surface area. This may be effected either by altering the shape of the melt during processing (e.g., spreading it as a thin layer on a substrate) or by physically disintegrating the melt into small droplets (e.g., by atomization). Fig. 2-1 presents combinations of three possibilities which are used in practice and also lists the major processes to which they apply. As can be clearly seen, the molten metal can be delivered in the form of either droplets, cylindrical stream or ribbon stream and the melt stream can be cooled by gas, liquid or solid.
2.3 Spray and Droplet (Atomization) Methods
Droplets
Gas
Cylindrical stream
Ribbon stream
Liquid
Solid
1. Atomization processes, rapid solidification rate (RSR), rotating electrode process (REP) 2. Water atomization, rapid spinning cup (RSC) 3. Duwez gun, piston-and-anvil, drum splat, electron beam splat quenching, controlled spray deposition, spray deposition, Osprey 4. Taylor wire, free-flight melt spinning
63
Rapid solidification methods have been classified into various categories by different authors. For example, Jones (1982) classifies them into three categories (i) spray methods involving fragmentation of melt into droplets prior to quenching, (ii) chill methods preserving continuity of the melt up to and during quenching, and (iii) surface methods which involve rapid melting and solidification of a limited depth at the surface of a more substantial thickness of material acting as the heat sink. On the other hand, Savage and Froes (1984) classify the methods into two groups, viz., Atomization and Non-Atomization, depending on whether atomization of a melt stream into droplets is involved or not. Although each of these classifications has its advantages, we shall follow the former classification and describe the techniques accordingly in the following pages.
5. Free-flight melt spinning, in-rotating water spinning 6. Melt extraction, pendant drop melt extraction 7. Melt spinning, planar flow casting, melt drag, melt overflow
Figure 2-1. Combinations of molten metal stream and cooling medium used in the principal rapid solidification processes.
The shape, size and nature of the products which result from the different combinations of molten metal stream and cooling medium depend essentially on what happens to the metal stream before it solidifies. For example, droplets which are allowed to solidify individually form powders or flakes, whereas droplets which land on top of one another before they have solidified coalesce into solid shapes. Similarly, a continuous cylindrical stream solidifies as wire, but a stream which is broken up solidifies as fibers.
2.3 Spray and Droplet (Atomization) Methods In these methods, a continuous stream of liquid metal is atomized, i.e., broken down into fine droplets by means of a gas or a liquid. The resultant product after solidification is powder. The mechanism of achieving atomization and the means of cooling can be different in different techniques. For large-scale applications of rapidlysolidified alloys, the preferred form for consolidation is powder, because of its nearly equiaxed geometry. The deformation processes which occur during the consolidation of powder are characterized by large amounts of relative motion between particles and the flow of individual particles. The high degree of interfacial shear that results helps to disrupt oxide films providing clean surfaces across which good inter-
64
2 Rapid Solidification
particle bonding can occur. Because of this, atomization techniques have become the preferred methods of rapid solidification for commercial exploitation (see Chap. 4 in this volume). Atomization is not a very new technique. The Merchant's Shot Tower in Baltimore, MD, USA, was continuously used between 1828 and 1892 to produce some half-million 25 lb bags of spherical lead shot each year (Adam, 1986). In recent years several versions of the atomization techniques have been developed and some overviews of these different techniques have been presented from time to time (Lawley, 1977, 1978,1981; Grant, 1983; Miller, 1983; Aller and Losada, 1990). 2.3.1 Gas Atomization Conventional gas atomization of liquid metals has been in continuous use since the 193O's to produce a variety of metallic powders for diverse applications. Each space shuttle launch, for example, consumes 160000 kg of atomized aluminum powder as part of the solid fuel propellent mixture (NASA, 1980). Gas atomization involves breaking down of a continuous stream of liquid metal by one or more high-velocity jets of air or any other gas. Atomization occurs by kinetic energy transfer from the atomizing medium to the metal. The small liquid particles solidify in flight by convection or radiation. The solidification rate depends on the particle size, with higher rates associated with smaller particle sizes. The number and geometry of gas/metal configurations used in commercial practice are very varied. Typically, two or more jets, or a ring, are positioned around the axis of the metal stream. The axes of the gas jets are equally inclined to the metal stream axis and intersect this axis at the geometrical
impingement point. A representative configuration is shown in Fig. 2-2 a. Details of recent developments of this process can be found in the annual conference proceedings of the various powder metallurgy societies. The overall process of gas atomization is governed by several interrelated operating parameters: jet distance, jet pressure, nozzle geometry, velocity and mass flow rate of gas and metal, metal superheat, angle of impingement, metal surface tension and metal melting range. Consequently, rigorous interpretation of experimental data is difficult even though empirical expressions have been proposed relating the particle size distribution to the atomization conditions (Lubanska, 1970; Kim and Marshall, 1971). The solidification rate depends on the melt droplet size and on the type of atomization medium. Higher solidification rates are achieved with smaller particle sizes and lighter gases. For example, helium provides more rapid cooling than argon. In practice, nitrogen, hydrogen, argon, or air is used even though mixtures of these gases and helium are also effective. If sufficient superheat is provided and the atmosphere is neutral to the alloy, the final powder product is a sphere. The range of powder sizes is broad with the mean particle diameter around 100 |im (Fig. 2-3), even though Unal (1990) reported a mean diameter of 12 to 15 jim for gas-atomized zinc powder. This technique has been employed for atomizing a wide range of alloys, including superalloys, high-alloy steels and aluminum alloys. With aluminum, however, a special problem exists. When aluminum powders are handled in moist air, hydrates tend to form on their surfaces. These, if not removed, tend to form blisters on subsequent solution annealing. Inert gas atomization is currently the primary process for the production of su-
2.3 Spray and Droplet (Atomization) Methods
65
Melt
Melt
Gas jets
Gas jets
Liquid jets
'/rWW
Powder slurry
^Atomized powder
Figure 2-2. Schematic showing the principle of (a) gas atomization and (b) gas-water atomization.
Argon atomized
Rotating electrode process
10
60 80100 Particle size
peralloy, titanium and other reactive metal powders (Moll and Yolton, 1986; Hohmann and Jonsson, 1990). This process suffers from a very low overall energy efficiency (~3%) and is expensive if inert gases other than nitrogen have to be used. The yield is about 80% for - 35 mesh powder. Typical solidification rates achieved are 10 2 -10 3 K s " 1 . Hopkins (1987) has recently given insight into the critical areas of atomizer design affecting powder morphology, fineness and purity.
Figure 2-3. Representative particle size distribution for powders prepared by various atomization methods.
400 600 800 1000
2.3.2 Water Atomization
This is similar to gas atomization except that water jets are used instead of gas jets to atomize the metal stream. Further, the water pressure used is higher than the pressure employed in gas atomization and the angle of impingement is smaller. Water atomization has been used extensively to make powders of tool and low-alloy steels, copper, tin, and iron. Excess oxide on the particle surfaces can be subsequently re-
66
2 Rapid Solidification
moved or minimized in amount by hydrogen reduction. Trial production of steels (Dunkley, 1982) and superalloys (Tracey and Cutler, 1981) has shown that reactive components tend to oxidize, making it difficult to achieve oxygen contents below 1000 ppm, as against 100 ppm typical of gas-atomized superalloy powders. Representative operating conditions for these two types of atomization are listed in Table 2-1. Attempts have also been made in recent times to achieve more effective atomization (Fortman and Ullman, 1984; Smith, 1985; Couper and Singer, 1985) to achieve higher yields of fine powders. A similar effect has also been claimed with the use of higher gas pressure (Ricks and Clyne, 1985). Water atomization requires large quantities of energy to supply the water at high pressures. Consequently, the overall energy process efficiency is less than about 4%. The particles are irregular in shape and the average solidification rate is in the range of 10 2 -10 4 K s " 1 . Higher solidification rates of 10 5 10 6 Ks~ 1 for 20 jam particles have been achieved by quenching the gas-atomized
Table 2-1. Typical operating conditions for gas and water atomization processes. Parameter
Gas atomization
Water atomization
Pressure (MPa) Velocity (m s"1) Superheat (°C) Angle of impingement Particle size (urn) Particle shape
1.4-4.2 50-150 100-200 15-90°
3.5-21 40-150 100-250 ^30°
50-150 Smooth and spherical 40% at — 325 mesh
75-200 Irregular and rough surface 60% at — 35 mesh
Yield
droplets in streams of high-velocity water (Miller and Murphy, 1979) (Fig. 2-2 b). By this combined technique an amorphous phase could be produced in 100 jim diameter Cu 60 Zr 40 powder particles. The process offers some degree of control over particle geometry by the relative placement of the gas and liquid jets. Oil atomization of iron-base alloys has recently been tried out (Kainer and Mordike, 1989) to produce powders in the size range of 10-150 jim. The solidification rates achieved (about 105 K s " 1 at 20 jim powder size) are comparable to those obtained in water atomization. However, a decisive advantage is the low oxygen content of the oil-atomized powders. 2.3.3 Ultrasonic Gas Atomization This method is again similar to gas atomization except that the impinging gas is at a very high velocity (up to Mach 2.5). Further, while the gas jets flow in a continuous manner in normal gas atomization, they are pulsating at a high frequency of 80-100 kHz in ultrasonic gas atomization. The pulsations are provided by the incorporation of a series of Hartman shock wave tubes in the gas passage of the atomizing die (Anand et al., 1980). This causes the molten stream to break up more efficiently, producing finer, more uniform particles quenched at higher average rates. It is claimed that droplet formation in ultrasonic gas atomization is a one-step process in contrast to the three-stage mechanism of conventional gas and water atomization. The average particle size is about 20 |im (with a narrow size distribution) and the cooling rates estimated are of the order of 105 K s " 1 . The yields are in excess of 90% and this technique has been successfully exploited for producing aluminum and superalloy powders (Grant, 1983) and Ti-Al
67
2.3 Spray and Droplet (Atomization) Methods
powders (Sastry et al., 1983), Because of the large specific surface area, care must be exercised in safe handling of these powders.
Cooling gas
2.3.4 Rotating Atomization Processes 2.3.4.1 Centrifugal Atomization
Centrifugal atomization is a technique, the different variants of which involve disintegration of liquid metal at the rim or edge of a rapidly rotating container (disk, cup, crucible, or plate which may be flat or concave). The liquid metal is supplied to the rotating part by pouring from a crucible or tundish, or by melting the end of a bar in situ. Owing to the centrifugal force, the molten metal is forced out to the edge of the container where it is atomized, ejecting particles of molten metal. The droplets spheroidize and solidify in flight. The whole process - melting, atomization and solidification - takes place in an inert atmosphere. In the Rapid Solidification Rate (RSR) Centrifugal Atomization process as practiced by Pratt and Whitney (Cox et al., 1976), the melt stream is usually bottom-poured on to the center of a watercooled horizontal disk, spun at speeds up to 35 000 rpm. The liquid metal is mechanically atomized and thrown off the edges of the spinning disk. Solidification takes place in flight, and can be speeded up by blasting the emerging particles with a stream of helium (Fig. 2-4). The helium atmosphere can also minimize oxidation. The particles are spherical and the solidification rates are of the order of 104 -10 6 K s ~1 for particle size of 25-80 jum (Fig. 2-3). Nickel, aluminum, titanium and superalloy powders have been produced by this technique. Rapid erosion of the crucible orifice at high temperatures results in coarser powders and lower solidification rates. The deformation and overall life of the spinning
Fine particles
Rotary atomizer disk
Figure 2-4. Schematic of the rapid solidification rate centrifugal atomization technique.
disk/cup presents other problems at very high centrifugal speeds. In the related Centrifugal Shot Casting (CSC) process, an electric arc between a stationary electrode and a rotating crucible containing the charge melts the metal (Fig. 2-5). Owing to the centrifugal force, the molten metal is forced out through the edge of the crucible where it is atomized, ejecting particles of the molten metal (Sutcliffe and Morton, 1976).
Electrode
Atomization rim (spin off)
Powder particles Alloy material Rotating crucible Electrode
Figure 2-5. Schematic of the centrifugal shot casting process.
68
2 Rapid Solidification
2.3.4.2 Laser Spin Atomization
This process (Konitzer et al., 1984; Peng etal., 1985 a) is similar to the Pratt and Whitney process, but a high-power laser beam is used to melt the top surface of a rod rotating at 10000-30000 rpm. The molten droplets expelled by centrifugal force from the circumference of the rod are cooled by flowing helium gas before they impinge on the walls of the apparatus. The droplets solidify predominantly as spherical particulates, although 10-30 wt.% of the yield was in the form of needles with diameters of 0.1 to 1.0mm (Peng etal., 1985 b). Typical solidification rates of 105 K s " 1 are achieved for 100 jum powders using this technique; different cooling rates can be obtained by controlling the rotational speed and gas flow which result in different powder diameters. Two particularly significant advantages of this process are (i) the high power laser beam causes high superheats so that most second-phase particles are dissolved in the melt and (ii) the technique is of the local melting type, so that contamination during atomization can be kept to a minimum. This process has been extensively used to produce titanium alloy powders (Sastry et al., 1983).
2.3.4.3 Electron Beam Rotating Disk Process
In this variation of the method (Wahlster etal., 1980) a vertically fed, slowly rotating cylindrical bar form is drip melted by several electron beams. The droplets run down the pencil-shaped bar and drop into the center of the rotating disk. Molten metal particles are thrown off the rim, preferably within a horizontal angle between 60° and 80° and are deflected downwards by a water-cooled copper shield
Electron beam guns
Alloy bar Electron beam gun
Rotary water cooled copper crucible
Figure 2-6. Schematic of the electron beam rotating disk process.
(Fig. 2-6). Using this process, it is possible to manufacture metal powders of highly reactive materials. The powder size is between 30 and 50 |im and the particle shape is spherical with clean surfaces. In a slightly different variation of this process known as Electron Beam Splat Quenching (EBSQ), an electron beam is focused on to the end of an alloy rod where it melts the surface and produces molten droplets (Sastry et al., 1984; Seshadri et al., 1988). The molten droplets fall onto a rotating disk and are stretched into thin flakes under the combined actions of a high angular velocity and the centrifugal force of the rotating disk. It has been recently reported (Seshadri et al., 1988) that as the substrate angle increases, the shape of the flake changes from elongated to nearly circular and that the lower substrate angles yield thinner flakes solidified at a higher cooling rate. The flake thickness and hence the solidification rate are controlled by varying the rotational speed of the copper disk. Typical cooling rates achieved vary from 104 to 106 K s " 1 .
2.3 Spray and Droplet (Atomization) Methods
2.3.4.4 Rotating Electrode Process (REP) In this process, a rod of the material to be atomized is rapidly rotated while it is simultaneously melted at one end by an electric arc issuing from a non-consumable tungsten electrode (Friedman, 1976; Loewenstein, 1981). Fig. 2-7 shows a schematic diagram of this process. Molten metal spins off the bar and solidifies before hitting the walls of the chamber filled with an inert gas. The process is used fairly widely for the atomization of reactive metals, e.g., high-purity, low-oxygen Ti, Zr, Nb, Ta, V and their alloys, and Ni and Co superalloys. Tungsten contamination from the stationary electrode is possible in REP, but can be avoided by replacement with a titanium cathode or by melting the bar by plasma arc (PREP), laser (LREP), or electron beam (EREP). REP powders are usually spherical, of high surface quality, but have a large average particle diameter, generally more than 200 jim (Fig. 2-3), and so the solidification rates are only about 102 K s " 1 . Typically, yields run 75% for —35 mesh powder. Vacuum
Inert gas Rotating consumable electrode
Non-rotating tungsten electrode
Collection port Figure 2-7. Schematic of the rotating electrode pro-
69
A method similar to the PREP method is Pulverisation Sous Vide (PSV) (or powder under vacuum) process, in which the ingot to be converted into powder is placed in an ingot mold with vertical axis, rotating at a slower speed (1000-4500 rpm) than in the PREP process. An electron gun melts the tip of the ingot, the droplets from which are ejected and solidify into spheres (Devillard and Herteman, 1980). Special measures are taken to prevent the powder particles from flattening against the cold walls of the chamber during solidification by use of carbonaceous shields. This process generates a larger particle size (>90% is + 125 jim) than the PREP process. The carbonaceous shield may cause slight contamination of the powder. 2.3.4.5 Rapid Spinning Cup (RSC) Process In this process, developed at Battelle, Columbus, OH, USA, the molten metal particles fall into a cup with a rotating liquid, usually water. The cup with the quenching liquid is rotated at high speeds (8000-16000 rpm) and the liquid forms a thick layer on the vertical interior side of the cup. The force of the rotating liquid layer improves the heat transfer conditions, increasing the solidification rates and acting also as an atomizer. A wide variety of steels, superalloys, aluminum, copper and many other alloys have been processed by this method. Fig. 2-8 shows a schematic of the rapid spinning cup process (Raman et al., 1984). The pendant drop rapid spinning cup (PDRSC) process is a variant in which a droplet is generated by melting of the rod-end and falls on the rotating liquid. The powder produced by this method has a narrow size distribution (Fig. 2-3) and a high yield of fine powder. Most of the particles are spherical in shape and the
70
2 Rapid Solidification Molten metal
Powder particles
Spinning cup
with sharper ends) are typically 10005000 jim long and 1000 jim in diameter, and are subsequently introduced into a rolling mill and consolidated directly as sheet at rates up to 60 m min" 1 . The process can be made continuous and has been employed to fabricate sheets of relatively low-melting alloys such as aluminum, lead, zinc, etc. Since the canning and outgassing steps during consolidation are eliminated, the economics might be very attractive. 23.5 Soluble Gas Atomization
Figure 2-8. Schematic of the rapid spinning cup pro-
particle surfaces are clean and oxide-free. The RSC process (i) produces rapidly solidified powders, including highly alloyed microcrystalline and glassy alloys, (ii) controls the size distribution in the particles produced from very narrow to very broad, (iii) varies the particle shapes from irregular-to-spherical-to-elongated, (iv) controls the oxygen, nitrogen and hydrogen contamination and (v) does not produce airborne fines. The solidification rates are in the range of 10 4 -10 6 K s " 1 (see also Section 2.4.5).
In this process, also known as the vacuum atomization process, and developed by Homogeneous Metals Co., a crucible of liquid metal supersaturated with a gas under pressure is suddenly exposed to a vacuum. Under these conditions, the gas expands, comes out of solution and causes the metal to be atomized (Wentzell, 1974). Various gases, including nitrogen, argon, and hydrogen have been used. Alloy powders based on Ni, Cu, Co, Fe, and Al have been vacuum atomized with hydrogen as the gas. The powders are mostly spherical (but can be a mixture of powders and flakes depending on the actual disintegration process), clean and are of high purity. The
2.3.4.6 Perforated Rotating Cup (PRC) Process An important commercial variant of the Centrifugal Atomization is the Perforated Rotating Cup process developed by Reynolds Metals Co. (Daugherty, 1964). In this method the molten metal is poured into a deep rotating cup with holes in the side (Fig. 2-9). Centrifugal force causes the molten metal to flow through the holes. The streams created break up in flight and solidify at rates of about 10-10 2 Ks~ 1 . These acicular powders (rice-shaped but
Collect and transfer
Coil
Figure 2-9. Schematic of the Reynold's perforated rotating cup process.
2.3 Spray and Droplet (Atomization) Methods
cooling rates are low (about 102 Ks *). A 1000 kg/run production plant is in operation. 2.3.6 Electrohydrodynamic Atomization (EHDA)
This method was developed to produce rapidly solidified powders and coatings for characterization studies (Perel et al., 1980, 1981). An intense electric field, typically 105 V/m, is applied to the surface of liquid metal contained in a capillary emitter. Under these high field strengths, surface tension forces are overcome and the droplets are emitted. The charged droplets are accelerated towards a collector and may impact while still liquid to produce foils, flakes or a coating, or may solidify in flight to give spherical powders. Particles ranging from 0.1 |nm to 100 |im in diameter can be produced and the solidification rates can be as high as 107 K s " 1 for very fine particles of 0.01 (im diameter (Perel et al., 1980). Amorphous and microcrystalline powders of Al-Si, Al-Cu, aluminumcoated Ti- and Fe-base alloys have been produced. The advantages of this process lie in it being a continuous process, having no vital moving parts, and it can include radiation, convection and splat cooling techniques. 2.3.7 The Drop Tube Method
There has been a revival of interest in recent times in this method (Steinberg et al., 1981; Evans et al., 1986; Vinet et al., 1991), wherein droplets generated at the top of a vertical column are allowed to fall freely to different distances and in different environments. Useful fundamental knowledge can be gained from these studies about the mode and kinetics of solidification from deeply undercooled melts.
71
2.3.8 The Spark Erosion Technique
This method, patterned after the electric discharge machining (EDM) method, involves maintaining a repetitive spark discharge between two electrodes immersed in a dielectric fluid. The electrodes are made of the material whose powder is to be produced. Each spark melts or vaporizes a minute quantity of the material which immediately freezes or condenses to powder particles (Yamaguchi and Narita, 1977; Berkowitz and Walter, 1980). The powder is usually contaminated with breakdown particles of the dielectric fluid and for this reason inert gas cryogenic fluids have been used (Cogan et al., 1978). Amorphous and microcrystalline powders have been produced by this technique, at a rate of 1 to 20 g per 6 h run. 2.3.9 Twin-Roll Atomization
This is a novel process to produce powder by atomizing the molten stream fed between two high-speed rolls rotating in opposite directions (Singer and Roche, 1977). The rolls are coated with a carbon paste to reduce heat transfer and prevent solidification in the nip of the rolls. As the liquid metal emerges from the far side of the rollers, it cavitates and is thrown off as droplets (Fig. 2-10). The droplets are quenched in a water bath, located within 25 mm of the roller nip. Elongated flakes of 200 jim thickness can be produced at solidification rates of 10 5 -10 6 K s " 1 (Durand etal, 1976; Murty and Adler, 1982; Ishii et al., 1982). This method is capable of converting a wide range of metals into flake, acicular, irregular, or spherical particles. At present the process is limited by its inability to produce small liquid metal droplets, as only 2 to 3% of the powder produced is below 37 jim in diameter (Ishii etal., 1982).
72
2 Rapid Solidification
2.3.11 The Duwez "Gun" Technique
Liquid/Solid interface Molten droplets
Figure 2-10. Schematic of the twin-roll atomization process.
2.3.10 Vibrating Electrode Atomization
In this process, an electrode having one end free is continuously moved between rollers towards a slowly rotating copper disk in a vacuum or inert-gas filled chamber. Atomization takes place in the arc struck between the water-cooled disk and the vibrating end of the electrode. The droplets are made to fall off by causing this electrode to resonate by means of a transducer (Matei et al., 1977). Spherical particles are formed whose size and size distribution can be controlled by changing the length of the resonant rod. The narrowest size distribution is obtained at the resonant frequency of the electrode. This method is not widely used and shows little potential for scale up.
This technique, used for the first time by Duwez and his associates in 1959 to quench small quantities of liquid alloys rapidly, remained the most popular of all RS techniques until the advent of continuous production techniques in the 1970's. Even to-day, it retains the distinction of being a simple, elegant and versatile laboratory technique for producing small (mg) quantities of rapidly solidified metals and alloys. Fig. 2-11 shows a schematic of the"gun" apparatus (Duwez and Willens, 1963). Essentially, a small quantity (up to 100 mg) of the metal or alloy is melted by induction or resistance heating in a graphite crucible with an orifice of about 1 mm in diameter at the bottom. Because of high surface tension the melt does not fall through the orifice. Ejected by means of a shock wave, the molten metal passes through the orifice, is atomized and cools rapidly on contact with a copper substrate in the form of a thin foil (up to about 15 jam in thickness). The shock wave is generated by the rupture of a thin mylar diaphragm located between the high- and low-pressure chambers, by an inert gas let into the high-pressure chamber with increasing pressure. Graphite appears to be the best material for the container for metals and alloys which do not react with carbon. In case of reactive and refractory metals like tantalum, tungsten and titanium, either a ceramic insert can be kept at the bottom of the crucible or the metal can be melted with concentrated r.f. induction heating over a silver hearth (Willens and Beuhler, 1966; Ruhl and Cohen, 1969). Because of the rapid heating rate, even alloys with high vapor pressure constituents can be melted with a minimum change in composition. Although the "gun" devices are gen-
2.3 Spray and Droplet (Atomization) Methods
73
Clamping pin High pressure chamber Helium 0-Rings Argon Mylar diaphragm Low pressure chamber Water — Tantalum
Clamp for graphite container
Water cooledi induction coil Graphite container
Figure 2-11. Schematic of the Duwez "gun" technique. 2 cm
erally operated in air, versions operating under controlled atmospheres or vacuum have been described (Shingu et al., 1968; Ruhl and Cohen, 1969; Lohberg and Muller, 1969; Davies and Hull, 1972). The shock wave can also be generated with the help of an explosive (Predecki et al., 1965). Cahn et al. (1976) produced the shock wave by means of vanes rotating rapidly at speeds up to 13 000 rpm and the alloy was levitation melted. The substrate should have a high thermal conductivity to extract the heat rapidly. Therefore, copper is most commonly used. In the earliest version (Duwez et al., 1960), the globule was received on the inside of a rotating cylinder located at about 1.2 cm from the exit of the graphite container. Because of the geometry of the crucible orifice, the full impact of the shock wave could not be realized. The substrate takes the form of a "Ski-Slide" in the later version (Duwez and Willens, 1963) and the
liquid globule strikes the substrate at a glancing angle to improve the thermal contact. The substrate can be maintained at low temperatures, if desired. This helps in retaining those metastable phases which decompose rapidly at ambient temperatures (Kane et al., 1966). Glass substrates have been used to quench oxide melts (Sarjeant and Roy, 1967) and diamond was used in another investigation (Ramachandrarao et al., 1972) to take advantage of its substantially increased conductivity at liquid nitrogen temperatures. The product of this technique, popularly called splat, the technique itself referred to as "splat cooling", is of varying thickness (0.1 to 15 |im in different portions of the same splat) and porosity and is at best a few cm in size. Although best suited for structural characterization by electron microscopy (near the edges and holes, the foils are electron transparent) and X-ray techniques, these splats are not suitable for
74
2 Rapid Solidification
evaluating the physical and mechanical properties. Solidification rates in the range of 106 to 108 K s " 1 and 104 to 1010 K s " 1 have been measured (Predecki et al., 1965) and calculated (Ruhl, 1967) for this technique. It was suggested that higher solidification rates can be achieved by improving the thermal contact between the splat and the substrate and decreasing the thickness of the splat. Quenching the melt in an inert atmosphere (Davies and Hull, 1974) or in vacuum (Boswell and Chadwick, 1976) has been reported to result in higher rates of solidification. Successful operation of this technique involves (a) restricting the quantity of the molten alloy, (b) atomizing it rapidly into very small droplets, and (c) expelling the atomized stream over a short distance and time interval to make high speed impact with the quenching substrate. 2.3.12 Spray Deposition Methods
In these methods, atomized melt particles get deposited on a cooled solid substrate, where the droplets solidify to make a more or less dense deposit, which may be several millimeters in thickness. The deposit can be removed from the substrate and subjected to further processing to densify it. The solidification rate experienced by the initial deposit is much higher because of conduction cooling, whereas the later deposits cool relatively slowly. Singer, who formulated the concept of spray forming (Singer, 1968) has reviewed the spray forming of solid materials (Singer, 1982, 1983). Its application presently includes (a) spray rolling, (b) spray forging, and (c) centrifugal spray deposition. 2.3.12.1 Spray Rolling
In this process (Singer, 1970), a gas-atomized molten stream is sprayed directly
Holding furnace Nitrogen
Substrate
Figure 2-12. Schematic of the spray rolling process with a rotating substrate.
on to the roughened surface of a wide, internally-cooled, slowly-rotating drum (Fig. 2-12). The droplets solidify on impact, with subsequent droplets building up the thickness. This layer is peeled off from the drum and fed directly into a rolling mill to produce a strip. Strips of 18 mm thickness and 0.5 m width have been produced so far. Solidification rates as high as 10 6 Ks~ 1 have been claimed for this process, which can be made continuous and where powder handling is completely avoided. Particle sizes vary widely depending on circumstances, but a typical size is 150 |tim. The main difficulty in spray rolling is in ensuring that a spray deposit is uniformly thick across the width of the strip. Variations in thickness should not be more than ± 2 % for industrial exploitation. A further refinement of spray rolling is spray peening, in which spray deposition is accompanied by simultaneous shot peening to consolidate the spray deposit as it forms (Singer, 1978). 2.3.12.2 Spray Forging
In this process, a spray of gas-atomized metal is directed into a mold or, in some cases, onto a flat substrate or collector
2.3 Spray and Droplet (Atomization) Methods
which is moved in an appropriate manner under the spray by a manipulator. In the case of a mold, the shape is replicated in reverse, whereas in the case of a movable flat collector, the preform shape is determined by the motions of the manipulator. A high spray density ensures that the porosity is only about 1 % and so the preform can be subsequently forged. This process has been adopted by Osprey Metals Ltd. in South Wales, U.K. to produce alloy steels and superalloys (Brooks et al., 1977). The spray-forged materials have been shown to have better mechanical properties and to be far more isotropic than conventional drop forgings. 2.3.12.3 Centrifugal Spray Deposition
This method involves centrifugally generating the spray to produce metal products having axial symmetry together with a central hole (Singer and Kisakurek, 1976). With this method, it is also possible to have very fine-grained non-segregated products to be produced under strict atmosphere control. This could be especially useful for materials such as titanium. A 2-ton commercial scale plant is already in operation to produce tool and high alloy steels (Rickinson et al., 1981). This process has now been renamed Liquid Dynamic Compaction (Ogata etal., 1986) presumably to associate it with the solid-state dynamic compaction technique practised on rapidly solidified powders (Morris, 1980 b). 2.3.12.4 The ALCOA Flake Process
In this process, gas atomized particles impinge on the cooled surface of a rotating drum (Fig. 2-13). Here, they form small splats that leave the drum surface after solidification and are subsequently collected and screened. Quench rates depend upon the size and velocity of the molten parti-
75
Molten droplets
Figure 2-13. The ALCOA process to produce splat particulate by impact of atomized droplets on a rotating cooled drum.
cles, and rates up to 106 K s i appear possible. The particle, or flake, size is quite variable. Handling and pouring are not as easy as with spherical powders, but these flakes are subjected to higher quench rates than most of the spherical particulates and therefore, a wide range of alloy compositions can be processed. 2.3.12.5 Plasma Spray Deposition
Very hot, highly ionized plasma jets are commonly employed to melt and to spray prealloyed powder onto cold substrates (Moss, 1968; Moss and Schuster, 1969; Jackson etal., 1981; Apelian et al., 1983; Herman, 1988). In one pass of the spray, a layer typically 100 jim thick may be deposited, and thicker deposits can be produced by continuing the deposition process. However, there is possibility of selfannealing and to minimize this effect a rotating water-cooled substrate may be advisable. A jet of inert gas can also be helpful in diverting the hot gases away from the substrate (Cahn, 1978). Metal droplet velocities can reach up to 1000 m s " 1 and cooling rates approximately 107 K s " 1 . All types of metastable effects achieved by
76
2 Rapid Solidification
rapid quenching from the melt have been obtained by flame or plasma spraying also (Krishnanand and Cahn, 1976; Giessen et al., 1977; Shingu et al., 1979; Miura et al., 1982). Further, oxides and ceramic materials can also be easily deposited by spraying. Nearly fully dense microstructures with much lower oxygen content than air plasma spraying have been produced using a low-pressure environment (Jackson et al., 1981). These showed substantial improvements in mechanical properties, but with limited ductility. Two major problems have been encountered in plasma spraying technology. First, the strength of the bond between the substrate and the deposit is not always acceptable in situations involving dissimilar materials. The problem is particularly acute when the coated part is subjected to thermal cycling, which rise to high stresses at the coating-substrate interface. Second, the mechanical properties of the as-deposited material are far from ideal owing to the presence of internal stresses and microporosity.
2.4 Chill Methods Rapid solidification is achieved in these methods by bringing the melt into contact with a chill substrate. The various methods involve injecting the melt into a die cavity, or forming the melt into a thin section by forging between a hammer or piston and anvil, or extruding the melt on to a chill surface or extraction of melt by contact with a rotating disc. These will be described below. 2.4.1 The Die Method These methods involve forcing a liquid metal into copper chill-mold cavities of
small cross section through vacuum (Groeber and Haneman, 1937), gravity (Serita et al., 1970) and pressure plus vacuum (Hinesly and Morris, 1970). A schematic representation of the process is shown in Fig. 2-14. Melt penetration is difficult in view of solidification at the entrance of the mold. However, the unique advantage of this method is that wires of predetermined cross-section (mostly circular) can be obtained. A tapering (wedge) mold has also been used to obtain controlled changes of section (and thus cooling rate) within a single sample (Falkenhagen and Hofmann, 1952; Esslinger, 1966; Burden and Jones, 1970 a; Ichikawa et al., 1971). The cooling rates measured are 104 K s " 1 at thickness of 0.7 mm and 107 K s~1 at 0.2 mm (Armstrong and Jones, 1979). 2.4.2 The Piston-and-Anvil Technique
Shortly before the Duwez gun technique was perfected, Salli (1958) (Miroshnichenko and Salli, 1959) proposed the method of thinning of a molten droplet by squeezing it between a stationary anvil and a moving piston. The melt was transferred to the anvil by means of a lever and spring device in a graphite crucible, allowing the melt droplets to be caught between a stationary anvil and a fast-moving piston. Several modifications to this basic technique have been tried out. Unlike in the "gun" technique, the foil produced in the piston-and-anvil apparatus has a uniform cross-section, is devoid of porosity and has a few cm2 in area. This technique has been used quite extensively for many scientific investigations. The crucible used for melting has been either graphite (sometimes a ceramic insert is introduced) or fused silica. The alloy can be melted by any conventional method - induction heating (Pietrokowsky, 1963),
2.4 Chill Methods
77
Vacuum outlet Insulated coolant container Coolant bath Copper mold Mold cavity
O-Ring Base plate Compression nut
Injection tube
Aluminum foil
levitation melting (Baker et al., 1969) (which also has the advantage of eliminating the crucible problems in the study of reactive alloys) or even resistance heating (Dixmier and Guinier, 1967; Laine et al., 1971). Gas flame (Pietrokowsky, 1963), electron beam (Galasso and Vaslett, 1966), plasma jet (Zboril and Posedel, 1970) and arc (Wang, 1970; Ohring and Haldipur, 1971) have also been used. The molten droplets can be released by switching off the r.f. generator (in levitation melting) or by applying a small gas pressure to the melt container. The falling droplet intercepts a beam of light between a source and a photocell and thus triggers the solenoid which in turn releases the piston, through an adjustable delay circuit. Although solenoids have been generally used to release the piston, hydraulic (Pietrokowsky, 1963; Dixmier and Guinier,
Figure 2-14. Schematic of the liquid quench injection mold used for rapid solidification studies.
1967; Baker et al., 1969; Blank, 1972), spring (Ohring and Haldipur, 1971) pendulum (Wang, 1970; Caryll and Ward, 1967), electromagnetic (Harbur et al., 1969) and magnetic yoke (Cahn et al., 1976) drives have also been employed. Other modifications have also been described in the literature (Booth and Charles, 1966; Beghi et al., 1969; Tonejc and Bonefacic, 1969; Bletry, 1970; Williams and Jones, 1974). Two moving pistons were employed in some cases instead of one stationary anvil and one moving piston. Fig. 2-15 shows the basic principle of the piston-and-anvil apparatus. The solidification rates achieved in this technique are typically about 10 5 Ks~ 1 . Even though cooling takes place from both the sides of the foil (and thus the cooling rate should have been higher), the thickness of the foil is generally large (up to
78
2 Rapid Solidification
tion. It was found that the higher the piston velocity and the higher the splat-substrate heat transfer coefficient, the faster are the attainable cooling rates, although there is a tendency to approach limiting values. 2.4.3 Twin-Roller Quenching
1. Fixed anvil 2. Fast-moving piston 3. Copper discs A. Silica tube
5. 6. 7. 8.
Susceptor Finger Light source Photo cell
Figure 2-15. Schematic of the piston-and-anvil apparatus.
300 |im) and therefore, the cooling rate is lower. The higher cooling rates of the gun technique and the uniform cross-section of the foils produced in the piston-and-anvil apparatus have been achieved simultaneously by introducing a pressure switch. The stream of molten metal ejected from the gun is caught between the piston-and-anvil assembly, by synchronizing the release of the piston with the rupture of the diaphragm (Ramachandrarao et al., 1970). The foils so produced are 6-7 cm in diameter and about 100 }im in thickness (Hanumantha Rao et al., 1985). Modeling of the piston-and-anvil quenching process has been done (Miyazawa and Szekely, 1979). Allowance has been made for fluid flow effects regarding the spreading of the slug, convective heat flow and the actual movement of the piston after the impact. In representing heat flow, allowance has been made for the two-dimensional transient temperature fields and the existence of a mushy zone on solidifica-
This process involves rolling of a melt stream between two conducting rolls rotating in opposite directions (Chen and Miller, 1970; Babic et al, 1970 a). The melt stream is directed vertically downwards into the nip of the rolls (Fig. 2-16). Short strips of about 10 to 200 |im thickness quenched at approximately 105 K s " 1 are generally produced. By controlling the parameters carefully, it is possible to produce very long filaments. Variables in the process include the material and diameter of the roller, the roller gap and the speed of rotation. The rolls should be hard enough for durability and the surface must be ma-
Heater Melt Rollers
Figure 2-16. Schematic of the twin-roll quenching apparatus.
2.4 Chill Methods
chined to close tolerances. They are generally made of hardened steel (Chen and Miller, 1970), chromium-plated steel (Babic etal., 1970 b), copper (Murty and Adler, 1982) or brass. The two rolls, 5-15 cm in diameter, are held together with a pressure ranging from 20 to 90 kg and so the ribbon is prone to deformation. This can be minimized by incorporating a silicon rubber damping pad (Lewis et al., 1979) between the core and periphery of the rollers or by spring loading (Murty and Adler, 1982) to allow the rollers to be run in mutual contact without application of excessive pressure. The rollers can be cooled by liquid nitrogen for continuous operation. The rollers can rotate at speeds ranging from 100 to 6000 rpm. Variations of this standard configuration have been reported. These help in producing splat flakes (Sankaran and Grant, 1980; Singer and Roche, 1980; Lakshmikumar et al., 1980) or continuous lengths of ribbon (Lewis et al., 1979; Murty and Adler, 1982; Fujiwara et al., 1982). The process has been carried out with the roller axes of rotation vertical instead of horizontal to permit lateral dispersal of furnace debris (Leontic et al., 1978) or excess liquid metal which otherwise accumulates between the rollers above the roll gap. A serious problem associated with this technique is the very short contact length between the melt and the rollers. It has been estimated that for a jet of 1 mm dia and a roller dia of 80 mm, the contact length is only 6.3 mm which, for a roll speed of 6000 rpm gives a contact time of about 0.25 ms (Davies, 1978). Because of this, the sample may still be hot and cools by radiation (at rates < 103 K s~x) (Davies et al., 1978), losing thereby the benefits of rapid quenching. Amorphous phases cannot be produced easily by this technique (Davies et al., 1978), although the very first
79
report (Chen and Miller, 1970) indicates that this technique is for the preparation of amorphous solids. However, thicker sections of non-equilibrium crystalline phases are easy to produce. This contact time can be increased by introducing a third roller (Budaira and Suito, 1979) on the exit side of the twin rollers or a flexible belt (Sakata and Ishibachi, 1982) running between the rollers. Although there is no limit on the quantity of melt that can be continuously rolled, normally only a small quantity is produced. Durand et al. (1976) reported a version designed to produce up to 4.5 kg lots per run via a tundish. For a successful operation of the twinroll quenching apparatus, very stringent control is necessary in comparison with single roller melt spinning. Surface cracking of the foils produced in the twin-roller apparatus appears to be a problem. Further, as in melt spinning and melt extraction methods, roller life is another problem. Process modeling of the twin-roller technique has been carried out (Miyazawa and Szekely, 1981; Murty and Adler, 1982) to optimize the parameters that yield the best results. It has been shown that the most important variables are (a) roller gap, (b) angular velocity of the rollers, (c) feed rate of the material, and (d) physical properties of the material. If all the parameters are considered, there exists only a narrow range of these parameters that gives a stable mode of operation. It was demonstrated that the diameter of the circular jet is a critical parameter governing to a great extent the nominal width and thickness of the resultant foil product. Further, to get a continuous foil, the ratio of the jet velocity to the roller velocity should be between 0.5 and 1.0.
80
2 Rapid Solidification
2.4.4 Melt Spinning Process
Melt spinning is the most commonly used method now-a-days to produce long and continuous ribbons. In fact, the development of this technique has been mainly responsible for the accelerated progress of rapid solidification technology since the 1970's. A wide variety of materials including steels, aluminum, copper, titanium-base alloys and superalloys have been successfully cast as filaments using this process. The melt spinning process derives its name from the fact that it involves the extrusion of molten metal to produce fine fibers in a way which is akin to that used for the manufacture of synthetic textile fibers. The melt spinning process can be broadly divided into two categories depending upon whether an orifice is involved or not. Each one of these processes, in turn can be further subdivided into two variants depending on whether the molten metal stream is solidified in flight or against a chill. These will be, in subsequent sections, referred to as free-flight melt spinning (FFMS) and chill-block melt spinning (CBMS), respectively (Hubert et ah, 1973). 2.4.4.1 Free-Flight Melt Spinning
This process (Pond, 1959, 1961) consists of creating and subsequently solidifying a stable liquid metal jet. Generally, the stable jet is created by ejecting the liquid through an orifice and the metal is then solidified by passage through a gaseous or liquid quenching medium. Fig. 2-17 a illustrates the principle. Owing to the relatively high surface tension and low viscosities of liquid metals, the small diameter jets typically used in the process will form droplets unless solidified and/or stabilized within a short distance from the orifice. Thus, the major question governing the successful performance of this technique is to deter-
mine the experimental conditions which will allow the fabrication of a continuous wire, i.e., to prevent the breakdown of the molten jet into droplets before it solidifies. Typically, the orifices used to create the liquid jets are circular with diameters ranging from about 50 to 1250 jim. The pressure required to eject the molten metal from the orifice increases as the orifice area decreases. This aspect, plus the inherent difficulties in keeping small orifices open and free streaming, sets the lower size limit for the orifices. The upper size limit is primarily set by the length of the column required to solidify large streams with most of the quenchants used. The orifices (nozzles) are made of a suitable refractory material, the exact nature of which depends on the type of metal to be processed. Successful long-term operation requires that the orifices (and crucible) be stable (i.e., non-reacting, non-eroding) with respect to the alloy being cast. For example, the best orifice material for aluminum and its alloys has been found to be alumina with a silica content of <0.5%. Table 2-2 presents a list of suitable orifice materials for different metals and alloys. The liquid metal jet is free-streamed into a chamber containing an atmosphere suitable to generate a very thin film around the
Table 2-2. Orifice materials for melt spinning. Metal
Orifice material
Aluminum Tin
Alumina Oxidized steel, pyrex glass, graphite Steel, pyrex glass, graphite, alumina Pyrex and vycor glass, mullite, alumina, SiC Armco iron Dense alumina, sapphire
Lead Zinc and alloys Magnesium Grey iron
2.4 Chill Methods Pressure
Pressure
— Molten metal
Molten metal
Liquid stream Orifice
81
Heat removal
Orifice
« — Liquid stream
Quenchant Liquid solid-Interface Fiber (wire) (a)
liquid jet, thereby preventing its break-up before it is fully solidified. A stopper rod or surface tension of the melt may be used to prevent pouring until melting is complete. The simplest technique uses ambient air, which is most efficiently introduced as a pressurized stream which accelerates the solidification process. Further, the high pressure, if it exceeds the partial pressure of the alloying elements, can prevent their vaporization. This was found to be the case in suppressing zinc evaporation during melt spinning of brass. Air-water fog quenchants have also been successfully used with cast iron. Liquid quenchants (Kavesh, 1978) and liquid nitrogen (Shepelskii and Zhilkin, 1969), inert gases (Mottern and Privott, 1973) and brine have also been used to both stabilize the molten metal jet and to accelerate its cooling rate. For a given diameter, experimental and theoretical considerations dictate a minimum jet speed which is required to obtain a continuous jet. For a given liquid and orifice diameter, the stable length of the jet increases with increasing velocity. Typically, a jet is ejected from the orifice with a velocity of 2 to 10ms" 1 , yielding filament production rates of about 1 to 10 g s " 1 per stream. Multiple orifices can be used to increase the productivity of a given stream. Available data suggest that for aluminum,
Fiber (ribbon) Chill block (b)
Figure 2-17. Schematic illustration of the (a) free-flight melt spinning and (b) chillblock melt spinning techniques.
the critical extrusion velocity decreases from 12 to 3 m s " 1 , as the orifice diameter increases from 50 to 300 jum. The temperature profiles in the solidifying liquid jet under FFMS have been analyzed (Ostroumov, 1959; Kavesh, 1978). As with all casting processes, the quench rates increase as the product dimension (in this case, diameter) decreases. Typically, the quench rate associated with 50 (im diameter metal jet is about 105 Ks~ x and decreases to 10 K s~1 when the jet diameter is 1250 jim. A cooling rate of about 104 K s~1 was claimed for 150 jim diameter glassy filaments of Fe 38 Ni 39 P 14 B 6 Al 3 (Kavesh, 1974). A detailed account of this process has been given by Pond et al. (1976). Both microcrystalline and amorphous filaments have been produced by this process. Microcrystalline wires generally exhibit a bamboo-type structure wherein a single grain extends across the entire crosssection of the wire. The filament surface exhibits a microscopically rough, dendritic-type pattern. The basic advantage of this technique, sometimes also known as melt extrusion, in addition to its obvious simplicity, is to allow the production of continuous filaments with a circular cross section, more or less directly usable in wire-type applications. Monsanto Corporation have devel-
82
2 Rapid Solidification
oped this technique for, among other applications, production of ferrous-base tire cord. 2.4.4.2 Chill-Block Melt Spinning (CBMS)
Originally patented by Strange and Pirn (1908), this process involves directing a molten metal jet onto a cold, moving heat sink where the jet is reshaped and solidified. The jet on impingement with the disk forms a melt puddle, of thickness approximately equal to and length about double that of the jet. As solidification begins, the ribbon is expelled from the surface of the wheel, as schematically illustrated in Fig. 2-17 b. Since the orifice can be placed very close to the chill block, and the jet streaming distance thereby minimized, jet stability is not quite as critical in CBMS as in FFMS. In the original version of Pond (1958), the chill block, rotating about a vertical axis, had a shallow concave impression. Long and continuous filaments with 1 to 100 |im thickness could be produced at speeds of 15 to 300 m s " 1 . Pond and Maddin (1969) changed the design by ejecting the melt through a fine orifice to impinge onto the inner surface of a rapidly rotating crucible. The orifice traverses rapidly parallel to the axis of rotation, thus producing a helix of rapidly solidified filament. The radial acceleration imparted by the crucible to the liquid results in efficient spreading of the stream and promotes good thermal contact between the melt and the heat sink. Following the successful application of this method to produce specimens suitable for mechanical testing (Masumoto and Maddin, 1971), this technique has been continuously used by Masumoto and his group in Japan. A crucible diameter of 100 mm and rotation speeds of about 5000 rpm yielded tapes of 20-40 \im thick-
ness, while thinner tapes could be produced at speeds of about 10000 rpm. Boswell and Chadwick (1976) used a cup of 380 mm diameter and the whole process was carried out in vacuum. It was claimed that the vacuum promoted formation of more uniform ribbons by minimizing aerodynamic disturbances. A further variant involves jetting on to the cylindrical surface of a disk or roll rotating about a horizontal axis (Liebermann and Graham, 1976). Glassy ribbons of about 10 to 40 |im thickness were produced in Fe-Ni-B alloys at speeds of 20 to 40 m s~*. Chen and Miller (1976), on the other hand, cast the jet on to the convex inner surface of a quench wheel rotating about a horizontal axis. This imparts radial acceleration to the ribbon which provides forced contact with the chill surface, but has only a limited contact time with the chill, allowing continuous operation. The process, however, suffers from the disadvantage that the ribbon solidifies with a curvature, reflecting the disk diameter and the angle of sloping surface. The cooling rates in all the above variants can reach up to 10 6 Ks~ 1 , thus making it possible to produce glassy phases in a large number of alloy systems. The most common method of CBMS, in which the molten metal stream impinges on the outer rim of a rotating disk, was developed by Bedell (1975). Here, the centrifugal force acts to throw the ribbon off the chill surface at rates of about 50 m s ~1 and by internally water cooling the disk, the process can be made continuous. Detailed accounts of the CBMS process have been given by several workers (Hilzinger and Hock, 1981; Jech et al., 1984; Liebermann, 1984 a). The following paragraphs deal with the methods of optimizing the several parameters involved in CBMS. Crucibles: Selection of crucible material is based on chemical compatibility with the
2.4 Chill Methods
melt, temperature capability, thermal shock resistance, low thermal conductivity and low porosity (gas permeability). Crucibles can be made by dipping a wax mandrel in commercial shell mold refractory slurry. Repeated dippings build the crucible wall to the desired thickness. Intricate nozzle geometries may be molded in wax and duplicated in the shell mold. Alumina shell mold crucibles are thermal shock-resistant, have low thermal conductivity and are compatible with most engineering alloys. The major disadvantage of the shell mold crucible is that it is not fully dense and therefore becomes permeable to pressurizing gases. Dense alumina is commercially available and nozzles can be shaped by ultrasonic abrasive machining. Fine holes with circular, clean and sharp edges can be produced. These dense alumina crucibles are gas tight and are usable up to temperatures in excess of 1600°C. The major disadvantage is susceptibility to cracking when internally heated at a high rate. This can be minimized by using a susceptor to heat the crucible externally or by jacketing the crucible to reduce thermal gradient. Quartz crucibles are useful for temperatures below 1400°C. They are thermally shock-resistant, relatively inexpensive and can be readily purchased with nozzles in place. Temperature limitations and possible reactivity with the melt are the major disadvantages. Attempts have been made in recent times to quench reactive melts, such as those based on titanium, molybdenum, chromium, and niobium. The system devised by Whang and Giessen (1983) (Whang, 1984) is similar to the one used to melt spin lanthanum-based alloys (Tutzauer et al., 1980). In this apparatus, the conventional arrangement of r.f. melting and pressurized expulsion from a quartz nozzle was re-
83
placed by arc melting the alloy sample on the orifice of a water-cooled copper anode. Expulsion of the melt on to the rotating chill block was achieved with applied pressure differences, preferably exceeding about 0.5 bar. The newer version incorporates an arrangement to feed, automatically, solid granules of alloy into the melt zone in order to process larger quantities continuously into ribbon or flake, depending on the conditions. Masumoto et al. (1980) used levitation melting to melt niobium, chromium and titanium-base alloys in a chamber filled with argon. When melting was complete, the alloy was ejected by switching off the current to the levitation coils and allowing the melt to drop because of the difference in pressure between this chamber and the one containing the wheel (Fig. 2-18). The ejection of the molten metal was achieved in some cases by melting a fuse plug (usually Mo over Ti) over the nozzle (Rowe and Amato, 1987). Other designs have also been reported (Ray and Clemm, 1986). Wheels: Wheels for melt spinning have been made from a variety of materials, although copper is the most popular one. Primarily, the aim is to select a wheel material which will extract heat from the ribbon as quickly as possible, while allowing the puddle to wet the wheel and form the ribbon. Cooling of the wheel is desirable for long runs such as would be necessary in commercial operations. Wheel surface texture and cleanliness influence both product quality and form (Huang and Fiedler, 1981). The wheel side of vacuum cast ribbon is almost an exact replica of the wheel surface. Progressive wear of the chill surface is a serious problem, but can be counteracted by refinishing the wheel during every turn. The casting equipment may also be designed to delay the wear of the wheel, by carrying out the
84
2 Rapid Solidification Specimen chamber
Argon gas inlet
Stainless steel vessel
Molten alloy (Cooling water)
Copper levitation coil
— Supporting stand of specimen
Shutter Quartz nozzle
Argon gas inlet
—
casting in such a way as to minimize the ratio of product tape area to the chill surface area. This can be done by employing a larger wheel diameter, or by continuously displacing the crucible-nozzle unit along the vertex of the cylindrical quenching body during the operation so that the worn track spirals along the wheel. Both these measures offer the additional benefit of reduced surface temperature. The hardness of the substrate is found to be one of the factors that influences the wear resistance because a roll made of agehardened Cu-2% Be alloy showed comparatively little wear even after much longer service than pure copper (Hilzinger and Hock, 1981). Moreover, all tapes cast on copper show perceptible traces of copper on their contact side, which is entirely absent in the tape cast on the Cu-Be wheel. Other wheel materials used to date are steel (Chen and Miller, 1970; Hubert
Stainless steel vessel
Figure 2-18. Schematic illustration of the melt-spinning apparatus used by Masumoto and his group in Japan.
et al., 1973; Mitera et al., 1979; Jech et al., 1984), chromium (Narasimhan, 1979) and molybdenum (Ray, 1979). Chamber Atmosphere: Spinning can be carried out in vacuum, air, inert atmosphere, or reactive gas depending upon the chemical and physical properties of the charge. Alloys susceptible to oxidation can be cast in vacuum or inert gas. Chamber atmosphere influences ribbon quality with respect to surface smoothness and edge definition. Many alloys cast in vacuum have regular, smooth edges and wheel-side surfaces which are almost exact replicas of the wheel surface. The same alloys cast in helium may have ribbon edges with a rough 'saw-tooth' appearance and wheelside surfaces with pits and irregularities. Casting in argon aggravates the condition (Jechetal., 1984). Ejection Pressure: Ejection of the melt from the crucible is accomplished by gas
2.4 Chill Methods
pressurization. Although an inert gas is generally used, any gas compatible with the melt can be used. Ejection pressures of 5-70 kPa have been used depending upon the desired melt delivery rate. The use of high ejection pressure results in improvement of the wetting pattern and hence better thermal contact between the melt puddle and the rotating substrate (Huang, 1982). Wheel Speed: Increasing the wheel speed leads to the formation of thinner ribbons. For example, an Fe 4 0 Ni 4 0 B 2 0 alloy cast on a 250 mm diameter copper wheel rotating at a substrate velocity of 26.6 ms""1 produced a ribbon of 37 jim thickness, while at a velocity of 46.5 m s " 1 , the ribbon thickness was only 22|im (Liebermann, 1980 a). The time of contact of the solidifying material on the chill block is of decisive importance in the fabrication of amorphous metallic filaments. If the filament detaches from the disk too soon, crystallization and phase decomposition may occur during secondary cooling from the melt in the solid state. In extended operations, this can be avoided by deliberately increasing the filament contact time by employing a spring-loaded auxiliary disk in contact with the main melt-spinning disk (Bedell, 1975). The nozzle tip is usually about 3 mm from the wheel surface although much shorter gaps (about 0.35 mm) have also been used. The optimum direction of melt jet impingement has been found to be 6 to 15° from the local substrate normal at the point of jet impingement. Modeling of the CBMS process has been done both empirically and theoretically (Kavesh, 1978; Pavuna, 1981; Huang, 1982; Katgermann and Van den Brink, 1982; Takeshita and Shingu, 1983,1986). Empirical relationships have been established, with little scatter, for the ribbon geometry
85
as a function of melt ejection pressure, temperature, flow rate, wheel velocity, atmosphere, etc. In all the models described in the literature, the ribbon is visualized as having formed by transport of momentum and/or heat between the melt puddle and the chill surface. Two limiting cases and a mixed case can be distinguished: (a) thermal transport control, (b) momentum transport control, and (c) coupled thermal and hydrodynamic regime. Further, the liquid puddle is visualized as consisting of a "boundary layer" zone in which the thermal and momentum effects of the chill surface are strongly felt and an outer zone where the effects of the chill surface have not yet penetrated. If thermal transport is much faster than momentum transport, a solid boundary layer will form adjacent to the chill surface and propagate into the melt puddle to form the ribbon. In this case, material within the "frozen" layer will be in motion at a velocity equal to that of the chill surface. A sharp transition in velocity occurs just outside of the frozen layer; the width of this transition zone depends on the viscosity-temperature characteristic of the alloy. In the outermost zone, the melt will experience cooling but only slight momentum transport (see Figure 3 of Kavesh, 1978). On the other hand, if momentum transport from the chill surface is much faster than thermal transport, a liquid boundary layer will be dragged out of the melt puddle by the moving substrate to solidify farther downstream. In this case, the velocity gradient will be continuous across the depth of the puddle. Determination of which of the limiting cases is likely to exist is based on the transport properties of the melt. Although till recently the thermal transport mechanism has been considered to be of importance
86
2 Rapid Solidification
because of the small value of the Prandtl number of liquid metals (which means the large thermal boundary layer thickness in comparison with the momentum boundary layer thickness), it is now recognized that a combined thermal-momentum transport model explains the experimental observations more satisfactorily. Although these approaches have partial success in describing the experimental results, the complex nature of the ribbon formation process and the factors which control it are still to be described by a physically consistent model. The solidification rates achieved in this process are typically about 1 0 6 K s " 1 (Shingu and Ozaki, 1975; Davies and Lewis, 1976) and the ribbon is only a few mm wide. Pavuna (1981, 1982) has suggested methods by which good quality tapes can be produced. CBMS has been modified to produce helical glassy alloy ribbons (Liebermann, 1981), composite alloys and multilayer deposits (Tenwick and Davies, 1984; Soderhjelm and Mandal, 1985; Liebermann, 1985).
ing rates. Thus, the centrifuge melt spinning technique has been developed, the merits and limitations of which have been summarized very recently (Baram, 1990). In this technique (Rosen et al., 1986 a, 1987), a rotating disk holds the rotating crucible. The melt is ejected from the crucible by centrifugal force, through an orifice in the crucible wall, and impinges on the inner surface of the copper rim, which serves as the quenching substrate and rotates in a direction opposite to the disk holding the crucible. The rotational velocities of both the disk and the rim can be adjusted, as also the crucible orifice diameter and the distance from the rim. In contrast to conventional CBMS, centrifuge melt spinning employs higher metal ejection pressures (up to 280 kPa) together with higher extraction velocities (up to 95 ms" 1 ) (Rosen etal., 1986b). Parametrization (Rosen etal., 1986a) and mathematical modeling studies (Baram, 1988 b, 1990) have shown that this new technique produces ribbons solidified at rates of 108 K s~1 and with better dimensional uniformity and improved wetting pattern than in conventional melt spinning methods.
2.4.4.3 Centrifuge Melt Spinning
All the variants of the chill block melt spinning techniques described above suffer from some inherent problems. Some of these are (i) short contact times between the wheel and the melt leading to poor heat transfer, (ii) restricted wetting of the casting rim by the liquid metal (at low ejection pressures) due to the formation of an air boundary layer (Baram, 1988 a) and (iii) "catastrophic sticking" (Liebermann, 1984 b) at high substrate velocities. Although these can be overcome by increasing the ejection pressure (Huang, 1982), this leads to an increased ribbon thickness (Kavesh, 1978; Liebermann, 1981) and hence lower cool-
2.4.4.4 Planar Flow Casting Process
The continuous production of tapes by the chill block melt spinning (CBMS) process has been widely used. However, since a circular orifice is used, the tapes so produced have a width not much greater than the jet diameter and necessarily there is a limit up to which the width can be increased. Accordingly, the width of the tapes produced by the CBMS process is limited to a few mm (typically about 5 mm). To produce wider ribbons of uniform cross-section, a rectangular melt puddle is required and that cannot be produced by a cylindrical jet (Anthony and Cline, 1978,
2.4 Chill Methods
1979). Further, rectangular jets degenerate rapidly in a complicated fashion (Cline and Anthony, 1979), owing to the low viscosities and high surface tensions of metallic melts. This problem was solved by Narasimhan (1979), who invented the Planar Flow Casting (PFC) Process. In this process, illustrated in Fig. 2-19, the molten metal is forced through a slotted nozzle in close proximity (about 0.5 mm) to a rotating chill substrate. The melt puddle is constrained to a stable, rectangular shape by the lips of the nozzle and the substrate. Flow is basically pressure controlled, but it is also critically dependent on the substrate surface speed, nozzle width (parallel to the direction of ribbon motion) and nozzlesubstrate gap. An important feature of the PFC process is that it is intrinsically scalable. It has been reported that ribbons up to 300 mm in width have been produced by this process (Heineman, 1985), which can also be used in an inert atmosphere or vacuum. The tape comes away from the roller at a speed of 900-1800 m min" 1 and so winding of this into a coil could be a problem. A system has now been devised (Narasimhan et al, 1981) to capture the ribbon in a winder in a matter of seconds after the initiation of a cast. PFC has a few distinct advantages over the CBMS process. Firstly, moving the nozzle closer to the substrate increases uniformity of quench rate and secondly, the melt puddle is stabilized and suffers less perturbation. The thickness of the tapes varies from 20 to 100 ^m and cooling rates achieved are about 10 6 Ks~ 1 . However, care must be exercised to maintain a fixed nozzle-substrate gap because it significantly affects ribbon dimensions, smoothness and quench rate. Some improvements have been effected to the basic PFC process. For example,
87
Nozzle wall
Substrate
Substrate motion
Figure 2-19. Schematic of the planar flow casting process.
preparation of wide tapes with teeth or other geometries have been made without the need to punch the teeth out with a die. The teeth were formed by casting a low-thermal-conductivity pattern onto the wheel, causing the ribbon to be brittle and easily separated from the very tough nonbrittle remainder of the ribbon (Luborsky et al., 1983). Another novel variation is the preparation of helical rather than straight ribbons (Liebermann, 1981). The helical ribbon can be wound without producing the degradation in properties associated with the bending stresses produced on winding up straight ribbon into a core. By combining this helical-ribbon technique with the technique for producing a cut-out pattern, electric motor stators can be prepared directly from the cast ribbon without the need for punching. Gutierrez and Szekely (1986) analyzed the PFC process using a mathematical model based on the principles of lubrication theory, capillary fluid dynamics and solidification heat transfer. The model suggests that the most important parameters of the PFC process are (a) the melt flow rate, (b) the wheel velocity, (c) the nature of the mechanical and thermal contact at the splat-wheel interface, (d) the geometry of the system and (e) the material properties. It has been shown that satisfactory performance of the PFC process depends on a delicate balance between capillary pro-
88
2 Rapid Solidification
cesses in the meniscus and solidification phenomena underneath the puddle. It should be noted that perturbation of the melt on the wheel by the gas boundary layer (Huang and Fiedler, 1981) and surface roughness of the wheel etc. are substantially reduced because of the constraint of the melt imposed by crucible. It has been calculated that a momentum, rather than a thermal, boundary layer controls the ribbon dimensions (Vincent et al., 1982). 2.4.5 In-Rotating Water Spinning (INROWASP)
This is a modification of the melt spinning technique in that instead of allowing the stream to impinge on to the interior of a rotating drum, the melt stream is ejected into rotating water. In this process (Ohnaka, 1985; Masumoto et al., 1981; Ohnaka et al., 1985), a jet of molten metal is ejected through a quartz nozzle (80 to 200 |am diameter) into a liquid cooling layer (usually water) formed by centrifugal force on the inner surface of a rotating drum of 400 to 600 mm dia. The speeds of the coolant and of the melt jet are controlled by the rotation of the drum and by the ejection gas pressure, respectively. In this process, which is somewhat similar to the Rapid Spinning Cup process described earlier for the production of powder (see Section 2.3.4.5), a wire can be formed by making sure that the jet solidifies before being subjected to disturbance. The distance between the nozzle tip and the coolant surface and ejection angle, the depth of the coolant layer and the coolant temperature are some of the important variables which need to be controlled and varied depending on the alloy being cast. The product of this technique is a continuous wire of round cross section. The
dendritic structure in these wires is often along the direction of the wire unlike in melt-spun ribbons where they tend to grow transverse to the casting direction. The cooling rates achieved are reported to be about l O ^ s " 1 (Ohnaka et al., 1985). A wide variety of ferrous and non-ferrous alloys have been cast into amorphous or microcrystalline wires. This process has the advantage of producing wires of even those alloys which are difficult to draw by the conventional methods. The wires produced by this process have been extensively characterized for their microstructural, mechanical and magnetic properties by many Japanese teams. 2.4.6 Taylor Wire Process
This technique can be used to produce fine wires of uniform cross section (Taylor, 1924). In this process, a metallic rod contained in a glass tube is melted. The glass is softened due to contact with the molten metal and it can then be drawn, while acting as a continuous mold crucible during the solidification of the metal, ensuring regular surface and diameter of the wire (Fig. 2-20). The composite product obtained, consisting of a metallic wire in a glass sheath, is collected on a rotating drum at speeds of about 5 m s~1. A critical aspect of this technique concerns the availability of sheath material having a sufficient chemical inertness towards the molten metal used, as well as a softening temperature consistent with the melting point of that metal. In addition, the necessity of removing the sheath by a relatively simple treatment might place further limitations upon the choice of the glass material to be used. Cooling in this process is predominantly by convection and the rates are in the
2.4 Chill Methods
89
2.4.7 Melt Extraction Process Metal rod Glass tube Heating coil
Tension Fiber and sheath Figure 2-20. Schematic illustration of the Taylor wire process.
range of 104 to 106 K s " 1 for wires of 50 down to 2 jim diameter (Bunge, 1976). Fibers from a wide range of metals have been produced in this way including steels, copper and noble metals as well as lowmelting metals (Manfre et al., 1974; Bunge, 1976; Pardoe et al, 1978). The production, properties and uses of "microwires" made by this process have been reviewed by Donald (1987). A modified version of this technique has been used by Goto and coworkers for a variety of alloys. A major problem in this technique is the contamination of the material by the glass sheath. This can be avoided by choosing a glass compatible with the material in terms of chemistry, viscosity and melting temperature. Microstructure, mechanical and superconducting properties of a variety of microcrystalline and amorphous alloys (copper, steels, stainless steel, iron, silver, gold and silvercopper alloys, Pd-Si alloys, F e - P - C - B alloys) have been evaluated (Goto et al, 1974, 1977).
As mentioned earlier, the melt spinning methods can produce fibers and ribbons continuously. However, the material compatibility problems (of melt with crucibles, orifice, or stream stabilizing sheaths) increase enormously as the melt temperature and reactivity increase. The melt extraction methods overcome some of these problems (Pond, 1959; Pond et al, 1976; Maringer and Mobley, 1974, 1978 a, b; Maringer et al, 1976; Collings et al, 1978; Robertson et al, 1978). Melt extraction differs substantially from melt spinning in that no orifice and no liquid jet is involved. It is a process wherein the periphery of an appropriatelyshaped, rotating heat-extracting disk contacts a source of molten metal. The metal solidifies over the area of contact and adheres to the disk for a short residence time (of the order of milliseconds), and then it is spontaneously released and thrown free in the form of a filament or staple fiber. There are several variants of the process. For example, if the melt is contained in a crucible, as schematically shown in Fig. 2-21 a, the process is called Crucible Melt Extraction (CME). On the other hand, if the crucible is eliminated by melting the end of a rod to produce a molten pendant drop, shown in Fig. 2-21 b, it is called Pendant Drop Melt Extraction (PDME). 2.4.7.1 Crucible Melt Extraction (CME) Crucible Melt Extraction is a process wherein the periphery of a rotating heat extraction disk is brought into contact with the surface of a source of molten metal (Fig. 2-21 a). The disk is normally water cooled and is rotated about an axis perpendicular to the plane of the disk and parallel to the plane of the molten surface. The melt surface is contacted by the disk either by
90
2 Rapid Solidification
Clean, quiet, flat melt surface
, Wiper
Pendant drop
Melt stock
Rotating extracting disc Filament
Heat source Filament
Induction furnace
(a)
Rotating extracting disc
Wiper
(b)
Figure 2-21. Sketches illustrating the principles of (a) crucible melt extraction and (b) pendant drop melt extraction.
lowering the disk or raising the crucible. As the periphery of the disk passes through the molten metal, the metal solidifies onto and adheres to the disk. This solidified metal is extracted from the melt, still clinging to the disk periphery, and continues to cool. As a result of thermal contraction and centrifugal force, the metal in the form of a fiber, is then spontaneously thrown free off the disk after a short residence time. Materials with high surface tensions (e.g. steels) give coarse fibers and large particulates, while finer fibers/particulates can be produced in alloy systems with low surface tensions (e.g. lead-base alloys). The solidification rate for stainless steel has been measured as about 5 x 104 K s " 1 . The important elements of the process are (i) the melting system, (ii) the disk drive, (iii) the disk geometry and properties, (iv) the feed system and (v) the surface chemistry of the melt. The disk of about 20 cm diameter is rotated smoothly at controlled speeds ranging from 200 to 2000 rpm. The fiber is thus produced at rates of 1.5 to 15ms" 1 . The disk has to be water cooled to extract the heat from the solidifying fibers and also from the bath radiation.
The disk has a wedge type periphery. The wedge angle is 90° and the periphery is normally slightly rounded. For the same area of contact between the melt and the disk, greater speed limits the time during which solidification can occur and, therefore, limits the thickness of the product. A decrease in the depth of penetration of the disk into the melt will lead to the production of a smaller (and rounder) product. The materials used for the disk include aluminum, copper, brass, various steels, nickel and molybdenum. Copper has been the most popular material in view of its high thermal conductivity, easy machinability, durability and relative cheapness. Fiber shapes can be controlled by disk geometry as well as by the extraction parameters of disk speed and depth of immersion. Fiber in the shape of ribbon, or with C-, D-, and L-shaped cross sections can be cast simply by altering the shape of the disk periphery. Staple fibers of any particular length can be cast by putting suitable indentations in the periphery at desired spacings. The source of molten metal should have a clean, stable surface to have uniform rapid solidification processing and for uni-
2.4 Chill Methods
form product dimensions. It is essential to use a wiper to remove the metal debris from the disk edge. 2.4.7.2 Pendant Drop Melt Extraction (PDME) Pendant Drop Melt Extraction is very similar to CME except for the way in which the molten metal is supplied to the disk periphery. Here, the end of a rod of material 6 to 25 mm in diameter is melted to form a pendant drop, as shown schematically in Fig. 2-21 b. When the droplet contacts the edge of the heat extracting disk, the filament is formed in the same way as in CME. Once fiber production has begun, the intensity of heating and the feed rate are adjusted to stabilize a continuous casting process. By adapting a carousel to the system (Stewart et al., 1974), the filament can be collected as a continuous coil. Like in CME, by notching the disk, staple fiber can be produced. Such fibers can be made of almost any length and diameter (L/D) ratio. These are know as L/D powders. If L/D is 10 : 1 or less, the particles flow freely and may be handled as powder. The greatest single advantage of PDME is the elimination of the crucible and thus it is possible to produce fibers of high-melting-point and highly reactive materials with little difficulty. Further, green powder compacts, either pre-alloyed or mixed as elemental powders, may also be used as starting materials. A great variety of metals and alloys, including titanium alloys and superalloys, have been converted to fiber by this method. The effective diameters range from 15 to 1500 jim. Quench rates associated with the melt extraction process depend strongly on the thickness of the cast filament. For steel filaments of 300 jiim effective diameter, the quench rate is 102 to 103 K s ~ \ while
91
for filaments having effective diameters below 25 (im, the quench rate can exceed 1 0 6 K s " 1 (Maringer and Mobley, 1979). (The effective diameter is the diameter of a circle of the same cross-sectional area as the irregularly shaped filament.) The major disadvantage of PDME is that it is unsuitable for bulk production since the process relies heavily on the equilibrium droplet size governed by surface tension forces. Although CME and PDME share the ability to produce fibers and filaments of a wide variety of metals and alloys, there are some recognizable differences. CME lends itself to the casting of heavier cross-sections because the massive melt supply enables one to exert considerable control over the area of contact between the melt and the disk surface. CME also lends itself to the use of multiple edges, since the extent of the melt surface can be easily expanded through the use of larger crucibles. PDME, on the other hand, easily produces fine fiber, is readily adaptable to vacuum or inert gas enclosed systems and is ideally suited to produce small quantities of reactive materials for which suitable crucibles are not readily available. However, it is very difficult to get a suitable droplet from alloys with a large liquidussolidus range, or with a miscibility gap in the liquid phase. It is also difficult to maintain a pendant drop with materials such as silver or copper, which have exceptionally high thermal conductivities. The interrelationship between various process variables and product dimensions has been evaluated largely on a predictive/ theoretical basis (Maringer and Mobley, 1978 b). Maringer et al. (1978) produced L/D types of powders of Ti-6 A1-4V by PDME and subjected them to vacuum hot pressing or hot-isostatic pressing to produce
92
2 Rapid Solidification
sound billets with acceptable mechanical properties. Thus, melt extracted staple fibers may provide an alternative to the types of metal or alloy powder currently in use. CME has been brought into commercial production. Some hundreds of tons of CME mild and stainless steels are used every year for reinforcing refractories and concrete. 2.4.8 Melt Drag Process The Melt Drag process takes its name from its main feature, viz., the molten metal is dragged from an orifice onto a cooled rotating cylinder, as shown schematically in Fig. 2-22. The process involves bringing the meniscus of liquid metal protruding from a slot nozzle into contact with the periphery of a rotating substrate. The meniscus is partially solidified by contact with the rotating drum which rapidly drags away the solidifying metal to form a continuous wire or ribbon (King, 1967; Hubert et al, 1973). After the solidification is completed, the final product is removed from the drum after about one-third of its circumference, and then coiled.
This method has several technological advantages over classical continuous casting. This requires no fabricated, lubricated or reciprocating molds nor any complex roller guide system for withdrawing the cast product. The width of the product is equal to the width of the slot and strips exceeding 25 cm in width have been cast. In this sense, this process is similar to the PFC process. However, in the latter, the thickness is somewhat lower. The thickness in the melt drag process can vary from 25 to 1000 jim, with the foils cooling at rates from 106 to 103 K s " 1 . Another important difference between the PFC and melt-drag processes is that while the melt is pressurefed to the nozzle in PFC, it is only gravity fed in the melt drag process. The dimensions of the wire or tape are determined in this process essentially by (a) circumferential speed of the drum, (b) viscosity of the molten metal, and (c) surface tension of the molten metal with respect to the drum, which in turn depends on the nature of protective atmosphere, if any, and the superheat of the molten metal. This process has been used to cast aluminum, steel, armco iron, solders and
Crucible Heating coil
Water-cooled rotating drum
Fiber
Figure 2-22. Sketch illustrating the melt drag process.
2.5 Laser Surface Treatment
others. It allows for a simple and inexpensive technology and is particularly suited to the requirements of relatively small scale production. 2.4.9 Melt Overflow Process
In this process, the molten material is allowed to flow over a lip in the side of a furnace on to a rotating water-cooled wheel. The melt is not constrained on the upper surface as in the orifice methods of planar flow casting and melt drag. Fig. 2-23 a shows a schematic of the technique. The nature of the shear zone (shown shaded in Fig. 2-23 b) is governed by the melt properties (viscosity, surface tension) and wheel speed. The rate of material rela) Filament
93
moval is primarily controlled by the melt flow rate which in turn is normally adjusted by displacing the liquid metal in the crucible behind the overflow lip. The fine fibers that are produced can have variable lengths by having indentations at different positions in the wheel. The fibers produced in the Melt Overflow process are apt to have a variable crosssection and have been described as having a sector shape. This results in a flat heat extraction surface and this, coupled with a longer dwell time on the wheel, leads to cooling rates of > 1 0 6 K s " 1 . The overall fiber thickness is between 20 and 25 jim. Details of this process can be found in the literature (Gaspar et al., 1986; Boulby and Wood, 1986; Wood and Boulby, 1986). 2.4.10 Comparison of Rapid Solidification Techniques
Table 2-3 presents a summary of the various rapid solidification techniques and their product parameters (Savage and Froes, 1984).
2.5 Laser Surface Treatment (b)
Figure 2-23. (a) Schematic illustration of the melt overflow technique, (b) Shows details of the melt pool shape and nature of the shear region at the overflow lip-wheel contact point.
All the techniques described in the preceding sections involve complete melting of the alloy prior to rapid solidification. An alternative is to locally melt the surface (the depth ranging from 10 to 1000 \im) of bulk material followed by rapid solidification and subsequent solid-state cooling (Breinan and Kear, 1976; Breinan et al., 1976). This process has also been referred to as self quenching, laser annealing, laser glazing, layer glazing, etc. However, in order to avoid possible confusion with other similar terms used in metallurgy, the term laser surface treatment has been suggested by Cahn (1983) (see also Chapter 3).
94
2 Rapid Solidification
Table 2-3. Rapid solidification techniques and product parameters. Technique
Product tpye
Typical dimensions (um)
Typical cooling
Comments
Gas atomization
Smooth, spherical powder
50-150 dia.
10 2 -10 3
Water atomization
Rough, irregular powder Smooth, spherical powder Smooth, spherical powder Elongated splats
75-200 dia.
10 2 -10 4
Inert gas (nitrogen, argon) used where oxidation is a problem Used on tonnage scale
10-50 dia.
up to 106
25-80 dia.
10 5
40-100 thick
10 4 -10 7
125-200 dia.
10 2
< 50 dia., possible 1000-5000 long x 1000 dia. 20-100 dia.
10 4 -10 6
Powder, flake or coating Spherical or irregular Variable, powder to flake Foils Thick deposit
0.01-100 dia.
up to 107
0.5-30 dia.
10 5 -10 6
200 thick flake
10 5 -10 6
0.1-10 thick > 1 mm thick
106_108
Plasma spray deposition Chill methods
Coherent deposit
> 1 mm thick
<10 7
Wedge or cylinder
10 4
Piston-and-anvil
Splat
Free-flight melt spinning Chill-block melt spinning Planar flow casting/ melt drag In-rotating water spinning Taylor wire
Circular section filament Rectangular section tape Wide rectangular section tape Circular section wire Circular section wire Filament of fiber
200-1000 thick possible 50 mm dia. x 5-300 thick 100-200 dia. wire 10-100 thick up to 3 mm wide 20-100 thick up to 300 mm wide 60-200 dia.
Ultrasonic gas atomization Centrifugal atomization Electron beam splat quenching Rotating electrode process Rapid spinning cup Perforated rotating
Smooth, spherical powder Variable, spherical to irregular Acicular granules
cup
Soluble gas atomization Electrohydrodynamic atomization Spark erosion Twin roll atomization "Gun" technique Spray deposition
Crucible melt extraction/pendant drop melt extraction Melt overflow
Spherical powder
Sector shape fiber
rateCKs"1)
10 1
lO1-^2
10 3 -10 6
10 4 -10 6
Narrow size distribution, high efficiency Low oxide, clean powder Applicable to reactive metals Low cooling rate, low contamination Narrow size distribution possible Very economical Very clean, satellite-free powder Excellent powder size control Little control of powder size Economical Widely applicable Short process route to thick section Has potential to produce near-net shape Widely applicable
10 2 -10 4
Widely used laboratory technique Restricted application
10 5 -10 7
Very widely used
10 5 -10 6
Widely applicable
10 5 -10 6
Widely applicable
20-100 dia.
10 3 -10 6
20-100 thick
10 5 -10 6
High risk of contamination of melt Applicable to reactive melts
20-25 thick
>10 6
Widely applicable
2.5 Laser Surface Treatment
The process of laser surface treatment is accomplished by concentrating on alloy surfaces extremely high power densities of about 10 6 Wcm~ 2 in a small spot, typically 0.1 to 1.0 mm diameter for short durations of about 10 5 s. Although lasers have been very commonly used, one can also use electron beams (Kadalbal et al, 1980; Lewis et al., 1980; Tucker and Ayers, 1980) to achieve similar results. Two different types of lasers have been used - pulsed ruby, neodymium glass or other solid-state lasers and continuouswave (CW)-gas (CO2). The former are used in stationary mode, whereas the latter are used in conjunction with a rapid scanning of the alloy surface perpendicularly to the beam. If such scanning is combined with slow transverse translation of the laser beam, then successive strips of melted material can cover the whole surface (Breinan and Kear, 1978). The melted layer solidifies immediately after the heat source moves on or after it is switched off because the liquid layer is in good thermal contact with the bulk of the material. Cooling rates of 1 0 6 - 1 0 8 K s " 1 have been reported (Tuli et al., 1978; Strutt et al, 1980; Kear et al, 1981) for these processes. Early studies on laser surface treatment were concerned with the production of metastable phases such as extension of solid solubility limits (Elliott et al, 1973) and formation of metastable crystalline phases (Laridjani et al, 1972). Later, it was also possible to produce quasicrystalline phases (Schaefer et al, 1986) and amorphous phases in "easy" glass-forming systems (Sepold and Becker, 1986; Lin and Spaepen, 1986). There are, however, two possible effects which may prevent the formation of a glassy phase in many systems. Firstly, the underlying crystalline material being in contact with the melt serves as a nucleation site for the crystalline phases
95
which may grow so rapidly that they disrupt the amorphous layer. Secondly, when vitrifying a large surface area with overlapping passes, recrystallization of the amorphous material starts in the heat-affected zones (Sepold and Becker, 1986). An important difference between rapid solidification from the melt and laser surface treatment is that in the latter technique, it has been possible to produce amorphous phases in metal-metalloid systems even at as low a concentration as 5 at.% of the metalloid (Lin and Spaepen, 1983). Practical applications of laser surface melting (Steen, 1985) include: (a) Surface remelting to reconstitute, homogenize, or refine the surface to achieve special properties at the surface, (b) Surface alloying in which preplaced or injected alloying additions are incorporated into the surface (Draper and Poate, 1985), (c) Particle injection in which a dispersion of second phase is incorporated, without being remelted, into the remelted zone at the surface (Ayers etal., 1980; Cooper and Ayers, 1985). (d) Surface cladding in which a metal or alloy is fused on to the underlying material with a minimum of dilution by the underlying material (Powell and Steen, 1982), and (e) Laser layer technique for joining amorphous materials to produce thick foils (Sepold and Becker, 1986). Alternate elemental layers (compositionally modulated films) totalling less than 1 \\m in thickness have been remelted and solidified on a surface to produce very high cooling rates of 10 1 0 -10 1 3 K s " 1 (von Allmen et al, 1984). This resulted in the formation of glassy phases in much wider composition ranges than had been possible with rapid solidification processing (Lin
96
2 Rapid Solidification
and Spaepen, 1983; von Allmen, 1983; von Allmen et al., 1984). Research in the area of laser surface treatment is prompted mainly by two considerations, viz., (i) it requires possibly expensive alloying elements to be present only in a very thin surface layer which can be easily achieved by laser surface treatment, and (ii) it offers a way of working with bulk materials, which is not possible with the usual rapid solidification techniques. Laser surface treatment has been found to increase fatigue, erosion and wear resistance. This process can also be used for coating a cheap base material with a more expensive, harder or corrosion-resistant surface. The technique of electron beam melting and solidification has been elegantly employed to study the basic principles of solidification structures and conditions for solid solubility extensions (Bendersky and Boettinger, 1981; Boettinger et al., 1984) and formation of quasicrystalline phases (Schaefer et al., 1986). (A separate chapter on Laser Surface Modification by B. L. Mordike follows the present chapter in this book, and may be consulted for further details.)
2.6 Cooling Rates in Rapid Solidification The rate of cooling of a metallic melt before, during, and after solidification can, in principle, be calculated and also measured experimentally. A knowledge of such cooling rates helps in the first instance in evaluating the relative efficiencies of different quenching techniques. An understanding of the variables affecting the rates of cooling enables one to optimize the conditions to achieve the maximum possible
cooling rates. Further, the magnitude of the cooling rate and its variation aids in interpreting the microstructures obtained. It is, therefore, not surprising that several attempts have been made to estimate, measure or calculate the cooling rates during rapid solidification in the laboratory or the industry. In most rapid solidification processes, the thin dimension of the product is typically ^100 |im, with a corresponding cooling rate of ^ 1 0 5 K s " 1 . Thus, the solidification and cooling are complete in a matter of a few milliseconds. Actual recording of temperature variations in such a short interval of time naturally poses serious experimental difficulties. All the same, some measurements have been carried out. On the other hand, microstructural features of the solidification product such as secondary dendrite arm spacing or eutectic lamellar spacing can be used to estimate the cooling rate at the point of solidification fairly easily. However, this approach fails when the product of rapid solidification is glassy in nature. Under conditions of non-equilibrium cooling many solid solution alloys solidify in a dendritic fashion. Relationships can be developed between the cooling rate (T) and the primary or secondary dendrite arm spacings. However, primary dendrite arm spacings are neither uniform nor at all suitable for estimating T, since the nucleation of a primary dendrite can take place rather randomly anywhere in the melt. However, once the primary dendrite is nucleated, the spacings of the secondary dendrite arms are dictated by the solidification conditions. Therefore, the secondary dendrite arm spacing can be used and has been widely used as a very convenient measure oft. At low cooling rates, the secondary dendrite arm spacing, X is related to t through
2.6 Cooling Rates in Rapid Solidification
the relationship (2-2)
where K and n are constants for a given material, n has been shown to have a value of about 1/3. A plot relating X and T up to about 3 x 104 K s ~1 was first developed by Dean and Spear (1966). By measuring the cooling rates in rapidly solidified alloys, this plot was extended to higher values of t (Matyja et al, 1968). This method has since been frequently used to measure and grade the cooling rates (Suryanarayana and Anantharaman, 1970; Mehrabian, 1978; Duflos and Stohr, 1982). For example, the cooling rate was measured as 2 x l O 6 K s ~ 1 when Al-Ge alloy melts were quenched by the "gun" technique, while it was only 4 x 104 K s " 1 when the melt was dropped on to a copper substrate (Suryanarayana and Anantharaman, 1970). The variation of dendrite cell spacing measured directly on rapidly solidified Ni-Al and Fe-Ni alloys with cooling rate were also found to follow the above equation (Hayzelden et al., 1983; Gillen and Cantor, 1985). Fig. 2-24 shows a plot relating the dendrite arm spacing to the cooling rate for stainless steels, maraging steels, copper,
97
and aluminum alloys. It may be noted that the slopes are different for different alloys. Burden and Jones (1970 b) used the relationship between eutectic lamellar spacing, d to estimate the cooling rates from the relation d=
(2-3)
where A is a constant (1.04 x 10 5 cm 3/2 s~1/2) and n2 = 0.5 for an Al-Cu eutectic alloy. A knowledge of the splat thickness is necessary for these estimates, which was avoided in estimation of cooling rates from the eutectic cell spacings (Chattopadhyay etal., 1980). Although very convenient for estimating the order of cooling rates, these indirect methods suffer from the following limitations: (a) they are not applicable to glass-forming alloy compositions, (b) they can only give an average cooling rate, (c) each family of alloys requires a separate calibration, and (d) the dendrites may coarsen after solidification and so yield misleading results. For these reasons, direct measurement of cooling rates is desirable.
1000 • 100 Copper alloys 10 Maraging 1 .steels Austenitic stainless steels
0.1
Figure 2-24. Secondary dendrite arm spacing as a function of cooling rate for steels, copper and aluminum alloys.
Aluminum alloys 0.01
10-3
10°
103 Solidification rate (K/s)
10e
109
98
2 Rapid Solidification
The earliest measurement of cooling rates in "gun"-quenched alloys was by Predecki etal. (1965). In their method, the splat was allowed to land on a pair of dissimilar metals kept close to one another in an insulating base, thus forming the hot junction of a thermocouple. The e.m.f. generated was recorded on an oscilloscope and making some assumptions (not all of them justifiable), they calculated the cooling rate to be 1 to 5 x 108 K s " 1 for silver and 106 K s" 1 for an Au-14 at.% Sb alloy. Measurements have been made on piston-quenched liquid alloys using either the above principle (Harbur et al., 1969) or by embedding a temperature sensor such as a photo-electric (Strachan, 1967) or a thin junction conventional thermocouple (Duflos and Cantor, 1982; Nishi etal., 1982). Pyrometry of the droplet has also been employed (Kattamis et al., 1973). In all these cases, it has been shown that the efficiency of the cooling process increases with increasing piston velocity. Cooling rate measurements have also been made on melt-spun ribbons using photographic techniques. A calibration between film density and surface temperature has been established, using which the temperature profiles along the ribbon have been evaluated using either color (Hayzelden et al., 1983; Gillen and Cantor, 1985) or black-and-white (Warrington et al., 1982; Cantor, 1986) photographic techniques. From these investigations also it could be established that the cooling rate increases with higher wheel velocities and also with lower specimen thickness. A remarkable feature noted by Cantor (1986) is that the cooling rates in melt spinning can be represented by a simple relation of the type T = aV
(2-4)
where t is the cooling rate in the vicinity
of the solidification point, V is the wheel velocity and a is a constant (=1.2x x 104 Km" 1 ). The above equation is virtually independent of melt-spun ribbon material or other melt spinning variables. Other methods for measuring the cooling rates in rapid solidification have also been used, e.g., recording of infra-red radiation emitted from the solidifying melt (Lohberg and Muller, 1970). Irrespective of the technique used to measure the cooling rates, it should be borne in mind that the cooling rate varies substantially between the freezing point and ambient temperature. Thus, at 1500°C, near the freezing point, the cooling rate for piston-quenched steel or iron samples 25-50 fim in half-thickness is 1 0 5 107 K s " 1 . The cooling rate at 1000°C is about 10 times slower than at 1500°C, and at 500°C it is 2 - 3 times slower again. The cooling rates in single-roller melt spinning appear to be lower than in piston-andanvil quenching (Hayzelden, 1984). It is generally accepted that the twin-roller quenching is even slower than single-roller melt spinning, in spite of the fact that heat is extracted from both the sides. This is because the ribbon is in contact with the rolls for a very short duration. However, the initial cooling rate is very high (Davies et al., 1978). Direct measurements of cooling rates using thermocouples are subject to the uncertain effects of response-time lag arising from imperfect contact between the melt and the thermocouple and, in the "gun" technique, due to the effects of material overlapping at different stages of cooling. These can result in reduced e.m.f. readings and the measured cooling rates can thus be substantially lower than the actual values. Cooling rates have also been calculated using the heat-flow models (Ruhl, 1967). These have been based on one-dimen-
2.7 Consolidation Methods
sional heat flow and depending on the nature of contact between the specimen and the substrate, one can consider "ideal cooling" (where the heat transfer coefficient h has a value of oo) or "Newtonian cooling" (where h is quite small) or "intermediate cooling". Using these concepts, for "gun"quenched 1 jim thick iron splat, the cooling rate has been calculated a s 8 . 1 x l O 9 K s ~ 1 for ideal cooling and 6.9xlO 7 Ks~ 1 for Newtonian cooling. Refined treatments of these models have also been published (Mehrabian, 1982; Levi and Mehrabian, 1982; Clyne, 1984; Cantor, 1986). In addition to the above-mentioned theoretical estimates or experimental calculations, one can also use structural parameters of the quenched product for comparing two different techniques. For example, if one technique can produce a homogeneous solid solution throughout the composition range in the Cu-Ag system and another technique cannot, then obviously the former technique can produce higher cooling rates (Franetovic et al., 1979). Similarly, the extent of supersaturation (Cahn et al., 1976) and the formation of an amorphous phase can also be used as criteria in this regard.
2.7 Consolidation Methods For rapid solidification technology to become commercially viable and a strong competitor to the existing technology, the rapid solidification product (powder, ribbon, flake, fiber, etc.) needs to be consolidated into larger, more usable, fully dense forms that are adequate in size and shape for required applications. Conventional consolidation techniques and processes used in powder metallurgy require application of pressure at relatively high temperatures. This high-temperature exposure may
99
result in coarsening of the ultrafine structures and/or elimination of other metastable effects produced by rapid solidification processing and this leads to reduced mechanical property levels. Thus, to realize the full potential of rapid solidification products (i.e., refined microstructures, supersaturation of solute atoms and formation of metastable intermediate phases and consequent improved physical and mechanical properties), the consolidation temperature must be relatively low. However, it is difficult to achieve the appropriate particle deformation in the highstrength powders with full densification and inter-particle bonding that is necessary to achieve the desired properties in the consolidated structure at low temperatures. Thus, the challenge is to achieve the necessary integrity at a relatively low temperature. The consolidation of rapidly solidified powders can be carried out in different ways depending on whether the chemical homogeneity or the metastable effects in the rapid solidification products is more important. If chemical homogeneity is the primary benefit sought, then time, temperature and pressure may be chosen in such a way as to produce a nearly porosityfree, fully dense and metallurgically wellbonded product. On the other hand, if the other virtues of rapid solidification (e.g., unique microstructures and metastable phases) have to be retained, then the consolidation parameters have to be much more carefully controlled. The temperatures may have to be low especially when metallic glasses are concerned and their crystallization has to be avoided at all costs. Several methods have been in use for consolidating powders in conventional powder metallurgy technology. These methods can also be used to consolidate rapidly solidified powders or flakes with
100
2 Rapid Solidification
, Sealed
I\container) / (
Heat
\
RamO pressure
Extruded
(b) Laser or electron beam Rapidly solidified deposit
Powder or wire feed Projectile
Powder
Gas pressure (c)
Gun barrel I
appropriate modifications. The most common methods used for consolidating rapidly solidified powders are hot isostatic pressing (HIP), hot pressing, vacuum hot pressing (VHP), and hot extrusion. Some other methods like cold isostatic pressing, hydrostatic extrusion, forging, rolling and dynamic compaction have also been employed. However, HIP seems to be the best method to get a near-net shape. Fig. 2-25 shows schematically four of the commonly used consolidation methods. Grant (1985) and Flinn (1985) have recently summarized the present status of consolidation methods for rapidly solidified products. 2.7.1 Shock-Wave (Dynamic) Compaction
Metallic glasses have been usually compacted by cold compaction methods. Very often, the glassy tapes are subjected to comminution before consolidation. Because of the high strength of these materials and the necessity of preserving the glassy structure, shock-wave (high-energyrate forming) techniques are preferred. These involve explosive forming, explosive bonding, electrohydraulic forming, electromagnetic forming, etc.
(d)
Figure 2-25. Four consolidation methods related to rapid solidification processing, (a) Hot isostatic pressing, (b) hot extrusion, (c) incremental solidification, and (d) dynamic compaction.
The shock-wave concept involves application of a very high pressure or stress pulse for a short duration. The wave travels through the medium at supersonic velocities (about 105 cms" 1 or Mach 3) and greatly accelerates matter in the shock front into a region of the medium which is in a static state. This produces a marked interaction of the atoms or particles. The important parameters to be considered in this process are pressure and pressure duration, stress state and history, particle size, shape and distribution, and precompaction density. Pressure-densification curves for different amorphous and microcrystalline materials have been reported (Morris, 1980 a). Amorphous metallic ribbons or strips have been consolidated by explosive compaction methods (Clauer et al., 1980; Murr et al., 1981) and the results indicate that bonding of successive layers of amorphous strips is rather difficult to achieve (Clauer etal., 1980). Powders produced by atomization techniques or obtained by comminution of strips/ribbons have also been compacted (Cline and Hopper, 1977; Cline etal., 1978; Morris, 1980b). Cline etal. (1978) were able to achieve full densifica-
2.7 Consolidation Methods
tion and also good bonding between amorphous metallic filaments and conventional crystalline material substrates. Rapidly solidified Al-6% Si alloy (Gourdin, 1984), MAR-M 200 (superalloy) (Meyers et al., 1981), and stainless steel powders were also compacted. Although some investigations have been carried out using shock waves, the results are not easy to analyze because the process is very complex. Optimization of structure/ property correlation in terms of consolidation parameters is still to be carried out systematically. However, this technique seems to hold great potential. 2.7.2 Hot Compaction
These techniques have been employed when chemical homogeneity is more important than retention of metastable phases and fine microstructures in the rapidly solidified products. Although several methods are possible, hot isostatic pressing (HIP) and hot extrusion are the most favored techniques. Hot isostatic pressing is favored because of its near-net-shape processing capabilities and potential for reducing manufacturing costs, while hot extrusion is favored for control of microstructures. The HIP process of rapidly solidified powders involves (i) complete filling of a container to achieve maximum possible packing efficiency, (ii) degassing to remove any residual gases, (iii) hermetical sealing of the container, and (iv) hot isostatic pressing by choosing the appropriate pressure, temperature and time cycles. Grain growth in HIP-consolidated powders is very slow compared to conventional heat treatments (Kelly et al., 1982). Recently, the Rapid Omnidirectional Compaction (ROC) process has been used to produce fully recrystallized structures in
101
a variety of titanium-base alloys (Mahajan et al., 1985). In this process the powder contained in a thick-walled container (made of a material which acts as an incompressible fluid when subjected to an external pressure) is subjected to isostatic pressure developed in a conventional forging press. Hot extrusion is much more commonly applied to consolidate rapidly solidified powders. Preparation of the powder is done in a manner similar to that for HIP (the can used has a thicker wall). The powder-filled billet is heated to approximately 0.6-0.7 Tm prior to extrusion. The principal variables in this process are (a) extrusion ratio, (b) deformation velocity, and (c) working temperature. This process has been used extensively for superalloys and aluminum alloys. Warm extrusion has been shown to be superior to warm uniaxial die pressing since extrusion confers substantial shear components resulting in breaking-up of surface oxide layer and improved adhesion between neighboring particles (Liebermann, 1980 b). Tables 2-4 and 2-5 show the process correlations with principal parameters for powder consolidation and the extent of retention of rapid solidification virtues, respectively (Flinn, 1985). The advantages and disadvantages of different consolidation methods are presented in Table 2-6. In addition to the methods mentioned above, one can also use incremental deposition methods like spray forming techniques to obtain large quantities of rapidly solidified material. This can be later subjected to rolling or extrusion. In some of the cold consolidation methods, addition of other powders can also be made before extrusion (Austen et al., 1986).
102
2 Rapid Solidification
Table 2-4. Process correlations with principal parameters for powder consolidation. Consolidation method Inert gas hot pressing Vacuum hot pressing Hot isostatic pressing Forging Rolling Extrusion Shock/Dynamic
Principal parameter characteristic Pressure
Bulk temperature
Time
Densification rate
Low-medium Low-medium Low-medium Medium-high Medium-high High Very high
High High High Medium-high Medium-high Medium-high Low-medium
Long Long Long Short Short-medium Short Very short
Low Low Low Medium-high Medium-high High Very high
Table 2-5. Expected retention of rapid solidification virtues for the various consolidation processes. Consolidation method
Rapid solidification virtues Chemical homogeneity
Refined microstructures
Extended solid solubility
Metastable phases
Inert gas hot pressing Vacuum hot pressing Hot isostatic pressing Forging Rolling Extrusion Shock/dynamic
High High High High High High High
Medium Medium Medium Low-medium Low-medium Medium-high High
Low-medium Low-medium Low-medium Low-medium Low-medium Medium High
Low Low Low Low Low Low High
2.8 Concluding Remarks The salient features of the remarkable range of techniques available for rapidly solidifying metallic melts have been reviewed. While the rapid solidification product can be powder, ribbon, tape, wire, etc., the solidification rates achieved can vary from as low as 102 to as high as 10 8 109 K s ~*. The exhaustive list of references at the end of the Chapter can be consulted for the specific details of any process. With a proper choice of the technique, the desired metastable effects can be easily achieved in a given alloy system. In fact, rapid solidification processing combined with a subsequent heat treatment can confer properties on a material which are otherwise difficult to attain or unattainable.
However, caution has to be exercised in comparing the properties of materials obtained by different methods, especially the rapid solidification and other non-equilibrium processing techniques, since the mechanisms of formation of the non-equilibrium structures are quite different in different techniques. It should also be realized that differences in the structure and properties of the same alloys have been noted even when they were prepared by rapid solidification techniques. These can be attributed to the differences in the quenching conditions, which in turn calls for a great control on the experimental variables for obtaining reproducible results. The large amount of experimental data available should make the rapid solidification process amenable to rigorous mathe-
2.8 Concluding Remarks
103
Table 2-6. Some apparent advantages and disadvantages of powder consolidation processes for rapidly solidified materials. Consolidation processing
Apparent advantages
Apparent disadvantages
Inert gas hot pressing
Containerless operation
Vacuum hot pressing
Containerless operation, in situ outgassing, less tendency for porosity Uniform pressure - homogeneous consolidations; more flexibility in parametric cycling Simple operation
High temperatures, tendencies for porosity, possible nonhomogeneous consolidation High temperatures, possible nonhomogeneous consolidation High temperatures, requires containerization, and substantial degassing, porosity
Hot isostatic pressing
Forging
Rolling
Established operation
Extrusion
Established one-step operation, very high-restrictive deformation
Shock/dynamic
Low bulk temperatures
matical modeling. Efforts have already been made in this direction with partial success. For the successful exploitation of the rapidly solidified alloys, tonnage quantity of the material is required. At the present, the different variants of the atomization process appear to be the only choice. The solidification rates achieved in these methods, however, are low and so the extent of non-equilibrium effects achieved is also limited. Development of innovative consolidation methods is also required to retain the rapid solidification virtues in bulk forms made from the as-quenched material. Rapid solidification technology has come a long way from the 1960's when only a few milligrams of the product could be produced at a time. The variety of alloy
Requires thick wall container and degassing, nonuniform consolidation and structure gradients, parametric control in gross, requires several billet reheats High temperatures, nonuniform consolidation and structure gradients, relative thick wall container and degassing, requires several workpiece reheats High temperature, requires containerization and degassing, some nonuniformity in consolidation and structure Shock physics and material response not understood, very high work-hardened conditions, size limitations
systems investigated, the quantity of material produced, the wide spectrum of techniques developed, and the commercial exploitation of rapid solidification products are indeed remarkable. The U.S. market for rapidly solidified materials is expected to reach $ 290 million by 1993. Significant improvements in combinations of tensile strength, fatigue, fracture toughness, electrical conductivity and corrosion resistance have been responsible for the increased acceptance of these alloys for high-performance applications. The cost of these products is also expected to come down with increased tonnage production. Thus, it can be concluded that rapid solidification technology will pave the way for new materials and metal processing technologies in the future.
104
2 Rapid Solidification
2.9 References Adam, C M . (1986), in: Mechanical Behavior of Rapidly Solidified Materials: Sastry, S.M.L., MacDonald, B. A. (Eds.). Warrendale, PA: Metall. Soc. AIME, pp. 21-39. Aller, A. I, Losada, A. (1990), Metal Powder Rep. 45, 51-55. Anand, V., Kaufman, A.J., Grant, N . I (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 273-286. Anantharaman, T.R. (Ed.) (1984), Metallic Glasses: Production, Properties and Applications. Aedermannsdorf (Switzerland): Trans. Tech. Publ. Anantharaman, T.R., Suryanarayana, C. (1971), /. Mater. Sci. 6, 1111-1135. Anantharaman, T. R., Suryanarayana, C. (1987), Rapidly Solidified Metals: A Technological Overview. Aedermannsdorf (Switzerland): Trans. Tech. Publ. Anthony, T. R., Cline, H. E. (1978), /. Appl. Phys. 49, 829-837. Anthony, T.R., Cline, H.E. (1979), J. Appl. Phys. 50, 245-254. Apelian, D., Paliwal, M., Smith, R.W., Schilling, W.F. (1983), Internat. Met. Rev. 28, 271. Armstrong, G.R., Jones, H. (1979), in: Solidification and Casting of Metals. London: The Metals Soc, pp. 454-459. Austen, A.R., Hutchinson, W.L., Feilbach, W.H. (1986), Metal Powder Rep. 41, 294-298. Ayers, I D . , Tucker, T.R., Schaefer, R.J. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 212-220. Babic, E., Girt, E., Krsnik, R., Leontic, B. (1970 a), /. Phys. E: Sci. Instrum. 3, 1014-1015. Babic, E., Girt, E., Krsnik, R., Leontic, B., Zoric, I. (1970 b), Fizika 2, Suppl. 2, Paper 2. Baker, IN., Mighton, C.E., Bitler, W.R. (1969), Rev. Sci. Instrum. 40, 1065-1066. Baram, J. (1988 a), J. Mater. Sci. 23, 405-409. Baram, J. (1988b), J. Mater. Sci. 23, 3656-3659. Baram, J. (1990), JOM 42, No. 1, 20-26. Beck, H., Guntherodt, H.-J. (1983), Glassy Metals II. Berlin: Springer-Verlag. Bedell, J. (1975), U.S. Patent No. 3 862 658. Beghi, G., Matera, R., Piatti, G. (1969), /. Nucl. Mater. 31, 259-268. Bendersky, L. A., Boettinger, W.J. (1981), in: Treatise on Materials Science and Technology, Vol. 20 - Ultrarapid Quenching of Liquid Alloys: Herman, H. (Ed.). New York: Academic Press, p. 887. Berkowitz, A.E., Walter, A.L. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M.
(Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 294-305. Bickerdicke, R.L., Clark, D., Easterbrook, J.N., Hughes, G., Mair, W.N., Partridge, P.G., Ranson, H.C. (1986), Internat. J. Rapid Solidification 2, 1-19. Blank, E. (1972), Arch. Eisenhuttenw. 43, 649-655. Bletry, J. (1970), J. Phys. Chem. Solids31,1263-1272. Boettinger, W.J., Shechtman, D., Schaefer, R.J., Biancaniello, F.S. (1984), Metall. Trans. 15 A, 55-66. Bonetti, E., Evangelista, E., Lanzoni, E. (1981), Scri. Metall. 15, 1067-1071. Booth, A.R., Charles, J. A. (1966), Nature 212, 750751. Borders, J.A. (1979), Ann. Rev. Mater. Sci. 9, 313339. Boswell, P.G., Chadwick, G.A. (1976), /. Phys. E: Sci. Instrum. 9, 523-526. Boulby, K.A., Wood, J.V. (1986), Powder Met. 29, 33. Breinan, E.M., Kear, B.H. (1976), in: Superalloys: Metallurgy and Manufacture: Kear, B.H., Muzyka, P.R., Tien, J.K., Wlodek, S.T. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 435-440. Breinan, E.M., Kear, B.H. (1978), in: Rapid Solidification Processing: Principles and Technologies: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 87-103. Breinan, E.M., Kear, B.H., Banas, C M . (1976), Phys. Today 29, No. 11, 44-50. Brenner, A. (1963), Electrodeposition of Alloys: Principles and Practice. New York: Academic Press. Brooks, R.G., Leatham, A.G., Coombs, I S . , Moore, C (1977), Metallurgia and Metal Forming 9, No. 4, 1. Bryant, W.A. (1977), /. Mater. Sci. 12, 1285-1306. Budaira, S., Suito, S. (1979), U.K. Pat. Appln. 2003 772 A. Bunge, H.-J. (1976), Z. Metallkde. 67, 720-728. Burden, M.H., Jones, H. (1970a), Metallography 3, 307-326. Burden, M.H., Jones, H. (1970b), /. lnst. Metals 98, 249-252. Cahn, R. W (1978), in: Rapid Solidification Processing: Principles and Technologies: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 129-139. Cahn, R. W. (1983), in: Physical Metallurgy, Vol. 2, 3rd ed., Cahn, R.W., Haasen, P. (Eds.). Amsterdam: Elsevier Sci. Pub., pp. 1779-1852. Cahn, R.W, Krishnanand, K.D., Laridjani, M., Greenholz, M., Hill, R. (1976), Mater. Sci. & Eng. 23, 83-86. Cantor, B. (1986), in: Science and Technology of the Undercooled Melt: Sahm, P. R., Jones, H., Adam, C M . (Eds.). Dordrecht, NL: Martinus Nijhoff, p. 3. Caryll, D.B., Ward, R.G. (1967), J. Iron and Steel lnst. 205, 28-31.
2.9 References
Chattopadhyay, K., Ramineni, A. P., Ramachandrarao, P. (1980), J. Mater. Sci. 15, 797-799. Chen, H.S., Leamy, H.J., Miller, C.E. (1980), Ann. Rev. Mater. Sci. 10, 363-391. Chen, H.S., Miller, C.E. (1970), Rev. Sci. Instrum. 41, 1237-1238. Chen, H.S., Miller, C.E. (1976), Mater. Res. Bull. 11, 49-54. Chopra, K.L. (1969), Thin Film Phenomena. New York: McGraw-Hill. Clauer, A.H., Raman, R.V., Carbonara, R.S., Maringer, R.E. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 399-403. Cline, C.F., Hopper, R.W. (1977), Scri. Metall. 11, 1137-1138. Cline, C.F., Mahler, I, Finger, M., Kuhl, W, Hopper, R.W. (1978), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 380-387. Cline, H. E., Anthony, T. R. (1979), J. Appl. Phys. 50, 239-244. Clyne, T.W. (1984), Metall. Trans. 15B, 369-381. Cogan, S.F., Rockwell, III, IE., Cocks, F.H., Shepard, M.L. (1978), J. Phys. E: Sci. Instrum. 11, 174-176. Cohen, M., Kear, B.H., Mehrabian, R. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 1-23. Collings, E.W., Mobley, C.E., Maringer, R.E. (1978), A.ICh.E. Symp. Ser. No 180, 74, 102. Cooper, K. P., Ayers, J. D. (1985), Surface Eng. 1, 263. Couper, M.J., Singer, R.F. (1985), in: Rapidly Quenched Metals V: Steeb, S., Warlimont, H. (Eds.). Amsterdam: Elsevier Sci. Pub., B.V., pp. 1737-1742. Cox, A. R., Moore, J. B., Van Reuth, E. C. (1976), in: Superalloys: Metallurgy and Manufacture: Kear, B.H., Muzyka, P.R., Tien, J.K., Wlodek, S.T. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 45-53. Dahlgren, S.D. (1978), in: Rapidly Quenched Metals III, Vol. 2: Cantor, B. (Ed.). London: The Metals Soc, pp. 36-47. Daugherty, T.S. (1964), /. Metals 16, No. 10, 827830. Davies, H.A. (1978), in: Rapidly Quenched Metals III, Vol. 1: Cantor, B. (Ed.). London: The Metals Soc, pp. 1-21. Davies, H.A., Hull, XB. (1972), Scri. Metall. 6, 241245. Davies, H. A., Hull, J. B. (1974), J. Mater. Sci. 9,101717. Davies, H. A., Lewis, B.G. (1976), Metall. Trans. 7 A, 310-311.
105
Davies, H.A., Lewis, B.G., Donald, I.W. (1978), in: Rapid Solidification Processing: Principles and Technologies: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 78-83. Dean, W. A., Spear, R.E. (1966), in: Proc. 12th Army Materials Research Conf. Syracuse, NY: Syracuse Univ. Press, pp. 268-271. Devillard, I, Herteman, J.P. (1980), in: Powder Metallurgy of Titanium Alloys: Froes, F. H., Smugeresky, J. E. (Eds.). Warrendale, PA: The Met. Soc. AIME, pp. 59-70. Dixmier, I, Guinier, A. (1967), Mem. Sci. Rev. Met. 64, 53-58. Donald, I.W. (1987), /. Mater. Sci. 22, 2661-2679. Draper, C. W, Poate, J. M. (1985), Internat. Met. Rev. 30, 85-108. Duflos, B., Cantor, B. (1982), Ada Metall. 30, 323356. Duflos, B., Stohr, J. F. (1982), J. Mater. Sci. 17, 3641 3652. Dunkley, J.J. (1982), in: Progress in Powder Metallurgy, Vol.37: Capus, J.M., Dyke, D.L. (Eds.). Princeton, NJ: MPIF. Durand, J.P.H.A., Pelloux, R.M., Grant, N.J. (1976), Mater. Sci. & Eng. 23, 247-256. Duwez, P. (1968), in: Techniques of Metals Research, Vol. I: Bunshah, R. F. (Ed.). New York: Interscience, pp. 347-358. Duwez, P., Willens, R.H. (1963), Trans. Met. Soc. AIME 227, 362-365. Duwez, P., Willens, R.H., Klement, W (1960), J. Appl. Phys. 31, 1136, 1137, 1500. Egami, T. (1984), Rep. Progr. Phys. 47, 1601-1725. Elliot, W.A., Gagliano, F.P., Krauss, G. (1973), Metall. Trans. 4, 2031-2037. Esslinger, P. (1966), Z. Metallkde. 57, 12-19. Evans, N.D., Hofmeister, W.H., Bayuzick, R.J., Robinson, M.B. (1986), Metall. Trans. 17A, 973981. Falkenhagen, G., Hofmann, W. (1952), Z. Metallkde. 43, 69-81. Fleetwood, M.J. (1987), Metals & Materials 3, 1420. Flinn, J.E. (1985), Rapid Solidification Technology for Reduced Consumption of Strategic Materials. Park Ridge, NJ: Noyes Publ. Fortman, W.K., Ullman, T.S. (1984), Metal Powder Rep. 39, 259-261. Franetovic, V, Milat, O., Ivcek, D., Bonefacic, A. (1979), J. Mater. Sci. 14, 498-500. Friedman, G. (1976), in: Advanced Fabrication Techniques in Powder Metallurgy and Their Economic Implications, AGARD Conf. Proc. No. 200, Structures and Materials Panel Meeting, Canada, p. 1. Fujita, H., Mori, H. (1988), Suppl. Trans. Jap. Inst. Metals 29, 37-40. Fujiwara, S., Nishikawa, Y, Hanafusa, H., Tamura, I. (1982), in: Rapidly Quenched Metals IV, Vol. II:
106
2 Rapid Solidification
Masumoto, T., Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 1497-1500. Galasso, K, Vaslet, R. (1966), Rev. Sci. lustrum. 37, 525. Gardiner, R. W, McConnell, M.C. (1987), Metals & Materials 3, 254-258. Gaspar, T., Hackman, L. E., Sahai, Y., Clark, W. A. T, Wood, J. V. (1986), in: Rapidly Solidified Alloys and Their Mechanical and Magnetic Properties: Giessen, B.C., Polk, D.E., Taub, A.I. (Eds.). Pittsburgh, PA: Mater. Res. Soc, pp. 23-26. Giessen, B.C., Madhava, N.M., Murphy, R.J., Ray, R., Surette, J. (1977), Metall Trans. 8 A, 364-366. Gillen, A.G., Cantor, B. (1985), Acta Metall. 33, 1813-1825. Goto, T., Yuki, Y, Nagano, M., Oda, T, Takai, H. (1974). Sen-i-Gakkaishi 30, 381-386. Goto, T, Nagano, M., Tanaka, K. (1977), Trans. Jap. Inst. Metals 18, 209-213. Gourdin, W.H. (1984), /. Appl. Phys. 55, 172-181. Grant, N. J. (1983), /. Metals 35, No. 1, 20-27. Grant, N. J. (1985), in: Rapidly Quenched Metals V: Steeb, S., Warlimont, H. (Eds.). Amsterdam: Elsevier Sci. Pub., B.V., pp. 3-24. Groeber, H., Haneman, H. (1937), Arch. Eisenhuttenw. 11, 199-202. Giintherodt, H.-J., Beck, H. (1981), Glassy Metals I. Berlin: Springer-Verlag. Gutierrez, E.M., Szekely, J. (1986), Metall. Trans. 17B, 695-703. Hanumantha Rao, M., Sridhar, G., Suryanarayana, C. (1985), Internat. J. Rapid Solidification 1, 199218. Harbur, D.R., Anderson, J.W., Maraman, W.J. (1969), Trans. Met. Soc. AIME 245, 1055-1061. Hayzelden, C. (1984), D. Phil. Thesis, University of Sussex, U.K. Hayzelden, C , Rayment, J. X, Cantor, B. (1983), Acta Metall. 31, 379-386. Heineman, W. A. (1985), in: Rapidly Quenched Metals V: Steeb, S., Warlimont, H. (Eds.). Amsterdam: Elsevier Sci. Pub., B.V., pp. 27-34. Herman, H. (1988), Mater. Res. Soc. Bull. 13, No. 12, 60-67. Hilzinger, H.R., Hock, S. (1981), in: Metallic Glasses: Science and Technology, Vol. 1: Hargitai, C , Bakonyi, I., Kemeny, T. (Eds.). Budapest, Hungary: Central Res. Inst. Phys., pp. 71-90. Hinesly, C.P., Morris, J.G. (1970), Metall. Trans. 1, 1476-1478. Hohmann, M., Jonsson, S. (1990), Metal Powder Rep. 45, 47-50. Hopkins, WG. (1987), Metal Powder Rep. 42, 706707. Huang, S. C. (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto, T, Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 65-68. Huang, S.C., Fiedler, H.C. (1981), Metall. Trans. 12 A, 1107-1112.
Hubert, J.C., Mollard, R, Lux, B. (1973), Z. Metallkde. 64, 835-843. Ichikawa, R., Ohashi, T, Ikeda, T. (1971), Trans. Jap. Inst. Metals 12, 280-284. Ishii, H., Naka, M., Masumoto, T. (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto, T, Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 35-38. Jackson, M.R., Rairden, J.R., Smith, J.S., Smith, P. W. (1981), /. Metals 33, No. 11, 23-27. Jech, R.W., Moore, T.J., Glasgow, T.K., Orth, N.W. (1984), /. Metals 36, No. 4, 41-45. Johnson, WL. (1986), Progr. Mater. Sci. 30, 81. Jones, H. (1981), in: Treatise on Materials Science and Technology, Vol. 20: Ultrarapid Quenching of Liquid Alloys: Herman, H. (Ed.). New York: Academic Press, pp. 1-72. Jones, H. (1982), Rapid Solidification of Metals and Alloys, Monograph No. 8. London: The Institution of Metallurgists. Jones, H., Suryanarayana, C. (1973), J. Mater. Sci. 8, 705-753. Kadalbal, R., Montoya-Cruz, X, Kattamis, T.Z. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 195-205. Kainer, K.U., Mordike, B.L. (1989), Metal Powder Rep. 44, 28-31. Kane, R.H., Giessen, B.C., Grant, N.X (1966), Acta Metall. 14, 605-609. Katgermann, L., Van den Brink, P.X (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto, T, Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 61-64. Kattamis, T. Z., Brower, W E., Mehrabian, R. (1973), /. Cryst. Growth 19, 229-236. Kavesh, S. (1974), U.S. Patent No. 3845805. Kavesh, S. (1978), in: Metallic Glasses. Metals Park, OH: Amer. Soc. Metals, pp. 36-73. Kear, B.H., Mayer, J.W., Poate, J.M., Strutt, P.R. (1981), in: Metallurgical Treatises: Tien, J.K., Elliott, XF. (Eds.). Warrendale, PA: The Met. Soc. AIME, p. 321. Kelly, T.F., Olson, G.B., Vander Sande, XB. (1982), in: Rapidly Solidified Amorphous and Crystalline Alloys: Kear, B. H., Giessen, B. C , Cohen, M. (Eds.). New York: Elsevier Sci. Pub. Co., pp. 343-348. Kim, K. Y, Marshall, W.R. (1971), A.I.Ch.E. Journal 17, 575. King, D. La W (1967), Metals 2, 32. Klement, Jr., W, Willens, R.H., Duwez, P. (1960), Nature 187, 869-870. Konitzer, D.G., Walters, K.W., Heiser, E.L., Fraser, H.L. (1984), Metall. Trans. 15B, 149-153. Krishnanand, K.D., Cahn, R.W. (1976), in: Rapidly Quenched Metals II: Grant, N.X, Giessen, B.C. (Eds.). Cambridge, MA: M.I.T. Press, pp. 67-75. Laine, E., Heikkila, E., Lahteenmaki, I. (1971), Rev. Sci. Instrum. 42, 1724-1725.
2.9 References
Lakshmikumar, S.T., Mallya, R.M., Gopal, E.S.R. (1980), Bull. Mater. Sci. (India) 2, 233. Laridjani, M., Ramachandrarao, P., Cahn, R.W. (1972), J. Mater. Sci. 7, 627-630. Lawley, A. (1977), Internat. J. Powder Met. & Powder Technol. 13, 169-188. Lawley, A. (1978), Ann. Rev. Mater. Sci. 8, 49-71. Lawley, A. (1981), /. Metals 33, No. 1, 13-18. Leontic, B., Lukatela, I, Babic, E., Ocko, M. (1978), in: Rapidly Quenched Metals III, Vol. 1: Cantor, B. (Ed.). London: The Metals Soc., pp. 41-48. Levi, C.G., Mehrabian, R. (1982), Metall. Trans. 13 A, 13-23. Lewis, B.G., Donald, I.W., Davies, H.A. (1979), in: Solidification and Casting of Metals. London: The Metals Soc., pp. 490-495. Lewis, B.G., Gilbert, D.A., Strutt, P. R. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 221-224. Liebermann, H.H. (1980 a), Mater. Sci. & Eng. 43, 203-210. Liebermann, H.H. (1980b), Mater. Sci. & Eng. 46, 241-248. Liebermann, H.H. (1981), Mater. Sci. & Eng. 49, 185-191. Liebermann, H.H. (1983), in: Amorphous Metallic Alloys: Luborsky, F. E. (Ed.). London: Butterworths, pp. 26-41. Liebermann, H. H. (1984a), /. Cryst. Growth 70, 497506. Liebermann, H. H. (1984b), Metall. Trans. 15B, 155161. Liebermann, H.H. (1985), Internat. J. Rapid Solidification 1, 103-113. Liebermann, H.H., Graham, Jr., C D . (1976), IEEE Trans. Magn. MAG-12, 921-923. Lin, C-I, Spaepen, F. (1983), Scri. Metall. 17, 12591262. Lin, C-J., Spaepen, F. (1986), Acta Metall. 34, 13671375. Liu, B.X. (1988), Suppl. Trans. Jap. Inst. Metals 29, 29-36. Loewenstein, P. (1981), Metal Powder Rep. 36, 5964. Lohberg, K., Muller, H. (1969), Z. Metallkde. 60, 231-237. Lohberg, K., Muller, H. (1970), Fizika 2, Suppl. 2, Paper 4. Lubanska, H. (1970), J. Metals 22, No. 2, 45-49. Luborsky, F. E. (Ed.) (1983), Amorphous Metallic Alloys. London: Butterworths. Luborsky, F.E., Livingston, J.D., Chin, G. Y. (1983), in: Physical Metallurgy, Vol.2: Cahn, R.W., Haasen, P. (Eds.). Amsterdam: Elsevier Sci. Pub., B.V, p. 1673. Mahajan, Y.R, Eylon, D., Kelto, C.A., Egerer, T, Froes, F. H. (1985), in: Titanium Science and Technology, Vol. 1: Liitjering, G., Zwicker, U., Bunk, W.
107
(Eds.). Oberursel, Germany: Deutsche Gesellschaft fur Metallkunde, pp. 339-346. Manfre, G., Servi, G., Ruffmo, C. (1974), J. Mater. Sci. 9, 74-80. Maringer, R., Mobley, C.E. (1974), J. Vac. Sci. & Tech. 11, 1067-1071. Maringer, R., Mobley, C.E. (1978a), in: Rapidly Quenched Metals III, Vol. 1: Cantor, B. (Ed.). London: The Metals Soc, pp. 49-56. Maringer, R., Mobley, C.E. (1978 b), in: Rapid Solidification Processing: Principles and Technologies: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 208-221. Maringer, R., Mobley, C. E. (1979), Wire J., January, 70-74. Maringer, R., Mobley, C.E., Collings, E.W. (1976), in: Rapidly Quenched Metals II. Grant, N. 1, Giessen, B.C. (Eds.). Cambridge, MA: M.I.T. Press, pp. 29-36. Maringer, R., Mobley, C.E., Collings, E.W. (1978), A.I.Ch.E. Symp. Ser. No. 180, 74, 111-116. Masumoto, T, Inoue, A., Sakai, S., Kimura, H.M., Hoshi, A. (1980), Trans. Jap. Inst. Metals 21, 115122. Masumoto, T, Maddin, R. (1971), Acta Metall. 19, 725-741. Masumoto, T, Ohnaka, I., Inoue, A., Hagiwara, M. (1981), Scri. Metall. 15, 293-296. Matei, G., Bicsak, E., Huppman, W.J., Clausen, H. (1977), in: Developments in Powder Metallurgy, Vol. 9: Hausner, H.H., Taubenblaut, P.W. (Eds.). Princeton, NJ: MPIF, pp. 153-159. Matyja, H., Giessen, B. C , Grant, N. J. (1968), J. Inst. Metals 96, 30-32. Mehrabian, R. (1978), in: Rapid Solidification Processing: Principles and Technologies: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 9-27. Mehrabian, R. (1982), Internat. Met. Rev. 27, 185208. Meyers, M.A., Gupta, B.B., Murr, L.E. (1981), J. Metals 33, No. 10, 21-26. Miller, S.A. (1983), in: Amorphous Metallic Alloys: Luborsky, F. E. (Ed.). London: Butterworths, pp. 506-521. Miller, S.A., Murphy, R.J. (1979), Scri. Metall. 13, 673-676. Miroshnichenko, I.S., Salli, I. V. (1959), Industr. Lab. 25, 1463-1464. Mitera, M., Masumoto, T., Kazama, N.S. (1979), /. Appl. Phys. 50, 7609-7611. Miura, H., Isa, S., Omura, K. (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto, T, Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 43-46. Miyazawa, K., Szekely, J. (1979), Metall. Trans. 10 B, 349-358. Miyazawa, K., Szekely, J. (1981), Metall. Trans. 12 A, 1047-1057.
108
2 Rapid Solidification
Moll, J.H., Yolton, C.F. (1986), in: Titanium: Rapid Solidification Technology: Froes, F. H., Eylon, D. (Eds.). Warrendale, PA: The Metallurgical Soc, Inc., pp. 45-56. Morris, D.G. (1980a), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R,, Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 372-384. Morris, D.G. (1980b), Metal. Sci. 14, 215-220. Moss, M. (1968), Ada Metall. 16, 321-326. Moss, M., Schuster, D.M. (1969), Trans. ASM 62, 201. Mottern, J.W., Privott, W.J. (Eds.) (1973), Spinning Wire from Molten Metal, A.I.Ch.E. Symp. Ser. No. 180, 74. New York: Amer. Inst. Chem. Engrs. Murr, L.E., Inal, O.T., Whang, S.H. (1981), Mater. Sci. Eng. 49, 57-64. Murty, Y.V., Adler, R.P.I. (1982), J. Mater. Sci. 17, 1945-1954. Narasimhan, M. C. (1979), U.S. Patent No. 4142571. Narasimhan, M. C , Boggs, B. A., Davis, L. A., Kushnick,, J.H. (1981), cited in Davis, L.A., DeCristofaro, N., Smith, C.H., in: Metallic Glasses: Science and Technology, Vol.1: Hargitai, C , Bakonyi, I., Kemeny, T. (Eds.). Budapest, Hungary: Central Res. Inst. Phys., p. 1. NASA (1980), Fact Scheet KSC 191-80: A Primer on Propellants, obtainable from Federal Duplication Service, Library of Congress, Washington, DC. Nishi, Y, Morohoshi, T., Kawakami, M., Suzuki, K., Masumoto, T. (1982), in: Rapidly Quenched Metals IV, Vol.1: Masumoto, T., Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 111-114. Ogata, K., Lavernia, E., Rai, G., Grant, N. J. (1986), Internat. J. Rapid Solidification 2, 21-35. Ohnaka, I. (1985), Internat. J. Rapid Solidification 1, 219-236. Ohnaka, I., Fukusaka, T, Tsutsumi, H. (1985), Trans. Jap. Inst. Metals 26, 52. Ohring, M., Haldipur, A. (1971), Rev. Sci. Instrum. 42, 530-531. Ostroumov, G.A. (1959), Zhur. Tekh. Fiz. 2, 2. Pardoe, G. W R, Butler, E., Gelder, D. (1978), /. Mater. Sci. 13, 786-790. Pavuna, D. (1981), /. Mater. Sci. 16, 2419-2433. Pavuna, D. (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto, T, Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 81-84. Peng, T.C., Sastry, S.M.L., O'Neal, J.E. (1985 a), Metall. Trans. 16A, 1897-1900. Peng, T.C., Sastry, S.M.L., O'Neal, J.E. (1985b), in: Titanium Science and Technology, Vol. 1: Liitjering, G., Zwicker, U., Bunk, W. (Eds.). Oberursel, Germany: Deutsche Gesellschaft fur Metallkunde, pp. 389-396. Perel, X, Mahoney, I F , Duwez, P., Kalensher, B.E. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 287-293.
Perel, J., Mahoney, I F , Kalensher, B.E., Duwez, P. (1981), in: Advances in Metal Processing: Burke, II, Mehrabian, R., Weiss, V. (Eds.). New York: Plenum, pp. 79-89. Pietrokowsky, P. (1963), Rev. Sci. Instrum. 34, 445446. Pond, R.B. (1958), U.S. Patent No. 2825108. Pond, R.B. (1959), U.S. Patent No. 2879566. Pond, R.B. (1961), U.S. Patent No. 2976590. Pond, R.B., Maddin, R. (1969), Trans. Met. Soc. AIME 245,2415-2416. Pond, R.B., Maringer, R.E., Mobley, C.E. (1976), in: New Trends in Materials Processing. Metals Park, OH: ASM, p. 128. Ponyatovsky, E.G. (1988), Suppl. Trans. Jap. Inst. Metals 29, 111-116. Powell, I, Steen, W.M. (1982), U.K. Patent Appln. No. 2 090 873 A. Predecki, P., Mullendore, A.W., Grant, N . I (1965), Trans. Met. Soc. AIME 233, 1581-1586. Ramachandrarao, P., Banerjee, D., Anantharaman, T.R. (1970), Metall. Trans. 1, 2655-2657. Ramachandrarao, P., Laridjani, M., Cahn, R.W (1972), Z. Metallkde. 63, 43-49. Raman, R.V., Patel, A.N., Carbonara, R.S. (1984), Metal Powder Rep. 39, 106-107. Ray, R. (1979), Internat. Patent Appln. No. W79j 01054. Ray, R., Clemm, P.C. (1986), Titanium: Rapid Solidification Technology: Froes, F. H., Eylon, D. (Eds.). Warrendale, PA: The Metallurgical Soc, Inc., pp. 57^-68. Rechtin, M.D., Vander Sande, IB., Baldo, P.M. (1978), Scri. Metall. 12, 639-643. Reddy, G.S., Rao, P.V., Sekhar, I A. (1986), Internat. J. Rapid Solidification 2, 37-45. Rickinson, B.A., Kirk, F. A., Davies, D.R.G. (1981), Powder Met. 24, 1. Ricks, R. A., Clyne, T.W. (1985), /. Mater. Sci. Lett. 4, 814-817. Robertson, S.R., Gorsuch, T.I, Adler, R.P.I. (1978), in: Rapid Solidification Processing: Principles and Technologies: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 188-207. Rosen, G., Avissar, I, Baram, I, Gefen, Y (1986a), Internat. J. Rapid Solidification 2, 67-82. Rosen, G., Avissar, I, Gefen, Y, Baram, I (1986b), in: Rapidly Solidified Materials: Lee, P.W., Carbonara, R.S. (Eds.). Metals Park: ASM, pp. 9-13. Rosen, G., Avissar, I, Gefen, Y, Baram, I (1987), /. Phys. E: Sci Instrum. 20, 571-574. Rowe, R.G., Amato, R.A. (1987), in: Processing of Structural Metals by Rapid Solidification: Froes, F.H., Savage, S.I (Eds.). Metals Park, OH: ASM Internat., pp. 253-260. Ruhl, R.C. (1967), Mater. Sci. & Eng. 1, 313-320. Ruhl, R . C , Cohen, M. (1969), Trans. Met. Soc. AIME 245,241-251.
2.9 References
Sakata, M., Ishibachi, T. (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto, T., Suzuki, K. (Eds.)Sendai, Japan: The Jap. Inst. Metals, pp. 39-42. Salli, I. V. (1958), J. Inorg. Chem. (SSSR) 3, No. 4, 136-150. Sankaran, K.K., Grant, N.J. (1980), Mater. ScL & Eng. 44, 213-227. Sarjeant, P.J., Roy, R. (1967), J AppL Phys. 38, 4540-4542. Sastry, S.M.L., Peng, T.C., Meschter, P.I, O'Neal, I E . (1983), J Metals 35, No. 9, 21-28. Sastry, S.M.L., Meschter, P.J., O'Neal, I E . (1984), Metall. Trans. 15 A, 1451-1463. Savage, S. I, Froes, F.H. (1984), J. Metals 36, No. 4, 20-33. Schaefer, R.J., Bendersky, L. A., Biancaniello, F. S. (1986), /. Physique (Paris) 47, C3-311-320. Sepold, G., Becker, R. (1986), in: Science and Technology of the Undercooled Melt: Sahm, P. R., Jones, H., Adam, C M . (Eds.), Dordrecht, NL: Martinus Nijhoff, p. 112. Serita, Y, Ishikawa, T., Kimura, F. (1970), J. Jap. Inst. Light Metals 20, No. 1, 1-6. Seshadri, R., Krishna Rao, R.V., Krishnan, R.V., Mallya, R.M. (1988), J. Mater. Sci. 23,1637-1642. Shechtman, D., Blech, I., Gratias, D., Cahn, J.W. (1984), Phys. Rev. Lett. 53, 1951-1953. Shepelskii, N. V., Zhilkin, V.Z. (1969), Soviet Powder Met. Mater. Ceram. 10, 813-818. Shingu, P.H., Ozaki, R. (1975), Metall. Trans. 6A, 33-37. Shingu, P.H., Shimomura, K., Ozaki, R. (1979), Trans. Jap. Inst. Metals 20, 33-35. Shingu, P.H., Takamura, J., Kawashima, M. (1968), Suiyokai-Shi 16, All. Singer, A.R.E. (1968), British Patent No. 1262471. Singer, A.R.E. (1970), Metals & Materials 4, 246257. Singer, A.R.E. (1978), British Patent No. 1517283. Singer, A.R.E. (1982), Powder Met. 25, 195. Singer, A.R.E. (1983), Materials & Design 4, 892. Singer, A. R. E., Kisakurek, S. E. (1976), Metals Technol. 3, 565-570. Singer, A.R.E., Roche, A.D. (1977), in: Modern Developments in Powder Metallurgy, Vol. 9: Hausner, H.H., Taubenblaut, P.V. (Eds.). Princeton, NJ: MPIF, pp. 127-140. Singer, A. R. E., Roche, A. D. (1980), Powder Met. 23, 81. Smith, P. (1985), Metal Powder Rep. 40, 159-161. Soderhjelm, H., Mandal, L. (1985), in: Rapidly Quenched Metals V; Steeb, S., Warlimont, H. (Eds.). Amsterdam: Elsevier Sci, Pub,, B,V,, pp. 107-110. Steen, W.M. (1985), Metals & Materials 1, 730. Steinberg, I, Lord, Jr., A.E., Lacy, L.L., Johnson, J. (1981), AppL Phys. Lett. 38, 135-137. Stewart, O.M., Maringer, R. E., Mobley, C.E. (1974), U.S. Patent No. 3812901.
109
Strachan, R.W. (1967), Sc.D. Thesis, M.I.T., Cambridge, MA. Strange, E.A., Pirn, C.H. (1908), U.S. Patent No. 905 758. Strutt, P.R., Kurup, M., Gilbert, D.A. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B. H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 225-236. Suryanarayana, C. (1980), Rapidly Quenched Metals: A Bibliography 1973-1979. New York: IFI Plenum. Suryanarayana, C , Anantharaman, T. R. (1970), J. Mater. Sci. 5, 992-1004. Suryanarayana, C , Jones, H. (1988), Internat. J. Rapid Solidification 3, 253-293. Sutcliffe, P.W., Morton, P.H. (1976), in: Advanced Fabrication Techniques in Powder Metallurgy and Their Economic Implications, AGARD Conf. Proc. No. 200, Paper SC.3; see also (1981), Metal Powder Rep. 36, 84. Takeshita, K., Shingu, P.H. (1983), Trans. Jap. Inst. Metals 24, 529. Takeshita, K., Shingu, P.H. (1986), Trans. Jap. Inst. Metals 27, 141. Taylor, G.F. (1924), Phys. Rev. 23, 655-660. Tenwick, M.J., Davies, H.A. (1984), Mater. Sci. & Eng. 63, L1-L4. Tonejc, A., Bonefacic, A. (1969), Trans. Met. Soc. AIME 245, 1664. Tracey, V.A., Cutler, C.P. (1981), Powder Met. 24, No. 1, 32. Tucker, T. R., Ayers, J. D. (1980), in: Rapid Solidification Processing: Principles and Technologies II: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 206211. Tuli, M., Strutt, P.R., Nowotny, H., Kear, B.H. (1978), in: Rapid Solidification Processing: Principles and Technologies: Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitor's Pub. Div., pp. 112-116. Tutzauer, H., Esquinazi, P., De La Cruz, M.E., De La Cruz, F. (1980), Rev. Sci. Instrum. 51, 546-547. Unal, A. (1990), Internat. J. Powder Met. 26, 11-21. Vincent, J. H., Herbertson, J. G., Davies, H. A. (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto, T, Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. Metals, pp. 77-80. Vinet, B., Cortella, L., Favier, J.J., Desre, P. (1991), AppL Phys. Lett. 58, 97. Von Allmen, M. (1983), in: Glassy Metals II: Beck, H., Giintherodt, H.-J. (Eds.). Berlin: Springer-Verlag, p. 261. Von Allmen, M., Huber, E., Blatter, A., Affolter, K. (1984), Internat. J. Rapid Solidification 1, 15-25. Wahlster, M., Stephan, H., Ruthard, R. (1980), Powder Met. Internat. 12, 113-111. Wang, R. (1970), Rev. Sci. Instrum. 41, 1233-1234. Warrington, D.H., Davies, H. A., Shohoji, N. (1982), in: Rapidly Quenched Metals IV, Vol. I: Masumoto,
110
2 Rapid Solidification
T., Suzuki, K. (Eds.). Sendai, Japan: The Jap. Inst. view. Aedermannsdorf (Switzerland): Trans. Tech. Publ. Metals, pp. 69-72. Cahn, R. W. (1983) in: Physical Metallurgy, Vol. 2, Wentzell, J.M. (1974), /. Vac. Sci. & Tech. 11, 1693rd ed., Cahn, R.W, Haasen, P. (Eds.). Amster171. dam: Elsevier Sci. Pub., pp. 1779-1852. Whang, S.H. (1984), /. Metals 36, No. 4, 34-40. Whang, S.H., Giessen, B.C. (1983), in: Rapid SolidiCochrane, R.W, Strom-Olsen, J.O. (Eds.) (1988), fication Processing: Principles and Technologies HI: Rapidly Quenched Metals 6. London and New York: Elsevier Appl. Sci. Mehrabian, R. (Ed.). Gaithersburg, MD: National Herman, H. (Ed.) (1981), Ultrarapid Quenching of Bureau of Standards, pp. 439-442. Liquid Alloys, Vol. 20; Treatise on Materials SciWillens, R.H., Beuhler, E. (1966), Trans. Met. Soc. ence and Technology. New York: Academic Press. AIME236, 111-114. Williams, C.A., Jones, H. (1974), Metals Technol. 1, Jones, H. (1982), Rapid Solidification of Metals and 202-203. Alloys, Monograph No. 8. London: The InstituWood, J. V., Boulby, K. A. (1986), Metal Powder Rep. tion of Metallurgists. 41, 299-301. Mehrabian, R., Parrish, P. A. (Eds.) (1988), Rapid Yamaguchi, T., Narita, K. (1977), IEEE Trans. Magn. Solidification Processing: Principles and Technologies IV. Baton Rouge, LA: Claitor's Pub. Div. MAG-13, 1621-1623. Yeh, X.L., Samwer, K., Johnson, W.L. (1983), Appl. Sahm, P.R., Jones, H., Adam, C M . (Eds.) (1986), Phys. Lett. 42, 242-244. Science and Technology of the Undercooled Melt. Zboril, J., Posedel, Z. (1970), Z. Metallkde. 61, 214Dordrecht, NL: Martinus Nijhoff. 217. Steeb, S., Warlimont, H. (Eds.) (1985), Rapidly Quenched Metals V. Amsterdam: Elsevier Science Pub., B.V. General Reading Suryanarayana, C. (1980), Rapidly Quenched Metals: Anantharaman, T. R., Suryanarayana, C. (1987), A Bibliography 1973-1979. New York: IFI PieRapidly Solidified Metals: A Technological Over-
3 Surface Modification by Lasers Barry L. Mordike Institut fur Werkstoffkunde und Werkstofftechnik, Technische Universitat Clausthal, Clausthal-Zellerfeld, Federal Republic of Germany
List of 3.1 3.2 3.2.1 3.2.2 3.2.2.1 3.2.2.2 3.2.2.3 3.3 3.4 3.4.1 3.4.2 3.4.3 3.4.4 3.4.5 3.4.6 3.4.6.1 3.4.6.2 3.4.7 3.5 3.5.1 3.5.2 3.5.3 3.6 3.6.1 3.6.1.1 3.6.1.2 3.6.2 3.7 3.8 3.9
Symbols and Abbreviations Introduction Fundamental Considerations Reflectivity and Absorption Methods of Increasing Absorptivity Preheating Surface Coatings Surface Roughening Applications of CO2 Lasers Surface Modification Processes Transformation Hardening in Iron-Base Alloys Surface Melting Constitutional Effects Surface Melting of Cast Iron Surface Melting of Aluminium-Silicon Alloys Surface Alloying and Surface Cladding Surface Alloying Laser Cladding Surface Treatment of Ceramic Materials Wear Properties of Laser-Treated Surfaces Transformation Hardening Surface Melting and Surface Alloying Cladding Other Effects of Laser Treatment Internal Stresses Transformation Hardening Stresses Cladding Stresses Effect on Bulk Properties, in Particular, Fatigue Life Industrial Applications Acknowledgements References
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
112 113 113 113 114 114 114 114 114 115 116 117 117 120 121 122 123 126 127 129 129 129 130 131 131 131 131 132 133 135 135
112
3 Surface Modification by Lasers
List of Symbols and Abbreviations A
c A ,c B D G
T T £
r HV On NdYAG
absorptivity equilibrium solubilities liquid and solid concentrations diffusion constant temperature gradient intensity distribution coefficient temperature melting temperature quenching rate Gibbs-Thomson coefficient Vickers hardness under load 0. neodymium aluminium garnet
113
3.2 Fundamental Considerations
3.1 Introduction Lasers have been used for several years in industry for cutting and welding applications. More recently, with the advent of high power lasers, not only CO 2 but also neodymium aluminum garnet (NdYAG) and excimer lasers, attention has been paid to developing methods of surface modification of components. The problems in introducing laser surface treatment into industry are three-fold; obtaining a surface structure capable of withstanding the conditions obtaining in service (mechanical loading, chemical attack, etc.); development of handling systems to treat the complicated shapes of components; establishing the deleterious effects of laser surface treatment, if any, on the bulk properties of the components, e.g., the effect of residual stresses on the fatigue life and treatment at an economical price. Nevertheless some applications have been successfully taken up by industry, although mainly for high cost components, turbine blades (Mac Intyre, 1986), ships' diesel cylinder liners (Amende, 1990). The present article discusses the applications of CO 2 lasers to the surface treatment of materials. There are three main aspects - fundamental considerations, surface treatment of components and properties of laser-treated surface and components. Some general aspects of laser surface treatment are also discussed in Chap. 2, Sec. 2.5.
to the lattice. Photons are absorbed and electrons are excited into higher energy states. Assuming that the incident beam energy is insufficient to expel an electron, it imparts its energy to the lattice and lattice defects, thus increasing the lattice vibrations or temperature. There are many factors which limit the overall efficiency of the photon energy -• heat energy transfer. The first stage is the absorption of the photons. Here losses may occur owing to reflectivity or plasma formation. 3.2.1 Reflectivity and Absorption
The reflection losses are often the limiting factor in overall efficiency in many laser applications in the processing of materials. The absorptivity A = {\-reflectivity). The heating effect of a beam with a known intensity Io (watts m~ 2 ) is AI0. Although there are methods of calculating the reflectivity (Johnson and Christy, 1975; Shavarew etal., 1978) and hence the absorptivity A, the simplest approach is to measure A directly during laser excitation. Figure 3-1 shows the wavelength dependence of the absorptivity A for some metals. The wavelengths of CO 2 , NdYAG and excimer lasers are also shown. It is immediately apparent that the absorption
0.7 0.6
C02Fe--
e-o.3
**v \ .
o U) _Q
The initial stage in all laser processing applications involves the coupling of laser radiation to the material. The majority of applications involve metallic materials and the electrons ensure the transfer of energy
- NdYAG
0.5
>
3.2 Fundamental Considerations
i
— Excimer
<
0.2
Stainless steel
l-Cu
0.1 0.0
0.1
1 A in pm
Figure 3-1. Wavelength dependence of absorptivity.
114
3 Surface Modification by Lasers
of CO 2 (infrared) light, of wavelength 10.6 jim is significantly less efficient than that of the NdYAG and excimer lasers. This is especially marked in the case of Cu and Al. The wavelength dependence of the absorptivity suggests that certain applications might better be carried out with NdYAG or excimer lasers. Now that higher power NdYAG and CO 2 lasers are becoming available, this possibility may be realizable. 3.2.2 Methods of Increasing Absorptivity 3.2.2.1 Preheating Increasing the absorptivity is necessary for many applications, particularly with CO 2 lasers. As A usually increases with temperature, preheating by conventional heat sources may improve the efficiency of laser treatment. Independently of this, there may be other advantages in preheating. 3.2.2.2 Surface Coatings Oxide coatings or deliberately added coatings have a significantly different composition from the metal substrate. The change in reflectivity can be calculated if the optical constants are known (Heavens, 1965). Such effects depend also on the wavelength. Nevertheless, several coatings have been found to be successful for IR wavelengths.
defects and not the surface topography per se and these anneal out at elevated temperatures (Weiting and De Rosa, 1979). A plasma on the surface dramatically increases the coupling of energy. If too high a power is used, however, the plasma lifts off from the surface and effectively shields the surface from the incident beam.
3.3 Applications of CO 2 Lasers Figure 3-2 shows the fields of application of lasers according to the energy density of the laser and interaction time. The surface treatments lie on a straight line ranging from transformation hardening to shock hardening. Transformation hardening and melting rely on the introduction of heat into the material whereas in shock hardening the laser pulse induces an elastic wave which causes local plastic deformation. The important parameters in laser treatment are the energy density and interaction time. There are limitations in the choice of parameters depending on the application required. This is illustrated in Figs. 3-3 and 3-4.
3.2.2.3 Surface Roughening It has been observed that the surface topography has an effect on the absorption. Roughening the surface by sandblasting or emery cloth increases A by an amount of the order of 30%. This improvement disappears at elevated temperatures. The improvement observed in A is probably due to the formation of surface
10~8
10~6 10"A Interaction time in s
10~2
10°
Figure 3-2. Applications of lasers according to energy density and interaction time.
115
3.4 Surface Modification Processes 3500
3500 A P . 6 0 0 0 W ; P// melted depth = 2.04nm vaporised = 40}jm
3000
A
2
o P= 3000 W ; P/A = 3820 W-cnrff melted depth= 0.96mm vaporised = 0 pm • P= 1000 W ; P/A = 1273 W • cm" 2 .melted depth = 0 mm vaporised = 0 pm
2500
2 2000
t = 10S melted depth = 3.31mm
3000 o t = 3.33 s melted depth = 0.18mm 2500 Df=1.0s melted depth = 0.0mm
M500
1000^
500 293 0.00
0.80
1.60 2.40 Depth x in mm
3.20
4.00
0.00
0.80
1.60 2.40 Depth x in mm
3.20
4.00
Figure 3-3. Temperature profile for cast iron for various energy densities at constant interaction times.
Figure 3-4. Temperature profile for iron for various interaction times at constant energy densities.
These figures show the temperature profile generated in a material by a laser beam 10 mm in diameter for various energy densities and interaction times. The calculations were for simplification made for onedimensional heat flow. Figure 3-3 shows the temperature profiles for cast iron for 3 different energy densities for an interaction time of 1 second. It is apparent that for interaction times of 1 second there is a limit to the power density which can be used. 40 jim of the surface have evaporated for the highest power. Reducing the energy density reduces the temperature gradient and also the depth to which the heat penetrates. The broken line shows the melting point of cast iron. The depth melted for the higher density was about 2 mm whereas at the lower energy density about 1 mm is melted. At lower energy densities still, no melting occurs. Figure 3-4 shows the influence of interaction time. A low energy density was chosen. Increasing the interaction time displaced the curves to higher temperatures.
The conclusions which can be drawn are: the penetration depth at high energy densities is limited by evaporation from the surface. Long exposure times and low energy densities are required for thick heat treated layers. Other low energy density sources are much cheaper than lasers and may be able to carry out such treatments satisfactorily. The laser is thus more likely to be applied for applications where relatively thin heat-treated layers are required.
3.4 Surface Modification Processes The following processes have been developed for treatment of surfaces: • transformation hardening in iron-base alloys • surface melting • surface alloying • cladding • amorphisation.
116
3 Surface Modification by Lasers
3.4.1 Transformation Hardening in Iron-Base Alloys
Transformation hardening of iron base alloys involves austenitization and selfquenching. For the purposes of discussion it will be assumed that the coupling of the laser energy into the workpiece is assured and that it is massive enough to ensure quenching. Laser treatment involves rapid heating with a short austenitization time. On subsequent rapid cooling the hardening will depend on the degree of austenitization which had been achieved. The results can thus be significantly different from those obtained for the alloy using conventional hardening processes. This is illustrated by the time-temperature austenitization curve (Fig. 3-5) (Orlitsch etal., 1973). For the composition chosen the areas for homogeneous austenite, inhomogeneous austenite, ferrite + pearlite + austenite are shown. It is obvious that the rapid rate of heating effectively
1300
2400
Heating rate in °C-s~ 1 260 65 13 2.6 0.55
0.05
Steel B 1200
homogeneous austenite
: 1100
B 1000
900
800
700 10 seconds Time
10°
1
10
102 minutes
Figure 3-5. Time-temperature austenitization curve for a low chromium steel. (After Orlitsch et al., 1973).
displaces the transformation temperatures to higher temperatures. This can be by as much as 100-150 °C. The time spent above A 3 is very short and this limits carbon diffusion and dissolution of carbides resulting in a heterogeneous structure. The initial micro structure is very important in determining the suitability for laser transformation hardening. Microstructural heterogeneities, coarse carbides or pearlite exacerbate the problem of homogenization. Ideally, the microstructure in addition to the composition should be specified for alloys which are to be laser transformation hardened. In practice, the beam must be defocused to obtain a spot of the required size. In some cases, special optics are necessary to obtain rectangular spots with a homogeneous energy distribution. The energy acts only for a short time and hence the beam must be moved accurately and with a constant speed. The depth of hardening and the temperature profile are determined by choice of energy density and spot speed. The method can be applied to all materials which can undergo a martensitic transformation with the limitation imposed by the effect of rapid heating. Steels or irons with largely pearlitic or heat treated matrices are better suited than those with high fractions of ferrite or stable carbides. Figure 3-6 shows the heating curves for different depths below the surface. These curves are dependent on the energy density. Figure 3-7 shows schematically the nature of the hardened zones. If higher energy densities are used, not only is there a bigger difference in heating rates between the surface and interior but where higher rates of heating are experienced the transformation temperatures are displaced to higher temperatures. Figure 3-8 shows the hardness profile for a transformationhardened grey cast iron. Transformation
3.4 Surface Modification Processes
117
1200 1100 1000
| \ ^
900 800 o c 700 0)
B 600 o
I 500
*~ U)0 300 200 100
i \m /
transformation temperature (rapid heating)
250 500 750 Distance from surface in pm
\
Figure 3-8. Hardness profile for a transformationhardened grey cast iron.
cycle
cooling cycle
3.4.2 Surface Melting
0 0
0.2
0.4
1000
0.6
0.8 1.0 Time in s
1.2
1.4
1.6
1.8
Figure 3-6. Heating curves for different depths below the surface.
Applications involving surface melting rely for improvement of the surface properties on one or more factors. In the case of cast materials, remelting the surface enables casting defects (e.g., porosity, cold shuts) to be eliminated. The main reason for laser surface melting is to obtain better properties as a result of the rapid solidification. The quenching rates can be as high as those achieved in other techniques, e.g., melt spinning, powder atomization (Pond et al., 1976; Lawley, 1977). The aim of surface melting is thus to reproduce the improvements obtained in rapidly quenched powders and ribbons on massive materials. 3.4.3 Constitutional Effects
Distance from surface
Figure 3-7. Micro structural changes at different depths for a given heating rate.
hardening has been successfully applied to carbon and alloy steels, tool steels, case carburized steels and cast irons (Amende, 1990).
Rapid quenching influences the nature and stability range of phases, also their size, morphology and distribution. The final microstructure depends on the process parameters and obviously on the nature of the system. The phases which are produced may be non-crystalline (amorphous) or crystalline (stable or metastable). Independently, there may be an extension in the stability range of solid solubility.
118
3 Surface Modification by Lasers
Equilibrium diagrams can only be employed as an aid when considering the possible outcome of laser surface treatment. The equilibrium diagram, Fig. 3-9, shows the relationship between free energy and composition and the resultant solubility limit for a chosen temperature 7\, for a eutectic system. The tangent defines the equilibrium solubilities CA and CB for the terminal solid solutions a and /?. The intersections of the liquid and solid free energy curves show the corresponding maximum possible metastable solubilities for a and /?. The locus of C c and CD provides the well known To curve. It represents the highest interface temperature at which the partition coefficient can be unity (Jones, 1973). Generally, further kinetic undercooling is necessary for partitionless solidification (solute trapping). Extension of solid solubility must be avoided if an amorphous structure is to be possible. Figure 3-10 shows the range of compositions of solid solution which can form from a liquid at the interface of composition CL. In case (a), partitionless solidification is possible but not in case (b) (Boettinger, 1981). How the solidification in fact proceeds depends on the kinetic conditions, the speed with which the solidification front moves. This is determined by the relative speed of laser beam and workpiece. The conditions in laser melting are such that growth occurs usually by cellularly or dendritic mechanisms, which involve segregation. At sufficiently high growth rates the solidification front remains planar and segregation-free growth is possible. The condition for this is that the speed exceeds D AT/FK0 where F is the Gibbs-Thomson coefficient (ratio of Interfacial Energy: Entropy of Melting), D is the diffusion coefficient, AT is the temperature difference between liquidus and solidus, Ko is the distribution coefficient (Coriell and Skeraka, 1980). Such growth
C
Composition
HCB 1 Q
Figure 3-9. Relationship between free energy and composition at the temperature T (schematic).
Composition
Figure 3-10. The shaded areas show the range of solid solubility Cs which can form from a liquid at the interface of composition CL at different interfacial temperatures. Solidification without segregation is possible in case (a) but not in case (b).
3.4 Surface Modification Processes
119
Figure 3-12. Effect of energy density on the microstructure for a Fe-13%Cr-2.1%C steel, a) Low energy density, b) High energy density. Transverse speed constant.
Figure 3-11. Different nucleation modes in lasermelted surfaces; a) from substrate (Mag. 25 x); b) within the melt (Mag. 100 x); c) from the surface (Mag. 50 x).
rates are possible with lasers or electron beams. The microstructure which is produced in laser surface melting is the result of many competing processes. There is nucleation, which can occur either heterogeneously on the substrate or within the melt, or homo-
geneously. Thereafter, there are the various possible forms of growth. Examples of different nucleation forms are given in Figs. 3-1 la, b and c (Bergmann et al., 1984). The solidification behaviour is influenced mainly by three parameters quenching rate, s, solidification rate, R, and the temperature gradient, G. These three parameters are interrelated in the expression 8 = RG. The temperature gradient in laser surface melting depends on the energy density and the solidification rate, and apart from stationary pulsing, on the displacement rate. Figure 3-12 shows the effect of increasing the energy density whilst maintaining the solidification rate constant.
120
3 Surface Modification by Lasers
The influence of increasing the solidification rate while maintaining the energy density constant is shown in Figs. 3-13 a, b, c, d. The type of growth changes from near planar to cellular to dendritic. The mechanical properties are correspondingly different. 3.4.4 Surface Melting of Cast Iron Iron-carbon alloys can solidify according to the metastable iron-Fe3C phase diagram or the iron-graphite equilibrium phase diagram. The relative stability of cementite and graphite is affected by alloying additions. Most cast irons are grey cast irons, containing graphite in a matrix which is usually pearlite, or ferrite. Increasing the cooling rate favours the formation of cementite, a fact used in the production of chill-hardened irons. If grey cast iron is surface-melted, the layer solidifies rapidly and thus in the form of ledeburite - i.e., austenite and cementite. The properties, particularly wear resistance, and strength are much better than those of grey cast irons. Simple surface remelting can thus convert a cheap material with relatively poor properties into a high performance material. There are many components made from cast iron, e.g., cam followers, brake drums, supports, engine blocks, camshafts, cylinder liners. The loading and operating conditions are very different but in each case surface melting can improve the wear resistance considerably. Figures 3-14 a and b show the structure obtained on remelting lamellar and spheroidal cast iron, respectively.
Figure 3-13. Different microstructures in lamellar cast iron surface remelted with increasing solidification rate and constant temperature gradient; (a) has the lowest solidification rate, and (b) the highest.
3.4 Surface Modification Processes
121
Depending on the conditions of melting, the thickness and hardness of the layer can be determined (Fig. 3-15). The heat-affected zone also depends on the parameters and also on the type of cast iron used, e.g., ferritic or pearlitic. This zone can show martensite which can induce cracking in the layer on cooling or during subsequent grinding operations. It is customary to preheat to about 350 °C before laser treatment to prevent martensite forming. 3.4.5 Surface Melting of Aluminum-Silicon Alloys
Figure 3-14. Surface-melted lamellar (a) and spheroidal (b) cast irons (Mag. 100 x).
Aluminum-silicon alloys are widely used as casting alloys in the automobile and related industries, in particular Al-13Si or Al-(8 Si + 3 Cu). They rely for their strengthening on the eutectic precipitation of silicon and primary aluminum. The silicon phase is coarse and acicular and to improve the properties it is customary to add sodium to the melt to "modify" the silicon morphology. Rapid cooling produces a similar effect.
Figure 3-15. Hardness profile of lamellar cast iron remelted with different laser parameters. 500
1000
1500
Distance from the surface in pm
2000
122
3 Surface Modification by Lasers
200
£00 600 800 1000 Distance from surface in urn
1200
Figure 3-17. Hardness profile corresponding to Fig. 3-16.
Figure 3-16. Surface-melted aluminium-silicon alloy (Mag. 50 x).
Surface melting produces a very fine eutectic on the surface of both modified and unmodified Al-Si alloys (Pelletier et al., 1989; Mordike, 1989). Figure 3-16 shows an example of surface melting with unmodified A1-13SL The surface was covered by a raster and the individual tracks can be discerned. No difficulties were encountered in laser treatment. The use of a line focus instead of a point focus improves the coupling and also the surface quality. The hardness profile is shown in Fig. 3-17. 3.4.6 Surface Alloying and Surface Cladding Surface melting alone is often insufficient to provide the properties required. Consequently it is necessary to change the composition of the surface. If this is done
by elements soluble in the bulk material and accomplished during laser surface melting, then it is termed laser surface alloying. Laser dispersion hardening is the injection of insoluble particles into the laser melted surface layer. If the properties are still insufficient then the surface must be clad, i.e., a different alloy must be melted onto the surface much in the same way as deposits are laid on surfaces using standard welding procedures. When a laser is used, this is termed laser-cladding. It is convenient to discuss these techniques together as the processes are very similar. A change in the surface composition can be brought about by melting a deposit on the surface into the surface (laser-alloying) or onto the surface with minimal substrate melting (laser-cladding). These are two-stage processes and involve first coating the surface and then laser-treating it. The initial coating can be applied by thermal spraying, electrodeposition or affixing foil to the surface, for example, The thickness of the coating is chosen so that when the desired depth is laser melted, the mixing with the substrate provides the desired composition in the
3.4 Surface Modification Processes
123
. laser \ beam
(a)
. , ,.
.
,
wderinjection mjeCTK
<'.y/:/
cladding ^^Uykv
nozzle
- Feed direction
yx\ laser , ^ Y/} beam ' / / ^— nozzle
(b)
y
y\ r
shield^ gas remelted layer
coating
^
^
' / 7 / , , , substrate /
molten pool /
/
Feed direction
Figure 3-18. Schematic drawing of a) one- and b) two-stage coating processes.
Figure 3-19. Powder injection head.
surface layer. The alternative is to add the alloying element or dispersed phase at the same time as the surface is melted as part of a single-stage process. The alloying element may be added in the gaseous form, powder or wire. Hard metals can be added by injecting powder into the surface with or without bond metal, with or without coating or enclosing it in wire. Figure 3-18 shows the two processes schematically. It is possible to coat or alloy virtually any substrate with a very wide range of materials. Laser coating/alloying is thus an extremely versatile and potent technique in matching component performance to service requirements. Success depends critically on the coating parameters and coating devices used. Figure 3-19 shows a powder injection head. The angle at which the powder strikes the surface can be varied.
This determines also the flight time in the laser beam and hence the degree of preheating. The rate of powder feed can be continuously varied. Typical rates are 2.5 g min" 1 . Powder injection in various forms has become a standard technique at least in the laboratory or small scale production. The alternative method of supplying the addition in the form of wire or filled wire is more elegant and less wasteful of material. The refinement of preheating the wire (Burchards and Hinse, 1990) provides the opportunity of a significant increase in deposit rate and hence economic viability. 3.4.6.1 Surface Alloying Examples will be discussed of alloying from the gas phase or by melting a deposited layer into the substrate. These examples
124
3 Surface Modification by Lasers Ti-N
suffice to illustrate the potential of the method. There are numerous other examples and the reader is referred to the literature.
3000
Laser Gas Alloying
2600 -
Laser gas alloying is the term applied to changing the composition of the surface by melting it in a stream of gas which will react with the substrate material. Solid state reactions are slow and hence it is necessary to melt the surface. Laser gas alloyed layers are much thicker than those obtained by solid state reactions even after long treatment times. This has particular advantages for some metals, e.g., titanium. Although titanium transforms martensitically on quenching and although this can be influenced by alloying, no improvement in properties is observed, as is the case for many iron-based alloys. An alternative means of improving the strength must be sought, e.g., by a dispersion of a second hard phase. By laser melting in the presence of nitrogen nitride dendrites precipitate out, producing a surface layer consisting of TiN in an a-Ti matrix (Mordike, 1987). The process of gas alloying can be employed for several systems, e.g., Ti-N, Ti-C, Zr-C, Fe-C. The precise mechanism of absorption/desorption depends in each particular case on the nature of the metalgas system. The behaviour can in general be described by the law of mass action. In applying this law to gaseous systems, the concentrations can be replaced by the partial pressures. In the case of nitriding titanium, nitrogen is mixed with an inert carrier gas. The partial pressure of nitrogen and the temperature and nitrogen concentration in the titanium will determine the rate and direction of the reaction. The phase diagrams for TiN is shown below Fig. 3-20. TiN forms at tempera-
2
5
10
15
22.6
2200 -
2 1800
uoo 1000
600 0
10
20
30
40
50
60
Figure 3-20. Phase diagrams for systems Ti-N.
Figure 3-21. Microstructure of laser-gas alloyed Ti.
3.4 Surface Modification Processes
0.66-105W cm 2 RT 40Vol.-% N2 60Vol.-% Ar overlap 75%
125
• 0.5 m/min A * D •
1 m/min 2 m/min 3 m/min 5 m/min
Figure 3-22. Relationship between TiN layer hardness and displacement rate. 100
200 300 400 500 600 700 800 Distance from the surface in pm
tures in excess of 2600 °C. The microstructure of the nitrided surface is shown in Fig. 3-21. The dendrite size depends on the partial pressure of nitrogen and interaction time, which can be the sum of several passes. Under appropriate conditions a continuous TiN layer can be produced. This can pose problems in practice with thick layers due to cracking. Figure 3-22 shows the relationship for a given nitrogen partial pressure between thickness of the layer as determined by the hardness and displacement rate. Surface Alloying of Aluminum Alloys Aluminum alloys, apart from Al-Si, suffer from poor wear resistance and even Al-Si alloys have insufficient load-bearing capacity. The problem is thus initially one of increasing the strength. It is found that the increase in strength obtained by simply remelting Al-Si alloys was only modest. There is no possibility of increasing this as changing the silicon content produces unacceptable changes in the bulk properties. Any increase in the number of dispersed particles must come from direct addition or reaction with aluminum. Many metals form intermetallic compounds with aluminum. Melting a metal coating, galvani-
900 1000
cally or plasma-deposited, into the surface of Al alloy will produce a fine dispersion of intermetallic precipitates throughout the melted layer. The addition of nickel has been investigated by several groups (Pelletier et al., 1989; Mordike and Veit, 1988; Bergmann etal., 1988). Three intermetallic phases are likely to form, Al3Ni, Al 3 Ni 2 and Al-Ni. Their structure and composition range are given in Table 3-1. Figure 3-23 shows the surface of AlSi9Cu surface alloyed with 19.6 at. %Ni. Increasing the nickel content to 25% increases the hardness in the surface layer to values between 900 and 1000 HV. This results in significant increase in load bearTable 3-1. Intermetallic compounds in Al-rich Al-Ni alloys. Intermetallic compound
Crystal Composition structure at.% Ni
Al3Ni
orthorhombic
25.0
Al3Ni2
hexagonal
36.3-40.8
B2 type
45-60
AINi
Melting point Melts incongruently at 854 °C Melts incongruently at 1120°C 1638°C
126
3 Surface Modification by Lasers
'"•"' -Vv<
Figure 3-23. AlSi9Cu layer alloyed with 19.6 at.% Ni.
Figure 3-24. Laser-melted deposited WC-Co layer. Substrate Ck45 steel (Mag. 100 x).
ing capacity and at the same time a reduction in abrasive wear. 3.4.6.2 Laser Cladding
The most commonly applied cladding materials are stellites or hard metals (Steen, 1987). They are applied by one of the methods previously discussed. The method initially adopted was to laser-melt plasmadeposited layers. Ideally the laser power should be set to melt the deposited layer onto the substrate so as produce a good bond without dilution of the layer. In the case of hard metals, it is unavoidable that the carbide particles are largely dissolved, which may deleteriously affect properties. Figure 3-24 shows the structure of a surface-melted WC-Co plasma-deposited layer. The structure apart from individual undissolved carbide is very fine, essentially free of pores and crack free. The final thickness of such layers is typically less than 1.0 mm. Increasing the thickness increases the tendency to cracking. This limits the maximum thickness possible with the technique. The average hardness of the layer is about 650HV 0<1 (Fig. 3-25). The layers are extremely well bonded to the substrate and exhibit good thermal shock and impact resistance.
500
1000
1500
2000
2500
Distance from the surface in urn
Figure 3-25. Hardness profile of a laser-melted WCCo layer.
Figure 3-26 shows a layer of TiC lasermelted on to a carbon steel substrate. The difference in structure from that obtained by melting predeposited coatings is apparent. The carbides are retained rather than dissolved. The structure is therefore a true composite with hard particles in an essentially ductile matrix. To prevent oxidation, the carbide particles can be coated, in the above case by nickel. The TiC particle size in the powder used was 45-90 jim. After spraying and laser melting the size measured (circular intercept) was about 10% smaller. This ignored the fine reprecipitated particles in the matrix which became visible at high magnifi-
3.4 Surface Modification Processes
127
3.4.7 Surface Treatment of Ceramic Materials
Figure 3-26. TiC particles in a Ni matrix, optimum melting conditions. Substrate, carbon steel (Mag. 100 x). Table 3-2. Hardness values of TiC-Ni layer. Microstructure TiC particles Ni-rich phase Heat-affected zone Substrate (CK45)
HV0.3
2300 470 800 230
cations. Table 3-2 lists the hardness values measured for the various phases and zones. Martensite has formed in the heat-affected zone. The macro hardness value of 650 HV0 3 in this layer shows this. Cobalt alloys are suitable layers to combat wear, particularly where high temperatures are involved. Stellites are easy coating materials and find wide application for valve seats. The layers produced have a fine dendritic structure with a hardness of ^470 HV (Mordike and Burchards, 1989). Injection of powder on to the surface or laser wire cladding of metals has similarly been studied and perfected by several groups (Steen, 1987; Amende, 1991; Becker etal., 1988; Burchards and He, 1991; Kahrmann, 1988; Mordike and Burchards, 1989). The reader is referred to the literature for detailed information.
Plasma-sprayed ceramic coatings on hot-section components of gas turbines offer significant advantages by increasing the operating temperatures and thus efficiency. Research is directed to improving these coatings. Most of the coatings are based on zirconia as it has a lower thermal conductivity and higher coefficient of thermal expansion than other ceramics. The main life-limiting factor of such plasma-sprayed coatings is oxidation of the metallic bonding coat by corrosive gases and molten salts able to penetrate the interconnected porosity of sprayed coatings. The metallic coat ensures a good bond to the ceramic coating and should also limit corrosion. Porosity in sprayed coatings is unavoidable and in many respects desirable since porosity increases the thermal insulation and resistance to thermal cycling. Surface melting of porous layers would improve the performance considerably. The densification of the surface layer would prevent access to gas whilst at the same time retaining the heat insulation of the porous substrate. Various workers have reported results on surface melting of ceramics (Galasso and Veltri, 1983; Zaplatiynsky, 1982; Iwamoto, 1982; Adamski and McPherson, 1986; Havrda et al., 1986). The problem is crack formation due to anisotropic thermal expansion combined with low toughness. Apart from sealing the pores, laser-melting can provide a smoother surface and hence reduce aerodynamic losses. Sivakumar and Mordike (1988) studied the problem of crack formation in surface melting of ceramics based on ZrO 2 , A12O3 and TiO 2 . The tendency to cracking can be reduced by using the pulsed mode and high energy densities as this is more suitable for melt-
128
3 Surface Modification by Lasers
ing a thin layer with a high temperature gradient. The rate of cooling during solidification is directly proportional to the energy density but an upper limit is imposed as the surface temperature should not exceed the boiling point. In the case of all three ceramics, surface melting on specimens at room temperature was not possible without some cracking. ZrO 2 was worst and A12O3 best. In all cases, melting thin layers produced transverse cracks which can possibly be tolerated. On increasing the thickness, a point is reached when longitudinal cracks form. These facilitate spalling and can not be tolerated under any circumstances. Figure 3-27 shows a cross-section of a surface-melted layer of ZrO 2 -7Y 2 O 3 . The problem of cracking arises mainly from the non-uni-
ta)
(b)
Figure 3-27. Cross sections of laser-melted ZrO27Y 2 O 3 . a) Thin layer melted on top. b) Entire coating melted (Mag. 25 x).
form cooling of the melted spot which produces elastic thermal stresses. At temperatures above ^0.5 Tm these stresses can be relieved by plastic deformation, but below this temperature they build up on cooling and since they are tensile in character cause fracture when the fracture strength is exceeded. On the basis of considerations of cooling of thin discs it is possible to estimate the maximum stress developed, using known or estimated values of the coefficient of thermal expansion, the modulus of elasticity and the range of temperature during cooling in which the stresses cannot be relieved. If the surface is preheated, the temperature range in which the stresses can build up can be reduced. For example, for A12O3, which is plastic above 1000°C (Dorre and Hubner, 1984), treatment with specimens at room temperature would result in a maximum stress of 1488 MPa, which is much higher than quoted values of the fracture strength (220 to 310 MPa). By preheating to 800 °C, the effective cooling range is reduced to 200 and the maximum stress developed is only « 305 MPa. No cracking is observed in surface melting of A12O3 preheated to 800 °C. Similar calculations for ZrO 2 show that preheating temperatures of 1050 °C must be employed which is too high for current substrate materials. Consequently, for ZrO 2 some modification to the ceramic composition is necessary to increase the fracture strength and/or fracture toughness. The benefits in practice which may be achieved by laser surface melting are shown by hot corrosion tests. In a test used by Sivakumar (1988), only crack-free specimens (i.e., made by laser treatment of preheated specimens) withstood the test. In fact, the specimen treated at room temperature was worse than the plasma-sprayed specimen although few cracks were present (Fig. 3-28). Zaplatiynsky (1982) showed
3.5 Wear Properties of Laser-treated Surfaces
0 1 2 3 Figure 3-28. Appearance of A12O3 coatings after hot salt corrosion. 0 As sprayed. 1 Lasered at room temperature, 2, 3 Lasered at 800 °C.
that in burner-rig tests, i.e. simulated jet engine conditions, where the salt deposition conditions were different and the probability of salt penetration lower due to the smaller number of cracks, laser-treatment at room temperature also improved the corrosion resistance. This one example suffices to demonstrate the problems and possible solutions associated with surface melting of ceramics.
3.5 Wear Properties of Laser-Treated Surfaces The main reason for surface treatment is to improve the wear resistance and hence useful service life of components. Components in general engineering use are subject to different types of loading simultaneously and indeed different parts of the component can experience completely different conditions. Wear is the result of frictional processes between components or components and materials in relative motion. Consequently wear is a function of the system and not a material constant. Improvement of the wear resistance of a component requires a knowledge of the system, the material, environmental conditions and the dominant wear mechanisms, e.g., adhesion, tribo-oxidation, abrasion and surface fatigue. It is against this background that an appropriate laser treatment must be selected or developed to reduce
129
component wear. In engineering systems, the tribological conditions are complex and more than one wear mechanism is operative which make compromise solutions necessary. The methods of laser treatment previously discussed, i.e., transformation hardening, surface alloying and cladding, involve structural changes which must be exploited in a particular wear situation. Hardness measurements are universally used as a mean of establishing the depth and extent of a laser surface treatment. It is wrong to believe, however, that to combat wear, hardening the surface of the endangered partner is all that is necessary. The microstructure and not the hardness is the prime consideration. 3.5.1 Transformation Hardening Steels and cast irons are often hardened by laser treatment to combat wear. Such treatment does not automatically imply an increase in wear resistance, although there is in general an increase in resistance to abrasion and also surface fatigue (pitting), the latter by increasing the load-bearing capability. The hardness has no influence, however, either on the coefficient of wear (Burwell and Strang, 1952) or on the rate of wear in the case of adhesion wear. For the latter it is better to produce multiphase microstructures with a lower hardness than a predominantly martensitic structure. Typical applications of transformation hardening are cylinder liners (Amende, 1990), cog teeth (Miller and Winemann, 1978) and shafts. 3.5.2 Surface Melting and Surface Alloying Melting the surface provides the chance of removing surface defects, modifying and refining the surface. Elimination of
3 Surface Modification by Lasers
3.5.3 Cladding
20
40 60 Time in h
80
Figure 3-29. Improvement in rolling wear in TiA16V4 achieved by gas alloying. Driving roll • •: laser gas nitrided, • not ground, Q ground, HV0 3 850-1010. Driving roll •: annealed ground, HV0 3 375. Driven roll: 42CrMo4, hardened ground, HV0 3 630. RPM 300 m~ 1 , surface pressure 63 N mm~ 2 , ship 10%, no lubrication.
surface defects has been shown to markedly improve rolling wear resistance (Mordike, 1988). Modifying and refining the surface microstructure also increases the resistance to rolling wear by increasing the yield stress. In addition, a fine distribution of phases as obtained by laser treatment improves the adhesive wear resistance. This has been observed for a wide range of materials. Figure 3-29 shows the improvement in rolling wear achieved on TiA16V4 by gas alloying to form TiN in the surface layer (Mordike, 1990). On increasing the contact pressure from 63 MPa to 89 MPa an abrupt increase in wear rate was observed. This is due to the fact that at 89 MPa the layer is no longer able to withstand the Hertzian pressure and hence the predominant mode of wear changes from abrasive wear to fatigue wear.
Abrasive wear is best combatted by producing a surface harder than the abrading medium. This is best achieved by coating the surface with hard particles, usually carbides. The carbide or other hard material can be chosen for ease of application and also the expected conditions. The wear occurs by the matrix material being removed until the carbide particle breaks out. If the mean free path of the carbide particles is less than the diameter of the abrading medium, then wear of the matrix is more difficult. The mean free path depends on the size and volume fraction of carbide particles. Figure 3-30 shows the abrasive wear measured on untreated TiA16V4 and TiC-coated TiA16V4. The wear was measured by the pin and disc method (Mordike, 1990). Different microstructures result for the single- and two-stage cladding processes and this affects the wear behaviour. The one-stage process is capable of producing much thicker layers
50
uncoated
TiC/Ni (70/30) TiC
10 20 30 40
50 60 70 80
Load in g -cm
Figure 3-30. Abrasive wear in TiA16V4 before and after cladding by TiC.
3.6 Other Effects of Laser Treatment
and is capable of being varied more by choice of particle size and composition. Erosive wear is similar to abrasive wear but very much more complex. Depending on the impact angle and impact velocity of the eroding particles, the nature of the wear can change. Abrasive wear predominates at small angles of incidence whereas for normal incidence, fatigue failure and ablation predominate.
3.6 Other Effects of Laser Treatment 3.6.1 Internal Stresses
In all laser processing, problems may be caused by the formation of residual stresses. These can cause distortion or even cracking. Although laser treatment is better in this respect than many other techniques, the dependence of residual stresses on the production procedure and their influence on the bulk properties of the component must both be understood if laser surface treatment is to be accepted in industry. There are currently two methods used for determining the residual stresses, X-ray methods (Maeder etal., 1981) and the Leluan method (Leluan, 1969). The X-ray method is difficult to apply if the surface is rough or if a texture is present (see Chap. 10, Sec. 10.3.7). The Leluan method involves sequential chemical machining and measuring the shape relaxation after each machining operation. The stress values obtained are therefore average values. The origin of the internal stresses lies in the nonuniform cooling together with any solid state transformations which may occur. Chabrol and Vannes (1986) have carried out a comprehensive study of the development of internal stresses in laser transfor-
131
mation hardened steels and also made some measurements on clad specimens. 3.6.1.1 Transformation Hardening Stresses In transformation hardening, the purpose of the treatment is to induce a phase change. The coarseness of the microstructure and the heating rate as already discussed determines the degree of austenitization and hence hardness. The intensity and distribution of the residual stresses reflect this (Fig. 3-31). Furthermore, in multipass-treated surfaces the previous pass is modified by subsequent passes. The degree of overlap is important in determining the magnitudes and distribution of residual stresses. Generally the surface is in compression as a result of the transformation. 3.6.1.2 Cladding Stresses In this case a material with different properties is applied to the surface. The elastic properties, thermal expansions and 600
£00 -
tempered £00 °C
annealed
200 -
Q_
-200
-£00
-600
Depth in mm
Figure 3-31. Residual stress distribution for different microstructures.
132
3 Surface Modification by Lasers
surface
Figure 3-32. Residual stress field of stellite-clad steel. substrate
cladding
temperature dependence of the mechanical properties are factors which must be considered when matching coating and substrate materials. Figure 3-32 shows a typical profile for a laser clad substrate. Stellite powder was injected into the laser beam as the substrate was a martensitic steel. The results show that tensile residual stresses are induced in the clad layer and again in the substrate. An intermediate zone is characterized by a peak in tensile strength. 3.6.2 Effect on Bulk Properties, in Particular, Fatigue Life
A problem to be resolved before a lasertreated component can be put into service is whether the treatment has affected the bulk properties and if so, how this would influence the performance. The remarks on residual stresses show that all laser treatments produce stresses and an influence on the properties is unavoidable. There are the obvious effects such as cracking of the layer which can arise during laser treatment or subsequently on mechanical working, e.g., grinding, and there is also the problem of the associated distortion which often arises. These associat-
ed stresses obviously influence the mechanical behaviour and corrosion properties. They must be added to any stresses which arise in service. The stresses arising from surface treatment can fluctuate wildly and also change in sign in different regions. It is impossible to predict such stresses and their influence. This is one of the aspects of laser treatment which has been insufficiently studied. It has been shown that the distribution of residual stresses depends on the microstructure of the substrate and the laser parameters. Methods are being developed of producing adaptive layers, i.e., layers designed to minimize the residual stresses whilst at the same time preserving the layer properties. Attempts are also being made to predict the formation of residual stresses by modelling. Ultimately, in any case, the components must be tested under operational conditions. It has been shown (Bergmann, 1986; Amende, 1990; Mordike, 1986) using standard fatigue specimens that the fatigue resistance is changed by laser treatment. In those cases where compressive stresses can be generated in the surface there is an improvement in fatigue resistance and in those cases where tensile stresses are in-
3.7 Industrial Applications
133
duced there is a reduction in fatigue resistance. If components have been mechanically worked or heat-treated, then there is the complicating effect of surface annealing which can affect the fatigue resistance. Fortunately, for many applications fatigue considerations are not critical (e.g., forming tools and dies) but for moving parts subjected to fluctuating stresses fatigue tests are necessary.
3.7 Industrial Applications Surface engineering applications of laser treated surfaces are confirmed to hardening and isolated cases of cladding. The breakthrough to mass production will be made in 1991 when Volkswagen (Germany) start to laser surface-treat camshafts following development (Mordike, 1988) at the Technical University, Clausthal, Germany. Components which are ideal for laser treatment are those with highly stressed regions which represent only a small fraction of the total volume of the component. It is then possible to treat the required surface without too much application of heat and hence little danger of distortion. MAN (Germany) Neue Technologie has undertaken a number of interesting applications. Hardening has been applied to hardening of cylinder liners for car, lorry and ships' diesels (Amende, 1990). Figure 3-33 shows a press tool in the process of being hardened. The hardening tracks are clearly visible. Figure 3-34 shows hardening of cast iron cylinder liners. The liners are large enough for the hardening optics to be inserted. By controlling the insertion of the optics and rotating the liner it is possible to achieve any desired hardening pattern. Other examples reported of transformation hardening in
Figure 3-33. Transformation hardening of a press tool (Courtesy of MAN Neue Technologie).
Figure 3-34. Transformation hardening of pearlitic cast iron cylinder blocks. (Courtesy of BIAS Bremen).
production are hardening of cylinder linear sealing surfaces (Mordike and Burchards, 1989), shafts and splines, racks for rack and pinions. The applications demonstrate that lasers are ideally suited to local treatment of large components not amenable to hardening by other methods, and local hardening of small components which show excessive distortion when hardened by conventional methods. Cladding is an expensive process and it is not surprising that applications have to be selected carefully. Figure 3-35 shows the localized cladding of a valve seat for a large diesel motor. Powder is sprayed into
134
3 Surface Modification by Lasers
Figure 3-35. Cladding of diesel motor valve seat. (Courtesy of MAN Neue Technologie).
Figure 3-36. Cladding of piston-ring grooves. (Courtesy of BIAS Bremen).
the laser beam and melts onto the surface with a minimum of mixing with the substrate. Another interesting application developed by BIAS Fig. 3-36 is the cladding of piston ring grooves in large pistons. In this application wire is fed down the layer
beam onto the cylinder ring groove. Other applications are internal surfaces of tubes (Sepold, 1990) and turbine blades (Rolls Royce (U.K.) and Pratt and Whitney (U.S.A.)). Such applications result in considerable saving owing to the high price of the component and absence of competitive techniques. This is not the case in the motor industry where economic considerations play a major role in determining which processes are employed. It has been shown that surface melting using lasers is feasible. The use of lasers depends on whether the required properties can be attained and whether the production brings technical and economic advantages. It is not sufficient to consider simply using the laser to replace an existing hardening process. The whole production process must be analyzed to see where it can be modified to benefit from the advantages offered by laser treatment. Another problem is the caution demonstrated by manufacturers with regard to the introduction of a new technology when failure could involve very high costs. Testing procedures are thus extensive and time-consuming. Many components have been surface-melted, e.g., camshafts, rocker arms, brake drums, spring supports and axle supports. The most demanding application is surface-melting of camshafts. This became possible only after lasers of sufficient power became available so that, with appropriate beam-forming, a track of the required width could be method in a single pass. (Previously, the surface had been tediously remelted by laying narrow tracks side by side until the required area had been treated. The production rate was too slow for the method to be economic.) Figure 3-37 shows a camshaft surface-melted using a 6 kW Heraeus laser equipped with a line focus optical system.
3.9 References
135
Figure 3-37. Surface-melted camshaft.
Production rates of camshafts higher than those of competitive processes assured its acceptance in principle. This was followed by a comprehensive laboratory and field testing programme before the final decision was made. This indicates that the acceptance of laser surface treatment is dependent on the application and industry involved. Wide acceptance in mass production will be a slow and tedious process.
3.8 Acknowledgements The author is indebted to Dr. Amende, MAN (Germany) Neue Technologie, and to Prof. Sepold, Bremer Institut fur Angewandte Strahlenphysik, for supplying photographs of the treatment of components which are in industrial service.
3.9 References Adamski A., Me Pherson, R. (1986), Proc. 11th Int. Thermal Spraying Conference 1986. Oxford: Pergamon Press, p. 555. Amende, W. (1990), in: Der Laser in der industriellen Fertigungstechnik: Treiber, H. (Ed.). Darmstadt: Hoppenstedt Verlag, pp. 193-233. Becker, R., Binruth, C , Sepold, G. (1988), ECLAT '88. Diisseldorf: DVS, pp. 121-124. Bergmann, H. W. (1986), in: Laser Surface Treatment of Metals: Draper, C. W., Mazzoldi, P. (Eds.). NATO Advanced Science Institutes Series. Dordrecht/ Boston/Lancaster: Martinus Nijhoff Publishers, pp. 351-368.
Bergmann, H. W, Barton G., Mordike, B. L., Fritsch, H. U. (1984), Rapidly SolidifiedMetastable Materials. Amsterdam: Elsevier, pp. 28, 29. Bergmann, H. W., Gundel, P. H., Kalinitchenko, A. S. (1988), Optoelektronik4, 510-517. Boettinger, W.J. (1981), in: Rapidly Solidified Amorphous and Crystalline Alloys, Vol. 8: Kear, B.H., Giessen, B.C., Cohen, M. (Eds.). New York/ Amsterdam/Oxford: North Holland, pp. 15-31. Burchards, D., He, X. (1991), Journal of Lasers in Materials Engineering (in press). Burchards, D., Hinse, A. (1990), in: ECLAT '90, Bergmann, H.W., Kupfer, R. (Eds.). Diisseldorf: DVS, pp. 439-450. Burwell, J.T., Strang, C D . (1952), J. Appl. Phys. 23, 18-28. Chabrol, C , Vannes, A.B. (1986), in: Laser Surface Treatment of Metals: Draper, C.W., Mazzoldi, P. (Eds.). NATO Advanced Science Institutes Series. Dordrecht/Boston/Lancaster: Martinus Nijhoff Publishers, pp. 435-450. Coriell, S.R., Skeraka, R.R (1980), in: Rapid Solidification Processing, Principles and Technology II, Mehrabian, R., Kear, B.H., Cohen, M. (Eds.). Baton Rouge, LA: Claitors, pp. 35-49. Dorre, E., Hiibner, H. (1984), Aluminia 103. Berlin: Springer Verlag. Galasso, F. S., Veltri, R. (1983), Ceramic Bull. 62, 253. Havrda, M. (1986), Proc. 11th Int. Thermal Spraying Conf. 1986. Oxford: Pergamon Press, p. 569. Heavens, O. S. (1965), Optical Properties of Thin Solid Films. New York: Dover. Iwamoto, N. (1982), Thin Solid Films 95, 563. Johnson, P.B., Christy, R. W. (1975), Phys. Rev. B 11, 1315. Jones, H. (1973), Rep. Prog, in Physics 36, 14251497. Kahrmann, W. (1988), ECLAT '88. Diisseldorf: DVS, pp. 119-120. Lawley, A. (1977), Int. J. Powd. Met. 13, 169-188. Leluan, A. (1969), Revue du Garni, Mechanique JuinJuillet, 19-24. Mac Intyre, R. M. (1986), in: Laser Surface Treatment of Metals: Draper, C. W, Mazzoldi, P. (Eds.). NATO Advanced Science Institute Series. Dordrecht/ Boston/Lancaster: Martinus Nijhoff Publishers, pp. 545-550.
136
3 Surface Modification by Lasers
Maeder, G., Lebrun, XL., Sprauel, J.M. (1981), Materiaux et Techniques Avril-Mai, 135-149. Miller, J.F., Winemann, J. A. (1978), Metal Processes III, 38. Mordike, B. L. (1986), in: Laser Surface Treatment of Metals: Draper, C. W, Mazzoldi, P. (Eds.). NATO Advanced Science Institutes Series. Dordrecht/ Boston/Lancaster: Martinus Nijhoff Publishers, pp 389-412. Mordike, B.L. (1987), Oberflachentechnik Surtec, Miinchen: Hanser, pp. 153-158. Mordike, B.L. (1988), Opto Elektronik 4, 482-490. Mordike, B.L. (1989), Oberflachentechnik Surtec, Miinchen: Hanser, pp. 571-578. Mordike, B.L. (1990), in: Laser 6, Mordike, B.L., Vannes, A.B. (Eds.). Paris: Int. Inst. of Techn. Transfer, 9-20. Mordike, B.L., Veit, S. (1988), ECLAT '88. Diisseldorf: DVS, pp. 95-96. Mordike, B.L., Burchards, D. (1989), in: Laser 5, Ghosh, S.K. (Ed.). Paris: Int. Inst. of Techn. Transfer, 157-172.
Orlitsch, J., Rose, A., Weist, P. (1973), Atlas zur Wdrmebehandlung der Stdhle. Dusseldorf: Verein Deutscher Eisenhiittenleute, Vol. 3. Pelletier, J.M., Bonnet-Jobez, S., Vannes, A.B., Gobin, P. F. (1989), in: Laser 4, Ghosh, S. K. (Ed.): Tech. Trans. Series. Paris: Int. Inst. of Techn. Transfer, 143-152. Pond, R.B., Maringer, R.E., Mobley, C.E. (1976), New Trends in Materials Fabrication. Cleveland, Ohio: Amer. Soc. Metals, pp. 128-164. Sepold, G. (1990). Bremer Institut fur Angewandte Strahlentechnik, Private Communication. Shavarew, K.M. (1978), High Temp. Res. 16, 441. Sivakumar, R., Mordike, B.L. (1988), Surface Engineering 4, 127-140. Steen, W.M. (1987), Oberflachentechnik Surtec. Miinchen: Hanser, pp. 159-166. Weiting, T.T., De Rosa, J.L. (1979), J. Appl. Physics 50, 1071. Zaplatiynsky, I. (1982), Thin Solid Films 95, (3), 275.
4 Powder Metallurgy V.S. Arunachalam * and R. Sundaresan2 1 2
Scientific Adviser to the Minister for Defence, New Delhi, India Defence Metallurgical Research Laboratory, Hyderabad, India
List of 4.1 4.2 4.2.1 4.2.1.1 4.2.1.2 4.2.2 4.2.2.1 4.2.2.2 4.2.2.3 4.2.2.4 4.2.3 4.2.4 4.2.4.1 4.2.4.2 4.2.4.3 4.2 A A 4.2.4.5 4.2.4.6 4.3 4.4 4.4.1 4.4.1.1 4.4.1.2 4.4.1.3 4.4.2 4.4.3 4.4.4 4.4.5 4.5 4.5.1 4.5.1.1 4.5.1.2 4.5.1.3 4.5.2
Symbols and Abbreviations Introduction Production of Powders Mechanical Methods Comminution Mechanical Alloying Chemical Methods Reduction Decomposition Precipitation Self Propagating High Temperature Synthesis Electrochemical Methods Physical Methods Principles of Atomization Gas and Water Atomization Centrifugal Atomization Ultrasonic Gas Atomization Vacuum Atomization Rapid Solidification Characterization of Powders Compaction of Metal Powders Pressureless Compaction Vibratory Compaction Slip Casting Injection Molding Unidirectional Compaction Cold Isostatic Compaction Explosive Compaction Powder Rolling Sintering Solid State Sintering Equations of Sintering Pore Closure and Development of Microstructure Sintering Mechanism Maps Liquid Phase Sintering
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
139 142 143 144 144 146 148 148 149 150 150 151 151 152 154 156 157 158 158 159 161 162 162 163 163 163 165 167 169 170 170 171 173 173 173
138
4.5.3 4.6 4.6.1 4.6.2 4.6.3 4.6.4 4.6.5 4.6.6 4.6.7 4.7 4.8
4 Powder Metallurgy
Activated Sintering Deformation Processing at High Temperatures Principles of Pressure Sintering Hot Isostatic Pressing Uniaxial Hot Pressing Powder Forging Powder Extrusion Special PM Processes Post-Consolidation Operations Applications References .
176 176 177 179 180 181 182 183 184 185 189
List of Symbols and Abbreviations
List of Symbols and Abbreviations a a,d A A, As
b B B Bs c C
c ps
d d dx D e E
f /a /. /v
F P
K 9 h H AH AH0 k kB K AL Lo m, n
Nu Pu2 PH2O
P P
particle radius constants constant creep constant surface area compression coefficient constant magnetic flux density microstructure based constant vacancy concentration slope of the deformation resistance against temperature curve heat capacity average particle size diameter diameter of the liquid metal stream melt pool diameter emissivity free energy propellant force flow rate of air flow rate of liquid cyclic force peak vibratory compaction force direct force amplitude of cyclic force acceleration due to gravity half distance of approach of the particle centers magnetic field droplet cooling enthalpy of formation constant Boltzmann constant constant linear shrinkage initial length sintering parameters creep constant Nusselt number partial pressure of hydrogen partial pressure of water vapor vapor pressure pressure applied pressure, effective pressure
139
140
4 Powder Metallurgy
extrusion pressure pressure of liquid metal, pressure of atomizing atmosphere particle radius principal radii of curvature R R Re s s, s0 t T Tad Td9 V
V AV
v0 K
w W
w X
extrusion ratio
gas constant Reynold's number Stefan-Boltzmann constant solubilities time temperature adiabatic temperature T w , I,g temperature of the droplet, temperature of the chamber wall, temperature of the gas velocity volume reduction of volume initial volume volume of propellant gases for unit mass propellant powder mass Weber number punch assembly mass radius of the neck
ag y Co.
s
fix, Zy> K
er9 e09 ez
n 0
6 jl V
Q ^app tccomp O
creff °y CO
Q
grain growth parameter surface tension charge density total creep rate strain rates in x9 y9 z direction radial strain, circumferential strain, axial strain coefficient of viscosity kinematic viscosity of liquid metal, kinematic viscosity of gas volume porosity densification rate chemical potential Poisson ratio density apparent density of powder density of compacted mass stress effective stress yield stress rotation rate atomic volume
List of Symbols and Abbreviations
APT CAP CSC EBRD FEM HIP MA PM PREP PSV REP ROC RRP RS RSP
ammonium paratungstate consolidation by atmospheric pressure centrifugal shot casting electron beam rotating disk process finite elements method hot isostatic pressing mechanical alloying powder metallurgy plasma rotating electrode process pulverisation sous vide rotating electrode process rapid omnidirectional compaction rotating rod process rapid solidification rapid solidification processing
141
142
4 Powder Metallurgy
4.1 Introduction Powder metallurgy is an ancient technology. Our ancestors used it extensively when they did not possess the required knowledge or wherewithal to melt metals, or even reduce ores. They therefore heated the metal nuggets they collected and beat them together to form the required shapes. Later they used chunks of ores that were reduced by charcoal under heat and bonded together. A classic example of such an effort is the Iron Pillar of Delhi which was fabricated during the reign of Emperor Ashoka in the second century B.C. As monuments go, it is a large one weighing over six tons and it has withstood the ravages of time rather well (Fig. 4-1). There are
Figure 4-1. Delhi iron pillar.
other instances where particulates were heated and bonded to form desired objects. The famous Damascus sword with cobonded layers of iron-carbon alloys of different compositions and the platinum jewelry of the Inca civilization typify this approach. Powder metallurgy (PM) received a major boost in this century from the growth of the electrical industry. Wires made from high melting metals and alloys were needed to replace the carbon heating elements of incandescent lamps; tungsten metal produced by powder metallurgy met the need. This metal was later replaced by tungsten alloys containing dispersoids to increase the life of the filaments. Other refractory metals needed for similar applications were also produced by powder metallurgy. In fact, until the middle of this century, PM was exclusively used either for producing parts from refractory metals and compounds, or for fabricating structures with controlled porosity or dispersoids. As in the case of conventional electrical power, the emergence of nuclear power also brought in new challenges to PM. Fissile materials for nuclear fuel were needed as ceramic compounds because of their high temperature strength and environmental compatibility. PM techniques were so successful that practically all nuclear fuels are currently produced using the PM route (see Volume 10, Chapters 1-3). With melting temperatures for all metals well within reach, PM techniques are now widely employed for their other attributes. PM is in itself a complete materials processing route. Often the process begins with the reduction of ores to produce powder and the mixing of appropriate alloying additions. The homogeneity of composition is also assured down to the size of the powder particle itself. The powder agglom-
4.2 Production of Powders
erates are then shaped into products with the application of pressure and temperature. When necessary, some additional processing steps such as forging are introduced as finishing operations. Today, PM is used for manufacturing a variety of products. Powders can often be the end products in themselves and are used in many industrial applications. Production of porous products such as selflubricating bearings and filters is made easy because of PM. Refractory metal products and composite structures such as cemented carbides are also produced using this route. A recent application is in the more demanding area of aerospace technology where PM is preferred not only for the attractive high-temperature properties and desirable microstructures, but for the near-net shape of the end products as well. PM is intrinsically economical in the use of raw materials as only minimum scrap is generated during processing. Further, the consumption of energy is less compared to conventional ingot metallurgy. This is especially true when products are needed in very large numbers. In spite of these attractions, there are some limitations imposed both by the processing equipment such as presses and furnaces, and by the large surface areas of powder aggregates. Largesized products are more easily produced by conventional metallurgical operations than by PM. There is also some resistance to the use of PM products in critical loadbearing applications; this is due more to the non-availability of reliable quality assurance techniques for detecting small residual pores and other defects than to the processing route. In many applications, this resistance has been overcome by finishing the PM processing with forging or extrusion. PM can broadly be divided into the following processing steps (Fig. 4-2): produc-
143
tion of powder; compaction and sintering, or hot pressing where compaction and sintering are combined into a single step. Where necessary, these steps are followed by further deformation processing. In the following Sections, we shall discuss these steps in some detail, as also the applications of PM in many areas of materials processing.
4.2 Production of Powders The size, shape and purity of the powder are important parameters in determining the quality of a PM product. Considerable effort has therefore gone into developing a variety of techniques to manufacture metal and alloy powders with the required properties. Even though it is possible to fabricate PM products from powders of any size, practical considerations dictate that the particles should be small and their size distribution appropriate to pack the volume of the preform efficiently. The voids in a powder agglomerate have a geometrical relationship to the particle size: the finer the particle size, the smaller is the size of the void. These voids are to be filled during compaction and sintering. As sintering is a diffusion-controlled process, small pores are more easily filled, thus making small particles attractive starting materials for PM processing. There is also an additional advantage: the specific surface energy of fine particles is large compared to that of coarse ones, and this provides an increased driving force for sintering. But there are also instances where large surface areas result in extensive formation of thin oxide films which may adversely affect the properties of the resulting PM products. Relatively coarse powders around 100-300 (im diameter are preferred in such cases.
144
4 Powder Metallurgy
Electrochemical Methods
Chemical Mechanical Physical Methods Methods Methods
Synthesis
POWDER
Lubricants -JBLENDINGU_ Additives
I COLD COMPACTION Die Cold Compaction Isostatic Pressing
Other Methods
HOT CONSOLIDATION Hot Pressing
Hot Isostatic Pressing
Other Methods Extrusion PseudoIsostatic
Rolling Extrusion Explosive CocnpocHon Injection Molding
SINTERING
FINISHING OPERATIONS Coining
FORGING
Infiltration
FINISHED PM PARTS
The shape of the powder is also an important parameter. Some applications such as porous filters specifically demand spherical particles while irregular particles are adequate for many other applications. Spherical particles are difficult to compact and are therefore used only with large compacting pressures or in hot pressing, where high temperatures and pressures are simultaneously applied. Metal powders are produced by a variety of techniques which include mechanical comminution, chemical synthesis, decomposition or precipitation, electrolytic depo-
Figure 4-2. Basic processing steps in powder metallurgy.
sition and atomization from the melt. The choice of the technique is determined by the properties of the metal, the application in view, the desired purity and economics. Table 4-1 gives a summary of the methods used in the manufacture of different metal powders and these methods are discussed in the following sections. 4.2.1 Mechanical Methods 4.2.1.1 Comminution
Mechanical comminution processes such as crushing and grinding are perhaps the
4.2 Production of Powders
Table 4-1. Processes used in the manufacture of powders, (a) Mechanical comminution; (b) Chemical: reduction; (c) Chemical: decomposition; (d) Chemical: precipitation; (e) Electrochemical; (f) Physical: atomization. Material
Process (a)
Al/Al alloys Be Co Cu Cu alloys Fe Steels Mo Ni Ni alloys Ag Ta Sn Ti Ti alloys W Zr
(b)
(c)
(d)
(e)
(f) X
x
X X X
X
X
X
X
X
X X
X
X
X X X
X
X X X
X
X
X
X
X X
X
X
X
X
X X
oldest ones for the production of powder and are still used in many PM plants. The metal to be powdered is usually milled in a vessel containing appropriate grinding elements for comminution. Depending on the
145
type of mill, the grinding elements can be hardened balls or even rods. Tumbler ball mills containing appropriately chosen grinding media are most commonly used for comminution. Comminution can be carried out either in wet or dry conditions. The limiting particle size that can be obtained depends on the size and shape of the grinding media, and other milling conditions. Several high-energy milling processes are also used effectively for size reduction. These include vibratory ball mills, attrition ball mills and large-diameter tumbler ball mills (see Chapter 5, Sec. 5.2). The comminution efficiency and the limiting particle size depend on the equipment used and these are shown in Fig. 4-3 (Kuhn etal., 1984). Even ductile metals can be powdered by mechanical comminution though they may first have to be made brittle before grinding. For instance, titanium is hydrided to the friable hydride, or stainless steel sensitized before comminution. The milling processes are also used for an effective mixing of the components in a powder blend. Particle shapes resulting from mechanical comminution are generally irregular, but the shape can be controlled to some
Jaw crusher Vibratory mills Ring roller mill Ball mill Hammer mill Attritor Fluid energy mills
105
104
103
102 Product size, urn
101
10°
Figure 4-3. Comminution range of different classes of equipment.
10'
146
4 Powder Metallurgy
extent by the milling conditions: wet milling, for instance, leads to more flaky powders than dry milling. There are also limitations to the level of purity that can be maintained in mechanical milling because of the wear of the grinding media. Therefore mechanical comminution is used only in those applications where the resulting shape and purity are acceptable. However, some special grinding processes are available to maintain purity during mechanical milling. Gas-jet pulverizing techniques lead to better purity by avoiding grinding media such as steel balls. These methods employ collision between powder particles themselves or between powder and targets of the same composition for grinding. Energy for collision is provided by gas jets. In such a fluidized-bed counterflow jet mill, particles are ground by mutual collision in a fluidized bed, the kinetic energy being provided by a number of high-speed gas jets (Joensson and Hohmann, 1987). This technique is often used in the comminution of high purity powders that are obtained from rapid solidification of melts (Sec. 4.2.4.6), which are generally too coarse to be used directly for PM processing. 4.2.1.2 Mechanical Alloying
A novel extension of mechanical comminution was developed by Benjamin (1976) for producing alloyed and dispersion hardened materials from powder blends. This process, known as mechanical alloying (MA) is carried out in attritors or other high energy ball mills. In these ball mills, the powder is welded, broken and rewelded repeatedly during comminution leading to a homogeneous dispersion of the constituents. The various stages of mechanical alloying are described by Gilman and Benjamin (1983) and are shown schematically in Fig. 4-4. The dispersion in MA
can reach such fine levels that interparticle spacing between dispersoids can be well below 0.5 jam, and alloying may even appear homogeneous under X-ray diffraction. The welding, rewelding and homogeneous blending take place on the surface of the grinding media during the impact of the powder with the media. This is seen by the change of particle size with increasing comminution time; it may also so happen that after prolonged comminution, the particle size increases, confirming the welding of impacted particles on the surface of the grinding media. As comminution and mechanical alloying take place in a high energy mill, a large quantity of heat is generated, and this in turn enhances diffusion among the components and promotes recrystallization of the deformed powder. Recent experiments (Bhagiradha Rao, 1989) have also confirmed the postulated mechanisms of diffusion and recrystallization. These experiments were carried out in a nickel-chromium alloy with and without dispersoids. Results of this study confirm that true alloying does indeed take place during mechanical alloying, because of enhanced diffusion due to cold work and recrystallization cycles (Fig. 4-5). As recrystallization is retarded in alloys containing dispersoids, mechanical alloying is found to be correspondingly slower for the same alloy with dispersoids. Because of the heavy cold work, mechanically alloyed powders are not easy to consolidate. They are therefore compacted at elevated temperatures by hot pressing, hot extrusion or hot rolling. This is further followed by thermomechanical processing to yield elongated grain structures that give high-temperature creep resistance. At present, mechanical alloying is employed mainly in the production of oxide dispersion strengthened nickel-, iron- and aluminum-base alloys. Because of their
147
4.2 Production of Powders
100 pm
Metal A
100 urn
ntermetallic
Concentration vX of metal B Dispersoid
-l ° ^ -
Metal B
•
•
a
0.5pm (a)
t
0 "JL * P fj
/ t Concentration of metal A
Remnant of intermetallic Equilibrium precipitate
Dispersoids 0.5 pm (c)
Heat (Consolidation) 100 |
Metal A Interdiffusion Dispersoids
100 pm
100pm
Metastable phase Dispersoids Intermetallic Dispersoids Precipitate phase
0.5|jm
(d) Figure 4-4. Stages in mechanical alloying: (a) intense cold welding in the first stage, (b) rapid fracturing in the intermediate stage and (c) moderate cold welding in the final stage lead to (d) extremely deformed structure in completely mechanically alloyed powder with fine grained structure and equilibrium distribution of dispersoids (Sundaresan and Froes, 1987).
148
4 Powder Metallurgy
(a)
(b)
Ni
Figure 4-5. Mechanical alloying of Ni-20Cr (a) with ThO2 addition and (b) without ThO 2 . EPMA line scans show low degree of alloying in the presence of ThO2 (Bhagiradha Rao et al., 1986).
Cr
high-temperature properties, mechanically alloyed superalloys find applications in aircraft and industrial gas turbines, in power generation equipment, and in heat treatment furnaces. Mechanically alloyed aluminum-base alloys are currently entering into some applications. These alloys incorporate organic process control additives, resulting in dispersions of A12O3 and A14C3. Lately several new facets of mechanical alloying have been recognized (Sundaresan and Froes, 1987). Liquid-immiscible and solid-immiscible systems which are difficult and often impossible to process by conventional methods of solidification processing can be obtained in the solid phase with a homogeneous distribution of the second phase by using mechanical alloying. Superconducting Nb 3 Sn, supercorroding Mg-Fe/Cu/Cr/Ti and hydrogenstoring Mg-Ti/Ce alloys are examples of this kind. Because of the heavy cold work involved in the process, mechanical alloying is known to produce amorphous phases and this is now being explored in some multicomponent systems. Mechanical alloying is also treated in Chapter 5, Sec. 5.5, of this volume.
Several factors have to be considered in selecting an appropriate reaction for powder production. Thermodynamics and kinetics of the reaction dictate the feasibility and economics, while nucleation and growth parameters determine the resulting powder characteristics.
4.2.2 Chemical Methods
4.2.2.1 Reduction
Chemical reactions often offer attractive routes to produce metal powders. These reactions can be in the reduction of metal-
One of the early developments in PM was the production of iron powder by the Hoganaes process (Goetzel, 1949). The
lic ores, or in the synthesis, precipitation or decomposition of metallic compounds. The governing chemical equations for these reactions are the following: Reduction by carbon monoxide or hydrogen: MO + H 2 ^ M + H 2 O
(4-1)
MO + CO -> M + CO 2
(4-2)
Decomposition of carbonyls or hydrides: Fe(CO) 5 -*Fe + 5Co
(4-3)
TiH 2 -+Ti + H 2 (4-4) Precipitation from liquid solutions or gaseous compounds: CuSO 4 + H 2 -> Cu + H 2 SO 4
(4-5)
ZrCl 4 + 4 Na -* Zr + 4 NaCl
(4-6)
Synthesis from constituents: MoO 3 + 2Al + 2 S i ^ M o S i 2 + Al 2 O 3 (4-7)
4.2 Production of Powders
process was developed using the high purity iron ore mined in Sweden, and produces good quality iron powder of high purity and excellent flow characteristics needed for good compaction and sintering. In this process, the magnetite concentrate Fe 3 O 4 is mixed with coke, and limestone to absorb sulfur, and reduced at temperatures around 1475 K. Part of the required heat is generated by burning CO produced during the reduction. Pure iron powder is produced by separating the sponge iron produced in the reduction, pulverizing the sponge and annealing the powder in hydrogen. Iron powder is also produced by the hydrogen reduction of millscale. This reaction takes place at a lower temperature compared to the Hoganaes process (1300 K), but the requirement of critical partial pressures demand that a continuous countercurrent flow of hydrogen is maintained when the oxide charge is continuously or semicontinuously fed. An industrially important hydrogen reduction reaction is in the production of tungsten powder (Borchers, 1979). This forms part of a complex processing route that includes purification of the ore concentrate and its conversion to high purity ammonium paratungstate (APT) by a series of dissolution and re-precipitation reactions. Subsequently, APT is thermally decomposed to a suboxide of tungsten (WO3_X) which is then reduced in a countercurrent of dry hydrogen in a tubular furnace with semicontinuous feed. The ratio of partial pressures (PU2/PH2O) provides an effective tool for controlling the reaction rate and the particle size of the reduced powder. This ratio is determined by the rate of gas flow, the reducing temperature, oxide bed height and its feed-rate. In some cases, where the oxides are not reducible by carbon or hydrogen, the re-
149
duction is carried out using akali or alkaline earth metals. For instance, titanium oxide can be reduced by a calciothermic reduction: TiO 2 + 2 Ca -• Ti + 2 CaO
(4-8)
The reduced metal comes out as powder because of CaO which separates metallic agglomerates. As this reaction is highly exothermic and difficult to control, calcium hydride is used as a reducing agent: TiO 2 + CaH 2 -*Ti + Ca(OH) 2
(4-9)
By using excessive reductant, it is possible to control the reaction and the powder particle size. However, these reactions are of limited importance as the products are generally impure and are therefore only useful where non-critical and non-load bearing applications are envisaged. Selective reduction of oxides is also employed to produce complex powders with oxides dispersed in a metallic matrix, as in thoriadispersed or yttria-dispersed nickel and nickel-chromium alloys. 4.2.2.2 Decomposition Both nickel and iron powders are produced by a decomposition reaction of their respective carbonyls. The process (Fig. 4-6) consists of first preparing the carbonyl by passing high pressure CO over the heated metal and condensing the vapor as liquid under pressure. The liquid is purified by fractional distillation and is then decomposed at temperatures around 400 K at atmospheric pressure to produce fine, high purity metal powder (Queneau et al, 1969). To ensure purity and suitable particle shape, reaction conditions are optimized so that the decomposition is spontaneous within the heated space and not nucleated on the walls of the container. Complex carbonyls such as Pt(CO) 2 Cl 2 and organo-
150
4 Powder Metallurgy
as hydrogen:
Iron Reaction Chamber
M+ + + H 2 ^ M ° + 2H+
Synthesis of (CO^
with an equilibrium constant given by +
Purification of carbonyl Fractionating Column
CO recycle to synthesis
Pure
I
Vaporization
NH
3 —I Decomposition
i
Carbonyl iron powder "2 ->|
Reduction
| |
j
(4-10)
Blending
|
i
I Soft grades | | Hard grades |
Figure 4-6. Flowsheet for the manufacture of carbonyl iron powder.
metallic compounds have also been used to produce metal powders by decomposition. Decomposition of the hydride is also an effective method for producing titanium and zirconium powders via the hydridedehydride route. The ductile metal or its alloy is first converted to its hydride by reaction with hydrogen at 700-900 K under pressure. The friable hydride is easily crushed to powder of the required fineness and then decomposed to the metallic powder by heating under vacuum. 4.2.2.3 Precipitation
Copper powder is sometimes produced by displacing the copper ion in the salt solution (generally copper sulfate) by iron. The extent of precipitation is controlled by factors such as the pH value of the solution. Precipitation of metals from salt solutions can also be achieved with gases such
]p H 2
(4-11)
Both nickel and copper powders are produced directly from the leached ores using this route. Various organic surface-active agents are used as additives to control the powder characteristics. An extension of this process is the precipitation of composite particles by suitably seeding the second phase particles in the solution. Precipitation from the vapor phase is also sometimes used to produce metal powder. A case in point is the production of fine, spherical powders of the refractory metals W and Mo from their gaseous hexafluorides. 4.2.2.4 Self-Propagating High Temperature Synthesis
Self-propagating high temperature synthesis of powders of ceramic and refractory compounds and alloys is a recent innovation. In this process, a porous compact made from a mixture of the reagents is brought into contact with a hot tungsten coil. This initiates an exothermic reaction which continues as a self-sustaining combustion wave propagating through the porous mass and converting it into the reaction products. The scheme of the reaction synthesis is shown in Fig. 4-7 (Crider, 1982). Depending on the temperature of the reaction, synthesis can be entirely in the solid state as gasless combustion, partly in the gaseous state as filtration combustion, or entirely in the gaseous state as condensed combustion. Of these, gasless combustion can be used for the production of powders suitable for PM applications (Munir and Anselmi-Tamburni, 1989).
4.2 Production of Powders
Rate of thermal release
Direction of wave
Initial substances
n
k y IK Zone of heating
151
\J 1 j I
Zone of synthesis
s— Temperature
\ - D e p t h of transformation Figure 4-7. Equilibrium adiabatic structure of a self-propagating high tem perature synthesis wave (Crider, 1982). Final product
The synthesis can be designed on the basis of the adiabatic temperature of combustion, the combustion wave velocity and activation energy for the compound formation. For a useful reaction, the adiabatic temperature Tad must be lower than the melting temperature of the reaction product (Munir, 1988). The value of Tad of a compound can be computed on the basis of its enthalpy of formation {AH0) and its heat capacity Cps and compared with the melting temperature to evaluate the feasibility of the reaction. The process has relatively low energy requirements and is highly cost-effective. With an exothermic reaction, the processing time at high temperatures is short and the purity of the powder can be maintained. Metastable phases such as cubic TaN and composites such as MoS 2 /Nb can be made by this process. The process is already in use in the USSR for the production of TiC abrasive powders, TiNi memory alloys and MoSi2 heating elements. 4.2.3 Electrochemical Methods
Metals of high purity can be precipitated from aqueous solutions on the cathode of an electrolytic cell in a spongy form to produce powder by subsequent comminution.
This method is mainly used in the production of copper and iron (Willis and Klugston, 1959). Fused salt electrolysis has also been used for the production of tantalum and beryllium powders (Miller, 1958). The factors that promote powdery deposits are high current densities, weak metal concentrations, low temperatures, and surfactants to promote high viscosities. All these factors also lead to low deposition rates and high costs. With increasing energy costs, electrochemical methods of powder production are being phased out (see also Chapter 11). 4.2.4 Physical Methods
Atomization is a process of breaking up liquid metal or alloy into fine droplets and allowing them to solidify as powder. Because of the versatility, purity and the control of powder shape inherent in the process, atomization is becoming increasingly popular as a route for large scale manufacture with attendant reduction in costs. Many techniques are available for breaking up a liquid metal stream into fine droplets which are then allowed to solidify by exchange of heat with the surroundings. Most common among these are the impingement of high speed jets of gas or wa-
152
4 Powder Metallurgy
ter on the liquid metal stream. This technique is limited to metals and alloys that do not chemically react with the impinging gas or water. There are also variations where the liquid metal stream is directly fragmented by centrifugal forces. These are used for producing powder from reactive metals and superalloys which are sensitive to gas contamination. There are also other techniques like vacuum atomization and ultrasonic atomization which are used in special cases. 4.2.4.1 Principles of Atomization
In conventional atomization, a liquid metal stream is produced by pouring molten metal through a tundish which is then broken into droplets by the impingement of high pressure gas or water. This disintegration of the liquid is shown schematically in Fig. 4-8. The interaction between the jets and the liquid stream begins with the initiation of small disturbances at the liquid surface which grow into shearing forces that fragment the liquid into ligaments. The energy of the jets is so large that the broken ligaments are further fragmented into droplets which can reach very
Stage I Growth of waves on liquid sheet
Stage II Fragmentation and formation of ligaments
small sizes (Shinde and Tendulkar, 1977; Lawley, 1977). The liquid metal stream has a velocity v given by =
A[2g(Pl-Pg)/Q] 0.5
(4-12)
where A is a geometric constant and g, the acceleration due to gravity; Px is the injection pressure of the liquid metal, Pg the pressure of the atomizing atmosphere and Q is the density of the liquid. The shearing forces in atomization that lead to the formation of ligaments from the liquid metal depend on the Reynold's number, Re, which in turn, is related to the size and velocity of the stream and the density and viscosity of the liquid metal. There are several empirical equations for calculating the average particle size in atomization. The most frequently used one for determining the average particle size d is given by d=A
y
0.45
/•M.5
1000 y
Stage III Breakdown of ligaments into drops
Figure 4-8. Mechanism of droplet formation in atomization (Lawley, 1981).
(4-13)
4.2 Production of Powders
where Q is the density of the liquid, y, the surface tension, rj, the coefficient of viscosity and v, the relative velocity between the liquid and gas jets; /, is the flow rate of liquid, / a , the flow rate of air and A and B are constants. Another empirical equation for the particle size is
153
for instance, even with high surface tension, aluminum and zinc tend to solidify as irregular particles because of oxidation effects. The mechanism of centrifugal atomization is different because of the process design. In this technique, the molten pool is rotated at high speed till the liquid flows over the rim of the container and is frag0.5 mented by the centrifugal force acting on (4-14) it. As the cooling rate in this process is slow Wr,. in the absence of gas jets, centrifugally atomized powders tend to be spherical. It where Wis the Weber number of the metal may also happen that the time of flight of stream, vg, the velocity of the atomizing the liquid droplet to hit the chamber walls medium, du the diameter of the liquid metal stream; rjx and r\% are the kinematic is smaller than the time required for solidification. In such instances, the particles viscosity of the liquid metal and gas, and appear as flat discs solidified on impact K, a constant. with the walls. Both these equations show the particle Because of the rotation of the molten size to decrease with decreasing surface pool, the liquid metal spreads towards its tension of the liquid metal and increasing periphery and forms a toroidal rim. Owing velocity of the atomizing medium. In practo instability at the rim, the sheet of the tice, for a given nozzle design, the average liquid breaks up into fine threads, which in particle size is controlled by the pressure of turn break into droplets surrounded by the atomizing medium and also by the smaller droplets known as satellites. The apex angle between the axes of the gas jets. mean diameter d of the main drop can be Higher apex angles lead to smaller particle estimated by balancing the surface tension sizes. force against the centrifugal force and is The particle shape in gas or water atomgiven by (Hodkin et al, 1973) ization depends largely on the surface tension and the solidification rate (Lawley, 1977). While the surface tension of the drop (4-15) tends to spheroidize it, the cooling rate limits the time available for this process. If, for instance, the time for solidification is larger where co is the rotation rate, y, the surface than the time required for spheroidization, tension and D, the melt pool diameter. The above equation is strictly valid only for the then spherical particles will result. For promain droplets, and it is difficult to calculate ducing fine spheres, it is necessary for the either the particle size or the distribution of solidification to be as late as possible bethe satellite sizes as the mechanisms underfore the droplets hit the chamber wall. Belying their formation are not fully undercause of this condition, iron and copper stood. solidify readily as fine spherical powders, Since the flight of the atomized droplets while lead and tin solidify as irregularly is entirely dictated by friction and gravity shaped powders. There are also other facopposing the centrifugal momentum, the tors that contribute to the particle shape:
154
4 Powder Metallurgy
trajectory can be easily calculated. The droplet cooling AH in time At is given by AH=-nKdNu(Td-Tg)At + T(t-T*)At
(4-16)
where Td, Tg and Tw are the temperatures of the droplet, the gas and the chamber wall respectively; e is the emissivity of the droplet, s the Stefan-Boltzmann constant and Nu the Nusselt number of the cooling gas flow. On the basis of Eq.(4-16), it is possible to design the cooling parameters and chamber size appropriately for centrifugal atomization. 4.2.4.2 Gas and Water Atomization
Inert gas atomization plant
Vacuum melting
A typical atomization unit consists of a melting facility, an atomizing chamber, and powder drying and collection unit (Gummeson, 1972) as shown in Fig. 4-9. Typical atomization parameters are given in Table 4-2, and some atomized powders are shown in Fig. 4-10. Any of the standard melting furnaces is acceptable for producing the liquid metal. However, complex alloys prone to contamination are usually melted in a vacuum induction furnace or by skull melting with an electron beam. The molten metal is then poured into a tundish with a top ceramic filter and a bottom nozzle which controls the shape and size of the falling metal stream. This stream passes through an atomizing nozzle sys-
Table 4-2. Typical atomization variables. Parameter Powder collection
Figure 4-9. Inert gas atomization unit (Courtesy DMRL).
Flow rate Flow velocity, m/s Pressure, MPa Superheat, °C
Liquid metal Water jet 4.5-90 kg/min — — 75-150
Gas jet
100-400 1-14 1/min m3/min 70-230 20-supersonic 5.5-21 0.3-8.5 -
4.2 Production of Powders
155
Figure 4-10. Atomized powder particles: (a) bronze, water atomized, (b) aluminum, water atomized, (c) high speed steel, water atomized, (d) aluminum, ultrasonic gas atomized, (e) Nimonic 80 A, argon gas atomized, and (f) Astroloy, argon gas atomized.
156
4 Powder Metallurgy
tern, in which the high velocity jets of the atomizing medium (gas/air/water) strike and disintegrate the stream into fine droplets. These droplets are carried at high velocity with the atomizing gas stream inside the chamber and solidify owing to heat loss by convection. It is possible to increase the solidification rate by cooling the atomizing gas, or even by pumping additional precooled gas from the bottom of the chamber. The selection of the atomizing jet medium is based mainly on the reactivity of the metal and the cost of the medium. Air and water are inexpensive, but react with many molten metals. Inert gases can be expensive and it may be necessary to recycle the gas when helium is used as the atomizing medium. The heat transfer characteristics of the gases are different as shown in Table 4-3 and these also determine the particle size and other characteristics; but the mechanism of atomization is independent of the medium. Nozzle design controls the efficiency of disintegration and thus dictates the final characteristics of the powder. The nozzle can be an annular concentric ring around the metal stream, or can be a set of discrete jets (Klar and Fesco, 1984). The nozzle design also incorporates the choice of a "free fall" or "confined" atomization. In the free fall atomization, disintegration takes place some distance ahead of the tundish orifice, while in confined atomization the disintegration is at the tundish orifice itself. Confined atomization with its higher efficiency of energy transfer is used for gas atomization. The apex angle for water atomization is smaller than for gas and is designed for free fall atomization. Spherical and fine powders are difficult to produce in water atomization and particles generally tend to be irregularly shaped. The process is also lim-
Table 4-3. Heat transfer coefficients for convective cooling of liquid metal droplets by different gases. Cooling gas
Argon Nitrogen Helium
Heat transfer coefficient, W x 10 3 /m 2 K for droplet diameter (in jim) 1
10
100
1000
49.0 70.8 350.0
11.8 16.8 60.0
3.3 4.7 10.0
1.02 1.45 3.00
ited only to metals and alloys from which the surface oxide can be easily removed by mechanical or chemical methods. 4.2.4.3 Centrifugal Atomization
The driving force for the development of centrifugal atomization processes has been the cleanliness of the product. These processes avoid gas jets and atomization nozzles, and therefore the liquid metal continues to remain pure during the entire process. Thus this method is eminently suited for atomizing reactive metals such as titanium and zirconium. Several centrifugal atomization processes have been developed, as listed in Table 4*4. On the Table 4-4. Centrifugal atomization processes. Mode of Mode energy of transfer melting Direct
Arc
Process
References
Rotating electrode Friedman process (REP) (1970) EB 'Pulverisation Decours sous vide' (PSV) (1976) Rotating rod Arunachalam process (RRP) etal. (1975) Plasma Plasma rotating Roberts rod process (PREP) (1984) Indirect Arc Centrifugal shot Hodkin casting (CSC) (1973) EB Electron beam Stephan rotating disc and Fischprocess (EBRD) hof(1976)
4.2 Production of Powders
basis of the mode of centrifugal energy transfer, the processes can be broadly divided into direct energy transfer processes where the liquid metal pool itself is rotated, and indirect transfer processes where the liquid metal is poured on to a rotating surface. Among the processes listed, only the rotating electrode process (REP) and plasma rotating electrode process (PREP) are commercially well developed. These are direct energy transfer processes, where a rod is rotated at high speed and simultaneously melted at one end. The melting is carried out with a tungsten electrode as in a vacuum arc furnace or with another electrode of the same material. The centrifugally atomized liquid droplets are cooled by convection with the helium gas present in the atomizing chamber and the powder is collected at the bottom. A schematic diagram of a REP atomization chamber is shown in Fig. 4-11. There are many variations to the design shown in the figure. In one variant, a long rod to be atomized is fed through a vacuum seal into the chamber, with the rotating mechanism kept outside. In another variant, small electrodes are kept within the chamber itself and are substituted into the rotating spindle as and when the molten electrodes are consumed. A glove compartment is provided to effect this substitution without breaking the furnace atmosphere. In an indirect energy transfer process, liquid metal is poured through a tundish on to a fast rotating disc which acts as a rotary atomizer. The centrifugally atomized droplets are cooled by precooled helium gas before solidification. By this, it is possible to solidify the droplets rather rapidly, generally with a rate higher than 105 K/s. Rapidly solidified nickel base alloy powders have been produced by this process and subsequently formed by iso-
Water cooling
157
Rotating consumable electrode
0 Stationary electrode Inert •£ gas (He)
Vacuum Collection port
Figure 4-11. Schematic of rotating electrode process (Lenel and Ansell, 1982).
thermal forging for the manufacture of high-temperature turbine discs for gas turbines. On the other hand, REP and PREP powders are generally large-sized and spherical with a cooling rate of about (102 K/s). It is therefore not possible to produce rapidly solidified powders by these routes. 4.2.4.4 Ultrasonic Gas Atomization Ultrasonic atomization is used for the production of powder with particle size finer than those obtained in conventional atomization. The process is similar to conventional gas atomization except for the discrete gas jets impinging on the metal stream in a pulsed fashion, with frequencies of 40-100 kHz. The high speed gas jets are accelerated by a shock wave tube to speeds up to Mach 2. The shearing action results in fine droplets, usually less than 30 jim and with very high cooling rates (up to 106 K/s) as well. Ultrasonic atomization has so far been used for low-melting alloys based on aluminum and is yet to be widely applied.
158
4 Powder Metallurgy
4.2.4.5 Vacuum Atomization
4.2.4.6 Rapid Solidification
Vacuum atomization, or soluble gas atomization, employs the difference in solubility of gas in liquid metal at different pressures to fragment the liquid into fine droplets. In this process, the molten metal is first supersaturated with a soluble gas under high pressure and the liquid stream is then suddenly exposed to vacuum. The supersaturated gas almost explodes out of the liquid fragmenting it into very fine droplets. A schematic diagram of the process is shown in Fig. 4-12. Alloys of iron, cobalt, nickel, copper and aluminum have been vacuum-atomized with hydrogen as the dissolved gas. The powders are generally fine and spherical. As this process depends exclusively on dissolved gases, it has only limited applicability and is not being used extensively.
Many techniques have been developed employing the advantages of rapid solidification processing (RSP) for utilizing metastable phase structures to advantage (see Chapter 2). Most of the conventional atomization processes discussed in the earlier sections lead to a normal cooling rate of 103 K/s. With some enhanced heat transfer techniques, several atomization techniques have been developed to achieve cooling rates of over 104 K/s. Such techniques with increased cooling produce rapidly solidified powders, flakes, wires and strips (Grant, 1983; Savage and Froes, 1984). Rapid solidification techniques developed (Table 4-5) are based on atomization, continuous casting, melt extraction and melt extrusion (Jones, 1981, 1982). In rapid solidification rate centrifugal atomization (Fig. 4-13), liquid metal is centrifugally atomized off a disc rotating at very high speeds (35 000 rpm) and the droplets cool by flight through high velocity helium gas streams to give powders in the size range 25-100 \im (Cox et al., 1976). An impinging melt stream contacts a moving chill surface in chill block melt spinning, planar flow casting and free jet casting to produce tapes with thickness 10-100 |im and widths 3-250 mm. In melt extraction processes, a rotating disc is brought into contact with a liquid metal pool for producing rapidly solidified fibers and tapes (see Chapter 2). While the rapidly solidified atomized powders can be consolidated directly, in the case of tapes and fibers, some comminution or shredding is necessary for further PM processing. Often the RS powders see application in the form of powder itself. For applications that call for consolidation of the RS powders, suitable consolidation techniques are employed with minimum
Vacuum
— Powder collection tank
Melting crucible
|— Powder drain Induction heater Vacuum
— Container
Figure 4-12. Schematic of vacuum atomization unit (Lawley, 1986).
4.3 Characterization of Powders
159
He gas input
H t
Figure 4-13. Schematic of rapid solidification rate atomization process (Joensson and Hohmann, 1987).
Air out
Table 4-5. Rapid solidification techniques. Process
Atomization Centrifugal Atomization Splat Extraction
Technique
Product type
Powder size, urn
Cooling rate, K/s
Application
Water atomization Ultrasonic gas atomization Rapid spinning cup Rapid solidification rate Electron beam splat quench Twin roll atomization CBMS PFC/MD CME/PDME
Irregular Spherical Variable Spherical Splat Flaky Tape < 3 mm Tape, wide Filament /Fiber
75-200 10-50 <50 20-100 40-100 200 10-100 20-100 20-100
10 2 -10 4 <10 6 106 105 4 10 -10 7 10 5 -10 6 105 107 10 5 -10 6 10 5 -10 6
Fe, Cu, Ag Fe, Cu, Ag, Al, Ni Fe, Cu, Ag, Al, Ni Fe, Cu, Ag, Al, Ni Fe, Al, Ni, Ti Fe, Al, Ni Fe, Al, Ni, Ti Fe, Al, Ni Fe, Al, Ni, Ti
exposure to high temperature, to retain the structural advantages of rapid solidification (Lee and Carbonara, 1985; Froes and Savage, 1986).
4.3 Characterization of Powders Characterization of powders is important because of their relevance in PM processing. Often the choice of a particular processing route is determined by the characteristics of the available powder. Powder
160
4 Powder Metallurgy
characterization is a complex procedure (Wang, 1982) as the properties reflect not only those of individual powder particles but of the powder aggregates and of the voids in the aggregates as well. These characteristics are listed in Table 4-6 (Poster, 1966). Particle size is considered one of the most important properties of the powder. It is customarily represented as a diameter, although not all particles are spherical. It is therefore usual in any measurement to consider the size and the shape together. A simple system of classification of particle shapes is shown in Fig. 4-14. The basic shapes include one-dimensional (large aspect ratio) acicular, rod-like and fibrous powders, two-dimensional (length and width much greater than the thickness) dendritic and flaky powders, and three-dimensional spherical, rounded, irregular and angular powders. The majority of the powders used in PM fall in the last category, and are represented by an equivalent particle diameter called simply "the particle size". A real powder aggregate does not have all particles of the same size, and One-dimensional Acicular
Irregular rod-like
Two-dimensional Dendritic
Three-dimensional Spherical
Irregular
Flaky
O
Rounded
o
Angular
O
Figure 4-14. System of particle shape characterization.
Table 4-6. Characterization of powders. Characteristics of a powder particle
Characteristics of a powder mass
Characteristics of porosity in a powder mass
Particle size
Average particle Total pore size volume Pore volume Particle shape Particle size between particles distribution Specific surface Pore volume Density within particles Surface Apparent density Number topography of pores Microstructure Tap density Average pore size Lattice defects Flowability Pore size distribution Pore shape Gas content Compactability (contained and adsorbed) Surface oxide Reactivity
therefore a powder description must also include the parameters defining the particle size distribution. The size distribution can be normal (unimodal), or bimodal, or with a broad band, or can be even irregular. Several techniques are used for the measurement of size and its distribution. Both direct and indirect measurement techniques can be used. Direct measurement techniques include sieving, microscopy, elutriation and sedimentation, while indirect techniques use turbidimetry, measurement of electrolytic resistivity, permeability and surface area. Sieve analysis consists of tapping powder through a set of standard sieves assembled in order in a mechanical sieve shaker. The quantity remaining on each screen is weighed to obtain the particle size distribution. The finest standard screen opening is 25 jim. For powders finer than this, the average particle size is obtained in a Fisher subsieve sizer. Essentially, the sizer determines the surface area of the powder by measur-
4.4 Compaction of Metal Powders
ing the resistance to the flow of air through a packed column of powder. This measurement is directly correlated with the average particle size. For sub-micron powders, the surface area itself is characterized directly, generally by gas adsorption technique in a BET apparatus. Apparent and tap densities of the powder are also important parameters in characterization and these depend on the particle shape, size and its distribution. Apparent density is defined as the weight of a unit volume of powder as poured through a standard Hall flowmeter. It varies widely from a high 50-60% of the specific gravity of the material in atomized spherical powders, to a low 5-10% in fine flaky powders. Tap density is the weight per unit volume of powder after tapping or vibrations to obtain maximum consolidation of the loose powder and is therefore higher than the apparent density. The frictional conditions of the powder which determine the apparent density are overcome during tapping and the tap density can be much higher. The flow rate of metal powders is a characteristic that may contribute nothing to the final part made from the powder, but a good flow is important for ease in compaction as it involves the movement of precise quantities of powder. This parameter is defined as the time taken for the flow of 50 g of powder through a standard Hall flowmeter. The flow rate is influenced by the particle shape, particle size and its distribution, the apparent density of the powder and its specific gravity. The powder is said to have no flow if the powder does not all flow out of the flowmeter.
4.4 Compaction of Metal Powders In conventional PM processing, the powder is generally consolidated into a
161
compact by shaping in a mold or a die, usually with the application of pressure. We use the term 'compaction' to describe consolidation without the application of heat. Compaction is a necessary step in powder processing as loose powder cannot be shaped without consolidation. This step also provides for the making of near-netshape parts which is an advantage in PM processing. Compaction is subsequently followed by sintering to make the final parts. Hot consolidation where external pressure and temperature are simultaneously applied is also a method of compaction and is now becoming increasingly popular for fabricating load bearing high temperature parts and ceramic components. Hot consolidation cannot be strictly defined as compaction as simultaneous sintering is also involved, and is described in a later section.
Table 4-7. Methods for forming compacts from powders. Methods with applied pressure
Methods without application of pressure
Loose powder sintering Unidirectional pressing in mold Single action pressing Double action pressing Vibratory compacting Isostatic pressing Powder rolling Slip casting Metal injection molding Stepwise pressing Explosive compaction
There are many methods of compaction (Table 4-7) and the choice is dependent on the application envisaged and the scale and economy of operation. Most of the methods depend on the application of external pressure on powder contained in a die or mold, while some use techniques like vibratory compaction or slip casting where the effect of external pressure is negligible.
162
4 Powder Metallurgy
4.4.1 Pressureless Compaction
Table 4-8. Packing of equal sized spheres.
In any agglomeration of powder, the powder particles acquire one of four geometric configurations. These configurations are shown in Fig. 4-15 and their packing densities given in Table 4-8 (James, 1972). As the size of the voids is determined by the particle size, it is possible to introduce particles of appropriate diameter in these voids. With such an addition the size of the voids reduces further, as also the total volume of the voids. The smaller voids can also be filled with particles and this sequence of operation can be carried till a high packing density of powder is achieved even before consolidation (Table 4-9). Though this system of extreme dense packing is not known to have been used in practical applications, it is a corn-
Packing
Coordination No.
Cubic Orthorhombic Tetragonal Rhombohedral
Porosity (%) 47.6 39.5 30.2 26.0
10 12
Table 4-9. Effect of size distribution on powder packing. Order Relative Relative Porosity pet. Weight pet. size of number in powder in final spherical of powder mixture mixture for particles particles with this optimum addition packing 1 2 3 4 5 6
1 0.414 0.225 0.175 0.117 fines
1 1 2 8 8 large
22.95 20.70 19.0 15.8 14.9 3.9
77.1 5.5 1.7 3.3 1.0 11.4
mon practice to use powder of appropriate size distribution to obtain high packing densities. 4.4.1.1 Vibratory Compaction
Figure 4-15. Basic coordination in packing of spherical powder: (a) cubic, (b) and (c) orthorhombic, (d) tetragonal, (e) and (f) rhombohedral (James, 1972).
An effective method of increasing the packing density of powder particles is by vibration. Mechanical vibration helps in settling particles in the voids present in agglomerates leading to the formation of nearly close packed powder configurations. This method has also an additional advantage of using only low pressures, thus avoiding generation of internal stresses within the compact. The process generally consists of vibrating a powder mass w contained in a mold, against a punch assembly with a mass W. A small pressure is usually superimposed by means of a pneumatic cylinder (Brackpool and Phelps, 1967). Compaction is effected
4.4 Compaction of Metal Powders
by both the direct force Fp and the cyclic force / v = iv • sin (co t). The peak vibratory compaction force Fc is given by W
(W+w)
(4-17)
Generally vibratory compaction is more effective than static pressure compaction for the less ductile powders. Such brittle materials as borides and carbides are prone to developing cracks when compacted under pressure, which can be avoided in vibratory compaction. But the handling strength of these compacts is low, as the particle bonding is only through mechanical interlocking. 4.4.1.2 Slip Casting
Slip casting is frequently used for compacting ceramic powders to complex and large shapes and for limited production runs. The slip is a suspension of the powder in water, which is poured into an absorbent (usually plaster of Paris) mold, dried and sintered. The formation of appropriate and consistent slip is important for successful slip casting and it is achieved by maintaining a proper control of particle size and size distribution, deflocculant addition, water content, viscosity and pH values. This technique is extensively used for ceramics but has also been extended to metals. 4.4.1.3 Injection Molding
Injection molding provides another method of consolidating metallic or ceramic powders without the application of stress (Billiet, 1985; Pease III, 1987). This method is very similar to the molding of plastics and consists of mixing powders and thermoplastic binders, and injection molding to required shapes. This is followed by debinding, sintering and post-
163
sintering operations. Injection molding is different from other compaction techniques in that it uses very fine particles: a typical particle size for injection molding is ten |im or less. The powder to be compacted is mixed with large quantities of thermoplastic binder. The volume of binder is determined by the rheological properties of the mix, linear shrinkage observed during processing and sintering and the uniformity of dispersion. Uniformity of the mix and a proper coating of all the powder particles with the binder are essential prerequisites for injection molding. Conventional plastic molding equipment is adequate for this compaction. Injection pressures and velocities are generally lower for powders than for plastics though mold temperatures tend to be high. The feedstock can be recycled a number of times in the molding machines so that the loss of material is avoided. An important part of the compaction is concerned with debinding, the stage at which the thermoplastic binder is removed before sintering. This is done by solvent extraction and evaporation. Many variations of this process are now available for injection molding and the process appears ideally suited for producing small but complex parts (Erickson and Amaya, 1984). 4.4.2 Unidirectional Compaction Unidirectional compaction is perhaps the most widely used consolidation process in powder metallurgy. In this process the powder is shaped in a die by pressing in a mechanical or hydraulic press. Die compaction is so well suited for process automation that today most of the large scale production of small sized PM components depend on this route. The only limitations countering still larger usage appear to be the size of the press and the mechanical
164
4 Powder Metallurgy
and geometrical constraints imposed by the size and shape of the compact. The densification mechanisms both in unidirectional and isostatic (Section 4.4.3) compaction depend on the material and the structural characteristics of the powder and also on geometrical factors such as the particle size, shape and size distribution. The surface characteristics of the powder are also important in compaction. Consolidation occurs in three overlapping stages: initially, under low pressures, interparticle friction is predominant and consolidation is therefore controlled by the geometrical aspects of the powder agglomerate. At intermediate pressures, elastic and plastic deformation of the powder occur and the material properties become important. At these pressures there is increased mechanical interlocking and cold welding of particles, leading to a good 'green strength' of the compact. ('Green strength' is the strength of a compact before any sintering or hot-pressing has begun.) In ductile powders, interparticle contact area increases with plastic deformation, while in brittle powders, massive fragmentation occurs at points of contact where the local pressures are high. Strainhardening in ductile powders also changes the rate of densification. Finally, at higher pressures, the forces transmitted within the compact become more isotropic and the particles tend to come into contact with 12 nearest neighbor spheres, forming a dodecahedral configuration with all faces in total contact. In practice, the compaction stops when the porosity becomes largely disconnected and when further deformation of the compact becomes merely elastic. Many mathematical equations are available to model unidirectional compaction. Among these, the following two have been shown to describe both unidirectional and
isostatic compaction rather well (James, 1983). The Kawakita equation relates the reduction A Fin the initial volume Vo of the powder mass with pressure P: (4-18) (A V/Vo) = 1 - (Qapp/Qcomp) = a b P/(l + b P)
where gapp and Qcomp are the apparent density of the initial powder and the density of the compacted mass. Eq. (4-18) can be simplified as P/(AV/V0) =
(4-19)
The constant a in the equation is related to @app, and b is a coefficient of compression related to the deformation characteristics of the powder. The Konopicky-Shapiro-Kolthoff equation takes the change in density to be proportional to pore fraction, leading to (dg/dP) = A (1 - £comp)
(4-20)
Two distinct regions of compaction are seen experimentally, the slope K in the first region being proportional to the inverse of yield stress. Above a critical pressure, the slope becomes proportional to the workhardening rate of the powder (Fig. 4-16). The constant A in Eq. (4-20) is given by In {1/(1 -£ a p p )}. The effect of initial particle rearrangement and restacking on @app can be obtained by extrapolation from the second region. In its simplest form, unidirectional compaction is done in a mechanical press with a suitable die for shaping the powder agglomerate and for ejecting the compact. Because of die wall friction and the nature of load distribution inside the die, densification of the compact is not uniform, particularly at the corners. This problem is minimized by using lubricants and also by using double action presses where the load is applied from the top and the bottom simultaneously.
4.4 Compaction of Metal Powders
165
as 4.5 GPa, in what has been termed as 'cold sintering' (Gutamanas, 1983). Tool steel dies and punches can be designed for use at pressures up to 3 GPa, while still higher pressures need cemented carbide inserts. Plastic deformation and fracture of powder particles in gradients of such high pressures lead to the bonding of freshly formed contamination-free surfaces. It may be possible to employ cold sintering to advantage in compacting composite and other multi-component materials, also rapidly solidified powders. 4.4.3 Cold Isostatic Compaction
Compacting pressure P Figure 4-16. Konopicky-Shapiro-Kolthoff relationship in powder compaction. Slope Kx is related to the material yield stress, and slope K2, to the work-hardening behavior of the material.
An important aspect of unidirectional compaction is the design of dies, which must take into account not only the shape of the compact, but also the ease of its ejection after pressing. This, in fact, limits the shapes that can be achieved by this method. Many factors are built into the design of dies to achieve correct dimensions and surface finish of the product (Goetzel, 1950; Kuhn, 1978). These include powder characteristics such as the shrinkage factor, lubricant characteristics, and also properties of the die material. The hardness, surface finish and dimensional tolerances that are achievable by machining determine not only the making of a die, but also the effectiveness of compacting. Typical compacting pressures are in the range of 100-400 MPa. Recent developments indicate the use of pressures as high
There are many limitations to die compaction. The size and shape of the compact are limited by the capacity of the press and also by the geometry of the part. The necessity of fabricating complex dies and provide for ejection after pressing, and die friction, makes unidirectional pressing unattractive for some applications. Cold isostatic pressing provides an alternative method for such applications. In this method, powder in a mold is subjected to a hydrostatic pressure, commonly referred to as isostatic pressure, of about 200 to 500 MPa for a short time. The pressure is then released and the compact removed from the mold. Sealed molds of powder are loaded in a liquid inside a high pressure chamber, and hydrostatic pressure is applied by pressurizing the liquid to the required level. Hydrostatic pressure does no work on a homogeneous solid. However, in a porous material, this pressure is resolved into shear stresses which are large near the pores but decrease with distance. With increasing density (decreasing porosity), these stresses become negligible and compaction ceases. The effective pressure Peff can be calculated on the basis of the applied pressure
166
4 Powder Metallurgy
Papp and the volume porosity 9 of the preform and can be shown to be app
1-/
Vent valve Sealing nut Cover
(4-21)
A detailed derivation of this equation is given in Section 4.6.1. Initially, the resolved shear stress densifies the compact by particle sliding and rotation. This stage is determined, as in unidirectional compaction, by the packing geometry and the flow properties of the powder aggregates. In the second stage, deformation of powder particles takes place. Experiments on particles with different geometries have confirmed these observations. Even though spherical particles initially have high packing densities, they do not mechanically interlock with one another and do not easily deform; higher pressures are therefore required for their compaction. This is not the case for irregularly shaped particles which interlock with one another and also deform during both the stages. A cold isostatic pressing unit consists of a thick pressure vessel with an appropriate top closure which can either be threaded to the vessel itself or axially loaded with appropriate seals from an external frame. The pressurizing medium in the vessel is generally a liquid. Wire wound external frames are advocated to support the pressure vessel which is subject to both radial and axial stresses. The process can be automated so that loading, pressurization, depressurization and unloading are carried out in a proper sequence to ensure almost continuous production. Two different types of isostatic equipment are available, depending on the nature of the mold chosen for isostatic pressing (Jackson, 1967). These are known as wet-bag tooling and dry-bag tooling (Fig. 4-17). In wet-bag tooling, the mold, made from natural rubber (latex) or flexi-
_ Back-up ring O-Ring - — Pressure vessel Plug Envelope Powder
Pressure relief
Cover plate
Pressure vessel
(b)
Perforated support
Powder
En\e|
Figure 4-17. Schematic of cold isostatic pressing: (a) wet bag technique, (b) dry bag technique (James, 1971).
4.4 Compaction of Metal Powders
ble elastomeric material such as neoprene, polyvinyl chloride or polyurethane, is filled with powder, evacuated if required and then sealed. A number of such molds are loaded into the pressure vessel and pressurized, and after the compaction dwell time has elapsed, the compacted molds are removed from the vessel. This method is versatile and appropriate for low volume production. Dry-bag tooling consists of molds made from rubber rigid enough to retain shape, firmly fixed as in conventional type presses. The powder is filled into the mold automatically from a hopper, the top punch of the press closes the mold, and isostatic pressure is applied. After pressing, the pressure is released, the mold returns to its original size and the part is removed by ejection using the top or bottom punch. The size of the isostatically pressed part is limited only by the volume capacity of the press and of the mold. Depending on the applications envisaged, parts are pressed as billets, forging preforms or nearnet shapes. The process parameters include the pressure, dwell time and depressurization rate. Isostatic pressing is widely applied in compacting porous filters, high speed steel billets, titanium and superalloy near-net shapes and large cemented carbide parts. A recent application is in the compaction of metal-matrix composites. 4.4.4 Explosive Compaction
High energy rate forming using explosives is also used to advantage for compacting powder aggregates. Explosive compaction is made possible by the detonation wave front generated by the shock wave, which can travel with a velocity as high as 7000 m/s, developing pressures up to 28 GPa and raising the local temperature up to 5000 K. Various parameters of
167
explosive compacting are determined either experimentally or calculated, using the properties of the explosive and the products of the explosion. There are many variants of this process of explosive compaction which can be broadly divided under three headings (Fig. 4-18): contact operations, where an explosive charge is located on the surface of the container with powder aggregates generating normal or tangential incidence of shock waves on explosion; remote operations, in which the high explosive charge is placed at a distance from the powder being compacted and the pressure transmitted through an intermediate medium like water, air or even a metallic flying plate or a punch; and explosive isostatic pressing, where a large isostatic pressure is generated in a closed volume by an explosive charge. In some recent developments, dynamic compaction is achieved by using a projectile accelerated by electromagnetic force instead of an explosive charge. In explosive isostatic pressing, the explosive charge is kept in a specially designed chamber on top of a pressure vessel. The ignition of the charge generates a pressure wave which moves the piston inside the pressure vessel pressurizing the medium (usually water) containing the powder mold. The dynamic pressure P can be large, around 0.6 GPa, and can be calculated using the following equation (Roman and Gorobtsov, 1989) P=
1-5K.
(4-22)
where / is the propellant force characterizing the efficiency of the propellant under isobaric expansion, Vx9 the volume of the propellant gases for unit mass propellant and S is the charge density i.e., the ratio of the mass of the propellant to the volume of the combustion chamber. The value of S is
168
4 Powder Metallurgy
Detonator Card board tube
Detonator assembly -Explosive
Wooden cone Projectile
Steel tube .Barrel
Explosive
. Die Powder . Anvil Die holder^ assembly Detonator
(b)
, Piston
, Pressure vessel . Powder
Figure 4-18. Schematic of explosive compaction: (a) contact compaction (Bhalla and Williams, 1976); (b) remote compaction (after Raybould, 1981); (c) explosive isostatic compaction (Roman and Gorobtsov, 1989). (c)
typically equal to 0.7 kg/dm3, for which the pressure can be as high as 2000 MPa. The microstructure of explosively compacted powder shows intense deformation of powder particles, often seen as bands. Sometimes the pressure can be very high, leading to local melting and even rapid solidification. The density of the compact can also be high, often in excess of 95%. Ductile materials show intense deformation, while brittle solids fracture into fine particles which are themselves filled with cracks.
There are also other interesting effects associated with the interaction of shock waves with powder. For instance, the shock wave compression of graphite powder for synthesis of diamond is well known. Similar results are available for producing cubic boron nitride and some other refractory compounds. Another effect is concerned with the shock-wave induced activation of powders. Shock-wave-induced deformation produces a very high dislocation density in powders and this in turn
4.4 Compaction of Metal Powders
(a)
169
(b)
Gravity-fed powder -.
Gravity-fed powder - Hopper
(c) powder forceHopper
Hopper
influences their subsequent sintering behavior. Many refractory compounds sinter better, often without binders, after the powder has been subjected to explosive treatment. Some advantages of this kind of activation are now only beginning to be realized in sintering multi-phase materials. 4.4.5 Powder Rolling
Roll compacting, or powder rolling, is a process of cold compacting metal powder continuously in a rolling mill. In this process, the powder is fed from a hopper to a set of rolls to produce a continuous green strip or sheet (Dube, 1981, 1982, 1983; Ro et al., 1982,1983). The strip is then sintered and re-rolled to produce the finished product. As in conventional die compaction, the starting material can be elemental, blended or prealloyed powder with good flowability and adequate compactibility. For powder rolling, the rolls can be positioned horizontally, vertically or even at an inclined angle (Fig. 4-19) (Ro and Toaz, 1982; Knopp et al., 1984). Horizontal roll
Figure 4-19. Powder rolling: (a) and (b) vertical powder feed, twohigh and four-high mills; (c) horizontal powder feed; (d) inclined angle powder feed (after Dube, 1981; Evans and Smith, 1959).
positioning using gravity feed offers a good flexibility in controlling powder feed and also in introducing different powders simultaneously for layered rolling. The important parameters in powder rolling are the control of powder feed, roll diameter, edge control, sintering, optional re-sintering and rerolling. The powder feed between the rolls from a hopper is controlled to ensure uniform thickness and uniform density of the rolled strip. Height of the powder bed in the hopper is kept constant, to ensure uniform pressure on the powder in the roll gap. In some cases, the powder feed to the roll gap is controlled by means of adjustable gates, to take into account variations in compacted density. Edge control is essential in powder rolling to ensure that the edge is as well formed and as dense as the center of the strip. This is achieved by floating flanges attached to one of the rolls and overlapping the other roll. A relatively high green density is required for handling the powder-rolled strip for sintering. It is mainly the roll diameter
170
4 Powder Metallurgy
that determines the strip density and thickness. Even though the nip angle in powder rolling is about the same as in conventional metal rolling, the arc included by the nip angle on the roll surface depends on the roll diameter. Consequently, a larger amount of powder is pulled into the roll gap by larger diameter rolls and a higher density strip is produced. The strip density is also determined by the coefficient of friction and the roll temperature. With higher friction, larger amount of powder is pulled into the roll gap to give a higher strip density. Higher temperature of the rolls also increases the friction with the same effect. In practice, the roll gap and the roll friction are adjusted to give a green density of 75 to 90% of the theoretical density. There are many options available in handling compacted powder strips for sintering. The compacted strip can either be coiled and sintered afterwards, or can pass directly to the sintering furnace as it comes out of the rolls. Studies on products manufactured from powder strips show that powder rolling has reached a level of technological maturity to meet the demands of many applications and of the forming processes usually encountered in manufacturing operations.
4.5 Sintering An important step in powder processing is the densification of compacted powder aggregates. This is carried out by heating the powder compacts at temperatures below their melting point and is known as sintering. The driving force for sintering is not large and is provided by the reduction in surface energy of the powder. Often, this energy is clearly inadequate to eliminate all the porosity present in the material. In addition, there is also a slight increase in the free energy of the compact because of the
creation of new grain boundaries at junctions where powder particles meet. To provide a larger driving force, sintering is sometimes carried out with very small additions of reactive elements which alter the balance between the surface energy of the particles and the grain boundary energy favorably for sintering; or, sintering with an additional component which becomes a liquid phase at the sintering temperature. External pressure which induces plastic flow can also be applied concomitantly with sintering for a fuller densification. There are also instances where sintering is followed by mechanical processing and annealing to eliminate any residual porosity. Such post sintering processing is usually required when load-bearing and other critical applications are envisaged. 4.5.1 Solid State Sintering
Solid state sintering can be broadly divided into the following stages (Thummler and Thomma, 1967): (i) bonding of powder particles forming necks; (ii) shrinkage of the compact due to decrease in porosity and change in pore geometry; and (iii) isolation of pores followed by homogenization of composition and grain growth, with the elimination of residual porosity. These processes do not occur in discrete or serial steps, but overlap, a particular process being predominant at a specific temperature and time. The driving force for mass transport that leads to solid state sintering is the change of surface energy by reduction of the surface area, dAs. The corresponding change d£ in the free energy of the system is then given by dE = y dAs
(4-23)
4.5 Sintering
where y is the surface tension. The variation in surface curvature which occurs with the growth of the neck that joins individual powder particles is thus reflected in the energy change dE with respect to the change in volume dV at the neck: /A . .. (4-24) dE/dV=y • d where rx and r2 are the principal radii of curvature. The term (d£/dF) formally corresponds to the stress a which is tensile on the concave surface of the neck and compressive on the convex side and is given by a = y[(l/r1) + (l/r 2 )]
(4-25)
Similarly, we can define a change in chemical potential between the two curved surfaces and this is seen as a gradient in vapor pressure p and vacancy concentration c: (p/p0)
(4-26)
and (c/c0)
Figure 4-20. Mass transport mechanisms in sintering: (A) lattice diffusion from grain boundary; (B) lattice diffusion from surface; (C) lattice diffusion from defects; (D) boundary diffusion from grain boundary; (E) surface diffusion from surface; (F) vapor phase transport.
Table 4-10. Mass flow mechanisms in solid state sintering. Mechanism
Mass flow source
Mass flow sink
Contribution to shrinkage
A. Plastic flow (viscous flow in non-crystalline materials) B. Vapor phase transport C. Surface diffusion D. Lattice diffusion E. Lattice diffusion
Matrix
Neck
Yes
Surface
Neck
No
Neck Neck Neck
No No Yes
Neck Neck
Yes Yes
(4-27)
These gradients lead to material transport from convex to concave surfaces by plastic deformation, or evaporation and condensation, or by vacany diffusion (Fig. 4-20). Part of the mass flow occurs merely around the neck region and does not result in shrinkage, while other parts readjust the entire particle surface, leading to the shrinking of pores (Table 4-10). A chemical driving force is also present in multicomponent systems due to differences in diffusion rates among the components. Initially, these differences can lead to inhomogeneous compositions at the neck, and porosites at sites from where the faster diffusing species have migrated. This phenomenon is known as the Kirkendall shift (see Vol. 5, Ch. 2, Sec. 2.2.2.4) and can lead to the creation of porosities at least during the initial stages of sintering. However, as sintering proceeds, re-homogenization
171
Surface Surface Grain boundary F. Lattice diffusion Dislocation G. Grain boundary Grain boundary diffusion
takes place in the entire powder compact, removing the initial inhomogeneities seen on the necks formed between particles. 4.5.1.1 Equations of Sintering
There have been many models, reviewed exhaustively by Exner (1979), to describe
172
4 Powder Metallurgy
the growth of necks and shrinkage of pores, but a quantitative model still remains to be developed because of difficulties associated with the modelling of spheres of different radii and with the constant changing of positions and particle contacts during sintering. It is possible, however, to develop an idealized twosphere model with equal sized spheres in contact with one another, the contact area becoming a neck and growing with time. A great deal of work has been done using this approach and this helps us in identifying the various mechanisms of sintering. In general, sintering of two spheres of the same diameter can be described by an equation of the type (Kuczynski, 1949)
i] =ka
(4-28)
where x is the radius of the neck, a, the particle radius and t, the time of sintering, while n, k and m are parameters which depend on the rate-controlling mechanism. Table 4-11 gives the range of values for two of these parameters for the different mech-
Table4-ll. Equations for neck growth for different mass flow mechanisms in solid state sintering. Basic form of equation: (xn/am) =F (T) • t
F(T)
Mechanism Viscous flow Vapor phase transport
3 y/2 r\ K y ft1-5 g°-5 Vjj'
AL h — = -
1
=ktn
(4-29)
Here, AL is the linear shrinkage for the original length L0,h, half of the distance of approach of the two particle centers and t is the time for sintering with k and n as appropriate constants. Another approach to analyze sintering is by evaluating the changes in surface area of the powder system. On the basis of statistical modelling, the following relations for changes in surface area As during sintering have been derived by German and Munir (1975, 1978) as (4-30)
dt 2
1
^
j
and
1
(RT) (
Mi 5-7
2-3
V VMR Surface diffusion
K y Q ws DJRT
Lattice diffusion
KyQ033VMDJRT KyQD
Grain boundary diffusion
anisms. Analysis of Eq. (4-28), using idealized experiments, show that diffusional creep (Nabarro-Herring and Coble creep), involving lattice and grain boundary diffusion, play a more predominant role than either vapor transport or plastic flow. However, the problem becomes very complex when we consider sintering of real powder particles with different particle sizes and geometries (Exner, 1980). There have been attempts to model densification in real systems by relating the compact shrinkage to center-to-center approach of the particles. A classical equation of this type is given by Kingery and Berg (1955)
RT Ky Q whDh RT
1
3
3-6
1-2
4-5
1-2
6
2
\ae+l
(4-31)
dt Eq. (4-30) is applicable during the initial stages of sintering and Eq. (4-31), for the intermediate stage. The exponent m is based on the initial density and sintering mechanism, while Bs is a microstructure based constant; As0 is the surface area at the start of the grain growth and ag, a grain
4.5 Sintering
173
growth parameter. However, these equations are difficult to verify experimentally because of constantly changing particle arrangements during sintering.
Table 4-12. Rate equations used in Ashby's sintering mechanism maps. F = yQ/kB T; Kl9 K2, K3: surface curvature difference of the pore.
4.5.1.2 Pore Closure and the Development of Microstructure
Stage 0 Adhesion x( 0 = ca2/x for x<(a2/10fis)033 Stage I Surface diffusion from surface x1=2wsDsFK3l Lattice diffusion x2 = 2DyFK\ from surface Vapor phase transport :i3=PF(Q/2 n Q0 kB T) 0 5 Kx Boundary diffusion from grain boundary x4 = 4whDhFK2/x Lattice diffusion from x5=4DvFK2 grain boundary Lattice diffusion x6 = (4/9) K2Nx2 D V Ffrom dislocations • {K2-(3 fisx/2ya)} Stages II and III (l/16)whDhFK33 Boundary diffusion from grain boundary Xl {\n(xfK3/2)-0J5)}
A green compact consists of particles and pores. With the beginning of sintering, grain boundaries form in the contact area between particles. The growth of necks leaves a continuous network of pores which has a well defined relationship with the grain size of particles surrounding them (Zener relationship) (Exner and Arzt, 1983). With further sintering, the continuity of the pore network breaks down and the pores get isolated at grain boundaries. Both densification and coarsening of pores as well as grains occur simultaneously during the final stage of sintering. The pores can coarsen by either diffusional mass flow between the pores, or by coalescence of the pores due to grain growth (Kaysser, 1988). With grain growth, it is also possible that pores get "trapped" inside grains. Such pores are not easily removed as they can be closed only by volume diffusion, which is a slow process. The relative values of surface and grain boundary diffusion determine the extent of such trapping of pores within grains. The choice of sintering temperatures and other parameters that determine the diffusion behavior is therefore critical in controlling the excessive grain growth and in minimizing the residual porosity. 4.5.1.3 Sintering Mechanism Maps
On the basis of the equations of sintering (Table 4-12), Ashby (1974) has developed a system of graphs which are known as sintering mechanisms maps. These maps are drawn for the normalized neck size (x: a), linear shrinkage or relative density against
Mechanism
Lattice diffusion from grain boundary
Neck growth rate equation
{1/16) DyFK33 *8
{\n(x{K3/2)-0J5)}
temperature. The regimes where a particular sintering mechanism predominates are marked on this map, as exemplified in Fig. 4-21. Although these maps have been generated for ideal conditions, they povide a convenient framework for appreciating the specific requirements for sintering. 4.5.2 Liquid Phase Sintering
While sintering multicomponent systems, it is possible to choose a component which will be in the liquid state at the sintering temperature, thereby accelerating the sintering. This has a strong practical interest as many products based on tungsten, tungsten carbides, ceramics and rare earth materials are commercially sintered this way. Liquid phase sintering takes place in three stages (Kingery, 1959; Kaysser and
174
4 Powder Metallurgy Temperature (°C) 600 700 800 I
Full I density
boundary diff. from boundary 2 / A Volume diff. from
Volume diffusion from boundary
Surface diffusion from surface Boundary diffusion from boundary I I 0.6 0.7 0.8 0.9 Homologous temperature T /Tm
(b) 400 I
500 T
Temperature (°C) 600 700 800
900
10-3
I
Full I density I Boundary diff^fromj>6undary 2 /3 ^ '\^ol.diff.f.boundary
Boundary diffusion from boundary
Surface diffusion from surface
Adhesion -2.0 0.5
I I J_ 0.6 0.7 0.8 0.9 Homologous temperature T /Tm
10"5
1.0
Figure 4-21. Sintering mechanism maps for (a) silver spheres of radius a = 100 urn, and (b) silver spheres of radius a = 10 um, both with an initial relative density of 0.8 (Ashby, 1974).
4.5 Sintering
175
Solution-reprecipitation can be rate controlled by either one of the two steps, dissolution and diffusion (Huppmann, 1979; Huppmann and Petzow, 1979). Diffusion controlled shrinkage in this stage follows: zlL\3_
(4-34)
For interface reaction control the shrinkage follows Time
Figure 4-22. Mechanisms of liquid phase sintering: Stage 1 - rearrangement; Stage II - solution-reprecipitation; and Stage III - final solid solution sintering (German, 1985).
Petzow, 1985). The effect of the three stages on densification is shown in Fig. 4-22. Initially, the liquid phase (for example, Co with WC or Cu with W) wets the particles leading to rearrangement of the particles and densification. With the capillary pore size decreasing with sintering, the stress from the surface tension of the liquid phase increases. Hence the shrinkage from the viscous flow in this stage is not directly proportional to sintering time t, and the relative linear shrinkage is given by AL
fALY = k4
(4-35)
The constants k± and k2 are related to diffusivity and reaction rate respectively of the solid in the liquid. The solution-reprecipitation sequence leaves a skeleton of solid phase with liquid squeezed out to larger voids in the solid and on to the surface. The microstructure of a liquid-phase-sintered heavy alloy system is shown in Fig. 4-23 and indicates the liquid phase component solidified in such a network. In the third stage of sintering, the solid skeleton sinters just as does a normal powder compact. The three stages are sequential, and the kinetics of liquid phase sintering is more sensitive to temperature than to sintering
(4-32)
where x is a small fraction. In the second stage, the solid phases dissolve in the liquid and reprecipitate elsewhere due to differences in solubility between small and large particles. This difference in solubility is given by the Gibbs-Thomson equation: S
Sr\ —
2ys0Q RT
(4-33)
where s and s0 are the solubility of a particle with radius r and that of a flat surface, respectively, and Q is the atomic volume.
Figure 4-23. Microstructure of a liquid-phase-sintered tungsten heavy alloy, showing the two phases (Courtesy: DMRL).
176
4 Powder Metallurgy
time. The initial volume fraction of the liquid phase can be as high as 35%. The use of fine particle size and large volume of liquid enhances sintering. Coarse particle sizes and high green densities sinter poorly as the compact swells instead of densifying. There is a particular case of liquid phase sintering where the liquid phase is transient (German, 1990). With the dissolution of the solid phase, the liquid composition changes, raising the melting point. Some instances of fragmentation of particles during the initial stage of sintering are also seen when alloy powders are sintered at supersolidus temperatures. This arises due to the formation of liquid phase in each particle which leads to fragmentation. The subsequent re-arrangement step would then be the rearrangement of fragments rather than that of particles.
lower melting point than the base metal, with a high solubility in the base metal and a low solubility for the base metal. The amount of activator addition is also important: excessive activator can lead to local clusters retarding grain boundary movement. Activation of sintering is also produced by other methods. These include controlling the sintering atmosphere, increasing defect structure by shock, and controlling the microstructure by stabilizing a higher diffusivity phase. Although the activated sintering of refractory metals has been studied extensively, it has not found industrial application. This is because the accompanying brittleness problems have not been overcome yet. However, activated sintering has found industrial application in the case of steels and superalloys in a process called
4.5.3 Activated Sintering
(CAP) (Dax, 1983). In this process, boric acid in a methanol carrier is used as the activating agent, and the powders are sealed in evacuated glass molds. Sintering is activated by the boric acid and is aided by the atmospheric pressure acting on the mold, to achieve over 95% density.
Uniform and monatomic layers of certain transition metals coated on powder particles lower the activation energy of sintering and this is known as activated sintering (Reshamwala and Tendulkar, 1970). Because of its potential in lowering sintering temperatures, this phenomenon has been studied extensively both in refractory metals and in ceramics. An analysis of experimental results shows that activators, by their presence on the grain boundaries, enhance the grain-boundary diffusion rate significantly, leading to better sintering (German and Munir, 1982). For instance, the presence of nickel, an activator, on the grain boundaries of tungsten increases the grain boundary self diffusion coefficient by a factor of 500 to 5000 at 1675 K (Kaysser, 1988). The choice of an activator for sintering is dependent on its meeting some specific requirements. For instance, it must have a
consolidation
by
atmospheric
pressure
4.6 Deformation Processing at High Temperatures Fully dense compacts of metal powder with controlled microstructure can be produced by hot consolidation, in which pressure and heat are applied simultaneously. Various techniques have been developed for the application of pressure: static or dynamic, uniaxial or isostatic pressures have all been used for one application or another. The available major techniques include (uniaxial) hot pressing, hot iso-
4.6 Deformation Processing at High Temperatures
177
static pressing, powder forging and powder extrusion. 4.6.1 Principles of Pressure Sintering
When a powder aggregate is subject to high pressure and temperature, densification occurs in three stages. In the first stage, the aggregate is almost instantaneously consolidated to high density by plastic deformation. No further closure of pores is possible by plastic collapse as the external pressure reaches equilibrium with the internally developed stresses. The extent of densification is of course dependent on the applied pressure: with sufficiently high pressures it is possible instantaneously to attain densities as high as 95%. In the second stage of densification, various creep mechanisms become operative. At higher stresses, dislocation creep is prominent. But with increasing pore closure the stresses become low and diffusional creep becomes more prominent. In the final stage, with the increased curvature of pore surfaces, diffusion as in the case of normal sintering becomes the dominant mechnism of densification. Thus densification is seen as the net effect of local strains which combine to give overall contractions; these strains arise because of an effective stress present in the material due to the externally applied pressure. An expression for the effective pressure has been proposed by a number of workers and can be derived from an analysis of the stress pattern when a hollow sphere is subject to an external pressure P. Results of this analysis show that the effective pressure Peff, or the average effective stress <7eff, is
where P is the applied pressure and 0, the porosity. Within the compact itself, the ef-
Figure4-24. Effective stress on particle contacts in isostatic pressing (Courtesy: DMRL).
fective stress varies, depending on the particle size and the contact characteristics (Fig. 4-24). Many expressions have been derived to describe the first stage when the plastic collapse takes place. A closed form solution derived by Carrol and Kim (1984) for the applied pressure P using a hollow sphere model has the form P = §
(4-37)
where 8 =
Here
(4-38)
which was derived by Torre. The plastic collapse stops once the pressure reduces below the critical value of P in Eq.(4-37). The second stage of densification is characterized by a time dependent deformation which is driven by the effective stress
178
4 Powder Metallurgy
power law creep processes, the net creep rate can be approximated to the sum of individual creep rates: ^Z^eff)1"
(4" 39 )
i
where s is the total creep rate, A{ and nt are the creep constants for a particular mechanism i. At this stage of densification, it can be easily shown that the overall densification rate 9 is related to the porosity as well as the overall strain rate due to creep and is given by, 9 = (s
+£
+s
)(l—0)
(4-40)
where ex0, £y0 and sz0 are the strain rates in three perpendicular directions over the compact as a whole. The rate of densification during diffusion creep was derived by Coble (1970) and for the entire sequence of pressure sintering by Wilkinson and Ashby (1975). Wilkinson and Ashby (1975) have derived separate sets of equations to describe pressure sintering at the initial, intermediate and the final stages of sintering. Caligiuri (1980) has derived a semiempirical equation for pressure sintering (a) 800
Temperature, °C 1000
based on experimental results on iron and steel powders which is found to be valid for copper, aluminum and alumina powders as well. In this equation the pore-closure rate is given by APn
where A and n are material constants. As with sintering (see Section 4.5.1.3), densification mechanism maps have been evolved by Arzt et al. (1983) for pressure sintering also. A typical mechanisms map is shown in Fig. 4-25. Such maps are useful in visualizing regimes of various mechanisms operative during the pressure sintering cycle, and design the cycle parameters for maximum densification. An equation for pressure sintering using finite element analysis has been given by Bhatt etal. (1981). This equation was derived by computer simulation of a powder aggregate under isostatic pressure and temperature. By measuring individual displacements at various points in the aggregate and by imposing the necessary boundary conditions, the strain rates can lbl
1200
Pressure, Mn/m 2 10 100
1.0
^ 0.9 c 0) T3
Power-law creep
(4-41)
a> 0.8
5
Ah >^
w7 f
1000
/creep^/ law / / N-H/C 1 Power-/ creep / / Yield
£ 0.7 —
_J 0.7 0.8 0.9 Homologous temperature, T/Tm
0.6 -2
i 1 -1 0 1 Normalized pressure, log(P/o y )
Figure 4-25. Pressure sintering mechanism maps for tool steel: (a) density-temperature map for particles of radius a = 25 jum and mean grain size 50 um at pressure P = 100 MN/m 2 ; (b) density-pressure map for particles of radius a = 50 urn and mean grain size 10 \xm at temperature T=1475 K (Helle et al., 1985).
4.6 Deformation Processing at High Temperatures
be calculated using Eq. (4-41). The corresponding effective pressure Peff for these boundary conditions is seen to be the same as that derived from Eq. (4-40). Using this effective pressure, the pore closure rate 6 is given by
where A and n are dense material creep parameters appropriate to the operating pressure. Fig. 4-26 shows a comparison of this "effective stress" model with the other two models described in this section. The above analysis is broadly valid for uniaxial pressing as well except for some differences: the effective pressure is marginally lower and varies exponentially with the coefficient of friction and with distance from the surface of the plunger. 1CT3
i
i
|
Commercial purity Fe powder
Wilkinson-Ashby eqn. ^ \
#v
Caligiuri eqn.
"5 I/)
Average stress ' 1 0 " b -- model
• Experimental data by Caligiuri
r=700°C I
0.20
1
0.16 0.12 Porosity 9
1
1
0.08
Figure 4-26. Comparison of the average stress model, Caligiuri model and Wilkinson-Ashby model for pore closure in pressure sintering of commercially pure iron (Bhatt et al., 1989).
179
4.6.2 Hot Isostatic Pressing Hot isostatic pressing (HIP) was originally developed for cladding nuclear fuel elements, but is now widely adapted for powder consolidation to complex shapes in many areas of technology. The advantages of HIP include its capability for forming complex shapes, for consolidation at a lower temperature and for handling radioactive and toxic materials. In its simplest form, HIP of powders is a process where the pressure is applied isostatically on a hermetically sealed powder can inside an autoclave. The powder to be compacted is tap filled in a shaped container, degassed under vacuum at elevated temperature and sealed. The canned components are loaded into the HIP pressure vessel, and processed through the pressure/temperature/time cycle for consolidation and sintering. After HIP, the can material is stripped off the component by acid leaching or machining. A schematic layout of the HIP system is shown in Fig. 4-27. Argon is the most common pressurizing gas. Gas pressures of over 200 MPa are possible, which is higher than the pressure normally used for uniaxial hot pressing. Complex sealing arrangements are employed for the pressure vessel. Temperatures of up to 1450 K are routine with Kanthal resistance heating and higher temperatures up to 2300 K are also attainable either with graphite or molybdenum heating elements. The various sequences in the process of hot isostatic pressing are shown in Fig. 4-28. There are modifications to this sequence when special processing steps are considered. An important commercial application of HIP is in the manufacture of cemented carbide parts. For this, the equipment is designed to combine both the dewaxing and sintering operations with HIP. In other instances, HIP is the primary con-
180
4 Powder Metallurgy
Pressure vessel
Heat Exchanger
Mantle Furnace Tooling Thermo couple —
Controls
Figure 4-27. Schematic of a hot isostatic press (Price, 1984). Pump
(2)
(5)
(3)
i m P
Ik t
Figure 4-28. Processing steps in hot isostatic pressing for powder consolidation: (1) fabricate can to fit formed part; (2) fill powder in can, evacuate, degas and seal; (3) HIP; (4) strip can to give (5) consolidated part (Meiner and McCall, 1981).
solidation step, followed by secondary metalworking operations. The economics of HIP for powder consolidation is a crucial concern in the choice of the technology. The high capital costs involved in the setting up HIP facilities can be justified only when there is a large output or when special processing conditions are envisaged as in the manufacture of nuclear fuels, jet engine turbine components of high temperature composite materials. There are also developments in automating the process and in reducing the cycle time to improve the production efficiency.
4.6.3 Uniaxial Hot Pressing Uniaxial hot pressing is a relatively old route and was first developed to consolidate refractory metals and alloys. In this technique, as in cold compaction, pressure is applied on the powder in a die either in a single acting or in a double acting press. In addition, the press has facilities for heating the die in controlled atmosphere or in vacuum. Depending on the application envisaged, hot pressing tools are made from TZM molybdenum, tungsten, stellites, hot die steels, cemented carbides, ceramics or graphite. Metallic dies and punches used have to be water cooled to avoid plastic deformation. The alternate materials graphite and ceramics are strong, but are prone to failure by brittle fracture.
4.6 Deformation Processing at High Temperatures
Hot pressing consists of charging the powder in the cold die, heating the die assembly to the pressing temperature and applying pressure. Sometimes preformed pellets are preferred to avoid cooling the die each time for charging powder. A controlled atmosphere is essential in hot pressing to protect the die and the charge material, and is provided either by vacuum or by bleeding an inert gas like argon into the furnace chamber. The die is heated either by resistance heating or by induction. There are also instances where high electric currents are passed directly through powders in ceramic dies. In some instances, the filled die assembly is heated separately in a furnace before transferring it to the press. Until recently this process was not considered economically viable because of the costs of the die, the complex furnace setup with protective atmospheres and the long process cycle. Currently the increased availability of ceramic and graphite dies, coupled with an awareness of the benefits of hot pressing, has brought a resurgent interest from PM manufacturers in this technology, which promises good growth in the coming decade.
181
4.6.4 Powder Forging
Forging a sintered preform to the required shape and density is yet another useful technique of deformation processing in PM. This has become a popular process for forging steels and aluminium alloys for automobile applications. In this method, the powder is first compacted to a preform shape, sintered and then press-forged to the final shape and full density. The processing steps are shown schematically (Brown, 1981) in Fig. 4-29. In powder forging, there are some conflicting process requirements with regard to the desired density and shape of the preform. For instance, a high preform density gives a better strength and oxidation resistance, but leads to greater wear on the compaction tooling. Accordingly, the preform density is maintained around 75-80% of the theoretical value. The preform shape is dictated both by the shape of the final part, and by the extent of the material flow desired. Good flow enhances the mechanical properties of the forged component, but increases die wear. In optimal preform designs, the shapes are so chosen that the
(2)
m m
i
i
I (3)
m Q .'ii',7':.*.'
8
Ii
i
m U)
— *
Figure 4-29. Steps in the powder forging process: (1) fill compacting die with powder and (2) compact to required density for preform; eject compact and sinter in controlled atmosphere in (3) preheat furnace; transfer preform rapidly to (4) forge press and press forge to full density part.
182
4 Powder Metallurgy
maximum material flow occurs in areas of the component that require maximum strength. In the absence of friction, the change in density Q during unconfined forging is given by (James, 1982): = dsr
+ dez
(4-43)
(b)
where sr, se and ez are the radial, circumferential and axial strain increments respectively. The strains are interrelated as dsr = dse = v dsz.
If the Poisson's ratio v is also incorporated, the densification equation becomes (c)
An
£rel
-Q°J2)
= zz
(4-44)
where grel is the relative density. For closed die forging (repressing), the densification expression simplifies to — = dez
(4-45)
Figure 4-30. Schematic of powder extrusion process: (a) powder is canned, evacuated and sealed for the extrusion billet; (b) with loosely packed powder on extrusion the can may fold; (c) penetrator technique used to avoid folding of can.
Q
Finite element methods (FEM) are also available for designing the forging preforms and for calculating the stresses and density distributions in various parts of the forging (Kobayashi et al., 1989). 4.6.5 Powder Extrusion
Powder extrusion differs from its conventional counterpart only in the need to have the extrusion preform canned first and dejacketed after extrusion. The process offers the advantages of shaped extrusions and complete densification in a single consolidation step. The process (Fig. 4-30) consists of canning the powder, evacuation and sealing, followed by extrusion (Gardner et al., 1967). Often some degree of cold or hot compaction on the powders is employed prior to extrusion. The can serves
as a container not only for degassing, dehydriding or annealing of powders before extrusion, but also to protect the powder from contamination from the atmosphere, extrusion lubricants and the tooling. The can material is selected on the basis of its hot plasticity, its reactivity and cost. Typical extrusion pressure and displacement curves (Fig. 4-31) show four regions (Hughes and Sellars, 1972): barrelling of the can to make contact with the container walls, hot upsetting of the billet to full density, stable extrusion at constant velocity and constant pressure and finally extrusion under unstable flow conditions at the end of the billet. The stable extrusion pressure PR is approximated by
PR = AKlnR
(4-46)
4.6 Deformation Processing at High Temperatures
Ram displacement Extrusion pressure
where R is the extrusion ratio, A and K are constants related to the system and to the deformation resistance of the material, respectively. A more rigorous analysis (Sheppard and McShane, 1976) takes into account the local temperature increase at the die exit due to pressure dissipation. This suggests
i
PR=AK(l-CAT)lnR
Time
Figure 4-31. Stem displacement and extrusion pressure developed during powder extrusion.
I
50
I
i
I
i
i
i
Product overV heating
Pressure limit: no extrusion
o o 30
I20
Sound extrusion
•D
\
-
X jj
10
z
Low ratio: unsound product ~~i
i
1050
T*
i
1100
183
i
i
1150
1200
Preheat temperature, °C
Figure 4-32. Typical extrusion limit diagram for a tool steel powder.
(4-47)
where C( = dX/dT) is the slope of the deformation resistance against temperature curve, and A T is the temperature rise during extrusion which is itself related to PR. The above equations, as also the extrusion limit diagrams (Causton and Dunkley, 1987) which show the permissible extrusion ratios against temperature (Fig. 4-32), provide the necessary guidelines for designing optimum extrusion conditions. Powder extrusion is widely used in the production of high speed steels, composites, beryllium and oxide dispersion strengthened nickel base superalloys. This route has a special attraction for beryllium processing as it produces fine grained structure with good ductility. In oxide-dispersion-strengthened alloys, powder extrusion results in elongated grain structure which can be further enhanced by annealing to produce zone aligned polycrystals. This microstructure (Fig. 4-33) is resistant to creep and can therefore be used for high temperature applications as in gas turbines. As there is minimum thermal exposure during processing, this route is being evaluated for producing components from rapidly solidified powders. 4.6.6 Special PM Processes
Figure 4-33. Elongated coarse-grained structure of (Courtesy: powder-extruded Ni - 20 Cr - 2 ThO 2 DMRL).
In addition to the above processes in PM, there are also a few others which have been developed with some specific applications in mind. We shall enumerate some of
184
4 Powder Metallurgy
these in this section, as also the processes which are used after sintering to tailor the PM products for specific applications. Ceracon process and rapid omnidirectional compaction are pseudo-isostatic hot consolidation processes that do not use complex and expensive gas pressurizing systems as in hot isostatic pressing. The Ceracon process (Ferguson et al., 1984) uses ceramic granules as a pressure transmitting medium. In this process a hot porous preform is inserted into the ceramic medium in a hot die and an axial load applied on the ceramic grains in a conventional forge press. The ceramic grains transmit the pressure to the preform nearly isostatically. Very short cycle times under pressure are employed and the processing is completely automated, using robotic handling in all steps. After consolidation, the fully dense part is ejected and the hot granular medium recycled. Because of the soft tooling and unrestricted metal flow, the process calls for a complete characterization of the ceramic grains, preform design and strict control of operating pressures and temperatures. Rapid omnidirectional compaction (ROC) uses a thick-walled powder container as a fluid die (Kelto, 1984). The powder is sealed in a die made of soft steel which is heated and placed in a pot die. Uniaxial pressure is transmitted by the fluid die to the powder as a pseudoistostatic pressure, enabling complete consolidation with very short pressure dwells. Unlike the Ceracon process, this process is amenable for consolidation to complex shapes. The preheat temperatures can be considerably lower than the HIP temperatures and thermal exposure times can also be very short. This offers a special advantage for processing metastable microstructures with their attendant property advantages.
Spray forming is an integral manufacturing process which employs atomization as the primary step (Singer, 1970; Leatham et al., 1984). The spray of the atomized liquid metal in an inert atmosphere is directed on to a mold. The particles flatten on impact, coalesce and solidify to produce a preform which can be further hot worked by forging or rolling to full density (Singer, 1970; Singer and Evans, 1983; Williams, 1987). Processing of rapidly solidified powders are possible by this route without a need to handle or store these powders (Singer et al., 1985). The Osprey process (Fig. 4-34) is a specific spray processing route which uses manipulators to move the mold inside the atomization chamber so that the powder is deposited uniformly to high densities. Further processing is done by forging to required shapes. Currently, Osprey process is used in the manufacture of rolls, disks, tubes and other shapes from high alloy steels, superalloys and high strength aluminum alloys. Other methods of atomization are also being used for spray forming. Among these, low pressure plasma deposition uses as feedstock prealloyed powders which are melted in a low pressure plasma and sprayed into a mold (Apelian, 1986). With a high temperature plasma, it is possible to spray-form refractory metals, carbides, borides, oxides and intermetallics. 4.6.7 Post-Consolidation Operations
There are some post-sintering processing steps for realizing the requirements of the envisaged product which are unique to PM. These are secondary cold pressing (coining / sizing / repressing) and infiltration, in addition to the standard metal finishing operations such as machining, heat treatment, hot working and surface treatment.
4.7 Applications Molten metal
185
Crucible Atomizing gas
Spray of particles
Spray deposited shape shape -Transfer mechanism
Collector
Discharge chamber
Spray chamber
Secondary cold pressing of the sintered parts is carried out to improve the dimensional accuracy (sizing), or to increase the density (repressing), or to change the top and bottom surface configurations (coining). Infiltration is a process by which a liquid metal is introduced into the pores of a sintered compact and is used to reach full density without secondary mechanical processing. It may also be used to form multilayer composite structure, or to obtain a surface layer compatible for welding or brazing. For subsequent infiltration, the sintering conditions are adjusted to produce compacts with the required interconnected porosity. The infiltrant, which has a lower melting point, is then injected into the pores by a variety of techniques which include capillary dip, contact infiltration, gravity feed and pressure or vacuum infiltration. Infiltration is widely used in refractory metal based and carbide based systems. Major infiltrated components include electrical contacts and electrodes (W/Mo/ WC-Cu/Ag), rocket nozzles (W-Ag), experimental cermet turbine blades (TiCsuperalloy), tools (TiC-steels), mechanical parts (iron/steel-copper alloy), and bearings (steel-babbitt alloy). Superconducting Nb 3 Sn tapes and wires are prepared by
Figure 4-34. Schematic of Osprey process (Mathur et al., 1989).
first infiltering tin in sintered niobium compact which is then heat treated to produce Nb 3 Sn intermetallic compound.
4.7 Applications Metal powders even in their elemental form find many industrial applications and also as feedstock for the manufacture of PM components. Metallic flakes of aluminum, copper, gold and bronze are used in pigments. Aluminum powder is extensively used in rocket propellants, pyrotechnics and explosives. Fine iron powder is used as a nutrient, food additive, and in magnetic particle inspection. Fillers and coatings on welding electrodes, plasma guns and laser equipment also use elemental or prealloyed powder for coating or hardfacing. Electrostatic imaging systems like xerox use powders for image transfer. In this application, the powder is used as a carrier core material for carrying electric charge and dry ink. Interconnected porosity that can be retained in a PM part has many industrial applications. Porous parts are used variously for filtering, damping and metering in fluid flow devices. Porosity in a self-lubricating bearing serves as a storage reservoir for the lubricant. Interconnected
186
4 Powder Metallurgy
porosity in a battery element increases the surface area, thus enhancing its chemical reactivity. Parts made from bronze, stainless steel, nickel-based alloys, titanium and aluminum are used in such applications. The compaction technique is chosen depending on the desired level of porosity and the shape of the component, with cold compaction and sintering being the most popular. Pressureless compacting is used for the manufacture of filters where porosity levels as high as 50% are required; for filters with large length-to-diameter ratios, however, isostatic compaction or even explosive compaction are used. Because of the shape of electrodes, batteries use powder-rolled sheets. Self-lubricating bearings with their extensive applications are produced in large numbers by die compaction. Bronze and iron-based journal bearings are compacted from elemental powder blends and sintered to retain about 15% interconnected porosity. Lubricating oil infiltrated into the pores provides life-time lubrication for the bearing by a unique mechanism. The oil is drawn out to the bearing surface by thermal and centrifugal effects when the shaft within the bearing rotates, and is drawn back into the pores by capillarity when the rotation stops. Cermets, or ceramic-metal composites, are another class of materials where PM processing appears to be the only optimal route. A classic example is cemented carbides which have been used for machining and wear applications for over sixty years. They are used mainly in the form of small throwaway inserts made from tungsten carbide with titanium and tantalum carbide additions and with cobalt as binder. The carbides are made by high temperature carburizing of mixtures of tungsten and carbon black. These carbides are then milled together with cobalt, compacted and sintered. Liquid phase sintering in
these alloys ensures a fully dense product. To improve the wear resistance, tungsten carbide inserts are surface coated with titanium carbide, titanium nitride and alumina. Friction materials are the other common cermet products which are used as aircraft brake pads and as clutch plates. These are sintered iron-based or copperbased alloys with silica and alumina additions. Recent experiments by L. McCandlish and B. H. Kear at Rutgers University have shown that tungsten carbide/cobalt cermets made from nanocrystalline components (see Chap. 13 of this volume) are considerably harder and more wear resistant than conventional cermets. Powder metallurgy products are also widely used in electrical and magnetic applications. Electrical contacts of silver or copper with tungsten, molybdenum, tungsten carbide or cadmium oxide are made by PM techniques. Liquid phase sintering or liquid infiltration are some of the techniques used in the manufacture of these contacts. A wide variety of magnetic materials with different BH values are produced by PM. These include the widely used hard ferrites such as barium and strontium ferrites, the cobalt-rare earth permanent magnets with high BH value, and the new compositions based on neodymium, iron and boron. This last type is capable of producing an energy product as high as 40 MG Oe, and this is made possible by compacting in a magnetic field which orients the magnetic domains favorably. A typical process chart for producing rare earth permanent magnets is given in Fig. 4-35. Multifilamentary tapes of superconducting Nb 3 Sn are produced by powder rolling niobium to tapes followed by liquid tin infiltration. Among the many processes considered for producing high temperature superconductor tapes and wires, PM offers some attractive routes. The YBa 2 Cu 3 O 7 _ ;c
4.7 Applications
composition for such superconductors is obtained from precursor oxide and carbonate powders by suitable reactions and consolidation. Hot isostatic and explosive compaction have been used with some success in the consolidation. Powder processing is a convenient method for manufacturing components of refractory metals and alloys. Practically all tungsten lamp filaments are drawn from sintered tungsten rods doped with alkali metals which prevent the degradation of microstructure by creep. The process involves the doping of tungsten oxide powder with aluminum, potassium and silicon prior to reduction. It is then pressed and sintered at very high temperatures by resis-
Alloy Preparation by Vacuum Melting or Co-reduction
Premilling
Fine Milling
<500/xm
<10 /xm
Composition Control and Adjustments
Compaction in Magnetic Field for Particle Alignment
Sintering and Heat Treatment
Machining
Coating
Magnetizing Figure 4-35. Production of cobalt-samarium permanent magnets (Ormerod, 1985).
187
tance heating using the compact itself as the resistor. Further processing is carried out by rolling, swaging and wire drawing. The bubbles formed by doping anchor the grain boundaries and prevent any large scale grain boundary migration and grain coarsening. The elongated grains formed by processing ensures that the microstructure remains stable even at high temperatures. (For further details, see Welsch and Walter, 1990.) The above applications are some examples where the special attributes of PM have been used with advantage. But there are also areas where PM is attractive because of the ease it provides for large scale economic production. The automobile industry is such an example and has been a large consumer of PM products manufactured mainly from iron based alloys (Mocarski et al., 1989). These components are necessarily small and add a maximum weight of only 16 kg per car. Various types of gears, pressure plates and small but complex parts of automatic transmission, power steering and power brake systems make up these parts. Several powder forged parts are also used in automobiles, including the carbon steel connecting rods. Powder forging is economically more attractive for this component than ingot forging because of the higher production rate and reduced machining. The automotive applications have been dictated by three main attributes of the PM process, viz., economy, large scale production and capacity to manufacture complex parts. PM is also used with advantage in manufacturing many ordnance items such as mortar bodies, complex machine gun and rifle parts and sintered iron driving bands infiltrated with lead. In addition, some major components are also produced only by the PM route. One such example is the manufacture of heavy alloy ( W - C u -
188
4 Powder Metallurgy
Figure 4-36. Composite carbon-carbon nosecone of a rocket.
Ni/Fe) anti-tank penetrators used in kinetic energy ammunition. Because of their high density and good ductility, these have replaced other materials. Another application is in the development of carbon-carbon composites for the reentry nose-cones of rockets and missiles (Fig. 4-36) and for high performance aircraft brakes. These are produced by first impregnating carbon filament-wound structures with pyrolytic carbon and hot isostatically pressing them at elevated temperatures to high densities. Even though the starting material for this application is not powder, the use of powder processing techniques has enabled the realization of high strength and high densities. Similar developments are also taking place for manufacturing metal matrix composites with good strength and adequate ductility. Nuclear fuel elements and control rods of power reactors are manufactured solely by powder metallurgy. The fuel is generally UO 2 which is produced either from natural uranium or enriched with U 2 3 5 . These fuel pellets are compacted from granulated uranium dioxide powder blended with appropriate binders, and sintered in reducing atmosphere at about 1900 K, in a slow cycle which incorporates dewaxing. After sintering and inspection, these pellets are
placed in zircaloy or stainless steel tubes to make fuel bundles. Strict dimensional control is necessary to ensure efficient operation and good burn-out. There are also plate type fuel elements which are used in special reactors as in nuclear submarines. For these applications, the U235-enriched nuclear fuel is dispersed in a metallic matrix, sintered and diffusion bonded in hot isostatic presses. With a large surface area, these fuels ensure efficient heat transfer enabling a rapid rise to peak power on demand. Control rods and shielding materials are also manufactured by PM techniques using neutron-absorbing materials such as Eu 2 O 3 and B4C dispersed in metallic matrix. Radioactive wastes are also contained by PM techniques. In this process, highly active wastes are encapsulated in glass or ceramic materials and then hot isostatically pressed to seal the contents effectively and provide the container with adequate mechanical strength. This process appears to be promising though it is not used extensively today (see also Vol. 10). Powder metallurgy processing has been developed for manufacturing highly alloyed tool steels which are prone to segregation. In the commonly used commercial processes, atomized tool steel compositions are canned and either hot extruded, or cold isostatically compacted and then hot isostatically pressed. The mechanical and wear properties of powder processed tool steels are shown in many cases to be superior to their conventionally produced counterparts. In recent times, powder metallurgy has made inroads into the processing of aerospace components. An integral rotor component of a rocket engine produced by PM processing is shown in Fig. 4-37. Many of the high temperature alloys are based either on nickel or on cobalt and are
4.8 References
heavily alloyed. They have poor ductility and the workability of these compositions is poor. In some cases, the alloy compositions have become so complex that casting appears to be the only option. Atomized powder eliminates segregation and thus provides an option for processing powder to required shapes and properties. There are some conditions though: the powder has to be very pure and free from ceramic and gas inclusions and must be sintered to almost theoretical densities. Even a small amount of residual porosity, acceptable in many other applications, cannot be tolerated in jet engine components. Such components are now produced from argon-atomized powder of nickel base superalloys which are freed from ceramic and inert gas inclusions, canned, evacuated and hot isostatically pressed to near-net shapes. The purity of the powder and its characteristics are important, as otherwise the sintered components show the particle boundaries decorated by oxides and nitrides. Though the economics of this process is shown to be very attractive, the industry still uses some supplementary hot forging to ensure total densification. Another limitation of the near net-shape process is imposed by the current non-destructive techniques. These techniques can examine the product only when it has some extra thickness for the shape of the 'sonic envelope'. In spite of this limitation, it is expected that the use of near-net shape powder processing will grow in the coming years. Oxide dispersion strengthened alloys described in an earlier Section (4.2.1.2) have been used for some years to fabricate combustor chambers. They are now entering service as materials for high temperature turbine blades and vanes in jet engines. Sheets of these materials are formed by hot extrusion followed by rolling and are then diffusion bonded to fabricate the required
189
Figure 4-37. PM integral rotor component for an impulse turbine.
components. The fabrication can also include drilling complex cooling channels required in turbine blades and vanes. As discussed in an earlier Section, for these applications, oxide dispersion strengthened materials are zone aligned to produce elongated grains. In addition to the above aerospace applications, PM components from aluminium, titanium and beryllium alloys are being used in aircraft and rocket structures. Beryllium is specially used for manufacturing some critical components in the inertial navigation system, heat shields and aircraft brakes. Some new compositions of beryllium alloys are now being investigated for their suitability as skin materials for aeorspace vehicles.
4.8 References Apelian, D. (1986), in: Progess in Powder Metallurgy, Vol. 42: Carlson, E. A., Gaines, G. (Eds.). Princeton, NJ: MPIF/APMI, p. 17. Arunachalam, V. S., Gopinathan, K.G., Krishnan, R.V., Raghuram, A.C., Sundaresan, R. (1975), Trans. PMA12, 1.
190
4 Powder Metallurgy
Arzt, E., Ashby, M. R, Easterling, K. E. (1983), Metall. Trans. 14 A, 211-221. Ashby, M.R (1974), Ada Metall 22, 275. Benjamin, J. S. (1976), Scientific American 234 (5), 40. Bhagiradha Rao, E. S. (1989), in: Powder Metallurgy - Recent Advances: Arunachalam, V. S., Roman, O. V. (Eds.). New Delhi: Oxford & IBH Publishing Co., pp. 27-43. Bhagiradha Rao, E. S., Mallya, R. M., Sastry, D. H (1986), Trans. Ind. Inst. Metals 39, 597-603. Bhalla, A. K., Williams, J. D. (1976), Powder Metall. 19, 31-37. Bhatt, T. B., Ramakrishnan, N., Arunachalam, V. S. (1981), Scri. Metall. 15, 339. Billiet, R. (1985), Int. J. Powder Metall. & Powder Technol. 21, 119-129. Borchers, P. (1979), in: Tungsten. London: Mining Journal Books Limited, 64-77. Brackpool, J. L., Phelps, L. A. (1967), in: New Methods for the Consolidation of Metal Powders, Perspectives in Powder Metallurgy, Vol. 1: Hausner, H.H., Roll, K.H., Johnson, P. K. (Eds.). New York: Plenum Press, pp. 239-255. Brown, G.T. (1981), Metal Powder Rep. 36, 210-215. Caligiuri, R. D. (1980), Ph.D. Thesis, Stanford University, California. Carrol, M. M., Kim, K.T. (1984). Powder Metall. 27, 153-158. Causton, R. J., Dunkley, J. J. (1987), in: New Perspectives in Powder Metallurgy: Powder Metallurgy for Full Density Products: Kulkarni, K. M. (Ed.). Princeton, NJ: MPIF, pp. 53-76. Coble, R.L. (1970), /. Appl. Phys. 41, 4798-4807. Cox, A. R., Moore, J. B., van Reuth, E. C. (1976), in: Super alloys: Metallurgy and Manufacture: Kear, B.H., Muzyka, D. R., Tien, J.K., Wlodek, S.T. (Eds.). Baton Rouge, LA: Claitor's Publishing Division, pp. 45-53. Crider, J. F. (1982), Ceram. Eng. & Sci. Proc. 3, 519528. Dax, F. R (1983), Metal Powder Rep. 38, 200-202. Decours, X, Devillard, J., Sainfort, G. (1976), in: Advanced Fabrication Techniques in Powder Metallurgy and Their Economic Implications, AGARD CP No. 200. Dube, R.K. (1981), Powder Metall. Int. 13, 188-190. Dube, R.K. (1982), Powder Metall. Int. 14, 45-48; 108-111; 163-165. Dube, R.K. (1983), Powder Metall. Int. 15, 36-40. Erickson, A. R., Amaya, H. E. (1984), in: Modern Developments in Powder Metallurgy, Vol. 15: Aqua, E.N., Whitman, C. L. (Eds.). Princeton, NJ: Metal Powder Industries Federation, pp. 145155. Evans, P. E., Smith, G. C. (1959), Powder Metall. 3, 1-25 and 26-44. Exner, H. E (1979), Rev. Powder Metall. and Phys. Ceram. 1, 7. Exner, H. E (1980), Powder Metall. 23, 203.
Exner, H. E., Arzt, E. (1983), in: Physical Metallurgy, 3rd ed.: Cahn, R.W., Haasen, P. (Eds.). Amsterdam: North-Holland Physics Publishing, pp. 1885-1912. Ferguson, B., Kuhn, A., Smith, O. D., Hofstatter, R (1984), Int. J. Powder Metall. & Powder Technol. 20, 131-139. Friedman, G.I. (1970), Int. J. Powder Metall. 6, 43. Froes, F. H., Savage, S. J. (Eds.), (1986), Processing of Structural Metals by Rapid Solidification. Metals Park, OH: American Society for Metals. Gardner, N.R., Donaldson, A.D., Yans, RM. (1967), in: New Methods for the Consolidation of Metal Powder, Perspectives in Powder Metallurgy, Vol. 1: Hausner, H. H., Roll, K. H., Johnson, P. K. (Eds.). New York: Plenum Press, pp. 169-194. German, R. M. (1985), in: Progress in Powder Metallurgy, Vol. 41: Princeton, NJ: Metal Powder Industries Federation. German, R. M. (1990), Int. J. Powder Metall. 26, 2 3 34; 35-43. German, R.M., Munir, Z.A. (1975), Metall. Trans. 6A, 2229. German, R.M., Munir, Z.A. (1978), /. Am. Ceram. Soc. 61, 212-21 A. German, R.M., Munir, Z.A. (1982), Rev. Powder Metall. and Phys. Ceram. 2, 9. Gilman, P. S., Benjamin, I S . (1983), Ann. Rev. Mater. Sci. 13, 279. Goetzel, C. G. (1949), Treatise on Powder Metallurgy, Vol. I. New York: Interscience Publishers Inc., pp. 64-71. Goetzel, C. G. (1950), Treatise on Powder Metallurgy, Vol. II. New York: Interscience Publishers Inc. Grant, N. J. (1983), J. Metals 35 (1), 20-27. Gummeson, P.U. (1972), Powder Metall. 15 (29), 67-94. Gutamanas, E. Y. (1983), Powder Metall. Int. 15, 129. Helle, A.S., Easterling, K.E., Ashby, M.R (1985), Acta Metall. 33, 2163-2174. Hodkin, D. I, Sutcliffe, P. W, Mardon, P.G., Russell, L.E. (1973), Powder Metall. 16 (32), 277-313. Hughes, K.E., Sellars, C. M. (1972), /. Iron & Steel Inst. 210 (9), 661. Huppmann, W. J. (1979), Z. Metallkde. 70, 792-797. Huppmann, WX, Petzow, G. (1979), in: Sintering Processes: Kuczynski, G. C. (Ed.). New York: Plenum Press, pp. 189-201. Jackson, H.C. (1967), in: New Methods for the Consolidation of Metal Powders, Perspectives in Powder Metallurgy, Vol.1: Hausner, H.H., Roll, K.H., Johnson, P. K. (Eds.). New York: Plenum Press, pp. 13-26. James, B. (1982), Metal Powder Rep. 37, 251-253; 291-293. James, P.J. (1971), Production Engr. 50, 515-520. James, P.J. (1972), Powder Metall. Int. 4, 82-85; 145-149; 193-198.
4.8 References
James, P.J. (1983), in: Isostatic Pressing Technology: James, P.J. (Ed.). London: Applied Science Publishers, pp. 1-27. Joensson, S., Hohmann, M. (1987), Metal Powder Rep. 42, 49-52. Jones, H. (1981), in: Ultrarapid Quenching of Liquid Alloys, Treatise on Materials Science and Technology, Vol. 20: Herman, H. (Ed.). New York: Academic Press, pp. 1-72. Jones, H. (1982), Rapid Solidification of Metals and Alloys, London: The Institution of Metallurgists Monograph No. 8. Kaysser, W. A. (1988), in: Advanced Metallic and Ceramic Materials: Rogers, M.D., Jovicevic, J. (Eds.). Institute of Advanced Materials, Commission of the European Communities Joint Research Centre, Petten, Ispra and Karlsruhe. Kaysser, W. A., Petzow, G. (1985), Powder Metall. 28, 145-150. Kelto, C.A. (1984), in: Metals Handbook: Powder Metallurgy, Vol.7, 9th ed. Metals Park, OH: American Society for Metals, pp. 542-546. Kingery, W.D., Berg, M. (1955), J. Appl. Phys. 26, 1205. Kingery, W. D. (1959), J. Appl Phys. 30, 301. Klar, E., Fesco, J.W. (1984), in: Metals Handbook, 9th ed., Vol. 7. Metals Park, OH: American Society for Metals, pp. 25-31. Knopp, W.V., Duncan, W. R, Moses, A.I, Toaz, M. W. (1984), in: Metals Handbook, 9th ed., Vol. 7. Metals Park, OH: American Society for Metals, pp. 401-418. Kobayashi, S., Oh, S.-L, Altan, T. (1989), Metal Forming and the Finite-Element Method. Oxford: University Press, 244-274. Kuczynski, G. C. (1949), Trans. AIME 185, 169. Kuhn, H. H. (1978), Int. J. Powder Metall & Powder Technol 14, 259-275. Kuhn, W. E., Friedman, I. L., Summers, W., Szegvari, A. (1984), in: Metals Handbook, 9th ed., Vol. 7. Metals Park, OH: American Society for Metals, pp. 56-70. Lawley, A. (1977), Int. J. Powder Metall. & Powder Technol 14, 259-275. Lawley, A. (1981), /. Metals 33 (1), 13-18. Lawley, A. (1986), /. Metals 38 (8), 15-25. Leatham, A. G., Brooks, R. G., Yaman, M. (1984), in: Modern Developments in Powder Metallurgy, Vol.15: Aqua, E.N., Whitman, C. L. (Eds.). Princeton, NJ: MPIF/APMI, p. 157. Lee, P.W., Carbonara, R. S. (Eds.) (1985), Rapidly Solidified Materials. Metals Park, OH: American Society for Metals. Lenel, F.V., Ansell, G. S. (1982), /. Metals 34 (2), 17-29. Mathur, P., Apelian, D., Lawley, A. (1989), Acta Metall. 37, 429-443. Meiner, K. E., McCall, J. C. (1981), Metal Powder Rep. 36, 437-446.
191
Miller, G. L. (1958), Powder Metall 1/2, 53-64. Mocarski, S., Hall, D.W., Khanuja, J., Suh, S.-K. (1989), Int. J. Powder Metall. 25, 103-124. Munir, Z. A. (1988), Am. Ceram. Soc. Bull. 67, 342349. Munir, Z.A., Anselmi-Tamburni, U. (1989), Mater. Sci. Rep. 3 (7, 8), 277-365. Ormerod, J. (1985), J. Less-Common Metals 111, 4969. Pease III, L. F. (1987), in: Progress in Powder Metallurgy, Vol.43: Freeby, C. L., Hjort, H. (Ed.). Princeton, NJ: Metal Powder Industries Federation, pp. 789-829. Poster, A. R. (Ed.) (1966), Handbook of Metal Powders. New York: Reinhold Publishing Corp. Price, P. E. (1984), Metal Powder Rep. 39, 28-29. Queneau, P., O'Neill, C.E., Illis, A., Warner, I S . (1969), J. Metals 21 (7), 35-45. Raybould, D. (1981), J. Mater. Sci. 16, 589. Reshamwala, A.S., Tendulkar, G.S. (1970), Powder Metall. Int. 2, 15 and 58. Ro, D. H., Toaz, M. W (1982), in: Progress in Powder Metallurgy, Vol. 38. Princeton, NJ: Metal Powder Industries Federation. Ro, D.H., Toaz, M.W, Moxson, V.J. (1983), /. Metals 35 (1), 34-39. Roberts, P. (1984), in: Metals Handbook: Powder Metallurgy, Vol. 7, 9th ed. Metals Park, OH: American Society for Metals, 39-42. Roman, O.V., Gorobtsov, V. G. (1989): in: Powder Metallurgy - Recent Advances: Arunachalam, VS., Roman, O.V. (Eds.). New Delhi: Oxford & IBH Publishing Co., pp. 83-111. Savage, S.J., Froes, F.H. (1984), J. Metals 36 (4), 20-33. Sheppard, T, McShane, H. (1976), Powder Metall 19, 121. Shinde, S. L., Tendulkar, G. S. (1977), Powder Metall Int. 9, 180-184. Singer, A.R.E. (1970), Metals & Materials 4, 246250. Singer, A. R. E., Evans, R. E. (1983), Metals Technol 10, 61-68. Singer, A.R.E., Hodkin, D.J., Sutcliffe, P.W., Mardon, P.G. (1985), Metals Technol 10, 105-110. Stephan, H., Fischhof, J. K. (1976), in: Modern Developments in Powder Metallurgy, Vol. 9: Hausner, H.H., Taubenblat, P.W. (Eds.). Princeton, NJ: Metal Powder Industries Federation, p. 183. Sundaresan, R., Froes, F. H. (1987), /. Metals 39 (8), 22-27. Thimimler, F., Thomma, W. (1967), Metall Rev. 12 (115), 69. Wang, F. F. Y. (1982), in: Advances in Powder Technology: Chin, G.Y. (Ed.). Metals Park, OH: American Society for Metals, pp. 39-51. Welsch, G., Walter, J.L. (1990), in: Encyclopedia of Materials Science and Engineering, 2nd supplementary Vol.: Cahn, R.W. (Ed.). Oxford: Pergamon Press, pp. 1007-1012.
192
4 Powder Metallurgy
Wilkinson, D.S., Ashby, M. F. (1975), Ada Metall 23, 1277-1285. Williams, B. (1987), Metal Powder Rep. 42, 712-716. Willis, F.W., Klugston, E.J. (1959), /. Electrochem. Soc. 106, 362-366.
General Reading Powder Metallurgy, General Arunachalam, V. S., Roman, O. V. (Eds.) (1989), Powder Metallurgy - Recent Advances. New Delhi: Oxford & 1BH Publishing Co., Bombay and Calcutta. ASM (1984), Metals Handbook: Powder Metallurgy, Vol. 7, 9th ed. Metals Park, OH: American Society for Metals. German, R. M. (1984), Powder Metallurgy Science. Princeton, NJ: Metal Powder Industries Federation. Hirschhorn, J. S. (1969), Introduction to Powder Metallurgy. New York: American Powder Metallurgy Institute. Klar, E. (Ed.) (1983), Powder Metallurgy - Applications, Advantages and Limitations. Metals Park, OH: American Society for Metals. Lenel, F. V. (1980), Powder Metallurgy Principles and Applications. Princeton, NJ: Metal Powder Industries Federation. Manufacture of Powders Beddow, J.K. (1978), The Production of Metal Powders by Atomization. London: Heiden. Frankhouser, W. L., Brendley, K. N., Kieszek, M. C , Sullivan, S. T. (1985), Gasless Combustion Synthesis of Refractory Compounds. Park Ridge, NJ: Noyes Publications. Jones, H. (1982), Rapid Solidification of Metals and Alloys, Monograph No. 8. London: The Institution of Metallurgists. Powder Characterization Beddow, J.K. (1980), Paniculate Science and Technology. New York: Chemical Publishing Company. Fayed, M.E., Otten, L. (1984), Handbook of Powder and Science and Technology. New York: Van Nostrand Reinhold Co. Consolidation German, R. M. (1989), Particle Packing Characteristics. Princeton, NJ: Metal Powder Industries Federation. James, P.J. (Ed.) (1983), Isostatic Pressing Technology. London: Applied Science Publishers. MPIF (1987), MPIF Compendium on Metal Injection Molding. Princeton, NJ: Metal Powder Industries Federation.
Sintering Ivensen, V. A. (1973), Densification of Metal Powders During Sintering. Studies in Soviet Sciences, New York: Consultants Bureau. Kolar, D., Pejovnik, S., Ristic, M. M. (Ed.) (1982), Sintering - Theory and Practice, Materials Science Monograph No. 14. Amsterdam: Elsevier Scientific Publishing Co. Kuczynski, G. C. (Ed.). Materials Science Research, Vol. 6 (1972), Vol. 10 (1975), and Vol. 13 (1979), New York: Plenum Press. Ristic, M. M. (Ed.) (1979), Sintering - New Developments, Materials Science Monograph No. 4. Amsterdam: Elsevier Scientific Publishing Co. Somiya, S., Moriyoshi, Y (Eds.) (1990), Sintering Key Papers. New York: Elsevier Applied Science. Liquid Phase Sintering Eremenko, V.N., Naidich, Y.V., Ivarienko, LA. (1970), Liquid Phase Sintering, Consultants Bureau, New York. German, R. M. (1985), Liquid Phase Sintering. New York: Plenum Press. Hot Isostatic Pressing Atkinson, H.V, Rickinson, B.A. (1991), Hot Isostatic Pressing. Bristol: Adam Hilger. Hanes, H. D., Siefert, D. A., Watts, C. R (1979), Hot Isostatic Pressing. Columbus, OH: Metals and Ceramics Information Center, Battelle Press. Metallography Huppmann, W.J., Dalai, K. (1986), Metallographic Atlas of Powder Metallurgy. Freiburg (i. Br.): Verlag Schmidt. Alloy Systems Froes, F.H., Smugeresky, I (Eds.) (1980), Powder Metallurgy of Titanium Alloys. Warrendale, PA: The Metallurgical Society of AIME. Gessinger, G. H. (1984), Powder Metallurgy of Superalloys, Butterworths Monographs in Materials. London: Butterworth. Book Series Materials Science Research. New York: Plenum Press. Modern Developments in Powder Metallurgy. Princeton, NJ: Metal Powder Industries Federation. New Perspectives in Powder Metallurgy. Princeton, NJ: Metal Powder Industries Federation. Perspectives in Powder Metallurgy. Princeton, NJ: Metal Powder Industries Federation. Progress in Powder Metallurgy. Princeton, NJ: Metal Powder Industries Federation.
5 Mechanical Milling and Alloying Carl C. Koch Materials Science and Engineering Department, North Carolina State University, Raleigh, NC, U.S.A.
List of Symbols and Abbreviations 5.1 Introduction 5.2 High Energy Ball Milling: Equipment and Process Variables 5.3 The Mechanics/Physics of Mechanical Milling and Alloying 5.3.1 Ductile/Ductile Components 5.3.2 Ductile/Brittle Components 5.3.3 Brittle/Brittle Components 5.3.4 Modeling the Kinetics of Mechanical Alloying in a SPEX Shaker Mill . . . . 5.3.5 Modeling the Mechanics of Mechanical Alloying in a Planetary Mill 5.3.6 Temperature Effects During Milling 5.3.7 Relationships to Wear and Erosion Processes 5.4 ODS Alloys by Mechanical Alloying 5.4.1 ODS Ni-Base Superalloys and Fe-Base High Temperature Alloys Produced by Mechanical Alloying 5.4.2 ODS Aluminium-Base Alloys Produced by Mechanical Alloying 5.4.3 Other ODS Alloys Produced by Mechanical Alloying 5.4.3.1 Ti-Base Alloys 5.4.3.2 ODS Intermetallics by Mechanical Alloying 5.4.3.3 ODS Refractory Metal Alloys 5.5 Synthesis of "Equilibrium" Phases by Mechanical Alloying 5.5.1 Solid Solutions 5.5.2 Intermediate Phases 5.5.3 Immiscible Alloy Systems 5.5.4 Synthesis of Materials for Special Applications 5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool 5.6.1 Extended Solid Solutions 5.6.2 Disordering by Mechanical Milling 5.6.3 Amorphization by Mechanical Alloying/Milling 5.6.4 Crystallization of Amorphous Alloys by Mechanical Milling 5.6.5 Demixing Reactions by Mechanical Milling 5.6.6 Nanocrystalline Materials by Mechanical Alloying/Milling 5.6.7 Quasicrystalline Materials by Mechanical Alloying 5.7 Chemical Reactions Induced by Mechanical Alloying 5.8 Summary 5.9 Acknowledgements 5.10 References Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
194 195 196 198 198 199 200 201 204 206 209 210 211 213 214 214 215 216 216 216 217 220 223 224 224 226 228 237 238 239 240 240 241 241 241
194
5 Mechanical Milling and Alloying
List of Symbols and Abbreviations a0 C%
cP
D E A£ b F F k0 P P* R r r S
T T. t
Vr QP
°n
Q,co DSC EXAFS GAR INCO LRO MA MG MM ODS PPD SAP SEM SRO SSAR TD TEM YAG
lattice parameter critical concentration specific heat diffusion constant total strain total energy released by a ball dissipated energy flux observed physical quantity (e.g. resistivity, volume change, enthalpy) thermal conductivity probability of ball impacting normalized power radius of a disk atomic radii ratio of atomic radii RA/RB radius of a container LRO parameter temperature glass transition temperature time time between impact events processing time total powder volume relative velocity powder particle density normal stress rotation speed of the disk and container relative to the disk differential scanning calorimetrie extended X-ray absorption fine structure grain aspect ratio International Nickel Company long range order mechanical alloying mechanical grinding mechanical milling oxide dispersion-strengthened parameter phase diagram sintered aluminum powder scanning electron microscopy short range order solid state amorphization reaction thoria dispersion transmission electron microscopy yttrium-aluminum garnet
5.1 Introduction
5.1 Introduction The synthesis of materials by high energy ball milling of powders was first developed by John Benjamin and his coworkers at the International Nickel Company in the late 1960's (Benjamin, 1970). The goal of this work was the production of complex oxide dispersion-strengthened (ODS) alloys for high temperature structural applications. It was found that this method, termed mechanical alloying, could successfully produce fine, uniform dispersions of oxide particles (A12O3, Y 2 O 3 , ThO 2 ) in nickel-base superalloys which could not be made by more conventional powder metallurgy methods. In addition, reactive alloying components such as Ti, could be incorporated from master alloy powders such that it was clear that "alloying" could occur on the atomic level during the ball milling process. Benjamin and his coworkers at INCO also explored the synthesis of other kinds of materials, e.g. solid solution alloys and immiscible systems, and pointed out (Benjamin, 1976) that mechanical alloying (MA), in addition to synthesis of dispersion-strengthened alloys, could make metal composites, compounds, and/ or new materials with unique properties. However, the major application of MA so far has been the production of oxide-dispersion-strengthened (ODS) alloys. Fundamental research on MA also was concentrated on ODS alloys for some years. In the early 1980's, however, an increase of interest in MA as a nonequilibrium processing method occurred, partly due to the discovery that amorphous alloys could be made by MA (Koch et al., 1983), and to the renewed interest in intermetallic compounds and other materials which were difficult to synthesize by conventional processing techniques. During the last several years there have been in-
195
ternational conferences devoted to materials synthesis by MA ("New Materials by Mechanical Alloying Techniques", 1989; "Solid State Powder Processing", 1990; and "Structural Applications of Mechanical Alloying", 1990). The subject of amorphization by MA has also played a prominent role in conferences devoted to solid state amorphization ("International Conference on Solid State Amorphizing Transformations", 1988; "International Symposium on Amorphization by Solid-State Reaction", 1990). A number of general reviews of MA have appeared (Benjamin, 1976; Gilman and Benjamin, 1983; Benn et al., 1984; Sundaresan and Froes, 1987; Koch, 1989), as well as reviews of specific topics (e.g., amorphization) or materials (e.g., refractory metal alloys). This chapter will cover much of the material presented in previous reviews but will emphasize the more recent work, in particular, the use of high energy ball milling as a nonequilibrium processing method which can be analogous to rapid solidification. The paper will first describe the equipment and some of the process variables used for high energy ball milling. Next, the physics of the milling processes will be described to the extent that it is presently understood, and models of the mechanics of several mills will be reviewed. An important parameter - the temperature of the powder during milling - will be discussed. The remainder of the chapter will be devoted to a description and analysis of the results of mechanical milling and/or alloying of specific classes of materials including ODS alloys, difficult-to-make alloys or compounds, immiscible alloy systems, and many metastable materials such as amorphous alloys, nanocrystalline materials, and quasicrystals.
196
5 Mechanical Milling and Alloying
5,2 High Energy Ball Milling: Equipment and Process Variables
balls material-
The milling of materials has been a major component of the mineral, ceramic processing, and powder metallurgy industries. The objectives of milling include particle size reduction, mixing or blending, and particle shape changes. The typical mill used for these purposes has been the tumbler ball mill (or for small charges the jar mill) which is a cylindrical container rotated about its axis in which balls impact upon the powder charge. The balls may roll down the surface of the chamber in a series of parallel layers or they may fall freely and impact the powder and balls beneath them. The tumbler ball mill is illustrated in Fig. 5-1. While a number of ingenious milling devices were developed early in the century, the one high energy ball mill that has been adopted by industry was invented by A. Szigvari in 1922, in order to quickly attain a fine sulfur dispersion for use in vulcanization of rubber. This mill is called an attritor or attrition mill and is illustrated in Fig. 5-2. Milling occurs by the stirring action of an agitator that has a vertical rotating shaft with horizontal arms. The motion causes a differential movement between the balls and the powder thus providing a much higher degree of surface contact than is achieved in tumbler ball mills. The kinetic energy imparted to the milling media in either of these kinds of mills also depends on the speed of rotation of the tumbler mill cylinder or the attritor shaft. Tumbler mills may also be considered high energy mills if of sufficiently large diameter (of the order of meters) and operated just short of the critical rotational speed at which centrifugal force "pins" the balls to the mill chamber wall. For large scale production of commercial al-
drive rollers
Figure 5-1. Schematic cross-section of a tumbler ball mill.
drive shaft
.rotating impeller stationary tank —
attritioning balls
Figure 5-2. Schematic of an attrition mill.
loys, attrition mills have been largely replaced by the large tumbler ball mills. Another type of mill that has been used for pilot-size production is the vibratory tube mill. The motion of the balls and particles in a vibratory mill is complicated. The cylindrical container is vibrated and the impact forces acting on the powders are function of the rate of milling, amplitude of vibration, and mass of the milling medium. High-energy milling forces can be obtained by using high vibrational frequencies and small amplitudes of vibration. The details of the tumbler ball mill, the attrition mill, and the vibratory mill construction and operation are given in the review by Kuhn et al. (1985). Several different kinds of laboratoryscale high energy mills have been used for research purposes. The SPEX model 8000 shaker mill has been used extensively for research on small batches of powder (^10cc), especially in the U.S.A. This high energy ball mill agitates the charge
5.2 High Energy Ball Milling: Equipment and Process Variables
of powder and balls in three mutually perpendicular directions at approximately 1200 rpm. It is highly energetic compared to the attrition and vibratory mills. A given reaction can typically take place an-order-of-magnitude faster in the SPEX mill than in the typical attritor or vibratory mill. Planetary ball mills have also often been used for research studies, particularly by European investigators (see Fig. 5-8, below). The commercially available Fritsch, Pulverisette 5 has typically been used (e.g. Hellstern und Schultz, 1986), or modifications there of (e.g. Gaffet, 1990). Several researchers have designed their own high energy milling devices, for example, the single large steel ball (6 cm) in a tungsten carbide bottom steel vial (6.5 cm diameter) attached to a vibrating frame used by Bakker and coworkers (Weeber et al., 1987). The energy of the milling media depends on the internal mechanics of the specific mill, the power supplied to drive the milling chamber, and the composition, size, and size distribution of the balls. Since the kinetic energy of the balls is a function of their mass and velocity, dense materials (steel or tungsten carbide) are preferable to ceramic balls, and the size and size distribution should be optimized for the given mill. Too dense packing of balls reduces the mean free path of the ball motion, while a dilute distribution minimizes the collision frequency. Empirically, ball mass-to-powder-mass ratios of 5 to 10 are typically used and are effective. One important process variable is the temperature of the mill and that induced in the powders by the kinetic energy of the milling media. This temperature may be critical for the reactions or transformations in the powders during milling. While there is still some doubt as to the precise temperatures that the powder surfaces
197
may experience, a good deal of experimental, albeit indirect, evidence as well as model calculations suggest modest temperature rises - ^ 100 to 200 °C - are likely under most conditions. This question will be addressed in more detail in the next section. One serious problem with the milling of fine powders is the potential for significant contamination from the milling media or atmosphere. If steel balls and containers are used, iron contamination can be a problem. It is most serious for the highly energetic mills, e.g., the SPEX shaker mill, and depends on the mechanical behavior of the powder being milled as well as its chemical affinity for the milling media. Extended milling times in a high energy shaker mill can result in iron contamination of >10 atomic percent in some refractory metal powders. In order to minimize such contamination, less energetic mills - such as the vibratory mill - may be used, or other more inert milling media can be used such as tungsten carbide. In extreme cases, balls and container lining can be made of the same or similar alloys as the composition of the powders being processed. The other potential source for contamination of the powders is the milling atmosphere. If milling is carried out in air, significant contamination with oxygen, or in some cases with nitrogen, can occur. Oxygen contamination is, of course, most severe in reactive metals such as Al, Ti, Zr, etc. Oxide concentrations can be more than tripled in aluminum powders after 3 h of milling (Gilman and Nix, 1981) and doubled in Ni3Al powders after 6 h of milling in a SPEX mill in an impure argon atmosphere (Jang and Koch, 1988). Oxygen concentrations of «10 atomic percent are not unusual for milling of reactive powders such as Zr- or Ti-base alloys. While the usual concern for atmospheric contamination is for oxygen, the possibili-
198
5 Mechanical Milling and Alloying
ty of nitrogen contamination should not be ignored in certain cases. Fu et al. (1990), have recently shown that nitrogen gas in the milling vial reacted with the powder in a mixture of Fe and Er powders to form an ErN phase. It was suggested that the steel vial did not remain hermetically sealed during the milling process and the nitride phase was formed due to the catalytic action of the Fe in dissociation of the N 2 molecules. Atmospheric contamination can be minimized or eliminated by milling in an inert atmosphere or vacuum. One approach is to load the ball/powder charge in an inert gas (argon) glove box and seal the vial with an "O" ring seal. It must then be assumed that the seal is maintained during milling. A more certain method, which is applicable to small experimental mills, is to enclose the mill in an inert gas glove box. Less contamination is also often observed with the less energetic mills which may be a consequence of the lower powder temperatures and/or the better maintenance of the vial seal. While most attention of experimentalists has focused on minimizing atmospheric contamination during milling, the handling of the powder after milling is also important. In many cases the final milled particle size is a few micrometers in diameter, presenting a large powder surface area for atmospheric contamination of reactive alloys. While mechanical milling/alloying is defined as a dry milling process, certain very ductile metals such as Al (Benjamin and Volin, 1974; Benjamin and Bomford, 1977; Gilman and Nix, 1981) and Sn (White and Nix, 1979), require the addition of a surfactant, process control agent, to obtain the proper balance between cold-welding and fracture. The process control agents are typically organic compounds such as Nopcowax-22DSP, (Diamond Shamrock
Co. trade name for ethylenebisdistearamide, C 2 H 2 _ 2 (C 18 H 36 ON)). They minimize excessive cold welding and allow mechanical alloying to occur in very ductile systems. This occurs even though the surfactant often disappears after short milling times ( < l h ) . Ovesoglu and Nix (1986) used differential scanning calorimetry to monitor the melting point of Nopcowax22DSP. They found that after 25 minutes of milling, the melting peak for the surfactant disappeared. It is assumed that the surfactant is incorporated into the metal powder or volatilizes by forming gaseous species. It can, therefore be a serious source of contamination of the metallic powders.
5.3 The Mechanics/Physics of Mechanical Milling and Alloying The central event in mechanical milling or alloying is the ball-powder-ball collision. Powder particles are trapped between the colliding balls during milling and undergo deformation and/or fracture processes which define the ultimate structure of the powder. The nature of these processes depends upon the mechanical behavior of the powder components, their phase equilibria, and the stress state during milling. It is convenient to classify the powder components into 1) ductile/ductile, 2) ductile/brittle, and 3) brittle/brittle systems. 5.3.1 Ductile/Ductile Components The phenomenological description of the mechanical alloying of ductile components (and ductile/brittle components) was first presented by Benjamin and Volin (1974). They discuss the mechanical alloying of ductile components in terms of a competition between cold-welding and fracture. They divided the mechanical al-
5.3 The Mechanics/Physics of Mechanical Milling and Alloying
loying process into five sequences as delineated by observations with optical microscopy of the powders at various stages of milling. Initially, micro-forging flattens ductile powders into plates and fragments more brittle components into finer particles. A schematic drawing of a ball-powder-ball collision of a powder mixture is given in Fig. 5-3. Extensive cold-welding follows the first stage with a composite lamellar structure of the ductile components in the form of plates. At longer milling times, the composite powder particles are further refined, the lamellar spacings decrease, and lamellae become convoluted. At this stage, alloying begins, aided by any heating that may be introduced by milling, the enhanced diffusion paths of the lattice defects created by the deformation, and the shortened diffusion paths as the lamellae become finer and more convoluted. Eventually, the lamellar spacing becomes too fine to be resolved by optical microscopy. With continued milling, components which exhibit complete solid solubility mix to the atomic level, i.e. a true alloy is formed. Benjamin (1976) demon-
Figure 5-3. Ball-powder collision of powder mixture during mechanical alloying.
199
strated that true atomic level alloying occurred by mechanical alloying of Ni-Cr alloys by showing that the magnetic-behavior of the mechanically alloyed Ni-Cr powder was identical to Ni-Cr alloys of the same composition prepared by conventional ingot metallurgy. 5.3.2 Ductile/Brittle Components The evolution of microstructure during mechanical alloying of ductile/brittle systems has also been described phenomenologically (Benjamin, 1970; Benjamin and Volin, 1974; Benjamin, 1976). The hard brittle powders are fragmented during milling, and the fragments are trapped at the boundaries between the ductile powders. As milling proceeds and the welds between ductile particles come closer together - and finally blend - a fine dispersion of the brittle phase results if it is insoluble in the ductile matrix, as inert oxides in ODS alloys. However, brittle intermetallics are also incorporated into, for instance, Ni-base ODS superalloys by mechanical alloying. The brittle intermetallics are fragmented but apparently alloy with the Ni-base matrix so that they are not resolvable by optical microscopy when mechanical alloying is complete. Another example of the alloying of ductile/brittle components by milling wherein a homogeneous alloy is formed is the mixture of Zr (ductile) with NiZr 2 (brittle) powder to form an amorphous Ni-24 at.% Zr alloy (Lee and Koch, 1988). Schultz et al. (1988) did not observe the alloying of brittle amorphous boron particles into Fe, Fe-Si, or Fe-Zr powders by ball milling. In all cases the boron particles were dispersed in the metal matrix. In Fe-Zr-B, the boron did alloy after annealing the dispersion below the crystallization temperature of the amorphous Fe-Zr alloy. Brittle Si did form
5 Mechanical Milling and Alloying
an alloy with Fe during mechanical alloying. These observations presumably reflect the differences in the mutual solid solubilities of the components. Boron has negligible solubility in Fe, while Si exhibits significant solid solubility in Fe. Likewise, the stable oxide dispersoids, such as A12O3, Y 2 O 3 , ThO 2 , are essentially insoluble in the metal matrices of ODS alloys in which they are incorporated by mechanical alloying. Thus, alloying of ductile/brittle components during milling requires not only that the brittle particles become fragmented so that short-range diffusion can occur, but the brittle component must also have some solubility - as indicated by either the stable or metastable phase equilibria - in the ductile matrix component. 5.3.3 Brittle/Brittle Components It might be expected that mechanical alloying of brittle/brittle powder systems would not occur and that milling would simply reduce the brittle powder components' size down to what is called the limit of comminution. The grinding of brittle mineral powders to achieve fine particles for subsequent processing reaches a limit on particle size below which further fracturing stops - the limit of comminution (Harris, 1967). It has been suggested that this limit may be due to extremely small particles deforming plastically rather than fracturing, the increased cohesion between fine particles leading to aggregation, or phase changes in surface layers. It has been shown, however, that milling of certain nominally brittle components can lead to homogeneous alloys. The brittle/brittle systems which have been studied in this regard include Si/Ge (which form solid solutions) (Davis and Koch, 1987), Mn/Bi (which forms the intermetallic com-
pound MnBi), (Davis et al., 1988) and various mixtures of intermetallic compounds such as NiZr 2 /Ni 11 Zr 9 (which forms amorphous alloys) (Lee et al., 1988a). Davis and Koch (1987) observed the formation of a Si-Ge solid solution on mechanical alloying of the pure components which are both nominally brittle at ambient temperature. The solid solution formation was detected by measurement of precise lattice parameters for Si and Ge as a function of milling time. As illustrated in Fig. 5-4, the lattice parameters of Si and Ge move towards each other with increasing milling time until they merge to a single value, which is identical to that for the solid solution of this composition (Si-28 at. % Ge) prepared by conventional metallurgical methods. Thus, it is possible to attain atomic level alloying by milling brittle components together. The mechanism for mechanical alloying of brittle components is not well understood as yet. The microstructural evolution during milling differs markedly from the lamellar morphology of ductile components. A granular morphology is observed during mechanical alloying as illustrated in Fig. 5-5 for Ge and Si powder milled for
a Ge ^ Si • ALLOY
D
0.57-
D
c meter
200
D
D
0.56-
2
(0 £L O O
•• 0.55-
•
•
i 0 54 •
2
4
6
8
10
milling time (h) Figure 5-4. Lattice parameter vs. milling time for Si and Ge powder for the composition Ge-72 at.% Si.
5.3 The Mechanics/Physics of Mechanical Milling and Alloying
Figure 5-5. SEM micrograph of Ge (dark) and Si (light) powder milled 2 h in a SPEX mill.
two hours in a SPEX mill. It would appear that the harder Si particles are embedded in the softer Ge. It was also noted in the Si-Ge system that mechanical alloying was suppressed by milling in a vial cooled by liquid nitrogen. In this brittle/brittle system, thermal activation - diffusion - is apparently a critical requirement for mechanical alloying. In contrast to this observation, mechanical alloying has been achieved at sub-ambient temperatures in ductile/ductile systems, e.g. at -40°C in Ni-Ti alloys (Schwarz et al., 1985), and ductile/brittle systems, e.g. Nb-Ge at 15 °C (Koch and Kim, 1985). This difference may reflect the longer diffusion distances required in the brittle/brittle granular vs. ductile/ductile lamellar geometry, and/or the enhanced diffusion paths provided by severe plastic deformation in ductile/ductile systems. While the mechanism for mechanical alloying of brittle/brittle systems is not yet well understood, it is clear that material transfer can occur between certain brittle components. Possible mechanisms which may contribute to material transfer during mechanical alloying of brittle components (Davis et al., 1988) may include plastic de-
201
formation which is made possible by (a) local temperature rises, (b) microdeformation in defect free volumes, (c) surface deformation, and/or (d) hydrostatic stress state in the powders during milling. A frictional wear mechanism, to be discussed in more detail later, may also be operable. The material transfer in the brittle powder agglomerates during milling may be related to the conditions that control the "limit of comminution", as described above. When the limit of comminution is reached in multicomponent brittle powders, the conditions may be favorable for the cold welding or other material transfer mechanisms that result in mechanical alloying. 5.3.4 Modelling the Kinetics of Mechanical Alloying in a SPEX Shaker Mill McDermott (1988) and Davis et al. (1988) examined the mechanics of the SPEX shaker mill. The following questions related to the kinetics of mechanical alloying in the SPEX mill were considered. What velocities do the milling media (balls) attain? How much energy is transferred to the environment (powders) during an impact? How much energy is transferred to the powders as heat? To help answer these questions, a computer simulation of the interior of the SPEX mill during operation was constructed to gain insight into the collision process as related to mechanical alloying. The model consists of a simplified version of the movements of the vial, combined with the inferred ball movements based on classical mechanics and analytic geometry. The vial motion was recorded on videotape after slowing its apparent velocity by use of a high precision stroboscope. This videotape was then analyzed by a "Motion Analysis" computer translation system. This converted the analog mo-
202
5 Mechanical Milling and Alloying
tion of the vial into digital coordinate displacements and velocities, based on appropriate time constants of length calibrations. The ball motion and velocity within the vial was monitored continuously by the program. The ball motion was studied by observation through a transparent lucite vial. If any two balls were within 1 diameter of their loci, or any one ball was within 1 radius of the vial interior surface, then an impact was recorded, the ball impact dissipated kinetic energy and mean free paths were recorded, and the ball's direction and velocity adjusted based on geometry of collision, vial impulse, and gravitational effects. The geometric and kinetic energy values were adjusted based on the restitution of the stainless steel powder-coated balls, as determined by separate experimentation. Table 5-1 shows the number of impacts occurring for 0.5 and 1.0 seconds of mill operation at several different ball loads. As can be seen, the vast majority of impacts occurred in the 1 x 104 to 105 ergs (10" 3 to 10 ~2 J) range of energy dissipated during the collision. This information can be used to estimate the temperature increases in the powder Table 5-1. Number of impacts for various ball and kinetic energy values. Number of balls used (2geach)
5 10 15
Kinetic energy of impact (J) 10" 7 10~4 10~3 10~2 10" 1 4 tol0~ tol0~ 3 tol0~ 2 tolO" 1 to 1.0
0 0 4
Mill operation for 0.50 s 43 297 3 78 505 13 124 928 24
0 1 0
Mill operation for 1.00 s 5 10 15
0 0 4
78 148 229
612 1201 1873
3 13 24
0 2 0
induced by the kinetic energy of the balls. An expression for temperature increases from localized shear of powder particles trapped between two colliding balls is given by Schwarz and Koch (1986) as
F
AT=
i
At (-1
1/2
(5-1)
2 Vt£o£p where F= dissipated energy flux = crni;r, where an = normal stress due to a head-on collision and vr = relative velocity of the balls before impact. At = stress state lifetime, gp = powder particle density, k0 = thermal conductivity of the powder, and Cp = specific heat of the powder. With the appropriate constants put into this expression for Ge (Davis and Koch, 1987; Davis et al., 1988) temperature rises of A r - 1 0 K and A r = 9 5 K are calculated for vT = 2 m s" 1 and t;r = 18.7 m s" 1 respectively. The value of i;r = 18.7 m s~ * was the maximum value determined by the computer model. Further discussion of the temperature induced in the powders by the kinetic energy of the balls will follow descriptions of other models of the ball milling process. Courtney and Maurice (1990) have constructed a model for the mechanical alloying of ductile powders which they have applied to both the SPEX shaker mill and the attritor mill. They treat the impaction of a porous powder cylinder, composed of many individual powder particles, between two flat surfaces. The surfaces are assumed flat since the powder cylinder radius is much smaller than the ball surface radius. The geometry for this model is given in Fig. 5-6. The authors developed an expression for the time for mechanical alloying based on this model. This processing time, tp9 is given by: V
Ft
r
(5-2)
203
5.3 The Mechanics/Physics of Mechanical Milling and Alloying
of 0°. This probability is given by P(<9) = l-cos9 and is plotted as curve "a" in Fig. 5-7, which represents a random distribution. Curve " b " represents a distribution biased toward glancing angles. The curve labeled "c" represents the expected distribution in a SPEX mill containing a few balls, while curve "d" is from the mod-
Porous powder cylinder
P
Figure 5-6. Schematic model of Courtney and Maurice (1990) for impaction of a porous powder cylinder.
Table 5-2. Estimated milling times for MA devices (after Courtney and Maurice, 1990). Device
where Vp = total powder volume, E=total strain needed for alloying, rc = time between impact events (which is determined by the "unit cell" of the mill). In the SPEX mill the unit cell is defined as the mill volume divided by the number of balls within it. In an attritor the unit cell is defined as the volume belonging to each ball, e.g. for a packing density of 0.5 the unit cell is « twice the volume of a ball. The authors describe how each of the parameters in the expression for tp are estimated. They then calculate the estimated milling times for mechanical alloying for the SPEX and an attritor mill. Their results are shown in Table 5-2. Qualitatively the results agree with experiment in that the attritor mill requires longer milling times than the SPEX mill. The magnitude of the milling times are too short compared to experiment, however, since the transient times for the development of a steady state particle distribution are not included, and, more significantly only "head on" collisions were treated. Courtney and Maurice (1990) describe how the impact angles for the collisions may be included in the model. The simplest approach is to consider the probability of a ball impacting a flat surface at an angle less than 0. A head-on collision corresponds to an impact angle
Material
Time (min)a
Time (min)b
Attritor (ball radius = = 2.4xlO~ 3 m)
Al Cu Fe
120-210 3 8 - 64 4 4 - 73
500-830 120-210 150-250
SPEX Mill (ball radius = = 2.4xlO~ 3 m)
Al Cu Fe
2.4-4.0 0.7-1.1 0.8-1.3
8.1-14.0 2.4- 4.0 2.7- 4.6
SPEX Mill (ball radius = = 6.4xlO~ 3 m)
Al Cu Fe
1.8-3.1 0.3-0.5 0.4-0.7
6.7-11.0 1.8- 3.1 2.1- 3.6
Ball/powder ratio = 10
b
Ball/powder ratio = 3
0.3 A a b c
0.2
Davis, Me D e r m o t t & K o c h Solid angle Biased glancing angle Spex mill
CD Q_
0.1
0
0
20
40 G (deg.)
60
80
Figure 5-7. Impact angle distribution functions. (After Courtney and Maurice, 1990). The triangles represent observed/simulated distributions in a SPEX mill containing a large number of balls (Davis et al., 1988). Curve (a) represents a random (solid angle) distribution. Curve (b) represents a distribution biased toward glancing angles. Curve (c) represents an expected distribution in a SPEX mill containing relatively few balls. Curve (d = ±) refers to a SPEX mill with many balls.
204
5 Mechanical Milling and Alloying
el of McDermott (1988) and Davis et al. (1988) for the SPEX mill with 15 balls. As the number of balls in the mill increases, the relative number of direct impacts decreases. The consideration of the impact angle distributions would result in decreasing the average velocities, and therefore kinetic energy of the balls, thus increasing the processing times listed in Table 5-2. 5.3.5 Modeling the Mechanics of Mechanical Alloying in a Planetary Mill A schematic drawing of a planetary mill is given in Fig. 5-8. The mechanics of this mill are characterized by the rotation speed of the disk, Q, that of the container relative to the disk, co, the mass, ra, size, and number of balls, the radius of the disk, R, and the radius of the container, r. Martin and Gaffet (1990) have shown that depending on the relative values of co/Q and r/R, two extreme regimes may be achieved: 1) the ball rolls on the inner surface of the container or 2) it escapes and impacts an opposite portion of the surface. For both cases the energy transferred per unit area
scales with m Q2, and the frequency of the occurrence of the impacts scales with co. The power induced to the powder therefore scales as P oc m I2 Q2 co where I2 is a characteristic area of the order R2 or rR for the rolling or impact regime respectively. Gaffet (1990) has designed an experimental planetary mill so that the parameters Q and co can be independently varied. He also incorporated a heater so that the vials can be heated to a temperature of up to 200 °C. Using 5 steel balls, he milled the intermetallic compound Ni 10 Zr 7 to milling times where a previous study (Gaffet, 1989) had indicated that a steady state structure was attained (about 40 h). The parameters Q and co were then independently varied and a "parameter phase diagram" (PPD) was then established as illustrated in Fig. 5-9. The final milled structures depend on the Q and co values and a domain in the Q — co plot for the amorphous phase is delineated. At low energies (Q) a partially amorphous phase forms, at intermediate energies a completely amorphous structure is seen, and again at high energies a mixed amorphous and crysE £800-1
H
CT700-
jj 6001 500-] mm
300G
€€
4 C
200.
0
Figure 5-8. Schematic drawing of a planetary ball mill. Q and co are the angular velocities of the ball mill plate and of the vials, respectively.
$m
240 480 720 960 1200 Co,Rotation speed (frequency)(r.p.m.) Figure 5-9. Ball-milling parameter phase diagram for Ni 10 Zr 7 (large circles) and Ni^Zrg (small circles). (After Gaffet, 1990). Filled circle: homogeneous amorphous end product structure; semi-filled circle: mixture of amorphous and crystalline structures; open circle: crystalline structure.
5.3 The Mechanics/Physics of Mechanical Milling and Alloying
205
0.03 AE! PW30
®
®
®
A
®
PW60
I -o—o—o—I—®
—n—o
U
®—<&j W5
®
<*>-U
PW90
0.003
Figure 5-10. Energy, A£J vs. input power, P* for a planetary ball mill. The map indicates zones where (A) only line broadening is observed; (B) amorphous phase formation, and (C) amorphous + intermetallic phase are formed for the Fe2Zr composition of elemental Fe and Zr powders (after Burgio et al, 1991).
P* in Wh' 1
talline structure exists. These results will be discussed later in the chapter. If the weight of the milling balls was decreased by a factor of a, Q is decreased by a factor oc1/2Q such that the PPD diagram boundaries are shifted. Changing the temperature of the mill over the range from 30 °C to 200 °C did not alter the PPD diagram for Ni 10 Zr 7 . Burgio et al. (1990) have also modeled the energy transfer in a planetary mill. They derived kinetic equations which describe the velocity and acceleration of a ball in a vial of a planetary mill. They derive an expression for the total energy released by the ball during a series of collision events, AEh. A normalized power in-
put is also defined as P*, which is a function of the energy released by the ball (A2?b), the angular velocities of the disk and vial, and their dimensions, the weight of the powder, and the milling time. They plot AEh vs. P* as shown in Fig. 5-10 for milled Fe and Zr powders at an average composition of Fe 2 Zr. The end products depend on the milling conditions and a correlation between the energy input and the structures observed is evident. The AE% — P* map separates the observed structure into domains where 1) only line broadening is seen, 2) only an amorphous phase is observed, 3) the crystalline Fe 2 Zr compound is observed, or 4) demixing is noted. The temperature of the
206
5 Mechanical Milling and Alloying
powders is not considered in this model. This model assumed that the energy transfer is derived from collisions of the balls, the chamber walls, and the powder, although the energy of interest is that imparted to the powder by the balls which is only a fraction of the total calculated energy. The authors recognize this restriction but find that the parameters derived give a useful separation of the powder structures under various milling conditions. 5.3.6 Temperature Effects During Milling
The temperature that the powders attain during milling is a critical parameter which may control the final structure. If milling is carried out under ambient conditions the mill and/or the powders typically absorb heat from the mill motor, friction in the mill bearings etc., and the kinetic energy of the milling media. The macroscopic temperature of mill and powder has been measured by several investigators. Davis (1987) and McDermott (1988) measured the exterior of a SPEX mill vial as a function of milling time with thermocouples. The temperature increased with initial milling time and reached an equilibrium value, i.e. saturated, after about 1 hr of milling. The saturation value was a function of the number of balls in the vial. With no balls in the mill, the vial temperature increased to an equilibrium value of about 40 to 42 °C. The maximum temperature rise observed by McDermott was up to about 50 °C (with 13 balls). Thus, in a SPEX mill, much of the heat comes from the motor and bearings. It is possible to significantly lower the macroscopic mill temperature in this case using forced air convection, i.e. a fan, and saturation temperatures of 35 to 40°C can be obtained even with the maximum kinetic energy of the balls operative. The macroscopic tem-
peratures of the milling vials are thus specific to the mill design as well as the kinetic energy of the milling media. In the SPEX mill, apparently about 2A of the temperature rise comes from the motor, bearings, etc.; not from the kinetic energy of the balls. Very large macroscopic temperature increases have been observed by Kimura and Kimura (1990) in a high speed attritor mill. At the fastest rotation speed used, 450 rpm, a maximum steady-state temperature of 172°C was measured. Substantial macroscopic temperature rises, to 120 °C, were also noted in a vibratory mill by Kuhn et al. (1985). However, the Invicta Vibrator Mill (Model BX 920/2, Grantham Electrical Engineering Co) exhibits little or no measurable macroscopic heating above ambient. Thus, the macroscopic heating, i.e. that heating which can be measured by thermocouples or other temperature sensors, is a strong function of the milling device. The maximum working temperatures measured for most common mills, under ambient conditions, are modest, <100°C. It is also possible, of course, to cool the mills. Attritor mills are typically run with water cooling of the mill chamber. Several researchers have cooled experimental mills with liquid nitrogen (e.g., Davis and Koch, 1987; Schwarz et al. 1985, Koch and Kim, 1985). The powder surfaces however, might exhibit significantly higher temperatures during milling than can be measured macroscopically, i.e. "microscopic" heating may be large. It has even been suggested that melting could occur during milling due to the severe localized plastic deformation of the powders (Ermakov et al., 1981). Clear evidence exists for melting of powder surfaces in shock-wave consolidation of powders (Cline and Hopper, 1977). How-
5.3 The Mechanics/Physics of Mechanical Milling and Alloying
ever, in typical shock-wave experiments, the velocities of the "flyers" are of the order of 500 to 1000 m s " 1 compared to the velocities of balls in typical ball mills; 1 to 2 0 m s " 1 . It is of course very difficult to measure the transient temperature of the powder surfaces during milling because of the microscopic and dynamic nature of the milling process. Two different approaches have been used to estimate the temperatures which may be attained in the milled powders. One method is to calculate the temperature from appropriate models, another is to infer the temperature from the structure/microstructure of the final powder product. One approximate calculation for A Tin a SPEX shaker mill was presented in Sec. 5.3.4 as given by Schwarz and Koch (1986). Temperature rises of <100°C were predicted for, e.g. Ge. Another approach is to consider microscopic heating due to sliding friction (Carslaw and Jaeger, 1959). This analysis considers a system in which a body is making contact with another over a limited area while moving over the surface at a constant velocity. If the contact area is assumed square, the microscopic temperature increase is given by: AT =
fwvr
(5-3)
where / = coefficient of friction, w = normal load, / = side half-length of the contact area, / = mechanical equivalent of heat, kl9 k2 = thermal conductivities of components 1 and 2. For the material constants for given components and a specified ball mass, the expression AT cc 1//. Thus the magnitude of AT depends on the contact area. For the smallest contact areas, e.g. 1=1 nm, AT values for Ge-Si, SiSi, or Ge-Ge combinations can be on the order of 500 °K. Clearly, A r i n the above
207
model depends critically on the volume involved. All the models developed to date for estimating AT are reasonable, but not definitive. There are too many unknown variables, for example the size of the volume element which is heated during a collision, as above, to prove by calculation the magnitude of the temperature rise in a given powder system. However, microstructural evidence from the final powder product tends to support the modest temperature rises predicted by the approximate calculations for the usual mill energetics. Mechanically alloyed ductile layered materials, for example, exhibit deformation bands, slip lines as revealed by optical microscopy (Benjamin, 1976) or high dislocation densities and deformation bands as seen in transmission electron microscopy (e.g. Schlump and Grewe, 1989). These observations indicate recrystallization has not taken place during the milling process so that the powder temperature must be below the recrystallization temperature of the given material. That is, a cold-worked microstructure not a hot worked microstructure is observed in milled powder. Similarly, mechanically alloyed amorphous alloys do not, in general, exhibit in-situ crystallization, suggesting the powder temperatures remain below the crystallization temperature of the amorphous alloy. There are several counterexamples to this general observation, in which crystalline phases are observed under highly energetic conditions (Schultz, 1988; Kimura and Kimura, 1990) and were presumed to originate from in situ crystallization because of the temperatures attained during milling. Some doubt has recently been cast on this interpretation (Trudeau et al., 1990; Gaffet, 1990) and this topic of ball-milling induced crystallization of amorphous alloys will be discussed later in this chapter.
5 Mechanical Milling and Alloying
208
In an effort to obtain an upper limit on the temperatures attained in the powders during milling, Davis et al. (1988) studied the tempering response of as-quenched tetragonal ("fresh") martensite in an Fe1.2 wt. % C alloy. Differential scanning calorimetry (DSC) was used to follow the decomposition of the martensite as s-carbide and then Fe 3 C precipitates from the martensite matrix at about 418 K and 575 K respectively. A DSC trace for the unmilled Fe-1.2wt. % C martensite powder is shown in Fig. 5-11. The percent of Fe 3 C converted from the martensite was calibrated by separate annealing experiments and in this way DSC could monitor the amount of martensite decomposition occurring during milling. This gave an upper bound to the temperature the powders reached during milling. It was determined to be 265 to 280 °C. The defects, dislocations and point defects, which are presumably introduced during milling, may accelerate the Fe 3 C precipitation reaction such that it could occur at lower temperatures. High-energy ball milling is unusual
sample:MARTENSITE size: 25.30 mg
DSC
in that plastic deformation can be introduced in nominally brittle materials like Fe-1.2 wt. % C martensite which fracture in the elastic regime under uniaxial deformation. The upper limit to the AT during milling of Fe-1.2 wt.% C martensite experimentally determined is consistent with the model for the mechanics of the SPEX mill presented above and the approximate AT calculations (Davis et al., 1988). Another experiment which suggests only modest (100 to 200 °C) increases in temperature occur during milling was conducted on Bi powder (Davis et al., 1988). Bismuth is a low melting (271 °C) but brittle material. It was assumed that if local powder temperatures rose above the melting point, microscopic examination would reveal this by the presence of smooth, glassy surface morphologies indicative of surface melting. No such morphology was observed, suggesting that milling did not result in temperature increases >271°C in Bi. In contrast to these results, larger AT values were measured by Miller et al. (1986) in impact experiments. Infrared emission was measured from several insulating material single crystals at impact from 0.5 kg loads dropped with various velocities. Examples of their results are given in Table 5-3. If all the kinetic energy of the impact was uniformly distributed over the samples, an overall AT of only about 0.2°C would be expected. The observed infrared Table 5-3. Maximum temperatures for impacted materials from Miller et al., 1986.
-2
350
400
450
500 550 600 Temperature in K
650
700
Figure 5-11. DSC analysis of martensite decomposition for an Fe-1.2wt.% C martensite powder. The peaks represent decomposition of the martensite to e carbide (418 K) and cementite (575 K).
Material
Impact velocity
NaCl Ammonium perchlorate Cyclotrimethylene trinitranine
18.5 5.5 5.5
Tm
380 400 200
5.3 The Mechanics/Physics of Mechanical Milling and Alloying
emission suggests very small localized volumes, e.g. protrusions, must be the source of the observed heating. The above experiments and calculations all suggest some heating of the powders occurs during mechanical milling in "conventional" high energy ball mills such as the SPEX shaker mill. However, the magnitude of the temperature rise is believed to be modest, <100 to 200 °C. This subject will be considered further in the discussions of crystallization of amorphous alloys. All the above discussion pertained to heat input from the mechanical energy of the milling media. It ignored any additional heat which may be generated by exothermic reactions of the powder components. In Sees. 5.5.2 and 5.7, very large temperature rises during mechanical alloying will be discussed which result from either large exothermic heats of formation of certain intermetallic compounds, or the reduction of metal oxides. In these special cases temperature rises can even exceed the melting temperature of the component powders. 5.3.7 Relationships to Wear and Erosion Processes
Wear processes in sliding systems have some features in common with mechanical alloying and milling which may help to elucidate the complex mechanisms of both phenomena. Rigney et al. (1984) have reviewed the phenomenology of the wear processes in sliding systems. The development of the sliding wear process first involves local contacts causing large plastic strains in either or both solid components. The plastic deformation modifies the nearsurface microstructure making it unstable to local shear. Material transfer can then occur, as in mechanical alloying, and the
209
components are further deformed and mixed. A very fine-grained microstructure is formed with nanocrystalline grain sizes. Heilmann et al. (1983) observed debris particles in the wear of Cu blocks with 440 C stainless steel rings which had grain diameters of 3 to 30 nm. Such nanocrystalline grain structures are also observed in ball-milled powder, as will be discussed later. The fine grain structure of this transfer material in wear is believed to be stabilized by the mixing in - i.e. mechanical alloying - of a second phase. The relative hardness of the transfer material and the adjacent substrate material affects the surface morphology and the nature of the wear debris. Loose debris is often derived from the transfer material. Adhesion is apparently an important influence at several stages of the sliding wear process. Sliding wear also is similar to mechanical alloying/ milling in that they both involve cyclic deformation. Each powder or wear surface area may be deformed many times. Solid-particle erosion also exhibits certain common features with mechanical alloying/milling. In particular, single-particle impact experiments may shed light on two aspects of mechanical alloying/ milling; that is 1) thermal effects, and (2) deformation of brittle materials. Impact velocities in single-particle impact experiments have been typically an order-ofmagnitude or more greater than in ballmilling (Sundararajan and Shewmon, 1983). Even so, the calculated AT values for various target materials at impact velocities of about 150 m s ~ l ranged from only 58 K for copper to 210 K for 301 stainless steel. Single particle impact experiments also provide information on the fundamental scaling relationship between fracture energy and plastic work for localized deformation conditions. The critical penetration depth of the particle-into-
210
5 Mechanical Milling and Alloying
target to initiate lateral cracks in the target depends on the geometry of the indentor and material or defect properties of the target material (Lawn and Evans, 1977). This implies that in small volumes, plastic deformation may occur instead of fracture. Therefore, in the milling of brittle materials, small asperities, sharp edges etc. of brittle irregularly shaped particles may indent other particles and cause plastic flow-cold-welding-instead of fracture, and thus allow mechanical alloying to take place.
5.4 ODS Alloys by Mechanical Alloying As pointed out in the Introduction, the driving force for the development of mechanical alloying by the International Nickel Co. was the goal of combining in a nickel-base superalloy the intermediate temperature strength of a precipitation hardened (gamma-prime, Ni3Al) material with the high temperature strength of an oxide dispersion strengthened (ODS) nickel (thoria dispersion (TD) strengthened Ni). The goal of dispersed-phase alloys is to strengthen the alloy matrix by impeding the motion of dislocations. The matrix remains the major load-bearing constituent. The metallic matrix will be strengthened in proportion to the effectiveness of the dispersion as a barrier to the motion of dislocations. While thoriated tungsten was developed early in the century (e.g. Coolidge, 1910), the first structural ODS alloy was sintered aluminum powder, SAP, developed by Irmann (Irmann, 1952). This material, Al containing A12O3 flakes, exhibited strength up to the melting point of the Al matrix. The success of SAP led to development of the higher melting
material thoriated nickel (ThO 2 in a Ni matrix, Anders, 1964). It was desired to develop ODS alloys with more complex matrices, such as the superalloys. However, the synthesis of fine, homogeneous oxide dispersions in the Ni-base superalloys which contain reactive metals such as Al and Ti by techniques such as internal oxidation, ignition surface coating, selective reduction, or powder mixing was not feasible (Benjamin, 1970). Thus the INCO effort involving high energy ball milling - mechanical alloying was required. This process involves highenergy ball milling of the starting powders together with the dispersoid particles, followed by consolidation and thermomechanical processing. In a similar process, termed "reaction milling", Jangg et al., 1975, succeeded in forming well-distributed dispersoid particles by chemical reaction of milling additions with the powder. The advantage of dispersion-strengthened alloys lies in their ability to retain useful strength up to a high fraction of their melting temperatures (>0.9 Tm) where other strengthening mechanisms such as strain-hardening, precipitation hardening, or solid solution strengthening rapidly lose their effectiveness. Most of the development of ODS alloys by mechanical alloying has been devoted to the Ni-base superalloys and Fe-base high temperature alloys. Much attention has also been given to Al-base ODS alloys. More recently, other ODS materials, such as Ti-base alloys and intermetallic compounds, have been explored. A number of review articles have described the synthesis, structure, and properties of specific ODS alloys prepared by mechanical alloying. These include the reviews of Gilman and Benjamin (1983), Benn et al. (1984), Sundaresan and Froes (1987), and a number of papers in the books "New Materials
211
5.4 ODS Alloys by Mechanical Alloying
by Mechanical Alloying Techniques", edited by Arzt and Schultz (1989) and "Solid State Powder Processing", edited by Clauer and de Barbadillo (1990). The several classes of mechanically alloyed ODS alloys will be briefly described in turn. 5.4.1 ODS Ni-Base Superalloys and Fe-Base High Temperature Alloys Produced by Mechanical Alloying
The major commercial dispersion strengthened Ni-base superalloys produced to date by mechanical alloying are INCONEL alloys MA 754, MA 758 and MA 6000. A new, but similar, ODS Nibase superalloy via mechanical alloying, TMO-2, is being developed in Japan (Yamazaki et al., 1990). The compositions of these alloys are given in Table 5-4. INCONEL MA 754 was the first mechanically alloyed ODS superalloy to be produced in large quantity (Weber, 1980). It is essentially a Ni-20%Cr alloy strengthened by about 1 vol.% Y 2 O 3 . The Ni, Cr, and Y 2 O 3 powders are milled until a homogeneous Ni-20 % Cr alloy is formed in which the Y 2 O 3 particles are uniformly distributed. Hot extrusion, often followed by hot rolling, is the typical consolidation process for mechanically alloyed ODS alloys. The powder is sealed in a steel can for these operations. A recrystallization step, often directional, follows consolidation that results in elongated, high-aspect-ratio grains that are very stable owing to the
inert oxide pinning. After the directional crystallization the grains have typical dimensions of «500 to 700 |im parallel to the working direction and «15 jam perpendicular to this direction (Stephens and Nix, 1985). The oxide dispersoid distribution is shown in Fig. 5-12. The fine particles are a uniform dispersion of stable yttrium aluminates formed by the reaction between the added Y 2 O 3 , excess oxygen in the powder, and aluminum added to getter oxygen (Benjamin et al., 1974). The larger particles are titanium carbonitrides. The dispersoids are typically 14 nm in diameter with an average spacing of 0.2 jim. The 1093 °C stress rupture properties of INCONEL MA 754 are compared to those of
-4-
•*> ^ ,
-i>
' ^ - i**#' -A * *F-^'
Er- ^-
Figure 5-12. TEM micrograph of INCONEL MA 754 showing uniform distribution of fine primary dispersion, the presence of coarser carbonitrides, and microtwins (Courtesy of John H. Weber, INCO Alloys International).
Table 5-4. Composition of Ni-base ODS superalloy by mechanical alloying (in wt.%). Alloy
Ni
Cr
Co
W
Mo
INCONEL MA 754 INCONEL MA 758 INCONEL MA 758 TMO-2
Bal. Bal. Bal. Bal.
20 30 15 6
_
_ 4 12.4
_ 2 2
9.7
Ta
Al
Ti
Y2O3
_ 2 4.7
0.3 0.3 4.5 4.2
0.5 0.5 2.5 0.8
0.6 0.6 1.1 1.1
212
5 Mechanical Milling and Alloying
50 30
Inconel alloy MA 754 bar
300 200
20 TDNiCr bar
£ 10 •£
6
100 a
a. Thoriated nickel bar Alloy 80 A
50 40
s £
{/)
in
MAR-M 509
CD
20 .fc
10 6 10 100 Rupture life in hours
other alloys in Fig. 5-13. MA 754, like other ODS materials, has a very flat log stresslog rupture life slope compared to conventional alloys. The strength of MA 754, about 100 MPa for 100 hours life, is somewhat higher than both of the other ODS alloys and several times greater than the conventional materials MAR-M alloy 509 and alloy 80 A. Thus, while MA 754 is comparable to TD NiCr it has a non-radioactive dispersoid and high strength, so it is suitable for applications such as gas turbine vanes. INCONEL MA 6000 combines precipitation strengthening (gamma prime, y', precipitates) from its Al, Ti, and Ta content for intermediate temperature strength with oxide dispersion strengthening from the Y 2 O 3 addition for strength and stability at very high temperatures. It typically contains about 52% of y' precipitates. The dispersoid dimensions are 30 nm average diameter and 0.1 ^im average spacing. As in MA 754, the Y 2 O 3 reacts with oxygen and aluminum to form uniform dispersions of yttrium aluminates, e.g. yttriumaluminum garnet 5 A12O3 • 3 Y 2 O 3 (YAG). The YAG dispersoid appears to be very stable. Negligible coarsening is observed for stress rupture tests at ^750°C and only small coarsening for tempera-
Figure 5-13. Stress rupture properties of INCONEL MA 754 at 1093 °C compared to other bar materials (after Weber, 1980).
1000
tures of 950 to 980 °C at rupture lives beyond 104 hours. These changes cause no serious loss in the loadbearing capability of the alloy for practical applications (Benn et al., 1984). The 1000-hour creep rupture strength for MA 6000 is illustrated in Fig. 5-14 along with several other MA ODS alloys and cast and wrought superalloys (Sundaresan and Froes, 1987). It is clear that MA 6000 has superior rupture strength at the highest temperatures 600
800
900 1000 Temperature in °C
Figure 5-14. 1000 h creep rupture strength for MA ODS alloys compared with cast and wrought superalloys: (1) MA 956, (2) MA 764, (3) MA 753, (4) MA 6000, (5) MAR-M 200, and (6) Ni80A (after Sundaresan and Froes, 1987), reprinted with permission from JOM (formerly Journal of Metals, Vol. 39, No. 8, p. 24, a publication of The Minerals, Metals & Materials Society, Warrendale, Pennsylvania 15086).
213
5.4 ODS Alloys by Mechanical Alloying
(>900°C) and comparable strength to MAR-M 200 at intermediate temperatures. MA 6000 has been used to a small extent so far in gas turbine engine blades. The iron-base INCOLOY alloy MA 956 contains about 20 % Cr, 4.5 % Al, 0.5 % Ti and about 0.5% Y 2 O 3 . It can be used at operating temperatures of over 1300°C in corrosive atmospheres (Sundaresan and Froes, 1987). MA 956 also has excellent fabricability. It can be cold-worked and can be joined by several welding techniques. MA 956 sheet can be bent more than 150° around a diameter equal to twice the sheet thickness (Weber, 1980). The high temperature mechanical properties of ODS alloys produced by mechanical alloying have been reviewed by Arzt (1989). In particular the effects of dispersoid particles and grain structure on creep were described. It was concluded that grain boundaries can be very vulnerable in ODS materials. In order to minimize the detrimental influence of grain boundaries, a high grain aspect ratio (GAR) in the loading direction must be achieved. Single crystals are most desirable.
( ^ 1 0 n m to 1 jim). In order to maximize the strength of A1-A12O3 ODS alloys it was desired to obtain a finer and more uniform distribution of the A12O3 dispersion in Al. For a given vol.% of dispersoid, mechanically alloyed A1-A12O3 exhibited much higher yield and tensile strength than SAP. This is illustrated for tensile strength in Fig. 5-15. It was found necessary to use a process control agent such as stearic acid or methanol to prevent excessive welding. The decomposition of the process control agent during mechanical alloying and subsequent powder compaction by hot pressing and extrusion resulted in significant contamination by carbon. The carbon was believed present as A14C3 dispersoids. Gilman and Nix, 1981, used Nopcowax-22 DSP (C 2 H 2 -2(C 18 H 36 OH)) as a process control agent. They found that during vacuum hot pressing at 773 K the Nopcowax22 DSP was reduced by the Al, yielding A1 4 C 3 ,N 2 ,O 2 and H 2 . The O 2 presumably recombined with Al to form additional A12O3 while the N 2 and H 2 were removed during the vacuum hot pressing. Thus, dispersion strengthening in these materials comes from both A12O3 and
5.4.2 ODS Aluminum-Base Alloys Produced by Mechanical Alloying
The first ODS material designed as a structural system was SAP (sintered Al powder) as developed by Irmann (1952). SAP displayed a greater strength than pure Al and no changes were observed after extended heating near the melting point. Lenel et al., 1957, among others, developed SAP further and verified Orowan's model for dispersion hardening (Orowan, 1948) in this material (see Vol. 6, Chap. 7). Bars of SAP were made by extrusion of a mixture of 1 to 10 vol. % of A12O3 powder in pure Al powder. The A12O3 particles in SAP exhibit a wide distribution of sizes
mechanically alloyed aluminum
400 £
en c
350 "£ SAP
300 £ a
250 30,-
i
200 2
U 6 8 10 12 Total volume % dispersoid
U
Figure 5-15. Room temperature tensile strength vs. dispersoid content for SAP and MA Al (after Benjamin and Bomford, 1977).
214
5 Mechanical Milling and Alloying
A14C3 dispersoids. The effects of a variety of processing additives on the processing, structure, and properties of mechanically alloyed Al has been presented recently by Weber, 1990. Mechanical alloying was extended to Al alloys such as Al-4% Mg (Benjamin and Schelling, 1981). This material, with uniform, equiaxed O, C based dispersoids and Mg solid solution strengthening, gave tensile strengths of 520 to 585 MPa along with excellent resistance to corrosion and stress corrosion cracking. Some Al-base mechanically alloyed materials developed by INCO are listed in Table 5-5. Al-9052 is
tain A12O3 and A14C3 dispersions and are processed without a process control agent but with carbon black and controlled additions of oxygen from the milling atmosphere. The properties and applications of these DISPAL aluminum alloys have been reviewed by Arnhold and Hummert, 1989. The AL 9052 and AL 9021 alloys are also under development as matrices for SiC and BN particulate reinforced metal matrix composites (Jatkar et al., 1985).
Table 5-5. Some aluminum-base ODS alloys produced by mechanical alloying (in wt.%).
The possibility of dispersion strengthening Ti by mechanical alloying was first explored by Wright and Wilcox (1974). The first attempt was not successful in that a homogeneous alloy with a uniform dispersion of Y 2 O 3 in Ti (3 vol. % Y 2 O 3 ) was not attained. Considerable agglomeration of Y 2 O 3 occurred in part of the alloy along with bands relatively denuded in Y 2 O 3 . As a result of this inhomogeneous microstructure very little dispersion strengthening was realized. More recently, Sundaresan and Froes (1988) have studied mechanical alloying of several classes of Ti-base alloys including dispersion strengthened beta-Ti alloys. The alloy T i - 1 % A l - 8 % V - 5 % Fe with 1 % Er was milled in order to try to produce finer, more uniform dispersions of Er than exist in the starting atomized powder. The Er internally oxidizes to form Er 2 O 3 dispersoids during consolidation. Milling times up to 48 hours were not sufficient to give a homogeneous structure. The structure on consolidation was not uniform, exhibiting sub-micron grains with 30 to 50 nm dispersions along with regions of course grains of about 10 jim diameter. Similar non-homogeneous microstructures were observed for Ti-24%
Alloy
Mg Cu
Li
C
O
Inco MAP Al-9052 Inco MAP-A1-9021 Inco MAP AL-905XL
4 1.5 4
4 -
1.5
1.1 1.1 1.2
0.8 0.8 0.4
currently being considered for application as torpedo hulls and other marine applications. Alloy A1-905XL is an Al-Mg-Li alloy for airframe applications (Schelleng et al., 1988). Quist et al., 1985, have compared Al-Li alloys processed by mechanical alloying with those made by either rotary atomization or pulverized melt-spun ribbon. With superior mechanical properties and corrosion resistance the mechanically alloyed Al-Li appears to have advantages over the alloys made by the other processing methods. The reaction milling of aluminum, developed by Jangg and coworkers (Jangg et al., 1975), is now used to produce commercial alloys under the tradename DISPAL. The materials have been jointly developed by Krebsoge, Erbsloh-Aluminum and Eckart-Werke. The DISPAL alloys con-
5.4.3 Other ODS Alloys by Mechanical Alloying 5.4.3.1 Ti-Base Alloys
5.4 ODS Alloys by Mechanical Alloying
V - 1 0 % C r - 5 % Er alloys also milled 48 hours. However, in the fine-grained regions of this alloy a very fine (^10nm) homogeneous dispersion of Er 2 O 3 was observed. It was concluded that fine dispersions are possible in these Ti-base alloys but that optimization of the processing parameters is needed to obtain a uniform structure. Suryanarayana et al. (1990) studied the microstructure of Er 2 O 3 dispersoids in a Ti-25% Al-10% N b - 3 % V - 1 % Mo alloy with 2 wt. % Er. This alloy is essentially the Ti3Al oc2 ordered h.c.p. (DO 19 ) phase. The powder, received as rapidly solidified powder, was milled for times of 18 to 72 hours in a SPEX shaker mill. It was found that significant differences in grain size and dispersoid distribution occur between small and large powder particles. As in the above Ti-base alloys optimization of the process variables are needed to achieve desirable mechanical properties for this dispersion-containing Ti3Al ordered intermetallic. 5.4.3.2 ODS Intermetallics by Mechanical Alloying There has been renewed interest in intermetallic compounds in recent years as possible elevated temperature structural materials. While improving ductility and fabricability have been major research thrusts, improvement in strength and creep resistance at high temperature has also been addressed. Preliminary work on dispersion strengthening of intermetallics has been reviewed by Koch (1987) and Benn et al. (1990). The Ni-, Fe-, and Ti-aluminides have been the intermetallics most studied. Jang et al. (1988) studied the influence of oxide additions (0.5 to 2.5 vol.% of either A12O3, Y 2 O 3 , or ThO2) to Ni 3 Al (Ni-23.5 at.% Al-0.5 at.% Hf-0.2 at.% B) pow-
215
ders by mechanical alloying. Mechanical properties were measured on samples consolidated by hot isostatic pressing. The dispersoid additions refined the grain size and increased the yield strength but decreased the ductility at room temperature. All samples failed in a brittle manner at 513 °C owing to dynamic oxygen embrittlement. Jang and Koch (1988) found that the major strengthening effect of the oxide dispersoids on Ni3Al (Hf, B) is in their control of the grain size during powder consolidation by hot isostatic pressing. A HallPetch relationship was observed such that the yield strength of the fine-grained samples (1 to 2 jam) was more than twice that for the larger grained samples (35 jam) without dispersoids. The dispersoids themselves provided a small (^13%) additional strengthening effect via the Orowan mechanism. Benn et al. (1990) studied mechanically alloyed Ni3Al-base intermetallics with 0 to 1 % Y 2 O 3 additions. They prepared « 7 kg charges in attritor mills and ^35 kg charges in larger conventional ball mills. Consolidation of the canned powder was by extrusion at temperatures in the range 1065 to 1150°C. The as-extruded grain sizes were in the range of 0.5 to 1 jam with a fairly uniform distribution of fine oxide particles. Gradient annealing heat treatments provided some grain growth but only slight directional secondary recrystallization even after additional thermomechanical processing by rolling. It was not clear why directional grain growth was inhibited in these materials. Primary recrystallization did increase the average grain size to about 5 to 10 |im diameter. A comparison of the yield strength at different test temperatures for these ODS Ni3Al alloys was made against Ni3Al alloys produced by other methods. It was shown that the mechanically alloyed samples, with
216
5 Mechanical Milling and Alloying
fine grain sizes, have superior strength at temperatures up to about 600 °C. Above this temperature the strength drops off rapidly, presumably because of the stable fine grain structure and the onset of grain boundary sliding etc. There was also evidence of dynamic oxygen embrittlement in the 600 to 800 °C temperature range as shown by ductility decreases. If ODS Ni3Al alloys are to be considered for elevated temperature applications, processing to provide large, elongated grains (as in ODS superalloys such as MA 6000) must be developed. Vedula and Strothers (1990) have studied the B2 FeAl intermetallic (40 at. % Al) with and without 1 vol. % Y 2 O 3 . The powder (prealloyed Fe-40 at.% Al-0.1 at.% Zr-0.2 at.% B + l vol.% Y 2 O 3 ) was mechanically alloyed in a high energy vibratory mill for 32 h. The room temperature ductility of this ODS FeAl intermetallic was as high as 9 % - the highest ductility observed in B2 FeAl compounds. The ODS alloys were also much stronger than the non-ODS alloys at room temperature. Mechanical tests were also carried out at 827 °C. These included tensile tests and compressive creep measurements. In contrast to the observations on ODS Ni 3 Al, the yield stress of the ODS Fe-Al alloy was higher than that of the alloy without oxide dispersions and the compressive creep rates were lower. These results were most pronounced for the finest grain size (4 jim) ODS alloys. Thus, in the ODS FeAl alloys grain boundary strengthening was maintained at temperatures up to at least 827 °C. The strengthening effects of the oxide dispersions are attributed mainly to their role as grain refiners. 5.4.3.3 ODS Refractory Metal Alloys Because of their reactivity, the milling of refractory metal alloys can result in the
formation of oxides or carbides if the milling environment allows for such contamination. The reports of planned additions of dispersoids to refractory metals for alloys are limited. Oxide dispersion hardening has been used in Nb to increase its strength for application as an implant material (Schider, 1986). A mixture of pure Nb and oxide powder was mechanically alloyed in an attritor mill, cold isostatically pressed, sintered, and extruded. It was shown that the ODS Nb is biocompatible, strong, and ductile.
5.5 Synthesis of "Equilibrium" Phases by Mechanical Alloying In this section examples are given of material synthesis by mechanical alloying of elemental component powders which produce the phase predicted by the equilibrium phase diagram. The word "equilibrium" is put in quotation marks in the section heading to indicate that there may be some non-equilibrium features associated with the equilibrium phase product of mechanical alloying. 5.5.1 Solid Solutions There are several examples of the formation of solid solutions by mechanically alloying the elemental component powders in alloy systems that exhibit complete solid solubility. Benjamin (1976), first showed that true atomic level alloying occurred by mechanical alloying in Ni-Cr solid solution alloys. He reported that the magnetic behavior of the Ni-Cr alloys prepared by mechanical alloying was identical to that of Ni-Cr alloys of the same composition prepared by conventional ingot metallurgy. Ge and Si exhibit complete solid solubility, which would be expected since they are isoelectronic, have the same crystal struc-
5.5 Synthesis of "Equilibrium" Phases by Mechanical Alloying
ture (diamond cubic), and their lattice parameters differ by only about 4.0%. Ge and Si are, however, both nominally brittle at room temperature and they were taken as components in a study of the MA of brittle materials (Davis et al., 1988). Davis and Koch (1987) had demonstrated that ball milling Ge and Si powder together resulted in material transfer and alloying. The lattice parameters of Ge and Si move together with milling time and after about 4 h to 5 h merge into the single lattice parameter for the Ge-Si solid solution. This change in lattice parameter is illustrated in Fig. 5-4 for a Ge-72 at.% Si alloy. It was found that in Si-rich compositions (> 50 at. % Si) the lattice parameters fell on the linear extrapolation between the lattice parameters for pure Ge and Si (Vegard's "law"). However, in Ge-rich alloys (>50 at.% Si) the lattice parameters of the as-milled powders were larger than predicted for Vegard's law. Annealing the alloys with the expanded lattice parameters serve to decrease them in the direction of the Vegard's law line - i.e. partial "recovery" was observed. Impurities are apparently not responsible for the expanded lattice parameters in Ge-rich MA Ge-Si alloys. The possibility of nonequilibrium (e.g., partially interstitial) solid solutions is being explored with further study of the recovery of the lattice parameters by annealing (Leonard and Koch, NCSU, work in progress). Another example of an apparent nonequilibrium solid solution due to MA was noted during MA of pure Nb and Sn powders to form initially the A15 intermetallic compound, Nb 3 Sn, which then was found to transform to the amorphous structure on further milling (Kim and Koch, 1987). The lattice parameter measured for the A15 phase which was MA with tungsten carbide milling media under
217
argon was 0.529 + 0.001 nm, in excellent agreement with literature values for bulk A15 Nb 3 Sn. The powder milled with steel media, however, exhibited expanded lattice parameters. The oxygen levels in both sets of samples were similar, but the samples milled with the steel balls and vial contained significant quantities of iron impurities incorporated from the milling media. The apparent expansion of the Nb 3 Sn lattice by Fe impurities is in contrast to the results of Caton (1985) who measured a small contraction of the A15 Nb 3 Sn lattice with iron additions in sintered material. The latter result is consistent with the smaller atomic radius at Fe (0.127 nm) compared to Nb (0.147 nm) and Sn (0.155 nm) (Pearson, 1972). Since it appears that the lattice expansion in MA Nb 3 Sn is due to incorporation of Fe, it is concluded that Fe must be present in the Nb 3 Sn lattice in a non-equilibrium, perhaps partly interstitial, arrangement. 5.5.2 Intermediate Phases
Intermediate phases and intermetallic compounds have been synthesized from the pure components by MA in several alloy systems. The equilibrium Hume-Rothery electron compounds, p'-brass (McDermott and Koch, 1986), y-brass, and e-brass (McDermott, 1988), were synthesized by MA of pure Cu and pure Zn powders mixed in the proper proportions. Thus the equilibrium intermediate phases were synthesized by MA of the elemental components. In the case of P'-brass, evidence for f.c.c. deformation-induced martensite was observed by X-ray diffraction as a minor phase in the p'-brass matrix. A nonequilibrium phase was therefore also present as a result of the plastic deformation which is an integral part of MA. Lee et al. (1990) have shown that the y-brass (Cu5Au8)
218
5 Mechanical Milling and Alloying
phase is very stable to milling. Milling for 60 h in a planetary mill only served to broaden the X-ray diffraction lines but did not alter their positions. Similar milling conditions easily lead to amorphization in other compounds such as NiZr 2 . The stability of Cu 5 Zn 8 was attributed to the modest value of stored energy of cold work obtained by milling (2 kJ mol" 1 ) compared to the free energy difference between the Cu 5 Zn 8 compound and its supercooled liquid at 300 K (9 kJ mol" l ). Kim (1987) and Koch and Kim (1985) synthesized the intermetallic compounds Nb 3 Ge, Nb 5 Ge 3 , and NbGe 2 by MA of elemental Nb and Ge powders. In the cases of Nb 3 Ge and Nb 3 Sn, continued milling eventually resulted in the formation of an amorphous phase. The topic of amorphization will be discussed in more detail in Sec. 5.6. Ivanov et al., 1988, studied the synthesis of Ni and Co aluminides and amorphous alloys by MA of the elemental powders. They showed the formation of the intermetallic compound Ni 2 Al 3 from mixtures of Ni and Al powders at a composition of Ni 40 Al 60 . Continued milling produced a metastable P'-NiAl phase which reverted to the rhombohedral Ni 2 Al 3 phase after annealing. Powder of composition Ni 30 Al 70 formed a metastable Ni 2 Al 3 intermetallic on MA that decomposed to two-phase Ni 2 Al 3 and Al3Ni by annealing. Ni 75 Al 25 powder composition formed an f.c.c. solid solution which ordered to the Ll 2 , Ni3Al compound on annealing. The formation of an amorphous alloy was observed in a narrow range of composition (27 to 35 at.% Al). Atzmon (1990) has made very interesting observations relevant to the mechanism for MA during the synthesis of Al3Ni and NiAl intermetallic compounds by MA of elemental Ni and Al powder in a SPEX shaker mill. In a powder mixture of
composition Al 75 Ni 25 MA produced the Al3Ni compound in a gradual manner consistent with layer diffusion. The reaction at 60 °C (ambient conditions) was faster than at 35 °C (cooling with a fan) presumably because of the higher interdiffusion coefficient. The formation of the NiAl compound was found to depend on the vial temperature, the atmosphere in the vial, and whether milling was interrupted or continuous. Gradual transformations to NiAl occurred for continuous milling (2.5 h) in pure argon at 35 °C and at 60°C in argon (2 h). Transformation to NiAl was seen at somewhat shorter milling times at 35 °C in a mixture of air and argon. The fascinating observation occurred after milling in pure argon at 35 °C for 1.7 to 2 h. Milling was then stopped for periods of 1 to 12 h, allowing the vial to cool to room temperature. Separate experiments showed that the powder at this stage was still elemental Al and Ni mixed lamellae. Shortly (30 to 60 sec later) after resumption of milling after such a pause, an exothermic reaction occurred. A thermocouple mounted in the vial bottom indicated temperature rises of 27 °C within about 1 sec while a thermocouple in the side wall showed a rise of 20 °C in 7 sec. Atzmon estimated that under adiabatic conditions this exothermic heat released from the sample as 49 kJ mol" 1 which is comparable to the heat of formation of NiAl (59 kJ mol" 1 ) given the approximate nature of the measurement. In fact, again assuming adiabatic conditions, it was estimated that the heat of formation of NiAl was sufficient to bring solid NiAl from 25 °C to 1738 °C (100 °C higher than its melting point). Atzmon speculated that the interruption and subsequent cooling (and possibly oxidation?) result in lower ductility and therefore resumed milling leads to energy concentration in smaller volumes, re-
5.5 Synthesis of "Equilibrium" Phases by Mechanical Alloying
suiting in higher local temperature upon impact. This explanation seems unlikely since Al and Ni do not normally exhibit ductility losses at low temperatures, and the average temperature difference (35 °C to 25 °C) is so small. More research is needed to understand this important experiment. Kumar and Mannan (1989) observed a similar rapid phase formation during MA of a mixture of pure Nb and Si powders at an average composition corresponding to Nb 5 Si 3 . After milling for 73 min in a SPEX mill the powders remained elemental Nb and Si as determined by X-ray diffraction. Only two more minutes of milling (perhaps less), i.e. X-ray examination at a total milling time of 75 min, resulted in complete transformation to crystalline Nb 5 Si 3 in three tetragonal allotropic forms: oc-Nb5Si3, the stable room temperature form; p-Nb 5 Si 3 , the high temperature form; and y-Nb 5 Si 3 , believed to be stabilized by the presence of impurities. Further milling increased the amount of P-Nb5Si3 at the expense of ot-Nb5Si3 and y-Nb 5 Si 3 . No temperature measurements were made, but since Nb 5 Si 3 is a high-melting point intermetallic with an estimated heat of formation o f - 5 4 kJ mol" 1 (de Boer et al., 1988) an exothermic reaction as in the case of NiAl is likely. A much larger exothermic heat effect has been observed during the reduction of CuO by Ca by mechanical alloying (Schaffer and McCormick, 1990). This will be discussed later in the Chapter. Therefore, it appears that in the case of compound formation or chemical reactions with large exothermic heats of reaction, significant heat input leading to high temperatures and even explosive reactions can occur by mechanical alloying. This is in contrast to the very modest ( < 1 0 0 200 °C) temperature rises usually attributed to the heat generated from the kinetic energy of the milling media.
219
Morris and Morris (1990) used mechanical alloying in a Fritsch Mini-Planetary ball mill to synthesize the Cr 2 Nb intermetallic compound from elemental Cr and Nb powders. After about 15 h of milling X-ray diffraction lines of the hexagonal Laves phase, Cr 2 Nb, were observed. This is the stable phase at high temperatures but it undergoes an allotropic transformation to the cubic Laves phase at about 1600°C which is the room temperature equilibrium phase. Further milling to 20 and 25 h resulted in the disappearance of the hexagonal Cr 2 Nb Laves phase and the reappearance of b.c.c. lines of Cr and Nb. Such "demixing" reactions due to milling will be discussed further in Sec. 5.6. However, Kumar (1990) did not observe the formation of the Cr 2 Nb intermetallic compound on milling elemental Cr and Nb powder at compositions from 49 to 52 wt. % Cr (within the Cr 2 Nb phase field). He observed the evolution from the elemental components to an amorphous structure during milling in a SPEX shaker mill. The differences in these results can not be assessed without more information. Kumar carried out his milling in an argon atmosphere in the milling vial while the milling atmosphere in the work of Morris and Morris was not reported. Vial temperatures were not reported for either experiment. These very different results on milling the same components point out how critical the milling variables such as atmosphere, vial temperature, and mill energetics must be with regard to the end products of the solid state transformations induced during milling. A number of intermetallic compounds have been produced with MA as the first step in a synthesis procedure. Benn et al. (1988) have reviewed several intermetallic compound systems where an intimate mixture of the elemental components was at-
220
5 Mechanical Milling and Alloying
tained by MA, as well as partial compound formation in some cases. The intermetallic synthesis was then completed during the thermomechanical treatments carried out for compaction of the powders. Larson et al. (1977) have produced the A15 structure intermetallic Nb 3 Al by MA of stoichiometric mixtures of Nb and Al powders followed by heat treatment. Similarly, (Benn et al., 1988), Ti3Al and TiAl have been synthesized by milling in the presence of a process control agent under inert atmosphere. Subsequent heat treatment at 540 °C (Ti3Al) or 600 °C (TiAl) produced the given intermetallics. The compound Al3Ti is difficult to prepare by conventional ingot metallurgy because of the high Al vapor pressures, the large differences in melting points between Al and Al3Ti, and the peritectic solidification behavior. It has, however, been synthesized by MA followed by annealing (Benn et al., 1988). 5.5.3 Immiscible Alloy Systems Mechanical alloying offers one of the few methods for producing a homogeneous mixture of two or more immiscible phases. This is the case for the ODS alloys where the oxides are essentially insoluble in the metallic matrices. More generally, MA may be applied to binary alloy systems that exhibit solid, or even liquid, immiscibility. Benjamin (1976) described the synthesis by MA of homogeneous mixtures of Fe-50 wt. % Cu, a system that exhibits limited solid solubility, as well as Cu-Pb alloys for which there is a liquid miscibility gap. Patel and Diamond (1988), have used MA, sometimes in conjunction with rapid solidification methods, to synthesize fine homogeneous phase distributions in immiscible alloys such as Cu-Cr, Al-In, CuW and in immiscible Cu-base bearing alloys such as Cu-Pb-Sn.
Green et al. (1984) have developed a novel electrical contact material by MA of Cu-15 vol.%Ru mixtures. Copper and ruthenium are mutually insoluble. A CuRu composite was produced by MA the elemental powders, annealing the MA powders, cold-pressing, and warm rolling the composite. Cold-rolling and annealing were used to obtain the final strip dimensions. Scanning electron microscopy (SEM) revealed the final size of the Ru particles to be about 1 to 2 jim in diameter. Removal of surface copper by etching produced a structure in which the hard, refractory, and conductive Ru particles protrude from the surface and serve as the electrical contacts, supported by the Cu matrix that provides electrical continuity. Iron and magnesium are immiscible in the solid state and form a large miscibility gap in the liquid state (Kubaschewski, 1982). Konstanchuk et al. (1987) studied the hydriding properties of a Mg-25 wt. % Fe composite produced by MA. The microstructure after MA consisted of a laminated mixture with iron dispersed in a magnesium matrix. Clean metallic contact was achieved at the Mg/Fe interfacial boundary; surface oxides of iron being reduced by magnesium to form a clean metallic surface. A separate experiment was conducted to show that the magnesium did indeed reduce Fe 2 O 3 to Fe during the MA process. The Mg-25 wt. % Fe composite exhibited a large hydrogen storage capacity (5.1 to 5.8 wt. %) and the rate of hydriding was considerably higher than that for pure magnesium. At hydriding temperatures above 625 K, a ternary hydride, Mg 2 FeH x , formed which inhibited dehydriding reaction rates. Thus, a temperature of about 615 K was deemed to be optimum for the hydriding reaction. A question regarding the phase distribution in mechanically alloyed immiscible
221
5.5 Synthesis of "Equilibrium" Phases by Mechanical Alloying
of the system, i.e., large positive heat of mixing. With sufficient mobility, the like atoms will segregate together. An order of magnitude estimate for this segregation, using *jDt as the characteristic diffusion distance, and ambient temperature diffusion coefficients, was consistent with the microstructural observations. A particularly interesting observation from the study of MM immiscible Ge-Sn and Ge-Pb systems was the depression of the melting point of Sn or Pb with milling time and Ge concentration. The first observation of the melting point depression was noted in DSC scans for a Sn-45 vol. % Ge mixture as a function of milling time (Koch et al., 1989). The magnitude of the melting point depression, ATm, increased with milling time, that is, with refinement of the dispersion. Melting point depressions were defined from the DSC endotherms as A7^ eak and A7^ail as illustrated in Fig. 5-16. The magnitudes of the AT^'s were found to reach constant values after 32 h of milling. After 32 h of milling the average diameters of the hard Ge particles embedded in the Sn (or Pb) matrix were
-rtail 1 M
T peak 1
M
T
t
uni
(f)
-
flow larbit
systems is: what is the lower size distribution limit to which immiscible particles can be milled? Since in solid solution alloys, MA can result in alloying at the atomic level, very fine particle distributions might be attainable by MA immiscible components. Mechanical alloying has recently been applied to the immiscible systems Ge-Sn, Al-Ge, and Ge-Pb with the goal of determining the limits of phase refinement (Gross, 1988). Mechanical alloying was carried out for milling times up to 60 h under an air or argon atmosphere in a SPEX mixer/mill. Some tests were conducted with the milling vial cooled with a stream of liquid nitrogen. X-ray diffraction of powder taken at various milling times showed the systems remained as twophase pure components and their precise lattice parameters remained constant with milling time, which indicated no alloying or contamination to the accuracy of the lattice parameter measurements. Optical microscopy, SEM, and transmission electron microscopy (TEM) were used to follow the progress of the microstructural refinement with milling time. The average center-to-center nearest neighbor distance between dispersed Ge particles was measured by the above metallographic techniques. The logarithm of Ge interparticle distance exhibits an inverse linear dependence on milling time. After 32 h of milling, the average Ge interparticle distance was about 20 nm. Stereology techniques determined that the Ge particles had a random Poisson distribution throughout the Sn matrix. Further milling did not appear to significantly change the particle sizes or distributions so the limit of refinement was attained for the given milling conditions in this system. The mixing of these insoluble components by mechanical attrition is balanced by the demixing driven by the thermodynamics
^
\
/
V
X
180
200 220 Temperature in °C
240
Figure 5-16. DSC scan for the melting of Sn in Sn45.5 vol.% Ge powder milled 32 h. 7geak and T£il are defined as shown (Jang and Koch, 1990 a).
222
5 Mechanical Milling and Alloying
approximately 10 nm. As Ge concentration was increased in each system, for a constant milling time of 32 h, the melting point of Sn (or Pb) decreased (Jang and Koch, 1990a). ATm for Sn is plotted against vol. % Ge in Fig. 5-17. The melting point depression increases with vol. % Ge. No ATm data are shown for Ge concentrations > 80 vol. % because no endothermic peak could be seen for the melting of Sn for samples containing > 80 vol. % Ge. This is not a problem of resolution. Unmilled powder of 88 and 95 vol. % Ge clearly showed a well-defined endothermic melting peak in the DSC at the Tm of pure Sn, and with the enthalpy of fusion for pure Sn, 60 J g" 1 . The enthalpy of fusion for Sn, determined from the area of the melting endotherms, is plotted against vol. % in Fig. 5-18. The enthalpy of fusion, A// m , decreases with Ge concentration and finally disappears for the Ge-rich mixtures (88 and 95 vol. % Ge). Only minor changes in Tm and AHm were observed after heating the samples in the DSC through the melting point, cooling to room temperature, and re-heating. These experimental results suggest that the premature melting is nucleated at the Ge/Sn interfaces. As the fraction of Sn atoms adjacent to the Ge particle surfaces increases, the melting point and enthalpy of fusion decreases. When the density of Ge particles becomes so large that all the Sn atoms are within a few atomic layers of the interface, the Sn may assume a disordered or amorphous structure. Selected area electron diffraction patterns were obtained on samples of 76.5 vol.% Ge and 88 vol. % Ge. The electron diffraction pattern for the 76.5 vol. % Ge sample shows a ring and discrete spots for the (101) reflection of crystalline Sn. This reflection is not seen for the 88 vol. % Ge sample but a ring of diffuse intensity is observed near the
50 m
40
• •
•
A
O
30
•
o
:20
•
9
A
o
•
10
20
o
o
•
A
40 60 vol.% Ge
80
100
Figure 5-17. Melting point depressions, A7^ eak and AT^ail for Sn as a function of vol.% Ge after milling 32 h. ATNJeak is also shown for samples milled 32 h and then cycled once to a temperature >505 K. • Ar,Jail: o ATfPeak: A A7^eak after heating cycle to >505 K (Jang and Koch, 1990 a). 70.0
40
60
80
100
vol.% Ge Figure 5-18. Enthalpy of fusion of Sn as a function of vol.% Ge after milling 32 h. Data for samples cycled once to above 505 K are also shown, o as-milled: A after heating in DSC to > 505 K: A unmilled Sn88 vol.% Ge (Jang and Koch, 1990a).
position where the (101) Sn line should appear. The electron diffraction data suggest that an amorphous structure has been induced in the Sn layers trapped between the Ge particles by milling for the 88 vol. % Ge sample. That mechanical milling of powder of immiscible components can result in a very
5.5 Synthesis of "Equilibrium" Phases by Mechanical Alloying
fine dispersion at the level of nanocrystalline dimensions has been demonstrated by Schlump and Grewe (1989) in systems such as Fe-W, Cu-Ta, TiNi-C, and W-NiC. Similarly, Shingu et al. (1989) found a fine grain structure in immiscible Ag-Fe powders at the nanometer level. Nanocrystalline structures prepared by ball milling will be discussed in more detail in Sec. 5.6.6. Fukunaga et al. (1990) give evidence for at least the partial amorphization of the immiscible system Cu-Ta. Iron impurities from steel milling media accelerated the amorphization reaction, but amorphization was still observed when milling was carried out with Cu-Be balls and vial. This experiment will also be discussed further in the Section 5.6.3 on amorphization by milling. 5.5.4 Synthesis of Materials for Special Applications Mechanical alloying/milling has been used to prepare unique materials for a variety of special applications. The excellent permanent magnet material, Nd 1 5 Fe 7 7 B 8 (see Vol. 3A, Chap. 5), has been prepared by MA of the elemental powders followed by annealing (Schultz etal., 1987). The milling was carried out under argon in a cylindrical steel chamber in a planetary ball mill. The Fe and Nd particles are milled (30 h) until the Nd Xray diffraction peaks are no longer visible and only broadened peaks for Fe are observed. Amorphous boron particles do not mechanically alloy with the Fe-Nd matrix. After annealing 1 h at 600 °C, the boron does dissolve in the Fe-Nd powder and the Nd 2 Fe 14 B phase is formed. The magnetically isotropic particles so formed have a very fine microstructure comparable to rapidly quenched samples, exhibit a do-
223
main wall pinning behavior and have excellent hard magnetic properties such as Hc up to 13 kOe and BHmax up to 12.8 MGOe. Ivanov and coworkers (Ivanov et al., 1987; Song et al., 1987; Stepanov et al., 1987; Konstanchuk et al., 1987) have developed Mg-base alloys for hydrogen storage by mechanical alloying. Mechanical alloying of Mg with either Ni, Fe, Co, or Ce was carried out in a planetary mill under an inert or hydrogen atmosphere for short times (3 to 15 min). X-ray diffraction of the powders after MA showed only Mg and the given metal. Hydrogen absorption and desorption experiments were then carried out on the MA powders. All the samples demonstrated relatively high reactivity with hydrogen. The elements added to Mg could be classified with regard to their hydrogenation behavior as follows (a) those, e.g. Ni, forming an intermetallic compound (Mg2Ni) capable of absorbing and desorbing hydrogen, (b) those, e.g. Ce, forming hydrides that can function as hydrogen pumps due to stoichiometric variations, and (c) those systems (Mg-Co, MgFe) that do not form hydrides. The intermetallic compound Mg2Ni was formed from the mechanically alloyed Mg-Ni powder during hydriding and dehydriding cycles. The reaction temperature was 583 K, and the hydrogen pressure during hydriding was 0.8 MPa, and 0.15 MPa during dehydriding. The volume fraction of Mg 2 Ni increased with hydriding/dehydriding cycling. The compound formation by interdiffusion of Mg and Ni at 583 K was presumably enhanced by the defects created by the expansion and contraction of the lattice during the process of hydriding/dehydriding cycling. Mechanical alloys, composites of Mg with 5 to 20 wt. % Fe (or Ni, Ti, or Cu), have also been used as supercorroding alloys. These alloys, developed by the Naval
224
5 Mechanical Milling and Alloying
Civil Engineering Laboratory, operate as short-circuited galvanic cells to react rapidly and predictably with seawater to produce heat and hydrogen gas (Black, 1979; Sergev et al., 1981) for several marine applications. Mechanical alloying provided the appropriate microstructures for these cells to function, which could not easily be attained by other, more conventional methods. Mechanical alloying has been used in the past to synthesize certain high magnetic field A15 superconductors such as Nb 3 Al (Larson et al., 1977) and Nb 3 Sn (White and Nix, 1979). It was used because of the difficulty in preparing such compounds by standard solidification techniques due to the very large differences in the components melting points and the peritectic solidification behavior. The mechanically alloyed powders formed the A15 compound during compaction and the resulting material exhibited good homogeneity. Inoue and Masumoto (1989) have prepared the new high Tc oxide superconductors by MA followed by oxidation heat treatment. They studied the possibility of forming a homogeneous phase in immiscible Ba-Ln-Cu alloys (Ln = Y, Gd, Ho, or Er) as a precursor to the high Tc oxides. Mixed powders of Ba, Cu, and Ln 80 Cu 20 (Ln = Y, Gd, Ho, or Er) were milled in a conventional laboratory ball mill with an argon atmosphere in the milling vial. After 10 h of milling the X-ray diffraction peaks of the starting powder mixture (Ba, Cu, Ln 80 Cu 20 ) disappear and are replaced by peaks for a non-equilibrium f.c.c. Cu solid solution phase. Thus, it was possible to produce a homogeneous metallic phase precursor for subsequent oxidation. However, the LnBa 2 Cu 3 O 7 _ 8 (Ln = Y, Gd, Ho, or Er) material formed by oxidation (920 °C in oxygen) had superconducting
properties no better than material prepared by the conventional sintering process.
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool The application of high energy ball milling that has stimulated the most research interest in recent years is its use as a nonequilibrium processing tool. It has been realized that mechanical alloying/ milling can be used to synthesize metastable structures, in analogy to other nonequilibrium processing methods such as rapid solidification and physical vapor deposition. However, the precursor phase in the case of MA/MM is typically a crystalline solid, or solids, rather than liquid or vapor. The thermodynamics and kinetic factors which govern metastable phase formation can therefore be very different. In this section the various, often competing, metastable structures that have been made by the high energy ball-milling of powder will be reviewed. 5.6.1 Extended Solid Solutions Equilibrium solid solubility limits are often exceeded by nonequilibrium processing methods such as rapid solidification. This is also true for MA. There are a number of examples of this effect in the literature, but, with the exception of recent work by Polkin et al. (1990), no systematic studies have been reported, to the author's knowledge, on solid solubility enhancement by MA. Extended solid solubilities have been noted in the process of studying amorphization in several alloy systems. Schwarz et al. (1985) found that the solubility limit of Ti in f.c.c. Ni was approxi-
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
amorph.cryst.- &
0)
c 0) CD if)
.a
mately 28 at. % on MA of Ti and Ni powders. This may be compared with a predicted solubility limit from the equilibrium phase diagram of only a few percent (Hansen, 1958). It was suggested (Schwarz et al., 1985) that this increased solid solubility could be explained by the metastable equilibrium between the oc-Ni f.c.c. solid solution and the Ni-Ti amorphous phase as opposed to the stable equilibrium between oc-Ni and Ni 3 Ti. The common tangents to the free energy curves calculated for these phases at 235 K (the milling temperature used) are consistent with the observed solubility limits. This is illustrated in Fig. 5-19. A similar result was obtained by Lee and Koch (1987) in a study of amorphization of Ni-Nb alloys by MA. The terminal Ni f.c.c. solid solution and the Nb b.c.c. solid solution were found to
225
Figure 5-19, Free energy of alloys of Ni and Ti. The heavy solid curve is the free energy of the amorphous phase. The thin curves are the free energies of the crystalline terminal solid solutions. The dotted curves are the free energies of the crystalline intermetallic compounds. The dashed lines are tangents common to the crystalline solid solutions and the amorphous phase. These define composition regimes a, b, c, and d. Regime (a) is the crystalline Ni-amorph. rich terminal solution. Regime (c) is the single-phase amorphous alloy. Regimes (b) and (d) are two_ ^ % -cryst. phase mixtures of amorphous alloy and the terminal solid solution of the major element. The symbols near the bottom of the figure denote the products obtained by MA of pure Ni and Ti powders: Solid and half-solid symbols denote single-phase and two-phase products, respectively (after Schwarz et al., 1985).
be approximately 10 at. % Nb and 10 at. % Ni respectively. This may be compared with solubility limits determined by Duerden and Hume-Rothery (1966) of 4.2 at. % Nb (at 987 °C) and 3.5 at.% Ni (at 1000 °C). The solubility limits at the milling temperature (nominally 60 °C) are expected to be much lower. Hellstern et al. (1988) found an extended solid solubility of Al in b.c.c. Nb of about 30 at. % for MA of elemental Nb and Al powder. The stable equilibrium diagram (Lundin and Yamamoto, 1966) for Al-Nb indicates Al solubility of <10 at.% for temperatures below 1000 °C. Similar to the above Ni-Ti and Ni-Nb systems, metastable equilibrium between the b.c.c. solid solution and an amorphous phase at Nb-50 at.% Al is apparently responsible for the extended solid solubility. Oehring
226
5 Mechanical Milling and Alloying
and Bormann (1990) have subsequently shown that ball milling the A15 structure Nb 3 Al intermetallic transforms it to the Nb (Al) b.c.c. solid solution. Ivanov et al. (1988) obtained a solid solubility of about 27 at. % Al in f.c.c. Ni by MA of elemental Ni and Al powders. The equilibrium solid solubility of Al in Ni at 500 °C is about 4 at.%. Again, the metastable f.c.c. solid solution with extended Al solubility is in metastable equilibrium with an amorphous phase found by Ivanov et al. (1988) over the composition range of 27 to 35 at.% Al. The equilibrium solubility of Fe in Al is nil ( < 3 x 10" 3 at.%) below 500°C (Edgar, 1949). However, Shingu and coworkers (Shingu et al., 1988; Shingu et al., 1989; Huang et al., 1990) have observed an amorphous phase for the composition range 17 to 33 at. % Fe in MA Al-Fe alloys along with an extended solid solution of Fe in Al up to about 10 at. % Fe. Thus, large extended solubility is again associated with the formation of a metastable amorphous phase. Polkin et al., 1990, have been systematically studying extended solid solutions by MA in several systems such as Al-Fe, Ni-Al, Ni-W, and Ni-Cr. Large extensions of solid solubility are seen in these studies. Sundaresan and Froes (1989) have carried out MA on the immiscible system TiMg. The solubility of Mg in Ti reported from studies of the phase equilibrium (Obimata et al., 1959) is <0.2 at.% for temperatures <700°C. However, Sundaresan and Froes (1989) report solid solubilities for MA powder of up to 6.0 at. % Mg. No amorphous phase was reported in MA of this system. Gaffet and Gaspard (1990) have recently reported some mutual solid solubility induced in the immiscible Cu-W system by MA. Preliminary X-ray diffraction studies show lattice parameter
changes that suggest extended solid solubilities along with the possibility of partial amorphization. Confirmation of these results awaits analysis by transmission electron microscopy. 5.6.2 Disordering by Mechanical Milling
It is well known that large strains in plastic deformation (>20%) can destroy long range chemical order in ordered alloys (Stoloff and Davies, 1966). Studies of such disordering are therefore usually limited to ordered alloys and intermediate phases which exhibit extensive ductility such as Cu3Au and Ni 3 Mn. High energy ball-milling extends the possibility for studying deformation-induced disorder to nominally brittle ordered alloys and intermetallic compounds. As in the case of extended solid solubilities, most of the studies of disordering by MM have so far been in conjunction with experiments on amorphization. Amorphization of intermetallic compounds by MM is analogous to amorphization by energetic particle irradiation in that defects introduced by deformation or irradiation must be responsible for raising the free energy of the crystalline compound and allowing the crystalline-toamorphous transformation to occur. Hellstern et al. (1989 a) studied the changes in structural and thermodynamic properties of Ru and the AlRu intermetallic with milling time. Milling the CsClstructure (ordered b.c.c.) AlRu compound induced a decrease in LRO from 5 ^ 1.0 to 5^0.7, which value did not change with further milling time. AlRu did not exhibit amorphization but showed the development of a fine nanocrystalline grain size. It was concluded that in AlRu most of the disordering was a consequence of the crystal refinement as opposed to the density of antiphase boundaries. The crystal size at
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
long milling times saturated at about 7 nm and the lattice strain at 3 % . The specific heat of AlRu increased by almost 15% suggesting changes in the vibrational and configurational part of the entropy due in part to anharmonic atomic vibrations at the grain boundaries and disordered lattice sites. Seki and Johnson (1990) reported the structural changes in ball-milled CuTi and CuTi2 intermetallic compounds. The change in the LRO parameter, S, was small for CuTi2 before amorphization occurred. However, S for CuTi decreased with milling time and saturated at a value of about 0.5. Amorphization did not occur for CuTi. In contrast to the above results, Jang and Koch (1990 a) observed complete disordering by ball-milling of the ordered Ni3Al intermetallic which has the LI 2 (ordered f.c.c.) crystal structure. The change in LRO parameter, S, with milling time in a vibratory mill is given in Fig. 5-20 for Ni3Al powder. A monotonic decrease of S with milling time is observed and *S = 0 after milling times of about 5 h. The microhardness of Ni3Al versus milling time is presented in Fig. 5-21. The microhardness is observed to rise rapidly with milling time and peak at about 0.5 h. At this milling time the structure is heavily deformed and the LRO parameter S has decreased to about 0.5. In previous studies of the influence of LRO parameter on mechanical behavior it was found that the flow stress peaked in the Fe 3 Al ordered intermetallic when the LRO parameter was at values of 5=0.4-0.6 (Stoloff and Davies, 1966). This effect has been rationalized as due to the transition from deformation by superlattice dislocations in ordered structures to deformation by unit dislocations in the disordered lattice. As disordering continues with milling
1 10 Milling time in h
227
100
Figure 5-20. Long-range-order parameter, S, vs. milling time at room temperature for Ni3Al (Jang and Koch, 1990 b).
time, the hardness decreases again to a minimum value for the completely disordered f.c.c. solid solution at 5 h, although the hardness here is still greater than the annealed, fully ordered powder. A small increase in hardness occurs on further milling and then the hardness values ap800
1 10 Milling time in h
100
Figure 5-21. Microhardness vs. milling time at room temperature for Ni3Al (Jang and Koch, 1990 b).
228
5 Mechanical Milling and Alloying
pear to saturate to a constant level thereafter. The disordering and hardness behavior as a function of milling presented above were for stoichiometric Ni 3 Al. Jang and Koch (1990 b) also studied Ni-rich (Ni 76 Al 24 ) and Al-rich (Ni 73 Al 27 ) compositions for the Ni 3 Al intermetallic. The milling time for complete disorder was the same for the Al-rich compound as for stoichiometric Ni 3 Al, but was shorter, i.e. S = 0 at 3 h, for the Ni-rich compound. Complete disorder was obtained at the same milling time, 5 h, for stoichiometric Ni3Al milled at vial temperatures of -10 °C and -50 °C, indicating no dependence of temperature on the disordering mechanism for this range of temperatures. Loeff et al. (1989) have studied disordering by ball milling in the CsCl-structure type compound CoGa. The structural evolution was followed by magnetization measurements along with X-ray diffraction and microscopy. The disorder in CoGa was explained by creation of a "triple defect" in which Co atoms substitute on the Ga sublattice and two vacancies form on the Co sublattice. If the stoichiometric CoGa compound is completely ordered the Co atoms are isolated from one another by Ga atoms, which results in non-magnetic behavior. However, a Co atom on the wrong sublattice is surrounded by 8 Co nearest neighbors and forms with these neighbors a small magnetic cluster. It was predicted that if ball milling induced disorder in CoGa, an increase in magnetization due to Co anti-site atoms should occur. This was indeed observed, with the magnetization increasing with milling time up to a constant saturation value. The maximum value of magnetization corresponds to material with a composition of 53.7 at. % Co, which suggests 3.7% of anti-site Co atoms in the 50%-50% alloy. This is consistent with measurements of the lattice parame-
ter which decreases to a constant value with milling time. The maximum relative decrease, Aao/ao=-0J% is in excellent agreement with the vacancy concentration (0.074) obtained from the magnetization measurements and a model given by Edelin (1979) for changes in lattice parameter due to anti-site disorder. The maximum vacancy concentration induced by ball milling in CoGa (0.074) is equivalent to the equilibrium thermal vacancy concentration at a temperature of about 1100°C. Oehring and Bormann (1990) have measured a decrease in LRO parameter S on ball milling the Al 5 compound Nb 3 Al. Although S only decreases from about 1.0 to 0.8 before the A15 structure transforms to a b.c.c. solid solution, they attribute the transformation to the increase of the free energy of the A15 structure by disordering. The resulting disordering energy estimated for this change in S amounts to 2 kJ mol" 1 , which is a larger value than the stored energy from the 43 nm diameter grain boundaries (0.7 kJ mol" 1 ) or elastic strain energy (0.085 kJ mol" 1 ).
5.6.3 Amorphization by Mechanical Alloying/Milling
The fundamental research using MA/ MM during the last several years has been dominated by the use of this technique for solid state amorphization. This is at present a very active field of research with many new results being reported. It has many similarities to the intense interest in metallic glasses in the 1970's. Some of the conclusions presented in this chapter regarding the rapidly moving field of solid state amorphization by MA/MM may be shown to be incorrect by the time of publication.
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
It is convenient to divide amorphization by ball milling into the categories of 1) mechanical alloying (MA) of elemental (or dissimilar alloy) powders involving material transfer between the components, and 2) mechanical milling (MM) of a single composition (e.g. an intermetallic compound, an immiscible mixture, or an element) such that no material transfer need occur. Some investigators prefer to use "mechanical grinding" (MG) for the latter process. Since "grinding" is normally thought of as an abrasive machining process that involves mainly shear stresses and chip formation, the author prefers the term "milling" to include the more complex triaxial, perhaps partly hydrostatic, stress states that can occur during ball milling of powders. Amorphous metallic phases have been of interest for many years and for the most part have been produced by rapid solidification or vapor deposition on cold substrates. Amorphization from the solid state gained renewed attention in 1983 with the discoveries of amorphization by hydrogenation of Zr 3 Rh by Yeh et al. (1983), amorphization by diffusion between thin film sandwiches of crystalline La and Au by Schwarz and Johnson (1983), and amorphization by mechanical alloying of crystalline Ni and Nb powders by Koch et al. (1983). Amorphization from the solid state by irradiation with energetic particles was first observed in 1962 by Bloch who amorphized U 6 Fe by fission fragments. Interest in irradiation amorphization has also increased in recent years, and the amorphization of intermetallic compounds by high voltage electrons has been reviewed by Luzzi and Meshii (1987). Johnson (1986) has reviewed the fundamental thermodynamics and kinetic factors of amorphization in metallic systems with emphasis placed on
229
the mechanisms for solid-state amorphization. The first examples of amorphization of intermetallic compounds by mechanical milling were presented by Ermakov et al. (1981, 1982) in the Y-Co and Gd-Co systems. Intermetallics such as YCo 3 , Y 2 Co 7 , YCo 5 and Y 2 Co 17 were milled and found to exhibit the broad diffraction peak and Mossbauer spectra of an amorphous alloy. Milling of GdCo 3 and Gd 2 Co 7 resulted in a two phase (amorphous + crystalline) structure. Subsequently a number of intermetallic compounds have been amorphized by milling and are reported in the review of Weeber and Bakker (1988). The first suggestion that MA might produce amorphous material was made by White (1979) during a study of the synthesis of superconducting Nb 3 Sn via MA and subsequent thermomechanical treatment. The first definitive study of amorphization by MA was carried out by Koch et al. (1983) in the easy-glass-forming alloy system Ni-Nb. Since then, a number of alloys have been amorphized by MA, and the details for the various systems are given in the review of Weeber and Bakker up to 1988. A recently investigated system in which MA has induced amorphization is 80 Al + 20Fe (Wang et al., 1991). Although the detailed mechanisms for amorphization by either MM or MA are not yet welldefined, working hypotheses have been developed from which to predict experimental results. First MA will be considered. Schwarz and Johnson (1983) reported the formation of amorphous metallic alloys by what they term a "solid state amorphization reaction" (SSAR) between films of two crystalline pure metals, A and B, that have a large negative heat of mixing in the amorphous state. They also showed, for thin film sandwiches of crystalline La and Au, that the products of the SSAR are
230
5 Mechanical Milling and Alloying
those predicted by a calculated free-energy diagram at the estimated reaction temperature. The free energy diagram for a hypothetical system A-B with a large negative heat of mixing in the amorphous alloy is schematically illustrated in Fig. 5-22. The driving "force" (free energy difference) for the crystalline-to-amorphous transition at composition AmBn is then the difference in free energy between state (1), the mixture of crystalline components A and B, and state (2), the amorphous alloy. The equilibrium stable phase(s) will, of course, always have a lower free energy than the amorphous alloy (e.g. state (3) for the intermetallic AmBn in Fig. 5-22). A kinetic constraint must therefore prevent the formation of the equilibrium phase(s) in competition with the amorphous alloy. Schwarz and Johnson (1983) suggested that such a constraint involves a large asymmetry between the diffusivities of the components. It was assumed that if only one species was mobile in the other, and in the amorphous phase, amorphization
Crystalline intermetallic Mixture of components A and B
Amorphous alloy
AmBn
B
Composition (% B) Figure 5-22. Schematic free energy vs. composition diagram for a binary system, AB, with a negative heat of mixing in the amorphous state.
would be kinetically favored over the formation of new crystalline phase(s) with unit cells different from either A or B that might require the cooperative motion of both atomic species. The criteria proposed for a SSAR were therefore: 1) a large negative heat of mixing for the amorphous alloys, and 2) a kinetic constraint on the formation of the equilibrium phases, which may involve a large asymmetry in the component diffusivities. These criteria seem to be generally obeyed for the systems studied. Details of the experimental results for SSAR are given in the review by Johnson (1986). It was reasonable to extend the model for SSAR to amorphization by MA. Schwarz et al. (1985) and Hellstern and Schultz (1986) have argued that the MA of elemental powders leads to an ultrafme composite in which a SSAR takes place during milling. Schultz (1984) and Atzmon et al. (1984) had previously produced bulk amorphous material from metallic composites which were prepared by mechanical codeformation followed by annealing. Thus, it could be assumed that during MA the refinement of the powder composite and possible enhancement of diffusion by the defects created by the severe plastic deformation allowed the SSAR to occur without additional thermal treatment. Schwarz and Petrich (1988) analyzed the progress of amorphization by MA of Ni and Ti powders at an average composition Ni 33 Ti 67 by carrying out differential scanning calorimetry measurements after different milling times. In the early stages of MA it was concluded that the rate of amorphization was controlled by a SSAR at the clean Ni/Ti interfaces created by milling. The DSC experiments showed an exothermic enthalpy corresponding to the formation of the amorphous phase. This enthalpy disappeared at longer milling
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
times, but well before the samples were completely amorphized. It was postulated that at this time the rate of amorphization becomes governed by the rate of mechanical attrition. Petzoldt (1988) used transmission electron microscopy to follow the amorphization of Ni-Nb alloys by MA. As a function of milling time he first observed crystalline lamellae, then crystalline lamellae surrounded by amorphous regions, then an amorphous matrix containing small crystalline particles, and finally a completely amorphous structure. This morphology suggests the possibility for a SSAR as the major mechanism for amorphization by MA. Dolgin et al. (1986) had observed a growth of the amorphous phase during MA of Ni 5O Ti 5O , Co 50 Ti 50 , and Fe5OTi5O to obey a square root of time law. This growth rate was determined by measuring a "peak ratio" of crystalline X-ray lines, defined as the amplitude of the line at a given milling time divided by the amplitude of the line from the unmilled powder. A linear dependence of this peak ratio was observed when plotted against the square root of milling time. This of course implies a diffusion controlled mechanism. This same t1/2 dependence is observed for aging times in SSAR in thin film diffusion couples (Johnson, 1988), after a brief initial transient, interface-limited regime. Finally, many of the same alloy systems which undergo SSAR also are found to amorphize by MA. On the other hand, the defects created by the severe plastic deformation induced by milling would be expected to influence both the thermodynamics and kinetics of the crystalline-to-amorphous transformation by MA. Thus a one-to-one comparison between SSAR of diffusion couples and amorphization by MA might not be gener-
231
al. In fact, it has been shown that several alloys can be amorphized by MA but not by SSAR of thin film diffusion couples. These included Nb5OAl5O (Hellstern et al., 1988) and Nb 7 5 Ge 2 5 (Koch and Kim, 1985; Bormann, 1987). In the case of Nb 7 5 Ge 2 5 the mechanism for amorphization can be classified with amorphization of intermetallic compounds by MM, since for Nb and Ge powders MA first forms the A15 structure Nb 3 Ge compound which subsequently transforms to the amorphous alloy with continued milling. A clear example of this behavior was observed in the Nb-Sn system (Kim and Koch, 1987). Elemental Nb and Sn powders were milled together at the composition Nb 75 Sn 25 . As a function of milling time the X-ray diffraction patterns change from a mixture of Nb and Sn lines to lines of the crystalline Al 5 structure and then finally to the pattern for an amorphous structure. The reaction An + Bm->Cryst. AnBm->amorph. AnBm with milling time is in effect a variation of first producing the equilibrium intermetallic compound which can then be amorphized. In order for amorphization to occur by MM of an equilibrium intermetallic compound, the free energy of the crystalline compound must be raised from state (3) to state (2) as illustrated in Fig. 5-22. As Johnson points out (Johnson, 1986), in this case no chemical energy is available in the initial crystalline compound to drive the amorphization reaction. Defects introduced by the deformation during milling must be responsible for raising the free energy of the crystalline compound. The critical question is; what defects are controlling a given reaction? Energy can be stored in a deformed material in the form of defects such as vacancies, interstitials, dislocations, grain boundaries, and atomic disorder. The energy stored during severe
232
5 Mechanical Milling and Alloying
cold working of metals or alloys in the form of dislocations is typically less than 1-2 kJ mol" 1 (e.g. Bever et al., 1973). A maximum value for the energy associated with a high dislocation density of ^ 1 0 1 4 cm ~ 2 in cold-rolled NiTi was estimated to be 2.2 kJ mol" 1 by Koike et al. (1990). Partial amorphization was claimed and ascribed to the dislocation energy which was comparable to (about 60 % of) the crystallization energy of amorphous Ni-Ti alloys. However, Hellstern et al. (1989 a) did not observe amorphization of NiTi by ball milling even though a larger, 5 kJ mol" 1 , stored energy was measured from the exothermic heat released in DSC experiments. Koike et al. (1990) rationalize this discrepancy by suggesting the temperature of the ball milled powders was higher than in cold rolling and therefore a higher probability might exist for dislocation accumulation during cold-rolling. In general then, dislocations have not been observed to be a significant contribution to the defect energy that promotes amorphization by ball milling. Deformation can induce chemical disorder by the creation of atomic site defects and/or antiphase boundaries. It has been demonstrated experimentally that atomic disordering of a crystal may precede the transition to the amorphous state in an irradiation experiment. Okamoto et al. (1988) observed a decrease of the longrange order (LRO) parameter S to values of ^0.3 prior to amorphization of Zr3Al with charged-particle beams. A decrease in the average shear elastic constant accompanied the loss of LRO in the compounds Zr3Al, Zr 2 Al, and Nb 3 lr. It was suggested, Okamoto et al. (1988), that a strong coupling between strain and order parameter may be responsible for the observed elastic softening and that strain accumulation is an important prerequisite for the amor-
phization. Linker (1986) had previously suggested that strain release was responsible for amorphization of Nb thin films implanted with B ions. His ideas of strain-induced amorphization were based on a model of Egami and Waseda (1984) which considers local atomic level stresses and the importance of the atomic size factor. Egami and Waseda (1984) argue that a crystalline solid solution becomes topologically unstable when the concentration of smaller atoms (A atoms) reaches a critical concentration C* which depends on the ratio of atomic sizes r = RA/RB, where RA and RB are the atomic radii of the two atoms. The Egami and Waseda criterion for instability is based on the existence of a critical level of strain disorder arising from atomic size mismatch. Luzzi and Meshii (1988) have reviewed experimental evidence that the energy increase due to chemical disordering is the major driving force behind the electron-irradiation induced amorphization in intermetallic compounds. The magnitudes of disordering energies are comparable to the energy differences between crystalline and amorphous phases in many alloy systems. It is likely that disordering can also provide the driving force for amorphization of intermetallics by MM in many cases. Seki and Johnson (1990) have suggested that the amorphization of the CuTi2 intermetallic by MM was driven mainly by energy storage in antiphase boundaries. Contributions from dislocations and grain boundaries were estimated to be 1 kJ mol" 1 and 1.6 to 2.9 kJ mol" 1 respectively. The remainder of the stored energy, assumed equal to the enthalpy of crystallization = 11 kJ mol~1, is believed to come from the disordering at antiphase boundaries. While disordering is a likely source of defect energy to induce amorphization by MM in ordered intermetallic compounds,
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
there is recent evidence of amorphization in elements (Gaffet and Harmelin, 1990) and disordered alloys (Jang and Koch, 1990 b) by MM which must result from another type of defect. Gaffet and Harmelin (1990) have recently reported on the partial amorphization of elemental Si by ball-milling. This is the first instance of amorphization by MM of a pure element. The total contamination in Fe and Cr from the steel milling media was less than 0.2 at. %. No oxygen analyses were reported. The existence of an amorphous phase was inferred from analysis of the X-ray diffraction patterns, selected area electron diffraction, and the heating responses in a differential scanning calorimeter. The X-ray diffracted intensity was fit by a computer program which modelled the spectrum by a polynomial background plus a set of peaks which used either a Gaussian, modified Lorentz, or Cauchy curve. The data were fit by assuming either an amorphous phase plus one size of crystalline grains, or an amorphous phase plus two sizes of crystalline particles. The analysis of these data suggest that an amorphous phase contributes to the intensity of the (111) peak (1st amorphous peak) and produces a broad peak between the (220) and (311) peaks (2nd amorphous peak). This observation of an amorphous diffraction pattern is clearer for the selected area electron diffraction patterns. A diffuse halo is seen around the (111) ring and a second diffuse halo is observed between the (220) and (311) rings. The DSC results give three contributions on heating (to 1000 K) which are interpreted as 1) an exothermic release of strain energy, 2) an endothermic effect associated with the amorphous phase, and 3) an exothermic crystallization peak. A slight expansion (0.1 to 0.2 %) of the lattice parameter of Si is also noted. The authors
233
suggest that the crystalline-to-amorphous transition occurs during a continuous refinement of the crystalline size during milling and the associated expansion of the Si lattice parameter. A dynamic equilibrium between the amorphous and microcrystalline state is assumed. A crystalline-to-amorphous transition in Si with decreasing grain size was previously noted by Veprek et al. (1982) in chemically vapor deposited Si films. An abrupt transition from microcrystalline to amorphous Si was observed when the grain size was reduced to about 3 nm. A lattice expansion was observed with decreasing grain size, starting at about 8 nm grain size, and increasing to « 1 % at 3 nm grain size when the crystalline-to-amorphous transformation occurred. The authors present a thermodynamic argument for the increase in free energy of crystalline Si as the grain size decreases. The driving force for the crystalline-to-amorphous transition is the increase of the elastic energy due to the lattice expansion which is in turn related to the increase of the specific surface energy of small crystallites. The ball-milled Si exhibited lattice expansions of only about 0.1 to 0.2%, but with a calculated grain size of about 8 nm this is consistent with the above results for Aao/ao vs. grain size. However, a partially amorphous structure was observed for the ball-milled Si which according to the average grain size should not have occurred. This implies an inhomogeneous distribution of crystallite sizes such that in some regions - those which transformed to the amorphous structure - a much finer crystallite size was reached than the average size determined by X-ray diffraction. More detailed TEM studies are needed to clarify the interesting results on ball-milled Si. The results of ball-milling the ordered intermetallic compound, Ni3Al, exhibit
234
5 Mechanical Milling and Alloying
similarities of the results for MM of Si discussed above. Jang and Koch (1990 b) studied the structural evolution of the Ni3Al ordered intermetallic with milling time. As described in Sec. 5.6.2, the LRO parameter 5* exhibits a monotonic decrease with milling time, reaching S = 0 after 5 hours (see Fig. 5-20). Transmission electron microscopy was used to study the microstructure at the longer milling times where X-ray diffraction indicates broad lines but still crystalline patterns. If the Scherrer formula is applied to the breadth of the major X-ray diffraction peak, the effective scattering length decreases with milling time and saturates at a value of about 8 nm. At milling times > 5 hours a fine microcrystalline microstructure is observed as illustrated in Fig. 5-23 for a sample milled 23 hours. The selected area electron diffraction pattern is clearly crystalline, and the microcrystalline grain size appears to be about 2 to 4 nm in diameter. This microcrystalline structure is also observed for longer milling times (i.e. 50 h) over parts of the sample, but in some areas an image and selected area electron diffraction pattern consistent with the amorphous structure is observed, as illustrated in Fig. 5-24. The structural development during milling of the f.c.c. disordered
Ni3Al solid solution was dominated by the formation and refinement of a dislocation cell structure into microcrystallites which eventually reached nanometer dimensions. It is assumed that the grain boundaries can provide the energy required to induce the crystalline-to-amorphous transformation when their dimensions reach < 2 nm diameter in certain regions. A comparison of the calculated interfacial free energy of the grain boundaries with the estimated difference in free energy between the crystalline f.c.c. and amorphous phases for Ni3Al was consistent with this assumption. The similarity between the results for ball-milled Si and Ni3Al can be summarized as follows: 1. Both systems reach a two-phase metastable equilibrium between amorphous and microcrystalline structures. 2. The major observable defect is the fine microcrystalline grain structure, i.e., grain boundaries. 3. It appears that when the microcrystalline grain size falls below some critical value the increase in grain boundary area, and therefore energy, can act as the driving force for the crystalline-toamorphous transformation. While new "amorphous" materials are typically revealed by diffraction experi-
I1I& 50 nm Figure 5-23. TEM micrograph; Ni3Al powder milled 23 h. Bright field (Jang and Koch, 1990 b).
Figure 5-24. TEM micrograph; Ni3Al powder milled 50 h. Bright field (Jang and Koch, 1990 b).
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
ments, there is often ambiguity in distinguishing between a "truly" amorphous i.e. continuous, non periodic - structure and a microcrystalline structure. In a diffraction experiment of such an "amorphous" structure, the intensity but not the phase of the scattered radiation is measured. Fourier inversion of these data can yield only the radial distribution function of the structure which cannot uniquely specify the atomic positions. To determine the structure, the experimentally determined radial distribution function must be compared with radial distribution functions calculated from the structural models being considered (Cargill, 1975) (e.g. microcrystalline vs. continuous dense random packing) (see Vol. 9, Chap. 4, Sec. 4.2.1.1). While the diffraction patterns of most amorphous metallic alloys can be better fit by continuous dense random packing models (Cargill, 1975) there are reports of microcrystalline structures (e.g. Wagner et al., 1968). The problem of whether a new disordered material is "only" microcrystalline or "truly" amorphous is particularly acute for alloys synthesized by ball milling since the evolution of the disordered structure often involves the broadening of diffraction lines with milling times (Koch et al., 1983; Politis, 1985). This implies a fragmentation of crystallites with milling time that might lead to a microcrystalline structure. However, Lee et al. (1988 b) have recently demonstrated that, within experimental error, the short-range atomic distribution was the same in a Cu 50 Zr 50 amorphous alloy prepared by rapid solidification, proton irradiation, and mechanical alloying. This conclusion resulted from the similarities of the reduced interference functions and reduced atomic distribution functions of the amorphous alloys so prepared as determined by X-ray diffraction.
235
In addition, extended X-ray absorption fine structure (EXAFS) spectroscopy has clearly indicated that the amorphous NiNb alloys prepared by MA are amorphous on the atomic level (Nasu et al., 1990). The appearance of a glass transition temperature, r g , as determined by calorimetry, is strong evidence for a liquid-like and therefore "truly" amorphous structure. Schwarz et al. (1985) reported a glass transition temperature in Ni 32 Ti 68 amorphized by MA. Similarly, Weeber and co-workers (Weeber et al., 1987; Weeber and Bakker, 1988) have reported Tg's in N i ^ Z r , alloys prepared by MA. Jang and Koch (1989) have observed 7^'s in amorphous Cux -^Zr^ alloys prepared either by MA of elemental Cu and Zr powder or by MM of CuZr and Cu 3 Zr 2 intermetallic compound powder. Differential scanning calorimetry clearly show the endothermic Tg's prior to the occurrence of the much larger exothermic crystallization peaks. It was found that if the iron and oxygen impurities which can occur in amorphous MA or MM Cu 1 _ x Zr x alloys are accounted for, essentially identical Tg values to those of amorphous rapidly solidified material are observed. In addition to the above evidence for the nature of amorphous alloys prepared by MA/MM, physical properties of amorphous alloys prepared by MA have been found to agree with those of the same alloys made by rapid solidification (e.g. Schultz and Hellstern, 1987; Hikata et al., 1988). This evidence indicates that amorphous alloys prepared by MA or MM can be "truly" amorphous as opposed to microcrystalline - that is, they can be indistinguishable from "metallic glasses". It has been convenient to classify the processes of amorphization by ball milling into: 1) "MA" - material transfer and
236
5 Mechanical Milling and Alloying
amorphization driven by a decrease in free energy as in a SSAR, and 2) "MM" - increase in free energy of a single phase crystalline material (element or compound) by the defects created during deformation. The "MA" process must dominate for amorphization of mixtures of elemental powders with negative heats of mixing even though "MM" effects are always present. For elements or compounds, "MM" is the only operative mechanism. There may be instances where "MM" may be more important than "MA" in mixtures of elemental powders with a positive heat of mixing. Recent work by Fukunaga et al. (1990) suggests that partial amorphization may occur on ball milling Cu-Ta powders even though this is an immiscible system with a positive heat of mixing. For complete amorphization, iron contamination was required (4 to 5 at.% Fe). However milling with Cu-Be vial and balls still resulted in partial amorphization. The authors conclude that the "MM" mechanism must be dominant in this case. While there is considerable experimental evidence to support the broad ideas of the mechanisms involved in amorphization by MA and/or MM, there is still much uncertainty in the details of such mechanisms. Amorphization by the variety of solid state reactions has been reviewed recently by Gerl and Guilmin (1990). They include amorphization of MA/MM in their review and discuss some features common to the various amorphization reactions. They point out that while the physical state of a material can be followed during amorphization by different means, the physical quantity observed, F (where F might be resistivity, volume change, enthalpy etc.) seems to vary with the perturbation of the material, 5, as shown schematically in Fig. 5-25. A relatively moderate perturbation (milling time for MA/MM), <51? is suf-
Figure 5-25. Variation of the physical quantity F (Q, AV/V, AH, ...) as a function of the perturbation (milling time, d.p.a. for irradiation processes, LRO parameter, ...).
ficient to approach a saturation value of F for which amorphization begins. It is not until the value d2 of the perturbation is reached that amorphization (or equilibrium between amorphous and crystalline phases) is completed. This implies a gradual transformation from the crystalline to the amorphous state and a first order nucleation and growth phenomenon reminiscent of the melting transition. In fact Cahn and Johnson (1986) have discussed the features common to the nucleation of "disorder" in several transitions including melting, order-disorder, and solid state amorphization (including MA/MM). While it is assumed that amorphization by MA/MM is usually a nucleation and growth phenomenon, Johnson (1986) has proposed an intriguing theory for destabilization of crystalline solids which may in certain cases occur by a homogeneous transformation that he has termed a "shear strain spinodal". As was previously the case for rapid solidification studies, the complexity of the ball milling reactions and the number of experimental variables makes comparison of different investigator's results difficult. The construction of special mills with the possibility for controlling the kinetic energy and temperature will be of great help in this regard (e.g. Gaffet, 1990). Better con-
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
trol of the experimental variables along with new theoretical approaches (Martin and Gaffet, 1990; Beke et al., 1990) should result in better understanding of the complex and fascinating phenomenon of amorphization by MA/MM. 5.6.4 Crystallization of Amorphous Alloys by Mechanical Milling
While many crystalline materials have been amorphized, there are relatively few reports of the crystallization of amorphous alloys by ball milling. Eckert et al. (1988) reported that if the milling intensity i.e. kinetic energy was too high, Zr-Ni alloys which were amorphized at a lower milling intensity would exhibit crystalline structures. It was assumed that the increased kinetic energy of the milling media raised the temperature of the powders so much that in-situ crystallization and/or direct formation of crystalline compounds could occur - that is, the kinetic constraint to the formation of the equilibrium phases is removed at sufficiently high temperatures. A similar result was reported by Matsuki et al. (1988) for MA of Nb and Sn powders in a rotary ball mill. At a rotation velocity of 180 rpm an amorphous structure was observed for 24 h of milling at the average Nb 3 Sn composition. However, at the higher rotational velocity of 250 rpm the crystalline A15 Nb 3 Sn equilibrium phase was formed. It was assumed that the temperature of the powders was increased at the higher rotational speed. Kimura and Kimura (1990) measured the temperature of their attritor mill during a study of the amorphization of NiTi by MM. At a rotational speed of 450 rpm the average temperature of the attritor rose to a constant value of about 180°C. At this rotational speed and temperature a crystalline product was observed in contrast to the amor-
237
phous structure obtained by milling at lower rotational speeds and temperatures (e.g. 300 rpm, 100 °C). While 180°C is several hundred degrees below the crystallization temperature for NiTi, the powder surfaces may be this amount hotter than the measured macroscopic temperature of the attritor. Mizutani and Lee (1990) milled elementary Ni and Zr powder in a planetary ballmill. The equivalent vial rotation of 710 rpm resulted in a temperature on the surface of the vial of <150°C. Powders were removed from the vial at 5 h intervals for analyses. Amorphization was observed over the range of compositions of 20 to 80 at. % Zr. Alloys of composition Ni 30 Zr 70 , Ni 5O Zr 5O , and Ni 70 Zr 30 were milled for times beyond those for complete amorphization. Crystallization was observed after long milling times and was attributed to contamination by the milling media and/ or atmosphere. It was believed that oxygen contamination was responsible for the crystallization in the Zr-rich Ni 30 Zr 70 alloy (in which the milling time for crystallization was shortest) while accumulation of Fe and Cr impurities beyond a critical concentration pushes the Zr concentration out of the glass-forming range and leads to crystallization for the Ni-rich Ni 70 Zr 30 amorphous powders. Crystallization on milling the Ni 30 Zr 70 alloy did not occur if milling was not stopped and the vial opened for sampling. This suggests contamination, not temperature, is the factor controlling the crystallization. Petzoldt et al. (1989) observed an interesting sequence of structures on ball milling mixtures of Ni-Zr intermetallic compounds of average composition Ni 61 Zr 39 or Ni 78 Zr 22 . The structures observed as a function of milling time were: mixture of crystalline intermetallics -• amorphous phase -• amorphous + metastable crys-
238
5 Mechanical Milling and Alloying
talline phase -> amorphous phase. This sequence was observed for the Ni 78 Zr 22 alloy while the Ni 61 Zr 39 alloy exhibited another step in the "reversible" sequence of amorphous phase <-> amorphous + metastable to a final amorphous phase structure. The metastable crystalline phase was not identified. An enrichment of Zr in the amorphous phase was observed by a shift of the amorphous X-ray diffraction peak and the magnitude of the crystallization temperature measured by DSC. The metastable crystalline phase is therefore assumed to be Ni-rich. The "reversible" amorphous <-» metastable transformations were discussed in terms of the free energy curves of the phases. No contamination levels were reported. Trudeau et al. (1990) have recently studied the structural changes of amorphous, melt-spun Fe-base metallic glasses induced by ball milling. They milled the amorphous metallic glass compositions FeggCo^SiiB^ (Metglas 2605 Co) and Fe 78 Si 9 B 13 (Metglas 2605 S-2). Both alloys crystallized on ball milling but the Cocontaining alloy was found to crystallize after shorter milling times. Since the Cocontaining alloy has a lower crystallization temperature (441 °C) than the 2605 S alloy (553 °C) it could be thought that ball milling induced local heating in the powders so that the crystallization temperature was reached, and that the alloy with the lower crystallization temperature was therefore more unstable to milling. However, they then added either Co or Ni powders to the 2605 S alloy and milled the mixture of powders to produce alloying of these elements into the amorphous alloy. Co additions induced a rapid and complete crystallization of the amorphous alloy whereas Ni additions stabilized the amorphous structure against crystallization for the longest milling times used (24 h). This
is significant because the alloy with the Ni additions has a lower crystallization temperature - 375 °C - than either of the original two alloys which both crystallized fully on milling. This result suggests that reaching the crystallization temperature via heat generated in the milling process is an unlikely explanation for the crystallization observed. Gaffet (1990) has varied the vial temperature in his well-controlled experimental planetary mill. He found that changing the vial temperature from room temperature to 200 °C did not influence the product of milling Ni 10 Zr 90 powder. However, variations of the mechanical milling parameters, i.e. the rotation speeds of the disk and of the vial holders, do determine the final powder structures. In the case of milling an amorphous structure the thermodynamics always favors more stable crystalline phases and the equilibrium crystalline structure. The transformation of the amorphous -> crystalline phase by ball milling means that the kinetics of the transformation must be influenced by ball-milling. Introduction of impurities - i.e. chemistry changes - and local temperature increases are possibilities for enhancement of the crystallization kinetics. However, these effects do not appear in all the experiments cited above, leaving the amorphous -• crystalline transformation not well understood. Perhaps milling introduces "defects", e.g. increased free volume, in the amorphous alloys which could accelerate diffusion and therefore crystallization. 5.6.5 Demixing Reactions by Mechanical Milling If the free energy increase due to the introduction of defects by ball milling can cause amorphization of equilibrium crys-
5.6 Mechanical Milling/Alloying as a Nonequilibrium Processing Tool
talline phases it may be possible to induce other metastable structures by milling. Loeff et al. (1989) have ball-milled powders of composition La 4 3 Ni 5 7 , La 3 4 Ni 6 6 , La 3 8 Co 6 2 , La 4 0 Co 6 0 , La 50 Ag 50 and La 45 Ag 55 . The starting powders consisted of the equilibrium phases, for example LaNi + La 2 Ni 3 intermetallics for La 4 3 Ni 5 7 . After milling for 100 to 200 h all these alloys showed the same type of behavior which was demixing into a mixture of the elemental components. That is, for example LaNi + La2Ni3->|3-La + Ni. The measured lattice parameter for p-La was identical (to within experimental error) to the literature value suggesting little solid solubility between Ni and La, as indicated in the equilibrium diagram. The same result was seen for the other alloys with milling resulting in elemental (3-La and Ni, Co, or Ag respectively. The particle size of the elemental powder mixtures was estimated from X-ray diffraction line broadening to be 10-20 nm. The authors assume that about 50 kJ mol~ x of enthalpy can be stored in these alloys (based on amorphization of Pd 49 Zr 51 ) and this enthalpy can raise the enthalpy of the crystalline equilibrium phases to above the value (enthalpy of mixing = 0 for demixing) needed for the metastable demixing reaction. More studies are needed to confirm these ideas and to further explore the generality of these fascinating observations. 5.6.6 Nanocrystalline Materials by Mechanical Alloying/Milling Nanometer-sized crystalline materials referred to as nanocrystalline materials were first studied by Gleiter and coworkers (Gleiter and Marquardt, 1984; Birringer et al., 1984; Gleiter, 1989). These materials are polycrystals with very small crystallite sizes of about 2 to 10 nm diameter.
239
They contain, therefore a very large grain boundary area and are apparently composed of randomly oriented high angle boundaries. The first nanocrystalline materials were produced (e.g. Birringer et al., 1984) by evaporation of the material in a high purity noble gas atmosphere followed by condensation and then compaction in an ultra-high vacuum. The structure and properties of this unique class of materials have been described (Schaefer et al., 1988; Karch et al., 1987; Horwath et al., 1987). It has been shown recently that nanocrystalline materials can also be synthesized by high energy ball milling of elemental powders (Hellstern et al., 1989a; Fecht et al., 1990; Gaffet and Harmelin, 1990), intermetallic compound powders (Hellstern et al., 1989 a; Hellstern et al., 1989b; Jang and Koch, 1990 b), or powders of immiscible alloy systems (Schlump and Grewe, 1989; Shingu et al., 1988; Koch et al., 1989). Fecht et al. (1990) developed nanocrystalline structures in various elemental b.c.c. and h.c.p. metal powders by ball milling. Transmission electron microscopy revealed that a random nanocrystalline grain structure evolved from dislocation cell structures within shear bands with increasing milling time. The deformation texture, as determined by electron diffraction, of the structure containing dislocation cells/low angle grain boundaries disappeared with longer milling times and was replaced by a nearly random texture consisting of nanocrystals with high angle grain boundaries. Jang and Koch (1990c) have shown that the hardness increases with decreasing grain size for nanocrystalline Fe produced by ball milling. A HallPetch relationship for hardness vs. (grain diameter) ~ 1/2 was observed. Nanocrystalline materials can be synthesized by ball milling in a wide variety of materials. The stability and properties of
240
5 Mechanical Milling and Alloying
such structures have not been studied systematically. Ball milling should be a practical method of producing large quantities of nanocrystalline materials. 5.6.7 Quasicrystalline Materials by Mechanical Alloying The discovery in 1984 by Shechtman and coworkers (Shechtman et al., 1984) of an Al-Mn phase that exhibited a sharp electron diffraction pattern with icosahedral point symmetry has attracted considerable interest in the materials community in such "quasicrystals". Although a few stable quasicrystals have been discovered, most of the known quasicrystals are thermodynamically metastable. They are usually produced by nonequilibrium methods such as rapid solidification (Shechtman et al., 1984), ion beam mixing (Follstaedt and Knapp, 1988) or transformation of the amorphous phase (Poon et al., 1985). Ivanov et al. (1989) and Eckert et al. (1989) synthesized icosahedral phases in Mg-Zn-Al, Mg-Cu-Al and Al-Cu-Mn alloys, respectively, by MA. Ivanov et al. (1989) found the icosahedral phases in compositions of Mg3Zn5_xAlJC, where x = 2 to 4, and in Mg 32 Cu 8 Al 41 . The structures were the same as in icosahedral phases obtained by rapid solidification. Ivanov et al. (1989) attribute the formation of the icosahedral phases during MA to the disclinations introduced by the deformation. Eckert et al. (1989) observed the icosahedral phase in Al 65 Cu 20 Mn 15 powder after MA of the elemental components. They believe the formation of the icosahedral phase during MA is due to an interdiffusion reaction between the components, as has been suggested for amorphization by MA of elemental component powders. Both groups suggested that the advantage of synthesizing quasicrystalline
materials by MA is the possibility of producing large quantities of powder which might be compacted to obtain bulk material for physical property studies or potential technical applications. Ivanov et al. (1989) in fact did compact icosahedral powder by shock wave compression into bulk pellets of 12 mm diameter and 10 mm height. The icosahedral phase remained in the compacted pellet.
5.7 Chemical Reactions Induced by Mechanical Alloying There are only a few reports in the literature regarding chemical reactions driven by MA of the starting constituents. Schaffer and McCormick (1989 a, b) and McCormick et al. (1990) have studied the reduction of metal oxides with reactive metals using MA. For example, CuO and Ca powders were milled in a SPEX shaker mill under toluene. As measured by the temperature of the vial the reaction began slowly, then increased steadily after about 600 s and was complete after about 6000 s. The progress of the reaction was followed by X-ray diffraction. The ratio of the principal CuO/Cu diffraction peak areas approach zero for milling times > 6000 s, in agreement with the temperature measurements. The reaction CuO + Ca->Cu + CaO: AH=-413 kJmol" 1 was thus induced by MA. When the CuO + Ca powders were milled without toluene, i.e., "dry", the vial temperature increased slowly at first but after about 360 s the vial temperature increased abruptly by 140 K followed by a relatively slow decrease. The temperature rise was virtually instantaneous indicating the occurrence of a combustion reaction. The self-combustion of the reactants during MA is presumably
5.10 References
due to the large negative enthalpy of reao toin(-473kJmol" 1 ). Schaffer and McCormick (1989) calculated the adiabatic temperature rise using the mean heat capacity of the reactants and their enthalpies of melting and obtained a value of about 4000 K. This is much higher than the temperature, ^2300 K, needed for a self-sustaining combustion reaction. A similar combustion reaction was noted for a number of reactions such as: CuO + Mg ^ Fe 2 O 3 + 3Ca 2V 2 O 5 + 5Ti
2Fe + 3CaO, 2
It was also found that simultaneous milling of CuO, ZnO, and Ca led to reduction of the oxides, and for equal proportions of the starting oxides, to formation of p'-CuZn. These reduction reactions may find industrial applications for production of metal and alloy powders. A novel way of producing the intermetallic compound, TiAl, by MA was described by Suryanarayana et al. (1990). They milled the compounds Al3Ti and TiH 2 so that the reaction: Al3Ti + 2TiH 2 ^3TiAl + 2H2T occurred. The reaction did not go to completion by milling (55 vol. % TiAl after 52 h) but went to 95 vol. % TiAl after consolidation by hot isostatic pressing.
5.8 Summary High-energy ball milling is a processing method which is presently attracting much interest. Commercialization of ODS alloys prepared by MA has occurred in recent years. The "academic" interest in ball milling as a nonequilibrium processing tool has exploded in recent years, especially with regard to solid state amorphization
241
by MA or MM. Other metastable structures such as quasicrystals, nanocrystalline materials, and disordered intermetallics have also been synthesized by ball milling. An understanding of the mechanics and physics of the MA/MM processes is still at an early stage of development. The process variables associated with the variety of milling equipment being used complicates comparisons of results from different laboratories. Experimentally designed mills with good control of the milling energies, temperature, atmospheres, etc. which have recently been built should help elucidate the physics of the milling processes. For most potential applications, the powder product of milling needs to be consolidated into bulk form. Only preliminary experiments have been done so far in compaction of the powder products of MA/ MM with metastable structures. The maintenance of desired metastable structures after compaction, and studies of the mechanical and physical properties of such bulk material remain fruitful areas for future research.
5.9 Acknowledgements The author wishes to thank all those researchers who have provided him with preprints of their recent work. The author's research on mechanical alloying/ milling has been supported by the U.S. Office of Naval Research (contract N-0001484-K-0253), and the U.S. National Science Foundation (contracts DMR-8318561 and DMR-8620394).
5.10 References Anders, F. J., Jr. U.S. Patent No. 3159 908, issued 8. Dec. 1964. Anders, F. J., Alexander, G. B., Wartell, W S. (1962), Metal Progress 82, 88.
242
5 Mechanical Milling and Alloying
Arnhold, V., Hummert, K. (1989), in: New Materials by Mechanical Alloying Techniques: Arzt, E., Schultz, L. (Eds.). Oberursel: DGM Informationsgesellschaft; p. 263. Arzt, E. (1989), in: New Materials by Mechanical Alloying Techniques: Arzt, E., Schultz, L. (Eds.). Oberursel: DGM Informationsgesellschaft; p. 185. Arzt, E., Schultz, L. (Eds.), (1989), New Materials by Mechanical Alloying Techniques: Oberursel: DGM Informationsgesellschaft. Atzmon, M. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, J. J. (Eds.). Warrendale (PA): TMS-AIME, p. 173; Phys. Rev. Lett. 64, 487. Atzmon, M., Verhoeven, J. D., Gibson, J. D., Johnson, W. L. (1984), in: Rapidly Quenched Metals: Steeb, S., Warlimont, H. (Eds.). Amsterdam: North-Holland, p. 1561. Beke, D. L., Bakker, H., Loeff, P. I. (1990), /. Phys. France (Colloques)
51, C 4—63.
Beke, D. L., Loeff, P. I., Bakker, H., Acta Metall. et Mater., to be published. Benjamin, J. S. (1970), Metall. Trans. 1, 2943. Benjamin, J. S. (1976), Sci. Amer. 234, 40. Benjamin, J. S., Bomford, M. J. (1977), Metall. Trans. 8 A, 1301. Benjamin, J. S., Schelleng, R. D. (1981), Metall. Trans. 12 A, 1827. Benjamin, J. S., Volin, T. E. (1974), Metall. Trans. 5, 1929. Benjamin, J. S., Volin, T. E., Weber, J. H. (1974), High Temperatures-High Pressures, 6, 443. Benn, R. C , Benjamin, X S., Austin, C. M. (1984), in: High Temperature Alloys: Theory and Design: Stiegler, J. O. (Ed.). Warrendale (PA), TMS-AIME; p. 419. Benn, R. C , Mirchandani, P. K., Watwe, A. S. (1988), APMIProc. Modern Developments in Powder Metallurgy 18, 479. Benn, R. C , Mirchandani, P. K., Watwe, A. S. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, J. J. (Eds.). Warrendale (PA): TMS-AIME; p. 157. Bever, M. B., Holt, D. L., Titchener, A. L. (1973), Prog. Mater. Sci. 17, 1. Birringer, R., Gleiter, H., Klein, H. P., Marquardt, P. (1984), Phys. Lett. A 102, 365. Black, S. A. (1979), Development of Super corroding Alloys for Use as Timed Releases for Ocean Engineering Applications. Civil Eng. Lab. (Navy), Port Hueneme (CA), p. 40. Bloch, J. (1962), /. Nucl. Mater. 6, 203. Bormann, R. (1987), private communication, GKSS Research Center, Germany. Burgio, N., Iasonna, A., Magini, M., Martelli, S., Padella, F. (1991), Nuovo Cimento, in press. Cahn, R. W, Johnson, W. L. (1986), J. Mater. Res. 1, 724. Cargill, G. S. Ill (1975): Solid State Physics 30: Ehrenreich, H., Seitz, E, Turnbull, D. (Eds.). New York: Academic Press, p. 227.
Carslaw, H. S., Jaeger, J. C. (1959), Heat Conduction in Solids: New York: Oxford University Press, p. 255. Caton, R. (1985), /. Less Common Metals, 109, 9. Clauer, A. H., de Barbadillo, J. J. (Eds.), (1990), Solid State Powder Processing: Warrendale (PA): TMSAIME. Cline, C. E, Hopper, R. W (1977), Scripta Metall. 11, 1137. Coolidge, W D. (1910), Proc. Am. Inst. Elec. Eng. 961. Courtney, T. H., Maurice, D. R. (1990), Solid State Powder Processing: Clauer, A. H., de Barbadillo, J. J. (Eds.): Warrendale, PA: The Minerals, Metals and Materials Society, p. 3. Davis, R. M. (1987), M. S. Thesis, North Carolina State University, Raleigh (NC). Davis, R. M., Koch, C. C. (1987), Scripta Metall. 21, 305. Davis, R. M., McDermott, B. T, Koch, C. C. (1988), Metall. Trans. 19A, 2867. de Boer, F. R., Boom, R., Mattens, W. C. M., Miedema, A. R., Nissen, A. K. (1988): Cohesion in Metals: Amsterdam: North-Holland, p. 391. Dolgin, B. P., Vanek, M. A., McGory, T, Ham, D. J. (1986), /. Non-Cryst. Solids 87, 281. Duerden, I. X, Hume-Rothery, W. (1966), /. Less Common Metals i/, 381. Eckert, X, Schultz, L., Hellstern, E. (1988), J. Appl. Phys. 64, 3224. Eckert, X, Schultz, L., Urban, K. (1989), Appl. Phys. Lett. 55, 117. Edelin, G. (1979), Acta Metall. 27, 455. Edgar, X K. (1949), Trans. AIME 180, 225. Egami, T, Waseda, Y (1984), /. Non-Cryst. Solids 64, 113. Ermakov, A. E., Yurchikov, E. E., Barinov, V. A. (1981), Phys. Met. Metall. 52, 50. Ermakov, A. E., Barinov, V. A., Yurchikov, E. E. (1982), Fiz. Metal. Metalloved. 54, 50. Follstaedt, D. M., Knapp, X S. (1988), /. Less Common Metals 140, 375. Froes, F. H., de Barbadillo, X X (Eds.), (1990): Structural Applications of Mechanical Alloying. Materials Park (OH): ASM Int. Fu, Z., Fecht, H. X, Johnson, W. L. (1990), MRS Symposium Proceedings, Vol. 186. Gaffet, E. (1989), Mater. Sci. and Eng. A 19, 185. Gaffet, E. (1990), Materials Science and Engineering A. Gaffet, E., Harmelin, M. (1990), /. Less Common Metals 157, 210. Gaffet, E., Gaspard, X-P. (1990), J. Phys France (Colloques), 51, C 4-205. Gerl, M., Guilmin, P. (1990), in: Diffusion in Materials: Laskar et al. (Ed.) Norwell (MA): Kluwer Academic Publishers, p. 625. Gilmann, P. S., Benjamin (1983), Ann. Rev. Mater. Sci. 13, 279. Gilman, P. S., Nix, W D. (1981), Metall. Trans. 12 A, 813.
5.10 References
Gleiter, H. (1989), Prog. Mat. Sci. 33, 223-315. Gleiter, H., Marquardt, P. (1984), Z. Metallkd. 75, 263. Green, M. L., Coleman, E., Bader, F. E., Sproles, E. S. (1984), Mater. Sci. Eng. 662, 231. Gross, S. S. (1988), M. S. Thesis, North Carolina State Univ., Raleigh (NC), p. 67. Hansen, M. (1958), Constitution of Binary Alloys, 2nd ed., New York: McGraw-Hill, p. 1049. Harris, C. C. (1967), Trans. Soc. Min. Eng. 238, 17. Heilmann, P., Clark, W. A. T, Rigney, D. A. (1983), Acta Metall. 31, 1293. Hellstern, E., Schultz, L. (1986), Appl. Phys. Lett. 48, 124. Hellstern, E., Schultz, L., Bormann, R., Lewe, D. (1988), Appl. Phys. Lett. 53, 1399. Hellstern, E., Fecht, H. X, Fu, Z., Johnson, W. L. (1989 a), J. Appl. Phys. 65, 305. Hellstern, E., Fecht, H. X, Garland, C , Johnson, W. L. (1989 b), Mater. Res. Soc. Symp. Proc. 132, 139. Hikata, A., McKcnna, M. X, Elbaum, C. (1988), /. Appl. Phys. 65, 305. Horvath, X, Birringer, R., Gleiter, H. (1987), Solid State Commun. 62, 319. Huang, B., Tokizane, N., Ishihara, K. N., Shingu, P. H., Nasu, S. (1990), J. Non-Cryst. Solids 117/118, 688. Inoue, A., Masumoto, T. (1989), in: New Materials by Mechanical Alloying Techniques: Arzt, E., Schultz, L. (Eds.). Oberursel: DGM Informationsgesellschaft, p. 327. Irmann, I. R. (1952), Metallurgia 46, 125. Ivanov, E., Konstanchuk, 1., Stepanov, A., Boldyrev, V. (1987), J. Less Common Metals 131, 25. Ivanov, E., Grigorieva, T., Gdubkova, G., Boldyrev, V., Fasman, A. B., Mikhailenko, S. D., Kalinina, O. T. (1988), Mater. Lett. 7, 51. Ivanov, E. Yu., Konstanchuk, I. G., Bokhonov, B. D., Boldyrev, V. V. (1989), Reactivity of Solids 7, 167. Jang, X S. C , Koch, C. C. (1988), Scripta Metall. 22, 611. Jang, X S. C , Koch, C. C. (1989), Scripta Metall. 23, 1805. Jang, X S. C , Koch, C. C. (1990a), /. Mater. Res. 5, 325. Jang, X S. C , Koch, C. C. (1990b), /. Mater. Res. 5, 498. Jang, X S. C , Koch, C. C. (1990c), Scripta Met. et Mater. 24, 1599. Jang, X S. C , Donnelly, S. G., Godavarti, P., Koch, C. C. (1988), Int. J. Powder Met. 24, 315. Jangg, G., Kutner, F, Korb, G. (1975), Aluminum 51, 641. Jatkar, A. D., Schelleng, R. D., Donochie, S. X (1985), in: Metal-Matrix Carbon and Ceramic-Matrix Composites: Buckley, X D. (Eds.). NASA CP2406, p. 119. Johnson, W. L. (1986), Prog, in Mater. Sci. 30, 81. Johnson, W. L. (1988), Mater. Sci. and Eng. 97, 1.
243
Karch, X, Birringer, R., Gleiter, H. (1987), Nature 330, 556. Kim, M. S., Koch, C. C. (1987), J. Appl. Phys. 62, 3450. Kimura, H., Kimura, M. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, X X (Eds.). Warrendale (PA): TMS-AIME, p. 365. Koch, C. C. (1987), Mat. Res. Symp. Proc. 81, Materials Pittsburgh (PA): Research Society, p. 369. Koch, C. C. (1989), Annu. Rev. Mater. Sci. 19, 121. Koch, C. C , Kim, M. S. (1985), /. Physique 46, C 8-573. Koch, C. C , Cavin, O. B., McKamey, C. G., Scarbrough, X O. (1983), Appl. Phys. Lett. 43, 1017. Koch, C. C , Jang, X S. C , Gross, S. S. (1989), J. Mater. Res. 4, 557. Koike, Jr., Parkin, D. M., Nastasi, M. (1990), J. Mater. Res. 5, 1414. Konstanchuk, I., Ivanov, E., Pezat, M., Darriet, B., Boldyrev, W., Hagenmuller, P. (1987), /. Less Common Metals 131, 181. Kubaschewski, O. (1982), Iron-binary Phase Diagrams, Berlin: Springer-Verlag, p. 59. Kuhn, WE., Friedman, I. L., Summers, W, Szegvari, A. (1985): ASM Metals Handbook Vol. 7, Powder Metallurgy. Metals Park (OH): ASM, pp. 56-70. Kumar, K. S. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, X X (Eds.). Warrendale (PA): TMS-AIME, p. 315. Kumar, K. S., Mannan, S. K. (1989), Mat. Res. Soc. Symp. Proc. 133, 415. Larson, X M., Luhman, T. S., Merrick, H. F. (1977), in: Manufacture of Superconducting Materials: Meyerhoff, R. W (Ed.). Metals Park (OH): ASM, p. 155. Lawn, B. R., Evans, A. G. (1977), /. Mater. Sci. 12, 2195. Lee, P. Y, Koch, C. C. (1987), J. Non-Cryst. Solids, 94, 88. Lee, P. Y, Koch, C. C. (1988), J. Mat. Sci., 23, 2832. Lee, D., Cheng, X, Yuan, M., Wagner, C. N. X, Ardell, A. X (1988 a), /. Appl. Phys. 64, 4772. Lee, P. Y, Jang, X, Koch, C. C. (1988 b), J. Less Common Metals 140, 73. Lee, C. H., Mori, M., Mizutani, U. (1990), J. NonCryst. Solids 117/118, 733. Lend, F. V., Ansell, G. S., Nelson, E. C. (1957), Trans. AIME 209, 117. Leonard, R., Koch, C. C. (1990), unpublished research, North Carolina State University. Linker, G. (1986), Solid State Comm. 57, 773. Loeff, P. (1990), Ph. D. Thesis, University of Amsterdam. Loeff, P. L, Bakker, H., de Boer, F. R. (1989), in: New Materials by Mechanical Alloying Techniques: Arzt, E., Schultz, L. (Eds.). Oberursel: DGM Informationsgesellschaft, p. 119. Lundin, C. E., Yamamoto, A. S. (1966), Trans. AIME 236, 863.
244
5 Mechanical Milling and Alloying
Luzzi, D. E., Meshii, M. (1987), Res. Mechanica 21, 207. Luzzi, D. E., Meshii, M. (1988), /. Less Common Metals 140, 193. Martin, G., Gaffet, E. (1990), J. Phys. France (Colloques) 51, C 4-71. Matsuki, K., Inoue, A., Kimura, H. M., Masumoto, T. (1988), Mater. Sci. and Eng. 97, 47. McCormick, P. G., Wharton, V. N., Schaffer, G. B. (1990): Int'l Symp. on the Physical Chemistry of Powder Metals Production and Processing: Warrendale (PA): TMS, p. 19. McDermott, B. T. (1988), M. S. Thesis, North Carolina State University. McDermott, B. T, Koch, C. C. (1986), Scrip. Metall. 20, 669. Miller, P. X, Coffey, C. S., De Vast, V. F. (1986), J. Appl. Phys. 59, 913. Mizutani, U., Lee, C. H. (1990), J. Mat. Sci. 25, 399. Morris, M. A., Morris, D. G. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, J. J. (Eds.). Warrendale (PA): TMS-AIME, p. 299. Nasu, T, Nagaoka, K., Sekinchi,T, Sakurai, M., Fukunaga, X, Itoh, F., Suzuki, K. (1990), /. Non Cryst. Solids 117/118, 725. Obinata, I., Takeuchi, Y, Kawanishi, R. (1959), Metall. 13, 392. Oehring, M., Bormann, R. (1990), J. Phys. France (Colloques) 51, C 4-169. Okamoto, P. R., Rehn, L. E., Pearson, J. (1988), /. Less Common Metals 140, 231. Orowan, E. (1948), in: Symposium on Internal Stresses in Metals and Alloys: London: Institute of Metals, p. 451. Ovecoglu, M. L., Nix, W. D. (1986), Int. Powder Metall. 22, 17. Quist, W. E., Narayanan, G. H., Wingert, A. L., Ronald, T. M. F. (1986): Aluminum-Lithium Alloys III. London: Institute of Metals, p. 625. Patel, A. N., Diamond, S. (1988), Mater. Sci. Eng. 98, 329. Pearson, W. B. (1972): The Crystal Chemistry and Physics of Metals and Alloys. New York: Wiley-Interscience, p. 151. Petzoldt, F. (1988), /. Less Common Metals 140, 85. Politis, C. (1985), Physica B 135, 286. Polkin, I. S, Kaputkin, E., Ja, Borzov, A. B. (1990), ASM Int'l Conf. of Structural Applications of Mechanical Alloying, Myrtle Beach (SC) March 2729, p. 25. Poon, S. X, Drehman, A. X, Lawless, K. R. (1985), Phys. Rev. Lett. 55, 2324. Rigney, D. A., Chen, L. H., Nayler, M. G. S., Rosenfield, A. R. (1984), Wear 100, 195. Schaffer, G. B., McCormick, P. G. (1989a), Appl. Phys. Lett. 55, 45. Schaffer, G. B., McCormick, P. G. (1989b), Scripta Metall. 23, 835. Schaefer, H. E., Wiirschurr, R., Birringer, R., Gleiter, H. (1988), J. Less Common Metals 140, 161.
Schelleng, R. D., Kemppinen, A. I., Weber, X H. (1988), Space Age Metals Technology. Azusa (CA): SAMPE, 177. Schider, S. (1986), International! Powder Metallurgy 22, 47. Schlump, W, Grewe, H. (1989), in: New Materials by Mechanical Alloying Techniques: Arzt, E., Schultz, L. (Eds.). Oberursel: DGM Informationsgesellschaft, p. 307. Schultz, L. (1984), in: Rapidly Quenched Metals, Steeb, S., Warlimont, H. (Eds.): Amsterdam: North-Holland, p. 1585. Schultz, L. (1988), Mat. Sci. & Eng. 97, 15. Schultz, L., Hellstern, E. (1987), MRS Symp. Proc. 80, Tenhover, M., Tanner, L. E., Johnson, W L. (Eds.). Bodston: MRS, p. 3. Schultz, L., Wecker, X, Hellstern, E. (1987), / Appl. Phys. 61, 3583. Schultz, L., Hellstern, E., Zorn, G. (1988), Z. Phys. Chem. 157, 203. Schwarz, R. B., Johnson, W. L. (1983), Phys. Rev. Lett. 51, 415. Schwarz, R. B., Johnson, W L. (1988), Int. Conference on Solid State Amorphizing Transformation. Lausanne: Elsevier Sequoia, S.A. Schwarz, R. B., Koch, C. C. (1986), Appl. Phys. Lett. 49, 146. Schwarz, R. B., Petrich, R. R. (1988), J. Less-Common Metals 140, 171. Schwarz, R. B., Petrich, R. R., Saw, C. K. (1985), /. Non-Cryst. Solids 76, 281. Seki, Y, Johnson, W. L. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, X X (Eds.). Warrendale (PA): TMS-AIME, p. 287. Sergev, S. S., Black, S. A., Jenkins, X F. (1981), US Patent No. 4,264,362. Shechtman, D., Blech, I., Gratias, D., Cahn, X W (1984), Phys. Rev. Lett. 53, 1951. Shingu, P. H., Huang, B., Nishitani, S. R., Nasu, S. (1988), Suppl. to Trans. JIM 29, 3. Shingu, P. H., Huang, B., Kuyama, X, Ishihara, K. N., Nasu, S. (1989), in: New Materials by Mechanical Alloying Techniques: Arzt, E., Schultz, L. (Eds.). Oberursel: DGM Informationsgesellschaft, 319. Song, M. Y, Ivanov, E., Darriet, B., Pezat, M., Hagenmuller, P. (1987), J. Less Common Metals 131, 71. Stepanov, A., Ivanov, E., Konstanchuk, I., Boldyrev, V. (1987), J. Less Common Metals 131, 89. Stephens, X X, Nix, W. D. (1985), Metall. Trans 16A, 1307. Stoloff, N. S., Davies, R. G. (1966), Progress in Materials Science 13, 11. Sundararajan, G., Shewmon, P. G. (1983), Acta Metall. 31, 101. Sundaresan, R., Froes, F. H. (1987), /. of Metals 39, 22. Sundaresan, R., Froes, F. H. (1988), Modern Developments in Powder Metallurgy 21, 429.
5.10 References
Sundaresan, R., Frocs, F. H. (1989), in: New Materials by Mechanical Alloying Techniques: Arzt, E., Schultz, L. (Eds.). Oberursel: DGM Informationsgesellschaft, p. 253. Suryanarayana, C , Sundaresan, R., Froes, F. H. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, J. J. (Eds.). Warrendale (PA): TMS-AIME, p. 55. Trudeau, M. L., Schulz, R., Dusault, D., Van Neste, A. (1990), Phys. Rev. Lett. 64, 99. Vedula, K., Strothers, S. D. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, J. J. (Eds.). Warrendale (PA): TMS-AIME, p. 213. Veprek, S., Iqbal, Z., Sarotte, F.-A. (1982), Phil. Mag. B45, 137. Wagner, C. N. J., Light, T. B., Haider, N. C , Lukens, W E. (1968), J. Appl. Phys. 39, 3690. Wang, G., Zhang, D., Chen, H., Lin, B., Wang, W, Dong, Y. (1991), Phys. Lett. A 155, 57. Weber, J. H. (1980), The 1980's - Payoff Decade for Advanced Materials, 25. Azusa (CA): S.A.M.P.E., p. 752. Weber, J. H. (1990), in: Solid State Powder Processing: Clauer, A. H., de Barbadillo, J. J. (Eds.). Warrendale (PA): TMS-AIME, p. 227. Weeber, A. W, Bakker, H. (1988), Physica B 153, 93. Weeber, A. W, Wester, A. J. H., Haag, W. I, Bakker, H. (1987), Physica B 145, 349. White, R. L. (1979), Ph. D. Thesis, Stanford University, p. 57. White, R. L., Nix, W D. (1979), in: New Developments and Applications in Composites: KhulmannWilsdorf, D., Harrigan, W C. (Eds.). Warrendale (PA): TMS, p. 78.
245
Wright, I. G., Wilcox, B. A. (1974), Battelle Columbus Laboratory Research on Metallurgical Synthesis for AFML, AD-781 133, p. 15. Yamazaki, M., Kawasaki, Y, Kusumoki, K. (1990), ASM International Conf on Struc. Appl. of Mech. Alloying. Myrtle Beach (SC). March, p. 33. Yavari, A. R., Desre, P. J. (1990). "Multilayer Amorphisation by Solid-State Reaction and Mechanical Alloying", in: /. Phys. France (Colloques) 51, C4-1-C4-310. Yeh, X. L., Samwer, K., Johnson, W. L. (1983), Appl. Phys. Lett. 42, 242.
General Reading Benjamin, J. S. (1976), Scientific American 234, 40. Clauer, A. H., de Barbadillo, J. J. (Eds.) (1990), Solid State Powder Processing. Warrendale, PA: The Minerals, Metals, and Materials Society. Arzt, E., Schultz, L. (Eds.) (1990), New Materials by Mechanical Alloying Techniques. Oberursel: DGM Informationsgesellschaft. Koch, C. C. (1989), "Materials Synthesis by Mechanical Alloying", in: Annu. Rev. Mater. Sci. 19, 121. Froes, F. H., de Barbadillo, J. J. (Eds.) (1990), Structural Applications of Mechanical Alloying. Metals Park, Ohio: ASM International. "International Symposium on Amorphization by Solid State Reaction" (1990), in: J. Phys. France (Colloques) 51, C4.
6 Ion Implantation and Ion-Beam Mixing David M. Follstaedt Sandia National Laboratories, Division 1112, Albuquerque, NM, U.S.A.
List of Symbols and Abbreviations 6.1 Introduction 6.1.1 Microstructures of Ion-Implanted Alloys 6.1.2 Irradiation-Induced Microstructure Changes 6.1.3 Ion-Beam Mixing 6.1.4 Reasons for Studying Ion-Beam Alloys 6.2 Basic Processes 6.2.1 Distribution of Implanted Atoms 6.2.2 High-Fluence Ion Implantation 6.2.3 Lattice Displacements 6.2.4 Solute-Defect Interactions 6.2.5 Radiation-Induced Changes in Microstructure 6.2.6 Ion-Beam Mixing 6.3 Metastable Phase Formation 6.3.1 Metastable Solid Solutions 6.3.2 Precipitated Phases 6.3.3 Metastable Crystalline Phases 6.3.4 Amorphous Phases 6.3.5 Icosahedral Phases 6.3.6 Al-Ni Ion-Beam Alloys 6.4 Basic Alloy Phenomena 6.4.1 Solubility and Diffusion of Sb in Fe 6.4.2 Deuterium Trapping at He Bubbles in Ni 6.4.3 Search for Cold Fusion: Pd-D 6.5 Surface Alloys with Improved Macroscopic Properties 6.5.1 Nitrogen Implantation to Reduce Wear of Steels 6.5.2 Implantation of Ti + C to Reduce Friction and Wear 6.5.3 High-Strength A1(O) Alloys 6.5.4 Aqueous Corrosion of Fe-Based Alloys 6.5.5 Reduced Corrosion and Wear of Ti-6A1-4V 6.6 Concluding Remarks 6.7 Acknowledgements 6.8 References Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
248 249 251 252 253 254 255 256 257 258 259 259 261 262 262 263 264 265 268 269 272 272 273 275 275 276 277 279 280 281 282 283 283
248
6 Ion Implantation and Ion-Beam Mixing
List of Symbols and Abbreviations a0 Cm Co D' D Ed Fd /s 1 x
AHS Kl,K2 N Nd Rp ARp S t Tc Q cj) AES dpa IBAD MWD PSII RBS SIMS TEM TRIM UHMWPE
distance between lattice points maximum concentration concentration of equilibrium chemically biased diffusivity thermal diffusivity energy needed to displace an atom energy deposited into elastic collisions channeling fraction cohesive energy of the alloy heat of mixing of the elements heat of solution empirically evaluated constants atomic density average number of displacements per target atom projected ion range width of distribution of Rp number of sputtered atoms time critical temperature for ion-beam mixing average atomic density ion fluence Auger electron spectroscopy displacements per atom ion-beam assisted deposition maximum wear depth plasma-source ion implantation Rutherford backscattering spectroscopy secondary ion mass spectroscopy transmission electron microscopy transport of ions in matter ultrahigh molecular weight polyethylene
6.1 Introduction
6.1 Introduction Metal alloys formed by ion-beam methods have been studied since the 1970's. These studies use either ion implantation to inject elements directly into a target, or a later technique, ion-beam mixing, to alloy deposited elemental layers. Both methods produce atomic-scale alloys, as opposed to mechanical mixtures of the species. The elements are alloyed together by athermal ballistic processes, which may allow some of the thermodynamic and kinetic constraints which usually govern alloy formation (i. e., liquid or solid solubilities, transformation rates) to be bypassed. This advantage permits metastable compositions and structures to be more readily achieved, although thermodynamic and kinetic considerations still determine alloy microstructures.
249
With ion implantation, individual atoms are accelerated electrostatically to kinetic energies of ~100keV and embedded into a host material. Any element can thus be placed into an intimate atomic mixture with any target. The implantations of Ta and W into solution in Cu (Cullis et al., 1976; Borders et al., 1976), are notable examples, since these elements are immiscible in both solid and liquid Cu. This ability to form homogeneous alloys with non-equilibrium structures can lead to useful and novel new materials. For example, implanting Ti and C together into Fe or steels forms an amorphous surface layer which reduces friction and wear (FoUstaedt, 1985 a). The dramatic benefits of this treatment are seen in Fig. 6-1: dry sliding friction is reduced to ~ 1/2 and wear depth is reduced down to 1/10 of unimplanted values (Pope etal., 1983). Secondly, the Alfa) WEAR SURFACE
a) FRICTION
0.8
Ti + C IMPLANTED
TYPE 304 STAINLESS STEEL |
I
500
1000
WEAR CYCLES Figure 6-1. a) Friction traces from 304 disks, unimplanted and implanted with 2 x 1021 Ti/m2, 180-90 keV, plus 2 x 1021 C/m2, 50 keV (440 C pin, 12 g load), b) Scanning electron micrographs of the wear tracks (FoUstaedt etal., 1989 b).
250
6 Ion Implantation and Ion-Beam Mixing
The thicknesses of surface alloys formed with ion beams are limited by the ion ranges to < 1 jim. Concentrations up to 10's at.% can be implanted if sputtering effects are not severe, and entire binary ranges are accessible with ion-beam mixing. Ion-beam analysis methods are often used to measure composition profiles of implanted species in a non-destructive manner with depth resolutions of ~ lOnm. For example, Rutherford backscattering spectroscopy (RBS) was used to obtain the profile of Ni implanted into Al in Fig. 6-3 (Picraux et al., 1980); RBS is especially useful for depth profiling a heavier element implanted into a light substrate. The use of ion beam analysis to characterize materials is discussed in Vol. 2 of this series. Two other techniques are also used to profile implanted elements: secondary ion mass spectroscopy (SIMS) (see Vol. 2 of this series) and sputter Auger electron spectros-
AKNi)
7xlO 1 5 Ni/cm 2 150 keV
Figure 6-2. a) Bright-field TEM micrograph of fine grains of icosahedral Al 80 Mn 20 formed by ion-beam mixing with 1 x 1020 Xe/m2, 400 keV, at 150°C b) Diffraction pattern of rings from the icosahedral layer and spots from sapphire substrate (Knapp and Follstaedt, 1985).
Mn phase which exhibits icosahedral symmetry in diffraction can be formed by ion-beam mixing, as seen in Fig. 6-2 (Knapp and Follstaedt, 1985; Lilienfeld et al., 1985), and by ion implantation (Budai and Aziz, 1986). Ion-beam alloys contributed significantly to understanding this new phase, whose symmetry was widely doubted to be possible before 1984.
500
1000
1500
2000
2500
DEPTH (A)
Figure 6-3. Concentration profile for 7 x 1019 Ni/m2 implanted at 150keV into <110> Al single crystal, determined by RBS. The depth profile of lattice disorder as determined by ion channeling is also shown, and seen to peak closer to the surface than the concentration (Picraux et al., 1980).
1.0
6.1 Introduction
copy (AES) (Singer et ah, 1981). The implanted element usually retains the distribution with which the atoms came to rest, unless the temperature is high enough to allow diffusion. In most instances, room temperature maintains the as-implanted profile. High-energy ions slow and come to rest by colliding with target atoms, which they displace from lattice sites. The displacements can be numerous for high fluences: implanting 10 21 atoms/m2 may produce a concentration of ~ 10 at.% and ~10 2 displacements per atom (dpa) in the implanted layer. Many of the vacancies and interstitials produced along the ion track recombine immediately, but some remain in the lattice. The temperature determines the mobility of the remaining point defects, and thereby affects the resulting alloy microstructure. The displacements disrupt the existing lattice structure and thus can promote the formation of new phases. In the remainder of Sec. 6.1, the types of microstructures found in ion-beam alloys and the ion-beam mixing process are introduced, followed by a discussion of the motivations for study of such alloys.
251
tron microscopy (TEM) for most metals. To resolve individual dislocations, weakbeam imaging is used, as shown for Al(Mo) (Bentley et al, 1984) in Fig. 6-4. With increasing fluence, the matrix often becomes a metastable solution whose concentration exceeds the equilibrium solid solubility. When single crystal hosts are used, ion channeling can be combined with ion-beam analysis techniques like RBS to determine the position of the implanted solute in the lattice (Feldman et al., 1982). For example, implanting 1 at.% W into Cu at 40-60 K results in a channeling fraction / s ^0.9 of W atoms substituting on f.c.c. lattice sites. The alloy thus contains ~0.9 at.% W in solution, in contrast to the negligible solubility of W in Cu. The small fraction, / s = 0.18, found for the Al(Ni) alloy of Fig. 6-3 implies a substitutional concentration ~0.2 at.% Ni, which nonethe-
6.1.1 Microstructures of Ion-Implanted Alloys
In this section, the changes in microstructure produced during implantation are considered in order of increasing fluence and composition change for typical metal alloys. Implantations are often done at room temperature, where point defects are mobile in metals and coalesce into dislocation loops. With increasing density, the loops interact and form tangled networks of dislocations. This sequence was observed in Al implanted with Ni (Al(Ni)) (Picraux et al., 1980). Dense dislocation networks are seen with transmission elec-
Figure 6-4. Weak-beam TEM image of dislocations in Al implanted with 4.4 at.% Mo. Reprinted by permission of the publisher from Second Phase Formation in Aluminum Annealed After Ion Implantation with Molybdenum, by J. Bentley, L. D., Stephenson, R. B. Benson Jr., P. A. Parrish, and J. K. Hirvonen, Mat. Res. Soc. Symp. Proc. 27, p. 152. Copyright 1984, by Elsevier Science Publishing Co., Inc.
252
6 Ion Implantation and Ion-Beam Mixing
less exceeds the solubility of Ni in Al by 10 times (Picraux et al., 1980). Further concentration increases can produce phase transformations such as precipitation within the host. A high density of small precipitates is typically observed. The precipitates in Fe implanted with 38at.%C have been characterized with TEM (Follstaedt, 1985b; also Pavlov et al., 1973). The diffraction pattern of the second phase is superimposed on that of the b.c.c. Fe matrix in Fig. 6-5 a, and identifies it as hexagonal e-Fe2C. The dark-field image in Fig. 6-5 b shows a high density (> 1 x 104/mn2) of precipitates 5-30 nm in diameter. If the implanted concentration reaches that of the precipitates, they can
impinge on each other and form a continuous layer of the phase (Musket et al., 1985). If precipitation does not occur, an amorphous phase may form with increasing concentration (Grant, 1978). Metalloid implantation (~20at.%) often forms amorphous phases as in melt-quenched alloys, such as Ni(P) (Ali et al., 1977). Results with Ni(C) demonstrate that temperature can determine the phase of the implanted alloy (Nastasi et al., 1988). At room temperature Ni 3 C precipitated, while at 77 K an amorphous phase formed. This example also illustrates the competition between the two microstructures: if the implanted element segregates into precipitates, it cannot amorphize the matrix. Amorphous metals can also be formed when metallic elements are implanted. 6.1.2 Irradiation-Induced Microstructure Changes
Figure 6-5. a) Electron diffraction pattern with weak 8-Fe2C reflections (outlined) and intense b.c.c. reflections «111> zone axis), b) Dark-field TEM image of £ precipitates obtained by imaging with the (10T0) reflection indicated in a) (Follstaedt, 1985 b).
To understand how atomic displacements affect microstructure, irradiations that do not significantly change composition are used: self-ions of the target element, inert-gas ions, or high-energy ions passing through the depth of interest. Pure metals usually retain their equilibrium structures, but Ni transforms from f.c.c. to h.c.p. (Johnson et al., 1983). The h.c.p. phase is thought to be stabilized by lattice damage and associated stress. Steels transform martensitically (f.c.c. -• b.c.c.) when irradiated, and stress is again believed to drive the change (Johnson, 1990). Phase changes are more often observed when binary alloys are irradiated, since atomic displacements not only produce lattice damage, but also disrupt the chemical order of compounds (Schulson, 1979). For example, atoms are interchanged between the Cu and Au sublattices of the ordered phase Cu3Au (Ll 2 structure), and
6.1 Introduction
a random solution is formed. Disorder is detected with diffraction by the disappearance of reflections associated with chemical order. The disorder and lattice damage produce amorphization of some alloys but not others, and irradiation studies thus indicate which phases are expected in ionbeam alloys (see also Chap. 9, Sec. 9.2.5). A phase which amorphizes during irradiation is not expected to form in ion-implanted or ion-beam mixed alloys, but a phase which persists may form in alloys with its composition. For example, the phases of Al-Ni alloys formed by the two methods are the same as those observed after ion irradiation of Al-Ni compounds (see Sec. 6.3). 6.1.3 Ion-Beam Mixing Ion-beam mixing was developed to form alloys with higher concentrations not achievable with implantation. During implantation, atoms near the surface can receive sufficient energy to escape the solid. If on average S atoms (either host or implanted) are sputtered from the target by each incident ion, the maximum concentration which can be implanted at a fixed energy reaches a steady-state value (Liau and Mayer, 1980): Cm = l/S
(S>1)
(6-1)
Sputtering often limits the concentrations to a few 10's of atomic percent, and binary compositions near 50 at.% may not be achievable by ion implantation. With ion-beam mixing, layers of two deposited elements are alloyed by atomic displacements during irradiation. High-energy, inert-gas ions are often used (e.g., 300-600 keV Xe) since they pass through 50 nm of metal while creating sufficient intermixing to form an alloy. Two configurations are typically used (Lilienfeld, 1987).
253
The bilayer structure of Fig. 6-6 a is often used to study the amount of intermixing with fluence, while multiple alternating layers in Fig. 6-6 b are used to form alloys of fixed composition determined by the relative thicknesses of the elemental layers. Ion-beam mixing also alloys more atoms per incident ion: doses of 10 1 9 -10 2 0 ions/ m 2 are typical for mixing, whereas > 1021 ions/m2 are often needed with implantation. Thus, high concentrations are achieved with less accelerator time. Heavy ions like Xe create numerous atomic displacements and mix more efficiently. The configuration in Fig. 6-6 a was used to alloy Al and Pt (Mayer et al., 1981) as shown in Fig. 6-7. With Xe ion mixing, RBS analysis indicates that the initial sharp interface between the layers (the nearly vertical solid lines at the right edge of the Pt signal and left edge of the Al signal) becomes graded as a Pt-Al alloy forms. In this example, a stoichiometric compound (PtAl2) forms, and the mixing proceeds at constant composition with increasing Xe fluence, as indicated by the steps in the two signals. The ion-beam mixed phases can differ from those produced by thermally reacting the layers, and
Ion Induced Phase Transformation Sample Configurations A)
Bilayer
j Substrate B)
Multilayer, Planar
3
2Z2 <
B
Substrate
Figure 6-6. Layer configurations for ion-beam mixing: a) bilayer and b) multilayer (Lilienfeld et al., 1987).
254
6 Ion Implantation and Ion-Beam Mixing
Si/Pt/AI,
2.0 MeV 4He* "• BACKSCATTERING
ION IMPLANTED
rUNIMPLANTED |x|0l5cm2 2 - 4 x | 0 1 5 cm
PtEDGE
0.6
1.2
ENERGY
1.4
1.6
1.8
I2.0
Figure 6-7. RBS spectra for Al/ Pt layers ion-beam mixed with 300 keV Xe. The steps show the formation of PtAl2 (Mayer et ah, 1981).
(MeV)
in this system, PtAl 4 forms when Pt/Al layers are annealed, instead of PtAl 2 . Some systems form solid solutions; Cu and Ag form metastable solutions across the composition range (Tsaur et al., 1982) instead of separating into the elemental phases with only limited solubilities as for equilibrium alloys. Solutions with wide composition ranges also form in alloys of Au with V, Fe, Co or Ni (Tsaur et al., 1981). Intermediate phases with simple crystal structures can form as well, such as B2 compounds (CsCl structure) as found with Pd-Al and Ni-Al (Hung et al., 1983). The phases of ion-beam mixed alloys are often the same as those of ion-implanted or irradiated alloys with the same composition, but immiscible alloy systems provide exceptions. For example, Cu/W layers show very little mixing and the f.c.c.-Cu and b.c.c.-W elemental phases are retained, even when irradiated with Kr at 77 K (Meissner et al., 1987). The lattice constants of Cu and W have been observed to shift from the pure-element values in irra-
diated layers, which indicates that some alloying occurs (Hiller et al., 1989); however, even very thin (0.6 nm) W layers could not be completely mixed into Cu at 10 K (Westendorp et al., 1987). When both phases are initially present, they persist during irradiation. In contrast, W atoms implanted into Cu at 40-60 K remain dispersed in a metastable f.c.c. solution for up to ~ 1 at.% W or an amorphous phase at - 1 0 at.% W (Cullis et al., 1976), instead of precipitating. A large positive heat of solution (AHs = 36 kJ/mol, Cu-W) is characteristic of immiscibile systems. Ion-beam mixing has produced alloying in some binary systems with AHS as large as +40 kJ/ mol (Ag-Cr), and amorphous phases were also formed (Liu et al., 1987). 6.1,4 Reasons for Studying Ion-Beam Alloys
Work with ion-beam alloys often belong in one of three categories. First, many studies attempt to form and understand meta-
6.2 Basic Processes
stable alloys. Compositions and structures not readily achieved with conventional alloying methods may be produced, and lattice damage can induce the formation of new phases. Amorphous Fe(Ti,C) alloys with ~20at.%Ti and C cannot be produced by liquid quenching, even with 30 ns laser pulses and cooling rates >10 1 0 K/s (Follstaedt et al., 1984 a), but are readily formed with implantation (Fig. 6-1). Icosahedral phases were discovered in meltquenched alloys (Shechtman et al., 1984), but ion-beam alloys formed in the solid state (Fig. 6-2) provided unique information about the formation, thermodynamics and structure of this new type of atomic order. Thus ion-beam alloys not only increase the array of metastable phases available for materials engineering, but also provide insight into these novel materials. A second category of studies characterizes basic solute processes in alloys. When the thermal evolution of implanted alloys with sub-micron dimensions is followed by ion-beam analysis methods having sufficient depth resolution, these processes can be examined at lower temperatures. In addition, ion implantation can be used to tailor composition profiles to exhibit the phenomena of interest. A typical example is the trapping of H at He bubbles in Ni (Besenbacher et al., 1982). With this approach, basic information important for bulk equilibrium alloys is obtained. The third type of investigation aims to produce surface alloys with improved properties for potential applications. Engineering alloys implanted with the appropriate elements can exhibit increased strength, reduced friction and wear rate, or increased corrosion resistance. Many steel components wear less rapidly when implanted with N, and Cr implantation is found to reduce the corrosion rates of steels. The amorphous phase of Fe im-
255
planted with Ti and C also forms when they are implanted into steels, and reduces friction and wear (Fig. 6-1). Ion implantation is offered commercially for treatment of engineering components (Armini, 1986; Dearnaley, 1987; Sioshansi, 1987). The basic atomic processes occurring during ion irradiation are examined in Sec. 6.2. This background allows a more complete understanding of composition profiles of implanted elements, the interactions of implanted atoms with lattice defects, radiation-induced microstructure changes, and the kinetics of ion-beam mixing. The following Sections (6.3-6.5) discuss experiments in the three categories discussed above. These discussions give examples of the types of alloys, information and benefits which can be obtained, but are not intended to be comprehensive reviews of these areas or classes of alloys. More thorough compilations of published work are found in review articles which are frequently referenced. Expected extensions of work with ion-beam alloys are discussed in Sec. 6.6.
6.2 Basic Processes The interaction of high-energy ions with the target is usually assumed to have two independent components (Ziegler et al., 1985): an elastic or nuclear interaction in which the ion collides with atoms and displaces them, and an inelastic or electronic interaction in which the ion excites electrons. The electronic interaction increases with ion energy and dominates at high energies, while nuclear scattering is strongest at low energies as the ion comes to rest. In general, the interactions are comparable, and together determine the average depth or projected range (Rp) of ions in the target and the width of the distribution (ARp)
256
6 Ion Implantation and Ion-Beam Mixing
about the peak at Rp. Displacements created by the nuclear interaction produce high concentrations of vacancies and interstitials, which can interact with the implanted atoms to form solute-defect complexes. If sufficient energy is transferred from the ion to atoms near the surface, they can be sputtered from the target. These effects combine to determine the concentration profiles of implanted atoms and their defect structures. The nuclear interaction can also produce high local densities of displaced atoms, termed collision cascades, which are responsible for ion-beam mixing at low temperatures. 6.2.1 Distribution of Implanted Atoms
A common approach is to use Monte Carlo simulations to calculate the distribution of implanted atoms in the target. Methods have been developed to evaluate the nuclear scattering interaction analytically, and to obtain the deflection angle of the ion and the energy lost to the host atom. This approach uses a random number generator to choose the impact parameter from probability distributions which correspond to random atomic positions in an amorphous target (Biersack and Haggmark, 1980). The ion is decelerated between collisions by the electronic interaction, which is evaluated from analytical expressions fitted to measured values. The distance between collisions is chosen to be as large as possible in order to speed the calculation, but subject to requiring many small-angle collisions to occur before a large-angle collision is probable. Each ion is followed until it stops; by obtaining final positions for many ions, Rp and ARp can be accurately determined. Both interactions have been fitted with expressions which account well for the observed slowing of ions (stopping power) for
many ion-target combinations and a wide range of ion energies (Ziegler et al., 1985). The calculated values agree within ±10% for 80% of the experiments, and +20% for 93%. The final result is a program that can be run on modern desk-top computers and gives useful information within an hour. A widely-used version is TRIM 87 (TRansport of Ions in Matter; Ziegler, 1987). Experience suggests that its Rp values are accurate to within ~ 15%, and ARp to within -30%. In Fig. 6-8, values of Rp calculated with TRIM are given for ions with a range of masses (He to Xe) implanted into Al, Fe and W. For lOOkeV, a typical implantation energy, Rp is 0.01-1.0 jam. The ARp values (not shown) are 1/3 to 1/2 of Rp. The range increases with energy, and decreases with the atomic mass of either the target or the ion. Thus, thicker alloy layers can be formed with light elements, such as B, C, N or O. The range of 600 keV Xe (-0.1 jim) is sufficient for a set of layers with total thickness <0.1 jam to be ion-beam mixed, and the Xe will pass through them. Highenergy He + used for ion-beam analysis
He
1.00
N
P
Ti Ni Mo Xe
•
0.50
A
Al
«
0.20
• Fe W
0.10
A
• Al • A
0.05
•
•
• o w •
A
•
0.02
A Fe
B
A
n m 2
5
10 20 50 100 200 Ion Mass (amu)
Figure 6-8. Ranges of ions implanted into Al, Fe and Ag as calculated with TRIM 87, plotted versus ion mass. Filled symbols - 100 keV; open symbols 600 keV.
6.2 Basic Processes
(typically 2 MeV, Rp ~ 2 jam) also passes through ion-beam alloys. 6.2.2 High-Fluence Ion Implantation
At high fluences (> 10 21 ions/m 2 ) and concentrations (>25 at.%), sputtering and stopping power changes can affect the implanted distribution. If sputtering removes thicknesses comparable to Rp, the distribution becomes asymmetric about the peak, instead of the nearly symmetric shape seen for Ni in Al in Fig. 6-3. The profile is then increased near the surface, as seen for Ti in 304 stainless steel in Fig. 6-9 (Follstaedt et al., 1989 a), because ions implanted early in the treatment are uncovered by sputter erosion during further implantation. The lighter Ti atoms were profiled in the steel matrix by using higherenergy (6 MeV) He + backscattering to separate the Ti signal from structure on the Fe signal for more accurate analysis (Knapp et al., 1985). The profile in Fig. 6-9 was obtained by fitting it to the He + energy spectrum with the RUMP simulation program (Doolittle, 1985). The Rp value from TRIM is shown for comparison. The TRIM program does not take account of sputtering nor changes in stopping power due to the composition change during high-fluence implantations. Stop-
•I
4. 2 x 1 0 " Ti
40 -
•
\
20 o o
• Ti + C x H
\
•
itV 0
50
100
150
DEPTH (nm)
Figure 6-9. Concentration profiles of Ti, C, and H in 304 stainless steel (Follstaedt et al., 1989a). The range of 180 keV Ti is indicated with an arrow.
257
ping power changes can be important for high concentrations of an implanted element with an atomic number significantly different from that of the host. A program which accounts for such effects gives better agreement with the profiles observed for N implanted into Ti and B in Ni, which are deeper than predicted by TRIM (Bunker and Armini, 1989). Another important effect occurring for high fluences of reactive species is seen in Fig. 6-9. Only Ti + was accelerated and implanted, but the resulting alloy contains significant C (20 at.%) and a lesser amount of H near the surface. Residual molecules in the vacuum are the source of such impurities (Singer and Barlak, 1983; Hoffmann etal, 1987). Notably, the C and H in Fig. 6-9 were incorporated during implantation in a high-vacuum system with a base pressure of only 3 x 10~ 7 Torr (Follstaedt et al., 1989 a). Profiling of the light C atoms was possible because their cross section is enhanced to 48 times the Rutherford value by using the high-energy He + (6 MeV). The H was profiled by another technique, elastic recoil detection. The amount of C incorporated into steels increases with fluence, and a high concentration of Ti at the surface is the apparent cause. A model which uses sputtering, a reaction of C with Ti, and inward diffusion of C can account for the profiles of Ti and C in 52100 steel (Farkas et al., 1984). The sputtering coefficient of Ti (S = 2) may have been reduced by its bonding to C. Using this value in Eq. (6-1) gives reasonable agreement with the Ti concentration seen in Fig. 6-9 (40 at.%). It should be emphasized that impurity incorporation does not generally occur during implantation, but only with reactive species. In addition to C and H, O can be incorporated, depending upon the residual gases in the vacuum (Hoffman et al.,
258
6 Ion Implantation and Ion-Beam Mixing
1987). Incorporation of C is important for steels implanted with Ti to reduce friction and wear (Singer et al., 1981). In these alloys C is not an unwanted impurity, but is essential for tribological benefits. In such work, additional C is often implanted into the deeper part of the Ti profile to produce thicker, longer-wearing amorphous layers (Follstaedt et al., 1984 b), as in Fig. 6-1. However, H is apparently an unimportant contaminant. 6.2.3 Lattice Displacements The number of atomic displacements created by the nuclear interaction can be estimated with the well-known, modified Kinchin-Pease relation (Sigmund, 1969): = 0.S(f>Fd/2EdN
(6-2)
where Nd is the average number of displacements per target atom (dpa), 0 is the ion fluence (ions/m2), Fd is the energy deposited into elastic collisions, Ed is the energy needed to displace an atom from its lattice site (~ 20 eV), and N is the atomic density. TRIM also keeps account of Fd during the simulations, and uses this relation to calculate the number of displacements. The number is substantial: ~1000 vacancies are expected along the ion track in Al for each 100 keV Ni ion, but most of these recombine with interstitials so that far fewer are retained. The host atoms are nonetheless displaced many times; for 1020 Ni/m 2 , calculated to produce a peak concentration of 2.6at.%Ni, individual atoms in the implanted layer are displaced 10-24 times. For the 40at.%Ti in 304 steel in Fig. 6-9, the damage varies from -1000 to 2000 dpa. TRIM also determines the depth distribution of displacements, which peaks closer to the surface than jRp. The distribution of lattice damage in Ni-implanted Al (see Fig. 6-3) was found
with ion channeling to peak closer to the surface than the concentration (Picraux et al., 1980), as expected. An important aspect of lattice displacements is that they are not uniform along an ion track. Instead, an ion displaces a host atom which recoils and produces a dense cascade of displaced atoms; it then travels in a straight line until it has another such collision. These localized displacements are termed collision cascades, and are believed to be well-defined entities in sufficiently heavy (Z > 20) materials irradiated with heavy ions (Johnson et al., 1985); the density of displaced atoms increases with ion and target masses. The collision cascade is a state of high kinetic energy (several keV), small size (<10nm), and short duration ( 1 0 " l x 10 ~ 10 s) (Rehn and Okamoto, 1989). These features make the processes occurring in it inaccessible to experiment; instead, they are explored with molecular dynamics simulations (de la Rubia et al., 1987). A cascade with 3 or 5 keV in Cu is found to have a structure like that of a liquid soon after the incoming ion collides and transmits energy to the zone. This structure, known as a thermal spike, lasts for several picoseconds before its kinetic energy dissipates into the surroundings. The simulations indicate that most atomic diffusion occurs during the thermal spike period. In this regime, the cascade is inferred to be molten from the radial distribution function of atoms (Harrison and Webb, 1983) and the Maxwellian distribution of kinetic energies (King and Benedek, 1983) found in simulations. These features of collision cascades allow several aspects of ion-beam alloying to be understood, including amorphous phase formation and the kinetics of ionbeam mixing.
6.2 Basic Processes
6.2.4 Solute-Defect Interactions
The degree to which implanted atoms occupy lattice sites can be explained with the point defects created when defect mobility and interactions are considered. The substitutional character of oversized atoms with positive heats of solution (AHS > 0; endothermic reaction) has been measured by in-situ ion channeling for Al (Kloska and Meyer, 1987), V (Turos et al, 1987) and Fe (Meyer and Turos, 1987). For example, high fractions (/ s ~ 1) of Au, Sb, Hg, Bi or Pb atoms implanted into Fe at 77 K occupy host lattice sites, while lower values (/s = 0.6-0.9) result at 293 K. These results are accounted for by the trapping of vacancies at the oversized solute atoms. The solute is assumed initially to annihilate a vacancy in the cascade and occupy a host lattice site. For implantation at a temperature below which vacancies become mobile (stage III = 200 K for Fe), the solute remains substitutional. However, if vacancies are mobile as in Fe at 293 K, they diffuse to the solute and form defect complexes in which the solute moves off the lattice site. Solute-vacancy binding increases with the heat of solution, and / s is found to be greatly reduced for AHS >250kJ/mol, e.g., implanting Cs {AHS = 512 kJ/mol) at 77 K gives / s = 0.35, and 0.0 at 293 K. The lower values (/ S <1) at 77 K for large AHS suggest that some binding occurs during the cooling phase of the thermal spike. Undersized solutes are similarly expected to bind to interstitials in mixed dumbell configurations. 6.2.5 Radiation-Induced Changes in Microstructure
In addition to producing lattice damage, ion irradiation can alter two features of alloy microstructures: crystal size and phase. The effects of irradiation on grain sizes in
259
polycrystalline layers have been examined using self-ions or inert-gas ions. Grains in free-standing Au films, which are stable at room temperature, increased in radius from < 40 nm to > 100 nm after 1 x 1019 Xe/m 2 at 200 keV (Atwater et al., 1988). The grain size increased with fluence, and the size distribution indicated that the graingrowth was driven by the reduction of grain-boundary energy. By changing ion mass and energy, the grain size was found to scale with the number of defects created in the film. The authors propose mechanisms involving migration of individual point defects across grain boundaries to explain the size evolution. Similar increases have been found for Ni, with an additional notable feature: the grain size saturated with increasing ion fluence (Liu and Mayer, 1987). The saturation was inferred to reflect the finite size of the collision cascades, which were hypothesized to remove small grains by disordering the entire grain followed by epitaxial regrowth on surrounding grains. The distribution would no longer increase when the smaller grain sizes exceeded the cascade size so that no grains remained which could be totally disordered. Recrystallization of grains is thus promoted by radiation, even at temperatures where thermal evolution is negligible. Lattice displacements are generally believed to be responsible, but the detailed mechanisms and the role of cascades are not yet conclusively identified. In some metal layers, texturing is also produced by ion irradiation (Wang et al., 1985). The evolution of precipitate microstructure during implantation has been examined for Sb in Al (Kant et al., 1979). Figure 6-10 shows two differing microstructures of oriented, cubic AlSb precipitates in an Al matrix obtained with dark-field TEM; both were produced by the same Sb fluence
260
6 Ion Implantation and Ion-Beam Mixing
Figure 6-10. Dark-field TEM images of AlSb precipitates formed in Al a) implanted at room temperature with 5xlO 1 9 Sb/m 2 , 50 keV, and then annealed at 300 °C, and b) implanted at 300 °C with the same fluence (Kant et al., 1979).
(5xlO 1 9 Sb/m 2 , 50keV). For Fig. 6-10a the implantation was done at room temperature with subsequent annealing at 300 °C to induce precipitation, which gave precipitate diameters ^ 3 nm. The ~ 10 nm precipitates in Fig. 6-10 b resulted from implanting the sample at 300°C. The size is seen to increase greatly when implantation is combined with elevated temperature.
This work indicates that greater precipitate sizes and decreased numbers are promoted by increased implantation temperature and fluence, and decreased ion flux. Increased sizes were also obtained by irradiating the precipitates at 300 °C with Al. These variations of precipitate size were argued to be due to: a) lower nucleation densities for elevated temperatures and lower fluxes, b) point-defect enhanced diffusion of Sb in the implanted layer, and c) dissolution of precipitates by collisions with incoming ions. An exact accounting of the precipitate evolution would require a numerical model with accurate treatment of these effects. Nonetheless, it is clear that radiation-induced processes can play a significant role in determining precipitate microstructure. The atomic displacements produced by irradiation can lead to chemical disorder as well as structural defects. The mixing of two ordered elements between their respective sublattices can transform the alloy to a disordered solid solution. In Cu3Au, TEM examination of alloys irradiated with lowfluences of 100 keV Cu + showed discrete zones of disorder (~10nm) which were identified as individual cascades (Wiedersich, 1985). Irradiation amorphizes some binary ordered phases. In a study of 16 phases irradiated with 2.5 MeV Ni + , most of the 10 compounds which became amorphous have narrow (< 2 at.%) equilibrium composition ranges (Brimhall et al., 1983 and 1984). The narrow range indicates that the free energy of the phase increases rapidly at compositions away from the central value. Irradiation produces local atomic arrangements like those away from the central composition and results in a higher free energy, which favors amorphous phase formation. The compounds which remained crystalline have composition
6.2 Basic Processes
ranges >10at.%. The compounds which became amorphous did so at damage levels < 1 dpa, while the others remained crystalline at > 10 dpa. It is notable that amorphization of NiTi (B2 structure) proceeded more slowly with ion fluence than expected for direct quenching within cascades to an amorphous structure. Overlapping cascades were inferred to be needed for amorphization. The fluence dependence and correlation with narrow composition range are thought to indicate that a buildup of disorder to a critical defect density is required for amorphization. Binary or higher-component alloys are generally required for amorphous phases, but pure a-Ga amorphizes when irradiated with 2xlO 1 8 Ar/m 2 (275 keV) at < 1 0 K (Holz et al., 1983). The low temperature reduces atomic mobility, which is thought to prevent amorphous zones from crystallizing epitaxially on the surrounding a crystals. Irradiating metastable /?-Ga did not form an amorphous phase, apparently because the local order of /? is similar to that of the liquid, which allows the zones to crystallize more readily. Irradiating a-Ga with He to the same damage level did not amorphize it, which indicates that dense cascades like those with Ne or Ar are needed. The Ga results are an example of the initial structure influencing the final phase of ion-beam alloys. To date, Ga is the only pure metal reported to be amorphized by irradiation.
time product, D t, which increases with ion fluence. The value of D t is generally found to be temperature independent below a critical temperature, Tc, but thermally activated above Tc. Both intervals have received considerable attention. First, the low-temperature interval is understood with the thermal spike concept, and a phenomenological model is used to account for results from a wide range of metals (Johnson et al., 1985). The mixing rates are well described by:
where <> / is the fluence, Q is the average atomic density of the elements, AHcoh is the cohesive energy of the alloy, AHmix is the heat of mixing of the elements evaluated at 50 at.% using Miedema's approach (1976), and K1 = 0.037 and K2 = 27 are constants evaluated from observed rates. The prefactor is derived with a thermal spike model (Vineyard, 1976) by using the energy deposited per unit length and the activation energy for jumps within cascades, which is proportional to AHcoh. The quadratic dependence on AHcoh was verified with binary combinations having AHmix~0 (van Rossum etal., 1985). The term in parentheses in Eq. (6-3) involving AHmix is a chemical bias on the diffusion, analogous to the Darken term used to correct thermal interdiffusion for this effect (Shewmon, 1963): D'= D (1-2 AHmJkBT)
6.2.6 Ion-Beam Mixing The rate at which two elements are alloyed by ion-beam mixing is measured by using ion backscattering to determine the concentration profiles across the alloyed zone. The profiles are fitted to an interdiffusion profile, and the alloyed thickness yields an effective value for the diffusivity-
261
(6-4)
The linear dependence on AHmix was verified for D t of transition elements with similar masses mixed into Au and Pt (Cheng etal., 1984). This term indicates that the mixing will be slower for binary combinations which react endothermically (AHmix >0), while exothermic systems will mix more rapidly. The dependence on AHmix
262
6 Ion Implantation and Ion-Beam Mixing
and the value for K2 allow the effective thermal energy for mixing in the cascade to be evaluated using Eq. (6-4): /cBTeff = l-2eV. This energy is consistent with diffusion occurring primarily during the thermal spike regime, and is cited as evidence for the applicability of that model (Johnson et al, 1985). Secondly, the increase of Dt with temperature above Tc is widely taken to be radiation-enhanced diffusion (Matteson et al., 1979), where mobile point defects in the alloy cause increased interdiffusion. The values of Tc have been shown to correlate with AHcoh, which has been justified on empirical grounds (Cheng et al., 1986; Cheng, 1989). The activation energy in this regime is generally observed to be less than that of thermal diffusion by a vacancy mechanism; reduction by 1/2 is expected for radiation enhanced diffusion. At sufficiently high temperatures, conventional thermal diffusion overcomes the radiationenhanced diffusion. Recently, discrepancies between the mixing of Ni/Zr layers above Tc and radiation-enhanced diffusion in large-grain Ni specimens have been noted (Rehn and Okamoto, 1989). This comparison indicates that Tc is too low, the activation energy is not 1/2 that for thermal diffusion, and the mixing dose-rate dependence is uncertain. It was shown that in deposited Au/Zr and Ni/Zr layers, thermal interdiffusion also exhibits a regime with a lower activation energy, which could be present in the ion-beam mixing experiments as well. In summary, the lattice damage created by ion beams is responsible for ion-beam mixing. Thermal spike models account quantitatively for the rate of mixing at low temperatures. An enhanced rate is seen at elevated temperature, for which some detailed questions remain. Notably, me-
tastable alloys (like amorphous phases) are more readily formed in the high-temperature, thermal-spike regime, which governs the reaction of the layers at low temperatures. Irradiating above Tc, where point defects are mobile, is more likely to produce phases which are thermally stable at that temperature (Sood, 1982; Johnson etal., 1985).
6.3 Metastable Phase Formation The types of metastable phases formed with ion beams are discussed in this section. Extensions of elemental phases are considered first, followed by precipitation within such phases. The formation of metastable crystal phases is then examined, and finally phases with non-crystalline atomic order are discussed. Examples are given for each type, some of which may fit into more than one category. Since implanted species usually remain localized, the materials are not uniform in composition and in that sense are always metastable. However, here we examine the phases present in the implanted layer and assume that a local equilibrium exists (Myers and Rack, 1978); the equilibrium is metastable if the phases are not those of equilibrium alloys at the implanted composition. The section ends by discussing the prototype system Al-Ni, which has been thoroughly studied across the composition range and exhibits many of the types of metastable phases. Reviews of metastable alloys include Poate and Cullis, 1980, Follstaedt, 1985 b, and Lilienfeld etal., 1987. 6.3.1 Metastable Solid Solutions Much of the early work on ion-implanted metals used ion channeling to ex-
6.3 Metastable Phase Formation
amine the degree to which implanted species substitute on matrix lattice sites, and has been reviewed (Poate and Cullis, 1980; Sood, 1982). Many elements implanted at room temperature into Fe, Cu and Ni to ~ 1 at.% occupy lattice sites ( / s >0.5), including insoluble ones (Sood, 1978), but fewer elements occupy lattice sites in Al (Sood and Dearnaley, 1976). Substitutionality was found to extend to solutes outside the Hume-Rothery limits: findings for Cu and Fe indicate that the allowable difference in atomic radius should be increased from ±15% to - 1 5 % - + 4 0 % to include more oversized solutes, and that of electronegativity increased from ±0.4 to ±0.7. These extensions clearly indicate that solutions are being formed with concentrations beyond the equilibrium solid solubilities (c0). For instance, 1 at.%W implanted into Cu is almost completely substitutional (Borders and Poate, 1976), but equilibrium Cu alloys do not contain W in solution, even in the liquid state (Hansen and Anderko, 1958). The more recent studies discussed in Sec. 6.2.4 (Turos et al., 1987) examined the effect of temperature on the channeling properties of implanted elements, and allow the substitutionality to be understood in terms of defect mobility and solute-defect interactions. The mobility of vacancies in Al at room temperature was earlier suggested to be responsible for the more restricted solubility range of solutes in Al (Sood, 1982). Electronegativity correlates with the heat of mixing used to discuss the later findings, and atomic size also influences the binding of point defects. The recent work also indicates that further extension to oversized atoms can be achieved by implanting at low temperatures where vacancies are immobile. Identifying the solution composition with the substitutional concentration is
263
rather restrictive. The usual approach is to identify the largest concentration for which a second phase did not form. This approach gives larger solution concentrations which may include defected solute configurations, and is used for ion-beam mixed alloys. Metastable solid solutions have been formed by mixing f.c.c. metals of well-characterized binary systems, including Cu-Ag, Co-Au and Ni-Ag (Mayer etal, 1981). The multilayer configuration of Fig. 6-6 b was used to examine alloys spanning the binary range by irradiating with 300 keV Xe + at room temperature or near 77 K. The Cu-Ag system is a well-known exception to the Hume-Rothery rules, which predict solubility for it, while Co-Au is predicted not to form solid solutions. However, ion-beam mixing at room temperature formed single-phase f.c.c. solutions in both systems for all compositions, with lattice constants which nearly obey Vegard's law for ideal solutions. Mixing near 77 K produced solid solutions of C u Ag, but amorphous alloys were formed with Co-Au. For the immiscible system Ag-Ni, the elemental phases were retained during mixing at all temperatures. However, 12 at.% Ag can be implanted into Ni with/ s ~ 0.5 (Buene et al., 1981). Thus, ionbeam mixing can form metastable solutions of higher concentration than implantation, but forms single-phase alloys of immiscible metals less readily, apparently because the equilibrium phases are present initially. 6.3.2 Precipitated Phases Insoluble inert gases have been implanted into metals because of their relevance to fission and fusion reactor technology, and to identify microstructures which might result from ions retained in alloy
264
6 Ion Implantation and Ion-Beam Mixing
layers during ion-beam mixing. Gas bubbles of He were studied first (Jager and Roth, 1981); subsequently, the heavier elements were discovered to form f.c.c. precipitates (vom Felde et al., 1984; Templier etal, 1984). The microstructures of Kr precipitates in Ni have been well characterized (Birtcher and Liu, 1989). The electron diffraction pattern from a specimen implanted with 2 x 1020 Kr/m 2 in Fig. 6-11 shows the intense spots of a [100] pattern of Ni, with superimposed weaker spots due to the Kr. The alignment of the two sets of reflections indicates that the cubic axes of the two phases are parallel, as seen for other f.c.c. metals. The alignment of precipitate lattice directions with matrix directions is common for precipitation in implanted alloys. The differing radial positions indicate that the lattice constants of the two phases are mismatched, by up to 55%. Implanting at room temperature produces precipitates a few nanometers in diameter. Annealing or implanting at higher temperatures pro-
#
•
duces larger precipitates, which at 500 °C grow more readily and develop facets; however, this growth coincides with melting of the f.c.c. Kr lattice, which is also observed during temperature cycling in the TEM. The inert gases are known to form crystals with van der Waals bonding at high pressures and low temperatures; closepacked structures are theoretically preferred, and f.c.c. is observed (Kittel, 1971). The lighter elements require increased pressures and lower temperatures to form solids. The cavities in the metal exert a pressure oc 1/r on the inert element. As precipitates grow, the internal pressures decrease and the solid structure cannot be maintained. The largest solid Kr particles observed in Ni are 6 nm, which corresponds to 1 GPa pressure (10 kbar). This value just exceeds the minimum needed to solidify bulk Kr at room temperature, 0.8 GPa (Birtcher and Liu, 1989). Although the equilibrium structure of solid inert gases is generally believed to be f.c.c, Kr implantation into h.c.p. Ti produced h.c.p. precipitates (Evans and Mazey, 1986). Apparently, the two structures are sufficiently close in energy (Kittel, 1971) that the metal matrix can impose its structure on the crystal. However, implantation of Kr into b.c.c. Mo (Evans and Mazey, 1985) and Xe into b.c.c. Fe (Templier et al., 1986) did not produce b.c.c. precipitates. In addition, the metal matrix can impose an ordering of precipitates (or gas bubbles) into regular positions, termed a bubble lattice (Mazey and Evans, 1986; Jager and Roth, 1981). 6.3.3 Metastable Crystalline Phases
Figure 6-11. Electron diffraction pattern from Ni implanted with 2 x 1020 Kr/m 2 showing intense Ni reflections and weak reflections from f.c.c. Kr precipitates (Birtcher and Liu, 1989).
If equilibrium phases are unable to form or are not stable under ion irradiation, other crystalline phases may form instead,
6.3 Metastable Phase Formation
such as hexagonal £-Fe2C seen in Fig. 6-5. This phase is not metastable cementite (Fe3C), which is common in bulk alloys (Hansen and Anderko, 1958); cementite is not expected due to its complicated crystal structure (Hohmuth et al., 1983). The e phase has a simple h.c.p. arrangement of Fe atoms with C atoms on octahedral interstitial sites (Pearson, 1958, 1967). The phase is less stable than cementite and transforms to it, but forms and persists in implanted alloys. The s lattice is aligned with the b.c.c. Fe matrix (Follstaedt, 1985b): (1210) || (111) (plane in Fig. 6-5) [0001] || [110] and [1010] || [112] (directions in the plane). The s phase is often referred to as Fe 2 C, but it exists over a wide composition range by varying the amount of interstitial C. The sputtering coefficient of C on Fe is low, which allows concentrations up to 80 at.% C to be implanted (Follstaedt and Knapp, 1986 a). Only s is seen for ^ 50 at.% C, while a diffuse ring is also observed at higher concentrations. This limit is readily accounted for by noting that there is one octahedral C site for each Fe atom. This result extends the e phase to higher C concentrations than previously reported. Implanting high concentrations of N into Fe or stainless steels produces the analogous phase (Rauschenbach and Kolitsch, 1983; Yost et al., 1983). The blistering observed on the surface of 304 stainless steel implanted with > 50 at.% N (Antilla et al., 1985) is then explained by N 2 bubble formation after all interstitial sites in the nitride have been filled (Follstaedt and Knapp, 1986 a). Ion-beam mixing has produced phases which were not previously reported, and are apparently metastable. Layers (Fig.
265
6-6 b) of Al-Mn were mixed at 30 and 150°C to form alloys with 21-44 at.% Mn, and a b.c.c. phase (a0 = 0.519 nm) labeled the F phase was identified (Knapp and Follstaedt, 1987 a). A b.c.c. phase (a0 = 0.382 nm) was formed in Al-V with 1526 at.% V (Karpe et al., 1989), and a f.c.c. phase (a0 = 0.423 nm) was found near Ti 70 Co 20 Ni 10 (Lilienfeld and Borgesen, 1990). These examples demonstrate that ion-beam mixing can not only form known phases, but also new structures not observed and perhaps not accessible with other alloying methods. 6.3.4 Amorphous Phases
One class of amorphous metals is formed by implanting metalloids to obtain phases with the same glass structures as produced by melt quenching (e.g., Ni(P), Ali et al., 1977). However, in many metals implanted with metalloids, crystalline compounds form instead. A systematic understanding of these results was achieved by noting that the crystalline phases form for smaller metalloid atoms, and relating the results to a relation for crystal-structure of metal-metalloid compounds (Hohmuth etal, 1983). The Hagg rule (1933) states that such compounds will have simple crystal structures if the ratio (metalloid/metal) of atomic radii is < 0.59 so that the metalloid can occupy interstitial sites in a close-packed arrangement of metal atoms. Larger metalloids form complex crystal structures instead. Figure 6-12 plots the structures observed in implanted alloys by the atomic radii. A slope of 0.59 is seen to separate the crystalline and amorphous alloys. Such interstitial compounds were argued to nucleate readily in implanted alloys because of their simple, close-packed structures, but complex phases cannot form and amorphous phases form instead.
266
6 Ion Implantation and Ion-Beam Mixing Cu Nj Co Fe MoPt Al hAu V Ta
= 0.861 <=
crystalline
0.12
-As amorphous
to
4O
o
-Si
0
-P
\R- 0.59|—'S
0.10
y
Q O
B
5 o.oe
y
hi
• C
*
crystalline
y
-N
y 0.06
y 0.11
0.13 METAL
0.15 ATOMIC
Figure 6-12. Diagram summarizing observations of amorphous (o, +) and crystalline ( —) alloys in metals implanted with metalloids, shown as a function of the two atomic radii (Hohmuth and Rauschenbach, 1983).
0.17 RADIUS
The Fe-C system (above) is a borderline case (# = 0.60) for which the interstitial compound e-Fe2C is observed. An amorphous phase forms for N i - C (# = 0.62) at low temperatures, but the hexagonal compound is observed at room temperature (Nastasi et al., 1988). Thus lattice temperature can decide between the two microstructures in borderline cases. Similar results are found for Fe-B (Hirano and Miyake, 1988). Both ion implantation and ion-beam mixing can form metal-metal amorphous alloys, but the latter has the advantage of allowing entire binary composition ranges to be investigated. Recently, the glass-forming ranges have been examined for systems of early and late transition elements (Bottiger et al., 1989; Andersen et al., 1990), and compared with those of other techniques. Shown in Fig. 6-13 for Ni-Ti are (a) the equilibrium phase diagram, (b) a map indicating the microstructures observed at the compositions and mixing
[nmj
temperatures investigated, and (c) a bar graph showing glass-forming ranges for other techniques. Ion-beam mixing forms amorphous phases at central compositions, even at moderate temperatures (<600°C). The range extends to ~70at.%Ni and is similar to that of the other techniques. This range includes the B2 phase NiTi, which is expected since this phase is amorphized by ion irradiation (Brimhall et al., 1983 and 1984). Some general results for these systems are notable (Bottiger et al., 1989). The glass-forming ranges are different for ionbeam mixing and melt quenching. In some cases, that of melt quenching is larger, even though the effective quench rate inferred for thermal spikes (~ 1012 K/s) is much faster. The ion-beam mixed alloys often have twophases, amorphous and elemental. These features indicate that cascades do not simply quench to amorphous zones; some diffusion and phase separation occur. The elemental phases present in the initial lay-
6.3 Metastable Phase Formation
1600 1500 (a)
-\ 1\
UOO •-
\
/I
\\ fV/\v \
11
nnn
,-
-
I
/
1\ \
1 /
/1
/
I
aD a
a
a
700
a
I
Ii
1
a
DO D
y
o
/
o
500
o -
o
g Z.00
.
.
.
.
300 200
_
1 1
600 (b)
o
o
L O
o
\
—
-
\
100 -
0
o i
i
i
i
ii
1
1
MQ VQ
(c)
MA i
i
i
i
I I
20
267
60
1
80
i
Figure 6-13. a) Ni-Ti phase diagram (after Hansen and Anderko, 1958). b) Phases observed after ion-beam mixing with 500 keV Xe at the plotted temperature: •, amorphous; o, f.c.c; n, unidentified crystalline, c) Glassforming ranges for other techniques: solid bar - amorphous; open bar - crystalline; hatched bar - crystalline + amorphous. Techniques: MQ - melt spun; VQ - sputter deposited; MA - ball milled (Bottiger et al., 1989).
100
at. % Ni
ers surround the cascades and can probably regrow into the spike zone without having to nucleate. In general, vapor and melt quenching give wider glass-forming ranges than ion-beam mixing. To account for the composition limits in systems with a negative heat of mixing, the free energy of an undercooled liquid was calculated and used for the amorphous phase, and the common tangent with the free-energy curve of the elemental phase was used to obtain the minimum concentration of that element in the amorphous phase (Andersen et al., 1990). An effective temperature of 700 K was used, which is not well justified, but the results are not very sensitive to it. This approach gives reasonable agreement with experiment:
typically, the theoretical value is within 5 at.% for an observed 20 at.%, except for Fe-Zr and Ni-Ti. Amorphous phases can also be formed in binary systems with small, positive heats of mixing (< +15 kJ/mole), but the calculations fail for such alloys. Amorphous phase formation has been examined in many systems, and it is useful to deduce rules based upon known alloy properties to predict their formation, or equivalently, to identify when competing crystalline phases form instead. Many rules have been suggested, including those based upon atomic sizes (Hohmuth et al., 1983), simplicity of crystalline structure (Hung et al., 1983), structures of the elemental phases (Liu et al., 1983), heats of mixing (Alonzo and Simozar, 1983;
268
6 Ion Implantation and Ion-Beam Mixing
Rauschenbach and Hohmuth, 1982), solid solubilities (Liu, 1987), or chemical disorder (Luzzi and Meshii, 1986). All of them address some aspect of phase formation and are successful to some extent. For instance, elements with the same crystal structure can form a continuous solid solution (like Cu-Ag) and are thus less likely to form amorphous phases (Liu et al., 1983). Also, compounds with simple structures like B2 (Hung et al., 1983) are expected to nucleate more readily; however, this does not guarantee their survival during irradiation (Brimhall et al., 1984). Exceptions are found to each of these rules, and at present, there is no generally accepted method that accurately predicts which phase will form in ion-beam alloys (B0ttiger et al., 1989). However, all the rules taken together provide a useful set of guidelines. There is also no definitive answer as to how amorphous phases form in ion-beam alloys. Initially, lattice damage was thought to build up with increasing fluence and be stabilized by ion-implanted species (Sood, 1982). However, the reasonable agreement with thermodynamic calculations for amorphous composition ranges supports models in which the amorphous state is quenched within the cascade, within the limitations noted above (Andersen etal., 1990). Radiation studies (Brimhall et al., 1983 and 1984; Luzzi and Meshii, 1986) indicate that compounds whose free energy increases with disordering will be amorphized. The apparent need for multiple, overlapping ion cascades may indicate that disordering must occur before amorphization can take place. Recent examinations of the elastic properties of irradiated compounds indicate that chemical disorder produces an instability for lattice shear, in close analogy to that observed for metals near their melting point (Rehn et al., 1987). The irradiation studies indicate
the concentrations at which amorphous phases are likely to form in ion-beam alloys, but a complete understanding of amorphization during ion implantation or ion-beam mixing is not yet available (see also Vol. 9, Chap. 10, Sec. 10.2.4). 6.3.5 Icosahedral Phases
Soon after its discovery (Shechtman et al., 1984), the icosahedral phase of AlMn was found to be formed as fine grains (< 30 nm) in layers ion-beam mixed at elevated temperature (150°C), as seen in Fig. 6-2 (Knapp and Follstaedt, 1985: Lilienfeld et al., 1985). Electron diffraction showed up to 18 sharp rings (Knapp and Follstaedt, 1986) which matched the atomic spacings found by high-resolution x-ray diffraction, thus indicating well-developed icosahedral order in the grains. Producing fine-grained layers on a sapphire substrate was a key step in determining the melting point of an icosahedral phase for the first time (Knapp and Follstaedt, 1987 b). Other transition elements mixed with Al also formed icosahedral phases, including 3d metals from V to Co (Lilienfeld et al., 1986). The icosahedral symmetry of the alloys is directly evident in the microdiffraction pattern in Fig. 6-14 taken from an Al 85 Cr 15 grain along a five-fold symmetry axis (Lilienfeld, 1990). Alloys mixed at 30 °C have only a few very diffuse diffraction rings and show no features >1 nm; they were thus identified as amorphous (Follstaedt and Knapp, 1986 b). This dependence of the icosahedral order and grain size on lattice temperature indicates that the structures observed at 150°C do not form in collision cascades. Instead, the ordered grains develop by diffusion in the alloy. This interpretation suggested that the phase might also form during solid-state interdiffusion of Al/Mn
6.3 Metastable Phase Formation
Figure 6-14. Five-fold microdiffraction pattern from Al 85 Cr 15 icosahedral grain formed by ion-beam mixing at 150°C. Kindly provided by D.A. Lilienfeld, 1990.
269
(Budai and Aziz, 1986). By using an Al single crystal, high-resolution x-ray diffraction patterns could be obtained from single orientations of the phase, whereas only random patterns could be examined with melt-quenched specimens. This technique allowed slight deviations from exact symmetry to be measured, and indicated that structural models based upon twinning of crystalline phases are incorrect (Budai et al., 1987). The grain sizes observed with ion-beam mixing also indicate that such twinning would have to occur on a very fine scale not possible for crystals with large unit cells. Thus, the icosahedral alloys formed by ion-beam alloying provided critical information on the novel atomic order of these phases.
6.3.6 Al-Ni Ion-Beam Alloys layers. The phase does form, but is not as well ordered; the diffraction patterns are not as sharp and weaker reflections are not observed (Follstaedt and Knapp, 1986 c). Formation in the solid state gave a new perspective on icosahedral phases, which previously had been formed only by melt quenching and whose structure was thought to be related to that of liquids. In addition, an equilibration of Mn between the icosahedral phase and f.c.c. Al could be inferred to occur during irradiation, which allowed the minimum Mn content of icosahedral Al-Mn to be identified, 15 ± 1 at.% (Follstaedt and Knapp, 1988). Recent work indicates that melt-quenched structures of Al-Mn which were initially identified as amorphous are in fact extremely fine grains with some icosahedral order (Chen and Spaepen, 1988). Further work is needed to determine the nature of such structures in ion-beam alloys. Ion implanting Mn into Al was found to produce oriented icosahedral precipitates
The Al-Ni system illustrates several features of metastable ion-beam alloys. The equilibrium phase diagrams (Hansen and Anderko, 1958; Singleton et al., 1986) contain four ordered phases of varying structural complexity, composition and composition width, which remain ordered up their decomposition temperatures; the later version also contains a fifth phase, Al 3 Ni 5 , found in more recent work. Alloys have been formed with several ion-beam methods, and low sputtering coefficients allow large composition ranges to be studied by ion implantation as well as ionbeam mixing. The phases observed for these treatments are summarized in Table 6-1, which includes results obtained at room temperature, 500 °C and cryogenic temperatures. The phase boundaries of the equilibrium compounds at room temperature are also given. The same phases are found in Al-rich alloys at room temperature for each method.
270
6 Ion Implantation and Ion-Beam Mixing
Table 6-1. Phases of ion-beam alloys of Al-Ni. Phase V Comp. Temp./ Method 20 °C Irradiate Ni->A1 Al->Ni Ion mix Cryogenic Irradiate Ion mix 500 °C Al->Ni
Al-f.c.c. 0-0.02%
Al3Ni 25.0%
Al 3 Ni 2 37-41%
AINi 45-59%
f.c.c. f.c.c.
amor amor amor amor
AINi AINi AINi
AINi AINi AINi AINi
amor amor
amor
amor amor
f.c.c.
h.c.p. f.c.c.
AINi
AINi 3
f.c.c.
Al 3 Ni 5 64-68%
AINi 3 73-75%
Ni-f.c.c. 93-100%
f.c.c, h.c.p.
f.c.c.
h.c.p. h.c.p.
f.c.c.
f.c.c.
f.c.c. f.c.c.
Al 3 Ni 2
Equilibrium phases, with composition in atomic percent.
At room temperature, the complex phase Al3Ni (DO 20 ) is fully amorphized by Xe + and Ni + irradiations, but is only partially transformed by the lighter ion, Ne + (Nastasi et al., 1986; Brimhall et al., 1984). Irradiation transforms Al 3 Ni 2 (D5 13 ) to the structure of AINi (/?') (Nastasi et al., 1985), which is stable as expected from its wide composition range. The structures of these two phases are related. In equilibrium, AINi extends to 42 at.% Ni (Hansen and Anderko, 1958). Its ordered b.c.c. structure (B2, like CsCl) accommodates substoichiometry with vacancies on the Ni sublattice. At 40 at.% Ni, the vacancies order on regularly spaced {111} planes of the b.c.c. lattice to form Al 3 Ni 2 (Bradley and Taylor, 1937). Ion irradiation disorders the vacancies, producing the B2 structure at this lower, metastable concentration (Nastasi et al., 1985). The AlNi 3 (/) phase has an ordered Ll 2 structure, in which Al atoms occupy one fourth the sites in an f.c.c. lattice. Ion irradiation initially disorders it to an f.c.c. solution (Nelson et al., 1972), but continued irradiation produces a two-phase alloy of f.c.c. and h.c.p.
phases, which have nearly the same free energies (Eridon et al., 1988). As discussed in Sec. 6.1, pure Ni (f.c.c.) transforms to h.c.p. during irradiation (Johnson et al., 1979). Implantation of Ni into Al at room temperature forms a metastable solution with 0.2 at.% Ni on the f.c.c. lattice sites (Picraux et al., 1980). At higher concentrations, amorphous zones form in the crystalline matrix, until at 20-25 at.%Ni the alloy is completely transformed (Follstaedt and Picraux, 1984). At 32 at.% Ni, AINi precipitates also form and were found to contain 36.5 + 2.5 at.%Ni (Follstaedt and Romig, 1985). This concentration is below the minimum equilibrium composition of AINi, and is consistent with the Al 3 Ni 2 ->AlNi transformation discussed above. Up to 75 at.% Al has been implanted into Ni, and the phases have been determined (Ahmed and Potter, 1985). A metastable f.c.c. solution forms first and extends to 30 at.% Al; increasing the Al to 32±2 at.% Al forms a h.c.p. phase. This h.c.p. phase forms at the composition of the orthorhombic compound Al 3 Ni 5 , but
6.3 Metastable Phase Formation
the relation (if any) between the two phases is not known. These Ni(Al) phases differ slightly from the irradiation results, where the h.c.p. phase is seen for pure Ni; however, the f.c.c. solution is expected from the irradiation of AlNi3. The AlNi phase forms for 40 to 60 at.% Al (40 at.% Ni), while at 75at.%Al (25 at.% Ni) the amorphous phase forms. Thus Ni(Al) and Al(Ni) implanted alloys have the same Al-rich phases. Ion beam mixing at room temperature was done for 0 to 50 at.% Ni (Hung et al., 1983), and f.c.c. Al, the amorphous phase at 25 at.% Ni, and AlNi were again observed. AlNi was observed at the composition of Al 3 Ni 2 , but some amorphous material was also present. Considering all the roomtemperature results, the same phases are found at the same compositions, except that the metastable h.c.p. phase is not seen in Ni implanted with low Al concentrations. Low Al contents appear to stabilize the f.c.c. phase, while at 32 at.% Al the h.c.p. phase forms. The agreement of phases in Al-rich alloys and the observation of two-phase alloys at intermediate compositions suggests that equilibria may be established between those phases during irradiation. The stability of several Al-Ni phases has been investigated at cryogenic temperatures where atomic mobilities are reduced and metastable phases are more likely. Irradiating Al3Ni at 100 K produced the amorphous phase as expected from the room temperature result (Nastasi et al., 1986). At 77 K, Al 3 Ni 2 transformed to NiAl, but further transformations are observed at 20 K (Nastasi, 1990): the phase first transformed to ordered AlNi (B2), then to disordered b.c.c. (A2), and finally to an amorphous phase. It is notable that chemical disordering of this phase after 0.2 dpa first produced the crystalline b.c.c.
271
phase, and an additional 0.6 dpa was needed for structural disorder and the amorphous phase. Irradiation also amorphized AlNi at 90 K (Thome et al., 1987). These low-temperature irradiation results agree with those for ion-beam mixing at 77 K, which also indicate that Ni-rich f.c.c. solutions are stable (Jaouen et al., 1987). The ordered alloy Ni3Al disorders to f.c.c. when irradiated at 80 K; the h.c.p. phase was not observed (Eridon et al.). Implantation of Al into Ni at 500 °C produces a microstructural effect not discussed previously (Ahmed and Potter, 1987). Instead of the high concentrations of Al obtained with high fluences (2.4 x 1022 Al/m 2 , 180 keV) at 25 °C, which extend to a depth of ~0.3 jim, the Al profile for 500°C is a plateau at ~25 at.% which extends much deeper. Similar results are obtained by implanting at 25 °C and annealing at 650 °C, for which the profiles extend as deep as 0.8 |im. Microstructure examination and the temperature dependence of Al diffusion during annealing indicate that AlNi3 forms and grows into the Ni substrate, which recrystallizes and provides grain-boundary paths for continued Al diffusion to greater depths. The recrystallization is critical for transporting Al to such depths since bulk diffusion (both thermal and radiation-enhanced) is insignificant at these temperatures. Thus at 500 °C, the f.c.c. and h.c.p. phases are replaced by the ordered AlNi3 phase. At lower Ni content (< 1.4 xlO 2 2 Al/m 2 , <40at.%Al), recrystallization does not occur and the Al remains at the implanted depth; AlNi and Al 3 Ni 2 are then observed in their equilibrium ranges. A structure which might have formed in Al-Ni alloys is an icosahedral phase, since this phase forms in Al with the transition elements V through Co (Lilienfeld et al., 1986). However, the phase apparently does
272
6 Ion Implantation and Ion-Beam Mixing
not form in Al-Ni. It was not reported for ion-beam mixing at 100 °C with 23 at.% Ni (Nastasi et al., 1986), and was not seen when amorphous Al implanted with 20 at.% Ni was annealed up to 500°C in the TEM (Follstaedt, 1990).
6.4 Basic Alloy Phenomena In this category of studies, the techniques of ion implantation, ion-beam analysis and TEM are used together to form and characterize a microscopic alloy designed to exhibit a physical effect of interest. In many cases, the thermal evolution of the system is monitored to measure alloy parameters of the implanted solute, such as its solid solubility, diffusion coefficient, and binding energy to traps, as illustrated by the examples on Fe-Sb and F e - H e - H . Ion-beam studies of solute phenomena have been reviewed by Myers, 1978, and Myers et al., 1987. In the final example, a unique, high-concentration, metastable palladium hydride phase was produced by implanting deuterium (D) to determine whether D-D nuclear reactions occur within it. 6.4.1 Solubility and Diffusion of Sb in Fe
Antimony is one of several metalloids which segregate to grain boundaries and cause loss of fracture toughness, termed temper embrittlement. The properties of these solutes at embrittling temperatures (650-850 K) can be difficult to obtain by conventional metallurgical techniques because of the relatively slow diffusion rates. However, ion-beam analysis has the resolution needed to explore Sb solubilities and diffusivities in Fe on a sub-micron scale in ion-implanted alloys (Myers and Rack, 1978). Furthermore, by implanting solutes directly into the metal, possible difficulties
of surface diffusion barriers, which can occur in conventional diffusion couples, are avoided. Pure Fe was annealed at 1130 K to remove residual lattice damage and to produce large-grained specimens (~ 0.1 mm) in which grain-boundary diffusion was not significant. The Fe was implanted with 6 x 1020 Sb/m 2 at 200 keV. After implantation at room temperature, the near-surface region was found by Rutherford backscattering to contain up to 12 at.% Sb. The alloy was probably highly disordered. Annealing for 0.5 hours at 873 K in vacuum produced the concentration profile shown in Fig. 6-15. The Sb profile still exhibits a peak, but its amplitude is now significantly reduced owing to Sb diffusion into the Fe substrate and Sb evaporation at the surface. The profile shows a break in slope between the peak and the low-amplitude tail extending into the substrate which occurs at 2.6 at.% Sb and a depth of ~ 0.05 |im. A second break at the same concentration is seen near the front surface. Annealing implanted specimens at 773, 823 or 873 K caused the diffusion tails to grow into the bulk and transport Sb out of the near-surface layer; however, the near-surface peak and the break in slope were always observed. The alloy layer of Fig. 6-15 was found with TEM to contain ~ 10 nm precipitates of the equilibrium phase, FeSb, in a b.c.c. Fe matrix. The concentration tail is due to diffusion of Sb in solution into the b.c.c. Fe substrate. At their intersection, the solution and two-phase regions are in equilibrium with each other since the anneals are sufficient to produce diffusion over distances Df~0.2|im. The concentration at their intersection is then the solid solubility, Co = 2.6 at.% for 873 K (±20% normalized error). A smaller value, C o = 2.05 at.%, was found at 773 K. These
6.4 Basic Alloy Phenomena
273
Sb - IMPLANTED Fe 0.5 HOUR AT 873 K
Figure 6-15. Concentration profile of 6xl0 2 0 Sb/m 2 , 200 keV, after annealing 0.5 hr. at 873 K (Myers and Rack, 1978). The areal density of atoms used for a depth scale with RBS was converted to the distance interval shown in the figure by using the atomic density of Fe. d(AT./cm 2 HO 17
values are in good agreement with extrapolations of solubilities obtained at higher temperatures. Concentration profiles were measured for several anneal times and were fit with numerical solutions to diffusion equations which accounted for precipitate dissolution and Sb evaporation at the surface. The fitted diffusion coefficients include 3.9 x 10~ 2 i m 2 /s (factor of 2 accuracy) for 773 K. This value is ~10~ 5 times lower than those obtained by mechanical sectioning. The measured diffusion coefficients were found to be ~ 10 times smaller than extrapolations from high temperature would predict. However, the measured values successfully explain the rate of Sb segregation during temper embrittlement, and the solid solubility was consistent with the level of segregation observed. Several experimental features allowed these solubility and diffusion data to be obtained. First, the depth resolution of the ion beam analysis (~ 11 nm) was sufficient to measure the Sb diffusion into the bulk and to locate accurately the break in the concentration profiles. Second, the precipitation occurs with a high density which
gives small particles (~10nm); the time needed to form them is ~10~ 2 times shorter than that required for significant diffusion into the bulk (~100nm). Thus precipitation occurs first, and dissolutionsupplied transport into the bulk at the solubility concentration is subsequent and slower. Finally, to measure an equilibrium solid solubility, the equilibrium phase must precipitate, as confirmed here by TEM. Solubilities with respect to metastable phases should also be obtainable when such phases precipitate. Similar studies of solubility and diffusion have been done with Be alloys (Myers, 1978). 6.4.2 Deuterium Trapping at He Bubbles in Ni
A similar approach is used to demonstrate and quantify solute trapping effects, such as D trapping at He bubbles in Ni (Besenbacher et al., 1982). In this study, 4 x 1020 He/m 2 was implanted first at lOkeV and room temperature. The He precipitates as bubbles ~ 1 nm in diameter as observed with TEM, and is then immobile at the temperatures used in this
274
6 Ion Implantation and Ion-Beam Mixing
study. The specimen was cooled to 100 K and implanted with 1.25 x 1020 D/m 2 at 10 keV. Figure 6-16 a shows the depth profiles for He and D calculated with TRIM (Ziegler et al., 1985), as well as the depth variation of the cross section for the nuclear reaction D( 3 He,p) 4 He. The cross section is constant over the implanted depths, but decreases significantly at depths beyond 0.5 jum. The proton yield thus measures the amount of D remaining in the implanted layer.
IMPLANTATION PROFILES IN Ni
4 7 b|3
0.0
0.05
0.10 DEPTH (/xm)
a) 1 0
[W |» 0 *
0.20
*'
•
ooo , 0.8
0.15
o
\ D IN He-IMPLANTED Ni \ THEORY: \ 0.55 eV He TRAPS"
o oo
\
o NO H e ^ ^
o
\ > WITH
O 0
\ \#
o
o
o
X)
^OCo 1
250
300
350 TEMPERATURE(K)
\
\ \
•
•
•
400
b) Figure 6-16. a) Calculated implantation profiles for 4 x 1O20 He/m2, 10 keV, and 1.25 x 1020 D/m2, 10 keV, and nuclear cross section for D( 3 He,p) 4 He as functions of depth in Ni. b) Fraction of D retained in traps during ramping of temperature at 2 K/min. (Myers et al., 1989).
In Fig. 6-16 b, the fraction of D remaining at the implanted depth is plotted versus temperature during warming of the specimen at 2 K/min. Also plotted are results for the same experiment but with no He. With He, D migrates from the implanted zone at 370 K, which is clearly higher in temperature than found when no He is present (310 K). This shift demonstrates the trapping of D at the He bubbles. In fact, the release temperature in the absence of He is higher than that at which D would leave due to diffusion alone, 200 K, which indicates that D is also trapped (but more weakly) by lattice damage created during the implantation. Nickel is well suited to these experiments because its surface is known to be quite permeable to hydrogen isotopes; thus when D is thermally released from the traps, it can diffuse to the surface and leave the specimen. The transport of D out of the implanted zone is treated by numerically solving a set of coupled differential equations which treat the binding to He traps, diffusion in the matrix, and release at the surface. The binding enthalpy is adjusted to fit the observed release, as shown by the theoretical curve through the data with 0.55 eV trap strength. This binding enthalpy and that of D to He in other metals agree with values expected for chemisorption on the walls of the He bubbles (Myers et al., 1989). The chemically inert He is insoluble in metals and forms cavities whose surfaces chemisorb hydrogen like those of a bare metal. The surfaces are attractive sites for D because of their low electron density. The binding of D to vacancies in metals has also been extensively characterized using similar methods. This trapping and other solute reactions in metals determined with ion-beam methods are reviewed by Myers etal. (1987 and 1989).
6.5 Surface Alloys with Improved Macroscopic Properties
6.4.3 Search for Cold Fusion: Pd-D In 1989, it was reported that electrochemical charging of Pd electrodes with D was able to induce D-D fusion, even at a level capable of significant heat output (Fleischmann and Pons, 1989). The reported work linked high densities of D to fusion; D/Pd ratios >1.0 were desirable. Ion implantation of D into Pd was identified as a method to explore high D concentrations based on previous work which achieved a D/Pd ratio of 1.3 + 0.2 by implanting at 35 K where D is immobile (Moller, 1982). The implantation produces a hydride phase, with D ordered in octahedral sites of an f.c.c. Pd lattice at D/Pd = 1.0 (Traverse and Bernas, 1987); electrochemical charging also forms the hydride, but with D/Pd ^ 0.9 (Knapp et al., 1990). In the recent investigation (Myers et al., 1990), D was implanted at 10 keV to 3.9 x 1022 D/m 2 at 41 K. The amount of D retained in the specimen was measured with the D(d,p)T (T-triton) reaction. The yield from the reaction was accounted for with a mathematical model which used the implanted depth profile from TRIM (Ziegler et al., 1987), the nuclear cross section variation with depth in the sample, and a saturated D concentration of D/Pd = 1.6±0.3. After implantation, the specimen was maintained at 41 K and the charged particle detector left on to count protons and tritons from possible fusion events. After 8 hours, only one count was observed against an expected background level of 1.4 counts. This negligible yield places a limit on the possible fusion rate of ^ 2 xlO~ 2 1 events/D-s. Fusion was also not observed at 81 K, nor during warming to room temperature. Similar work with Zr and Ti placed comparable limits on their fusion rates. This limit on the fusion rate in Pd is not as low as that of other work (e.g., ^ 10~ 24
275
events/D-s; Knapp et al., 1990), but the experiment is unique in examining such high D/Pd ratios. The D in excess of D/Pd = 1.0 could also be identified during warming of the specimen by its release at 120 K, as compared with the release of the remaining D at ~ 200 K. This superstoichiometric D is thought to occupy tetrahedral sites and to diffuse more rapidly due to its avoiding the tightly binding octahedral sites, which are filled for D/Pd > 1.0 (Richards, 1989).
6.5 Surface Alloys with Improved Macroscopic Properties This section examines ion-beam alloys with improved surface-related macroscopic properties. Two general areas being widely examined are mechanical properties and corrosion resistance. Important examples of alloys with reduced friction and/or wear, increased hardness and strength, reduced aqueous corrosion rates or with combined reductions in corrosion and wear rates are discussed below. These and other surface-related properties modified with ion beams have been reviewed (Herman, 1981; Picraux, 1984; Clayton, 1987). Benefits observed in both areas are promoting the commercial ion-beam treatment of components made of several steels and alloys of Be, Ti, Zr, Co, or Ni (Armini, 1986; Dearnaley, 1987; Sioshansi, 1987). Ion implantation has several attractive features for use on engineering components. The treatment can be applied to finished products with no change in dimensions. It modifies only a surface layer and not the bulk alloy. Components can be implanted at room temperature to avoid changing the mechanical properties of the substrate, unlike some coatings which require elevated temperatures. The implant-
276
6 Ion Implantation and Ion-Beam Mixing
ed layer is formed by injecting atoms beneath the surface and is thus more adherent than a coating. However, there are also negative aspects: implantation is usually a line-of-sight process, forms a very thin alloy, and becomes more limited by sputter erosion for glancing incident angles. The process is also relatively expensive. Treated components are therefore usually highercost, precision parts which must closely retain their dimensions in mild-wear or corrosive environments. 6.5.1 Nitrogen Implantation to Reduce Wear of Steels
The implantation of N is widely used to reduce wear rates for a variety of steels and components (Picraux, 1984; Hirvonen, 1984). In general, N implantation is believed to harden the surface and thus increase wear resistance, but in some cases other mechanisms are important. The phases observed in N-implanted Fe and steels form a basis to interpret the wear behavior. The phases of N-implanted pure Fe have been identified with TEM and Mossbauer spectroscopy (Rauschenbach and Kolitsch, 1983; Moncoffre, 1987). For concentrations up to 11 at.%, N stabilizes the austenitic y phase of Fe (f.c.c), which replaces the ferritic a (b.c.c.) phase. With increased N content, a' martensite (b.c.t.) and oc"-Fe16N2 (b.c.t.) form, while at the highest fluences, e-Fe2N (hex) forms. The e phase was found to contain 25-33 at.% N, which is in the concentration range often used to improve mechanical properties. For Fe, the implanted concentration saturates at 33 at.% N (Singer, 1984b). The microstructures of N-implanted steels are similar to those of Fe but differences in depth profiles and compositions result from other elements in the steels. In stainless steels with >10at.%Cr, the N
profiles have the expected gaussian shape (Singer, 1984 b), but in Fe and low-alloy steels, N can migrate toward the surface, especially if the temperature rises (> 50 °C) during implantation (Moncoffre, 1987). This difference is believed due binding of N to Cr, which immobilizes it, whereas N can redistribute during implantation in the absence of Cr. Nitrogen also stabilizes austenite (Carbucicchio et al., 1981) and forms nitrides at high concentrations, including the e phase. However, the phases may be carbonitrides which incorporate C from the steel into their lattice structures (dos Santos et al., 1983). The reduction in wear obtained with N implantation is generally thought to result from its strengthening of the metal near the wearing surface. Several mechanisms have been considered, but the observed microstructures noted above and other experimental comparisons identify two as being applicable (Hubler, 1982): 1) strengthening by interstitial N, which can be expected to accumulate at dislocations and impede their motion, and 2) strengthening by nitride precipitates, which inhibit dislocation motion. Furthermore, at high N concentrations, the surface layer may also be almost completely converted into the hard nitride phase. Thus the benefits are believed to be directly related to phases containing N, as opposed to lattice damage or compressive stresses which would also be present when other elements are implanted. Type 304 stainless steel has a metastable f.c.c. structure which transforms to b.c.c. when the surface is mechanically polished (Singer et al., 1988) or undergoes sliding wear (Follstaedt et al., 1983), and requires special consideration. The transformed surface layer is brittle, and breaks away during sliding wear. Implanting N at low concentrations stabilizes austenite, which
6.5 Surface Alloys with Improved Macroscopic Properties
results in a softer surface that wears more quickly during abrasive testing (Singer et al., 1988), but nonetheless reduces sliding wear (Fayeulle and Treheux, 1987). High concentrations (45 at.%) can be implanted into 304 to form a layer with predominantly £-(Fe,Cr)2N; reduced sliding wear is also observed with this treatment (Yostetal., 1983). In summary, wear reductions are observed by many workers for N implantation into steels. It is notable, however, than benefits are not obtained for fully hardened bearing steels (Carosella et al., 1980; Pope etal., 1984). The benefits for softer steels are generally believed to result from strengthening the surface. However, questions remain about one aspect of the treatment: reduced wear rates are observed at depths beyond the implanted thickness (Lo Russo et al., 1979). It was initially proposed that N migrates deeper into the steel ahead of the wearing surface (Hartley, 1979), but evidence for such migration is disputed (Singer et al., 1984 a). It has also been proposed that N implantation initiates a mild, oxidative wear mode between the contacting parts, which persists even after the implanted layer is worn away (Hale et al., 1987). Although extended wear reduction is not fully understood, its existence seems not to be in question. 6.5.2 Implantation of Ti + C to Reduce Friction and Wear
A second treatment used to improve the tribological properties of steels is the implantation of Ti and C at the same depth. This treatment has some significant advantages over the more widely used implantation of N (Follstaedt, 1985 a). However, Ti is more difficult to implant, and its heavier mass (Fig. 6-8) results in thinner surface alloys. One advantage is that Ti + C im-
277
plantation reduces both friction and wear, whereas N generally reduces only wear. Secondly, the Ti + C treatment provides these reductions for all steels examined to date, whereas N provides no benefits in hard bearing steels like 52100 and 440 C. Implantation of Ti + C is being evaluated for treatment of 440 C ball bearings in the U.S. Space Shuttle's main rocket engines (Ng and Naerheim, 1987). The results for 440 C with Ti + C are illustrated in Fig. 6-17, which compares (a) maximum wear depths (MWD) and (b) friction coefficients for pin-on-disk tests of implanted disks with those of unimplanted disks (Pope et al., 1984). At the top of Fig. 6-17 a, the calculated maximum Hertzian stresses exerted on the disk are given for the pin loads on the horizontal axis. Reduced friction and wear are found to persist for 1000 cycles at loads as high as 600 g and Hertzian stresses up to 1.5 times the yield strength of 440 C steel at maximum hardness (1840 MPa). For 304 steel, benefits are obtained even at 3.5 times the yield stress (Follstaedt, 1985 a). In addition to improving the performance of steels, Ti + C reduces friction and wear of Co alloys (Dillich et al., 1984), and such benefits are expected for Ni alloys (Follstaedt et al., 1989 b). These benefits have been closely linked to the amorphous phase which forms when Fe is implanted with Ti and C. All the steels are amorphized, and the amorphous layer has been observed intact across wear tracks (Follstaedt et al., 1984 c). The amorphous phase in Fe implanted with Ti and C is a ternary phase for typical implanted concentrations; e.g., for 20 at.% Ti, 4 at.% C is required to form the phase (Knapp etal., 1985). The benefits and amorphous phases were observed in steels with > 20 at.% of both Ti and C, which are within the ternary composition limits. For
278
6 Ion Implantation and Ion-Beam Mixing
HERTZIAN STRESS (MPa) 1850 2300 2650 2900 3150 1.00 0.50
0.05 A,T -IMPLANTEp 200 a)
400 600 800 NORMAL LOAD (g)
1000
1.0
0.8
h
W
FFl
o
LJJ
0.6
ION
o o
0.4
hO
£
0.2 ^•-IMPLANTED
200
b)
400 600 800 NORMAL LOAD (g)
1000
Figure 6-17. a) Average maximum wear depth (MWD) and b) friction coefficient at the end of 1000 cycle pin-on-disk test of unimplanted 440 C and implanted with 2xlO 2 1 Ti/m 2 , 180-90 keV, plus 2x x 1021 C/m2, 30 keV, plotted versus load on 440 C pin (Pope et al., 1984).
high Ti concentrations, small precipitates (10-20 nm) of mixed Ti + Cr carbides are found within the amorphous layer on stainless steels (Follstaedt et al., 1989 a) and may contribute to improved performance.
Two recent studies indicate the essential roles played by both Ti and C in providing extended tribological benefits. At high Ti concentrations (50at.%), binary amorphous Fe-Ti alloys can be formed by ionbeam mixing (Hirvonen et al., 1986), as expected from irradiation studies (Brimhall et al., 1984). The friction and wear behavior of amorphous Fe-Ti on 304 was compared to that of the same alloy implanted with a high fluence of C (3 x 10 21 C/m 2 , 50 keV), and the C was found to significantly increase the number of cycles with low friction (Hirvonen et al., 1987 and 1990). Second, the implantation of C into stainless steels has recently been found to produce an amorphous alloy, in which Cr acts like Ti to stabilize the amorphous phase instead of the £-Fe2C phase that forms in pure Fe (Follstaedt et al., 1989 c). Benefits are observed with C alone, but do not persist for as many cycles and are not observed at the high pin loads as with Ti + C. These two comparisons indicate that tribological performance is improved somewhat for amorphous layers with either Ti alone or C alone, but extended benefits require both elements. Comparisons with other amorphous alloys indicate that the Ti + C treatment gives superior performance (Follstaedt et al., 1989 b). Current research is aimed at identifying how the amorphous Fe(Ti,C) layer reduces friction and wear, and two factors appear to be important. First, the phase is harder than the steel substrates. Hardness was inferred from earlier work (Follstaedt, 1985 a), but has now been directly observed for the layer on 440 C steel with indentation at depths of ~100nm (Bourcier, 1990). Second, oxidation of the wear surface appears to be important in determining friction levels and wear rates (Pope et al., 1988; Fayuelle and Singer, 1989).
6.5 Surface Alloys with Improved Macroscopic Properties
279
6.5.3 High-Strength A1(O) Alloys
Implantation of O into Al has recently been shown to produce hard surface layers with very high flow stresses (Bourcier et al., 1990 a and 1990 b). In this work, O was implanted at five energies (200-25 keV) to give a nearly constant composition of 20 at.% O extending to a depth of 500 nm. An ultra-low load indenter was used to obtain penetration-versus-load curves at depths ^lOOnrn. The curves were fitted with numerical simulations obtained by large-strain, finite-element modeling of the deformations. The simulations took account of the indenter shape, the compressive yield strength of the pure Al substrate (41 MPa) and the higher flow stress of the implanted layer, which was adjusted to obtain agreement with experiment. The penetration depths observed in the implanted specimen were only ~ 1/12 of those in Al at the same loads, and required a flow stress of 2900 + 400 MPa for the layer. A smaller but still quite high value, 1200 + 400 MPa, was found after annealing the alloy 1/2 hour at 550 °C. The flow stress of the asimplanted alloy exceeds the yield strength of fully hardened 440 C bearing steel, 1840 MPa, and is several times those of the strongest commercial Al alloys. The microstructure of the as-implanted alloy is shown in Fig. 6-18. Electron diffraction patterns must be tilted off the zone axes in order to detect low-intensity, diffuse spheres surrounding each (sharp) Al matrix spot, as seen in Fig. 6-18 a. Darkfield imaging with a diffuse sphere shows a high density of fine precipitates 1.5-3.5 nm in diameter, as seen in Fig. 6-18 b. Annealed alloys have also been examined (Myers and Follstaedt, 1988; Bourcier et al., 1990 a), and their microstructures are useful for understanding the crystal structure of the precipitates in the as-implanted
Figure 6-18. a) Diffraction pattern of Al implanted with 20 at.% O, tilted off the <110> zone axis, b) dark-field image of oxide precipitates imaged with the diffuse reflection circled in a) (Bourcier et al., 1990b).
alloy. Annealing 1/2 hour at 550°C produced larger (4-10 nm diameter) precipitates of y-Al2O3, a cubic spinel phase. The cubic axes of y-Al2O3 align parallel to those of the f.c.c. Al matrix, and a spinel reflection almost overlaps each Al reflection. The phase has a f.c.c. sublattice of O~ 2 ions, with Al + 3 ions ordered on {111} planes. This ordering doubles the sublattice cell size to 0.790 nm, and gives addi-
280
6 Ion Implantation and Ion-Beam Mixing
tional reflections which are isolated from those of Al and readily observed. These ordering reflections are absent for the asimplanted alloy, implying that its precipitates are disordered y-Al2O3 with Al + 3 ions at random interstitial sites in an f.c.c. O ~2 lattice of nearly the same spacing as AL The precipitate sizes and coherency have been used in well established models of hardening, with the assumption that all of the O was precipitated (Bourcier et al., 1990 a and b). The unusually high strengths are well accounted for, even though the models are being applied at very high precipitate densities and volume fractions (20%). The higher strength of the as-implanted alloy is due to the smaller size and correspondingly higher precipitate density. Aluminum has previously been found to be hardened by implanting O or N (Ohira and Iwaki, 1987). The above studies extend earlier work by quantifying the strength of the layers and accounting for it with conventional hardening mechanisms. The numerical modeling is an essential part of the evaluation; the increase in strength to 80 times that of Al is not fully reflected in the reduced indentation depths because the soft substrate influences deformation, even for penetrations less than 1/5 of the implanted layer thickness. 6.5.4 Aqueous Corrosion of Fe-Based Alloys
The use of ion implantation to reduce the corrosion rates of Fe and its alloys in acidic or chloride solutions has been examined by many workers, and is summarized in recent reviews (Clayton, 1987 and 1989). Anodic polarization of metals in these solutions is often used to examine the formation of passivating surface layers which reduce further dissolution of the metal, and also to measure the degree of protection
against pitting corrosion offered by the passivating layer. Increased corrosion resistance has been obtained by implanting Cr, Cr + Mo, Cr + P, Ta, N, P or B, but in some cases improvements depend upon the composition of the bulk alloy. Ion-beam mixing is also being used to form corrosion-resistant surface layers. Increased corrosion resistance is widely observed for Cr implantation of Fe, and the concentration dependence agrees with that observed for bulk Fe-Cr alloys. Implantation thus creates a surface layer with the corrosion resistance of a stainless steel. The effect of Cr concentration was more pronounced than any other effects of the implantation, such as those due to lattice damage. These benefits are also found with Cr for bearing steels (Wang et al., 1979) and dual implantations of Cr + Mo or Cr + P are even more protective, again in agreement with bulk alloy results. The increased protection against pitting corrosion for these two steels has potential for application to jet engine bearings, which corrode during operation at sea as well as during shelf storage. The benefits obtained with implantation of bearing steels led the U.S. Navy to initiate a program to adapt the treatment to the balls and races of bearings (Smidt et al., 1987). Facilities have been developed for rotating and heat-sinking these curved pieces during implantation while mitigating the effects of sputtering. The cost of treatment on a production scale was projected to be $ 82.50/bearing. An additional consideration favoring surface alloying is that costly or strategic elements like Cr which are needed only to improve surface properties can be used in much smaller quantities instead of being distributed throughout the bulk. Some of the more corrosion-resistant Fe-based alloys are metallic glasses con-
6.5 Surface Alloys with Improved Macroscopic Properties
taining both Cr and P. Sorenson et al. (1986, 1988) have used P implantation of bulk Fe-Cr alloys to form passive surface alloys and to study the effects of P content and alloy structure on corrosion behavior. The results with Fe 6Cr, Fe-lOCr and Fe-18Cr illustrate the mechanisms responsible for improved electrochemical properties when P and Cr are present together. The alloys were implanted with 100 keV P to form layers - 1 0 0 nm thick. Corrosion currents were examined for 600 mV polarization in 0.1 N H 2 SO 4 solutions with 500 ppm Cl. The implantation of P decreased the corrosion current of Fe-lOCr, increased the current of Fe-6Cr and increased pitting, and had essentially no effect on Fe-18Cr. These seemingly contradictory results are understood with current models of passi vating film formation. The Fe-6Cr alloy has marginal Cr for forming a protective passivating layer. The P is believed to enhance the Cr dissolution rate and speed the formation of the layer for solutions without Cl, but in more aggressive Cl solutions, it also increases the rate of pitting. The Cr content of Fe-18Cr alloys is sufficient to form passivating layers readily, and P has little effect for this alloy, which is already very corrosion resistant. Implantation of P improves the intermediate alloy, Fe-lOCr, because it has enough Cr to form a protective layer, and Cr dissolution from the metal into that layer is enhanced by the P. Examining several P fluences indicates that maximum benefits are achieved at high concentrations (20-30 at.% P), where the alloy is transformed from the crystalline b.c.c. phase to an amorphous phase. At these P concentrations, a thicker passivating layer is formed which reduces the corrosion current by ~ 10 ~ 4 relative to the bulk Fe-lOCr alloy to produce behavior like that of Fe-18Cr.
281
The P-implantation results are the first to show implantation-enhanced passivity of low Cr Fe-based alloys. The studies use ion implantation to alter composition near the surface in a controlled manner and to form an amorphous surface alloy. The mechanisms identified are applicable to bulk alloys, and these studies thus also fall in the category of Sec. 6.4. 6.5.5 Reduced Corrosion and WearofTi-6AI-4V The implantation of N also significantly improves the mechanical properties of Ti alloys. Substantial reductions in friction and wear of the engineering alloy Ti-6 Al4 V (weight percentages) have been found (Oliver et al., 1984): the friction coefficient was reduced from 0.48 to 0.15, and the wear volume by a factor of more than 100. The strong chemical reaction of Ti and N produces TiN precipitates in N-implanted Ti-6A1-4V (Vardiman and Kant, 1982); improved fatigue properties were also observed with the treatment. Titanium alloys are generally thought to wear poorly, and N implantation is thus of interest for treating such components. One of the more successful commercial applications of ion beams is N implantation of hip and knee replacement joints made of Ti-6A1-4V, which are treated on a production basis (Sioshansi, 1987). The prostheses use a Ti component, for instance, to replace the ball of the femur, which moves against a mating component made of ultrahigh molecular weight polyethylene (UHMWPE). The Ti alloy was chosen for its biocompatibility and bulk mechanical properties; N is also an acceptable element for components placed in the body. Basic research investigating the use of N implantation for this wear couple has been done by Williams and Buchanan
282
6 Ion Implantation and Ion-Beam Mixing
(1985). They devised an electrochemical cell with a cylindrical Ti-6A1-4V electrode rotating between two UHMWPE pads pressing against it. The electrolyte is a 0.9%NaCl solution with 10% bovine serum added to simulate body fluids. The corrosion currents observed during the sliding wear with the electrode at — lOOmV anodic potential are shown in Fig. 6-19 for unimplanted and 20 at.% Nimplanted electrodes. Despite scatter, the data separate clearly and show that the corrosion current is reduced more than a factor of 100 with N implantation. Examination of the sliding surfaces indicates that wear is also reduced by the treatment (Williams, 1985). Unimplanted electrodes are heavily blackened and scored by the rotating contact with the pads, while implanted ones have a smooth, mirror finish like that before the test. Profilometry of ORNL-DWG 8 3 - 4 8 3 8 2
10
Figure 6-19. Corrosion current vs. time for unimplanted and N-implanted (20 at.%) Ti-6A1-4V electrodes. I M. Williams and R. A. Buchanan, Ion Implantation of Surgical Ti-6Al-4V Alloy, Materials Science and Engineering 69 (1985) 237-246, Figure 2.
the worn electrodes shows that the maximum depths of wear grooves are reduced from 30-80 jim to ^ 1 Jim. Moreover, the UHMWPE pads also show reduced wear when the electrode is implanted. The Ti-6A1-4V and UHMWPE elements of the prostheses are eroded by a combination of mechanical wear and chemical corrosion. Nitrogen implantation reduces both mechanisms in this example.
6.6 Concluding Remarks Many aspects of surface-alloy formation by ion implantation and ion-beam mixing are relatively well understood, as are the basic atomic processes occurring during these treatments. The microstructures of new alloys can be interpreted against a broad background of accepted results. Frequently, the initial materials are transformed into metastable solids, including supersaturated solutions, disordered crystalline phases, metallic glasses and quasicrystals. Systematic trends are found for the occurrence of these phases and have been formulated into predictive guidelines. In this chapter, additional guidance for predicting phases was obtained by examining ion-irradiated alloys. The microstructures of irradiated alloys are also of interest for radiation damage studies, for instance, to determine the evolution of alloys used in reactors. By combining the special capabilities of ion implantation with the depth resolution of ion beam analysis (often ~ 0.01 ^m), submicron-scale alloys can be tailored to exhibit solute phenomena, such as trapping, precipitation and dissolution, and diffusion. These processes can be quantified to provide basic information of interest for bulk alloy phenomena at temperatures lower than may be otherwise inves-
6.8 References
tigated. The methods presented in Sec. 6.4 can be extended to other alloys, for instance, to study solute trapping at dislocations, or to obtain solid solubilities for phase diagrams, including ternary systems (Myers, 1978). The ability of ion-beam treatments to improve tribological and electrochemical properties has led to their being applied to commercial products. A number of service companies in the U.S.A., the U.K. and Europe now offer ion-beam treatments for engineering components, and the industry can be described as steadily growing. Commercial acceptance of ion-beam treatments depends upon technical merit and economic considerations with respect to competing treatments. In general, the processes are being accepted for specialized applications, such as treatment of medical components placed in the body. Two recent developments appear important for the future of ion-beam alloying. First, ion beams are being combined with vapor deposition in a process termed ionbeam assisted deposition (IBAD) (Hubler, 1989). Implantation is carried out during deposition to produce a layer which is more adherent and can have new properties derived from the ion beam. Second, a new technique termed plasma-source ion implantation (PSII) is being developed for gaseous species (Conrad, 1989). A plasma is formed around the target, which is biased to accelerate the ions into it. This method injects ions into the target from all directions and at normal incidence to curved surfaces. The sputtering of curved surfaces is thus reduced, and the process is stated to be substantially less expensive. Both developments are likely to find commercial uses. This chapter has focussed on metallic alloys formed by implanting into metal substrates or by ion-beam mixing two
283
metals. In addition to the well-known uses of ion beams with semiconductors, other materials are being modified. Ion-beam treatment of ceramics is now being examined (White etal., 1989) for improvement of surface-mechanical properties and optoelectronic properties. Polymers are also being treated to form new materials with different optical and electronic properties (Davenas et al, 1989). These examples demonstrate that ion-beam treatments are not limited to metal and semiconductor applications, but find uses in the full range of modern materials.
6.7 Acknowledgements The author wishes to thank his many colleagues who have contributed to this chapter through valuable discussions of their work or by collaborating with him. This work was supported by the U.S. Department of Energy under contract number DE-AC04-76DP00789.
6.8 References Ahmed, M., Potter, D.L (1985), Acta Met. 33, 2221. Ahmed, M., Potter, D.I. (1987), Acta Met. 35, 2341; Mat. Sci. Eng. 90, 127. Ali, A., Grant, W.A., Grundy, P. X (1977), Rad. Eff. 34, 215; also (1978), Phil. Mag. B37, 353. Alonzo, J. A., Simozar, S. (1983), Sol. State Commun. 48, 765. Andersen, L.-U.A., Bottiger, X, Dyrbye, K. (1990), Mat. Res. Soc. Symp. Proc. 157, 187; also, Nucl. Inst. Meth. B51, 125. Antilla, A., Keinonen, X, Uhrmacher, M., Vahvaselka, S. (1985), /. Appl. Phys. 57, 1423. Armini, A. (1986), Industrial Heating (January), p. 17. Atwater, H. A., Thompson, C. V., Smith, H.I. (1988), /. Appl. Phys. 64, 2337. Bentley, X, Stephenson, L.D., Benson, R.B. Jr., Parrish, P. A., Hirvonen, XK. (1984), Mat. Res. Soc. Symp. Proc. 27, 151. Besenbacher, R, Bottiger, X, Myers, S.M. (1982), J. Appl. Phys. 53, 3547.
284
6 Ion Implantation and Ion-Beam Mixing
Biersack, J.P., Haggmark, L.G. (1980), Nucl. Inst. Meth. 174, 257. Birtcher, R.C., Liu, A.S. (1989), /. Nucl. Mat. 165, 101. Borders, J. A., Cullis, A. G., Poate, J. M. (1976), Inst. Phys. Conf. Ser. 28, 204. Borders, J.A., Poate, J.M. (1976), Phys. Rev. BIS, 969. Bottiger, I, Dyrbye, K., Pampus, K., Poulsen, R. (1989), Phil. Mag. A 59, 569. Bourcier, R.J. (1990), private communication. Bourcier, R.J., Myers, S. M., Polonis, D.H. (1990a), Nucl. Inst. Meth. B44, 278. Bourcier, R. X, Follstaedt, D. M., Myers, S. M., Polonis, D. H. (1990 b), Mat. Res. Soc. Symp. Proc. 157, 801. Bradley, A. I, Taylor, A. (1937), Proc. R. Soc. London 159, 554; Phil. Mag. 29, 1175. Brimhall, J. L., Kissinger, H. E., Chariot, L.A. (1983), Rad. Eff 77, 273. Brimhall, J. L., Kissinger, H. E., Pelton, A. R. (1984), Mat. Res. Soc. Symp. Proc. 27, 163. Budai, J. D., Aziz, M. J. (1986), Phys. Rev. B33, 2876. Budai, J. D., Tischler, J. Z., Habenschuss, A., Ice, G.E. (1987), Phys. Rev. Lett. 58, 2304. Buene, L., Jacobson, D.C., Nakahara, S., Poate, J.M., Draper, C.W., Hirvonen, I K . (1981), Mat. Res. Soc. Symp. Proc. 1, 583. Bunker, S.N., Armini, A.J. (1989), Nucl. Inst. Meth. B39, 1. Carbucicchio, M., Bardani, L., Tosto, S. (1981), /. Appl. Phys. 52, 4589. Carosella, C. A., Singer, I. L., Bowers, R.C., Gossett, C. R. (1980), in: Ion Implantation Metallurgy: Preece, C M . , Hirvonen, J. K. (Eds.). Warrendale, PA: TMS-AIME, p. 103. Chen, L.C., Spaepen, F. (1988), Nature 336, 366. Cheng, Y.-T., van Rossum, M., Nicolet, M.-A., Johnson, W.L. (1984), Appl. Phys. Lett. 45, 185. Cheng, Y.-T., Zhao, X.-A., Banwell, T, Workman, T.W., Nicolet, M.-A., Johnson, W.L. (1986), /. Appl. Phys. 60, 2615. Cheng, Y-T. (1989), Phys. Rev. B40, 7403. Clayton, C. R. (1987), in: Surface Alloying by Ion, Electron and Laser Beams: Rehn, L. E., Picraux, S.T., Wiedersich, H. (Eds.). Metals Park, Ohio: American Society for Metals, p. 325. Clayton, C. R. (1989), in: Environmental Degradation of Ion and Laser Beam Treated Surfaces: Was, G.S., Grabowski, K. S. (Eds.). Warrendale, PA: The Minerals, Metals & Materials Society, p. 33. Conrad, J. R. (1989), Mat. Sci. Eng. A116, 197. Cullis, A. G., Poate, J. M., Borders, J. A. (1976), Appl. Phys. Lett. 28, 316. Davenas, I, Xu, X. L., Boiteux, G., Sage, D. (1989), Nucl. Inst. Meth. B39, 754. Dearnaley, G. (1987), Nucl. Inst. Meth. B 24/25, 506. Dillich, S. A., Bolster, R. N., Singer, I. L. (1984), Mat. Res. Soc. Symp. Proc. 27, 637.
Doolittle, L.R. (1985), Nucl. Inst. Meth. B9, 344. dos Santos, C.A., Behar, M., de Souza, J. P., Baumvol, I. J. R. (1983), Nucl. Inst. Meth. 209/210, 907. Eridon, J., Was, G. S., Rehn, L. (1988), /. Mater. Res. 3, 626. Evans, J.H., Mazey, D.J. (1985), Scr. Met. 19, 621. Evans, J. H., Mazey, D.J. (1986), /. Nucl. Mat. 138, 176. Farkas, D., Singer, I. L., Rangaswamy, M. (1984), Mat. Res. Soc. Symp. Proc. 27, 609. Fayeulle, S., Treheux, D. (1987), Nucl. Inst. Meth. B19/20, 216. Fayeulle, S., Singer, I. L. (1989), Mat. Sci. Eng. A115, 285. vom Felde, A., Fink, J., Muller-Heinzerling, Th., Pfliiger, X, Scheerer, B., Linker, G. (1984), Phys. Rev. Lett. 53, 922. Feldman, L.C., Mayer, XW, Picraux, S.T (1982), Materials Analysis by Ion Channeling. New York: Academic Press. Fleischmann, M., Pons, S. (1989), /. Electroanal. Chem. 261, 301. Follstaedt, D.M. (1985 a), Nucl. Inst. Meth. B10/11, 549. Follstaedt, D. M. (1985b), Nucl. Inst. Meth. B7/8,11. Follstaedt, D.M. (1990), previously unpublished results. Follstaedt, D.M., Yost, F.G., Pope, L.E., Picraux, S.T, Knapp, X A. (1983), Appl. Phys. Lett. 43, 358. Follstaedt, D. M., Knapp, X A., Peercy, P. S. (1984a), /. Non-Cryst. Sol. 61 & 62, 451. Follstaedt, D.M., Knapp, J.A., Pope, L.E., Yost, EG., Picraux, S.T. (1984b), Appl. Phys. Lett. 45, 529; erratum: 46, 207. Follstaedt, D.M., Yost, F.G., Pope, L.E. (1984c). Mat. Res. Soc. Symp. Proc. 27, 655. Follstaedt, D.M., Knapp, XA. (1986a), Mat. Res. Soc. Symp. Proc. 51, 473. Follstaedt, D.M., Knapp, XA. (1986b), J. Appl. Phys. 59, 1756. Follstaedt, D.M., Knapp, XA. (1986c), Phys. Rev. Lett. 56, 1827. Follstaedt, D. M., Knapp, X A. (1988), J. Less Comm. Met. 140, 375. Follstaedt, D.M., Knapp, XA., Pope, L.E. (1989a), Nucl. Inst. Meth. B42, 205. Follstaedt, D.M., Knapp, J.A., Pope, L.E. (1989b), Mat. Res. Soc. Symp. Proc. 140, 133. Follstaedt, D.M., Knapp, J.A., Pope, L.E. (1989c), /. Appl. Phys. 66, 2743. Follstaedt, D. M., Picraux, S.T. (1984), in: Alloy Phase Diagrams: Extended Abstracts: Bennett, L.H., Giessen, B.C., Massalski, T.B. (Eds.). Pittsburgh: Materials Research Society, p. 94. Follstaedt, D. M., Romig, A. (1985), in: Microbeam Analysis-1985: Armstrong, XT. (Ed.). San Francisco: San Francisco Press, p. 173. Grant, W.A. (1978), /. Vac. Sci. Technol. 15, 1644.
6.8 References
Hagg, G. (1931), Z. Phys. Chem. B12, 33. Hale, E. B., Reinbold, R., Kahser, R. A. (1987), Mat. Sci. Eng. 90, 273. Hansen, M., Anderko, K. (1958), Constitution of Binary Alloys, 2nd ed. New York: McGraw-Hill. Harrison, D. E., Webb, R. P. (1983), Nucl. Inst. Meth. 218, 727. Hartley, N.E.W. (1979), Thin Solid Films 64, 111. Herman, H. (1981), Nucl. Inst. Meth. 182/183, 887. Hiller, W, Buchgeister, M., Eitner, P., Kopitzki, K., Lilienthal, V., Peiner, E. (1989), Mat. Sci. Eng. A 115, 151. Hirano, M., Miyake, S. (1988), Appi Phys. Lett. 52, 1469. Hirvonen, J.K. (1984), Mat. Res. Soc. Symp. Proc. 27, 621. Hirvonen, J.-P., Nastasi, M., Mayer, J.W. (1986), J. Appl. Phys. 60, 980. Hirvonen, J.-P., Nastasi, M., Mayer, J.W. (1987), Appl. Phys. Lett. 51, 232. Hirvonen, J.-P., Nastasi, M., Zocco, T.G., Jervis, T.R. (1990), J. Appl. Phys. 67, 7292. Hoffmann, B., Baumann, H., Rauch, R, Bethge, K. (1987), Nucl. Inst. Meth. B28, 336. Hohmuth, K., Rauschenbach, B., Kolitsch, A., Richter, E. (1983), Nucl. Inst. Meth. 209/210, 249. Holz, M., Ziemann, P., Buckel, W. (1983), Phys. Rev. Lett. 51, 1584. Hubler, G. K. (1982), Mat. Res. Soc. Symp. Proc. 7, 341. Hubler, G.K. (1989), Mat. Sci. Eng. A115, 181. Hung, L.S., Nastasi, M., Gyulai, J., Mayer, J.W. (1983), Appl. Phys. Lett. 42, 672. Jager, W, Roth, J. (1981), Nucl. Inst. Meth. 182/183, 975. Jaouen, C , Riviere, J.P., Delafond, J. (1987), Nucl. Inst. Meth. B19/20, 549. Johnson, E., Wohlenberg, T., Grant, WA. (1979), Phase Trans. 1, 23. Johnson, E., Sarholt-Kristensen, L., Johansen, A. (1983), Nucl. Inst. Meth. 209/210, 289. Johnson, E. (1990), Mat. Res. Soc. Symp. Proc. 157, 759. Johnson, W L., Cheng, Y. T., Van Rossum, M., Nicolet, M.-A. (1985), Nucl. Inst. Meth. B7/8, 657. Kant, R.A., Myers, S. M., Picraux, S.T. (1979), J. Appl. Phys. 50, 214. Karpe, N., Andersen, L.-U., Dyrbye, K., Bottiger, J., Rao, K.V. (1989), Phys. Rev. B39, 9874. King, WE., Benedek, R. (1983), J. Nucl. Mater. 117, 26. Kittel, C. (1971), Introduction to Solid State Physics, 4th ed. New York: John Wiley. Kloska, M.K., Meyer, O. (1987), Nucl. Inst. Meth. B19/20, 140. Knapp, I A., Follstaedt, D. M. (1985), Phys. Rev. Lett. 55, 1591. Knapp, J.A., Follstaedt, D.M., Doyle, B. L. (1985), Nucl. Inst. Meth. B7/8, 38.
285
Knapp, J. A., Follstaedt, D. M. (1986), Mat. Res. Soc. Symp. Proc. 51, 415. Knapp, J.A., Follstaedt, D. M. (1987a), Nucl. Inst. Meth. B19/20, 611. Knapp, J.A., Follstaedt, D. M. (1987b), Phys. Rev. Lett. 58, 2454. Knapp, J.A., Guilinger, T.R., Kelly, M.J., Doyle, B.L., Walsh, D., Tsao, S.S. (1990), /. Fusion Energy 9, 371. Liau, Z. L., Mayer, J. W (1980), in: Ion Implantation: Hirvonen, J. K. (Ed.), Treatise on Materials Science and Technology, vol. 18: Herman, H. (Ed.). New York: Academic Press, p. 17. Lilienfeld, D.A. (1990), private communication. Lilienfeld, D.A., Nastasi, M., Johnson, H.H., Ast, D.G., Mayer, J.W. (1985), Phys. Rev. Lett. 55, 1587. Lilienfeld, D.A., Nastasi, M., Johnson, H. H., Ast, D.G., Mayer, J.W. (1986), Phys. Rev. B34, 2985. Lilienfeld, D.A., Hung, L.S., Mayer, J.W. (1987), Nucl. Inst. Meth. B19/20, 1. Lilienfeld, D. A., Borgesen, P. (1990), Mat. Res. Soc. Symp. Proc. 157, 789. Liu, B.-X., Johnson, W. L., Nicolet, M.-A., Lau, S. S. (1983), Appl. Phys. Lett. 42, 45; Nucl. Inst. Meth. 209/210, 229. Liu, B.-X. (1987), Mat. Lett. 5, 322. Liu, B.-X., Ma, E., Li, J., Huang, L.J. (1987), Nucl. Inst. Meth. B19/20, 682. Liu, J.C., Mayer, J.W. (1987), Nucl. Inst. Meth. B19/ 20, 538. Lo Russo, S., Mazzoldi, P., Scotoni, I., Tosello, C , Tosto, S. (1979), Appl. Phys. Lett. 34, 627. Luzzi, D. E., Meshii, M. (1986), Scripta Met. 20, 943. Matteson, S., Roth, X, Nicolet, M.-A. (1979), Rad. Eff. 42, 217. Mayer, J.W, Tsaur, B.Y., Lau, S. S., Hung, L.-S. (1981), Nucl. Inst. Meth. 182/183, 1. Mazey, D.J., Evans, J.H. (1986), /. Nucl. Mat. 138, 16. Meissner, J., Kopitzki, K., Mertler, G., Peiner, E. (1987), Nucl. Inst. Meth. B19/20, 669. Meyer, O., Turos, A. (1987), Nucl. Inst. Meth. B19/ 20, 136. Miedema, A.R. (1976), Philips Techn. Rev. 36, 217. Moller, W, Besenbacher, F., Bottiger, J. (1982), Appl. Phys. A 27, 19. Moncoffre, N. (1987), Mat. Sci. Engr. 90, 99. Musket, R.G., Brown, D.W., Hayden, H. C. (1985), Nucl. Inst. Meth. B7/8, 31. Myers, S.M. (1978), /. Vac. Sci. Technol. 15, 1650. Myers, S.M., Rack, H.J. (1978), J. Appl. Phys. 49, 3246. Myers, S. M., Follstaedt, D. M. (1988), J. Appl. Phys. 63, 1942. Myers, S.M., Follstaedt, D. M., Besenbacher, F. (1987), in: Surface Alloying by Ion, Electron and Laser Beams: Rehn, L. E., Picraux, S.T., Wieder-
286
6 Ion Implantation and Ion-Beam Mixing
sich, H. (Eds.). Metals Park, Ohio: American Society for Metals, p. 223. Myers, S. M., Richards, P.M., Wampler, W.R., Besenbacher, F. (1989), /. Nucl. Mat. 165, 9. Myers, S. M., Follstaedt, D. M., Schirber, I E., Richards, P. M. (1990), /. Fusion Energy 9, 263; also, Phys. Rev. B, to be published. Nastasi, M. (1990), /. Less-Corn. Met., to be published. Nastasi, M., Hung, L. S., Johnson, H.H., Mayer, J. W, Williams, J. M. (1985), /. Appl. Phys. 57,1050. Nastasi, M., Johnson, H. H., Mayer, J.W., Williams, J.M. (1986), /. Mat. Res. 1, 268. Nastasi, M., Hirvonen, J.-P., Jervis, T. R., Pharr, G.M., Oliver, W C. (1988), /. Mater. Res. 3, 226. Nelson, R. S., Hudson, X, Mazey, D.X (1972), /. Nucl. Mat. 44, 318. Ng, L., Naerheim, Y. (1987), presented at 1987 Ball Bearing Technical Symposium, Orlando, Florida, 1987. Ohira, S., Iwaki, M. (1987), Mat. Sci. Eng. 90, 143. Oliver, W.C., Hutchings, R., Pethica, J. B., Paradis, E.L., Shuskus, A.J. (1984), Mat. Res. Soc. Symp. Proc. 27, 705. Pavlov, P.V., Zorin, E. I., Tetelbaum, D. I., Lesnikov, V.P., Ryzhkov, G. M., Pavlov, A.V. (1973), Phys. Stat. Sol. 19, 373. Pearson, W B. (1958, 1967), in: Handbook of Lattice Spacings and Structures of Metals, Vol. I, II. New York: Pergamon Press, p. 919, 1339. Picraux, S.T., Follstaedt, D.M., Baeri, P., Campisano, S.U., Foti, G., Rimini, E. (1980), Rad. Eff 49, 75. Picraux, S. T. (1984), Ann. Rev. Mater. Sci. 14, 335. Poate, J. M., Cullis, A. G. (1980), in: Ion Implantation: Hirvonen, J. K. (Ed.), Treatise on Materials Science and Technology, vol. 18: Herman, H. (Ed.). New York: Academic Press, p. 85. Pope, L.E., Yost, F. G., Follstaedt, D. M., Knapp, J.A., Picraux, S. T. (1983), in: Wear of Materials 1983: Ludema, K. C. (Ed.). New York: ASME, p. 280. Pope, L.E., Yost, EG., Follstaedt, D.M., Picraux, S.T., Knapp, J.A. (1984), Mat. Res. Soc. Symp. Proc. 27, 661. Pope, L.E., Knapp, J.A., Follstaedt, D. M. (1988), Surf & Coat. Technol. 36, 361. Rauschenbach, B., Hohmuth, K. (1982), Phys. Stat. Sol. (a) 72, 667. Rauschenbach, B., Kolitsch, A. (1983), Phys. Stat. Sol. (a), 80, 211. Rehn, L. E., Okamoto, P. R., Pearson, I, Bhadra, R., Grimsditch, M. (1987), Phys. Rev. Lett. 59, 2987. Rehn, L. E., Okamoto, P. R. (1989), NucL Inst. Meth. B39, 104. Richards, P.M. (1989), private communication. van Rossum, M., Cheng, Y-T., Nicolet, M.-A., Johnson, WL. (1985), Appl. Phys. Lett. 46, 610. de la Rubia, T.D., Averback, R. S., Benedek, R., King, WE. (1987), Phys. Rev. Lett. 59, 1930.
Schulson, E.M. (1979), /. Nucl. Mat. 83, 239. Shechtman, D., Blech, I., Gratias, D., Cahn, J.W. (1984), Phys. Rev. Lett. 53, 1951. Shewmon, P.G. (1963), Diffusion in Solids. New York: McGraw-Hill, p. 126. Sigmund, P. (1969), Appl. Phys. Lett. 14, 114. Singer, I. L., Carosella, C.A., Reed, J.R. (1981), Nucl. Inst. Meth. 182/183, 923. Singer, I.L., Barlak, T. M (1983), Appl. Phys. Lett. 43, 457. Singer, I. L. (1984 a), Mat. Res. Soc. Symp. Proc. 27, 585. Singer, I. L. (1984b), Vacuum 34, 853. Singer, I.L., Vardiman, R.G., Bolster, R.N. (1988), /. Mater. Res. 3, 1134. Singleton, M. F, Murray, XL., Nash, P. (1986), in: Binary Alloy Phase Diagrams: Murray, J. L., Bennett, L.H., Baker, H. (Eds.). Metals Park, Ohio: American Society for Metals, p. 140. Sioshansi, P. (1987), Mat. Sci. Eng. 90, 373. Smidt, FA., Sartwell, B. D., Bunker, S.N. (1987), Mat. Sci. Eng. 90, 385. Sood, D. K., Dearnaley, G. (1976), in: Applications of Ion Beams to Materials 1975: Carter, G., Colligon, I S . , Grant, W.A. (Eds.). London: Institute of Physics, p. 196. Sood, D.K. (1978), Phys. Lett. 68A, 469. Sood, D.K. (1982), Rad. Eff 63, 141. Sorensen, N. R., Diegle, R. B., Picraux, S.T. (1986), /. Mater. Res. 1, 752. Sorensen, N.R., Diegle, R.B., Picraux, S.T., Nelson, G.C. (1988), Proc. Symp. on Corrosion, Electrochemistry, and Catalysis of Metallic Glasses: Diegle, R.B., Hashimoto, K. (Eds.) Pennington, NJ: The Electrochemical Society, p. 264. Templier, C , Jaouen, C , Riviere, J.-P., Delafond, X, Grilhe, J. (1984), C. R. Acad. Sci., Paris 299, 613. Templier, C , Garem, H., Riviere, X-P. (1986), Phil. Mag. A 53, 661. Thome, L., Jaouen, C , Riviere, J.-P., Delafond, X (1987), Nucl. Inst. Meth. B19/20, 554. Traverse, A., Bernas, H. (1987), /. Less-Comm. Met. 129, 1. Tsaur, B.-Y, Lau, S.S., Hung, L. S., Mayer, XW (1982), Appl. Phys. Lett. 36, 823. Tsaur, B.-Y, Lau, S.S., Hung, L. S., Mayer, XW. (1981), Nucl. Inst. Meth. 182/183, 67. Turos, A., Azzam, A., Kloska, M. K., Meyer, O. (1987), Nucl. Inst. Meth. B19/20, 123. Vardiman, R.G., Kant, R.A. (1982), /. Appl. Phys. 53, 690. Vineyard, G. H. (1976), Rad. Eff 29, 245. Wang, P., Thompson, D.A., Smeltzer, W. W (1985), Nucl. Inst. Meth. B 7/8, 97. Wang, Y.F., Clayton, C. R., Hubler, G. K., Lucke, W H., Hirvonen, X K. (1979), Thin Solid Films 63, 11. Westendorp, J.F.M., Saris, F.W., Koek, B., Viegers, M.P.A., Fenn-Tye, I. (1987), Nucl. Inst. Meth. B26, 539.
6.8 References
White, C. W., McHargue, C. I, Sklad, P. S., Boatner, L. A., Farlow, G. C. (1989), Mat. Sci. Reports 4,41. Wiedersich, H. (1985), Nucl. Inst. Meth. B7/8, 1. Williams, J. M. (1985), Nucl. Inst. Meth. B10/11, 539. Williams, J. M., Buchanan, R. A. (1985), Mat. Sci. Eng. 69, 237. Yost, F.G., Picraux, S.T., Follstaedt, D.M., Pope, L.E., Knapp, J.A. (1983), Thin Solid Films 107, 287. Ziegler, J. E, Biersaek, J. P., Littmark, U. (1985), The Stopping and Range of Ions in Solids. New York: Pergamon Press. Ziegler, I F . (1987), private communication.
General Reading Clayton, C.R. (1989), in: Environmental Degradation of Ion and Laser Beam Treated Surfaces: Was, G.S., Grabowski, K.S. (Eds.). Warrendale, PA: The Minerals, Metals & Materials Society, p. 33.
287
Follstaedt, D.M. (1985), Nucl. Inst. Meth. B7/8, 11. Follstaedt, D.M., Knapp, J.A., Pope, L.E. (1989), Mat. Res. Soc. Symp. Proc. 140, 133. Johnson, W. L., Cheng, Y.-T., van Rossum, M., Nicolet, M.-A. (1985), Nucl. Inst. Meth. B7/8, 657. Meyer, O., Turos, A. (1987), Mat. Sci. Reports 2, p. 371, Myers, S.M., Richards, P.M., Wampler, W.R., Besenbacher, F. (1989), J. Nucl. Mat. 165, 9. Picraux, S.T. (1984), Ann. Rev. Mater. Sci. 14, 335. Poker, D.B., Withrow, S.P. (Eds.) (1991), Proceedings of the Seventh International Conference on Ion Beam Modification of Materials 1990, to be published in: Nuclear Instruments and Methods of Physics Research B. Was, G.S. (1990), Prog. Surface Sci. 32, 211. Ziegler, J.F., Biersaek, I P . , Littmark, U. (1985), The Stopping and Range of Ions in Solids. New York: Pergamon Press.
7 The Epitaxy of Metals Donald W. Pashley Department of Materials, Imperial College of Science, Technology and Medicine, London, U.K.
List of 7.1 7.2 7.3 7.3.1 7.3.1.1 7.3.2 7.4 7.4.1 7.4.1.1 7.4.1.2 7.4.1.3 7.4.2 7.4.2.1 7.4.2.2 7.5 7.5.1 7.5.2 7.5.3 7.5.3.1 7.5.3.2 7.5.3.3 7.6 7.6.1 7.6.2 7.6.2.1 7.6.2.2 7.6.2.3 7.6.3 7.7 7.8
Symbols and Abbreviations Introduction Deposition Methods General Characteristics of Epitaxy Orientation Relationships Deposition Conditions for Obtaining Epitaxy The Role of Lattice Misfit The Modes of Growth of Epitaxial Metal Films The Nucleation Modes of Growth Monolayer Growth The Volmer-Weber Nucleation Mode The Stranski-Krastanow Mode of Growth Post-Nucleation Growth Processes Liquid-Like Coalescence Reorientation and Recrystallization Effects Elastic Strains and Misfit Dislocations Changes in Elastic Strain with Increasing Thickness Misfit Dislocations The Formation of Misfit Dislocations Formation During Monolayer Growth Formation During Volmer-Weber Growth Formation During Stranski-Krastanow Growth Lattice Imperfections in Layers Grown by Epitaxy Imperfection Structures Observed Modes of Formation of Lattice Defects Copying from the Substrate Defects Linked with Misfit Dislocations Defects Resulting from Coalescence of Nuclei Changes in Imperfection Structure as Growth Proceeds Summary and Conclusions References
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
290 291 292 293 293 294 295 296 296 299 301 302 303 303 306 308 309 310 312 312 317 319 320 320 320 320 321 322 325 326 327
290
7 The Epitaxy of Metals
List of Symbols and Abbreviations a a0 b b
K
9 I m n N
spacing of atom (or ion) rows in the substrate surface lattice constant spacing of atom (or ion) in the parallel rows in the deposit total Burgers vector of the dislocation edge component in the interface spacing of the parallel planes in the deposit film spacing of the substrate planes perpendicular to the surface and parallel to the misfit dislocations reciprocal lattice vector length of misfit dislocations lattice misfit in % integer number of threading dislocations spacing of the parallel array of dislocations periodicity of the moire pattern critical thickness growth temperature melting temperature
Ay
surface energy of the deposit film interfacial energy surface energy of the substrate surface energy of materials A and B change in surface energy
2D,3D AES CVD MBE RHEED TEM UHV
two-, three-dimensional Auger electron spectroscopy chemical vapour deposition molecular beam epitaxy reflection high energy electron diffraction transmission electron microscopy ultra-high-vacuum
7 A > 7B
7.1 Introduction
7,1 Introduction The subject of epitaxy has its origins in observations made well over a century ago, when mineralogists became interested in the natural intergrowth of two different mineral species, in such a way that the two had a clear crystallographic orientation relationship. Such observations caused mineralogists to study the growth of one substance on another in the laboratory, mainly by growth from solution. The first known paper on the subject concerned the growth of sodium nitrate on calcite (Frankenheim, 1836). Initially, most observations were made by optical microscopy but once X-ray diffraction techniques became available the scope of such studies was much increased, and by the 1930's many examples of epitaxy had been reported. The term epitaxy was introduced by Royer (1928) to denote the phenomenon, and was derived from the Greek to mean "arrangement on". The scope for studying epitaxy was again increased once electron diffraction techniques became available, allowing extremely thin surface growths of substance B on a single crystal of substance A to be examined for orientation. As a result, it was found that well defined epitaxy occurs for a wide range of combinations of substrate and deposit. There has, over the years, been controversy concerning the correct form of the adjective and adverb to be used in connection with epitaxy, in view of its Greek derivation. Whilst words such as epitactic can be fully justified on this basis, the adjective epitaxial is now so widely used that it is impractical to attempt to bring about any change, even though epitaxial cannot be justified on etymological grounds. Therefore epitaxial is used as the accepted adjectival form in this chapter.
291
A new tool was applied to the study of epitaxy when high quality transmission electron microscopes became available in the 1950's. Detailed studies of the mode of growth and internal structure of epitaxial deposits provided much new knowledge, and theoretical approaches to the understanding of epitaxy could be initiated and developed with much more realism than hitherto. Considerable progress was made towards scientific understanding, resulting from studies which included many on epitaxial growth of metals. Until about 20 years ago most work on epitaxy was carried out as a piece of basic scientific research, but then strong interest developed in the possibility of using epitaxy as a means of making single crystal thin layers of semiconductor materials for use in electronic devices. In the last 15 years, and particularly the last 10 years, considerable effort has been devoted to studies of the epitaxy of semiconductors, and epitaxial growth is now established as a unique processing step in the manufacture of electronic and opto-electronic devices. The demonstration by Esaki and Tsu (1970) that artificial superlattice structures of semiconducting materials can be produced by epitaxial growth has stimulated much new research. Studies of the epitaxial growth of other substances has diminished over the same period. The epitaxy of single and multilayer semiconductor layers is treated in Vol. 4, Chap. 8. Interest in the epitaxy of metals has been revived in recent years by the developments with artificial metallic superlattice structures. These have applications as gratings for soft X-rays, but current interest centres on their potential as magnetic materials or as superconducting materials. A recent review of the properties of artificial metallic superlattices has been given by Jin and Ketterson (1989) (see also
292
7 The Epitaxy of Metals
Vol. 3, Chap. 6). Metallic superlattices are treated in Chapter 8 of this Volume. It is clear that epitaxial growth now has the potential of becoming an important processing route for certain specialised metallic materials, hence the justification for the inclusion of this chapter in a volume on processing. The purpose of this chapter is to provide a summary of the most important aspects of the epitaxial growth process relevant to the growth of single crystal metal films. Inevitably, because of the dominance of work on semiconductor epitaxy during the last decade, some of the recent understanding of epitaxy is based upon such work. Thus it is necessary to refer to some of the semiconductor research, but as far as possible examples are restricted to epitaxial deposits of metals. Further discussion of ultrathin films and superlattices of both metallic and ceramic systems will be found in Vol. 3 of this Series. The first important area of the subject concerns the conditions which have to be satisfied if epitaxial growth is to take place. In practice, experimental conditions can be found to allow many materials to be grown epitaxially on a range of different substrates. Ever since Royer (1928) put forward his rules for epitaxy, the role of the lattice misfit, or mismatch, between the substrate and the deposit at their interface, has been considered to have an important influence. However, although it is commonly believed that a small percentage misfit is essential for epitaxy to occur, it is well established that very large percentage misfits do not necessarily prevent epitaxy. The misfit does have an important influence on both the initial stages of growth of an epitaxial deposit and the way in which the structure of the deposit changes as deposition continues. Much has been learned about the modes of growth involved in epi-
taxy, from both theoretical and experimental studies. These modes of growth determine the structure of an epitaxial deposit, including the degree of structural perfection. Transmission electron microscopy has been of great importance in providing evidence of the nature of the lattice imperfections present in epitaxial deposits, and because a number of applications of these deposits require low densities of such imperfections, great interest centres on how the imperfections are formed, and how their formation can be prevented or their existence eliminated.
7.2 Deposition Methods Any technique which provides a means of depositing a thin metal film on a substrate surface, in a well controlled manner, can be used for producing epitaxial deposits on a single crystal surface. It can be important to prepare a clean flat surface of the substrate, with a particular crystal plane parallel to the surface, and to maintain the cleanness of the substrate surface during the initial stages of the deposition. For this reason, growth under conditions of a good clean vacuum is commonly used, either by the simple method known as vacuum evaporation or by its more modern version known as molecular beam epitaxy (MBE) (see this Volume, Chap. 8). Nevertheless good epitaxy can also be achieved by the simpler technique of electrodeposition (e.g. of one metal upon another) provided the substrate is electrically conducting and is stable in the required electrolyte. (Epitaxy in electrodeposited metallic coatings is treated in this Volume, Chap. 11, and in Chap. 10, Sec. 10.4.5.) A further technique is known as chemical vapour deposition (CVD). The substrate is heated in an atmosphere containing a
7.3 General Characteristics of Epitaxy
gaseous compound which decomposes as its molecules impinge on the substrate surface causing metal atoms to stick onto the surface. The technique is used widely for the growth of epitaxial layers of semiconductors. An example is the formation of a layer of silicon by the decomposition of SiH 4 . Alternatively, metal atoms are released onto the heated substrate surface by the chemical interaction between two different gaseous species which are passed over the substrate. The interaction only occurs on, or in close proximity to, the substrate surface. An example is the formation of gallium arsenide by the interaction between (CH 3 )Ga and AsH 3 : (CH 3 ) 3 Ga+AsH 3 -*GaAs + 3CH 4 With both of these CVD techniques, the rate of deposition can be varied by changing the substrate temperature or by changing the gas pressures or the gas flow rates. There is commonly only a limited range of substrate temperatures which can be used for a given rate of deposition of metal atoms. Undoubtedly the most common method used for metal epitaxy is that of vacuum evaporation. The source of metal vapour is simply a small volume of molten metal, the temperature of which is varied to control the rate of deposition of metal atoms on a substrate surface placed at a distance of a few centimetres away. The substrate temperature can then be varied quite independently of the rate of deposition of metal atoms. Whilst good epitaxy can be obtained if the deposition is carried out in a relatively poor vacuum, e.g., 10~ 3 Pa, much work is now carried out in a much higher vacuum (e.g., 10 ~9 Pa) in order to ensure that the substrate surface remains clean during the deposition process. The MBE deposition method also benefits from a clean vacuum environment, and
293
additionally uses metal sources consisting of Knudsen cells which allow highly stable rates of deposition to be used. This is of considerable importance for the formation of alloy layers of particular compositions from two or more sources as well as for the growth of artificial superlattices, where alternating layers of two metals, or alloys, are required with accurately controlled thicknesses. One important advantage of both the vacuum evaporation and the MBE techniques is that a reflection high energy electron diffraction (RHEED) system can be incorporated in the growth chamber to allow the structure of the epitaxial deposit to be observed during its growth.
7.3 General Characteristics of Epitaxy Epitaxy covers a range of deposits with wide variations in their characteristics. The essential requirement is the occurrence of preferred orientation in the deposit, directly related to the three-dimensional orientation of the substrate. The deposit crystals vary from large facetted crystallites, readily visible to the naked eye, such as were observed by the mineralogists, to thin uniform films of deposit which are only detectable by techniques such as electron diffraction or electron microscopy. 7.3.1 Orientation Relationships
It is normal, although not always observed, for a plane of atoms or ions in the deposit to be parallel to such a plane in the substrate, usually including the surface plane of the substrate. The parallel planes are usually aligned in such a way that a row of atoms or ions in one plane is parallel to a similar row in the other plane. Symmetry matching at the interface plane (i.e.
294
7 The Epitaxy of Metals
the substrate surface plane) is common, resulting in all atom rows in the substrate surface having a parallel row in the deposit. Since matching at the interface plays a dominant role in determining the orientation relationships, it follows that epitaxy can occur even if the substrate and deposit do not have the same crystal structure. Thus face-centred cubic (f.c.c.) metals such as gold and silver can be deposited epitaxially on the cleavage surface of mica, which has a monoclinic structure. The (001) cleavage face of mica contains a hexagonal network of ions, and the (111) plane of the face-centred cubic metals which also consists of a hexagonal network of atoms grows parallel to the mica cleavage surface. The control of epitaxy by alignment at the interface plane also results in the threedimensional orientation relationship between substrate and deposit being different for growth on different crystallographic surfaces of the substrate. This normally applies in all cases except where the substrate and deposit have the same crystal structure (e.g., both are face-centered cubic), and where the two are in parallel orientation on all substrate plane. Thus f.c.c. metals tend to grow on each other in parallel orientation, for all substrate surface planes. Comprehensive compilations of observed epitaxial orientations have been published by Seifert (1953) and Pashley (1956). More recently, Grunbaum (1975) has produced an updated and much more extensive list of known epitaxial systems, including references to all of the sources of information, but space limitations prevented inclusion of the details of the epitaxial relationships. It does not always happen that only one deposit orientation occurs on a given sub-
strate plane under a particular set of conditions of deposition (e.g., rate of deposition, substrate temperature). Often, a mixture of apparently unrelated orientations is observed. For example, Honjo et al. (1977) observed mixtures of (001) and (110) orientations of nickel and copper on (001) faces of magnesium oxide. Sometimes, a mixture of orientations occurs in the early stages of growth, but only one of these persists in thicker layers. Thus Matthews and Grunbaum (1965) found that, under certain conditions of deposition, (111) orientations of gold formed initially on the (001) cleavage face of rocksalt, but did not persist into thicker layers (but see Sec. 7.4.2.2). The (001) parallel orientation of gold, also present initially, dominated the thicker layers. Another form of multiple orientations is known as multiple positioning. This arises in situations where there are two or more equivalent ways of arranging a particular deposit orientation on the substrate surface, due to differences in symmetry. The simplest example is double positioning which occurs with (111) orientations of face-centred cubic metals (e.g. silver) on substrates such as mica (see Fig. 7-1 a). The two distinct orientations are: (111) Ag//(001) mica, with either [lTO] or [110] Ag//[100] mica. These are related to each other by a rotation of 180 degrees about the [111] axis, i.e., they are twin related orientations. In the case of (111) gold orientations on (001) rocksalt, as quoted above, there are the eight orientations made up of two sets of four equivalent orientations as shown in Fig. 7-1 b. 7.3.1.1 Deposition Conditions for Obtaining Epitaxy
In many of the reported cases of epitaxy, good quality orientation has been ob-
7.3 General Characteristics of Epitaxy
[110]
Figure 7-1. (a) Double positioning represented by the two twin-related orientations which occur for f.c.c. structures on a substrate of hexagonal symmetry, such as the mica cleavage surface, (b) The eight orientations of gold on the (001) rocksalt cleavage surface. Four equivalent (111) orientations have <110> directions parallel to NaCl <100> directions and the other four equivalent orientations have <110> directions parallel to NaCl <110> directions.
tained only over a limited range of deposition parameters. The relevant parameters appear to be: 1. The cleanness of the substrate surface 2. The smoothness of the substrate surface 3. The presence of contamination in the deposit 4. The rate of deposition 5. The substrate temperature during deposition. Because, in many experiments, there is inadequate information on items 1 and 3, and how they are influenced by items 4 and 5, it is not possible to draw any firm general conclusions. This applies particularly to much of the older evidence which relates to growth in poor quality vacuum systems. It is generally assumed that the presence of contamination is detrimental to the occur-
295
rence of good quality epitaxy (i.e., good alignment of all of the deposit in one single orientation). This is probably so, although Matthews and Grunbaum (1965) found that contamination can favour the establishment of good quality epitaxy of gold on rocksalt. Many investigators have found that substrate temperature has an important influence on the occurrence of epitaxy, following the early work of Bruck (1936), who found that a minimum temperature was required for the epitaxy of metals on rocksalt. This became known as the epitaxial temperature. When growth is carried out by the vacuum evaporation or MBE techniques, it is normal for the substrate to be held at an elevated temperature during deposition. Apart from the obvious possibility that substrate temperature is likely to modify the state of cleanliness of both the substrate and the deposit, it will also influence the various kinetic factors in film growth such as the surface mobility of the atoms arriving on the substrate surface. The rate of deposition will also influence the kinetic factors. 7.3.2 The Role of Lattice Misfit Ever since the classic work of Royer (1928), lattice misfit, or mismatch, at the interface between the substrate and the deposit has been considered as having an important influence on whether or not epitaxy occurs. The percentage misfit is simply defined as: 100 (b- a) m=
(7-1)
where a is the spacing of atom (or ion) rows in the substrate surface and b is the equivalent spacing in the parallel rows in the deposit. For some orientations, m varies with direction in the interface. Royer put forward a number of rules for the oc-
296
7 The Epitaxy of Metals
currence of epitaxy, mainly based upon the extensive studies of his school on the growth of surface deposits from solution. The most widely quoted rule is that epitaxy only occurs if the misfit is less than about 15%. Extensive observations since have shown quite clearly that a small misfit is not a necessary requirement for epitaxy. This has been shown particularly well by systematic studies of the growth of a series of similar compounds (e.g., alkali halides), with a range of lattice parameters, by evaporation onto the same substrate. For a review of this evidence see Pashley (1956). Frank and van der Merwe (1949 a, b, 1950) introduced the idea that the growth of an epitaxial deposit depends upon the initial growth of a monolayer of the deposit which is strained elastically to match (i.e., have zero misfit with) the substrate surface. This was based upon the concept of pseudomorphism introduced by Finch and Quarrell (see Sec. 7.5). Frank and van der Merwe calculated that the pseudomorphic monolayers would only form for natural misfit values less than some limiting value in the region of 10-15%. Whilst this criterion for epitaxy is not sufficient, because other growth modes occur (see Sec. 7.4.1), the ideas introduced by Frank and van der Merwe have had a profound influence on the understanding of epitaxy, especially in relation to lattice strain and misfit dislocations (see Sec. 7.5). Because good quality epitaxy can occur with large misfit values, and because the orientation which occurs in a particular case is not necessarily that which has the best possible fit between the deposit and the substrate, it is difficult to provide a systematic framework for describing the role of misfit in epitaxy. However, there are some systematic studies which provide convincing evidence. Thus, Honjo et al. (1977) have carried out extensive observa-
tions on the orientation of three series of materials, including f e e . metals, on a magnesium oxide substrate in (001) orientation. They find that several distinguishable orientations of the f e e . metals occur with (001), (110) or (111) planes parallel to the surface, and that there is a systematic dependence of these orientations on the ratio of the lattice parameters of the substrate and the deposit, as summarized in Table 7-1. The best fit of the (001) orientation requires the ratio to be unity. For the other two orientations there are good onedimensional fits, corresponding to parallel rows of atoms having the same spacings in the substrate ^nd the deposit, when the ratio has the value v ^3/2 for the (110) orientation and 2/^/3 for the (111) orientation. The three orientations seem to occur for those f e e . metals which have lattice parameters giving ratios fairly close to these values. Thus there are examples where good lattice matching does appear Table 7-1. Epitaxial orientations of metals on MgO (from Honjo et al., 1977). Metala
Ni Cu Pd Pt Al Fe Au Ag In Pb a
Ratio R b
0.84 V3 ^0.85 2 "^0.92 0.93 0.96 0.96 0.97 . 0.97 ^ 1.091 2 ^ 1.17 j y 3 ~^ i.i8
Orientations (110)
(001)
**c **
*d * ** ** **
(111)
*
**
all metals are f.c.c. except In, which is face-centred tetragonal and Fe which is b.c.c. and is referred to its f.c.c. base. b R = ratio ametal/aMgO. c **: strong preference. d *: less strong preference.
7.4 The Modes of Growth of Epitaxial Metal Films
to determine particular epitaxial orientations, but this is not generally true. The misfit does seem to have significant influence on the details of the growth of an epitaxial layer.
7.4 The Modes of Growth of Epitaxial Metal Films Extensive studies by electron microscopy and electron diffraction and Auger electron spectroscopy have shown that there are wide variations in the way in which thin expitaxial deposits grow on a substrate. The initial stages of growth have received particularly strong attention because it has often been considered that epitaxy is determined during the initial stages, although post-nucleation processes are now known to have an important influence in some cases. It is generally recognised, following the work of Bauer (1958), that the initial stages can be classified into three idealised kinds of nucleation as described by Bauer and Poppa (1972). A useful historical summary of the basis for this classification has been given by Markov and Stoyanov (1987). 7.4.1 The Nucleation Modes of Growth The three distinct modes are illustrated in Fig. 7-2. The first mode follows from the theory of Frank and van der Merwe (1949 a, b, 1950) (and hence is called the Frank-van der Merwe mode, or monolayer mode) and consists of monolayer or twodimensional (2D) growth. The deposit grows monolayer by monolayer. Once one monolayer has been completed, a new monolayer is nucleated on top of it and when this monolayer is completed the process repeats itself.
297
(a)
(b)
(0 Figure 7-2. The three modes of growth of epitaxial layers (a) The Frank and van der Merwe monolayer (2D) mode; (b) the Volmer-Weber (3 D) mode; (c) the Stranski-Krastanow mode involving monolayer (2 D) growth followed by 3D growth.
The second mode involves the initial formation of a surface distribution of threedimensional (3D) nuclei, separated by uncovered regions of the substrate surface. The size and number of these three-dimensional nuclei changes as further deposition continues, until the nuclei coalesce and eventually form a continuous deposit film. This mode is normally known as the Volmer-Weber mode of growth. The third mode of growth is a combination of the other two. 2D growth occurs first, but after a few monolayers have formed the mechanism changes and 3D nuclei form on the uppermost layer. Further growth then occurs as for the VolmerWeber mode. This third mode is known as the Stranski-Krastanow mode. Whether 2D or 3D growth occurs initially can be considered in terms of surface energies, as has recently been summarised by Bauer and van der Merwe (1986). If the surface energy of the substrate is ys, the surface energy of the deposit film is yf and
298
7 The Epitaxy of Metals
the interfacial energy is yin, monolayer growth is expected if the change in surface energy Ay, resulting from the deposition, is given by:
Table 7-2. The surface energy of various substances (from Kern et al., 1979). Substance
Face
Ay = yf + y i n - y s ^ 0
KBr KC1 NaCl CaCO 3 LiF CaF2 Cd Mg Zn Pb Pb MgO Si Al Al Ag Ag Au Au Cu Cu Cr Cr a-Fe a-Fe Ni Ni Pt Pt W Diamond
(100) (100) (100) (1010) (100) (111) (0001) (0001) (0001) (111) (100) (100) (111) (111) (100) (111) (100) (111) (100) (111) (100) (110) (100) (110) (100) (111) (100) (111) (100) (110) (111)
(7-2)
If this continues to apply as deposition continues, monolayer growth will continue. Since the Frank and van der Merwe theory involves the elastic straining of the monolayers, the elastic strain energy is included in these y values. If the strain energy contribution is large and increases as the number of monolayers increases, a point can be reached where Eq. (7-2) no longer applies and 3D nucleation takes place. This is the Stranski-Krastanow mode. If Eq. (7-2) does not apply at the beginning of growth, the Volmer-Weber 3 D nucleation process takes place. The main difficulty in applying the surface energy criterion represented in Eq. (72) is the lack of reliable data on the three surface energies. The problems have been considered by Kern et al. (1979), who tabulate a limited amount of data including surface energies of particular crystal faces of some alkali halides and some metals. An extract from these data is given in Table 7-2. In addition to the difficulty of making measurements, there is the problem of ensuring that measured values are not seriously influenced by the presence of contamination. It is well known that small traces of impurities can have a significant effect on surface and interfacial energies. Kern et al. (1979) have also given some values for the interfacial energies between a few pairs of metals. These values are small in comparison with the majority of the surface energies given in Table 7-2. Thus if ys is significantly greater than yf, 3 D nucleation is expected. If the materials of the substrate and deposit film are interchanged, ys becomes significantly less than
Surface energy ys (nJ mm ~ 2) 137 152 170 230 340 450 624 739 909 774 892 1200 1240 1692 1941 1693 1944 2218 2547 2554 2932 2775 3644 3032 4010 3246 3720 3294 3781 ^3000 5650
yf with the result that 2D nucleation is expected. When the values of ys and yf are nearly equal, the type of nucleation is determined by yin. If yin is small in comparison with ys and yf, it is possible for Eq. (7-2) to be satisfied, or nearly satisfied, both for growth of a particular deposit on a particular substrate and when the substrate and deposit are interchanged. Bauer and van der Merwe (1986) suggest that close to monolayer growth could then occur in both cases.
7.4 The Modes of Growth of Epitaxial Metal Films
7.4.1,1 Monolayer Growth
The interfacial surface energy yin is dependent upon the nature and the strength of the bonding at the interface. It also depends upon the value of the misfit, through the contribution of the elastic strain energy. Thus sufficiently low values of yin to allow Eq. (7-2) to be satisfied are likely to be confined to substrate/deposit combinations for which the misfit is small, although there are some exceptions to this. In accordance with the Frank and van der Merwe theory, pseudomorphic monolayer growth will normally only apply to cases of low misfit, although it would not be expected that a small misfit is a sufficient condition for monolayer growth. Many examples of strained pseudomorphic monolayer growth have been reported, and these generally involve misfits of no more than a few percent, perhaps up to values approaching ten percent. Because the surface energies of metals tend to be high in relation to many nonmetals (see Table 7-2), growth of metals on non-metals commonly occurs by the Volmer-Weber 3 D mode. The known cases of monolayer growth of metals are largely confined to the growth of metals on each other, although monolayer grown on some other substrates has been observed. The detection of monolayer growth can be achieved by several techniques. Reflection high energy electron diffraction (RHEED) (Milne, 1990) distinguishes monolayer growth from 3 D nucleation, but it can only be done with certainty if the growth is carried out simultaneously with the diffraction analysis by an appropriate in-situ technique. For growth of semiconductor layers by MBE, it is now common practice to include a RHEED system in the growth chamber. This has had particular value following the observation of oscillations in
299
the intensity of the specular reflected electron beam, as well as the diffracted beams, from surfaces on which monolayer growth is taking place (Harris et al., 1981a, b). These RHEED oscillations are interpreted (Neave et al, 1983) in terms of the changes in reflected intensity which occur as each monolayer of the growth is completed, to be followed by the nucleation of monolayer islands which coalesce to form the next monolayer. Although there is not complete agreement as to the detailed interpretation of the intensity oscillations, it is generally accepted that the periodicity does correspond to a thickness increase of one monolayer. Also it is generally accepted that the occurrence of the oscillations is direct evidence of monolayer growth. The technique has been applied to a number of metal depositions. For homoepitaxy, i.e., the growth of a metal on itself, RHEED oscillations have been observed (Steigerwald and Egelhoff, 1987) for copper grown on copper and f e e . iron grown on f e e . iron, When, however, f e e . iron is grown on copper there are no RHEED oscillations initially, but they do occur after the first few layers are deposited. This is explained as due to the initial growth of 3 D nuclei of iron followed by monolayer growth once a continuous layer of iron has formed. Jalochowski and Bauer (1988 a, b) have observed RHEED oscillations for silver and lead grown on silicon at 100 K. Auger electron spectroscopy (AES) can be used to detect monolayer growth (see Sec. 7.4.1.3 and Fig. 7-4), if in-situ observations are made during the growth of the deposit. Transmission electron microscopy also allows monolayer growth to be identified, if in-situ growth is carried out inside the electron microscope. Alternatively, if transmission electron microscopy is carried out after the growth of a deposit below the critical thickness tc (see Sec.
300
7 The Epitaxy of Metals
7.5.3.1), the absence of any misfit dislocations at the interface is evidence of pseudomorphic monolayer growth. There is now considerable interest in the growth of artificial superlattices of metals, such as are formed by the alternate epitaxial deposition of two different metals, or alloys, on a substrate. It is possible to grow the superlattice as effectively a single crystal if the two components have the same structure and the same lattice parameter. In the case of semiconductor III-V compounds this is commonly achieved by adjusting the lattice parameters of the two components by alloying between two or more III-V compounds. This approach can be extended by using two components with only a small lattice parameter difference, giving a misfit of perhaps one percent or less. The pseudomorphic strain associated with the monolayer growth mechanism then results in the superlattice also effectively being a coherent single crystal. This is then known as a strained layer superlattice. The periodicities in the interface plane match perfectly, but the periodicity perpendicular to the interface will be different in the two components since each component will be elastically strained in the direction away from the average of the natural periodicities. Such strained layer superlattices can be grown provided the thickness of each pseudomorphic component does not exceed the value of the critical thickness tc at which strain relief occurs (see Sec. 7.5.1; see also Chap. 8 of this Volume). Bauer and van der Merwe (1986) have considered the requirements for the growth of metallic superlattices, including the need for just one orientation to occur for growth of both A on B and B on A. For growth of superlattices with a small periodicity it is necessary for monolayer growth, or near monolayer growth, to take
place. This can be achieved with the appropriate combination of surface and interfacial energies (see Sec. 7.4.1). When Eq. (7-2) is nearly, but not completely, satisfied, the formation of a layer of uniform thickness can be achieved if a high rate of deposition and a low substrate temperature are employed. This results in a high density of thin plate-like nuclei which join together with the minimum of liquid-like coalescence (see Sec. 7.4.2.1), similar to the case illustrated in Fig. 7-8. Superlattices of metals have been grown by MBE for a number of different combinations. Durbin et al. (1982) prepared superlattices of niobium and tantalum which are both b.c.c. structures with a lattice mismatch of less than 0.2%. They were grown on single crystal sapphire substrates, with superlattice periodicities of 2-100 nm. Kwo et al. (1985) succeeded in growing good quality superlattices of gadolinium and yttrium, which are rare earth metals with hexagonal structures. Since both of these metals react chemically with many possible substrates, such as sapphire, a buffer layer of niobium on sapphire was used as the substrate. The structures of the two components do not have to be the same. Thus Cunningham and Flynn (1985) have grown superlattices of f.c.c. iridium and h.c.p. ruthenium. One of the problems in making superlattices of metals is to avoid significant alloying of the two components at the interfaces by being able to keep the temperature of the substrate sufficiently low during the deposition. Hsieh and Chiang (1986) have shown that it is possible to maintain sharp interfaces of silver and gold. Flynn (1988) has considered both the effect of alloying, and the maintenance of a smooth surface during growth, on the allowed growth temperatures. He maintains that RHEED oscillations are associ-
7.4 The Modes of Growth of Epitaxial Metal Films
ated with periodic surface roughening (see, e.g. Pautikis and Sindzingre, 1987), which is undesirable for the formation of a good superlattice. As the growth temperature is raised, the RHEED oscillations disappear, and this is explained as due to the increased surface mobility of the deposited atoms leading to ledge growth at steps (Fig. 7-3 a) with no formation of isolated monolayer nuclei or other clusters (such as shown in Fig. 7-3 b) on the ledges in between steps. He concludes that the growth temperature Tg must be greater than about 3/8 Tm, where Tm is the melting temperature of the metal being deposited, to avoid roughening. However, to avoid interdiffusion or alloying at the interfaces he estimates that the value of Tg needs to be less than 3/8 Tm. Thus there is only a very narrow range of possible values for Tg, if good quality superlattices are to be produced. He concludes, taking account of the limited accuracy of his estimates, that this optimal range is Tg = 0.35-0.40 Tm. This also means that it is not likely that good metal superlattices can be grown for two metals with very different melting temperatures, unless they are immiscible.
(a)
(b)
Figure 7-3. (a) Atoms diffusing to surface steps, resulting in growth of ledges without formation of new ledges, (b) Nucleation on ledges in between steps leading to periodic variation in surface roughness, causing RHEED oscillations.
301
7.4.1.2 The Volmer-Weber Nucleation Mode
A considerable amount of experimental study of 3 D nucleation has been made by transmission electron microscopy (TEM). Much of this work has involved deposits of metals on non-metallic substrates for which yf is generally much greater than ys. Gold and silver grown by vacuum evaporation onto rocksalt surfaces were by far the most widely studied systems before studies of semiconductor epitaxy became widespread in recent years. This arose because of the ease of making clean smooth surfaces by cleavage, and because the deposit layers could readily be detached from the substrate, for examination in the electron microscope, by dissolving the sodium chloride substrate in water. However, it is important to take into account a wide range of systems in order to obtain a balanced view of the nucleation mode of growth. The size and shape of the initial nuclei will also be determined by surface energies. This can be considered in terms of an extension of Wulff s theorem which determines the equilibrium shape of an isolated three-dimensional crystal (see Kern et al., 1979). For low values of the interfacial energy, the nuclei forming on a foreign substrate will have small ratios of height to width, and will therefore tend to be platelike. For high values of the interfacial energy, the nuclei will be much more equiaxed. Often they will have well developed crystallographic shapes. Nuclei are observed down to the smallest size which can be detected in the electron microscope, approximately 10 A in diameter. They are commonly distributed randomly over the substrate surface, with average separations which can be up to about a hundred diameters during the initial stages of deposi-
302
7 The Epitaxy of Metals
tion. However, particularly for metals deposited onto alkali halide surfaces, the initial nuclei form preferentially at surface steps on the substrate, even when those steps are only one atom layer in height. This was first observed by Bassett (1958), and many examples of the use of this decoration technique to provide information about the arrangement of steps on alkali halide surfaces, treated in different ways, have been reported by Bethge et al. (1968). Considerable efforts, both experimental and theoretical, have been devoted to the understanding of the kinetics of nucleation in the Volmer-Weber mode. Good reviews have been published by Venables and Price (1975) and Venables et al. (1984). Interest has centred upon topics such as (i) the smallest size of stable cluster, which would determine the initial nuclei; (ii) the rate of formation of stable nuclei for given deposition conditions; (iii) the geometrical distribution of nuclei over the substrate surface; and (iv) the change in size and numbers of nuclei as deposition continues, particularly to take account of the onset of the coalescence of nuclei leading to a reduction in numbers. Although these factors have considerable scientific interest, they do not seem to have had any major impact on the understanding of the factors controlling the occurrence of epitaxy. Some attempts have been made to determine the most energetically favourable orientation of a small stable nucleus on a single crystal substrate, in the hope that this could provide a means of developing a theory of epitaxy which would predict which orientation(s) will occur in a given situation. The best known treatment is that of Walton (1962), who considered two mechanisms by which a particular orientation may be favoured. Either a nucleus of a critical size, and in a particular orientation, is adsorbed more strongly than any other, or it is more able
to grow because nuclei or clusters of atoms in other orientations require the addition of atoms in unfavourable positions in order to grow. Whilst this approach helped to explain the orientation of f.c.c. metals on alkali halides, even though the misfit values are very high (e.g., 27%), it has not resulted in the prediction of orientations in many other systems. For this reason, no further consideration is given to the nucleation kinetics, and the reader is referred to the above-mentioned reviews for further information. It is important to consider how growth proceeds beyond the early nucleation stage, and how both the orientation and the structure of a deposited thin film are influenced. This is discussed in Sec. 7.4.2. 7.4.1.3 The Stranski-Krastanow Mode of Growth
Following the classification of the growth modes by Bauer and Poppa (1972), and their identification of the StranskiKrastanow mode as intermediate between the other two modes, various examples of the Stranski-Krastanow mode have been observed. In addition to examples involving the growth of semiconductor films, there are now an increasing number concerning epitaxial metal deposits. These seem to occur mainly either on substrates of molybdenum and of tungsten or on substrates of silicon and of germanium. The deposits include f.c.c. metals such as silver, gold and lead. Extensive reviews of the evidence have been published by Kern et al. (1979) and Venables et al. (1984). Much of the early evidence was based upon the application of Auger electron spectroscopy because the Auger signal originates from only the top few atom layers of a surface, so that monolayer growth causes the signal from the substrate to be reduced strongly
7.4 The Modes of Growth of Epitaxial Metal Films
for growth of just a few monolayers (Fig. 7-4b). For 3D nucleation, the Auger signal from the substrate remains strong for relatively large amounts of deposit (Fig. 7-4 a). The intermediate StranskiKrastanow mode causes the Auger signals to change as shown in Fig. 7-4c. The initial
(a)
No. of monolayers
(b)
No. of monolayers
(c)
No. of monolayers
Figure 7-4. The effect of different modes of growth on the Auger signal from the surface during AES analysis, (a) Three-dimensional nucleation according to Volmer-Weber mode, (b) Monolayer growth, (c) Stranski-Krastanow growth, with three-dimensional nucleation occurring after the formation of two monolayers. S = substrate; D = deposit film.
303
rapid decrease in the Auger signal from the substrate ceases immediately the 3D nucleation stage is reached. The scanning electron microscope, in conjunction with Auger electron spectroscopy and reflection high-energy electron diffraction, has also been used to provide direct evidence on the Stranski-Krastanow mode (Venables et al., 1980), by in-situ deposition. The nuclei, or islands, of the deposit could be observed in the microscope, and the presence of monolayer growth remaining between the islands could be determined by means of the Auger signals. A further electron microscope technique which has been applied to studies of this growth mode is that of reflection electron microscopy (Osakabe et al., 1980). The images are formed from diffracted beams obtained when the surface of the sample is illuminated by an electron beam at a glancing angle, and the resolution is adequate to allow monatomic surface steps to be seen. The technique has been applied to studies of copper, silver and gold on silicon surfaces (Takayanagi, 1985). For example, the growth of gold is seen to occur by the nucleation of monolayers at steps on the silicon surface and, after these had grown to complete one monolayer, the formation of 3D islands was observed. Detailed information was obtained on the reconstruction of the silicon substrate surface (see Vol. 1, Chap. 8) and on the superlattice structures of the deposited monolayers and the associated domain arrangements. 7.4.2 Post-Nucleation Growth Processes 7.4.2.1 Liquid-Like Coalescence Once a two-dimensional distribution of three-dimensional nuclei has been established by the Volmer-Weber mechanism (see Sec. 7.4.1.2), further deposition results in the growth in size of these nuclei. The
304
7 The Epitaxy of Metals
shape of the nuclei will remain the same if the surface diffusion of the newly arriving deposit atoms over the surfaces of the nuclei is sufficient to allow equilibrium shapes of the nuclei to be maintained. The nuclei, or islands, will continue to grow in size, in this way, until they grow sufficiently large to cause neighbouring islands to touch each other, and coalesce. The coalescence process has been observed when the epitaxial deposition process has been studied by in-situ electron microscopy. Bassett (1960) was the first to observe the so-called liquid-like coalescence behaviour of silver islands on a thin film substrate of molybdenite (MoS2). Two islands were shown to coalesce just as though they were liquid droplets, with the apparently instantaneous shape changes illustrated in Fig. 7-5. Pashley et al. (1964) extended the studies to include deposition of gold on molybdenite, and showed that the material in two coalescing islands remained solid, giving normal electron diffraction patterns, during the entire liquid-like coalescence. The initial observations were carried out in the normal poor vacuum conditions of electron microscopes of the day, leading to some doubts concerning the validity of the observations. Significantly improved vacua were used subsequently for in-situ microscopy of epitaxial deposition (Valdre et
Figure 7-5. The sequence of shape changes during the liquid-like coalescence of two non-crystallographic nuclei.
al., 1970; Poppa, 1969; Takayanagi et al., 1978 a), and the general characteristics of the initial observations were confirmed. The liquid-like coalescence is explained as due to self-diffusion of the deposit metal, e.g., gold, driven by surface energy. It seems likely that surface diffusion provides the predominant process, and Pashley et al. (1964) made some estimates of the times required to form necks between coalescing spheres of a range of different sizes, based upon the theory of particle sintering given by Kingery and Berg (1955). The results showed that the time required to produce a neck (e.g., as shown in Fig. 7-5b) of a given proportion of the initial size of the coalescing spheres is proportional to the fourth power of the initial radius of the spheres. Hence, in accord with the observations, smaller islands coalesce very much more rapidly than larger islands, although the effect of the non-spherical shape of the islands would need to be considered in any exact quantitative comparison. The general characteristics of the liquidlike coalescence depend upon the rate of surface diffusion, and the two extreme cases are illustrated schematically in Fig. 7-6 and Fig. 7-8. For fast surface diffusion and small islands (say less than about 100 nm across, in the case of gold at 300 °C), the sequence shown in Fig. 7-6 applies. Although the example illustrated is for triangular-based islands, the same kinds of sequence occur for islands of other crystallographic shapes. As soon as the two islands of Fig. 7-6 a touch, surface self-diffusion provides the possibility of rapid transfer of material to bring about a shape change in the compound island. In the initial stages of coalescence, rapid changes in total surface energy are brought about by significant rounding at the edges of the islands, due to the transfer of material from regions most remote from the neck to the
7.4 The Modes of Growth of Epitaxial Metal Films
Figure 7-6. The sequence of shape changes during the liquid-like coalescence of two triangular-based nuclei.
neck region which forms at the point of contact between the two islands (Fig. 76 b). This results in a dumb-bell shaped island which has lost its crystallographic shape (Fig. 7-6 c). Further reduction in surface area, coupled with a general thickening of the compound island, causes a more equiaxed island to form. A final reduction in total surface energy is brought about by the formation of a crystallographically shaped island (Fig. 7-6 d) which is just a larger version of each of the initial islands before coalescence commenced.
305
This coalescence process continues in the same way, thereby causing a continuous decrease in the number of islands per unit area, until a stage is reached when the total process of Fig. 7-6 is incomplete before a compound island is involved in coalescence with a neighbouring (compound) island. This changed situation arises because (i) the coalescence time increases rapidly as the average island size increases and (ii) the longer is the coalescence time the more do newly arriving deposit atoms cause an increase in the size of a compound island before the coalescence process is completed. Consequently, a deposit which starts off as nuclei with well developed crystallographic shapes develops into a distribution of much larger and fewer islands with rounded shapes. These rounded islands become joined together to form a network structure, the holes in which eventually become filled-in to produce a continuous hole-free deposit film. Fig. 7-7 illustrates the general characteristics of the growth sequence. When the surface self-diffusion, together with any volume self-diffusion, of the deposit material is sufficiently small, no significant change in the size and shape of islands can take place by the mechanisms
Figure 7-7. Transmission electron micrographs showing the stages of growth of a continuous epitaxial film by the Volmer-Weber 3 D mode, when pronounced liquid-like coalescence takes place (from Pashley etal., 1964). (a)
306
7 The Epitaxy of Metals
given in Fig. 7-6. Shape changes are then dependent upon the selective deposition of newly arriving deposit atoms, as illustrated schematically in Fig. 7-8. The extent to which the newly arriving atoms, diffusing over the uncovered regions of the substrate surface, deposit in the re-entrant regions, such as R in Fig. 7-8 b, determines how large the compound island grows before an equilibrium shape is re-established. Even with strong selective deposition at R, such as is shown for the sequence of Fig. 7-8, the compound island will be much larger than results from the coalescence of two islands of equal size by the mechanism of Fig. 7-6. A feature of the latter mechanism is that the area of coverage of the substrate surface is reduced by the liquid-like coalescence, but this does not happen in the selective deposition mechanism of Fig. 7-8. In general it is to be expected that for the coalescence of two islands, a combination of selective deposition and liquid-like coalescence will be involved in the shape changes which take place. The precise sequence of shape changes which occur, when two crystallographically shaped is-
lands coalesce, will depend upon just where they contact. For example, if the two islands contact at their apices, the sequence will be quite different from that which occurs when the apex of one touches the mid-point of the side of the other. The case illustrated is intermediate between the two. The liquid-like coalescence, resulting from surface self-diffusion of the deposit material, occurs commonly with a number of metals and it is mainly of significance when 3D nucleation takes place. During the early stages of the growth sequence, the liquid-like coalescence delays the formation of a continuous deposit film, since it causes a decrease in the actual area of coverage of the deposit, as shown in Fig. 7-5 and Fig. 7-6. During the later stages of the sequence a network structure is formed, as illustrated in Fig. 7-7. Further deposition occurs preferentially to fill in re-entrant regions such as that at R, driven by the reduction in surface area and hence surface energy. Hence, surface diffusion aids the formation of a continuous hole-free deposit film during the later stages of the sequence. 7.4.2.2 Reorientation and Recrystallization Effects
Figure 7-8. The sequence of shape changes during the coalescence of two triangular based nuclei, in the absence of liquid-like behavior and involving only selective deposition of newly arriving atoms.
In a number of systems, notably the growth of f.c.c. metals on alkali halides such as rocksalt, the initial nuclei can occur in a mixture of different orientations. In other systems the initial nuclei, although in the same nominal orientation, have deviations from exact alignment so that there is a spread of up to several degrees in orientation of the nuclei. In at least some of these latter cases it is believed that the poor alignments are related to contamination. When either mixed orientations or misalignments occur, it is possible for con-
7.4 The Modes of Growth of Epitaxial Metal Films
tinuous deposit films, resulting from further deposition, to be good quality single crystal films. The reorientation or recrystallization involved can take place either during coalescence of nuclei or following coalescence. The basic recrystallization process has been observed by in-situ transmission electron microscopy. Jacobs et al. (1966) studied the double positioning structure (see Sec. 7.3.1 and Fig. 7-1 a) in gold deposited on molybdenite, and found that reorientation occurred on coalescence. When two nuclei, or islands, in the two twin-related (111) orientations (or positions) coalesced to form a compound island containing a twin boundary, not necessarily a coherent twin boundary, the boundary sometimes migrated out of the island to leave an island in one single orientation (Fig. 7-9). Clearly, the driving force for this migration is the reduction and removal of the energy of the boundary. When the two coalescing islands were very different in size, the smaller island converted to the orientation of the larger island as the rapid liquidlike coalescence took place. The equivalent processes can occur for any general misorientation between two coalescing nuclei or islands. The epitaxial growth of gold on rocksalt cleavage surfaces contains a mixture of nuclei in eight different (111) orientations and nuclei in the parallel (001) orientation. As growth proceeds there is considerable recrystallization, but the result depends upon whether the rocksalt surface is exposed to air before the deposition is carried out (Matthews and Grunbaum, 1965). If growth takes place on a surface cleaved in ultra-high vacuum (UHV), the (111) oriented nuclei consume the (001) nuclei which recrystallize so that the continuous gold film contains a mixture of (111) oriented grains. If the rocksalt surface is ex-
307
posed to air before deposition of gold is carried out in UHV, the (001) nuclei consume the (111) nuclei and the continuous gold film is a single crystal in (001) orientation. Matthews (1965) explains this as due to the more numerous and larger (001) nuclei gaining over the smaller, less numerous (111) nuclei. When the substrate is not exposed to air, fewer nuclei form during the deposition of gold and the (111) nuclei grow quite large before coalescence takes place. Consequently the (111) nuclei seem to gain over the (001) nuclei during coalescence. Once a continuous deposit film is produced, there is still the possibility that any small grain of one orientation, which has become surrounded by material in another orientation, can be removed by grain boundary migration. Just as for the effect illustrated in Fig. 7-9, the driving force for the migration is the reduction in grain boundary area. This process would allow any remaining isolated grains of (111) orientation, in the example of the previous paragraph, to be eliminated. The important conclusions from the above evidence are: (1) significant changes
Figure 7-9. The liquid-like coalescence of two twinrelated triangular-based nuclei, leading to a double positioning twin boundary which then migrates out of the compound island which is formed. I and II represent the two positions.
308
7 The Epitaxy of Metals
in orientation can occur as a result of the coalescence of nuclei having different orientations; and (2) the final orientation of an epitaxial layer grown by the VolmerWeber 3D nucleation mechanism can be determined during the post-nucleation growth processes, rather than being completely determined by the orientation of the initial nuclei. Conclusion (2) is well illustrated by the fact that considerable grain growth can be achieved in polycrystalline films deposited onto single crystal substrates, so that large grains with well defined epitaxial orientations are produced (Thompson et al., 1990). The effect was obtained for the growth of f.c.c. metals deposited on alkali halides or mica, at relatively low temperatures for normal epitaxial growth, followed by annealing at higher temperatures. In this situation, the recrystallization is not related to coalescence since it takes place in a continuous film of uniform thickness. The technique provides a possible alternative method for growing epitaxial thin films, especially for film thicknesses lower than the minimum thickness for growth of continuous films by the normal Volmer-Weber mechanism.
the interface between them. This strain is accompanied by strain of the opposite sign normal to the interface, so as to maintain a nearly constant atomic volume. Their evidence was shown (Pashley, 1956) to be invalid, but their idea has proved to be correct for many substrate/deposit combinations studied during the last 25 years. The idea was incorporated into the classic theoretical treatment of Frank and van der Merwe (1949 a, b, 1950), in which the effect of the magnitude of the misfit on the formation of pseudomorphic monolayers was considered. It was shown that such monolayers would form for misfits below a certain limiting value (see Sec. 7.3.2). It was suggested that thickening of the deposit by successive growth of monolayers would lead to relaxation of the elastic strain and the formation of edge dislocations at the substrate/deposit interface. The way in which these dislocations accommodate the misfit between the substrate and the deposit is illustrated in Fig. 7-10. These are what are now well known as misfit dislocations. The above use of the term 'pseudomorphic' must be distinguished from its sometimes other use to denote the occurrence of Deposit
7.5 Elastic Strains and Misfit Dislocations The existence of elastic strains in epitaxial deposits is most commonly related to the formation of layers which are pseudomorphic. The term basal plane pseudomorphism was introduced by Finch and Quarrell (1933 and 1934), who interpreted observations on thin films of zinc oxide on zinc, and aluminium on platinum, as demonstrating that oriented deposits are strained elastically so that their lattice spacings match those of the substrate at
T
f
1F 1 1
Substrate Figure 7-10. The formation of misfit dislocations between the substrate and the epitaxial deposit to accommodate the misfit between the two.
7.5 Elastic Strains and Misfit Dislocations
an abnormal crystalline structure in an epitaxial deposit, to match the structure of the substrate (e.g., a normally h.c.p. material growing as f.c.c. on a substrate of f.c.c. material). 7.5.1 Changes in Elastic Strain with Increasing Thickness
For sufficiently small misfit values, the lowest energy state of a monolayer is a pseudomorphic monolayer. As further monolayers are added a stage is reached at which the deposit film should revert to its normal crystal spacings. This occurs at what is commonly known as the critical thickness, tc, and various approaches have been taken to the calculation of the critical thickness. The first treatment was published by van der Merwe (1963), and extended by Ball (1970) and Ball and van der Merwe (1970), who considered the minimum total strain energy by taking account of both the pseudomorphic elastic strain and the strain field of misfit dislocations. They defined the critical thickness as that at which it becomes energetically favourable for misfit dislocations to be present at the interface. This approach has been extended by taking account of the possible mechanism which can lead to the formation of misfit dislocations once tc has been exceeded. These mechanisms are summarised in Sec. 7.5.3.1. Jesser and Matthews (1967, 1968 a, b) and Matthews and Crawford (1970) considered the critical thickness as being determined by the misfit strain being adequate to exert a sufficient force on a threading dislocation to cause it to move according to the mechanism illustrated in Fig. 7-13. Thus, on this basis, the value of tc is determined by a force balance rather than an energy balance. The approach was extended to the case of the generation of
309
half-loops by the mechanism of Fig. 7-14, and the force required for their nucleation was calculated by Matthews (1975 a). He suggested that a dislocation loop will have a critical radius for expansion under the action of epitaxial stresses, and produced expressions for this. Cherns and Stowell (1976) extended this approach and included consideration of the nucleation and expansion of partial dislocation loops together with the associated stacking fault. A detailed theoretical discussion of the interpretation of the critical thickness has been provided by Willis et al. (1990) and Jain et al. (1990), with particular reference to semiconductor epitaxy. In the case of epitaxial metals, it is generally found that the theoretical values for tc agree fairly well with the observed thicknesses at which the onset of formation of misfit dislocations occurs, whereas epitaxial semiconductor layers commonly remain strained for thicknesses much greater than the theoretical values. This is believed to be due to the difficulty of nucleating misfit dislocations in these materials. An interesting confirmation of this conclusion has been provided by Hull et al. (1988) who annealed, in-situ in the transmission electron microscope, a layer of germanium-silicon alloy grown on silicon. They showed that there was a large increase in the number of misfit dislocations when the annealing temperature was well above the growth temperature, illustrating that such epitaxial layers have a metastable structure above tc. A good example of the determination of tc for a metallic system is that of gold grown on (111) palladium surfaces (Kuk et al., 1983). Low energy electron diffraction was used to measure the lattice constant of the gold as a function of thickness for growth which was shown to occur by the monolayer mechanism. Fig. 7-11 shows
310
7 The Epitaxy of Metals
Lattice
No. of monolayers
Figure 7-11. The change in the lattice parameter of an epitaxial gold layer on a (111) palladium surface as a function of the amount of deposit, as determined by low-energy electron diffraction (Kuk et al., 1983).
the result, with pseudomorphism extending only for the first two monolayers, to give a tc value of about 0.45 nm, followed by the gradual relief of the pseudomorphic strain during the growth of the next eight monolayers. This strain relief was assumed to be due to the gradual formation of misfit dislocations. 7.5.2 Misfit Dislocations The most efficient dislocations for accommodating the misfit between an epitaxial deposit and its substrate are pure edge dislocations with their Burgers vector in the plane of the interface. For the common situation of a two-dimensional misfit in the plane, a two-dimensional dislocation network is required. For metal deposits, the examples where the misfit dislocations have been observed and their Burgers vector determined is rather limited, but it is normally found that the Burgers vectors are of the same kind as those found within bulk crystals. However, they are not free to move by slip unless the interface plane happens to be a slip plane. Thus for cubic structures having (111) slip planes, the ideal misfit dislocations cannot slip unless the
deposit is growing in (111) orientation, and then slip can only occur in the interface plane. Such misfit dislocations cannot cross-slip and pass through the thickness of the film. The spacing of these misfit dislocations can be determined with reference to the spacing of the moire patterns obtained from two overlapping crystals. Examples of moire patterns are given in Fig. 7-26. Let the spacing of the substrate planes perpendicular to the surface and parallel to the misfit dislocations be ds9 and the spacing of the parallel planes in the deposit film be df. The misfit is m in %, where 100(d f -d s ) m=-
(7-3)
These two parallel sets of planes will go in and out of register with the periodicity of the Moire pattern, Sm. This is given by Bassett etal. (1958): d{ds d(-ds
100 d{
(7-4)
m
If the component of the Burgers vector of the misfit dislocations, parallel to the interface and perpendicular to the dislocation line, is be (i.e., the edge component in the interface), this will be given by be = ndf
(7-5)
where the dislocation is assumed to be in the deposit film. This assumption is based upon the substrate being thick so that all elastic and/or plastic strain resulting from the misfit occurs in the deposit. For perfect dislocations n is an integer, which is given by n=g b
(7-6)
where b is the total Burgers vector of the dislocation and g is the reciprocal lattice vector perpendicular to the planes of spacing d{. Thus if these are the (hkl) planes g
7.5 Elastic Strains and Misfit Dislocations
311
is given by ghkl. It follows that the spacing Sd of the parallel array of dislocations which completely accommodates the misfit is given by 100 ndf 100 (g • o
(7-7)
(a)
m m For the common situation of f.c.c. metals growing in parallel orientation on the (001) face of another f.c.c. metal, the normal arrangement is a square network of misfit dislocations along [110] and [110] with Burgers vectors of [f \ 0]and[f \ 0] respectively. The spacing of the dislocation network required to accommodate all the elastic strain is given by putting when g • b = 2 in each case and S, = l00aL/2m Eq. (7-7) is more commonly given in the form 100 be
(b) Figure 7-12. A schematic representation of the slip involved in forming misfit dislocations at the interface between a substrate and an epitaxial deposit which forms initially by the pseudomorphic monolayer mechanism, (a) Positive misfit producing pseudomorphic compressive stresses, (b) Negative misfit producing pseudomorphic tensile stresses.
(7-8)
m In general, as implied above, it is not necessary that the misfit dislocations have their Burgers vectors in the interface plane. If the interface contains an array of dislocations which have Burgers vectors with significant components lying in the interface plane, the misfit can be accommodated. However, the Burgers vector components normal to the interface plane results in the interface also being a small angle tilt boundary (see Sec. 7.5.3.1 and Fig. 7-12). Any Burgers vector components parallel to the misfit dislocation lines, giving them a screw component, result in the interface being a low-angle twist boundary. It is therefore possible that the interface plane is a low-angle boundary consisting of a mixture of tilt and twist components. The consequences of this can be illustrated by reference to the above example of the par-
allel orientation of one f.c.c. metal crystal on another. A possible alternative to the ideal misfit dislocations is dislocations along [110] and [110] with, for example, Burgers vectors of [\ 0 \] and [0 \ \\ respectively. Such misfit dislocations are commonly observed with the epitaxy of III-V semiconductor compounds on each other, where they are called 60° dislocations because their Burgers vectors are at 60° to the dislocation line. These dislocations are free to slip on (111) and (111) planes respectively. The edge component of their Burgers vector within the (001) interface plane is a/y/8, giving an effective value of n as 1, so that their spacing is halved relative to that of the ideal misfit dislocations, and is 100 a 2 . /2 m
312
7 The Epitaxy of Metals
The edge component perpendicular to the (001) plane is a/2, which would cause each of the two perpendicular sets of misfit dislocations to give a tilt of a/2
i00a/{2^2m)
2m
100
radians
2m The net result is a tilt of ^j- radians about [100]. This angle increases linearly with m and has a value of 1.1 degrees for ra = l%. Such tilt angles should be observable but none appear to have been reported. The screw component of the Burgers vector along the dislocation lines is a/y/8, and the square network results in a twist of the deposit lattice, about [001], of m radians
Too
i.e., 0.55 degrees per 1% misfit. 7.5.3 The Formation of Misfit Dislocations
The way in which misfit dislocations are formed in an epitaxial deposit depends upon (i) whether or not the deposit is elastically strained (i.e., pseudomorphic), and at what stage the elastic strain is relieved by misfit dislocations, and (ii) the mode of growth of the film, i.e., monolayer growth or nucleated growth. 7.5.3.1 Formation During Monolayer Growth
Generally, Frank-van der Merwe monolayer growth is confined to the lower range of misfit values, say m < 1 0 % , and nucleation will commonly occur for higher misfit values. If monolayer growth occurs for the higher misfit values, the theory indicates that pseudomorphism will not occur and the first monolayer will have close to its natural spacing, so that misfit disloca-
tions are expected to form at the interface right from the earliest stages of growth. The most energetically favourable dislocations would be expected to have the same Burgers vectors as those which normally occur in bulk material of the deposit. If the contact plane of the deposit contains required directions for these Burgers vectors, the formation of ideal misfit dislocations with their Burgers vectors parallel to the interface is likely. However, if the contact plane of the deposit does not contain these directions, either misfit dislocations with abnormal Burgers vectors would form or the misfit dislocations would have normal Burgers vectors not parallel to the interface plane. Most systematic evidence, including the relatively extensive observations on semiconductor epitaxy, has been obtained on low-index surfaces which do contain the directions along which normal Burgers vectors occur (i.e., the (100) and (111) surfaces of f e e . materials). Thus there is little evidence on what happens when there are no normal Burgers vectors parallel to the interface, or when there are insufficient (i.e., there is only one in the {110} surface off.ee. materials). The most important consideration is how dislocations are introduced once a film of large area growing by the Frank and van der Merwe monolayer mechanism exceeds its critical thickness tc. Spontaneous nucleation of misfit dislocations at the interface, without any passage of dislocations through the film, is neither likely nor an adequate requirement. In order to relieve the elastic strain of m in % which is present in the plane of the film, and which exists uniformly through the thickness of the film, it is necessary to bring about plastic deformation. This can be done either by slip or by climb, or by a combination of the two. The basic requirement for slip is that it occurs on an inclined slip plane, with the
7.5 Elastic Strains and Misfit Dislocations
Burgers vector of the dislocation having a non-zero component parallel to the interface and perpendicular to the line of intersection between the interface and the slip plane. This ensures that the misfit dislocation formed at the interface has an edge component in the interface and that linear plastic strain occurs in the plane of the film. Fig. 7-12 a illustrates the requirement for a film initially under elastic compression, since it involves plastic elongation of the film. Fig. 7-12 b illustrates the requirement for a film initially under elastic tension, which is compensated by plastic compression. The diagrams of Fig. 7-12 illustrate that a tilt boundary is produced as a result of the slip occurring on a parallel set Of regularly spaced slip planes. Thus the deposit layer would become detached from the substrate in the absence of an accompanying tilt. Two basic kinds of mechanism for introt
. ~
,. .
.
,
, - i
ducing misiit dislocations by slip have been proposed, and supported by experimental evidence, The first, proposed by Matthews (1975 b) to explain some observations on the growth of certain f.c.c. metals on each other, involves the movement of dislocations which thread continuously through the substrate and the deposit (see Fig. 7-13). Glide of the dislocation on an inclined slip plane, through the deposit only, leaves a trailing dislocation in the interface. This results in a non-ideal misfit dislocation being formed, provided the above basic requirement for the Burgers vector is fulfilled. The mechanism is likely to apply mainly for cases of rather small misfits, perhaps significantly less than 1%, because large numbers of threading dislocations with appropriate Burgers vectors would be needed to provide the high density of misfit dislocations required to accommodate larger misfits. Also, it is necessary to consider mech-
313
Deposit Substrate i t
I ~T
Figure7 . 13>
The formation of a misfit dislocation by the glide of a dislocation which threads through both the substrate and the epitaxial deposit. The Burgers vector of the dislocation must contain a component n o r m a l t0 the i n t e r f a c e t h e r e f o r e the misflt disloca ' " tion produced is not an (ideal) edge dislocation.
anisms which apply in the absence of threading dislocations, as would be likely for growth up to the critical thickness on a dislocation free substrate, Such a mechanism was also proposed by Matthews (1975 a and 1976 b), and involved the nucleation of dislocation loops at the growing surface of the epitaxial deposit. The loops expand by glide (see Fig. 7-14) and provide the misfit dislocations as they reach the interface, together with a pair of threading dislocations. If this mechanism operates so as to produce the complete interface misfit dislocation network required to relieve all of the epitaxial elastic strain, it results in the interface also being a low-angle boundary, and twist and/or tilt will occur as discussed in Sec. 7.5.2. There is little experimental evidence that such tilts or twists do actually occur in practice, even though the angles
314
7 The Epitaxy of Metals
Substrate
Figure 7-14. The nucleation of dislocation loops at the surface of an epitaxial deposit, and their expansion to form misfit dislocations at the interface, together with a pair of threading dislocations AB and CD.
involved for misfits of 1 % or more should be detectable. In fact, almost none of the papers dealing with the occurrence of misfit dislocations which are not ideal misfit dislocations contain any reference to the angular tilts and twists, which seem to have been largely ignored. One exception is the observation of a rotation of about 2 Vi degrees of an epitaxial deposit of cadmium selenide on germanium (Gejji and Holt, 1978). These authors showed that this rotation is consistent with the presence of a network of misfit dislocations having a screw component. Observations of small angular misorientations of cadmium sulphide grown on several different substrates have been made by Igarashi (1971), who explains them as due to the misfit dislocation arrangements. One possible reason for the lack of much evidence for tilts or twists is that the occurrence of non-ideal misfit dislocations is not common, at least for metal deposits. Alternatively, it is possible to avoid the tilts or twists by a suitable combination of non-ideal misfit dislocations of different Burgers vectors. Any mechanism for introducing ideal misfit dislocations must involve at least some elements of dislocation climb. This was first demonstrated by Yagi et al. (1971) who observed the formation of ideal misfit
dislocations during the growth of gold films on palladium inside the electron microscope. The simplest mechanism is for Frank dislocation loops to nucleate at the surface, and to move to the interface by climb, as shown in Fig. 7-15. If the plane of the loop is perpendicular to the interface, ideal misfit dislocations are produced. More generally, the plane will be inclined at some other angle and the dislocations will have Burgers vectors with a component normal to the interface, resulting in the interface becoming a tilt boundary. Also, for f e e . metals the Frank sessile dislocation forms on a (111) plane where it produces a stacking fault, so that a regular arrangement of stacking faults would extend through the epitaxial layer, on (111) planes. Cherns and Stowell (1975 a) have made a detailed study of the growth of palladium on (001) gold, by evaporation inside an electron microscope, and find that Frank sessile dislocations do climb as loops generated at the palladium surface. However, this follows the nucleation of Shockley partial dislocations, see Fig. 716, which produce stacking faults on the inclined (111) plane. These faults are removed as the Frank dislocation follows on the same (111) plane, and the two dislocations join together at the interface to pro-
Deposit
Figure 7-15. The relief of compressive pseudomorphic strain by the nucleation of a dislocation loop which expands by climb, to produce a misfit dislocation at the interface.
7.5 Elastic Strains and Misfit Dislocations
Frank partial Deposit
Stacking fault
Ideal misfit dislocation
Shockley Partial
Substrate Figure 7-16. The nucleation, and expansion by climb, of a half-loop of a Shockley partial dislocation in an epitaxial layer of palladium on gold. This produces a stacking fault on the (111) slip plane which is removed by the nucleation, and expansion by slip, of a half-loop of a Frank partial, which combines with the Shockley partial at the interface to produce an ideal misfit dislocation.
duce an ideal misfit dislocation:
- [1 1 2] + \ [1 1 1] -> i [1 1 0] o
Shockley
3
z
Frank
The misfit dislocation forms along [110] in the (001) interface, and is a pure edge dislocation with its Burgers vector in the interface. This particular sequence applies only for a negative misfit (i.e., the lattice parameter a0 for palladium is 4.7 % smaller than a0 for gold), where the Frank dislocation must produce an extrinsic stacking fault since it must introduce an extra (111) plane of atoms to relieve the elastic tension in the palladium. If the Frank dislocation were to nucleate first, followed by the Shockley partial, an ideal misfit dislocation would form at the interface but an arrangement of overlapping stacking faults would remain on the (111) plane. However, for a positive misfit, it is necessary to nucleate a Frank dislocation (involving an intrinsic fault) first followed by a Shockley partial, although this particular mechanism does not appear to have been reported. Reversal of this order is not possible since it would
315
involve a non-allowed stacking sequence on the (111) plane. Thus a combination of slip and climb, involving two (or more) dislocation loops, provides a mechanism for introducing ideal misfit dislocations at the interface, without any remaining planar faults. It seems that in some cases, at least, more complex slip and climb sequences apply. Cherns and Stowell (1975 b) find that when palladium is grown on a (111) gold surface, trigons of ideal misfit dislocations are produced by a sequence of interactions following the simultaneous nucleation of Frank partials on the three inclined {111} planes. Similar trigons of misfit dislocations have been observed by Honjo et al. (1977), who studied the in-situ growth of iron on a gold (111) surface. In this case the iron has the f.c.c. structure (y-phase) and is pseudomorphic up to a thickness of about 0.3 nm, after which the misfit dislocations are nucleated. The splitting of these dislocations into partials was revealed by a detailed contrast analysis. More recently, evidence for the mechanisms by which misfit dislocations are introduced during the growth of some epitaxial semiconductor films has been sought. For the growth of germanium-silicon alloys, a new misfit dislocation source has been observed by Eaglesham et al. (1989). This involves the formation of a diamond-shaped stacking fault which acts as a source for loops of 60° dislocations which can glide into the interface. No climb is involved in this mechanism. It is necessary for more careful experimental studies to be made for a range of substrate/ deposit combinations which result in pseudomorphic monolayer growth. Improved understanding can only follow from more direct evidence of the actual mechanisms which occur, and the conditions under which they operate.
316
7 The Epitaxy of Metals
The existing evidence on misfit dislocations and their formation during monolayer growth has been obtained predominantly with deposits of either (a) f.c.c. metals or (b) silicon and germanium and their alloys or (c) III-V semiconducting compounds with the cubic zinc sulphide structure. These three classes of material have similar slip systems and dislocations, involving dislocations with < | \ 0> Burgers vectors slipping on {111} planes. The growth of deposits in (001) orientation has featured strongly, with growth of metal layers in (111) orientation also having been fairly extensively studied. Little work has been done on deposits in other orientations, including (110). In the case of (001) deposits, the situation is simple because a square network of misfit dislocations along the [110] and [lTO] directions can be formed, either of ideal misfit dislocations or of the so-called 60° dislocations. In either case, complete relief of the pseudomorphic elastic strain is possible by this network, although tilt or twist should occur with the 60° dislocations (see Sec. 7.5.2). Likewise, for deposits in (111) orientation, misfit dislocations along [110], [101] and [011] form a triangular network of either ideal or 60° type. For most other orientations the pos-
[110]
[110] Figure 7-17. The slip planes on which dislocations can glide into the interface between an epitaxial layer of a f.c.c. metal and the (110) surface of another f.c.c. metal, to form misfit dislocations along AB. ABCD is the (111) plane and ABEF is the (111) plane.
sibilities are more limited, with no more than one <| \ 0> direction in the interface plane. Therefore the pseudomorphic strain can be relieved by ideal misfit dislocations, either in one direction only in the interface plane, or not at all. Complete relief of the elastic strain requires the formation of non-ideal misfit dislocations with their Burgers vectors inclined to the interface plane. The lack of much reported experimental evidence on misfit dislocation formation in deposits in orientations other than (001) and (111) makes it difficult to come to any firm conclusions, but what evidence there is highlights the need to study such orientations further. Postnikov et al. (1976) have studied the formation of misfit dislocations in thin layers of platinum on (001), (110), (102), (103), (111), (112) and (123) surfaces of gold. They show an example where all of the misfit dislocations in (110) interfaces lie along the [110] direction, with inclined Burgers vectors of < | \ 0> type (see Fig. 7-17). These are assumed to have formed as a result of slip on the (111) and (111) planes. Slip on the (Til) and (111) planes would not relieve any of the pseudomorphic elastic strain because these planes are perpendicular to the interface. Therefore the observed misfit dislocations relieve strain only in the [001] direction, and it is not possible to relieve strain in the [110] direction by slip involving the normal {111} i\ \ 0> system. Growth of platinum on the (102) gold surface demonstrates another important aspect. The first misfit dislocations, produced as strain relief commences, lie along the [211] and [2TT] directions, with inclined Burgers vectors (see Fig. 7-18). These two directions are at an angle of 48 degrees to each other, and it is not possible to relieve the strain equally in all directions in the interface with such an arrangement. Such an-
7.5 Elastic Strains and Misfit Dislocations
isotropy arises unless the two sets of dislocations are perpendicular. This probably explains why further relief of strain involves the formation of misfit dislocations along [23T] and [23T], since these do help to reduce the strain anisotropy. These two examples demonstrate that strain relief by pure slip will result in an anisotropic strain distribution in the plane of an epitaxial layer, for most orientations of the interface. There is little experimental evidence as to how complete strain relief can be achieved in such cases. On the basis of known mechanisms, it seems that either abnormal slip systems must operate or dislocation climb must take place. It is also of interest to note that whilst there are many examples of TEM images of misfit dislocations produced following the formation of pseudomorphic layers, there is a lack of systematic evidence showing the increase in density of the misfit dislocations as the deposit layer is thickened and elastic strain is fully relieved. The first misfit dislocations which form tend to be in small groups, or clusters, such as is shown in Fig. 7-18. The non-uniform spacing is often much greater than that required to relieve all of the misfit strain. Further increase in thickness is required to raise the stress to a level at which further misfit dislocations are formed, so that complete relief of stress occurs at a thickness well above tc, as shown in Fig. 7-11. It seems to be a common experience with semiconductor epitaxial layers, such as those of the III-V compounds, that pseudomorphic layers often grow to thicknesses considerably greater than the theoretical critical thickness tc (see Sec. 7.5.1), and that the reduction in strain occurs gradually as growth continues. It is widely believed that this arises because of the difficulty of nucleating and moving dislocations in these structures. Since dislocations
317 [211]
[211] Figure 7-18. Groups of misfit dislocations formed in the interface between a platinum deposit film and a (102) gold substrate, as observed by Postnikov et al. (1976).
move easily in many of the metal deposits under consideration, especially at the elevated growth temperatures, it seems likely that the main reason for any lack of misfit dislocations with the deposits of these metals relates to the difficulty of nucleation. Jesser and van der Merwe (1989) have recently reviewed the theory involved in predicting the critical thickness in epitaxy. 7.5.3.2 Formation During Volmer-Weber Growth
For many examples of the Volmer-Weber growth mode (e.g., metals deposited on non-metallic substrates), the misfit is sufficiently high to prevent growth of pseudomorphic nuclei. Consequently misfit dislocations can form during the initial stage of nucleation simply by the local distribution of the misfit at the interface in accord with the classic diagram for the formation of misfit dislocations (see Fig. 7-10). For very high values of the misfit, requiring misfit dislocations very closely spaced together (say by three atomic spacings or less), the concept of misfit dislocations becomes meaningless. But for smaller values of misfit requiring spacings of dislocations of (say) 5 atomic spacings or more, the local concentration of misfit is likely to occur so
318
7 The Epitaxy of Metals
that misfit dislocations are produced. Once the misfit dislocations have been formed at the interface between the nuclei and the substrate, further increase in the lateral size of the nuclei automatically causes the misfit dislocation network to extend as the interface extends. Once a continuous holefree deposit film is formed, a continuous misfit dislocation network is present, although the coalescence of nuclei can result in irregularities in the network coupled with the formation of threading dislocations (see Sec. 7.6.2). In some systems, the initial nuclei of an epitaxial deposit can be pseudomorphically strained. An example is the growth of tin on tin telluride (Vincent, 1969). In analogy with the monolayer growth mechanism, the elastic strain of the nuclei should be relieved by the formation of misfit dislocations. This can occur by the nucleation of dislocations at the edge of the nuclei (Fig. 7-19), and their movement along the interface. This limited amount of interfacial slip seems possible even if the interface plane is not a normal slip plane, because the nuclei are not constrained by any surrounding material from moving slightly as slip occurs, so that the stresses required for such slip could be significantly less than for bulk material. Vincent (1969) produced evidence, based upon the spacing of moire patterns, that the residual elastic strain in the tin nuclei varied in a sawtooth manner with the lateral dimension of the nuclei (Fig. 7-20). He interpreted this as due to
•l/i Substrate (a) (b) ' Figure 7-19. The nucleation of ideal misfit dislocations at the edge of nuclei, and their movement along the interface as in (a), resulting in the formation of an array of misfit dislocations as in (b).
build-up of elastic strain as the nuclei increased in size, followed by a significant drop in the residual strain as a new misfit dislocation moved into the interface from the edge of the nucleus. Takayanagi et al. (1975) pointed out that the tin nuclei should be molten at the growth temperature of 200 °C, raising doubts as to the validity of this interpretation. They failed to reproduce the sawtooth variation found by Vincent, but by carrying out in-situ deposition in the electron microscope they did obtain evidence for a sawtooth variation in the moire pattern spacing in images of individual solid nuclei during growth at 80 °C. A different explanation for this sawtooth variation was put forward (see also Honjo and Yagi, 1980), partly because no direct evidence could be found of the presence of misfit dislocations. This explanation is based upon the existence of linear atomic chains of tin, which are strained uniformly without the concentration of misfit required for the model of Fig. 7-10. The sawtooth variation is explained as a result of the changes of the linear chain between two different stable configurations. The movement of dislocations from the edge of a nucleus to provide misfit dislocations to relieve elastic strain was observed in the early work of Matthews (1966), for example with the growth of gold nuclei on platinum. Cabrera (1964, 1965) concluded that the amount of pseudomorphic elastic strain in a nucleus should decrease as the size of the nucleus increases, and that this would be accompanied by an increase in the number of misfit dislocations. This decrease is illustrated in Fig. 7-20, for tin on tin telluride which has a misfit of 7.8 % in the direction in which the measurements were made. The important difference between the mechanisms for introducing misfit dislocations at the interfaces of nuclei,
7.5 Elastic Strains and Misfit Dislocations
Strain
No. of misfit dislocations 4
5
6
7
8
9
10
3.0-
20
30 40 Island width (nm)
Figure 7-20. The sawtooth variation in the strain in nuclei of tin grown on a substrate of tin telluride, as a function of the width of the tin nuclei and as measured from the spacing of moire patterns (Vincent, 1969).
319
and those mechanisms required with deposits formed by monolayer growth, is that there is no requirement for any plastic deformation through the thickness of a nucleus. When misfit dislocations are introduced at the interface of the nucleus with the substrate, the elastic strain in the nucleus automatically becomes relaxed, whereas no such relaxation is possible in a continuous deposit film without associated plastic deformation (see Sec. 7.5.3.1). A further mechanism, which involves no plastic deformation through the thickness of the nucleus, is the introduction of misfit dislocations by existing grown-in threading dislocations, illustrated in Fig. 7-21, as proposed by Jesser and Matthews (1968 b). 7.5.3.3 Formation During Stranski-Krastanow Growth
Substrate
Figure 7-21. The stages in the generation of non-ideal misfit dislocations at the interface with nuclei, due to the motion of a substrate dislocation which is extended into the nucleus. The dislocation remaining in nucleus A can move out of the nucleus to complete the formation of a misfit dislocation below nucleus A.
There seems to be little or no experimental evidence on the mode of formation of misfit dislocations following the monolayer growth stage of the Stranski-Krastanow mechanism. This is because little electron microscopy has been carried out on metal deposits produced by this mode. Possibilities can be deduced by analogy with what occurs during Frank van der Merwe growth and Volmer-Weber growth. If relaxation of pseudomorphic elastic strain occurs before the 3D nucleation stage commences, the introduction of misfit dislocations should occur by the mechanisms discussed in Sec. 7.5.3.1. If nucleation occurs on top of a pseudomorphic layer, it seems likely that misfit dislocations will nucleate at the edges of the nuclei, as in Fig. 7-19, either immediately the nuclei form or during their subsequent growth. These misfit dislocations would have to propagate through the continuous layer beneath the nuclei, perhaps in the form of expanding dislocation loops.
320
7 The Epitaxy of Metals
7.6 Lattice Imperfections in Layers Grown by Epitaxy 7.6.1 Imperfection Structures Observed In addition to any misfit dislocations, epitaxial layers commonly contain a variety of lattice defects, although the layers can be largely defect free under ideal conditions of growth. The most common defect is the dislocation line which starts at the interface with the substrate and ends at the free surface of the layer. This is known as a threading dislocation, which can either follow a fairly straight path joining the two surfaces of the film or follow a more complex path between the two surfaces. In the latter case, there is likely to be an intersection between two or more dislocations giving rise to some kind of irregular network, although such arrangements have been more commonly observed in semiconductor films. This is partly because more electron microscope observations have been carried out on thick (say > 1 jim) semiconductor layers than on metal films. Dislocation densities up to 108 mm~ 2 have been reported. Stacking faults are found in many epitaxial layers, especially in f.c.c. metals grown on non-metallic substrates such as alkali halides. These planar faults normally extend through the thickness of the film
Partial Partial
Stacking fault Figure 7-22. A stacking fault of limited length produced by coalescence, and bounded by two partial dislocations.
and are terminated within the film by partial dislocations as shown in Fig. 7-22. Overlapping stacking faults can also occur, as well as microtwins. 7.6.2 Modes of Formation of Lattice Defects The main modes of formation of lattice defects can be classified as follows: (1) copying of defects from the substrate; (2) formation of defects linked with misfit dislocations; (3) introduction of defects during the coalescence of nuclei. In addition, it is known that defects can be introduced as a result of contamination of the substrate surface, but what follows relates to mechanisms which can operate under nominally clean conditions. 7.6.2.1 Copying from the Substrate If the substrate contains lattice imperfections which emerge at the surface on which the epitaxial growth takes place, extension of those imperfections into the epitaxial layer can occur. It would not be expected that imperfections would extend into a layer which has its own lattice spacings and is not pseudomorphic with the substrate, whether it grows as monolayers or as 3 D nuclei. This is because the relative positions of atoms in the deposit, at the interface, should be determined largely by their binding with the deposit. If, on the other hand, the epitaxial layer is pseudomorphic, it would be expected that imperfections would extend into the deposit, because the positions of the deposit atoms (e.g., in monolayer growth) should be determined entirely by the position of the atoms in the substrate surface. Whilst it is known that copying, or extension, of substrate dislocations into the deposit does oc-
7.6 Lattice Imperfections in Layers Grown by Epitaxy
cur in some cases (e.g., Matthews, 1975 b) and not in others, there is no systematic experimental evidence to support the kind of arguments given above. It would be helpful to know whether the occurrence of copying depends upon other material properties, in addition to pseudomorphism in the deposit, or upon the Burgers vector of the substrate dislocation. 7.6.2.2 Defects Linked with Misfit Dislocations Whether or not the formation of misfit dislocations involves the formation of defects threading through the deposit thickness, depends upon the mechanism by which the misfit dislocations are formed. For ideal pseudomorphic monolayer growth the formation of misfit dislocations, as the deposit film increases in thickness beyond the critical thickness tc, occurs in one of two ways (see Sec. 7.5.3.1). The first way involves the movement of threading dislocations copied from the substrate, and this mechanism introduces no new threading dislocations although rearrangement of the existing threading dislocations is involved. The second, and more important, class of mechanisms involves the nucleation of dislocation loops at the surface of the growing pseudomorphic deposit film. The basic process is illustrated in Fig. 7-14, which shows that as the loop expands and causes a misfit dislocation to form at the interface it leaves two threading dislocations AB and CD. These move apart as expansion of the loop continues, but they remain as threading dislocations unless they move to the edge of the film. In general, this is unlikely to happen since loops would be nucleated at many points on the surface of the film and each would produce a misfit dislocation of a limited length. Further expansion would cease ei-
321
ther as a result of the reductions in local stresses due to the generation of other misfit dislocation loops in neighbouring regions of the film, or possibly as a result of an obstacle such as a local region of impurity or another dislocation. Thus the number of threading dislocations produced per unit area will be given by twice the number of loops generated per unit area. This, in turn, will depend upon (a) the misfit, and the extent to which it is relieved; (b) the Burgers vectors of the dislocation loops, and (c) the average length of the misfit dislocations produced. On this simple basis, the number of threading dislocations, N is given by: N=
4/cmlO 10
per
(7-9)
where m in % is the misfit, a fraction k of which is accommodated by a square array of misfit dislocations of length / in nm and edge component of Burgers vector in the interface plane b e in nm. This is derived from Eq. (7-8). However, in practice, there is likely to be considerable interaction between the different expanding dislocation loops. For two loops on the same inclined plane, the interaction would result in annihilation of a pair of threading dislocations, as illustrated in Fig. 7-23. For two loops on near neighbouring parallel inclined planes, separated by a distance less than the spacing of misfit dislocations required to relieve all of the elastic strain, a pair of close dislocations of opposite sign would be produced. In general they would have mixed edge and screw character, and there would be a force of attraction between them. They could therefore annihilate each other by a combination of climb and crossslip. Consequently N might be very considerably reduced from the value given by Eq. (7-9). In fact, there is little experimental evidence of the number of threading dislo-
322
7 The Epitaxy of Metals
/\
~7
Deposit
J
Substrate
j Figure 7-23. The interaction between two expanding dislocation loops, of the same Burgers vector and on the same slip plane, resulting in the reduction of the number of threading dislocations.
cations remaining in a thin epitaxial layer, due to this mechanism of formation. This includes studies on epitaxial semiconductor layers, although it is generally assumed that the mechanism does lead to the formation of threading dislocations. In the same way, the more complex mechanisms summarised in Sec. 7.5.3.1 would also be expected to result in the formation of threading dislocations, and possibly also stacking faults.
lier, Matthews (1959) had proposed that stacking faults can be produced as two nuclei coalesce and Matthews and Allinson (1963) concluded that the often extensive formation of microtwins in f.c.c. metals grown on rocksalt result from coalescence. The effects can be considered by reference to Fig. 7-24, which represents the displacement misfit between three coalescing nuclei in terms of just one set of crystal planes normal to the substrate surface. According to the model, as each of the three pairs of nuclei join together as they grow larger, the planes shown will join so as to minimise the elastic strain energy. For certain values of the displacement misfits between the coalescing pairs of nuclei, such as for the example of Fig. 7-24, the joining will occur so as to produce an incipient threading dislocation in the hole between the nuclei (Jacobs et al., 1966). As the hole becomes filled in, a real threading dislocation is produced. Whilst the compound nucleus, or island, remains free from any further coalescence it is possible for the dislocation to move out of the island. Once a
7.6.2.3 Defects Resulting from Coalescence of Nuclei When nuclei form with their natural crystal spacings, and there is a misfit at the interface with the substrate, it follows that a line joining an interface atom in one nucleus to an interface atom in a neighbouring nucleus will not necessarily be an integral multiple of the normal interatomic distance. This arises because the substrate and deposit lattices are not commensurate, so that there will be a small random displacement of the lattices of the two nuclei. Jacobs et al. (1966) were the first to propose that these displacement misfits can lead to the formation of dislocations. Ear-
Figure 7-24. The formation of a threading dislocation by the coalescence of three nuclei which are not pseudomorphic with the substrate, so that a random displacement misfit exists between adjacent nuclei. The vertical lines represent a parallel set of crystal planes.
7.6 Lattice Imperfections in Layers Grown by Epitaxy
continuous network structure is formed, following further coalescences (see Sec. 7.4.2), any threading dislocations moving out of the islands effectively result in incipient threading dislocations in the holes into which they move. It follows that real threading dislocations are inevitably created once the holes become filled in. Liquidlike coalescence might prevent the formation of threading dislocations by the mechanism of Fig. 7-24 during the early stages when the nuclei and islands are quite small. It would not, however, prevent the mechanism operating during the later stages leading up to the network stage. Thus displacement misfit between islands, which is a necessary consequence of the misfit between substrate and deposit, does lead to the formation of threading dislocations. The experimental evidence in support of this conclusion has been obtained by insitu transmission electron microscopy and examples have been presented in earlier reviews (Pashley, 1965; Honjo and Yagi, 1980). The mechanism shown in Fig. 7-24 applies where there is a misfit between substrate and deposit, and no pseudomorphism. It applies irrespective of the presence of misfit dislocations. If the misfit is not too large for the formation of misfit dislocation (see Sec. 7.5.3.2) an alternative, and largely equivalent, mechanism of the formation of threading dislocations can be put forward, in terms of the joining up of misfit dislocations (Fig. 7-25). The misfit dislocations at the interface of neighbouring nuclei with the substrate will not necessarily be aligned in register with each other, so that as such nuclei coalesce some local bending of the misfit dislocations is required to allow them to join together. Depending upon the extent of disregistry between the misfit dislocations associated with each pair of the three coalescing is-
323
Figure 7-25. The formation of a threading dislocation as a result of the joining of misfit dislocations at the interface with three adjacent nuclei. The vertical lines represent a parallel set of misfit dislocations.
lands, one of the misfit dislocations (e.g. pq in Fig. 7-25) can be left terminating at the centre of the coalescence. Since dislocations cannot terminate inside a crystal, the misfit dislocation pq must be linked at q to a threading dislocation which forms during coalescence. In this situation the threading dislocation must have the same Burgers vector as that of the misfit dislocation pq. Thus, by this mechanism, the threading dislocations produced would be confined to those with Burgers vectors matching the Burgers vectors of the misfit dislocations, and there is no evidence as to whether or not this is normally the case. Subsequent interactions between threading dislocations could lead to the formation of dislocations with other Burgers vectors. Coalescence between two nuclei or islands can lead to the formation of a stacking fault if the displacement misfit between them has an appropriate value. For the f.c.c. system the exact displacement required is | <211 >. With parallel growth occurring on a (001) substrate there is no such vector in the surface plane, but there are <211 > directions at angles of 24 degrees to the surface, so that the appropriate dis-
324
7 The Epitaxy of Metals
placements in the surface plane can have large components parallel to a <211> direction. In the most favourable situation a displacement misfit of d along the [210] direction in the (001) surface has a component of 0.91 d along [211], so that an appropriate value of the displacement d would be largely accommodated by the formation of a stacking fault on coalescence. For growth on a (111) substrate there are <211> directions in the surface, so that appropriate displacements can be completely accommodated by stacking faults being formed during coalescence.
These examples demonstrate that some of the random displacements of nuclei due to misfit are likely to result in the formation of stacking faults during coalescence, as is actually observed (Jacobs et al., 1966; see Fig. 7-26). The evidence shows that such stacking faults can be eliminated by slip, whilst the deposit continues to consist of isolated nuclei or islands. During the later stages of growth the elimination of stacking faults results in the formation of incipient threading dislocations, just as for the above case of the removal of dislocations at the network stage.
Figure 7-26. The growth of gold on MoS 2 , inside the electron microscope, showing the formation and elimination of stacking faults, F 1 , F 2 , F 3 , F 4 , following coalescence (From Jacobs et al., 1966). 25nm
7.6 Lattice Imperfections in Layers Grown by Epitaxy
In addition to the formation of thin microtwins threading through an epitaxial deposit, and also believed to result from coalescence of nuclei (Matthews and Allinson, 1963), the presence of multitwinned particles of f.c.c. metals formed during the early stags of epitaxial deposition has been observed by transmission electron microscopy. They have been identified by several workers for growth on alkali halide cleavage surfaces (e.g., Ino and Ogawa, 1967), and have also been observed for growth on mica cleavage surfaces (Allpress and Sanders, 1964). In their simplest form they consist of a tetrahedron bounded by (111) surfaces together with a primary twin and a secondary twin on each of two of the faces of this tetrahedron, as illustrated in Fig. 7-27. This produces a compound particle which has five-fold symmetry when viewed along OA. More complex particles, involving larger numbers of tetrahedral components, appear to have hexagonal symmetry. Other particles containing five tetrahedral components appear as rhombic shapes. The number of multiply twinned particles present in the early nucleation stages decreases as growth continues. It is believed that this is largely due to recrystallization occurring during coalescence (see Sec. 7.6.3). The size of the remaining particles increases during continued deposition. There is some evidence that the multiply twinned particles are formed initially by coalescence, although they can also be considered as originating from earlier stages of nucleation, in the form of spontaneously strained structures (Ogawa and Ino, 1972). 7.6.3 Changes in Imperfection Structure as Growth Proceeds
Once a continuous hole-free and largely strain free epitaxial layer has been formed,
B
325
C (a)
(b)
Figure 7-27. A model for the formation of five-fold multiply twinned particles, (a) OABC is the basic (111) oriented tetrahedron, on two faces of which the primary twin tetrahedra OACL and OABF are formed. The secondary twin tetrahedra are OALD and OAFE. Closing of the gap ED gives the five-fold symmetry of the particle when viewed along OA, as shown in (b).
further growth can proceed by continuing the orientation of the deposit by a monolayer process. It is of interest to consider how the lattice defects present in the layer will develop as the layer thickness increases. It would not be expected that any formation of new independent defects would take place, since none of the mechanisms discussed in Sec. 7.6.2 should apply to homoepitaxy or growth with a zero misfit.
326
7 The Epitaxy of Metals
There seems to be little evidence, in the case of metal epitaxy, as to how the imperfection structure changes with increasing thickness. Many studies have been made of thicker layers of epitaxial semiconductors, since layers of 1 jim to 5 jam in thickness are commonly used for the investigation of optical and electronic properties. It is commonly found that the dislocation density in the epitaxial layer decreases with distance from the interface, and that the surface regions of such layers can be relatively perfect. Existing threading dislocations must extend into the epitaxial layer as growth proceeds but, under the action of local stresses, they can change their direction as they extend in length. In so doing, they can interact with other dislocations, joining with them and possibly forming irregular networks as several dislocations join together. As this joining process occurs, the further propagation of the joining dislocations into the growing layer ceases. In this way the number of threading dislocations reduces with thickness, and irregular networks containing dislocations running approximately parallel to the surface are left below the growing surface. These networks can be modified, and simplified, by slip and/or climb processes if growth continues at a suitably high temperature. This general picture seems to be consistent with the observations made on semiconductor layers, and it seems likely that similar changes will occur with epitaxial metal layers, so that a reduction in defect density occurs as thickness increases.
7.7 Summary and Conclusions Epitaxy has been observed with a range of metals, but much emphasis has been put on the growth of f e e . metals. When f e e . metals are grown on single crystal metals
as substrates, the pseudomorphic monolayer growth mechanism applies in a number of cases, whereas the formation of 3 D nuclei by the Volmer-Weber mechanism occurs in others. For a few cases the Stranski-Krastanow mechanism applies. When f e e . metals are grown on non-metallic single crystals the most common mode of growth is 3 D nucleation. Much of the existing evidence relates to growth on alkali halide substrates. It seems likely that epitaxy of metals can be obtained on a wide range of substrates, under appropriate deposition conditions. A recent summary of some of the factors which are inadequately understood, together with a discussion of the present understanding of epitaxy, has been published by Bauer et al. (1990). In common with other epitaxial thin films, especially those of semiconductors which have been studied rather extensively over the last 10 to 15 years, epitaxial deposits of metals commonly contain high densities of lattice defects threading through the thickness of the films. This applies generally for growth by the YolmerWeber 3 D nucleation mechanism, and only for growth by the Frank van der Merwe monolayer mechanism do the epitaxial metal films contain really low defect densities. In these latter cases the defects extending through the thickness of the film appear to be formed mainly as a result of the mechanisms which operate to provide the interfacial misfit dislocations required to accommodate the misfit between the substrate and deposit. There is a lack of systematic evidence on the development of both interfacial misfit dislocations, and defects extending through the thickness of an epitaxial deposit, as the growth of an epitaxial layer proceeds. Various models exist for explaining and predicting how the various kinds of defect structure are formed, and these
7.8 References
models are supported by much experimental evidence, but a number of detailed aspects are not covered by the models and the limitations and range of applicability are not adequately understood. If, as seems likely, the use of epitaxial deposits of metals and alloys for studies of, for example, magnetic and superconducting properties increases there will be need for an increase in knowledge on defect structures and their formation. This need would increase further as technological applications of epitaxial metal layers develop. One area of considerable current interest is the growth of superlattice structures, including strained layer superlattices (see this Volume, Chap. 8). The MBE technique has allowed such superlattices to be grown successfully, but there are likely to be limitations on the metallic pairs which can be used because monolayer growth will often not occur both for metal A on metal B and for metal B on metal A. It will be interesting to see whether the use of alloys, rather than pure metals, increases the scope for growth of good quality superlattices.
7.8 References Allpress, J. G., Sanders, I V. (1964), Phil. Mag. 10, 645. Ball, C. A. B. (1970), Phys. Status Solidi 42, 357. Ball, C. A. B., van der Merwe, J. H. (1970), Phys. Status Solidi 38, 335. Bassett, G. A. (1958), Phil. Mag. 3, 1042. Bassett, G. A. (1960), in: Proc. Eur. Conf. Electron Microsc. Delft: Houwink, A. L., Spit, B. J. (Eds.). Nederlandse Vereniging voor Electromenmicroscopie, p. 270. Bassett, G. A., Menter, J. W., Pashley, D. W. (1958), Proc. Roy. Soc. A246, 345-368. Bauer, E. (1958), Z. Kristallogr. 110, 372. Bauer, E., Poppa, H. (1972), Thin Solid Films 12,167. Bauer, E., van der Merwe, J. H. (1986), Phys. Rev. B33, 3657. Bauer, E., Dodson, B. W., Ehrlich, D. X, Feldman, L. C , Flynn, C. P., Geis, M. W, Harbison, J. P., Matyl, R. I , Peercy P. S., Petroff, P. M., Phillips,
327
J. M.? Stringfellow, G. B., Zangwill, A. (1990), J. Mater. Res. 5, 852. Bethge, H., Keller, K. W., Ziegler, E. (1968), /. Cryst. Growth 3, 184. Bruck, L. (1936), Ann. Phys. 26, 233. Cabrera, N. (1964), Surf. Sci. 2, 320. Cabrera, N. (1965), Mem. Sci. Rev. Mat. LXII, 205. Cherns, D., Stowell, M. J. (1975a), Thin Solid Films 29, 107. Cherns, D., Stowell, M. J. (1975 b), Thin Solid Films 29, 127. Cherns, D., Stowell, M. J. (1976), Thin Solid Films 37', 249. Cunningham, J. E., Flynn, C. P. (1985), J. Phys. F: Met. Phys. 15, L221. Durbin, S. M., Cunningham, J. E., Flynn, C. P. (1982), J. Phys. F: Met. Phys. 12, L75. Eaglesham, D. J,, Kvam, E. P., Maher, D. M., Humphreys, C. I , Bean, J. C. (1989), Phil. Mag. A 59, 1059. Esaki, L., Tsu, R. (1970), IBM J. Res. Develop. 14, 61. Finch, G. I., Quarrell, A. G. (1933), Proc. Roy. Soc. A 141, 398.
Finch, G. L, Quarrell, A. G. (1934), Proc. Phys. Soc. Lond. 46, 148. Flynn, C. P. (1988), J. Phys. F: Met. Phys. 18, L195. Frank, F. C , van der Merwe, J. H. (1949a), Proc. Roy. Soc. A 198, 205. Frank, F. C , van der Merwe, J. H. (1949b), Proc. Roy. Soc. A 200, 125. Frank, F. C , van der Merwe, J. H. (1950), Proc. Roy. Soc. A 201, 261. Frankenheim, M. L. (1836), Ann. Phys. 37, 516. Gejji, F. H., Holt, D. B. (1978), J. Mater. Sci. 13, 2048 Grunbaum, E. (1975), in: Epitaxial Growth: Matthews, J. W. (Ed.). New York: Academic Press, pp. 611-673. Harris, J. X, Joyce, B. A., Dobson, P. X (1981 a), Surf. Sci. 103, L90. Harris, J. X, Joyce, B. A., Dobson, P. X (1981 b), Surf. Sci. 108, L444. Honjo, G., Takayanagi, K., Kobayashi, K., Yagi, K. (1977), /. Cryst. Growth 42, 98. Honjo, G., Yagi, K. (1980), in: Current Topics in Materials Science Vol. 6. Amsterdam: North Holland Publishing Company, pp. 195-307. Hsieh, T. C , Chiang, T. C. (1986), Surf. Sci. 166, 554. Hull, R., Bean, X C , Warder, D., X, Leibenguth, R. E. (1988), Appl. Phys. Lett. 52, 1605. Igarashi, O (1971), J. Appl. Phys. 42, 4035. Ino, S., Ogawa, S. (1967), J. Phys. Soc. Japan 22, 1365. Jacobs, M. H., Pashley, D. W, Stowell, M. X (1966), Phil. Mag. 13, 129 Jain, S. C , Willis, X R., Bullough, R. (1990), Adv. Phys. 39, 127. Jalochowski, M., Bauer, E. (1988 a), Phys. Rev. B37, 8622. Jalochowski, M., Bauer, E. (1988 b), Phys. Rev. B38, 5272.
328
7 The Epitaxy of Metals
Jesser, W. A., Matthews, X W. (1967), Phil. Mag. 15, 1097. Jesser, W. A., Matthews, J. W. (1968 a), Phil. Mag. 17, 461. Jesser, W. A., Matthews, J. W. (1968 b), Phil. Mag. 17, 595. Jesser, W. A., van der Merwe, J. H. (1989), in: Dislocations in Solids. Vol. 8: Nabarro, F. R. N. (Ed.). Amsterdam: North Holland, pp. 421-496. Jin, B. Y, Ketterson, J. B. (1989), Adv. Phys. 38, 189. Kern, R., Le Lay, G., Metois, J. J. (1979), in: Current Topics in Materials Science Vol. 3. Amsterdam: North Holland Publishing Company, pp. 131-419. Kingery, W. D., Berg, M. (1955), J. Appl. Phys. 26, 1205. Kwo, J., McWhan, D. B., Hong, M., Gyorgy, E. M., Feldman, L. C , Cunningham, J. E. (1985), in: Layered Structures, Epitaxy and Interfaces, Pittsburgh: Materials Research Society; Symposia Proceedings Vol. 37, pp. 509-515. Kuk, Y, Feldman, L. C , Silverman, P. J. (1983), Phys. Rev. Lett. 50, 511. Markov, I., Stoyanov, S. (1987), Contemp. Phys. 28, 267. Matthews, J. W. (1959), Phil. Mag. 4, 1017. Matthews, J. W. (1965), Phil. Mag. 12, 1143. Matthews, J. W. (1966), Phil. Mag. 13, 1207. Matthews, J. W. (1975 a), /. Vac. Sci. Technol. 12, 126. Matthews, J. W. (1975 b), in: Epitaxial Growth, Matthews, J. W. (Ed.) New York: Academic Press, pp. 560-609. Matthews, J. W, Allinson, D. L. (1963), Phil. Mag. 8, 1283. Matthews, J. W, Grunbaum, E. (1965), Phil. Mag. 11, 1233. Matthews, J. W., Crawford, J. L. (1970), Thin Solid Films 5, 187. van der Merwe, J. H. (1963), /. Appl. Phys. 34, 117. Milne, R. H. (1990), in: Supplementary Volume 2 of the Encyclopedia of Materials, Science and Engineering: Cahn, R. W. (Ed.). Oxford: Pergamon, pp. 1231-1234. Neave, J. H., Joyce, B. A., Dobson, P. J., Norton, N. (1983), Appl. Phys. A31, 1.
Ogawa, S., Ino, S. (1972), J. Cryst. Growth 13/14, 48. Osakabe, N., Tanishiro, Y, Yagi, K., Honjo, G. (1980), Surf Sci. 97, 393. Pashley, D. W. (1956), Adv. Phys. 5, 173. Pashley, D. W. (1965), Adv. Phys. 14, 327. Pashley, D. W., Stowell, M. X, Jacobs, M. H., Law, T. X (1964), Phil. Mag. 10, 127. Pautikis, V., Sindzingre, P. (1987), Physica Scripta T 19, 375. Poppa, H., (1969), /. Vac. Sci. Technol 2, 42. Postnikov, V. S., Ijevlev, V. M., Solovjev, K. S. (1976), Thin Solid Films 32, 173. Royer, L. (1928), Bull. Soc.frang. Min. 51, 1. Seifert, H. (1953), in: Structure and Properties of Solid Surfaces: Gomer, R., Smith, C. S. (Eds.). Chicago: University Press, p. 218. Steigerwald, D. A., Egelhoff, W. F. (1987), Surf. Sci. 192, L887. Takayanagi, K. (1985), in: Layered Structures and Epitaxy, Pittsburgh: Materials Research Society; Symposium Proceedings Vol. 56, pp. 129-138. Takayanagi, K., Yagi, K., Kobayashi, K., Honjo, G. (1975), J. Cryst. Growth 28, 343. Takayanagi, K., Yagi, K., Kobayashi, K., Honjo, G. (1978 a), /. Phys. E: Sci. Instrum. 11, 441. Thompson, C. V., Floro, X, Smith, H. I. (1990), /. Appl. Phys. 67, 4099. Valdre, U., Robinson, E. A., Pashley, D. W, Stowell, M. X, Law, T. X (1970), /. Phys. E: Sci. Instrum. 3, 501. Vincent, R. (1969), Phil. Mag. 19, 1127. Venables, X A., Price, G. L. (1975), in: Epitaxial Growth: Matthews, X W. (Ed.) New York: Academic Press, pp. 382-436. Venables, X A., Derrien, X, Janssen, A. P. (1980),
Surf Sci. 95,411. Venables, X A., Spiller, G. D. T., Hanbuchen, M. (1984), Rep. Progr. Phys. 47, 399. Walton, D. (1962), Phil. Mag. 7, 1671. Willis, X R., Jain, S. C , Bullough, R. (1990), Phil. Mag. A62, 115. Yagi, K., Takayanagi, K., Kobayashi, K., Honjo, G. (1971), J. Cryst. Growth 9, 84.
8 Metallic Multilayers A. Lindsay Greer and Robert E. Somekh
Department of Materials Science and Metallurgy, University of Cambridge, Cambridge, U.K.
List of 8.1 8.2 8.2.1 8.2.2 8.2.3 8.2.4 8.2.5 8.3 8.3.1 8.3.2 8.3.3 8.3.4 8.3.5 8.3.6 8.3.7 8.3.8 8.4 8.4.1 8.4.2 8.4.2.1 8.4.2.2 8.4.2.3 8.4.2.4 8.4.3 8.4.3.1 8.4.3.2 8.4.3.3 8.4.3.4 8.4.3.5 8.4.3.6 8.4.3.7 8.4.4 8.4.4.1 8.4.4.2
Symbols and Abbreviations Introduction Structure Classification Composition Modulation Two-Phase Multilayers Interfacial Structure Stability of Multilayer Structure Properties Origins of Special Properties X-Ray and Neutron Reflectivities Normal-State Electron Transport Superconductivity Magnetic Properties Mechanical Properties Other Properties Summary Preparation Introduction Thin Film Growth Processes General Considerations Surface Mobility Development of Microstructure and Morphology Internal Stress Deposition Technologies General Considerations Evaporation Sputtering Chemical Vapor Deposition Electrolytic Deposition Atomic Layer Epitaxy Mechanical Reduction Process Control Layer Thickness Size and Uniformity
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
331 332 332 332 333 333 333 336 337 337 338 339 340 341 343 345 345 346 346 346 346 347 348 349 350 350 350 351 351 352 352 352 352 352 354
330
8.4.4.3 8.4.4.4 8.4.4.5 8.4.5 8.4.5.1 8.4.5.2 8.4.5.3 8.5 8.5.1 8.5.2 8.5.3 8.5.4 8.5.5 8.5.6 8.6 8.7 8.8
8 Metallic Multilayers
Deposition Rate Buffer Layers Level of Vacuum Summary, Issues and Outlook Comparison of Evaporation and Sputtering Issues in Multilayer Deposition Outlook for Ideal Multilayers Structural Characterization Introduction X-Ray Diffraction Neutron Diffraction Transmission Electron Microscopy Techniques for Chemical Profiling Probes of Local Structure Applications and Outlook Acknowledgements References
355 355 356 356 356 358 360 360 360 361 365 365 367 368 368 368 369
List of Symbols and Abbreviations
List of Symbols and Abbreviations a D D Do HC2 kB Mr, Ms Q T t Tc Tm Y
lattice parameter diffusion coefficient interdiffusivity pre-exponential term of the diffusion coefficient upper critical field Boltzmann constant remanent magnetization, saturation magnetization activation energy temperature time for growth of one monolayer superconducting transition temperature melting temperature biaxial in-plane elastic modulus
9 0 9k X v
coverage in monolayers Bragg angle rotation angle in magneto-optic Kerr effect radiation wavelength lattice vibration frequency
ALE b.c.c. c.c.p. CVD EXAFS LUCS MBE MOCVD NMR PACVD RHEED RKKY TEM UHV XHV YSZ
atomic layer epitaxy body-centered cubic cubic close-packed chemical vapor deposition extended X-ray absorption fine structure layered ultrathin coherent structures molecular beam epitaxy metal-organic chemical vapor deposition nuclear magnetic resonance plasma-assisted chemical vapor deposition reflection high-energy electron diffraction Ruderman-Kittel-Kasuya-Yosida (exchange interaction) transmission electron microscopy ultra-high vacuum extra high vacuum yttrium-stabilized zirconia
331
332
8 Metallic Multilayers
8.1 Introduction Metallic multilayers epitomize the refinement of metal thin film deposition processes. They represent what might be called materials engineering on an atomic scale, with structures made up of layers only two to three atomic monolayers thick. From initial work half a century ago which was aimed at using metallic multilayers to calibrate X-ray wavelengths, there are now established applications in X-ray optics. There are other potential applications, notably as thin film magnetic and magnetooptic recording media. The commercial applications of metallic multilayers so far are belittled by the development of semiconductor multilayers (Volume 4, Chapter 8), which have created the new field of bandstructure engineering and revolutionized semiconductor device design. Nonetheless, the properties of metallic multilayers have been explored in a wide variety of materials research studies which can take advantage of the precise artificial, finescale layering which is possible. Metallic multilayers remain a particularly fruitful area for research because of their unique properties, and work on them is aided by recent advances in materials characterization. In this chapter we first describe the structural types of these new materials in Section 8.2, before introducing their wideranging properties in Section 8.3. Section 8.4 is the core of the chapter, covering preparation methods and highlighting the recent developments which now permit the routine deposition of these fine-scaled structures. Characterization methods which have aided the development of multilayer research are outlined in Section 8.5. Metallic multilayers, while as yet less important than semiconductor multilayers,
have great potential, and Section 8.6 briefly considers their applications and the outlook for them.
8.2 Structure 8.2.1 Classification In essence a multilayer is made by the alternating deposition of two different materials. After perhaps the first few layers, it is assumed that the structure of all the layers of one material are the same. The structure of each material is clearly of importance for the properties of the multilayer, not only in itself, but also for the influence it can have on the structure of the other material; each material acts as a substrate for the deposition of the other. In metallic multilayers, the materials may be amorphous, polycrystalline, or monocrystalline - a much wider range of structures and combinations of structures than is common in semiconductor multilayers. Yet not all combinations of structures are possible. Single crystal thin films can normally be obtained only by epitaxial growth on a monocrystalline substrate; the thin film will then have a well defined orientation relationship with the substrate (see Chapter 7). Polycrystalline deposits normally exhibit a degree of preferred orientation. A polycrystalline thin film, even on an amorphous substrate, can often be annealed to achieve a large grain size (much greater than the film thickness) with a particular crystallographic axis perpendicular to the substrate. Such a mosaic structure may have many of the properties of a monocrystalline film, though the orientations of the grains about the normal to the substrate are not defined. A monocrystalline layer will contain defects of a type and density dependent on deposition conditions and the perfection of the substrate.
8.2 Structure
Depending on the degree of preferred orientation and grain size in comparison to layer thickness, there can thus be a range of structures and properties between the random polycrystal and the single crystal. If a mosaic layer is used as the basis for successive deposition of epitaxial crystalline layers, then for most purposes the multilayer will behave as though it were monocrystalline. Thus it is possible to have multilayers of the following types: amorphous/ amorphous, polycrystalline/polycrystalline, amorphous/polycrystalline, monocrystalline/monocrystalline (including mosaic/ mosaic), and amorphous/mosaic. The other combinations are unlikely. In monocrystalline/monocrystalline multilayers it is expected that there is an epitaxial relationship between the layers defining the orientation relationship between successive layers and ensuring that all layers of the same type are in identical orientation. In amorphous/mosaic multilayers, on the other hand, there is not expected to be any orientation relationship between the mosaic layers (other than the common direction normal to the substrate). 8.2.2 Composition Modulation The simplest multilayer structures to describe and analyse are those which consist of a composition modulation imposed on a single structure. In almost all cases of this type, intermixing can lead to a uniform single phase of the starting structure. Compositionally modulated amorphous alloys do not exhibit the special effects associated with the relationship between the repeat distance of the multilayer and crystallographic spacings. Such effects can arise when a composition modulation is imposed on a single crystal (a monocrystalline film, or one grain in a mosaic film). Such multilayers, in which each layer has
333
the same structure in the same orientation, may properly be termed superlattices. Semiconductor multilayers have mostly been of this type, and often approximate to the special case in which each layer consists of an integral number of crystallographic periods. Polycrystalline/polycrystalline multilayers in which the two materials have the same crystal structure may also be regarded as compositionally modulated, though the orientations of the grains and relationship between the orientations in successive layers may not be well defined. 8.2.3 Two-Phase Multilayers While compositionally modulated multilayers may be regarded as single phase, there are clearly examples of two-phase multilayers, in which the two materials have different structures and in which simple homogenization is not possible. If twophase multilayers are annealed, one material may dissolve in the other or react with it to yield a third phase. Alternatively, the two materials may be stable in contact with each other, in which case the only possible effect of annealing would be coarsening of the layer pattern. The two materials, though of different structure, may have an orientation relationship. The resulting multilayers (monocrystalline/ monocrystalline, or mosaic/mosaic) have been termed layered ultrathin coherent structures (LUCS) (Schuller, 1980). In these there may be a full epitaxial relationship, as in a superlattice, or only the crystallographic directions normal to the substrate may be defined. The degrees of order in a multilayer have been discussed by McWhan (1985). 8.2.4 Interfacial Structure In multilayers of whatever type, the nature of the interfaces is of great impor-
334
8 Metallic Multilayers
tance. Even if the average repeat distance is well defined, there may be local variations, i.e., the interfaces may be rough. While roughness can be controlled by varying production parameters, keeping it to low values is not straightforward. For a smooth thin film to be obtained on a substrate, the material must wet the substrate, or, if not, the atomic mobility in the deposit must be insufficient for agglomeration to occur. But (at least for macroscopic film thicknesses) if one material "A" wets another material "B", B cannot wet A; thus deposition of a high quality multilayer relies at least partly on lack of mobility. Simply lowering the substrate temperature to ensure lack of mobility is not acceptable, because there must be enough mobility to allow the desired structure to be achieved in both materials (unless they are both amorphous). Since the temperature scale for atomic mobility is normalized with respect to the melting point Tm of the material, problems are most likely to arise for multilayers in which the two materials have widely differing Tm. If we suppose that A wets B (and therefore that B does not wet A), then a good multilayer can still be obtained if T m (A)< r m (B). If T m (A)> Tm(B), however, a substrate temperature low enough to stifle the agglomeration of B into islands is likely to be too low to permit successful deposition of monocrystalline A. A possible solution is to change substrate temperature from layer to layer, as discussed in Section 8.4.5.2. Distinct from interfacial roughness, yet experimentally difficult to distinguish from it by many techniques, is interfacial diffuseness. This may arise from interdiffusion during deposition or subsequently. Since diffuseness corresponds to material of intermediate composition, which roughness per se does not, it can lead to distinct effects on multilayer properties.
When the two materials in the multilayer are monocrystalline, the structure of the interface is important. The considerations to be discussed here apply also to the single, layered grains in a mosaic multilayer. When the two materials are of the same phase (i.e., in a compositionally modulated single crystal), the interface may be fully coherent, or partially coherent (Fig. 8-1). In the former case, the interfaces are free of dislocations and lattice matching is achieved by having compensating average strains in the layers, compressive in one material, tensile in the other. Poisson's ratio effects lead to an accentuation of the lattice parameter difference in the direction normal to the substrate. The strains may be considerable and can substantially alter the material properties. When the interfaces are partially coherent through the presence of interfacial dislocations, there is local strain around the dislocations but the average strain in the layers is zero. There may be significant differences in multilayer properties between the coherent and partially coherent cases. For example, in a compositionally modulated alloy, the diffusional mixing which occurs on annealing is much faster in the coherent multilayer, because of the extra driving force provided by the average layer strains (Philofsky and Hilliard, 1969) (Fig. 8-2). The degree of coherency may be governed by strain energy minimization or by the critical strain necessary for dislocation nucleation or motion. Details may be found in Chapter 7. For thin layers, coherency is always preferred; in multilayers with layer thickness or repeat distance greater than a critical value, there is partial coherency. A special case of two materials with the same structure in a multilayer is when one of the materials adopts a non-equilibrium structure in order to match the other. This pseudomorphism arises because the free en-
8.2 Structure
335
Figure 8-1. Schematic drawings of (a) coherent, and (b) partially coherent layered structures. In the coherent structure the planes perpendicular to the substrate are of constant spacing and alternating layers are under tension and compression. In the partially coherent structure the strains are relieved by the introduction of misfit dislocations. [From McWhan(1985).]
ergy of the multilayer is lowered when the interfaces can be coherent or partially coherent as is possible with matching structures (and similar lattice parameters). The non-equilibrium structure can be a known metastable phase. For example, in Nb/Zr multilayers, contact with the b.c.c. niobiX (nm) 16
^E
4 3 1 1
14-
o
1 2--
N
2
1.5
1
I
1.2 1
1.0 1
-
10-
^O ?o
6-
o
4I
1
i
I
I
I
10
15
20
25
30
35
B 2 (10 18 m' 2 )
Figure 8-2. The variation of the effective interdiffusivity D at 389 °C with repeat distance X in compositionally modulated Cu/Pd mosaic multilayer. The quantity B is defined by B = (2/d2) [1 - cos(2 n d/X)], where d is the spacing of the atomic planes parallel to the substrate (in this case (111)). The layers were fully coherent for X < 2.8 nm (giving enhanced D), and incoherent for X > 3.8 nm. [From Philofsky and Hilliard (1969).]
um stabilizes the b.c.c. high temperature allotrope of zirconium (Lowe and Geballe, 1984). But in other cases, the non-equilibrium phase is quite novel: in Mo/Ge multilayers the germanium adopts a b.c.c. structure to match the molybdenum (Wilson and Bienenstock, 1988). The interfaces in a multilayer may be ordered or disordered (Clemens and Gay, 1987). An example of an ordered interface is the fully coherent interface between two layers with the same crystal structure and a small lattice mismatch. But ordered interfaces can also exist when there is epitaxy between materials with different crystal structures, e.g., b.c.c. niobium and c.c.p. copper. When the interface is ordered, the fluctuations in layer thickness must be a multiple of a crystal plane spacing, giving long-range coherence through the thickness of the multilayer. This coherence is readily detectable by X-ray diffraction (see Section 8.5.2) which shows high-angle superlattice lines. When the interface is disordered, there can be a continuous distribution of layer thickness, long-range coherence is lost, and the high-angle superlattice lines are absent (though those at low angle remain). Disordered interfaces are
336
8 Metallic Multilayers
most likely for materials with a large structural mismatch. The structure of the interfaces in a multilayer will determine their energy. The interfacial energy will in most cases be a major component of the amount by which the multilayer's free energy exceeds that of the equilibrium state of the same average composition. It is also possible that the interfaces may possess an interfacial stress, i.e., an effective tensile stress capable, in the absence of substrate effects, of holding the multilayer in a state of biaxial in-plane compression (Cammarata and Sieradski, 1989). 8.2.5 Stability of Multilayer Structure Although multilayered structures can be found in equilibrium in natural systems (e.g., dichalcogenides), most artificial metallic multilayers have free energies far in excess of equilibrium and are susceptible to some type of transformation if there is sufficient atomic mobility. Contributing to the excess free energy are the interfacial free energy, the strain energies and excess chemical energy relative to a mixed composition. Stability is clearly important if the special properties of multilayers are to be exploited. The simplest type of structural change in a multilayer is diffusional mixing at the interfaces. The increased interfacial diffuseness and the reduced amplitude of the composition modulation may affect many properties. The repeat distance can also change. The individual layer materials may show changes in structure. Amorphous layers may crystallize. The crystallization temperatures may be raised or lowered by interaction with the surrounding layers, possibly as a result of interdiffusion (Sevenhans etal., 1988). Polycrystalline layers
and mosaic layers may show grain growth. The moving grain boundaries are paths for fast diffusion, and as they sweep through the multilayer the composition modulation is destroyed. Thus for retention of a composition modulation it is essential to have a large stable grain size, or amorphous phases (Greer and Spaepen, 1985). The melting point of a material in a multilayer can be significantly depressed, as observed for example for lead layers only a few nm thick in a Pb/Ge multilayer (Willens etal., 1982). Both interfacial energy and chemical mixing can contribute to this effect. A further type of structural change is reaction between the materials of the multilayer to give one or more new phases. If the multilayer is composed of elements with a strongly negative enthalpy of mixing, the heat released when the reaction starts may be sufficient to allow it to proceed explosively. This has been observed in transition metal/amorphous silicon multilayers (Clevenger et al., 1988). The phase which forms by reaction in a multilayer may itself be metastable. Even if the two materials in the multilayer are stable in contact with each other, the multilayer configuration may not be stable. On annealing, some coarsening may be expected in which the density of interfaces will be lowered. This coarsening may take the form observed in lamellar eutectics (Graham and Kraft, 1966), but it has not been studied in artificial multilayers. While stability is essential for applications, the structural changes which can occur in multilayers are of considerable scientific interest. Multilayers are particularly useful for studies of interfacial reactions and of interdiffusion. Most of the quantitative work so far has been on interdiffusion (Volume 5, Chapter 2, Section 2.2.2.3). This has mostly been achieved
8.3 Properties
by measurement of X-ray satellite intensities during annealing. The technique is useful because interdiffusivities down to ~ 10~ 2 7 m 2 s" 1 can be measured. This sensitivity, possible because of the short diffusion distances, is at least one thousand times better than for techniques based on composition profiling (Greer and Spaepen, 1985), and permits measurement at low temperature in metastable materials and unrelaxed structures. A potential complication for determination of interdiffusivity, Z>, is the dependence of D on the repeat distance of the multilayer, an effect which becomes apparent at the very high concentration gradients achievable in deposited multilayers. Yet this dependence of D is useful, for it is related to the mixing thermodynamics of the system. Strain effects, mainly due to coherency, can also have a marked effect on D (Fig. 8-2). Interdiffusivity in both metals and semiconductors (in each case either crystalline or amorphous) has been studied using multilayers, as reviewed by Greer and Spaepen (1985).
8.3 Properties 8.3.1 Origin of Special Properties The fine repeat distances in artificial multilayers can give rise to special properties and the precise control of layer thickness leads to a wide range of possibilities for tailoring of properties. The special properties, which may be improvements of normal properties or may be unique to multilayers, are in some cases of interest for applications; in other cases they are of interest in providing opportunities to test the theory of the origin of the property. The properties of semiconductor/semicon-
337
ductor superlattices are already widely exploited. Although not so developed in their applications, metallic multilayers have aroused much interest; their physical properties are treated elsewhere in this Series (Volume 3, Chapter 6). In this chapter, however, it is important to pay some attention to the origin of the special properties of multilayers in order to assess the demands placed on processing to achieve a desired property. Most theories of physical properties contain at least one characteristic length. When the repeat distance of a multilayer becomes comparable with that length, effects due to the layering can be expected. Examples of characteristic lengths are: the wavelength of incident radiation for reflectivity; the electron mean free path and the de Broglie wavelength for electrical resistivity and other normal-state transport properties; the exchange interaction length for magnetism; and the penetration depth and coherence length for superconductivity. These lengths vary from about 5 nm to 1 jim, setting quite varied requirements for multilayer structure, particularly for the perfection of the interfaces. In detail, the possible origins of the distinctive properties of multilayers are (classification after Schuller (1988)): - thin film effects, due to the limited thickness of one or other type of layer; - interface effects, arising from the interactions between neighboring layers; - coupling effects between layers of the same type, acting through the intervening layers; and - periodicity effects from the overall periodicity of the multilayer. Without attempting a comprehensive review, some examples are given below of the properties of metallic multilayers; where possible, reference is made to published reviews.
338
8 Metallic Multilayers
8.3.2 X-Ray and Neutron Reflectivities The reflection of X-rays provided the motivation for most of the early studies of metallic multilayers. The aim was to use multilayers of known repeat distance to calibrate the X-ray wavelength, and later to develop useful monochromators. The first successful work was by Deubner (1930). A more detailed analysis followed in the work of DuMond and Youtz (1940). They produced a multilayer with a repeat distance of about 10 nm by modulating the gold content in an evaporated polycrystalline copper deposit. A first-order Bragg reflection was attributable to the modulation, but the control of repeat distance was not sufficient to be useful for calibration. An important observation was the decay of the reflection due to interdiffusion in the multilayer at room temperature. At present, metallic multilayers are quite widely used (and commercially available) as optical elements for X-rays (in particular soft X-rays with wavelength 2 = 5 nm). They extend the capabilities of conventional long-period crystals for use in spectrometers. Fig. 8-3 shows a reflectivity typ-
0
2 4 6 Grazing angle (degrees)
8
Figure 8-3. The measured reflectivity for X-rays (1 = 0.154 nm) at grazing incidence on a W/Si multilayer (200 layers, repeat distance 2.44 nm), available commercially as an X-ray mirror. [Adapted from Spiller (1985).]
ical for such multilayers. In addition, the principles of multilayer design used for visible light can now be extended to lower wavelengths. In this way, elements with controlled reflectivity (for a range of wavelengths) at normal incidence can be produced. The field has been reviewed by Spiller et al. (1980). For good reflectivity, the layers must have a large difference in scattering potential; for X-rays, materials of low artd high electron densities must be combined. In the high-electron-density layers there is strong absorption of the Xrays. (Absorption, by comparison, is hardly significant in multilayers for the reflection of visible light.) In the presence of absorption, "quasiperiodic" multilayers, in which the repeat distance is constant but the fraction of high electron density material increases with distance from the top surface, give higher reflectivities than normal periodic designs. Still better performance, both in range of reflected wavelengths and in integrated intensity, are obtained with aperiodic designs in which the repeat distance also is altered slightly with depth. Low-electron-density materials which have been used include carbon, boron, B4C and silicon, in each case amorphous. A widely used high-electron-density material is tungsten (which is polycrystalline or, with a sufficient impurity level, amorphous). The smoothness of the interfaces is very important and seems easiest to achieve when the structures are amorphous (Spiller et al., 1980). Tungsten, however, is very strongly absorbing, and better results may be obtained with a lighter element such as nickel which permits deeper penetration of the multilayer and consequently better spectral resolution. For applications it is important to know the stability of the X-ray mirror structures not only to thermal annealing but also to in-
8.3 Properties
tense irradiation (as experienced, for example in a monochromator on a synchroton source) (Kortright et al., 1991). As interdiffusion occurs it is expected that the reflectivity will decrease, with the intensities of higher order Bragg reflections particularly affected as the interfaces become more diffuse. These effects are found in W/Si multilayers, where a decrease in repeat distance is also noted. In W/C multilayers, however, all these effects are reversed, a surprising result which may be attributed to compound formation at the interfaces. Stability provides another reason for designing X-ray multilayers to be amorphous; the fast interdiffusion possible in fine polycrystals would be a problem. Kortright et al. (1991) have shown that prior thermal annealing can stabilize a multilayer structure against further changes under irradiation. Recently there has been work on using periodic multilayers as elements in larger structures to achieve higher efficiency in wavelength dispersion. With a spacer deposited between two multilayers, a FabryPerot etalon is formed (Barbee, 1985), and lateral patterning by lithography has also been used (Barbee, 1988). In a different application, Bionta et al. (1988) have used a cross-sectional slice of an Al/Ta multilayer as a linear zone plate for focussing an X-ray point source to a line. Multilayers are also of interest for the reflection of neutrons (k = 0.05 to 5 nm). There are uses as monochromators and as polarizing "supermirrors" with reflectivity over a range of angles. In the latter case aperiodic designs are used, combining a ferromagnetic and a non-magnetic material (e.g., Fe/Ge and Fe/W) (Majkrzak, 1989).
339
8,3.3 Normal-State Electron Transport
The theory of non-superconducting electron transport in multilayers is well advanced, and many novel effects have been predicted. While for semiconductor 'superlattices' there is experimental verification of the effects, some of which are now exploited in commercial devices, for metallic multilayers it is fair to say that theory has outstripped experiment. The requirements for some of the novel transport effects to be observed in metallic multilayers are particularly stringent, and producing materials of sufficient quality is one of the major challenges in multilayer deposition. There are two reasons why the effects are more difficult to observe in metallic than in semiconductor multilayers. The interface quality in semiconductor multilayers (deposited typically by MBE, see Section 8.4.3.2) seems to be higher than for metals and often approximates well to the ideal expected in a superlattice. Metallic multilayers have been made with slightly less sharp interfaces (two mixed monolayers at the interface, rather than one) as might be expected from the easier diffusion in metals. More importantly, however, for metals some of the characteristic lengths are substantially less than for semiconductors. An example of this is the de Broglie wavelength. When the thickness of a thin film approaches the de Broglie wavelength, the quantum size effect alters the transport properties. In semiconductors the de Broglie wavelength can be 5 to 10 nm, and the quantum size effect can be readily observed in multilayers of comparable repeat distance. In metals, on the other hand, the wavelength is such that the effect could be observed only for atomic monolayers. The field of semiconductor superlattices has been reviewed by Esaki (1985). The possible origins of the electron
340
8 Metallic Multilayers
transport properties of metallic multilayers have been analysed by Jin and Ketterson (1989) who also provide a review of experimental results. The main observation for metallic multilayers is the resistivity increase due to electron scattering at the interfaces. The resistivity has a component which is inversely proportional to repeat distance, and rises to the usual limit, for strongly scattering materials, of ~150|iQcm. As the resistivity rises, its temperature coefficient becomes less positive, and near the upper limit of resistivity it becomes negative (Fig. 8-4). The multilayers with a vanishing temperature coefficient of resistance can be exploited as temperature-independent thin film resistors. The anisotropy in resistivity which would be expected in a multilayer has not been observed because of the difficulty of making through-thickness measurements in a low resistivity material. The predicted negative differential conductivity in the through-thickness direction has been verified for semiconductor superlattices (Esaki, 1985), but not for metals. Other effects which should occur in metallic multilayers even when the repeat
515 0.12
distance is quite large (i.e., when it exceeds the electron mean free path) are the integral quantum Hall effect and hopping conduction; these also have not been observed, presumably because the required atomically smooth interfaces were not achieved. When a thin film thickness is comparable with the mean free path, the resistivity rises according to the classical size effect. This has proved useful in interpreting some results on metallic multilayers. On the other hand, as discussed above, the quantum size effect which would set in at lower values of layer thickness is probably not observable in metals. When the mean free path becomes very short, comparable with the atomic diameter, localization and interaction effects can arise. Metallic multilayers provide a useful probe of these effects for two-dimensional conduction (as reviewed by Jin and Ketterson (1989)) (see also Vol. 3, Chap. 6). Many of the novel transport properties of metallic multilayers remain to be demonstrated. The challenge is to produce multilayers which, in particular, have higher quality interfaces, showing better coherency, smoothness and sharpness. Even with the better materials, however, it must be doubted whether the technological potential can be as great as for semiconductor superlattices.
^0.08 o
8.3.4 Superconductivity
^0.04 ^
0 0.04 20
n /c
100 200 Temperature (K)
300
Figure 8-4. Normalized electrical resistivity as a function of temperature for a series of Nb/Cu multilayers. The individual layer thickness (in nm) is marked in each case on the curve. [Adapted from Werner et al. (1982).]
In marked contrast, the novel superconducting behavior of metallic multilayers can be readily observed without stringent requirements for multilayer quality. The characteristic length is the superconducting (Ginzburg-Landau) coherence length in the through-thickness direction (normal to the layers), and this is often in the range 5 to 100 nm. Multilayers with repeat dis-
341
8.3 Properties
tances on this scale are relatively easy to fabricate, and furthermore the superconducting properties will often be quite insensitive to interfacial quality. Multilayers of two superconductors, a superconductor and a normal metal, and a superconductor and an insulator have been studied. There are naturally occurring layered superconductors: dichalcogenides, intercalated graphite and high transition temperature ceramics based on copper oxide. The artificial superconductor/non-superconductor multilayers are of interest for comparison with these. Thin superconducting layers in such a material exhibit the proximity effect in which the critical temperature is reduced as the layer thickness is reduced. This effect is observed in the superconducting/non-superconducting multilayers, and the critical temperature in such multilayers is reduced also if the thickness of the non-superconducting spacer material is increased. Such multilayers can show three-dimensional or two-dimensional conduction behavior, depending on whether or not there is coupling between the layers of superconductor. Coupling becomes insignificant as the thickness of the non-superconducting layers exceeds the coherence length (Volume 3, Chapter 4), and the "crossover" from three-dimensional to two-dimensional behavior can be observed by increasing the thickness of the non-superconducting layer (Fig. 8-5) or, in a single sample, by lowering the temperature to decrease the coherence length. The upper critical field in layered superconductors is anisotropic and also shows clear dimensional effects. The dependence of the parallel critical field on the repeat distance arises from the interaction of the vortex lattice and the multilayer; this is an example of an effect attributable to the periodicity (Section 8.3.1). Superconducting properties of metallic
N
bs£] NV
75 -
Nv
2-D Nb: 4.5 nm Ge: 5.0 nm
\ 2-D/3-D Crossover Nb: 6.5 nm Ge: 3.5 nm
50 -
25 -
\V \\ 3-D \\ Nb: 4.5 nm V V
^v
0.7 nm
\ 0
0.8
\
0.9 T/Tc
\ \
\ 1.0
Figure 8-5. Upper critical fields HC2 (measured parallel to the layers) of Nb/Ge multilayers as a function of reduced temperature (7^ is the superconducting transition temperature). The niobium and germanium layer thicknesses are shown in each case. As the germanium thickness is increased, there is a crossover from anisotropic three-dimensional behavior to twodimensional behavior. The solid lines are calculated from Josephson coupling theory. [Adapted from Ruggiero etal. (1980).]
multilayers have been comprehensively reviewed by Ruggiero and Beasley (1985) and by Jin and Ketterson (1989) (see also Vol. 3, Chap. 6). 8.3.5 Magnetic Properties Of all multilayer properties, the magnetic behavior is the most widely studied. Band structure calculations for quantitative ab initio prediction of magnetic properties are now becoming possible, and the control of multilayer deposition is sufficient to permit the theory to be tested. Of great interest is the exchange interaction between magnetic spins. The decay length of the interaction is of the order of 1 nm and therefore accessible, albeit with some difficulty, to experimental test. In addition to this fundamental scientific interest,
342
8 Metallic Multilayers
there is great technological potential in magnetic multilayers whose properties can be tailored in a remarkably complete way. In addition to layer thickness and repeat distance, the magnetic properties may be affected by the structure of the individual layer materials (for example, iron could be b.c.c. and ferromagnetic, or pseudomorphically c.c.p. and antiferromagnetic), by the stresses in the layers (magnetostrictive effects) and by the nature of the interfaces. Important properties susceptible to control in this way are the coercivity and the anisotropy. Technologically, the greatest current interest is in polycrystalline multilayers of cobalt and platinum or palladium. These are alternatives to rare earth-transition metal thin films for magneto-optic recording, and have the advantages of greater oxidation resistance and a larger Kerr effect at the shorter incident wavelengths with which higher density recording is possible. Careful selection of deposition conditions is necessary for these multilayers as a typical Co-layer thickness is only ~0.4 nm, with Pd or Pt layers of — 1 nm. The film in total has <10 bilayers. For high density recording, perpendicular anisotropy (preferred magnetization parallel to the film normal) is required; a high coercivity and a square magnetic hysteresis loop are also desired. Carcia etal. (1990) have analysed the effects of deposition methods on coercivity. They point out that during deposition, a thin film is bombarded by energetic species which may promote interfacial mixing. This bombardment is least for deposition from the vapor (evaporation or molecular beam epitaxy (MBE)), intermediate for magnetron sputtering and greatest for ion-beam sputtering (for discussion for these techniques, see Section 8.4.3). They show that the largest coercivity correlates with the least bombardment,
the degree of which can be controlled in magnetron sputtering by varying the sputtering gas. Carcia et al. (1990) concluded that low bombardment is beneficial because less interfacial mixing is induced. Later work by Highmore et al. (1991) has confirmed that low bombardment does give high coercivity (and also squareness of hysteresis loop) (Fig. 8-6), but their direct structural studies show that the interfaces under low bombardment are less (not more) smooth and sharp than under high bombardment (Section 8.5.4, Fig. 8-19). When there is low bombardment, the film will be more likely to be in tension, will grow with less texture and with a less dense, more columnar microstructure, and will have a less smooth surface with lower optical reflectivity. Highmore et al. suggest that the stress plays a role in giving the desired coercivity and anisotropy, that the columnar microstructure promotes anisotropy, and that interfacial roughness raises the coercivity.
1.0 2.0 Sputtering pressure (Pa)
3.0
Figure 8-6. Coercivity and magnetization ratio MJ Ms as a function of sputtering pressure for Co/Pd multilayers (0.3 nm Co/1.2 nm Pd). The ratio of remanent magnetization Mr to saturation magnetization' Ms indicates the squareness of the hysteresis loop. The sputtering pressure controls the degree of bombardment of the growing film; both coercivity and squareness are greater for lower bombardment. [Adapted from Highmore et al. (1991).]
8.3 Properties
The Co/Pt and Co/Pd multilayers are examples of ferromagnetic/non-ferromagnetic combinations. Such combinations are also of interest because the ferromagnetic layers, if far enough from each other not to show coupling, can show two-dimensional ferromagnetism, with its characteristic temperature-dependence of the magnetization. The layers can also show a reduced magnetic moment due to magnetic "dead layers" at the interfaces; for Ni in Mo, for example, these can be as much as four atomic planes (Khan et al., 1984). As the non-magnetic spacers get thinner there is coupling between the layers, with the Curie temperature being dependent on the spacer thickness. The oscillatory nature of the RKKY (Ruderman-Kittel-Kasuya-Yosida) exchange interaction between spins permits dramatic manipulations of properties. Chromium is antiferromagnetic, but it appears that in a Cr/Au multilayer the interaction distance can be increased so that the spins in the chromium order ferromagnetically (Fu et al., 1985). Ferromagnetic/ non-ferromagnetic multilayers can show rather low coercivities in comparison with uniform thin films; in contrast ferromagnetic/anti-ferromagnetic combinations can have high coercivities. Multilayers with rare-earth ferromagnets as one material have been used widely for studies of magnetic interactions. In Gd/Y multilayers with constant Gd-layer thickness, variation of the Y-layer thickness can give parallel or antiparallel coupling between the layers (Majkrzak et al., 1987). In Dy/Y multilayers there is coherent helical coupling of the magnetic layers (Hong et al., 1987). There has also been scientific interest in multilayers, for example Fe/V, which permit the co-existence of ferromagnetism and superconductivity. There are examples of magnetic behavior which can be attributed to the periodicity of multilayers. In
343
Ni/Mo multilayers, for example, the dependence of the magnon (quantized spin wave) frequency on the magnetic field can be altered by altering the repeat distance. The results are in agreement with the predicted development of bands for magnons when there is coupling between the magnetic layers. Comprehensive reviews of both theoretical and experimental aspects of the magnetic properties of multilayers have been published by Schuller (1988) and by Jin and Ketterson (1989). (The topic is also treated rather fully in Volume 3, Chapter 6.) 8.3.6 Mechanical Properties As analysed by Koehler (1970), any multilayer of materials of different stiffness will impede dislocation motion. Enhanced yield stress has now been found for a number of metallic multilayers (Cammarata etal., 1990). Typically the yield stress increases as the repeat distance is reduced according to a Hall-Petch relation, but saturates when the layer thickness for each metal becomes less than the critical thickness for dislocation generation (of order 50 nm) (Tsakalakos and Jankowski, 1986). Enhanced yield stress may be technologically significant. Bickerdike etal. (1985) have shown that a 1 mm thick multilayer of aluminum (20 to 1600 nm thick) and transition metal (0.1 to 21 nm thick) can be produced by high-rate evaporation and rolling. This sheet material shows high strength and good high temperature stability. In contrast to plastic behavior, elastic properties cannot usually be altered much by manipulation of microstructure. Against this background it is remarkable that the elastic modulus for biaxial stretching of some multilayered thin films can
344
8 Metallic Multilayers
show a large enhancement, of 100% or more, at small repeat distances. Most of the studies have been of mosaic films of epitaxial c.c.p. metals in (111) orientation. The modulus enhancement is proportional to the square of the amplitude of the composition modulation, and when normalized to constant amplitude, the supermodulus effect is found to peak at ~ 2 nm repeat distance. Figure 8-7, from the work of Tsakalakos and Hilliard (1983) shows the effect in the most studied system, Cu/Ni. The enhanced in-plane modulus is much greater than would be expected for the composite material or for the homogeneous alloy of the same overall composition. Clemens and Eesley (1988) have shown that even for epitaxial c.c.p./c.c.p. metal combinations (e.g., Pt/Ni) there can be an expansion in the direction normal to the thin film, localized at the interfaces. This interfacial expansion becomes evident as an average expansion, and correlates with a reduction in elastic modulus perpendicular to the layers, as the repeat distance is reduced. Other modulus reduc-
1
2
3 A (nm)
U
5
6
Figure 8-7. Biaxial in-plane elastic modulus Y as a function of repeat distance X for (111) orientation epitaxial mosaic multilayers of Cu/Ni. The average multilayer composition is equiatomic and the values of Y have been extrapolated (assuming modulus enhancement proportional to the square of the amplitude) to a composition amplitude of +50at.%. [Adapted from Tsakalakos and Hilliard (1983).]
tions, parallel to the layers, have been found in epitaxial multilayers of c.c.p. and b.c.c. metals (e.g., Ni/Mo) and also correlate with interfacial expansion. The elastic properties of metallic multilayers have been reviewed by Tsakalakos and Jankowski (1986) and by Cammarata (1988). Understandably, most interest has focused on modulus enhancement, but the supermodulus effect remains controversial as the evidence is disputed (Moreau et al., 1990), and the origins remain to be established. Suggested origins based on electronic effects or non-linear elastic effects have difficulties (Cammarata, 1988). Indeed, recent calculations predict that in coherent multilayers the in-plane modulus should be reduced, not enhanced (Mei and Fernando, 1991). A model which can account for both enhanced in-plane modulus and reduced perpendicular modulus is that of Cammarata and Sieradzki (1989). They point out that the interfaces in a multilayer will exert an interfacial tensile stress causing a biaxial compression. If, as seems reasonable, the interfacial stress (variation of interfacial free energy per unit area with strain) is of the same order as the interfacial free energy per unit area, the modulus changes can be quantitatively explained. An important point about this model is that interfacial free energies (and presumably stresses) are insignificant for coherent interfaces. Thus it is predicted (in agreement with experiment) that for repeat distances below a critical value (when coherency is preserved) the supermodulus effect will not be found. For partially coherent or incoherent interfaces, at repeat distances greater than the critical value, there is a supermodulus effect which decays as the repeat distance is increased (decreasing the density of interfaces). While this model remains to be fully tested, it does suggest the importance of coherency.
8.3 Properties
8.3.7 Other Properties As stated in Section 8.2.5, reactions at the interfaces in metallic multilayers can yield metastable phases. Of particular interest in recent years has been the production of metastable amorphous alloys by the reaction of two polycrystalline metals (Johnson, 1986, and Volume 9, Chapter 9, Section 9.2.6). The metals are typically an early transition metal (e.g., Zr) and a late transition metal (e.g., Ni); combinations of this type have a strongly negative heat of mixing so that there is a large driving force for reaction. Amorphization appears to result because fast diffusion of the small atom (e.g., Ni) through the growing amorphous phase permits mixing, while the slow diffusion of the large atom (e.g., Zr) inhibits nucleation and growth of the stable intermetallic phases. While solid state amorphization of this type is of scientific interest in deposited multilayers (Greer, 1991), it is of technological interest for the prospect of producing amorphous phases in bulk form from mechanically reduced multilayers (Atzmon etal., 1985). In a multilayer system such as Ni/Ti, the competition between amorphization and formation of crystalline intermetallics is finely balanced. Clemens (1987) has shown for multilayers deposited at high sputtering pressure, with low bombardment of the growing film, that: the growth is by the island mechanism (Section 8.4.2.3); there is no preferred orientation; the interfaces are disordered; and that annealing results in amorphization. In multilayers deposited under greater bombardment, however, the crystalline layers are of higher quality with preferred orientation, and there is no amorphization on annealing. This demonstrates that deposition conditions, by altering the microstructure of the multilayer (specifically the degree of disorder at the
345
interfaces and at high-angle grain boundaries) can change the transformation behavior. While almost all of the multilayers considered in the above discussion of properties have consisted only of metals, there are clearly possibilities for more diverse combinations of materials. For example, techniques are being developed for the deposition of metal/ceramic multilayers which are of interest for their mechanical properties (Moustakas et al., 1988). Fully ceramic multilayers have also been made (McKee et al., 1988) and may yet offer a range of new properties as wide as those found for semiconductor and metallic multilayers. An example of a new property is found in CuCl/CuBr multilayers. The distortions in each type of layer change the symmetry, making the materials artificially ferroelectric; the resulting multilayer is usefully piezoelectric (Wong et al., 1982). 8.3.8 Summary From the overview given in the preceding sections, it can be seen that there is considerable scope for tailoring of multilayer properties, but that, particularly in some cases, the multilayer structure must be very tightly controlled. The characteristics of the multilayer which may affect the properties are: layer thickness and its spread (either periodic or non-periodic designs may be desirable); interface smoothness and sharpness; interfacial structure, including coherency; the crystal structure and crystallographic orientation (or amorphicity) of the constituent materials; the grain size in crystalline layers; and the stresses in the layers. Although the control of deposition has advanced to permit many predicted multilayer properties to be realised, there is yet room for improvement; for example, additional electron
346
8 Metallic Multilayers
transport phenomena should be observable if metallic multilayers could be refined to the stage of atomically sharp alternating monolayers. The present state of the art of multilayer processing is described below, together with an assessment of the possibilities for improvement.
8.4 Preparation 8.4.1 Introduction
The deposition of metallic multilayers is an immature field in comparison with the deposition of semiconductor multilayers. Overall, about ten times as many papers have been published on semiconductor multilayers, and it is natural that this work has been the basis from which much of the understanding of metallic multilayers has evolved. However, in many ways metallic multilayers in general differ from their semiconductor cousins. The main difference is that nearly all work on semiconductor multilayers has been performed on single crystal material, whereas metallic multilayers of all degrees of crystaUinity are of importance, and all have been studied (see Section 8.2.1). In considering the modes of growth during multilayer deposition, different mobilities and kinetics may be imagined, depending on the material being deposited and the base on which it is growing. It is of interest to note that the most ideal multilayers produced to-date are either lattice-matched single crystalline or totally amorphous. This is because the surface on which deposition is occurring is more homogeneous for these structures than for a polycrystalline material on which, inevitably, the surface mobility is not uniform. In view of the variety of structures of interest for metallic multilayers the discussion of multilayer preparation begins with a consideration of general
thin film growth processes (Section 8.4.2). This is followed by a description of the variety of preparation methods (Section 8.4.3) and of the methods of process control which are particularly important for multilayer deposition (Section 8.4.4). Finally, in Section 8.4.5, the choice of deposition method is discussed, together with some current issues in multilayer growth and the outlook for more ideal multilayers. 8.4.2 Thin Film Growth Processes 8.4.2.1 General Considerations
The most important multilayer preparation methods involve growth from a supersaturated vapor phase; the atoms arrive from the vapor, condense on a surface (either the substrate or the growing film), and aggregate to form a deposited layer. It is important to stress that such growth occurs far from equilibrium and is governed mainly by kinetics. The phases and structures which result are often themselves far from equilibrium. While all materials processing exhibits deviations from equilibrium, for the growth of solid structures from the vapor the deviations are the most extreme. The non-equilibrium nature of physical vapor deposition causes the structure of thin films to be highly dependent on the deposition conditions. There are several fundamental processes which occur during the growth process (Venables et al., 1984): formation of nuclei from adsorbed atoms; coalescence of nuclei; and, arising from the coalescence, continuous growth during which a threedimensional microstructure develops. These three processes are separately affected by the deposition temperature (usually written as the homologous temperature T/Tm9 where Tm is the melting temperature of the material), by the purity and cleanliness of the deposition environment, and
8.4 Preparation
finally by the energy flux impinging on the growing surface. A complete description of the development of microstructure in thin films is still being developed, and much work in this field lacks reproducibility because even minute levels of impurity can affect the structures produced. The arrival of atoms during depositions is statistical, often with a range of angles, and the roughness of the films produced is on a scale which depends on the degree of surface mobility. Surface morphologies of thin films have many of the attributes of fractal structures generated by accumulation processes. Messier and Yehoda (1985) have used the idea of a fractal of slowly varying dimension to describe the development of roughness, though a complete description by fractal geometry has not been demonstrated. Smooth flat films (and multilayers) occur only under relatively special circumstances such as epitaxial layer-bylayer growth, or growth under bombardment. Central to any discussion of growth processes is surface mobility, which is considered next. 8.4.2.2 Surface Mobility When an atom arrives at a growing surface its mobility will be determined mainly by the activation energies for the various movements it might make. The literature often gives a range of values for surface diffusion (Rhead, 1989); this is partly a reflection of the difficulty of preparing very clean surfaces, but is also partly because there are many different types of surface mobility that can be measured. Some of the most clearly specified surface mobilities refer to atoms moving either along or perpendicular to the channels of a (211) surface plane of a b.c.c. structure (Wang and Ehrlich, 1988). On the other hand, a considerable amount of surface diffusion data has been assembled from traditional
347
grooving experiments in which atoms not only travel across terraces but then have to jump across ledges or grain boundaries. Whatever the mechanism of the surface migration, it is often characterized by a diffusion coefficient D = Do exp (— Q/kB T), where the activation energy Q = nkBTm, n is of order 3 to 6, kB is Boltzmann's constant and Tm is the melting temperature of the solid. The pre-exponential term Do is often taken to be a2v/4 where a is the lattice parameter and v is the lattice vibration frequency. A typical v of 10 13 to 10 14 Hz gives D0 = 10-1m2s-1 (Wang and Ehrlich, 1988). Surface self-diffusion (A atoms on A atoms) and the surface diffusion when a new layer is being started (B atoms on A atoms) can be distinguished. The effective diffusion distances are determined by the structure of the deposit, the temperature, the deposition rate, and the level of impurities. The role of temperature is particularly important because of the exponential temperature-dependence of D. For non-ideal surfaces, the roles of different contributions to the surface diffusion process (including surface vacancy creation and kink sites) are poorly understood. Besides the thermal element of the surface mobility there is also a further element associated with the arrival of energetic particles at the surface of a growing film. In general, the kinetic energy of an atom approaching the surface will be increased by the binding energy as it finally adheres to the growing surface. Simple sputtering and ion beam sputtering (Section 8.4.3.3) are both methods in which the energy of the arriving species, which plays an important role in producing extra mobility, may be modified in several ways. In evaporation processes there are many examples in which ion beams are used to enhance the bombardment of the growing film and thus promote the surface mobilities.
348
8 Metallic Multilayers
A simple example can be given of the importance of surface diffusion length in relation to ledgewise growth. Taking Q^3kBTm and Do « 10~7 m 2 s" 1 gives a surface diffusion coefficient D of - 3 x l O " 1 4 m 2 s ~ 1 at 0.2 Tm. The diffusion length is given by {AD t)i/2, where t is the time for the growth of one monolayer and thus also the average lifetime of the free surface. Taking a typical value of / = 1 s, the diffusion length is ~350nm. We can conclude that ledgewise growth at 0.2 Tm would be possible at acceptable rates only if the ledges were separated by distances less than this length. Of course for polycrystalline material, or material grown in poor vacua, the length reached would be much reduced, as is discussed in Section 8.4.4.5.
8.4.2.3 Development of Microstructure and Morphology
The rudiments of an understanding of the development of the microstructure and morphology of thin films are to be found in the work of Movchan and Demchishin (1969); their general morphological description for evaporated films was extended to sputtered films by Thornton (1977)
Densely packed fibrous grains Tapered Crystallites separated by voids
(Fig. 8-8). In these early works, three zones, I, II, III, were defined in temperature ranges T/Tm of 0-0.3, 0.3-0.5 and >0.5, respectively; Thornton introduced a transitional zone between I and II, found in his sputtering experiments. These models reflect primarily the role of atomic mobility: as the mobility increases, there is a change from very porous low-density columns in zone I, to tightly packed columns in zone II, and finally to equiaxed structures in zone III. While the concept of different zones is correct, the exact temperatures of the transitions from one morphology to another may be affected by the higher surface mobilities found nowadays in cleaner deposition systems. Recently the models for thin film morphology have been refined by Messier and Yehoda (1985). In cleaner systems the films may often be rougher. Indeed, surface smoothness is an exception rather than the rule. Special measures must be taken to produce smooth multilayers over large lateral distances. The films must be grown single crystal, or at least highly textured (mosaic structure), or if they are amorphous or polycrystalline then best results are obtained if the surfaces are bombarded in some way during deposition.
Columnar grains Recrystallized grain structure
Figure 8-8. Schematic representation of the microstructure of metallic thin films as a function of substrate temperature and working gas pressure in magnetron sputtering. [From Thornton (1977).]
349
8.4 Preparation
There appears to be a difference between growing single crystal multilayers and other types. On a single crystal surface the mobility is a strict function of temperature. In all other systems, however, the density of surface defect states is more likely to control the mobility. The effective diffusion length is then related to the original distribution of nucleation sites, and the structure that has grown as the film has thickened, rather than to a limiting ideal surface diffusional process. The effective diffusion length can usually be enhanced by bombardment. Diffusion on amorphous surfaces has still not been studied, but the similarity of observed growth morphologies suggests that the surface mobilities are not greatly different from those on microcrystalline surfaces. It is clear that there are different problems in optimizing deposition conditions for single crystal films and for other types. Single crystal films grow by a range of processes on a given substrate. The processes fall into three categories (illustrated in Fig. 8-9): Volmer-Weber (island growth, as with droplets on a non-wetting surface), Frank-van der Merwe (layer-by-layer growth), and Stranski-Krastanov (growth of islands on a thin layer). These different processes are all described in detail in Chapter 7, to which the reader is referred for a complete description. 8.4.2.4 Internal Stress A general feature of thin films is that they have an internal stress. The stress in a thin film deposited at room temperature is due to the non-equilibrium nature of the growth process. For films grown on hot substrates there may be added the stress arising from the differential thermal contraction of the substrate and the film. Two factors contribute to the deposition stress:
1<0<2 i
(c)
Figure 8-9. Three modes of crystal growth from the vapor, illustrated schematically at increasing coverage 6 in monolayers: (a) layer-by-layer, or Frankvan der Merwe; (b) layer plus island, or StranskiKrastanov; (c) island, or Volmer-Weber. [Adapted from Venables et al. (1984).]
annealing out the defects associated with limited atomic mobility during growth leads to a tensile stress; bombardment by energetic species gives an "atomic peening" effect and compressive stress. In sputtering, the amount of bombardment can be controlled by altering the gas pressure, and it is possible to achieve a degree of compensation in which the residual stress level is low. In evaporation, however, films will normally be in tension unless extra bombardment (e.g., by an ion beam) is introduced. It is important to have a low stress to prevent the film from peeling from its substrate. This will occur when the stored energy per unit area in the film equals the interfacial bonding energy per unit area at the interface of film and substrate. The probability of loss of adhesion increases with increasing total film thickness and also with the level of dirt at the interface. For films < 0.3 jim thick there are few problems with stress, but for thicknesses > 1 jum there can be serious problems unless care is taken to reduce the stress or to improve the interfacial bonding (e.g., by prior surface cleaning, usually by ions). Reduction of stress is also important if the film is to be removed from the substrate to be used unsupported in an experiment (e.g., tensile
350
8 Metallic Multilayers
testing, or transmission electron microscopy (TEM) examination); because there is a stress gradient, the free-standing film will curl if no care has been taken to reduce the stress. Unfortunately, the deposition conditions which give minimal overall stress may not yield the best flatness and interface sharpness.
tures. Techniques such as electrodeposition and mechanical reduction are included in the review of techniques; while these may not be capable of the precision of vapor deposition, they may offer the prospect of larger production rate at lower cost.
8.4.3 Deposition Technologies
Conventionally, evaporation is from an open source, in the liquid or solid state, heated in vacuo resistively or by an electron beam. In MBE the sources are Knudsen cells in which the source is resistively heated in a small enclosure, yielding a beam of atoms from a small orifice. For any form of evaporation, rate monitoring is essential as the evaporation rate is sensitive to small changes in source temperature. The vapor pressure above the source varies as exp (Q/kB T) in which the energy of evaporation Q is typically 25 kBTm (Tm9 melting temperature); it follows that a temperature change of 0.1 % at 1.1 Tm gives a rate change of 2.2%. To keep the rate constant to better than 0.3% (necessary for some applications) the temperature needs to be controlled to within 0.2 K. The more refractory the source element, the more difficult it is to maintain a very low base pressure during evaporation, because the power dissipated to the vacuum chamber warms it up so much that severe outgassing occurs even if a previous bake-out has been performed. A load-locked system, in which sources can be introduced and deposits removed without exposing the main vacuum chamber to air, can be very beneficial in improving the overall quality of the deposition. High temperature Knudsen cells, capable of reaching 1600°C to 1900°C are becoming available.
8.4.3.1 General Considerations In the last thirty years many techniques have evolved which are capable of precision multilayer fabrication, with considerable technological development even in the last few years. The techniques are outlined in the following sections, in which relative advantages and disadvantages will be highlighted. All the important techniques for metallic multilayers are based on physical vapor deposition; these are: evaporation, molecular beam epitaxy (MBE - evaporation in a ultra-high vacuum (UHV) environment using Knudsen cells) and sputtering. Sputtering can be from many different types of sources (r.f. or d.c, planar or cylindrical magnetron, triode or diode) operating at modestly high pressures 0.1 to 1 Pa, or can be ion-beam sputtering in which the target is bombarded by an ion beam generated independently and which must operate at a pressure much lower than in ordinary sputtering, typically only 1 0 ~ 2 P a t o l 0 ~ 3 Pa. Chemical vapor deposition (CVD) techniques, including photon-assisted CVD and metal-organic CVD (MOCVD), are widely used for deposition of semiconductor multilayers, but their use is not widespread for metals, partly because the chemical precursors are not yet so readily available. In principle, they can be equally used for metallic multilayers so long as precursors can be prepared and can decompose at sufficiently low tempera-
8.4.3.2 Evaporation
8.4 Preparation
351
8.4.3.3 Sputtering Sputtering is the removal of an atom from a surface resulting from the impact of an energetic ion. This process can be applied in deposition in two distinct ways. Most commonly, an inert gas pressure of 0.1-10 Pa is maintained in the deposition chamber and the target is the cathode of a gas diode. A typical system is illustrated in Fig. 8-10. A plasma is generated around the target, and ions (generated in the plasma by accelerating electrons) bombard the target and sputter the target atoms away while releasing sufficient secondary electrons to maintain the discharge. There are tens of different configurations for such sources. Because of its relative reliability, efficiency and reproducibility, the most widely used is the planar magnetron source, which uses a suitably arranged magnetic field greatly to enhance the plasma close to the target. Quite high deposition rates may be achieved with such a system - from O.lnms" 1 up to lOnms" 1 , depending on the target material and the capacity for target cooling. For precision deposition of multilayers, however, slower rates may often be preferred. In the second type of sputter deposition process, an ion beam (usually of argon) with an accelerating voltage of 0.5 to 5.0 keV is generated in a separate gun and is directed towards the target in a chamber pressure of 10 ~2 to 10~ 3 Pa. To achieve a uniform deposition rate, accurate power and pressure control are required. For conventional sputtering, power to the target can be controlled to 0.1% or even 0.01%, and the pressure control can easily be to 0.1 %, which lead to a 0.3% to 0.15% variation in deposition rate depending on the material being sputtered. An additional complication is that the deposition rate changes slowly
Substrate table indirectly heated ; j — - i — * Residual gas '*---)—•* analyzer
1
Power supplies Mass flow controllers Shutters
Ar-gas
Figure 8-10. Schematic illustration of a sputter deposition system for fabricating metallic multilayers. [Adapted from Fernandez and Falco (1985).]
with the erosion and faceting of the target, but changes in power can be programmed to compensate for this. The erosion rate can vary by a few tens of percent over the lifetime of a target; the shorter a deposition run in relation to the target life, the less the effective drift in deposition rate. 8.4.3.4 Chemical Vapor Deposition In chemical vapor deposition (CVD) a stream of reactive gas molecules is passed over a heated substrate on which the molecules decompose or react. Applied mainly to the deposition of semiconductors or of refractory carbides and nitrides, the process is attractive for these applications in which the relatively high deposition temperature (800 °C to 900 °C) can be tolerated. Recently, efforts have been made to reduce the working temperature of the process, either by using metallo-organic precursors (MOCVD) which decompose at much lower temperatures, or by using some form of activation of the de-
352
8 Metallic Multilayers
composition process so that it can occur at much lower temperature (for example, plasma-assisted (PACVD) or photon-assisted CVD). These techniques are still under development for commercial application, but show great promise. While they have been widely used in research and development on the simpler semiconductor multilayers, they have not, to-date, been much used for metallic multilayer deposition. This is because of the deposition temperatures, which are still relatively high, the consequent interdiffusion at the interfaces being a particular problem for metal multilayers. For precision deposition, accurate control is required of pressure, gas composition and temperature. These parameters affect the surface chemical reactions. Rapidly acting valves and purge gases to effect rapid changes in composition can yield very sharp interfaces. 8.4.3.5 Electrolytic Deposition Electrolytic deposition is of interest as a low-cost fabrication method. It has been demonstrated that it can produce fine scale multilayers with a limited range of metals. The process and its results are reviewed in Chapter 11, Section 11.4.3. 8.4.3.6 Atomic Layer Epitaxy A variant of MOCVD, atomic layer epitaxy (ALE) is a technique developed for the semiconductor industry which achieves monolayer deposition (Suntola, 1990). In ALE, the adsorption of the precursor molecules on the substrate surface is a separate process from their decomposition. In the adsorption stage exactly one monolayer of atoms is deposited, each with an attached organic ligand. Subsequently another gas is introduced to sweep away the ligand, leaving behind the monolayer of
the desired species on the surface. The process has not yet been demonstrated for metals, but in principle it holds great promise if suitable precursor molecules can be developed. 8.4.3.7 Mechanical Reduction When two metals have compatible elastic and plastic deformations, it is possible to preserve an initial macroscopic layering (obtained by stacking metallic sheets or foils) even after substantial mechanical reduction. For example, by a combination of folding and rolling it is possible to reduce a composite of 25 |im thick nickel (or copper) foils and 20 \mv thick erbium foils to a multilayer with a total thickness of up to 200 |im in which the layers are 5 nm thick (Atzmon et al., 1985). The layer thickness in multilayers produced this way is rather uneven, yet the average spacing is small enough for some useful effects to be achieved (solid-state amorphization in the case of Ni/Er and Cu/Er). Mechanical reduction offers the possibility, in some special cases, of a low cost fabrication route. 8.4.4 Process Control 8.4.4.1 Layer Thickness The great majority of metal multilayers have been made by evaporation or sputtering, and the discussions of process control given here apply mainly to these techniques. Because in these techniques the generation of the atomic species and their growth on a substrate are totally separated, there are many possibilities for independently changing the deposition conditions. Despite the significant differences in deposition rates between the various techniques, there are some general considerations for thin film growth and layer thickness control.
8.4 Preparation
Layer thickness control has two aims: achievement of the desired average layer thickness, and constancy of layer thickness. Achievement of these aims depends on long- and short-term stability, respectively, of deposition rate. The rate can be controlled in two general ways: "open loop" control in which the deposition conditions are maintained as stable as possible, without monitoring of deposition rate; and "closed loop" control in which feedback is established, using monitoring of the rate to regulate the power to the source. Open loop control works well with sputtering because the deposition rate is approximately linear with power to the source and can readily be held constant. Closed loop control is often used for evaporation, particularly electron beam evaporation in which a very hot region has to very carefully regulated to maintain the deposition rate constant. In MBE, the temperature of the sources can be very precisely regulated (to better than 0.1 K), and the stability of the deposition rate makes the need for closed-loop control intermediate between the need in the case of evaporation and that of sputtering. Nonetheless, monitoring is usually employed. The means of measuring deposition rate include: quartz balance/oscillators, mass spectrometers, electron impact monitoring and optical emission control. Only optical emission and quartz balances can work for sputtering because of the presence of the sputtering gas. Careful choices have to be made about how close the monitoring is to the source; the closer it is, the greater the measured rate and the greater the accuracy with which it can be measured, but the less the monitor represents conditions on the substrate. Clearly, an open-loop system will give a gradually worsening thickness error if there is a drift in deposition rate. It is possible in a closed-loop system to trig-
353
ger shutter movements on the basis of total deposited thickness, but this suffers from errors, particularly at smaller layer thicknesses. An alternative is to measure not thickness but deposition rate and to average over some time before the signal is used to adjust the rate. Direct monitoring of the multilayer itself is possible when it has a high contrast in electron density from layer to layer. This may be expected if the multilayer is intended for use as an X-ray mirror, but the technique is not confined to that case. The reflectivity at a given angle is determined by the X-rays used for monitoring and the intended periodicity. The intensity is followed and the shutter movements (to permit deposition from one or the other source) are made to coincide with the extrema of the intensity oscillations as has been discussed by Spiller (1985). In this way, small thickness errors are to some extent compensated for in successive layers. The method is best suited to the control of large layer thickness, the X-ray data being used for occasional adjustment of deposition rate rather than for direct shutter control. The method can work for any vapor-phase deposition technique. Another monitoring technique which can be widely applied is ellipsometry (Houdy, 1988). The ellipsometric signal is particularly sensitive to the surface of the deposit and can be used to obtain information on the growth process. Both X-ray reflectivity and ellipsometry as elements of a feedback loop can potentially suffer from jitter; the triggering of the shutter change is at an extremum in the property being measured and therefore subject to maximum uncertainty. However, if the atomic emission from the sources is sufficiently stable on a short timescale, this problem can be circumvented by introducing a phase shift
354
8 Metallic Multilayers
between the property observation and the shutter movement in response. At present there has been no claim of better than 0.3% stability in deposition rate over a full run. There is room for improvement in this figure. The deposition rates for precision multilayers are typically 0.1 to 0.3 n m s " 1 , and the accuracy with which these rates can be measured is limited by noise to the range 0.3 % to 1 %. With high quality X-ray mirrors, a very accurate measure of the repeat distance can be obtained after the film has been prepared. This can be used as a means of correcting the computer control of the deposition rates in a sputter deposition system. 8.4.4.2 Size and Uniformity In the previous section the concern was with the accuracy and constancy of layer thickness at a given point on a substrate. Also important for most applications is the uniformity of layer thickness across the substrate, and the difficulty of achieving this must depend on the lateral size, which can vary widely from application to application. Most multilayers for research need only be 1-25 mm in lateral dimension, depending on the type of characterization and physical property measurement that is to be used. On the other hand, a recording medium may be required to be made on a 150 mm diameter substrate for which specially designed cylindrical sputter sources may be particularly useful. X-ray mirrors used in bench top spectrometers are typically 30 mm by 100 mm to 300 mm and have strict requirements for uniformity. These sizes can easily be achieved by the use of sputtering (possibly with a long magnetron source). In sputtering, it is appropriate to scale the target size with the substrate, maintaining a low target erosion rate and consequently low drift. A further
problem with large size can be substrate heating. If a single crystal multilayer is to be made, then the scale of the heating required, especially if the epitaxy temperature is high (say>600°C), may limit the size. Heating also poses technical problems when the substrate must be rotated. In this case indirect heating such as an electron beam, quartz lamps or other infra red sources are best used to eliminate the need for sliding current connections. Typically, the variation of deposition rate across a substrate is related to the target-substrate distance roughly according to the inverse square law. Greater uniformity can then be achieved by increasing the target-substrate distance. If a high level of uniformity is required, however, this policy is very inefficient in usage of target material; for evaporation, in particular, there are considerable problems with the reduction in deposition rate and the increased incorporation of impurities. A relatively simple means of achieving better uniformity is to introduce shielding, that is, small sheets of metal (possibly with a pattern of perforations) in the path of the depositing material which, by shadowing, make the distribution of deposited species more uniform. Carefully designed shielding, combined with substrate rotation, is an efficient means of achieving uniformity. For cold depositions, the appropriate rotation stages can be scaled up quite easily. In evaporation systems, it is customary to rotate substrates about the axis of the deposition system which often has two (or more) sources offset from this central axis; a rotation under these circumstances evens out the layer by removing the approximately linear variation of the deposition rate across the substrate due to the offset of the source. In general, however, a rotation only removes asymmetric spatial variation and specially placed shields must be intro-
8.4 Preparation
duced to reduce symmetric non-uniformities. The greater the complexity of the shielding, the more material-specific the deposition process becomes. The largest multilayers made by evaporation are the X-ray mirrors (250 mm diam.) of Spiller and Golub (1989), in which good uniformity was achieved by a combination of large source-substrate distance and several special shields. Still larger mirrors, like the aperiodic neutron mirrors of Mezei (1976), have been made using a linear magnetron with accurately controlled travel of the substrate beneath. 8.4.4.3 Deposition Rate The choice of deposition rate for multilayer fabrication is a matter of compromises. Normally, the faster the film is deposited the fewer the impurities, but there is a limit to this effect when a source driven harder generates more heat and leads to increased impurity outgassing. With faster deposition, control is more difficult, both because shutter timing is more critical and because the deposition rate shows greater variation on short timescales. In practice, most deposition rates are of the order of one monolayer per second. 8.4.4.4 Buffer Layers The control of the quality of single crystal multilayers often relies on choice of a buffer layer, i.e., a layer deposited on the substrate before the multilayer itself. For a good multilayer to be obtained, the growth must be by the layer-by-layer (Frank-van der Merwe) mechanism. Yet often this mechanism will not operate initially. For copper on rocksalt or niobium on sapphire, for example, a layer thickness of 0.1 to 0.2 \xm is needed before the initial island growth (Volmer-Weber) gives way to layer
355
growth. A buffer layer of one of the elements in the multilayer can be grown first to establish a single crystalline thin film with good epitaxy. Buffer layers may also be of a material not occurring in the multilayer. One of the best examples of this pioneering technique was the use of niobium on sapphire as a buffer for the growth of Dy/Gd multilayers (Kwo et al., 1985). If a buffer layer does not have the same lattice spacing as the multilayer to be grown then it may be appropriate to grade the composition of the buffer so that it is latticematched (Vook, 1988), and this may require more sources in the deposition chamber. The number of demonstrations of epitaxial growth is increasing rapidly as highquality substrates (for example: sapphire, cubic yttrium-stabilized zirconia (YSZ), and magnesia) become more widely available and the possibilities of buffer layers are realised. It seems that almost any metallic element, if pure enough, can be grown as a single-crystal epitaxial layer (Matthews, 1975). One problem is that a large buffer layer may interfere with measurements of the properties of the multilayer. There may be ways to alleviate this: for example, it has been reported that deposition of an alloy base layer can facilitate measurement of electrical resistance (Huang etal., 1990). Another potential role of a buffer is to achieve greater smoothness. It has been reported, for example, that X-ray mirror performance can in some circumstances be improved if an initial layer of carbon is deposited prior to the growth of the W/C multilayer. How much the observation of this smoothing effect depends on the original roughness of the underlying substrate has not been unequivocally established.
356
8 Metallic Multilayers
8.4.4.5 Level of Vacuum
As suggested above, the deposition process is often in reality complex. The quality of the deposition environment is very important (and often not fully monitored). Whereas most of the individual phenomena associated with thin film growth such as surface mobility, re-nucleation, impurities and temperature effects are understood in general, most multilayers are made under circumstances in which the contributions of the various factors are not well established. The vacuum cleanliness plays an important and not fully understood role in the quality of the layers. The experience with the growth of semiconductor multilayers has shown that over several years, as the quality of thin film deposition has improved, the low-temperature electron mobility in GaAs/AlGaAs multilayers has continued to improve, even though the impurity levels concerned may be at the 1 ppb level. The cleanliness of metallic multilayers has never been claimed to be near that level and there is clearly room for improvement. In some cases this may require the availability of purer target materials. In most cases, however, the impurity level is largely a function of the deposition system. For better control, systems permitting loading without exposure of the main chamber to air (i.e., with load-locks) are being used. These are very useful if a reasonably large charge of material is available in the evaporation source or in the thickness of a sputtering target. For research instruments, in which the sputtering or evaporation source may be of small capacity, the short lifetime of the source may make load-locking not worthwhile. The level of cleanliness is of considerable importance for the realisation of ideal surface properties. Taking the ideal surface mobility as in Section 8.4.2.2, and assum-
ing a deposition rate of ~ 1 monolayer per second, the surface diffusion lengths at 0.2, 0.3, and 0.4 Tm would be respectively 350 nm, 5 |im, and 17 |im. In traveling these distances an atom would actually jump 4 x 106, 6 x 108, 7 x 109 atomic sites, whereas at a base pressure of 10 ~ 10 mbar only ~ 3 x 104 jumps would be possible before the surface migrating atom comes across an impurity atom. A base pressure of 10" 1 2 mbar would be needed to allow the diffusion length to reach its intrinsic limit at 0.2 Tm9 and 10" 1 4 mbar at 0.3 Tm. In practice, the atoms on the surface of a growing film encounter many impurities of varying reactivity and this will greatly reduce the surface mobility. In some cases, of course, impurities may be beneficial. For example, Spiller and Golub (1989) have reported that tungsten forms a desirable amorphous structure when deposited in an evaporation system held at 10~ 3 Pa. Even for single crystal growth it is not totally clear that the optimum deposition conditions will be those of highest purity. If an impurity accidentally incorporated is beneficial, however, it is always better in the long term to add the impurity in a controlled way in a system which is clean enough to permit this. 8.4.5 Summary, Issues and Outlook 8.4.5.1 Comparison of Evaporation and Sputtering
While MOCVD is clearly full of potential for the deposition of metallic multilayers, the great majority of multilayers have been made by conventional evaporation or sputtering, though MBE is increasingly being used. It is between the physical deposition methods that a choice will most often have to be made. In practice, there is little difference between a good conventional
8.4 Preparation
evaporation system using electron-beam guns and Knudsen cells, and a standard MBE system using Knudsen cells throughout. Ultimately the levels of deposition rate control and of vacuum depend on the investment in the control systems and in pumping. It has not been demonstrated that any one deposition method is always to be preferred. For both evaporation (conventional or MBE) (Section 8.4.3.2) and sputtering (Section 8.4.3.3), the quality of the vacuum varies from one deposition system to another. Over recent years, continual improvements have been made in the vacua and cleanliness achieved in all types of deposition system. The two general methods may be compared by considering what can be purchased for a given price and what feature is desired in the deposited layers. To deposit alloys and very refractory metals, sputtering has many advantages over evaporation, whilst for high levels of cleanliness and in-situ surface analysis, particularly for pure metals, evaporation is clearly superior. Most techniques of in-situ surface analysis, e.g., RHEED (Arrot and Heinrich, 1988), require a base pressure of 10 ~6 Pa, or less, and cannot easily be applied in sputtering. MBE in a high-quality UHV system clearly has the potential to achieve very high levels of purity; the pressure can be monitored during deposition, and the distillation process by which the Knudsen cells operate also helps. Sputtering has the disadvantage that all impurities are transferred almost directly from target to substrate, though gettering on cryocooled walls may reduce the impurity content by a factor of up to three. On the other hand, the sputtering gas itself may contribute to the impurity level. There is considerable current effort to produce hyperpure argon, with levels reaching the 1 ppb level, for use as the sputtering gas. Such
357
cleanliness levels combined with a UHV system would be worthwhile only if the targets to be used have contents of gaseous impurities on or below the 1 ppm level. Although it has not been established that gettering can clean up the sputtering gas to levels which make sputtering as good at present as MBE, the best niobium films (with extremely high resistance ratios) are as easily produced by sputtering as by MBE. This is partly because the sources used for evaporation are electron-beam guns operating at high powers not ideal for the best vacuum conditions during deposition. Sputtering is technically far simpler to control accurately, though there are now several (expensive) systems which will allow the control of deposition rate in evaporation to better than 1%. In contrast, with sputtering a calibration of the deposition rate with time will permit the control of deposition rate to 0.1 % relatively easily. The problem with control at this level for evaporation is that the integration times for the monitoring equipment are long in comparison to the time require to deposit one layer, and hence active feedback is always difficult to achieve for short period multilayers. Modern effusion cells and electron beam sources, however, are always increasing in stability. Sputtering has some advantage over evaporation in that a gentle atomic bombardment ("atomic peening") is intrinsic to the process. However, an ion source may be easily incorporated in an evaporation system to achieve the same effect (sometimes termed "ion-polishing" (Spiller, 1989)). An effect of such polishing is shown in Fig. 8-11. The extremely high levels of purity needed in the most advanced research in the semiconductor industry have driven the development of Knudsen cells and MBE. While great ef-
358
8 Metallic Multilayers
21 22 23 2k Grazing angle (degrees)
Figure 8-11. Reflectivity for X-rays (/l = 48nm) as a function of grazing angle for two similar (repeat distance = 6.2 nm, ~ 50 layers) Rh/C multilayer X-ray mirrors, showing the considerable (>2x) improvement gained from ion polishing. [Adapted from Spiller (1989).]
fort has been made to develop MBE as the prime method for precision deposition, less effort has been made in developing an equivalent sputter source. One reason for this is the problem of target purity, but this is an area where improvements are being made; manufacturers are becoming more aware of the gas contents of the target materials. If the same investment had been made in sputter deposition as has gone into MBE, then many developments would be foreseeable: for example, in-situ melting and thus cleaning of targets, and self-sustaining targets which use the sputtered species as the sputtering ions. For simplicity and versatility UHV sputtering has advantages, but for in-situ surface analysis during deposition MBE is clearly to be preferred. 8.4.5.2 Issues in Multilayer Deposition Whilst the general features of single crystal metal and semiconductor multilayers are very similar, there are also many important differences (Volume 4, Chapter 8). The similarities of the means of deposition and characterization are to be con-
trasted with the differences in the degree of covalency of the bonding, the role of defects and the range of applications and systems that can be formed. The greater covalency (i.e., directionality of the bonding) of the semiconductors is an important feature. Although metals can be strongly bound, if not so directionally, lower homologous substrate temperatures are generally used during deposition. Another significant difference rests in the range of different metal systems which can be studied and are potentially useful, whereas in the semiconductor case the need for near perfect dislocation- and defect-free structures limits the element and alloy combinations to more finely matched alloy systems. Semiconductor multilayers are almost all used for their potentially useful charged particle mobility, and any extended defect has a deleterious effect on such a property. The applications of metallic multilayers, on the other hand, do not appear to be so affected by such defects. In contrast, as discussed in Section 8.3.3, the semiconductor systems are more tolerant of interfacial imperfections as the characteristic de Broglie wavelength is so much longer than for metals. As discussed in Section 8.3, the various properties of multilayers can depend in quite different ways on their structure. The necessary degree of structural perfection depends on the application for which the multilayer is intended. This applies to the ideality of the interfaces produced, the uniformity of thickness over the area of the film, and the constancy of the period throughout the thickness of the film. A sample for a diffusion experiment, for example, can be quite small (3 mm square) and have a repeat distance of as much as 5 nm; even if this period drifts by 1 to 2%, the low order superlattice lines necessary for the experiment will still be easily seen.
8.4 Preparation
An X-ray mirror, on the other hand, with a repeat distance of 2.5 nm, needs a constancy of period of better than 0.5% to take advantage of the 150 to 400 layers which typically are needed. Each of these examples requires some regularity of layering throughout the sample. For many properties, however, only the local repeat distance is important. The magnetic perpendicular anisotropy in Co/Pt multilayers, for example (Section 8.3.5) varies smoothly with the local repeat distance, and the level of control needed is relatively small. An important parameter in deposition from the vapor is the substrate temperature, and there is much current debate about the optimum temperature to obtain high quality metallic multilayers. This may be an area in which the lessons from semiconductor deposition have to be modified. The growth temperature for a multilayer which is to have reasonably sharp interfaces must be one at which the atoms arriving at the surface have relatively little energy compared to the activation energy for surface diffusion, and even less compared to the energies necessary for bulk diffusion or for re-evaporation. For single crystal films, if a sharp interface is required, the deposition temperature should be as low as possible consistent with relatively defectfree growth; the preferred reduced temperature is typically <0.25 r m . The deposition temperature for a single crystal film must be high enough to permit the surface diffusion necessary for the layer-by-layer growth. It is becoming clear that as deposition systems improve, the growth temperatures can be lowered as the number of impurities on the surfaces decrease (Bauer et al., 1988). One way to achieve low defect densities at a lower temperature is to allow time for the atoms to find their ideal positions. This technique is already in practice
359
with semiconductors, and has been attempted for metallic multilayers. The optimum deposition temperature may well differ for the two materials in a multilayer (the problem of materials with widely differing melting temperatures has already been discussed in Section 8.2.4), and it may even change as a given layer is deposited. For example, to inhibit intermixing it would be useful to keep it low at the beginning of each new layer. The lack of consensus in the literature about the ideal deposition temperature for a given element is at least partly because the quality of the vacua varies from one deposition system to another. There are now reports of RHEED oscillations being observed during deposition at temperatures of just above Vs Tm, implying that at this temperature layer-by-layer growth is becoming possible (Bauer et al., 1988). Together with other results, this suggests that between V6 and V4 Tm is an adequate temperature to grow good single crystal films. A major remaining issue is how perfect a single crystal metallic multilayer can be. Is it possible to deposit a multilayer at a temperature at which the interfaces are perfectly atomically sharp as well as flat? If a multilayer were grown under these conditions, would be inevitable ledges produced during growth prevent observation of desired phenomena? These questions for single crystal multilayers remain. At the other extreme of atomic mobility is the growth of amorphous films. Although surface mobility is often taken to be the sole criterion for the formation of an amorphous phase, it is important to recognize that the nucleation of crystalline phase limits the upper temperature for amorphous phase growth. Particularly for alloys and intermetallics, effective diffusion lengths do not have to be very short for this nucleation stage to be difficult, as
360
8 Metallic Multilayers
the competing phase may have a large unit cell and hence a large activation energy for nucleation (Thornton, 1977). As with single crystal films, the issue is whether interfaces in an amorphous multilayer can be made both atomically sharp and smooth. There is the possibility that the surface of an amorphous phase could be smooth in a manner analogous to a liquid surface. Bombardment by energetic species during deposition could aid the attainment of this ideal, whether by the ballistic process of "peening" or by raising the local effective temperature. There is some evidence to suggest that in clean vacuum systems during the growth of amorphous films, atoms are able to diffuse over the surface to seek their lowest energy positions. Between these extremes of growth conditions yielding single crystal and amorphous films, polycrystalline materials are obtained which can have a wide variety of morphologies. Despite the general tendency for polycrystalline films to have inherently rough growth fronts owing to the range of orientations of crystallites in the front, many such systems show remarkably sharp interfaces. The study of such features may prove an interesting area for future research. 8.4.5.3 Outlook for Ideal Multilayers
As the deposition of metallic multilayers is developed, the importance of impurities is increasingly being recognized. One development which has occurred recently, and will take some time to affect routine thin film deposition, is the introduction of extra high vacuum (XHV) chambers. The systems have specially treated aluminumwalled chambers which reduce the outgassing rates so considerably that base pressures of ~10~ 1 4 mbar can be envisaged. At such a vacuum level, slower depo-
sition rates may be used and, if the source material can be pure enough, even lower temperatures will suffice for epitaxial growth processes.
8.5 Structural Characterization 8.5.1 Introduction
The structural characterization of multilayers has been essential in developing an understanding of the link between deposition conditions and properties. It has also been a stimulus to the production of multilayers which are ever more precise and close to ideal. Indeed, the advances in deposition capabilities have to a large extent been matched by advances in characterization. Examples of the latter are: a clearer understanding of the origins of X-ray scattering effects from multilayers, the availability of intense X-ray (from synchroton sources) and neutron (from spallation sources) beams for diffraction studies, high resolution electron microscopy including image simulation, and the wider use of Rutherford backscattering and ion channeling. It is the properties of metallic multilayers (Section 8.3) which make them of interest both scientifically and technologically, and the most straightforward type of characterization is that of property measurement. In this way, the most direct link can be established between the property required for an application and the basic multilayer design (compositions, thicknesses and crystal structures of the layers) and deposition conditions. Such an empirical approach may well be attractive (an example is shown in Fig. 8-12), because except in a few cases (band-structure-based calculations of magnetic properties may be the most notable exception, e.g., Oguchi and Freeman (1986)), ab initio prediction
8.5 Structural Characterization
361
Figure 8-12. Magneto-optic Kerr effect hysteresis curves (rotation angle 9k versus applied perpendicular magnetic field) for a series of Co/Pd multilayers with constant Pd-layer thickness, but different Co-layer thicknesses. This illustrates how a technically useful property may be used to characterize a multilayer. [Adapted from Engel and Falco (1990).]
of multilayer properties has not been achieved. To achieve a desired property by design in general requires understanding both of the dependence of structure on deposition conditions and of the dependence of the property on structure. This understanding, which can only be developed by comprehensive structural characterization, is essential for the realisation and optimization of properties more efficiently than by trial and error. This section is concerned with the methods of structural characterization. All the standard methods for materials characterization can be applied; the aim here is to discuss the particular types of characterization most useful for multilayers and to give examples of the main types of information which can be obtained. The structure of multilayers (Section 8.2) has several components: the nature of the individual layers (crystal structure and orientation, amorphicity or grain size, defect density, stress, composition), the nature of the layering (periodicity, individual layer thickness, imperfections in these), and the nature of the interfaces (roughness, diffuseness, coherence, composition profile). All of these can be studied by Xray diffraction, which is by far the most widely used characterization technique.
8.5.2 X-Ray Diffraction X-ray diffraction is so widely used because of the range of information which can be obtained, of its comparative ease of use, and of its non-destructive nature (permitting, for example, studies of multilayers still on a substrate). The application of Xray diffraction to multilayers has been treated by McWhan (1985), Fujii (1987) and Cargill (1989). In most studies a diffractometer is used in symmetrical reflection geometry; the wave vectors of the incident and diffracted beams make equal (Bragg) angles 6 with the multilayer surface, and the scattering vector (wave vector of diffracted beam minus wave vector of incident beam) is parallel to the film normal. In this type of experiment (a 9 — 26 scan) the scattering vector (at fixed orientation relative to the sample) changes length, and only the structure of the multilayer in the direction normal to the substrate is directly probed. A compositionally modulated structure (Section 8.2.2) can be regarded as the multiplication of the electron density distribution in a material structure by a modulation function describing the variation in scattering potential. The diffraction pattern of a compositionally modulated mate-
362
8 Metallic Multilayers
rial is then the diffraction pattern of the (perhaps idealized) uniform, non-modulated structure convoluted with the diffraction pattern (Fourier transform) of the modulation function. A compositionally modulated single crystal shows the Bragg peaks expected for the uniform crystal, each surrounded by a set of satellites with a spacing related to the modulation wavelength. For an amorphous material, there are no Bragg peaks, but the satellites are still observed about the zero-order (undiffracted) beam, i.e., typically at a low diffraction angle 6 of a few degrees. Information may be obtained from a diffractometer scan as follows: the crystal structure (or amorphicity) from the positions of the Bragg peaks (or their absence); preferred orientation in a polycrystal from the relative intensities of the Bragg peaks; the modulation wavelength from the positions of the satellites; the film thickness or structural coherence in the direction normal to the substrate from the width of the Bragg peaks; and the variation in modulation wavelength from the width of the satellite peaks (in relation to the width of the Bragg peaks). In addition, information can be obtained on the amplitude and form of the composition modulation and of the associated strain modulation from the intensities of the satellites. Unfortunately, the usual phase problem in diffraction prevents direct determination of the composition and strain profiles, which are estimated by trial-and-error fitting to diffraction data. The strength of diffraction from metallic multilayers is such that often it is necessary to take dynamical effects into account in analysing peak intensities (Fujii, 1987) (see Vol. 2, Chap. 8). The basic nature of the diffraction pattern from a multilayer remains for twophase multilayers (Section 8.2.3). When the layer thicknesses are large, the diffrac-
tion pattern is essentially the superposition of the usual diffraction patterns of the two materials, but as the thicknesses are reduced, a distinctive pattern characteristic of a multilayer evolves which is most easily interpreted as an average diffraction pattern convoluted with a modulation function as in the compositionally modulated case. Fig. 8-13, from the work of Schuller 5 4
(a)
3 2 1 0 4 (b) 3 2 1
3.4 nm
5 0 § 3
£2
(0
o
i? I*
(d)
3 2 1 0 5
(e)
4 3 2 1 50 48 46 44 42 40 38 36 34 32 30 20 (degrees)
Figure 8-13. X-ray diffractometer traces (symmetrical 9-29 geometry, CuKa radiation) for a series of Nb/Cu multilayers with layer thicknesses (in nm) as shown. [Adapted from Schuller (1980).]
8.5 Structural Characterization
(1980), shows 6 — 26 scans for multilayers of niobium (with (110) planes parallel to the substrate) and copper (in (111) orientation). At the largest multilayer repeat distance, the two Bragg peaks (110)Nb and (11 l) Cu a r e visible, each with satellite peaks around it at a spacing related to the repeat distance. As the repeat distance is further reduced the pattern evolves toward one in which there is a central average Bragg peak with satellites. As has been demonstrated by Clemens and Gay (1987), at larger repeat distances the satellites around the Bragg peaks can arise only if there is coherence (in the direction normal to the substrate) between the atomic planes in successive layers of the same type. Fig. 8-14 from their work shows diffraction patterns from Mo-Ni and Ti-Ni multilayers of similar repeat distance. The Ti-Ni has a large' structural mismatch between the two materials and disordered interfaces; there is no coherence between the layers and consequently the Bragg peaks are observed without satellites. So far, only diffraction experiments with the scattering vector normal to the substrate have been considered. It is also possible to oscillate a scattering vector of given length about the substrate normal. In this type of experiment a "rocking curve" is obtained which shows the width of diffraction peaks in directions parallel to the film surface. The width of Bragg peaks gives information on the preferred orientation of the crystals. The rocking curves from satellite peaks give information on the flatness of the layers and may be useful in distinguishing between the effects of undulating layers and of interfacial diffuseness which are superposed in 6 — 26 scans. Experiments with the scattering vector lying in the plane of the thin film are useful in further characterizing multilayers. In such experiments the X-rays are transmit-
363
8000 -
6000 -
4000 -
- 2000 -
4000
3000 -
2000 -
1000 -
40 50 20 (degrees) Figure 8-14. X-ray diffractometer traces (symmetrical 0-2 6 geometry, CuKa radiation) for (a) Mo/Ni, and (b) Ti/Ni multilayers with repeat distance of 8 nm prepared by sputtering under similar conditions. The Ti/Ni multilayer has disordered interfaces and no satellite peaks. [Adapted from Clemens and Gay (1987).]
ted through the multilayer and it is necessary to remove the substrate. An example of the type of data which may be obtained is given in Fig. 8-15. This shows the (220) Bragg peak for a Cu/Ni multilayer in which the crystal orientation throughout is with (111) planes parallel to the surface. At small repeat distance, the copper and nickel layers are fully coherent and there is a single well defined (220) spacing of planes perpendicular to the surface. At larger repeat distance full coherency is lost as inter-
3
8 Metallic Multilayers
5
1.95
2.00 Miller Index (h)
2.05
Figure 8-15. The (220) average Bragg peak for transmission X-ray diffractometry (scattering vector in the plane of the thin film) scans from Cu/Ni epitaxial mosaic multilayers with (111) parallel to the substrate. The positions of the (220) reflections for pure copper and nickel are shown. Curve (a), from a multilayer with repeat distance 1.6 nm, shows the well defined lattice parameter of fully coherent layers; curve (b), from a multilayer with repeat distance 6.0 nm, shows the peak broadening arising from partial coherency and interfacial dislocations. [Adapted from Gyorgy etal. (1982).]
facial dislocations are introduced, and the (220) spacing can vary causing a broadening of the Bragg peak. Information on the degree of coherency can be obtained, albeit less directly, also from the standard reflection 9 — 26 scan. A composition modulation in a material not only causes a variation in scattering potential because of the atomic number variation itself, but also because of the variation in atomic size (and therefore lattice parameter). In a fully coherent multilayer the variation in plane spacing (measured normal to the substrate) is amplified, because the material of larger (smaller) lattice parameter is compressed (stretched) in-plane and, through Poisson ratio effects, expanded (contracted) out-of-plane. The satellites due to a composition modulation about the zero-order beam have intensities dependent only on the variation in atomic number. Intensities of satellites about Bragg peaks, however, depend also on the variation in lattice parameter. The
latter dependence introduces an asymmetry in the intensity profile as illustrated in the calculations of McWhan et al. (1983), reproduced in Fig. 8-16. The degree of asymmetry can then be related to the degree of coherence at the interfaces between the layers. New possibilities for X-ray diffraction studies are opened up by the use of intense synchroton radiation. It is possible, for example, to select an X-ray wavelength from
\ .J. /
(a)
/
i i i i
ntensity (arbitrary unit
364
(b) / i
j^
, /
,
1.
t
i i i i \ \ \ i
\\
1.0 q (arbitrary units)
1.1
Figure 8-16. Calculated 6-26 diffractometer scans for the first order Bragg reflection (average of (110)Nb and (111)A1) from a Nb/Al multilayer. Curve (a) takes the plane spacing dNb(110) = dA1(111); for (b), dNb(110) = 1.02dA1(111); and for (c), dNb(110) = 1.04 dA1(111). The satellites arising from the layering are symmetrical if they arise only from the modulation in scattering potential, but become asymmetrical if there is a variation in plane spacing (as would be true in practice). [Adapted from McWhan et al. (1983).]
8.5 Structural Characterization
the radiation to be close to an absorption edge for one of the elements in a multilayer. In the anomalous scattering which results, the scattering power of that element is altered. This is useful to obtain high contrast between two elements of neighboring atomic numbers. Nakayama et al. (1990) have used this technique for studying multilayers of iron (atomic number 26) and manganese (atomic number 25). At the very high incident intensities possible with synchroton radiation, it is possible to detect the weak X-ray magnetic scattering and in this way investigate magnetic ordering (often with better resolution than is possible using neutrons) (Vettier et al., 1986). Time-resolved studies of structural evolution would be of interest. In the symmetrical reflection geometry, an X-ray standing wave is set up in the multilayer. By adjusting the scattering vector, the antinodes of the standing wave pattern can be moved through the multilayer. At the antinodes, X-ray fluorescence is greatest, and the variation of fluorescence with scattering vector can be used to determine the distribution of an impurity. For example, using this technique, Matsushita et al. (1986) showed that argon impurity in a W/Si multilayer was concentrated mainly in the silicon layers. 8.5.3 Neutron Diffraction
Neutron diffraction may be used to characterize multilayers in the same way as X-ray diffraction. The relative difficulty of access to neutron diffraction facilities means, however, that it will be used only if there is a particular advantage. As for general structural studies, neutron diffraction is preferable to X-ray diffraction for obtaining contrast between elements of neighboring atomic number (though X-ray anomalous scattering can be used, Section
365
8.5.2), and for scattering from light elements. The particular importance of neutron diffraction for multilayers is in characterization of magnetic ordering. Since neutrons have a magnetic moment, diffraction of a polarized beam can be used, for example, to determine in detail the magnetization profile through the thickness of a multilayer. The classic studies in this area are of the magnetic ordering in Gd/Y (Majkrzak et al., 1987) and Dy/Y (Hong et al., 1987) multilayers as discussed in Section 8.3.5 (see also Vol. 3, Chap. 6). Because isotopes of an element may have quite different neutron scattering lengths it is possible to conceive of isotopically layered, chemically homogeneous multilayers which could be used for very sensitive self-diffusion studies, but there are no reports as yet. 8.5.4 Transmission Electron Microscopy
Thinning samples to be electron-transparent is necessarily destructive and sample preparation for transmission electron microscopy (TEM) is not straightforward (Newcomb e t a l , 1985; Lepetre et al., 1987). Nonetheless, the information which can be obtained is in many ways complementary to that obtainable from X-ray diffraction, and TEM is a widely used and important characterization technique. TEM of thin films with the electron beam parallel to the film normal has been used in studies of epitaxy (Chapter 7) and for general characterization of micro structure. Although such "plan-view" TEM is not of particular use for multilayers, sections at a slight angle to the multilayer surface are of interest; the sectioning gives an expanded view of the layering and interfacial defects (e.g. dislocations and voids) can be characterized. For multilayers most interest
366
8 Metallic Multilayers
has centered on "cross-sectional" TEM in which the electron beam is transmitted in a direction perpendicular to the film normal. The main driving force for the development of cross-sectional TEM has come from the semiconductor industry, but the technique has now been widely applied to metallic multilayers, as reviewed by Stobbs (1988). In contrast to X-ray and neutron diffraction, phase information is preserved in the electron microscope and it is possible to form images of the specimen. Consequently, the major feature of TEM multilayer studies is the characterization of local defects, such as dislocations, twins and unevenness in the layers, which cannot be directly studied by X-ray diffraction. A problem with TEM is that the image shows a superposition of effects from viewing the structure in projection. For this reason near-plan-view sections are more useful for characterizing interfacial defects (e.g., misfit dislocations) than are cross-sections. On the other hand, dislocations threading through the layers, and other defects, such as twins, parallel to the growth direction, can best be studied in cross-section (Fig. 817). One of the most direct applications of cross-sectional TEM is to study unevenness in the layers, though for roughness on a small lateral scale the projection problem may prevent useful characterization. Fig. 8-18 shows unevenness originating from the substrate (decreasing with distance from the substrate), while Fig. 8-19 shows more marked unevenness arising from the microstructural development in a polycrystalline multilayer (increasing with distance from the substrate). For flat interfaces (and most interfaces in useful multilayers are at least locally flat) viewed in cross-section, there has been much work in developing TEM characterization methods. It is important to realise
j y t nil
Figure 8-17. Cross-sectional dark-field transmission electron micrograph of twin boundaries inclined to the growth direction in a Au/Ag multilayer of 5 nm repeat distance with (001) planes parallel to the substrate. [From Baxter et al. (1984).]
Figure 8-18. Cross-sectional bright-field transmission electron micrograph of a W/Si X-ray mirror (repeat distance 6.5 nm), showing substrate-derived unevenness in the layers. The vertical defect may be due to unevenness in the direction normal to the plane of view. [From Shih (1991).]
that in TEM, the imaging conditions (e.g., level of defocus) have significant effects on the image. Thus the image is not always capable of direct interpretation; comparison with computer-calculated simulated images is necessary. An example of this is shown in Fig. 8-20 for a Cu/Ni-Pd multi-
8.5 Structural Characterization
Defocus:-5?0nm
-220nm
-20nm
367
+200nm
Figure 8-19. Cross-sectional bright-field transmission electron micrographs of Co/Pd multilayers (0.3 nm Co/1.2 nm Pd) made under sputtering pressures of (a) 2.41 Pa and (b) 0.41 Pa. The polycrystalline materials show a high defect density and uneven layers arising from the growth microstructure. The higher sputtering pressure during deposition gives lower bombardment, a lower quality of layering, but better magnetic properties for magneto-optic data storage. See also Fig. 8-6, from the same set of samples. [From Highmore et al. (1991).]
Figure 8-20. Fresnel contrast in transmission electron micrographs can be used to estimate interface sharpness by comparison with simulated images. For a Cu/Ni-Pd multilayer with repeat distance 4.4 nm are shown for the defocus values marked: (a) a through-focal series of electron micrographs; (b) the best-fit simulated images, calculated assuming intermixing over two atomic planes at the interfaces; and (c), for comparison, simulated images obtained when it is assumed that four atomic planes are intermixed. [From Baxter and Stobbs (1985).]
layer. A series of experimental images is shown at varying defocus together with two sets of simulated images for different assumed composition profiles. In this way, essentially making use of the Fresnel fringes due to the interfaces, the composition profile can be determined. From such images it may also be possible to determine the rigid body displacement between the lattices on either side of an interface (Stobbs, 1988).
composition of the removed material can be analysed by secondary ion mass spectrometry. An alternative, non-destructive technique is Rutherford backscattering. All these techniques have been described by Feldman and Mayer (1986). The depth resolution of the techniques is such that the composition modulation is detectable in multilayers with repeat distance greater than about 5 nm. Associated with Rutherford backscattering is the phenomenon of ion channeling, in which the yield of backscattered ions is dramatically reduced for particular crystal orientations (Feldman and Mayer, 1986). In these orientations the ions are channeled through the crystal, but the channeling effect is severely disrupted by lattice defects or interstitial atoms. As a sensitive measure of crystal perfection, ion
8.5.5 Techniques for Chemical Profiling
Several techniques suitable for fine scale chemical profiling have been applied in thin film studies. As the surface of a thinfilm sample is eroded by ion-milling, the composition of the surface can be measured by Auger electron spectroscopy or the
368
8 Metallic Multilayers
channeling is useful for characterizing epitaxial single crystal multilayers.
8.6 Applications and Outlook
8.5.6 Probes of Local Structure
As outlined in Section 8.3, metallic multilayers have many interesting properties. The majority of metallic multilayers have been used in research on these properties, and many possibilities remain to be explored. Fundamental materials and physics research, rather than technological application, is likely to be the main driving force for the production of multilayers which are more precisely designed and closer to perfection. The main technological application of multilayers so far is as optical elements for X-rays and neutrons (Section 8.3.2), but this is likely to remain on a small commercial scale. The largest potential application readily foreseeable for metallic multilayers is as media for magnetic or magneto-optic data storage. The properties of Co/Pt and Co/Pd multilayers, for example, are particularly well suited for these applications (Section 8.3.5). Although electron transport phenomena in metallic multilayers are of great scientific interest, they seem unlikely to be the basis of large scale commercial application as for semiconductor multilayers. It is early as yet in multilayer studies, however, to identify the main future applications, especially given the wide range of special properties as described in Sec. 8.3.
The high energy side of an X-ray absorption edge has an oscillatory modulation and analysis of this extended X-ray absorption fine structure (EXAFS) yields information on the local environment of the absorbing atoms (number and identity of neighboring atoms). The technique has been widely applied to amorphous alloys for which the same information is difficult to obtain by conventional X-ray scattering measurements (Wong, 1981). For multilayers, EXAFS, particularly at glancing angle (Lamble et al., 1988) is useful for its sensitivity to interface structure. In particular, any amorphous phase at the interfaces can readily be detected. Mossbauer spectroscopy uses the interaction between the nuclear quadrupole moment of a probe atom and the local electric field gradient to provide information on the symmetry of the local environment. The most widely used probe species is 57 Fe, and for multilayers the technique is mainly of use when one type of layer is iron (Shinjo, 1987). In addition to magnetic measurements (e.g., determining the directions of spontaneous magnetization), determination of the hyperfine field distribution in the multilayer can be used to find the composition profile, quantitative estimates of the interfacial mixing being possible in some cases. Nuclear magnetic resonance (NMR) has also been applied to metallic multilayers. It is particularly useful for studies of magnetism at the interfaces, enabling conclusions to be reached about the degree of interfacial mixing (Yasuoka, 1987). The techniques of characterization briefly mentioned here are discussed in more detail in various chapters of Volume 2.
8.7 Acknowledgements The authors are grateful to the Science and Engineering Research Council (UK) for support of their research in this area. Useful discussions with Dr C. S. Baxter are gratefully acknowledged.
8.8 References
8.8 References Arrot, A. S., Heinrich, B. (1988), in: Metallic Multilayers and Epitaxy: Hong, M., Wolf, S., Gubser, D. C. (Eds.). Warrendale, PA: The Metallurgical Society, pp. 147-165. Atzmon, M., Unruh, K. M., Johnson, W. L. (1985), /. Appl. Phys. 58, 3865-3870. Barbee, T. W. (1985), in: Synthetic Modulated Structures: Chang, L. L., Giessen, B. C. (Eds.). New York: Academic Press, pp. 313-337. Barbee, T. W. (1988), Mater. Res. Soc. Symp. Proc. 103, 307-314. Bauer, E., Jalochowski, M., Koziol, C , Lilienkamp, G. (1988), in: Proc. MRS Int. Meeting on Advanced Materials Vol. 10: Doyama, M., Somiya, S., Chang, R. P. H. (Eds.). Pittsburgh: Mater. Res. Soc, pp. 505-519. Baxter, C. S., Somekh, R. E., Stobbs, W. M. (1984), Proc. 42nd. Annual Meeting of the Electron Microscope Society of America: Bailey, G. W. (Ed.). San Francisco: San Francisco Press, pp. 588-589. Baxter, C. S., Stobbs, W. M. (1985), in: Inst. Phys. Conf. Ser. No. 78 EMAG '85, Newcastle upon Tyne: Tatlock, G. I (Ed.). Bristol: Adam Hilger, pp. 387390 Bickerdike, R. L., Clark, D., Easterbrook, J. N., Hughes, G., Mair, W. N. Partridge, P. G., Ranson, H. C. (1985), Int. J. Rapid Solidification 1, 305-325. Bionta, R. M., Jankowski, A. R, Makowiecki, D. M. (1988), Mater. Res. Soc. Symp. Proc. 103, 257-263. Cammarata, R. C. (1988), Mater. Res. Soc. Symp. Proc. 103, 315-325. Cammarata, R. C , Schlesinger, T. E., Kim, C , Qadri, S. B., Edelstein, A. S. (1990), Appl. Phys. Lett. 56, 1862-1864. Cammarata, R. C , Sieradski, K. (1989), Phys. Rev. Lett. 62, 2005-2008. Carcia, P. F., Shah, S. I., Zeper, W. B. (1990), Appl. Phys. Lett. 56, 2345-2347. Cargill, G. S. (1989), Mater. Res. Soc. Symp. Proc. 151,231-242. Clemens, B. (1987), J. Appl. Phys. 61, 4525-4529. Clemens, B. M., Gay, J. G. (1987), Phys. Rev. B 35, 9337-9340. Clemens, B. M., Eesley, G. L. (1988), Phys. Rev. Lett. 61, 2356-2359. Clevenger, L. A., Thompson, C. V., Cammarata, R. C , Tu, K. N. (1988), Mater. Res. Soc. Symp. Proc. 103, 191-196. Deubner, W. (1930), Ann. Phys. 5, 261-324. DuMond, J., Youtz, J. P. (1940), /. Appl. Phys. 11, 357-365. Engel, B. N., Falco, C. M. (1990), Mater. Res. Soc. Bull. 15, 34-37. Esaki, L. (1985), in: Synthetic Modulated Structures: Chang, L. L., Giessen, B. C. (eds.). New York: Academic Press, pp. 3-41.
369
Feldman, L. C , Mayer, J. W. (1986), Fundamentals of Surface and Thin Film Analysis. New York: North Holland. Fernandez, F. E., Falco, C. M. (1985), Proc. Soc. Photo-Opt. Instrum. Eng. 563, 195-200. Fu, C. L., Freeman, A. J.,, Oguchi, T. (1985), Phys. Rev. Lett. 54, 2700-2703. Fujii, Y (1987), in: Metallic Super lattices: Artificially Structured Materials: Shinjo, T, Takada, T. (Eds.). Amsterdam: Elsevier, pp. 33-80. Graham, L. D., Kraft, R. W. (1966), Trans. Met. Soc. AIME 236, 94-102. Greer, A. L. (1991), Mater. Sci. Eng. A 132, in press. Greer, A. L., Spaepen, F. (1985), in: Synthetic Modulated Structures: Chang, L. L., Giessen, B. C. (eds.). New York: Academic Press, pp. 419-486. Gyorgy, E. M., McWhan, D. B., Dillon, J. F., Walker, L. R., Waszczrak, J. V. (1982), Phys. Rev. B25, 6739-6747. Highmore, R. I, Shih, W C , Somekh, R. E., Evetts, J. E. (1991), J. Vac. Sci. Technol. A, in press. Hong, M., Fleming, R. M., Kwo, X, Schneemeyer, L. F., Waszczak, J. V., Mannaerts, J. P., Majkrzak, C. F., Gibbs, D., Bohr, J. (1987), /. Appl. Phys. 61, 4052-4054. Houdy, Ph. (1988), Rev. Phys. Appl 23, 1653-1659. Huang, K. H., Blamire, M. G., Somekh, R. E. (1990), Vacuum 41, 1237-1240. Jin, B. Y, Ketterson, J. B. (1989), Adv. Phys. 38, 189-366. Johnson, W. L. (1986), Prog. Mater. Sci. 30, 81-134. Khan, M. R., Roach, P., Schuller, I. K. (1984), Thin Solid Films 122, 183-189. Koehler, J. S. (1970), Phys. Rev. B 2, 547-551. Kortright, J. B., Joksch, St., Ziegler E. (1991), J. Appl. Phys. 69, 168-174. Kwo, J., McWhan, D. B., Hong, M., Gyorgy, E. M., Feldman, L. C , Cunningham, J. E. (1985), Mater. Res. Soc. Symp. Proc. 37, 509-515. Lamble, G. M., Heald, S. M., Sayers, D. E., Ziegler, E. (1988), Mater. Res. Soc. Symp. Proc. 103, 101107. Lepetre, Y, Ziegler, E., Schuller, I. K. (1987), Appl. Phys. Lett. 21, 1480. Lowe, W. P., Geballe, T. H. (1984), Phys. Rev. B29, 4961-4968. Majkrzak, C. F. (1989), Physica B 156 & 157, 619626. Majkrzak, C. E, Cable, J. W, Kwo, I, Hong, M., McWhan, D. B., Yafet, Y, Waszczak, J. V., Grimm, H., Vettier, C. (1987), J. Appl. Phys. 61,4055-4057. Matsushita, T, Iida, A., Ishikawa, T., Nakagiri, T., Sakai, K. (1986), Nucl. Instrum. Meth. A 246, 751757. Matthews, J. W. (Ed.) (1975), Epitaxial Growth. Orlando: Academic Press. McKee, R. A., List, F. A., Walker, F. J. (1988), Mater. Res. Soc. Symp. Proc. 103, 35-40. McWhan, D. B. (1985), in: Synthetic Modulated
370
8 Metallic Multilayers
Structures: Chang, L. L., Giessen, B. C. (Eds.). New York: Academic Press, pp. 43-74. McWhan, D. B., Gurvitch, M., Rowell, I M., Walker, L. R. (1983), J. Appl. Phys. 54, 3886-3891. Mei, I, Fernando, G. W. (1991), Phys. Rev. Lett. 66, 1882-1885. Messier, R., Yehoda, J. E. (1985), J. Appl. Phys. 58, 3739-3746. Mezei, R (1976), Commun. Phys. 1, 81-85. Moreau, A., Ketterson, J. B., Mattson, I (1990), Appl. Phys. Lett. 56, 1959-1961. Moustakas, T. D., Koo, J. Y, Ozekcin, A. (1988), Mater. Res. Soc. Symp. Proc. 103, 41-46. Movchan, B. A., Demchishin, A. V. (1969), Fiz. Met. Metalloved. 28, 653-660. Nakayama, N., Moritani, I., Shinjo, T, Fujii, Y, Sasaki, S. (1990), Phil. Mag. A 59, 547-567. Newcomb, S. B., Boothroyd, C. B., Stobbs, W. M. (1985), /. Microsc. 140, 195-207. Oguchi, T, Freeman, A. J. (1986), /. Magn. Magn. Mater. 54-57, 797-798. Philofsky, E. M., Hilliard, J. E. (1969), J. Appl. Phys. 40, 2198-2205. Rhead, G. E., (1989), Int. Mater. Reviews 34, 261276. Ruggiero, S. T., Barbee, T. W, Beasley, M. R. (1980), Phys. Rev. Lett. 45, 1299-1302. Ruggiero, S. T., Beasley, M. R. (1985), in: Synthetic Modulated Structures: Chang, L. L., Giessen, B. C. (Eds.). New York: Academic Press, pp. 365-417 Schuller, I. K. (1980), Phys. Rev. Lett. 44,1597-1600. Schuller, I. K. (1988), Mater. Res. Soc. Symp. Proc. 103, 335-341. Sevenhans, W., Vanderstraeten, H., Locquet, J. P., Bruynseraede, Y, Homma, H. Schuller, I. K. (1988), Mater. Res. Soc. Symp. Proc. 103, 217-222. Shih, W.-C. (1991), Ph. D. Thesis, University of Cambridge. Shinjo, T. (1987), in: Metallic Super lattices: Artificially Structured Materials: Shinjo, T., Takada, T. (Eds.). Amsterdam: Elsevier, pp. 107-149. Spiller, E. (1985), Proc. Soc. Photo-Opt. Instrum. Eng. 563, 367-375. Spiller, E. (1989), Proc. Soc. Photo-Opt. Instrum. Eng. 1160, 271-279. Spiller, E., Golub, L. (1989), Appl. Optics 28, 29692974. Spiller, E., Segmuller, A., Haelbich, R.-P. (1980), Ann. N. Y. Acad. Sci. 342, 188-198. Stobbs, W. M. (1988), Mater. Res. Soc. Symp. Proc. 103, 121-131. Suntola, T. (1990), Mater. Sci. Reports 4, 261-312. Thornton, J. A. (1977), Ann. Rev. Mater. Sci. 7, 239260.
Tsakalakos, T, Hilliard, J. E. (1983), /. Appl. Phys. 54, 134-137. Tsakalakos, T, Jankowski, A. F. (1986), Ann. Rev. Mater. Sci. 16, 293-313. Venables, J. A., Spiller, G. D. T, Hanbucken, M. (1984), Rep. Prog. Phys. 47, 399-459. Vettier, C , McWhan, D. B., Gyorgy, E. M., Kwo, J., Buntschuh, B. M., Batterman, B. W (1986), Phys. Rev. Lett. 56, 757-760. Vook, R. W (1988), Mater. Res. Soc. Symp. Proc. 103, 3-11. Wang, S. C , Ehrlich, G. (1988), Surf. Sci. 206, 451474. Werner, T. R., Banerjee, I., Yang, Q. S., Falco, C. M. Schuller I. K. (1982), Phys. Rev. B 26, 2224-2226. Willens, R. H., Kornblit, A., Testardi, L. R., Nakahara, S. (1982), Phys. Rev. B25, 290-296. Wilson, L., Bienenstock, A. (1988), Mater. Res. Soc. Symp. Proc. 103, 69-78. Wong, H. K., Wong, G. K., Ketterson, J. B. (1982), /. Appl. Phys. 53, 6834-6838. Wong, J. (1981), in: Glassy Metals I: Giintherodt, H. X, Beck, H. (Eds.). Berlin: Springer, pp. 45-77. Yasuoka, H. (1987), in: Metallic Superlattices: Artificially Structured Materials: Shinjo, T, Takada, T, (Eds.). Amsterdam: Elsevier, pp. 151-186.
General Reading Barbee, T W, Spaepen, R, Greer, A. L. (Eds.) (1988), "Multilayers: Synthesis, Properties and Non-Electronic Applications", Mater. Res. Soc. Symp. Proc. Vol. 103. (General References on Metallic Multilayers). Chang, L. L., Giessen, B. C. (Eds.) (1985), Synthetic Modulated Structures. Orlando: Academic Press. (Structure, Synthesis, Electronic, Magnetic and Superconducting Properties of, and Diffusion in, Metallic Multilayers). Hong, M., Wolf, S., Gubser, D. C. (Eds.) (1988), Metallic Multilayers and Epitaxy. Warrendale, PA: The Metallurgical Society. (General References on Metallic Multilayers). Jin, B. Y, Ketterson, J. B. (1989), Adv. Phys. 38, 189-366. (Physical Properties of Metallic Multilayers). Shinjo, T, Takada, T. (Eds.) (1987), Metallic Superlattices: Artificially Structured Materials. Amsterdam: Elsevier. (Characterization of Metallic Multilayers).
9 Recrystallization and Recovery John F. Humphreys Manchester Materials Science Centre, University of Manchester and UMIST, Manchester, U.K.
List of 9.1 9.2 9.2.1 9.2.2 9.3 9.3.1 9.3.2 9.3.3 9.3.3.1 9.3.3.2 9.3.4 9.3.5 9.4 9.4.1 9.4.1.1 9.4.1.2 9.4.2 9.4.3 9.4.4 9.4.5 9.4.5.1 9.4.5.2 9.4.5.3 9.4.5.4 9.5 9.5.1 9.5.1.1 9.5.1.2 9.5.2 9.5.3 9.5.3.1 9.5.3.2 9.5.4
Symbols and Abbreviations Introduction The Deformed State The Stored Energy of Cold Work The Deformed Microstructure Recovery Basic Mechanisms of Dislocation Recovery Subgrain Formation Mechanisms and Kinetics of Subgrain Growth Subboundary Migration Subgrain Rotation and Coalescence Extended Recovery Effect of Recovery on Mechanical Properties Recrystallization of Single-Phase Alloys Driving and Dragging Forces During Recrystallization Driving Force Boundary Curvature Kinetics of Primary Recrystallization Rate of Recrystallization Grain Size After Recrystallization Nucleation of Recrystallization Strain-Induced Grain Boundary Migration Preformed Nucleus Model Nucleation Sites The Role of Twinning Growth of Grains During Primary Recrystallization Experimental Observations The Effect of Temperature Effects of Orientation and Purity on Boundary Migration Effect of Solute on Grain Boundary Mobility Theories of Grain Boundary Migration Group-Process Theories Single-Process Theories Computer Modelling of Primary Recrystallization
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
373 374 376 376 376 377 378 380 381 382 383 386 386 387 388 388 388 388 389 390 391 391 392 392 393 394 394 395 395 395 397 397 397 398
372
9.6 9.6.1 9.6.2 9.6.2.1 9.6.2.2 9.6.2.3 9.6.3 9.7 9.7.1 9.7.2 9.8 9.8.1 9.8.2 9.8.3 9.8.3.1 9.8.3.2 9.8.3.3 9.8.4 9.8.5 9.8.5.1 9.8.5.2 9.8.6 9.8.7 9.8.8 9.9 9.9.1 9.9.1.1 9.9.1.2 9.9.1.3 9.9.1.4 9.9.2 9.9.3 9.10. 9.10.1 9.10.2 9.10.3 9.10.4 9.10.5 9.10.6 9.10.7 9.10.7.1 9.10.7.2 9.10.7.3 9.10.7.4 9.11
9 Recrystallization and Recovery
Grain Growth After Primary Recrystallization Factors Affecting Grain Growth Theories and Models of Grain Growth Statistical Models of Grain Growth The Modelling of Boundary Movement Atomistic Simulation of Grain Growth Secondary Recrystallization The Recrystallization of Ordered Alloys The Interaction of Ordering and Recovery The Effect of Ordering on Recrystallization Kinetics Recrystallization of Two-Phase Alloys Recrystallization Kinetics The Deformed Microstructure Particle-Stimulated Nucleation of Recrystallization (PSN) Mechanisms of Nucleation The Efficiency of PSN Factors Affecting PSN Pinning Effects of Particles (Zener Drag) Bimodal Alloys and Prediction of Grain Size Site-Saturated Nucleation Johnson-Mehl Kinetics Paniculate Composites Interaction of Precipitation and Recrystallization Grain Growth in Two-Phase Alloys Recrystallization Textures Orientation Effects in Nucleation Transition Bands Shear Bands Prior Grain Boundaries Second-Phase Particles Orientation Effects in Growth Competition Between Texture Components Annealing Processes During Hot Deformation The Flow Stress During Hot Working Dynamic Recovery The Characteristics of Dynamic Recrystallization The Conditions for Dynamic Recrystallization Microstructural Evolution During Dynamic Recrystallization Metadynamic Recrystallization Mechanisms of Dynamic Recrystallization Dynamic Recrystallization in Single Crystals Grain Boundary Nucleation Grain Boundary Impingement Dynamic Recrystallization by Lattice Rotation References
398 399 400 400 400 400 401 402 402 402 403 404 405 407 407 409 409 410 411 412 412 412 413 414 415 417 417 417 417 417 417 418 419 419 420 520 421 422 423 423 423 423 424 424 425
List of Symbols and Abbreviations
List of Symbols and Abbreviations b
e r r d D EB Ed Es EH G G kB M N P Pz Q QW.QB
r R S t
T
Burgers vector solute concentrations diameter of spherical particles grain or subgrain diameter grain boundary energy energy of low-angle boundary stored energy of the deformed metal energy of high-angle boundary volume fraction shear modulus grain growth rate Boltzmann constant boundary mobility nucleation rate force on a boundary particle pinning force activation energy activation energy for volume/boundary diffusion radius of particle radius of grain coefficient of self-diffusion time temperature ordering temperature
Q
shear strain, rotation angle true strain angle Poisson's ratio dislocation density
£G £s G
density of geometrically necessary dislocations density of statistically stored dislocations flow stress
HVEM ODF PSN TEM
high voltage electron microscope orientation distribution function particle-stimulated nucleation transmission electron micrograph
y 8
9 V
373
374
9 Recrystallization and Recovery
9.1 Introduction The free energy of a crystalline material is raised by the presence of dislocations, boundaries or interfaces. Therefore a material containing such defects is thermodynamically unstable. If such defects are introduced into the material by, for example, plastic deformation, or by irradiation, then the energy of the material is increased. Although thermodynamics would suggest that these defects should spontaneously disappear, in practice the kinetics of the diffusion processes by which these defects migrate are often very slow at low homologous temperatures, with the result that such thermodynamically unstable defect structures are retained after deformation. If the material is subsequently heated to a high temperature or annealed, then solid state diffusion provides a mechanism whereby the defects may be removed or alternatively arranged in lower energy configurations. This chapter will be concerned with such annealing processes in metals and alloys. The defects may be introduced in a variety of ways. However, we will confine ourselves to a consideration only of those introduced by plastic deformation. Although point defects as well as dislocations are introduced during deformation, the point defects anneal out at low temperatures and generally have little effect on the mechanical properties of the metal. In summary, then, we will be primarily concerned in this chapter with the annealing behavior of dislocations in deformed metals. The mechanical and physical properties of the metal are dependent on the dislocation content of the material, and therefore these properties are altered during plastic deformation. On annealing at an elevated temperature, these properties may be par-
tially restored to their original values by recovery, a process which involves annihilation and rearrangement of the dislocations. The changes in microstructure are shown schematically in Figure 9-1 a and b. Similar softening processes may also occur during the deformation, particularly at high temperatures, and this dynamic recovery plays an important role in the creep and hot working of materials. Recovery generally involves only a partial restoration of properties because the dislocation structure is not completely removed, but reaches a metastable state. There is a further annealing process called recrystallization in which new dislocationfree grains are formed within the deformed or recovered structure. These then grow and consume the old grains, resulting in a new grain structure with a low dislocation density. (Figure 9-1 c and d). Recrystallization may take place during deformation at elevated temperatures and this is termed dynamic recrystallization. Although recrystallization removes the dislocations, the material still contains grain boundaries, which are thermodynamically unstable. Further annealing may result in grain growth, in which smaller grains are eliminated, larger grains grow, and the grain boundaries assume a lower energy configuration. In certain circumstances this uniform grain growth may give way to the selective growth of a few large grains, a process known as abnormal grain growth or secondary recrystallization. The understanding and control of recrystallization is of particular importance in the mechanical processing of alloys. To a large extent the mechanical properties and behavior of a metal depend on the dislocation structure, the size of the grains and the orientation or texture of the grains. The first of these parameters is dependent on the amount of deformation,
9.1 Introduction
T
X
T|- h K
\
i
I
375
'"
Figure 9-1. Schematic sketch of recovery and recrystallization. a) Deformed micro structure, b) Recovered structure, c) Partial recrystallization. d) Fully recrystallized specimen.
the temperature of deformation and on the subsequent annealing treatment as well as other material and micro structural parameters. The grain size and texture are controlled by the recrystallization process. There are numerous examples of the need to control grain size. For example, a small grain size increases the strength of a steel and may also make it tougher (see Volume 7, Chapter 2). However, a large grain size is desirable in a high-temperature nickel-based superalloy in order to reduce creep rates. Superplastic forming, in which alloys are deformed to large strains at low stresses, is becoming an important technological process for the shaping of advanced materials (see Volume 6, Chap-
ter 9). As superplasticity is only found in materials with grain sizes less than about 10 jinn, great ingenuity must be exerted in producing and maintaining such a small grain size. The control of texture is vital for the successful cold forming of metals, a particularly important example being the deep drawing of aluminium or steel beverage cans (see Chapter 10, Sections 10.3.1 and 10.3.2). The use of annealing treatments to produce a texture which gives the required magnetic properties in Fe-Si alloys used for transformer cores is well established (see Chapter 10, Section 10.3.5).
376
9 Recrystallization and Recovery
9.2 The Deformed State The nature of the deformed material is a key element in determining the annealing behavior, because the stored energy of deformation provides the driving force for recovery and recrystallization, whilst the mechanisms of recrystallization are determined by the microstructure of the deformed material. An understanding of the recrystallization process is therefore dependent on a complete understanding of the deformed state. As the latter has not yet been achieved, it is not surprising that recrystallization cannot as yet be fully quantified. 9.2.1 The Stored Energy of Cold Work
Most of the work expended in deforming a metal is given out as heat with typically only some 1% being stored in the form of dislocations and, at low deformation temperatures, point defects. The amount of plastic strain in a metal is related to the numbers of mobile dislocations and the distances which they move. However, in a single phase single crystal, dislocation storage depends on the trapping of dislocations by others and this is not easy to predict from first principles - hence the difficulties in the formulation of good theories for work hardening. The density of such statistically stored dislocations (QS) is typically 1012 m~ 2 for a lightly deformed metal and 10 1 6 m~ 2 for a heavily cold rolled metal. If the metal is constrained to deform inhomogeneously by the presence of non-deformable second-phase particles, then the plastic incompatibility creates additional geometrically necessary dislocations (QG as shown by Ashby (1970)). The density of these dislocations depends on the strain and the particle size, shape and spacing. For example, for a volume frac-
tion Fy of spherical particles of diameter d, Ashby shows that QG is given by: QG
=
(9-1)
where b is the Burgers vector and y the shear strain. For large volume fractions or small particles, the predicted geometric dislocation density may be much larger than QS. However, this may in reality be significantly lowered by dynamic recovery, even at ambient temperatures. For similar reasons, the presence of high-angle grain boundaries leads to inhomogeneous deformation and hence to geometrically necessary dislocations (Ashby, 1970). The line energy of a dislocation per unit length, is given approximately by Gb2 where G is the shear modulus. Therefore the stored energy of the deformed metal (Es) is given approximately by: Es = Gb2 (QG -\-QS)
(9-2)
A highly deformed metal typically has a stored energy of about 105 J m ~ 3 . Note that this is very much smaller than the energies associated with phase transformations, e.g. only about 0.1 % of the latent heat of fusion of the metal. 9.2.2 The Deformed Microstructure
The annealing behavior of a deformed metal is dependent not only on the number of dislocations stored in the material, as discussed above, but also on the way in which the dislocations are distributed - the homogeneity of the microstructure. As we will discuss later, recrystallization originates in regions of the material where the microstructure is heterogeneous and thus a material with a very homogeneous dislocation distribution is unlikely to recrystallize during annealing as readily as one in which the same number of dislocations are inho-
9.3 Recovery
mogeneously distributed. The distribution of the recrystallized grains and their orientations (texture) depends on the nature of the site at which recrystallization starts, e.g., grain boundary, shear band, secondphase particle, etc. We shall look at the effects of the deformation micro structure on annealing behavior in more detail in later sections.
9.3 Recovery Before considering the details of recovery, it should be noted that recovery and recrystallization are competing processes, both driven by the stored energy of the deformed state. Once recrystallization has occurred and the deformation substructure consumed, then clearly no recovery can occur. The extent of recovery may therefore depend on the ease with which recrystallization occurs. Conversely, as recovery lowers the driving force for recrystallization, a significant amount of recovery may influence the kinetics of recrystallization. During recovery, many physical and mechanical properties change, and recovery may conveniently be measured indirectly by a variety of experimental techniques such as hardness or electrical resistivity. Measurements of the stored energy by calorimetric methods provides the most direct method of measuring recovery, and modern high-sensitivity differential scanning calorimeters are now extensively used (see Volume 2 A, Chapter 6). Because of the small amount of stored energy in a deformed material, calorimetric measurements of recovery and recrystallization can only be made on materials in which no phase transformations occur over the temperature range of the experiment. It is now thought that many early measurements of
377
stored energy are suspect because of inadequate characterization of the material. Figure 9-2 from the work of Schmidt and Haessner (1990), shows the annealing of high-purity aluminium which was deformed at 77 K and maintained at that temperature until the calorimetry was performed. The peak at 200 K is due to the
~ 1.0 c Z) >.O8 o ~ 0.6
•o
S.0.2
co 4 0 V)
*213K
•o 30
±db *? wo 7
6 160 | 140 "120 100 i/>
80
S 60 2 20-
-150
-120
-90
150
160
-60 210
-30 240
30 [ • c ] 6 0 270 300 330 Temperature [K]
Figure 9-2. Comparison of the differentiated isochronous resistance, hardness and flow (shear) stress, as a function of the annealing temperature of Al 99 999, deformed to a strain of 6.91 at 77 K (from Schmidt and Haessner, 1990).
378
9 Recrystallization and Recovery
recovery of point defects and the peak at 260 K is due to recrystallization. A separate peak due to dislocation recovery is not detected. The high purity aluminium recrystallizes at such a low temperature that either no dislocation recovery occurs prior to recrystallization or else the peak is too close to the recrystallization peak to be detected. The data from calorimetry are compared in Figure 9-2 with those from other techniques. Complete recovery can usually only occur in lightly deformed samples, although single crystals of hexagonal close-packed metals can be deformed to large strains on one slip system and in this case the original structure and properties can be completely restored by recovery. It is usually found that the greater the degree of deformation, the smaller is the fraction of the property change which can be recovered. This is largely due to the increasing ease of recrystallization at larger strains. Figure 9-3 shows annealing curves for deformed iron. It can be seen that at higher temperatures, recovery is more complete. The nature of the material itself also determines the extent of recovery. One of the most important parameters is the stacking fault energy, which, by affecting the extent to which dislocations dissociate, deter-
0
50
100
150
200 250 TIME (min)
300
mines the rate of dislocation climb, which is usually the rate controlling process during recovery. In metals of low stacking fault energy such as copper, oc-brass and austenitic stainless steel, climb is difficult, and little, or no recovery of the dislocation structure or of mechanical properties occurs prior to recrystallization. However, in metals of high stacking fault energy such as aluminium and oc-iron, climb is rapid, and significant recovery may occur as seen in Figure 9-3. 9.3.1 Basic Mechanisms of Dislocation Recovery During recovery the stored energy of the material is lowered by dislocation movement. There are two primary processes, these being the annihilation of dislocations and the rearrangement of dislocations into low energy configurations. Both processes are achieved by glide, climb and cross-slip of dislocations. A schematic drawing of a crystal containing an array of edge dislocations is shown in Figure 9-4. The elastic stress fields of the dislocations will exert forces on the dislocations (e.g. Hull and Bacon, 1984). The direction of these forces will depend on the Burgers vectors and relative
350
400
450
Figure 9-3. Isothermal recovery of iron after 5 % prestrain at 237 K (after Leslie et al., 1963).
9.3 Recovery
X T
A
X B T
T X
C JL
X T
D T
Figure 9-4. Schematic diagram of a crystal containing edge dislocations.
positions of the dislocations. For example, dislocations of opposite sign on the same glide plane, e.g. A and B, may annihilate by gliding towards each other. Such processes can occur even at low temperatures, lowering the dislocation density during deformation and leading to dynamic recovery. Dislocations of opposite Burgers vector on different but parallel glide planes, e.g., C and D, can annihilate by a combination of glide and climb. As climb requires thermal activation, this can only occur at high homologous temperatures. A similar configuration of screw dislocations would recover by annihilation of dislocations by cross-slip. This would occur at low temperatures in a material such as aluminium with a high stacking fault energy, but at high temperatures for a material of lower stacking fault energy. The crystal of Figure 9-4 contains dislocations of only one type of Burgers vector and it contains equal numbers of dislocations of the two signs. Thus complete recovery by annihilation is possible, as in the case discussed earlier of crystals deformed in easy glide. If unequal numbers of dislocations of the two signs are produced during deformation, as shown in Figure 9-5 a, then, the excess cannot be removed by annihilation (Figure 9-5 b). On annealing,
379
these excess dislocations will arrange into lower energy configurations in the form of regular arrays or low-angle grain boundaries. The simplest case is that shown in Figure 9-5 c in which dislocations of only one Burgers vector are involved. This process, first described by Cahn (1949) is often known aspolygonization, and the excess dislocations form tilt boundaries. If the vertical spacing of the dislocations (Burgers vector b) is h, then the parts of the crystal on either side of the boundary are misoriented by a small angle 9, where = b/h
(9-3)
The energy of such a boundary, E09 is given (Shockley and Read, 1950) as: = E09(A-\n6)
(9-4)
where E0 = Gb/4n(l - v ) and A=l + -\-ln(b/2nr0) and r 0 is the radius of the dislocation core, often taken as b. According to Equation (9-4), the energy of a tilt boundary increases with increasing misorientation (decreasing h) as shown in Figure 9-6. However, it can be seen from Figure 9-6 that the energy per dislocation decreases with increasing misorientation. Therefore there is a driving force to form fewer, more highly misoriented boundaries as recovery proceeds. In the more general case, dislocations of two or more Burgers vectors react to form two-dimensional networks whose character depends on the types of dislocation involved. For example a twist boundary in the form of a square network of dislocations may be formed by two sets of screw dislocations with orthogonal Burgers vectors. Details of the reactions involved in forming low-angle grain boundaries may be found in textbooks on dislocation theory (e.g., Hull and Bacon, 1984).
380
9 Recrystallization and Recovery
x
X
X
X
X
X
T
x
x
X T
x
X
X
x
X
X X
x
X
x
X
X
X
T
X
X
X
X
x
X
X
T
x
X
X
X
X
X
X
X
th
X
X
X
X
X
X
X
Figure 9-5. Recovery of a crystal containing unequal numbers of edge dislocations of the two signs, a) Deformed state, b) After annihilation of dislocations, c) Rearrangement of the remaining dislocations into low angle tilt boundaries. The misorientation of the lattice planes across the boundaries is not shown.
9.3.2 Subgrain Formation
10-
8-
o U-
A 6 8 Misorientation (degree)
10
Figure 9-6. The energy of a tilt boundary and the energy per dislocation, as a function of the angle of misorientation.
In the case of a polycrystalline material subjected to large strains the dislocation structures produced on deformation and on subsequent annealing are more complex than the simple case shown in Figure 9-5 because dislocations of many Burgers vectors are involved. In an alloy of medium or high stacking fault energy, the dislocations are typically arranged after deformation in the form of a three dimensional cell structure, the cell walls being complex dislocation tangles. The size and shape of the cells depends on the material and the strain. The transmission electron micrograph of Figure 9-7 a shows equiaxed cells of diameter about 1 jim in deformed aluminium. By anneal-
9.3 Recovery
a)
381
b)
Figure 9-7. Transmission high voltage electron micrographs (HVEM) of aluminium deformed 10 % and annealed in-situ. a) Deformed structure, b) Same area after 2 min anneal at 500 K.
ing this specimen in situ in a high voltage electron microscope (HVEM), the recovery processes can be followed directly. Figure 9-7 b shows the same area after annealing. The tangled cell walls, e.g., at A have become more regular dislocation networks or low angle grain boundaries and the number of dislocations in the cell interiors has diminished. The cells have now become subgrains, and at this stage there is little change in the scale of the structure. Dislocation rearrangements of this kind are not usually observed in metals of lower stacking fault energy such as copper or stainless steel, because, as discussed above, recrystallization occurs before significant recovery can occur. However, if recrystallization is inhibited by the presence of finely dispersed second-phase particles, then recovery will occur at a sufficiently high temperature, leading to the formation of
well defined subgrains (Humphreys and Martin, 1967). 9.3.3 Mechanisms and Kinetics of Subgrain Growth
The stored energy of a recovered substructure such as that of Figure 9-7 b is still large compared with that of the fully recrystallized material, and can be further lowered by coarsening of the substructure, which leads to a reduction in the total area of low-angle boundary in the material. In addition, as discussed above, the energy of the dislocations is lowered if boundaries of higher angle can be formed. As recovery processes play an important role in the nucleation of recrystallization, there has been particular interest in the mechanisms of recovery. Two different mechanisms by which coarsening of the
382
9 Recrystallization and Recovery
substructure can occur have been proposed - subboundary migration and subgrain rotation. 9.3.3.1 Subboundary Migration The subgrain structure shown in Figure 9-7 b is topologically rather similar to a foam, and the nodes of the structure will be subject to forces akin to surface tension. Thus, in the two-dimensional case shown schematically in Figure 9-8, the triple point A is subjected to forces from three boundaries, F±, F 2 and F 3 . If the specific energies of the boundaries are equal, then the triple point will be in equilibrium when the boundaries make angles of 120° with each other. The boundaries are therefore forced to become curved (dotted lines), and will tend to migrate in the direction of the arrows so as to minimize their length. Thus the node will migrate, and will become stable if it reaches a position (A') in which the boundaries are straight (dashed lines) and the angles are 180°. This process is very similar to the migration of high angle grain boundaries, and is discussed in more detail in Section 9.6. By means of this Y-junction migration, nodes will migrate, small subgrains will be consumed by larger ones, and the average subgrain size will increase. As the energy of low-angle
Figure 9-8. The migration of low-angle boundaries.
boundaries depends on the angle of misorientation of the adjacent subgrains (Equation (9-4)), the equilibrium angle will generally not be 120°. The velocity (v) at which a low angle grain boundary will move is given by: v = Mp
(9-5)
where p is the net driving force and M is the boundary mobility. Although simple tilt boundaries of the type shown in Figure 9-5 can migrate by glide, movement of low-angle boundaries in general requires both climb and glide of the dislocations and is thus a thermally activated process. The mobility of a lowangle grain boundary is therefore usually given by an expression of the form:
= M0Qxp-Q/kBT
(9-6)
where Mo is a constant dependent on the material and the character of the boundary and Q is the activation energy for boundary migration. For copper, Viswanathan and Bauer (1973) have shown that the activation energy is considerably higher for low-angle boundaries than for high-angle boundaries. The intrinsic mobility of a low-angle boundary is known to depend on the type and misorientation of the boundary. For symmetrical tilt boundaries the mobility is high and may occur at low temperatures for the reasons discussed above (Parker and Washburn, 1952). The mobility of other low angle boundaries is lower. Viswanathan and Bauer (1973) showed that for boundaries of misorientations larger than 2° in copper, the mobility increased with increasing misorientation, whilst the activation energy decreased. The mobilities of the low angle boundaries was several orders of magnitude less than for high angle boundaries in the same material. Qualitatively, these effects are ex-
9.3 Recovery
383
plained by the fact that boundary migration requires diffusion of atoms, which is easiest in disturbed regions of a crystal. A boundary with low misorientation has widely spaced dislocations and thus less disturbed area and a lower diffusivity than a boundary of greater misorientation. The kinetics of subgrain growth by boundary migration have been modelled by Li (1966) and by Sandstrom (1977), who points out the similarity to the growth of high angle grain boundaries. In Sandstrom's model, the boundary energy is held constant, and the resultant parabolic growth law is: D2 = D20 + Kt
(9-7)
where d0 is the initial subgrain diameter, K a temperature dependent constant and t the time. Dillamore et al. (1972) have used a similar form of Equation (9-7), which was originally derived for grain growth by Hillert (1965). There is some limited experimental support for such kinetics (Varma, 1986), although there is scope for the development of more accurate models which take into account both the variation of boundary energies and the subgrain size distributions in real microstructures and for more experimental evidence. 9.3.3.2 Subgrain Rotation and Coalescence
An alternative recovery mechanism was proposed by Li (1962). He suggested that subgrains might rotate by diffusional processes until adjacent subgrains were of similar orientation. The two subgrains involved would then have coalesced into one larger subgrain with little boundary migration, the driving force arising from a reduction in boundary energies. The process is shown schematically in Figure 9-9. Con-
Figure 9-9. Li's model for the rotation and coalescence of sub-boundaries.
sider two adjacent low angle boundaries, and let rotation occur in such a direction as to increase the misorientation of the higher angle boundary (AB) and reduce the misorientation of the lower angle boundary (BC). It can be seen from Equation (9-4) and Figure 9-6 that the change in boundary energy, dE9 jdQ is greater the lower the angle of the boundary. Therefore it is energetically favorable for rotation to occur in such a way as to decrease the misorientation of boundary BC and increase the misorientation of boundary AB, leading eventually to coalescence. The kinetics of rotation have been analyzed by Li (1962) who showed that the time for a coalescence event (t) is given by: D2kBT 3SEob3j
(9-8)
where E0 — Gb/4n(l — v) and j is the jog density.
384
9 Recrystallization and Recovery
Although it is clear that coalescence is thermodynamically feasible, there has been extensive debate as to whether it actually occurs. Comparison of the measured kinetics of subgrain growth with the predictions of the two theories, by Smith and Dillamore (1970), has suggested that subgrain rotation is too slow by several orders of magnitude to account for subgrain growth in iron. However, as the kinetics of neither model has been adequately verified, this result cannot be regarded as conclusive. Experimental evidence for coalescence has not been accepted as unambiguous (see, e.g., Doherty (1978)). There have been two types of experiments carried out - a) direct observation of the annealing of thin foils in the TEM (transmission electron microscope) or b) 'post-mortem' observation in the TEM of material annealed in the bulk and then thinned. a) In-Situ HVEM Observations This technique is suspect because events occurring in thin foils are not necessarily representative of bulk behavior. Most such experiments, particularly those in which the mechanisms of recrystallization were being investigated, have found unambiguous evidence of subgrain migration, e.g., Ray et al. (1975), Bay and Hansen (1979), Humphreys (1977), but no evidence of subgrain rotation. Clear evidence of subgrain rotation in deformed Al-6 wt. % Ni at very high temperatures was obtained by Chan and Humphreys (1984) using in-situ annealing in a HVEM. This alloy contains a dispersion of second phase particles which coarsen but which do not dissolve at high temperatures. On annealing, well defined subgrains are formed (Figure 9-10 a), and at lower temperatures of annealing these
were seen to grow by subboundary migration at a rate controlled by the coarsening of the particles (see below for a discussion of this phenomena). However, at temperatures in excess of 550 °C, subgrain reorientation was found to occur. Clear microscopic evidence of the change in the dislocation content of the subboundaries is seen in Figure 9-10 b at boundary X. In addition, measurements of the orientations of boundaries before and after reorientation clearly established that the contrast changes in the subgrains were due to changes in relative orientations and not to any artifacts such as bending of the thin foil. Although the experiment described above showed that rotation occurred, it should be emphasized firstly that this only occurred at very high temperatures, far above those at which subboundary migration occurs, and thus the relevance to normal recovery at relatively low temperatures is questionable. Secondly, it is quite possible that this is a thin foil effect; reorientation is easier for a two-dimensional subgrain since dislocations can escape easily to the free surfaces. b) Evidence from Bulk-Annealed Specimens The difficulty with this method is that only the structure before or after the event can be seen and the evidence is therefore indirect. Evidence of coalescence is usually a micrograph of a large subgrain containing a very low angle boundary. This latter boundary is assumed to be in the process of disappearing by coalescence. Several clear instances of such structures have been obtained, for example by Hu (1962), Faivre and Doherty (1979) and Jones et al. (1979). The problem with such observations is that, during deformation, sub-
9.3 Recovery
385
Figure 9-10. Rotation of subgrains during the in-situ annealing of a thin foil of Al-6 % Ni in a HYEM (Chan and Humphreys, 1984).
grains or cells are continually being formed and reformed, so that new subgrain boundaries will often form within old subgrains. Therefore micrographs of poorly developed low-angle boundaries within well defined subgrains are as likely to be showing a boundary forming as a boundary disappearing. As pointed out by Doherty (1974), the driving force for subgrain rotation is greatest near to high angle grain boundaries, and evidence of apparent coalescence events near to high-angle boundaries has been obtained by Faivre and Doherty (1979) and Jones et al. (1979). The latter authors have shown that the dislocation spacing in a low angle grain boundary increases close to a high angle grain boundary (Figure 9-11), and this is consistent with the coalescence model. In summary, most observations show that subgrain growth by boundary migra-
tion is an important mechanism. The importance of subgrain rotation as an annealing mechanism in bulk specimens has, in the opinion of the author, yet to be established.
02pm
D
Figure 9-11. Transmission electron micrograph of annealed aluminium showing that the spacing of dislocations in the low-angle boundary AB becomes larger as the boundary approaches the high-angle boundary CD (Jones et al., 1979).
386
9 Recrystallization and Recovery
9.3.4 Extended Recovery On annealing a cold worked metal, subgrain growth may occur as discussed above. In most cases, recrystallization will intervene before extensive subgrain growth can occur. However, if recrystallization is prevented, for example by the formation of precipitate particles on the substructure, then a more extensive homogeneous subgrain growth, whose kinetics are controlled by the coarsening of the particles, may occur. This was first studied in Al-Cu alloys by Hornbogen and colleagues (e.g., Koster and Hornbogen (1968), Hornbogen (1970)), and is shown schematically in Figure 9-12. The micrograph of Figure 910 a is an example of this process. If this process continues, then the scale of the substructure may become comparable
I XI I
r r rT
r
r
Figure 9-12. Extended recovery controlled by the coarsening of second-phase particles.
with that of a normal grain structure. A similar recovery process has been observed within certain texture components of rolled aluminium by Hjelen et al. (1990). This process has been referred to as continuous recrystallization or recrystallization in situ. However, there is little evidence that the resulting boundaries are predominantly high-angle boundaries, and as the process is clearly one of recovery which is uninterrupted by the onset of recrystallization, it has been suggested that the process should be termed extended recovery, and this more rational terminology is adopted here.
9.3.5 Effect of Recovery on Mechanical Properties As the dislocation density decreases and the substructure coarsens, the strength of the material is lowered, and an example of this for iron is seen in Figure 9-3. Further examples of the effect of recovery on strength may be found in the review by McElroy and Szkopiak (1972). The strength of the material depends primarily on two factors, the intrinsic strength (aY) which is the strength of the material without the dislocations introduced during deformation, and the substructure strength (crs). If we make the assumption, which is not necessarily justified, that these two factors are linearly additive, the flow stress of the material is given by: a = ox + as
(9-9)
As we have discussed above, the changes in substructure during annealing are complex, and therefore an evaluation of as is not straightforward. The problem is discussed in detail by McElroy and Szkopiak (1972) and Thompson (1977).
9.4 Recrystallization of Single Phase Alloys
For many deformed metals, cxs is found to be related to dislocation density (Q) by: <js = aGb^o
(9-10)
where a is a constant of the order of 0.5. This well known relationship follows from any work hardening theory and applies strictly to relatively uniform distributions of dislocations. The application to dislocations in networks and tangles is less clear. A deformed structure often consists of unbound dislocations and dislocation-tangled networks forming cell boundaries (Figure 9-7 a). During recovery, the free dislocations are removed and the cells recover to become subgrains. The parameter of greatest importance is the cell or subgrain size (d). It is generally found that as is proportional to D~m where m is a constant. Thus the strength of the material is given by: (9-11) where k is a constant. As the subgrain size increases during recovery, then the strength falls. However, as there is a transition from Equation (9-10) to Equation (9-11) during recovery, the application of such relationships may be problematical. The constant m is found to depend on the material and on the perfection of the subgrains. For cells, m is often 1/2, so that Equation (9-11) is analogous to the HallPetch equation for strengthening by highangle grain boundaries. The constant k is however smaller for subgrains than for grains. For well recovered subgrains, the constant m is often found to be closer to 1. An additional complication is that the strength of a subgrain wall is likely to depend on its angle of misorientation (9) and this may depend on the amount and type
387
of deformation and may not remain constant during recovery. It is clear from the above discussion that a decrease in strength is predicted as the dislocation density falls and the subgrain size increases during recovery. However, a full quantitative account of the change in strength during recovery is not yet available.
9.4 Recrystallization of SinglePhase Alloys Recovery is a relatively homogeneous process in terms of both space and time. When viewed on a scale which is larger than the subgrain size, most areas of a sample are changing in a similar way. Recovery progresses gradually with time and there is no readily identifiable beginning or end of the process. By contrast, recrystallization involves the formation of new strain free grains in certain parts of the specimen and the subsequent growth of these to consume the deformed or recovered microstructure (Figure 9-1 c, d). The microstructure at any time is divided into recrystallized or nonrecrystallized regions, and the specimen progresses from 0% to 100% recrystallization as a function of time. Recrystallization of the deformed microstructure is often called primary recrystallization in order to distinguish it from processes of exaggerated grain growth which may occur in fully recrystallized material and which are sometimes called secondary recrystallization (Section 9.6.3). The unqualified term "recrystallization" is always taken to mean primary recrystallization. It is convenient to divide recrystallization into two regimes, the nucleation of recrystallization which is the formation of
388
9 Recrystallization and Recovery
the new grains, and grain growth during recrystallization, which follows it. Although these two events occur consecutively for any particular new grain, throughout the specimen, nucleation and growth may be occurring all through the period of recrystallization. The kinetics of recrystallization are therefore superficially similar to those of a phase transformation which occurs by nucleation and growth. 9.4.1 Driving and Dragging Forces During Recrystallization
The boundary of a recrystallized grain growing into the deformed matrix is subjected to various forces {pt) either tending to promote or retard recrystallization. These forces act in a direction normal to the boundary plane. The net force on the boundary (p) is given by £ A and the boundary will move with a velocity (v) given by Equation (9-5). A detailed account of the forces acting on boundaries is given by Stiiwe (1978).
formed structure, then we must consider the opposing force which comes from the curvature of the high-angle grain boundary of specific energy EH. Grain boundary area will be reduced thus energy lowered if the grain were to shrink. There is thus a retarding force on the boundary given by: = 2EH/R
(9-13)
This force is only significant in the early stages of recrystallization, although it becomes very important for grain growth following recrystallization (Section 9.6) when the driving force for growth is low. For a grain boundary energy of 0.5 J m~ 2 and the driving force discussed above, we find that/? 2 is equal to px when R is about 1 jim. Below this grain size there would therefore be no net driving force for recrystallization. During the recrystallization of alloys, other retarding forces arising from solutes or second-phase particles may be important, and these will be discussed later. 9.4.2 Kinetics of Primary Recrystallization
9.4.1.1 Driving Force
The driving force for recrystallization is provided by the dislocation density (Q), which gives rise to a stored energy, qGb2, as was discussed in Section 9.2.1. The resultant driving force for recrystallization is then given approximately as: (9-12) 15
For a dislocation density of 10 m" 2 , typical of a heavily deformed metal, and with Gb2 typically 10 ~ 8 N , the driving force for primary recrystallization is around 10 MPa. 9.4.1.2 Boundary Curvature
If we consider a small spherical new grain of radius R, growing into the de-
The progress of recrystallization with time is commonly represented by a plot of the fraction of material recrystallized (X) as a function of time. This plot usually has a characteristic sigmoidal form which is shown schematically in Figure 9-13. This typically shows an incubation time before recrystallization is detected, followed by an increasing rate of recrystallization, a linear region, and finally a decreasing rate of recrystallization. This type of curve is typical of many transformation reactions (e.g., Shewmon, 1969), and can be described phenomenologically in terms of the constituent nucleation and growth processes. Nuclei are formed at a rate N=dN/dt and grains grow into the deformed material at a rate
9.4 Recrystallization of Single Phase Alloys
Impingement of growing grains
Log Time
389
with two variables. In particular, the inhomogeneity of the distribution of stored energy (Ryde et al., 1990) and of the nucleation events (Doherty etal., 1986) have been shown to cause significant deviations from Avrami kinetics. Vandermeer and Rath (1990) have recently developed an improved phenomenological model which it is claimed will provide information on the microstructural processes of recrystallization from an analysis of the kinetic data.
Figure 9-13. The effect of time on the fraction recrystallized.
9.4.3 Rate of Recrystallization G. The volume of spherical grains varies as the cube of their diameter, and the recrystallized fraction rises rapidly with time. However, the new grains will eventually impinge on each other and the rate of recrystallization will decrease, tending to zero as X approaches 1. Such an approach leads to a kinetic equation of the form: =
l-exp(-Btk)
(9-14)
which is generally known as the Avrami (Avrami, 1939) or Johnson-Mehl (Johnson and Mehl, 1939) equation. The constants B and k depend on the details of the model. For example they will depend on how the nucleation and growth rates vary during recrystallization, and on what are the shapes of the growing grains. Although such an equation gives a broad description of recrystallization, it is usually found that the constants cannot be clearly related to physical parameters and that the equation cannot be used to gain insight into the mechanisms and details of recrystallization. Careful experiments have shown that the equation is rarely accurately obeyed in practice. The reason for this is that the recrystallization process is far too complex to be described by an equation
Because recrystallization is a thermally activated process, the temperature of annealing has a profound influence on the rate of recrystallization. Measurements of the effect of temperature on the rates of nucleation and growth enable activation energies for these processes to be measured. However it is difficult to interpret these in terms of the atomistic processes involved. Figure 9-14 shows the effect of annealing temperature on the isothermal recrystallization kinetics of copper. For material which is isochronally annealed, it is often convenient to measure the ease of recrystallization in terms of a recrystallization temperature, the temperature at which recrystallization is complete (or 50 % complete) in a fixed time,, e.g. 1 hour. The rate of heating of the specimen to the annealing temperature is also important. During rapid heating, there is less recovery and the driving force for recrystallization is greater. Thus recrystallization occurs more rapidly. The amount, and to some extent the type of deformation, affect the rate of recrystallization, because the deformation alters both the amount of stored energy and the number of effective nuclei. There is a minimum amount of strain, generally a
390
9 Recrystallization and Recovery
FG CG
1.0 -—•-200°C --A-- -A- 225°C
g
—o- -•-250°C
73
o
-T-275°C
CD N
A /
/
f
/
/ /
f
S 0.5 0)
?
a: c o "o
o' / O/
D
0 -
> x ^ ^ y
1 10
/
_/
O
-*-
A A
C
/
/ /
*
/
i
/
I
/
/
-1
_J
7 °/ 1 100 Time [s]
a)
/
f
/ /
i.
I 1000
10000
Figure 9-14. The effect of temperature and inital grain size on the recrystallization kinetics of copper. The dotted curves show the behavior of material with a 15 (im grain size (FG) and the full curves show the behavior of material with a 50 urn grain size (CG). (Rydeetal., 1990).
few percent, below which recrystallization will not occur. Above this strain the rate of recrystallization increases, levelling out to a maximum value at true strains of around 2-4. The initial grain size may affect the stored energy of deformation as well as the number of nucleation sites (see Section 9.4.5), and it is found that a fine grained material will recrystallize more rapidly than a coarse grained material, as seen in Figure 9-14. Measurements of the overall recrystallization kinetics are often difficult to interpret because in many cases different regions of a specimen recrystallize at different rates because of differences in the dislocation structure and stored energy of different texture components (Dillamore etal., 1972; Ryde etal., 1990). Figure 915 a, from the work of the Trondheim group (Hjelen etal., 1990), clearly shows the different states of recrystallization in different texture components of high purity aluminium. 9.4.4 Grain Size After Recrystallization
Figure 9-15. Recrystallization of high purity aluminium after cold rolling, a) Partly recrystallized. b) Fully recrystallized (Hjelen et al., 1990).
The magnitude of the final grain size can be rationalized in terms of the effects of
9.4 Recrystallization of Single Phase Alloys
|
0.12
Initial grain size 0.5 mm
N 0.10
I 0.08 -o 0.06 N
Initial grain size 0.05 mm
o
0
20 40 60 Percent rolling deformation
80
Figure 9-16. The effect of deformation and initial grain size on the recrystallized grain size in oc-brass (Channon and Walker, 1953).
various parameters on the nucleation and growth processes. Any parameter favoring a larger number of nuclei or a rapid nucleation rate, such as a high strain or a small initial grain size will lead to a small final grain size. This is illustrated for oc-brass in Figure 9-16. The grain size may not be constant throughout a specimen. Just as the different texture components within a specimen recrystallize at different rates, as discussed above, so the final grain size and shape within each texture component may be different, as shown in Figure 9-15 b.
391
Although the term nucleation of recrystallization is universally used, it is known that recrystallization does not and indeed cannot occur by a true nucleation event such as occurs in phase transformations (see for example Doherty, 1978). Recrystallization starts by the growth of a small volume or crystallite which was present in the deformed microstructure. 9.4.5.1 Strain-Induced Grain Boundary Migration
This mechanism, an example of which is seen in Figure 9-17 was first reported by Beck and Sperry (1949). It involves the bulging of part of a pre-existing grain boundary leaving a dislocation-free region behind the migrating boundary, as shown schematically in Figure 9-18. The driving force for such an event arises from a difference in dislocation density on opposite sides of the grain boundary, and the process has been analyzed by Bailey and Hirsch (1964). The force opposing the migration is due to the radius of curvature of the boundary (Equation (9-13)). This mechanism is particularly important after low strains. The characteristic feature seen from the contrast in Figure
9.4.5 Nucleation of Recrystallization
A considerable research effort has been put into trying to understand the processes by which recrystallization originates. Although we now understand a great deal about the subject, our knowledge is still far from complete. The importance of recrystallization nucleation is that it is a critical factor in determining the size and orientation (texture) of the resulting grains. In order to control recrystallization effectively, we need to understand the mechanisms and the parameters which control it.
Figure 9-17. Strain induced grain boundary migration in aluminium deformed 40% by compression (Bellier and Doherty, 1977).
392
9 Recrystallization and Recovery
a)
b)
Figure 9-18. Schematic diagram of strain-induced grain boundary migration.
9-17 is that the new grains have similar orientations to the old grains from which they have grown. The origin of the differential dislocation density required for the operation this mechanism is not entirely clear. It could result directly from the deformation process, because it is known that the dislocation storage rate is dependent on grain orientation. However, it could also arise from preferential recovery in the vicinity of a grain boundary, increasing the subgrain size and hence lowering the dislocation density on one side of a region of boundary, and Jones et al. (1979) present evidence in support of such a process. 9.4.5.2 Preformed Nucleus Model In the process of strain induced grain boundary migration discussed above, the high angle grain boundary, which is a prerequisite for recrystallization was already in existence. However, in many cases, recrystallization originates in other regions of the material, and we need to consider how a nucleus is formed in such circumstances. It is now established beyond reasonable doubt that recrystallization originates from dislocation cells or subgrains which
are present after deformation. Although there are still uncertainties about how these pre-existing subgrains become nuclei, several points are now clear. - The orientation of the nucleus is present in the deformed structure. There is no evidence that new orientations are formed during or after nucleation, except by twinning. - Nucleation occurs by the growth of subgrains by the mechanisms discussed in Section 9.3.3. Although all direct TEM observations have shown the mechanism to be one of sub-boundary migration, subgrain coalescence may also occur. A characteristic of nucleation is that the recovery in that area is significantly faster than in the majority of the material. - In order for a high angle grain boundary to be produced by this rapid recovery, there must be an orientation gradient present. This is shown schematically in Figure 9-19 which shows two subgrain structures with similar misorientations between the individual subgrains. Recovery of the structure shown in Figure 9-19 a produces large subgrains but no high-angle boundary, whereas the same amount of recovery in the presence of an orientation gradient (Figure 9-19 c) results in the formation of a high-angle grain boundary (Figure 9-19 d). This point was first clearly stated by Dillamore et al. (1972). The site of the nucleus is very important in determining its viability, and it is clear that the general mechanism discussed above can occur at a variety of sites. 9.4.5.3 Nucleation Sites The grains of a polycrystal do not deform homogeneously, but tend to split into regions of different orientations. This is a consequence of the constraining effects of neighboring grains and of the different slip
9.4 Recrystallization of Single Phase Alloys
2
3
2
1
2
•j
2
1
2
4
5
6
8
0
1
2
3
1
L
3
9
10
12
d)
systems which operate in different grains (see, for instance, Dillamore and Katoh (1974)). Following the terminology used by Hatherly (1982) we note that the regions of different orientation are known as deformation bands. The size and number of the bands depends upon the material, the deformation conditions and the amount of strain, and an example is shown in Figure 9-20. The boundary of a deformation band is a small region of continuous orientation change which is called a transition band. In certain cases the orientation is identical on either side of a deformation band, and in that case it is called a kink band. The transition bands at the edge of the deformation bands are ideal sites for nucleation of recrystallization, having a large dislocation density and a large orientation gradient. Doherty and coworkers (see Doherty, 1978) have carried out extensive work on nucleation in such regions in aluminium and iron. Recent detailed studies of the orientation of subgrains in deformed aluminium (Hjelen et al., 1990) have shown that very large misorientations exist within a grain, as illustrated in Figure 9-21. These are fine scale deformation bands, which are potential nucleation sites for recrystallization (cf. Figure 9-19). During rolling, another form of deformation heterogeneity called a shear band is
13
15
15
16
393
Figure 9-19. Recovery in materials with orientation gradients. The numbers are the orientations of the subgrains in degrees, with respect to the left edge of the microstructure. a) Misorientations average out and there is no long-range gradient, b) Recovery of (a) leads to no high-angle boundary, c) There is a long range orientation gradient, d) Recovery of (c) leads to the formation of a high-angle boundary.
formed. This is a planar region oriented at around 30° to the rolling plane in which extensive shear has occurred locally. These bands are another favored site for nucleation, particularly in metals of medium and low stacking fault energy (Duggan etal., 1978). 9.4.5.4 The Role of Twinning
As discussed above, nucleation of recrystallization requires the formation of a high-angle grain boundary, capable of growth into the deformed material. There is evidence in materials of low stacking fault energy such as a-brass, that such
Figure 9-20. Fine scale deformation bands in Al5 % Mg deformed in compression to a strain of 0.9 (Drury and Humphreys, 1986).
394
9 Recrystallization and Recovery
AO'
n t
30 CD T> OIID
C
C CD
10^1171
k M \
20 .a \ 10 ' \ b n u
O U)
h
\c
1
1
/
A
I
A
( :
\
-10 -
\d
-20 -
6
f
y
I/. oU^
g
boundaries may be formed by a twinning process. Hatherly and coworkers (e.g. Huber and Hatherly, 1980) have shown that small twins form during recovery {recovery twins) in these materials, and that these may develop into recrystallization nuclei. Such a mechanism is fundamentally different from those discussed above because it generates new crystallographic orientations during the nucleation process. Haasen and coworkers (e.g. Berger et al., 1988) have found that the formation of multiple annealing twins often occurs at a very early stage in recrystallization, and suggest that this can be considered to be part of the nucleation process. Their insitu high-voltage TEM work has even shown such twinning to occur in metals of high stacking fault energy such as aluminium, although the possibility of this being an artefact of the experimental technique cannot yet be ruled out.
/
^
\\ •MM,I, V
\
\
Ste
P" SCQn =
Direction.ND
Figure 9-21. The orientations of subgrains in deformed and recovered aluminium measured from electron backscattering patterns. A step scan was carried out across a grain and the orientations are given relative to the Cu texture component (Hjelen et al., 1990).
ary, capable of migrating into the deformed material is produced. In order to complete the recrystallization process, the deformed material is consumed by these migrating boundaries, and we examine this process in this section. Despite the importance of grain boundary migration not only during primary recrystallization, but also during post-recrystallization grain growth (Section 9.6) and in other microstructural transformations, the details of this process are not well understood. This is partly because boundary migration involves atomistic processes occurring rapidly and at high temperatures. There is as yet, no single theory which is accepted as valid, and although there is a wealth of experimental evidence, in many cases no clear pattern emerges, although it is clear that boundary mobility depends very strongly on the purity of the material (see Babcock and Balluffi, 1989). 9.5.1 Experimental Observations
9.5 Growth of Grains During Primary Recrystallization In the nucleation stage of recrystallization discussed above, a high angle bound-
It is generally accepted that the velocity of a migrating boundary is given by Equation (9-5). If the driving force is known, then the boundary mobility can be determined. Experiments in which the velocity
9.5 Growth of Grains During Primary Recrystallization
of a boundary during recrystallization is measured are however very difficult to interpret because the driving force, arising from the stored energy, and typically 10 MPa, is not easy to measure accurately, varies throughout the microstructure and does not remain constant with time, decreasing as recovery proceeds. For this reason, many measurements of boundary mobility have been carried out on materials with a better characterized driving force, e.g., an as-cast substructure. Another commonly used technique is use the boundary energy itself as the driving force by using an undeformed bicrystal whose geometry is such that the boundary area is reduced as the boundary migrates. Further details of the available techniques may be found in the reviews by Haessner and Hofmann (1978) and Grant et al. (1984). In such experiments however the driving force is much lower (around 103 Pa). It is not clear to what extent measurements of boundary mobility in specimens with such low driving forces are directly applicable to migrating boundaries in recrystallizing material. In addition, there is evidence that boundaries interact with dislocations, and that defects affect boundary mobility, indicating that there may be a real physical difference between boundary migration in deformed and undeformed materials.
395
9.5.1.2 Effects of Orientation and Purity on Boundary Migration
As boundary migration involves diffusion processes in and across the boundary, it is not surprising that the structure of the boundary should affect its mobility. Boundaries may be classified in part by reference to the number of coincidence sites, which are atom sites common to the grains on either side of the boundary (see Volume 1, Chapter 9). Special boundaries with a large number of coincidence sites are known to have properties different from those with few coincident sites. Aust and Rutter (1959) measured the migration rate of boundaries in zone refined lead under a low driving force, as a function of both orientation and impurity level. Their results, shown in Figure 9-22, indicate that in very pure lead, the boundary mobility is almost independent of orientation. How-
r?
10.0 u
"Random" grain boundaries
9.5.1.1 The Effect of Temperature
As was the case for low angle boundaries (Section 9.3.3.1), the mobility of high-angle boundaries is very temperature dependent and is given by Equation (9-6). The activation energy for pure metals is typically about half that for self diffusion (see Haessner and Hofmann, 1978). However, the activation energy is higher for impure materials.
0.002 0.004 0.006 Weight percent of tin
Figure 9-22. The rate of grain boundary migration at 573 K into zone refined lead crystals doped with small amounts of tin (after Aust and Rutter, 1959).
396
9 Recrystallization and Recovery
ever, as the impurity level rises, the mobilities of the randomly oriented boundaries fall much more than those of the coincident site lattice boundaries, suggesting than impurities are adsorbed more easily on the more open and disordered structure of the random boundaries. Liebmann et al. (1956) found that during recrystallization of an aluminium single crystal, the boundary migration rate was highest for boundaries misoriented about a <111> axis by about 40° (Figure 9-23). Rapid growth of grains of certain orientations will lead to the development of a recrystallization texture, and the results above have often been cited in favor of the model of oriented growth of textures (see Section 9.9.2). 9.5.2 Effect of Solute on Grain Boundary Mobility As may be seen from Figure 9-22, small amounts of solute have a very large effect
on the mobility of boundaries. Lixcke and Detert (1957) were the first to formulate a quantitative theory of the effect of solute atoms on boundary mobility. This theory has been extended by others, and the developments are reviewed by Grant etal. (1984). The theory is based on the idea that atoms in the region of a boundary have a lower energy than those in the grain interior, because the open structure of a boundary allows more relaxation of the elastic misfit stresses associated with a solute atom. Therefore there is an attractive force between boundary and solute. As the boundary moves, the solute atmosphere is dragged with the boundary. The velocity of a boundary as a function of the driving pressure is shown schematically in Figure 9-24, for three different solute concentrations C1? C2 and C 3 . For low boundary velocities the solute atmosphere is dragged with the boundary, the boundary mobility being lower for larger solute content. At large boundary velocities, the boundary
— 3
o 2 '6 o -a c
Twin orientation
•D
U O 20 40 60 Orientation difference (degs) about common <111>
Figure 9-23. Growth rates of new grains into a deformed aluminium crystal at 888 K (after Liebmann and Liicke, 1956).
Driving Force
Figure 9-24. Schematic diagram showing the grain boundary velocity as a function of the driving force for different solute concentrations C3>C2>C1 according to the theory of Liicke and Detert (1957).
9.5 Growth of Grains During Primary Recrystallization
breaks away from its atmosphere, behaving like a boundary in the pure material. For large solute concentrations the change from a loaded to a free boundary results in a discontinuity in the curve. There is evidence that the theory is qualitatively in agreement with experiments. 9.5.3 Theories of Grain Boundary Migration
Theories of grain boundary migration need to take into account the structure of grain boundaries and the nature and kinetics of the atomistic processes involved in boundary migration. Comprehensive reviews of the structure of boundaries can be found in Haessner and Hofmann (1978) and Grant et al. (1984) and in Volume 1, Chapter 9 of this series. The structure of a grain boundary is a function of the misorientation between the grains. The coincidence site lattice model is concerned with certain special orientations in which there are a number of lattice points which are common to both of the adjacent grains. Such boundaries are defined in terms of Z, where 1/S is the fraction of atoms which occupy coincidence lattice sites. Small deviations from the exact orientation can be accommodated by imposing a network of intrinsic dislocations on the boundary. Such a theory describes the geometry of the boundary, but not the details of the atom arrangements in the boundary, which are determined by energetic considerations. Detailed calculations and computer simulations (see Haessner and Hofmann (1978) for details), have been developed to predict grain boundary structures. However, little is known about the structure of the moving boundaries, relevant to our present discussion, and in which equilibrium structures may not be present.
397
Theories of boundary migration are based on reaction rate theory in which atoms are continually detached from the grain and are able to move into the boundary. For a static boundary, the atom flux from the two adjacent grains is the same, but for a moving boundary they are different. The main differences in theoretical approach is as to whether migration occurs by single atom movement or by the collective movement of a group of atoms. 9.5.3.1 Group-Process Theories In the earliest group-process theory by Mott (1948), groups (islands) of atoms move from one grain into the boundary region and similar groups attach themselves to the other grain. Developments of the model took into account the variation of grain boundary energy with orientation. Although more recent work has concentrated on single-process theories, Haessner and Hofmann (1978) have suggested that the high mobilities of boundaries in pure materials indicate that some sort of cooperative processes, such as is envisaged in the group-process theories, may be occurring. 9.5.3.2 Single-Process Theories In the earliest single-process theories (Turnbull, 1951) it was assumed that every atom in the boundary could jump by thermal activation from one grain to the other, i.e. the boundary was narrow. Later theories have taken the boundary structure into account and considered the three steps - detachment of an atom from one grain, movement in the boundary region and attachment to the other grain. Gleiter (1969) proposed a detailed atomistic model in which boundary migration occurred by the movement of steps or kinks in the boundary in a similar manner
398
9 Recrystallization and Recovery
to that occurring during the growth of crystals from a vapor. Other models based on the movement of boundary dislocations (e.g. Smith and Rae, 1979) have also been proposed. The role of vacancies in grain boundary structure and mobility has been extensively discussed, and it is thought that boundary mobility rises with increasing "porosity" of the boundary. The vacancy concentration of a boundary is likely to be a function of the boundary velocity, and this again suggests that theories developed for static boundaries may not be entirely applicable to migrating boundaries. Accurate dynamic atomistic simulations of moving boundaries are likely to provide useful insight into many aspects of grain boundary migration. 9.5.4 Computer Modelling of Primary Recrystallization
There have recently been several attempts to simulate primary recrystallization by computer modelling (see e.g. Mahin et al. (1980); Saetre et al. (1986) Doherty et al. (1986) and Marthinsen et al. (1990)). The main advantage of such models is that they can allow for realistic spatial distribution of nuclei and for complex variations of nucleation and growth rates. Such models are capable of predicting grain size distributions as well as recrystallization kinetics. However, as pointed out by Marthinsen et al. (1990), the models are not yet sophisticated enough to give any detailed insight into the mechanisms of recrystallization.
9.6 Grain Growth After Primary Recrystallization When primary recrystallization, which is driven by the stored energy of cold work,
is complete, the structure is not yet stable, and further growth of the recrystallized grains may occur. The driving force for this is a reduction in the energy stored in the material in the form of grain boundaries. The driving pressure for grain growth is some two orders of magnitude less than that for primary recrystallization, and consequently, grain growth will be slower than during primary recrystallization and will be more affected by solutes and particles which pin grain boundaries. General reviews of grain growth have been undertaken by Higgins (1974), Doherty and Martin (1976) and Randle et al. (1986). Consider first a two-dimensional grain structure as shown in Figure 9-25. If the grain boundaries are assumed to have equal energies, then the triple points are in equilibrium when the boundaries make an angle of 120° with each other. This is possible for an array of regular hexagonal grains (Figure 9-25 a) and in this case we have achieved a metastable structure which will not coarsen. However, if we take a more realistic, less regular array of grains as shown in Figure 9-25 b, then the boundaries will become curved to achieve the required triple point angles. Curved boundaries will be unstable and tend to migrate in the direction of the arrows, so as to shorten their length. The result is that grains of less than 6 sides will tend to shrink and eventually disappear, whilst those of more than 6 sides will tend to grow, and the average grain size therefore increases with time. Extension of these ideas to a three dimensional grain structure leads to similar results except that a truly stable three dimensional array, equivalent to Figure 9-25 a, does not appear to exist. Burke (1949) showed that the velocity of a boundary during grain growth would be
9.6 Grain Growth After Primary Recrystallization
399
where the constant n has a maximum value of 0.5. However, most experimental measurements show n to be less than this (see Cotterill and Mould (1976), Haessner and Hofmann (1978) and Randle et al. (1986) for details). The highest values, typically 0.4 are found for pure metals. However, reliable experimental measurements on well characterized material are not easily obtained because of the low driving force and the strong effects of small amounts of impurities. The main problem of formulating a theory of grain growth is clear from a comparison of Figures 9-25 a and 9-25 b where it can be seen that the driving forces for grain growth are local and depend in detail on the distribution of grain sizes in the material, which in turn will depend on the history of the specimen. In a detailed study of the morphology of grains during growth, Rhines and Patterson (1982) showed that the distribution of grain sizes was determined by the strain before recrystallization, and that this distribution persisted throughout the subsequent grain growth.
Figure 9-25. Growth of a two-dimensional grain structure, a) An array of regular hexagonal grains is stable, b) In an irregular grain structure the boundaries are curved, c) Secondary recrystallization.
inversely proportional to the radius of curvature, so that: (9-15) from which it can be shown that: (9-16) Experimental measurements suggest that Equation 9-16 should really be written as: D2-D20
=
(9-17)
9.6.1 Factors Affecting Grain Growth
Grain growth is inhibited by a number of factors, and the effect of solutes on boundary migration has already been discussed in Section 9.5. The rate of grain growth diminishes when the grain size becomes greater than the thickness of a sheet specimen (Burke, 1949). This is because the grains are now curved only in one direction rather than two, and thus the driving force is diminished. Thermal etching grooves may also be formed on the sample surface by diffusion, and these will also impede grain growth.
400
9 Recrystallization and Recovery
Grain growth may also be affected by the presence of a sharp crystallographic texture (Beck and Sperry, 1949). This arises at least in part from a large number of grains of similar orientation leading to more low-angle and hence low-energy boundaries. Thus the driving force for growth is reduced. The texture may also alter during grain growth (Koppenaal et al., 1960; Abbruzzese and Liicke, 1986), thereby affecting the kinetics (see Chapter 10, Section 10.4.2). Undoubtedly the most important features affecting grain growth are second-phase particles, and their effect will be considered in Section 9.8.8. 9.6.2 Theories and Models of Grain Growth 9.6.2.1 Statistical Models of Grain Growth Topological theories of grain growth, in which a regular lattice of 6-sided grains is perturbed by "defects" such as a 7-sided and a 5-sided grain have been developed. These do not describe the behavior of individual grains, and allow only statistical averages of the behavior to be derived. However, because they are analytical, they incorporate physical models of the processes involved. Early theories were due to Hillert (1965) and by Cahn and Pedawar (1965), and these predict growth kinetics according to Equation (9-6). In a recent series of papers, Liicke, Abbruzzese and colleagues (see Abruzzese and Liicke 1986, Liicke et al., 1990) have extended this approach to allow the effects of texture, boundary energy and mobility to be taken into account. Because of the difficulty in formulating an analytical theory which will take realistic complex grain size distributions into effect, there has recently been much interest in computer modelling of grain growth and the two main approaches are de-
scribed in the following sections. For further details and references the reader is referred to the review by Anderson (1986). 9.6.2.2 The Modelling of Boundary Movement In these approaches, which have been carried out for two or three dimensional cases, an initial microstructure is assumed and the forces acting on the boundaries are calculated. The movement of the boundaries on the basis of the local topology and forces is then calculated and the boundary is moved appropriately. In this way, the development of microstructure is followed (e.g. Hunderi et al., 1979, Weaire and Kermode, 1983). 9.6.2.3 Atomistic Simulation of Grain Growth In a recent series of papers, Anderson and colleagues (e.g. Anderson et al., 1984; Anderson, 1986) have described Monte Carlo simulation of grain growth in two and three dimensions. The material is divided into a number of discrete points (at the centre of areas or volumes), each of which is given a number, corresponding to a grain orientation as shown in Figure 9-26. The grain boundary energy is then specified by the interaction between the grid sites. For example, a certain value of energy might be assigned to a 4-6 boundary and the same or a different one to a 3-9 boundary. The numbers of adjacent points are then swapped and the energy change (AE) measured. If the energy is lower, then there is a probability of exp( — AE/kBT) that the change will be accepted. Transitions occur at grain boundaries and the grains grow, showing most of the features of grain growth. If this technique is applied to three dimensions, the size of the array which can be used is
9.6 Grain Growth After Primary Recrystallization 4
4
4
4
4
4
,
9
9
9
9
9
9
401
Figure 9-26. The type of "microstructure" used for
several millimeters or larger. This process is known as secondary recrystallization, and an example is shown in Figure 9-27. Detailed discussions of this phenomenon were published by Detert (1978) and Cahn (1983). The rapidly growing grains may be larger than the average or they may have an orientation which is more favorable for growth. If the latter is the case then a pronounced texture may result on secondary recrystallization. The best known examples of this are the iron-silicon alloys used for transformer cores. In these a pronounced (110) [001] texture is produced on
atomistic simulation of grain growth. The integers denote orientations and the lines are grain boundaries (Anderson, 1986).
secondary recrystallization (see Chapter j Q S e c ti O n 10 3 5) J ' '
2
2
2
/
4
2 /4
4
S
4
4
5 / 4 5
k
4
5/8 8
8
1 8 8
4
4
4
4
4
4
8 8
4
4 4
4
4 ^4
4
4 4
4
I
8 ^ X 4 / 6 8
8
8
8
8
8
8
8
8
8
8
8
)
8 8
8
8 8 8
5
6
6
9 9
9
9
9
3
6 \3
3
6 \ 3 6
9
3
3
^
9 9
9^3"~~3
6
6
9 9
6 ^ 3
6
9
3
3
3
3 3
3
6\3
3 3
3
3
8*
8 8 8
9
8 / 6 6 6 6 8
9 9
9
6 6
9 9
9
9
6
4/6
8^ 8
9
4 / 6
4
8 8
9 9
4 \
4
9
9 4
9
4
6
8
9 4
8 8
/ Q 9
/ 4
4
4 4
8
8 8 8
4
4
8
1
4
4
8
4
4
4 8
8
4
4
4 4
5 W . 8
4
4
~ 5 | 4 S
4
6
6
8
8\6 \
6
6
6
(
6
6
6
8
b\A
6 \
6
6/7 6/7
/ 1
7 i \
7 7
7 7
7 7 7 7
7
7
limited to around 100 x 100 x 100 points. If the discrete points are equivalent to atoms (which are the unit which jump in a real material), then this means that the technique can be criticized for dealing only with very small numbers of very small grains. Liicke et al. (1990) have also pointed out that the method does not reveal general relationships or the physical meaning of the results, thus lacking the predictive power of the statistical theories. There continues to be great interest in this area of research which is just one example of the powerful molecular modelling approach now used in many branches of physics and chemistry. 9.6.3 Secondary Recrystallization In the previous section we discussed the growth of grains after recrystallization, and such growth is relatively uniform throughout the material. However, in certain circumstances, a few grains, such as the large grain in Figure 9-25 c may grow excessively, consuming the recrystallized grains. This may lead to a grain size of
"V " ' l ^ f c l ^'
Figure 9-27. Secondary recrystallization in Fe-3 % Si during an anneal at 1373 K (Detert, 1978).
The driving force for secondary recrystallization is the energy of the grain boundaries and therefore the process is more likely to occur when pinning of boundaries by second-phase particles restricts normal grain growth (Section 9.8.8). If, during a subsequent anneal, the particles coarsen or dissolve, then secondary recrystallization commonly occurs (Detert, 1978).
402
9 Recrystallization and Recovery
9.7 The Recrystallization of Ordered Alloys
280 in
As was discussed in Section 9.5, solute atoms have a significant effect on recrystallization. If the solid solution is ordered, rather than random, then there may be further effects on the recrystallization behavior. There is increasing interest in the use of ordered intermetallic alloys as materials for high temperature structural applications, and an understanding of their recrystallization behaviour is therefore of more than academic interest. The subject has recently been comprehensively reviewed by Cahn (1990). Although there is more commercial interest in alloys which are permanently ordered, solid solutions which are ordered at low temperatures, but which become disordered at higher temperatures have been studied for many years. There are differences in behavior between various alloy systems, but some general trends and effects are now established.
9.7.1 The Interaction of Ordering and Recovery
During deformation, an initially ordered alloy becomes partly disordered. For example, in Cu 3 Au, the long range order parameter was reduced from 0.85 to 0.5 during a cold roll of 60% (Roessler et al., 1963). During a subsequent anneal, the material re-orders as well as recrystallizing, and as may be seen from Figure 9-28, there is significant hardening {strain-agehardening) during the recovery stage. This is believed to be due in part to the formation of small antiphase domains during the reordering, and also due to the rapid work hardening rate as order develops in the material.
L_ (D
"E 260 • Ordered, worked A Disordered, worked • Disordered, unworked
Q_ O O 240 C in U)
cu c
T3 i_
a 150 o
si
\ 130 r Unannealed 0.1
1
10 Time [h]
100
1000
Figure 9-28. The microhardness of rolled Cu3Au as a function of annealing time (Roessler et al., 1963).
Deformation also has an effect on the kinetics of ordering. Figure 9-28 shows that for Cu3Au, initial ordering is more rapid in the deformed than in the undeformed material. This is thought to be due to the effect of dislocations on the ordering reaction. However, in FeCo, ordering is slowed by cold work (Smith and Rawlings, 1976). 9.7.2 The Effect of Ordering on Recrystallization Kinetics
An ordered solid solution is found to recrystallize more slowly than a similar but disordered alloy. This was clearly demonstrated by Hutchinson etal. (1973), as shown in Figure 9-29. The curves have the normal sigmoidal shape both above and below the ordering temperature Tc (400 °C), but below Tc the rate of recrystallization is very much slower. In these specimens re-ordering of the deformed material is complete before recrystallization commences. These authors showed that the slow recrystallization was primarily
9.8 Recrystallization of Two-Phase Alloys
Figure 9-29. Recrystallization kinetics of rolled Cu3Au, showing a marked increase in recrystallization rate above Tc (400 °C) (Hutchinson et al., 1973).
0 -
102
10
103 Time Is]
10*
so 100 \
i\
1j 10 1.0 800
92
Temperature. °C 775 750 725 700
96
100
10*
region, and therefore of higher relative energy than in a disordered material. There is also evidence (see Cahn, 1990) that a narrow layer adjacent to the boundary may be disordered, and that there may be solute segregation in this region. This would lead to similar dragging effects on a boundary to those discussed in Section 9.5.2. Cahn (1990) has suggested that the occurrence of equiaxed antiphase domain boundaries following recrystallization of an ordered alloy is evidence of the existence of such a disordered layer adjacent to the boundary.
1000 -
0.1
403
675 1(K
1/T(K)x10 5
Figure 9-30. Plot of the grain growth rate constant against temperature in Fe-Co-V (Davies and Stoloff, 1966).
due to retardation of grain boundary mobility. A similar effect is found in FeCo-V as shown in Figure 9-30, in which a 10-fold increase in the rate constant occurs at Tc. It is not surprising that a grain boundary is less mobile when the adjacent grains are ordered. The boundary is a disordered
9.8 Recrystallization of Two-Phase Alloys As most alloys of commercial importance are multiphase, an understanding of the recrystallization behavior of such materials is of practical as well as of scientific interest. The second phase may be in the form of dispersed particles, which are present during the deformation, or the particles may form during the subsequent anneal. There are also alloys in which the volume fractions of the two phases are
404
9 Recrystallization and Recovery
similar - duplex alloys. In this section we will be primarily concerned with alloys containing stable dispersions of particles. Particles have three important effects on recrystallization: 1. Particles may increase the stored energy and hence the driving force for recrystallization. 2. Large particles may act as nucleation sites for recrystallization. 3. Particles, particularly if closely spaced, may exert a significant pinning effect on grain boundaries. The first two effects tend to promote recrystallization, whereas the last effect tends to prevent recrystallization. Thus the recrystallization behavior, particularly the kinetics and the resulting grain size, will depend on which of these effects dominate. The final grain size tends to be small when recrystallization is accelerated and coarse when recrystallization is retarded. 9.8.1 Recrystallization Kinetics
The recrystallization kinetics strongly depend on both the particle size and the interparticle spacing, as was first clearly demonstrated by Doherty and Martin (1962). It is difficult to separate the effects of these two parameters because there are few investigations in which they have been independently varied. Nevertheless it is clear that by comparison with a single phase alloy, recrystallization is retarded or even completely inhibited by closely spaced particles, and is accelerated by widely spaced particles, as shown in Figure 9-31 (see, e.g., Doherty and Martin (1976), Humphreys (1979 a) for further discussion). The strong effect of particle size on the recrystallization kinetics, for a constant (large) interparticle spacing is shown in Figure 9-32.
Time for 50% Recrystallization (s) 10000000 1000000 100000 10000
single-phase alloy
1000
•H-+
100, 10, 0
0.5
1 1.5 2 2.5 Interparticle Spacing (^m)
3
3.5
Figure 9-31. The effect of interparticle spacing on the time for 50% recrystallization in Al-Cu single crystals. (Data from Doherty and Martin, 1964).
Time for 50% RecrystaUization [s]
10000
1000
2 3 Particle Diameter [jiml
Figure 9-32. The effect of particle size on the time for 50% recrystallization in Al-Si crystals with a large interparticle spacing. (Data from Humphreys, 1977).
The effects of both particle size and interparticle spacing on the kinetics are shown schematically in Figure 9-33. Curve A, in which only retardation is found, is typical of aluminium alloys containing particles which are below the critical size for particle stimulated nucleation (Section 9.8.3). Curve B is typical of copper alloys with small particles. Curve C shows the behavior of alloys with large particles. The reason why small particles appear to accelerate recrystallization in copper but not in aluminium may be a result of the lower recovery rate and hence larger driving force for recrystallization in copper alloys. Thus in aluminium, many of the geometri-
9.8 Recrystallization of Two-Phase Alloys
Interparticle Spacing
Figure 9-33. Schematic diagram showing the effects of both particle size and spacing on the recrystallization kinetics. See text for details.
cally necessary dislocations produced during deformation (Section 9.2.1) may be dynamically recovered, so that the stored energy of the 2-phase alloys is not significantly greater than that of single-phase alloys. 9.8.2 The Deformed Microstructure
Particles have a large effect on the microstructure developed during deformation, and this in turn affects the recrystallization behavior. If the particles deform during the deformation, then they do not significantly alter the dislocation density compared to a single phase alloy. However, they cause the slip to be inhomogeneous, and this may affect the subsequent recrystallization behavior as demonstrated by Kamma and Hornbogen (1976). These authors showed that at lower strains the inhomogeneous slip accelerated recrystallization, but at larger strains, the dispersion was refined by repeated particle cutting, leading to retarded recrystallization. If the particles do not deform with the matrix, then geometrically necessary dislocations are generated at the particles (Section 9.2.2). The form and distribution of these dislocations is primarily a function of strain and particle size, although other fac-
405
tors such as shape, interface strength and matrix are known to be important (see for instance Humphreys, 1985). The dislocations which accumulate at the particles are a result of the incompatibility between the plastically deforming matrix and the non-deforming particle, and the nature of the dislocation structures depends on how the resulting stresses are relieved by plastic relaxation. For small particles and low strains, plastic relaxation generally involves the generation of prismatic loops. However, for particles of diameter greater than around 0.1 jim, stress relaxation may also occur by local lattice rotation. The effect of strain and particle size on the relaxation mechanisms in aluminium single crystals is summarized in Figure 9-34. For small particles, at which prismatic loops are formed, the important effect on the microstructure is the increased dislocation density which was discussed in Section
Primary Prismatic Loops
Particles deform
0.2 0.3 Shear strain
(U
Figure 9-34. Deformation mechanisms at particles in aluminium as a function of shear strain and particle radius normalized with respect to dislocation Burgers vector (after Humphreys, 1979 b).
406
9 Recrystallization and Recovery
9.2.2. Measurements of stored energy in particle-containing copper alloys (Baker and Martin, 1983) show an increase in stored energy with strain which is in approximate agreement with Equations (9-1) and (9-2). The effect of particles on the homogeneity of the deformation, for example on the formation of deformation or shear bands is however less clear (see Humphreys, 1985). At particles larger than around 0.1 |im, Figure 9-34 shows that plastic relaxation results in the formation of rotated structures or deformation zones near the particles. Detailed study of the zones in aluminium single crystals deforming by single slip (Humphreys, 1979 b) has shown that for large particles, the maximum misorientation occurs at the particle surface, and is equal to t a n - 1 y , where y is the shear strain. The axis of rotation is perpendicular to the primary Burgers vector and to the slip plane normal. The misorientation falls rapidly with distance from the particle, and the size of the deformation zone is related to the particle size. A simple model of plastic relaxation by lattice rotation, proposed by Humphreys and Kalu (1990a) is shown in Figure 9-35. Figure 9-35 c shows how rotation of the particle and the adjacent matrix, together with the generation of secondary dislocations (not shown), allow the relaxation of elastic
i
ill
1 a)
b)
c)
d)
Figure 9-35. Schematic diagram of plastic relaxation at a large particle, a) Undeformed. b) Unrelaxed plastic deformation, c) Relaxation by rotation and generation of secondary dislocations (not shown), d) Formation of impenetrable zone (dotted) at larger strains (Humphreys and Kalu, 1990 a).
stresses at the particle. On further straining, the relaxation debris forms an impenetrable zone, preventing glide dislocations from approaching close to the particle (Figure 9-35 d), and further relaxation results in both the particle and the impenetrable region rotating. Such a mechanism leads to a maximum rotation angle of y and a decrease in misorientation with increasing distance from the particle which is in accord with that found experimentally (Humphreys, 1979 b). Although the above model is applicable only to crystals deforming by single slip, Humphreys and Kalu (1990 a) have extended it to polycrystals deformed in compression. In this model, based on the Taylor theory of polycrystalline plasticity (see Volume 6, Chapter 3), several deformation zones, resulting from activity on the various slip systems, may be formed at a single particle, although significant rotations are not predicted in more than 3 or 4 such zones. Figure 9-36 shows the maximum misorientations in different zones ( P I PS) predicted for particles in a grain of initial orientation I and final orientation M, for a grain in a polycrystal. Experimental measurements of the orientations of deformation zones in compressed Al-Si polycrystals are consistent with this type of model. Although some relaxation by local lattice rotation does occur at particles as small as 0.1 Jim, other relaxation mechanisms also operate, and in aluminium, lattice rotation does not become the dominant relaxation mechanism until the particles are larger than around 2-3 \im (Humphreys, 1979 b). For particles in this intermediate size range the lattice rotations are a function of both strain and particle size, as may be seen from Figure 9-34. Although most work has been carried out on equiaxed particles, there is evidence
9.8 Recrystailization of Two-Phase Alloys
Figure 9-36. Predicted orientations (P1? P 2 and P3) of the deformation zones near particles in a grain of initial orientation I in a compressed f.c.c. polycrystal (Humphreys and Kalu, 1990 a).
(Herbst and Huber, 1978; Humphreys, 1979 b) that lattice rotations at the ends of elongated particles are particularly large. 9.8.3 Particle-Stimulated Nucleation of Recrystallization (PSN) Nucleation of recrystallization may occur within the deformation zones discussed above, and this is commonly referred to as particle-stimulated nucleation (PSN). 9.8.3.1 Mechanisms of Nucleation Although specimens annealed to produce recrystallization nuclei and which are subsequently examined metallographically, can provide information about the kinetics of recrystallization and the orientation of the new grains, there is little that can be deduced from these specimens about the mechanisms of nucleation, and most direct evidence has come from in-situ observations in the HVEM. From a study of recrystallization at particles in rolled
407
aluminium, Humphreys (1977) concluded that: - Recrystallization originates at a preexisting subgrain within the deformation zone, but not necessarily at the particle surface. - Nucleation occurs by rapid subboundary migration. - The grain may stop growing when the deformation zone is consumed. Later in-situ work (Bay and Hansen, 1979), together with work on bulk annealed specimens (Herbst and Huber, 1978) has supported these general conclusions. A sequence of electron micrographs from an in-situ annealing sequence, Figure 9-37, shows the occurrence of particlestimulated nucleation in aluminium. The kinetics of nucleation, as determined by in-situ HVEM annealing of A l Si are illustrated in Figure 9-38. The rapid annealing of the deformation zone, shown in Figure 9-38 a is due to the high dislocation density and small subgrain size, compared to the matrix. The drop in the maximum misorientation within a deformation zone shown in Figure 9-38 b is consistent with the fact that the nucleus may not originate in the region of highest misorientation at the interface, but elsewhere in the deformation zone. There is no evidence that PSN occurs by a mechanism different from nucleation of recrystallization at heterogeneities in single-phase materials, and the models of PSN are consistent with the earlier model of recrystallization at transition bands proposed by Dillamore et al. (1972). Although the nucleus usually is misoriented by at least 10 ° from the matrix, it has been suggested (Humphreys, 1979 a) that an alternative annealing process might involve the growth of a matrix subgrain into the deformation zone, thus producing a
408
9 Recrystallization and Recovery
Figure 9-37. In-situ HVEM annealing of Al-Si. Recrystallization originated in the deformation zone near the particle (arrowed).
10 Minutes
(a) (b) Figure 9-38. Changes in subgrain size and misorientation within the deformation zone of Si particles in Al, as determined by in situ annealing, a) Growth kinetics, b) The maximum misorientation within the deformation zone (Humphreys, 1980).
9.8 Recrystallization of Two-Phase Alloys
nucleus which was not highly misoriented. 0rsund and Nes (1988) have found evidence of such a process. In deformed AlMn alloys, they found that although PSN occurred after annealing at all temperatures, at high annealing temperatures the resultant texture was consistent with nuclei growing from the core of the deformation zone, whereas after low temperature annealing, the texture was consistent with nuclei originating in the outer regions of the deformation zones, which are only slightly misoriented from the matrix. 9.8.3.2 The Efficiency of PSN An important application of PSN is in controlling the grain size of recrystallized alloys, and in particular, the production of fine grained material. If each particle nucleates one grain, then the resultant grain size will be directly related to the number of particles per unit volume. If several grains nucleate at a particle, then an efficiency (grains/particles) of greater than 1 may occur. However, as multiple nucleation generally only occurs for particles larger than 5-10 jam, this is rarely achieved. If the particle size is close to the minimum, then the nucleation efficiency is very low, and the expected fine grain size is not achieved (Wert et al., 1981). As nuclei are essentially in competition with each other, we only obtain a high efficiency if all nucleation events occur simultaneously (site saturation), and if they grow at similar rates. Although this condition may be met for alloys with widely spaced particles (e.g., Humphreys, 1977), if the particles are closely spaced, or if growth of nuclei is affected by for example a fine dispersion of particles, then the nucleation efficiency may decrease markedly. It is possible that particles in different situations, e.g., at grain bound-
409
aries, shear bands, etc., may nucleate at different rates. However, there is as yet little direct evidence of this effect. The main parameters known to affect PSN are discussed in the following section. 9.8.3.3 Factors Affecting PSN Strain and Particle Size The main parameters which determine whether or not PSN occurs are the strain and the particle size. This is clearly seen in Figure 9-39 for rolled aluminium containing Si particles. Two criteria must be fulfilled for growth of a nucleus beyond the deformation zone (Humphreys, 1977). Firstly, a deformation zone with sufficient rnisorientation to create a high angle boundary must be created on deformation. The conditions for this are seen in Figure 9-39, and comparison with Figure 9-34 shows that this is a necessary but not a sufficient criterion for nucleation, because the nucleus must also be able to grow into the surrounding matrix which has a stored energy Es. On an energy balance, Humphreys showed that this condition approximated to: d =4EH/Es = 4EH/p1 100
> o
80
>
0
O
i
I
#
o \
60 > o
#* »# v ••
o
o
(9-18)
Nucleation at Particles
oo
*
20 n u
0
2
U 6 8 Particle Diameter [|im]
10
Figure 9-39. The effect of rolling reduction and particle size on the occurrence of particle-stimulated nucleation (Humphreys, 1977).
410
9 Recrystallization and Recovery
where dc is the critical particle diameter and En is the grain boundary energy. It is difficult to accurately predict values of Es in deformed alloys. However, if reasonable values are taken (Humphreys, 1977) or if experimental subgrain data are used, then this equation gives agreement with the results shown in Figure 9-39, showing that the growth criterion is the critical one. The Effect of Particle Distribution There is evidence that nucleation occurs preferentially at pairs or groups of particles, even if the individual particles are below the critical size for nucleation. This has been detected statistically (Gawne and Higgins, 1971), from metallographic observations (Herbst and Huber, 1978) and from in-situ annealing (Bay and Hansen, 1979). A recent detailed study of the distribution of recrystallization nuclei in Al-Si specimens containing particles of diameters close to the critical size of Figure 9-39 (Koken etal., 1988) has also shown that under these conditions, nucleation is favored in sites of particle clustering. If the spacing of the large particles becomes small, which will happen with large volume fractions, then the recrystallization may be inhibited by particle pinning effects. As this effect is of importance in the recrystallization of metal-matrix composites, we will discuss this aspect of PSN later. The Effect of Deformation Temperature If the temperature of deformation is raised, then PSN may become less viable. We need to consider the effect of deformation temperature on the two criteria for PSN - the formation of deformation zones and growth of the nucleus beyond the particle.
At high temperatures, dislocations may be able to bypass particles without forming deformation zones. Humphreys and Kalu (1987) have shown that critical strain rate for the formation of a deformation zone is given by: + K2exp(-QB/kBT)/Td3
(9-19)
where K± and K2 are derived constants and Qy and QB are the activation energies for volume and boundary diffusion. For large particles the second term will generally be negligible. For a constant strain rate, the critical particle size for zone formation thus increases with temperature. The growth criterion (Equation (9-18)) is also affected by temperature, because the stored energy (E) is reduced at elevated temperatures, although this effect is more difficult to assess (Kalu and Humphreys, 1986). An investigation of aluminium alloy AA3004 by Oscarsson et al. (1987), in which E was calculated from measurements of subgrains suggests that in the temperature range at which the alloy is hot-rolled, the growth criterion is the critical condition for PSN. 9.8.4 Pinning Effects of Particles (Zener Drag) A dispersion of particles will exert a retarding force, on a grain boundary. The effect is known as Zener drag after the original analysis by Zener which was published by Smith (1948). When a grain boundary, of specific energy EH, migrates onto a spherical incoherent particle, when it can be assumed that the interfacial energy of the particle is unchanged by the passage of the grain boundary, then the particle, of radius r, effectively removes an area of the bound-
9.8 Recrystallization of Two-Phase Alloys
ary equal to 71 r2, and the energy is reduced
by AE: AE=nr2
(9-20)
To move the boundary past the particle a force p must be applied and the maximum value of p is: p = nrEH
(9-21)
If the boundary is planar, and randomly intersects particles, then the number of particles (TV) per unit area of boundary is:
N=3FJ2nr2
(9-22)
The pinning force exerted on the boundary is then given by: = 3FvEH/2r
(9-23)
The Zener drag force has been examined by several authors, and the reader is referred to the reviews by Nes et al. (1985), Hillert (1988) and Doherty et al. (1989) for further details. These authors conclude that the more sophisticated calculations do not lead to relationships which differ significantly from Equation (9-23). If the particles are initially coherent, then they will lose choherence during the passage of a grain boundary, and therefore the energy will be significantly higher than beforehand. It is therefore clear that coherent particles will be more effective in pinning boundaries than will incoherent particles. The increase in energy may be sufficiently great to induce the particle either to rotate into a coherent orientation in the new grain, or to dissolve in the boundary and to re-precipitate in a coherent orientation (Grant etal., 1984; Randle etal., 1986; Ringer etal., 1989). The particle shape, if not spherical, will have some effect on the pinning force, and non-uniform particle distributions, particularly planar arrays of inclusions will make grain boundary mobility anisotropic
411
in the material (Nes et al, 1985; Ringer etal., 1989). In certain circumstances, the force of the boundary on the particle may actually drag the particle with the boundary (see Ashby, 1980). However, this effect, which may be controlled by diffusion within the matrix, particle or interface, is unusual, and is only likely to occcur for low volume fractions of small particles at high temperatures. Although Equation (9-23) has been applied extensively to grain growth in fully recrystallized material, because of the difficulty of determining the driving force accurately, it has not been properly verified for primary recrystallization. Nevertheless, particle pinning is undoubtedly the cause of retarded recrystallization in alloys containing closely spaced small particles, and plays a dominant role in grain growth after recrystallization (Section 9.8.8). 9.8.5 Bimodal Alloys and the Prediction of Grain Size Many commercial alloys contain distributions of both large (> 1 jim) particles which will act as nucleation sites and small particles which will pin the migrating boundaries. In such a situation, the driving force is offset by the Zener pinning force and the critical particle size for nucleation (Equation (9-18)) now becomes: Pl-3FvEH/2r
Thus, as the Zener pinning force increases, the critical particle diameter for PSN increases. As there will be a distribution of particle sizes in a real alloy, this means that less particles are able to act as nuclei, and that the recrystallized grain size will increase.
412
9 Recrystallization and Recovery
The number of particles capable of acting as nuclei (TV) is the number of particles of diameter greater than dc. The grain size, D N , will be given approximately by: DN = N-''3
Nucleation limit
(9-25)
In an alloy containing large particles, then, if other nucleation sites are neglected, TV is the number of particles larger than Fv/r
There is considerable interest in being able to predict the grain size of particlecontaining alloys as a function of the particle parameters and the thermomechanical processing route, and Nes (1976,1986) and Wert (Wert and Austin, 1988) have developed models, based on the mechanisms discussed above for recrystallization in commercially important aluminium alloys. There are two important situations which may be considered, site-saturated nucleation in which all nucleation events occur at the start of the anneal, and JohnsonMehl kinetics, in which the nucleation rate is low, and constant with time. 9.8.5.1 Site-Saturated Nucleation
If all nuclei are formed at the same time, and grow at the same rate, then the final grain size depends only on the number of viable nuclei (TV), as given by Equation (9-25), assuming that the nuclei are evenly distributed. This is likely to be the situation when the alloy contains a large fraction of particles of diameter greater than dc.
Figure 9-40. Schematic diagram showing the effect of the particle dispersion (FJr) on the recrystallized grain size. As FJr increases, both the limiting grain size and the number of viable nucleation sites are reduced.
At high temperatures, as discussed in Section 9.6, grain growth may occur after recrystallization is complete, and the limiting grain size, DG, is then given by Equation (9-27). The recrystallized grain size as a function of the volume fraction and size of the small particles is then as shown schematically in Figure 9-40. At small values of Fv/r the grain size is determined by growth (Equation (9-27)), but at large values of Fy/r, it is determined by nucleation (Equation (9-26)). In reality, the situation is more complicated than this. Nevertheless, from a combination of fundamental theory and semi-empirical parameters, useful models, capable of predicting grain sizes in commercial alloys undergoing realistic thermomechanical processing schedules have been developed.
9.8.5.2 Johnson-Mehl Kinetics If there a limited number (TV) of viable nucleation sites in the material, and if the nucleation rate is TV, then the average grain size is given (Johnson and Mehl, 1939) as: (9-26) where G is the growth rate.
9.8.6 Particulate Composites
Particulate metal matrix composites, such as aluminium alloys containing around 20 vol. % of SiC particles are of increasing interest in applications such as automotive and aerospace ones, where a high strength and stiffness combined with
9.8 Recrystallization of Two-Phase Alloys
413
dicted conditions for this are shown in Figure 9-41. Experimental work has broadly confirmed these predictions (Humphreys et al., 1990) although other factors such as grain growth after recrystallization (Section 9.8.8) and small particles introduced during the manufacture of the composite, may also affect the annealing behavior.
25
9.8.7 Interaction of Precipitation and Recrystallization
0.1
0.2 0.3 Volume Fraction
Figure 9-41. The predicted recrystallized grain size in participate composites as a function of particle size and volume fraction. The dotted lines show conditions under which pinning effects may prevent recrystallization (Humphreys et al., 1990).
a low density is often required. One of the advantages of particulate composites is that they can be mechanically processed in the same way as conventional alloys, and the recrystallization behavior is therefore important (see Humphreys et al., 1990). In these materials we have the unusual situation of alloys containing both a large volume fraction and large particles, typically 3-10 jim. The particles play an important role in controlling the grain size by particle stimulated nucleation. If each particle produces one recrystallized grain, then the grain size as a function of particle diameter and volume fraction will be as shown in Figure 9-41, in which it can be seen that very small grain sizes are achievable in these materials. With large volume fractions of smaller particles, the interparticle spacing is small, pinning effects will become important, and recrystallization may not occur. The pre-
The mutual interaction of precipitation and recrystallization has been extensively studied by Hornbogen and colleagues in Al-Cu and Al-Fe alloys, and has been comprehensively reviewed by Hornbogen and Koster (1978). If a supersaturated solid solution is deformed and annealed, then unless recrystallization is complete before precipitation begins, the two processes affect each other. Precipitation on low angle or high angle boundaries will hinder recovery and recrystallization, and dislocations will themselves promote the nucleation of precipitates. The recrystallization behavior as a function of time and temperature is summarized schematically in Figure 9-42. Temperature Tx is the solvus temperature, above which no precipitation occurs.
Range I
Range II
Range III
Log Annealing time
Figure 9-42. The interaction of recrystallization and precipitation (after Hornbogen and Koster, 1978).
414
9 Recrystallization and Recovery
Curve R shows the onset of recrystallization in the absence of precipitation. Curve RP shows the onset of recrystallization in the presence of particles. Curve P shows the onset of precipitation in the absence of a deformed structure. Curve PR shows the onset of precipitation in the presence of a deformed structure. The solid lines show the events that actually occur. The recrystallization behavior depends on the temperature, and we can identify three regimes. I) T>T1 There is no precipitation possible, and so the recrystallization behavior is that of a solute containing alloy. II) T1>T>T2 Recrystallization occurs before precipitation, therefore the recrystallization is similar to I. III) T< T2 Precipitation occurs before recrystallization. The particles control the recovery rate (see extended recovery in Section 9.3.4). Eventually recrystallization may occur. In some cases discontinuous precipitation may occur on migrating high-angle grain boundaries. 9.8.8 Grain Growth in Two-Phase Alloys
During grain growth of a fully recrystallized alloy, the driving force is much smaller than during primary recrystallization. Therefore the pinning effects of particles on grain boundaries discussed in Section 9.8.4 are relatively more important. The pinning force (pz) exerted on a boundary by an array of particles is given by Equation (9-23). During grain growth, the driving force arises solely from the curvature of the boundaries as discussed in Section 9.4.1, and the magnitude of this driving force is given by Equation (9-13). On average, we expect the grains to stop
growing at a diameter D (the limiting grain size or Zener limit) when pz = p2, that is: Dz=4r/3FY
(9-27)
More sophisticated treatments of the pinning of grain boundaries by particles (Gladman, 1966; Nes etal., 1985; Hillert, 1988; Doherty etal., 1989) yield rather similar results. The pinning effect of very low volume fractions of small particles is sufficient to limit the grain size significantly. For example, a volume fraction of 5 x 10 ~3 of particles of diameter 50 nm gives Dz = 67 jim. This analysis averages the particle pinning force over the boundary, which may be valid if the average number of particles per grain (NPG) is very large. However, if NPG is small, then we might expect the particles to lie predominantly on the grain boundaries (Hellman and Hillert, 1975; Hutchinson and Duggan, 1978). In this case, Pz is given approximately by: Pz = 3FyD/8nr2
(9-28)
From Equations (9-13) and (9-28) we find: (9-29) with k ~ 6. For the particle dispersion above, this gives Dc ~ 4 |im. The number of particles per grain is now small and it is doubtful whether Equation (9-29) which neglects the correlation of particles and boundaries is valid. A more extreme case arises if the particles are at the grain corners (Hellman and Hillert, 1975). We see that the derivation of an accurate value of limiting grain size is not straightforward in circumstances where there is strong correlation between particles and boundaries. However, it is reasonable to assume that there will be a small limiting grain size Z>c, given by a more accurate
9.9 Recrystallization Textures
form of Equation (9-29), when the number of particles per grain is small. We should perhaps think in terms of two limiting grain sizes. If the material crystallises to a grain size less than Dc then we expect the grains to stop growing at about this size. However, if primary recrystallisation produces a grain size larger than Dc then we might expect the grains to be able to grow to the larger limiting size Dz. The experimental evidence necessary to test Equations (9-27) and (9-29) is not particularly clear, as has been discussed by Cahn (1983) and Doherty et al. (1989). Computer simulations of grain growth have been used extensively to study this problem (see Anderson (1986) and Doherty et al. (1989) for reviews). Two-dimensional simulations give results in agreement with Equation (9-29), and three-dimensional simulations suggest a dependence on iC 0 " 3 3 .
9.9 Recrystallization Textures When a metal is recrystallized, the crystallographic orientations of the new grains may be quite different from those of the old ones. If this is the case, then the recrystallization texture will be different from the deformation texture. As textures have a significant effect on the mechanical and physical properties of a material, the understanding of the origin of textures is of great importance. The control of textures during the mechanical processing of metals is discussed in more detail in Chapter 10 of this volume. The published literature on the subject of textures is extensive, and there are several contentious issues which continue to arouse much debate. For further general information the reader is referred to reviews by Hatherly and Dillamore (1975),
415
Grewen and Huber (1978), Nes and Hutchinson (1989) and Hatherly (1990). The series of trienniel international conferences on texture of which the most recent was in Avignon in 1990 also provide much reference material. A texture can be described in terms of the ideal orientation or group of orientations to which it approximates. We will confine ourselves to textures produced on rolling and on subsequent annealing, and to materials of cubic symmetry. In this case the ideal orientation is given as (hkl) [uvw], where (hkl) is the plane parallel to the rolling plane and [uvw] is the lattice vector parallel to the direction of rolling. In some cases the texture has a single component, for example (113 [2lT]). However, more commonly the texture of the material corresponds to several ideal orientations. Thus the texture of deformed aluminium may contain the components (112) [111], (Oil) [211] and (427) [232] (see also Chap. 10, Sec. 10.2.2). Although such descriptions are convenient and concise, they give no indication of how closely the material approximates to these orientations, and if the material has only a weak preferred orientation then these descriptions are of little use. A graphical representation of the texture spread is provided by a pole figure, which is obtained directly from X-ray diffraction data. A pole figure is essentially a stereographic projection in the rolling plane of the specimen in which the orientations from all the grains in the material, of a particular type of crystallographic pole, e.g. <111), are plotted. Pole densities are normally represented by plotting contours of equal pole density. Figure 9-43 shows the <111> pole figures for deformed and annealed copper. Pole figures do not give a complete description of the orientation of the material,
416
9 Recrystallization and Recovery RD
Figure 9-43. (Ill) pole figures for copper, a) Cold rolled 90%. b) Fully recrystallized. (Ridha and Hutchinson, 1982).
and a better representation of texture is via orientation distribution functions (ODFs). These are three-dimensional plots in which the orientation of the three axes of a crystallite are plotted with respect to three standard axes. ODFs are determined by mathematical analysis of the data from several pole figures, and they are usually displayed as a series of two-dimensional
sections. For an introduction to the subject, refer to Hatherly and Hutchinson (1979). The topic is also treated in Chapter 10, Section 10.2.3. The texture after annealing is primarily determined by two factors: - the orientations which are produced during nucleation, - the relative growth velocity of grains of different orientations. For many years there has been heated debate as to which of these two factors control the final texture. It has been claimed that the recrystallization texture has its origin in either the preferred nucleation of grains with the required orientation {oriented nucleation theory) or the preferred growth of such grains from a more randomly oriented array {oriented growth theory). See Hatherly (1990) for a critical discussion. The experimental evidence in favor of these theories has been provided mainly from textures measured on bulk specimens. This is a particularly blunt instrument for processes which occur on a submicron scale and is about as sensitive as trying to identify the micromechanisms of deformation solely from mechanical tests. Recent advances in microtexture techniques which enable detailed correlation of microstructure and local texture are now providing the types of information necessary for obtaining a better understanding of the formation of annealing textures (see Chapter 10, Sections 10.2.5 and 10.3.4). It is clear that there are some instances in which nuclei of a limited range of orientations are produced, and this factor then will predominantly determine the final texture. Similarly, there are cases when grains of certain orientations clearly grow more rapidly. In general we must expect the final texture to be determined by both factors in proportions which will vary for different
9.9 Recrystallization Textures
materials, deformations and annealing conditions. 9.9.1 Orientation Effects in Nucleation
As discussed in Section 9.4.5, a recrystallized grain originates in a small region of the deformed material and will therefore have the orientation of that region. Although this orientation is present in the deformed material, it may not be present in sufficient quantity to be detected by bulk texture measurements and therefore is unlikely in general to be an important component of the deformation texture. In Section 9.4.5 we discussed several different sites for recrystallization, and each of these may lead to one or more texture components. 9.9.1.1 Transition Bands
A transition band separates different parts of an old grain which have undergone different types of deformation and where the individual new parts have rotated towards different, but stable end orientations (Hu, 1962). A transition band therefore is a region with a sharp orientation gradient bridging the two neighboring texture components. Dillamore and Katoh (1974) predicted that in f.c.c. metals, transition bands containing the cube orientation (001) [100] would develop. This was confirmed experimentally by Ridha and Hutchinson (1982) who showed that thin pancake-shaped regions of this orientation developed into grains of the cube orientation. The success of these nuclei in developing was undoubtedly enhanced by their shape. Similar cube nuclei have also been found in aluminium by Hjelen and Nes (1986), although only in transition bands within (112) [111] grains.
417
9.9.1.2 Shear Bands Shear bands are also important sites for recrystallization. However it is more difficult to predict the resulting textures. For example in steels, it has been shown (Haratani et al., 1984) that the Goss component (110) [001] originates at shear bands. However, in copper, the nuclei which form from shear bands are of widely spread orientations (Ridha and Hutchinson, 1982). 9.9.1.3 Prior Grain Boundaries Strain induced grain boundary migration, which is an important mechanism after lower amounts of deformation results in texture components which are present in significant amounts in the deformed material. For example the (001) [110] and (112) [110] orientations in steel are thought to originate in this way (Inokuti and Doherty, 1977). 9.9.1.4 Second-Phase Particles
Nucleation in the vicinity of large second phase particles gives rise to grains whose orientation reflect those in the deformation zones. In single crystals, where a very restricted range of orientations are produced, sharp textures may be formed on recrystallization (Humphreys, 1977). However, in polycrystals, the large spread of orientations available near the particles results in a severe weakening of the deformation texture with no new components being formed (Humphreys and Juul Jensen, 1986; Humphreys and Kalu, 1990 b). 9.9.2 Orientation Effects in Growth
The effects of orientation on grain growth were discussed in Section 9.5. In summary, there is clear evidence, particu-
418
9 Recrystallization and Recovery
larly from lightly deformed single crystals that the growth rate of grains is a maximum when there is a misorientation about a <111> axis. In aluminium the maximum growth rate occurs for a misorientation of ~40°, whereas in copper the maximum is closer to 30°. For highly deformed polycrystals there is less information, and the situation is complicated because the orientation of the deformed material changes over short distances and so no precise orientation relationship exists between the old and new grains. As pointed out by Hatherly (1990), some of the confusion over the various results may well be due to the strong effects of small amounts of impurity. Another factor which must be taken into account is that the deformation structure is heterogeneous and different regions of the microstructure contain different components of the deformation texture. Therefore a model in which oriented growth is taken into account must consider the local nuclei and their environment. Microtexture work such as that by Hjelen
• cube texture 50 _ • rolling texture A random
10
102 Time Imin]
103
Figure 9-44. The average size of grains with different orientation, measured for a series of partly recrystallized specimens of aluminium of commercial purity (Juul Jensen et al., 1985).
et al. (1990) is now beginning to provide the information which will enable more complete theories of texture development to be formulated. The role of twinning in recrystallization was discussed in Section 9.4.5. Twinning will provide new orientations, and may also be important in producing grains of suitable orientation for rapid growth. There is still some debate about the conditions under which twinning during annealing is important in determining textures, particularly in metals of higher stacking fault energy. A critical review by Hatherly (1990) gives a good summary of the current situation. 9.9.3 Competition Between Texture Components
The final texture is the result of the selection of orientations during nucleation and growth. In many materials several components are formed, originating from different microstructural sites. Nucleation and growth rates for the different components may also alter with time, making it therefore very difficult to predict the final texture. As an example, we can consider the recrystallization of commercial purity aluminium, which contains some large second-phase particles. The growth of various texture components was followed during the annealing process and the results are shown in Figure 9-44. Initially, microstructural observation showed nucleation to be predominantly at particles (random component). However, during the anneal, these grains grew slowly, whereas grains of the rolling texture (formed at the old grain boundaries) grew more rapidly. Cube grains grew very rapidly, and the final texture contained a high percentage of this component. This exam-
9.10 Annealing Processes During Hot Deformation
pie illustrates the way in which both selection of orientations, different growth rates, and competition between the various components plays a significant role in determining the final texture.
9.10 Annealing Processes During Hot Deformation During the hot working of metals, the softening processes recovery and recrystallization may occur during the deformation. In this case we call the processes dynamic recovery and dynamic recrystallization in order to distinguish them from the processes which occur during static annealing, which we have discussed above. Although the static and dynamic processes have many features in common, the simultaneous occurrence of deformation and softening leads to some interesting and important effects. These mechanisms occur during operations such as hot rolling, extrusion or forging. They are important because they lower the flow stress of the material, thus enabling it to be deformed more easily and they may also determine the texture and the grain size of the worked material. The difference between hot working and creep is essentially one of strain rate. Hot working is generally carried out at strain rates in excess of 1/s, whereas typical creep rates are below 10~5/s. Nevertheless, in many cases similar mechanisms occur in both types of deformation. Both the basic and applied aspects of hot working of metals have been extensively studied in recent years, and the reader is referred to the reviews by Sellars (1981), Roberts (1984), Sakai and Jonas (1983), Sellars (1986), Jonas (1990) and Sellars (1990), for further information.
419
9.10.1 The Flow Stress During Hot Working During plastic deformation, dislocations move, multiply and interact, and as the dislocation density increases, the flow stress rises. If dislocation recovery (Section 9.3) occurs during the deformation, then the flow stress is reduced. In practice, dynamic recovery occurs even at low temperatures when the driving force - the dislocation density - is sufficiently high. However during deformation at a high homologous temperature, extensive recovery occurs and the flow stress tends to saturate, often reaching a steady state flow stress as shown in Figure 9-45. Such a curve is typical of aluminium, which softens dynamically only by recovery. If a material undergoes dynamic recrystallization, however, then there is generally a peak in the flow stress (Figure 9-46), which coincides approximately with the onset of recrystallization. In some cases multiple peaks are found as shown in the lower curve of Figure 9-46. The steady state flow stress during hot work (crs) depends on the temperature and strain rate of deformation, and these parameters may be linked by an empirical equation of a similar form to that which is
Strain
Figure 9-45. Restoration by dynamic recovery alone.
420
9 Recrystallization and Recovery
Strain
Figure 9-46. Restoration by dynamic recrystallization. The stress strain curve has either a single peak or multiple peaks.
Figure 9-47. Grain boundary serrations formed during dynamic recovery of an Al-Mg alloy.
often applicable to creep:
state is reached (McQueen and Jonas, 1975). However, in other materials it is not so clear as to whether the subgrain misorientation actually saturates or continues to increase with strain. During dynamic recovery at high temperatures, the grain boundaries become serrated, with a wavelength closely related to the subgrain size, as shown in Figure 9-47. This local grain boundary migration is thought to be driven by differences in the local dislocation density.
O
(9-30)
where K is a constant. Many other relationships are also commonly used (Roberts, 1984). In some cases, for example aluminium, the activation energies for both creep and hot work are similar and are close to that for self-diffusion. However in most steels, the activation energy for hot work is considerably higher than that for creep and self-diffusion. 9.10.2 Dynamic Recovery The basic mechanisms of dynamic recovery are dislocation climb and glide the same as in static recovery as discussed in Section 9.3. Dislocations form subgrains, which grow and are further deformed, and new subgrains form within the old ones. The subgrains tend to remain equiaxed, and after a strain of perhaps 1, the structure often appears to achieve a steady state. The subgrain diameter is related to the flow stress via a relationship similar to Equation (9-11). In some cases, for example pure aluminium, the misorientation between subgrains is found to remain constant at a few degrees when the steady-
9.10.3 The Characteristics of Dynamic Recrystallization The process of dynamic recrystallization is complex and not fully understood. It is also clear that there are different types of dynamic recrystallization. However, a simplified view of the phenomenon is that during dynamic recrystallization, new grains are formed and these consume the deformed matrix. As the material continues to deform, the dislocation density of the new grains increases, thus reducing the driving force for growth until growth stops. The general characteristics of dynamic recrystallization are as follows (e.g., Roberts, 1984):
9.10 Annealing Processes During Hot Deformation
421
- A critical deformation ec is necessary in order to initiate dynamic recrystallization. This is somewhat before the peak of the stress strain curve. - ec decreases steadily with decreasing stress (as), although at low (creep) strain rates the critical strain may increase again (Sellars, 1978). - The size of dynamically recrystallized grains (DR) increases monotonically with decreasing stress and the empirical relationship q
(j=k1+k2DR-
(9-31)
has often been reported, q being 0.5 to 0.8. Grain growth does not occur and the grain size remains constant through the microstructure. - (is and DR are almost independent of the initial grain size (Do), although the kinetics of dynamic recrystallization are accelerated in specimens with smaller initial grain sizes. - Dynamic recrystallization is usually initiated at pre-existing grain boundaries (Figure 9-48). For very low strain rates and large initial grain sizes intragranular nucleation becomes more important. 9.10.4 The Conditions for Dynamic Recrystallization
Several models of dynamic recrystallization have been proposed. Regardless of the details of the mechanism of nucleation, growth of a dynamically recrystallized grain depends on the distribution and density of dislocations, which provide the driving force for growth. Figure 9-49, based on the model of Roberts and Ahlblom (1978), shows schematically the dislocation density expected in the vicinity of a migrating boundary. The boundary at A is moving from left to right into unrecrystallized material which has a high disloca-
Figure 9-48. Dynamically recrystallized grains at the old grain boundaries in magnesium (Ion et al., 1982).
Direction of boundary movement
Figure 9-49. Schematic diagram of the dislocation density during dynamic recrystallization according to the model of Roberts and Ahlblom (1978).
tion density Q0. As the boundary moves, it reduces the dislocation density in its wake to zero by recrystallization. However, the continued deformation raises the dislocation density in the new grain, so that it builds up behind the moving boundary (QX at a distance X behind the boundary), tending towards a value of Q0 at large distances. If a bulge model of nucleation is assumed (Bailey and Hirsch, 1964), then in order for the grain to nucleate, the driving force (QO — QX) must reach a critical value over a minimum distance Xc. If we simplify the analysis of Roberts and Ahlblom (1978), then dynamic re-
422
9 Recrystallization and Recovery
crystallization occurs at a critical value of the parameter Q3Me~1, where M is the boundary mobility. In materials such as aluminium and pure (b.c.c.) iron, recovery occurs readily and this parameter, strongly dependent on dislocation density, never reaches the critical value. Therefore only dynamic recovery occurs. However in materials of lower stacking fault energy such as copper, nickel and stainless steel, recovery is slow and the dislocation density increases to the value necessary for dynamic recrystallization to occur. 9.10.5 Microstructural Evolution During Dynamic Recrystallization
Dynamic recrystallization generally starts at the old grain boundaries as shown in Figure 9-50 a. New grains are nucleated at the boundaries of the growing grains, and in this way a thickening band of recrystallized grains is formed as shown in Figure 9-50 b. If there is a large difference between the initial grain size and the recrystallized grain size, then a "necklace" structure of grains may be formed (Figures 9-50b and 9-51). At sufficiently large
strains, the material will become fully recrystallized (Figure 9-50c). Whether the size of the recrystallized grains is determined by their cessation of growth owing to the reduction in driving force as deformation proceeds or by the nucleation of new grains at the moving boundary is not entirely clear. As shown in Figure 9-46, the stress strain curves of a dynamically recrystallizing material may be characterized by a single peak or by several oscillations. Luton and Sellars (1969) have explained this in terms of the kinetics of dynamic recrystallization. At low stresses, the material recrystallizes completely, before a second cycle of recrystallization begins. This process is then repeated. The flow stress, which depends on the dislocation density, therefore oscillates with strain. At high stresses, subsequent cycles of recrystallization begin before the previous ones are finished. The material is therefore always in a partly recrystallized state after the first peak, and the stress strain curve is smoothed out, resulting in a single peak. Sakai and Jonas (1983) have suggested that the shape of the stress strain curve
Figure 9-50. The microstructure developed during dynamic recrystallization. (a)-(c) show the finegrained structure developed at high stresses and (d) shows the larger grain size formed at low stresses. The initial grain structure is shown.
9.10 Annealing Processes During Hot Deformation
423
matrix. This is known as metadynamic recrystallization. Unrecrystallized regions may also subsequently recrystallize statically and therefore both static and metadynamic recrystallization may coexist. There is some dispute in the literature (see Roberts, 1984) as to whether metadynamic recrystallization results in a smaller or a larger grain size than that which is formed by dynamic recrystallization alone. Figure 9-51. "Necklace" microstructure developed in zinc during dynamic recrystallization.
depends on the ratio of the recrystallized and starting grain sizes (D0/DR). If (Do/ DR) > 2 then the microstructure develops as shown in Figures 9-50 a-c, the material is always partly recrystallized and a smooth curve with a single peak if found. However, if D0/DR<2 then the new grains develop at about the same time because there are enough sites (i.e., old boundaries) for recrystallization to be complete in one cycle as shown in Figure 9-50 d. In this case an oscillatory stress strain curve is predicted. Models of dynamic recrystallization in which the kinetics and the final grain size can be predicted with some accuracy are now finding extensive application in the commercial thermomechanical processing of steels (e.g. Sellars 1986, 1990; Jonas, 1990). 9.10.6 Metadynamic Recrystallization Whenever the critical strain for dynamic recrystallization (ec) is exceeded, recrystallization nuclei will be present in the material. If straining is stopped, but annealing continued, then these nuclei will grow with no incubation period into the heterogeneous, partly dynamically recrystallized
9.10.7 The Mechanisms of Dynamic Recrystallization The problems of investigating the micromechanisms of a dynamic process occurring at high temperature are formidable, and there is still an incomplete understanding of the mechanisms of dynamic recrystallization. 9.10.7.1 Dynamic Recrystallization in Single Crystals A considerable research effort has been expended in studying the fundamental aspects of the dynamic recrystallization of single crystals, and this is reviewed by Mecking and Gottstein (1978). In copper single crystals, recrystallization is nucleated by a large subgrain, and this is followed by repeated annealing twinning until a high mobility boundary is formed. However, as dynamic recrystallization in materials of industrial importance is always associated with grain boundaries, we will not consider this aspect of dynamic recrystallization in detail. 9.10.7.2 Grain Boundary Nucleation Dynamic recrystallization occurs at high-angle boundaries, either the original grain boundaries, boundaries of dynamically recrystallized grains or boundaries such as deformation bands which are ere-
424
9 Recrystallization and Recovery
ated on straining. Bulging of grain boundaries is frequently observed as a prelude to dynamic recrystallization, and it is probable that a mechanism closely related to strain-induced grain boundary migration (Section 9.4.5) operates. 9.10.7.3 Grain Boundary Impingement As seen in Figure 9-47, grain boundaries develop serrations during dynamic recovery (Figure 9-52 a). At high rolling strains, if the wavelength of these serrations is comparable with the thickness of the grains, then interpenetration of the scalloped boundaries will occur (Figure 952 b). The result of this is an equiaxed structure of what appear to be small re-
Figure 9-53. Microstructure developed by grain boundary impingement in Al—Mg (Drury and Humphreys, 1986).
crystallized grains (Figure 9-53). However, no recrystallization has occurred and this should not be regarded as dynamic recrystallization. Humphreys (1982) has shown that the strain at which impingement occurs is given approximately by: s=K-ln(Doa)
(9-32)
9.10.7.4 Dynamic Recrystallization by Lattice Rotation al Low strain
b) High strain Figure 9-52. Schematic diagram showing impingement of serrated grain boundaries during hot working, a) Low strain, b) High strain.
There is considerable evidence that in certain materials, new grains with high angle grain boundaries may be formed by the progressive rotation of subgrains during straining. This is a strain-induced phenomenon which should not be confused with the static subgrain rotation discussed in Section 9.3.3.2. In this mechanism, subgrains adjacent to grain boundaries are progressively rotated as the material is strained. The old grains develop a gradient of misorientation from centre to edge. In the centre of the old grain, subgrains may not be well developed or may have very low misorientations. Towards the grain boundary, the
9.11 References
misorientations increase, and at high strains may become high-angle boundaries. This mechanism was first found in minerals (Poirier and Guillope, 1979), and has since been found in a variety of nonmetallic crystalline solids. There is evidence that it occurs in metals containing significant solute additions, particularly Al-Mg alloys (Drury and Humphreys, 1986) and Al-Zn alloys (Gardner and Grimes, 1979). Figure 9-54 shows the microstructure of an Al-5%Mg alloy in which a mantle of well developed subgrains has developed at the original grain boundaries. There is some evidence that a similar process may occur in alloys containing finely dispersed second-phase particles. The particles pin the subgrains and prevent extensive growth, and there is evidence (Nes, 1979) that during straining the subgrain misorientation progressively increases. It has been suggested that such a mechanism is responsible for the development of the fine grained microstructure during the thermomechanical processing of superplastic aluminium alloys (Watts et al., 1976; Stowell, 1983).
Figure 9-54. Development of misorientations adjacent to the grain boundary by lattice rotation during deformation of Al-Mg (Drury and Humphreys, 1986).
425
9.11 References Abbruzzese, G., Liicke, K. (1986), Acta Metall. 34, 905. Ahlborn, H., Hornbogen, E., Koster, U. (1969), J. Mater. Sci. 4, 944. Anderson, M. (1986), Proc. 7th Int. Riso Symposium, Riso, Denmark: Hansen, N. (Ed.), p. 15. Anderson, M. P., Grest, G. S., Srolowitz, D. X, Sahni, P. S. (1984), Acta Metall 32, 783. Ashby, M. F. (1970), Phil. Mag. 21, 399. Ashby, M. F. (1980), Proc. 1st Int. Riso Symp., Riso, Denmark: Hansen, N. (Ed.), p. 325. Aust, K. X, Rutter, J. W. (1959), Trans AIME 215, 119 and 608. Avrami, M. J. (1939), Chem. Phys. 7, 103. Babcock, S. E., Balluffi, R. W. (1989), Acta Metall. 37, 2367. Bailey, X E., Hirsch, P. B. (1964), Proc. Royal Soc. A267, 11. Baker, 1., Martin, J. W. (1983), Metal Sci. 17, 459. Bay, B., Hansen, N. (1979), Metall. Trans. A 10, 279. Beck, P. A., Sperry, P. R. (1949), Trans. AIME 180, 240. Beck, P. A., Sperry, P. R. (1950), J. Appl. Phys. 21, 150. Bellier, S. P., Doherty, R. D. (1977), Acta Metall. 25, 521. Berger, A., Wilbrandt, P. X, Ernst, F., Klement, U., Haasen, P. (1988), Prog. Mater. Sci. 32, 1. Burke, X E. (1949), Trans. AIME 180, 73. Cahn, X W., Pedawar, G. E. (1965), Acta Metall. 13, 1091. Cahn, R. W. (1949), /. Inst. Metals 76, 121. Cahn, R. W. (1983), in: Physical Metallurgy. Cahn, R. W., Haasen, P. (Eds.). New York: Elsevier Science Pulishers, 3rd edition, p. 1595. Cahn, R. W. (1990), in: High Temperature Aluminides and Intermetallics. Whang, S. H., Liu, C. T., Pope, D. P., Stiegler, X O. (Eds.). TMS, p. 245. Chan, H. M., Humphreys, F. X (1984), Proc Electron. Micr. Soc. of America 476. Channon, S., Walker, H. (1953), Trans. AIME 45, 200. Cotterill, P., Mould, P. R. (1976), Recrystallization and Grain Growth in Metals. Surrey, U.K.: Surrey Univ. Press. Davies, G. G., Stoloff, N. S. (1966), Trans. AIME 236, 1605. Detert, K. (1978) in: Recrystallization of Metallic Materials: Haessner, F. (Ed.). Stuttgart: Dr. Riederer Verlag, p. 97. Dillamore, I. L., Katoh, H. (1974), Metal Sciences 8, 73. Dillamore, I. L., Morris, P. L. Smith, C. X E., Hutchinson, W. B. (1972), Proc. Royal Soc. 329 A, 405. Doherty, R. D. (1974), Metal. Sci. J. 8, 132. Doherty, R. D. (1978), in: Recrystallization of Metal-
426
9 Recrystallization and Recovery
lie Materials: Haessner, F. (Ed.)- Stuttgart: Dr. Riederer Verlag, p. 23. Doherty, R. D., Martin, J. W. (1962), /. Inst. Metals. 91, 332. Doherty, R. D., Martin, J. W. (1964), Trans. ASM 57, 874. Doherty, R. D., Martin, J. W. (1976), Stability of Micro structure in Metallic Systems. Cambridge: Cambridge University Press. Doherty, R. D., Rollett, A. R., Srolovitz, D. J. (1986), Proc. 7th Int. Rise Symposium, Ris0, Denmark: Hansen, N. (Ed.), p 53. Doherty, R. D., Li, K., Kashyup, K., Rollett, A. D., Anderson, M. P. (1989), Proc. 10th Int. Riso Symp. Riso, Denmark: Bilde-Sorensen, J. (Ed.), p. 31. Drury, M. D., Humphreys, F. J. (1986), Acta Metall. 34, 2259. Duggan, B. X, Hutchinson, W. B., Hatherly, M. (1978), Scripta Metall. 12, 1293. Faivre, P., Doherty, R. D. (1979), /. Mater. Sci. 14, 897. Gardner, K. X, Grimes, R. (1979), Metal Sci. 13, 216. Gawne, D. T., Higgins, R. A. (1971), /. Mater. Sci. 6, 403. Gladman, T. (1966), Proc. Royal Soc. A 294, 298. Gleiter, H. (1969), Acta Metall. 17, 853. Grant, E., Porter, A., Ralph, B. (1984), /. Mater. Sci. 19, 3554. Grewen, X, Huber, X (1978), in: Recrystallization of Metallic Materials: Haessner, F. (Ed.). Stuttgart: Dr. Riederer Verlag, p. 111. Haessner, F , Hofmann, S. (1978), in: Recrystallization of Metallic Materials: Haessner, F. (Ed.). Stuttgart: Dr. Riederer Verlag, p. 63. Haratani, T., Hutchinson, W. B., Dillamore, I. L., Bate, P. (1974), Metal Science 18, 57. Hatherly, M. (1982), Proc. 6th Int. Conf on Strength of Metals and Alloys. Melbourne: Pergamon. Hatherly, M. (1990), in: Recrystallization '90: Chandra, T. (Ed.). Wollongong, Australia: TMS, p. 59. Hatherly, M., Dillamore, I. L. (1975), /. Australian Inst. Metals. 20, 71. Hatherly, M., Hutchinson, W. B. (1979), An Introduction to Textures in Metals. London: The Institute of Metals. Hellman, P., Hillert, M. (1975), Scand. J. Metall. 4, 211. Herbst, P., Huber, X (1978), in: Texture of Metallic Materials. Gottstein, Lucke (Eds.). Berlin: Springer, p. 452. Higgins, G. T. (1974), Metal Sci. J. 8, 143. Hillert, M. (1965), Acta Metall. 13, 227. Hillert, M. (1988), Acta Metall. 36, 3177. Hjelen, X, Orsund, R., Nes, E. (1990), Acta Metall., in press. Hjelen, X, Nes, E. (1986), Proc. 7th Int. Rise Symposium. Riso, Denmark: Hansen, N. (Ed.), p. 367. Hornbogen, E. (1970), Praktische Metallographie 9, 349.
Hornbogen, E., Koster, U. (1978), in: Recrystallization of Metallic Materials, Haessner, F. (Ed.). Stuttgart: Dr. Riederer Verlag GmbH, p. 159. Hu, H. (1962), Trans. Met. Soc. AIME 224, 75. Huber, X, Hatherly, H. (1980), Z. Metallkde. 71, 15. Hull, D., Bacon, D. X (1984), Introduction to Dislocations, 3rd edn. New York: Pergamon. Humphreys F. X (1977), Acta Metall. 25, 1323. Humphreys, F. X (1979 a), Metal'Science 13, 136. Humphreys, F. X (1979b), Acta Metall. 27, 1801. Humphreys, F. X (1980), Proc. 1st Int. Riso Symp., Riso, Denmark: Hansen, N. (Ed.), p. 35. Humphreys, F. X (1982), Proc. 6th Int. Conf. on Strength of Metals and Alloys, Vol. 1: Gifkins, R. (Ed.). Melbourne: Pergamon Press, p. 625. Humphreys, F. X (1985), in: Dislocations and Properties of Real Materials: Loretto, M. H. (Ed.). London: Inst. Metals, p. 175. Humphreys, F X, Juul Jensen, D. (1986), Proc. 7th Int. Riso Symposium, Riso, Denmark: Hansen, N. (Ed.). Humphreys, F. X, Kalu, P. N. (1987), Acta Metall., 35, 2815. Humphreys, F. X, Kalu, P. N. (1990a), Acta Metall. 38, 917. Humphreys, F. X, Kalu, P. (1990 b), Proc. 9th Int. Conf. on Textures, Avignon, France. New York: Pergamon, in press. Humphreys, F. X, Martin, X W. (1967), Phil. Mag. 16, 927. Humphreys, F X, Miller, W. A., Djazeb, M. R. (1990), Materials Science and Technology, in press. Hunderi, O., Ryum, N., Westengen, H. (1979), Acta Metall. 27, 1131. Hutchinson, W. B., Besag, F. M., Honess, C. V. (1973), Acta Metall. 21, 1685. Hutchinson, W. B., Duggan, B. X (1978), Metal Sci. 12, 372. Inokuti, Y, Doherty, R. D. (1977), Texture 2, 143. Ion, S. E., Humphreys, F. X, White, S. H. (1982), Acta Metall. 30, 1909. Johnson, W. A., Mehl, R. F. (1939), Trans. AIME 135, 416. Jonas, X X (1990), in: Recrystallization '90: Chandra, T. (Ed.). Wollongong, Australia: TMS, p. 27. Jones, A. R., Ralph, B, Hansen, N. (1979), Proc. Royal Soc. A 368, 345. Juul Jensen, D., Hansen, N., Humphreys, F. X (1985), Acta Metall 33, 2155. Kalu, P. N., Humphreys, F X (1986), 7th Int. Riso Symp., Riso, Denmark: Hansen, N. (Ed.), p. 385. Kamma, C , Hornbogen, E. (1976), J. Mater. Sci. 11, 2340. Koken, E, Chandrasekaran, N., Embury, X D., Burger, G. (1988), Mater. Sci. Eng. A104, 163. Koppenaal, T K., Parsatharathi, M. N., Beck, P. (1960), Trans. AIME 218, 98. Koster, U., Hornbogen, E. (1968), Z. Metallkde. 59, 792.
9.11 References
Leslie, W. C, Mikalak, J. T., Aul, F. W. (1963), in: Iron and Its Dilute Solid Solutions: Spencer, Werner (Eds.). New York: Interscience, p. 119. Liebmann, B. Liicke, K., Masing, G. (1956), Z. Metallkde. 47, 57. Li, J. C. M. (1962), /. Appl. Phys. 224, 75. Li, J. C. M. (1966), in: Recrystallization Grain Growth and Textures: Margolin (Ed.). ASM, p. 45. Lucke, K., Abbruzzese, G., Heckelmann, I. (1990), in: Recrystallization '90: Chandra, T. (Ed.). TMS, p. 37. Lucke, K., Detert, K. (1957), Acta Metall. 5, 628. Luton, M. 1, Sellars, C. M. (1969), Acta Metall. 17, 1033. Mahin, K. W, Hanson, K., Morris, J. W. (1980), Ada Metall. 28, 443. Marthinsen, K., Lohne, O., Nes, E. (1990), Acta Metall. in press. McElroy, R. J., Szkopiak, Z. C. (1972), Int. Metall. Reviews 17, 175. McQueen, H. X, Jonas, J. J. (1975), in: Treatise on Materials Science and Technology, Vol. 6: Arsenault, R. (Ed.). New York: Academic, p. 393. Mecking, H., Gottstein, G. (1978), in: Recrystallization of Metallic Materials: Haessner, F. (Ed.). Stuttgart: Dr. Riederer Verlag, p. 195. Mott, N. F. (1948), Proc. Phys. Soc. 60, 391. Nes, E. (1976), Acta Metall. 24, 391. Nes, E. (1979), Scri. Metall. 13, 211. Nes, E. (1986), Proc. Symp. on Microstructural Control During Processing of Aluminium Alloys. New York: AIME-TMS, p. 95. Nes, E., Hutchinson, W. B. (1989), Proc. 10th Int. Riso Symp. Riso, Denmark. Bilde-Sorensen, J. (Ed.), p. 233. Nes, E. Ryum, N., Hunderi, O. (1985), Acta Metall. 33, 11. 0rsund, R., Nes, E. (1988), Scri. Metall., 22, 671. Oscarsson, A., Hutchinson, W. B., Karlsson, A. (1987), 8th ILMT. Vienna, Leoben, p. 531. Parker, E. R., Washburn, J. (1952), J. Trans. AIME 194, 1076. Poirier, J. P., Guillope, M. (1979), Bull. Mineral. 102, 67. Randle, V, Ralph, B., Hansen, N. (1986), Proc. 7th Riso Int. Symp., Riso, Denmark: Hansen, N. (Ed.). p. 123. Ray, R. K., Hutchinson, W. B., Duggan, B. J. (1975), Acta Metall. 23, 831. Rhines, F. N., Patterson, B. P. (1982), Metall. Trans. 13A, 985. Ridha, A. A., Hutchinson, W. B. (1982), Acta Metall. 30, 1929. Ringer, S. P., Li, W. B., Easterling, K. (1989), Acta Metall. 37, 831. Roberts, W. (1984), in: ASM Seminar: Deformation, Processing and Structure. Krauss, G. (Ed.). St. Louis, Missouri: ASM, p. 109. Roberts, W, Ahlblom, B. (1978), Acta Metall. 26, 801.
427
Roessler, B., Nowick, D. T, Bever, M. B. (1963), Trans. AIME 227, 985. Ryde, L., Hutchinson, W. B., Jonsson, S. (1990), in: Recrystallization '90: Chandra, T. (Ed.). Wollongong, Australia: TMS, p. 313. Sakai, T., Jonas, J. J. (1983), Acta Metall 32, 189. Saetre, T. O., Hunderi, O., Nes, E. (1986), Acta Metall. 34,981. Sandstrom, R. (1977), Acta Metall. 25, 897. Schmidt, 1, Haessner, F. (1990), Condensed Matter 81,215. Sellars, C. M. (1978), Phil. Trans. Royal Soc. 288,147. Sellars, C. M. (1981), Metals Forum 4, 75. Sellars, C. M. (1986), Proc. 7th Int. Riso Symposium, Riso, Denmark: Hansen, N. (Ed.), p. 167. Sellars, C. M. (1990), Materials Science and Technology, in press. Shewmon, P. G. (1969), Transformations in Metals. New York: McGraw-Hill. Shockley, W, Read, W. T. (1950), Phys. Rev. 78, 275. Smith, A. W, Rawlings, R. D. (1976), Phys. Stat. Sol. 34 A, 117. Smith, D. A., Rae, C. M. F. (1979), Metal Sci. 13, 101. Smith, C. J. E., Dillamore I. L. (1970), Metal Science Journal, 4, 161. Smith, C. S. (1948), Trans AIME 175, 47. Stowell, M. J. (1983), Proc, 4th Int. Riso Symposium, Riso, Denmark. Bilde-Sorensen, J. (Ed.), p. 119. Stuwe, H. P. (1978), in: Recrystallization of Metallic Materials. Haessner, F. (Ed.). Stuttgart: Dr. Riederer Verlag, p. 11. Thompson, A. W. (1977), Metall. Trans. 8A, 833. Turnbull, D. (1951), Trans. AIME 191, 661. Vandermeer, R. A., Rath, B. B. (1990), in: Recrystallization '90. Chandra, T. (Ed.). Wollongong, Australia: TMS, 511. Varma, S. K. (1986), Mater. Sci. Eng. 82, 19. Viswanathan, R., Bauer, C. L. (1973), Acta Metall. 21, 1099. Watts, B. M., Stowell, M. J., Baikie, B. L., Owen, D. G. E. (1976), Metal Sci. 3, 189 and 198. Weaire, D., Kermode, J. (1983), Phil. Mag. 48, 245. Wert, J. A., Austin, L. K. (1988), Metall. Trans. 19 A, 617. Wert, J. A., Paton, N. E., Hamilton, C. H., Mahoney, M. W. (1981), Metall. Trans. 12A, 1267.
General Reading Cahn, R. W (1983), in: Physical Metallurgy: Cahn, R. W, Haasen, P. (Eds.). New York: Elsevier Science Publishers, 3rd edition, p 1595. Chandra, T (Ed.) (1990), Recrystallization '90. TMS. Haessner, F. (Ed.) 1978), Recrystallization of Metallic Materials. Stuttgart: Dr. Riederer Verlag. Martin, J. W, Doherty, R. D., (1976), Stability of Microstructure in Metallic Systems. Cambridge
428
9 Recrystallization and Recovery
Solid State Series. Cambridge: Cambridge University Press. "Recrystallisation in the Development of Microstructure". Proceedings of a conference held at Leeds in 1978. Published in: Metal Science 13, 101-267 (1989). "Microstructure and Mechanical Processing". Proceedings of a conference held in Cambridge in 1990. Published in: Materials Science and Technology (in press) (1991).
Proceedings of the Riso Symposia on Metallurgy and Materials Science, held annually since 1980 at Ris0 National Laboratory, Roskilde, Denmark. The following are particularly relevant: "Recrystallisation and Grain Growth of MultiPhase and Particle-Containing Materials" (1980). "Annealing Processes, Recovery, Recrystallization and Grain Growth" (1986). "Materials Architecture" (1989).
10 Measurement and Control of Texture Robert W. Cahn Department of Materials Science and Metallurgy, University of Cambridge, Cambridge, U.K.
List of 10.1 10.1.1 10.1.2 10.1.3 10.1.4 10.2 10.2.1 10.2.2 10.2.3 10.2.4 10.2.5 10.3 10.3.1 10.3.2 10.3.3 10.3.4 10.3.5 10.3.6 10.3.7 10.4 10.4.1 10.4.2 10.4.3 10.4.4 10.4.5 10.4.6 10.5 10.6 10.7
Symbols and Abbreviations Introduction The Nature of Texture The Examination of Texture Sources of Texture Texture and Properties The Measurement of Texture Pole Figures and Their Determination Principal Texture Types Orientation Distribution Functions and Their Determination Micro textures (Mesotextures), Their Determination and Use On-Line Measurement of Textures The Influence of Texture on Properties Ductility, Deep Drawing and r-Factor Earing Strength, Toughness and Fatigue Resistance Microtextures and Mechanical Properties Magnetic Properties and Texture Other Properties Effect of Texture on the Measurement of Residual Stress The Control of Texture Hot and Cold Rolling; Primary Recrystallization Grain Growth and Secondary Recrystallization Tertiary Recrystallization Transformation Textures Textures in Intermetallics Magnetic and Stress Annealing Prognosis Acknowledgement References
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
430 432 432 433 433 437 438 438 439 443 447 455 455 455 457 458 460 460 461 462 465 465 469 471 472 474 475 476 476 476
430
10 Measurement and Control of Texture
List of Symbols and Abbreviations a,b,c Alm D,D0 d, d^ E
f(u,(j)),f{
(hkl) J K
l,m M
P(u,4>) r
rm (or R) R t
t,t0 u
(uvw} w o ,w
crystal axes Fourier transform final and initial grain diameters lattice plane spacing Young's modulus 2) orientation distribution function crystallite orientation indices of the lattice plane quantitative texture parameter (or texture index or texture sharpness parameter) constant integers Taylor factor pole figure function continuous orientation distribution function plastic strain ratio (or plastic anisotropy parameter) average of r in plane of sheet vector in the direction of the rotation axis time final and initial sheet thickness = cosy crystal vector initial and final sheet width
(j>i,4>2 0 W
angle between a vector and a crvstal axis angle between a vector and b crystal axis angle between a vector and c crystal axis shear on 7th slip system macroscopic tensile or compressive strain associated Legendre function glide transfer number Poisson's ratio principal stress reciprocal of the fraction of lattice point of one grain tan" 1 (cosa/cos/?); azimuth angle in X-ray measurement of residual stresses two of three Euler angles used in defining ODF one of three Euler angles used in defining ODF inclination of X-ray beam to sample surface normal
CR CSL EGBD EBSP
cross direction (= TR, q.v.) coincidence site lattice extrinsic grain boundary dislocations electron back-scattering pattern
a
P y yj s 0lm X v
List of Symbols and Abbreviations
GBCD GMT l.d.r. N OCF OCR ODF RF RD SACP SAD SEM SFE STEM TEM TR
grain boundary character distribution grain misorientation texture limiting draw ratio sheet normal orientation coherence function orientation correlation function orientation distribution function Rodrigues-Frank rolling direction selected area channeling patterns selected area diffraction scanning electron microscope stacking fault energy scanning transmission electron microscopy transmission electron microscopy transverse direction (= CR, q.v.)
431
432
10 Measurement and Control of Texture
10.1 Introduction
10.1.1 The Nature of Texture
Texture, in the sense in which the word is used in this chapter, denotes a preferred orientation, that is, a statistical tendency for the constituent crystal grains of a polycrystalline material to be oriented in a particular way: a different way of putting this is to say that the mean orientation is equivalent to that of a single crystal with its crystallographic axes set at specific angles with respect to the surface and edges of the specimen, accompanied by a greater or lesser degree of scatter about that "ideal" orientation. A texture in this sense is the rule rather than the exception in practical materials: in fact, when a material with a random texture is needed (as it was in the early days of nuclear fuel design) it may be quite difficult to produce it. The objective of this chapter is to outline the origins of texture without going into detail (since such detail can be found elsewhere in the Series), to describe the methods of measuring and graphically depicting texture, to explain the effects that texture has on the properties and processing of materials (mostly metals and alloys) and, finally, to outline the available methods of controlling texture in industrial materials. The generation of textures by intense plastic deformation is described in Vol. 6, Chap. 3 (and further treated with specific respect to steels in Vol. 7, Chap. 6, Sec. 6.4.1), while texture formation during annealing of deformed metals and alloys is discussed in Chap. 9, Sec. 9.9 of this volume. Specific topics relating to texture can also be found elsewhere in the series (e.g., Vol. 7, Chap. 7, Sees. 7.3 and 7.4, concerned with the effect of texture on the formability of steel).
The first distinction to be made is between macrotexture and microtexture. In describing a macrotexture, the grains in a polycrystal are regarded as constituting a single statistical population without regard to the spatial location of any particular grain or its relation to its neighbors. By contrast, a microtextural description involves determining the orientation of each grain of population and determining the nature and degree of its misorientation with respect to its immediate neighbors. A microtextural description is more complete than a macrotextural one, and the former already contains the information required to describe the latter. - The more nearly all the grains approximate to a single ideal orientation (or, sometimes, to two or three alternative ideal orientations), the stronger the texture is said to be. A second distinction needs to be made between different types of macrotexture. The first type is called & fiber texture and is found in some castings, electrodeposits, evaporated thin films and especially in cold-drawn wires or in extrusions. Here there is always one direction in the material which is functionally distinct from all others: the wire axis, the direction of heat flow in a casting or of current flow in an electrodeposit. A particular crystal vector, (uvw}, tends to align itself along the unique material direction, but all azimuths about that vector are equally probable. The fiber texture is the stronger, the larger the strain in a wire-drawing operation; in the case of castings or electrodeposits, the matter is not so straightforward. - Something formally equivalent to a fiber texture is also found when anisotropy of surface free energy induces grains to develop preferentially at a free surface with a particular (hkl) plane parallel to the surface. - Fiber
10.1 Introduction
textures may be modified by becoming cylindrical textures, in which a particular lattice plane tends to lie parallel to the wire surface at all azimuths round the circumference (e.g., Rieck, 1957). The second type of macrotexture is a sheet texture, and refers either to unidirectionally rolled sheet (a rolling texture) or to sheet of this kind after it has been annealed and recrystallized (an annealing texture). An ideal rolling texture consists of a specific (hkl) plane parallel to the sheet plane and a specific (uvw} vector parallel to the rolling direction: such an ideal texture is simply denoted "(hkl) (uv w>". An annealing texture is generally as sharp as the preceding rolling texture, but is often qualitatively different. - Heavy emphasis is placed on the study of rolling and annealing textures because of the great importance of metallic sheet, steel sheet especially, in industrial practice, and because a rolling texture affects both the strength and the formability of sheet. 10.1.2 The Examination of Texture A microtexture can only be determined by measuring the individual orientations of numerous grains. This can be done by etching the polished surface to generate sharp crystallographically defined etchpits and using a two-circle goniometer to reflect light from the etchpit walls. Geologists who work with transparent materials such as quartzite use a polarizing geological microscope with a (rotating) universal stage as a means of mapping the orientations of individual grains, but this falls outside our concern here. This kind of optical approach is not often used nowadays. - The alternative approach is to use diffraction: the alternatives include back-reflection Laue X-ray diffraction, transmission or back-reflection Kossel X-ray dif-
433
fraction, electron diffraction or electron channeling. A macrotexture is determined only by exploiting diffraction, usually of X-rays but sometimes of neutrons. There are two drastically different ways of determining and graphically depicting such a texture. In the first, the intensity of diffraction from one particular family of lattice planes in the polycrystal is measured as a function of direction in space and the results are collected in the form of a pole figure, which refers exclusively to one particular {hkl}. The alternative is to record many such pole figures for different {hkl} and combine them by a process of pole figure inversion to produce an orientation distribution function (ODF), which gives fuller information about the nature of the texture than a single pole figure can. The information in such an ODF can only be depicted in the form of a three-dimensional diagram, generally shown on the page in terms of a number of parallel plane sections. This second type of measurement is nowadays referred to as three-dimensional texture analysis. 10.1.3 Sources of Texture In practical terms, by far the most important source of texture is unidirectional plastic deformation, most commonly wiredrawing or rolling, and recrystallization following such deformation. It has been known ever since the studies by Schmid and Boas (1935) that the orientation of the axis of a long metallic single crystal rotates when it is plastically deformed in tension; the axis rotates towards the slip vector and the slip planes become progressively more nearly parallel to the tensile axis. When double slip (on two planes simultaneously) begins, the rotation is abruptly slowed and its character changes.
434
10 Measurement and Control of Texture
The individual grains of a polycrystal behave in an analogous way. Because each grain has to fit in with the overall strain of the assemblage, such grains are much more severely constrained than a free single crystal. Taylor (1934) made a renowned study of this problem, based on von Mises' (1928) proof that any one grain requires the uniform operation of at least 5 independent slip systems to achieve an arbitrary imposed strain; 'independent' here is rather a subtle concept which places restrictions on shared slip directions and planes. Taylor introduced a criterion of minimum work: that combination of 5 systems, out of the many possible combinations, will function which minimizes M = S yy/e. Here M is the Taylor factor, yj is the shear in the jth slip system and e is the macroscopic tensile or compressive strain. On this basis, Taylor was able to show how grains of different initial orientations, relative to the externally imposed stress, become reorientated in different ways, and thus he was the first to show how a preferred orientation, or texture, can come into existence. Early ideas, and some more recent developments, are summarized in many papers, such as those by Chin (1969) and Cahn (1978). Since Taylor's original work, much has been done to refine his ideas. The most important work was done by Bishop and Hill (1951), who showed that the Taylor/ von Mises theory satisfies the yield criterion, i.e., that the shear stress reaches the critical level for slip on the active systems identified by Taylor without exceeding it on the inactive slip systems - although some other derivative theories do not satisfy the yield criterion. Both Taylor and Bishop/Hill assume strain continuity from one grain to its neighbors - i.e., each grain deforms as the assemblage does. But this assumption is incompatible with stress
continuity, and indeed it is known that the internal stress varies from one grain to another (see Sec. 10.3.7), and indeed is not even uniform in any one grain. The implication of this is that Taylor's assumption of uniform strain in each grain is unsustainable. Varying strains in adjacent grains implies varying rotation of orientation from one grain to another, and also variations in rotation in different parts of the same grain. Indeed, microtexture determinations in deformed iron (Inokuti and Doherty, 1977 and 1978) have shown that such variations indeed exist (Fig. 10-1); this figure refers to a single grain in a polycrystalline sheet. (The observation that a single grain, on rolling, splits into regions, or bands, with two quite distinct groups of orientation, as seen in Fig. 10-1, is quite common, and is readily explained in terms of slip on multiple systems.) A further consequence of non-uniformity of stress and of strain within individual grains is that the
411
100
110
Figure 10-1. Local orientations (surface normals) of various positions in a single grain of polycrystalline iron deformed 40% in compression. Orientations were determined by individual X-ray transmission Kossel diffraction photographs. The two groups of orientations correspond to the matrix and to a series of deformation bands (after Inokuti and Doherty, 1978).
10.1 Introduction
435
80 o o
Figure 10-2. The variation in the relative proportions of (111> and <100> fiber textures in drawn wires of various f.c.c. metals and alloys, in relation to y/Gb, where y is the specific stacking fault energy, G is the shear modulus and b, the magnitude of the Burgers vector (after English and Chin, 1965).
bo
a
§ 40 PL,
16
24
10-3 yjGb
von Mises assumption that at least 5 slip systems are required in each grain also becomes unsustainable, and Fleischer (1987) has recently established experimentally that in polycrystalline oc-brass, some grains show four or even fewer distinct slip systems. The nature of slip in face-centered metals and alloys, particularly, is strongly affected by the specific energy of the stacking faults which connect the partial dislocations in a dissociated perfect dislocation (a common feature of such metals and alloys), and this energy (SFE) in turn has a strong effect on both wire and sheet textures. Fig. 10-2 shows this very clearly with respect to wire-drawing textures in a range of such metals and alloys. The texture is duplex, i.e., <111> and <100> fiber axes coexist. The ratio of grains belonging to each component texture is clearly correlated with the SFE. Similar effects are familiar in relation to sheet rolling textures. The above indicates that a realistic simulation of texture development in a polycrystal during its deformation has neces-
sarily become quite complicated, and elaborate computer codes have been developed to achieve this objective. A recent conference report includes 300 pages of papers on modelling and measurement of deformation textures (Kallend and Gottstein, 1988, pp. 265-562). - Further details will be found in Volume 6, Chap. 3. The formation of annealing textures by recrystallizing a textured, deformed polycrystal has been the topic of sustained controversy. In essence, the argument has been between two rival models. One model maintains that annealing textures are formed by oriented nucleation - i.e., the creation of nuclei by the formation of a recrystallized nucleus in a particular orientation relationship (but not one of identity) to the region of a deformed grain in which it appears. Much of the evidence in support of this model comes from microtextural studies (e.g., Doherty and Cahn, 1972). - The other model maintains that nuclei are formed at random, but only particular orientations among all these nuclei are capable of growing rapidly at the
436
10 Measurement and Control of Texture
expense of deformed material {oriented growth). It is of course possible that both models are applicable, i.e., that a wide (though not random) range of orientations appear among nuclei, and that these do not grow at quite equal rates. The report of a recent representative panel set up to attempt a measure of agreement between the rival views (Doherty et al., 1988) indicates that agreement is as far off as ever (see also Sec. 10.4.1, and Vol. 7, Chap. 6, Sec. 6.4.2). A further complication arises from the well established fact that the specific stored energy of cold work of different grains in a deformed polycrystal varies according to the orientations of individual grains; the consequence is that those grains which have the highest energy density recrystallize first and impose the consequent recrystallized orientations (presuming oriented nucleation) on the entire specimen. This has been firmly established for rolled iron by Dillamore et al. (1967). Another complication is the role of deformation twins (especially at low deformation temperatures) in f.c.c. as well as b.c.c. metals, also uranium, in determining both deformation and annealing textures. Further, the generation of annealing twins during heat-treatment has been shown to generate an especially large range of recrystallized orientations (Wilbrandt, 1988; Berger et al., 1988). More detailed evidence concerning the mechanisms by which annealing textures come about will be found in Chap. 9 of this volume, and some implications are discussed below (see Sec. 10.4.1). Textures formed during casting are apparently restricted to columnar grains, and are not found in the equiaxed grains often seen near the mold surface and at an ingot center. Weld metal often grows epitaxially on the surfaces of the pieces being joined and can thus mimic their texture (e.g.,
Babu et al., 1991). Details of actual textures and the influence of experimental variables such as freezing rate will be found in a very full exposition by Wassermann and Grewen (1962); a summary in English of the results presented there will be found in a book by Barrett and Massalski (1966, Chapter 19). One striking feature is that the nature of the texture can be drastically altered by small concentrations of dopant (Fig. 10-3). This feature is in fact reminiscent of the observation that the migration rate of grain boundaries of specific orientations during recrystallization is also highly sensitive to dissolved impurities, and this has been linked with a dragging influence of "atmospheres" of dopant travelling along with the moving interface (Cahn, 1983, pp. 1630-1636). Since the fiber texture of columnar grains sharpens progressively during the growth of grains, it can be concluded that the texture forms because differently oriented grains grow competitively at different rates. Whether this should be attributed to a difference in thermal conductivity does not appear to be known. Electrodeposits often have fiber textures, the fiber axis being the direction of current flow normal to the surface. Early results, with copious details of specific metals, are presented by Wassermann and Grewen (1962) and by Barrett and Massalski (1966, pp. 538-540). According to the plating conditions, the deposit may begin by being epitaxially related to the substrate or may have a spontaneous fiber texture from the start. Barrett and Massalski report on attempts which have been made to establish generalizations, but these have not proved very satisfactory. More recent results and models are presented by Weil (1989). It is clear that there is competition between epitaxial growth (continuation of the orientation of the substrate grains) and
10.1 Introduction (Hi)
437
(111)
10°,
10V
Zone purified Pb
Pb +0.0005 %Ag
20
(100)
(110) (100)
(a)
(110)
(b)
Figure 10-3. Stereographic projection of orientations of columnar grains in lead after unidirectional solidification, (a) Zone-refined, high-purity lead; the fiber texture is <111>. (b) Pb + 0.0005 wt.% Ag; the fiber texture is predominantly <100>. The fiber axis coincides with the heat flow vector (after Rosenberg and Tiller, 1957).
formation of a fiber texture. The circumstances favoring epitaxy are rather fully treated; a high current density seems to be a principal factor in epitaxial growth. Even when growth starts epitaxially, it often fades out in the sense that grains of other orientations grow faster and a fiber texture gradually forms. Rashkov et al. (1955) have shown that the applied overpotential determines which orientation has the lowest work of formation; selectivity is pronounced for low overpotentials but disappears when it becomes high. It has also been proposed that orientation affects the ease of adsorption of impurities such as colloidal solutes, and grains without such adsorbates grow fastest and form a preferential texture (Reddy, 1963). Many extraneous influences can affect fiber textures formed in electroplating. One of potential importance is the role of a magnetic field applied during plating, which reportedly can alter the texture so that the direction of easy magnetization becomes the fiber axis (Chiba et al., 1986). For further details, Weil's review should be consulted.
Preferred orientations of surface grains associated with anisotropy of surface energy will be discussed in Sec. 10.4.3. 10.1.4 Texture and Properties
The principal reason for the very large amount of experimental attention which has been devoted to the formation and nature of textures arises from the effect of textures on properties. Such effects are predominantly of four types, and apply especially to sheet material: (1) Modification of strength and ductility, so that both these properties are functions of direction in a sheet. (2) Changes in the ratio of width strain to thickness strain during the deformation (deep drawing in particular) of sheet metal; this is denoted as the r-ratio. The r-ratio affects how much the material can be strained before it fails locally; width strain is beneficial, thickness strain makes for premature fracture. (3) Anisotropy in the r-ratio means that the elongation in different directions during deep drawing leads to the formation of ears around the periphery of a drawn object:
438
10 Measurement and Control of Texture
this is inconvenient in that machining is needed to restore the drawn object to the desired shape, and is also undesirable because earing implies azimuthal fluctuations of sheet thickness. (4) Ferromagnetic properties, such as permeability, coercivity and saturation magnetization, are a function of crystal direction in all ferromagnetic materials, and therefore preferred orientation affects magnetic performance. This is of great industrial importance, especially in connection with the production of transformer laminations. In the later parts of this chapter, these influences are described in some detail and methods of modifying texture so that properties are optimised will also be discussed.
10.2 The Measurement of Texture 10.2.1 Pole Figures and Their Determination The traditional way of depicting a texture is by means of a pole figure. This is a stereogram which (for a rolled sheet) has the sheet normal vector (N) at the center, the rolling direction (RD) at the top and bottom and the direction in the sheet transverse to the rolling direction (TR, or CR for cross-direction) at left and right. Any pole figure refers only to a single population of lattice planes, such as {111} or {200}. ('200' is cited here rather than '100' because pole figures are determined by means of diffraction and hence absent reflections, such as '100' from a face-centered cubic metal, are not mentioned). If the orientation of a population of grains is random, the corresponding pole figure shows no features; if however there is a preferred orientation, then (say) {111} poles cluster in certain locations of the
stereogram, and this is depicted by the use of contours calibrated in multiples of the pole density that corresponds to a random texture. Fig. 10-4 shows three examples of such pole figures. Two of them refer to pure copper, heavily rolled, and show {111} and {200} pole figures together with various ideal textures. It can be seen that for this metal, the preferred orientation approximates to a combination of (123) [412] + (146)[21T]. Here, round brackets denote the lattice plane that tends to lie parallel to the rolling plane, and the square brackets denote the lattice vector that lies parallel to the rolling direction. Fig. 10-4c shows an oc-brass texture, again after heavy rolling, and it can be seen that it is quite different from the copper texture: brass approximates to a (110) [112] texture. This kind of distinction has been studied for many f.c.c. metals and alloys and it is clear that the change-over from copper type to brass type is correlated with the specific stacking-fault energy (which is much lower for a-brass than for copper). Rolling temperature also affects the texture. - The changeover is primarily associated with the linkage between the ease of cross-slip and SFE. Clearly, this changeover is analogous to that depicted for drawn wires in Fig. 10-2. Details of the available information and its interpretation can be found in the books by Barrett and Massalski (1966, Chap. 20) and by Wassermann and Grewen (1962, Chaps. 3 and 4) and in review articles such as that by Dillamore and Roberts (1965) or that by Dillamore and Stoloff (1969). Interest in the SFE-texture correlation continues, as evidenced by a study of the effects of strain, temperature and grain size for cold-rolled brass (Duggan and Lee, 1988) and a very recent paper linking textures in hammered foil with SFE (Kitagawa, 1989).
10.2 Introduction
Pole figures were first determined by exploiting diffraction photographs made with monochromatic X-rays, such as that exemplified in Fig. 10-5. (Sometimes monoenergetic neutrons are used instead, because of their much greater penetrating power, which allows a larger sample volume to be averaged (Bunge and Tobisch, 1972); this is specially beneficial for coarsegrained materials.) Instead of uniformly blackened diffraction rings, a textured material gives strong local maxima and other parts of the rings are entirely absent. There is a one-to-one relationship between, say, the configuration of the {111} ring and the contour distribution in the corresponding {111} pole figure. The transfer from film to pole figure required a number of diffraction photographs to be made with the incident beam in different orientations relative to the wire or sheet. Nowadays, this technique has been entirely displaced by one exploiting diffractometry with a counter instead of film, originally introduced by Schulz. The sample is mounted on a goniometer head like that sketched in Fig. 10-6: the incident beam and the slit in front of the counter (on the left) are held fixed, while the specimen (at the center of the ring) is rotated, usually in such a way that the corresponding point on the pole figure traces out a spiral. The counter readings are turned directly into a series of pole density readings and the contours are interpolated by means of a computer programme; the various needed corrections are performed automatically. - The technique works either in reflection or transmission. Each yields only a partial pole figure: the reflection method leaves the outer zone blank, the transmission method leaves a central circle blank, but the overlap between the two approaches permits a seamless, complete pole figure to be determined. - The details
439
of the method are particularly clearly explained by Barrett and Massalski (1966, Chap. 9). A readily accessible elementary account is in a pamphlet by Hatherly and Hutchinson (1979), while a detailed discussion of error correction, scanning procedures, etc. can be found in an article by Bunge (1982 a). An alternative way of depicting the information inherent in a pole figure is the inverse pole figure. Here, the stereogram shows a single crystal in standard orientation (often only a single cubic stereographic triangle bounded by 100, 110 and 111 poles), and the pole the distribution of which is shown is the sheet normal or the rolling direction. Fig. 10-1 is an example of an inverse pole figure. A frequently cited instance of an inverse pole figure applied to an overall deformation texture is that for rolled uranium bars (Harris, 1952). This mode of depiction has special benefits for metals of low crystallographic symmetry, such as orthorhombic uranium. It is also quite commonly used to depict fiber textures. 10.2.2 Principal Texture Types
There is an enormous mass of detailed information about both deformation textures and annealing textures, as well as fiber textures in electrodeposits, etc., which it would be profitless to survey here. Variables such as composition of alloys; trace impurities; presence of a second phase; degree, mode and temperature of deformation; differences between surface and interior layers; annealing temperature; thermal history before annealing; rate of heating to temperature - all these play a role in determining texture, and clearly the complexities can become awesome. (An extreme example is the effect of explosive deformation, which can generate a ran-
440
10 Measurement and Control of Texture
A—CD
• Near (123)[412] A Near (146)[211]
A(ll0)[00l]
• Near (123)[412] aNear(146)[2ll]
ffl(110)[00l]
Figure 10-4. (a) {111} pole figure of electrolytic copper rolled to 96.6% reduction at room temperature. The unity contours corresponds to randomness (after Hu and Goodman, 1963). (b) {200} pole figure for same material.
441
10.2 Introduction
Figure 10-4. (c) {111} pole figure for the texture in the center of sheet of 70/30 brass rolled to 95% reduction. The triangles represent the ideal (110) [112] texture (after Hu et al., 1952).
222
22Cf 200
Figure 10-5. Monochromatic X-ray diffraction pattern of heavily cold-drawn aluminum wire (wire axis is vertical). The film is flat and placed normal to the incident beam (after Barrett and Massalski, 1966).
it i if
1!
• !
•' / if
Figure 10-6. Slits and goniometer head for Schulz's reflection method of texture determination in an Xray goniometer (after Barrett and Massalski, 1966).
442
10 Measurement and Control of Texture
dom annealing texture even though the starting material was strongly textured (Chojnowski and Cahn, 1973).) - Here the aim is only to cite a few specific textures which are of special scientific or industrial concern (especially those to which we shall return later) and to give a very few specific illustrations of the effects of some of the above-cited variables. Comprehensive overviews of textures in specific materials and for specific processing variables will be found, among many other places, in Barrett and Massalski (1966, Chaps. 19-21), Wassermann and Grewen (1962; this is the most massive compilation), Underwood (1961), Dillamore and Roberts (1965), Merchant and Morris (1985), and in the proceedings of the successive Texture Conferences, of which the most recent were those held in 1987 (Kallend and Gottstein, 1988) and in 1990 (not yet published). It is clear from the dates of most of the refer-
ences just cited that the bulk of systematic work on textures was done more than 25 years ago. The principal rolling textures of f e e . metals such as copper and brass have already been indicated (see Sec. 10.2.1). Annealing of such rolled sheets can produce an annealing texture that mimics the deformation texture; generate one or more quite new textures; produce approximate randomness; or generate the extraordinary cube texture. The cube texture, (001) [100], has generated sustained interest especially among academic students of textures, because it has been so difficult to establish the mechanism that produces it. Fig. 10-7 shows this texture as it is formed on annealing very heavily rolled copper. Heavy strain and a small penultimate grain size favors formation of cube texture in copper; some alloying elements, even in trace quantities, especially those that cause
RD
Figure 10-7. (a) {111} pole figure for the annealing texture of tough-pitch copper strip, rolled as for Fig. 10-4 and annealed 5 min at 200 °C. The pole figure refers to the central layer of the rolled strip.
A (100) [001] cube texture; A (122) [212] twin orientation of the cube texture, (b) {200} pole figure; • cube texture; • twin texture (after Beck and Hu, 1952).
10.2 Introduction
strong solution-hardening, prevent its formation. - Cube texture can also form in steels, and its formation in aluminum has recently been discussed by Hutchinson and Ekstrom (1991). The cube grains form in narrow transition bands which separate principal components of the deformation texture (Dons and Nes, 1986). These transition bands are so narrow that the cube texture that they contain after rolling is too slight to be detectable by pole figure determination. For further discussion of the cube texture, see Sec. 10.4.1 and this volume, Chap. 9. 10.2.3 Orientation Distribution Functions and Their Determination
The concept of an orientation distribution function, and the name to go with it, appear to have been originated by Sturcken and Croach (1963). They worked in the nuclear energy industry and were concerned to predict physical properties, such as the thermal expansion coefficient, for uranium which had been mechanically worked and had therefore acquired a measure of preferred orientation (texture). Orthorhombic oc-uranium is extremely anisotropic (the thermal expansivity is even negative along one crystal axis) and properties are thus exceptionally sensitive to texture: as Sturcken and Croach pointed out, pole figures or inverse pole figures give textural information only in graphical form (and moreover in a form, a stereographic projection, which is not "area-true") and it is not possible to derive accurate relations between such a projection and the physical variable of interest. To overcome this, they proposed to determine a continuous orientation distribution function, P(u, 0), synthesized by a least-squares procedure from a discrete set of diffracted intensities. (We shall call this function f(u, (/>), to follow
443
more recent practice.) Here, u = cosy and tan cf) = cos a/cos /?, where a, /?, y are the angles between the normal to the specimen surface and the a, b and c crystal axes, respectively, a, /? and y (and therefore also u and 0) determine the indices of the lattice plane, (hkl), which lies parallel to the specimen surface. The intensity of diffraction from an (hkl) plane, suitably normalized with respect to the (hkl) diffraction intensity from a randomly oriented specimen (devoid of texture), is proportional to/(w, (j>) for h, k, I. The crucial innovation was to express f(u, (/>) as a sum of discrete terms by the use of spherical harmonics, @im(u)eim(j). The 0lm(u) are the associated Legendre functions and /, m are integers. - Spherical harmonics have a number of technical mathematical properties which make them particularly useful for the present purpose (just as Fourier summations are used to express electron densities in crystals in terms of coefficients related to structure factors). - f(u, 0) can be expressed as a summation in the form 00
f(u,4>)= Z
+1
I
AUmQUm(u)em*
(io-i)
/ =0 m = - 1
The authors decided that ten terms of this expansion (allowing for trigonometrical simplifications, this takes the expansion up to l=m = 6) give sufficient precision, taking into account experimental accuracy in determining diffraction intensities. Each term of the expansion has an algebraically well defined form, but the coefficients of the terms have to be determined empirically. The next step is to minimize the difference between calculated and measured values of/(w, 4>) for a given pair of u and 0, using suitable normalization with respect to randomly oriented material to obtain the proportionality constant between/and diffracted intensity, by
444
10 Measurement and Control of Texture
solving a series of least-squares equations. As many such equations are needed as there are terms in the summation, and so measured diffraction intensities for ten distinct (hkl) are needed. This procedure (the details of which are much too complex to set out here) then gives a series of numerical coefficients for the ten terms of the summation for /, and thus the optimum magnitude of/itself is determined. This is repeated for many combinations of u and (j) and in this way a distribution of/values is obtained for a range of different h, k, I. It is important to note that in this original treatment of the problem, the plane parallel to the surface is defined by u and (/>, but that plane is free to rotate about its normal without altering the values of u and cj). The addition of a third angular variable to fix the orientation without any degree of freedom came later. Sturcken and Croach then showed how the thermal expansion coefficient can be estimated by calculating it for each orientation in terms of the known coefficients parallel to the three crystal axes, assuming no interaction between differently oriented grains, and compared expansion coefficients thus calculated with those measured directly, with reasonable success. Sturcken and Croach introduced one other concept, a quantitative texture parameter, J. This is a measure, in the form of a single number, of the degree of preferred orientation, without concern for the crystallographic nature of that preferred orientation. / is defined by: (10-2) 0 0
rolling reduction, 2.58 for 45 % and 3.44 at 70% reduction. This was for rolling at 300 °C; during rolling at 600 °C, which entails continuous (dynamic) recrystallization during rolling, less texture is able to develop, and / i s only 1.98 for 70% reduction. It is clear that the quantitative texture parameter is a most useful single figure to denote the gradual development of texture during a progressive working progress; nevertheless, the concept has not been much used hitherto. - A more thorough discussion of how / (also termed a texture index or a texture sharpness parameter) can be determined will be found in a treatment by Bunge (1982b). As mentioned above, Sturcken and Croach's pioneering invention left a degree of freedom, in that the (hkl) plane parallel to the specimen surface was deemed free to rotate about its normal. In modern usage, an orientation distribution function is wholly determinate, without degrees of freedom, and so three angles, not two, are required to define any particular orientation. These are the Euler angles, and there are a number of rival definitions. The most common set of Euler angles is l9
Urn
For a randomly oriented assemblage of grains, J= 1; for increasing degrees of preferred orientation, / progressively increases. Sturcken and Croach found /for oc-uranium to increase from « 1 to 1.29 for 10 %
Figure 10-8. Definition of the orientation of a grain in rolled sheet by means of the three Euler angles, 4>l,
10.2 Introduction
90°
445
thogonal crystal axes. An orientation distribution function (ODF), then, can only be expressed in the form of a three-dimensional plot, with the three Euler angles as axes. An example is shown in Fig. 10-9, which refers to copper unidirectionally rolled by 95% reduction (Bunge and Tobisch, 1968). The figure shows both a perspective sketch of the ODF (which is easily visualisable but of little practical use) and a series of sections at different values of 2, which is the usual way of
(b)
Figure 10-9. (a) Perspective view of the surface of halfmaximum density in the orientation distribution function for copper rolled 95 % (one of the branches has been omitted for clarity), (b) The same ODF, complete, depicted in a series of sections (after Bunge, 1982c).
446
10 Measurement and Control of Texture
displaying an ODF. This ODF was determined with the aid of neutron diffraction, which allowed the texture throughout the thickness of an unetched rolled sheet to be properly averaged. The ODF should be compared with the pole figures in Fig. 10-4 a and 10-4b for similar material. Fig. 10-4 shows a number of "ideal textures", denoted by distinctive symbols: if the texture is a simple one (e.g., the (110) [112] texture for brass shown in Fig. 10-4c), then the ODF is a single small, roughly spherical blob in the middle of the ODF 'cell', (Fig. 10-10), instead of being, as it is for copper, an irregular 'worm'. A single crystal would be represented by single point in ODF space. This particular, early ODF was determined with the aid of just four measured pole figures, for (111), (200), (220) and (311). The mathematical procedures used nowadays are complex, and have been fully set out in many publications by Bunge (1982d; 1987 a and 1987 b). His review article (Bunge, 1987 a) is titled "Three-dimensional texture analysis". This title partly denotes the trivial fact that an ODF can only be displayed in three dimensions, but it also denotes something more subtle: a pole figure, which is projected in two dimensions, can give only partial textural
Figure 10-10. ODF of heavily rolled oc-brass (after Bunge, 1987 a).
information. It is really a combinaton of millions of sets of three orthogonal poles (for a (100) pole figure) and there is no way of marrying (correlating) the sets with each other, so that complete orientations of myriads of grains could be known. Any one pole figure can be shown to be a projection, in effect, of a complete ODF on to a curved surface in ODF space, and any projection necessarily conveys less information than the three-dimensional entity which has been projected. The ODF is computed from a series of experimental pole figures, by a process termed pole figure inversion. The pole figure function, which simply expresses the variation of pole density for a particular (hkl) according to direction in space, is related to the ODF,/(0 1 , (P, (j)2) by a master equation i ^ i («,/*)=
(10-3) ',
2kl hk
kl
The pole figure function is thus directly related to the ODF,/, which is to be determined. To turn Eq. (10-3) round in such a way that f(4>1, 2) is explicitly determined as a function of the Euler angles i.e., pole figure inversion - any of a number of mathematical tricks can be used. The most commonly used is a series expansion method (Bunge, 1987 a) in which both P and / are expressed as expansions in terms of the relevant angular variables (a, /?) and the Euler angles, respectively and Eq. (10-3) is turned into a relation between the two sets of coefficients. The expansion used for / is the spherical harmonic expansion which we met above. The coefficients of the pole figure expansion are determined first from the measured pole figures and then the coefficients of the ODF expansion can be computed from the set of pole figure coefficients. The
10.2 Introduction
procedure replaces the least-squares process introduced by Sturcken and Croach. The whole modern procedure is only feasible with the help of high-speed computers. This outline, which is the best that can be done without a long mathematical essay, hides a number of complications. There are termination errors, akin to those encountered in crystal structure determination: in that case, there is only a limited number of distinct (hkl) intensities experimentally available, so a notionally infinite series has to be terminated. In the present case, only a limited number of distinct (hkl) pole figures is available. There is also another source of errors, akin to the phase problem in crystal structure determination (in its simplest form, the impossibility of knowing whether a coefficient is positive or negative): the effect of this is to create "ghosts" in the computed ODF. Another way of regarding ghosts is that they are a consequence of the pole correlation problem in ordinary pole figures, mentioned above. - Ghosts can be largely eliminated, however, by computations that take into account the crystal symmetry and remember that an ODF cannot have a negative value for any combination of Euler angles. In the last few years, a number of other methods of achieving pole figure inversion have been tested and improved. They are surveyed in the report of a panel discussion held at the 8th International Conference on Textures of Materials (Wenk et al., 1988). Several rival Fourier methods have been introduced (thus sharpening the analogy with the procedures of crystal structure determination) and also a number of "direct methods" which work in real, not Fourier space. Critical comparisons of the resolution and general accuracy of the many rival methods are beginning to be implemented. Alternative methods of depicting ODFs, distinct from the Cartesian
447
method universally used up to now, are also being proposed. For further details of these arcane matters, the report should be consulted. The various complications are mentioned only to indicate that the practical task of determining ODFs is not one to be lightly undertaken. Nevertheless, the benefits flowing from the use of a three-dimensional texture description, the ODF, can be considerable, as we shall see, and many investigators have found the effort of determining full ODFs worthwhile. 10.2.4 Microtextures (Mesotextures), Their Determination and Use Up to this point, we have only been concerned with experimental methods which average the texture over volumes of the order of cubic millimeters (when neutron diffraction is used, the averaging may be over cubic centimetres). Pole figures and ODFs give no indication of where, in the specimen, different texture components are located. If, say, surface layes and the sheet interior differ in mean texture, only elaborate series experiments with progressively etched sheet can establish this. Pole figures and ODFs can in principle give no information about the distribution of types and magnitudes of misorientations across the myriads of grain boundaries in a sheet or wire. In this Section, we address these issues. - As mentioned earlier, the term microtexture has been applied to a state in which the distribution of the type and magnitude of misorientations between neighbouring grains is non-random. Another concise term recently proposed for this situation is mesostructure (Adams et al., 1987), while Randle (1990a) introduced the terms grain misorientation texture (GMT). Watanabe (1984) prefers the term grain boundary character distribution
448
10 Measurement and Control of Texture
(GBCD). Misorientation distribution has also been used. This luxuriance of nomenclature is what one might expect from a rather new field of investigation. We shall use 'microtexture' and 'mesotexture' interchangeably. In Section 10.1.3 and Fig. 10-1, we saw that orientations vary substantially even within the confines of a single grain in substantially cold-worked polycrystals. However, here we shall be concerned mainly with changes in orientation between adjacent grains, neglecting fluctuations within individual grains; thus we shall be concerned mainly with lightly deformed or recrystallized materials. The methods used, however, are the same for both types of problem. The available experimental techniques for microtexture determination involve either X-ray or electron diffraction. Until recently, the traditional Laue method of X-ray diffraction (see Vol. 2 A, Chap. 9) was generally avoided, since hour-long exposures are necessary when microbeams (typically, a few tens of microns in diameter) are used (Cahn, 1951). However, recently Gottstein (1988) has demonstrated that by the use of ultrabright synchrotron X-ray sources (see Vol. 2 A, Chap. 10), a Laue photograph can be obtained with a 5 x 5 jim beam in about one second, and this technique may come to be more widely used as synchrotron sources become more accessible. Humphreys (1988) has reviewed the range of techniques (other than the Laue method) now used for microtexture determination. Another recent study which exemplifies microtextural studies in several metal processing situations, using three distinct techniques, is by Juul Jensen and Randle (1989). The principal X-ray method used is the Kossel method (see Vol. 2 A, Chap. 9): here,
an electron beam is focused to a fine point (which can be as fine as 20 nm with a STEM, scanning transmission electron microscope, or 1-2 jam if a SEM, scanning electron microscope, is used) on the surface of the specimen: characteristic X-rays are generated at this point and spread in a cone. Some rays in this cone are diffracted from different crystal planes in the grain under investigation, and are detected either in transmission or in back-reflection. The pattern consists of a series of conic sections which are easy to analyse in terms of orientations, with the use of automated computerised methods. The Kossel method was first applied to microtextural problems in a study of the recrystallization of aluminum polycrystals, by Ferran et al. (1971) and extensively used thereafter for similar studies on aluminum (Bellier and Doherty 1977) and iron (Inokuti and Doherty, 1978). Exposures are of the order of a few minutes and the method has the great advantage that large areas of specimen are readily available for examination. Electrons are used in the familiar selected area diffraction (SAD) method in TEM. This method, however, is restricted to very small areas which are sufficiently thin for electron transmission. Other electron methods include the use of Kikuchi lines, electron analogues of Kossel patterns (see Vol. 2 A, Chap. 1), electron back-scattering patterns (EBSP) and selected area channeling patterns (SACP), outlined in Vol. 2 A, Chap. 6. In EBSP, the intensity of backscattering has a crystallographic component (such patterns are in effect high-angle Kikuchi patterns), while channeling depends upon the ability of electrons to travel further in a crystal when moving parallel to atom rows. The application of EBSP to microtextural studies was pioneered by Venables and Harland (1973) and developed by Dingley (1984).
449
10.2 Introduction
Table 10-1. Microtextural examination: pros and cons of available methods. Instrument
Highest spatial resolution
Angular resolution (°)
Usable for very imperfect grains?
Transmission (electrons): SAD SAD Micro-diffraction Micro-diffraction
HVEM TEM TEM/STEM FEG/STEM
0.4 urn 1.5 um 10 nm 1 nm
2 2 2 5
Yes Yes Yes Yes
Kikuchi (electrons): SAD Micro-diffraction
TEM TEM/STEM
1.5 um 20 nm
0.2 0.2
No No
SACP
SEM SEM/STEM
10 um 1-2 um
>0.5 0.5
No No
EBSP
SEM
0.5 um
<0.5
No
Kossel (X-rays)
SEM
10 um
0.5
Yes
Laue (X-rays)
Synchrotron source
5 um
<0.5
Yes
Pattern
Table 10-1, modified from one published by Humphreys (1988), summarises the pros and cons of the various microtextural orientation methods. Results of microtextural studies are sometimes displayed in inverse pole figures (see Sec. 10.2.1). - However, when interest, as it often does, focuses on the orientation relationship of neighboring grains, a display method which combines a micrograph with orientational information is highly desirable. The most effective display method developed to date is due to Inokuti et al. (1987). Fig. 10-11 shows an example of this method. The authors had undertaken a prolonged study of the development of industrially useful Goss texture, (110) [001], in silicon-iron used for transformer laminations (see Sec. 10.3.5). The microtextural method used was the transmission Kossel method, and for the sample displayed in Fig. 10-11, ^1500 Kossel photographs had been taken of individual recrystallized grains and interpreted by computer analysis. The microstructure was then stored in an image analyser memory
(see Vol. 2B) and the orientations, transformed into angles suitable for stereographic depiction, fed into the memory. The end-result was a print-out in which color has a one-to-one correlation with orientation, as shown in the (cubic) unit stereographic triangle at the bottom of each illustration. Separate printouts are needed for the distribution of rolling plane and rolling direction. - The band of large grains in the centre of the figure are secondary grains with the Goss texture. - A refinement inherent in this display technique is that the computer draws the grain boundaries in a range of thicknesses according to the magnitude of the misorientation across them. Another, more abstract method of displaying microtextural information has recently been proposed by Plege (1987) and by Adams et al. (1987). This is an orientation correlation function or OCR (Plege), or alternatively orientation coherence function (OCF) (Adams et al., 1987). This is defined as the probability density for the joint occurrence of crystallite orientation g
10.2 Introduction
at point p and orientation gr at point p', where p and p' are separated by a vector r. An OCF is plotted in 1? 3>, c/>2 space, like on ODF. The OCF approach is more quantitative but seems less illuminating than Inokuti's approach. - Another innovation is the introduction of the concept of a glide transfer number, A, which is a measure of the hindrance to the transfer of dislocation glide across a population of grain boundaries (Prantl and Werner, 1990); it can be applied both to grain boundaries and to interphase boundaries. For grain boundaries, X is defined so as to be maximum for a small mean misorientation. There are two distinct aspects to the misorientation between two contiguous grains. First, there is the simple magnitude of the rotation that will convert one orientation into the other (the magnitude of misorientation) and secondly, the orientation of the axis about which that rotation takes place. With cubic grain pairs, this information is quite tricky to work out, because any pair of crystals are related by numerous alternative axes and rotation angles. The normal procedure (e.g., Randle, 1990 b; Prantl and Werner, 1990) is to choose the axis which gives the smallest rotation angle, and to quote that angle and axis. The procedures involved were set out long ago by Mackenzie (1964). One
Figure 10-11. Computer-generated color map of orientations related to microstructure: (a) rolling plane, (b) rolling direction. The maps refer to transformer sheet steel containing 3.35 wt.% Si, 0.013 wt.% Mo, and several other constituents. The sheet was coldrolled by 70% reduction and given a secondary recrystallization anneal for 50 h at 850 °C in argon. The microstructure refers to a layer parallel to the sheet surface, 1/10 of the sheet thickness below the surface. The colonies of grains which are red in (a) and blue in (b) have the Goss texture. (110) [001] (photographs by courtesy of Dr. Y. Inokuti).
451
Figure 10-12. Microtexture: data for (f.c.c.) stainless steel, recrystallized, lightly strained and reannealed, initiating surface-induced grain growth. The stereographic unit triangle shows the non-random distribution of rotation axes for pairs of contiguous grains (after Randle and Brown, 1989).
can display a histogram of rotation angles (which can be compared with a corresponding histogram, theoretically constructed, for a purely random array of grain boundaries). Another procedure is to plot the orientations of the rotation axes (also known as misorientation axes) on a stereographic unit triangle, and it will be clear to inspection whether these axes are randomly oriented (Randle and Brown, 1989). For comparison, a random distribution of rotation axes can be constructed using the methods of Mackenzie (1964). In the experiment illustrated in Fig. 10-12, referring to grain growth in stainless steel, it was found that the distribution of rotation axes becomes progressively more nonuniform as grain growth continues (Randle and Brown, 1989). In an attempt to rationalise the description of inter-grain rotations (as well as orientation mapping), Frank (1988) has resurrected a nineteenth-century mode of representation of crystal orientations and rotations between crystal pairs, due to O. Rodrigues; for mathematical details of this
452
10 Measurement and Control of Texture
involved subject, the reader is referred to Frank's account. The method, now referred to as the Rodrigues-Frank method (Randle, 1990) allows both the direction of the rotation axis and the magnitude of rotation to be depicted in a single three-dimensional diagram. Fig. 10-13 shows the "fundamental zone" for cubic crystals in "Rodrigues-Frank" space, with the three crystal axes indicated. R is a vector the direction of which represents the rotation axis and the length, the rotation angle (Randle, 1990 b). The zone marks the maximum feasible rotation angles. Randle (1990 a) has demonstrated the utility of the RF method in analysing the recrystallization process in (austenitic) stainless steel. The RF plot of the mesotexture of such recrystallized material was compared with the simulated mesotexture plot for grains which have passed through two generations of twinning, and good agreement was established (see also Chap. 9, Sec. 9.4.5.4, for a discussion of the role of twinning in recrystallization). A further aspect of misorientation is the distinction between ordinary and special grain boundaries (this distinction is treated
Figure 10-13. Shape of the fundamental zone for cubic crystals in Rodrigues-Frank (RF) space. The origin is at the centre of the truncated cube and the axes x, y, z are parallel to the crystal axes a, b, c. An orientation or intergrain misorientation is specified by a vector R in RF space, where the rotation axis and angle are defined by the direction and length of R, respectively (after Randle, 1990).
in Vol. 1, Chap. 9). Special boundaries are those for which the fit between the contiguous lattices at the boundary is particularly good, and they have a local minimum in grain boundary energy. (For this reason they are sometimes referred to as low-energy boundaries; another commonly used name is CSL boundaries, because they are defined by a coincidence site lattice between the two adjacent grains . . . i.e., a proportion of the two sets of lattice points which lie on a single common coincidence lattice.) The degree of specialness is expressed by the I number, where Z is the reciprocal of the fraction of lattice points of one grain which coincide, or nearly coincide, with the lattice points of the neighboring grain; a lower value of Z represents a 'more special' boundary. (A very low-angle boundary is a particular form of special boundary.) A polycrystal with a largerthan-random population of special boundaries (allowing a defined small deviation from an ideal orientation relationship) is expected to have unusual properties, particularly with respect to mechanical behavior, but also with respect to grain boundary mobility, resistance to corrosion and electrical properties. This influence of special boundaries has been intensively studied for some years by Watanabe (1984 and 1988), who proposes the development of improved (especially, more ductile) polycrystalline materials by grain boundary design. A recent illustration of his concerns is found in a paper by Watanabe et al. (1989), devoted to the study of a Fe-6.5mass% Si polycrystal. Arai and Ohmori (1986) had developed this alloy family for transformer sheet; by rapid solidification processing followed by careful heat-treatment, they were able to render this concentrated alloy ductile (as well as giving it a magnetically favorable texture), a remarkable achievement in view
10.2 Introduction
453
(a) 24.1 10.1
10
Fe-6.5mass% Si (Twin-roller method) 1363 K. 3.6 ks Annealed. (100) Texture Total G.B. =387 2T1 (Low-angle) = 24.8% =20.2%
0) D
cr
2.1
1
3
5
Him n Fl n
7
11
9
13
15 17
19 21
23 25 27
29
Z (b)
40
81 27
9
29 13
• 42.5% 41% Steel , (Don and Majumdar, J
1986)
30
ONi (Lim and Raj. 1984)
/ ' I
Fe-6.5 mass% Si rapidly solidified and annealed 1363 K, 3.6 ks
2 20
Fe-3 mass% Si 74% rolled and annealed, 1333 K
10
Pure Fe rolled annealed 973 K. 360 ks I Random orientn Distribution I 0.5
1.0
1.5
Figure 10-14. (a) The frequency of special boundaries as a function of Z in rapidly solidified and annealed ribbon of Fe-6.5 mass % Si alloy, in the state of optimum heat-treatment, (b) The same data, together with data for other materials, showing frequency or special boundaries as a function of I. Hatched bars represent experimental values, open bars represent frequencies expected if the texture were random (after Watanabe et al., 1989).
of the normal brittleness of this kind of alloy. Watanabe et al. examined the alloy in various states of heat-treatment, culminating in that found to be optimal, and they found a much higher than random prevalence of special boundaries, as indicated in Fig. 10-14 a. In their paper, they examine in great detail the functional relationship between I and boundary frequency, and find a theoretical linear relationship between Z~ 1 / 3 and boundary frequency. This relationship is shown in Fig. 10-14 b and shows the very much higher special boundary frequency for the ductile Fe-Si alloy than for a material with a random mesotexture. (Z~ 1/3 = l represents a small-angle boundary.) Recently, Lim and Watanabe (1990) have simulated in detail the fracture process of a simplified polycrystal and have demonstrated a mechanism by which a high fraction of fracture-resistant low-angle boundaries can enhance the overall fracture toughness of the polycrystal. Low-angle boundaries more easily permit the transfer of slip from one grain to the next (as expressed by the glide transfer number, A, mentioned above).
454
10 Measurement and Control of Texture
Lartigue and Priester (1988) have determined the mesotexture of magnesiumdoped polycrystalline alumina and found a substantial fraction of special boundaries after heat-treatment at 1500°C. They also showed that such boundaries are poor sources and sinks for extrinsic grain boundary dislocations (EGBD) and therefore their prevalence enhances the creep resistance of the material. An even subtler aspect of the mesotexture has recently been analyzed in a remarkable series of papers by Nichols (1991) and Nichols et al. (1991). They develop a kind of percolation theory to analyse the size and shape of clusters of grains which are each separated from at least one nearest neighbor by grain boundaries with a small misorientation (an arbitrary cutoff value of misorientation can be imposed, 10° in their analysis). The work was done entirely by simulation for an initially randomly oriented two-dimensional array of regular grains and then the relation between grain boundary energy and misorientation is called into play to allow the simulated microtexture to develop by a postulated process of grain rotation (see Chap. 9, Sec. 9.3.3.2). Clusters of the kind referred to develop progressively. An applied constraint such as a directional field or stress, acting through a simple algorithm, enhances this process of cluster formation and constrains the clusters to become elongated and oriented parallel to the constraint. (A good example of this is the influence of stress in modifying mesotexture in quartzite because of elastic anisotropy, see a review by Cahn, 1978.) The authors conclude from all this that attempts to rationalise properties by reference to simple concepts such as mean grain size are sometimes too unsophisticated; for some purposes, a measure of the mean cluster size is more appropriate. Watanabe
would no doubt agree! - Their study appears to have been motivated by the experimental observation by Dimos et al. (1990) that the critical current in a bicrystalline Ba 2 Cu 3 O 7 _ 5 ceramic superconductor is intimately linked to the magnitude of the misorientation at the grain boundary (Fig. 10-15); it follows that in a polycrystal, the generation of low-misoriented grain clusters could be of capital importance in determining the overall critical current. In fact, Nichols and Clarke (1991) have applied their approach to an interpretation of the data in Fig. 10-15. - A similar
(b) 10L A
• •
5 -
o
• O
•
A
c -
(001) tilt
• (100) tilt (100) twist
o a A
5 -
o
•
o o o
2 --
io"
2
o
1
0
10
20
o i
30
Ltt
Misorientation angle (deg) Figure 10-15. Critical current across a grain boundary for a bicrystalline YBa 2 Cu 3 O 7 _ 5 ceramic superconductor at 5 K (after Dimos et al., 1990).
10.3 The Influence of Texture on Properties
mesotexture (cluster) dependence should arise with regard to the electrical resistivity of polycrystalline metal films. 10.2.5 On-Line Measurement of Textures As outlined below (see Sec. 10.3.1), the formability of steel sheet is intimately linked to the type and degree of macrotexture. Quality control of sheet intended for deep drawing or press-forming (as for automobile bodies) would be much facilitated if some appropriate measure of the texture could be continuously obtained from a moving steel sheet in the steel mill. An instrument specifically for this purpose has recently been developed (Kopineck and Bunge, 1988). A detailed analysis showed that the rm value (see Sec. 10.3.1), a measure of deep drawability, is closely correlated with the intensity of (220) and (221) diffractions from planes parallel to the sheet plane; according to the texture type, rm is either proportional to the (220) intensity or to a weighted function of the sum of the two intensities. The instrument uses an X-ray tube which generates strong white radiation, combined with a pair of energyresolving germanium detectors cooled by a compact cooling device (such detectors need to be kept permanently cold to avoid irreversible damage). The principle is that of energy-dispersive X-ray diffraction (see Vol. 2, Chap. 9). The intensities of both reflections are measured continuously at each of two positions (i.e., two different Bragg angles); this is possible because the detectors can resolve the energy and thus the wavelength of the components of the white radiation, and thus the reflected intensity for a particular wavelength, corresponding to a particular lattice plane, can be selected from the total signal. This geometry allows both the texture type and its strength to be continuously monitored,
455
and this in turn means that a continuous derived value for rm is recorded. - What this instrument in effect does is to record continuously something resembling the quantitative texture parameter, /, explained in Sec. 10.2.3. Because of the time constants of the recording instrument, the rm value derived from the measurement is averaged over ^10 seconds, which corresponds to 100 m length of steel strip. It turns out that the texture quality and intensity varies little over such distances, so the spatial resolution is acceptable. The instrument can be used on-line to modify or arrest the operation of the rolling-mill which produces the steel sheet.
10.3 The Influence of Texture on Properties Texture affects a range of properties of practical importance: the most intensely studied are deep drawing and press-forming (especially of steel sheet), and ferromagnetic anisotropy as it affects transformer laminations. We offer here a summary overview of these and a few other, more specialised aspects of texture-property correlation, with some references to the very copious literature. 10.3.1 Ductility, Deep Drawing and r-Factor The plastic strain ratio or plastic anisotropy parameter, r, for a tensile test, defined as
= ln(w/wo)/ln(t/to)
(10.4)
where w0, w are initial and final sheet widths, t, t0, final and initial sheet thicknesses, was first seen to be closely correlated with deep drawability in a classic study
456
10 Measurement and Control of Texture
by Lankford et al. (1950); accordingly, r is often called the Lankford value. What really matters in this connection is the mean, rm, of r measured for a number of different directions in the sheet (sometimes this parameter is denoted by R). Some years later, it was found (Burns and Heyer, 1958) that the rm value, in turn, could be calculated from a knowledge of the major texture parameter. The reasons for these correlations, and the nature of the deep drawing process, are fully explained in the chapter on formable steels (see Vol. 7, Chap. 7, Sees. 7.3.4 and 7.3.5) and will not be further discussed here, beyond pointing out that rapid thickness reduction favors premature fracture in deep drawing, so that a low rm value necessarily implies limited drawability. It is also important to emphasize that the press-forming processes widely used in industry (for instance, to form automobile bodies and domestic consumer durables) can be regarded as particular forms of deep drawing, and thus proper control of the plastic strain ratio affects a huge sector of manufacturing industry. Fig. 10-16 a shows the limiting draw ratio (before local fracture) as a function of rm and Fig. 10.16b shows how rm depends on a simplified aspect of texture, the proportions of grains with (111) and (100) planes parallel to the sheet plane. Fig. 1016 c is a remarkable plot: it shows the variation in limiting draw ratio for a range of different metals and alloys as a function of rm, and shows a very wide range of l.d.r. (1.9-3.0) as rm ranges from ^0.1 to 5 (Wilson, 1966). This graph is particularly clear evidence for the major role played by texture-induced plastic strain anisotropy in controlling drawability. A more detailed, up-to-date discussion of the role of texture in connection with the drawing and stretching of steel sheets can be found in an overview by Bleck et al.
(1988). The Japanese in recent years have made particular advances in controlling and improving deep drawing steels (for instance, by the addition of titanium to create extra deep drawing steel, Fukuda and Shimizu, 1972), and some details of these advances can be found in Chap. 7 of Vol. 7, which is by a Japanese author. Some further remarks about methods of controlling texture in steels to achieve high drawability will be found in Sec. 10.4.1 below. Aluminum alloys also need an optimised texture to give the best drawability, but here it is probably more important to reduce anisotropy as evidenced by earing (see next section). Other non-ferrous metals have also been treated to optimise texture for drawing, for instance, pure tantalum sheet which is used for the manufacture of high-quality electrolytic condensers (Pokros, 1989). Texture is of particular importance in designing titanium and zirconium alloys with reasonable drawability, and makes all the difference between an acceptable and a wholly unusable material (e.g., Tome et al., 1988). Texture in relation to hot working of titanium alloys has recently been discussed by Flower (1990). The role of texture in determining the feasibility of deformation twinning has been examined by Chirkin et al. (1991). Various attempts have been made at a rigorous theory of the plastic (deep drawing) behavior of anisotropic, textured sheets. A recent one is due to Lequeu and Jonas (1988): they applied continuum plasticity theory to rationalize the effect of different types of texture and are able to explain in detail why textures with (111) in the sheet plane confers the best properties on mild steel.
10.3 The Influence of Texture on Properties
457
2.45
2.40 -
2.35 -
2.30 ^ 1
10~
10°
101
102
103
3.0
Figure 10-16. (a) The deep drawability of mild steel, represented by the limiting drawing ratio (l.d.r. = ratio of the diameter of the largest blank which can be completely drawn without fracture to the diameter of the punch) as a function of the mean r value (rm, here R) for the material (after Atkinson and McLean, 1965). (b) R value of mild steel as a function of the volume ratio, /(in)//(ioo)> of grains having (111) and (100) approximately parallel to the
sheet plane (after Held, 1967). (c) Limiting drawing ratios for a range of metals and alloys: 1 - rolled and annealed zinc; 2 - rolled aluminum; 3 - cube-textured tough-pitch copper; 4 - rolled and annealed aluminum; 5 — balanced-texture tough-pitch copper; 6 — 70/30 brass, rolled and annealed; 7 and 8 - mild steel variants; 9 - rolled and annealed titanium alloy (after Wilson, 1966).
10.3.2 Earing
cup has ears, in the sense that what should be a smooth rim of the drawn cup is in fact wavy, and shows a series of lobes (ears), at azimuth intervals typically of 90° round the rim. This matters because (a) it increases the machining or other processing which has to be done on the drawn product, and (b) it implies an azimuthal variation in the thickness distribution of the product, which in turn requires a thicker gauge than
The r value, as we have seen, differs according to the direction in the sheet plane in which tension is applied in a tensile test. The consequence of this anisotropy is that, in deep drawing, the balance between thickness and width strain varies with direction, and so the drawing strain is not isotropic. The consequence is that a drawn
458
10 Measurement and Control of Texture
would be necessary if the strain were isotropic. Ears are a problem particularly in connection with beverage cans drawn from aluminum alloy, a very large industry indeed and one in which very small variations in quality are economically crucial. The anisotropy which leads to earing is usually expressed as a planar anisotropy, Ar, defined by = l/4(ro+r9O-2r45)
(10-5)
where the suffixes refer to measurements of 0°, 45° and 90° with respect to the rolling direction. The various steps that have been taken to control texture so as to minimise or eliminate earing are discussed in Sec. 10.4.1, below. A particularly detailed discussion of earing in aluminum alloys is available in several papers in a conference report (Merchant and Morris, 1985); a paper in this report by Bunge (1985) exemplifies the calculation of r-values as a function of direction in steel sheet from experimentally determined ODFs, and excellent agreement was obtained. ODFs can therefore be used as a means of predicting earing behavior, though one may surmise that it can be found out more expeditiously by means of a deep drawing test. - A possibly more useful approach is that due to Nelson et al. (1988): they were able to show, by empirical statistical correlation tests, that the earing behavior of various batches of an aluminum alloy could be very accurately predicted from the ODFs derived from the alloy after the hot-rolling stage and before the final annealing and cold-rolling. Here, ODFs could be used as a quality control test at an early stage of production and indicate which batches should not be further processed.
10.3.3 Strength, Toughness and Fatigue Resistance
A suitable texture, obtained typically by an appropriate deformation and annealing schedule, can generate sheet material with an exceptionally high strength in certain sheet directions. This is notably so for materials of low symmetry, such as titanium, zirconium, beryllium, magnesium and uranium. This phenomenon is known as texture hardening.
The masters of this field of metallurgy were Hosford and Backofen, who in the 1960s published a stream of experimental and theoretical papers (e.g., Hosford and Backofen, 1964; Hosford, 1969). They applied continuum mechanics, particularly Hill's theory of metal plasticity, and were able to predict yield loci for all possible forms of stressing from measurements of a few yield stresses for specific directions. They were then able to interpret findings such as those by Babel et al. (1967), that suitably texture-strengthened Ti-5A12.5Sn alloy could be fashioned into spherical pressure vessels with yield levels ^ 4 0 % higher and burst strengths ^ 7 5 % higher than for the isotropic alloy. Hosford (1969) also analyses in general terms the behavior of hexagonal metals. Some of these form very strong textures with the basal plane parallel to the sheet plane, which gives them high rm values and consequently substantial texture strengthening under conditions of biaxial tension. Others, especially titanium and zirconium, with their multiple deformation mechanisms, form more complex textures. According to Hosford, any rigorous analysis of their behavior is particularly difficult, and this is borne out by the literature which in general avoids analysis of texture strengthening in hexagonal metals. The sensitivity of such metals to texture can be
10.3 The Influence of Texture on Properties
vanishing point in recent years, and this is certainly something that needs putting right. It has been shown that a strong texture can also have a spectacular effect on fatigue life. Thus, Fig. 10-18 shows the fatigue curves of samples of a Ti-4A1-4V alloy prepared in such a way that basal planes were normal to the rolling plane and parallel to the rolling direction. Curves labelled "T" in the figure were stressed in the transverse direction, i.e., normal to the basal planes, those labeled "L" were stressed parallel to the basal planes. This effect appears to be due to the suppression of mechanical twinning in the "L" specimens. - Creep resistance of magnesium has also been found to be the better, the stronger the basal plane sheet texture (Roberts, 1960). Texture-strengthened alloys generally have reduced formability compared with isotropic material, and texture control is thus essential for successful forming, especially of the intrinsically less ductile materials, beryllium in particular. Toughness control, e.g. of textured Ti-6A1-4V alloy (Bowen, 1978), is an underdeveloped field: there is very little literature on this important topic. Thus the proceedings of the
2 (mm 2)
Figure 10-17. Effect of texture on the Hall-Petch plot for extruded and annealed magnesium bars: 1 - extrusion ratio 54:1, parallel to extrusion axis. 2 - extrusion ratio 7.5:1, parallel to extrusion axis. 3 - as "2", at 35° to extrusion axis. 4 - (extruded sheet) - extrusion ratio 54:1, parallel to extrusion axis (after Wilson, 1966).
gauged from Fig. 10-17 (Wilson, 1966), which shows how the Hall-Petch plot, relating yield stress to grain size, varies with direction in the sheet, and also with the texture type for a fixed direction. The variations are very large. Studies of the kind pioneered by Hosford and Backofen have diminished to
(a)
(c)
•
750
\ a
700 "
R-OA
:
o
n
\
• •
e A
650
\
^ \
\
a
•-T ) D
600
L
~ jplat e
B
•v O D\O
1 10
3
459
o—•
, 5
10 Number of cycles
,,,,1 10 6
,
,,,,,,,1 107
Figure 10-18. Effect of specimen orientation on fatigue life of textured Ti-4A1-4V alloy (see text) (after Larson and Zarkades, 1976).
460
10 Measurement and Control of Texture
Eighth Texture Conference, in 1988, have some 300 pages on the interpretation of deformation texture but only two papers on the fracture toughness of textured hexagonal metals, neither very specific. 10.3.4 Micro textures and Mechanical Properties All that has been said in the three preceding Sections refers to macrotextures. Mechanical properties can also be analysed in terms of microtextures (mesotextures), as we saw in the discussions in Sec. 10.2.4, and the role of special boundaries was highlighted there. The reader is referred back especially to the account of Watanabe's notion of grainboundary design and the effect of special boundaries in enhancing fracture toughness and ductility. - It might prove beneficial to apply similar examination to the ductility of hexagonal metals and to attempt grain-boundary design on such metals! Studies of corrosion and stress-corrosion behavior of polycrystals, including the effect of attack by liquid metals, has generally shown that low-angle grain boundaries and twin boundaries are relatively immune to attack of these kinds. A recent study by Ortner and Randle (1989) has examined the sensitisation, by aging at 650 °C, of austenitic stainless steel to stress-corrosion cracking. They found that, while small-angle (and annealing twin) boundaries are not sensitised by this treatment, large-angle boundaries are sensitised, irrespective of whether they are general or CSL type. 10.3.5 Magnetic Properties and Texture Soft magnetic alloys (i.e., those which have high permeability and low coercive field) behave quite differently according to
the relative orientation of the exciting field and the crystal axes. This has been particularly thoroughly examined for b.c.c. Fe-Si alloys which are used in large tonnages for transformer laminations. Texture in such alloys is of crucial practical importance. The essential facts are displayed in Fig. 10-19. Fig. 10-19 a displays magnetization curves for iron crystals magnetized along [100], [110] and [111], determined by R. M. Bozorth many years ago.
(a) 2000
8 1600 D a en
1200 €
800
200
400 H(0e)
800
f
(b) 10
600
:
M-19 (non oriented) -(0.36mm t h i c k ) /
: JV^
///^^^ -
l^^thwt%
Si-Fe -
^ ^ - - - ^ , ( 0 . 0 6 mm thick)
< 1.0 >
XLO /
/
Metglas R 2605 S2 - 0.1 .g (0.03 mm thick) =
0.1 -
Exciting power Core toss
0.01 0.6
0.01
0.8
1.0
1.2 U 1.6 Induction B in T
1.8
2.0
Figure 10-19. (a) Magnetization along different crystallographic axes in single crystals of pure iron, (b) Core loss and exciting power vs. induction at 60 Hz for annealed grain-oriented Fe-6.5mass% Si, nonoriented Fe-3.5mass% Si alloy (M19), and Metglas 2605 S2 (after Das et al., 1985 a).
10.3 The Influence of Texture on Properties
Clearly, magnetization along the 'easy' [100] direction is by far the most favorable, because of the very high permeability and consequent very small hysteresis loss. The power (core) loss in a transformer is caused partly by the hysteresis loss and partly by eddy currents: the latter is reduced by enhancing the resistivity of the alloy, which is achieved by alloying with silicon (although that is not the sole reason for alloying with that element), the former is reduced by achieving a macro texture. The effect of a texture in reducing core loss is shown in Fig. 10-19 b (together with core loss in metallic glass laminations, which are beginning to displace even the best FeSi crystalline laminations) (see Vol. 9, Chap. 9, Sec. 9.6.1). A transformer lamination will normally encounter magnetizing fields predominantly in two directions at right angles, constituting a closed field circuit. Clearly, a (100) [001] texture would be ideal, since then both orthogonal directions are easy directions. Such a texture can be obtained (see Sec. 10.4.3) but only at unacceptable cost. What is always sought in practice is the Goss texture or cube-on-edge texture, named after its originator (Goss, 1934). This is (110) [001]. Here the major part of the field circuit is parallel to [001] but a minor part is along [110], which is not so favorable. For many decades the standard transformer laminations contained 3.03.5 mass % Si, but the challenge offered by metallic glass led to efforts to manufacture crystalline sheet containing as much as 6.5 mass % Si, still with texture; this alloy has a considerably higher electrical resistivity and thus reduced eddy currents. 3 % Si is sufficient to suppress the oc/y transformation and thus allows the high-temperature anneals needed to create the Goss texture. Higher silicon contents used to
461
produce highly brittle alloys which could not be rolled, but rapid solidification procedures have successfully generated ductile high-silicon material. The production methods used to make textured Fe-Si alloys are examined in Sec. 10.4.2. Permanent magnets, especially the new N d - F e - B magnets, are also highly anisotropic magnetically and to achieve the highest possible remanent magnetization, texture has to be achieved in the polycrystalline product. This is generally done by magnetic annealing, discussed in Sec. 10.4.6. 10.3.6 Other Properties
Irradiation or thermal cycling of nuclear fuels has a range of effects, one of which may be shape distortion. This was a serious problem in early fission reactors and is still a matter of concern with reactors such as the 'Magnox' generation which use metallic uranium fuel. Metallic uranium has an orthorhombic unit cell, and rodshaped single crystals of this metal behave very differently under neutron irradiation in the range 0-200 °C, according to the orientation: irradiation to 0.1 % 'burn-up' at 100 °C creates a length reduction of ^ 4 2 % if the axis is [100], a 42% length increase for [010] but no change if the axis is [001] (Mclntosh and Heal, 1960). (The phenomenon is caused by a complex redistribution of point defects under irradiation, the details of which are not important in the present connection.) - Since rolling and extrusion textures in uranium are apt to have [010] along the rolling direction (Holden, 1958; Dillamore and Roberts, 1965), polycrystalline uranium grows unacceptably if it has even quite a weak texture, and only randomly oriented material behaves acceptably because of the statistical cancellation of the growth or shrinkage
462
125
10 Measurement and Control of Texture
A-300°C B-600°C C-300°C D-300°C
rolled rolled rolled, beta-quenched rolled, beta-quenched, recrystallized
| 75 o 50
25
0.25
0.50
0.75 1.00 1.25 Burn-up (at.%)
1.50
1.75
Figure 10-20. Effect of different fabrication histories on the lengthwise growth of metallic uranium under neutron irradiation (after Mclntosh and Heal, 1960).
of individual grains. Fig. 10-20 shows this, and also shows that beta-quenched metal shows low, acceptable growth rates. Such metal has been heated above the oc/(3 transition temperature, 662 °C; p-uranium has a quite distinct, tetragonal structure and the transition back to the orthorhombic structure generates a random orientation distribution. - Fuel elements thermally cycled within the orthorhombic stability range also undergo a form of growth because of the highly anisotropic thermal expansivity, by a ratcheting mechanism, unless the metal has a random grain orientation distribution. Irradiation-enhanced creep of uranium, another important process, is also sensitive to texture. - It was this kind of problem, so closely related to texture, that stimulated the invention of orientation distribution functions by Sturcken and Croach 28 years ago (see Sec. 10.2.3). Another unexpected function of texture is in the production of electrolytic capacitors. These consist of layers of aluminum foil, successively anodes (which are oxidised) and cathodes, separated by layers of fluid electrolyte which assures that the cathodes effectively reach out to be in con-
tact with the free side of the oxide coatings on the anodes. It is the oxide which functions as dielectric. To achieve a good capacitance per unit volume of foil stack, it is important to enhance the effective surface area of the anode foils by etching. It is found that a dramatic increase of surface area, typically 70-fold, can be achieved by tunnel etching', the etchant follows [100] directions in the aluminum lattice, and to achieve the best results, the fine [100] tunnels must penetrate normal to the foil surface . . . i.e., the foils need to have a cube texture, (001) [100]. Considerable ingenuity has been applied to get an optimum texture for a maximum capacitance per unit volume and Ibe (1988) has described in detail how this can be achieved. Copper foils used for printed circuit boards also require special etching methods, so as to achieve very finely defined removal of narrow channels of metal during the preparation of miniaturised circuits. This requires anisotropic etching, of the kind made familiar by the etching of silicon crystals. It turns out that under some circumstances textured copper polycrystalline sheet with (111) planes preferentially parallel to the surface favor anisotropic etching of the kind that is required (Nelson et al., 1988). 10.3.7 Effect of Texture on the Measurement of Residual Stress
X-ray diffraction is routinely used to determine the intensity and nature of the residual (internal) stress in a polycrystal. The standard approach is the so-called sin2 W method, described in a number of publications (Barrett and Massalski, 1966, pp. 466-485; Macherauch, 1966; Jones, 1986). Fig. 10-21 shows the essential geometry: a series of back-reflection diffraction patterns is taken (photographically or by dif-
10.3 The Influence of Texture on Properties
463
1986) but the situation is different if there is a strong texture, when the d/sin2W plot may become curved. This arises, for instance, with electroplated deposits which generally have both high residual stresses and strong fiber textures: the result is that determination of residual stresses by X-ray diffraction becomes difficult (see Chap. 11, Sec. 11.3.2). A (Fig. 10- which are uniform on a scale much larger than the grain diameter. This difficulty is 21) to determine both principal stresses not specifically linked with the presence of and their directions in the surface. In prac2l a texture. - The problem as it pertains tice, plots of d vs. sin P prove to be accuspecifically to textures has more recently rately straight in most instances (Jones, been analysed both theoretically and experimentally by a number of authors; the most helpful analysis is probably that by Dolle and Cohen (1980). Their analysis, like others in recent years, focuses on the problem of elastic and plastic anisotropy, especially the former. The "X-ray elastic constants" vary with W and (/> in a material with strong texture, because grains with different orientations are sampled at each such tilt (a fact which is at the heart of Greenough's analysis also). Interestingly, calculations of these effective elastic constants as they pertain to X-ray measurements show that Figure 10-21. Diagrammatic representation of the the X-ray elastic constants for (/z00) and sin2lF method, assuming a biaxial state of residual (hhh) reflections are independent of W and stresses. The X-ray beam is directed along the vector <j), and linear d vs. sin2lP plots are again XO; in one series of experiments W is varied while 4> is kept constant; for another series is changed and obtained even in the presence of texture. Fig. 10-22 shows some experimental rea series of photographs is again taken with varying XP.
fractometer) with varying incidence angles (see caption); a high Bragg angle is used and the spacing, d, of a particular family of lattice planes roughly normal to the X-ray beam is determined. Theory shows that, ideally, the following equation should be obeyed by the lattice spacings for a series of different W values:
464
10 Measurement and Control of Texture
Figure 10-22. Lattice strain, 8 33 , versus sin2lF, for 75% rolled mild steel sheet, in different azimuths,
suits on low-carbon mild steel cold-rolled to 75 % reduction (all experimental studies of this problem have been done with mild steel); the strain s33 is plotted instead of d, which is equivalent to the usual plot. These plots confirm the above important generalization. The results show two things: (1) residual stresses can still be determined in the presence of a strong texture - although not necessarily the complete stress state by selecting appropriate reflections (which will in general mean selecting an appropriate X-ray wavelength to ensure that the chosen plane reflects at a sufficiently high Bragg angle); (2) even if unfavorable lattice planes are chosen, there will generally be a 4> angle for which d will be proportional to sin2W. A recent theoretical analysis of the problem (Brakman, 1988) sought to achieve a rigorous analysis by incorporating ODF's to represent the actual (rather than an idealised) texture, but the difficulties proved insuperable.
However, more recently still, Behnke and Hauk (1991) have succeeded in interpreting the varying internal stresses in populations of differently oriented grains in brass; there stresses were followed by X-ray diffraction, with precautions as explained above, and compared with stress values calculated from known ODFs for the material. A recent survey (Krawitz and Holden, 1990) treats the use of neutron diffraction to determine residual stresses. Neutrons penetrate much deeper than X-rays and thus average the texture over a much larger volume. This penetration also causes complications because the bulk residual stress systems have three principal stresses rather than two, as do surface states. Nevertheless, it was asserted that by an appropriate selection of (hkl) and incidence angles, calculated with reference to the texture, internal stresses can be successfully determined by this method even in the presence of texture.
10.4 The Control of Texture
10.4 The Control of Texture In this section, attention is focused on the various approaches by which textures have been controlled in pursuit of desired properties, especially mechanical and magnetic ones. All involve the use of heattreatment, and accordingly the various processes of recrystallization and grain growth are of particular concern. Little is said here about the fundamentals of recrystallization and grain growth, since these issues are thoroughly covered in Chapter 9 of this volume. 10.4.1 Hot and Cold Rolling; Primary Recrystallization
An enormous amount of systematic work has been done on the effect of a range of metallurgical variables on deformation and annealing texture formation in steels, formable steels in particular, and in aluminum alloys, because of the great industrial importance of these two alloy families. In this Section, we shall concentrate on aluminum alloys (with a sidelong glance at copper also) because steels are very thoroughly covered in Volume 7. Variables such as rolling temperature; rolling reduction; annealing temperature; continuous annealing; carbon, phosphorus, nitrogen, aluminum content; low metal alloy additions to steels, as they influence formability and textures, are covered in that volume, in Chap. 6, Sees. 6.4-6.7, and in Chap. 7, Sees. 7.3-7.5. Current ideas on mechanisms that determine annealing textures are covered in Chap. 9 of this volume, esp. Sec. 9.9, and also in Vol. 7, Chap. 6, Sec. 6.4.2. Agreement is steadily hardening that oriented nucleation and orientation-dependent growth rates both have major parts to play in determining the final annealing texture, especially in
465
steels, and a very recent paper by one of the acknowledged leaders in this field (Hutchinson, 1989) comes to the conclusion that "the texture evolution is interpreted in terms of a compromise between the frequencies of potential nuclei and their respective growth velocities". Aluminum alloys in industrial use are divided into non-hardenable and hardenable, or high-strength, categories (see Vol. 8, Chap. 3). Their texture characteristics are rather different, and each category has recently received an outstanding review, by Hutchinson and Ekstrom (1990) for textures of non-hardenable alloys and by Bowen (1990) for those of high-strength alloys. An important conference proceedings was also devoted some years ago to textures in non-ferrous metals and alloys, and the bulk of this (Merchant and Morris, 1985) is devoted to textures of aluminum alloys and their control. Non-hardenable alloys, such as Al-Mg and A l - M g - M n - F e - S i types, are widely used for deep-drawing and thus r-factors and earing behavior are paramount. Much effort has gone into the proper description of earing and into attempts to predict this from known textures and such prediction is now feasible with a useful degree of accuracy (Rodrigues and Bate, 1985). The rolling textures of this kind of alloy are generically depicted in terms of an ODF in Fig. 10-23, which actually refers to pure aluminum. The texture 'tube' extends from the Goss texture at lower right, via the brass texture, past an intermediate texture known as " S " to the copper texture. (We have already met the brass and copper textures in Figs. 10-4, 10-9 and 10-10.) The use of ODFs allows the incidence of these various texture components to be accurately compared, and it also allows the sharpness of the texture to be assessed accurately. For instance, it is established that
466
10 Measurement and Control of Texture
-Goss
Figure 10-23. ODF of texture of heavily cold-rolled aluminum, showing the characteristic 'tube' (after Hutchinson and Ekstrom, 1990).
the rolling texture is sharper for a given strain when the initial grain size is fine than when it is coarse, because coarse grains evince a higher level of strain heterogeneity. The annealing textures in non-hardenable aluminum alloys are complex, and generally contain a measure of residual deformation texture (itself complex), some Goss texture and some cube texture. Hutchinson and Ekstrom refer to a "wide flora of different annealing texture components", and as has become usual, they accord roles to both oriented nucleation and oriented growth. A crucial variable, to which much attention has been paid, is the role of secondphase particles; both their concentration and their size distribution affects texture (Hutchinson and Ekstrom, 1990). In general, unless they are very fine, particles are surrounded by a "turbulent" zone which favors the retention of a nearly random deformation texture in those parts of the
micro structure. Fig. 10-24 illustrates this: for a 3004 (Al-Mg-Mn-Fe-Si) alloy in different states of pre-treatment, as the concentration of large second-phase particles increases, three texture components each progressively weaken and the random component ist strengthened. To minimize earing in this alloy in its final state, it is necessary to create a strong cube texture at the hot-rolling stage; on subsequent cold rolling, the 0/90° earing tendency associated with the cube texture gradually gives way to a 45° tendency resulting from the deformation-induced reorientation of texture and when the two are properly played off against each other, earing virtually disappears. To ensure that cube texture can form at the hot-rolling stage, second phase particles must be rigorously controlled. The procedure by which earing can be eliminated can be gauged from a schematic diagram, Fig. 10-25, which shows that with the right pretreatment before cold-
500
50 1000 1500 Number of supercritical sized particles (mm2)
Figure 10-24. Variation of various texture components in hot-rolled and annealed 3004 sheet as the frequency of large (> 3 urn diameter) second phase particles increases (after Hutchinson and Ekstrom, 1990).
10.4 The Control of Texture
rolling (followed by a final anneal), earing can be eliminated. This diagram is not specific to any particular alloy. Much attention has been devoted to the generation of the cube texture on annealing; as we have just seen, it is of practical importance as well as of purely scientific fascination. Apart from its importance in the procedure leading to earing control, it is also sometimes desirable to finish up with a final cube texture, as in pure aluminum used for the production of condenser foil (see Sec. 10.3.6). Rodrigues et al. (1985) survey the influence of cube texture, desirable and otherwise, on the technological properties of aluminum sheet and foil. - Hutchinson and Ekstrom (1990) conflate a large and sometimes confusing literature by making the following points: cube texture, (100) [001], is present on a small scale in the deformation structure, in the form of fine transition bands separating large volumes of the major deforma-
Microstructure before cold rolling (T) Fine subgrains (2) Coarse subgrains
0/ 90c
o c
(3) Partially recrystallised © Fully recrystallised 2nd generation cube nuclei
©
'a
o UJ
Cold rolling reduction
Figure 10-25. Schematic diagram showing the effects of initial microstructure and cold rolling reduction on earing behavior, as measured after a final anneal (after Hutchinson and Ekstrom, 1990).
467
tion texture components (as seen in Fig. 10-23). This very small volume fraction of cube texture in the as-rolled texture, identified by TEM, is not sufficient to show up in the ODE This cube texture component has a fairly favorable growth rate during annealing, but its growth is apparently inhibited unless the transition bands which serve as recrystallization nuclei are smooth and continuous. "Turbulent" flow, as arises when second-phase particles are prevalent, breaks up the transition bands and cube texture is then inhibited during subsequent annealing. This is why high-purity aluminum is best for achieving 100% cube texture. - A recent study by Kitagawa (1989) of texture formation in aluminum fabricated by the traditional pack-hammering method, long used to make gold foil (in which a pack of alternate sheets of foil and paper are beaten with a hammer) has elicited the unexpected finding that the initial rolling texture of the starting sheet is turned into a texture with an almost perfect alignment of (100) with the surface. The most complex variable in affecting texture is the iron content in aluminum alloys, which in turn is interrelated with the silicon content (because of the formation of an AlFeSi phase, as well as the familiar Al3Fe). The literature is full of conflicting information. Hutchinson and Ekstrom demonstrate that, according to Fe level and Fe/Si ratio, iron can have five distinct effects on texture formation at the hot-rolling, cold-rolling and annealing stages. This topic is too complicated to go into here and the reader is referred to Hutchinson and Ekstrom's review. Bowen's (1990) survey of textures in high-strength alloys demonstrates, in several ways, the utility of ODFs in interpreting the fine details of textures. One example is shown in Fig. 10-26, which shows
468
10 Measurement and Control of Texture
16
U 12 1.0 Thickness (mm)
Figure 10-26. Variation in intensity of the major texture components as a function of depth, from the surface to the center-line of 1.6 mm thick recrystallized 8090 Al-Li alloy (after Bowen, 1990).
how the incidence of various texture components changes through the thickness of rolled and recrystallized 8090 Al-Li alloy sheet. The texture sharpness parameter, / , which we met in Sec. 10.2.3, can also be put to good use. Fig. 10-27 shows the texture sharpness parameter of metal matrix composites, in the form of a range of 8090 4.00Q
3.00 Q. O
2.00 2 mm sheet
/
1.00
6 mm plate
0.00 0.00
5.00
10.00 % SiC
15.00
20.00
Figure 10-27. Reduction of texture sharpness parameter, J, in a series of rolled metal matrix composites consisting of 8090 Al-Li alloy with various proportions of silicon carbide (after Bowen, 1990).
alloys reinforced with various volume fractions of silicon carbide, in the form of particles > 1 mm in diameter, and rolled (to an unspecified reduction). As we have seen above, the presence of relatively large dispersed particles leads to turbulent flow which weakens texture formation, and Fig. 10-27 shows this very clearly - especially for the first 2 vol. % of silicon carbide. This kind of information can only be obtained by use of ODFs. Another application of ODFs described by Bowen is the prediction of the in-plane variation of yield stress of rolled sheet as a function of orientation relative to the rolling direction; this can be calculated directly from the ODF coefficients. A further use is in connection with superplastic forming of aluminum alloys, an important industrial process (see Vol. 6, Chap. 9): it was shown that when the starting material has elongated grains as a consequence of unidirectional rolling, the texture changes sharply in the early stages of superplastic forming because at this stage, dislocation slip is the main mechanism. Later, when true diffusional (superplastic) flow sets in, grains become equiaxed and the texture ceases to change. For this kind of study, again the / parameter has proved a useful mode of summarising experimental findings. Cube texture was intensively studied in rolled and annealed copper and copper alloys before it became of interest in aluminum, and the mechanism of its formation has long been a metallurgical puzzle (see, for instance, Barrett and Massalski, 1966, pp. 570-572). A detailed study by Ridha and Hutchinson (1981) first showed the operation of thin transition bands in the rolled structure as nucleation sites (as later confirmed in aluminum), and also established the reasons, in terms of dislocation geometry, why these sites were able to
10.4 The Control of Texture
recover rapidly to the point where they became recrystallization nuclei. Unfavorable changes in composition, initial grain size or rolling temperature favor the generation of coarse shear bands which inhibit the formation of the transition bands essential to cube texture formation. This long-standing issue in metallurgy seems now to be satisfactorily resolved. Texture control has proved important in a number of other alloys for specific loadbearing uses. An interesting example is the treatment of copper-based spring materials as used in telephone relays. Chin et al. (1969) showed that a range of spring alloys, including phosphor bronze, nickelsilver, beryllium-bronze and a modified cupronickel with tin, could all be improved by abnormally heavy rolling, to reductions as high as 97 %, followed by a low-temperature anneal (at 150-350°C). The rolling generates a sharp (110) [112] texture and imparts a very high yield strength; the annealing presumably (this was not reported) preserves a sharp texture and further increases the yield strength above the already high value generated by rolling. This last effect, if it exists, is to be attributed to a strong plastic anisotropy associated with a strong annealing texture. 10.4.2 Grain Growth and Secondary Recrystallization
This section is concerned with the microstructural changes which follow the completion of primary recrystallization. Normal grain growth (see Chap. 9, Sec. 9.6) involves the progressive increase of mean grain size, driven purely by the associated drop in total grain boundary energy, in a process which keeps the grain size distribution at any stage quite narrow; secondary recrystallization, also called abnormal or exaggerated grain growth, is a variant of
469
this, in which a few grains grow very large while the remainder hardly change in size and are eventually consumed by the rampant giant grains. Until recently, any possible change of the annealing texture due to the grain growth process was largely ignored, though it had been established experimentally that in low-alloyed iron, the Goss, (110) [001], texture is progressively sharpened during normal grain growth. The issue has now been clarified by Lucke and his coworkers (Brickenkamp and Lucke, 1981; Abbruzzese and Lucke, 1986). The former first showed with the help of ODFs that, in a-brass, a principal texture component disappeared during grain growth and was replaced by a quite different one. Abbruzzese and Lucke then formulated a theory of grain growth in the presence of texture, taking into account the dependence of grain boundary energy and mobility on misorientation, and were able to predict a change of texture and also a discontinuity in grain growth kinetics, with a drastic fall in the exponent n in the relation D2 — Dl = Ktn (where D, Do are the grain diameter at time t and the initial grain diameter, respectively) at an early stage in the process, which was experimentally confirmed. This theoretical treatment was further developed by Abbruzzese et al. (1988) and compared with observations on grain growth in an AM % Mn alloy, heavily rolled and annealed. Again, good agreement was obtained between model and observations for the changes in the proportions of the texture components and in the changes of grain growth kinetics, in the early but not the late stages of grain growth. In this last study, the authors also succeeded in determining the partial grain size distributions of the different texture components and compared this with theory also.
470
10 Measurement and Control of Texture
A recent study has claimed that, in general, alloys in which recrystallization and grain growth is controlled by a dispersion of precipitates tend to preserve the deformation texture, unlike precipitate-free alloys. It remains to be seen how well this broad generalization stands up to further study (McQueen and Mecking, 1987). Microtextural techniques have also been applied to the problem, by Randle and Brown (1988 and 1989). They found that in the course of grain growth, the proportion of special low-energy (CSL) boundaries increased steadily; in a linked set of experiments, they studied grain growth in samples of stainless steel which had been lightly re-strained after primary recrystallization was complete and before grain growth had got under way. Under these circumstances, the incidence of CSL boundaries was particularly high (notably so at the free surface) and this led to secondary recrystallization. In the author's view, a proper understanding of secondary recrystallization, which is still elusive, will come from studies of this type. Discussions of secondary recrystallization will be found in treatments by Cahn (1983) and by Detert (1978). Essentially, the disruption of normal grain growth and the selective growth of a sub-population of the grains which constitutes secondary recrystallization, requires the inhibition of normal grain growth, and this can be brought about either by a distribution of dispersed particles, by working with thin foils where most grains are prevented from growing to a size exceeding the foil thickness, by the presence of impurities which segregate to and pin some boundaries but not others (this is where special boundaries become important), or through the creation of a suitable compound texture (in which case one texture component grows at the expense of the others). Randle and
Brown (1988 and 1989) have also shown that a small strain just before grain growth begins can induce secondary recrystallization. By far the most important application of secondary recrystallization is to the formation of coarse grains, several millimeters in size, with the Goss texture, (110) [001] in thin sheets of Fe-Si alloy to be used as transformer laminations (see Sec. 10.3.5). The technology of this process, as it was a few years ago, is concisely described by Luborsky et al. (1983). A number of different firms, in Japan and America, have developed somewhat different process strategies; however, all are exceedingly elaborate. All depend on the use of dispersed phases as grain growth inhibitor, mainly MnS, MnSe (+ Sb) and A1N. The inhibitor is removed at the end of the process by dissolving it in a glassy surface film which is also used as an electrically insulating separator between laminations. Developments have continued since the 1983 overview; for instance, Inokuti et al. (1985) have shown that Mo together with S, P, Se and Sb in the form of a surface coating forms a population of fine precipitates in the near-surface region after annealing, and the Goss texture forms a little way below the surface from a localised group of highly oriented primary grains, as shown in the color photographs of Fig. 10-11, above. - The role of the surface coatings used in the processing of transformer laminations is more complicated than indicated above; in addition to acting as a solvent or as a source of precipitates, or both, it appears that constituents of such coatings also diffuse along primary grain boundaries and thereby affect the evolution of texture (Gangli and Liicke, 1988). A kind of microtextural approach has been made to the process of Goss texture generation by secondary recrystallization
10.4 The Control of Texture
in Fe-3mass% Si alloy, by Shimizu et al. (1990). They did not actually determine the orientations of individual grains, but computed the probability of different specific nearest-neighbor misorientations by a statistical process, starting from experimental ODFs. They postulated that the product of the intensity of a given orientation and the expected frequency of highly mobile CSL boundaries that this orientation is expected to have with all its neighbors has to exceed a critical value, and on this basis rationalised the generation of the Goss texture. More recently, as explained in Sec. 10.3.5, oriented laminations of Fe6 mass % Si alloys have been prepared, using rapid solidification techniques to induce ductility. Sometimes such sheets are rolled, but there have also been experiments to modify texture by annealing of the as-solidified material. (There have been several reports of fiber textures in meltspun ribbons of various metals.). It was found (Cunha and Johnson, 1990) that abnormal grain growth took place (even though there had been no mechanical working at any stage) and the major texture component in the coarse grains was (100) [0 k I], which in fact was a constituent of the as-cast structure. A process which is much used in the production of superalloy components for aero engine components, turbine blades in particular, is directional recrystallization; the object is passed slowly through a steep temperature gradient or a narrow hot zone, and the result is the formation of very elongated, thin grains which is a favorable geometry for creep resistance. A recent study has been performed with an ODS (oxide dispersion strengthened) alloy, MA6000, in a hot-extruded form (therefore recrystallized during extrusion). On zone-annealing, directional secondary
471
recrystallization to very long, « 1 mm wide grains took place, with a strong [100] texture (Marsh and Martin, 1991). Fiber textures like this are normal in directional recrystallization.
10.4.3 Tertiary Recrystallization
A process called surface-controlled secondary recrystallization or, alternatively, tertiary recrystallization, was discovered independently by Detert (1959) and Walter and Dunn (1959). When a thin (<0.1 mm thick) cold-rolled sheet of Fe-3mass% Si alloy is recrystallized at a high temperature, 1000-1200 °C, it sometimes develops coarse (secondary) grains with a cube texture (as opposed to the normal Goss texture). This only happens if the annealing atmosphere contains some oxygen. When such a cube texture has been formed, if the annealing is continued but the atmosphere is changed to very dry, pure hydrogen, or vacuum, the cube texture disappears and is replaced by the Goss texture. This change can be repeatedly worked, backwards and forwards. It has been firmly established that this happens because of an anisotropy of surface energy: when the strip is clean, (110) has a lower specific surface energy than does (100), while in the presence of adsorbed oxygen, the reverse is true. This anisotropy can only be effective for very thin sheet, because otherwise the role of the grain boundary energies swamps the surface effect. The detailed facts are set out by Dunn and Walter (1966) and by Cahn (1983, pp. 1662-1665). This process has never been applied to produce cube-textured transformer laminations, probably because the critical thickness is too small to be convenient for such laminations and the cost of the necessary rigorous process control puts the tech-
472
10 Measurement and Control of Texture
nique beyond the range of economic feasibility. Earlier, McLean and Mykura (1965) had discovered a rather similar process of tertiary grain selection in highly pure platinum annealed at 1500°C: grains with (111) planes parallel to the surface grew at the expense of others. Dunn and Walter (1966) analysed these experiments, especially the amount of deviation from ideal orientations, in some detail. 10.4.4 Transformation Textures
Some forms of phase transformation entail an orientation relationship between parent and daughter phases, either through the mechanism of martensitic transformation, via epitaxial precipitation of Widmanstatten plates from a matrix, or some other mechanism. This implies that a textured polycrystal of the parent phase should generate a textured polycrystal of the daughter phase. However, in general the daughter texture should be quite different from, and less sharp than the parent texture, because there are usually multiple variants of the daughter orientation resulting from a single crystal of the parent phase. Typically, a cubic crystal undergoing a martensitic transformation is expected to generate up to 24 crystallographically equivalent variants of the daughter phase. This relationship between parent and daughter textures has been extensively examined for metal systems. An early example was a study by Davies et al. (1976) of an oc/p brass. This was treated at high temperature, 870 °C (which leads to a singlephase (3 structure) so as to generate a texture, quenched in the single-phase form so that the ODF could be determined, then heat-treated at an intermediate temperature to generate, by a diffusion-controlled
transformation, the a/(3 duplex structure. Renewed texture determination showed that the a phase had formed from the |3 according to the classic KurdyumovSachs orientation relationship. More extensive studies have been undertaken with Fe-30mass% Ni alloy which undergoes a martensitic y(f.c.c.)-a(b.c.c.) transformation, but can, like brass, be quenched to preserve the y phase metastably at room temperature before it is transformed to a by further cooling to liquid nitrogen temperature. In this way, both the y and the oc textures can be determined. Alternatively the material can be rolled at room temperature, then heated and quenched to metastabilize the high-temperature phase. The close orientation relationship between the two textures can be seen in Fig. 10-28. Such studies can also cast light on "texture memory" in the heattreatment of industrial steels which undergo 0C-7 transformations (see Vol. 7, Chap. 6, Sec. 6.4.4). Studies of iron single crystals subjected to a oc->7-»oc cycle showed that the final texture and grain-boundary character distribution was interpretable on the basis of a Kurdyumov-Sachs orientation relationship between a and 7 (Harase et al., 1990). Elaborate theories to predict the daughter texture in such a transformation have been proposed, depending on an application of the possible transformation variants to the starting texture as expressed by an ODF. When this is done (Wagner et al., 1981; Humbert, 1988) it is found that a daughter texture computed on the assumption that all martensitic variants are equally probable leads to poor agreement between predicted and observed daughter textures. Agreement can be obtained if a minority of possible martensitic variants are assumed to be favored. This is quite reasonable, because martensitic transfer-
10.4 The Control of Texture
473
RD (a)
Figure 10-28. (a) (110) a
pole figure of Fe-30mass% Ni sheet, rolled to 67 % reduction in the b.c.c. (a) phase, (b) (lll) y pole figure of the same material after heating to 500 °C to transform it to the f.c.c. structure (y), followed by quenching to room temperature (after Esling and Bechler-Ferry, 1982).
mations are accompanied by local stress fields resulting from the martensitic distortion of the parent lattice, and these stress fields back-react to inhibit some variants (see Vol. 5, Chap. 6, Sec. 6.3.6). (It was demonstrated long ago that an external stress applied during a martensitic transformation reduces the range of variants which form, sometimes to a single one (Kochendorfer and Muller, 1955).) - A study of the reverse, oc->y transformation (Liu et al., 1988) shows that at austenitization temperatures only slightly above the transition temperature, the original coldrolled y texture is restored . . . i.e., textures prove reversible. This is no longer true for higher austenitization temperatures. A recent study of texture memory in NiCo alloys (Chapellier et al., 1990) has shown that in this family of alloys, there is no variant selection. Hot-rolling textures in titanium deformed above 880 °C, when the metal has a b.c.c. structure, followed by cooling to room temperature which leads to a martensitic transition to a h.c.p phase, show again that the texture is predictable from the known martensitic relationship under the proviso that there is variant selection (Inagaki, 1990). Similar texture "memory" has been observed in both oxidation and reduction re-
actions. In general, chemical reactions in which the product phase has one or more defined orientation relationships to the parent phase are termed topochemical reactions', these have been surveyed by Thomas (1974). The epitaxy of oxide layers on metals is a large subject and cannot be reviewed here in detail. One of the most famous sets of experiments are those due to Gwathmey, who oxidised spherical copper crystals and examined the orientational dependence of the oxidation process. This work was reviewed by Gwathmey and Lawless (1960), and later work was covered by Bardolle (1975). - Only one study seems to have been done on the textural implications of the reduction of a metal oxide. This was the work of Revcolevschi and Dhalenne (1985): they generated a lamellar eutectic consisting of NiO and ZrO 2 and heat-treated it to reduce the NiO while leaving the ZrO 2 unaffected. The two oxides initially are exactly oriented with respect to each other, so that the eutectic is in effect a giant interleaved monocrystal. It was found that the metallic nickel resulting from the reduction remained in parallel orientation with respect to the NiO from which it was formed. It was not clear whether the ZrO 2 acted as a template or merely as an inert separator. Up to now, such studies have been per-
474
10 Measurement and Control of Texture
formed with single crystals rather than polycrystals, so textures in the narrow sense have not been examined. However, textures have entered into a series of studies of the topochemical reduction of polycrystalline haematite, Fe 2 O 3 , to magnetite, Fe 3 O 4 , en route to the totally reduced metallic state. Here again, clear texture memory is found. For details, the reader is referred to the latest review of inherited texture in phase transformations, by Humbert (1988). 10.4.5 Textures in Intermetallics
There is currently an intensive research effort, worldwide, to make intermetallic compounds {intermetallics for short) into useful high-temperature materials (see Vol. 8, Chap. 9). The principal difficulty with most of these materials is brittleness, certainly at room temperature and often at all temperatures. The most widely investigated category of intermetallics consists of the aluminides, notably Ni 3 Al, NiAl, FeAl, Fe3Al, Ti3Al, TiAl, TiAl 3 , Ti 2 AlNb, Ti2NiAl. These phases are investigated both in single-phase and multiphase forms. - It has long been known that the formability of metals and alloys of intrinsically limited ductility such as magnesium, zirconium, zircalloys, etc., is greatly affected by texture and so is strength (see Sees. 10.3.1 and 10.3.3). It is therefore surprising that up to now, so little research has been done on textures of intermetallics and their effect on plastic characteristics. One reason, no doubt, is precisely the limited deformability of many of these materials, so that even in hot-working, ductility is not sufficient to allow strong textures to be developed. What little has been done is all of very recent date. Khadkikar et al. (1990) have examined textures of NiAl (B2 structure type), FeAl
(B2) and Ni3Al (Ll 2 ) in hot-extruded form. Weak [111] fiber textures were observed in the first two when extruded from powder, but FeAl when extruded from cast material had a [110] texture. Ni3Al showed nearly random orientation (a most unusual characteristic!). There was evidence that the extruded textures were actually annealing textures and so could not be interpreted in terms of high-temperature slip crystallography. Recrystallization textures in 2-phase /(Ni 3 Al) + y (disordered) alloys have been examined by Inoui and Inakazu (1991). The hexagonal a2 phase Ti3Al, in Ti-Al alloys has been much studied over the past decade, notably in alloys with ternary additions of niobium (Christodoulou, 1990). The influence of texture on the plastic deformation of a Nb-bearing oc2 alloy has very recently been examined (Hon et al., 1991). The alloy was hot-rolled and annealed, and a sharp (0001) basal-plane texture was established. The thickness reduction was found to be much less at all testing temperatures than the width reduction . . . i.e., the rm factor was enhanced. The yield stress is somewhat reduced but the ductility significantly improved in comparison with a 2 alloys with or without niobium which did not have a texture. This appears to be the first account of oc2 textures in the open literature. The tetragonal y (Ll 0 ) phase, TiAl, deforms largely by twinning (Christodoulou, 1990). It is known from studies of hexagonal metals like zirconium and magnesium that texture strongly affects the ability to twin in unidirectional plastic deformation. A beginning has been made by a study of texture of hot-compressed TiAl, and a double [202] + [220] fiber texture was observed. This texture was believed to have formed as a result of dynamic recrystallization. Twins were observed in the resulting
10.4 The Control of Texture
microstructure but it was not clear whether they were deformation or annealing twins. Textures in this type of alloy deserve much more detailed study, with the use of a variety of modes of hot deformation. - The relation between plastic anisotropy, slip mechanisms and the hot-rolling texture of y-TiAl has recently been examined in detail by Hartig et al. (1992). 10.4.6 Magnetic and Stress Annealing
Permanent magnet alloys are routinely treated to maximise the macroscopic magnetic anisotropy of the material and thus optimise their magnetic properties as permanent magnets. This has long been done with the important Alnico family of alloys (which, in spite of their name, contain Fe, Cu and Ti in addition to Al, Ni and Co). These alloys separate out into a two-phase structure by spinodal decomposition and the two phases are usually elongated along the three <100> directions. Annealing in a magnetic field, below the Curie temperature of at least one of the phases, causes preferential growth of the [100] direction most nearly aligned with the field. The alignment can be strengthened by using directional solidification before magnetic annealing. Similar techniques are used for the more recently developed family of F e - C r - C o alloys, which also depend on spinodal decomposition in the Fe-Cr miscibility gap. All this is outlined by Luborsky et al. (1983, pp. 1689-1691). The new Nd-Fe-B magnets, which are the strongest permanent magnets known, also depend upon alignment of individual grains in the end-product to ensure that the excellent properties of individual grains 'pull together' in the polycrystalline product. The methods used are analysed in Buschow's (1986) definitive review. One way to align the grains is to start with pow-
475
der and align them in a magnetic field, press them sufficiently to keep them aligned and then sinter. An alternative method, used with meltspun and fragmented flakes, is to "die-upset" (a form of hot-pressing without use of a die), which mechanically aligns the particles in the correct orientation (Fig. 10-29). The effect seems to be purely mechanical and is connected with the flake shape of the particles (see Buschow's review for details). A very recent report (Dulis, 1990) indicates that similar results can be achieved with atomized (roughly spherical) alloy powder by using hot extrusion. Magnetic annealing has recently been applied to a high-ternperature superconductor, even though this is not ferromagnetic or even ferrimagnetic (de Rango et al., 1991). The anisotropy which is exploited here is only paramagnetic and thus very weak. It is necessary that the anisotropy energy associated with the interaction of the paramagnetic anisotropy with the \.z 1.0 0.8 -
0.4
„
0.2 0 -1600
/
-1200
I
-800 H (kA/m)
I
-400
Figure 10-29. Room-temperature demagnetization characteristics of die-upset Nd 0 13 (Fe 0 9 5 B 0 0 5 ) 0 parallel (||) and perpendicular (_L) to the press direction during die-upsetting (after Lee et al., 1985, reproduced by Buschow, 1986).
476
10 Measurement and Control of Texture
aligning field is large enough, at the annealing temperature, to swamp kT. The objective was to align the grains in polycrystalline YBa 2 Cu 3 O 7 . Starting from a somewhat off-stoichiometric composition, the ceramic was heated to 1050 °C at which temperature part of the material is molten. The mixture was cooled in a 5 Tesla magnetic field and cooled at 20°C/h while the desired phase was formed by a peritectic reaction. X-ray analysis showed that the resultant grains hat their c-axes strongly aligned with the field. - Presumably the objective of this work, which was not stated, was to exploit the texture to maximise the critical current, since the superconducting parameters of YBa 2 Cu 3 0 7 are very anisotropic with respect to the crystal axes. The authors point out that this technique may prove to be widely applicable to paramagnetic materials. It has recently been discovered (Atzmony et al., 1987) that bimetallic (Ni/Cu) multilayers made by electrodeposition (see Chap. 11, Sec. 11.4.3.1, and Chap. 8), which have strong fiber textures of several different types, according to deposition parameters, show a substantial magnetic after-effect in the nickel interlayers. The magnitude of the effect was strongly texture-dependent. The mechanism of this after-effect, which has not previously been observed in nickel, is as yet obscure. It has also been reported that a magnetic field applied during electroplating can alter the resultant texture (Chiba et al., 1986). The study of the general effects of texture in metallic multilayers is as yet in its infancy.
10.5 Prognosis The measurement and control of macrotextures is a mature discipline, as can be gauged from the number of references cited from the 1950s and 1960s. It underwent
a renaissance through the invention and exploitation of orientation distribution functions, associated with fast modern diffractometers and computers. Macrotextures are closely linked with the control of formability and, to a lesser degree, with the anisotropy of strength. The concept of micro- or meso-texture is of far more recent date and investigators are still feeling their way in developing the definition, measurement, statistics and interpretation of this kind of texture. It is probable that future advances in the linkage of properties to texture and, in due course, in the control of textures for practical purposes, will stem from a better understanding of microtextures; this is likely to be the most productive area for textures research in the next few years.
10.6 Acknowledgement I am grateful to Sir Alan Cottrell for a critical reading of this chapter.
10.7 References Abbruzzese, G., Liicke, K. (1986), Acta Metall. 34, 905. Abbruzzese, G., Liicke, K., Eichelkraut, H. (988), in: ICOTOM — Eighth International Conference on Textures of Materials: Kallend, I S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 693-704. Adams, B. L., Morris, P. R., Wang, T. T., Willden, K. S., Wright, S. I. (1987), Acta Metall. 35, 2935. Arai, K. L, Ohmori, K. (1986), Metall. Trans. A 17A, 1295. Atkinson, M., McLean, I. M. (1965), Sheet Metal Ind. 42, 290. Atzmony, U., Swartzendruber, L. J., Bennett, L. H., Dariel, M. P., Lashmore, D., Rubinstein, M., Lubitz, P. (1987), J. Magn. Magn. Mater. 69, 237. Babel, A. W, Eitman, D. A., Mclver, R. W (1967), Trans. Amer. Soc. Mech. Engrs. 89, 13. Babu, S., Bhadeshia, H. K. D. H., Svensson, L.-E. (1991), J. Mat. Sci. Lett. 10, 142. Bardolle, J. (1975), in: Interfaces et Surfaces en Metallurgie: Martin, G. et al. (Eds.). Aedermannsdorf: Trans Tech, p. 483.
10.7 References
Barrett, C. S., Massalski, T. B. (1966), Structure of Metals, 3rd Edition. New York: McGraw-Hill. Beck, P. A., Hu, H. (1952), Trans. AIME 194, 83. Behnke, H., Hauk, V. (1991), Z. Metallkde. 82, 151. Bellier, S. P., Doherty, R. D. (1977), Acta Metall. 25, 521. Berger, A., Wilbrandt, P.-J., Ernst, R, Klement, U., Haasen, P. (1988), Prog. Mater. Sci. 32, 1. Bishop, J. W. R, Hill, R. (1951), Phil. Mag. 42, 414, 1298. Bleck, W, Lotter, U., Strassburger, C. (1988), in: Directional Properties of Materials: Bunge, H.-J. (Ed.). Oberursel: DGM Informationsgesellschaft Verlag, pp. 103-114. Bowen, A. W. (1978), Acta Metall. 26, 1423. Bowen, A. W. (1990), Mater. Sci. Tech. 6, 1058. Brakman, C. M. (1988), in: Theoretical Methods of Texture Analysis: Bunge, H.-J. (Ed.). Oberursel: DGM Informationsgesellschaft Verlag, pp. 377390. Brickenkamp, W, Liicke, K. (1981), in: IC0T0M-6Sixth International Conference on Textures of Materials: Nagashima, S. (Ed.). Tokyo: The Iron and Steel Institute of Japan, pp. 570-580. Bunge, H.-J. (1982 a), in: Quantitative Texture Analysis: Bunge, H.-J., Esling, C. (Eds.). Oberursel: Deutsche Gesellschaft fur Metallkunde, pp. 85128. Bunge, H.-J. (1982b), Texture Analysis in Materials Science - Mathematical Methods. London: Butterworths, pp. 88-90. Bunge, H.-J. (1982c), see Bunge (1982b), pp. 229230. Bunge, H.-J. (1982d), see Bunge (1982b), pp. 47-116. Bunge, H.-J. (1985), in: Textures in Non-Ferrous Metals and Alloys: Merchant, H. D., Morris, J. G. (Eds.). Warrendale: The Metallurgical Society, pp. 145-171. Bunge, H.-J. (1987a), Int. Mater. Rev. 32, 265. Bunge, H.-J. (Ed.) (1987 b), Theoretical Methods of Texture Analysis. Oberursel: DGM Informationsgesellschaft Verlag. Bunge, H.-J., Tobisch, J. (1968), Z. Metallkde. 59, 471. Bunge, H.-J., Tobisch, J. (1972), /. Appl. Cryst. 5, 27. Burns, R. S., Heyer, R. (1958), Sheet Metal Ind. 35, 261. Buschow, K. H. J. (1986), Materials Science Reports i, ICahn, R. W. (1951), J. Sci. Instr. 30, 201. Cahn, R. W. (1978), Phil, Trans. R. Soc. Lond. A 288, 159. Cahn, R. W. (1983), in: Physical Metallurgy: Cahn, R. W. (Ed.). Amsterdam: North-Holland, pp. 1630-1636, 1658-1662. Chapellier, P., Ray, R. K., Jonas, J. J. (1990), Acta Metall. 38, 1475. Chiba, A., Kitamura, K., Ogawa, T. (1986), Surf Coat. Techn. 27, 83.
477
Chin, G. Y. (1969), in: Textures in Research and Practice: Grewen, X, Wassermann, G. (Eds.). Berlin: Springer, pp. 51-80. Chin, G. Y, Hart, R. R., Wonsiewicz, B. C. (1969), Lecture presented to the Fall Meeting of the Metallurgical Society of AIME, unpublished. Chirkin, A. V, Al-Nakow, A. S., Sherif, S. M. (1991), /. Nucl Mater. 178, 27. Chojnowski, E. A., Cahn, R. W. (1973), in: Metallurgical Effects at High Strain Rates: Rohde, R. W, Butcher, B. M., Holland, J. R., Karnes, C. H. (Eds.). New York: Plenum Press, pp. 631-644. Christodoulou, L. (1990), in: Supplementary Volume 2 of the Encyclopedia of Materials Science and Engineering: Cahn, R. W. (Ed.). Oxford: Pergamon Press, pp. 1346-1354. Cunha, M. A., Johnson, G. W. (1990), J. Mater. Sci. 25, 2481. Das, S. K., DeCristofaro, N. X, Davis, L. A. (1985), in: Rapidly Quenched Metals (Proc. 5th Int. Conf): Steeb, S., Warlimont, H. (Eds.). Amsterdam: North-Holland, pp. 1162-1164. Davies, G. X, Kallend, J. S., Morris, P. P. (1976), Acta Metall. 24, 159. Detert, K. (1959), Acta metall. 7, 589. Detert, K. (1978), in: Recrystallization of Metallic Materials: Haessner, R (Ed.). Stuttgart: Dr. Riederer Verlag, pp. 97-109. Dillamore, I. L., Roberts, W. T. (1965), Metall. Rev. 10,211. Dillamore, I. L., Stoloff, N. S. (1969), in: Textures in Research and Practice: Grewen, X, Wassermann, G. (Eds.). Berlin: Springer, pp. 110-119. Dillamore, I. L., Smith, C. J. E., Watson, T. W (1967), Met. Sci. J. 1, 49. Dingley, D. J. (1984), Scanning Electron Microscopy 2, 569. Dimos, D., Chaudhari, P., Mannhart, J. (1990), Phys. Rev. B 41, 4038. Doherty, R. D., Cahn, R. W (1972), J. Less-Common Met. 28, 279. Doherty, R. D., Gottstein, G., Hirsch, H., Hutchinson, W. B., Liicke, K., Nes, E., Wilbrandt, P. J. (1988), in: ICOTOM - Eighth International Conference on Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale, PA: The Metallurgical Society, pp. 563-572. Dolle, H., Cohen, J. B. (1980), Metall. Trans. A 11 A, 831. Don, J., Majumdar, S. (1986), Acta Metall. 34, 961. Dons, A. L., Nes, E. (1986), Mater. Sci. and Tech. 2, 8. Duggan, B. X, Lee, W. B. (1988), in: ICOTOM Eighth International Conference on Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 625-630.7. Dulis, E. X (1990), Materials and Processing Report 5/7, 7.
478
10 Measurement and Control of Texture
Dunn, C. G., Walter, J. L. (1966), in: Recrystallization, Grain Growth and Textures: Margolin, H. (Ed.)- Metals Park: American Society for Metals, pp. 461-521. English, A. T, Chin, G. Y. (1965), Acta Metall. 13, 1013. Esling, C , Bechler-Ferry, E. (1982), in: Quantitative Texture Analysis: Bunge, H.-X, Esling, C. (Eds.). Oberursel and Paris: Deutsche Gesellschaft fur Metallkunde and Societe Francaise de Metallurgie, pp. 427-458. Ferran, G. L., Doherty, R. D., Cahn, R. W. (1971), Acta Metall 19, 1019. Fleischer, R. L. (1987), Acta Metall 35, 2129. Flower, H. M. (1990), Mater. Sci. Tech. 6, 1082. Frank, F. C. (1988), MRS Bulletin, 13/3, 24; also in: ICOTOM, Eighth International Conference on Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 3-16. Fukuda, N., Shimizu, M: (1972), /. Japan Soc. for Technology of Plasticity 13, 841. Gangli, P., Liicke, K. (1988), in: ICOTOM - Eighth International Conference on Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 735-740. Goss, N. P. (1934), U.S. Patent 1 965559. Gottstein, G. (1988), in: ICOTOM - Eighth International Conference on Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 195-202. Greenough, G. B. (1949), Proc. Roy. Soc. 197A, 556. Gwathmey, A. T., Lawless, K. R. (1960), in: The Surface Chemistry of Metals and Semiconductors: Gatos, H. C. (Ed.). New York: Wiley, 483. Harase, J., Shimizu, R., Nakamura, Y, Takahashi, N. (1990), Materials Forum 14, 276. Harris, G. B. (1952), Phil Mag. 43, 113-123. Hartig, Ch., Fang, X. F , Mecking, H., Dahms, M. (1992), Acta Metall, in press. Hatherly, M., Hutchinson, W. B. (1979), An Introduction to Textures in Metals. London: The Institution of Metallurgists. Held, J. F. (1967). Trans. Met. Soc. Amer. Inst. Min. Metall Engrs. 239, 573. Holden, A. N. (1958), Physical Metallurgy of Uranium. Reading: Addison-Wesley, pp. 95-104. Hon, W P., Wu, S. K., Koo, C. HJ. (1991), Mater. Sci. Eng. A131, 85. Hosford, W. F. (1969), in: Textures in Research and Practice: Grewen, X, Wassermann, G. (Eds.). Berlin: Springer-Verlag, pp. 414-443. Hosford, W F, Backofen, W. A. (1964), in: Fundamentals of Deformation Processing. Syracuse University Press, 259. Hu, H., Goodman, S. R. (1963), Trans. AIME 227, 627. Hu, H., Sperry, P. R., Beck, P. A. (1952), Trans. AIME 194, 76.
Humbert, M. (1988), in: Directional Properties of Materials: Bunge, H.-J. (Ed.). Oberursel: DGM Informationsgesellschaft Verlag, pp. 223-238. Humphreys, F. J. (1988), in: ICOTOM - Eighth International Conference on Textures of Materials: Kallend, S. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 171-182. Hutchinson, W. B. (1989), Acta Metall. 37, 1047. Hutchinson, W B., Ekstrom, H.-E. (1991), Mater. Sci. and Tech. 6, 1103. Ibe, G. (1988), in: Directional Properties of Materials: Bunge, H.-J. (Ed.). Oberursel: DGM Informationsgesellschaft Verlag, pp. 145-156. Inagaki, H. (1990), Z. Metallkde. 81, 282. Inokuti, Y, Doherty, R. D. (1977), Textures Crystall Solids 2, 143. Inokuti, Y, Doherty, R. D. (1978), Acta Metall 26, 61. Inokuti, Y, Sujita, S., Tanaka, T. (1985), J Iron and Steel Inst. Japan 25, 1141. Inokuti, Y, Maeda, C , Ito, Y (1987), Trans. Iron and Steel Inst. of Japan 27, 139, 302 (in English). Inoui, H., Inakazu, N. (1991), in: Proceedings of International Symposium on Intermetallic Compounds (VIMIS-6): Izumi, O. (Ed.). Sendai: Japan Institute of Metals, pp. 785-790. Jones, A. M. (1986), Harwell Report AERER 12275, London: Her Majesty's Stationery Office. Juul Jensen, D., Randle, V. (1989), in: Tenth Riso Symposium: Materials Architecture: BildeSorensen, J. B., Hansen, N., Juul Jensen, D., Leffers, T., Lilholt, H., Pedersen, O. B. (Eds.). Riso, Denmark: Riso National Laboratory, pp. 103126. Kallend, J. S., Gottstein, G. (Eds.) (1988), ICOTOM - Eighth International Conference on Textures of Materials. Warrendale: The Metallurgical Society. Khadkikar, P. S., Michal, G. M., Vedula, K. (1990), Met. Trans. A 21 A, 279. Kitagawa, K. (1989), Z. Metallkde. 80, 648-652. Kochendorfer, A., Miiller, H.-G. (1955), Archiv Eisenhuttenw. 26, 291. Kopineck, H.-J., Bunge, H. J. (1988), in: Directional Properties of Materials: Bunge, H.-J. (Ed.). Oberursel, DGM Informationsgesellschaft Verlag, pp. 251-262. Krawitz, A. D., Holden, T. M. (1990), MRS Bulletin 15/11, 57.
Lankford, W. T., Snyder, S. C , Bausher, J. (1950), Trans. Amer. Soc. Metals 43, 1195. Larson, F. R., Zarkades, A. (1976), in: Texture and the Properties of Materials (Proc. 4th Texture Conference): Davies, G. X, Dillamore, I. L., Hudd, R. C , Kallend, J. S. (Eds.). London: The Metals Society, pp. 210-216. Lartigue, S., Priester, L. (1988), /. Am. Ceram. Soc. 71, 430. Lee, R. W, Brewer, E. G., Schafel, N. A. (1985), IEEE Trans. Magn. MAG-21, 1958.
10.7 References
Lequeu, P., Jonas, I J. (1988), in: ICOTOM - Eighth International Conference on Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 1091-1096. Lim, L. C , Raj, R. (1984), Acta Metall 32, 1183. Lim, L. C , Watanabe, T. (1990), Acta Metall. 38, 2507. Liu, W. P., Sun, L. X, Bunge, H.-X (1988), in: ICOTOM - Eighth International Conference on Textures of Materials: Kallend, X S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 749-754. Luborksy, F. E., Livingston, X D., Chin, G. Y. (1983), in: Physical Metallurgy (Third Edition): Cahn, R. W., Haasen, P. (Eds.). Amsterdam: North-Holland, pp. 1689-1691, 1703-1704. Macherauch, E. (1966), Experimental Mechanics 6, 140. Mclntosh, A. B., Heal, T. X (1960), Materials for Nuclear Engineers: London: Temple Press, pp. 37, 61. Mackenzie, X K. (1964), Acta Metall. 12, 233. McLean, M., Mykura, H. (1965), Acta Metall. 5, 628. McQueen, H. X, Mecking, H. (1987), Z. Metallkde. 78, 387. Marsh, X M., Martin, X W. (1991), Mater. Sci. Tech. 7, 183. Merchant, H. D., Morris, X G. (1985), Textures in Non-Ferrous Metals and Alloys. Warrendale: The Metallurgical Society. von Mises, R. (1928), Z. angew. Math. Mech.
8,161.
Nelson, K. D., Adams, B. L., Fricke Jr., W G. (1988), in: ICOTOM - Eighth International Conference on Texture of Materials: Kallend, X S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 1097-1102. Nelson, N. X, Martens, P. A., Battey, X E, Wenk, H.-R., Zhong, Z. Q. (1988), in: ICOTOM - Eighth International Conference on Texture of Materials: Kallend, X S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 933-938. Nichols, C. S. (1991), Mat. Res. Soc. Symp. Proc. on Defects in Materials, vol. 209, in press. Nichols, C. S., Clarke, D. R. (1991), Acta Metall. 39, 995. Nichols, C. S., Cook, R. F., Clarke, D. R., Smith, D. A. (1991), (2 papers), Acta Metall. 39, in press. Ortner, S. R., Randle, V. (1989), Scripta Metall. 23, 1903. Plege, B. (1987), in: Theoretical Methods of Texture Analysis: Bunge, H.-X (Ed.). Oberursel: DGM Informationsgesellschaft Verlag, pp. 393-403. Pokros, C. (1989), JOM (formerly Journal of Metals) 41/10, 46. Prantl, W, Werner, E. (1990), Z. Metallkde. 81, 672. Randle, V. (1990 a), Mater. Sci. Tech. 6, 1231. Randle, V. (1990b), Proc. Roy. Soc. 431 A, 61, Randle, V., Brown, A. (1988), Phil. Mag. A58, 111. Randle, V., Brown, A. (1989), Phil. Mag. A59, 1075.
479
de Rango, P., Lees, M,, Lejay, P., Sulpice, A., Tournier, R., Ingold, M., Germi, P., Pernet, M. (1991), Nature 349, 770. Rashkov, S., Stoichev, D. S., Tomov, I. (1955), Electrochim. Acta 17, 1955. Reddy, A. K. N. (1963), J. Electroanalyt. Chem. 6, 141. Revcolevschi, A., Dhalenne, G. (1985), Nature 316, 335. Ridha, A. A., Hutchinson, W. B. (1981), in: ICOTOM 6 - The Sixth International Conference on Textures of Materials: Nagashima, S. (Ed.). Tokyo: The Iron and Steel Institute of Japan, pp. 112-131. Rieck, G. D. (1957), Philip Res. Rep. 12, 423. Roberts, C. S. (I960), Magnesium and its Alloys, 191. Rodrigues, P. M. B., Bate, P. S. (1985), in: Textures in Non-Ferrous Metals and Alloys: Merchant, H. D., Morris, J. G. (Eds.). Warrendale: The Metallurgical Society, pp. 173-187. Rodrigues, P. M. B., Bichesel, H., Furrer, P. (1985), in: Textures in Non-Ferrous Metals and Alloys: Merchant, H. D., Morris, X G. (Eds.). Warrendale: The Metallurgical Society, pp. 45-59. Rosenberg, A., Tiller, W. A. (1957), Acta Metall. 5, 565. Shimizu, R., Harase, X, Dingley, D. X (1990). Acta Metall. 38, 973. Schmid, E., Boas, W. (1935), Kristallplastizitdt. Berlin: Springer-Verlag. Sturcken, E. F , Croach, X W (1963), Trans. Metall. Soc. AIME 227, 934. Taylor, G. I. (1934), /. Inst. Metals 62, 307. Tome, C , Pochettino, A., Penelle, R. (1988), in: ICOTOM - Eighth International Conference on Texture of Materials. Warrendale: The Metallurgical Society, pp. 985-990. Thomas, X M. (1974), Phil. Trans. Roy. Soc. A227, 251. Underwood, F. A. (1961), Textures in Metal Sheets. London: Macdonald. Venables, X A., Harland, C. X (1973), Phil. Mag. 27, 1193. Wagner, F , Bergmann, H. W, Humbert, M., Esling, C. (1981), in: ICOTOM 6 - Sixth International Conference on Textures of Materials: Nagashima, S. (Ed.). Tokyo: The Iron and Steel Institute of Japan, pp. 714-719. Walter, X L., Dunn, C. G. (1959), Acta Metall. 7, 424. Wassermann, G., Grewen, X (1962), Texturen Metallischer Werkstoffe, 2nd Edition. Berlin: Springer, pp. 120, 127, 415 et seq. Watanabe, T. (1984). Res Mechanica 11, 47. Watanabe, T. (1988), Materials Forum 11, 284. Watanabe, T., Fujii, H., Oikawa, H., Arai, K. I. (1989), Acta Metall. 37, 941. Weil, R. (1989), Ann. Rev. Mat. Sci. 19, 165-182. Wenk, H.-R., Bunge, J.-X, Kallend, X S., Liicke, K., Matthies, S., Pospiech, X, van Houtte, P. (1988), in: ICOTOM - Eighth International Conference on
480
10 Measurement and Control of Texture
Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 17-30. Wilbrandt, P.-J. (1988), in: ICOTOM - Eighth International Conference on Textures of Materials: Kallend, J. S., Gottstein, G. (Eds.). Warrendale: The Metallurgical Society, pp. 573-584. Wilson, D. V. (1966), J. Inst. Metals 94, 84.
General Reading For many years there have been triennial International Conferences on Textures of Materials. The last one published was of ICOTOM 9, held in America in 1988, and frequently referred to in the above text. Bunge, H.-J. (1987), "Three-Dimensional Texture Analysis". Int. Mater. Rev. 32, 265. Bunge, H.-J. (1982), Texture Analysis in Materials Science - Mathematical Methods. London: Butterworths. Bunge, H.-J. (1988), Directional Properties of Materials. Oberursel: DGM Informationsgesellschaft Verlag.
Bunge, H.-J., Esling, C. (Eds.) (1982), Quantitative Texture Analysis. Oberursel: Deutsche Gesellschaft fur Metallkunde; Paris: Societe Francaise de Metallurgie. Dillamore, I. L., Roberts, W T. (1985), "Preferred Orientation in Wrought and Annealed Metals", Metallurgical Reviews 10, 271-380. Hatherly, M., Hutchinson, W B. (1979), An Introduction to Textures in Metals. London: The Institution of Metallurgists (now Institute of Metals). Hutchinson, W. B., Ekstrom, H.-E. (1990), "Control of annealing texture and earing in non-hardenable aluminium alloys", Mater. Sci. Tech. 6,1103-1112. Merchant, H. D., Morris, J. G. (Eds.) (1985), Textures in Non-Ferrous Metals and Alloys. Warrendale: The Metallurgical Society. Teuckhoff, E. (1990), "Texturen und mechanische Anisotropie von Zirkoniumlegierungen", in: Beitrdge zur Materialkunde. Asbeck, O. W, Matucha, K. H. (Eds.). Oberursel: DGM Informationsgesellschaft Verlag. Wassermann, G., Grewen, J. (1962), Texturen metallischer Werkstoffe, 2nd ed. Berlin: Springer. Wilson, D. V. (1966), "Plastic Anisotropy in Sheet Metals", J. Inst. Metals 94, 84-93.
11 Electrodeposition of Metals and Alloys Jef R. Roos, Jean-Pierre Celis, Marc De Bonte Katholieke Universiteit Leuven, Department of Metallurgy and Materials Engineering, Leuven, Belgium
List of Symbols and Abbreviations 11.1 Introduction 11.2 Fundamentals of Electrodeposition 11.2.1 Electrolytic Cells 11.2.2 Thermodynamics of Electrodeposition Reactions 11.2.3 Kinetics of Electrodeposition Reactions 11.2.4 E-pH Diagrams 11.2.5 Summary 11.3 Electrodeposition-Related Properties of Coatings 11.3.1 Crystallographic Texture in Electrolytic Coatings 11.3.2 Internal Stresses in Electrolytic Coatings 11.3.3 Ductility of Electrolytic Coatings 11.3.4 Porosity of Electrolytic Coatings 11.3.5 Corrosion and Wear Protection by Electrodeposits 11.4 Materials and Process Developments in Alloy Electrodeposition 11.4.1 Alloy Plating 11.4.1.1 Effect of Plating Bath Hydrodynamics 11.4.1.2 Optimization of Deposition Parameters 11.4.2 Composite Plating 11.4.2.1 Electrolytic Codeposition of Solid and Liquid-Containing Particles 11.4.2.2 Mechanism of Electrolytic Codeposition 11.4.3 Compositionally Modulated Alloys 11.4.3.1 Electrolytic Production of Compositionally Modulated Alloys 11.4.3.2 Properties of Electrolytic Compositionally Modulated Alloys 11.4.4 Electroless Plating 11.4.4.1 Control of the Deposition Rate 11.4.4.2 Properties of Electroless Coatings 11.5 References
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
482 485 487 487 488 490 493 493 494 495 496 499 503 507 511 511 513 516 518 519 521 524 525 527 529 530 533 535
482
11 Electrodeposition of Metals and Alloys
List of Symbols and Abbreviations a «x
b C
cd cion
cT
D d0 d Dion E°" AE AE e~ F /o G AG
K
i »o i AI 'ox *p
ired
h k K K' kB k* L m M N* P' P P Pi
Q
geometric factor activity of element x agitation rate concentration double layer capacitance bulk concentration of the ion vol.% suspended particles diffusion coefficient interatomic distance in the nucleus thickness of deposited layer diffusion coefficient electrode potential potential field at the cathode potential variation electron Faraday's constant resonant frequency weight increase due to metal deposition free energy change parameter in theory of alloy codeposition current exchange current density current density imposed current variation oxidation current density plating rate reduction current density Faradaic current at time t number of reduced ions number of adsorbed ions reaction rate constant Boltzmann constant Langmuir isotherm constant bond strength constant atomic weight number of collisions Avogadro number pressure coating porosity probability coefficient probability for one ion to be reduced at a current density i quantity of electric charge
List of Symbols and Abbreviations
R R r
resistance gas constant particle radius roughness average critical radius polarization resistance entropy temperature time transfer number volume wear volume volume fraction of particles in the deposit quantity of matter weight of one particle number of electrons
K rc R
v
s
T
t t*
V V
w W z
v
a' a
A,
PB, £P
P
j8 a , j8 c
7 5 8 5{
n
rjcon
nCr
n fJLq
gp Qm Q* Qq
6 ap9
ah
AFHG ASTM CMA CMM CVD
roughness factor measure of interaction current efficiency, voltage efficiency, power efficiency symmetry factor anodic slope, cathodic slope edge energy thickness diffusion layer thickness frequency change activation overpotential concentration overpotential crystallization overpotential overpotential at time t shear modulus of the deposited material chemical potential of phase 1 or 2 shear modulus for quartz particle density density of deposited metal stress ratio density of quartz effective stress plane strain flow stress, balanced biaxial flow stress electrical potential additive free hard gold plating bath American Society for Testing Materials compositionally modulated alloys compositionally modulated multilayers chemical vanor deoosition
483
484
d.c. DTA IUPAC MOCVD PVD RCE RDE SEM STP TFS SHE PTFE
11 Electrodeposition of Metals and Alloys
direct current differential thermal analysis International Union of Pure and Applied Chemistry metallo-organic chemical vapor deposition physical vapor deposition rotating cylinder electrode rotating disc electrode scanning electron microscopy special technical publication tin-free steel standard hydrogen electrode polytetrafluoroethylene
11.1 Introduction
11.1 Introduction Recent developments in process technology have opened new possibilities in the field of surface treatment of materials. The availability of new electronic devices allowed the introduction in electroplating of more sophisticated on-line process controls, the use of pulsed current instead of direct current, and a better characterization of the coating composition and structure by the introduction of scanning electron microscopy and surface analysis methods like Auger electron and laser Raman spectroscopy. These and other advances in material characterization techniques also allow accurate measurement of structural features, such as composition distribution, morphology, texture, internal stress, etc., in thin coatings or films. As a result, the number of kinds of coatings and surface films that can be deposited in a controlled way on either metallic, ceramic or organic substrates has markedly increased. Recent research focussed on the unraveling of the trivalent relationship between production parameters, structural properties of the coating and functionality of the coating (see Fig. 11-1) has shown that surface treatment of materials, and electroplating in particular, can no longer be considered only as an art but is emerging as a science opening totally new avenues of application. The improvement of existing materials and the development of radically new ones is widely recognized by society. Recent surveys made in the U.S.A., Japan, Great Britain and Germany have led to the attribution of top positions to surface treatments and coating technology in the list of future priorities in research and development. The fact that electroplating can generally be carried out at room temperature and atmospheric pressure, and that it
485
is greatly flexible, makes it a very attractive technology. Today a large number of materials are used with electrolytic coatings. The reason can be that this is the only way to achieve a desired material property, or it can be that by combining a substrate material with a coating, the most economic product is achieved. A coated material can be viewed as a composite material system, the behavior of which depends on the properties of the bulk material, of the surface material and of the inherent interfaces. The importance of the right selection of plating parameters then becomes evident. In the electroplating process three major steps have to be considered, namely the pretreatment of the substrate, the deposition of the electrocoating and sometimes a post-treatment of the coated material. As a result, a surface layer is obtained which has a specific crystallographic structure resulting from complex processes consisting of nucleation, grain growth and influences from foreign species or contaminants. In this respect, a fundamental understanding is based on a better insight into the thermodynamics and kinetics of these processes (see Sees. 11.2.1 to 11.2.4). The surface-modified layer that is obtained has to be characterized. Two aspects are of importance here, namely, the definition of the structural characteristics of the coating or film and the development of appropriate measurement techniques. Con-
Figure 11-1. The coating-R&D triangle.
486
11 Electrodeposition of Metals and Alloys
cerning structural properties, a differentiation should be made between bulk properties and interface characteristics. Bulk properties can be classified into chemical properties like composition (see Sec. 11.3), physical properties like crystallographic structure, surface topography, texture (see Sec. 11.3.1) and internal stresses (see Sec. 11.3.2), porosity (see Sec. 11.3.4), electrical and thermal conductivity, and finally mechanical properties like elongation, ductility (see Sec. 11.3.3), tensile strength, fatigue strength, etc. Interfaces which should be considered in coated materials are of different kinds, namely, the outermost surface layer of the coating, the "transition" layer between the substrate material and the coating on which the adhesion between coating and substrate will depend largely, and finally some artificial interfaces deliberately introduced into the coating, as for instance in composite coatings (see Sec. 11.4.2) and in compositionally modulated coatings (see Sec. 11.4.3). At this stage, the electroplater can establish some specific relationships between the process parameters selected and the related structural characteristics of the coatings. This approach opens the possibility of tailoring the production method towards coatings having specific characteristics, as for instance in alloy plating where hydrodynamics can play an important role (see Sec. 11.4.1). Finally one should be concerned with the purpose of a surface treatment on a material, which is always the enhancement of the functionality of a component, as for instance the corrosion or wear resistance (see Sec. 11.3.5). For coated systems, the measurement of the functional or application-related properties is a very difficult task. One can use some simulation and/or accelerated tests on a laboratory scale and extrapolate the
results towards field service conditions, or perform expensive and tedious field tests. In this way, relationships between the structure of a coating and its functionality can be worked out. The availability of trivalent relationships between production, structure and functionality of coatings will surely support and stimulate future developments in the field of surface modification of materials. Especially in relation to electroplating, a number of new and ingenious process developments have already been introduced successfully into industrial practice, as for instance selective reel-to-reel plating, high current density electrodeposition and nonaqueous deposition of reactive elements like aluminium, while others are still more or less at a development stage, as for instance laser-enhanced electroplating and laser-induced deposition on non-conducting surfaces such as plastics. Finally, electroless plating (see Sec. 11.4.4) should be mentioned as an interesting development. Electroless plating can be considered as a modified electroplating process where the required electrons are no longer made available through an external power source but result from some autocatalytic electrochemical reactions. Research and development activities as well as technology transfer in the field of the electrolytic deposition of coatings on materials are growing very rapidly, and the authors believe that we are gradually approaching an era where tailoring of coating systems to the desired application will be within our reach. The following overview of the science and technology of the electrodeposition of metals and alloys is not intended to be complete. It is aimed primarily at showing the tremendous potential offered by a better insight into relationships between process parameters, structural properties and functionality of coat-
11.2 Fundamentals of Electrodeposition
ings, for the creation by electrodeposition of improved and also radically new materials.
487
metallic connector
anode
cathode
11.2 Fundamentals of Electrodeposition For a meaningful discussion of the electrodeposition of metals and alloys, the fundamentals of electrochemistry have to be understood. The introduction of electric current transforms chemical reactions into electrochemical reactions, and allows the generation of reaction products which could not be achieved without this electric current. Electrochemical reactions are realized in electrochemical cells. Reversible cell thermodynamics describes electrochemical cells where no net electric current flows, and tells us, from a thermodynamic point of view, which electrodeposition reactions are possible and which are not. An essential characteristic of electrodeposition is of course that a net electric current flows through the cell. In order to produce this current flow, the electrodes have to be brought out of equilibrium by imposing upon them an externally controlled potential that differs from their equilibrium potential. The difference between the two potentials is termed the overpotential. The study of the relationship between overpotential and net current is the domain of electrode kinetics. Insight into electrode kinetics will for instance allow us to understand why the electrodeposition of certain metals and alloys from aqueous solutions, which according to reversible thermodynamics is not possible, can be realized in practice. 11.2.1 Electrolytic Cells Electrodeposition is realized in an electrolytic cell as depicted in Fig. 11-2. The
Figure 11-2. Schematics of an electrolytic cell.
electrodes (anode and cathode) and the metallic connector conduct electricity electronically, i.e., by a net movement of electrons in the structure of the conductor when an electrical potential is applied. This flow of electricity is not accompanied by a significant movement of matter. In electrolytic conductors, or electrolytes, electricity is conducted ionically, i.e, by ions of non-negligible mass. Positive ions (A+) (cations) will move towards the cathode, and may produce a layer of A (electrodeposit) on the electrode surface. Aqueous solutions are by far the most important electrolytes used in electrodeposition, but fused salts and organic solutions are also used. Certain solids are also electrolytic conductors, but owing to the restricted movement of ions in solids they are not suitable for electrodeposition purposes. They are very useful however as solid electrolytes in electrochemical sensors. At the electrolyte-electrode interface the mechanism of conduction changes from electrolytic to electronic. The cathode is defined as the electrode where reduction reactions (like A + + e -• A) take place, where electrons are consumed by discharge of the positively charged ions (the cations). The anode is the electrode where oxidation reactions, such as B~-> B + e take place, where electrons are produced by discharge
488
11 Electrodeposition of Metals and Alloys
of the negatively charged ions (the anions). Electrolytes obey Ohm's law, where I = E/R, E being the potential difference applied across the electrolyte, R the resistance of the electrolyte and / the current flowing.
(11-1)
can take place. The ions are present as complex ions, in an aqueous solution as solvated ions. The metal electrode consists of metal ions in a crystal structure. The transfer of a metal ion from the electrolyte structure to the metal structure will be accompanied by a free energy change AG. Thermodynamics tells us that in a system with only PV work, the free energy changes for the transport of n moles of matter i from phase 1 to phase 2 according to dG = V dP -SdT+/d?
dn - fif dn
tfti
tf
(11-4)
or ..El, 1
Pi
..El, 2
(11-5)
= Pi
A(j)e = (j)M — cf)s
At the cathode a reaction of the type •M
ri
For the electrochemical reaction, Eq. (11-1), taking place at the electrode:
11.2.2 Thermodynamics of Electrodeposition Reactions Mz+ + z e "
The condition for equilibrium is now
(11-2)
where pf and fif are the chemical potentials in phases 1 and 2, respectively. The equilibrium criterion, with P and T constant, is that the free energy is minimum, thus dG = 0, and consequently fif = fif. In an electrochemical system, not only PVwork, but also electrical work - the work accompanying the transport of electric charge over a potential difference - must be taken into account. Eq. (11-2) then transforms into
(11-6)
where A(j)e is the potential difference between the metal at the electrode and its ions in solution: AGm = AG° — RT\naMZ
+
+ zFA(j)e
(11-7)
where AG° is the standard free energy of reaction, Eq. (11-1), and aMZ+ the activity of Under standard conditions (aMZ+ = 1) at equilibrium: AG°=-zFA(l)oe
(11-8)
where A(j)°Q is the standard electrode potential. Combination of Eqs. (11-7) and (11-8) leads to: Acf)e=A(l)oe+—-\naMZ + (11-9) zz tt In general for a redox reaction this equation turns into: (11-10) red
where n is the number of electrons exchanged in the reaction and A<j>® is the standard electrode potential of the redox reaction. Eq. (11-10) is known as the Nernst equation. It must be remembered (11-3) that Eqs. (11-2) to (11-10) refer only to therwhere z F (f — <j>\) is the electrical work modynamically reversible processes. This means that A(j>e (also denoted E) is the reinvolved in the transport of the electrical versible electrode potential, and zl° (also charge of 1 mol of ions from phase 1 to phase 2, (f) is the electrical potential and F denoted £°) is the reversible standard electrode potential. is Faraday's constant (96485 C mol~*).
11.2 Fundamentals of Electrodeposition
The problem with these single electrode potentials is that they cannot be measured. The potential difference across a metalelectrolyte interface cannot be measured without the introduction of at least one extra interface, so that the measured potential will also contain the contribution by the latter. E and E° can therefore only be known in a relative way. Any electrode can be taken as reference electrode, but the standard hydrogen electrode (SHE) is the one generally accepted. The SHE consists of an active platinum surface submerged in an HC1 solution with average ionic activity of unity at 298 K, which corresponds to a 1.18 molal HC1 solution. Hydrogen, saturated with water vapor, at a pressure of 1 atm is led over the platinum surface where the electrode reaction e-=l/2(H 2 ) ( g )
(11-11)
takes place reversibly. The SHE is a nonpolarizable electrode, which means that its potential hardly changes when the electron exchange reaction at its surface takes place at a finite speed. The SHE can be combined with another electrode in an electrochemical cell, as depicted in Fig. 11-3. Those cell potentials are electrode potentials relative
H2(1atm)
Figure 11-3. An electrochemical cell composed of a SHE and a metal electrode M submerged in a solution containing Mz + -ions.
489
to SHE, and denoted ESHE or in short E, E measured under standard conditions of the second electrode is then equal to E°, the standard electrode potential relative to SHE. Referring to the IUPAC convention, standard electrode potentials can be measured with this set-up. The electrochemical cell can be noted as: (11-12) P t / H 2 , H + (aH+ = 1) || Mz+ (aMZ+ =
The cell potential measured is then equal to £°(M Z+ /M), and the sign is determined by the polarity of the right-hand side electrode. Electrode reactions are always written as reduction reactions. The E° value is measured with a voltmeter with infinitely high resistance. Measured in this manner E° (Cu + + /Cu) = + 0.34 V and £°(Zn + + /Zn) = -0.76 V. Table 11-1 gives the E° values for a few selected electrode systems. An element from the standard electrode potential series behaves, in standard conditions, anodically (i.e., tends to be oxidized) vis-a-vis all other elements of the series that have a higher E°, and cathodically (i.e., tends to become reduced) vis-a-vis all elements having a lower E°. This also means that if the standard electrodes of two elements are connected with an electronic conductor, electrons will flow through this conductor from the cell with the lower E° to the cell with the higher E°. It also means, remembering the relation AG° = — nFE°, that when electrons are supplied from an external source, i.e., electrolysis, those reduction reactions with the highest E°, and consequently those oxidation reactions with the lowest £°, will be thermodynamically favored. The relationship between the quantity of electric current and the quantity of chemical transformation at the electrodes is
490
11 Electrodeposition of Metals and Alloys
given by Faraday's laws which state the following: 1. The electrochemical reactions take place only at the surface of an electrode. 2. The quantity of matter (W) that reacts at the electrode is directly proportional to the quantity of electric charge (Q) flowing through the cell. 3. The quantity of matter that reacts at the electrode as a result of the passage of a given electric current is directly proportional to the chemical equivalent (molar mass divided by the number of electrons (z) involved in the reaction) of the reacting substance. This leads to M M (11-13) It -zF zF M/zF is called the electrochemical equivalent and for C u + + is equal to 0.3293 x 1 0 " 3 g C " 1 . Eq. (11-13) allows the calculation of the quantity of matter that can be deposited theoretically with a given amount of electricity. As such, this formula is only valid for one defined ion. If more than one ion can react at an electrode, they can each consume part of the electricity. For a reaction at a cathode, for instance one can then define the cathodic current efficiency (jS7) as the ratio of the actually deposited amount of material to the theoretically possible amount. As an example: the cathodic current efficiency for the electrodeposition of Cr from a Cr 6 + solution is lower than 20%, largely owing to the consumption of electrons for the reduction of H + to H 2 and Cr 6 + to Cr 3 + . As already indicated above, the standard electrode potential series is a valuable instrument in understanding electrodeposition reactions, but it has inherent limitations. A first limitation is that it is confined to reactions where the participating species are in their standard state, i.e., at unit activity: pure metals, and ionic activities equal
to unity (e.g. 1 molal concentration of ions in an ideal solution). However, if the real activities are known, then the E values can be calculated through the Nernst equation, Eq. (11-10), and an adapted E-series can be generated for one's specific purposes, and interpreted in the same way as was done for the E° series (see also E-pH diagrams in Sec. 11.2.4). A second limitation lies in the reversible nature of the standard electrode potential series. Practical electrodeposition will be done under strongly irreversible conditions, i.e., with appreciable current flowing through the cell. Under these circumstances the electrodes cannot be at their reversible equilibrium potential. Here electrochemical kinetics comes into play. 11.2.3 Kinetics of Electrodeposition Reactions
From the moment that a current flows through an electrode, the electrode acquires a potential different from the reversible equilibrium potential. The relationship between the current intensity (or current density if expressed per unit area) and the deviation from the equilibrium potential is the object of the study of electrode kinetics. It is well to note here that equilibrium means dynamic equilibrium, and that although no net current flows through the electrode, oxidation and reduction reactions will take place simultaneously, such that the oxidation current density iox is equal in magnitude to the reduction current density ired. Both are at equilibrium equal to the so-called exchange current density, and the equilibrium potential across the electrode is Acj)c. In order to realize electrodeposition at a cathode, a net current / = (iox — zred) < 0 has to flow through the cathode (and thus
11.2 Fundamentals of Electrodeposition
through the cell). Therefore the potential Acj) over the electrode interface has to deviate from A(pe. This deviation (rj) is called the activation overpotential: A(j) =A(j)e + ri
(11-14)
The term 'activation' refers to the fact that the electrode reaction is a kinetic process, in which substances must acquire a certain activation energy before they can jump through the electrode interface. Deviation from the equilibrium potential is called polarization. The Butler-Volmer Equation (11-15) is a kinetic equation that relates the net current density i through the electrode interface with the activation overpotential rj: i -—
IQ
[e
e
j
yi i -1 DJ
where i0 is the exchange current density and fi is a symmetry factor with a value between 0 and 1, referring to the position of the maximum in the profile of the activation energy. The first term in Eq. (11-15) represents the oxidation current density iox, the second term the reduction current density ired. Fig. 11-4 gives the global i-rj graph for /? = 0.5.
- A a -log/a
anodic current cathodic current i*
' o " RT
6 c -log/ c
Figure 11-4. Graphic representation of the ButlerVolmer equation (/? = 0.5).
491
At large Y\ values Eq. (11-15) can be approximated by the simple logarithmic relationship: = a + b log i
(11-16)
At very small r\ values (i.e. close to reversible equilibrium) it approximates to a linear i-rj relationship: i« i0 n F/R T
(11-17)
From the above, it is clear that the activation overpotential is co-determined by the desired current density /, the temperature, the electrode material and the electrode reaction (via io\ and the electrolyte composition which determines the structure of the electrical double layer at the electrode (viajS). The activation overpotential is of major importance for the gas-producing reactions at an electrode. The electroplating of Zn, for example, from aqueous solutions is only feasible because the reduction of the more noble (higher £°, E) H + -ion is much hindered by a large hydrogen activation overpotential. As a result of the electrode reaction, the ionic concentration tends to change in the vicinity of the electrode, resulting in a concentration overpotential rjcon. For a reduction reaction at a cathode, for example, the concentration of positive ions is lowered, and the equilibrium potential shifts according to the Nernst equation. If, as a result of an increasing cathodic polarization, the reduction current density is increased, positive ions have to be brought to the cathode surface faster and faster. This can occur by migration under influence of the electrical field, by diffusion due to the created concentration differences, and through natural or forced convection (e.g., agitation of the electrolyte or vibration of the cathode). But a very thin layer adjacent to the cathode surface will never be
492
11 Electrodeposition of Metals and Alloys
reached by this convection, although the thickness itself of the layer can be influenced. Transport through this thin layer is mainly dependent on the mechanism of diffusion. With increasing cathodic current density, ions can at a certain point no longer be supplied fast enough and the solution adjacent to the cathode surface gets depleted completely, and the current density reaches a limit, the limiting current density (iL): I,
=
nFDC
(11-18)
where D is the diffusion coefficient, C the bulk concentration, t* the transfer number and S the thickness of the hydrodynamic double layer. From that point on, an increased polarization of the electrode will no longer lead to an increase in current density. Fig. 11-5 illustrates this in an i-rj diagram. Activation and concentration overpotential are the most important components
of the total overpotential. Another component is the crystallization overpotential (rjcr) originating from difficulties encountered with nucleation and growth (for a cathodic process). Minute concentrations of foreign substances can drastically increase the crystallization overpotential, and thus effectively slow down the electrodeposition process. Sometimes, organic additives are deliberately added to a plating bath. They adsorb at the electrode surface and thus lower the number of active sites. In doing so they lower the limiting current density, but they have the beneficial effect of creating smoother deposits. The global electrode overpotential is: ftot = n + *?con + >7cr
An electrodeposition cell contains two electrodes, a cathode and an anode. Both contribute to the total cell potential according to Eq. (11-19). In addition there is the resistance of the electrolyte leading to a potential IR, and various other resistances in the cell circuit creating a potential ER. The total cell potential is then: ,** 20) ^tot
ni-
(11-19)
=
^reversible + */tot + *ltot + ^ ^ + ^R
The practical cell potential is therefore sometimes much higher than the reversible cell potential calculated from purely thermodynamic considerations. The voltage efficiency fiE is defined as the ratio of the reversible potential to the practical potential: 0(O/)
=
R
Reversible •'-'practice
The power efficiency jSp is defined as: 100 Figure 11-5. Activation and concentration overpotential, and the occurrence of a limiting current density iL.
(11-22)
Note that the power efficiency is zero for an electrodeposition cell where the metal
11.2 Fundamentals of Electrodeposition
493
Cu 2.0 •
§P
1.0
corrosion -1.0 1
1
1
passivation
2.0
immunity
1.0
50 -1.0
- water ^stability " region
!
Figure 11-6. £ pH diagrams for some metal-water systems (298 K).
14pH
ions are supplied from a soluble anode. For such a process £ rev is indeed equal to zero. For a more detailed treatise on electrochemistry which is beyond the scope of this chapter, the reader should refer to textbooks (Bockris and Reddy, 1977; Newman, 1973).
An adapted version of E-pH diagrams for some metal-water systems of interest in electroplating are given in Fig. 11-6. The diagrams clearly show the potential-pH conditions where water is stable and where metals and their compounds can exist. Superposition of diagrams for different metal-water systems can be helpful in understanding alloy electrodeposition.
11.2.4 ZT-pH Diagrams
In addition to the effect of an applied potential on the tendency of a metal to deposit cathodically (or to dissolve anodically), the pH of the aqueous electrolyte can have a profound effect. This effect is neglected in the standard electrodepotential series, where the pH is fixed at zero (aH+ = 1, and pH = — loga H+ ) by the choice of the SHE. The combined effect of pH and applied potential is summarized in £-pH, or Pourbaix, diagrams (Pourbaix, 1974). These diagrams are equilibrium diagrams, and as such do not contain kinetic information. £-pH diagrams can be easily calculated using the Nernst equation.
11.2.5 Summary
The essential features of the fundamentals of electrodeposition, as discussed above, can be summarised in the following sequence of schematic plots (Fig. 11-7). The standard electrochemical series (see Fig. 11-7 a) ranks different metals with regard to their noble/less noble character. It is only valid for standard conditions, but gives a first indication of the possibility of electrodepositing metals. Deviations from standard conditions can be taken into account via Nernst's equation, and lead to practical electrochemical series as schematically shown in Fig. 11-7 b. Adding a
494
11 Electrodeposition of Metals and Alloys
+
Cu 7Cu • Cu + 7Cu • H7H 2 -
Zn + 7Zn-
Zn + 7Zn • 0
Ci
b
2
4
6
8 pH
C
Figure 11-7. Summary of electrodeposition fundamentals.
electrolysis has evolved towards the electrorefining of metals and the electrodeposition of metallic coatings from aqueous, non-aqueous or molten salt baths (Rudzki, 1983). With the development of improved electrolytes and process technology, new types of electrolytic coatings have been developed. With regard to the composition of the coating matrix one can differentiate between:
11.3 Electrodeposition-Related Properties of Coatings
- elements, e.g., Al, Cr, Fe, Co, Ni, Cu, Zn, Ru, Rh, Pd, Ag, Cd, In, Sn, Sb, Ir, Pt, Au, Pb, - binary alloys, e.g., Cu-Au, Cu-In, Cr-Co, Pb-Sn, Ag-Cd, Au-Co, Pd-Ni, Ni-P, Ni-Mo, - ternary alloys, e.g., Cr-Ni-Mo, Cr-Ni-P, Ni-Cu-P, Sn-Pb-Cd, - quarternary alloys, e.g., Cr-Fe-Co-Ni, Co-Ni-Fe-P, Au-Pd-Cu-As, - composites, e.g., Cr-graphite, Cu-alumina, Ni-SiC, Ni-diamond, Co-Cr 2 O 3 , Ni-PTFE, - duplex layers, e.g., Ni + Cr, Ni + Ni(S), - compositionally modulated multilayers, e.g., Cu/Ni, Ag/Pd, Ni/Ni-P, - liquid-containing layers obtained, e.g., by the incorporation of liquid-containing microcapsules.
The large variety of electrodeposition parameters which can be brought into play, undoubtly offers unexpected potentialities for the production of electrolytic coatings. Developments in this technology over the last two decades support this view. Thus, present electroplating technology offers the possibility of depositing electrolytic coatings with a large variety in composition and metallographic structure. Over the years, electrolysis has been able to follow the general efforts in the development of new materials. Historically developed as a convenient method for the winning of metals from dilute solutions,
More detailed information on the electrodeposition-related properties of some of these systems is given in Sec. 11.4, where recent materials and process developments in alloy electrodeposition are discussed. Notwithstanding the fact that the preceding list only aims to indicate the width of the spectrum, it is evident that the choice offered to users today in regard to composition is impressive. In connection with the structure of electrolytic coatings, one actually has the possibility of depositing coatings with a crystalline, a microcrystalline or even an amorphous structure, depending on the plating
pH-axis as in Fig. 11-7 c gives the very useful E-pH diagram where the influence of the actual pH can be directly read out. All the above series and diagrams are valid for thermodynamic equilibrium conditions, i.e., with no current flowing. The introduction of a current axis, as in Fig. 11-5, leads to E-i diagrams, from which the situation existing at a specific current (i.e., electrodeposition rate) can be deduced. The information described above is also very useful for the study of wet corrosion of metals, which is electrochemical in nature.
495
11.3 Electrodeposition-Related Properties of Coatings
conditions used. With the introduction of new plating techniques like pulsed and pulse-reverse plating, coarse-grained metallic coatings can be converted into finegrained structures, and even lamellar (compositionally modulated) deposits can be grown from a single plating bath. The controlled incorporation of solid particles into the coating matrix during electrodeposition now offers the possibility of creating real composite coatings. Short fibers and irregular particles of different size incorporated in the cathodic deposit can largely modify the physical and mechanical properties of the coating. From the preceding, it can be concluded that electroplating has developed over the years under the impulse of new developments in the field of the so called new or advanced materials. The old-fashioned art of electrolysis has evolved, thanks to a better scientific understanding of the art of electroplating, into a genuine high technology. It is well to stress however at this point that, whatever plating conditions are used, the appropriate pre-conditioning of the substrate surface to achieve the desired cleanliness and surface roughness is a conditio sine qua non for the realization of a successful electrolytic coating. 11.3.1 Crystallographic Texture in Electrolytic Coatings
orientations are predominant and this causes a residual anisotropy. The origin of texture in an electrodeposited coating is closely linked to the deposition process itself: the formation of a coating is governed by the two distinct processes of nucleation and the subsequent growth of grains. The development of a texture may be considered as being related to the competition between the two processes of nucleation and grain growth. The overpotential and as a consequence every process parameter that indirectly influences the overpotential such as hydrodynamics and additives, in this connection play a primary role. The dependence of the crystallographic texture on current density and pH is shown in Fig. 11-8 (Amblard et al., 1979) for the case of nickel plating. With respect to the development of texture in thin electrolytic coatings, the surface condition of the substrate material is also important. In Fig. 11-9 the development of texture in electrolytic nickel coatings at increasing thicknesses is illustrated. On electropolished brass substrates, by way of example, the very first nickel layers grow in an epitaxial way so that the metallographic structure of the coating reproduces
\
4.5
\
(211)
\
3.5
The crystallographic texture of materials expresses the way how the thousands of constituent crystals which form materials are orientated. Each of these crystals is anisotropic, which means that the properties vary according the direction in which they are measured. A convenient way to measure such a crystallographic texture is by X-ray diffraction. In most engineering materials and generally also in electrodeposited coatings, some well-defined crystal
Ni
/
\ /
(211)
\
2.5 |=-'
(
1.5 -
(100)
(110)
/
(210)
\ 0.5 i
0.04
i
i \
JI
21 0.1 0.2 0.4 1 Deposition rate in ^jm/min
1
Figure 11-8. Texture stability diagram for nickel coatings (from Amblard et al., 1979; reproduced by permission of J. Appl. Chem.).
496
11 Electrodeposition of Metals and Alloys
\\ \\ \\ \\ .y 3 --
tant control problem for the plating industry (see also this volume, Chap. 10, Sec. 10.1.3).
mechanically polished electropolished
—
\\ \ \\
^^-
(200)
11.3.2 Internal Stresses in Electrolytic Coatings
(200)
Internal stress is an inherent feature of electrodeposited coatings. Its sign and value depend as well on the coating composition being deposited as on the electroplating conditions selected. Sometimes the presence of such a stress can be deleterious, e.g., when it causes distortion of electroformed parts, the appearance of cracks (see Fig. 11-10) or poor adhesion resulting in blistering and peeling. Hereafter, the origin of internal stress is described as well as the principal methods for its measurement. Concerning the origins of internal stresses in electrolytic coatings, different theories exist which all give a different and partial view of the reality. So, for instance, internal stresses in electrodeposits have been related to the absorption of atomic hydrogen during electrolysis and the subsequent escape of molecular hydrogen. The codeposition of foreign products and impurities, which is mainly dependent on the electrolysis conditions, can result in a rather complex stress condition. Another reason for the development of internal stresses in electrodeposits is linked to the driving force of the process itself. The reduction of cations requires an external energy input, part of which can be stored into the growing metallic layer in the form of defects in the atomic arrangement. The dependence of internal stress on plating conditions is clearly illustrated in Fig. 11-11 for additive-free hard gold coatings (De Doncker et al., 1984). At increasing current densities, the internal stresses increase gradually from compressive to tension ones within each current range
y^\
\
^
\
2-
N
^ -
^ ^ ^ (220) 1""*"-" 0 0
^
(111)
1 ^ < 1 2 3 Thickness of Ni-layer in pm
(220)
Figure 11-9. Effect of the mode of substrate preparation on the development of texture in nickel coatings (from Roos et al., 1987).
that of the substrate material. On mechanically polished substrates, the lattice misfit at the interface is too large and inhibits epitaxy. On the basis of such data it becomes possible to determine the stability region of a given texture as a complex function of different plating parameters. The high sensitivity of texture to process parameters however also creates an impor-
Figure 11-10. Microcracked deposited coatings.
zinc-nickel
electro-
11.3 Electrodeposition-Related Properties of Coatings
Fine
0
10
20 30 40 Current density in A/dm 2
50
60
Figure 11-11. Variation of internal stresses in additive free hard gold deposited at different current densities (from De Doncker et al., 1984). Positive stresses are tensile.
wherein the morphology and the texture are constant. Another cause of internal stress in electrodeposits can be the occurrence of a structural transformation induced either by an ageing effect or by a thermal effect. Internal stresses in electrolytic coatings can be linked to a series of parameters such as the structure and the surface condition of the substrate material, the nature of the material to be deposited, the composition of the plating bath, and the electrolysis conditions. Internal stress in thin coatings can be determined either by measuring the macroscopic deformation of a coated sample or by using a non-destructive X-ray diffraction technique. The easiest way to measure internal stress is by using a flexible cathode during electrodeposition. A long small strip used as cathode will tend to deform under the influence of the internal stress present in an electrodeposited coating deposited on top of it. The stress level can be calculated either from the deflection, as in the Brenner-Senderoff contractometer (Brenner and Senderoff, 1949), or from the
497
elongation. In the case of the BrennerSenderoff contractometer, the mean stress in a coating of known thickness can be calculated from the deflection of a spirally shaped metal strip induced by the deposition, on top of the strip, of an electrolytic coating. The stress in the outermost layer of the coating is obtained when any additional deflection is solely caused by the stress in the added layer, and the stress distribution in the underlying deposit is no longer influenced by that added layer thickness. Fig. 11-12 shows some mean stress values determined in this way in Watts' nickel coatings deposited at different current densities. Each mean stress versus coating thickness curve drops towards a plateau value which corresponds to the intrinsic internal stress in the coating. This method allows, in general, the determination of stress during electrodeposition and is thus in addition a convenient in-situ control method. A major drawback of this technique is, as can be seen from Fig. 11-12, the need to deposit thick layers. Any extrapolation towards thin coatings is thus risky.
Watts' bath /•=60°C ; pH = 3.8 Slightly stirred bath
~ 250 / = 5.8 A/dm 2 / = 9.7 A/dm 2 = 11.7 A/dm2,
200 150 0
10
20 30 40 50 Coating thickness in pm
60
Figure 11-12. Stress state in Watts' nickel coatings determined with the Brenner-Senderoff contractometer.
498
11 Electrodeposition of Metals and Alloys
Figure 11-13. Schematic representation of the effect of triaxiality, stress gradient, and texture on the expected linear relationship sin2 \jf versus lattice distance. Triaxiality sin
In that respect the development of X-ray diffraction techniques offers new perspectives, although these techniques are unavoidably off-line methods. Here, the internal stress is calculated from the change in the interatomic plane distance. Thus, the sin2 \\i method based on the Bragg law, predicts a linear dependence between the lattice distance in a given direction and the square of sin \j/ in the case of biaxial stress (the stress normal to the surface being zero) and of a material that has an isotropic crystallographic texture. Measurements are required both in the normal direction and at an oblique angle to determine the stress in the surface at an arbitrary azimuth, and several exposures are needed to determine the magnitude and directions of the principal stresses. For a more detailed treatise the reader should refer to textbooks (Barrett and Massalski, 1966). Complications in using the X-ray method are possible in presence of stress triaxiality, stress gradients and texture, factors which cause deviations from the expected relationship between sin2 \jj and the lattice distance (see Fig. 11-13). For those cases, specific numerical calculation methods have been developed allowing the calculation of the internal stress (Serruys, 1988; Serruys et al, 1987; DeBuyser et al, 1990). It should be emphasized that the calculated stress value is a mean value representative for a coating thickness which corresponds to the penetration depth of the X-rays. For thin to very thin coatings
the use of specific diffraction techniques like the low-angle diffraction, is recommended in order to limit the depth of the X-ray penetration. A comparison of the internal stress values for Watts' nickel coatings deposited from a Watts' plating path (nickel sulfate + nickel chloride + boric acid) determined by using either the contractometer or the diffraction technique is shown in Fig. 11-14. The systematically lower stress levels calculated by the sin2 \fj method could have different origins. A relaxation of the strip coating system between electroplating and the diffraction measurement can be one. Another reason can be an induced stress modification due to a difference in thermal expansion coefficient of strip material and coating material since the diffraction measurements are gen-
Watts' bath
300 "\ \ \» \
E zz
| 250
Slightly stirred bath # contractometer • sin2 ^-method
N, \^
cz
% 200
1 150 — 100 i
0
5
|
10 Current density in A/dm 2
i
15
Figure 11-14. Comparison of internal stresses in Watts' nickel coatings determined with the BrennerSenderoff contractometer and by X-ray diffraction.
11.3 Electrodeposition-Related Properties of Coatings
erally performed at room temperature. In the case of a nickel coating deposited on a stainless steel strip from a plating bath operated at 60 °C, it can be calculated that this effect can account for at most ION/ mm2. Another factor can be a difference due to the thickness of the coating to which the internal stress is related: thick coatings in the case of the contractometer, thin coatings in the case of the X-ray diffractometer. Finally it has to be pointed out that the strip material used in the BrennerSenderoff technique should not have a specific crystallographic texture. Otherwise an estimated error of up to 30% can result from the use of Young's modulus values deduced for textureless materials. (For a fuller discussion of the influence of texture on the diffractometric determination of internal stress, see Chap. 10, Sec. 10.3.7.) 11.3.3 Ductility of Electrolytic Coatings The definition of ductility as "the ability of a material to undergo substantial amounts of general deformation" seems relatively straightforward (Rogers, 1967). The ductility of a metal is however not a unique material property, but rather a system property that varies with stress state, specimen shape, gauge length, temperature, strain rate and environment (Dieter, 1967). Most evidence further suggests that fracture not only depends on the instantaneous stress state but also on the history of the stress development. To understand the extent to which an electrolytic coating can be plastically deformed before fracture, it is essential to have a convenient way of measuring the ductility of coatings. The various methods that have been proposed can be grouped, according to the way the specimens are deformed, as tension tests, bend tests, fatigue tests, hydraulic and mechanical
499
bulge tests. The main drawbacks of the tension test are the difficult of preparation and handling of, typically, 25 |im thin coatings and the non-uniform deformation over the specimen width. As a result, standard testing methods such as ASTM E345 and E879 are not applicable (Parente and Weil, 1971) and the development of appropriate testing methods is still necessary. In order to perform tensile tests on thin coatings detached from their substrate, dog-bone specimens can be prepared with a supporting frame to avoid deformation during handling and mounting (Kim and Weil, 1987). The relative spread, defined as the standard deviation divided by the mean ductility value, can in case of the tensile test be as large as 15% for a ductile homogeneous copper foil. The bend test and the low-frequency fatigue test are considered to be even less reproducible and less adapted to yield quantitative results. In a bulge test, which is in fact a biaxial tensile test, strength and ductility of thin foils can be determined. This test is done by tightly clamping a detached coating over the opening of a test chamber. A force is imposed on the test sample either by applying hydraulic pressure to form a circular bulge, or by a punch head (Fig. 11-15). This bulge test has several advantages over the uniaxial tension tests in that specimen preparation and alignment are simpler, and that the bulge test is less sensitive to defects along the test sample edges. Depending on the experimental set up, the relative spread can be as low as 3% for a ductile copper foil (Xingpu, 1990). However, the bulge test is less suitable for testing brittle materials, owing to the clamping procedure. Furthermore, the strain rate is difficult to control and the determination of the mechanical properties in well-specified crystallographic directions is impossible.
500
11 Electrodeposition of Metals and Alloys
Figure 11-15. Schematic presentation of a stretching test.
Nevertheless, bulge testers have proven to be extremely useful and to be versatile tools for the study of the mechanical properties of electroless and electrolytic deposits (Nakahara et al., 1987). Owing to the attainable low standard deviation, this technique is very well suited for the comparison of commercially available plating baths, or for monitoring the loss of ductility as a bath ages (Rolff, 1987). As regards the bulge test, it should be pointed out that strict control of the test geometry is a prerequisite. Ductility values measured with the mechanical bulge tester are also dependent on the film thickness. Therefore a comparative study of ductility values for various specimens must be carried out on a series of specimens having a constant film thickness. This precaution applies in general for both uniaxial and biaxial test methods. The existence of a definite relationship between ductility, structural characteristics and process parameters has been amply documented over the last few years. That relationship is illustrated here for the case of electrolytic copper coatings (Xingpu, 1990). These 25 jum thick copper coatings were electrodeposited on rotating disk electrodes from acid copper sulphate solutions. The influence of chloride addition, temperature, and different substrate pre-
treatments was studied by polarization curves registered as discrete points. Polarization curves are a very interesting source of information since they are a kind of fingerprint indicating the effect of plating parameters on the electrochemical reduction process at the electrode. The cathodic depolarizing effect of up to 80 ppm chlorides can be clearly deduced from Fig. 11-16. This means that in practice the use of a constant current density in solutions with different chloride contents leads thus to a
0.40 0.42
° A • + •
0.44 0.46
0 mg/L 20 mg/L 80 mg/L 10 mg/L 40 mg/L
Cl Cl Cl Cl Cl
0.48 0.50 0.52 0,54 0.56
3.0
2.5
2.0
1.5
1.0
0.5
Current density in A/dm 2
Figure 11-16. Galvanostatic cathodic polarization curves for solutions with different chloride content.
501
11.3 Electrodeposition-Related Properties of Coatings
deposition of copper at different overpotentials. The ductility of copper foils, detached from their substrates, was measured in a mechanical stretching tester and expressed as the displacement of the punching head through the clamping die. Another series of 25 jam thick copper coatings was deposited at selected and constant overpotentials from a pretreatment with activated carbon. If the ductility is plotted versus the overpotential at which the copper coatings are deposited (see Fig. 11-17), The important role of the overpotential on the ductility of the coatings becomes evident. The use of a low overpotential is apparently an interesting way to optimize the ductility of electrolytic copper deposits. In order to try to understand the dependence of ductility on overpotential, it should be remembered that the driving force for the reduction of the copper ions is the overpotential, and that the extent of this overpotenial will largely determine the growth path of the deposit. Crystallographic texture measurements are a convenient way of quantifying the growth pattern. Texture can be deduced by recording the four most intense diffraction lines, i.e. (Ill), (200), (220) and (311), from which normalized peak intensities can be calculated. From Fig. 11-18 it becomes clear that a strong (220) texture is obtained at higher overpotentials. The presence of this preferential texture can explain the lower ductility. Much more information on the texture of materials is however contained in the orientation distribution function (see Chap. 10, Sec. 10.2.3) from which the yield locus can be calculated (Van Houtte, 1987). A yield locus for an electrolytic copper specimen with a typical (220) texture is shown in Fig. 11-19. The stress ratio Q*, which is the ratio between the plane strain flow stress (ap) and the balanced biaxial strain flow stress (
1.20 1.10
i 1.00 -
g °-
90
i |
-
• •
1 0.80 I 0.70 -
m
+
I
+
1
0.60 n RD
°-5°o.oo 4•
0.10 0.20 Cathodic overpotential in V
without C-pretreatment
t •
0.3
• with C-pretreatment
Figure 11-17. Ductility of 20 am copper electrodeposits as a function of the overpotential at which the layer was deposited.
Figure 11-18. Relationship between overpotential, normalized peak intensity of (220) and ductility.
of the coating, can be calculated from the yield locus and correlated with the ductility. Barlat's theory (Barlat, 1987) lays it down that under biaxial strain conditions the deformability should increase with the stress ratio. Figure 11-20 shows that indeed such a correlation exists between the stress ratio Q* and the ductility of electrolytic copper coatings. Similar observations were also reported for other systems such as electrolytic Pd-Ni on brass substrates (Van Vooren, 1989).
502
11 Electrodeposition of Metals and Alloys
R 60 g*: 1.03 R68p*:1.0
Figure 11-19. Typical yield locus calculated from ODF for a specimen with a (220)-texture (R60) and for a specimen with a (111) texture (R68) (a is the effective stress).
1.03
1.04
1.05
+ a'< 0.04
1.06 1.07 Stress ratio g*
1.08
1.09 1.10
• a'> 0.04
Figure 11-20. Relation between penetration and stress ratio Q* for specimens classified into two different roughness groups.
The dependence of ductility on structural properties is not limited to the crystallographic texture and appears to be rather complex. Thus, the ductility of copper coatings is higher for smoother specimens as shown in Fig. 11-20, where specimens are classified into two groups based on their roughness factor a'. This factor is defined as R a , the roughness average, divided by the coating thickness.
Grain size, or rather cell size, is another structural factor that can influence the ductility of electroplated coatings. An interesting experiment in this respect consisted in the deposition of an electrolytic copper coating from an additive-free acid copper sulfate solution on electroless copperplated ABS plastic and on large-grained copper substrates. Under identical deposition conditions, a very fine-grained (0.5-2 [im) deposit is obtained on the electroless copper layer, whereas large (10-40 \im) grains are induced in the coating deposited on the copper substrate, owing to epitaxial growth. The ductility of the fine- and large-grained deposits is 1 % and 16%, respectively (Nakahara et al., 1987). Finally, some attention should be paid to the way in which ductility is measured. Coatings attached to their substrates will deform in a different way than when they are detached from their substrate. Thus, for nickel, ductilities were reported of up to 12% for coatings attached to their substrates as compared with 1-2% for detached coatings (Vatakhov and Weil, 1990). These examples show how the ductility of electrolytic coatings can be related to a process parameter like the overpotential and to structure characteristics like texture, grain size, surface roughness. How complex the interaction between these parameters and ductility can be becomes evident when one looks at the effect of chloride ions on ductility. Chloride ions at low concentrations act not only as a depolarizing agent and as a leveller which is beneficial in obtaining better ductility, but also promote a preferential crystallographic texture which may cause a decrease in ductility. On the other hand, too high a concentration of chlorides can lead to the formation of pinholes in the coating, and as a consequence to a very low ductility.
11.3 Electrodeposition-Related Properties of Coatings
In conclusion it should be emphasized that in order to gain a real control over the ductility of electrodeposits, process related structural characteristics will need further investigation.
11.3.4 Porosity of Electrolytic Coatings A common problem with functional and decorative electrodeposits is the occurrence of pores, leaving some area of the substrate material uncovered. The porosity of electrodeposited coatings thus has a decisive bearing on the quality and durability of coatings. Knowledge concerning cause and the extent of porosity in coatings is by consequence of paramount importance in coating technology. On the basis of a large number of publications devoted to this subject over the years, porosity can be classified as either inclusion porosity or crystallographic porosity. Inclusion porosity (Leeds, 1969) may arise from small areas on the substrate surface on which, during the early stage of electrodeposition, metal deposition does not occur, although a bridge-over may arise at a later stage. Such small areas which initiate pores, also known as pore precursors, can be non-metallic inclusions such as oxides or sulfides, mill scale, polishing abrasive, grease and dirt present on the substrate surface. Another kind of precursor is minute cavities on a substrate surface. Deposition does not occur on such precursors, either because they are poor electrical conductors or because their electrochemical properties differ so much from these of the clean substrate material that different electrode reaction kinetics favor another process, such as hydrogen evolution, over metal deposition. Porosity arising from structural defects, whether they are caused by the base metal
503
or by electrolysis parameters, is known as crystallographic porosity. In this respect pseudomorphism, which is the continuing of grain boundaries and microgeometrical features of the substrate into the overlying deposit, and epitaxy, which is the orderly relation between the atomic lattices of substrate and deposit at their interface, may induce a lack of coherence between parts of the growing deposit and the substrate. The effect of substrate roughness (Garte, 1966), zoning effect (Clarke and Leeds, 1965) and base metal pretreatment such as mechanical grinding, chemical and electrochemical polishing (Clarke and Chakrabarty, 1970), on the development of porosity has been clearly identified by numerous researchers and through field practice. Zoning refers to substrate materials on which porosity varies periodically with the removal of successive layers from the substrate material before plating. As the coating thickens, the influence of the substrate through pseudomorphism and epitaxy is gradually lost and eventually only the plating conditions at the electrolyte/deposit interface control the coating growth. Indeed, plating parameters such as bath composition, current density and current form, bath temperature and bath agitation also play quite a large role in the development of porosity in electrolytic coatings. Thus, the thickness at which pore-free coatings are obtained varies from a maxium for uninhibited baths and annealed substrate materials to a minimum for bright plating baths and cold-worked surfaces. Apart from the understanding of the origin of porosity, an even more important task is to develop ways to determine the porosity itself. Here an overview is given of methods allowing the determination of porosity in thin metallic coatings. Porosity test methods can be classified into two main categories:
504
11 Electrodeposition of Metals and Alloys
- Pore detection tests which make individual pores observable by visual or microscopic examination. In situ pore counting can be carried out. Examples are the photographic, autoradiographic and microscopy porosity tests, as well as corrosion tests. The photographic test involves the insertion of a coating foil into a frame containing a photographic film. After being exposed to diffuse light and developed, the film shows black spots where pores exist in the foil. In the autoradiographic method (Wolff et al., 1955) a coating is plated over an electrodeposit containing radioactive iron and the radiation emanating through the top coating is measured by exposing a photographic film. Corrosion tests are based on the formation of corrosion products originating from a corrosion reaction of the substrate material with a corrosive medium through pores. Examples are the ferroxyl test for steel substrates and the sulfur dioxide test for gold-plated copper. - Porosity index tests which provide a quantitative measurement of the total porosity of a coating, such as the gas permeability method, chemical analysis and electrochemical porosity measurements. The gas permeability apparatus (Ogburn and Benderley, 1954) consists primarily of two individually controlled sections of a vacuum system which are connected through a coating foil holder. The rate at which the pressure difference changes across the foil is a measure of the permeability of the foil. In the electrochemical techniques one can distinguish between open circuit potential, anodic current flow and polarization resistance measurements. The open circuit potential measurement is based on a linear relationship (Mansfeld, 1971) between the corrosion potential and the logarithm of the fraction of exposed substrate area in the case of a binary gal-
vanic couple controlled by activation polarization. In the anodic current flow measurement a coated substrate is anodically polarized in a solution in which the substrate material becomes much more active than the coating. As a result, the current flowing from the coated sample is approximately proportional to the exposed substrate area, and the current can be used as a porosity index (Ehrhardt, 1960). Many techniques are thus available for measuring porosity in electrodeposits, either as individual pores or as a porosity index. Each method has its own advantages, but also contains certain limitations and none of the available porosity tests is entirely satisfactory. Which porosity test should be used to obtain reliable data is greatly dependent on the specific coating/ substrate system investigated, the coating thickness, the extent of the coating porosity, the size of the pores and so on. Further development of new test methods and improvement of existing tests is thus necessary. The following example illustrates the benefit obtained from a better detection of porosity in thin crystalline and amorphous nickel electrodeposits related to the technological development of pore-free thin coatings. The method used is a modified anodic current flow measurement test, called the coulometric porosity test (Roos et al., 1990). The difference in electrical charge consumed during the anodic polarization of a substrate coated with a porous coating and that of a pore-free coating is considered to be proportional to the exposed substrate area under the porous coating. This difference in electrical charge can be used as a quantitative expression of the porosity, i.e., as a porosity index.
11.3 Electrodeposition-Related Properties of Coatings
As a result it now becomes possible to develop relationships between coating porosity and process parameters related either to the initial surface condition of the substrate material or the plating parameters selected. Thus, the effect of the substrate condition on the porosity of Watts' nickel is shown in Fig. 11-21. For nickel coatings plated onto electropolished substrates (Ra = 0.04 urn), the porosity falls sharply as the coating thickness increases from 1-10 jam. Above 10 }im, the electrical charge consumed is practically constant, so that it is reasonable to assume that Watts' nickel coatings on electropolished substrates are pore-free. On rough substrates such as mechanically ground ones CRa = 1.37 pin), 20 (am thick Watts' nickel layers are necessary to obtain an integrated electrical charge comparable to that consumed on pore-free nickel deposited on electropolished substrates. The beneficial effect of a brightener like saccharin on the lowering of porosity is also demonstrated in Fig. 11-21 (see • curve). The general trend that porosity decreases at increasing coating thickness is also valid for amorphous nickel-phosphorus electrodeposits. The coulometric measurements moreover reveal that at comparable thicknesses below 10 jim, amorphous Ni-P is much less porous than crystalline Watts' nickel coatings. Pore-free amorphous Ni-P coatings are already obtained above 1 |im. The extent of pores in these thin electrodeposits was investigated by transmission electron microscopy and some results are shown in Fig. 11-22. From these micrographs it is evident that the size of the pores is much smaller in amorphous coatings than in crystalline deposits. Mechanisms involving nucleation are generally invoked in the description of metal electrodeposition theory (Bockris and Damjanovic, 1964) and a critical radius of
Watts' Ni on mechanically ground bronze Watts" Ni on electropolished bronze bright Ni on electropolished bronze amorphous NiP17 on electropolished bronze
1000
•s 100
I
505
10
0.3 0.05
0.1
1
10
Apparent coating thickness in urn
Figure 11-21. Effect of coating thickness on the porosity of different nickel coatings deposited on mechanically ground or electropolished bronze substrates.
Figure 11-22. Bright field transmission electron-micrographs of 0.2 jim thick (a) Watts' nickel and (b) amorphous NiP coatings.
506
11 Electrodeposition of Metals and Alloys
a nucleus in an electrodeposit has been identified. This critical radius of a nucleus (rc) can be expressed as: (11-23)
r, =
where d0 is the interatomic distance in the nucleus, y is the edge energy per unit length of the circumference of the nucleus, kB is the Boltzmann constant, T the absolute temperature, and C, C o are the adion concentration at the electrode surface respectively at a given overpotential and at zero overpotential. The adion concentration on the deposited surface in the steady state can be derived (Damjanovic and Bockris, 1963) as:
0.2pm Uc) Watts' nickel on electropolished bronze 6.0
\ (0.03)\
5.0
(11-25) = mrc where p is the coating porosity and m is a proportionality constant. Combining Eqs. (11-23) and (11-25), one obtains:
10y3dgrJV.ro 2zFr]
(11-26)
where Na is Avogadro's number. The role of the activation overpotential in the development of porosity predicted by Eq. (1126) as an inverse dependence, can indeed be deduced from Fig. 11-23 (Chonglun, 1990). This figure shows results of coulometric porosity tests performed on thin
(
N
4.0
\^(0.13) ^ v .
3.0 h-
(0.93)
(0.67) 2.0 400
i
.
-^ i
•
600 800 1000 Activation overpotential in mV
1200
0.2}jm(/(0 Watts' nickel on mechanically ground bronze
35.0 (b)
I 30.0
\(0.02) . \(0.04) L 25.0
I 20.0 S 15.0
where z is the number of charges involved in the electrochemical reaction, F is Faraday's constant, R is the gas constant and rj is the cathodic activation overpotential during plating. If the porosity of thin electrodeposits is assumed to follow a linear relationship with the critical radius of the nuclei in electrodeposition, one should have:
(a)
10.0 400
(0.09) " V
(i/i'i) • (0.62) ^
\
600 Activation overpotential in mV
1200
Figure 11-23. Porosity index vs. activation overpotential for 0.2 |um thick Watts' nickel coatings deposited on (a) electropolished and (b) mechanically ground bronze substrates.
nickel coatings deposited at different current densities. In conclusion it can be stated that although porosity in electrodeposits has been recognized long ago as being dependent on the surface condition of the substrate material as well as on the plating parameters used for electrodeposition, actual insight into the relative importance of these parameters for the development of porosity is far from being well developed. The further development of more efficient techniques for the determination of pores in thin coatings will probably open new possibilities of depositing in a controlled way pore-free coatings at decreasing layer thicknesses. Optimization then becomes feasible.
11.3 Electrodeposition-Related Properties of Coatings
507
113.5 Corrosion and Wear Protection by Electrodeposits
Concepts such as "wear resistant coatings" and "corrosion resistant coatings" are used not only in everyday life, but also in technical reports. It must however be stressed that wear and corrosion resistance are not material properties, but system properties. This approach has been systematically developed, as in case of tribology (Czichos, 1978) and also provides the philosophy for the fundamental standard on wear mentioned in DIN 50320 (DIN 50320, 1979). Herein wear is defined as follows: "Wear is the progressive loss of substance from the surface of a solid body due to mechanical action, i.e., contact and relative motion of a solid, liquid or gaseous counterbody". This definition reflects the fact that wear is always a result of the interaction of different structural parts or technical components. In Fig. 11-24 the description of a tribological system is visualized. The directly involved parts and materials are called the elements of the tribosystem; these are the elements (1) and (2), the interacting bodies, the interfacial medium (3), and the surrounding medium (4). The external parameters acting on the elements of the tribological system, such as load and relative velocity, represent the operating variables. Wear is the result of the action of the operating variables on the structure of the tribological system. The principal characteristics of the operating variables can be summed up as: type of motion (impact, sliding, rolling, etc.), time dependence of the motion (continuous, oscillating), stress distribution, velocity, temperature and time. The characterization of the structure of a tribological system requires a great deal of knowledge. The four elements presented in Fig. 11-24 have to be clearly identified and
r
Structure of the tribological system
Figure 11-24. Description of a tribological system according to DIN 50320. (1) and (2) wear couples, (3) interfacial medium, (4) surrounding medium.
characterized with regard to materials properties and geometry. Furthermore, the physical and chemical processes occurring at the interface between the elements have to be described, so that the prevailing wear mechanism(s) can be identified. Four distinct wear mechanisms have been described (cf. DIN 50320): adhesion, i.e., the formation and rupture of interfacial adhesive bonds; abrasion or the removal of material by micro-cutting processes; surface fatigue; and tribochemical reactions, i.e., the development of reaction products as a result of chemical reactions taking place between the wear couple and the interfacial medium. From the foregoing, it becomes clear that the analysis of a wear problem requires the compilation of extensive data. A checklist for describing a tribological system, such as the one in DIN 50320, is then quite helpful for clearly stating the problem and for setting up wear tests. In relation to the wear testing of coated structures, two different options can be pursued. In the first one the coater examines whether the deposit fulfils the requirements imposed by a standard method, or
508
11 Electrodeposition of Metals and Alloys
not, or checks the wear resistance characterized by a numerical quantity such as a weight loss (ASTM D4060-84, 1984) or a maximum applicable load (ASTM G83-89, 1989). If the testing procedure is strictly controlled, reliable comparative results can be obtained, and it becomes possible to classify different materials. It is however clear that the results of these tests only seldom give a true classification and prediction of the performance of a given coated material in a specific industrial application. Indeed, the latter requires a simulation of the real wear system; this is the second option. Owing to the complexity of wear, numerous test methods have been developed. More than a hundred friction and wear test devices have been described (Benzing, 1973). As can be deduced from Fig. 11-24, a discussion on wear tests and wear test procedures requires at least the characterization of the coating and the coating/substrate interface. First of all, good adherence of the coating to the substrate is required. Applications involving wear therefore require an adequate, adapted pretreatment of the surface. In most cases, the substrate roughness will also play a very important role since it influences the morphology and the topography of the coating. Especially in the case of a cyclic or impact load, the best performance is obtained when chemical and mechanical properties of substrate and coating material are matched. In some cases one can employ multilayers, as for instance a two-layer system consisting of an electrolytic nickel and an electrolytic chromium layer, an electroless nickel and an electrolytic chromium layer or a double layer composed of a microcracked hard chrome layer on top of a more ductile, nonporous but less hard chrome plating. This last layer can be obtained by operating the chromium plating bath at a higher temper-
ature, or by the use of pulsed currents. The resulting difference in the structure of chromium deposits is illustrated in Fig. 11-25. For applications where high loads are applied, the selection of a substrate having the required mechanical strength is a prerequisite, so that the deformation of the loaded component does not lead to a cracked coating. A striking example of how the wear behaviour is dependent on the structural properties of a deposit is worked out further in Sec. 11.4.1.2, where it is demonstrated how an optimization of the wear properties of cobalt-hardened gold electrodeposits as top coating for electronic connectors can be realized.
Figure 11-25. Structures of chromium coatings plated from a bath operated at (a) 55 °C and 60 A/dm2, (b) at 70 °C and 100 A/dm2.
11.3 Electrodeposition-Related Properties ot Coatings Table 11-1. Standard electrode potential series relative to SHE (T= 298.15 K). Electrode
Eo (Volt)
Reaction
Au++/Au Ag + /Ag Cu++/Cu
+ 1.35 + 0.80 + 0.34
Au + "' +2e--+ Au + e~ - • Ag Ag + Cu + "" + 2 e ~ - • Cu
H+/H2 ++
Pb /Pb Sn++/Sn Ni++/Ni Fe++/Fe Cr + + + /Cr Zn + + /Zn A1+ + +/A1 Mg++/Mg Na+/Na
0.00 -0.12 -0.14 -0.24 -0.44 -0.60 -0.76 -1.67 -2.39 -2.71
H+
+ e~
Pb + H h Sn + + Ni + + Fe+4• Cr + + + Zn + H r Al + + + Mg + + Na +
+ 2 e " ^ Pb + 2 e " - > Sn + 2 e ~ ^ Ni + 2 e ~ ^ Fe +3e~ -> Cr + 2 e " - ^ Zn +3e~ - Al + 2 e " - • Mg + e~ - • Na
~^ 1/2 H 2
Figure 11-26. Principle of the galvanic protection of steel by zinc.
Corrosion is usually defined as the deterioration that occurs when a material reacts chemically or electrochemically with its environment. From this definition the system related character is immediately revealed. The investigation of a corrosion problem thus requires, in analogy with wear, a careful description of the elements involved, i.e., the structure of the corrosion system, the operating variables and the corrosion mechanism. If the system is well described, test procedures can be developed to simulate the actual corrosion process and to evaluate the performance of different coated materials. Unfortunately there is no widely accepted guide like the DIN mentioned previously for wear. On
509
the contrary, many standard corrosion test procedures are available, even for coated systems (ASTM Bl 17-85, 1985; DINTaschenbuch 175,1989). This means that a careful selection of the testing procedure should be carried out for each application. As regards their corrosion resistance, electrodeposits have to be distinguished according to whether the coating material is more or less noble than the substrate material. A rough idea as to which situation applies can be obtained from the electrode potential series as shown in Table 11-1. It should however be emphasized that coatings which behave anodically to the substrate in a given environment, may become cathodic in another one. Furthermore, the environment itself may change from one on the surface of a coating to one that exists within defects such as scratches, pores or cracks in a coating. In atmospheric conditions, zinc or cadmium will protect a steel substrate by cathodic protection (Fig. 11-26). The quality of the protection offered will now strongly rely upon the thickness of the coating, the corrosion resistance of the protective coating itself and its ability to form a self-healing surface film. In the case of anodic protection, a real barrier layer has to be formed between environment and substrate by interposing a more corrosion resistant layer. A first requirement for the coating then is its capacity to act as a barrier in the corrosive environment in question. Secondly, the coating has to be exempt from porosity or cracks. This situation has to be maintained during the actual lifetime of the deposit. As a consequence, the ductility of the deposit must be high enough to follow a deformation of the substrate induced either by a thermal expansion or a mechanical interaction. How properties like porosity and ductility are related to process parameters was dis-
510
11 Electrodeposition of Metals and Alloys
cussed in Sees. 11.3.3 and 11.3.4. For further information, including the intrinsic corrosion resistance in specific environments, interested readers are refered to the extensive literature on this topic (Metals Handbook, 1987). As a case study, chromium plating is discussed here. Chromium is very resistant to atmospheric corrosion, but is soluble in HC1 or in alkaline solutions, as can be deduced from Pourbaix diagrams (cf. Sec. 11.2.4). This corrosion resistance is related to the formation of an amorphous oxide that protects the otherwise strongly electronegative metal. Electroplated chromium is mainly applied for three types of applications, namely as TFS or tin-free steel, as decorative coatings and on components requiring high wear and/or corrosion resistance. Chromium coatings less than 0.05 |im thick are applied to steel sheet for packing purposes. The chromium layer of this TFS is further supplemented by an oxide and/or an organic coating. A second application is as a decorative coating deposited on top of a nickel undercoating. The chrome coating with a thickness of about 0.5 jam is used for its scratch and tarnish resistance, and brightness. The corrosion resistance relies mainly on the nickel undercoat since the chrome itself is not only porous at these low thicknesses but also shows an extended crack pattern. Through the use of microdiscontinuous chromium with even more pores or cracks, it is possible to distribute the corroding currents more evenly over the surface and minimize in this way the anodic current density at each point of attack. Microcracked chromium is obtained by the use of additives in the plating bath that increase the internal stress in the deposit or by using a highly stressed nickel undercoat which induces the microcracking. Mi-
croporous chromium is obtained when chromium is deposited on a duplex nickel layer consisting of a bright nickel layer on which a second thin nickel layer is deposited containing a multitude of very fine non-conductive particles. This last process is known as composite plating and is extensively treated in Sec. 11.4.2. A pore density of up to 8000 to 16000 pores/cm2 can be obtained in this way (Lowenheim, 1974). Chromium is also electrodeposited at greater thicknesses of 2.5 |^m and more. The basic solution for the deposition of chromium contains 250 g/L CrO 3 and 2.5 g/L H 2 SO 4 . How the abrasive wear expressed as the wear volume (V) of the chromium layer is related to deposition process parameters like current density and temperature is shown in Fig. 11-27 (Morisset, 1982), while differences in structure were already revealed in Fig. 11-25. From this last figure it is clear that a porefree chromium layer can act as a protective barrier against corrosion. On the contrary, a chromium layer with a microcracked
60
40
l/ = volume [mm3-10~3] Cr03 = 250g/L S04""= 2.5g/L
20
0
0
20
40
60
80
100
120
Current density in A/dm2
Figure 11-27. Wear resistance of chromium coatings as function of current density and temperature (from Morisset, 1982; reproduced by permission of CETIM).
11.4 Materials and Process Developments in Alloy Electrodeposition
structure, typically showing 400 cracks/ cm, is preferred for applications involving thorough lubrication. Indeed, the wettability of chromium is rather poor. By introducing microcracks, the wettability of the chromium plating is improved through capillarity effects, and a homogeneous distribution of the lubricant over the whole surface of e.g. a piston can be realized in this way. Another interesting coating often used as a substitute for chromium, or in conjunction with chromium is electroless nickel. One of the main drawbacks of chromium is the very poor macrothrowing power of the chromium plating solution. Macrothrowing power is the ability of a bath to produce deposits of more or less uniform thickness on cathodes having macroscopic irregularities. Electroless nickel plating based on a heterogeneous autocatalytic reaction does not suffer from uneven deposition rate along surfaces of complex shape. This plating process is extensively discussed in Sec. 11.4.4.
11.4 Materials and Process Developments in Alloy Electrodeposition 11.4.1 Alloy Plating The possibility of codepositing simultaneously different elements from aqueous solutions is the basis of alloy plating. Major advances in electroplating technology over the last decade have accompanied progress in the electronics and automotive industries. Developments associated with applications such as electronic contacts, printed circuits, thin-film technology, bearings and corrosion protection of steel plate are linked with the evolution of electroplating science and technology. New processes and on-line control procedures have
511
been developed to provide coatings which meet the high performance levels required by mass production and environmental compatible discharge of wastes. Alloy plating is arousing more and more interest in the present electroplating technology developments. For more than a century, alloy plating has been of interest to the metal finisher. From a literature survey, one will discover several hundred alloy plating systems reported, and partly patented. Using the widest definition of "alloy", several classes of alloy coating systems can be recognized: Class I: Coatings are applied as successive layers of single metals to the surface. Typical examples are nickel/chromium plate, chromium top layer on dual or even triple nickel deposits, nickel underlayer/nickelpalladium/hard gold flash top layer. Through a subsequent heat treatment, an alloyed layer may arise by interdiffusion or may even result from a diffusion with substrate components. Class II: Coatings are directly deposited as alloys, the composition being almost constant from the start at the original substrate surface. Examples are brass coatings on steel cord used in order to improve the adhesion between steel and rubber, zincnickel as corrosion protection layer on steel plates, and cobalt-hardened gold as wear resistant layers. Through a subsequent heat treatment, intermetallic phases can be formed, as for instance in nickelphosphorus electrodeposits, so that precipitation hardening can be obtained. Class III: Coatings having an heterogeneous structure can be obtained through the incorporation of solid second phase species in a metallic matrix. Examples are composite and occlusion plating to produce, for example, nickel-alumina and
512
11 Electrodeposition of Metals and Alloys
cobalt-chromium oxide coatings. This class of alloy coatings is treated separately in Sec. 11.4.2. The mechanisms of alloy deposition for the three classes of alloy coatings mentioned above are quite different. Class I plating can be described in terms of general electroplating of pure metals: M
alloy
/ A
/r
B
(11-27)
Intermediate pretreatment operations to clean and activate the surface play an important role in coating initial growth and final coating adhesion. The cathodic current is related to the overpotential which is the driving force for reaction, Eq. (11-27), to proceed. The importance of the type of overpotential, namely charge-transfer or concentration overvoltage, appears fully when surface morphology and effectiveness of addition agents like brighteners and stress relievers are considered. In Class II plating, it is essential that two or more elements should codeposit at the same electrode potential. The simplest case is the one where the polarization curves of the single elements can be added up to give the polarization curve for the alloy deposition (Fig. 11-28). However, sometimes it is required to use selective complexants in order to bring the deposition potential of the elements closer to each other or to influence the polarization of the elements at the operating deposition current. As a consequence, in practice the relative position of the polarization curves for the single elements and for the alloy differs greatly from the situation shown in Fig. 11-28. From an engineering point of view the relationship between the coating composition and the plating conditions is of importance. So, for instance, five types of alloy electroplating solutions can be distinguished (Fig. 11-29) (Brenner, 1963). The normal solutions are those which behave
metal b
metal a
C AD = BD + CD Mh
MQ D Cathode potential +
^~ -
Figure 11-28. Schematic representation of the polarization curve of an alloy considered as the sum of the partial polarization curves of the base elements (from Brenner, 1963).
20
40 60 80 Metal a in electrolyte in %
100
Figure 11-29. Relationship between bath composition and coating composition for different types of alloy plating.
according to a simple theory of potential and current density prediction, e.g., regular solutions having simple ions and diffusion control, irregular solutions having one or more elements present as complex ions and equilibrium solutions for which the element ratio in the bath is equal to the element ratio in the deposit. The abnormal solutions are those in which the more noble element does not deposit preferentially as, for instance, in anomalous solutions
11.4 Materials and Process Developments in Alloy Electrodeposition
introduction and breakthrough of new plating technologies such as pulsed plating and high-speed plating. Pulsed plating allows, by acting on the double layer composition, the production of denser deposits and a more constant composition over the coating thickness. In order to solve some permanent problems in alloy plating, like the strong dependence of structural characteristics of coatings on small variations in plating conditions, and a limited deposition rate, the use of high fluid flow in electrolytic cells can be very attractive. Recent major technological improvements in electroplating on electronic connectors were indeed related to the enhancement of the hydrodynamic fluid flow as shown schematically in Fig. 11-30 (Roos et al., 1986). However, the introduction of such techniques requires a good insight into the effect of hydrodynamics on the electrolytic deposition process. Basically, the enhancement of the fluid flow is related to the larger mass transport achieved in this way towards the electrode. As a result, the lim-
where the less noble element deposits preferentially owing to a differential polarization or a complexing effect, and induced systems where elements which do not deposit individually may deposit as an alloy. The acquisition of a better insight into the different mechanisms of alloy plating has been very beneficial for the wider introduction of alloy plating in plating shops over the past decade. It is not our intention to give here a complete review of alloy plating systems which have already reached full industrial development. Only two specific aspects of great importance in alloy plating will be discussed, namely the effect of hydrodynamics on alloy plating and the optimization of alloy plating based on the unraveling of deposition-structurefunctionality relationships. 11.4.1.1 Effect of Plating Bath Hydrodynamics The future development of alloy plating will probably be linked to the commercial
Electroplating process
Barrelplating
partial immersion
0
i
spraying
selective spraying
513
Relative gold consumptions (%)
Rate of deposition (pm/min)
100
0.2
80
1.2
25
12
8
16
1
Figure 11-30. Recent developments in the process technology of electroplating on connectors.
514
11 Electrodeposition of Metals and Alloys
iting current density (see Eq. (11-18)) is increased, allowing electrodeposition of smooth deposits at higher nominal current densities. A judicious control of both mass transport and potential field conditions will result in an electroplating process exhibiting combined high speed and high selectivity. Some attractive cell geometries for high speed electroplating are schematically shown in Fig. 11-31. For some of these cell geometries, e.g., the rotating disc (RDE) and cylinder electrodes (RCE), mathematical equations describing the potential and concentration fields are well established (Levich, 1962; Gabe, 1974). For industrially more attractive high speed cells, these equations are rather difficult to solve, although, by identifying the important controlling factors (Landau, 1981) and by simplifying the equations accordingly, it is possible to derive useful process describing equations. Thus, quantitative calculations of the three-dimensional current dis-
RDE
RCE
;t© ;©!! ! Jet cell
\ cathode area
motion
•* fluid flow
Figure 11-31. Some attractive cell geometries for electroplating under high fluid flow conditions.
tribution on electrode surfaces by computer (Shih and Pickering, 1987) are available as well as dimensionless contour mapping, indicating generalized conditions under which high-speed electroplating can be accomplished with submerged and unsubmerged fluid jet cells on flat surfaces (Alkire and Chen, 1982) or in a circular through-hole (Alkire and Ju, 1987). Related to the electrodeposition of alloys under conditions of high mass transport, modelling is still rather limited, mainly because of a limited insight into the mechanism of the interaction between metals during the plating process. For the anomalous deposition of zinc-nickel alloys on a rotating electrode, the possibility of a fundamental modelling over a wide range of process variables has already been demonstrated (Mathias and Chapman, 1987). Experiments have demonstrated that the amount of nickel in the deposits shows a minimum at increasing current densities. This minimum is less pronounced once the fluid flow is increased. For regular copper-nickel alloy plating, the effect of plating bath hydrodynamics is illustrated here. A certain interest exists in the electrodeposition of copper-nickel alloy as decorative, anti-fouling, and corrosion-resistant coatings. The standard electrode potential of copper, which is +0.37 V vs. SHE, lies about 0.6 V higher than that of nickel. Therefore Cu-Ni alloy plating requires the use of complexing agents such as citrates. Appearance of typical Hull cell plates obtained in unstirred citrate solutions (Fig. 11-32) reveals the current density range at which copper-nickel alloys are deposited. Any imposed bath agitation however causes a disappearance of the gray-colored nickel bands. The reason becomes evident when one considers the polarization curves recorded on rotating cylindrical electrodes.
515
11.4 Materials and Process Developments in Alloy Electrodeposition
0.2 NCu/0.2NNi 0.18
0.51
0.05NCu/1.0NNi 0.18
|
pH: 4.37 0.97
pH: U 1 0.51
0.97 1.76 A/dm 2
| bright nickel-like semi-bright copper-like
|
1.76 A/dm
2
non-bright copper-like burnt
| bright copper-like
Figure 11-32. Appearance of Hull cell plates prepared at 20 °C in unstirred Cu-Ni solutions containing 75 g/L Na-citrate.
Concerning the mass transport phenomena on such electrodes (Gabe, 1974), it is known that no convective mass transport occurs towards the cylinder surface under laminar flow conditions, while at increasing rotation speeds Taylor vortices appear first and further on turbulent flow will cause a drastic increase of the mass transport towards the surface of the electrode which is then uniformly accessible. Polarization curves recorded on such rotating cylinder electrodes at different rotation speeds are shown in Fig. 11-33. Starting at potentials of about — 0.05 V vs. SCE a reduction of copper ions occurs, resulting in an increasing current density at decreasing electrode potential until when low rotation speeds are used a limiting current density (see a in Fig. 11-33) for the
reduction of copper is reached. This corresponds respectively to the bright and semibright copper-like zones in the Hull cell plates (Fig. 11-32). For such low rotation speeds, once the reduction potential of nickel is reached (see b in Fig. 11-33), a second current increase is noticed. From there on, copper-nickel alloy deposition is realized which corresponds to the bright nickel-like zone on the Hull cell plates. At much lower potentials the reduction of hydrogen causes bath agitation near the electrode resulting in a deposition of copper owing to an increased limiting current density. On Hull cell plates, this corresponds to the non-bright copper-like zone. Evidence here is found from the polarization curves recorded at higher rotation speed (see c in Fig. 11-33) where no limiting current for the copper reduction appears within the tested current range. In conclusion, it can be said that coppernickel alloy plating from citrate solutions cannot be achieved under high fluid flow regimes unless the relative concentration of the more noble element in the plating solution is drastically reduced.
1.00 mol/L Ni 0.05 mol/L Cu -0.2
75g/L Na-citrate
5 -0.4 Z-0.6 "-0.B -1.0 0.12
0.10 0.08 0.06 0.04 Current density in A/dm 2
0.02
Figure 11-33. Polarization curves recorded at different rotation speeds of a RCE in a copper-nickel citrate solution.
516
11 Electrodeposition of Metals and Alloys
11.4.1.2 Optimization of Deposition Parameters The link existing between the electrolysis parameters and the structural properties of coatings has been demonstrated in Sec. 11.3. In alloy plating it is evident that such an inter-relationship also exists and should be identified in order to optimize plating procedures. As a case study, the optimization of gold deposits for electrical connectors through the unraveling of inter-relationships between processing, structure, and functionality of electrodeposited coatings, will be discussed. Since pure gold coatings suffer from galling wear, attempts were made to develop hardened gold electrodeposits, e.g., by codepositing cobalt. Recent research (De Doncker and Vanhumbeeck, 1985) has demonstrated that this cobalt is incorporated either as metallic cobalt (Com) or as complex cobalt (Cok) species. The latter are codeposited together with potassium, carbon, nitrogen, oxygen and hydrogen. However, cobalt- hardened gold coatings are not necessarily wear-resistant nor have the required low and stable contact resistance. On the basis of an in-depth characterization of the electrodeposition conditions and the structural characteristics of the coatings (Roos et al., 1987; Celis et al., 1989), a quantitative criterion was developed which is helpful in producing cobalthardened gold deposits with optimum wear and contact resistance. In order to generate the required data, a large number of plating conditions were used, ranging from d.c. plating to pulsed plating, using low and high amounts of cobalt and gold salts in the plating solution, low to high current densities and temperatures, rotating disc electrodes and jet plating cells operating under moderate and high fluid flow conditions.
The structural investigation comprised morphology observed by scanning electron microscopy, texture and internal stresses measured by X-ray diffraction and hardness determined by micro-Vickers indentation. The potassium content was determined by flame emission spectroscopy, the grain size by transmission electron microscopy and the metallic and complex cobalt contents were determined by differential pulse polarography. As for functional properties, the wear resistance was tested by the flat-on-rider method, and the contact resistance was determined by a 4-point method before and after heat treatment. From a comparison of rotating disc and jet cell results it appeared that structural blocks similar to the one shown in Fig. 11-34 can be constructed. The dominating role of the overpotential on the structural properties is evident. Three kinds of coatings were observed, namely, - wear-resistant cobalt-hardened gold coatings with a low friction coefficient which remains stable during wear tests; - cobalt-hardened gold coatings, with a high friction coefficient, that suffer from galling wear; - cobalt-hardened gold coatings, with a moderate friction coefficient which slightly increases during wear tests, that suffer from abrasive wear. No correlation with either the plating conditions or some structural parameters generally recognized as wear-determining, namely hardness and potassium content, could be found. Regardless of the electrodeposition conditions used, it was however possible, as shown in Fig. 11-35, to deduce a wear criterion solely based on three structural parameters, viz., the ratio of the incorporated amount of complex to metallic cobalt, the internal stress, and the
11.4 Materials and Process Developments in Alloy Electrodeposition
20pm
3.5 p m
21pm
Morphology
Fine grained
Transition
Lenticular
Nodular
Dendritic
Appearance
Dull
Dull
Dull
Bright
Bright
(220) • (311)
Textureless or (1111 + 1311)
Texture
(111)
I2201 + I311)
Increasing overpotential
grain size. From this figure it is evident that a prerequisite for making wear-resistant cobalt-hardened gold coatings is that the ratio of complex to metallic cobalt should be larger than 1, otherwise the coatings will suffer from either galling or abrasive wear. The reason why the Co k /Co m ratio correlates so well with the wear behaviour becomes clearer when the other relevant structural parameters are considered. A low Co k /Co m ratio can be obtained in two ways: either the complex cobalt amount in the deposits is low, or the amount of metallic cobalt incorporated is high. Coatings with a high metallic cobalt content appeared to have, regardless of the deposition conditions used, besides a small grain size, high internal stresses which promote abrasive (or brittle) wear. On the other hand, coatings with a low amount of complex cobalt are characterized by a large grain size and a low internal stress, conditions which favour galling promoted by a high
517
f\
Figure 11-34. Structural characteristics of cobalthardened gold coatings deposited on a rotating disc electrode.
wear resistant
ff| abrasive wear ifl m HI W
adhesive wear industrial bath with Co Industrial bath without Co AFHG
Figure 11-35. Wear behavior versus structural properties for jet-plated cobalt-free and cobalt-hardened gold coatings (from Celis et al., 1988). (AFHG = additive free hard gold plating bath.)
518
11 Electrodeposition of Metals and Alloys
ductility and a lack of lubrication due to the absence of complex cobalt cyanide species. From this case study it can be concluded that it is indeed possible to derive relationships between electrolysis conditions and relevant structural properties. Such relationships open the way to a straightforward but still bath-dependent optimization of the deposition parameters to achieve coatings having the desired functional properties. Further studies on the role of overpotential may eventually deliver a process optimization criterion independent of the bath type used. 11.4.2 Composite Plating Composite plating refers to the ability to codeposit in a controlled way during the electrolytic deposition of metals, secondphase particles which are suspended in the plating solution. This technique has been recognized over the past decade as a technique which offers the possibility of producing a wide variety of coatings with unique properties. The term 'codeposition' is in fact a quite common expression used in plating technology, but it has different meanings. Codeposition is used in alloy plating to denote the simultaneous electrochemical reduction of different metallic ions present in a plating solution, such as copper and zinc ions in brass plating. Codeposition is also used in relation to the incorporation into electrolytic deposits of additives such as stress-relievers and brighteners added to the plating solution. The incorporation of complex species in deposits is also referred to in the literature as codeposition. This was at the basis of the development of cobalt-hardened gold deposits showing an increased wear resistance. Such species, present as charged colloids, are formed during the metal deposi-
tion process as, for instance, a consequence of a local pH increase near the cathode. Here, codeposition is used specifically to denote the process of incorporation of small solid particles added deliberately to the plating solution. These particles are kept in suspension either by mechanical or chemical means. The chemical approach generally requires the addition to the plating solution of appropriate surfactants, while suspension by mechanical means is realized by the use of a vibrating perforated bottom plate or by the use of air agitation. A typical morphology of such a composite Cu-SiC coating is shown in Fig. 11-36. The mean size of SiC particles in that figure is 3 ^im. From a technological point of view, two electrolytic techniques allowing the codeposition of particles can be distinguished, namely, occlusion plating and composite plating. In occlusion plating, particles are generally allowed to settle down onto cathode surfaces while some intermittent vigorous mechanical stirring of the plating solution is applied. Advantages of this technique include the possibil-
Figure 11-36. SEM of the surface morphology of an electrolytic Cu-SiC composite coating.
11.4 Materials and Process Developments in Alloy Electrodeposition
ity of incorporating a much larger volume percentage of particles in comparison with the composite plating process on vertical electrodes and of producing stratified coatings by imposing periodically a controlled bath agitation. A major limitation is the fact that the amount of codeposited particles is strongly determined by the shape of the cathode, and only applicable for almost horizontally flat surfaces. The first time research laboratories became interested in the codeposition of particulate matter during electrolysis was in the early sixties. Since that time a lot of research and development work has been done and in recent years some interesting industrial applications of composite plating have been realized. Four major fields of interest can here be indicated, namely, self-lubricating coatings such as silvergraphite and nickel-PTFE composite coatings (Helle and Opschoor, 1980), dispersion-strengthened coatings such as copperalumina coatings, wear and oxidation containing silicon carbide particles or cobalt layers containing chromium oxide particles (Ruimi and Martinou, 1989; Thoma, 1985), and decorative coatings like satin nickel. The use of composite plating is spreading from the high tech environment of aviation and aeronautics, where cost is not of decisive importance, to the general consumer industries. Another trend is the noticeable transition at present from limited series applications, e.g., selective composite plating on shafts, towards mass production, e.g., the current development of composite zinc coatings on steel sheets for the automotive industries (Kazumi et al., 1987). Such composite zinc-silica coatings have a lower corrosion rate than pure zinc, and an increased adherence with organic toplayers is achieved. Actual know-how on the galvanic composite plating allows one
519
to describe different aspects of that process, e.g., the influence of the plating parameters on the amount of codeposited particles, problems linked with the composite plating practice and the mechanism of particle incorporation during electrolysis. All this information is essential for successful deposition of high-quality composite coatings and to ensure reproducible production in an industrial environment. 11.4.2.1 Electrolytic Codeposition of Solid and Liquid-Containing Particles
With conventional electrolysis from aqueous solutions, two types of metallic matrices can be obtained, namely, a pure metal or an alloy. Every metal or alloy that can be electrolytically deposited can be used as the matrix of an electrolytic composite coating. Two remarks have to be made in this respect. First of all, when alloy plating is performed in the presence of particles in the vicinity of the cathode, the composition of the alloy matrix may differ from that obtained in pure alloy plating. Secondly, electrolytic codeposition offers the possibility to form alloys which are difficult to obtain by straight electrolysis, e.g., stainless steel coatings. By diffusion of chromium from chromium particles codeposited with an iron-nickel matrix, a stainless steel matrix is obtainable. Besides the choice of the metal matrix, one also has to select the second phase particles, which can be in powder or fiber form. Two types of powders can be distinguished, namely conductive or non-conductive ones. As practical particle size an upper limit of about 40 jam is imposed by practical problems of keeping larger particles in suspension. The question as to which electrolysis conditions have to be chosen in order to obtain composite coatings with well-de-
520
11 Electrodeposition of Metals and Alloys
fined characteristics is difficult to answer. The main reason is that the mechanism of the electrolytic codeposition of particles with metals is far from being fully understood, so that the influence of electrolysis parameters on the codeposition cannot yet be predicted. Notwithstanding this, from the available information on composite plating practice some rules can be derived about the influence of some important plating parameters on the amount of particles codeposited with metals during electrolysis (Celis and Roos, 1984). A schematic representation of the influence of four main plating parameters is given in Fig. 11-37. The relative position of the curves in these graphs is principally determined by the combined effect of the characteristics of the particles and the composition of the plating solution. Thus, codeposition with copper is lower from sulfate plating baths than from cyanide baths. Conductive particles give much higher codeposition than non-conductive
Concentration of particles in suspension
Current density
Hydrodynamic flow
Figure 11-37. Schematic representation of the dependency of the amount of codeposited particles versus four main electrolysis parameters: (a) charge transfer overvoltage; (b) concentration overvoltage; (1) laminar flow; (2) transition zone; (3) turbulence (from Celis and Roos, 1984).
ones. Large particles give (per embedded particle) much higher volume percent codeposition than do small particles. A pretreatment of particles in appropriate ionic solutions can improve the codeposition to a large extent, while addition of anionic surfactants generally reduces or even prevents the codeposition of particles and low pH-values increase the adsorption of protons onto particle surfaces at the cost of metallic ions. One of the few general prevailing rules is that an increased codeposition of particles is obtained at increasing concentration of particles suspended in the plating solution. As a conclusion it can be put forward that composite plating is a reproducible process on the condition that important electrolysis conditions like local current density, bath composition and fluid flow along the cathode are kept constant. A last remark is related to a possible interaction between particles and some constituents present in the plating bath which sometimes can be at the origin of an ageing effect. The latest addition to the growing list of industrially used composite coatings is the development of liquid-containing composite coatings. Such a codeposition of liquids is a logical extension of the classical codeposition of solids and creates possibilities for the production of novel engineering coatings. The technique relies on the codeposition of liquid-containing microcapsules (Fig. 11-38) (Fransaer et al., 1989). The production of the microcapsules can be based on a coacervation technique wellknown from pharmacology. The microcapsules consist of a liquid core surrounded by a membrane. The liquid can be an aqueous solution containing specific additives, organic solvents or even an oil. The only prerequisites for a successful codeposition of such microcapsules are
11.4 Materials and Process Developments in Alloy Electrodeposition
Figure 11-38. SEM of water-containing microcapsules embedded in a copper matrix.
that the microcapsules should be chemically resistant to the plating conditions and should withstand the mechanical stresses during plating. The membrane should not necessarily consist of conductive material but must be wetted by the plating solution. 11.4.2.2 Mechanism of Electrolytic Codeposition
In the early days of composite plating, three possible mechanisms of codeposition were put forward, namely electrophoresis, mechanical entrapment and adsorption of particles on the cathode. No reliable evidence that electrophoresis plays a major role has yet been obtained since most electrophoretic measurements are done in dilute solutions and a hazardous extrapolation towards plating solutions of high ionic strength has to be carried out. Concerning a possible adsorption effect on particles by van der Waals attractive forces, the effect of monovalent cations on the codeposition has been pointed out (Tomaszewski et al., 1969). Thallium ions, for example, drastically enhance the amount of codeposited particles from copper sulfate plating baths. As more and more practical data concerning the effect of electrolysis parameters on the codeposition of inert particles with
521
metals were gathered, a few attempts were made to cast the experimental data into models. An important step forwards was the mathematical model derived by Guglielmi (1972). On the basis of a striking similarity of the curves relating the volume percent of codeposited particles (vol.%) to the volume percent of particles suspended in the plating bath (Cv) with the well-known adsorption isotherm, a two-step mechanism was postulated in which the combined effect of adsorption and electrophoretic attraction is held responsible for the encapsulation of particulate matter in a growing electrodeposited layer. A loose adsorption, which has an essentially physical character, results in a higher degree of coverage of the cathode by particles. No real contact exists at that time between cathode and particle. A subsequent strong adsorption, which is thought to be field-assisted and therefore electrochemical in nature, permits the entrapment of particles in the growing metal layer. The formula so deduced is (11-28) C
vol.%
/ • iB/A o
nFQmv0
1 k*
where F is Faraday's constant; n is the valence of the electrodeposited metal; /c* is the Langmuir isotherm constant, largely determined by the intensity of interaction between particles and cathode; M is the atomic weight of the electrodeposited metal; i0 is the exchange current density; gm is the density of electrodeposited metal and vo,A,B are constants. The validity of Guglielmi's model has been verified for different codeposition systems like nickel with silicon carbide or titania from sulfamate baths, copper with alumina from acidic sulfate baths with and without addition of thallium and nickel with alumina from Watts-type electrolytes.
522
11 Electrodeposition of Metals and Alloys
Although this substantiation proves the soundness and the significance of this mathematical model of the electrolytic composite plating, some major objections have been raised. Important process parameters such as size, type and pretreatment of the particles, composition, temperature and pH of the plating bath and hydrodynamical effects are only empirically or at best semi-empirically taken into account. Guglielmi's model indeed does not allow prediction of which way these parameters will affect the electrolytic codeposition. A mathematical expression that could describe the effect of hydrodynamics on codeposition was proposed by Foster and Kariapper (1974). The rate of codeposition is arbitrarily associated with the following expression: dt
(11-29)
where N* = number of collisions of particles suitable for codeposition per second and Vp = volume fraction of particles in the deposit. The parameter hq is further associated with the adsorbed charge density (q) on a particle, the potential field at the cathode (AE\ the rate (i) at which the metal is deposited, the bond strength (L) of the metal/particle interface per surface area, the shape, size and density of the particles (a), and finally the rate of agitation (b), according to: hq = h* (q-AE-\-Li2 — ab)
timated, a Leuven model has been proposed by Celis et al. (1987) that contains measurable parameters so that the prediction of the amount of codeposited particles for a given system becomes feasible. This mathematical model is developed on the basis of two fundamental postulates related to the mechanism of codeposition, namely: - an adsorbed layer of ionic species is created around particles at the time these particles are added to the plating solution or pretreated in ionic solutions, - the reduction of some of these adsorbed ionic species is required for the incorporation of particles in the metallic matrix. Going from the bulk of the plating solution to the site of incorporation at the cathode, the particle has to proceed through five stages, schematically visualized in Fig. 11-39, viz., the adsorption of ionic species upon the particle, the movement of the particle by convection towards the hydrodynamic boundary layer at the cathode, the diffusion of the particle through the diffusion double layer, the adsorption of
Electrode deposit adsorption + reduction diffusion layer
(11-30)
diffusion hydrodynamic
in which h* is a constant. Because of the complex interrelationship between some of these factors, only a limited amount of quantitative work has been done to prove the validity of Eq. (11-30). In contrast to Guglielmi's and Foster's formulae which contain several parameters which can hardly be calculated or even es-
boundary layer convection Bulk of the solution •
formation of ionic cloud
Figure 11-39. The five different steps in the electrolytic codeposition of a particle (from Celis et al., 1987).
11.4 Materials and Process Developments in Alloy Electrodeposition
523
in which pt is the probability for one ion to be reduced at current density i and can be calculated from Pi =
"0
5
10
Current density in A/dm2
Figure 11-40. Probability for the incorporation of one particle versus current density for K = 315 (from Celis etal, 1987).
the particle with its adsorbed ionic cloud at the cathode surface and finally the reduction of some adsorbed ionic species by which the particle becomes irreversibly incorporated in the metal matrix. Here the primary hypothesis of the model is involved: a particle will be embedded only if a certain number k out of the K adsorbed ions on the particle's surface are reduced. As a consequence of this hypothesis, another one is added: no distinction is made between free and adsorbed ionic species so that both are considered equal with respect to transport and reduction processes. The quantitative description of the codeposition of a particle into an electrolytic deposit is then derived from a statistical approach. The probability coefficient for the incorporation of one particle based on the reduction of fc-ions out of K-ions at a current density i is calculated from a binomial distribution: (11-31)
(fc/K, i) z =k
wt.% =
3Mt
l
—T2
(H-32)
where 5 is the diffusion layer thickness, Dion is the diffusion coefficient of the ion and Cion the bulk concentration of metal ions. Fig. 11-40 shows this probability for K = 315 with given values of S, C and D ion . The volume fraction of embedded particles can then be expressed as: W PN vol.% = — ^ r-z-T7
(U-33)
in which Wp is the weight of one particle, G is the weight increase due to metal deposition per unit of time and surface area and N the number of particles arriving at the cathode per unit of time and surface area. The value of JV can be derived from: (11-34) where Nion is the number of ions crossing the diffusion layer per second and per unit of surface area, C p is the number of particles in the bulk of the solution, C?on is the number of ions in the bulk of the solution, itT is the transition current density from charge transfer to concentration overvoltage control and a is a measure of the interaction between free and adsorbed ions due to current density effects. By combination of the preceding formulae, the following expression for the weight percentage (wt.%) of embedded particles as a function of the current density i is obtained:
(11-35)
524
11 Electrodeposition of Metals and Alloys
for the case of spherical particles. The factor H takes into account hydrodynamic effects. On the basis of previous literature, this model has been demonstrated to be valid for the codeposition of alumina particles with copper from acidic sulphate plating baths on rotating disk electrodes (see Fig. 11-41) and on vertical cathodes, as well as for the gold-alumina deposition from cyanide solutions (Buelens et al., 1983). For the codeposition systems for which the necessary data for the calculation of the probability coefficient P are not available, the model parameters can be determined by comparing theory and experiments. In conclusion, it can be said that models have been developed which yield a quantitative description of the codeposition process under well defined experimental conditions. The exact nature of the particlecathode interactions are still under debate but hopes are high that they can be further uncovered in the near future.
2.0 —• experimental —•• theoretical
0
1
2
3 4 5 6 7 Current density in A/dm 2
8
9
Figure 11-41. Comparison of experimental results and predicted codeposition based on the Leuven model for Cu-Al2O3 codeposition on RDE (from Celis et al., 1987).
11.4.3 Compositionally Modulated Alloys
In recent years new materials synthesized in the form of thin films by depositing subsequently very thin layers of two different metals, have aroused considerable interest. These materials are usually designated as compositionally modulated multilayers (CMM) or compositionally modulated alloys (CMA) (see Chapter 8). The thickness of the individual layers is of the order of a few nanometers and the number of such layers varies from 10 to a few 100. Important characteristics of CMA are the modulation wavelength, which is the combined thickness of two successive layers, the modulation amplitude which is the amplitude of the composition wave and the concentration gradient at the interface. The stacking of two different materials at a near atomic dimension level can induce fundamental changes in the material properties owing to interlayer electronic interactions. Outstanding properties of CMA can be a supermodulus effect, and interesting superconductive, magnetic, electrical, optical and mechanical properties. Technological applications of CMA are found as dispersing devices in wavelength dispersive spectrometers, in magneto-optical recording, as materials having increased yield and/or tensile strengths and in thin film resistors as materials having an extreme stability against temperature fluctuations. Although vapor deposition techniques like CVD, PVD, MOCVD and molecular beam epitaxy have been almost exclusively used for the production of such CMA, the electrolytic deposition of such alloys can be an interesting alternative. A survey of the state of the art related to the electrodeposition process and an overview of the properties of actually electrodeposited CMA are given here.
11.4 Materials and Process Developments in Alloy Electrodeposition
11.4.3.1 Electrolytic Production of Compositionally Modulated Alloys
The primary requirement for a process to be suitable to produce CMA is for it to be an atomistic deposition process. Vapor deposition and electrodeposition techniques naturally meet this criterion. Other important criteria for the successful deposition of CMA are the ability to produce very thin layers which requires a slow growth rate, highly smooth surfaces which requires a planar growth process as well as abrupt interfaces which demands a low deposition temperature and a fast process response to changes in deposition conditions. Many of these requirements can be met to a reasonable extent by electrodeposition. Specific advantages of electrodeposition in comparison with vapor deposition processes are that electrodeposition is a less expensive method, that massive composition-modulated alloys can be obtained by electroforming, that it is a room temperature process minimizing the risk of diffusion during deposition and that a specific crystallographic texture can be induced by selecting appropriate electrodeposition conditions. Electrodeposition also has certain disadvantages linked to difficulties in achieving short wavelengths and sharp composition transitions, for instance, owing to the time-limited redistribution of solutes at the electrode/solution boundary layer. Another important limitation of electrodeposition from aqueous solutions lies in the difficulty of depositing reactive metals (cf. Sec. 11.2). Regarding the possibility of producing CMA by electrodeposition, two techniques are in competition, namely, the so-called single-plating bath technique and the two-plating baths technique. In the single-plating bath technique, a controlled periodical change of any deposi-
525
tion parameter that affects the composition of the deposited material can result in the deposition of CMA. Such a plating parameter can be current, voltage, temperature or mass transport. Thus, a potentiodynamic or galvanodynamic variation of the plating conditions can result in the deposition of different elements having separated deposition potentials (see Fig. 11-42). The single-plating bath technique is in fact a pulse-plating process in which at higher overpotential or current the more active constituent (element B in Fig. 11-42) is deposited while at lower overpotential or current the more noble constituent (element A) is deposited. Compositionally modulated alloys already produced in this way include e.g. Ag/ Pd, Co/W, Cu/Ni and NiP^/NiP, with varying phosphorus content. It should however be emphasized that these CMA consist of successive layers of relatively pure elements. In order to avoid alloying within each layer, a very low concentration of the more noble metal in the bath is required. At low overpotentials, layers containing only the more noble metal will be de-
—
/ /' A
_ Time"
Figure 11-42. A scheme for the electrodeposition of CMA from a single bath.
526
11 Electrodeposition of Metals and Alloys
posited. Although at higher overpotentials both the more noble and the more reactive metal will codeposit, the concentration of the more noble metal in the deposit will be negligible as a consequence of its very low concentration in the plating bath. The concentration of the more noble metal ion is recommended to be only 1% of the less noble metal ion concentration in the bath. The use of a modulated mass transport towards the cathode in combination with a current/voltage modulation can be used to improve the quality of the electrolytic CMA. By increasing the mass transport during the low overpotential cycle the quality of the more noble metal deposit can be improved and a higher deposition rate can be achieved owing to higher limiting current densities. On the other hand, if the mass transport during the high overpotential cycle is low, then the codeposition of the more noble metal will be reduced. The sharpness of the transition at the interfaces between consecutive layers in the single-plating bath technique obviously depends upon the current/potential pulse response obtained. Success in achieving shorter wavelengths also lies in ensuring sharp potential transitions at short pulse durations. Electrochemical measurements offer in this respect a very useful in-situ control of the overpotential variation required to achieve CMA with sharp composition transitions. Figures 11-43 and 11-44 (Celis et al., 1988) show some potential variations during the pulse deposition of CMA from a copper-nickel citrate plating bath. Galvanostatic pulses with a duty cycle of 50% were used. Figure 11-43 shows that the development of a stationary potential variation requires a certain number of pulses. This phenomenon occurs when the off-time is shorter than the time required to wipe out the composition gradient created over the
Cu-Ni citrate ammoniacal plating bath pH = 6 ; 20 °C
no agitation
Ji = OA/dm2
7on = /off = 2s
72= 0.5 A/dm 2
-1 .
lAA/l 7i = OA/dm 2 7 2 = 0.75 A/dm 2
7on = 7"Off = 2s
1
AAAJ
°- o 7i = OA/dm2
/on = ^ f = 0.5 s
72= 0.75 A/dm 2
mmm
0 Time •
Figure 11-43. Variation of the cathodic potential at the start of a galvanodynamic controlled deposition of nickel from a Cu-Ni citrate plating solution (from Celis et al., 1988). Cu-Ni citrate ammoniacal plating bath pH = 6;.20°C no agitation -0.5
i = 0A/dm 2
7 2 = 0.175 A/dm 2
i = 0A/dm 2
7 2 = 0.5 A/dm 2
7i = OA/dm 2
72 = 0.75 A/dm 2
_ *min
,-1.0
•-0.5
-1.0
-0.5
0
10
2 0.5 0.1 Pulse duration in s
0.02
0.005
Figure 11-44. Maximum and minimum cathode potential in a Cu-Ni citrate plating solution versus the pulse duration of galvanodynamic controlled pulses with a duty cycle of 0.5 (from Celis et al., 1988).
11.4 Materials and Process Developments in Alloy Electrodeposition
diffusion double layer during the onetime period. Maximum and minimum potentials reached during the on- and off-periods at different frequencies are shown in Fig. 1144. For the production of Cu/Ni CMA from this citrate bath pulse, durations of at least 0.5 sec are required. From Faraday's law it can be calculated that a minimum layer thickness of 1.33 nm is reached under these plating conditions. To sharpen the transition, a triple-pulse technique which consists in the introduction of a short zero current pulse just after the high current pulse, can be used. In-situ computer control and very high speed coulometers allow a further improvement of the quality of electrodeposited CMA deposited from a single-plating bath. The smallest wavelength in Cu/Ni CMA experimentally achieved with the single-plating bath technique up to now is 1.6 nm (Yahalom and Zadoc, 1987). In the two-plating bath technique, a relative movement of the cathode with respect to the plating baths, or vice versa, is required. Some automated movement of the cathode can therefore be used (Fig. 11-45). Any combination of elements capable of being electrodeposited can be selected for the production of CMA. The Bath 1 Anode 2
527
thickness of the individual layers is controlled by the current used and the residence time in front of each plating bath. An important technological drawback of this technique in comparison with the singleplating bath technique, is the need to avoid leakage or mutual contamination of the two plating baths. Wavelengths down to 2 nm for Ni/Ni-P CMA produced by the two-plating bath technique have been reported (Goldman et al., 1989). 11.4.3.2 Properties of Electrolytic Compositionally Modulated Alloys
The electrolytic production of CMA is still in an early stage of development. Systematic studies on the effect of plating parameters, such as type or pH of the plating bath, concentration of ions and presence of additives like brighteners, on the properties of electrodeposited CMA are not yet available. For a successful deposition of quality CMA, planar nucleation and growth conditions are necessary. In that respect the composition of the substrate and its initial surface condition play a very important role. Substrates having a very low surface roughness are highly recommended, and electropolishing is useful in that respect. Furthermore, the ability of Bath 2
Solution 1
• Insulator - Tensioning bar
Figure 11-45. Two-bath plating cell (from Goldman et al., 1986; reproduced by permission of J. Appl. Phys.).
528
11 Electrodeposition of Metals and Alloys
each deposited layer to act as an appropriate substrate material is quite essential in achieving good CMA. Problems are frequently encountered with the more active constituent which can be simply chemically displaced or electrochemically dissolved during the electrodeposition of the more noble metal. Layering in CMA can be demonstrated by either transmission electron microscopy, Auger depth profiling or X-ray diffraction. So at short wavelengths, CMA should be coherent and the high-angle diffraction pattern should show a central peak at position intermediate to the ones of the metallic constituents, flanked by satellite peaks due to the periodicity in the structure. The intensity of the satellite peak can be indicative of the sharpness of the interphases and/or of the variation in layer thickness within the CMA. The presence of (000) satellites in low-angle diffraction patterns also confirms the layered nature of deposits. Some examples from the literature are shown in Figs. 11-46 and 11-47 (Lashmore and Dariel, 1988; Goldman etal, 1986) (see also Chap. 8, Sees. 8.3.2 and 8.4.4). Interesting engineering properties of CMA include strength, electrical resistivity, magnetic properties and wear properties. A comparison of the tensile strength of electrodeposited and vapor-deposited Cu/ Ni CMA is shown in Fig. 11-48 (Tench and White, 1984; Ogden, 1986; Menezes and Anderson, 1990; Baral et al., 1984). In vapour deposited CMA, the maximum tensile strength is obtained at a wavelength of about 2 nm for which the thickness of both the Cu and Ni layers was 1 nm. In electrodeposited CMA, a maximum tensile strength occurs at a wavelength of about 20 nm for which the Cu and Ni layer thickness were respectively 2 and 18 nm. Maximum strength of the electrodeposited
42
46 50 26 in degrees
54
Figure 11-46. High-angle X-ray diffraction of electrodeposited Cu/Ni CMA with a wavelength of 5.7 nm (from Lashmore and Dariel, 1988; reproduced by permission of the publisher, The Electrochemical Society, Inc.).
4.0 I® in degrees
Figure 11-47. Presence of (000) satellites in low-angle diffraction patterns of electrodeposited Ni/NiP CMA with a wavelength of 3.9 nm (from Goldman et al., 1986; reproduced by permission of J. Appl. Phys.).
Cu/Ni CMA (1900 MPA) was 70% higher than that of vapor-deposited CMA. Possible explanations given therefore are the presence of some oxide films at the interfaces between the Cu and Ni layers, and the codeposition of hydrogen during plating. The fall of strength of CMA below a certain wavelength can on the other hand not be explained solely by a Hall-Petch strengthening mechanism. The electrical resistivity of electrodeposited Cu/Ni CMA with different wavelengths (Menezes and
11.4 Materials and Process Developments in Alloy Electrodeposition 2000
1500
vapour deposited Cu/Ni CMM:
0.3
0.6
0.9
• + A •
Trench 1984 Ogden 1986 Menezes 1990 Baral 1984
1.2
1.5
1.8
(Cu layer thickness)"172 in nm" 1/z
Figure 11-48. Variation of tensile strength of electroand vapor deposited Cu/Ni CMA versus Cu-layer thickness.
Anderson, 1990) shows a similar trend as the tensile strength. The maximum in electrical resistivity and in tensile strength occur at the same wavelength. Magnetic properties of Cu/Ni (Bennett et al, 1987) and Co/Cu (Dariel et al., 1987) CMA have been investigated by vibrating sample magnetometer and ferromagnetic resonance techniques. As in vapor-deposited CMA, a reduced magnetization was found although the presence of nonuniformities in the films was detected. Related to the wear and antifriction properties of electrodeposited CMA, the performance of Ag/Pb as a coating on magnetic hard discs has been reported (Cohen and Tan, 1987). The striction/friction value of sputtered carbon coatings tested in unlubricated conditions reached after 10000 contact start/stop cycles a value of 0.6 to 1.0 or above which caused 'freezing' of the drive. Under the same conditions, the electrodeposited CMA maintained a striction/ friction value of 0.3 to 0.5 even after 20000 cycles. It can thus be stated that substantial progress has been achieved recently in the
529
electrodeposition of CMA. Main attention has been given to the single-plating bath technique, but the advantages of the twoplating bath technique should become more apparent in the near future. The structure of electrodeposited CMA is not yet as good as that obtained by vapor deposition techniques. However the structure of electrodeposited CMA is improving, and recent work has already demonstrated the interesting capabilities of electroplating for the production of compositionally modulated alloys. 11.4.4 Electroless Plating Electroless deposition of metals has received over the last decade full recognition as a valuable industrially applicable coating technology. Heat-treated electroless nickel-phosphorus is, for certain wear problems a convenient alternative to electrolytic hard chromium. It is expected that further development of the electroless plating technology will result in new coating materials with unique physical, chemical and mechanical properties. In electroless plating it is assumed that different reactions occur simultaneously at the electrode surface. Thus, in case of electroless nickel plating from hypophosphite-containing plating solutions, the following reactions are generally proposed: H 2 PO 2
H 2 O -> H + + HPO 3
2H abs (11-36) (11-37) (11-38)
H2PO2+H2O (11-39) A part of the hydrogen will be absorbed on the electrode surface and will reduce the
530
11 Electrodeposition of Metals and Alloys
nickel ions. At the same time absorbed hydrogen will also reduce some hypophosphite to phosphor that can be incorporated into the growing nickel coating. Reaction, Eq. (11-39), occurs independently of the coating deposition process and is responsible for the very low efficiency of the electroless process. 11.4.4.1 Control of the Deposition Rate Although electrolytic and electroless plating are both based on electrochemical processes such as reduction and oxidation reactions, the process control of both plating procedures is basically different. For electrolysis, the on-line deposition rate control is quite easy and relies on a current measurement in the external electrical wiring between electrolysis cell and power supply unit by means of ammeters. For electroless plating, since no net current flow is induced, platers require adapted rate control procedures. A classical procedure consists in the measurement of the weight increase after a given time on small coupons immersed in the plating tank. The biggest drawbacks of that procedure are that any on-line deposition rate control is impossible, that only a mean deposition rate value is obtained and that variations in the process conditions can neither be detected nor be suppressed. Alternative methods for the in-situ estimation of the electroless metal deposition rate have been proposed in recent literature. These methods are based either on the analysis of electrochemical transients on parts being coated or on the detection of any change in physical property at an electrode induced by the deposition of a thin coating. Here, three in-situ rate control procedures will be discussed, namely, rate estimation from a measured electrode potential, coulostatic monitoring and real-time monitoring using quartz resonators.
One possible approach is to link the electrode potential that can be measured by using a reference electrode, to the metal deposition rate. Some comparison between a theoretical estimate and experimental results has been done (Oni et al., 1987). The starting point is the theory of mixed potential (Bindra et al., 1984) which postulates that the rate of a Faradaic process is independent of other Faradaic processes occuring simultaneously at the electrode. This rate thus depends only on the electrode potential. The following expression for the plating rate (7p) is then derived ip = nFK'exp{(l -/3)FE/RT}
(11-40)
where E is the electrode potential, /? is a symmetry factor, n is the numbers of electrons involved in the reaction and Kf a reaction rate constant. As theoretically predicted, an increase in the electroless deposition rate was noticed in practice as the potential became more noble. However, large discrepancies were experienced between theoretical predicted values and experimental deposition rates obtained. This was attributed to some difficulties encountered in obtaining very accurate measurements under operating conditions. Small variations in plating temperature and bath pH have indeed a large effect on the potential of an electrode immersed in an electroless plating solution, as shown in Fig. 11-49. It is important here to notice that an increase in plating bath temperature and pH result in opposite changes in electrode potential, although field practice has demonstrated that with increasing bath temperature and pH, the plating rate increases in both cases. A simple relationship between electrode potential and deposition rate thus does not prevail in case of electroless plating owing to the complex electrochemical and chemical reactions occuring at the electrode surface. Thus, the
531
11.4 Materials and Process Developments in Alloy Electrodeposition Temperature in °C 85 90
the interface between the electrode surface and the plating solution can be represented by a Faradaic resistance Rp in parallel with a double layer capacitance C d . At the time a definite amount of charge Q is impulsed, the double layer at the interface will be charged instantaneously and an overpotential is reached equal to
95
(11-42)
Vo = Q/Cd
Figure 11-49. Effect of pH and plating temperature on the electrode potential in a nickel electroless plating bath.
electrode potential appears to be influenced by the amount of oxygen present in the electroless solution, the bath agitation and the nickel bath composition as nickel concentration, presence of inhibitors such as lead, and accelerators. As an interesting alternative for the measurement of the in-situ electroless deposition rate, coulostatic monitoring can be used. Electrochemical transient techniques such as polarization curves and electrode impedance measurements during electroless nickel plating have been shown to be also largely affected by dissolved oxygen, so that a correction for this is necessary (Gabrielli and Raulin, 1971). Such a problem does not arise with the coulostat method based on the Stern and Geary equation (Stern and Geary, 1957) derived for corrosion processes: AE
PJo 2.3/ corr (/? a +/y
(11-41)
where AI is imposed current variation, AE is resulting potential variation, / corr is corrosion current and /?a, jSc are anodic and cathodic slopes of the polarization curves. The electrochemical equivalent circuit diagram for the electroless plating reaction at
Subsequently the impulsed charge will be consumed by a discharge through the Faradaic resistance. The electrode potential will gradually return to its original value. The capacitive behavior of the double layer is assumed to remain constant as long as the resulting overpotential is small (<10mV). The amount of charge consumed at time t is given by t
AD — C (n —n\ — ( i dt \£t
—
d v 10
It)
—
I 0
Ml -X\\ V
t
)
where it is the Faradaic current at time t and Y\t is the overpotential at time t. In combination with the Stern and Geary equation, one obtains 1
(11-44)
and — t
(11-45)
n% = no • e x p
The calculation of the anodic and cathodic slopes of the polarization curve can be done from the following equation based on three overpotential measurements made at different times: 2.3
2.3
2.3
2.3
ti-t2
(11-46)
532
11 Electrodeposition of Metals and Alloys
Assuming that dr] = rj1 — rj2 = rj2 — ri3, the
previous equation can be simplified to (11-47) &
=
•
Table 11-2. Double layer values determined by the coulostatic method for a hypophosphite nickel electroless plating bath. Temperature
pH
The validity of this deposition rate monitoring has been investigated by several researchers over the last few years (Sato et al., 1983; Matsuoka et al., 1987; Schillebeeckx, 1985). The potentialities of this method are illustrated hereafter for electroless nickel deposition from a hypophosphite solution. A recorded potential decay curve and the related calculated curve by computer fitting are shown in Fig. 11-50. Typical double layer resistance and capacitance values obtained under different sets of plating temperature and bath pH values are summarized in Table 11-2. This Table shows that a large scatter in Revalues is obtained at plating conditions where low deposition rates are expected, e.g., at low bath temperature and high lead content. Under plating conditions resulting in high deposition rates, the values of Rp are much better defined. A similar dependence is not found for the Cd-characteristic. This is in accordance with the literature, wherein generally a range rather than an exact value is cited for the double layer capacitance. The final check of the validity of this coulostatic technique is represented in Fig. 11-51 showing the polarization resistance Rp versus the experimentally measured deposition rate at bath temperatures between 90 and 96 °C. The deposition rate measured gravimetrically is in accordance with the Stern and Geary equation (cf. Eq. (11-41)) for an electroless deposition at temperatures below 92 °C. For electroless plating at temperatures above 92 °C a linear relationship with a different slope is
P
(fi cm2)
log h-t2 78.5 91.7 43.6 96.6*
6.64 4.81 4.72 4.23
3.1-10.8 2 . 1 - 2.4 2 . 1 - 2.5 9.5-21.1
18-54 25-48 37-59 29-40
* with 3 ppm Pb instead of 1 ppm Pb as in the other cases
experimental data calculated curve
150 Time in ps
200
Figure 11-50. Experimental and calculated potential decay curves after an imposed current pulse of 1.2 \xC in an electroless hypophosphite nickel solution (from Roosetal., 1988).
0
10 15 Deposition rate in (jm/h
20
Figure 11-51. Relationship between polarization resistance and the electroless nickel deposition rate determined gravimetrically (from Roos et al., 1988).
11.4 Materials and Process Developments in Alloy Electrodeposition
found. This slope change could indicate a modified electroless deposition reaction mechanism. It should be noticed that these experiments were conducted under plating conditions where pH variations were smaller than 0.01 and temperature variations smaller than 0.1 °C. Another way to realize a real-time monitoring is based on the use of quartz resonators. A quartz microbalance, which has been used successfully in vacuum deposition, can also be used in a slightly modified way to measure the electroless deposition rate (Kanazawa and Doss, 1987). The resonator takes the form of a flat disk having metallic electrodes on opposing faces. At the time an electrical potential is applied, a shear wave is generated between the electrodes. The resonant frequency (/0) of the quartz resonator is sensitive to lateral forces on its surface. Such forces can result from the deposition of material on one electrode. As a consequence the frequency is lowered by an amount <5f given by 2 Jof2
-
J
(11-48)
where gq is density of quartz, Q is density of the deposited layer, d is thickness of the deposited layer and jnq is effective shear modulus for quartz. For thick films deposited, a modified formula was proposed (Kanazawa and Doss, 1987) tan 7i -jr I = \ Jo/
ta £q ^
(11-49) where \L is the shear modulus of the deposited material. A typical electroless deposition rate obtained in this way for nickel is in the range of 30 A/s. Such deposition monitors have become available recently and allow computerized control of the electroless deposition process (Rico and Martin, 1987).
533
11.4.4.2 Properties of Electroless Coatings
Statements claiming the excellent properties of electroless deposits as corrosion and wear resistance of coatings are legion in technical publications. Such generalities always oversimplify reality, because the properties of electroless deposits, such as the amount of incorporated phosphorus from hypophosphite nickel plating solutions, are strongly dependent on the plating conditions used. The selection of an appropriate electroless coating requires therefore a thorough knowledge of the trivalent relationships existing between process parameters, chemical, physical and mechanical properties of coatings and their functional properties. At the processing stage, the pretreatment of the substrate will largely determine the adhesion of the coating, while deposition conditions like pH, temperature and bath composition, should be kept constant to ensure equal coating characteristics throughout the plating sequence. Thus, crystalline, microcrystalline or amorphous nickel coatings will be obtained according to the amount of phosphorus codeposited and lamellar or uniform cross-sectional structures will be obtained depending on how accurately the main deposition parameters were kept under control. Post-treatments consisting of mechanical finishing and heat treatments can be used to modify the surface finish and the structure of the deposits. As a result, electroless coatings produced from different batches can show large variations in chemical, mechanical and physical properties (see Table 11-3). It is thus evident that the functionality of such coatings for corrosion or wear resistant purposes will largely depend on the process parameters used.
534
11 Electrodeposition of Metals and Alloys
Table 11-3. Summary of some properties of electroless NiP coatings. Property
Coating type 1-7 wt.%P
7 - 1 2 wt.% P As-plated Heat-treated
As-plated
Heat-treated
120000
190000
120000
190000
Tensile strength (N/mm2)
390-450
Max. at 350 °C
450-860
Max. at 350 °C
Hardness (HV)
200-400
Max. at 400 °C
400-600
Max. at 400 °C
50-450
Increased
E-modulus (N/mm2)
Adhesion (N/mm2) Elongation Structure
2%
6%
2%
6%
Microcrystalline
Crystalline
Amorphous
Crystalline
The information available at present allows only a limited assessment of the relationship between functional properties of electroless coatings and their production conditions. Thus, Ni-P coatings with a layered structure produced from a bath stabilized with a heavy-metal thiocompound are much more corrosion-resistant than coatings with a columnar structure produced from an organic-divalent sulfur bath (Van Gool et al., 1987). Some prerequisites for depositing optimal coatings for a given application are a strict control of the process parameters like pH and bath temperature that allows the deposition of coatings having the required composition and having no layered structure, and a good understanding of the mechanism of phosphorus incorporation necessary to optimize deposition conditions. The variation in Rp value measured by a coulostatic method as a function of pH and plating temperature (Fig. 11-52), can be very useful in that respect. Thus, it is evident (from Figs. 11-52 and 11-53) that a high phosphorus content coincides with a high polarization resistance, indicating a possible physical hindrance of the electro-
chemical process by phosphide species generated in the double layer at the electrode. Information on the structure of electroless deposits in the as-plated condition and after heat-treatment are also interesting. Thus, the high density of non-crystalline Ni-P alloys suggests that these al-
9
Nickel sulfate
fj>| Nickel sulfate
ft\
sodium hypophosphite
LJJ sodium hypophosphite • 8g/L SiC
HI
E n l a t e Ni 18
P
^
Figure 11-52. Variation of the polarization resistance value versus electroless plating conditions (from Roos etal., 1988).
535
11.5 References
loys have a continous structure rather than one in which internal boundaries separate well-ordered regions (Cargill, 1970). Information on the internal stress present in electroless coatings as a function of process parameters and thickness is still very limited. Due to the amorphous structure of some electroless deposits, alternatives to the X-ray diffraction technique such as the proximity measurement technique can be useful (Deckert and Andrus, 1978). In order to optimize heat-treatment procedures, differential thermal analysis measurements can be useful. A representative DTA curve for Ni-P is shown in Fig. 11-54. Two exothermic transformations are observed the first of which occurs at 380 °C and the second at 460 °C. Metallographic structure, size, distribution and composition of phosphide precipitates formed during the heat-treatment will be largely influenced by the heat-treatment conditions selected. In respect to wear, the formation of these precipitates is very in-
250
300 350 400 Temperature in °C
450
500
Figure 11-54. Differential thermal analysis curve of a Ni-P deposit.
teresting as well as the use of composite electroless coatings. A reduction in the friction coefficient can be obtained by incorporating dry lubricating PTFE-particles in electroless nickel and a major improvement in wear resistance can be obtained once 18 vol. % PTFE is incorporated (Ebdon, 1987). Apart from electroless nickel-PTFE, nickel-graphite can also be used as a wear resistant coating. In conclusion, it can thus be stated that electroless coatings have unique functional properties, mainly due to the incorporation of foreign elements such as phosphorus or boron in nickel. By using well controlled deposition conditions, it is possible to optimize the composition and the structure of such coatings for a wide variety of desired functions in numerous industrial applications.
11.5 References
Nickel sulfate sodium hypophosphite
Nickel sulfate sodium hypophosphite
fTj Enplate Ni- 418
• 8g/L SiC
Figure 11-53. Variation of the phosphorus content versus electroless plating conditions (from Roos et al., 1988).
Alkire, R.C., Chen, T.J. (1987), /. Electrochem. Soc. 129 (11), 2424-2432. Alkire, R.C., Ju, J. B. (1987), /. Electrochem. Soc. 134 (5), 1172-1180. Amblard, X, Epelboin, I., Froment, M. (1979), J. Applied Electrochem. 9, 233-242. ASTM (1985), Method of Salt Spray (Fog) Testing, B117-85. Philadelphia: ASTM.
536
11 Electrodeposition of Metals and Alloys
ASTM (1984), Test Method for Abrasion Resistance of Organic Coatings by the Taber Abraser, D406084. Philadelphia: ASTM. ASTM (1989), Test Method for Wear Testing with a crossed Cylinder Apparatus, G 83-89. Philadelphia: ASTM. Baral, D., Ketterson, J.B., Hilliard, I E . (1984), in: Modulated Structure Materials, NATO ASI Series: Tsakalakos, T. (Ed.). Dordrecht: Martinus Nijhoff, pp. 465-473. Barlat, F. (1987), Mat. Sci. and Engin. 91, 55-72. Barrett, C.S., Massalski, T.B. (1966), Structure of Metals. New York: McGraw-Hill, pp. 466-485. Bennett, L.H., Lashmore, D.S., Dariel, M.P., Kaufman, M.J., Rubinstein, M., Lubitz, P., Zadok, O., Yahalom, J. (1987), /. Magn. Mat. 67, 239-245. Benzing, R. (1973), A Catalog of Friction and Wear Devices, American Society of Lubrication Engineers. Bindra, P., Light, D., Rath, D. (1984), IBM J. Res. Develop. 28 (6), 668-678. Bockris, O'M. I , Damjanovic, J. (1964), Modern Aspects of Electrochemistry, Vol. 3. London: Butterworths. Bockris, O'M. J., Reddy, A.K.N. (1977), Modern Electrochemistry. New York: Plenum Press. Brenner, A. (1963), Electrodeposition of Alloys: Principles and Practice. New York: Academic Press. Brenner, A., Senderoff, S. (1949), J. Res. Nat. Bureau of Standards 42, 89. Buelens, C , Celis, J.P., Roos, J.R. (1983), /. Appl. Electrochem. 1, 541-548. Cargill, G.S. (1970), /. App. Phys. 41, 12-29. Celis, I P . , Roos, J.R. (1984), Proc. AES SUR/FIN '84, New York, paper 0-1. Celis, I P . , Roos, J.R., Buelens, C. (1987), J. Electrochem. Soc. 134 (6), 1402-1408. Celis, I P . , Roos, J.R., Blanpain, B., Gilles, M. (1988), Proc. Interfinish '88, Paris, Vol. 2, pp. 435-445. Celis, I P . , Roos, J.R., Van Vooren, W, Vanhumbeeck, I (1989), Trans. Inst. Metal Finish. 67, 70-72. Chonglun, F. (1990), Ph.D. Thesis, Dept. MTM, Katholieke Universiteit Leuven (B). Clarke, M., Leeds, J.M. (1965), Trans. Inst. Metal Finish. 43, 50. Clarke, M., Chakrabarty, A.M. (1970), Trans. Inst. Metal Finish. 48, 99. Cohen, U., Tan, M. (1987), Proceed. Electrochem. Soc. Meeting, Honolulu, Abs. No. 1018. Czichos, H. (1978), Tribology - a System Approach to the Science and Technology of Friction, Lubrication and Wear. Amsterdam: Elsevier. Damjanovic, A., Bockris, O.M.I (1963), /. Electrochem. Soc. 110 (10), 1035. Dariel, M., Bennett, L.H., Lashmore, D.S., Lubitz, P., Rubinstein, M., Lechter, W.L., Harford, M.Z. (1987), J. Appl. Phys. 61, 4067-4070.
De Buyser, L., Van Houtte, P., Aernoudt, E. (1990), Proc. Tagung Eigenspannungen, DGM, Darmstadt (FRG). Deckert, C.A., Andrus, I (1978), Plat. Surf. Finish. 65 (11), 43-48. De Doncker, R., Vangaever, R, Vanhumbeeck, I, Celis, I P . , Roos, J.R. (1984), Proc. Interfinish '84: Zahavi, I (Ed.). Tel Aviv, pp. 52-60. De Doncker, R., Vanhumbeeck, I (1985), Trans. Inst. Metal Finishing 62 (2), 59-63. Dieter, G.E. (1967), Ductility, Proc. ASM Seminar, London: Chapman and Hall, pp. 1-30. DIN-Taschenbuch 175 (1989), Prufnormen fur metallische und anorganische nichtmetallische Uberziige. Berlin: Beuth-Verlag. DIN 50320 (1979), Verschleiss - Begriffe, Systemanalyse von Verschleissvorgdngen, Gliederung des Verschleissgebietes. Berlin: Beuth-Verlag. Ebdon, P.R. (1987), Trans. Inst. Met. Finish. 65, 43-48. Ehrhardt, R. A. (1960), Proc. Am. Electroplaters' Soc. 47, 78. Foster, I, Kariapper, A.M.I (1974), Trans. Inst. Met. Finish. 52, 87-91. Fransaer, I , Celis, IP., Roos, J.R. (1989), Metal Finishing 87 (6), 107-109. Gabe, D.R. (1974), /. Appl. Electrochem. 4, 91-108. Gabrielli, C , Raulin, F. (1971), J. Appl. Electrochem. 1, 167-177. Garte, S.M. (1966), Plating 53, 1335. Goldman, L.M., Blanpain, B., Spaepen, F (1986), J. Appl. Phys. 60 (4), 1374-1375. Goldman, L.M., Ross, C.A., Ohashi, W, Wu, D., Spaepen, F. (1989), Appl. Phys. Letters 55 (21), 2182-2184. Guglielmi, N. (1972), /. Electrochem. Soc. 119, 10091012. Helle, K., Opschoor, A. (1980), Proc. 10th World Congr. Metal Finishing, Kyoto, pp. 234. Kanazawa, K., Doss, S. (1987), Plat. Surf. Finish. 74 (7), 52-55. Kazumi, N., Yasuhika, M., Hada, T. (1987), Proc. Symp. Corrosion Protec. by Organic Coatings, pp. 140-152. Kim, I., Weil, R. (1987), STP947. Philadelphia: ASTM, pp. 11-18. Landau, U. (1981), American Institute Chemical Engineering, Symp. Series 204 (77), 75-87. Lashmore, D. S., Dariel, M. P. (1988), J. Electrochem. Soc. 135 (5), 1218-1221. Leeds, J.M. (1969), Trans. Inst. Metal Finish. 47, 222. Levich, V. G. (1962), Physicochemical Hydrodynamics. Englewood Cliffs: Prentice Hall. Lowenheim, F A . (1974), Modern Electroplating. New York: Wiley. Mansfeld, F (1971), Corrosion 27 (10), 436-442. Mathias, M.R, Chapman, T.W (1987), /. Electrochem. Soc. 134 (6), 1408-1416. Matsuoka, M., Iwamoto, K., Hayashi, T. (1987), Proceed. SURF/FIN '87, Chicago, paper 0-3.
General Reading
Menezes, S., Anderson, D.P. (1990), J. Electrochem. Soc. 137 (2), 440-444. Metals Handbook (1987), Corrosion, Vol. 13, ASM International Handbook Committee, American Soc. Metals. Morisset, P. (1982), Chromage dur et decoratif 2nd ed. Senlis: Cetim. Nakahara, S., Okinaka, Y, Strashil, H.K. (1987), Special Technical Publication 947. Philadelphia: ASTM, pp. 32-57. Newman, J.S. (1973), Electrochemical Systems. Englewood Cliffs: Prentice Hall. Ogburn, R, Benderley, A. (1954), Plating 41, 61. Ogden, C. (1986), Plat. Surf. Finish. 73 (5), 130-134. Oni, A., Cottis, R., Thompson, G.E. (1987), Trans. Inst. Met. Finish. 65, 105-107. Parente, M., Weil, R. (1971), Plating and Surface Finishing 5, 114-117. Pourbaix, M. (1974), Atlas of Electrochemical Equilibria in Aqueous Solutions. Houston: National Association of Corrosion Engineers. Rico, A.J., Martin, S.J. (1987), Proc. Electrochem. Soc. Meet., Honolulu. The Electrochemical Society, pp. 711-712. Rogers, H.C. (1967), Ductility, Proc. ASM Seminar. London: Chapman and Hall, pp. 31-61. Rolff, R. (1987), Special Technical Publication 947. Philadelphia: Amer. Soc. for Testing Materials, pp. 19-31. Roos, J. R., Celis, J. P., Heerman, M., Vanhumbeeck, J. (1986), Het Ingenieursblad55, 573-577. Roos, J.R., Celis, I P . , Van Vooren, W, Buelens, C. (1987), Proc. 11th IPMI, Brussels, pp. 93-107. Roos, J.R., Celis, IP., De Bonte, M. (1988), Proc. Int. Conf. Surf Modific. Warrendale, PA: The Metallurgical Society, pp. 215-235. Roos, J.R., Celis, I P . , Chonglun, F. (1990), /. Electrochem. Soc. 137 (4), 1096-1099. Rudzki, G.I (1983), Surface Finishing Systems. Ohio: Amer. Soc. Metals. Ruimi, M., Martinou, R. (1989), Gahano-Organo 595, 387-392. Sato, N., Suzuki, M., Sato, Y. (1983), /. Electrochem. Soc. 130 (7), 1485-1488. Schillebeeckx, P. (1985), Eng. Thesis, Dept. MTM, Katholieke Universiteit Leuven. Serruys, W. (1988), Ph.D. Thesis, Dept. MTM, Katholieke Universiteit Leuven (B). Serruys, W, Van Houtte, P., Aernoudt, E. (1987), Residual Stresses in Science and Technology, Vol. 1: Macherauch, E., Hauk, V. (Eds.). Oberursel (FRG): DGM-Verlag, pp. 417-424.
537
Shih, H., Pickering, H.P. (1987), J. Electrochem. Soc. 134 (3), 551-558. Stern, M., Geary, A.L. (1957), /. Electrochem. Soc. 104, 56. Tench, D., White, I (1984), Metall. Trans. 15A (11), 2039-2040. Thoma, M. (1985), Proceed. SURTEC, Berlin, pp. 597. Tomaszewski, T.W., Tomaszewski, L. L., Brown, H. (1969), Plating 56, 1234. VanGool, A.P., Boden, P.I, Harris, S.J. (1987), Trans. Inst. Met. Finish. 65, 108-114. Van Houtte, P. (1987), Textures and Microstructures 7, 29-72. Van Vooren, W. (1989), Ph.D. Thesis, Dept. MTM, Katholieke Universiteit Leuven (B). Vatakhov, P., Weil, R. (1990), Plating Surf. Finish. 3, 58-61. Wolff, R.H., Henderson, M.A., Eisler, S.L. (1955), Plating 42, 537. Xingpu, Y. (1990), Ph.D. Thesis, Dept. MTM, Katholieke Universiteit Leuven (B). Yahalom, I, Zadoc, O. (1987), /. Mater. Sci. 22, 499503.
General Reading Bockris, O'M. I, Reddy, A. K. N. (1977), Modern Electrochemistry. New York: Plenum Press. Durney, L. (1984), Electroplating Engineering Handbook. New York: Van Nostrand Reinhold Co. Harding, W, Bari, A. D. (1987), Testing of Metallic and Inorganic Coatings, STP 947. Philadelphia: ASTM. Lowenheim, F. (1974), Modern Electroplating. New York: Wiley. Riedel, W (1989), Funhtionelle Chemische Vernicklung, Saulgau: E. Leuze Verlag. Sard, R., Leidheiser, H., Ogburn, F. (1975), Properties of Electrodeposits: Their Measurements and Significance. Princeton: The Electrochemical Society. Silman, H., Isserus, G., Averill, A. (1978), Protective and Decorative Coatings for Metals. Teddington: Finishing Publ. Ltd. Wood, WG. (1982), "Surface Cleaning, Finishing, and Coating", in: Metals Handbook, Vol. 5. Ohio: ASM.
12 Solidification Processing under Microgravity Peter R. Sahm GieBerei Institut der RWTH Aachen, Aachen, Federal Republic of Germany Manfred H. Keller DLR Koln, Koln, Federal Republic of Germany
List of Symbols and Abbreviations Space Flights Incorporating Materials Experiments 1975-1992 12.1 Microstructures 12.2 Driving Forces in Microstructure Formation 12.3 Nucleation and Undercooling 12.3.1 Theoretical Background 12.3.2 Procedures, Methods, and Apparatus 12.3.3 State of the Art 12.4 Solidification-Front Dynamics 12.4.1 Solidification-Front Morphology 12.4.1.1 Theoretical Background 12.4.1.2 Procedures, Methods, and Apparatus 12.4.1.3 State of the Art 12.4.2 Single Crystal Growth 12.4.2.1 Theoretical Background 12.4.2.2 Procedures, Methods, and Apparatus 12.4.2.3 Crystal Growth: State of the Art 12.4.3 Solid-liquid Interface Interaction with Particulate Matter, Including the Consideration of Alloys Immiscible in the Liquid State: Engulfment of Dispersoids 12.4.3.1 Theoretical Background 12.4.3.2 Procedures, Methods, and Apparatus 12.4.3.3 State of the Art 12.4.4 Eutectics: In situ Composites 12.4.4.1 Theoretical Background 12.4.4.2 Procedures, Methods, and Apparatus 12.4.4.3 State of the Art 12.4.5 Ripening Phenomena 12.4.5.1 Theoretical Background 12.4.5.2 Procedures, Methods, and Apparatus 12.4.5.3 State of the Art 12.5 Acknowledgements 12.6 Bibliography of Space Results 12.7 References Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
540 542 543 543 544 544 545 548 552 554 554 554 556 561 561 562 566
570 570 571 572 576 576 576 576 578 578 579 579 580 580 580
540
12 Solidification Processing under Microgravity
List of Symbols and Abbreviations
Co C,d tC2,C3,C4 d,d0 D
m
AT;
Ao ACCESS AGHF
concentration concentration of alloying elements eutectic composition starting composition constants average spacings at time t, t0 average atomic diffusion coefficient in melt effective diffusion coefficient temperature gradient equilibrium distribution coefficient effective distribution coefficient slope of liquidus line corresponding to solidifying alloy particle radius quenching time glass temperature liquidus temperature melting point real temperature of the melt solidus temperature cooling rate constitutional undercooling undercooling to glass temperature kinetic undercooling curvature-dependent undercooling thermal undercooling undercooling critical growth rate solidification front velocity, growth rate volume increment at the melting point, when turning from liquid to solid system-dependent constant thickness dynamic melt viscosity primary arm spacing lamellar spacing ratio of particle-to-liquid heat conductivities density differences thermally and/or solutally induced density differences surface energy differences Aachen Center for Solidification in Space (Aachener Centrum fur Erstarrung unter Schwerelosigkeit) advanced gradient heating facility
List of Symbols and Abbreviations
AMF ASTP CHF CVD CVT D l ; SL-D1 D2 ELLI emf ESA ESTEC EURECA FSLP GFQ GHF HOLOP HFT HTT IHF IML1 MEPHISTO micro-g, jug NASDA OSIRIS PCF PVD PVT SEM SL-J SL-3 SPAR TEMPUS TEM TEXUS TCS WPF lg
541
automatic mirror furnace Apollo-Soyus test project continuous heating facility chemical vapor deposition chemical vapor transport first German spacelab mission second German spacelab mission Einellipsoid-Spiegelofen (monoellipsoid mirror furnace) electromotive force European Space Agency European Space Research and Technology Center European retrievable carrier first spacelab payload gradient furnace with quenching gradient heating facility holographic optics laboratory heating facility for turbine blades high temperature thermostat isothermal heating facility first international microgravity laboratory materiel pour l'etude des phenomenes interessant la solidification sur terre et en orbite reduced gravity, microgravity National Aeronautic Space Development Agency of Japan oxide dispersion strengthened single crystal improved by resolidification in space protein crystallization facility physical vapor deposition physical vapor transport scanning electron microscope Japanese Spacelab mission Spacelab mission 3 space processing applications rocket project Tiegelfreies und elektromagnetisches Prozessieren unter Schwerelosigkeit (facility for containerless processing by electromagnetic levitation) TEXUS experiment module Technologische Experimente unter Schwerelosigkeit (technological experiments under reduced gravity) triglycine sulfate Wissenschaftliche Projektfuhrung (der Dl Mission) normal gravity
542
12 Solidification Processing under Microgravity
Space Flights Incorporating Materials Experiments 1975-1992 July 1975 December 1975 1981 08.05.1981 29.04.1982 08.05.1982 15.05.1984 1983 28. 11. 1983 29. 04. 1985 30. 10. 1985 16. 05. 1991 17. 06. 1991 24. 09. 1992
Apollo-Soyuz test project (ASTP) SPAR-I (space processing application rocket) SPAR VI TEXUS4 TEXUS 5 TEXUS 6 TEXUS 10 SPAR-IX FSLP/Spacelab 1/SL-l: Spacelab +1 pallet Spacelab 3/SL-3: Spacelab + mission-peculiar support structure German Spacelab mission Dl: Spacelab + support structure EURECA: materials science payload with automated facilities Spacelab-J(apan): Spacelab German Spacelab mission D2: Spacelab + unique support structure
543
12.2 Driving Forces in Microstructure Formation
12.1 Microstructures Microstructures describe the microscopic state of condensed matter short of its crystal structure, in other words, down to resolutions of several atoms across. The classical ingot microstructure, such as shown in Fig. 12-1, encompasses its various forms in one piece. Nucleation is responsible for the number of grains and solidification-front morphology for grain shape. Fig. 12-2 indicates basic possibilities for the latter. Of particular interest are transitions between equiaxed (globular) and columnar grains, the latter of which may also lead to single crystals if grain boundaries are eliminated.
Figure 12-1. This quenched (at lower end) and decanted (upper end) Al-base alloy ingot shows the typical equiaxed (globular) microstructure at the rapidly cooled bottom part, transforming into columnar microstructure above, and finally revealing a dendritic growth interface at the decanted top (Sahm, 1986).
VTs
12,2 Driving Forces in Microstructure Formation The driving forces for solidification are various types of undercooling. The total undercooling, ATU, is given by: ATU = ATt + ATk + ATC + ATr
(12-1)
A melt will feel thermal undercooling, ATt, if a sufficiently strong cooling process is superimposed, normally by applying very high cooling rates f. This will unavoidably lead to the event of nucleation, unless ATU > ATg where ATg is the undercooling to the glass temperature Tg. Kinetic undercooling, ATk, plays a role in determining the formation of certain facets in single crystal growth. Constitutional undercooling, ATC, represents the outstanding phenomenon in morphological stability of crystallization fronts in alloys, e.g., in materials which consist of several components. The curvature-dependent undercooling, ATr, essentially characterizes equilibrium conditions between the crystallization-front morphology and the other
Q) O
Planar
5 0
rrrrm
"c
.... T
i
o
CD
Cellular
r
O O
T,
&
c
—T-
s J
Dendritic o
e
T
& B CO CD
"5 ° —T
% o
Globulitic Figure 12-2. The basic types of growth-interface morphologies determine the resulting microstructure. The growth morphology is determined by process parameters such as growth rate v, temperature gradient G, concentration ci of alloying elements. Tx is the liquidus temperature, Ts the solidus temperature (Sahm, 1986).
544
12 Solidification Processing under Microgravity
I
p |a:
(a)
(b)
(c)
(d)
Figure 12-3. The scale of various solid-liquid interface morphologies. Solidification morphologies are determined by the interplay of two effects acting at the solid-liquid interface. These are the diffusion of solute (or heat), which tends to minimize the scale of the morphology (maximize curvature) and the capillarity effect, which tends to maximize the scale. The crystal morphologies actually observed are thus a compromise between these two tendencies, and this can be shown with respect to nucleation (a), interface instability (b), dendritic growth (c), and eutectic growth (d) (from Kurz and Fisher, 1984).
driving forces just outlined (Fig. 12-3). The two particularly microgravity relevant driving forces are ATt and ATC, although ATk may also be important in this context, if further investigations uncover information that was not heretofore known (see Chap. 1, Sees. 1.1-1.5 and Vol. 5, Chap. 10, Sec. 10.2).
12.3 Nucleation and Undercooling 12.3.1 Theoretical Background
The non-equilibrium state of an undercooled melt can be achieved following two different paths, namely, rapid quenching or containerless undercooling, both of which may result in high undercooling temperatures prior to nucleation or glass formation. In particular, the brute force method of applying very high cooling rates, for example, in a melt-atomization or melt-spinning approach, will reach high ATt values according to ATt = f t q .
(12-2)
Here, the quenching time tq must be kept extremely short, normally presupposing samples with very large surface-to-volume ratios, S/V, to secure high t values. Containerless processing, however, will allow experimentation of samples of larger interconnected volume with decreased S/V and lower cooling times (i.e., increased tq values). Utilizing containerless processing, one is able to thermally undercool molten material without the influence of container walls which normally cause heterogeneous nucleation and thus make it difficult to study the behavior of undercooled melts, even if the melt is otherwise very clean. An interesting implication within this context may be the following question: Will it be possible to produce larger volumes of amorphously solidifying metallic material? A controversial working hypothesis, which states that in metallic melts there is no homogeneous nucleation, but rather only heterogeneous nucleation, can be pursued by microgravity experiments with containerless melting and solidification, thus presenting an intriguing scientific goal (Sahm, 1983).
12.3 Nucleation and Undercooling
545
Nozzle RF-Coil
Vacuum measurement Temperature measurement RF-Generator Pumping system
(a)
Figure 12-4. Various types of drop facilities: (a) laboratory drop tube, height 3 m: 1/3 s low gravity (Kallien and Sahm, 1986); (b) drop tower Bremen, height 110 m, 4.7 s free-fall time (courtesy of ZARM, Bremen, FRG); (c) underground drop tube, depth 101.7 m, 4.3 s free fall time (courtesy of NASA-MSFC, Huntsville, USA) (see next pages).
The potential for studying a host of phenomena in nucleation and micro structure formation from the under cooled melt by means of experiments in microgravity appears to be great. 12.3.2 Procedures, Methods, and Apparatus
A large amount of work relevant to undercooling and nucleation has been and is being done utilizing experimental approaches such as melt-spinning or planar
flow casting (see, for example, Hug and Sahm, 1988), high-cooling-rate directional solidification (see, for example, Kiminami and Sahm, 1986), slag-envelope large-interconnected-volume undercooling, as previously reported by Bardenheuer and Blechmann (1939) or Kiminami et al. (1989), undercooling of fine-droplet dispersions (see Perepezko, 1980), and laser or particle beam treatments (see, for example, Spaepen, 1986 or Mordike, 1990). Two methods allowing containerless processing without
546
12 Solidification Processing under Microgravity
146 m
Top details 119m
Release mechanism
Top details
Capsule elevator Drop capsule
Drop shaft
Elevator
Compensator Ground details Hanger bracket Deceleration chamber Deceleration container Hanger bars Foundation of drop shaft Ground details Multipurpose area 13 m 10m
Laboratories and service rooms
Acceleration shaft -JJ m
Foundation
-16m
Figure 12-4 (b)
going into space are to utilize drop tubes or towers (Fig. 12-4) or levitation techniques. Evacuated drop tubes or towers, with drop heights ranging from 3 m to > 100 m, can provide microgravity levels down to as
little as 10" 6 g for 1 to 10 s. Such facilities are, for example, described in an ESA Report (1989), and are normally used to study undercoolability and nucleation, metastable phases or glass formation (e.g. Lacy
12.3 Nucleation and Undercooling
Pyrometer
Turbomolecular pump Instrument port
Helium
Roughing pump Instrument port 5th . floor' View port
Instrument port
4th floor Instrument 3rd floor
View port
View
Instrument port
Turbomolecular pump
Ground floor Roughing pump Figure 12-4 (c)
et al., 1981; Lacy et al., 1982; Drehman and Turnbull, 1981; Vinet et al., 1991). Other methods for studying nucleation and undercooling are levitation melting and solidification. These techniques suspend the sample to be studied by non-contact forces such as acoustic fields, air streams (acoustic or aerodynamic levitators), or electromagnetic forces (electrostatic or electromagnetic levitators). In acoustic levitators (see, for example, Wang, 1979; Lierke and GroBbach, 1975; Lierke et al., 1983), standing waves are created within a gaseous ambient. The sample will experience a restoring force if it is
547
placed in the vicinity of a standing-wave mode. Acoustic levitation is particularly suitable to process glasses. In contrast to other methods, the materials to be processed do not have to be metallic or electrically charged and, in the case of oxidic glasses, it is highly desirable to melt them in a gas atmosphere to maintain stoichiometry. Different approaches have been pursued so far: - Single-axis acoustic levitation by creating a cylindrical cavity utilizing a single acoustic source (Lierke, 1976). - Three-axis levitator: Standing waves are produced in a rectangular cavity using three orthogonal drivers to produce an energy well at the center of the cavity (Wang et al., 1974); this device allows low-density samples to be levitated in unit gravity. - Single-axis interference levitator (see Whymark, 1975). A completely different approach to achieve levitation involves the use of air or gas in aerodynamic levitators (see Oran et al., 1982). The samples are supported by high-speed flow of gases provided by properly arranged nozzles. Alternatively, levitation may also be obtained by utilizing electrostatic forces. Because these forces have to be dynamically controlled, an arrangement (see, for example, Rhim et al., 1985) is usually applied. The specimen position is monitored and controlled through the electrode potentials. A very appropriate method for metals is the electromagnetic levitation technique. The metallic sample is placed into a nonhomogeneous electromagnetic field created by an RF-coil. This induces eddy currents in the sample which, in turn, create a magnetic field opposite to the external field. Consequently, the sample is repelled against the direction of the external coil field. The repulsive forces are then bal-
548
12 Solidification Processing under Microgravity
anced out by the sample weight. This method was patented by Much (1923) and first experimentally verified by Okress et al. (1952). A theoretical analysis has been reported by Rony (1964). Electromagnetic levitation is a very powerful tool for studying both liquid and solid metals. Numerous physical properties of undercooled molten metals may be investigated, e.g., surface tension and viscosity (see Mogi et al., 1986), density (see Shiraishi and Ward, 1964), undercoolability (see Willnecker et al., 1988), solidification-front dynamics (Schleip et al., 1988), and growth-front morphology (Walker, 1961). The disadvantages of earth-bound electromagnetic levitation are that both levitating and heating are caused by the same eddy currents. Thus, - a lower temperature limit for earthbound experimentation of approximately 1273 K normally exists for metals; - temperature measurement and control are difficult; - dynamic nucleation may result on account of electromagnetic stirring; - to expand the processable temperature regime, the use of cooling gases such as helium or hydrogen is required; this involves purity problems. To circumvent these shortcomings a microgravity environment would be helpful. Toward this end, a containerless processing unit named TEMPUS (tiegelfreies elektromagnetisches Positionieren unter Schwerelosigkeit) has been proposed and constructed, Fig. 12-5 a (see Piller et al., 1986). TEMPUS permits - processing of liquid metals within an extended temperature regime from 660 to 2700 K; - independent control of heating and positioning; a two-frequency concept is realized using a quadrupole field for posi-
tioning and a dipole field for heating, Fig. 12-5b,c; - processing at highly reduced positioning forces, resulting in diminished electromagnetic stirring; - application of an ultrahigh vacuum. The system permits the measurement of undercoolability, viscosity, surface tension, solidification-front velocity, and other phenomena within the undercooled state as a function of temperature. 12.3.3 State of the Art
On account of experiments by Turnbull (1950), it was believed for a long time that the maximal undercoolability of metallic melts would be ATu(max) = 0.18 Tm where Tm is the melting point (Turnbull and Cech, 1950). They subdivided a melt into numerous small droplets (a few micrometers in diameter) thereby diminishing heterogeneous nuclei on statistical grounds (smaller volumes contain fewer defects). They investigated a series of metals which showed the relationship mentioned above, Fig. 12-6, by posing them on a heated ceramic substrate and observing nucleation events under a microscope while cooling down the substrate. This basic idea was extended significantly by Perepezko (1984), who substantially increased the undercoolability of liquid metals by eliminating contacts with container walls (e.g., a ceramic substrate, sometimes used by Turnbull) utilizing liquid encapsulants in the form of oils or molten salts. Later, electromagnetic levitation added more data under still further improved conditions, particularly for the higher-melting metals. While Fig. 12-6 illustrates undercooling data for pure metals, Table 12-1 lists such data for several alloys. Electromagnetic levitation presents the only known technique to date that allows
12.3 Nucleation and Undercooling
CCD video camera
549
Two-color pyrometer
Vertical transfer mechanism
Window drive mechanism
Water pump Compensating volume
Motor
RF heating generator
RF positioning generator
(a) ultra high vacuum
positioning coils f : frequency
ultra pure gases
two colour pyrometer high speed device
heating coils
^
A
sample
I (b)
specimen manipulator
(c)
Figure 12-5. (a) The electromagnetic levitator TEMPUS, Germany, which utilizes a two-coil system, for independent positioning and heating - precise temperature measurements and ultrahigh vacuum enable well-controlled investigations into molten melt behavior (courtesy of Dornier Deutsche Aerospace, FRG); (b) the two-coil system (Walter, 1987); (c) the levitator in action (Walter, 1987).
550
2000
Tm IK) j
12 Solidification Processing under Microgravity
Dispersion of the melt in small droplets
Fe
1500 * Containerless undercooling by el. magn. le vitation
1000
Sb
•
500
= 0.18 Embedding small droplets in an inert emulsion
i
Sn •— Hg 100
200
_L ^ 300 rmoxlKl
Table 12-1. Maximal undercooling attained for alloys. Alloy Sn - Pb (> 10 at. % Sn) Sn - Pb (< 10 at. % Sn) Sn - Bi (< 45 at. % Bi) Pb-Bi (<41 at.% Bi) Pb-Sb(3.7at.%Sb) Bi-Cd (30-60 at.% Cd) Te-Cu (< 12.5 at.% Cu) Te-Cu (19-39 at.% Cu) Nb-Ge (13-25 at.% Ge) Al-10%Sn Fe-0.32%C Al-Si 4330 Steel (4330) Steel, 440 C Fe-Ni (0-100 at.% Ni) Cu-Ni Stainless steel (316) Ni-32.5% Sn Pd 40 Ni 40 P 20
T[K] AT/TL
Method
160 180 225 160 144 150 264
0.26 0.38 0.40 0.35 0.26 0.32 0.36
small particles small particles small particles small particles small particles small particles small particles
495 99 281 200 150 355 268 475 380
0.22 0.11 0.16 0.11 0.09 0.20 0.18 0.28 0.27
drop tube small particles slag envelope slag envelope slag envelope levitation levitation slag envelope slag envelope
the experimenter to measure thermophysical properties of undercooled metallic melts. Very promising results have been obtained so far. For example, Fe-Ni alloys have been undercooled to levels never be-
400
Figure 12-6. Maximal undercooling temperatures ATu(max), found experimentally as a function of melting temperature, here Tm, for pure metals (Zarzycki et al, 1988).
500
fore attained (see Willnecker et al., 1986). Still, solidification has always started at heterogeneous surface sites, and homogeneous nucleation has not yet been verified (Fig. 12-7). As concerns microstructures forming at high undercoolings, the solidification-front velocity v and resulting grain sizes were examined as a function of ATU by the same investigators. The two alloy systems Cu 70 Ni 30 and Cu 6 9 Ni 3 0 B 1 deviated from the expected potential behavior ( D ^ A T J at certain ATU values, namely, at ATU = = 193 K for Cu 70 Ni 30 and at ATU = 225 K for CugQl^oBj. These values exactly correspond to the temperatures at which a sudden decrease of grain size is observed (Fig. 12-8). It was also observed that, although the degree of undercooling was different for the two alloys, the corresponding solidification velocities were the same, namely, v = 20 m/s, indicating that the reason for the grain refinement appears to be a kinetic effect rather than a matter of undercooling.
12.3 Nucleation and Undercooling iyuu
1
1
1
1
I I i horn . nucleation - het. nucleation
I ~~~
—— ^
1700
—
— 1500
T [K]
1300
Fe
1900 1800 1700 1600 1500 1400
1100
66
Ni
35
MAT
(D
120
60
I
140 240 300 t fs] ,
20
0
—
•
,
I
40
<=
I
I 80
60
Fe
551
I
Figure 12-7. Phase diagram of Fe-Ni with results of undercooling experiments utilizing bulk samples processed by electromagnetic levitation (closed dots). Dashed lines indicate calculated curves for concentration-dependent undercooling relative to different nucleation processes. The insert delineates a typical temperature-time profile with the steep rise of temperature after the onset of solidification (Willnecker et al., 1986).
100 Ni
at. % Ni
300
200
200
100
AT[K] AT*= 225 K
AT*= 193 K
I c
1 1
.
*
30 -
£
%
20
V*
10 "-
100
f »
/ .
**
V
i AT*
,AT*
! ,
i
1 f
200
'
200
Undercooling
AT [K]
i
300
Figure 12-8. Grain sizes and solidification-front velocities versus undercooling temperatures (after Willnecker et al., 1986).
552
12 Solidification Processing under Microgravity
12.4 Solidification-Front Dynamics
cooling rate [K/s]
5000
1700
950
600
420
i
i
0.8 -
31
-
Q. CO 3
0.6 -
\ \
0.4 0.2 i
0 300
500
i
700 900 grain size [urn]
1100
1300
Figure 12-9. Glassy fraction in Pd 77 5 Si 12 5 Cu 6 particles solidified without a container utilizing a laboratory drop tube (after Hug et al., 1986).
The experiments just described were performed under normal gravity. Concerning experiments under microgravity, the easiest way to produce weightlessness is the use of free wall, which can be achieved in drop tubes. An example for free-fall containerless solidification was conducted with Pd 77 5 Si 16 5 Cu 6 droplets (see, for example, Hug et al., 1986). Partly amorphous or completely crystallized particles were thus produced, depending on the particle sizes chosen and cooling rates realized (Fig. 12-9). Unfortunately, drop-tube experiments only allow the solidification of small samples. Only a few experimental results from prolonged space flights have become known so far. These investigations, mainly on glasses and oxides, have been published in various proceedings (e.g. NASA, 1980; ESA, 1983; ESA, 1987). The TEMPUS facility (see Fig. 12-5) is planned for launch on the IML-2 Mission in 1992.
The central purpose of all solidificationlinked microgravity research is an increase in our understanding of interactions between fluid mechanics and the liquid-solid phase transition. This may be realized by separating gravity-induced transport mechanisms such as buoyancy, sedimentation, and natural convection from gravity-independant processes such as diffusion and specific types of convection, in order to study the contribution of individual mechanisms to the overall heat and mass transport and to determine growth-morphology transitions. The following basic microgravity-relevant effects pervade nearly all of materials-processing phenomena: Gravity-dependent transport processes (1) Buoyancy and sedimentation induced (2) Natural convection induced by AQ(T) and A£ (c). Gravity-independent transport processes (1) Convection caused by A^ liquid . solid
VA (2) Marangoni convection induced by A<7(T) and Aa(c). (3) Diffusion with AQ = density differences Ad = surface energy differences T = temperature c = concentration AQ (7^ c) = thermally and/or solutally induced density differences AF m ? H s = volume increment at the melting point, when transforming from liquid to solid. The field of research called solidification-front dynamics deals with the interaction between the solidification front and
12.4 Solidification-Front Dynamics
553
Figure 12-10. Constitutional supercooling in alloys. The steady-state diffusion boundary layer for a given growth rate is shown in the upper diagram. It corresponds to the concentration pileup ahead of the solidification front. The temperature profile 7J taken from the phase diagram (lower right), has been transferred to the diagram in the lower left. The actual temperature Tq, imposed by the temperature gradient arising from the heat flow in the casting is shown by the dotted line. Directly at the solidification line Tq is lower than T,. The cross-hatched region is called zone of constitutional supercooling (Kurz and Fisher, 1984).
macroscopic or microscopic convection which strongly affects the mass- and heattransport processes, including the interaction with discrete particles present at solidliquid interfaces. When working with alloyed melts, constitutional undercooling ATC becomes a determining factor in microstructure formation. It is due to thermal gradients too low to compensate for concentration pileups ahead of solidification fronts and results in regimes of supersaturation, Fig. 12-10. Constitutional supercooling allows the study of morphological stabilities interacting with thermal or solutal buoyancy driven convection in close vicinity to the solidification front. Thus, ATC is one of the essential parameters in solidification-front dynamics. Effects connected with supersat-
uration are also responsible for observations made in [ig vapor-phase growth. Simplifying, ATC can be derived from the stability criterion for growth front interfaces:
ATc=GD/v
(12-3)
where ATC is the constitutional undercooling temperature, D the average atomic diffusion coefficient in melt, G temperature gradient with liquid at the crystallization line, and v the growth rate. Equation (12-3) has been obtained from the better-known expression
G/v = mco(l-ko)/koD
= (12-4)
in which m is the slope of liquidus line corresponding to solidifying alloy, k0 the
554
12 Solidification Processing under Microgravity
equilibrium distribution coefficient, TY the liquidus temperature corresponding to real (supersaturated) melt composition, and Tq the real temperature of the melt at the same locality. 12.4.1 Solidification-Front Morphology 12.4.1.1 Theoretical Background
The two undercooling terms ATC and ATr equalize each other, emphasizing the strong iteration of solutal effects and growth-front morphology curvature. Its relevance to microgravity research lies in the fact that the effective distribution coefficient depends on a boundary layer. In the stagnant-film model of Coriell and Sekerka (1981) the film adjacent to the solid-liquid interface has the thickness Sx. It helps to define an effective distribution coefficient ett
ko + (l-ko)exp(vSJD)
(12-5)
As the dx value sensitively depends on convective conditions at the growth front, the comparison of results obtained in the normal-gravity laboratory with those found under jig indicates that jig renders an ideal environment for verifying models of solidification theory. Obviously, conditions generated by thermosolutal convection can thus be studied in much more detail and with more accuracy than on earth. (The term "thermosolutal convection" refers to natural convective moments caused by thermal and/or solutal (i.e., concentration) gradients, which, in turn, cause buoyancydriven flow. It has become an accepted term in the microgravity community.) 12.4.1.2 Procedures, Methods, and Apparatus
Precise and reliable parameters on heat and mass transport in liquids can only be
obtained if the convection contributions strongly masking these processes on earth are eliminated. The microgravity environment, which suppresses such gravitydependent processes, is the ideal experimental opportunity to obtain the exact experimental data essential for any theoretical modeling for solidification-front investigations. Thus, determination of diffusion coefficients in metallic melts (Sn/In) in microgravity has revealed a precision to better than 1%. Highly accurate thermophysical data are without doubt a prerequisite for such modeling, as are welldefined and optimized solidification conditions. The high-temperature thermostat (HTT), specially developed to investigate selfdiffusion and heterodiffusion mechanisms in liquid metal, allows isothermal treatment of eight samples between 525 K and 1673 K. Each sample cartridge equipped with two heating filaments enables a defined temperature gradient for heating up and cooling down and a defined isothermal temperature profile over the holding phase. As suggested by Fig. 12-10, a very effective procedure to study thermosolutal effects is to measure concentration profiles at the solid-liquid interface. Because temperature gradients are also important (Fig. 12-10), the furnaces to be utilized for this task have to be able to yield a variety of temperature profiles and be equipped with a built-in quenching capability. A furnace developed for experiments of this type, the gradient furnace with quenching (GFQ) is shown in Fig. 12-11. It has yielded several extremely interesting results during the German spacelab mission Dl. Local information on the solidification process by in situ monitoring of the crystallization front is of prime importance in fundamental studies. Current pulses inducing periodical striations, thus delineating
12.4 Solidification-Front Dynamics
555
Processing Chamber H2O Filter / Sample Holder
Guide Rail
^Sample
nut- and spindle drive
Figure 12-11. The GFQ, Germany, is a Bridgmantype furnace with two individually controlled heating zones and a water cooled zone. The solid-liquid interface is located in the adiabatic zone to minimize radial temperature gradients. For quenching, the entire furnace is displaced rapidly, spraying the liquid section to be quenched with water (courtesy of Dornier Deutsche Aerospace).
the growth front, have been proposed. A Peltier demarcation technique has been successly used in a sounding rocket (TEXUS) furnace (TEM 02-3) and is the gradient heating facility (GHF), see below. In addition the Peltier demarcation measurements, which make use of the Seebeckeffect, another in situ diagnostic technique is being offered by the MEPHISTO facility, Fig. 12-12 a. The determination of the emf at the two solid-liquid interfaces of the
same sample makes it possible to obtain continuous feedback about the solid-liquid-interface temperature (for the measurement principle, see Fig. 12-12 b). A third path offering insights into solidification-front dynamics involves the employment of transparent model systems of low-melting organics. The methods focus on investigations of the interface morphology by observing macro- and microconvective flows by tracers and non-invasive
556
12 Solidification Processing under Microgravity
Temperature (up to 900°C)
50°C
Liquid
Moving furnace
Cu wire
Figure 12-12. The purpose of MEPHISTO, France, is to perform fundamental studies on the effect of growth parameters and convection on the growth mechanisms and characteristics of the solidified sample, (a) The system consists of a mirror-symmetrical arrangement of two Bridgman-type furnaces in which the center section of a long sample can be partially remelted. (b) By keeping one of the solid-liquid interfaces stationary and by remelting or solidifying at the opposite interface, an electromotive force (emf) is generated between the cold ends (Seebeck voltage). From this voltage the actual temperature of the non-stationary solid-liquid interface can be determined with high resolution (ESA, 1989).
determination of temperature and concentration fields utilizing holography, interferometry, and polarimetry. The holographic optics laboratory for D2 (see Fig. 12-13) a substantially improved facility derived from the precursor Dl model, deals with studies of the kinetics of nucleation, growth of nuclei, phase separation, diffusion processes, and the like in transparent media.
12.4.1.3 State of the Art
The stability of a planar unidirectionally solidifying solid-liquid interface is governed the G/v criterion of Eq. (12-3). Instabilities of the planar interface will result in the formation of cellular, or even globular interface morphology (Fig. 12-2). Interface instabilities are mirrored by temperature and concentration profiles at
12.4 Solidification-Front Dynamics Thermoplastic-camera
557
Video-camera head
HV Laser supply
Experiment drawer (exchangeable) Lens Beam steering module Beam tilting module
Mirror
Object beam
Beam splitting module
V2-plate (adjustable)
Figure 12-13. Optical setup of the holographic optics laboratory (HOLOP), Germany. By two-wavelength doubleexposure holography, refractive-index variations due to thermal or concentration gradients or phase discontinuities can be observed in transparent media. A television link to the ground will allow teleseience operations (courtesy of Dornier Deutsche Aerospace, FRG).
the solid-liquid interface, and these will reflect thermosolutally caused convective motions. Transparent model materials allow direct visualization of such effects. For example, in a TEXUS 10 experiment and in the German Dl spacelab mission a holographic interferometer (HOLOP) was used to visualize the boundary layers, as previously reported by Ecker et al. (1987). The model system, succinodinitrileethanol, displays numerous characteristic properties of metallic alloys and thus has served as a model substance for studying metallic solidification (see Glicksman, 1981).
On earth, succinodinitrile-ethanol has an undisturbed concentration boundary layer only at high solidification velocities. At lower velocities, rising droplets of the ethanol-rich phase destroy the boundary layer. In space, a stable concentration boundary layer is observed for all solidification velocities. The thickness of this boundary layer agrees well with theoretical predictions (Ecker et al., 1987). Tensi and Schmidt (1987) measured the thickness of the concentration boundary layer of an Al-0.3wt.%Cu sample under 1 g and |ig conditions. The alloy was solidified in the GFQ (Fig. 12-11) during the Dl
558
12 Solidification Processing under Microgravity
Figure 12-14. Comparison between corresponding cross sections of Pb-30 wt.% Tl samples grown vertically (a) on earth and (b) in space. On earth the shape of the cells is affected by four convective rolls. In space the cells are more regular and are coarser (Billia et al., 1987). (a)
spacelab mission. The thickness and the concentration profile were in close agreement with calculated profiles using the pure diffusion theory. The characteristic thickness of the solute boundary layer was 23.5 times larger in space than in the terrestrial reference experiments. The reason for the increased boundarylayer thickness in the space experiments, both for the organic model system and the metallic alloy, is the absence of convection in the melt: convection inhibits the establishment of purely diffusion-determined concentration profiles. Billia et al. (1987) conducted experiments on the morphological stability of directionally solidified Pb-Tl samples in the GHF (see below) during the Dl mission. Owing to solutal convection in the melt at the interface in concentrated alloys (25, 30 and 40 wt.% Tl), the morphology of the cellular pattern was strongly affected. For instance, a comparison of earth and space samples of the Pb-25 wt.% Tl alloy shows several convective-induced rolls to have developed in the ground experiment, destroying the regularity of the cells, Fig. 12-14. Also the average cell size is larger in space than on earth.
(b)
102
Al Mg G = 150K/cm
o°
10"
growth
[cm/s]
Figure 12-15. Computations were made by Coriell (1987) for an Al-0.3 wt.% Mg alloy to determine the stability boundaries. The diagram shows the critical concentrations as a function of growth velocity, for the onset of morphological instability (G/u-criterion) and convective mixing. The calculations were made for earth conditions at the 10 ~A g acceleration level, which is representative for conditions in spacelab. The circle indicates the experimental parameters chosen for the Dl experiment (Rex and Sahm, 1987).
12.4 Solidification-Front Dynamics
ug 1.2 0.9
0.6 0.3
n 40
50
60
70
80
90
100
x in mm
(a) 0.9 r
0.6
0.3
30
(b)
60
90
120
x in mm
Figure 12-16. The macrosegregation profile of the jag solidified sample (a) differs substantially from that of a 1 g-reference sample. Without convection, only a 4 mm initial transient is needed here to reach nominal composition. The ground sample (b) remains at the low level given by the distribution coefficient. Furthermore, regarding the total loss of Mg, this seems to be clearly enforced by convective mixing in the ground experiment (Rex and Sahm, 1987).
The effect of thermosolutal convection on macrosegregation was reported by Rex and Sahm (1987). A sample of Al-0.3 wt.% Mg was solidified in a Bridgman configuration on earth and in the GFQ furnace during the Dl mission. For the low Mg concentration, a convectively unstable boundary layer was predicted only for 1 g conditions. According to the stability diagram, Fig. 12-15, as calculated by Coriell (1987), heat and mass transport in the melt
559
were purely diffusion controlled under jig conditions. Fig. 12-16 shows the macrosegregation profiles of the 1 g and jig specimens in the longitudinal direction. In the ground experiments, convective melt motion diminished the Mg pile-up at the solidification front. The concentration profile of the flight sample, (a), fulfilled the classic case of segregation without mixing in the melt as predicted by Coriell (1987). Only a 4 mm initial transient was needed to reach the normal position at 0.3 wt.%Mg. At higher undercooling, the solidification front becomes morphologically unstable, and dendrites appear. Tensi and Schmidt (1987) and Tensi et al. (1989) crystallized an Al-7 wt.% Si alloy to study the influence of gravity on the spacing of the dendrite arms. The experiments were done at growth velocities of 5 and 8 mm/min. Fig. 12-17 shows a longitudinal meridian section of the sample in the area of coarsening. A comparison of the experimental results for 1 g and jig conditions shows the dendrite-arm spacing to be nearly the same for v = 8 mm/min; however for v = 5 mm/ min the dendrite-arm spacing of the \ig
Area of coarsening
II. SF
I. SF
Figure 12-17. Metallographic longitudinal meridian section of an Al-7 wt.% Si specimen in the area of coarsening between the two solidification fronts (I. SF and II. SF). On the right of I.SF: quenched residual melt; on the left of II. SF: unidirectionally solidified eutectic volume (Tensi et al., 1989).
560
12 Solidification Processing under Microgravity
D-1 experiment Space
I
1CT 6
icr5
icr4
10"3
icr2
10-
Solidification rate (cm/s)
Figure 12-19. Primary interdendritic spacing versus growth rate: (solid line) theoretical, (•) ground and Dl experiment (Al-26 wt.% Cu, G = 25Kcm~ 1 ; Camel etal., 1987).
(b) Figure 12-18. Comparison between corresponding cross-sections of Al-26 wt.% Cu samples, solidified (a) on the ground and (b) in space, shows the presence of radial segregation in the first case and its absence in the second; the latter also displays coarser and more regular dendrites (Camel et al., 1987).
sample was higher than that of the 1 g sample. This behavior can be explained by the influence of gravity-driven convection on the dendrite spacings, which appears to stay for low crystallization velocities. Dendritic growth of a hypo-eutectic Al26 wt.% Cu alloy under 1 g and jug conditions were studied by Camel et al. (1987). The results of the microgravity experi-
ments were in excellent agreement with theoretical predictions: no radial nor longitudinal segregation in the hypo-eutectic samples, except in the final transients (i.e., diffusion-controlled conditions), is prevalent. The primary-arm spacing Xal is found to be five times larger than in the ground experiments, Fig. 12-18, the ideal regularity and large size of the dendritic array of the space sample are obvious at first glance. In Fig. 12-19 the variation of Aal with growth velocity is plotted for this alloy. The space result corresponds with the classical diffusion law, whereas the ground result confirms the scaling-law prediction of the growth regime affected by convection. This space experiment validates the modeling of solutal convection in dendritic growth. It could be useful to optimize dendritic-solidification processes on the, ground. A complete study of dendrite-arm morphology has been undertaken. A recon-
561
12.4 Solidification-Front Dynamics
Figure 12-20. Three-dimensional reconstruction of an aluminium dendrite grown in space: primary, secondary, and tertiary arms become visible (Camel etal., 1987).
struction of the three-dimensional shape of a singular dendrite array has been analyzed by preparing successive cross sections (in 20jim steps) and superimposing the images, Fig. 12-20. A stability analysis of the three-dimensional shape of primary, secondary, and even tertiary arms thus became possible, an important step toward a qantitative analysis of Ostwald-ripening phenomena within interdendritic regimes (see below). 12.4.2 Single Crystal Growth 12.4.2.1 Theoretical Background
In contrast to normal polycrystallinemetallic-alloy solidification in which incongruent melting behavior prevails, sin-
t. V Ol/Tlo) ATK
facetted growth / / ;
transitional regime / ;
ATk (facetted growth) = l/C1 In (v/C2) where Cx and C2 are constants. Assuming, however, that growth takes place with an atomically diffuse interface, the kinetic undercooling term is to be described by ATk (diffusive growth) = v/C3
ATk (transitional regime) = C 4 v1/2
ATK
t / i
ATK ATK
ATK' —
b.)
(12-8)
Crystal-growth phenomena may effect the ATk term via the generation of various crystal-lattice defects and dopant distributions and may thus possibly shift growth regimes within the indicated plot of Fig. 12-22.
V(T1/T1O)
; diffusive growth
(12-7)
in which, again, C 3 is a constant. The regimes may be distinguished by consulting Fig. 12-21. The transitional regime between the two (i.e., screw-dislocation nucleated growth) is given by:
' i '/T" ; /
a.)
gle-crystal growth mostly concerns either one- or multicomponent single-phase materials with congruent melting points. Nevertheless, single-crystal growth may be considered a special case in solidificationfront dynamics. The undercooling term most essential for single-crystal growth is kinetic undercooling, the ATk of Eq. (12-1). Several forms of kinetic undercooling are known. For example, presupposing growth of strongly facetted crystals (in which twodimensional nucleation is predominant), ATk is described by (12-6)
/
/
Al K ATK
Figure 12-21. Three regimes of crystal growth are shown as a function of kinetic undercooling and plotted against a viscosity dimensionalized growth rate (Tiller, 1971).
562
12 Solidification Processing under Microgravity
(a)
(b)
(c)
Figure 12-22. Micrographs of phosphorus-doped Si crystal: (a) grown under gravity with free-melt surface; (b) grown under microgravity with free-melt surface; (c) grown under gravity with SiO2 coated surface (Crolletal., 1990).
For growing striation-free crystals with methods involving free-melt surfaces, Marangoni convection must be avoided. Marangoni convection is induced by surface tension gradients with temperature gradients along a free surface (liquid-gas or liquid-liquid interface); thermocapillary flows will arise from such gradients. Because Marangoni convection is not caused by buoyancy forces, it will be unaffected by a change from a 1 g to a jo,g environment. It is mainly important in the floating-zone growth of crystals (Walter, 1984). Component separation in Marangoni convection
and crystal growth were studied in spacelab (FSLP, Dl) and in several rocket (TEXUS) flights. In most of the experiments, phosphorus-doped silicon was used. The result common to all of these experiments was that dopant striations essentially disappeared in jig, except for the Marangoni component, Fig. 12-22 a, b. To avoid Marangoni-convection-induced striations, a thin SiO2 coating can be used, which gets rid of free surfaces and thus also eliminates Marangoni convection, Fig. 12-22 c. 12.4.2.2 Procedures, Methods, and Apparatus
Presently, crystal growth from the liquid phase mostly employs zone-melting and Bridgman techniques. In addition, the diverse containerless processes now under development are promising a new and potential field of interest for wall-free crystal growth. Inherent in all the developments of heating facilities for space use is the restricted availability of power. Therefore mirrorheating methods appeared particularly appropriate as heating concepts, fulfilling the diverse requirements of the crystal growers' community. On the basis of a technique where radiation from a halogen lamp is focused on the sample, the attainable processing temperatures have turned out to be sufficient for growth experiments by floating-zone and traveling-heater methods. Double ellipsoidal mirrors, and later single ellipsoidal mirrors, became the basis for a family of furnaces in sounding rocket flights (monitored by telecommand), for spacelab and for automatic operation on retrievable carriers such as EURECA. The automatic mirror furnace (AMF) as part of the core payload for EURECA and the recent development for
12.4 Solidification-Front Dynamics
D2, ELLI, are depicted in Figs. 12-23 and 12-24. Pioneering work on such radiationheating systems mainly stems from Germany and Sweden. Bridgman-type gradient-heating facilities have been built and successfully employed in manned and unmanned space missions worldwide (see overview by Steinborn, 1986). These gradient furnaces are mainly for directional solidification or Bridgman crystallization. Variable growth rates can be achieved either by sustaining a defined temperature gradient ahead of the solid-liquid interface by independently controlled heaters or by displacing the furnace assembly relative to the sample. Both concepts have been put into effect by the French gradient-heating facility (GHF) (Fig. 12-25) and were operated on the first spacelab payload, FSLP, in 1983 and the spacelab mission Dl (1985). It is now being prepared for the spacelab mission D2. For generating even higher temperature gradients by coupling into a liquid metal cooler, the advanced gradient heating facility (AGHF) (see Fig. 12-26) should also be mentioned. Its basic concept consists of a heating section with two individually controlled heaters acting on a common thermal diffuser separated from the cooled heat-extraction zone by a sizeable adiabatic zone. It is to be employed in future missions. Crystal growth from the vapor phase under microgravity also promises benefits over ground-based techniques. This particularly concerns materials which decompose prior to melting. Vapor growth is also the only suitable method for obtaining crystallized materials for which suitable solvents do not exist. On earth, the production of large crystals (up to 10 cm) requires large reaction chambers, which, in turn, increases convective instabilities and leads to
563
undesired defect densities and crystal inhomogeneities. Of the four main vapor phase techniques, PVD (physical vapor deposition) and CVD (chemical vapor deposition) require open systems and, for safety reasons, have not been further pursued for utilization in space. In contrast, PVT (physical vapor transport) and CVT (chemical vapor transport) appear to be quite adaptable to |ig requirements. The PVT process is based on the sublimation of a source material followed by transport of the gaseous substance via the phase until it finally nucleates on the sink material. The method presupposes a material with sufficiently high vapor pressure and a temperature gradient along which the gas-solid equilibrium is shifted. PVT (and PVD) processes are mainly applied to growing bulk crystals. Figure 12-27 shows the scheme of an apparatus flown on the spacelab mission SL-3 in 1985. From the source material, Hgl 2 , which was placed around the chamber wall, the crystallization process starts at the cooled pedestal at the bottom of the growth chamber. The CVT methods (and CVD methods) do not require volatile solids and are used in particular for depositing thin (epitaxial) layers. Here, a gaseous reaction product of the source crystal and a reactive carrier gas that decomposes at the growing crystal interface determine the process. In a cyclic process, the carrier gas will be set free again and will re-react with the starting material. A characteristic of a large number of substances being processed by vaporgrowth techniques is their low melting point. Consequently, these growth procedures may not only be applied to appropriate inorganic substances but also to the interesting group of organic nonlinear optical materials such as urea, nitroaniline, and their derivatives..
(a) A Structure B Ellipsoidal mirror furnace with ring focus on the sample side C Sample conveyance drive D Thermal control and vacuum/gas system E Lamp storage and exchange mechanism F Sample storage and exchange mechanism G Electric unit incl. data processing system and power conditioning G
(b)
Lamp
Sample
(c)
Figure 12-23. The automatic mirror furnace (AMF), ESA, is the first unit dedicated for long-duration missions on free-flying carriers with fully automated operation equipped with automatic sample and lamp exchange. The total configuration with dummy samples is depicted in (a). Section drawings (b), (c) represent the main parts of the facility. The AMF provides the opportunity for interface demarcation by lamp pulsing. About 20 crystal-growth experiments can be perfomed during a six-month space mission (courtesy of Dornier Deutsche Aerospace).
12.4 Solidification-Front Dynamics
565
Biological materials which could potentially benefit from crystallization in microgravity, namely, proteins, DNA, proteinDNA complexes, and organelles, are gaining increasing interest among scientists and industry. In a reduced-gravity environment the following improvements are expected: larger, better crystals, useful for crystallographic structure analysis. The most-favored techniques of protein and macromolecular crystallization are vapor growth, dialysis, and free interface diffusion. Early promising results obtained by Littke and John (1986) on rocket (TEXUS 5) and spacelab (SL-1) flights stimulated the development of the protein crystallization facility (PCF) for EURECA. Figure 12-28 shows the PCF assembly. The freeinterface method chosen allows the formation of free interfaces between protein, buffer, and salt solution, causing the salt
Temperature Profile within the Specimen As-Reservoir s Heater
(b) Figure 12-24. The parabolic-ellipsoid mirror furnace (ELLI), Germany, is an advanced version of the single-ellipsoid mirror furnace, flown on SL-D1. The mirror geometry, shown here, has been optimized for zone-melting experiments, particularly for GaAs growth. The ellipsoidal and paraboloidal segments provide rotational and axial symmetry of heat radiation at the sample. Two video cameras allow in-flight control of the experimental and telescience procedures (courtesy of Dornier Deutsche Aerospace).
566
12 Solidification Processing under Microgravity
molecules to diffuse through the buffer into the protein solution, and thus initiating the crystallization process. 12.4.2.3 Crystal Growth: State of the Art
Growth from the Melt Crystal growth from the melt is significantly affected on earth by gravity, which is responsible for buoyancy-driven convection in the melt and for the existence of hydrostatic pressure. In the study of single-crystal growth under jig, InSb, Si, and Ge have been used as testing materials. The experiments in space
show a decrease in the dislocation density by nearly two orders of magnitude and an axial dopant distribution corresponding to the diffusion-controlled case without convection. In crystals grown from confined melts in Bridgman geometry, where Marangoni convection and rotation can be excluded, dopant striations disappear completely. Walter (1983) was the first to perform a melt-growth experiment on a sounding rocket. The directional solidification was carried out on a seeded Ga-doped Ge melt providing six minutes of microgravity. The etched sample, Fig. 12-29, clearly delineates the transition between the jig and several g phases. Melt growth by the floating-zone method has proved useful for elucidating the importance of surface-tension-driven Marangoni convection for solute segregation in semiconductors and for containerless processing in general. During the spacelab mission Dl, Kolker (1987) solidified a drop of Si in the mirror furnace (Fig. 12-30). The striation pattern observed on the crystallized drop is similar to that seen in terrestrial samples. They are ascribed
(a) FURNACE EXPLODED VIEW SUPER-INSULATION HEATER 1 HEATER 2 HEATER 3
GUIDE-BUSH GUIDE-PLATE RING
Figure 12-25. The gradient heating facility (GHF), France, consists of three independently controlled heaters allowing directional processing while sustaining a defined temperature gradient at the solid-liquid interface. Three such heating systems are integrated in the GHF, allowing three samples to be simultaneously processed (ESA, 1989).
12.4 Solidification-Front Dynamics
567
Operating Panel Data Electronics
Control Electronics Module
Process Chamber Core Facility Module
Turbomolecular Vacuum Pump
Gas Storage Module
Vacuum Ballast
Sample Port
Gas Bottles
Gas Shut Off
(a)
Super Insulation Heat Diffuser Furnace Housing
Temperature Control Thermocouples
Cartridge Liquid Metal Ring \ Process Chamber
Heater Windings
(b) Figure 12-26. The configuration of the advanced gradient heating facility (AGHF), ESA, showing (a) the external appearance and (b) a schematic drawing of the furnace as intended for semiconductor crystal growth (large sample diameter, maximum thermal stability, and pulse marking) (courtesy of Dornier Deutsche Aerospace).
568
12 Solidification Processing under Microgravity
D HEAT DISTRIBUTION RING REFLECTOR
ATMOSPHERE AT NORMAL PRESSURE
to Marangoni convection (see also Fig. 12-22b). Growth from Solutions For semiconductors of III-V and II-VI compounds and for organic materials, solution growth is the favored technique to obtain single crystals. Again, the microgravity environment enables one to virtually eliminate gravity-driven convection. In the high-temperature regime, no fluxgrowth experiments have been performed so far in microgravity although many electronic materials such as Si, GaAs, GaP, SiC, GaSb, HgTe, ZnTe, CdTe, CuGaS 2 , InP, InSb, or PbTe are being grown from solutions (Walter, 1987). Benz et al. (1987) grew InP single crystals during the Dl spacelab mission, utilizing the travelling-heater method in the mirror furnace (ELLI). The main results: - An increase in the growth rate was observed. The length of the space-grown crystal was about five times larger than the 1 g reference sample; the material transport in the solution zone may have been hampered by buoyancy effects in 1 g due to the low specific weigth of the phosphorus.
r
SOURCE MATERIAL
Figure 12-27. Apparatus for vapor growth of a-HgI2 used in the SL-3 mission by van den Berg et al. (1986): (a) an overall view; (b) a detail of the container with source material on the wall.
- The InP crystal was nearly striation free; Fig. 12-31 shows a section of the seedcrystal interface. These results demonstrate that dopant striations in space-grown InP crystals, if present at all, are less intense and of smaller density than in earth-grown crystals. Solution growth in space, therefore, offers the possibility of comparing kineticsrelated phenomena with effects that are influenced by the growth solution. In the low-temperature regime, inorganic material with high and low solubilities have also been grown in space. Typical examples were CaCO 3 , PbS, and TGS (triglycine sulfate). A special case in this context is the growth of protein single crystals under microgravity. Littke and John (1986) crystallized lysozyme and /?-galactosidase during the FSLP-mission. The single crystals produced were significantly larger than those grown in a reference experiment on earth. The yield in terms of volume was 27-fold for jS-galactosidase and 1000-fold for lysozyme compared with earth-grown crystals. The increased size of the crystals allowed for the first time the determination of crys-
12.4 Solidification-Front Dynamics
569
Experiment Module consisting of • 12 Freezers loaded with 12 Process Reactors • Optical & Video System
Service Module consisting of • Dedicated Computer • Data handling System • Data Management System • Controls
Cooling Systems • Primary Cooling System • Secondary Cooling System • Power Supply (Part of Service Module)
Silicone Diaphragm Protein Chamber
Buffer Chamber
Salt Chamber
Silicone Diaphragm
Figure 12-28. The protein crystallization facility (PCF) for EURECA, ESA, allows the individual processing of 12 samples, each contained in a dedicated process chamber. The process reactors are designed as transportable and removable inserts of the freezer, (a) A video system allows monitoring of the growth process, (b) The process reactor consists of a stack of three coaxial quartz glass slices, each of them provided with caAdties for accommodation of protein, buffer, and salt solution. The reaction is to be started by rotating the center slice to bring the protein and salt chambers into contact (Schoen and Seifert, 1987).
Figure 12-29. Longitudinal section of a Ga-doped Ge crystal obtained from crucible-confined solidification on a sounding rocket (TEXUS project): on the right, striation-free material grown under ug (evenly spaced lines are time markers); on the left, striated material partly grown under acceleration during the despin phase (Walter, 1983).
570
12 Solidification Processing under Microgravity
Fig. 12-28). The potential of protein crystal growth is considered to be extremely interesting, both scientifically and possibly also from the standpoint of practical applicability. Growth from the Vapor Phase
M9
Figure 12-30. Striation pattern of jig-processed Si is due to time-dependent Marangoni flow in unconfined melts; the top shows the earth-grown Si bar by floating zone; at the bottom the drop, which was solidified during the Dl mission, is shown with a similar striation pattern (Kolker, 1987).
Figure 12-31. Etched cross-section of the space-grown InP crystal: both seed and grown crystal are shown (Benz et al, 1987).
tallographic data such as unit cell, symmetry, space group, number of asymmetric units, approximate molecule size, etc., with a precision not previously attained. The apparatus is to be reflown on D2 and IML1. Meanwhile, on the basis of experience with this apparatus, the advanced and automated PCF has been built, for more extensive missions such as EURECA (see
Vapor growth of bulk single crystals of electronic materials is a method with great potential, which has not yet been adequately exploited. Its most important advantages are as follows: - A great number of important electronic materials cannot be grown by other techniques since they decompose prior to melting or react with crucibles or solutions. - Vapor-grown crystals can have considerably lower defect densities than crystals grown with other (high-temperature) methods. This has become particularly important for the ever-increasing production of highly perfect substrates for thin film applications. - Structural improvements cause an increased mobility of both holes and electrons, a higher lifetime for holes, and greater electrical uniformity compared to earth-grown crystals. On earth, large crystals (up to 10 cm) have so far only been prepared of a-HgI 2 . Details of a-HgI2 grown in space on SL-3 have been reported by van den Berg et al. (1986): High growth rates never observed before have been reported. 12.4.3 Solid-liquid Interface Interaction with Particulate Matter, Including the Consideration of Alloys in the Liquid State: Engulfment of Dispersoids 12.4.3.1 Theoretical Background
Interactions of the solid-liquid interface with discrete particles, be they gaseous, liquid, or solid, are of great significance for
571
12.4 Solidification-Front Dynamics
the entire field of solidification, but in particular for metal-matrix ceramic-dispersion composites. An essential question in this context, relevant to jig, is whether the particles will be engulfed by the crystallization front or not, and what the critical parameters are. In one of the first treatments of the subject, Uhlmann et al. (1964) introduced a critical growth rate, vc, which yields a quantifiable entity for deciding on engulfment, and thereby described corresponding mechanisms. As an introduction to the engulfment problem, the physical background is sketched in Fig. 12-32. Potschke and Rogge (1989) proposed the following simple formulation for vc: vc=C Aao/rj Rojnn
Growth Direction _ Solid
Liquid
Fluid Flow
(a)
v///////////////////////// Substrate
(12-9) CT
SP
CD
c en
a
SL
+
CT
LP
Surface
where C is a constant, and Aao = = a p,s — (G'P.I + G'S.I) W ^h ° a s the interface energy difference between particle p, liquid 1, and solid s; r\ is the dynamic melt viscosity, and Ro the particle radius; JLI = AP/AU the ratio of particle-to-liquid heat conductivities; and n is an exponent. Speeds above vc would engulf the growth front, and those below vc would push particles away from the growth front.
0
Separation d
•
(b)
Coppermelt
12.4.3.2 Procedures, Methods, and Apparatus
Experimental approaches investigating the engulfment of dispersoids for composites and immiscible alloys are principally faced with two aspects. One concerns undesired agglomeration mechanisms of the dispersoid in the melt, demanding an isothermal mode of operation; thermal gradients would include particle movements and thus present an increased danger for coalescence. The other concerns particle interactions with the advancing solidification front, requiring well-defined gradient modes of operation.
Figure 12-32. Excluding buoyancy and sedimentation by applying a microgravity environment, the behavior of particles at a solid-liquid interface (a) is determined by surface energies, flow forces, and heat flow conditions. The surface energy relation is a function of distance (b) and in particular of the wetting behavior between melt, particle, and possible gaseous phase (c).
572
12 Solidification Processing under Microgravity
A heating facility serving both the isothermal and the gradient condition is represented by the isothermal heating facility (IHF) (Fig. 12-33), which has so far been flown twice and has been rebuilt for its third deployment on the spacelab mission D2. The two-chamber furnace allows simultaneous heating and cooling cycles to be performed. Another isothermal facility is the continuous heating furnace (CHF) assigned for the Japanese spacelab mission in 1991, SL-J (Fig. 12-34). Another experimental setup related primarily to the gradient mode is the heating facility for turbine blades (HFT), Fig. 12-35. Intentions here are twofold, i.e., to process both improved material and complex part geometry. Thus a single crystal strengthened by oxide dispersion (improved material) is to be processed within a sample coated with a thin skin and shaped in the form of a turbine blade (complex part). The thin skin, replacing the conventional thick-shell mold ("skin technology"), is intended to stabilize the shape in the microgravity environment. Otherwise, the molten metal would assume an ideally spherical appearance. 12.4.3.3 State of the Art Interaction of Particles with a Solidification Front
SAMPLE FLANGE (b)
POSITION
1
2
Figure 12-33. The improved IHF configuration, Germany, provides within its housing resistance heaters (a) in which samples can be processed under a wide range of conditions. Besides isothermal treatment, directional melting and solidification can also be accomplished using a heat sink attached to the sample cartridge. Rapid cooling is achieved by transfer of the samples from the heater into a cooler cavity (b). This operation, which also greatly increases the number of samples that can be processed within a given time, is accomplished by rotation and translation of the furnace and the cooler cavity, while the sample remains stationary to avoid disturbances due to gravity (ESA, 1989).
The behavior of particles at an advancing solidification front without gravity-induced sedimentation has initially been studied in a rocket experiment (SPAR), where a transparent cylindrical tube filled with naphthalene and dispersions of glass beads and zinc oxide particles was used. The particles were rejected by the solidification front and large pileups were observed. Displacement of dispersoids could also be demonstrated with very small
573
12.4 Solidification-Front Dynamics
ROTATION MOTOR
ROTATION PLATE
PROTECTION COVERS
LINEAR TRANSFER MOTOR
COOLING CHAMBER
VACUUM CHAMBER HEATING CHAMBER SAMPLE CARTRIDGE
BACKWARD
ROTATION
FOREWARD
Figure 12-34. The continuous heating furnace (CHF), Japan, consists of two vacuum-heating chambers and two He-gas-cooling chambers to simultaneously process four sample cartridges under programmed temperature profiles. Two samples are thus heated while the other two are cooled (NASDA, 1986).
(0.1 |Lim) Y2O3 particles in a Ni-base alloy by Sprenger (1987). To test the theory on particle interactions with an advancing solidification front, Potschke and Rogge (1989) studied specimens of copper matrices with alumina dispersions (< 20 |im diameter). They were melted and solidified directionally in several TEXUS flights and also during the Dl spacelab mission. In accordance with the theory, the particles were displaced by the growing copper crystal at growth rates of v<\ mm/min (Fig. 12-36).
Composites Engulfment of particles in a moving solidification front is of technical interest in manufacturing composite materials, in which ceramic or metallic particles are combined with a metallic matrix into a macroscopically homogeneous microstructure. On earth, sedimentation and agglomeration prevent finely dispersed particle distributions. During the Dl mission, other composites with a copper matrix were also pro-
574
12 Solidification Processing under Microgravity
Figure 12-35. The heating facility for turbine blades (HFT), Germany, provides two processing chambers equipped with heating and cooling zones. Only one chamber can be processed while the second one serves as spare furnace in case of malfunction (courtesy of ERNO, Bremen, FRG).
cessed in the IHF. The main result was that the flight samples had a more homogeneous particle distribution, in the case of both coarse and fine particles. A typical example is illustrated in Fig. 12-37 (Froyen and Deruyttere, 1986), which shows the distribution of 0.1-0.5 jim diameter A12O3 particles in Cu after processing on earth and in space. An example for producing composite materials with high-temperature resistance
and high mechanical strength is found in the OSIRIS project (Amende, 1991). A Ni-base superalloy single-crystal turbine blade strengthened by oxide dispersion is planned to be solidified in D2, in an attempt to pool microgravity-relevant basic scientific work with technological endeavors (skin technology) and new furnace hardware (compare Fig. 12-25).
12.4 Solidification-Front Dynamics
Alloys Immiscible in the Liquid State Numerous mechanisms govern microstructure formation in the solidification of alloys that are immiscible in the liquid state (Langbein, 1980). For example, Ahlborn and Lohberg (1986) performed spacelab experiments (on FSLP and Dl) utilizing several Z n - P b and Zn-Bi alloys. They observed that Zn-rich droplets concentrated at the hotter side of the crucible. Similar behavior had been reported for Zn-Bi by Frederikson (1984). Space solidification of Al alloys containing Pb or In (Walter, 1984) did not show a statistical distribution of the dispersions as expected for no-gravity conditions. Conglomeration
575
tendencies are explained by droplet transport due to Marangoni convection. This flow develops where temperature gradients are present and interfacial energies between the two immiscible liquids also establish gradients. The droplets then migrate to the high-temperature regions.
(a)
A 1.2 Vs (mm/min) 0, i - _ _ _ (b)
«••-•
'•"«/•
•#
••
0,5 mm
Figure 12-36. Longitudinal section of a copper sample with A12O3 particles processed during the Dl spacelab mission. The molten sample section A shows particles that were pushed away by the solidification front (arrow) upon resolidification (Potschke and Rogge, 1987).
Figure 12-37. The distribution of A12O3 particles in a Cu matrix after processing on earth, (a), and in space, (b), indicating a homogenizing effect in the latter (Froyen and Deruyttere, 1986).
576
12 Solidification Processing under Microgravity
12.44 Eutectics: In situ Composites 12.4.4.1 Theoretical Background
In certain cases eutectic solidification is affected by gravitationally induced transport processes. The lamellar spacing Xe and the growth rate v are linked by the following well-known relationship: X2ev=C (D, m, ce, a (a, /J,m))
(12-10)
The constant C was derived from first principles by Jackson and Hunt (1966). Among others, it is a function of the melt diffusion coefficient D (very probably an effective D value in the ground laboratory), the liquidus slope m at ce (the eutectic composition), and the three interfacial energies, a, between the phases a, /?, and the melt m. 12.4.4.2 Procedures, Methods, and Apparatus
Eutectic solidification has been and can be studied effectively in any one of the fur-
naces displayed earlier, preferably mostly gradient heating facilities.
12.4.4.3 State of the Art
Several eutectic alloys were directionally solidified utilizing differing growth velocities at 1 g and jig. To study the influence of convection on solidification morphology, interfiber spacings were compared between 1 g and jig experiments. As of yet no simple correlations have been possible. Larger rod separations of a space-grown Al-Al 3 Ni eutectic were detected by Favier and de Goer (1984), in samples solidified in rocket flights (TEXUS VI) and spacelab (FSLP) (see Fig. 12-38). The same behavior was observed by Barbieri and Patuelli (1988) in eutectic Al-Cu samples (Fig. 12-39). Sprenger (1987) solidified an eutectic Ni/Al-Mo alloy. He also found a coarser eutectic microstructure (here eutectic cells) which was
Figure 12-38. Cross sections of 1 g- and |igdirectionally solidified Al 3 Ni-Al eutectic fibers (Favier and de Goer, 1984).
ig
Figure 12-39. SEM image of a rod-like eutectic structure in Ag-Cu specimens after melting and solidification in |uig, (a), and in 1 g, (b) (Barbieri and Patuelli, 1988).
577
12.4 Solidification-Front Dynamics
larger in length and diameter than on earth. Investigations with CuAl 2 -Al (TEXUS IV and VI and spacelab: FSLP) were reported by Favier and de Goer (1984). The interlamellar distance of the eutectic was not affected by the g level. Investigations with InSb-NiSb (spacelab: FSLP) yielded quite reproduceably lower mean interfiber distances for growth rates ranging from 0.15 to 0.5 mm/min (Fig. 12-40). These results were then confirmed by a sounding rocket experiment (TEXUS X) in which pulse marking was used to precisely determine the microscopic solidification rate (here 0.9 mm/ min). A second confirmation and extension of the data was obtained again in another spacelab Dl experiment (see Miiller and Kyr, 1987). A summary of all measurements is shown in Fig. 12-40. Smaller fiber separations in space had also been observed in MnBi-Bi eutectics during earlier American flights (orbital mission ASTP and rocket missions SPAR I, VI, and IX; Pirich et al, 1980). Tensi et al. (1989) observed a similar behavior in the Al-Si eutectic system. At low crystallization velocities (5 mm/min), the jig sample yields smaller interfiber distances than the 1 g sample; at higher velocities, the microstructure appears to become independent of the g level (Fig. 12-41). In summary, three different apparently contradictory observations have been gathered in eutectic jig solidification. The following explanations have been offered: 1. A perturbation of the diffusive concentration profile, caused by convection at the eutectic solidification front, is assumed for earthbound conditions. Eisa et al. (1986) modified the diffusive model of Jackson and Hunt (1966) and studied the influence of thermal convection, and possibly thermosolutally driven convection, on eutectic
15 x10 4
r-.
10"-
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Rate of solidification v [mm/min]
Figure 12-40. Density N of fibers versus growth rate v for microgravity (filled symbols) against 1 g reference experiments (open symbols) with bottom seeding (Miiller and Kyr, 1987).
Figure 12-41. Comparison of Al-Si-eutectic microstructure solidified under jag and 1 g conditions for different growth rates at a constant temperature gradient (G~15 K/mm), carried out on a spacelab-Dl experiment using the GFQ (Tensi et al., 1989).
578
12 Solidification Processing under Microgravity
solidification. They decided that the absence of convection in space is not sufficient to understand the experimental results. However, Favier and de Goer (1984) extended this idea by allowing slightly offeutetic solidification. This assumption makes for a drastic increase in the thickness of the diffusive concentration layer, leading to considerably stronger convection currents. According to their theory, for fibrous systems which are characterized by different liquidus slopes on both sides of the eutectic, the fiber separation should increase for hypo-eutectic compositions (e.g., the Al side of Al-Al 3 Ni), whereas in hypereutectic alloys it should decrease when buoyancy-driven convection is eliminated. On the strength of this model, the experimentally measured fiber spacings in InSbNiSb eutectic solidified under jig and 1 g were compared with calculations of Kyr and Miiller (1987), and agree quite well. In lamellar eutectics where volume percentages of the two phases as well as the liquidus slopes are approximately equal the lamellar distance should be independent of convection. Indeed, the lamellar distance in the Al 2 Cu-Al eutectic alloy is unaffected by the g level, as reported by Favier and de Goer (1984). 2. Another model hypothesizes that D should be replaced by an effective diffusion coefficient Deff (g) which is larger at 1 g and smaller under jig conditions (Sahm, 1977). The lateral mass transport in the liquid is caused by diffusion and convection. The mass transport by convection is subdivided into a term for gravity-driven convection and a term for convection, caused by the volume jump during the phase transition from liquid to solid (Sahm and Rittich, 1983). The strength of this microconvection depends on the difference of the
volume jumps of the two eutectic phases. The lower spacings 2e of Al-Si, InSbNiSb, and MnBi-Bi eutectics solidified in space could possibly be explained by the less turbulent microconvection under jig, which is overlaid by thermosolutally driven convection on earth. 12.4.5 Ripening Phenomena 12.4.5.1 Theoretical Background
Ripening is an important phenomenon with respect both to the coarsening mechanisms of coagulation in liquid-immiscible systems and to the development of interdendritic-arm spacing during alloy solidification. The evolution of dendrite-arm spacings or the coarsening of particle dispersions in a liquid matrix may be termed an annealing process of a system which is not in thermodynamic equilibrium. Ostwald ripening is responsible for the disappearance of small dendrite arms or fine dispersions by reducing the total surface energy. The time dependance of ripening phenomena has been described by Lifshitz and Slyzov (1961) and Wagner (1961) (the LSW theory). This theory predicts coarsening according to d2-dl=a-t
(12-11)
where d and d0 are average spacings at times t and 10 and a is a system-dependent constant. Ostwald-ripening phenomena are convection dependent and thus affected by microgravity, in that convection plays an important role in determining coarsening kinetics, A convectionless environment, such as in jig, should ideally support the LSW theory. The opposite condition of very strong convection has shown the exponents to be shifted to d7/3 (Wang and Sahm, 1989). A normal 1 g environment at a solidifying, typically dendritic
579
12.4 Solidification-Front Dynamics
solidification front should thus be found to lie between jug and strongly convected conditions. 12.4.5.2 Procedures, Methods, and Apparatus
Facility families serving investigations on ripening phenomena are isothermal and gradient heating furnaces (see above: IHF, GFQ). Studies in transparent media are also relevant. Here, the holographic laboratory HOLOP is also of interest. i l l ,
T—r->
i i i i i - — [jg experiment S GSF = 15.25 y ' y y < " - vSF = 8.06 .XX
- pg experiment GSF = 15.93
/
/'*
"VSF=5.22
£T\ -
I
I
I
I
1 g ref. experiment G SF = 15.80
V
" SF = 5.26
I yS
%S
g f
s — ,
-
\
i
I
i
1 g ref. experiment GSF = 15.34 V
s ' " ' SF =
789
i
i , s^/^/ X "
l s^S '
J &
'
; & ' S'\ 0
I 1
I
I 2
I
I 3
0
1
2
3
Coarsing time t ^ in s1/3
Figure 12-42. Comparison of dendrite-arm coarsening under ug and 1 g conditions for different growth velocities (Tensi et al., 1989).
12.4.5.3 State of the Art
So far, only two successful space experiments have been performed in the study of ripening phenomena. Tensi et al. (1989) used the GFQ during the spacelab mission Dl to measure coarsening of Al-Si dendrites under jag conditions. Comparison of the time-dependant spacing of the dendrite arms at two solidification velocities show a slower coarsening in space samples (Fig. 12-42). This behavior was explained by Tensi et al. (1989) in terms of the gravity dependance of a in Eq. (12-11), which is lower in the case of microgravity, where convection is largely suppressed. Another space experiment was performed utilizing IHF during the spacelab 1 mission (FSLP) (see Kneissl and Fischmeister, 1984). They studied coarsening of immiscible Z n - P b alloys at different concentrations. Figure 12-43 presents particle size distributions for two samples. At Pb concentrations lower than 4wt.% the experimental results agree well with the theoretical predictions. But at higher Pb concentrations, collision of droplets with subsequent coagulation results in larger particles than those predicted by theory.
Zn + 4% Pb-flight
f\
2.0-
1.5-
i
1.0-
k
0.5
06
12
18
Circle diameter [pm]
24
30
0
^ 9
1 18
. 27
36
Circle diameter [pm]
Figure 12-43. Normalized circle diameter distribution of two Zn-Pb alloys after flight. The drawn-in curve is the predicted LSW distribution for pure diffusion without convection (Ratke et al., 1987).
580
12 Solidification Processing under Microgravity
12.5 Acknowledgements For providing a literature survey and critical discussions, we are thankful to Dr. J. Laakmann and Dr. G. Zimmermann of the Aachen Center for Solidification in Space, ACCESS.
12.6 Bibliography of Space Results Proceedings of the 5th European Symposium on Materials Science under Microgravity Conditions. Schlofi Elmau, Germany, 1984. ESTEC Noordwijk: ESA Publication Division. Sahm, P. R., Jansen, R., Keller, M. H. (Eds.) (1986), Proceedings of the Norderney Symposium on Scientific Results of the German Space lab Mission Dl, Norderney, Germany. Koln: DLR. Materials Processing in the Reduced Gravity Environment of Space. Symposia Proceedings, Vol. 87 (1986). Boston: Materials Research Society. Proceedings of the 6th European Symposium on Materials Science under Microgravity Conditions, Bordeaux, France, 1986. ESTEC Noordwijk: ESA Publication Division. Kaldeich, B. H. (Ed) (1989), Proceedings of the 7th European Symposium on Materials and Fluid Sciences in Microgravity. Oxford, England. ESTEC Noordwijk: ESA Publication Division. Final Reports ofTEXUS 1-10 (1990), 11-12 (1988), 13-16 (1989) (all in German language). Bonn: DARA (German Space Agency). Summary Review of ESA-Experiments on TEXUS 13-20 (1989). Paris: ESA. Feuerbacher, B., Hamacher, H., Naumann, R. J. (Eds.) (1986), Materials Science in Space. A Contribution to the Scientific Basis of Space Processing. Berlin: Springer Verlag. Walter, H. U. (Ed.) (1987), Fluid Sciences and Materials Science in Space. Berlin: Springer Verlag.
12.7 References Ahlborn, H., Lohberg, K. (1986), Naturwissenschaften 73, 378. Amende, W. (1991), in: Research Program of the German Spacelab Mission D2: Keller, M. H., Sahm, P. R. (Eds.). In press Barbieri, F., Patuelli, C. (1988), in: Second Int. Symposium on Experimental Methods for Microgravity Materials Science Research: Schiffman, R.A. (Ed.). Warrendale, Pennsylvania: The Mineral, Metals and Materials Society, p. 87. Bardenheuer, P., Blechmann, R. (1939), Mitteilung KWIfur Eisenforschung 21, 201.
Benz, K.W., Danilewsky, A., Notheisen, B., Nagel, G. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 275. van den Berg, L., Schnepple, W. F , Schieber, M. M. (1986), Poster no 471, ICCG-8, 13-18 July, York, England. Billia, B., Jamgotchian, H., Favier, J. I, Camel, D. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 230. Camel, D., Favier, J. J., Dupony, M. D., LeMaguet, R. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 236. Coriell, S. R., Sekerka, R. F. (1981), Effect of convective flow on morphological stability, PCH, 2, 281. Coriell, S. R. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 222. Croll, A., Miiller-Sebert, W, Nitsche, R. (1990), in: Proc. of the Vllth European Symposium on Materials and Fluid Sciences in Microgravity, Oxford, 1989. ESA SP-295, pp. 263-270. Dornier Broschiire (1990), Microgravity: Experiment Facilities Instrumentation: Dornier/DASA, Friedrichshafen. Drehman, A. X, Turnbull, D. (1982), in: Materials Processing in the Reduced Gravity Environment of Space: Rindone, G. R. (Ed.), p. 81. Ecker, A., Schmitz, G. I , Sahm, P. R. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 212. Eisa, G. F., Wilcox, W R., Busch, G. (1986), J. Crystal Growth 78, 159. ESA (1983), 4th European Symposium on Material Sciences in Space, Madrid: ESA/ESTEC, Nordwijk (Ed.), Paris, SP-191. ESA (1987), 6th European Symposium on Material Sciences under Microgravity Conditions, Bordeaux: ESA/ESTEC, Nordwijk (Ed.). Paris, SP-256. ESA (1989), Experiment Facilities for Material and Fluid Sciences embarked on Spacelab: ESA, Paris, SP-1120. ESA-Report (1989), Microgravity Science and Applications Program Tasks: NASA Technical Memorandum 4097. Favier, J. J., de Goer, J. (1984), in: 5th European Symposium on Material Sciences under Microgravity, Schlofi Elmau, 1984. ESA SP-222, p. 127. Frederiksson, H. (1984), in: Proc. Workshop on The Effect of Gravity on the Solidification of Immiscible Alloys, Stockholm, 1984. ESA, SP-219, p. 25.
12.7 References
Froyen, L., Deruyttere, A. (1986), Naturwissenschaften 73, 384. Glicksman, M. E. (1981), in: Erstarrungsfrontdynamik - Workshop Proceedings, Giefierei-Institut RWTH Aachen, March 12/13, 1981: Sahm, P. R. (Ed.). Aachen: RWTH, p. 79. Hug, W, Kallien, L., Sahm, P. R. (1986), Giefiereiforschung 38 (2), 73. Hug, W, Sahm, P. R. (1988), Giefiereiforschung 40, 56. Jackson, K. A., Hunt, J. D. (1966), Transact. MetalL Soc. AIME 236, 1129. Kallien, L., Sahm, P. R. (1986), Conference Proceedings, Science and Technology of the Undercooled Melt: Sahm, P. R., Jones, H., Adam, C. M. (Eds.). Dordrecht: Martinus Nijhoff Publishers, p. 243. Kiminami, C. S., Sahm, P. R. (1986), Acta MetalL 34, 2129. Kiminami, C. S., Axmann, W, Sahm, P. R. (1989), /. Mat. Sa. Letts. 8, 201. Kneissel, A., Fischmeister, H. (1984), in: 5th European Symposium on Material Sciences under Microgravity, Schlofi Elmau, 1984. ESA SP-222, p. 63. Kolker, H. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 264. Kurz, W., Fisher, D. J. (1984), Fundamentals of Solidification. Aedermannsdorf, CH: Trans. Tech. Publishers. Kyr, P., Miiller, G. (1987), Abschlufibericht, Inst. f Werkstoffwissenschaften 6, Kristallabor, Univ. Erlangen-Niirnberg. Lacy, L. L., Robinson, M. B., Rathz, T. J. (1981), /. Cryst. Growth 51, 47. Lacy, L. L., Rathz, T. I , Robinson, M. B. (1982), J. Appl. Phys. 53 (1), 682. Langbein, D. 1980), Theoretische Untersuchungen zur Entmischung nicht-mischbarer Legierungen, Final Report, BF-R-64.152, Battelle, Frankfurt. Lierke, E. G., GroBbach, R. (1975), Study on positioning of molten material in zero gravity environment by ultrasonic methods - Phase II. ESA contract No. SC/67/HG. Lierke, E. G. (1976), Hardware development and performance of an acoustical positioning device. ESA Special publication No. 114. Lierke, E. G., GroBbach, R., Flogel, K., Clancy, P. (1983), in: Proceedings of the Ultrasonics Symposium, Vol. 2: McAvoy, B. R. (Ed.), p. 1129. Lifschitz, I. M., Slyozov, V. V. (1961), J. Phys. Chem. Solids 19, 35. Littke, W, John, Chr. (1986), J. Crystal Growth 76, 663. Malmejac, Y, Walter, H. U. (1980), Materials Science Research with Sounding Rockets. ESA, Paris, MAT (79) 5 revl. Mogi, K., Ogino, K., McLean, A., Miller, W. A. (1986), Met. Trans. 17B, 163.
581
Mordike, B. L. (1990), Laser Surface Modification. Much, D. (1923), German Patent No. 422004, Oct. 30. Miiller, G., Kyr, P. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, pp. 246-259. NASA (1980), 3rd Space Processing Symposium, Skylab Results: Malmejac, Y, Walter, H. U. (Eds.). NASA, Marshall Space Flight Center, Alabama. NASDA (1986), First Materials Processing Test: Engineering Information Manual. Tokyo: NASDASU-186070A. Okress, E C , Wroughton, D. M., Comenetz, G., Brace, P. H., Kelly, J. C. R. (1952), /. Appl. Phys. 23, 545. Oran, W. A., Berge, L. H., Berge, A. (1982), Rev. ScL, Instrum. 53, 851. Perepezko, J. H. (1980), in: Rapid Solidification Processes, Principles and Technologies II, p. 56. Perepezko, J. H. (1984), Mat. Sci. Eng. 65, 125. Piller, I , Knauf, R., Preu, P., Lohofer, G., Herlach, D. M. (1986), Proc. 6th Eur. Symp. Mat. Sci. under Microgravity Conditions, Bordeaux, p. 437. Pirich, R G , Larson, D.I, Busch, G. (1980), in: AIAA 18th Aerospace Science Meeting, Jan. 1980, paper 80-0119. Potschke, J., Rogge, V. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P.R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 304. Potschke, J., Rogge, V. (1989), J. Crystal Growth 94, 726-738. Ratke, L., Thieringer, W K., Fischmeister, H. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 332. Rex, S., Sahm, P. R. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P.R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 222. Rhim, W.K., Collender, M., Hyson, M. T, Sims, W T, Elleman, D. D. (1985), Rev. Sci. Instrum. 56 (2), 307. Rony, P. R. (1964), in: Trans. Vac. Met. Conf: Cocca, M. A. (Ed.). Boston: American Vac. Soc. Sahm, P. R. (1977), in: Haus der Technik, Vortragsveroffentlichungen 391, p. 33. Sahm, P. R. (1983), in: Nucleation - Rapid Solidification, Workshop-Proceedings, Giefierei-Institut I RWTH Aachen, March 14/15, 1983: Sahm, P. R. (Ed.). Aachen: RWTH. Sahm, P. R., Rittich, M. (1983), Adv. Space Res. 3, 103.
582
12 Solidification Processing under Microgravity
Sahm, P. R. (1986), in: Scientific Results of the German Spacelab Mission Dl: Sahm, P. R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 20. Schleip, E., Willnecker, R., Herlach, D. M., Gorier, G. P. (1988), Mat. Sci. Eng. 98, 39. Schoen, E., Seifert, E (1987), in: 38th Congress oflAF, Brighton. Paris: IAF. Shiraishi, S., Ward, R. G. (1964), Can. Met. Quarterly 3 (1), 117. Spaepen, F. (1986), Metastable States in Pulsed Laser Quenching, Ref. 17, 187. Sprenger, H. J. (1987), in: Sixth European Symposium on Material Sciences under Microgravity Conditions, Bordeaux, 1986. ESA SP-256, p. 349. Steinborn, W. (1986), Furnaces, Materials Sciences in Space: Feuerbacher, B., Hamacher, H., Naumann, R. J. (Eds.). Berlin: Springer-Verlag, pp. 227-265. Tensi, H. M., Schmidt, J. J. (1987), in: Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission Dl, Norderney, 1986: Sahm, P.R., Jansen, R., Keller, M. H. (Eds.). Koln: WPF, p. 216. Tensi, H. M., Schmidt, J.J., Mackrodt, C. (1989), Mat. Science Forum 50, 45. Tiller, WA. (1971), in: Solidification, ASM, Metals Park, Ohio. Turnbull, D., Cech, R. E. (1950), /. Appl. Phys. 21, 804. Uhlmann, D. R., Chalmers, B., Jackson, K. A. (1964), J. Appl. Phys. 35, 2986. Vinet, B., Cortella, L., Favier, J. I, Desre, P. (1991), Appl. Phys. Lett. 58, 97.
Wagner, C. (1961), Z. Electrochemie 65, 581. Walker, J. L. (1961), in: Physical Chemistry of Process Metallurgy. New York: Interscience, p. 845. Walter, H. U. (1983), Z. Flugwiss. Weltraumforsch. 7, 372. Walter, H. U. (1984), in: Proc. Workshop on Effect of Gravity on Solidification in Immiscible Alloys, Stockholm, 1984. ESA SP-219, p. 47. Walter, H. U. (1987), Fluid Sciences and Material Sciences in Space. Berlin, New York: Springer Verlag. Wang, T. G., Saffran, M. M., Ellemann, D. D. (1974), Acoustic chamber for space processing, AIAA paper No. 74-155. Wang, T. G. (1979), Acoustic levitation and manipulation for space application, IEEE, Ultrasonic Symposium Proceedings. Whymark, R. R. (1975), Ultrasonic 13, 251. Willnecker, R., Herlach, D. M., Feuerbacher, B. (1986), Appl. Phys. Lett. 49 (20), 1339. Willnecker, R. (1988), Messungen zur Unterkuhlung, Keimbildung und schnellen Erstarrung metallischer Systeme, Ph. D. Thesis Bochum, DFVLR-FB 8839. Willnecker, R., Herlach, D. M., Feuerbacher, B. (1988), Mat. Sci. Eng. 98, 85. Zarzycki, I , Frischat, G. H., Herlach, D. M., Glasses (1988), in: Fluid Sciences and Materials Science in Space: Walter, H. U. (Ed.). Berlin: Springer-Verlag, p. 599-636. ZARM brochure (1989).
13 Cluster Assembly of Nanophase Materials Richard W. Siegel Materials Science Division, Argonne National Laboratory, Argonne, IL, U.S.A.
List of Symbols and Abbreviations 13.1 Introduction 13.1.1 Background 13.1.2 Advantages of Cluster Assembly 13.2 Synthesis and Processing 13.2.1 Basic Principles 13.2.2 Conventional Gas-Condensation Method 13.2.3 Improvements in the Gas-Condensation Method 13.2.4 Other Physical Methods 13.2.4.1 Spark Erosion 13.2.4.2 Mechanical Attrition 13.3 Characterization 13.3.1 Structure 13.3.2 Stability 13.4 Properties 13.4.1 Sintering and Diffusion 13.4.2 Electrical Properties 13.4.3 Mechanical Behavior 13.4.3.1 Ceramics 13.4.3.2 Metals 13.5 Future Directions 13.6 Acknowledgements 13.7 References
Materials Science and Technology Copyright © WILEY-VCH Verlag GmbH & Co KGaA. Allrightsreserved.
584 585 586 587 587 587 588 595 596 596 597 598 598 601 602 602 606 606 606 609 611 612 612
584
13 Cluster Assembly of Nanophase Materials
List of Symbols and Abbreviations d Db D Ti Hy 12 kB m N (r) q S(q) t Tm x
grain diameter grain-boundary diffusivity diffusivity of titanium Vickers microhardness intensity of positron annihilation lifetime signal T 2 Boltzmann constant strain-rate sensitivity size distribution of scattering centers of radius r momentum transfer absolute scattered intensity annealing time melting temperature in K median particle diameter
e 0 a ay T2
strain rate Bragg angle geometric standard deviation yield stress positron lifetime in voids
BET EXAFS FWHM HREM LNDF MECS PAS SANS UHV
Brunauer-Emmett-Teller technique extended X-ray absorption fine structure full width at half maximum high-resolution electron microscopy log-normal distribution function multiple expansion cluster source positron annihilation spectroscopy small-angle neutron scattering ultra-high vacuum
12.4 Solidification-Front Dynamics
565
Biological materials which could potentially benefit from crystallization in microgravity, namely, proteins, DNA, proteinDNA complexes, and organelles, are gaining increasing interest among scientists and industry. In a reduced-gravity environment the following improvements are expected: larger, better crystals, useful for crystallographic structure analysis. The most-favored techniques of protein and macromolecular crystallization are vapor growth, dialysis, and free interface diffusion. Early promising results obtained by Littke and John (1986) on rocket (TEXUS 5) and spacelab (SL-1) flights stimulated the development of the protein crystallization facility (PCF) for EURECA. Figure 12-28 shows the PCF assembly. The freeinterface method chosen allows the formation of free interfaces between protein, buffer, and salt solution, causing the salt
Temperature Profile within the Specimen As-Reservoir s Heater
(b) Figure 12-24. The parabolic-ellipsoid mirror furnace (ELLI), Germany, is an advanced version of the single-ellipsoid mirror furnace, flown on SL-D1. The mirror geometry, shown here, has been optimized for zone-melting experiments, particularly for GaAs growth. The ellipsoidal and paraboloidal segments provide rotational and axial symmetry of heat radiation at the sample. Two video cameras allow in-flight control of the experimental and telescience procedures (courtesy of Dornier Deutsche Aerospace).
586
13 Cluster Assembly of Nanophase Materials
der high-vacuum conditions. The structure, stability, and properties of nanophase metals and ceramics synthesized by this method are considered. This is the most generally applicable of the available methods for producing size-selected atom clusters in the less-than-100-nm (nanophase) regime, and thus appears to have the broadest technological potential. Two other physical methods for the synthesis of ultrafme-grained materials are also briefly considered: these are spark erosion and mechanical attrition, both of which (along with the gas-condensation method) are also applicable to metals and alloys. The various chemical methods for the synthesis of nanophase materials, such as sol-gel synthesis and the spray conversion method, will not be covered in this chapter as they are essentially outside the scope of this volume. Appropriate references to all of these methods can be found under General Reading at the end of this chapter.
Figure 13-1. Schematic representation of a nanophase material. The black circles represent atoms in regular lattice positions within the grains; the white circles, atoms that may be expected to relax in the grain boundaries. No chemical differences between the atoms are implied. After Birringer et al. (1986).
100
-
80 \ c 6 0 --
0.5 nm
\ \
\ \
13.1.1 Background
The modern synthesis of ultrafmegrained materials by the in situ consolidation of nanometer size gas-condensed ultrafine particles or atom clusters was first suggested by Gleiter (1981). By consolidating clusters in this manner, materials with a large fraction of their atoms in grain boundaries could be formed, as shown schematically in Fig. 13-1. The degree to which this effect varies with grain size in the nanophase regime ( < 100 nm) is shown in Fig. 13-2. The considerable body of earlier research into the production of ultrafine particles by means of the gas-condensation method (Kimoto et al., 1963; Granqvist and Buhrman, 1976; Tholen, 1979), as well as the previously assembled knowledge on powder metallurgy and ceramics, provided a solid basis upon which this sug-
-
\
-
20 -
-
-
\ \
\
10 d in nm
100
Figure 13-2. Percentage of atoms in grain boundaries of a nanophase material as a function of grain diameter, assuming that the average grain boundary thickness ranges from 0.5 to 1.0 nm (ca. 2 to 4 atomic planes wide). From Siegel (1991).
gestion could grow to fruition. The application of this idea in recent years (Birringer et al., 1984, 1986; Siegel and Hahn, 1987; Hahn et al., 1988; Birringer and Gleiter, 1988; Gleiter, 1990; Siegel, 1990 a) to the synthesis of a variety of nanophase metals and ceramics has built upon this base.
13.2 Synthesis and Processing
While the gas-condensation method for the synthesis of nanophase metals and alloys, and other materials as well, has the greatest flexibility and control for engineering new forms of bulk nanostructured materials, other physical methods, such as spark erosion and mechanical attrition, can complement this approach. The first of these methods is an alternative to gas-condensation in that it can yield individual clusters which can be subsequently assembled via consolidation. The second method, however, produces its nanostructures by means of what is essentially a mechanical decomposition of coarser-grained structures, so that the individual grains are never isolated clusters and much synthesis and processing flexibility is lost. Nevertheless, each of these methods can have their particular technological applications and all are capable of commercial scale-up. 13.1.2 Advantages of Cluster Assembly Some unique advantages of the assembly of nanophase materials from gas-condensed clusters, and the nanophase processing method based on this, are as follows: (1) The ultrafme sizes of the atom clusters and their surface cleanliness allow conventional restrictions of phase equilibria and kinetics to be overcome during material synthesis and processing, by the combination of short diffusion distances, high driving forces, and uncontaminated surfaces and interfaces. (2) The large fraction of atoms residing in the grain boundaries and interfaces of these materials allow for interface atomic arrangements to constitute significant volume fractions of material, and thus novel materials properties may result. (3) The reduced size scale and large surface-to-volume ratios of the individual
587
nanophase grains can be predetermined and can alter and enhance a variety of physical and chemical properties. (4) A wide range of materials can be produced in this manner including, in addition to metals and alloys, intermetallic compounds, ceramics, and semiconductors. It is also apparent that they can be formed to contain crystalline, quasicrystalline, or amorphous structures. (5) The possibilities for reacting, coating, and mixing in situ various types, sizes, and morphologies of clusters create a significant potential for the synthesis of a variety of new multicomponent composites with nanometer-sized microstructures and engineered properties. Most of the research carried out to date, nevertheless, has concentrated on single-phase metals and ceramics. The following section describes the method of material synthesis and processing via gas-condensation that leads to ultrafme-grained polycrystalline metals and ceramics with controlled mean grain sizes below 100 nm and a number of the attributes just cited.
13.2 Synthesis and Processing 13.2.1 Basic Principles In any of the methods for the synthesis of nanostructured materials, control of the size or sizes of the phases being assembled is paramount. Beyond this, chemical control of the phases and cleanliness of the interfaces between phases need to be carefully addressed. In bulk nanophase materials, with their large volume fraction of grain boundaries, these issues take on considerable importance. The predominant feature of nanophase materials, as depicted in Figs. 13-1 and 132 and shown in the electron micrograph of
588
13 Cluster Assembly of Nanophase Materials
Figure 13-3. High resolution transmission electron micrograph of a typical area in nanophase palladium. From Thomas et al. (1990).
nanophase palladium in Fig. 13-3, is their ultrafme grain size and, hence, the large fraction of their atoms that reside in grain boundaries or interfaces. In this regard they can be thought of as the three-dimensional analogues of one-dimensionally modulated multilayered materials. For example, as indicated in Fig. 13-2, a nanophase material with a 5 nm average grain size will have from about 27 to 49 % of its atoms associated with grain boundaries, assuming a simple grain boundary picture and an average grain boundary thickness of about 0.5 to 1.0 nm (ca. 2-4 nearestneighbor distances). This percentage falls to about 14-27 % for 10 nm grain size, but is as low as 1-3 % for a 100 nm grain size. The interface volume fraction is, of course, essentially negligible for conventional grain sizes of 1 jim and above. Therefore, the properties of nanophase materials can be expected to be strongly influenced by their grain sizes and the nature (atomic and electronic structure) of their internal boundaries, simply because of the very large number density of these boundaries.
As such, the control over the various processes involved in the synthesis of nanophase materials takes on an even greater role than usual. A high degree of control is available with the gas-condensation method. 13.2.2 Conventional Gas-Condensation Method
Research carried out on the gas-condensation method and on the resulting atom clusters during the 1960s and 1970s (Kimoto et al., 1963; Granqvist and Buhrman, 1976; Tholen, 1979) essentially defined the various parameters that control the sizes of the clusters formed in the conventional gas-condensation method (primarily type of gas, gas pressure, and evaporation rate) that are used to synthesize nanophase materials. The pioneering work of Uyeda and coworkers (Kimoto et al., 1963) demonstrated that a wide range of metallic ultrafme particles (between 10 and 100 nm in diameter) could be condensed in a low pressure Ar atmosphere and that
13.2 Synthesis and Processing
their sizes could be controlled by varying the gas pressure in the range of about 1-30 Torr (0.13-4 kPa). Subsequent work by Tholen (1979) extended these investigations to additional metals and to a study of the nucleation and coalescence of the clusters. The most detailed study to date of the conventional gas-condensation process for forming ultrafine metal particles or clusters via condensation in inert gases (He, Ar, or Xe) was carried out by Granqvist and Buhrman (1976). Some of their results will be discussed further below. However, it was these early studies that elucidated the essential parameters controlling the formation of gas-condensed atom clusters,
589
which have made possible the synthesis of cluster-assembled nanophase materials. A typical apparatus for the synthesis of nanophase materials via the in situ consolidation of gas-condensed clusters is shown schematically in Fig. 13-4. It comprises an ultrahigh-vacuum (UHV) system fitted with two resistively-heated evaporation sources, a cluster collection device (liquidnitrogen filled cold finger) and scraper assembly, and in situ compaction devices for consolidating the powders produced and collected in the chamber. Before making the powders, the UHV system is first evacuated by means of a turbomolecular pump to below 10" 5 Pa and then back-filled with
Liquid nitrogen
Main vacuum chamber
Gas inlet
Vacuum pumps Fixed piston
Bellows Low pressure compaction unit
Slide Sleeve Piston
Piston
High pressure compaction unit
Figure 13-4. Schematic drawing of a gas-condensation chamber for the synthesis of nanophase materials. Precursor material evaporated from sources A and/or B condenses in the gas and is transported via convection to the liquid-nitrogen filled cold finger. The clusters are then scraped from the cold finger, collected via the funnel, and consolidated first in the low-pressure compaction unit and then in the high-pressure compaction unit, all in vacuum. From Siegel and Eastman (1989).
590
13 Cluster Assembly of Nanophase Materials 1
'
E200
TT-J—
j
,
s
,
: : ! i j
j
i
i
j
i
i
i . |
Al: Open symbols Cu: Filled symbols 1 , 1 1 1 1 1
Qi
•
1
O
g. 20
\ D He gasj | ° Ar gas j | A Xe gasj
^ ^
c o 10 -•
I !
!
0.5
2
:
1
,
.
. . I
!
.
a controlled high-purity gas atmosphere at pressures of, typically, a few hundred Pa. For producing metal powders this is usually an inert gas, such as He, but it can alternatively be a reactive gas or gas mixVapor source temperature in K Mg 800 Zn
850
700
900 750
-r-T-q
950 800
1
1
1000 850
1 | '' ' ' |
900 1—
1000 A
500 E c c ,.200
Mg/2.5 Torr A Mg/3.5 • Zn/2.5 o Zn/3.5
§100 o 50 20 10
5 JL
0.2
.
1
5 10 20 50 Inert gas pressure in Torr
0.5 1 2 5 10 Metal vapor pressure in Torr
20
Figure 13-6. Median particle diameter versus vapor pressure at the metal surface (or source temperature) for Mg and Zn evaporated and gas-condensed in two different Ar pressures (2.5 and 3.5 Torr), From Granqvist and Buhrman (1976).
.
~ _
. i i I
100
200
Figure 13-5. Median particle diameter versus pressure of He, Ar, or Xe gas for clusters of Al and Cu formed via gas-condensation. The straight lines only serve as a guide to the reader. From Granqvist and Buhrman (1976).
ture if, for example, clusters of a ceramic compound are desired. During evaporation of the starting precursor material (or materials) from which the nanophase material will be synthesized, atoms condense in the supersaturated region close to the Joule-heated source. The clusters are continuously transported via entrainment in the naturally convecting gas to the liquid-nitrogen filled cold finger, where they are collected via thermophoresis. The type and pressure of the gas and the precursor evaporation rate, all of which are readily controlled, determine the resulting particle-size distributions in such an apparatus. This is clearly seen in reference to the results from Granqvist and Buhrman (1976) shown in Figs. 13-5 and 13-6. The smallest cluster sizes for a given metal are obtained for a low precursor evaporation rate and condensation in a low pressure of a light inert gas, such as He. These conditions lead to a lower supersaturation of precursor atoms in the gas, slower removal of energy from the evaporated atoms, via the lighter gas atoms at lower pressure, and more rapid convective gas flow owing also to the lower gas pressure. The latter is significant, since it guarantees more rapid removal of the
591
13.2 Synthesis and Processing
condensed clusters from the supersaturated region in which, if they remained, they could grow further. It is important to realize that there are just three fundamental rates which essentially control the formation of the atom clusters in the gas-condensation process. They are: (1) the rate of supply of atoms to the region of supersaturation where condensation occurs; (2) the rate of energy removal from the hot atoms via the condensing medium, the gas; and (3) the rate of removal of the clusters once nucleated from the supersaturation region. There are other factors that can also affect the clusters finally collected, particularly those that result in significant cluster coalescence, but these three rates represent the core of the process. As such, since each of these rates can be easily increased well beyond the values utilized in the present research apparatus shown schematically in Fig. 13-4, it can be concluded that the commercial scale-up of the gas-condensation process for synthesizing nanophase materials can be readily accomplished. The clusters that are collected on the surface of the cold finger form very open, fractal structures as seen by transmission electron microscopy. A view of such a fractal collection of nanophase TiO 2 taken from the cold finger is shown in Fig. 13-7. The clusters are held there weakly and can be easily removed from this collection surface by means of a Teflon scraper. Upon removal, the clusters are funneled into a set of compaction devices (see Fig. 13-4) capable of consolidation pressures up to about 1-2 GPa, in which the nanophase compacts are formed at room temperature, or at elevated temperatures if needed. An example of the grain morphology that results from the consolidation of powders such as those shown in Fig. 13-7 is presented in Fig. 13-8. The grain size distri-
X Figure 13-7. Transmission electron micrograph of ascollected and oxidized TiO2 clusters synthesized in the apparatus shown in Fig. 13-4. From Siegel and Eastman (1989).
Figure 13-8. Transmission electron micrograph of nanophase TiO2 (rutile) after in situ consolidation at room temperature and 1.4 GPa pressure in the apparatus shown in Fig. 13-4, followed by sintering in air for 0.5 h at 500 °C. From Siegel and Hahn (1987).
bution for the as-consolidated TiO2 is shown in Fig. 13-9. It is quite narrow and has the log-normal shape typical of clusters formed via gas-condensation (Granqvist and Buhrman, 1976), as shown in Fig. 13-10. This shape is rather typical for the grain size distribution in any of the nanophase materials thus far produced by the gas-condensation method. The pellets formed in the conventional research apparatus depicted in Fig. 13-4 are typically about 9 mm in diameter and 0.1 to 0.5 mm thick. The sizes of these re-
13 Cluster Assembly of Nanophase Materials
592 I
200 -
h
150 -
I
N
100 -
50 -
— -n
Tio2
Jl
V
n
I H^T 10 15 20I 25 Grain diameter,d in nm
1 30
Figure 13-9. Grain size distribution for a nanophase TiO2 (rutile) sample compacted to 1.4 GPa at room temperature, as determined by dark-field transmission electron microscopy. From Siegel et al. (1988).
•
150 :-
100 -
" 50 -
/
search samples have been more a matter of convenience and history than any real limitation of the gas-condensation method itself. The scraping and consolidation are performed under UHV conditions after removal of the inert or reactive gases from the chamber, in order to maximize the cleanliness of the particle surfaces and the interfaces that are subsequently formed. Also, any possibility of trapping remnants of these gases in the nanophase compact is minimized by consolidation in vacuum. Since the as-collected clusters are generally aggregated in rather open fractal arrays (Kimoto et al., 1963; Tholen, 1979), their consolidation at pressures of 1-2 GPa is easily accomplished. The difficulties in consolidating the hard equiaxed agglomerates of fine powders resulting from conventional wet chemistry synthesis routes are mostly avoided. The sample densities resulting from cluster consolidation at room temperature have ranged up to about 97 % of theoretical for nanophase metals and up to about 75-85% of theoretical for nanophase oxide ceramics. This green-state porosity will be considered further below, but it probably represents (at least in part) a manifestation of powder agglomeration leading to void-like flaws. Fortunately, these appear to be capable of being removed by means of cluster consolidation at elevated temperatures and pressures without significant attendant grain growth. If a metal or alloy precursor is evaporated in an inert gas atmosphere, then the collected clusters are the same material, only in a reconstituted cluster form. The same would also be true for any nonmetallic elemental precursor evaporated in an inert atmosphere. However, if clusters of a compound such as a ceramic oxide are desired, the situation can become somewhat more complex.
^
•
•
•
i
•
•
•
LNDF's x = ^.1nm a- 1.60 cr = 1.55 Al data
\
I \• 15 5 10 Particle diameter in nm
-
20
Figure 13-10. Log-normal distribution functions (LNDF's) for two values of the geometric standard deviation o are compared with experimental Al cluster size distribution data (dots), The arrow indicates the median particle diameter x. From Granqvist and Buhrman (1976).
13.2 Synthesis and Processing
In order to produce the nanophase TiO 2 with 12 nm average grain size shown in Figs. 13-7 and 13-8, for example, Ti metal clusters condensed in He were first collected on the cold finger and subsequently oxidized by the introduction of oxygen into the chamber (Siegel et al., 1988). A similar method has been used to produce a-Al 2 O 3 (Eastman et al., 1989) with an 18 nm average grain size after oxidizing Al clusters in air at 1000 °C. If the vapor pressure of a compound is sufficiently large, as in the cases of MgO and ZnO, for example, it is possible to sublime the material directly from the Joule-heated source in a He atmosphere containing a partial pressure of O 2 . This method has been used (Eastman et al., 1989) to produce such nanophase oxides with average grain sizes down to about 5 nm. The O 2 partial pressure was utilized in an attempt to maintain full oxidation. However, this is not necessarily straightforward. Fig. 13-11 shows X-ray 0-20 scans from nanophase ZnO for three states of the nanophase samples thus produced. Fig. 13-11 a is a scan from a sample which was consolidated from the sublimed powder using only the low-pressure compaction unit of Fig. 13-4. Only Zn lines are observed. No ZnO lines are seen in this scan, although about 40% oxygen was present in the sample, indicating that an amorphous Zn-O phase may be present. After annealing the lightly compacted sample, shown in Fig. 13-11 a, in air for 3 h at 300 °C, complete conversion to ZnO occurred with little accompanying grain growth (indicated by the still broadened lines), as shown in Fig. 13-11 b. After this full oxidation step, the sample can be readily consolidated further at elevated compaction pressures. The need for performing the full ZnO oxidation step while the sample is still in a lightly compacted, and
593
(a)
(b)
(c)
JJJ 20
40 60 2-0 in degrees
80
Figure 13-11. X-ray 0-20 scans from three samples in a study of the synthesis of nanophase ZnO: (a) a sample consolidated from as-collected gas-condensed clusters using only the low-pressure compaction unit of Fig. 13-4; (b) the sample from (a) after annealing in air for 3 h at 300 °C; and (c) a sample consolidated at 1.4 GPa prior to annealing for 6 h at 600°C in air. From Eastman et al. (1989).
hence highly porous, state is evident from Fig. 13-11 c, which shows the importance of annealing samples in an oxygen atmosphere prior to the high-pressure compaction. The scan in Fig. 13-11 c was taken on a sample consolidated at 1.4 GPa prior to annealing for 6 h at 600 °C in air. The sample still contains both Zn and ZnO phases, even after such a long high-temperature oxidizing treatment. Grain growth has also occurred, as seen by the sharpening of the diffraction peaks. Fortunately, such a severe situation is not always encountered with nanophase oxides made by the conventional gascondensation method. In the case of nanophase TiO 2 cited above (Siegel et al., 1988), the oxygen deficiency, while still
594
13 Cluster Assembly of Nanophase Materials
present, is rather small and easily remedied. Raman spectroscopy has been a useful tool in studying the oxidation state of nanophase TiO 2 owing to the intense and well studied Raman bands in both the anatase and rutile forms of this oxide and the observation that these bands were affected in nanophase samples (Melendres et al., 1989). Raman spectra of as-consolidated nanophase TiO 2 samples and fully annealed and oxidized TiO2 are shown in Fig. 13-12. The band broadening observed in the nanophase samples, and band shifting as well in both the anatase and rutile phases, were confirmed (Parker and Siegel, 1990 a) to be the result of an oxygen deficiency in these samples. A subsequent calibration of this deviation from stoichiometry (Parker and Siegel, 1990 b), shown in
450
1.92 1.96 2.00 0/Ti ratio Figure 13-13. Variation with O/Ti ratio of the peak position of (a) the rutile "447 cm" 1 " vibrational mode and (b) the anatase "143 cm" 1 " vibrational mode, as well as (c) this anatase mode's full width at half maximum (FWHM). From Parker and Siegel (1990 b).
E o
or
(a)
200
800 400 600 Wavenumber in cm"1
Figure 13-12. Raman spectra of (a) an as-consolidated nanophase TiO2 sample with average grain diameter ca. 10 nm, (b) an as-consolidated nanophase TiO2 sample with average grain diameter ca. 100 nm, and (c) the sample shown in (b) after annealing in air for 0.5 h at 900 °C. The numbers shown for each spectrum are the approximate mean grain sizes in the respective samples. From Melendres et al. (1989).
Fig. 13-13, indicated that TiO1>89 was the actual material produced in the apparatus of Fig. 13-4, but that it could be easily oxidized to fully stoichiometric TiO 2 , if desired, without sacrificing its small grain size (12 nm). Also, if intermediate deviations from stoichiometry were sought, in order to select particular properties of this material, they could be readily accessed as well.
13.2 Synthesis and Processing
13.2.3 Improvements in the Gas-Condensation Method
By means of the conventional gas-condensation method described in Section 13.2.2, which utilizes natural convective gas flow, the average cluster diameters produced presently range down to about 5 nm, yielding nanophase materials with such average grain sizes and the type of size distribution shown in Fig. 13-9. However, transport of the condensed atom clusters via natural gas convection can be improved upon by using instead the motion of a forced gas. The use of a forced gas allows the gas flow rate to be independent of gas pressure (in contrast to natural convection), giving greater freedom of control of the important cluster condensation parameters. Forced gas flow is already used in more sophisticated cluster synthesis methods employed by cluster chemists and physicists to produce low yields of even smaller atomic clusters with very narrow size distributions and even monosized clusters (Andres et al., 1989). It can be expected that new cluster sources based on similar principles will be available in the future for the controlled generation of larger
595
amounts of material than are normally produced (less than a few hundred milligrams) in the type of apparatus shown in Fig. 13-4. Such new sources will enable the scale-up of cluster production to a level useful for the technological exploitation of nanophase materials and their special properties. An example of such a source (Bowles et al., 1981) that can produce significant numbers of metal clusters with a narrow size distribution is shown schematically in Fig. 13-14. A version of this cluster source has recently been used to produce smooth dense metal coatings on room-temperature substrates (Ramachandra et al., 1991). Another high-yield gascondensation cluster source for nanophase metals and alloys that utilizes forced gas flow, but generates broader size distributions, has been developed by Uda (1991). A variety of related methods for producing ultrafine particles have been described by Hayashi (1987), and one of these methods (Oda et al., 1991) for producing cluster beams has already been commercially developed for the deposition of cluster-assembled films and coatings. Most of the atom clusters assembled into nanophase materials to date have been generated from Joule-heated evaporation
Metal feed and carrier
Skimmer Beam
Oven
Quench region (mixing)
Mechanical pump
Main vacuum chamber
Figure 13-14. Idealized representation of a multiple expansion cluster source (MECS) for metal clusters. From Bowles et al. (1981).
596
13 Cluster Assembly of Nanophase Materials
sources. However, such sources have limitations that need not be accommodated to, since a wide variety of other sources are also available. The primary limitations are source-precursor incompatability, temperature range, uniformity and control, and dissimilar evaporation rates for different constituents in an alloy or compound precursor. Each of these limitations can be avoided by a host of alternative sources that have been developed over the years of ultrafme particle research, but are only now beginning to enter the field of nanophase materials synthesis. Among the alternative energy sources for bringing atom supersaturations into a condensing gas medium that have been successfully used to produce clusters or ultrafine particles are sputtering, electron beam heating, laser ablation, and plasma methods. Sputtering sources, for example, have been used in low-pressure inert or reactive environments to produce a variety of clusters including Ag, Cr, Fe, and Si (Oya et al., 1982) and Ag, Al, Cu, W, TiH 2 , WCX_X and crystalline Pd 80 Si 20 (Yatsuya et al., 1985, 1986). Recently, sputtering sources have been applied by two groups to the synthesis of nanophase materials. Hahn and Averback (1990) have used magnetron sputtering to produce clusterassembled nanophase ZrO 2 from a coarsegrained ZrO 2 precursor. Following this work, Chow et al. (1990), by alternately sputter-depositing Al films and Mo clusters with diameters below 15 nm, have formed what is effectively a three-dimensionally modulated nanostructured material with dispersed clusters. Electron-beam heating has been successfully applied by Iwama et al. (1973, 1982, 1984, 1985) to the formation of clusters of a wide range of metals and their nitrides. Metal clusters of Mo and W were formed by evaporation in He or Ar, while A1N,
TiN, ZrN, HfN, VN, NbN, CrN, Mo 2 N, and W 2 N clusters between 2 and 10 nm in size were condensed in low pressures of N 2 or NH 3 . Nevertheless, nanophase materials assembled via this promising route have been rather limited thus far (Giinther and Kumpmann, 1991). Laser ablation with a pulsed Nd: YAG laser (Matsunawa and Katayama, 1985) has also been used to produce a wide variety of metal, oxide, and nitride clusters in Ar, He, O 2 , or N 2 , but the method has apparently not yet been applied to nanophase materials. As a final example of the range of methods applicable to providing requisite atom supersaturations for the gas-condensation method, an rf plasma torch using an ArN 2 gas mixture was employed by Baba et al. (1989) to produce nanophase A1N from Al powder and NH 3 gas precursors. The material produced, at a rate of more than 200 g h~ 1 , had average particle sizes down to 20 nm and excellent sinterability in a N 2 atmosphere without the need for any sintering aids. Further discussion of the enhanced sinterability of nanophase ceramics will be found in Section 13.4.1. It should be clear that this wide variety of evaporation methods will allow for greatly increased flexibility in the use of refractory or reactive precursors for clusters, and will be especially useful as one moves toward synthesizing technological quantities of more complex multicomponent or composite nanophase materials in the future. 13.2.4 Other Physical Methods 13.2.4.1 Spark Erosion Spark erosion (Berkowitz and Walter, 1987) appears to be an alternative to the gas-condensation method in some cases in that it can yield individual clusters or particles which can be subsequently assembled
13.2 Synthesis and Processing
via consolidation. A schematic representation of a high-yield spark-erosion cell is shown in Fig. 13-15. The spark erosion process, familiar to users of electric-discharge machining, is based upon the spark-induced local melting and vaporization of electrode material in a dielectric liquid which then cools in part in the form of spark-eroded particles. Since it is estimated that only about 10% of the molten material produced in the discharge ends up in particle form, the remainder rejoining the electrodes, the process may not be very efficient. Also, only conducting materials can be used as electrodes, and hence as powder precursors, but this leaves a wide range of applicability, especially in the area of metals and alloys. While the need for a dielectric medium might seem like a drawback, providing an unavoidable source of surface contamination for the clusters prior to consolidation, one can use dielectric media such as liquid N 2 that should be essentially inert in many cases. In any event, clusters down to 5 nm in size have been synthesized by this method, and an example is shown
597
in Fig. 13-16. Further research on this promising method for ultrafine powder production and nanophase materials made by consolidating them should lead to interesting results. 13.2.4.2 Mechanical Attrition Mechanical attrition produces its nanostructures, not by cluster assembly, but by means of the mechanically induced structural decomposition of coarser-grained structures. Nanometer size grains nucleate within the shear bands of heavily deformed materials converting a coarse-grained structure to nanophase. The heavy deformation is usually induced by means of high-energy ball milling (Hellstern et al., 1989; Luton et al., 1989; Jang and Koch, 1990; Trudeau et al., 1990, 1991; Koch and Cho, 1991) but can result as well from surface wear phenomena (Ganapathi and Rigney, 1990). Thus, the individual grains are never isolated clusters and much synthesis and processing flexibility is lost. Nevertheless, ultrafine grain sizes can be readily accessed by this rather straightforward (brute-force) method, albeit with the
Charge
Screen Spark eroded particles
Figure 13-15. Schematic representation of a highyield spark-erosion cell for fine powder synthesis. From Berkowitz and Walter (1987).
Figure 13-16. Transmission electron micrograph of clusters of Fe 75 Si 15 B 10 spark eroded in pentane. After Berkowitz and Walter (1987).
598
13 Cluster Assembly of Nanophase Materials
50
1
1
Ru • AlRu
o \
E -
•- 30 N
c 20 o 10 1
1
10
0.1
100
Time in h
Figure 13-17. Grain size, determined by X-ray line broadening, of Ru and AlRu produced via mechanical attrition as a function of ball-milling time. From Hellstern et al. (1989).
probable contamination from the sources of mechanical work. The variation with grain size as a function of ball-milling time in Ru and AlRu from the work of Hellstern et al. (1989) is shown in Fig. 13-17. This clearly shows that grain sizes down into the nanometer regime can be accessed, at least in relatively hard materials. Applying the method at low temperatures (so-called cryomilling) can, however, extend the range of applicability, as shown by Luton et al. (1989) in their work on dispersion-strengthened Al. This technique is more fully treated in Chapter 5.
13.3 Characterization 13.3.1 Structure The structures of nanophase materials, both metals and oxides, have been investigated by a number of direct and indirect methods including transmission electron microscopy, X-ray and neutron scattering, and Mossbauer, Raman, and positron annihilation spectroscopy (Gleiter, 1990; Siegel, 1991). It has been found (as dis-
cussed in Section 13.2.2) that the grains in nanophase compacts are typically rather equiaxed, as are the clusters from which they were assembled, and retain the narrow log-normal size distributions representative of the clusters formed in the gascondensation method. In addition to their ultrafine grain sizes, all of the nanophase materials consolidated at room temperature to date have invariably posessed a degree of porosity ranging from about 25% to less than 5%, with the larger values for ceramics and the smaller ones for metals. Clear evidence of this porosity has been obtained by positron annihilation spectroscopy (Siegel et al., 1988; Schaefer et al., 1987,1988) and precise densitometry and porosimetry (Nieman et al., 1991; Hahn et al., 1990a) measurements. Consolidation at elevated temperatures, however, can remove this porosity without sacrificing the ultrafine grain sizes in these materials. Since a large fraction of their atoms reside in the grain boundaries of nanophase materials, the interface structures can play a significant role in determining the properties of these materials. A number of investigations on nanocrystalline metals by Gleiter and coworkers (Birringer and Gleiter, 1988), including X-ray diffraction (Zhu et al., 1987), Mossbauer spectroscopy (Herr et al., 1987), positron lifetime studies (Schaefer et al., 1987), and most recently extended X-ray absorption fine structure (EXAFS) (Haubold et al., 1988,1989), have been interpreted in terms of grain boundary atomic structures that may be random, rather than possessing either the short-range or long-range order normally found in the grain boundaries of coarser-grained polycrystalline materials. This randomness has been variously associated (Gleiter, 1990) with either the local structure of an individual boundary (as
13.3 Characterization
seen by a local probe such as EXAFS or Mossbauer spectroscopy) or the structural coordination among boundaries (as seen by X-ray diffraction). A somewhat confusing picture has emerged that needs further clarification, particularly with respect to the atomic relaxations that pertain to conventional high-angle grain boundaries. Recent investigations of nanophase TiO 2 by Raman spectroscopy (Melendres et al., 1989; Parker and Siegel, 1990 a) and of nanophase palladium by atomic resolution, transmission electron microscopy (Thomas et al., 1989, 1990; Thomas and Siegel, 1991) indicate that the grain boundary structures in these materials are rather similar to those in coarser grained, conventional materials. These studies indicate that the nanophase grain boundaries contain short-range ordered structural units representative of the bulk material and distortions that are localized to about + 0.2 nm on either side of the grain boundary plane. These conclusions are also consistent with the results from complementary small-angle neutron scattering
599
measurements (Epperson et al., 1989, 1990), and with the expectations for conventional grain boundaries from condensed matter theory (Wolf and Lutsko, 1988; Phillpot et al, 1990). Typical grain boundaries in nanophase palladium are shown in Fig. 13-3; a higher magnification view of one such boundary is shown in Fig. 13-18. This high resolution electron microscopy (HREM) study (Thomas et al., 1989, 1990; Thomas and Siegel, 1991), which included both experimental observations and complementary image simulations, indicated no manifestations of grain boundary structures with random displacements of the type or extent suggested by earlier X-ray studies on nanophase Fe, Pd, and Cu (Zhu et al., 1987; Haubold et al., 1988, 1989). HREM investigations of grain boundaries in nanophase Cu (Ganapathi and Rigney, 1990) and Fe alloys (Trudeau et al., 1991) produced by surface wear and high-energy ball milling, respectively, appear to support this view, while an earlier study of gas-condensed nanocrystalline Pd (Wun-
Figure 13-18. High resolution transmission electron micrograph of a grain boundary in nanophase palladium from an area as shown in Fig. 13-3. The magnification is indicated by the lattice fringe spacings of 0.225 nm for (111) planes. From Thomas et al. (1989, 1990).
600
13 Cluster Assembly of Nanophase Materials
derlich et al., 1990) was less conclusive regarding this issue. Indeed, very recent X-ray studies on nanophase Pd (Eastman et al., 1991; Fitzsimmons et al., 1991) suggest that the diffuse scattering effects and reduced EXAFS intensities previously attributed to such random disorder may have other sources. Atomic relaxations from normal lattice sites, which in the case of as-consolidated nanophase materials involve the atoms in grain boundaries, at void surfaces, and in grain boundary junctions, and hence a significant fraction of the atoms in the material, can be expected to give rise to observations of 'non-lattice' contributions to several types of experimental observations that are sensitive to such relaxations, such as EXAFS and Mossbauer spectroscopy (Herr et al., 1987; Haubold et al., 1988,
(a)
(b)
Figure 13-19. Image simulations for a E5 symmetric <001> tilt boundary in 7.6 nm thick palladium using microscope parameters and imaging conditions consistent with those used during HREM experimental observations (Thomas et al., 1989, 1990): (a) 'perfect' structure with no atomic displacements; (b) randomly disordered structure near the grain boundary with a maximum displacement of 0.25 of the nearest-neighbor distance.
1989; Ramasamy et al., 1991). It is clear that quantitative theoretical predictions of the effects to be expected from such relaxations need to be made and compared with experimental observations in order to ascertain whether any unexpected structural behavior remains to be attributed to nanophase grain boundaries (Siegel and Thomas, 1991). The HREM image simulations, examples of which are shown in Fig. 13-19, indicate that random atomic displacements of average magnitude greater than about 12% of a nearest neighbor distance, if present, could be readily observable by HREM for the assumed contrast conditions. The localized nature of the experimentally observed contrast changes found argues against the possibility that the interface atomic structure in nanophase materials can be fundamentally different from that observed in coarser-grained polycrystals. Such a fundamental difference could only be caused if atom positions were determined by their interactions with more than one boundary. Since the displacements observed by HREM appear to fall off rapidly for distances much smaller than even the small grain diameters of the nanophase materials thus far investigated, the atomic relaxations must be dominated by the influence of only the closest boundaries, as they are in conventional polycrystals. This also implies that the action of thinning the HREM foil, and hence removing the grains above and below the volume under observation, does not in itself significantly affect the structures observed. Recent calculations (Mills and Daw, 1990) of the possible effects of HREM specimen surfaces on grain boundary structure further support the reliability of such structural observations. Indeed, as shown in Figs. 13-3 and 13-18, the nanophase grain
13.3 Characterization
boundaries appear to be rather low energy configurations exhibiting flat facets interspersed with steps. Such a structure could only arise if sufficient local atomic motion occurred during the cluster consolidation process to allow the system to reach at least a local energy minimum. Such observations suggest at least two conclusions: first, that the atoms that constitute the grain boundary volume in nanophase materials have sufficient mobility during cluster consolidation to accommodate themselves into relatively low-energy grain boundary configuratons; and second, that the local driving forces for grain growth are relatively small, despite the large amount of energy stored in the many grain boundaries in these materials.
601
300
200 £00 600 800 Sintering temperature in °C
1000
Figure 13-20. Variation of average grain size with increasing sintering temperature (0.5 h at each) for a nanophase TiO2 (rutile) sample compacted to 1.4 GPa at room temperature, as determined by darkfield transmission electron microscopy. After Siegel et al. (1988). 1000
13.3.2 Stability A technologically important aspect of nanophase materials assembled from atom clusters, and scientifically interesting as well, is their apparently inherent stability against grain growth. Their grain sizes, as measured by transmission electron microscopy, remain rather deeply metastable to elevated temperatures. For example, as shown in Fig. 13-20, the 12 nm initial average grain diameter for the distribution shown in Fig. 13-9 changes little with annealing to elevated temperatures until about 40-50% of the absolute melting temperature (Tm) of TiO 2 is reached. This behavior appears to be rather typical for the nanophase oxides already investigated (Eastman et al., 1989) and for nanophase metals as well (Hort, 1986), as shown in the Arrhenius plot in Fig. 13-21. In the case of the TiO 2 , rapid grain growth only develops above the temperature at which the mean bulk diffusion distance (DTit)1/2 of Ti becomes comparable to the mean grain size, at which temperature any local
100
1 0.000
0.001
0.002 1/7 in K"1
0.003
0.004
Figure 13-21. Arrhenius plot of the variation of average grain size, measured by dark-field transmission electron microscopy, with sintering temperature for nanophase Fe (Hort, 1986), TiO2 (Siegel et al., 1988), MgO/WOx (Eastman et al., 1989), and ZnO (Eastman et al., 1989), The oxide samples were annealed for 0.5 h in air at each temperature; the iron for 10 h in vacuum. From Siegel (1990 b).
barriers to grain growth would cease to be significant. Given the observations of the grain size distributions, grain morphologies, and grain boundary structures in nanophase
602
13 Cluster Assembly of Nanophase Materials
materials, it seems likely that the resistance to grain growth observed for nanophase materials results primarily from frustration (Siegel, 1990b). It appears that the narrow grain size distributions normally observed in these cluster-assembled materials, coupled with their relatively flat grain boundary configurations (and also enhanced by their multiplicity of grain boundary junctions), place these nanophase structures in a local minimum in energy from which they are not easily extricated. They are thus analogous to a variety of closed-cell foam structures, which are stable (really, deeply metastable) despite their large stored surface energy. Under such conditions, only at temperatures above which bulk diffusion distances are comparable with or greater than the grain size, as in the case of nanophase TiO2 cited above, will this metastability give way to global energy minimization via rapid grain growth. Such diffusion-controlled grain-growth behavior is apparent in Fig. 13-21. The effective activation energy of this high-temperature limiting behavior, however, is only about 9 kBTm, approximately one half that for self-diffusion. Exceptions to this frustrated grain growth behavior could be expected if considerably broader grain size distributions were present in a sample, which would allow a few larger grains to grow at the expense of smaller ones, or if significant grain boundary contamination were present, allowing enhanced stabilization of the small grain sizes to further elevated temperatures. One could, of course, intentionally stabilize against grain growth by appropriate doping or composite formation. It has also been recently suggested (Hofler and Averback, 1990) that the porosity present in nanophase ceramics may assist in their stability against grain growth.
13.4 Properties 13.4.1 Sinterability and Diffusion
Nanophase materials have a variety of properties that are different and often considerably improved in comparison with those of conventional coarse-grained structures. For example, nanophase TiO 2 (rutile) exhibits significant improvements in both sinterability and resulting mechanical properties relative to conventionally synthesized coarser-grained rutile (Siegel et al., 1988; Averback et al., 1989; Hahn et al., 1990a; Mayo et al., 1990). Nanophase TiO 2 with a 12 nm initial mean grain diameter has been shown (Siegel et al., 1988) to sinter under ambient pressures at temperatures 400 to 600 °C lower than conventional coarser-grained rutile, and without the need for any compacting or sintering aid, such as polyvinyl alcohol, which is usually required. This behavior is shown in Fig. 13-22. More recently, it has been demonstrated by Hahn et al. (1990 a) that sintering the same nanophase material under pressure (1 GPa), or with appropriate dopants such as Y, can further reduce the sintering temperatures, while suppressing grain growth as well (Fig. 13-23). The resulting fracture characteristics (Li et al., 1988; Averback et al., 1989; Hofler and Averback, 1990) developed for sintered nanophase TiO2 are as good as (or in some aspects improved relative to) those for conventional rutile. It may not be terribly surprising that nanophase ceramics, with their ultrafine grain sizes in the nanometer regime, clean cluster surfaces, and high grain boundary purity, will sinter at much lower temperatures than conventional coarser-grained ceramics. However, it is unique that they can also retain their ultrafine grain size after sintering to full density and can con-
13.4 Properties I
I
I
1600
TiO9 • 1200
12nm.UGPa o 800
1*00
* 200
400
,
p
/*-
600 800 1000 Temperature in °C
1.3 urn. 0.1 GPa 1200
tinue to exhibit superior mechanical properties as well. Positron annihilation spectroscopy (PAS) has been a useful tool in the study of the ultrafine-scale porosity inherent in as-consolidated nanophase compacts (Schaefer et
O
• p-V
A
A p=
a
• p = 0.(Ti982Yl8)0
1|im at 1000°C
100 ~
c
& 50
Q
0
603
500 1000 Sintering temperature in °C
Figure 13-23. Density (open symbols) and grain size (closed symbols) of nanophase TiO2 as a function of sintering temperature. Included are data for sintering at atmospheric pressure, pressure-assisted sintering, and Y-doped nanophase TiO 2 . After Averback et al. (1989).
Figure 13-22. Vickers microhardness of TiO2 (rutile) measured at room temperature as a function of 0.5 h sintering at successively increased temperatures. Results for a nanophase sample (filled squares) with an initial average grain size of 12 nm consolidated at 1.4 GPa are compared with those for coarser-grained samples with 1.3 urn initial average grain size sintered with (diamonds) or without (circles) the aid of polyvinyl alcohol from commercial powder consolidated at 0.1 GPa and 1.4 GPa, respectively. After Siegelet al. (1988).
U00
al., 1987, 1988; Siegel et al., 1988). Such porosity can be probed to advantage by PAS, as a function of sintering temperature, to observe densification via the removal of voids. An example of PAS lifetime results (Siegel et al., 1988) used to follow the sintering behavior of nanophase TiO 2 is shown in Figs. 13-24 and 13-25. The positron lifetime spectra in Fig. 13-24 exhibit the long tails representative of void-trapped positrons until annealing above 500 °C significantly reduces the porosity in the nanophase TiO 2 . In the deconvoluted data shown in Fig. 13-25, the intensity I2 of the lifetime (T 2 ) signal corresponding to positron annihilation from void-trapped states in the nanophase sample is seen to decrease rapidly during sintering above 500 °C as a result of the densification of this ultrafme-grained ceramic, even though rapid grain growth does not set in until above 800 °C. Furthermore, the variation of T2 with sintering indicates that there is a redistribution of void sizes accompanying this densification. Similar behavior is also observed for the coarser-grained samples, but as expected, the densification proceeds more slowly in these latter samples and the pore sizes are
604
13 Cluster Assembly of Nanophase Materials
105
1
1
• TiO 2
10'c
#
c
o
D
o o
D
10
3 _
A
-
•
o
•
as compacted 500°C 700°C 900°C
Oo° o O
* «»#^
°°0d>
## *^#« °O°0O° *.• °ooo0 %• CXX3O
102
1 1.00 800 Time in ps
275 250 L 225
: 200 7
175
-5"
Single crystal
150
200
400 600 Temperature in °C
800
1000
1200
Figure 13-24. Positron annihilation lifetime spectra for nanophase TiO2 as a function of sintering temperature. The successive sintering anneals were for 0.5 h in air. After Siegel et al. (1988).
1600
larger according to the larger values of T2 . The redistribution of void or pore sizes can also be monitored by means of BET (Brunauer, Emmett, Teller) measurements as demonstrated by the work of Hahn etal. (1990a) shown in Fig. 13-26. The structure of nanophase TiO2 in its as-consolidated state and as a function of sintering in air has also been followed by small-angle neutron scattering (SANS) (Epperson et al., 1989) (see Volume 2, Chapter 22). SANS can yield further information regarding the nature of the intergrain nanophase boundaries, particularly their average density vis-a-vis the grain density (Epperson et al., 1989, 1990; Jorra
Figure 13-25. Results of two-component (T1? T2) lifetime fits to positron annihilation data, similar to those shown in Fig. 13-24, from three TiO2 samples as a function of sintering temperature. A 12 nm grain size nanophase sample (filled circles) compacted at 1.4 GPa is compared with 1.3 urn grain size samples compacted at 1.4 GPa (open circles) and 0.32 GPa (triangles) from commercial powder. The PAS data were taken at room temperature; no sintering aids were used. From Siegel et al. (1988).
13.4 Properties 1CT
as prepared 2h;350°C 15h;700oC
10"
£ 10"
10 100 Pore diameter in nm
1000
Figure 13-26. Pore size distributions in nanophase TiO2 in the as-prepared state (compaction at 150°C and 2 GPa for 2 h) and after sintering in air at atmospheric pressure at 350 °C and 700 °C. From Hahn et al. (1990a).
et al., 1989). It can also yield valuable information about the presence of voids (or porosity), and also void removal during the sintering process. Scattering data from nanophase TiO2 as a function of sintering time at 550 °C are shown in Fig. 13-27. The results of a maximum entropy analysis of these SANS data are shown in Fig. 13-28,
605
where the size distributions of scattering centers for various times during sintering are presented. At least the first peak in these distributions sippears to result from voids, whose number diminishes with sintering. The presence of such voids, which has also been clearly demonstrated by PAS (Schaefer et al., 1987; Siegel et al., 1988), can reduce the apparent average density of grain boundaries deduced from SANS data, and may lead to erroneous conclusions regarding their structure. The low densities in nanophase TiO2 and Pd grain boundaries deduced from SANS measurements (Epperson et al., 1989,1990; Jorra et al., 1989) may thus be partly a result of sample porosity. A clear separation of these effects has not yet been made, but will need to be done before reliable comparisons between experiment and theory can be made. Atomic diffusion in nanophase materials, which can have a significant bearing on their mechanical properties, such as creep and superplasticity, and electrical properties as well, has been found to be
Figure 13-27. Absolute SANS intensity S(q) as a function of the momentum transfer q = (4n/X) sin 0 for as-consolidated nanophase TiO2 and after sintering at 550 °C in air for the times indicated. From Epperson et al. (1989).
-8.0 -1.2
-3.1
-2.6
In (q)
-1.8
606
13 Cluster Assembly of Nanophase Materials
45.0 D - as received o = 15m A = 120m
41.0-
37.0-
33.0-
29.0-
25.0 0.0
30.0
60.0
90.0
120.0
150.0
Figure 13-28. Semi-logarithmic plot of the size distribution N(r) of scatterers obtained via a maximum entropy analysis of the SANS data in Fig. 13-27 from nanophase TiO2 compacts sintered at 550 °C in air for the indicated times. From Epperson et al. (1990).
r in A
very rapid. Measurements of self-diffusion and impurity-diffusion (Horvath et al., 1987; Horvath, 1989; Hahn et al., 1989; Averback et al., 1989; Schumacher et al., 1989) in as-consolidated nanophase metals and ceramics indicate that atomic transport is orders of magnitude faster in these materials than in coarser-grained polycrystalline samples. However, the very rapid diffusion in as-consolidated nanophase materials appears to be intrinsically coupled with the porous nature of the interfaces in these materials. The diffusion can be suppressed back to conventional values by sintering samples to full density (Averback et al., 1989), as shown by a comparison between the Hf diffusion observations shown in Figs. 13-29 and 13-30 before and after pressure-assisted sintering of nanophase TiO 2 with a grain size of about 12 nm. Nonetheless, there exist considerable possibilities for efficiently doping nanophase materials at relatively low temperatures via the rapid diffusion available along their ubiquitous grain-boundary networks, with only short diffusion paths remaining into their grain interiors, to synthesize materials with tailored optical, electrical, or mechanical properties.
13.4.2 Electrical Properties Little work has been carried out so far on the electrical properties of nanophase materials. However there appear to be interesting prospects, if the results shown in Fig. 13-31 are any indication. Nanophase TiO2 was doped at about the 1 % level with Pt diffused in from the surface. After annealing in air for 4 h at about 500 °C, the AC conductivity of the sample was measured as a function of temperature. The strongly nonlinear, and reversible, electrical response shown in Fig. 13-31, caused presumably by the Pt doping into the band gap of this wide-band-gap (3.2 eV) semiconductor, suggests that the rather easy doping of nanophase electroceramics may lead to a wide range of interesting device applications in the future. However, much work remains to be done in this area. 13.4.3 Mechanical Behavior 13.4.3.1 Ceramics It has been observed that nanophase ceramics are easily formed, as is clearly evident in the sample compaction process (Siegel et al., 1988) and from demonstra-
607
13.4 Properties
Room temp. 38820 s; I5O°C 30360s 300°C
-
"A
600
J
j\
10°
I.J \5nm Hf
400o
f 1
-
Hf
o 00
200
0
\
I:
U
-
950
940
960
970
10
0
100
200 300 400 Temperature in °C
500
600
Figure 13-31. The AC conductivity of Pt-doped nanophase TiO2 as a function of temperature. The sample was pre-annealed in air at about 500 °C for 4 h prior to the conductivity measurements; the electrical response is reversible with temperature. From Narayanasamy, Eastman, and Siegel, unpublished results.
Channels Figure 13-29. Diffusion profiles of Hf in nanophase TiO2 measured by Rutherford backscattering after sintering in air at atmospheric pressure at 100 °C and subsequent Hf deposition on the sample surface. From Hahn et al. (1989).
800
. ^ 6 0 0 --
555 °C 654 °C 764 °C
jB 400 S
n-Tiu 2 200--I—• U 1.5 nm Hf
/n
,J \
i 10°
i
V
850
900
950
1000
Channels Figure 13-30. Diffusion profiles of Hf in nanophase TiO2 measured by Rutherford backscattering after pressure-assisted sintering in air at 1 GPa at 550 °C and subsequent Hf deposition on the sample surface. From Averback et al. (1989).
tions via deformation (Karch et al., 1987; Karch and Birringer, 1990) as well. However, the degree to which nanophase ceramics are truly ductile is only beginning to be understood. Nanoindenter measurements on nanophase TiO 2 (Mayo et al., 1990) and ZnO (Mayo et al., 1991) have recently demonstrated that a dramatic increase of strain rate sensitivity occurs with decreasing grain size, as shown in Fig. 13-32. Since this strong grain-size dependence is found for sets of samples in which the porosity is changing very little, it appears to be an intrinsic property of these ultrafme-grained ceramics. The strain rate sensitivity (m) values at the smallest grain sizes yet investigated (12 nm in nanophase TiO 2 and 7 nm in ZnO) thus indicate ductile behavior of these nanophase ceramics, as well as a significant potential for increased ductility at even smaller grain sizes and elevated temperatures. The maximum strain rate sensitivities measured in these
608
13 Cluster Assembly of Nanophase Materials
0.045^ 0.0401
w c
0.035-
•
ZnO
•
TiC>
2 0.0302 0.025•| 0.0200.0150.010 10
50
100
500
Grain size in nm
Figure 13-32. Strain rate sensitivity, at room temperature, of nanophase TiO2 (Mayo et al., 1990) and ZnO (Mayo et al., 1991) as a function of grain size. The strain rate sensitivity was measured by a nanoindentation method (Mayo et al., 1990) and the grain size was determined by dark-field transmission electron microscopy. After Siegel (1991).
studies, about 0.04, are already approximately one-quarter that for lead at room temperature, for example. However, no superplasticity has yet been observed in nanophase materials at room temperature, which would yield m values about an order of magnitude higher than the maximum observed (Volume 6, Chapter 9). Nevertheless, it already seems clear that in the future, at smaller grain sizes and/or at elevated temperatures, superplasticity of these materials will indeed be observed. The possibilities for plastic forming nanophase ceramics to near net shape appear to be well on their way to realization. Karch and Birringer (1990) have recently demonstrated that nanophase TiO 2 could be readily formed to a desired shape with excellent detail below 900 °C, and the fracture toughness was found to increase by a factor of two as well. The ability to extensively deform nanophase TiO 2 at elevated temperatures (ca. 800 °C) without cracking or fracture is shown dramatically in Fig. 13-33 from the work of Hahn et al.
Figure 13-33. Nanophase TiO2 sample before and after compression at 810 °C for 15 h. The total true strains were as high as 0.6; this represents a deformation to a final thickness of less than 2 mm from an initial length of 3.5 mm at about 0.5 of its melting temperature (1830°C). The grain size increased from 40-50 nm to about 1 um. The small rule divisions are millimeters. From Hahn et al. (1990 b).
(1990b) (Hahn and Averback, 1991). An additional attribute of this nanophase ceramic is the possibility of machining it with conventional tools, as shown in Fig. 13-34. This material, still in its green state at 75 % density, could be subsequently sintered to full density without sintering aids or large amounts of shrinkage.
Figure 13-34. Core drilled as-consolidated nanophase TiO2 sample with an average grain size of 12 nm. The small rule divisions are millimeters. From Narayanasamy, Eastman, and Siegel, unpublished results.
13.4 Properties
13.4.3.2 Metals The dominant mechanical property change resulting from reducing the grain sizes of nanophase metals is the significant increase in their strength. While the microhardness of as-consolidated nanophase oxides is reduced relative to their fullydense counterparts, owing to significant porosity in addition to their ultrafine grain sizes, the case for nanophase metals is 400' nanocrystalline Pd 300-
200c T3 O
100-
o
0
£00 800 Temperature in °C
1200
Figure 13-35. Microhardness of three nanocrystalline (5-10 nm) palladium samples and two coarsegrained (100 um) palladium samples as a function of annealing temperature. All samples were annealed for 100 min in 0.16 Pa vacuum and then measured at room temperature. From Nieman et al. (1989).
3.0
—i—i—i—|—i—i—
•
1 i i
•
1
i
1
•
l
2.5 X •
<^'
+-
2.0
-i—i—i—|—i—i—i—
cu
:
6 nm 8 nm
"
15 nm
Q_
m :
-
*
/
\ m_
-+
^ ^ ^ ^ ^ - ^
w 0.5
0.0
;
J
•
<2. 1.5
1.0
quite different. Figs. 13-35, 13-36 and 1337 show recent microhardness and stressstrain results for nanophase Pd and Cu and their coarser-grained counterparts (Nieman et al., 1989, 1990, 1991). In the as-consolidated state, nanophase palladium samples with 5-10 nm grain sizes exhibit up to about a 500 % increase in hardness over coarser-grained (ca. 100 Jim) samples (Fig. 13-35), with concomitant increases in yield stress oy (Fig. 13-37). Similar results have been observed in nanophase copper as well (Fig. 13-36). As shown in Fig. 13-35, the hardness of nanophase Pd at room temperature falls only slowly with annealing up to about 50% of its absolute melting temperature, commensurate with the rather deep observed grain size metastability in these materials cited in Section 13.3.2. Nanophase metals and alloys produced via mechanical attrition also exhibit significantly enhanced strength. For example, Koch and coworkers (Jang and Koch, 1990; Koch and Cho, 1991) have found hardness increases of factors of 4 to 5 in nanophase Fe and a factor of about 1.2 in nanophase Nb 3 Sn when the grain size drops from 100 nm to 6 nm. On the other
\
"o
3?
609
25 nm
x^ •
7" • — • —•- - —m-..ilt. (D
2
6 8 Measurement No.
50 um
10
12
Figure 13-36. Vickers microhardness measurements at a number of positions across several nanophase Cu samples ranging in grain size from 6 to 50 nm, compared with similar measurements from an annealed conventional 50 um grain size Cu sample. After Nieman et al. (1991).
610
13 Cluster Assembly of Nanophase Materials
300
- Unm grain size -50|jm grain size
1.0 1.5 Strain in %
2.5
Figure 13-37. Stress-strain curve for a nanophase (14 nm) Pd sample compared with that for a coarsegrained (50 |iim) Pd sample. The strain rate £%2x 10~5 s" 1 . After Nieman et al. (1990).
hand, Chokshi et al. (1989) have reported an apparent softening with decreasing grain size in the nanometer regime for cluster-assembled Cu and Pd. The increased strength in ultrafmegrained nanophase metals, although analogous to conventional Hall-Petch strengthening observed with decreasing grain size in coarser-grained metals, must result from fundamentally different mechanisms. The grain sizes here are, after all, smaller than the necessary critical bowing lengths for Frank-Read dislocation sources to operate and smaller than the normal spacings between dislocations in a pile-up, as well. An adequate description of the mechanisms responsible for the increased strength observed in nanophase metals will clearly need to accommodate to the grain-size scale in these materials. It appears that as the grain-size scale is reduced, the energetic hierarchy of microscopic deformation mechanisms or paths is successively accessed, with easier paths (such as dislocation generation from Frank-Read sources) becoming frozen out at sufficiently small grain sizes and more costly paths becoming necessary to effect deformation.
The rapid atomic diffusion observed in nanophase materials, along with their nanometer grain sizes, has suggested that a large creep enhancement might result, even at room temperature (Karch et al., 1987). However, recently completed constantstress creep measurements on nanophase Pd and Cu (Nieman et al., 1990, 1991) show that the observed creep rates at room temperature are at least three orders of magnitude smaller than predicted on the basis of a Coble creep model, in which the creep rate varies as Dh/d3, where Dh is the grain boundary diffusivity and d is the mean grain size. Such creep resistance will need to be explored further at elevated temperatures. The enhanced strain rate sensitivity at room temperature found in nanophase TiO2 and ZnO (Mayo et al., 1990, 1991) appears to result from increased grain boundary sliding in this material, aided by the presence of porosity, ultrafine grain size, and probably rapid diffusion as well. The increased strength of nanophase metals, on the other hand, indicates that dislocation generation, as well as dislocation mobility, may become significantly difficult in ultrafine-grained metals. It may thus be that the increased strength of nanophase metals and the increased ductility of nanophase ceramics indicate a convergence of the mechanical response of these two classes of materials as grain sizes enter the nanometer size range. In such a case, grain boundary sliding mechanisms, accompanied by short-range diffusion-assisted healing events, would be expected to increasingly dominate the deformation of nanophase materials, and superplasticity in a wide range of nanophase materials, including metals and alloys, intermetallic compounds, ceramics, and even semiconductors, could result. Accordingly, increased opportunities for high-deforma-
13.5 Future Directions
tion or superplastic near-net-shape forming of a very wide range of materials could result.
13.5 Future Directions We are just beginning to take advantage of the opportunities for synthesizing nanophase materials via the assembly of atom clusters. The future appears to hold great promise for these materials, on the basis of the limited knowledge that has already been accumulated. The cluster sizes accessed to date indicate that the high reactivities and short diffusion distances available in cluster-assembled materials can have profound effects upon their processing characteristics. These characteristics should be further enhanced as even smaller and more uniformly sized clusters become available in sufficiently large numbers to effect their assembly into usable materials. The enhanced diffusivities available along the grain boundary networks of nanophase materials, with only few atomic jumps separating grain interiors from grain boundaries, should enable efficient impurity doping of these materials. Nanophase insulators and semiconductors, for example, could be easily doped with impurities at relatively low temperatures, thus allowing efficient introduction of impurity levels into their band gaps and control over their electrical and optical properties. Moreover, the ability to produce, via cluster assembly, fully dense ultrafine-grained nanophase ceramics that are formable and exhibit ductility can have a significant technological impact in a wide variety of applications. Near net-shape forming of nanophase ceramic parts with complex and ultrafine detail would seem to be possible. Subsequent controlled grain
611
growth can then be used to alter the grainsize dependent properties of these ceramics. Research on cluster-assembled nanophase materials is currently being carried out in only a few laboratories, and considerable work still remains to be done. An understanding of the structure of these interfacial materials is now being developed and a number of their interesting properties are being disclosed. It can be expected that numerous relationships between their atomic and electronic structures and properties will soon become clear, and that these relationships will lead to new materials science. For this to happen, the nature of the local structure of the grain boundaries and interfaces in these materials, which comprise such a large fraction of their volumes and affect their properties so dramatically, must become better understood. While progress has been made in this area recently, further work is still needed to fully understand which interface characteristics and properties are specific to those in nanophase materials and which are simply representative of interfaces in general and only observable in nanophase materials owing to the uniquely large volume fraction of material that they command. The contributions from atomic scale porosity must be clearly elucidated in this regard. With porosity removed, nanophase materials may offer a unique environment for the study of the properties of internal interfaces. Finally, further research on the synthesis of a broader range of nanophase materials, encompassing metals, alloys, ceramics, semiconductors, and composites, is needed. Also, a variety of measurements of the electrical, optical, magnetic, and mechanical properties of these new materials will certainly help in developing an understanding of how great an impact cluster-
612
13 Cluster Assembly of Nanophase Materials
assembled nanophase materials will eventually have on materials technology.
13.6 Acknowledgements This work was supported by the U.S. Department of Energy, BES-Materials Sciences, under Contract W-31-109-Eng38. The author wishes to thank his many collaborators at Argonne National Laboratory and elsewhere, without whose efforts and contributions this work would not have been possible.
13.7 References Andres, R. P., Averback, R. S., Brown, W. L., Brus, L. E., Goddard III, W. A., Kaldor, A., Louie, S. G., Moskovits, M., Peercy, P. S., Riley, S. J., Siegel, R. W., Spaepen, R, Wang, Y (1989), /. Mater. Res. 4, 704. Averback, R. S., Hahn, H., Holler, H. X, Logas, J. L., Chen, T. C. (1989), Mater. Res. Soc. Symp. Proc. 153, 3. Baba, K., Shohata, N., Yonezawa, M. (1989), Appl. Phys. Lett. 54, 2309. Berkowitz, A. E., Walter, J. L. (1987), J. Mater. Res. 2, 277. Birringer, R., Gleiter, H. (1988), in: Encyclopedia of Materials Science and Engineering, Suppl. Vol. 1: Cahn, R. W. (Ed.), Oxford: Pergamon Press, p. 339. Birringer, R., Gleiter, H., Klein, H.-P., Marquardt, P. (1984), Phys. Lett. 102A, 365. Birringer, R., Herr, U., Gleiter, H. (1986), Suppl. Trans. Jpn. Inst. Met. 27, 43. Blander, M., Abdel-Gawad, M. (1969), Geochim. Cosmochim. Acta 33, 701. Blander, M., Katz, J. L. (1967), Geochim. Cosmochim. Acta 31, 1025. Bowles, R. S., Kolstad, J. I, Calo, J. M., Andres, R. P. (1981), Surface Science 106, 111. Chokshi, A. H., Rosen, A., Karch, I, Gleiter, H. (1989), Scripta Metall. 23, 1679. Chow, G. M., Holtz, R. L., Pattnaik, A., Edelstein, A. S., Schlesinger, T. E., Cammerata, R. C. (1990), Appl. Phys. Lett. 56, 1853. Eastman, J. A., Liao, Y X., Narayanasamy, A., Siegel, R. W. (1989), Mater. Res. Soc. Symp. Proc. 155, 255. Eastman, J. A., Fitzsimmons, M. R., Muller-Stach, M., Wallner, G., Elam, W. T. (1991), Scripta Metall. et Mater., in press.
Epperson, J. E., Siegel, R. W, White, J. W, Klippert, T. E., Narayanasamy, A., Eastman, J. A., Trouw, F. (1989), Mater. Res. Soc. Symp. Proc. 132, 15. Epperson, J. E., Siegel, R. W., White, J. W., Eastman, J. A., Liao, Y. X., Narayanasamy, A. (1990), Mater. Res. Soc. Symp. Proc. 166, 87. Fitzsimmons, M. R., Eastman, J. A., Muller-Stach, M., Wallner, G. (1991), Phys. Rev. B, in press. Ganapathi, S. K., Rigney, D. A. (1990), Scripta Metall. et Mater. 24, 1675 Gleiter, H. (1981), In: Deformation of Poly crystals: Mechanisms and Microstructures: Hansen, N., Horsewell, A., Leffers, T., Lilholt, H. (Eds.). Roskilde, Denmark: Ris0 National Laboratory, p. 15. Gleiter, H. (1990), Progress in Materials Science 33, 223. Granqvist, C. G., Buhrman, R. A. (1976), /. Appl Phys. 47, 2200. Grossman, L. (1972), Geochim. Cosmochim. Acta 36, 597. Gunther, B., Kumpmann, A. (1991), Acta Metall. et Mater., in press. Hahn, H., Averback, R. S. (1990), J. Appl. Phys. 67, 1113. Hahn, H., Averback, R. S. (1991), J. Amer. Ceram. Soc, to be published. Hahn, H., Eastman, J. A., Siegel, R. W. (1988), In: Ceramic Transactions, Ceramic Powder Science, Vol. 1, Part B: Messing, G. L., Fuller, E. R., Jr., Hausner, H. (Eds.), Westerville: American Ceramic Society, p. 1115. Hahn, H., Hofler, H. X, Averback, R. S. (1989), Defect and Diffusion Forum 66—69, 549. Hahn, H., Logas, J., Averback, R. S. (1990a), J. Mater. Res. 5, 609. Hahn, H., Logas, J., Hofler, H. J., Kurath, P., Averback, R. S. (1990b), Mater. Res. Soc. Symp. Proc. 196, 71. Haubold, T., Birringer, R., Lengeler, B., Gleiter, H. (1988), J. Less-Common Metals 145, 557. Haubold, T., Birringer, R., Lengeler, B., Gleiter, H. (1989), Phys. Lett. A135, 461. Hayashi, C. (1987), J. Vac. Sci. Technol. A5, 1375. Hellstern, E., Fecht, H. 1, Fu, Z., Johnson, W. L. (1989), J. Appl. Phys. 65, 305. Herr, U., Jing, X, Birringer, R., Gonser, U., Gleiter, H. (1987), Appl. Phys. Lett. 50, 472. Hofler, H. X, Averback, R. S. (1990), Scripta Metall. et Mater. 24, 2401 Hort, E. (1986), Diploma Thesis, Universitat des Saarlandes, Saarbrucken. Horvath, X, Birringer, R., Gleiter, H. (1987), Solid State Commun. 62, 319. Horvath, X (1989), Defect and Diffusion Forum 6669, 207. Iwama, S., Hayakawa, K. (1985), Surface Sci. 156, 85. Iwama, S., Shichi, E., Sahashi, T. (1973), Jpn. J. Appl. Phys. 12, 1531.
General Reading
Iwama, S., Hayakawa, K., Arizumi, T. (1982), J. Cryst. Growth 56, 265. Iwama, S., Hayakawa, K., Arizumi, T. (1984), /. Cryst. Growth 66, 189. Jang, J. S. C, Koch, C. C. (1990), Scripta Metall et Mater. 24, 1599. Jorra, E., Franz, H., Peisl, I, Wallner, G., Petry, W, Birringer, R., Gleiter, H., Haubold, T. (1989), Phil. Mag. B60, 159. Karch, 1, Birringer, R. (1990), Ceramics International 16,291. Karch, X, Birringer, R., Gleiter, H. (1987), Nature 330, 556. Kear, B. H., Cross, L. E., Keem, J. E., Siegel, R. W, Spaepen, R, Taylor, K. C , Thomas, E. L., Tu, K.-N. (1989), Research Opportunities for Materials with Ultrafine Microstructures. Washington, DC: National Academy, Vol. NMAB-454. Kimoto, K., Kamiya, Y, Nonoyama, M., Uyeda, R. (1963), Jpn. J. Appl. Phys. 2, 702. Koch, C. C , Cho, Y. S. (1991), Scripta Metall. et Mater., in press. Li, Z., Ramasamy, S., Hahn, H., Siegel, R. W. (1988), Mater. Lett. 6, 195. Luton, M. J., Janath, C.S., Disko, M. M., Matras, S., Vallone, J. (1989), Mater. Res. Soc. Symp. Proc. 132, 79. Matsunawa, A., Katayama, S. (1985), In: Laser Welding, Machining and Materials Processing, Proc. ICALEO '85: Albright, C. (Ed.), IFS Publ. Ltd., p. 205. Mayo, M. J., Siegel, R. W, Narayanasamy, A., Nix, W. D. (1990), /. Mater. Res. 5, 1073. Mayo, M. J., Siegel, R. W, Liao, Y X., Nix, W. D. (1991), /. Mater. Res., to be published. Melendres, C. A., Narayanasamy, A., Maroni, V. A., Siegel, R. W. (1989), J. Mater. Res. 4, 1246. Mills, M. I , Daw, M. S. (1990), Mater. Res. Soc. Symp. Proc. 183, 15. Nieman, G. W, Weertman, J. R., Siegel, R. W. (1989), Scripta Metall. 23, 2013. Nieman, G. W, Weertman, J. R., Siegel, R. W. (1990), Scripta Metall. et Mater. 24, 145. Nieman, G. W, Weertman, J. R., Siegel, R. W (1991), J. Mater. Res. 6, 1012. Oda, M., Fuchita, E., Tsuneizumi, M., Kashu, S., Hayashi, C. (1991), Scripta Metall. et Mater., in press. Oya, H., Ichihashi, T., Wada, N. (1982), Jpn. /. Appl. Phys. 21, 554. Parker, J. C , Siegel, R. W. (1990 a), /. Mater. Res. 5, 1246. Parker, J. C , Siegel, R. W. (1990b), Appl. Phys. Lett. 57, 943. Phillpot, S. R., Wolf, D., Yip, S. (1990), MRS Bulletin XV(10), 38. Ramachandra, A., Vaziri, M., Andres, R. P. (1991), Mater. Res. Soc. Symp. Proc. 206, in press. Ramasamy, S., Jiang, I , Gleiter, H., Birringer, R., Gonser, U. (1991), Solid State Commun., in press.
613
Schaefer, H. E., Wurschum, R., Scheytt, M., Birringer, R., Gleiter, H. (1987), Mater. Sci. Forum 15-18, 955. Schaefer, H.-E., Wurschum, R., Birringer, R., Gleiter, H. (1988), Phys. Rev. B38, 9545. Schumacher, S., Birringer, R., Straub, R., Gleiter, H. (1989), Acta Metall. 37, 2485. Siegel, R. W. (1990 a), MRS Bulletin XV (10), 60. Siegel, R. W (1990 b), Mater. Res. Soc. Symp. Proc. 196, 59. Siegel, R. W (1991), Ann. Rev. Mater. Sci. 21, 559. Siegel, R. W, Eastman, J. A. (1989), Mater. Res. Soc. Symp. Proc. 132, 3. Siegel, R. W, Hahn, H. (1987), in: Current Trends in the Physics of Materials: Yussouff, M. (Ed.), Singapore: World Scientific Publ. Co., p. 403. Siegel, R. W, Thomas, G. J. (1991), Mater. Res. Soc. Symp. Proc. 209, 15. Siegel, R. W, Ramasamy, S., Hahn, H., Li, Z., Lu, X, Gronsky, R. (1988), J. Mater. Res. 3, 1367. Tholen, A. R. (1979), Acta Metall. 27, 1765. Thomas, G. J., Siegel, R. W, Eastman, J. A. (1989), Mater. Res. Soc. Symp. Proc. 153, 13. Thomas, G. X, Siegel, R. W, Eastman, X A. (1990), Scripta Metall. et Mater. 24, 201. Thomas, G. X, Siegel, R. W. (1991), J. Mater. Res., to be published. Trudeau, M. L., Dussault, D., Van Neste, A., Schultz, R. (1990), Phys. Rev. Lett. 64, 99. Trudeau, M. L., Van Neste, A., Schultz, R. (1991), Mater. Res. Soc. Symp. Proc. 206, in press. Uda, M. (1991), Acta Metall. et Mater., in press. Wolf, D., Lutsko, X F. (1988), Phys. Rev. Lett. 60, 1170. Wunderlich, W, Ishida, Y, Maurer, R. (1990), Scripta Metall. et Mater. 24, 403. Yatsuya, S., Yamauchi, K., Kamakura, T, Yanagada, H., Wakaiyama, H., Mihama, K. (1985), Surface Sci. 156, 1011. Yatsuya, S., Kamakura, T, Yamauchi, K., Mihama, K. (1986), Jpn. J. Appl. Phys. 25, Part 2, L42. Zhu, X., Birringer, R., Herr, U., Gleiter, H. (1987), Phys. Rev. B35, 9085.
General Reading Andres, R. P., Averback, R. S., Brown, W. L., Brus, L. E., Goddard III, W A., Kaldor, A., Louie, S. G., Moskovits, M., Peercy, P. S., Riley, S. X, Siegel, R. W, Spaepen, F, Wang, Y. (1989), "Clusters and Cluster-Assembled Materials", J. Mater. Res. 4, 704. Averback, R. S., Nelson, D. L., Bernholc, X (Eds.) (1991), "Clusters and Cluster-Assembled Materials", Mater. Res. Soc. Symp. Proc. 206. Birringer, R., Herr, U., Gleiter, H. (1986), "Nanocrystalline Materials - a First Report", Suppl. Trans. Jpn. Inst. Met. 27, 43.
614
13 Cluster Assembly of Nanophase Materials
Birringer, R., Gleiter, H. (1988), "Nanocrystalline Materials", in: Encyclopedia of Materials Science and Engineering, Suppl. Vol. 1: Cahn, R. W. (Ed.), Oxford: Pergamon Press, p. 339. Gleiter, H. (1990), "Nanocrystalline Materials", Prog. Mater. Sci. 33, 223. Kear, B. H., Cross, L. E., Keem, J. E., Siegel, R. W, Spaepen, E, Taylor, K. C , Thomas, E. L., Tu, K.-N. (1989), Research Opportunities for Materials with Ultrafine Microstructures. Washington, DC: National Academy. Vol. NMAB-454. Kear, B. H., Siegel, R. W. (Eds.) (1991), "Proc. Acta Metallurgica Conf. on Materials with Ultrafine
Microstructures", Scripta Metall. et Mater., in press. McCandlish, L. E., Polk, D. E., Siegel, R. W., Kear, B. H. (Eds.) (1989), "Multicomponent Ultrafine Microstructures", Mater. Res. Soc. Symp. Proc. 132. Siegel, R. W, Hahn, H. (1987), "Nanophase Materials", in: Current Trends in the Physics of Materials: Yussouff, M. (Ed.), Singapore: World Scientific Publ. Co., p. 403. Uyeda, R. (1991), "Studies of Ultrafine Particles in Japan: Methods of Preparation and Technological Applications", Prog. Mater. Sci. 35, 1.
