ADHESION ASPECTS OF THINFILMS VOLUME 2
Edit or: K.L. Mittal
///vs P/// UTRECHTBOSTON 2005
VSP (an imprint of Brill Academic Publishers) P.O. Box 346 3700 AH Zeist The Netherlands
Tel: +31 30 692 5790 Fax: +31 30 693 2081
[email protected] www.vsppub.com www .brill.nl
0VSP 2005 First published in 2005 ISBN 90-6764-421-8
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner.
Printed in The Netherlands by Ridderprint bv, Ridderkerk
Adhesion Aspects of Thin Films. Vol. 2, pp. vii-viii Ed. K.L. Mittal 0VSP 2005
Preface This volume documents the proceedings of the International Symposium on Adhesion Aspects of Thin Films (including Adhesion Measurement and Metallized Plastics) held under the aegis of MST Conferences in Orlando, FL, December 1516, 2003. The premier symposium in this series under the exclusive title “Adhesion Aspects of Thin Films” was held in Newark, NJ, in 1999, the proceedings of which were properly documented in a hard-bound book [l]. So essentially the present event was the second symposium on this topic except that we decided to blend Adhesion Measurement and Metallized Plastics topics with the Adhesion Aspects of Thin Films, as these two topics fall within the broad purview of thin films. Apropos, in the past we had organized separate symposia on Adhesion Measurement and Metallized Plastics. Because of the tremendous current interest in thin films, the next symposium is planned for some time in December 2005. Thin films are used for a myriad of applications ranging from mundane (e.g., potato chips bags) to hightech. Thin films of all kinds of materials are used for a variety of functions and purposes in microelectronics, optics, space, nanotechnology, tribology, biomedical and so on. Irrespective of the intended function of a thin film it must adhere well to the underlying substrate. So it becomes imperative to understand the factors dictating adhesion of thin films and devise ways to control it to the desired level. Lack of required adhesion between a thin film and the substrate can lead to deleterious effects, e.g., delamination. Currently, there is tremendous R&D activity in the arena of thin films and this high tempo will continue unabated in the future, as this particular class of materials offers a great potential and in certain applications thin films are the only answer. This high level of interest can be succinctly depicted as “Thin is in”. The technical program for this symposium comprised 26 papers addressing many aspects pertaining to adhesion of thin films: mechanisms of adhesion, factors influencing adhesion, adhesion measurement, and adhesion of a host of thin film materials on a variety of substrates. There were lively (not exothermic) discussions throughout the symposium, both formally and informally. The comments from the attendees were quite positive which is a testimonial to the success of the symposium. As for this book, it contains a total of 17 papers (others are not included for a variety of reasons) addressing the latest developments relative to adhesion aspects of thin films. As recorded in the Preface to the previous volume [ 11, it must again be underscored that all manuscripts for this volume were rigorously peer reviewed, appropriately revised (some twice or thrice), meticulously edited before
viii
Preface
inclusion in this volume. So this volume is not a mere collection of unreviewed papers - which generally is the case with many, if not all, so-called proceedings volumes - but reflects the highest standard of publication. This volume is divided into three parts. Part 1 “General Papers”; Part 2 “Metallized Plastics”; and Part 3 “Adhesion Measurement”. The topics covered include: Factors influencing adhesion of thin films; relevance of stresses in thin film adhesion; surface effects on intrinsic thin film stresses; adhesion improvement; hemocompatibility of DLC coatings; thin films of various materials on a host of substrates; thin polymer films; surface modification of polymers; adhesion of metal films on polymers; investigation of metal-polymer interactions; adhesion measurement: scratch test; microscratch test; and abrasion and durability of thin films. I fervently hope this and its predecessor volume [l] will serve as a reference source for the latest information on adhesion aspects of thin films. Furthermore, yours truly hopes that anyone interested (centrally or peripherally) in thin films will find these volumes useful. Acknowledgements
Now comes the pleasant task of thanking those who helped in many and varied ways. First, I sincerely extend my “thank you” to my friend and colleague, Dr. Robert H. Lacombe, for taking care of the myriad details relative to the organization of this symposium. The comments from the reviewers (individuals behind the scenes) were extremely valuable as these most definitely improved the quality of manuscripts. I am profusely thankful to the authors for their interest, enthusiasm and cooperation without which this book could not be born. In closing, my appreciation goes to the staff of Brill (publisher) for their cooperation and efforts in producing this book.
K. L. Mittal P.O. Box 1280 Hopewell Jct., NY 12533 1. K. L. Mittal (Ed.), Adhesion Aspects of Thin Films,Vol. 1, VSP, Utrecht (2001).
Contents
Preface
vii
Part 1. General Papers Surface effects on intrinsic thin-film stresses R. C. Cammarata
3
Adhesion properties of functionally gradient diamond-like carbon nanocomposite films R. J. Narayan
13
Adhesion improvement of magnetron-sputtered amorphous carbon coating on cemented carbide S. Zhang and X . L. Bui
37
Characterization of polyethylene-metal composite thin films deposited by evaporation S. Iwarnori, F. Tateishi, Y. Ono and Y. Yamada
49
Selection of efficient coatings for milling Inconel 7 18 based on their adhesion properties 0. Knotek, E. Lugscheider, K. Bobzin, C. Piiiero, F. Klocke, D. Lung and J. Grams
57
Investigation of tissue compatibility and hemocompatibility of DLC and CN, coatings D. J. Li and L. F. Niu
69
A study on structural characterization of and cell attachment to Ti-containing coatings Y. Liu, S. Liu, Q. X . Liu and D. J. Li
79
Adhesion issues with polymer/oxide barrier coatings on organic displays D. W. Matson, P. M. Martin, G. L. Gra8 M. E. Gross, P. E. Burrows, W. D. Bennett, M . G. Hall, E. S. Mast, C. C. Bonham, M . R. ZumhofJ; N. M. Rutherford, L. M. Moro, M. Rosenblum, R. F. Praino and R. J. Visser
91
Contents
vi
Part 2. Metallized Plastics Surface modification of polymers by ion-assisted reactions: An overview J. S. Cho, S. Hun, K. H. Kim, Y. G. Hun, J. H. Lee, C. S. Lee, J. W. Sung, Y. W. BeagandS. K. Koh
105
Contribution of chemical interactions between A1 atoms and different types of functional groups to the adhesion of AI-polymer systems R. Mix, G. Kiihn and J , Friedrich
123
Deposition of aluminum on three-dimensional polymeric substrates 0. Knotek, E. Lugscheider, K. Bobzin, M. Maes and A. Kramer
145
Improvement of metal adhesion to silicone films: A ToF-SIMS study A. Delcorte, S. Befahy, C. Poleunis, M . Troosters and P. Bertrand
155
Mechanical stability of a Ti02 coating deposited on a polycarbonate substrate M. Ignat, S. Ge'tin, B. Latella, C. Barbe' and G. Triani
167
Part 3. Adhesion Measurement Advances in adhesion measurement good practice: Use of a certified reference material for evaluating the performance of scratch test instrumentation N. M. Jennett, R. Jacobs and J. Meneve
179
Film hardness effect on adhesion strength of Ti02 film on a glass substrate measured by the scratch test A. Kinbara, E. Kusano and H. Nanto
195
Two critical events observed on Cu films on glass substrate in the microscratch test S. Baba, Y. Yamaguchi, M. Ogawa and T. Nakano
203
Abrasion life and scratch durability of sputtered PTFE thin film S. Iwamori, Y. Nagayama, Y. Yamagata and Y. Yamada
215
Part 1 General Papers
Adhesion Aspects of Thin Films, Vol. 2. pp. 3-12 Ed. K.L. Mittal 0VSP 2005
Surface effects on intrinsic thin-film stresses ROBERT C. CAMMARATA1'*'* 'Department of Materials Science and Engineering, Johns Hopkins University, Baltimore, MD 21218, USA 2Department of Mechanical Engineering, Johns Hopkins University, Baltimore, MD 21218, USA
Abstract-The origins of intrinsic stresses in thin films are discussed with emphasis on those mechanisms associated with surface and internal boundaries. Such stresses can be quite large, leading to a variety of effects, including de-adhesion. The structure and thermodynamics of surfaces in thin films are briefly reviewed, and then it is shown how surface thermodynamic parameters can be used to describe a variety of intrinsic stress behaviors.
Keywords: Intrinsic stress; surface thermodynamics: thin films.
1. INTRODUCTION
A thin solid film grown on a solid substrate is generally deposited in a state of stress. This stress can be quite large, often exceeding the yield stress of the material in bulk form, and can lead to deleterious effects such as cracking, spalling and de-adhesion. However, it is sometimes necessary, or even desirable, for a thin film to be under stress. For example, it is generally required for electronic material applications that a semiconductor film be grown epitaxially, i.e., as a single crystal deposited on a single crystal substrate with defect-free lattice matching at the filmsubstrate interface. If the in-plane equilibrium lattice spacings of the film and substrate are different, the film will be under stress in order to achieve this lattice matching. Another example where an intrinsic stress is desirable concerns a material that has a thin film coating in a state of compressive stress that can result in enhanced fracture and fatigue resistance compared to the uncoated material. 2. SURFACES AND INTERNAL BOUNDARIES IN THIN FILMS
A principal microstructural feature of thin films is the high density of surfaces relative to conventional bulk materials. In addition to the film-substrate interface and the free surface (solid-vapor interface) of the film, there can be grain bounda"Tel.: (1-410) 516-5462: Fax: (1-410) 516-5293: e-mail:
[email protected]
4
R. C. Cammarata
ries in polycrystalline films and interlayer interfaces in multilayered thin films. These surfaces can have a significant effect on the mechanical behavior of thin films in general and the intrinsic stress in particular. As mentioned above, there can be an epitaxial relationship between a thin film and the substrate, leading to lattice matching at the film-substrate interface. If the lattice matching is perfect, resulting in a defect-free interface, the interface is referred to as coherent. If the film and substrate have different equilibrium lattice spacings, and the substrate is much thicker than the film, the film will have to be coherently strained in order for it to be in perfect atomic registry with the substrate. Let af and a, denote the bulk equilibrium in-plane lattice spacings of the film and substrate, respectively. The misfit between the film and substrate is defined as m = ( a , - af)laf.For a film perfectly lattice matched to the substrate, the in-plane coherency strain is equal to the misfit. As long as the misfit is not too large, it is generally possible to grow a coherently strained epitaxial film, at least at smaller thicknesses [l].There is a critical thickness above which it is thermodynamically favorable for the film to elastically relax, resulting in a loss of the perfect lattice matching at the film-substrate interface. One way in which this can occur is by the formation of an array of dislocations at the film-substrate interface that can accommodate some or all of the misfit. As long as the spacing of these misfit dislocations is not too small, so that there is still a significant amount of residual lattice matching at the film-substrate interface, the interface is said to be semicoherent. If the misfit dislocation spacing is less than a few lattice spacings, or if there is no epitaxial relationship between the film and substrate, the interface is called incoherent. 3. SURFACE THERMODYNAMICS
Consider a solid-vapor interface such as the free surface of a thin film. There are two thermodynamic quantities associated with the reversible work to change the area of the surface [2]. One of these is the surface free energy which can be defined by setting the reversible work to create new surface of area A equal to @. The other surface thermodynamic quantity is the surface stress tensor &, which can be defined by setting the reversible work to introduce a surface elastic strain d q on a surface of area A equal to AA;,deJ.For simplicity, it will be assumed that the surface stress is isotropic and can be taken as a scalarf(this is valid for a surface which displays a three-fold or higher rotational symmetry). The surface free energy and surface stress are related to each other by the Shuttleworth-Herring equation [2]: where & is the in-plane linear surface strain. Unlike the surface free energy, which must be positive (otherwise solids would spontaneously cleave) the surface stress can be positive or negative. Experimental measurements and theoretical calcula-
Surface effects on intrinsic thin-film stresses
5
tions for the low index surfaces of many metals, semiconductors, and ionic solids give positive values for the surface free energy and surface stress of order 1 N/m. For finite size solids in mechanical equilibrium, the surface stress will induce a volume elastic strain relative to a bulk solid [2]. For a spherical solid of radius r, this will result in a pressure difference AP (called the Laplace pressure) between the solid and the surrounding vapor given by
AP= P, - P , = 2flr,
(9) where P , and P, denote the pressures of the solid and vapor, respectively. Similarly, for a thin disk of thickness t, a surface stressfacting on the top and bottom surfaces will result in a radial Laplace pressure of 2flt. Because of this Laplace pressure, the lattice spacing in the interior of the solid at equilibrium will be different from the equilibrium bulk spacing. Using Hooke’s law, this difference in lattice spacing can be described in terms of the in-plane elastic strain: E = -2flYt,
(2)
where Y is the biaxial elastic modulus, equal to E/(1 - v), where E is the Young modulus and vis the Poisson ratio. As with a free solid surface, a solid-solid interface, such as that between a thin film and the substrate, has an associated surface free energy that will be referred to as the interface free energy. Since the phases on either side of a solid-solid interface can be independently strained, resulting in different strain states at the interface, there are two surface stresses that can be associated with this interface [2]. These surface stresses for solid-solid boundaries will be referred to as interface stresses. For a thin film-substrate interface, it is convenient to define these interface stresses in the following manner. Let E’ represent an interface strain associated with an in-plane deformation of the film keeping the substrate fixed. Such a strain would lead to a change in the misfit dislocation density at a semicoherent interface. An interface stress g associated with this type of deformation can be defined by talung the reversible work to strain an interface of area A by an amount dE‘ equal to Ag dE‘ . Consideration is now given to an interface strain resulting from deforming the film and substrate by the same amount in the plane of the interface; let this strain be denoted as e ” . An interface stress h associated with this type of deformation can be defined by taking the reversible work to strain an interface of area A by an amount de” equal to Ah de“. Suppose a film-substrate system with a semicoherent film-substrate interface has a misfit m. Let af* be the in-plane lattice spacing of the strained film. The coherency strain can be defined as gC= (af* - af)/af.If the film is fully relaxed, so that it has its bulk equilibrium lattice spacing af, = 0; if the interface is completely coherent, gC = m. A simple model [1] for the interface free energy T o f a semicoherent interface of misfit m and coherency strain leads to the expression
6
R. C. Cammarata
Figure 1. Thin film growth modes. (a) Volmer-Weber (island growth); (b) Frank-van der Merwe (layer-by-layer growth); (c) Stranski-Krastanov (layer-by-layer followed by island-like growth).
r=To(1 - Glm),
(3)
where rois the interface free energy when the film is completely relaxed. Based on this model, the following approximate expression for the interface stress g has been given [ 2 , 31: g = -ro/2m.
(4)
It should be noted that for Iml << 1, Igl >> To.A similar analysis for the interface which is stress h indicates that for a semicoherent interface, h is of order -10 ro, consistent with experimental measurements for metal-metal semicoherent interfaces [2]. 3.1. Growth modes Three basic thin-film growth modes [4] have been identified (see Fig. 1) that can be associated with relationships involving the values of surface free energies. VolmerWeber growth involves three-dimensional growth of islands that eventually coalesce to form a continuous film. This growth mode is favored when < 2 + C
x
Surface effects on intrinsic thin-film stresses
7
where 7.; and are the surface free energies of the free surface of the film and substrate, respectively. This growth mode is often encountered when there is no epitaxial relationship between the film and substrate (for example, crystalline metal films deposited on amorphous substrates). The Frank-van der Merwe mode is a two-dimensional, layer-by-layer growth mode and represents “ideal” epitaxial growth. The Stranski-Krastanov mode involves two-dimensional growth for one or two monolayers, followed by three-dimensional island-like growth. The switch-over from two-dimensional to three-dimensional growth appears to be related to effects of stress relaxation. Frank-van der Merwe and Stranski-Krastanov growth modes are favored when 7.: 2 E + C and are often associated with epitaxial growth. 4. PHYSICAL ORIGIN OF THIN FILM STRESS
4.1. Epitaxial growth For layer-by-layer epitaxial growth, it is expected that the film stress is principally a result of coherency strains when a film with an in-plane equilibrium lattice spacing different from that of the substrate is completely latticed-matched with the substrate at the film-substrate interface. As long as the lattice spacing misfit is not too large, the film-substrate interface will be completely coherent during the initial stage of growth. From a thermodynamic point of view, this is because the work to form misfit dislocations at the interface is greater than the work to elastically strain the film to accommodate the misfit. Since the volume strain energy of the coherent film is proportional to the film thickness, there will eventually be a critical thickness above which it is thermodynamically favorable to introduce misfit dislocations at the interface to relieve some of the elastic strain energy. An expression for the critical thickness t, can be given in terms of the surface thermodynamic quantities and thin film elastic modulus as [2, 31:
wheref, and 7.: are the surface stress and surface free energy of the film, respectively. Usually the contribution of the free film surface has been ignored, which in many (but not all) cases is an acceptable approximation. Assuming this to be true, it is possible to calculate the critical thickness by substituting equations (3) and (4) into equation ( 5 ) and employing an expression for r, that involves the selfenergies for an array of interface dislocations which completely accommodates the misfit. If the elastic moduli for the film and substrate can be taken as approximately the same, such a model leads to the following expression for the critical thickness [1-4]: tc = [b ln(t,lb
+ l)]/Sz(l + v)m,
where b is the Burgers vector of the misfit dislocations.
(6)
R. C. Cammarata
8
L
cI
7.5
t
C
~~
a n 5.0 c3
0a 5
E /Th = 0.16
-0.5
i
U
0
.i4
44
b
b
-1 .o -1 =5
0
25
50
75
Thickness (nm) (d) h
E cI 1 a n o
E c
-
a -
(3 44
b
10
Thickness (nm)
(c ) h
-
5
0
-1
w
b
-2
-1
0
10
20
Thickness (nm)
0
100
0
4000
Thickness (nm) Time(s)
Figure 2. Real-time wafer curvature measurements during ultrahigh-vacuum deposition onto amorphous SiOz [ 5 ] : (a) polycrystalline Ag; (b) amorphous Ge; (c) polycrystalline Si; (d) polycrystalline Al. Th is the ratio of the deposition temperature to the melting temperature. For Al, the stress generation as a function of thickness during growth and the stress relaxation as a function of time after deposition was halted at a film thickness of 200 nm are shown.
The process of stress relaxation can involve the nucleation of misfit dislocations at the interface or slip of pre-existing dislocations. It is often found experimentally that it is possible to grow a completely lattice matched film to thicknesses greater than the critical thickness. This is presumably a result of kinetic limitations associated with the stress relaxation process. 4.2. Nonepitaxial island growth Results [ 5 ] from recent experiments investigating the development of thin film stress during Volmer-Weber (island) growth by ultrahigh vacuum evaporation onto an amorphous substrate for a variety of film materials are shown in Fig. 2. It
Surface effects on intrinsic thin-film stresses
9
is seen that the stress behavior, which occurs for both crystalline and amorphous films, generally involves an initial compressive stress, followed by a tensile jump. and then back to a compressive stress. This has been termed “compressivetensile-compressive’’ (CTC) behavior. The initial compressive stress regime occurs during the formation and growth of islands before coalescence. The rapid tensile rise initiates around the onset of coalescence and reaches a maximum when the film becomes continuous. The final compressive stage occurs during further growth of the continuous film. Different mechanisms that have been proposed for each regime that lead to permanent (static) contributions to the film stress will be reviewed below. It should be noted that there is often a large dynamic contribution to the stress during growth that relaxes when the deposition is halted (see Fig. 2d). This effect, which has been attributed to adatom effects [5], is not discussed here. Consideration is first given to the early stage of island growth [5-71. An isolated island can be modeled as a disk of diameter d and thickness t. As discussed earlier, surface stresses acting on this disk exert a size-dependent Laplace pressure that results in an equilibrium lattice spacing different from the bulk equilibrium lattice spacing. Let do and to represent the size of an island when it first becomes firmly attached to the substrate so that a film stress can be generated. As the island grows, the equilibrium spacing will change, but the island is constrained by the substrate not to deform laterally. Thus, the change in equilibrium spacing leads to a latent strain that is manifested as a film stress. A simple elasticity analysis gives the following approximate expression for the island stress [6, 71:
o= cfs+ h)(llt - l/to)+ pfi
(6)
where fi is the free surface stress for the curved surface of the island and p depends on the elastic constants of the disk; for an elastically isotropic island, p = (1 - 3 v)/(1 - v). It is generally expected that the first term on the right-hand side of equation (6) will dominate and that (f, + h) will be positive. As a result, the firmly attached islands will produce a compressive film as they grow, consistent with experiment. It is generally agreed that the rise in tensile stress observed in the different systems shown in Fig. 2 is correlated with the coalescence of islands. This is consistent with a model that has been popular over the years that involves tensile stress generation resulting from relaxation of grain boundaries formed during coalescence [8-111 and which has been the subject of recent reinterpretations and reformulations [12-141. When two islands impinge upon each other, a grain boundary is formed and two free surfaces disappear. This results in a lowering of the surface free energy by an amount Ay= (27.; - ygb), where X b is the interface free energy for the grain boundary. As the distance between neighboring growing islands is reduced, there will be a critical interaction distance where it will be thermodynamically favorable for the islands to elastically deform and impinge to form a grain boundary. The elastic strain energy created when the islands impinge
10
R. C. Caminarata
is compensated by the surface free energy reduction A y If the islands are modeled as having impinging faces that are flat, the contribution to the film stress can be approximated as [ 1 1 ,121 D = (EAjd,)lt2,
(7)
where d , is the size of the grain boundary formed at impingement. Using E = 100 GPa, A y= 1 Nlm and d , = 100 nm, equation (7) gives a value for the stress contribution of about 1 GPa. A more sophisticated analysis employing Hertzian contact mechanics [ 131 gives the following expression for the average stress contribution resulting from the coalescence of a square array of hemispherical islands of diameter d,: D = 4Ajd,.
(8)
Substituting the values given above into equation (8) gives D = 40 MPa, consistent with results from a finite element model [ 151 and is of order what is often observed experimentally. The values that can be obtained from equations (7) and (8) should be viewed as approximate upper and lower limits for the stress contribution from the grain boundary relaxation mechanism. Significant grain growth is sometimes observed during and after island coalescence [16, 171. This process is driven by the reduction in grain boundary area per unit volume of the film because the surface free energy of a grain boundary is positive, and thus reduction in the grain boundary area reduces the free energy of the film. Since a region near a grain boundary is expected to display a lower atomic density than the interior of the grain, the elimination of grain boundaries leads to a negative latent strain and, therefore, a tensile film stress. An approximate expression for the stress generated by this mechanism in a film as a function of the average grain size d is [ 16, 171 D = Ew( lld, - lld),
(18) where w is the excess volume per unit area associated with a grain. When d >> d,, this stress approaches the value of Ewld,. Using values of E = 100 GPa, d, = 100 nm and w = 0.1 nm, this leads to a value of Ewld, = 100 MPa, suggesting that grain growth can be a significant contributor to the tensile component of the film stress. It should be noted that an increased grain size often leads to a reduced flow stress, so that increasing the grain size may allow for stress relaxation by plastic flow. Unlike the initial compressive stress regime during island growth and the tensile stress stage during coalescence, where plausible models have been proposed to explain these behaviors, the origin of the compressive stress generated after coalescence is less clear. One possibility is that it is a continuation of the initial compressive stress that was being generated during island growth that was temporarily masked by the tensile jump during island coalescence. Considering only the
Surjace effects on intrinsic thin-film stresses
11
first term on the right-hand side of equation (6), it is noted that this mechanism leads to an asymptotic stress value of -(fs + h)/t,. If this asymptotic value is larger in magnitude than the tensile stress generated by the grain boundary relaxation mechanism, a superposition of the general compressive behavior resulting from surface stress effects with a step-function-like tensile jump at coalescence can qualitatively explain the CTC behavior [ 5 ] .Another proposed mechanism for this late stage compressive stress involves the incorporation of surface adatoms into grain boundaries [ 5 ] .The diffusion of atoms into the grain boundaries is driven by the supersaturation of adatoms that exists during deposition. More studies need to be conducted before a clear understanding of the late stage compressive behavior is understood. 5. CONCLUSIONS
Surfaces play an important role in the growth and stress behavior of thin films. The growth mode is determined by a balance of surface free energies, and will greatly influence the generation of intrinsic stresses. The formation of a coherent (lattice-matched) interface between a film and substrate with different equilibrium lattice spacings will result in a coherency stress characteristic of epitaxially grown films. In the case of films displaying (nonepitaxial) island growth behavior, several mechanisms have been proposed to explain the “compressive-tensilecompressive” stress evolution observed during vacuum evaporation. While a large amount of the stress results from dynamic processes during deposition, and can be relieved when deposition is halted, there is generally a significant residual stress. The early stage compressive residual stress, generated prior to island coalescence, can be understood as resulting from a Laplace pressure owing to surface stresses. The tensile jump, which occurs during island coalescence, is generated when islands that are separated by a certain critical distance elastically deform in order to impinge on each other and form a grain boundary. The driving force for this impingement is the lowering in surface energy when a grain boundary is formed and two free surfaces disappear. Grain growth, driven by the resulting reduction in grain boundary energy, can also contribute to a tensile stress contribution. The origin of the compressive stress generated after the tensile jump is not completely clear, but may be due to a variety of proposed mechanisms, including ones involving surface stresses or incorporation of adatoms at grain boundaries. Acknowledgements
The author gratefully acknowledges support from the National Science Foundation as administered through the Materials Science and Engineering Center at Johns Hopkins University.
12
R. C. Cammarata
REFERENCES 1. J. W. Matthews, in Epitaxial Growth, Part B, J. W. Matthews (Ed.), p. 559. Academic Press, New York, NY (1975). 2. R. C. Cammarata, Prog. Surj? Sci. 46, 1 (1994). 3. R. C. Cammarata, K. Sieradzki and F. Spaepen, J. Appl. Phys. 87, 1227 (2000). 4. M. Ohring, The Materials Science of Thin Films, p. 413. Academic Press, Boston, MA (1992). 5. J. A. Floro, E. Chason, R. C. Cammarata and D. J. Srolovitz, MRS Bull. 27, 19 (2002). 6. R. C. Cammarata, T. M. Trimble and D. J. Srolovitz, J. Mater. Res. 15, 2468 (2000). 7. R. C. Cammarata, in Adhesion Aspects of Thin Films, Vol. 1, K. L. Mittal (Ed.), p. 31. VSP, Utrecht (2001). 8. R. W. Hoffman, Phys. Thin Films 3, 21 1 (1966). 9. R. W. Hoffman, in Physics of Non-Metallic Thin Films, NATO Advanced Study Institute Series B, Vol. 14, C. H. S. Dupuy and A. Cachard (Eds.), p. 273. Plenum, New York, NY (1976). 10. F. A. Dojack and R. W. Hoffman, Thin Solid Films 12,71 (1972). 11. H. K. Pulker, Thin Solid Films 89, 191 (1982). 12. W. D. Nix and B. M. Clemens, J. Mater. Res. 14,3471 (1999). 13. L. B. Freund and E. Chason, J. Appl. Phys. 89,4866 (2001). 14. B. W. Sheldon, A. Lau and A. Rajamani, J. Appl. Phys. 90,5097 (2001). 15. S. C. Seel, C. V. Thompson, S . J. Hearne and J. A. Floro, J. Appl. Phys. 88,7079 (2000). 16. P. Chaudhari, J. Vac. Sci. Technol. 9, 520 (1972). 17. M. F. Doerner and W. D. Nix, CRC Crit. Rev. Solid State Muter. Sci. 14, 225 (1988).
Adhesion Aspects of Thin Films, Vol. 2. pp. 13-36 Ed. K.L. Mittal 0VSP 2005
Adhesion properties of functionally gradient diamond-like carbon nanocomposite films ROGER J. NARAYAN* School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0245, USA
Abstract-Diamond-like carbon (DLC) is an amorphous material with a high fraction of sp3-hybridized carbon atoms. DLC exhibits hardness, wear resistance and chemical inertness properties close to those of diamond. IJnfortunately, DLC films delaminate due to internal compressive stress. This paper describes processing and characterization of functionally gradient diamond-like carbonmetal nanocomposite films on Ti-6A1-4V alloy, which is commonly used in biomedical and aerospace applications. Internal stresses in diamond-like carbon thin films were reduced via incorporation of elements that form carbides (e.g., silicon and titanium), as well as incorporation of elements that do not form carbides (e.g., copper and silver). These materials were produced using a novel pulsed laser deposition process that incorporates a multicomponent rotating target. In addition, functionally gradient DLC-silver and DLC-titanium films of approx. 1 pm thickness were deposited on Ti-6A1-4V alloy. Transmission electron microscopy of the DLC-metal nanocomposite films revealed that these films self-assembled into particulate or layered nanocomposite structures that possessed a high fraction of sp'-hybridized carbon atoms. Scratch testing demonstrated good adhesion of the DLC-metal nanocomposite films to Ti-6A1-4V substrates. Nanoindentation testing of the DLC-metal nanocomposite films demonstrated that these films possessed high hardness and Young's modulus values of approximately 35 GPa and 350 GPa, respectively. Wear testing using a CSM Linear Tribometer demonstrated wear lifetimes in excess of 300 000 cycles. These DLC-metal nanocomposite films can be optimized for specific medical applications: for example, DLC-silver nanocomposites have been shown to possess antimicrobial properties. Keywords: Diamond-like carbon: pulsed laser deposition; functionally gradient materials; antimicrobial coatings.
1. INTRODUCTION
Biomedical researchers have created advanced materials over the past three decades by selecting bulk materials with appropriate fracture toughness, bulk modulus and durability, and performing surface modification to improve biocompatibility, wear resistance and corrosion resistance. One biomaterial coating with tremendous potential is diamond-like carbon (DLC). -Tel.: (1-404) 894-2823; Fax: (1-404) 894-9140; e-mail:
[email protected]
14
R. J. Narayan
The term diamond-like carbon (DLC) describes hydrogen-free hard carbon solids that possess a cross-linked, non-crystalline network of sp2- and sp3-hybridized carbon atoms [l]. Films of DLC are found to have outstanding hardness, in the range of 15 GPa. Friction and wear coefficients of DLC are lower than those of diamond, and are among the lowest recorded to date (static coefficient of friction = 0.006). DLC also offers transparency to light ranging from deep ultraviolet to far infrared. In addition, DLC films are amorphous, atomically smooth and do not contain open corrosion paths to the underlying substrate. The transparency, hardness, Young’s modulus, adhesion and residual compressive stress in diamond-like carbon thin films can be tailored by controlling the sp3/sp2ratio. DLC has been accepted as an ideal coating material for use in implantable medical devices over the last 10 years. Allergy, carcinogenicity, wear, corrosion, oxidation and metal ion release can all be eliminated by coating a metal or polymer implant with DLC. DLC thin films have current and potential applications in cardiovascular (cardiac stent and heart valve), orthopaedic (knee joint and hip joint) and ophthalmic (intraocular lens and artificial retina) areas. DLC has been shown by a number of investigators to be fully biocompatible with all cell types. Human fibroblasts, human macrophages, human monocytes, human placenta endothelial cells, neural cells, and 3T3 Balb/c cloned cells grown on DLC coatings exhibit normal cellular growth activities [2-51. DLC does not cause blood coagulation. For example, Jones et al. [6] deposited DLC coatings on Ti-6A1-4V alloy using plasma-assisted chemical vapor deposition, and they subsequently placed rabbit blood platelets on these coatings. DLC coatings did not cause hemolysis, platelet activation, or thrombus formation. In addition, fibrinogen adsorption on DLC is much lower than fibrinogen adsorption on metals or polymers commonly used in blood-contacting applications. Tissues generally react to non-living implants by forming a non-adhering fibrous capsule that “walls off’ or isolates the implant from the host. Several vascular and cellular processes, including fibroblast proliferation, collagen synthesis and blood vessel proliferation, lead to the formation of an avascular connective tissue capsule. The connective tissue capsule consists of several different cellular layers, including an inner layer of macrophages, a concentric layer of fibrous tissue and fibroblasts (30-100 pm) and an outer vascularized tissue layer. Many metals (including steel, Co-Cr-Mo alloy and Ti-6A1-4V alloy) and polymers (including poly(methy1 methacrylate), polyurethane and polyethylene) trigger the formation of relatively thick interfacial layers [7, 81. DLC films elicit a minimal or nonexistent fibrous capsule; such minimal encapsulation ensures that medical device function will not be diminished. In addition, a DLC film hermetically seals an implant, preventing the release of metal ions or monomer to surrounding tissues. Poor adhesion is the sole practical limitation preventing widespread application of DLC thin films. DLC films commonly possess large internal compressive stresses that can exceed 10 GPa. The magnitude of these internal stresses can be correlated with the fraction of sp3-hybridized carbon atoms [ 11. Both internal com-
Adhesion properties of DLC-metal nanoconiposire films
15
pressive stress and diamond-like bonding are thought to result from the shallow implantation or “subplantation” of energetic carbon ions during DLC film growth. Friedmann et al. [9] recently reported preparation of thick stress-free DLC films with hardness values close to those of diamond. DLC thin films (100-200 nm) were deposited at room temperature and subsequently annealed at 600°C for 2 min to relieve internal compressive stress. These films were then cooled to room temperature to allow for further deposition. Adherent films of 1.2 pm thickness with residual stresses less than 0.2 GPa were produced using this method. Both Raman spectroscopy and electron energy loss spectroscopy data from single-layer annealed specimens revealed only subtle microstructural and chemical changes compared with unannealed films. The main advantage of this thermal annealing technique for compressive stress reduction is that pure DLC films can be prepared. However, there are several drawbacks to high-temperature annealing of DLC films. First, no polymer substrate can undergo this annealing process. Also, Kustas et al. [lo] have pointed out problems associated with overtempering of metals used in tribological and medical applications. Alternatives to high-temperature annealing must be found in order to deposit DLC thin films on metals and polymers used in medical devices. We propose the use of functionally gradient films, in which the concentration of metal systematically varies from one film interface to the next (Fig. 1). A low concentration of metal is desired near the film surface to maximize hardness, Young’s modulus and wear resistance. On the other hand, a high concentration of metal is desired near the film-substrate interface to improve adhesion with the substrate and to reduce internal compressive stress. We have developed functionally gradient DLC-metal nanocomposite films using a novel single-target pulsed laser deposition technique. Functionally gradient DLC-metal nanocomposite films were developed with biofunctional metals. For example, silver exhibits anti-microbial and antiinflammatory properties. Nanocrystalline silver has demonstrated an unsurpassed anti-microbial spectrum, with anti-microbial function against 150 different pathogens. In addition, nanocrystalline silver provides broad-spectrum fungicidal action [ l l , 121.
DLC film
Ti-6AI-4V substrate Figure 1. Functionally gradient DLC film design. The high metal atom Concentration at the filmsubstrate interface provides improved adhesion to the underlying substrate. The low metal atom concentration at the film surface pro\ides maximal hardness and Young’s modulus values at the load bearing interface.
16
R. J. Narayan
Transmission electron microscopy was used to determine the film microstructure. Rutherford backscattering spectroscopy, X-ray photoelectron spectroscopy, visible Raman spectroscopy and electron energy loss spectroscopy were used to assess carbon-bonding characteristics. Scratch adhesion testing, nanoindentation testing and wear testing were performed in order to determine the mechanical and tribological properties of these films. Anti-microbial testing was performed in order to assess the anti-microbial properties of the DLC-silver nanocomposite film. 2. EXPERIMENTAL
Ti-6A1-4V substrates were prepared by polishing 2 cm x 2 cm x 2 mm pieces of stock alloy. The substrates were then cleaned with acetone and methanol for 10 min each in an ultrasonic cleaner. The Ti-6A1-4V substrates were placed onto the substrate holder of the pulsed laser deposition (PLD) ultrahigh vacuum (UHV) chamber. Silicon (100) substrates were cut from 4-inch wafers, cleaned successively with acetone and methanol in an ultrasonic cleaner, and etched in 10% hydrofluoric acid for 5 min to remove surface silicon oxide. The silicon substrates were placed alongside the Ti-6A1-4V substrates, allowing for simultaneous deposition on both materials. The sample holder was loaded into the PLD chamber, shown schematically in Fig. 2. A high-purity graphite pellet was used as the target, and its surface was partially covered by 1-3 small pieces of various metals, including silver, copper, titanium and silicon (Fig. 3). A Lambda Physik LPX 200 KrF excimer laser was used for target ablation (A = 248 nm, pulse duration = 25 ns). The depositions were conducted for 40 min at room temperature at a chamber pressure of Torr. The target-to-substrate distance was maintained at 4.5 cm. The graphitemetal target was rotated at 5 rpm. The energy density of the laser pulse was approx. 3-5 J/cm2 and the laser repetition rate was 10 Hz. In initial depositions, small amounts of metal ( ~ 5 % were ) alloyed with diamond-like carbon. Later, several functionally gradient DLC-metal nanocomposite films were deposited on polished Ti-6A1-4V alloy. The position of the laser spot on the target was adjusted manually over the course of the deposition, such that a gradual change in the graphite/metal ablation ratio was achieved. The different deposition conditions and time intervals are displayed in Table 1. The sample names are denoted by the metal component of the DLC-metal nanocomposite. The number of metal pieces on the graphite target and the relative amount of metal ablation are provided. Much larger amounts of metal were introduced in these functionally gradient DLC-metal films than in the initial films. Structural characterization was performed on the thin films deposited on silicon substrates. Several cross-sectional transmission electron microscopy samples were prepared. These samples were examined in a Topcon 002B unit with a point resolution of 0.19 nm at 200 kV. Radial distribution function (RDF) analysis from
Adhesion properties of DLC-metal nanocornpositefilms
Thermocouple Feedthrough Target Controls and N;I feedthrough (cooling) +
1
4 -
Substrate Shutter and Controls
El e c t r i c a l Feedthr ough
Ionizab on
To High Vacuum Pump
Excim er Laser Beam
Figure 2. Pulsed laser deposition system.
Metal Piece
Graphite Target
Figure 3. Schematic of target configuration used in this study.
17
18
R. J. Narayan
the electron diffraction pattern was used to obtain short-range structural information. Film morphology and crystallinity were determined by high-resolution transmission electron microscopy. Scanning transmission electron microscopy (STEM) was carried out in a VG HB501 UX unit with a point resolution of 0.13 nm at 100 kV. Heavy metal atoms in the DLC matrix can be studied in detail, since scattering power or contrast depends upon atomic number squared (Z2).Parallel electron energy loss spectroscopy was used to obtain information about the carbon bonding in the DLC-metal nanocomposite films; loss spectra were collected from zero up to 1000 eV energy loss. The chemical composition of the DLC-metal nanocomposite films was determined using Rutherford backscattering spectrometry (RBS) and X-ray photoelectron spectroscopy (XPS). Raman spectroscopy was performed to assess bonding configuration and internal compressive stress within the DLC and DLC-metal nanocomposite films. Mechanical and tribological testing was performed on functionally gradient DLC-silver and functionally gradient DLC-titanium nanocomposite films. An MTS Nanoindenter XP system was used for nanohardness and Young’s modulus measurements on the functionally gradient DLC films. A DCM (dynamic contact module) head was used in these tests. The adhesion of the functionally gradient DLC-metal nanocomposite films was determined using a CSM Microscratch Instrument (MicroPhotonics, Irvine, CA, USA). These scratch tests were performed under a linearly increasing load; the maximum load used was 1.5 N. The scratch length was set to 3 111111, and the scratch speed was set to 3 d m i n . A diamond tip (20 pm tip diameter, Rockwell C geometry) was used for these tests. Wear properties of the functionally gradient DLC-metal nanocomposite films were obtained using a CSM Instruments Linear Tribometer. Alumina and 100Cr6 bearings of 6 mrn diameter were used as static wear partners. The amplitude of the wear track was set to 6 mm, and the scratch speed was set to 3 c d s . Normal loads of 3, 7 and 10 N were applied. Testing was performed under dry conditions (in ambient atmosphere) and in Ringer’s lactate USP solution (Baxter Health-
Table 1. Functionally gradient DLC-metal nanocomposite film deposition parameters Sample
Target
Target set-up
Metal fraction during deposition
FGAgl FGAg2 FGAg3 FGAg4 FGTi1 FGTi2 FGTi3
Graphite/Silver Graphite/S ilver Graphite/Silver Graphite/Silver Graphite/Titanium Graphite/Titanium Graphite/Titanium
1 large piece 1 large piece 2 small pieces 2 small pieces 2 small pieces 3 small pieces
50%-45%+40%-35%-32%-25% 40%-36%-33%” 29~-2S%~20% 40% -36% +33 % -28% -23% +20% 30%+23%+20%+13%+10% 60%+42%+38%+32%+25%+ 12% 60%+46% +42%+30%-+27%+20% 30%+26%-22%-20%-1S%-7%
1 small piece
Adhlcsiorz properties of DLC-metal nanocornposite fililzs
19
care). Finally, anti-microbial testing was performed on a DLC-silver nanocomposite film and a silicon (100) piece using a modified disk diffusion test. 3. RESULTS AND DISCUSSION
3.I . Microscopy An optical micrograph of a DLC film on Ti-6A1-4V is shown in Fig. 4. The bright contrast corresponds to the buckling edges. A regular, sinusoidal pattern was noted throughout the film. The buckling appears to originate at the specimen edges and other defects, and spreads quickly over the entire film. Buckling appeared at film thicknesses exceeding 50 nm and buckling size increased with film thickness. The generation of buckling in DLC films is also related to post-deposition environmental factors. No buckling occurred as long as the films were kept in a vacuum. Exposure of the film to humidity or other gaseous species initiated the buckling process. This delamination process was also accelerated by an increase in ambient humidity. It has been proposed that gas atoms diffuse into the interface between the film and the substrate, and initiate the delamination process. These factors suggest internal compressive stress is the source of DLC film delamination. It is interesting to note that the DLC film on silicon (100) does not immediately delaminate, while DLC film on Ti-6A1-4V delaminates immediately after exposure to ambient humidity. DLC on silicon (100) forms a silicon carbide interfacial
Figure 4.Optical micrograph showing the buckling pattern of DLC film on Ti-6A1-4V alloy.
20
R. J. Narayan
layer and DLC on Ti-6A1-4V forms a titanium carbide interfacial layer. The silicon carbide interfacial layer provides better DLC film adhesion than the titanium carbide interfacial layer. This result suggests the importance of interfacial bonding in promoting adhesion of a DLC film to a given substrate. Optical micrographs of DLC-copper, DLC-titanium and DLC-silicon films are shown in Fig. 5. No buckling is observed in these films, which were placed in ambient humidity for an extended period of time (>24 h). It appears that internal compressive stresses are minimized in these DLC-metal films. These DLC-metal films are atomically smooth; however, some micrometer-sized particulates are observed in the DLC-titanium and DLC-silicon films. Transmission electron microscopy was also performed on the DLC-metal nanocomposite films. The DLC-copper composites were notable in that they appeared speckled (Fig. 6). This speckling indicates segregation of copper into a separate phase. DLC-silver and DLC-platinum composites also demonstrated separate phases. On the other hand, DLC-titanium and DLC-silicon composites did not demonstrate separate phases.
Figure 5. Optical micrographs of (a) DLC-copper, (b) DLC-titanium and (c) DLC-silicon nanocomposite films.
Figure 6. (a) High-resolution transmission electron micrograph of DLC-1.4 at% copper nanocomposite film, (b) corresponding electron diffraction pattern.
Adhesion properties of DLC-metal nanocomposite films
21
3.2. Z-contrast scanning transmission electron microscopy Z-contrast scanning transmission electron microscopy provides unique information on nanostructured composite materials. In the VG HB501 UX scanning transmission electron microscope, an image is formed by scanning a 2.2 A probe across the sample. The Z-contrast signal is collected from a high angle annular detector, and the electron signals scattered through large angles (typically 75 to 150 m a d ) are analyzed. Contrast is proportional to the atomic number (Z) squared. For example, the silverkarbon contrast is over 60: 1. Non-carbide-forming elements, such as silver, platinum and copper, were dispersed as nearly spherical metal clusters in the DLC matrix (Figs 7 and 8). Nanodiffraction and STEM imaging reveal that silver, platinum and copper form nanocrystalline particles, with an average crystal size that varies between 3 and 5 nm. Figure 8 demonstrates atomically sharp boundaries between the silver nanoparticle and the hard carbon matrix. The large random particles that are observed in these micrographs are artifacts of the ion milling process used in transmission electron microscopy sample preparation. Dark field cross-sectional images of functionally gradient DLC-titanium and functionally gradient DLC-silver nanocomposites are shown in Fig. 9. The bright regions correspond to the higher atomic number titanium and silver regions, while the dark regions correspond to the DLC matrix. From kinetic considerations, the formation of coherently-strained nanometersized metal particles having a narrow size distribution is favored, because Ostwald ripening is not favored under these conditions. From thermodynamic considerations, the total energy within the DLC-metal nanocomposite system includes elastic energy, surface energy, interface energy and the edge energy of nanometer-sized metal particles. This DLC-metal nanocomposite system minimizes its total energy when a periodically ordered array of three-dimensional, coherently-strained nanometer-sized dots is formed.
Figure 7. Bright field Z-contrast image of DLC-silver nanocomposite film.
22
R. J. Narajan
Figure 8. Dark-field Z-contrast image of DLC-silver nanocomposite film.
Figure 9. (a) Z-contrast dark-field image of functionally gradient DLC-titanium nanocomposite film (FGTi 1) and (b) Z-contrast bright field image of functionally gradient DLC-silver nanocomposite film (FGAg1).
Carbide forming elements, such as silicon, did not form a separate phase at low concentrations (less than 5 at%). The presence of metal carbides in DLC-metal composites containing carbide-forming metals was corroborated with election energy loss spectroscopy. For example, the low loss titanium carbide peak was observed at 22 eV. Segregation of the titanium carbide phase in an alternating DLC-metal carbide nanocomposite is observed at higher metal concentrations (>5 at%). The Zcontrast image of functionally gradient DLC-titanium nanocomposite demonstrates alternating layers of titanium carbide and DLC (Fig. 9).
3.3.Electron energy loss spectroscopy Electron energy loss spectra between 280 to 3 10 eV were acquired. The sp3 fraction was determined from the K edge loss spectra using an empirical technique [13]. The peak in the region from 285 to 290 eV results from excitation of electrons from the 1s ground state to the vacant n* antibonding state. The peak in the
Adhesion properties of DLC-metal nanocornposite films
23
region above 290 eV results from excitation to the higher o* state. The ratio of the integrated areas under these two energy windows is approximately proportional to the relative iiumber of IT and o* orbitals. The atomic fraction of sp2 bonded carbon (x) was determined using the expression: (l(IT)/Z(
O))J(Z(
n)/Z(o)),= 3x44-x),
(1)
where Z(7c) is the intensity in the range from 284 to 289 eV and Z(o) is the integrated intensity in the range from 290 to 305 eV. The subscripts s and r refer to the ratio determined for the DLC specimen and a reference material with 100% sp2 bonding, respectively. The sp3 content was determined to be 63% for a DLC film on silicon (100). The sp3 content was determined to be 47% for DLC-silver nanocomposite FG Agl and 40% for DLC-silver nanocomposite FG Ti2.
3.4. Radial distribution function Electron diffraction provides high intensity beams and large scattering cross sections, which assist in the characterization of amorphous and nanocrystalline materials. Radial distribution function (RDF) analysis of the electron diffraction pattern provides short-range structural information on amorphous materials. The radial distribution function, G(r),is given by: CC
G(r)= 4*9r[,p(R) ( r ) - p ( O ) ] = 8*n? JQ(s)*sin(2*9s*r)ds
(2)
0
where p(R) ( r ) is the atomic density at position r, p(0) is the average total atom density and s the scattering vector. The function G(r)gives the most probable distances between atoms in a sample. Specifically, radial distribution function analysis provides the first and second coordination numbers, and the first and second nearest atomic neighbor distances. The values for amorphous carbon can be compared with those of diamond and graphite. Since DLC possesses both fourfold and threefold atomic coordinations, the first and second coordination spheres in the radial distribution function have values between those of graphite (3, 6) and those of diamond (4, 12). Figure 10 shows the radial distribution function for the DLC-1.2 at% copper nanocomposite. The shaded areas in the radial distribution function represent the data used in calculating the second and first nearest neighbor distances. Background correction was used to aid analysis. Figure 11 contains normalized G(r) values as a function of distance for the DLC-1.2 at% copper nanocomposite. Using the best-fit data interpolation, the first and second neighbor distances are 1.50 A and 2.54 A, respectively. The ratio of the second to the first nearest neighbor for the DLC-1.2 at% copper nanocomposite is 2.84. As seen in Table 2, these values are similar to those for pure DLC and those for pure diamond. This result indicates that metal atoms have a minimal influence on the sp3bonding in DLC-metal nanocomposite films.
R. J. Narayan
24
.
"V
Average density
5
0 0
1
3
2
5
4
Figure 10. Radial distribution function G ( r ) as a function of distance of DLC-1.2 at% copper nanocomposite. n1 and n2 are the first and second nearest atomic neighbor distances, respectively.
5 4
3 h
L
v
c3
2 1
0 -f
-2
-3 -4
-5 0
1
2
3
4
5
6
7
8
Figure 11. Normalized radial distribution function G(r) as a function of distance for DLC-1.2 at% copper nanocomposite. Table 2. Results of RDF analysis of DLC-copper nanocomposite film and comparison with other carbon forms
Graphite a-C (sputtered) Glassy Carbon a-C (evaporated) Diamond
1.42 1.46 1.43 1.43 1.54
2.45 2.49 2.45 2.53 2.5 1
Pure DLC DLC-1.2 at% copper
1.51 1.50
2.52 2.54
2.0 2.1 2.1 2.6 3.0 2.8 2.8
rl and r2 are the first and second nearest atomic neighbor distances, respectively. nz/nl is the ratio of the first and second nearest atomic neighbor distances. n2/n1values were derived from the intensity analysis of the diffraction trace. rJrl and n2/n1ratios often are not directly correlated, since there are variations in the crystal structure. For example, the r2/r1values for both graphite and diamond are 1.732. On the other hand, n2/nlis 3 for diamond and 2.66 for crystalline graphite.
Adhesion properties of DLC-metal nanocomposite films
25
3.5. Rutherford backscattering spectroscopy and X-ray photoelectron spectroscopy
The composition of DLC-metal nanocomposite films can be estimated using a simple geometric consideration. This estimate can be obtained by taking a ratio of the length of the arc of the ablation path on the metal piece to the perimeter of the circle navigated by the laser beam (Table 3). Unfortunately, this estimate neglects the differences in the ablation rates of graphite and metal. The true metal composition will be less than the geometric estimate, because metals possess higher reflectivity and smaller ablation rates than graphite. The composition of DLC-metal nanocomposite films was obtained experimentally by Rutherford backscattering spectroscopy and X-ray photoelectron spectroscopy. Rutherford backscattering spectroscopy determined the concentration of copper in the DLC-copper nanocomposite film Cu-1 to be 1.4 at% (Fig. 12b) and the amount of titanium in the DLC-titanium nanocomposite film Ti-1 to be 2.7 at% (Fig. 12c). X-ray photoelectron spectroscopy data corroborated the Rutherford backscattering spectroscopy findings. X-ray photoelectron spectroscopy gave a concentration of 1.44 at% for DLC-copper nanocomposite film Cu-1, and 2.66 at% for DLC-titanium nanocomposite film Ti-1 (Fig. 13). DLC-copper nanocomposite films are quite smooth and nearly entirely free from particulates. On the other hand, a number of particulates of variable size can be observed in DLC-titanium and DLC-silicon nanocomposite films. The morphology and size of these particulates suggest that they were formed from condensed liquid droplets. The presence of these particulates is usually attributed to a ‘splashing’ mechanism during pulsed laser deposition. Splashing takes place in most materials through subsurface boiling or shock-wave ejection of particulates. The large amount of particulates in DLC-titanium and DLC-silicon nanocomposite films results in the geometric estimations being lower than the experimental values obtained through Rutherford backscattering spectroscopy and X-ray photoelectron spectroscopy. Table 3. Geometric (apparent) and true concentrations (%) of metals in DLC-metal nanocomposite films Copper
Titanium
-~
Apparent concentration ( 7 ~ ) True concentration (%)
Silicon
cu-1
cu-2
cu-3
Ti-1
Ti-2
Ti-3
Si-1
Si-2
Si-3
1.5
1.6
2.7
1.4
2.1
4.0
1.4
2.8
3.5
1.4
1.5
2.5
2.7
4.0
7.7
-
-
-
R. J. Narayan
26 ,
25110,
I
,
,
,
1
.
*
1
i1
t o
-
t
L
4
04
10
08
00
14
1 2
16
Backscattered energy (MeV)
- sunulation
n in
5
211011
(DLC 1.2% Cu)
3
3
v
2
s2
1Ylo
1000
5 5
gm
3(1U
0 6
Backscattered energy (MeV) 2500
- dmulation { DLC 2.75% Ti)
h
Y
c 5
2000
3
v
'D
?w
2u
1500
inoo
-.1 4
8
z
500
0 04
06
08
10
I2
14
Backscattered energq (MeV) Figure 12. RBS spectra of (a) pure DLC, (b) DLC-1.4 at% copper nanocomposite, (c) DLC-2.7 at% titanium nanocomposite.
Adhesion properties of DLC-metal nanocomposite films
27
Figure 13. CIScore level XPS spectra of (a) DLC, (b) DLC-copper nanocomposite film (Cu-1). (c) DLC-titanium nanocomposite film (Ti- 1) and (d) DLC-silicon nanocomposite film (Si- 1).
R. J. Narayan
28
3.6. Raman spectroscopy The Raman spectra of functionally gradient DLC-silver nanocomposite films contain broad peaks, because selection rules for optical transitions are relaxed. All of the spectra show the following: (1) a broad hump centered in the 1510-1557 cm-' region, which is known as the G-band, and (2) a small shoulder at 1350 cm-', which is known as the D-band. The G-band is the optically allowed EZgzone center mode of crystalline graphite, and is typically observed in DLC films. The Dband is the A,, mode of graphite. High quality DLC films demonstrate the following: (1) a relatively symmetrical G-band and ( 2 ) a lesser D-band, suggesting an absence or a low amount of graphite clusters. The visible Raman spectra for the DLC-titanium nanocomposite reveals increased asymmetry and an increased D-band. These features suggest that the DLC-titanium nanocomposite possesses less tetrahedrally bonded carbon than high-quality DLC. Raman spectroscopy was also used to study the internal stress conditions within the DLC-metal nanocomposite films. Interatomic separation is correlated with the interatomic force constant, which, in turn, is correlated with the atomic vibrational frequency. The principle by which this data interpretation technique operates is as follows: when a material is stressed, the equilibrium separation between its constituent atoms is altered in a reversible manner. If the tensile load on the material increases, bond lengths increase, force constants decrease and vibrational frequencies decrease. On the other hand, if the material is subjected to mechanical compression, bond lengths decrease, force constants increase and vibrational frequencies increase. The scale of the Raman shift is related to the residual stress, 0,as follows:
+ v)/( 1- V)] [A U / W O ] ,
0 = 2G[ ( 1
(3)
in which a i s the shift in Raman wavenumber, w0 is the wavenumber of reference, G is the shear modulus and vis the Poisson ratio. The visible Raman spectrum G-peak positions for DLC-metal nanocomposites were determined by multiple Gaussian fittings. The G-peak for DLC was at 1568.3 cm-', the G peak for DLC-copper film Cu-1 was at 1563.6 cm-', the G peak for DLC-titanium film Ti-1 was at 1560 cm-' and the G peak for DLC-silicon film Si-1 was at 1551 cm-'. Raman spectroscopy of functionally gradient DLCsilver composites also revealed that films with larger silver concentrations exhibited larger shifts in both G- and D-peaks (Table 4). These values suggest that DLC-metal films with higher metal concentrations possess lower amounts of internal compressive stress. 3.7. Nanoindentation, adhesion and tribological properties
Nanoindentation revealed significant substrate effects, due to the presence of a relatively soft substrate and a relatively hard coating. During nanoindentation, the modulus of the coated sample approached that of the uncoated sample at roughly
Adhesion properties of DLC-metal nanocomposite films
29
Table 4. G-peak position (cm-') and internal compressive stress reduction (A@ as obtained from visible Raman spectra Titanium
Copper
Silicon
~~
AC7 (GPa) (G-peak position
Cu-1
Cu-2
Cu-3
Ti-1
Ti-2
Ti-3
Si-1
Si-2
Si-3
0.77
1.80
2.31
1.64
4.11
7.20
1.54
1.64
1.67
52.03
45.64
42.44
46.60
31.23
11.99
47.24
46.60
46.44
+1500) (cm-') The G-peak of pure DLC: is located at 1556.83 cm-'.
Table 5. Average modulus and nanohardness from unloading curves during nanoindentation testing Sample
Modulus (GPa)
Hardness (GPa)
FG Agl FG Ag2 FG Til FG Ti2
299 288 274 25 3
32 32 29 27
400 nm (approx. 2/3 of the film thickness). Substrate effects are observed even at indentation depths of 75 nm. Table 5 illustrates average hardness and modulus values for several films at 100 nm maximum indentation depth. In general, functionally gradient DLC-silver nanocomposite films demonstrated slightly higher modulus and hardness values than functionally gradient DLC-titanium nanocomposite films. These differences can be primarily attributed to the slightly higher concentration of sp3-hybridized carbon observed in functionally gradient DLCsilver nanocomposites. The values observed are comparable to values reported by Voevodin et al. [ 141, who reported nanoindentation hardness values between 30 and 32 GPa and elastic modulus values between 350 and 370 GPa for a 0.5-ymthick titanium carbide-DLC nanocomposite film on steel prepared using a hybrid magnetron sputtering/pulsed laser deposition technique. Evaluation of the adhesion between coating and substrate was performed using a CSM Microscratch instrument. Scratch adhesion testing is commonly used for determining the integrity of coated substrates [ 15, 161, The coating-substrate response to scratch testing may be separated into three regimes. In regime one, mild plastic deformation is observed up to tensile cracking. In regime two, higher loads produce both regular and irregular crack patterns. In particular, regular cracking oblique to the loading direction (cohesive failure) is often observed.
30
R. J. Nrrrrryrrn
Cracks can extend outside the scratch border, in a phenomenon known as external transverse cracking. Cracks may also remain within the scratch track, in a phenomenon referred to as internal transverse cracking. The crack pattern often becomes highly irregular just before the critical load for coating removal is reached. In regime three, coating removal by buckling, delamination, or flaking occurs. First, small amounts of coating debris are observed at the scratch track border. Flaking is enhanced by large friction forces and compressive stresses in the coating ahead of the indenter. Buckling failure may be observed if plastic deformation occurs in the substrate. These scratch adhesion behaviors are highly dependent on the presence of internal compressive stress within the film. A plot of the scratch tip indentation depth versus applied normal load for the functionally gradient DLC-titanium nanocomposite film FGTi 1is shown in Fig. 14. The plot can be divided into three regimes. Initially, no cracking occurred and the curve appears relatively smooth. Next, the roughness of the curve increases as crack formation occurs within and at the sides of the wear track. An increase in lateral force onto the scratch tip is responsible for this effect. As the lateral force increases, the normal force applied to the scratch tip will not be sufficient to maintain the current depth. Since the normal load is continuously increased, the lateral force will eventually be large enough to cause delamination of material ahead of the scratch path. Cycles of abrupt up and down motion of the scratch test signify cycles of coating delamination ahead of the moving tip. This effect becomes even more apparent in the third regime, when the indenter tip scratches into the soft Ti-6A1-4V substrate. 5 4.5 4
3.5
5
3
5- 2.5 p.
z
z li.5 '1
0.5 0
0
0.2
0.4
0.6
0.8
I
'1.2
? .4
I.6
Load, H
Figure 14. Depth vs. load obtained from scratch testing of the functionally gradient DLC-titanium nanocomposite film (FGTi1) at 1 N normal load.
Adhesion properties of DLC-metal nanocomposite films
31
The scratch tip diameter has a strong influence on the maximum and mean Hertzian pressures. A load of 1.5 N applied to a circular diamond tip of 20 pm diameter in contact with a flat DLC surface results in a maximum Hertzian pressure of 34.5 GPa. The commonly used 0.2 mm diamond tip would require a load of 150 N to reach the same maximum Hertzian pressure. Micrographs and load curves were developed from the microscratch adhesion testing data. Figure 15 illustrates the scratch track on the functionally gradient DLC-silver nanocomposite film sample FGAg1 at 0.8 N normal load. The scratch direction is from left to right. Spallation can be observed, which is caused by poor adhesion of functionally gradient DLC-silver film to the substrate under given loading conditions. The size of the delaminated areas along the scratch track indicates the quality of the thin fildsubstrate interface. At a load of 0.9 N, the scratch tip reaches the substrate and plastic deformation of the Ti-6A1-4V substrate can be observed. Figure 16 illustrates the track for a scratch test performed on the functionally gradient DLC-titanium nanocomposite film sample FGTi 1. Formation of forward
Figure 15. Scratch on the functionally gradient DLC-silver nanocomposite film (FGAg1) at 0.8 N normal load.
Figure 16. (a) Scratch on the functionally gradient DLC-titanium nanocomposite film (FGTi1) at 0.7 N normal load. (b) Scratch on FGTi1 film at 1 N normal load.
R. J. Narayan
32
chevron cracks in the scratch direction can be observed. These cracks form as a result of high stress concentrations produced by the lateral motion of the scratch tip. Furthermore, delamination of the functionally gradient DLC-titanium nanocomposite film does not occur at the onset of crack formation. This result indicates strong adhesion of the functionally gradient DLC-titanium film to the Ti6A1-4V substrate. The region near the interface of the functionally gradient DLCtitanium film is titanium-carbide-rich and is capable of forming a strong interface with the underlying Ti-6A1-4V substrate. The functionally gradient DLC-titanium films also form smaller wear particles due to better adhesion to the Ti-6A1-4V substrate. It is interesting to note that crack formation is more common in the functionally gradient DLC-titanium films than in the functionally gradient DLC-silver films. It can be concluded that internal compressive stress is lower in the functionally gradient DLC-silver films. The functionally gradient DLC-silver film structure, in which low modulus nanoparticles are present within the high modulus DLC matrix, effectively relieves internal compressive stress. Coefficients of friction obtained during wear testing are shown in Table 6. Typical coefficient of friction values for pure unhydrogenated DLC against 100Cr6 steel and alumina are 0.12 and 0.08, respectively [17-191. The coefficient of friction for functionally gradient DLC-titanium nanocomposite against steel appears quite similar to values for pure DLC that appear in the literature. The coefficient of friction of functionally gradient DLC-titanium nanocomposite against alumina shown here also matches values for DLC reported in the literature. The coefficient of friction for the functionally gradient DLC-titanium nanocomposite film against steel was significantly decreased when testing was performed in Ringer’s lactate solution. This process was even more drastic in the case of functionally gradient DLC-silver sliding against steel. In this case, the coefficient of friction dropped from 0.149 under ambient (30%) humidity to 0.074 in Ringer’s solution. Table 6. Friction coefficients for functionally gradient DLC-metal nanocomposite films Sample
Load (N)
Static partner
Cycles
Velocity (cds)
Friction coefficient
FG Til FG Til FG Til
3 3
10 000 10 000 10 000
FG Til FG Ti2 FGAgl FGAgl Ti-6A1-4V
3 3 3
Steel, dry Steel, Ringer’s lactate Steel, dry Alumina, dry Steel, dry Steel, dry Steel, dry Steel, Ringer’s lactate
3 3 3 3 3 3 3 3
0.107 0.072 0.136 0.078 0.112 0.149 0.409 0.740
I
3
3
10 000 10000 10 000 10 000 10 000
Adhesion properties of DLC-metal nanocomposite films
33
These results agree with the work of Ronkainen et al. [17] and Erdemir et al. [ 191, who demonstrated that the friction coefficient of unhydrogenated DLC decreased as relative humidity increased. Humidity reduces the coefficient of friction in unhydrogenated DLC, because both oxygen and hydrogen occupy the dangling bonds on the surface of the diamond-like films. These species reduce the probability of chemical bond formation between DLC film and the opposing surface. The coefficient of friction for functionally gradient DLC-silver films under dry friction conditions was found to be higher than the coefficient of friction for functionally gradient DLC-titanium films under identical conditions. This result may be attributed to the graphitization mechanism of DLC under wear conditions. The extent of graphitization of DLC under pressure is correlated with the presence of internal compressive stress in the film. Functionally gradient DLC-silver films possess reduced internal compressive stress, as shown by Raman spectroscopy. As such, the functionally gradient DLC-silver films do not readily graphitize under pressure. Thus, the amount of lubrication provided by a graphite surface layer is reduced, and the coefficient of friction is increased.
3.8. Antimicrobial properties of DLC-silver nanocomposite The engineering of hybrid implant materials in order to achieve added biological functionality is an area of biomaterials research undergoing rapid development. The sustained delivery of antimicrobial agents into the local micro-environment of an implant avoids systemic side-effects and exceeds typical systemic concentrations by several orders of magnitude. DLC-silver coatings have been designed to slowly deliver antimicrobial silver ions to the immediate environment of an implant biomaterial. Metal ions and metal compounds have been used for many centuries as disinfectants for fluids and tissues. Silver, in particular, was employed as a germicide well before the invention of modem antibiotics. It has been well documented that silver was used in ancient Greece to disinfect water and other beverages. In ancient India, Ayurvedic healers used silver as an elixir for patients debilitated by age or illness. The biocidal effect of silver, with its broad spectrum of activity against bacterial, fungal and viral agents, is particularly well known and the term “oligodynamic activity” was coined for this phenomenon. Nanocrystalline silver has demonstrated an unsurpassed antimicrobial spectrum, with efficacy against 150 different pathogens. In addition, nanocrystalline silver also provides broadspectrum fungicidal action. The concentrations required for microbicidal activity are in the range mol/l [20]. The current evidence for antimicrobial activity of nanocrystalline silver runs counter to earlier thinking that metallic silver would only have slight antibacterial effects because it is chemically stable. It is possible that elemental silver within nanocrystals is converted into an active ionic or oxide species via the cell metabolism.
34
R. J. Narayan
Figure 17. Antimicrobial susceptibility testing of silicon (100) control surface. The presence of Staphylococcus aureus growth on the silicon surface obscures surface imaging.
Testing of microorganisms recovered from clinical specimens is not a simple undertaking. The inhibitory activity of an antimicrobial agent is assessed either by dilution testing, which is a quantitative measure, or by disk diffusion testing, which is a qualitative measure. Disk diffusion allows classification of antimicrobial agents. The procedure has been standardized. Its use has been validated for testing action against aerobic and facultative bacteria, such as Gram-positive cocci, Enterobacteriaceae and Pseudomonas species; in addition, it can be modified to allow consistent findings for fastidious bacteria, which include Streptococcus pneumoniae and Haemophilus influenzae. The antimicrobial susceptibility testing performed on these coatings is a variant of the disk diffusion test. Diffusion of metal ions from DLC-metal nanocomposites is relatively slow as compared to the diffusion of typical pharmacologic agents embedded in a paper or resorbable polymer matrix. As such, assessment of the coated surface itself, as opposed to assessment of adjacent regions in the agar medium, is far more predictive of antimicrobial behavior. The antimicrobial testing on the DLC-silver nanocomposite surface was performed as follows. A tryptic soy agar plate was inoculated with Staphylococcus aureus. The coating or surface of interest was placed in direct apposition to the inoculated surface. The agar plate was turned upside down to allow direct observation of the agar medium-coating interface. The incubation period was 24 h in ambient air at 35°C. The agar plates were imaged using a CanoScanN670U scanner. The first result shown is that for the control surface, which is 1 cm x 1 cm silicon (loo), Staphylococcus aureus grows easily over the silicon surface, as seen below. Figure 17 illustrates that bacterial streaks on the silicon surface are of the same strength as those in the surrounding agar medium. The next result shown is that, for the DLC-silver nanocomposite surface, Staphylococcus aureus does not grow over the DLC-silver nanocomposite surface, as seen in Fig. 18. These data
Adhemsionproperties of DLC-metal nanocomposite films
35
Figure 18. Antimicrobial susceptibility testing of DLC-silver nanocomposite surface. The absence of Staphylococcus aureus growth allows direct, unimpeded observation of the coated surface.
suggest that silver within the DLC-silver nanocomposite demonstrates antimicrobial efficacy against Staphylococcus aureus bacteria. 4. CONCLUSIONS
Several DLC-metal nanocomposite films were prepared via a novel target design during pulsed laser deposition. DLC-metal nanocomposites do not immediately delaminate, and Raman spectroscopy data suggest these films contain reduced internal compressive stresses. Radial distribution function results demonstrate that the presence of metal atoms does not significantly change the short-range environment of carbon atoms from that observed in DLC. The modified disk diffusion test demonstrates the unusual antimicrobial properties of DLC-silver nanocomposite surfaces. DLC-silver nanocomposites exhibit significant antimicrobial efficacy against Staphylococcus aureus. The in vitro antimicrobial susceptibility of several bacterial and fungal pathogens to DLCsilver coatings is currently being assessed. The concentration of the metal component, either titanium or silver, was found to be crucial. The functionally gradient DLC-silver and DLC-titanium nanocomposite films exhibited promising wear characteristics. We have observed normalized wear rates for DLC-silver nanocomposite coatings on Ti-6A1-4V alloy of approximately 10-7-10-8 mm3 N per m; these values are quite similar to those observed for multilayered DLC coatings on hardened steel substrates [ 141. Metal alloying is a good alternative to thermal annealing for the creation of adherent, wear-resistant DLC thin films. These hard carbon coatings have a variety of medical and tribological applications, including use in orthopaedic implants, cardiac implants, cutting tools and wear-resistant magnetic disks. Finally, the electronic properties of the DLC-metal nanocomposite films may be suitable for electronic applications, including field emission.
36
R. J. Narayan
REFERENCES 1. J. Robertson, Mater. Sci. Eng. R 37, 129-281 (2002). 2. C. Donnet, M. Belin, J. C. Auge, J. M. Martin, A. Grill and V. Patel, Su$ Coating. Technol. 68, 626-63 1 (1994). 3. A. Singh, G. Ehteshami, S . Massia, J. He, R. G. Storer and G. Raupp, Biomaterials 24, 50835089 (2003). 4. M. Allen, B. Myer and N. Rushton, J. Biomed. Mater. Res. 58, 319-328 (2000). 5 . D. O’Leary, P. Dowling, K. Donnelly, T. P. O’Brien, T. C. Kelly, N. Weill and R. Eloy, Key Eng. Mater. 99-100,301-308 (199.5). 6. M. I. Jones, I. R. McColl, D. M. Grant and K. G. Parker, Diamond Relat. Mater. 8, 457-462 (1999). 7. T. Kuwahara, M. Markert and J. P. Wauters, Art$ Organs 13,427-431(1989). 8. M. I. Jones, I. R. McColl, D. M. Grant, K. G. Parker and T. L. Parker, J. Biomed. Mater. Res. 52,413-521 (2000). 9. T. A. Friedmann, J. P. Sullivan, J. A. Knapp, D. R. Tallant, D. M. Follstaedt, D. L. Medlin and P. B. Mirkarimi, Appl. Phys. Lett. 71,3820-3822 (1997). 10. F. M. Kustas, M. S. Misra, D. F. Shepard and J. F. Forechtenigt, Su$ Coating. Technol. 48, 113-119 (1991). 11. R. 0. Darouiche, Clin. Infect. Dis. 29, 1371-1377 (1999). 12. D. J. Stickler, Curr. Opin. Infect. Dis. 13, 389-393 (2000). 13. J. Bruley, D. B. Williams, J. J. Cuomo and D. P. Pappas, J. Microsc. 180, 22-32 (1995). 14. A. Voevodin, S. Walck and J. Zabinski, Wear 203-204,516-527 (1997). 15. S. J. Bull, SUI$ Coating. Technol. 50,25-32 (1991). 16. F. Attar and T. Johannesson, Su$ Coating. Technol. 78, 87-102 (1996). 17. H. Ronkainen, S. Varjus, J. Koskinen and K. Holmberg, Wear 249,260-266 (2001). 18. Y. Liu, A. Erdemir and E. I. Meletis, Su$ Coating. Technol. 94-95, 4 6 3 4 6 8 (1997). 19. A. Erdemir, C. Bindal, J. Pagan and P. Wilbur, Sur$ Coating. Technol. 77,559-563 (1995). 20. V. V. Ramana and K. Saraswathi, J. Ind. Chem. Soc. 70,274-275 (1993).
Adhesion Aspects of Thin Films. Vol. 2, pp. 37-47 Ed. K.L. Mittal 0VSP 2005
Adhesion improvement of magnetron-sputtered amorphous carbon coating on cemented carbide SAM ZHANG" and XUAN LAM BUI School of Mechanical and Production Engineering, Nanyang Technological Universiw, 50 Nanyang Avenue, Singapore 639798, Singapore
Abstract-Magnetron sputtering was used to deposit amorphous carbon coatings on cemented carbide (WC-6% Co) with extremely high adhesion strength. Plasma cleaning was used to remove cobalt from the surface of the substrate to enhance the adhesion strength. After plasma cleaning at 200 W (RF power), constant bias deposition, bias-graded deposition and metal-doped deposition were carried out to produce three different categories of carbon coatings. At constant bias of -150 V, the lower critical load in the scratch test was 145 mN, compared to 312 mN for the bias-graded deposition which creates sp3 structural grading within the coating with less sp3 at the interface and more towards the coating surface. In the case of carbon coating produced via the metal-doped method, the coating could not be remobed from the substrate in scratch tests. In this coating, doping (Ti and Al) creates a nanocrystalline T i c phase embedded in an amorphous Al-doped carbon matrix (denoted as nc-TiC/a-C(A1)) whereby adhesion strength drastically improved through reduction of residual stress and improvement in toughness. Keywords: Amorphous carbon; adhesion strength; residual stress; scratch test; pre-sputtering; biasgraded deposition; doping.
1. INTRODUCTION
Hydrogen-free (or non-hydrogenated) amorphous carbon (a-C) has found many engineering applications for more than two decades [ 1 , 21. Various techniques have been used to deposit a-C coatings on Si wafers, steels or tungsten carbide tools. These techniques include magnetron sputtering [3], pulsed laser deposition [4], cathodic vacuum arc deposition [ 5 ] , etc. The applications of an a-C coating strongly depend on its adhesion to the substrate. Adhesion strength dictates the ability of a coating to remain attached to the substrate under operating conditions. A coating adheres to the substrate due to interfacial forces comprising valence and interlocking forces. The residual stress in the coating and at the interface between the coating and the substrate, and the coating toughness all exert strong in=Towhom correspondence should be addressed. Tel.: (65) 6790-4400; Fax: (65) 6791-1859; e-mail: msyzhang 63ntu.edu.sg
38
S. Zhang and X . L. Bui
fluence on the adhesion strength [6, 71. Under physical vapor deposition conditions, a-C coatings grow under energetic ion bombardment, thus a compressive residual stress is developed in the coating [SI. High values of compressive stress, up to 10 GPa, have been reported [S-IO]. Such a high compressive stress limits the coating thickness and contributes to poor adhesion between an a-C coating and substrate. Both an enhancement of bonding and reduction of residual stress improve the adhesion. Plasma cleaning (or pre-coating sputtering) is a simple, yet important, technique to increase the bonding between a coating and a substrate by removing contaminants for better intimacy, by activating the surface for better chemical bonding and by producing a desirable surface morphology for better interlocking. Both bias-graded [ 111 and metal doping [ 121 during deposition contribute to reduce the residual stress and enhance the toughness through modification of the coating structure. In this study, we have combined plasma cleaning with constant bias deposition, bias-graded deposition or metal doping to produce pure a-C, bias-graded a-C and metal-doped a-C coatings. Extraordinary adhesion strength is obtained on cemented carbide substrate for bias-graded and metal-doped a-C coatings. 2. EXPERIMENTAL
2.1. Substrate preparation
Cemented carbide (WC-6% Co) disks (polished to 30 nm R,) of diameter 55 mm and thickness 5.5 mm were used as substrates. In order to measure the residual stress in the coating, Si wafers (100 mm in diameter and 450 pm thick) were also employed as substrates. Substrates were ultrasonically cleaned for 20 min in acetone followed by 10 min cleaning in ethanol, and then loaded into the chamber of E303A magnetron-sputtering system (Penta Vacuum, Singapore). Details of the system were described elsewhere [13]. The system was pumped down to 1.33 x Pa, and the substrates were heated to 150°C for 20 min for outgasing. To study the effect of plasma cleaning on the surface morphology of the cemented carbide substrate, argon gas was introduced to generate an intense plasma to sputter the substrate under a constant negative potential. The pressure during plasma cleaning was kept constant at 0.4 Pa. The bias powers of 100, 200, 300, 400 and 500 W were used and the cleaning duration was 20 min. The cleaned substrates were then examined at SEM for morpholgy. 2.2. Coating deposition The plasma cleaning procedure as described above was applied and followed by three different coating procedures: pure a-C deposited at -150 V bias, pure a-C deposited using bias-graded, and a-C doped with metallic Ti and AI. The bias grading was from -20 to -150 V at a rate of -2 V for every 100 s. The substrate was not heated during deposition of pure a-C coatings. Graphite target (99.999%
Adhesion improvement of magnetron-sputtered amorphous carbon coating
39
purity, 100 mm in diameter) operated at a power density of 10.5 W/cm2. Ti (99.995% purity, 100 mm in diameter) and A1 (99.995% purity, 100 mm in diameter) targets were used for doping. During metal-doped deposition, the substrate temperature was maintained at 150°C, bias voltage at -150 V and the power density of the graphite target was 10.5 W/cm2, whereas those of Ti and AI targets were varied to achieve composition variation.
2.3. Surface and coating characterization The morphologies of the surfaces of cemented carbide substrates before and after plasma cleaning were characterized using SEM. The hardness of the coatings was assessed using Nanoindenter XP (MTS, USA) with a Berkovich diamond indenter and analyzed with the continuous stiffness measurement technique [14]. The indentation depths were set not to exceed 10% of the coating thickness to avoid possible substrate effect. The residual stress in the coatings was determined from the change in the radius of curvature of the Si substrate before and after deposition, which was measured by a Tencor laser scanner. The Stoney equation was employed for stress, a, evaluation:
where E, /( 1- v,) is the substrate biaxial modulus (180.5 GPa for Si( 100) wafers [15]); t, and t, are wafer and coating thicknesses, respectively; R1 and R2 are, respectively, the radii of curvature of Si wafer before and after deposition. The coating adhesion was studied with a scanning micro-scratch tester (Shimadzu SST101). In this system, a diamond stylus of 15 pm radius is drawn on the coating surface at a gradually increasing load from 0 to 500 mN. Meanwhile, the stylus oscillates up to a distance of 50 pm (the scanning amplitude) in the direction perpendicular to the drawing direction. The scanning capability of the system allows to scan a larger area (50 pm times the scratching distance) as compared to the traditional “single line” scratch. This greatly improves the reliability of the test. In this experiment, the scanning amplitude was set at 50 pm at a scratch speed of 10 p d s for all samples. As widely accepted, the lower critical load (the load at which coating damage was observed via a sudden increase of coefficient of friction) was used as a measure of adhesion strength. 3. RESULTS AND DISCUSSION
3.1. EfSect ofplasma cleaning on morphology of substrate su$ace
The effect of plasma cleaning on cobalt removal was reported in Ref. [16], where XPS binding energy profiles of Co 2p3” and W 4f were obtained at different depths from the sputter-cleaned surface. Subsequently, the composition of Co and
40
Bias Power During Plasma Cleaning, VV Figure 1. Dependence of Co removal on RF power during plasma cleaning [16].
W was obtained as a function of depth from the surface of the substrate, from which the “cobalt loss” was evaluated. The cobalt loss was found to increase with bias power used in plasma cleaning, until about 200 W. After that, a further increase in bias power did not increase the amount of cobalt loss (cf., Fig. 1 [16]). Figure 2 compares the SEM morphologies of the substrate surface before and after plasma cleaning at 200, 300 and 400 W bias powers. It is easily seen that with increasing bias power, pitting formation becomes increasingly severe, which is not seen before plasma cleaning. A higher power results in a rougher surface with larger and deeper pits. Since the sputter yield of Co is much higher than that of W, for example, at a bombarding energy of Ar’ of 600 eV, the sputter yield of Co is 1.4 atomshon, whereas that of W is 0.6 atomshon [17]; therefore, Co-rich areas erode more, thus leading to formation of pits. Sputtering of a material with a uniform etch rate will not result in such etch pits (as in stainless steel samples [ 111). As the plasma cleaning power increases, the materials removal becomes more severe for both W and Co even though the cobalt loss remains unchanged after 200 W. 3.2. EfSect of bias power and metal doping 3.2.1. Hardness and residual stress The hardness and residual stress values for the coatings are summarized in Table 1. As the deposition bias increased, the hardness of the coating increased drastically together with increase of residual stress. At a deposition bias of -60 V, the coating hardness was about 19 GPa. As bias increased to -150 V, a high hardness
Adhesion improvement of magnetron-sputtered amorphous carbon coating
41
Figure 2. Surface morphology of cemented carbide substrates before (a) and after plasma cleaning at RF power 200 W (b), 300 W (c) and 400 W (d).
S. Zhang and X . L. Bui
42
Figure 2. (Continued).
Table 1. Hardness of and residual stress in coatings Coating
Hardness (GPa)
Residual stress (GPa)
a-C (-60 V bias) a-C (- (50 V bias) a-C (bias-graded) nc-TiC/a-C(A1)
18.6 31.5 25.1 19.6
1.15 4.10 1.46 0.38
(about 32 GPa) was achieved, but accompanied by a high residual stress of 4.1 GPa. This can be understood through the relationship between the bombarding energy of ions to the substrate and the bias voltage [ 181: E..
Vb p1l2
-
where E, Vb and P are the bombarding ion energy, bias voltage and process chamber pressure, respectively. An increase in bias voltage causes an increase in the energy of ions coming to the substrate. When this energy exceeds the critical value for atomic displacement, the ions penetrate deep into the interior of the coating structure, leading to denser and smoother a-C coating with higher compressive stress and higher sp3 fraction. At lower bias, the bombarding ions have lower kinetic energy and the diffusion in surface layers becomes dominant. Surface diffusion tends to generate ordered clusters with a graphite-like structure. The coatings, therefore, have lower residual stress but also lower hardness (due to higher sp2 fraction). The relationship between bias voltage and sp3/sp2ratio has been reported in our previous papers [13, 191.
Adhesion improvement of magnetron-sputtered amorphous carbon coating
43
When bias-grading is used, Le., the deposition bias voltage is gradually increased as the deposition proceeds, the resultant a-C coating becomes slightly “softer” as a whole. In this study, the bias was increased from -20 to -150 V, an overall hardness of about 25 GPa was obtained, with a residual stress of only 1.46 GPa. Understandably, in the bias-graded deposition, the sp3 fraction was lowest at the interface between the coating and the substrate, and gradually increased to a maximum value at the surface [ 111, effectively creating a coating structure with graded sp3 fraction (increasing sp’ from the interface). Therefore, both hardness and residual stress varied through the coating thickness. Since the nanoindentation measured the average hardness of the whole coating, a lower overall hardness was observed. Metals have been used as dopants for stress relaxation of a-C coatings [20]. Among the metal dopants, AI is most effective in relaxation of residual stress. However, the hardness of the coating suffers: only about 40% of the hardness remains when 10 at% AI is doped in an a-C coating [21]. The hardness of the coating can be restored while keeping the residual stress low. This was accomplished by simultaneously doping Ti and A1 into a-C to form a nanocomposite a-C coating [12]. In this study, Ti and A1 doped coating consisted of 47 at% C, 40 at% Ti and 13 at% AI, where elemental A1 existed in the a-C matrix, whereas the Ti mostly bonds with carbon to form nanocrystalline TIC. The nanocrystalline T i c phase helped to maintain the coating hardness at an adequately high level (about 20 GPa), while the existence of AI in the matrix as clusters of A1 atoms reduced the residual stress to an extremely low level of 0.38 GPa due to an increasing sp2 fraction. Detailed explanation for the formation of nanocrystalline structures was reported in Refs [12, 221. 3.2.2. Adhesion strength Figure 3 illustrates the influence of plasma cleaning on the adhesion strength in the case of constant (-150 V) bias and bias-graded deposition. The adhesion strength of both coatings follows the same trend: when the plasma cleaning power is lower than 200 W, the critical load increases with increase of applied power. This was due to the effect of plasma cleaning through removal of surface cobalt contamination together with other contaminants such as oxides. A better interlocking from an increase of the surface roughness also helped. However, when the cleaning power exceeded 200 W, a drop in critical load resulted. This was attributed to the high residual stress at the substrate surface due to an excessive ion bombardment. When comparing the bias-graded coating and the coating deposited at constant bias (-150 V), the adhesion strength of bias-graded coating became much higher than that of the constant-bias coating, regardless of the power level during plasma cleaning. This considerably higher adhesion strength was the result of low residual stress combined with high toughness resulted from biasgraded deposition. Much higher critical load was also observed in bias-graded coating compared to that for the constant bias coating in our previous work, when stainless steel was used as the substrate [ l 11.
S. Zhang and X . L. Bui
44
350 I
300 250 -
200 -
150-
/o\o
100-
50
-150 V bias
0
1 -
I 0
100
200
300
400
500
Plasma cleaning power (W) Figure 3. Adhesion strength (in terms of critical load) of a-C coatings deposited under bias-grading and constant bias, as a function of plasma cleaning power.
Figure 4 compares the scratch tracks from (a) a-C coating deposited under -150 V bias, (b) bias-graded a-C coating and (c) metal-doped a-C coating. The substrates for all these coatings underwent the same plasma cleaning at 200 W for 20 min before deposition. In the a-C coating deposited under -150 V bias (constant), the lower and the higher critical loads were not distinguishable, i.e., a total and catastrophic failure occurred as the coating started to fail when the applied load reached 145 mN. This is a typical brittle fracture of coatings having high residual stress and poor toughness. In the bias-graded a-C coating, the first failure occurred at 312 mN. Note that at this load the coating peel-off was only partial and the scanning amplitude remained 50 pm. In the metal-doped a-C coating, no material removal or damage was observed even when the applied load increased to about 400 mN. However, at very high load, the scratching tip ploughed into the coating, which resulted in decreasing scanning amplitude as seen in the micrograph (Fig. 4c). Figure 5 shows the scratch profile of the metal-doped a-C coating. As the load increased, the coefficient of friction (here in terms of relative output voltage) increased gradually. As the diamond tip ploughed into the coating, the coefficient of friction increased more and also some vibration was observed. However, there was no sudden rise in the coefficient of friction. This is typical of the plastic behavior expected of an extremely tough material. This high toughness was the result of the nanocomposite structure combined with amorphous matrix doped with Al. The fact that there was no interfacial failure suggested that an excellent coating-substrate bonding resulted from the combination of plasma cleaning and extremely low residual stress (only 0.38 GPa) due to doping.
Adhesion improvement of magnetron-sputtered amorphous carbon coating
45
Figure 4. Scratch tracks on (a) a-C coating deposited under a bias voltage of -150 V, (b) a-C coating deposited under bias-grading and (c) nc-TiC/a-C(A1) coating. The substrates underwent the same RF plasma cleaning at 200 W for 20 min.
46
-
s
S. Zhang and X . L. Bui
100
W
b m
c , I
0
Tip radius: 15 pm Scanning amplitude: 50 pm 80Scratch speed: 10 pm/s
>
60-
3 Q 3 0
40-
c, .c,
c ,
m
I
*
d o0
O
S
50 100 150 200 250 300 350 400
Load (mN) Figure 5. Friction coefficient in terms of relative output voltage as a function of normal load in the scratch test on the nc-TiC/a-C(A1) coating.
4. CONCLUSIONS
Aside from traditional pre-coating plasma cleaning, adhesion of a-C on cemented carbide can be further improved through two other effective ways: (1) bias-graded deposition and (2) Al-doped deposition. The bias-graded deposition creates sp3 structural grading within the coating with less sp3 at the interface and more towards the coating surface. This effectively reduces the residual stress and increases toughness. As a result, the adhesion strength improves considerably. Aldoped deposition embeds metallic A1 in the amorphous carbon and forms a-C(A1) that effectively reduces residual stresses and enhances adhesion, but at the expense of hardness. Co-sputtering with Ti forms nanocrystalline TIC phase to imbed in the a-C(A1) matrix and form nanocomposite nc-TiC/a-C(Al), which brings back adequate hardness while maintaining the enhanced adhesion strength. Acknowledgements
This work was supported by Nanyang Technological University’s Research Grant No. RG12/02. REFERENCES 1. A. Matthews and S. S . Eskilsen, Diamond Relat. Mater. 3, 902-91 1 (1994). 2. Y. Lifshitz, Diamond Relat. Mater. 8, 1659-1676 (1999).
Adhesion improvement of magnetron-sputtered amorphous carbon coating
47
3. N. Savvides and B. Window, J. Vue. Sci. Technol. A3, 2386 (1985). 4. A. A. Voevodin and M. S . Donley. Surf: Coating. Technol. 82, 199-213 (1996). 5. B. K. Tay, D. Sheeja. S. P. Lau, X. Shi, B. C. Seet and Y. C. Yeo, Surf: Coating. Technol. 108, 72-80 (1998). 6. K. Holmberg and A. Matthews, in Tribology Series, 28: Coatings Tribology: Properties, Techniques and Applications in Surj%ce Engineering, D. Dowson (Ed.), Elsevier, Amsterdam (1994). 7. A. A. Voevodin and J. S. Zabinski, J. Mater. Sei. 33,319-327 (1998). 8. E. Mounier and Y. Pauleau, Diamond Relat. Mater. 6, 1182-1 191 (1997). 9. S. Zhang, H. Xie, X. T. Zeng and P. Hing, Surf: Coating. Technol. 122,219-224 (1999). 10. D. Sheeja, B. K. Tay, S. P. Lau and X. Shi, Wear 249,433-439 (2001). 11. S. Zhang, X. L. Bui, E'. Q. Fu, D. L. Butler and H. J. Du, Diamond Relat. Mater. 13, 867-871 (2004). 12. S. Zhang, X. L. Bui and Y. Q. Fu, Thin Solid Films, in press (2004). 13. S. Zhang, X. L. Bui and E'.Q. Fu, Surf: Coating. Technol. 167, 137-142 (2003). 14. G. M. Pharr, Muter. Sci. Eng. A253, 151-159 (1998). 15. W. A. Brantley, J. Appl. Phys. 44, 534 (1973). 16. S . Zhang and H. Xie, Surf: Coating. Technol. 113, 120-125 (1999). 17. R. J. Hill, Physical Vapor Deposition, Temescal, Berkeley, CA (1986). 18. Y. Catherine, Diamond and Diamond-like Films and Coatings, Plenum Press, New York, NY (1991). 19. S. Zhang, X. T. Zeng, H. Xie and P. Hing, Surf: Coating. Technol. 123, 256-260 (2000). 20. A. Grill, Wear 168, 143-153 (1993). 21. B. K. Tay, Y. H. Cheng, X. Z. Ding, S. P. Lau, X. Shi, G. F. You and D. Sheeja, Diamond Relat. Mater. 10, 1082-1087 (2001). 22. S. Zhang, X. L. Bui, Y. Q. Fu and H. J. Du, Znt. J. Nanosci, in press (2004).
Adhesion Aspects of Thin Films,Vol. 2, pp. 49-56 Ed. K.L. Mittal 0VSP 2005
Characterization of polyethylene-metal composite thin films deposited by evaporation SATORU IWAMORI,* FUMINORI TATEISHI, YOUHEI O N 0 and YOSHINORI YAMADA Faculty of Engineering, Kanazawa University, 2-40-20, Kodatsuno, Kanazawa 920-8667, Japan
Abstract-Polyethylene (PE), PE/gold (Au) composite (PE-Au) and PE/aluminum (Al) composite (PE-AI) thin films were deposited onto a glass slide and aluminum substrates by conventional vapor deposition process. Although the PE thin film deposited at high temperature (higher than 8OOcC) and high pressure (higher than 10 mTorr) was hazy and white in color, the PE thin films deposited at low temperature (lower than 500°C) or low pressure (lower than 1 mTorr) were transparent. The resistivity of the PE-Au thin films decreased gradually with increase of pressure, but that of the PEA1 thin film decreased dramatically with increase of pressure. This dramatic decrease was due to the oxidation of A1 in the PE-A1 thin film. The PE-Au and PE-A1 thin films were introduced between the PE thin film and the AI substrate. Although the scratch durability of the PE thin film deposited onto the A1 substrate deteriorated with the introduction of the PE-Au layer, it was improved with the introduction of the PE-.41 layer. The PE-A1 layer acts as a functionally gradient material layer.
Keywords: Vapor deposition; polyethylene; aluminum; gold; composite layers.
1. INTRODUCTION
Polymer thin films, such as polyethylene (PE), have been deposited by vacuum evaporation since the 1970s and their crystallinity, chemical structure and molecular weight have been characterized [l, 21. Pogrion and Tirnovan [3] analyzed the structure of the PE thin films before and after irradiation with electron beams using transmission electron microscopy and electron diffraction. In addition, polymer thin films have been deposited by the ionization-assisted deposition (IAD) method and their chemical structure, molecular weight and crystallinity have been characterized [4]. The morphologies of Au-containing PTFE thin films prepared by co-evaporation were analyzed by TEM [5, 61. Ultrahigh-molecular-weight polyethylene (UHMWPE) has been used as an artificial knee joint material, because it has excellent abrasion durability and is safe
*To whom correspondence should be addressed. Tel./Fax: (81-76) 234-4950; e-mail:
[email protected]
50
S. Iwamori et al.
for use in a human body. Many researchers have focussed on improving the mechanical friction and abrasion properties of the UHMWPE. In order to improve the durability of the PE, the PE-metal composite layers were introduced between the PE thin film layer and a metal substrate by the conventional vacuum evaporation process and their tribological properties were evaluated. 2. EXPERIMENTAL
2.1. Materials A low-density polyethylene pellet (Asahi Kasei, Japan) was used as the deposition material for PE thin films and PE/metal composite thin films. A gold wire (0.3 mm diameter) and an aluminum 1050 plate (0.3 mm thickness) were used as materials for Au thin films, A1 thin films and PE-metal composite thin films. A glass slide substrate was used for measuring the deposition rate and electrical resistance, for evaluating the topography of these thin films and for analysis of the composition of these thin films by X-ray photoelectron spectroscopy (XPS). An aluminum substrate and a copper substrate were used for the scratch test.
2.2. Apparatus A conventional vacuum evaporation apparatus equipped with a tungsten boat (for the metals) and tungsten boat coated with aluminum oxide (for the PE), as shown in Fig. 1, was used for the deposition of these thin films. In order to measure the temperature in these boats, a platinum (Pt)-rhodium (Rh) thermocouple with an amperemeter was used. A shutter was placed between the substrate and the boats. The chamber was evacuated by a rotary pump and a diffusion pump. After evacuation to a pressure of 5 x Torr, Au and A1 were heated by heating these boats and evaporated at 1430°C and 11l O T , respectively. The PE was gradually heated up to 460°C at a heating rate of 7.5Wmin after evacuation to a pressure of 5 x 1 0 - ~TO^.
2.3. Evaluation The electrical resistance of the PE-Au and PE-A1 thin films deposited on the glass slide substrate was determined according to the test method for resistivity of conductive plastics with a four-point probe measurement in order to eliminate the effect of the contact resistance [7]. The chemical bonding states of the PE-metal composite thin films were determined by XPS. A pin-on-disk type friction and scratch test apparatus was used for the evaluation of friction and scratch properties. A steel bearing ball (1 mm in diameter) was used as the slider with 0.2 N load. The sliding speed of the sample was 10 revolutionshin and the sliding diameter was 30 mm. The scratch life of these thin films on the substrates was evaluated by optical microscopic observations.
PIC-metal composite films deposited by evaporation
51
Figure 2 shows schematic diagrams of the friction and scratch durability test samples. The PE thin films were deposited onto the AI substrate (Fig. 2a), onto the PE-AI composite thin film on the A1 substrate (Fig. 2b), and onto the PE-Au composite thin film on the AI substrate (Fig. 2c). ,Bel
jar
Substrate . Shutter
Thermocouple Evaporation source
/Rotary
Diffusion pump Figure 1. Schematic diagram of the evaporation apparatus.
(a) PE/Al
3. 5 p m
PE thin film
AI substrate (b) PEE'E-MA1 .
I
A1 substrate
(c) PE/PE-Au/AI
Al substrate
110.
3mm
Figure 2. Schematic diagrams of the friction and abrasion durability test samples.
S. Iwarnori et al.
52
3. RESULTS AND DISCUSSION
3.1. Transparency of the PE thin films
Before depositing PE-Au and PE-A1 composite thin films onto the substrates, the deposition rates of the PE, Au and A1 thin films were determined [8]. These deposition rates increased with increase of the evaporation temperature. The deposition rate of the PE thin film was two orders of magnitude higher than that of the Au and A1 thin films [8]. Although the deposition rates of the Au and A1 thin films did not change at pressures between 0.1 and 10 mTorr, that of the PE decreased with increase of pressure [ 81. Figure 3 shows optical micrographs of the PE thin films deposited on glass slides under various conditions. The thickness of all PE thin films was 2.0 pm. Although the PE thin film deposited at high temperature (higher than 800°C) and high pressure (higher than 10 mTorr) was hazy and white in color, the PE thin films deposited at low temperature (lower than 500°C) or low pressure (lower than 1 mTorr) were transparent.
3.2. Relationship between resistivity, composition of the films and pressure Figure 4 shows the resistivity of the PE-Au and PE-A1 composite thin films deposited at various pressures. The resistivity of the PE-Au thin film decreased with increase of pressure between 0.05 and 10 mTorr. This decrease is attributed to the decrease of the PE deposition rate [8]. The Au content for the PE increased with increase of pressure. Table 1 shows the elemental composition of the PE-Au thin
Temperature (High) 4
890°C , 10 mTorr
890°C , 0.1 mTorr
Pressure (High)
460°C , 0.1 mTorr
I
460°C , 10 mTorr
(Low) 4
k
76mm Figure 3. Photographs of the PE thin films deposited on glass slides at various conditions.
PE-metal composite films deposited by evaporation
53
film prepared at a pressure of 1 mTorr. It was almost the same as the composition calculated from the deposition rates of the PE and Au. Figure 5 shows the relationship between the resistivity and the Au content in the PE-Au thin film. The resistivity decreased with the increase of the Au content between 14 and 36 wt%. The resistivity of the PE-A1 thin film decreased dramatically with increase of pressure between 0.05 and 10 mTorr compared to that of the PE-Au thin film (Fig. 4). Table 1 also shows the elemental composition of the PE-A1 thin film prepared at a pressure of 1 mTorr. Because the PE-A1 thin film contained large amounts of oxygen, the aluminum in the PE-A1 thin film would be expected to be oxidized. Although gold is difficult to be oxidized, aluminum is easily oxidized. This is the reason why the resistivity of the PE-A1 thin film showed an initially higher value than that of the PE-Au thin film. The PE-A1 thin film would be oxidized during vacuum evaporation due to the residual oxygen in the chamber. A tungsten boat coated with aluminum oxide was used as the evaporation boat for the PE coating. The chamber pressure increased with increase of the temperature of the boat. This increase was due to outgassing from the boat, because this in-
0.01
0.1
10
1
100
Pressure (mTorr) Figure 4.Resistivity of the PE-Au and PE-A1 composite thin films deposited at various pressures.
Table 1. Elemental composition of the PE-Au and PE-AI composite thin films
PE-AU PE-A1
All (wt7G)
A1 (wt8)
0 (wt8)
C (wt8)
30
-
-
20
2 54
68 26
S. Iwamori et al.
54
.~
40 n
E
c > , ..c, ' .-
l l
I
I
v)
2
a_-
20
lo
A -
,
0:
, 0
0
I
0 0
10
20
30
40
50
Au content (wtX) Figure 5. Relationship between resistivity and the Au composition in the PE-Au composite thin film.
crease was observed when the boat was heated without the PE pellets. We think the outgassed material would contain a large amount of oxygen, and the partial pressure of oxygen during vacuum evaporation at the low pressure would be much higher than that at the high pressure.
3.3. Evaluation of the PE thin film deposited on the A1 substrate Figure 6a-c shows the optical micrographs of the frictional tracks after the 10 min durability test on the PE thin film deposited onto the A1 (PE/Al), the PE-A1 thin composite film on the AI (PE/PE-Al/Al) and the PE-Au thin composite film on the A1 (PE/PE-Au/Al), respectively. Although the frictional tracks can be seen in Fig. 6a and 6b, these thin films did not peel off from the A1 substrate. On the other hand, the PEPE-Au thin films peeled off from the substrate (Fig. 6c). The PE-Au thin film reduced the adhesion strength between the PE thin film and the A1 substrate. Daudin and Martin [9] reported that the Au thin film prepared by vacuum evaporation did not adhere to the A1 substrate, but the adhesion strength was improved by ion beam treatment. We think that the friction and scratch durabilities of the PEPE-Au/A1 system were low because of the low adhesion between the Au thin film and the A1 substrate. Figure 7a and 7b shows the optical micrographs of the frictional tracks for the PE/A1 and PE/PE-Al/Al systems, respectively, after the durability test for 30 min. Although the frictional tracks can be seen in Fig. 7b, these thin films did not peel off from the A1 substrate. The PE thin film, however, peeled off from the substrate (Fig. 7a). The PE-A1 composite thin film improves the adhesion strength between the PE thin film and the A1 substrate. Fukuda [ 101 reported that introduc-
PE-metal composite films deposited by evaporation
(a) PE/Al
55
(b) PEPE-AVAl
(c) PEPE-AdAl
Figure 6. Optical micrographs of the frictional tracks in (a) PE/Al, (b) PE/PE-Al/Al and (c) PE/PEAu/Al systems after durability test for 10 min.
(a) PE/Al
(b) PE/PE-Al/Al
Figure 7. Optical micrographs of the frictional tracks in (a) PE/A1 and (b) PE/PE-Al/Al systems after durability test for 30 min.
tion of functionally gradient materials between the niobium (Nb) oxide layer and Nb substrate improved the adhesion strength. In this case the PE-A1 composite thin film apparently acted as a functionally gradient material layer, and improved the adhesion strength between the PE thin film and the A1 substrate.
56
S. Iwamori et al.
4. CONCLUSIONS
Polyethylene (PE), PE-Au and PE-A1 thin films were deposited by a conventional vapor deposition process and the following conclusions are drawn. 1. Although the PE thin film deposited at high temperature (higher than 800°C) and high pressure (higher than 10 mTorr) was hazy and white in color, the PE thin films deposited at low temperature (lower than 500°C) or low pressure (lower than 1 mTorr) were transparent. 2. The resistivity of the PE-Au and PE-A1 composite thin films decreased with increase of pressure. But the decrease in the resistivity of the PE-A1 thin film was much more than that of the PE-Au thin film. 3. PE-Au and PE-A1 composite thin films were introduced between the PE thin film and the A1 substrate. Although the scratch durability of the PE thin film deposited onto the A1 substrate deteriorated by the introduction of the PE-Au layer, this durability improved with the introduction of the PE-A1 layer. The PEA1 layer apparently acted as a functionally gradient material layer. REFERENCES 1. P. P. Luff and M. White, Thin Solid Films 6, 175 (1970). 2. Y. Hattori, M. Ashida and T. Watanabe, J. Chem. SOC.Jpn. 496 (1975). 3. N. P. Pogrion and M. Tirnovan, Thin Solid Films 317, 232 (1998). 4. H. Usui, Thin Solid Films 365, 22 (2000). 5 . K. P. Gritsenko and A. M. Krasovsky, Chem. Rev. 103,3607 (2003). 6. V. V. Petrov, A. A. Kriuchin and K. P. Gritsenko, Dokl. Akad. Nauk. Ukr. SSR 12, 64 (1989) 7. Japanese Industrial Standard JIS K7194 (1994). 8. S. Iwamori, F. Tateishi and Y. Yamada, Mater. Sci. Technol. 40,42 (2003). 9. B. Daudin and P. Martin, Nucl. Instrum. Methods Phys. Res. B34, 181 (1988). 10. R. Fukuda, Mater. Sci. Technol. 35, 10 (1998).
Adhesion Aspects of Thin Film.r, Vol. 2, pp. 57-68 Ed. K.L. Mittal 0VSP 200s
Selection of efficient coatings for milling Inconel 718 based on their adhesion properties 0. KNOTEK,’ E. LUGSCHEIDER,’ K. BOBZIN,’ C. PINERO,’.’; F. KLOCKE,* D. LUNG2 and J. GRAMS2 ‘Materials Science Institute, Aachen University, Augustinerbach 4-22, 52062 Aachen, Germany ’Laborator?;f o r Machine Tools and Production Engineering, Aachen University, SteinbachstraJe 53, 52074.4achen, Germany
Abstract-The main goal of the co-operative research program “Environmentally Compatible Tribosystems” (SFB 442) is to develop tribological systems which can be used in machine tools without causing environmental damage. The main focus of this program is to avoid the use of additives in lubricants, such as anti-wear and extreme-pressure additives, which possess high toxicity and noxiousness. This aim is mainly achieved by the application of a single fluid family (synthetic esters in this case) instead of lubricants that usually are used in the tribological systems of a machine tool. In this study, the focus was on cutting processes. It was observed that the cooling effect produced by applying biodegradable synthetic esters in cutting operations was lower than that produced using water-based emulsions due to different physical properties. For this reason, the diminished cooling during cutting processes is expected to be compensated by the application of PVD coatings. Nialloys, like Inconel 718, are rated to be difficult to machine, due to their low thermal conductivity and their tendency to strain harden and to adhere to the cutting tool. Wear-resistant coatings can improve the tribological properties of cutting tools for milling Inconel 718, but their efficiency depends considerably on the adhesion between the coating and the tool substrate. In this work, two different methods were employed for determining the adhesion properties of promising PVD and CVD coatings. Special attention was paid to impact testing, which is a suitable method for characterization of coating fatigue properties, as well as of interfacial and cohesive failure modes of the coating, at dynamic loads. The results of this work can be used to predict the cutting properties of coated cutting tools. Furthermore, coatings were characterized for their tribological functions by cutting and pin-on-disc tests.
Keywords: Coating adhesion; cutting tool; Inconel 718; face milling.
1. INTRODUCTION
Ni-alloys, like Inconel 718, are difficult to machine. This is mainly due to their special material properties (low thermal conductivity, tendency to strain harden and to adhere to the cutting tools). The high tensile strength (R,) of nickel-based -To whom correspondence should be addressed. Tel.: (49-241) 80-95340;
Fax: (49-24 1) 80-92264; e-mail:
[email protected]
58
0. Knotek et al.
materials at elevated temperatures (approx. tensile strength R, = 1100 N/mm2 at 700°C) causes a high mechanical load on the tools. Especially, the abrasive resistance (hardness) and plastic deformation of the tool material are important. Because of the complex loads imposed on cutting tools, a tool design for machining Inconel 71 8 must offer high toughness, high hardness, high chemical resistance and an excellent resistance against Inconel 7 18 adhesion to the cutting tool at the same time [ 11. The application of coatings on tool surfaces can improve the performance of cutting tools. The efficiency depends considerably on the coating adhesion. Tool surface pretreatments can also be used to influence the coating/tool interaction. The main goal of this co-operative research work is the development of environmentally compatible cutting processes (including turning, drilling and milling of Inconel 718). This is achieved by using a synthetic ester instead of an emulsion as the cutting fluid. But the use of a synthetic ester introduces new tribological stresses and strains on the cutting tool, which should be compensated by using coated cutting tools. Although the application of coated tools in milling Ni-alloys is not the state of the art, a previous face milling test against Inconel 718 with PVD-coated tools showed that coatings had a significant influence on the in-process behavior of a milling tool. Figure 1 shows the tool wear after milling Inconel 718 by two differently coated tools. Both coatings are based on TiAlN, but they have different structures and, consequently, different properties. On the left-hand side, a cutting tool coated with a TiAlN-monolayer coating, which possesses a homogeneous structure, is shown. And, on the right-hand side is shown a cutting tool coated with a TiAlNsuperlattice; this coating is composed of multiple nanolayers, thus it has different properties such as higher hardness and lower Young’s modulus compared to the monolayer coating. The TiAlN-superlattice-coated tools showed less cutting edge wear and also less coating flaking compared to the monolayer coating. Likewise, previous investigations on turning and drilling Inconel 7 18 with coated tools using a synthetic ester as the cutting fluid showed that the properties of the coating material modified the adhesion behavior of the tool to the workpiece during cutting [ 11. Thus, appropriately coated milling tools should offer excellent mechanical properties and a low adhesion tendency, based on the results of the milling test and the results of the previous work on drilling tools. Therefore, PVD and CVD coatings, e.g., TiAlN, WC/C+TiAlN and A1203, were investigated. These were deposited on both ground and polished surfaces, in order to determine the influence of the coating-substrate adhesion on the efficiency of coated cutting tools in milling Inconel 7 18.
Efficient coatings for milling Inconel 718
59
Figure 1. Tool wear after milling of two differently coated tools. The figure shows SEM pictures of cutting inserts (geometry: SDFT 1204 AEFN) taken by WZL, RWTH Aachen. *Emulsion used as cooling medium consists of 9 4 8 lubricant and 6% water.
2. EXPERIMENTAL
2.1. Sample preparation
Cutting inserts made of tungsten carbide HW-K10, with a WC-grain size of 0.50.8 pm, were used as substrates. This is a typical tool material for cutting operations for nickel-based alloys. Some samples were coated as delivered. This means that the ground surface of the samples was not pretreated mechanically before coating. The other samples were coated after a polishing treatment. Both substrate variants were coated under the same process conditions. The coated samples were tested in tribological and coating adhesion tests. 2.2. Coating deposition
TiAlN coatings were deposited by magnetron sputter ion plating (MSIP) at a base pressure of 1 Pa. Typical deposition temperatures were around 450°C. The WC/C+TiAlN coatings were also deposited by PVD techniques. The A1203coatings were deposited by CVD techniques. CVD coatings are usually deposited in a temperature range of 680-1 100°C. 2.3. General coating cJzaracterization
The samples coated in as-delivered conditions were characterized for thickness, hardness, Young’s modulus, roughness and surface energy. Coating thickness was
60
0. Knotek et al.
a) arrangement of srunple and grinding sphcrc
b) determination of coating thickness
Figure 2. Description of the calo test used to determine coating thickness. Coating thickness is determined using the following geometrical equation: thickness = (X* Y)/ (sphere diameter).
determined by the calo test. This test involves a rotating sphere, with a known diameter, that grinds the coated surface. Both the position of the sphere relative to the sample and the contact load are constant. Upon adding an abrasive slurry (in this case a diamond dispersion of 1 pm particle size was used) to the contact zone, a depression with the shape of a spherical cap is abraded into both the coating and the substrate (see Fig. 2a). The parameters X and Y (see Fig. 2b) were measured by light optical microscope examinations. The thickness of the coating was calculated by a simple geometrical equation: thickness = ( X * Y)/ (sphere diameter). Hardness and Young's modulus were measured by nano-indentation with a nanoindenter XP@(MTS Systems, USA). The mean roughness (R,) was measured in both parallel and perpendicular directions. R, is calculated by measuring the vertical distance from the highest peak to the lowest valley for five sampling lengths, and then averaging these distances. R, averages only the five highest peaks and the five deepest valleys. The surface free energy was estimated from contact angle measurements using the sessile drop method. The surface free energy including the polar and dispersion components was calculated using the Owens-Wendt approach. Five probe liquids were chosen to cover the widest possible range from very polar liquids, like water, to almost completely dispersive liquids like diiodomethane [2, 31. Five drops of each liquid were used to statistically guarantee the results of the measurements and approximately 60 contact angle measurements were made for each drop.
2.4. Tribological tests The samples coated in as-delivered conditions were analyzed by pin-on-disk tests. The pin-on-disk tribometer used was built by the Laboratory for Machine Tools and Production Engineering (WZL) at the RWTH Aachen University in order to simulate tribological conditions in cutting processes. This test device belongs to the wear testing category VI according to the German standard DIN 50322. By
Efficient coatings f o r milling Inconel 718
61
this test device it is possible to vary the complex set of tribological stresses and strains during the process by making relatively simple modifications, which allows examination of tribological processes in detail. The scatter in the results obtained using the pin-on-disk tribometer is less than that in real cutting processes, because of its more stable conditions. However, it is possible to extrapolate the results obtained to real cutting processes, which would otherwise be more expensive and difficult to carry out. 2.5. Coating adhesion tests The coating adhesion of both variants (samples coated after polishing and coated in as-delivered conditions) was evaluated by scratch and impact tests in order to determine the effect of substrate surface treatment on the coating adhesion. Also, the coating performance for milling operations was investigated by impact tests. 2.5.1. Scratch test A REVETESTer (CSM Instruments, Switzerland) was used as the scratch device. In this device, a diamond tip is drawn along the sample surface with a predefined normal load. The applied load is increased until a critical load is reached. The coating is seldom completely removed within the track. Therefore, it is necessary to define a critical load (related to coating adhesion). Here the critical load was defined as the load at which the coating was removed along the whole track length. All coated samples were evaluated up to a scratch load of 90 N. 2.5.2. Impact test Impact tests were carried out with an impact tester designed and built by CemeCon (Wurselen, Germany), in close collaboration with the Laboratory for Machine Tools and Manufacturing Engineering of the Aristotle University of Thessaloniki. During the impact test a carbide ball with a diameter of 6 mm periodically impacted the coating at a predefined load. After the tests, the surface of the impacted sample was investigated with optical techniques in order to analyze the coating failure. The samples were tested in a range of 1000-1 000 000 impacts using different loads: 400, 600, 800 and 1000 N. 3. RESULTS AND DISCUSSION
3.1. General coating characterization
In order to evaluate the performance of the coated tools, the as-deposited coatings were characterized as described above. The thickness values were around 4 pm. All coatings had good and acceptable nanohardness and Young’s modulus values, which are shown in Table 1. The hardness and Young’s modulus of the WC/C+TiAlN coating were not reported because the heterogeneous hard/soft combined structure of the WC/C top layer caused a high standard deviation in the measured values. Since the WC/C top layer is very thin (<1 pm), it was assumed
0. Knotek et al.
62
Table 1. Mechanical properties of coatings tested Coating
Hardness @Pa)
Young’s modulus @Pa)
TiAlN
35
A1301
11.2
406 295
45 40 35
30 25 20 15 10 5 n
Figure 3. Surface properties of the coated samples.
that the values of hardness and Young’s modulus of the WC/C+TiAlN coating were about the same as the values of the TiAlN coating. The R, values vary between 1 and 5 pm. In Fig. 3a it can be seen that the samples coated with TiAlN and A1203 exhibit a considerable difference between the R, values measured in parallel and perpendicular directions. This implies that the grinding grooves had a considerable influence on the surface topography of the deposited TiAlN- and especially of the A1203-coatings.Although the surface energy is defined for ideal surfaces (no surface roughness), it is acceptable to evaluate the surface energy of non-ideal surfaces for coated tools [2, 31. However, surface roughness has to be taken into account in the interpretation of the surface energy. The surface energy and its polar and dispersion components determined for the coated samples, uncoated cemented carbide and workpiece material Inconel 7 18 are shown in Fig. 3b. As expected, Inconel 718 possesses a very high surface en-
Efficient coatings for milling Inconel 718
63
Figure 4. Wear of the samples after pin-on-disk tests against Inconel 718. The figure shows SEM pictures of cutting inserts taken by WZL, RWTH Aachen.
ergy as well as polar component. Whereas the AI2O3-coated samples exhibit the lowest polar component of surface energy, and the uncoated HW K10 samples have the highest polar component of all samples measured. Since the polar component is mainly responsible for adhesion, it can be supposed that the coatings which possess the lowest polar component of the surface energy will theoretically offer the best protection against adhesion. This has been tested in a pin-on-disk friction test (see Fig. 4).
3.2. Tribological tests As can be seen from the test parameters (force and speed), this test was set up in order to simulate the process of a chip flowing through a chip flute. Uncoated and TiAIN-coated samples showed more wear caused by adhesion to Inconel 7 18 than the other tested samples (see Fig. 4a and 4b). The best anti-adhesion properties were achieved by the WC/C+TiAlN and A1203 coatings. The solid lubricant effect of the WC/C top layer in the WC/C+TiAlN-coated samples delayed the formation of deposits of Inconel 718 on the surface of the samples. Therefore, Inconel adhesion could be seen only in a few areas, where presumably the WC/C top layer was
0. Knotek et al.
64
worn out (Fig. 4c). Since A1203exhibited a very low polar component of surface energy, it was expected to see almost no Inconel adhesion in the A1203-coated samples. However, some coating wear failure caused by adhesion to Inconel 7 18 was observed. Probably the high roughness of the A1203-coatedsamples produced high friction, which was responsible for the detected Inconel adhesion (see Figs 4d and 3a).
3.3. Coating adhesion tests 3.3.1. Effect of substrate surface treatment on coating adhesion As stated above, cemented carbide inserts were coated in both ground and polished states. The basic roughness data on the inserts before coating are shown in Table 2. As expected, the polished substrate exhibited lower roughness. Coating adhesion depends ultimately on the interfacial bond strength to the substrate, as well as on the interface microstructure. Scratch and impact tests are classified as mechanical methods for coating adhesion measurement, because the adhesion is determined by the application of a force to the coatinghubstrate system [4].
3.3.1.1. Scratch test. All coatings show critical loads higher than 90 N in scratch testing. The results, however, depend on factors such as coefficient of friction between the tip and the coating as well as mechanical properties of the coating. According to the optical photographs in Fig. 5 , the surface roughness of the samples exerted a strong influence on scratch behavior because of the different friction coefficients between tip and coating surface. Due to the ductile character of TiAlN, a plastic deformation of the substrate was clearly observed in TiAlNcoated samples which were coated after polishing. In the WC/C+TiAlN-coated samples, the WC/C top layer was partly removed but the TiAlN-film remained. The high friction coefficient between the tip and surface of the A1203coatings deposited on ground substrates resulted in cracks formation because of the brittle character of the A1203. However, the coating was not completely removed. Meanwhile, in the A1203 coatings deposited on polished substrates, almost no crack formation was observed. 3.3.1.2. Impact test. To determine the performance of the coated tools for milling operations, it is necessary to evaluate the coating fatigue endurance. It has been shown that high stress levels appear under deformation conditions and the highest stresses are concentrated at the coating-substrate interface [5]. Therefore, the abil-
Table 2. Mean roughness values of the cutting inserts before coating Rz (Pm)
Ground substrate
Polished substrate
R z (parallel)
0.8 1.6
0.6 0.6
Rz (Demendicular)
Efficient coatings for milling Inconel 718
TiAlN
W C/C+TiAlN
ground substrate
I
500 pm
65
I
500 pm
t-----------l
500 pm
polished substrate
Figure 5. Scratch patterns at a 90 N load.
ity of a coated tool to resist high stresses depends on the strength of the coatingsubstrate interface. In case of insufficient interface strength, single or repeated loading can result in breaking of the interfacial bonds and delamination of parts of the coating. According to the optical photographs of impacted coated specimens after 100000 impacts at an 800 N impact force (Fig. 6), it is evident that the polishing treatment diminished the fatigue endurance of the coatings. It could be a consequence of the diminution of the coating adhesion due to larger spacing between roughness peaks (only a few roughness peaks on the polished substrates) in comparison to the ground substrates [6]. 3.3.2. Impact test performance of TiAlN, WC/C+TiAlN and A1203coatings Due to the plastic deformation that develops during the loading stage, the contact area does not fully recover to its initial plane shape, thus forming a permanent concave imprint [6,7]. The hardlsoft WC/C top layer delayed the appearance of fatigue failure in the WC/C+TiAlN-coated samples. It was observed that although there was a rather quick removal of the WC/C-upper layer, the base coating TiAlN withstood a higher number of impacts until fatigue failure appeared. As far as the initiation of fatigue failure is concerned, the experiments did not reveal any significant difference between TiAlN and WC/C+TiAlN coatings. In most cases the stresses in CVD coatings can be attributed to thermal stresses generated by the difference in thermal expansion coefficients between the coating and the substrate. The stresses can be either tensile or compressive, but tensile stresses are, in general, most damaging [4]. As the thermal expansion coefficient
0. Knotek et al.
66
WC/C+TiAlN
TiAlN ground substrate
I
,
500 prn
500 prn
,
I
500pm
I
polished substrate
Figure 6. Optical photographs of the coated samples after 100000 impacts at an 800 N impact force.
1000000
L Q
c,
E
10000 400
500
600
700
800
900
1000
impact force [N] Figure 7. Number of impacts before coating failure by impact tests at a predefined impact force. The letters p and g were used to indicate the samples coated after Eolishing and coated in asdelivered conditions (Le., ground surface).
Efficient coatings for milling Inconel 718
61
of the coating material is usually larger than that of the cemented carbide, the thermal stresses in the CVD coatings deposited on cemented carbide after cooling from the deposition temperature are normally tensile. Different studies have clearly indicated that embrittlement occurring in CVD-coated cemented carbides, which can originate coating failures under extreme stresses produced by interrupted cutting, can be avoided with PVD coatings, since they possess higher compressive stresses [4]. Nevertheless, the CVD-AI2O3-coatedsamples showed a very good coating adhesion and the best performance in impact testing as compared to other coatings investigated (Fig. 7). Coating failure occurred after 800 000 impacts at a 1000 N impact force for the A1203coatings deposited on ground substrate specimens, and after 600000 impacts on polished substrates. In addition, the A1203 coating has a very low affinity for the more commonly used materials because of its very low polar component of surface energy. Therefore, Al2O3coatings are very promising for cutting operations, such as milling Inconel 7 18. 4. CONCLUSIONS
Based on face milling of Inconel 7 18 with PVD-coated tools it was shown that the coating design as well as the surface treatment before coating had a significant influence on the in-process behavior of the milling tool. It can also be stated that a cutting tool for milling Inconel 7 18 must have a high resistance against thermal loads and abrasive wear attack. Furthermore, these tools need to withstand the high wear caused by adhesion of Inconel 718 to the tool surface. A good coating adhesion and a low roughness also are of a vital importance. This set of requirements leads us to imagine that a PVD-coated tool in combination with the application of synthetic esters as a coolant might offer a significant improvement in the in-process behavior of cutting tools for machining Incone1 7 18. Therefore, promising coating variants have been analyzed with respect to their adhesion tendency to Inconel 7 18, their resistance against dynamic load and their adhesion to the substrates. The results show that CVD-AI2O3coatings offer a good potential for improved in-process behavior of face milling tools for machining Inconel 718. This will be analyzed through face milling tests with different A1203variants. PVD-A1203and combinations of PVD-TiAlN/Al2O3look very promising coatings to improve the tool performance for milling Inconel 7 18 because they can offer a combination of the good properties of A1203and higher compressive stresses at the same time. Acknowledgements
The authors gratefully acknowledge the financial support of the Deutsche Forschungsgemeinschaft (DFG) within the co-operative research program SFB 442 “Environmentally Compatible Tribosystems”.
68
0. Knotek et al.
REFERENCES 1. F. Klocke, D. Lung, J. Grams, E. Lugscheider, K. Bobzin and C. Colmenares, in: Proceedings of the F i f h International Conference & Awareness Workshop on Behavior of Materials in Machining, Chester (2002). 2. E. Lugscheider and K. Bobzin, Tribol. Schrnierungstech. 48(2), 32-36 (2001). 3. E. Lugscheider and K. Bobzin, Su$ Coating. Technol. 165, 51-57 (2003). 4. R. Bunshah (Ed.), Handbook of Hard Coatings, William Andrew Publishing, New York, NY (200 1). 5 . B. Bhushan (Ed.), Modern Tribology Handbook, Vol. 2, CRC Press, Boca Raton, FL (2000). 6. K.-D. Bouzakis, N. Michailidis, S. Hadjiyiannis, K. Efstathiou, E. Pavlidou, G. Erkens, S. Rambadt and I. Wirth, Su$ Coating. Technol. 146-147,443450 (2001). 7. K.-D. Bouzakis, N. Michailidis, A. Lontos, A. Siganos, S . Hadjiyiannis, G. Giannopoulos, G. Maliaris and G. Erkens, Zschr.5 Metallk. 92, 1180-1185 (2001).
Adhesion Aspects of Thin Films, Vol. 2, pp. 69-78 Ed. K.L. Mittal
0VSP 2005
Investigation of tissue compatibility and hemocompatibility of DLC and CN, coatings D. J. LIx and L. F. NIU College of Physics and Electronic Information Science, Tianjin Normal University, Tianjin 300074, P.R. China
Abstract-Despite promising studies of carbon films supporting their biomedical applications, only a few results on their bio- and hernocompatibility have been reported, especially for CN,. In this work, CN, and DLC coatings were prepared using dc magnetron sputtering. The effects of CN, and DLC coatings on cultures of mouse fibroblasts and human endothelial cells were determined by scanning electron microscopy. The results showed that the coatings caused no adverse effects to the cells. CN, coating provided a comparable or better surface for the normal cellular attachment, growth, and morphology as compared to DLC. CN, coating also showed longer blood coagulation time and recalcification time compared to DLC. In addition, less platelet adhesion and less platelet activation were found on the CN,-coated surface. These results supporting the tissue compatibility and hemocompatibility of CN, should initiate an interest in the biomedical applications of CN, coating. Keyw3ords: Carbon nitride; diamond-like carbon: biocompatibility.
1. INTRODUCTION
Biomaterials must possess adequate surface and bulk characteristics when they are used in biological environments in order to fulfill the dual requirements of desirable biocompatibility and mechanical properties for special applications. Currently, however, most synthetic materials only display good bulk characteristics. Therefore, it is necessary to find suitable approaches to improve the surface properties of materials, which possess otherwise desirable bulk characteristics, for use in biological environments. One good way to achieve this is through thin films deposition. The deposition of diamond-like carbon (DLC) has received considerable attention in the biomedical field due to its excellent mechanical and tribological properties, as well as demonstrated biocompatibility. Attractive results have been reported for applications such as total joint replacement [ 1, 21, orthopedic pins and *To whom correspondence should be addressed. Tel./Fax: (86-22) 2354-0278; e-mail: dejunli @eyou.com
70
D. J. Li and L. F. Niu
screws [3], dental prostheses [4], and medical guidewires [5], despite very limited information about the hemocompatibility of DLC [6, 71. Carbon nitride (CN,) is similar to DLC in structure, as well as in mechanical and tribological properties, To date, however, CN, has not been investigated in terms of biocompatibility, especially for hemocompatibility. In previous studies [6, 8, 91, we investigated the biocompatibility of carbon films prepared by ion-beam-assisted deposition. In this work, our aim was to synthesize CN, and DLC coatings using magnetron sputtering under identical deposition conditions and compare the effects of these coatings on in vitro tissue compatibility and hemocompatibility. Their mechanical and tribological properties were also investigated. 2. EXPERIMENTAL
2.1. Coating preparation
A standard dc magnetron-sputtering system with a high-purity graphite (99.995%) target was used to synthesize CN, and DLC coatings. The chamber was evacuated by a turbomolecular pump to a pressure of 4 x Torr before beginning film deposition. The magnetron sputtering system was operated at a pressure of 2 mTorr in different process gases (Ar for DLC; Ar/25% N2 mixture for CN,r). Silicon (001) wafers were used as substrates. Substrates were cleaned ultrasonically in acetone and methanol before being introduced into the vacuum chamber. Before deposition, substrates were sputter-cleaned in an Ar plasma at 80 mTorr with -500 V bias for 3 min. The graphite target power was fixed at 200 W. A 2 kHz pulsed dc bias of -100 V was applied to the substrate for low-energy ion bombardment.
2.2. Cell culture Two types of cells were chosen in this experiment. 3T3 mouse fibroblasts were cultured in Iscove’s Modified Dulbecco’ s Medium (IMDM) containing 10% fetal bovine serum (FBS). The fibroblast suspension contained 1 x lo5 cells/ml. The endothelial cell (EC) suspension (1 x lo5 cells/ml) was from the umbilical cord of a healthy human fetus. The detailed preparation process for cell suspension has been described in our previous paper [lo]. Uncoated and coated silicon samples were sterilized in ethylene oxide prior to use. The same amounts of cell suspensions were cultured on 0.6 x 0.6 cm2 sample areas, some of which were coated with 50-nm-thick CN, and DLC, while others were not coated, in a 24-well culture plate. The cultures were performed in an incubator with a humidified atmosphere containing 5 % COz in air. After 72 h of 3T3 fibroblasts and ECs in the incubator, the medium was removed and the cell monolayer washed several times with phosphate-buffered saline (PBS) and fixed in methanol for scanning electron microscopy (SEM) observation. Six independ-
Tissue- and hemocompatibiliiy of DLC and CNx coatings
71
ent samples were used for each coating with each cell type to measure cellular attachment, growth and morphology. 2.3. Platelet adhesion and blood coagulation and recalcification times Blood drawn from a healthy adult volunteer, who was kept free of aspirin or other drugs that might interfere with platelet functions, was centrifuged for about 10 min to obtain platelet-rich plasma. Platelet-rich plasma was incubated on the surfaces for 15 min. After rinsing, fixing and drying, the morphology and statistical results on platelets attached to the surfaces coated with gold-palladium were investigated by SEM. The measurement of blood coagulation time of the coatings was performed using a shaker. 1.5 ml blood of a rabbit was injected into a sterile test-tube containing the sample and then the test-tube was placed immediately in the shaker. The measured interval from injection of the blood into the test-tube to the appearance of blood coagulation was taken as the blood coagulation time. To measure the recalcification time, 4 ml blood of a rabbit and 36 ml sodium citrate (3.8%) were placed in a sterile container and centrifuged at 3000 rpm for 10 min to obtain the plasma. 0.5 ml plasma was injected into a sterile test-tube containing the sample, and then 0.5 ml CaClz solution was injected. The test-tube was shaken in the shaker. The measured interval from injection of the CaC12 solution into the testtube to appearance of white particles in the plasma was taken as the recalcification time. The hemocompatibility results are summarized in Table 1. 2.4. Coating properties measurements X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES) were employed to examine the chemical composition of the coatings. Surface morphology was measured by atomic force microscopy (AFM). The surface rootmean-square roughness was obtained over a sampling area of 10 x 10 ,urn2. The hardness was measured using a Hysitron nanoindentor. In the measurements, the penetration depth was kept to less than 10-15% of the coating thickness to minimize substrate effects. The coefficient of friction as a function of the sliding time was monitored by moving a 10-mm-diameter sliding ball (AIS1 52100) back-and-forth on the surface of the coatings over a track length of 2 mm using a ball-on-disk tribometer. The tests were performed at a load of 1 N at reciprocating frequency of the ball of 10 Hz. Table 1. A summary of hemocompatibility results obtained Hemocompatibility measured
Blood donor
Incubation temperature
Platelet adhesion Blood coagulation time
Healthy adult volunteer Rabbit
37°C 37°C
Recalcification time
Rabbit
37°C
D. J. Li and L. F. Niu
12
3. RESULTS AND DISCUSSION
3.1. XPS and AES results Figure 1 shows the XPS survey spectra of DLC and Cl., coatings. It is clear that the coatings contain mainly carbon, nitrogen, and oxygen atoms. In order to obtain atomic concentrations, AES analysis was carried out after 3 min of Art ion sputtering. AES spectra reported in Fig. 2 clearly show that no oxygen peak is present, which means that the oxygen signal in the XPS spectra was due to air exposure during sample transport. The N/C ratio for CN, was found to be 15% by converting Auger intensities into atomic percentages.
200
0
400
600
800
1000
Binding energy (eV) Figure 1. XPS survey spectra of DLC and CN, coatings
-
C
After Ar' ion sputtering
Electron energy (eV) Figure 2. Auger electron spectra of DLC and CN, coatings after A r' ion sputter cleaning.
Tissue- and hernocompatibility of DLC and CNx coatings
13
3.2. Coating properties
Figure 3 shows the hardness and coefficient of friction of both DLC and CN,y coatings. The higher hardness (20 GPa) of CN, coating may be due to a higher fraction of sp3 bonding in its structure. At the same time, CN, coating also has a lower coefficient of friction than DLC. The combination of higher hardness and lower coefficient of friction is advantageous to improve the wear life of CN, when used as a wear-resistant coating for biomedical implants. Figure 4 shows the roughness of both coatings over a sampling area of 10 x 10 ym2. It demonstrates that the smoother surface of CN, coatings might be the main reason for its lower coefficient of friction. 24
0.12
22
0.11
20
0.10
18
0.090
16
0.080
14
0.070
n El
a
13 W
%I %I
0
c a
Figure 3. Hardness and coefficient of friction of DLC and CN, coatings.
0.5 h
E
0.45
I
I
tt
.............
W
i
I
........................................
4
%,
0.4
............
-I
AFM scan size: 10 pm
1.
..........................................
---I
i
I
.................................................
j 3 0.25
I
DLC Figure 4. RMS roughness of DLC and CN, coatings.
I
1
CN
1
74
D. J. Li and L. F. Niu
3.3. Tissue compatibility
The 3T3 fibroblasts on the surface are round in shape during initial incubation, and then gradually stretch out, with spindle-shape morphology. After 72 h incubation, SEM images (Fig. 5a, b) confirm that more filopodia are present between cells on CN, coating than on DLC, which suggests that CN, provides a much better condition for cellular proliferation than DLC. Figure 5c also shows that the cells growing on CN, coating attach to the surface and exhibit a normal cellular morphology of polygons, maintaining physical contact with each other through filopodia (arrows). There appeared to be no evidence of extensive cell death on the surfaces of both coatings.
Figure 5. SEM images of mouse 3T3 cells cultured on (a) CN, and (b) DLC coatings for 72 h at 5 0 0 ~ magnification; (c) is an image of 3T3 cells on the surface of CN, coating at 1500x magnification.
Tissue- and hemocompatibilih of DLC and CNx coatings
7s
Interesting results were obtained for both coatings, when the attachment, spreading and morphology of human endothelial cells (ECs) on the surfaces were observed by SEM. It is clear from Fig. 6 that many tiny flagellums adhere tightly to the surface, which implies a strong ability of ECs to attach to both CN, and DLC coatings. Also, the attached ECs on CN, coating exhibit more filopodia than those on DLC. The endothelial cell is a valuable cell for biocompatibility, as it has a strong ability to decrease blood coagulation when it attaches and grows normally on the surface of biomedical implants. 3.4. Hemocompatibilitql
Blood coagulation time and recalcification time, which are related to hemocompatibility, were obtained using the blood of a rabbit. Figures 7 and 8 show the results for both coatings. The values in the figures represent averages of five measurements. Significant differences between CN, and DLC coatings can be observed. It is clear that CN, coating exhibits long blood coagulation time, as well as recalcification time, which affirms its biomedical significance. Scanning electron micrographs of platelets fixed on CN, and the control Si, after 15 min incubation time, are shown in Fig. 9. It is clear that the number of individual platelets on CN, coating is lower than that on the control Si. The number of activated platelets with a certain degree of spreading is seen on the surface of the control Si (Fig. 9a). Many deformed platelets, such as pseudopodium, are observed on the surface of Si. However, the platelets attached to CN,, are disc-shape cells, despite that a few of the platelets exhibited the early stages of activation (Fig. 9b). The decrease in platelet deposition on CN, coating is beneficial to hemocompatibility .
Figure 6. SEM images of human endothelial cells cultured on the surfaces of (a) CN, and (b) DLC coatings at 2 0 0 0 ~magnification.
D. J. Li and L. F. Niu
16
20 18
I
I
-........................................
.-
4
I .......
16 -
t
..............
I
Y
14
I
A. T
........
......
..................................................
I
I
DLC
1'
I
CN
Figure 7. Blood coagulation time on DLC and CN, coatings.
50 45
40
I
- ...................................................
c
35f
I
30 -
25
% 41 f9
-
...............................................
-
0 i
i
........................................................
-I
20 %
DLC
CN
Figure 8. Blood recalcification time on DLC and CN, coatings.
It is well known that high platelet adhesion to the surface is the main reason for blood coagulation. Further, it will accelerate the deposition of Ca2' on the surface. It is likely that N atoms bonded to C in the structure make the surface hydrophilic, inhibiting platelet adhesion and Ca2+deposition, which, in turn, increases both the blood coagulation and recalcification times. At the same time, this surface also provides a proper plate for normal cellular attachment and growth of endothelial cells.
Tissue- and hemocompatibilih of DLC and C N , coatings
I1
Figure 9. SEM images of platelets on (a) Si and (b) CN,xcoatings at 3000x magnification.
4. CONCLUSIONS
The main clinical purpose of DLC or CN, is to provide a durable, wear- and corrosion-resistant coating with desirable biocompatibility for biomedical implants. This study shows that CN, and DLC coatings could provide surfaces for normal cellular attachment and morphology of mouse fibroblasts and human endothelial cells. Their surfaces also strongly inhibit blood coagulation. platelet adhesion, and platelet activation. It seems that CN, coating has a better surface for tissue compatibility and hemocompatibility as compared to DLC. Besides, CN, coating shows a slight improvement in smoothness and wear resistance. It is conceivable that CN, coating has a great potential as an attractive biomaterial and should find clinical applications in the future. Acknowledgements
This work is supported by a Joint Project of Tianjin Municipal Universities and Nankai University and Tianjin University under Grant No. GJDFOl and ProjectSponsored by SRF for ROCS, SEM.
REFERENCES 1. M. Allen, B. Myer and N. Rushton. J. Biomed. Mater. Res. 58. 319-328 (2001). 2. J. Narayan, W. D. Fan, R. J. Narayan, P. Tiwari and H. H. Stadelmaier, Mater. Sci. Eng. B25, 5-10 (1994).
78
D. J. Li and L. F. Niu
3. E. Mitura, S. Mitura, A. Niedzielski, Z. Has, R. Wolowiec, A. Jakubowski, J. Szmidt, A. Sokolowska, P. Louda, J. Marciniak and B. Koczy, Diamond. Relat. Mater. 3, 896-898 (1994). 4. A. Olborska, M. Swider, R. Wolowiec, A. Niedzielski, A. Rylski and S. Mitura, Diamond. Relat. Mater. 3, 899-901 (1994). 5 . J. McLaughlin, B. Meenan, P. Maguire and N. Jamieson, Diamond. Relat. Muter. 5 , 486-491 ( 1996). 6. D. J. Li, J. Zhao and H. Q. Gu, Sci. China 44,427-431 (2001). 7. M. I. Jones, I. R. McColl, D. M. Grant, K. G. Parker and T. L. Parker, J. Biomed. Mater. Res. 52, 413-421 (2001). 8. D. J. Li, F. Z. Cui and H. Q. Gu, J. Adhesion Sei. Technol. 13, 169-177 (1999). 9. D. J. Li, S. J. Zhang and L. F. Niu, Appl. Surj%ce Sei. 180,270-279 (2001). 10. D. J. Li and L. F. Niu, Nucl. Instrum. Methods Phys. Res. B192, 393-401 (2002).
Adhesion Aspects of Thin Films, Vol. 2, pp. 79-89 Ed. K.L. Mittal 0VSP 2005
A study on structural characterization of and cell attachment to Ti-containing coatings Y. LIU,' S. LIU,* Q. X. LIU' and D. J. L113x College of Physics and Electronic Information Science, Tianjin Normal Universih, Tianjin 300074, P.R. China 'Tianjin Electronic Information Vocational Technology College, Tianjin 3001 10, P.R. China
Abstract-Ti-containing coatings. such as TiN, T i c and TiCN, are appropriate materials for use as protective and biocompatible materials on artificial implants, due to their good wear performance, high hardness, low coefficient of friction and high adhesion to a range of materials. TIN, TIC and TiCN coatings were synthesized by multi-arc ion plating. The Auger analysis gave the constituent surface elements of TiN, T i c and TiCN coatings and their depth distributions. XRD analysis indicated a predominantly TiN( 111)-preferred orientation in the structure of TIN and TiCN coatings. The attachment of 3T3 mouse fibroblasts and human endothelial cells to TiN, T i c and TiCN coatings was investigated. The results showed that these coatings provided surfaces for normal cellular attachment, growth and morphology, which supports the tissue compatibility of the Ti-containing coatings. Keywords: Ti-containing coatings: multi-arc ion plating: biocompatibility.
1. INTRODUCTION
A protective coating on man-made implants should possess desirable biocompatibility in the environment of biological fluids, as well as suitable mechanical and tribological properties and corrosion performance for a given application. In other words, it must not have an adverse response, such as excessive deposition of fluid components, abnormal cellular growth and toxic or allergic reactions, or thrombosis in the case of blood-contacting devices. Such a coating should also possess satisfactory resistance to compression, tension, shear, wear, or corrosion when it is used in specific applications, such as total joint replacement, orthopedic pins and screws, dental prostheses, intraocular lens, etc. [ 1-41. Ti-containing ceramic coatings have been widely used in mechanical industry and surface decoration for commercial goods due to their chemical inertness, hardness, wear resistance, corrosion resistance and unique color [5-71. However, -To whom correspondence should be addressed. Tel./Fax: (86-22) 2354-0278: e-mail:
[email protected]
80
Y. Liu et al.
only little attention has been focused on their biocompatibility and comparison of different Ti-containing ceramic coatings so far [8-111. In previous studies [12-141, we used the PVD technique to synthesize DLC, CN, and TiN coatings and investigated their influence on certain cells. In this study, we have synthesized TIN, T i c and TiCN coatings using multi-arc ion plating, and have investigated the effects of surface chemical composition and structure on cell attachment. 2. EXPERIMENTAL
A DC multi-arc ion plating system with three high-purity titanium (99.95%) targets was used to synthesize TiN, T i c and TiCN coatings. The chamber was evacuated by a diffusion pump to a pressure of 6.7 x Pa before beginning coating deposition. A 3 17L medical grade steel substrate was mechanically polished to a mirror finish, and then was cleaned ultrasonically in ethanol and acetone for 10 min each before being loaded into the chamber. Before deposition, the substrates were sputter cleaned in an Ar plasma at a pressure of 2.5 Pa with -600 V bias for 5 min. DC power supplies were run at an arc-current of 55 A and arcvoltage of 20 V applied to all three Ti targets for all depositions. The main deposition parameters are shown in Table 1. The elemental compositions and their depth profiles for the coatings were determined by Auger electron spectroscopy (PHI-61OBAM) using a 3 kV electron gun. Sputter etching was carried out by 3 keV Ar' at a rate of 30 n d m i n . Relative intensities were determined from peak heights and atomic concentrations were calculated from these intensities using the appropriate sensitivity factors. To identify the phases and orientations in the coatings, X-ray diffraction was used for structural analysis of the coatings using a D/MAX 2500 diffractometer, ran at a voltage of 40 kV and a tube current of 100 mA, with Cu K, radiation at 1.54 A. The wettability of the coatings was investigated by measuring contact angles of distilled water using a contact angle analyzer at 25°C. Table 1. Typical experimental deposition parameters Coating TiN Working pressure (Pa) Nitrogen flow (sccm) Acetylene flow (sccm) Substrate bias (-V) Deposition time (min)
TIC
TiCN
1.1
1.1
2.3 200 30
-
1.1 1.7 0.6 200 30
2.3 200 30
Characterization of and cell attachment to Ti-containing coatings
81
The fibroblast is the most-often used cell in biocompatibility tests. Our 3T3 mouse fibroblasts were provided by the Institute of Hematology, Chinese Academy of Medical Science. Culture of endothelial cells on biomedical implants is an efficient way to enhance their hemocompatibility. The endothelial cell (EC) suspension was from umbilical cord of healthy human fetus. Two types of cell suspensions containing 1 x lo5 cells/ml were prepared according to our previous work [15]. All samples were sterilized in ethylene oxide prior to use. The same amount of cell suspension was cultured on 12 x 4 cm2 areas in 24-well culture plates. Cultures were performed in a 37°C incubator with a humidified atmosphere containing 5 % C 0 2 in air. Three independent samples were used for each coating with each cell type. After 1 and 2 days for 3T3 mouse fibroblasts and 72 h for endothelial cells of growth on different surfaces, the medium was removed and the cells were washed several times with phosphate-buffered saline (PBS), and fixed in methanol for 2 h at 4°C. Then the cells were washed with PBS to ensure a complete removal of methanol. After sputter coating with a layer of gold, they were observed with a scanning electron microscope (SEM). 3. RESULTS AND DISCUSSION
3.1. Chemical composition and structure
The surface element compositions for the three coatings were determined by AES. The AES spectrum in Fig. l a indicates the presence of Ti, N, C, 0 elements on the surface of TiN coating. After Art sputter cleaning for 8 min, the 0 and C peaks became very weak (see Fig. lb), which means that the 0 and C signals from the surface were due to the air exposure and oil contamination from diffusion pump. TiN coating contains mainly Ti and N elements. The distribution of elements along the depth was characterized by AES. Figure 2 shows that the concentrations of Ti and N elements in TiN coating are constant throughout the coating thickness except in the surface region. Similar results are seen for TiCN and T i c coatings in Fig. 3. The concentrations of Ti, C and N in TiCN coating and of Ti and C in T i c coating are almost constant within the thickness region examined. The results above imply that chemical bonding states of Ti with N and C are dominant in the structures of these three coatings. In order to check theAES results, XRD analysis was performed for all coatings. Figure 4 shows the XRD pattern of TiN coating. TiN coating shows a predominantly (1 11)-preferred orientation. Other peaks shown in this pattern are associated with the 317L steel substrate, except for a weak (222) orientation around 77". Compared with TiN, TiCN coating shows a reduced peak intensity of TiN( 11 1) preferred orientation (see Fig. 5 ) , which is due to lower N2 partial pressure during coating synthesis, implying that the chemical reaction of Ti and N plays a dominant role in the formation of TiCN coating. In addition, a new peak at TiN(220) orientation is produced in TiCN coating. However, no obvious crystalline phases
Y. Lit{ et al.
82
0
200
400
600
800
1000
Kinetic energy (eV)
I 0
TitN 200
400
600
800
1000
Kinetic energy (eV) Figure 1. AES spectra of TiN coating (a) before and (b) after sputter cleaning.
can be observed in the structure of T i c coating. A broad and weak band around 36" indicates a dominant amorphous structure in T i c coating.
3.2. Cell attachment and wettability The fibroblast is a cell type often used in biocompatibility tests. Its normal growth and attachment on the surface are evidence of good biocompatibility of biomaterials in the environment of biological fluids. Figure6 shows SEM images of the fibroblasts attaching to different coatings. For comparison, the result of the con-
Characterization of and cell attachment to Ti-containing coatings
h
0
200
100
300
400
Sputter time (sec.) Figure 2. Sputter depth profiles of TiN coating.
-
-
i
60S e,
-
-
Nl+Til
L)
S 0
.-
L)
Ti2
C
0 -
-
01
50
0
100
150
Sputter time (sec.) h
$.
80
.-
3
picr
..
I~
. . . .
'.
r~
.
.
.................
(b);
'
4
- ....................................................
Y
3S
60 . . . .
Q)
0
8
.-0
E 3
?
40
~
-....................
20
. L
fl
-rT........
......... ~
I
d
Til
~I
. . . . . . . . . . .-
0 L
0
I
50 100 Sputter time (sec.)
Figure 3. Sputter depth profiles of TiCN (a) and T i c (b) coatings.
1
150
83
Y. Liu et al.
84
--
n 3
n
W
.-z
3
cd
k
W
20
30
40
50
60
80
70
28 (degree) Figure 4. XRD pattern of TiN coating.
e,
Y
e
Y
p”
c
f
m
3
cd
W
,x I
.r(
8c
TiCN
-I
c, 3
20
30
40
50
60
70
80
28 (degree) Figure 5. XRD patterns of TiCN and T i c coatings.
trol 317L steel sample is also shown in Fig. 6. A low number of cells are seen on all surfaces after 1 day incubation. But the cells have the tendency to stretch out on Ti-containing coatings as compared to the cells on the control 317L steel. After 2 days incubation, compared with the low number of cells and filopodia on the control 317L steel, a large number of 3T3 fibroblasts with spindle-shape morphology are observed on Ti-containing coating (see Fig. 7). SEM pictures of TIN, TiCN and T i c coatings show that cells attached to their surfaces exhibit normal cellular morphology of polygons, maintaining physical contact with each other
Characterization of and cell attachment to Ti-containing coatings
85
Figure 6. SEM images of mouse 3T3 cells cultured on (a) the control 317L steel, (b) TiN, (c) TiCN and (d) T i c coatings for 1-day incubation at 500x magnification.
through filopodia, suggesting a favorable condition for normal cellular growth and proliferation. Compared with TiN and TiCN coatings, T i c coating exhibits a more obvious increase in the number of filopodia between the well-shaped polygons cells (see Fig. 7d). At the same time, there is no evidence of extensive cell death.
86
Y. Liu et al.
Figure 7. SEM images of mouse 3T3 cells cultured on (a) the control 317L steel, (b) TiN, (c) TiCN and (d) T i c coatings after 48 h incubation at 5 0 0 ~ magnification.
Interesting results were obtained for all coatings, when the growth of human endothelial cells on the surfaces was observed by SEM. A higher cell density and obvious tendency to stretch out on Ti-containing coatings as compared to 317L steel surface are observed in Fig. 8. This result means that Ti-containing coatings provide good surfaces for normal cellular growth and proliferation of endothelial cells. The endothelial cells attaching to the surfaces of biomaterials are able to inhibit blood coagulation in clinical application.
Characterization of and cell attachment to Ti-containing coatings
87
Figure 8. SEM images of human endothelial cells cultured on (a) the control 317L steel, (b) TiN, (c) TiCN and (d) TiC coatings after 72 h incubation.
It is well known that high wettability is supposed to be one of the primary factors enhancing cell attachment. Generally, biomaterials with higher wettability are compatible with the environment of biological fluids and are suited for growth of cells. In order to investigate the wettability of the coatings, water contact angle measurements were performed. Figure 9 shows water contact angles on all surfaces. Each point is an average of five readings. The contact angle on the control 317L steel sample is about 83". Ti-containing coatings show a decrease in the
Y. Liu et al.
88
90
1
I
I
T
T
317L
TiN
I
1 J
70
60
i t TiCN
Tic
Figure 9. Water contact angles on different surfaces.
contact angle. Its value for T i c coating is down to 57", exhibiting an improved wettability, which results in a positive effect on enhancement of cell attachment. In addition, similar to diamond-like carbon coating with desirable biocompatibility, a higher carbon atomic concentration on the surface of T i c coating may also enhance cell attachment. 4. CONCLUSIONS
A study of the chemical structure of and cell attachment to Ti-containing coatings has been made. This work shows that Ti-containing coatings provide good surfaces for normal cellular attachment, growth, and morphology of mouse fibroblasts and human endothelial cells. A more obvious enhancement in cell attachment is obtained on the surface of T i c coating, which is of great value for its clinical applications. The presence of Ti-C bonding on the surface leads to a higher polarity of the surface, which is known to enhance wettability. This enhanced wettability may be responsible for the enhancement of cell attachment to Ti-containing coatings. Acknowledgements
This work is supported by the Joint Project of Tianjin Municipal Universities and Nankai University and Tianjin University (Grant No. GJDFO1). This work is also supported by Project-Sponsored by SRF for ROCS, SEM and Tianjin Municipal University Science and Technology Development Foundation (Grant No. 200 10304).
Characterization of and cell attachment to Ti-containing coatings
89
REFERENCES 1. L. A. Thomson, F. C. Law, N. Rushton and J. Franks, Biomaterials 12, 37-40 (1991). 2. H. B. Skinner, Clin. Orthoped. 369, 83-91 (1999). 3. M. Allen, B. Myer and N. Rushton, J. Biomed. Muter. Res. 58, 319-328 (2001). 4. D. W. Murray and N. Rushton, J. Bone J. Surg. Br. 72, 988-994 (1990). 5. V. Valvoda, J. Mater. Res. 5. 2490-2495 (1990). 6. W. J. Meng and G. L. Eesley, Thin Solid Films 271, 108-116 (1995). 7. W. Li, X. He and H. Li, J. Appl. Phys. 75,2002-2006 (1994). 8. M. I. Jones, I. R. McColl, D. M. Grant, K. G. Parker and T. L. Parker, J. Biomed. Mater. Res. 52,413-421 (2001). 9. M. Browne and D. J. Gregson, Biomaterials 15, 894-898 (1994). 10. H. Behrndt and A. Lunk, Mater. Sci. Eng. A139, 58-60 (1991). 11. P. V. Kola, S. Daniels, D. C. Cameron and M. S. J. Hashmi, J. Mater. Proc. Technol. 56, 422430 (1996). 12. D. J. Li, F. Z. Cui and H. Q. Gu, J. Adhesion Sci. Technol. 13, 169-177 (1999). 13. D. J. Li, S. J. Zhang and L. F. Niu, Appl. Sutface Sci. 180,270-279 (2001). 14. D. J. Li, J. Zhao and L. F. Niu, J. Adhesion Sci. Technol. 17,751-761 (2003). 15. D. J. Li and L. F. Niu, Nucl. Instrum. Methods Phys. Res. B192, 393-401 (2002).
Adhesion Aspects of Thin Films, Vol. 2, pp. 91-102 Ed. K.L. Mittal 0VSP 2005
Adhesion issues with polymer/oxide barrier coatings on organic displays D. W. MATSON,' P. M. MARTIN,"* G. L. GRAFF,' M. E. GROSS,' P. E. BURROWS,' w. D. BENNETT,' M. G. HALL,] E. s. MAST,' C. C. BONHAM,' M. R. ZUMHOFF,' N. M. RUTHERFORD,2 L. M. MORO,' M. ROSENBLUM,*R. F. PRAIN02 and R. J. VISSER2 'Pacific Northwest National Laboratory, PO Box 999, MSIN K3-59, Richland, WA 99352, USA 'Vitex Systems, Inc., 3047 Orchard Parkway, San Jose, CA 95134, USA
Abstract-Multilayer polymer/oxide coatings are being developed to protect sensitive organic display devices, such as organic light-emitting devices (OLEDs), from oxygen and water vapor permeation. The coatings have permeation levels of approx. g/m2 per day for water vapor and << cc/m2 per day for oxygen, and are deposited by vacuum polymer technology. The coatings consist of either a base A1203 or acrylate polymer adhesion layer followed by alternating A1203/polymerlayers. The polymer is used to decouple the 30-nm-thick A1203barrier layers. Adhesion of the barrier coating to the substrate and display device is critical for the operating lifetime of the device. The substrate material could be any transparent flexible plastic. The coating technology can also be used to encapsulate organic-based electronic devices to protect them from atmospheric degradation. Plasma pretreatment is also needed for good adhesion to the substrate, but if it is too aggressive, it will damage the organic display device. We report on the effects of plasma treatment on the adhesion of barrier coatings to plastic substrates and the performance of OLEDs after plasma treatment and barrier coating deposition. We find that initial OLED performance is not significantly affected by the deposition process and plasma treatment. as demonstrated by luminosity and current-voltage curves. Keywords: Organic light-emitting device; barrier coating; permeation; adhesion; flat-panel display; organic display.
1. INTRODUCTION
The development of organic materials for electronic applications provides the promise for allowing improvement in existing electronic components, as well as for the production of novel devices not amenable to fabrication using traditional materials and methods. Organic-based electronics offer the potential for lighterweight yet more rugged devices, higher device efficiencies and vivid color dis'To whom correspondence should be addressed. Tel.: (1-509) 375-2076; e-mail:
[email protected]
92
D. W. Mntson et al.
plays. Among the organic-based electronic devices that are currently under development are organic light-emitting devices (OLEDs) [ 1-51, photovoltaic devices [6-lo], photodetectors [ 111, chemical detectors [ 121 and organic-based lasers [13-171. Despite the exciting potential for electronic components based on organic molecular species, they suffer from a potentially fatal characteristic: the organic molecules used in electronic devices are susceptible to environmental degradation from reactions with atmospheric oxygen and/or water [ 18-20]. Consequently, for these organic-based electronic devices to be practical and commercially viable, they must be hermetically encapsulated or otherwise protected to prevent atmospheric exposure [ 19-23]. Environmental degradation is an issue for organic electronics in all areas under development, but is especially pronounced in visual display applications where at least one side of the device must allow light transmission. Organic-based displays (OLEDs) can be produced on transparent impermeable glass substrates, but are then constrained by weight, breakage and geometry issues. The use of lightweight, transparent, flexible plastic substrates for OLED production would allow the potential benefits of the organic species to be more fully utilized. Unfortunately, bare plastics are far too permeable to both water and oxygen to allow their usage for this application. We have reported a transparent multilayer polymer/metal oxide barrier coating suitable for protecting organic-based electronics from oxygen and water permeation [24-261. This barrier coating can be applied to a polymer sheet using vacuum web-coating technology, with the coated polymer then being used as the substrate on which to build an organic-based display. A similar coating, applied using an in-line deposition tool, can be used to encapsulate devices built either on coated plastic or rigid glass substrates. In this paper we begin with a discussion of OLEDs and their permeation barrier requirements, followed by a discussion of the barrier coating structure and deposition technology. Specific attention will be paid to adhesion issues related to the deposition of the multilayer barrier coating on polymer substrates and directly onto OLEDs. 2. OLEDs AND THEIR BARRIER REQUIREMENTS
Light-emitting diodes (LEDs) are formed by producing a junction between n-type and p-type semiconductors and applying a voltage across it. Electrons flow into the p-type layer and holes flow into the n-type layer. Energy is released when holes and electrons combine, and light is emitted. A number of small organic molecules and polymers have been developed that, when properly doped, act as ptype or n-type semiconductors and can be used to produce organic LEDs or OLEDs. Progress in the development of OLEDs, first produced in the late 1980s [I], has been extremely rapid. This has been due, at least in part, to the vast variety of suitable organic and polymeric molecules that can be synthesized for this
Adhesion of barrier coatings to OLEDs
93
( ( (Emitted Light ) ) ) Figure 1. Basic structure of an organic light-emitting device (OLED).
application. Currently, high-resolution, high-contrast color OLED displays are being introduced into commercial markets. The basic structure of a simple OLED stack is shown in Fig. 1. In this example, the stack is produced on a transparent glass substrate and the anode is a thin layer of a transparent conductive oxide (indium tin oxide or ITO). A stabililitypromoting passivation layer is applied to the I T 0 anode, followed by p-type and n-type organic layers and a metallic cathode. A small voltage (typically 2-10 V), applied between the anode and cathode, is sufficient to drive the device. As noted in the Introduction, OLED displays are limited in terms of their resistance to environmental effects, limiting their wide-scale commercial application. There are several causes of environmental degradation of OLEDs, leading to lifetime limitations. The primary cause of damage is exposure to atmospheric water and oxygen. This exposure can lead to the oxidation and delamination of the metal cathode, as well as chemical changes within the organic layers. The conventional method used to combat this problem when using a glass substrate is to seal the device in a dry inert environment. Typically, a glass (or metal) lid is fixed in place over the device in an inert-atmosphere glovebox, using a UV-cured epoxy resin to seal the lid in place. A ‘getter’ material is also commonly incorporated within the package in order to eliminate any residual water and oxygen left within the encapsulated space. In this sealing method, the glass behaves as a perfect moisture barrier and the only method of ingress into the OLED package is through the epoxy edge seal. The problem incurred when using a flexible plastic substrate material is that polymers are very poor barriers to the diffusion of water and oxygen. In order to
94
D. W. Matson et al.
achieve the long life-times (approx. 10000 h) needed for displays, the OLED package must have an estimated water vapor permeability of less than g/m2 per day at 25°C [24]. Typical plastic substrate materials have permeabilities of lo2to lo-’ g/m2 per day at 25°C and are unsuitable for a commercial OLED product. One approach to overcome this issue is to use thin film permeation barrier coatings of dense dielectric materials on the polymer surface to inhibit diffusive processes. A similar coating can be used to encapsulate or seal the back side of an OLED built on an impermeable substrate.
3. MULTILAYER POLYMEWOXIDE BARRIER COATINGS
The multilayer polymer/metal oxide barrier coatings that were developed at Pacific Northwest National Laboratory (PNNL) and that are being commercialized by Vitex Systems, are referred to as BarixT” coatings. When a BarixTMcoating is applied to a plastic substrate, the coated product is referred to as Flexible GlassTM engineered substrate. The coating itself consists of multiple thin layers of a metallic oxide interspersed with thicker polymer layers (Fig. 2). Details of the BarixTMcoating deposition process have been presented elsewhere [24-261 and are only briefly summarized here for the sake of this discussion. The entire multilayer deposition process is performed under vacuum using
Figure 2. SEM cross-section of multilayer barrier coating. Oxide and polymer layers of the barrier coating are labeled as “0” and “P”. respectively.
Adhesion oj'barrier coatings to OLEDs
95
an in-line deposition tool or a web coater. Polymer layers (typically an acrylate composition) are applied by flash evaporation of a proprietary monomer mixture which is subsequently condensed in the liquid phase on the relatively cool substrate surface [27]. This evaporation technique enables extremely high deposition rates of up to 20 p d s . The as-deposited liquid monomer layer tends to self-level, filling depressions and covering small protrusions on the surface. Immediately following the monomer layer deposition, it is UV cured to a flexible, transparent acrylate polymer film. In Fig. 3, we show the effect of the polymer layer planarization on the surface of a commercially available poly(ethy1ene terephthalate) (PET) film. An atomic force microscopy (AFM) image of the untreated surface shows spikes higher than 15 nm. Figure 3b shows the planarization effect resulting from a 0.25-pm-thick polymer layer addition using the flash evaporation/condensation process on a sample of the same material. The addition of the polymer layer reduces the RMS surface roughness to <1 nm. The oxide layers in the BarixTMstack (20-30 nm) are applied by a reactive sputtering process from a metal (e.g., aluminum) sputtering target. The initial polymer deposit in BarixT" coatings is applied as a relatively thick layer (0.5-2.0 pm) to provide a smooth base on which to build the remainder of the barrier coating. In addition to smoothing imperfections in a substrate surface, the initial polymer smoothing layer can be used to smooth surface features on a fabricated OLED. Polymer layers deposited between the oxide layers are thinner
Bare PET
Flexible GlassTM
15 nm peaks (a) Figure 3. Atomic force microscopy images of 10 km square areas of (a) bare PET and (b) Flexible GlassTMsubstrate. illustrating the planarization effect of the polymer smoothing layer.
96
D. W. Matson et al.
(0.1-0.5 pm) and act to decouple individual layers of the dense, impermeable oxide, as well as to provide additional smoothing of imperfections developed in the coating stack resulting from normal deposition processes. By decoupling the individual oxide layers using the intermediate acrylate layers, the overall efficiency of the oxide as a barrier material is improved. The multilayer stack provides a considerably more effective barrier coating than a coating of comparable overall thickness, but consisting of single thick oxide and polymer layers. Since molecular permeation through the oxide layers occurs at defects (e.g., pinholes), by separating the barrier into multiple layers in which defects are offset from layer to layer, the pathlength through which the gas molecules must travel is dramatically increased [28]. Based on encapsulated OLED performance and reactive metal testing [29, 301, estimated O2 and H 2 0 permeation rates of <5 x g/m2 per day have been achieved for multilayer barrier coatings. Oxygen permeation rates are also well below the measurement limits of the industry-standard MOCONTMpercc/m2 per day). meation measurement technique In the case of Flexible GlassTMsubstrate, the multilayer deposition process is performed in a roll-coater operation that allows the alternating layers to be deposited sequentially. A single polymer/oxide pair, or dyad, can be deposited on each pass of the substrate around a central drum in the roll-coating chamber [27]. An in-line deposition tool has also been developed that produces a similar alternating polymer/oxide layer structure on flat substrates as they are shuttled above the monomer and oxide deposition sources. The in-line tool is suitable for encapsulating OLEDs already built on glass or Flexible GlassTMsubstrates. 4. ADHESION ISSUES RELATED TO THE B A R I X ~COATING ~ PROCESS
Because of the multilayer nature of the BarixTMstructure, adhesion issues are important considerations in establishing the barrier performance and other important properties of the coatings. Adhesion of the overall coating to the substrate, as well as adhesion at each of the ceramic/polymer interface layers within the coating is critical to the performance of the coating as an effective permeation barrier. Interlayer separation within the coating would tend to negate or reduce the decoupling effect of the intermediate polymer layers in the stack geometry, thereby reducing the effectiveness of the barrier. Also, in OLED applications where light must be transmitted through the BarixTMcoating, poor adhesion between any of the layers can result in formation of delaminated regions or bubbles, negatively affecting the aesthetic appeal of the OLED. We have found that in many cases adhesion issues are related, in either a positive or negative fashion, with other characteristics that can affect the overall performance of the barrier coating. Adhesion tests are routinely performed on Flexible GlassTMsamples. The standard adhesion acceptability test for these samples consists of tape pulls on crosshatched coatings, both as-produced in the dry state (dry pull test) and after soaking the sample overnight in deionized water (wet pull test). Adhesion grading is
Adhesion of barrier coatings to OLEDs
91
performed on the basis of the percentage of cross-hatched squares removed during the tape pull test. Samples that are top-graded for adhesion exhibit a 4%removal during the pull test. Corresponding adhesion testing is not performed on the coatings produced using the in-line deposition tool to encapsulate OLEDs. Certain factors have been designed into the BarixTMcoating process that are specifically meant to promote optimal adhesion, both between the overall coating and the substrate, and between inter-coating layers. First, wetting agents and surfactants and coupling agents are included in the standard monomer formulation to assist in promoting adhesion of the polymer layers to the underlying substrate. Second, the coating deposition chambers are outfitted with an argon plasma exposure station to chemically activate surfaces on which subsequent layers are deposited. For Flexible GlassTMthe bare polymer substrate is plasma activated, as are each of the oxide layer surfaces prior to depositing the overlying intermediate polymer layer. Often adhesion concerns must be balanced with barrier performance or other considerations when producing the BarixTMcoating stacks. For example, adhesion of the oxide-barrier-coating layer to underlying polymer layers could be improved by a plasma treatment of the polymer layer surface. However, such a treatment would also tend to roughen the polymer surface, producing a more irregular subsequent oxide-barrier layer and, thus, likely reducing the effectiveness of the oxide barrier layer and the overall efficiency of the coating. To produce the optimal oxide layer having minimal defects that might promote oxygen or water permeation, the underlying surface must be as smooth and regular as is feasible. We have also found that thin initial polymer smoothing layers tend to promote adhesion of BarixTMcoatings applied to polymer substrates. This is probably related to a positive stress/thickness relationship in the smoothing layer with thicker polymer smoothing layers possessing greater intrinsic stress and, therefore, being more susceptible to debonding. However, initial smoothing layers must be sufficiently thick to cover substrate surface irregularities and provide as smooth a platform as possible on which to build the remainder of the coating. Consequently, coating adhesion may need to be sacrificed in the interest of optimal permeation barrier performance. Coating/substrate adhesion for the Flexible GlassTMproduct has been found to depend on the plastic material used for the substrate, its as-received surface properties and the composition of the initial layer applied to the substrate. For example, PET is a common substrate material used to produce Flexible GlassTMengineered substrate. We have found that coating adhesion to this base material is optimal when a thin (approx. 30 nm) oxide layer is sputtered directly on the PET surface before the smoothing layer is applied. For other base materials (e.g., poly(ethy1ene naphthalate), PEN) the best adhesion results were obtained when the polyacrylate smoothing layer was applied directly to the substrate surface. Another substrate-related variable that was determined to affect coating/substrate adhesion was whether a scratch-resistant hardcoat had been applied to the polymer surface. Generally, the optimal adhesion for coatings on different substrate
98
D. W. Matson et al.
materials, or for the same substrate materials having different surface characteristics, must be determined through an iterative deposition and testing process. BarixTMcoatings produced directly onto OLEDs using the in-line deposition tool may be deposited with a thin (approx. 50 nm) layer of a buffer material before the remainder of the normal polymer/oxide coating is applied. Rather than an adhesion consideration, however, this buffer layer is intended to protect the highly-sensitive OLED from damage that may result from exposure to the adhesion-promoting plasma, chemical interactions resulting from direct exposure to the monomer mixture, and exposure to the UV source used to cure the polymer layers. In Fig. 4 we show current-voltage curves of OLEDs before and after BarixTMencapsulation, both with and without the inclusion of the protective buffer layer. The deviation between the before and after encapsulation curves for the device without the buffer layer is an indication that the OLED was damaged during the encapsulation process. An issue closely (or perhaps even directly) related to adhesion in multilayer BarixTMcoatings is the problem of bubble formation. The oxide layers within the coating act as highly effective gas barriers. Consequently, minute quantities of gas generated under or within the coating as a result of some aspect of the coating process or by some subsequent post-coating deposition process (e.g., a process step in building an OLED on a Flexible GlassTMsubstrate) may generate sufficient pressure to separate individual layers and form bubbles. Figure 5 shows bubbles formed in the coating on a piece of Flexible GlassTM.We have found that one coating process variable that can affect bubble formation is the power applied to the argon cleaning plasma used to treat the oxide surfaces immediately before application of the polymer layers of the coating. Application of high plasma power tends to increase bubble formation tendency within the coating. Too high a plasma power may degrade the polymer layer under the oxide layer being treated, either through a thermal process or through absorption of UV produced by the plasma source. We also know that bubble formation occurs in BarixTM-coated samples that are overexposed to UV/ozone surface cleaners commonly used for pre-treating surfaces on which OLEDs are built. Excessive exposure in the UV/ozone cleaning units tends to produce bubbles within the top couple of layers of the coating as the UV light is strongly absorbed within the polymer layers, breaking bonds that produce small volatile molecular species. One additional observation that we have noted related to bubble formation in the BarixTMcoatings is that bubbles are often associated with particles that are found within the bubbled area. We speculate that small particles generated within the coating chamber get embedded in one of the layers of the growing coating and can act as a nucleation site for bubble formation because of excess stress in the coating at that point as a result of the particle’s presence.
-
0.30 -
0.25
-
030--
025--
~
020-
aE
-c
+ 0
020-
aE
-
15--
+ 0
15--
S
P
P
L
L
010--
010--
0.05 --
0 05
0.00
-!
0.00
-r
---II 7
3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2 6.4 6.6
Voltage (V)
c
With protective layer
No protective layer
1
3.8 4.0 4 2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2 6.4 6.6
irl
2
Figure 4. Current-voltage curves of OLEDs before and arter barrier coating. The curves on the left were produced from a dcvice encapsulated without using a protective buffcr layer bcfore depositing the initial oxide barrier laycr. The curves in the right-hand side panel were generated from a device on which the protectivc buffer layer was used. The lack of change in current-voltage curves of the device on which the buffcr laycr was used indicates that no performance degradation rcsulted from deposition of the barricr coating.
W W
100
D. W. Matson et al.
Figure 5. Bubble formation in a Flexible GlassTMbarrier coating resulting from a 10 min exposure in a Jelight model 42 UV/ozone cleaner.
5. SUMMARY AND CONCLUSIONS
OLEDs and other organic-based electronic devices require protection from atmospheric oxygen and water to prevent rapid degradation and make them viable for commercial applications. Optically transparent multilayer polymer/oxide BarixTMcoatings have been demonstrated to reduce 0 2 and H20 permeation rates, making them suitable as barrier coatings in OLED applications. Application of the BarixTMcoatings to polymer substrates using a roll-coater process allows the opportunity to develop OLEDs and other organic electronic devices on robust flexible media. Adhesion of the BarixTMcoating to the substrate on which it is deposited, as well as adhesion between layers of the multilayer coating, is important to maintain the properties of the coating as a permeation barrier, as well as to maintain the appearance of the coating. Specific coating process parameters aimed at improving adhesion of the coatings to substrates and between layers of the coatings are chemical additives in the monomer mixture and a plasma cleaning stage designed to activate surfaces prior to deposition of the polymer layers. Other factors affecting adhesion of the multilayer barrier coatings include substrate composition, the order of layer deposition, post-deposition processing and particulate inclusion in the coating structure. The goal of optimal adhesion must often be balanced with other considerations that can affect performance of the coatings in specific applications.
Adhesion of barrier coatings to OLEDs
101
Acknowledgements
The authors gratefully acknowledge the support of Battelle Memorial Institute and Vitex Systems, Inc. Pacific Northwest National Laboratory is operated for the United States Department of Energy by Battelle Memorial Institute under contract DE-AC06-76RLO 1830. REFERENCES 1. C. W. Tang and S . A. VanSlyke,Appl. Phys. Lett. 51,913-915 (1987). 2. G. Gu, P. E. Burrows, S . Venkatesh. S . R. Forrest and M. E. Thompson, Opt. Lett. 22, 172-174 (1997). 3. H. Y. Low and S . J. Chua, Mater. Lett. 53, 227-232 (2002). 4. L. S. Hung and C. H. Chen, Mater. Sci. Eng. R 39, 143-222 (2002). 5. Z. Y. Xie, L. S. Hung and F. R. Zhu: Chem. Phys. Lett. 381, 691-696 (2003). 6. P. Peurnans, V. Bulovic and S. R. Forrest, Appl. Phys. Lett. 76, 2650-2652 (2000). 7. C. J. Brabec, S. E. Shaheen, T. Fromherz, F. Padinger, J. C. Hummelen, A. Dhanabalan, R. A. J. Janssen and N. S . Sariciftci, Synth. Met. 121, 1517-1520 (2001). 8. G. Hadziioannou, MRS Bull. 27, 456-460 (2002). 9. J. M. Nunzi, Compt. Rend. Phys. 66, 523-542 (2002). 10. B. A Gregg, J. Phys. Clzem. B 107,4688-4698 (2003). 11. J. Natali, M. Sampietro, M. Arca, C. Denotti and F. A. Devillanova, Synth. Met. 137, 1489-1490 (2003). 12. V. Savvate’ev, Z. Chen-Esterlit. J. W. Aylott, B. Choudhury, C. H. Kim, L. Zou, J. H. Friedl, R. Shinar, J. Shinar and R. Kopelman. Appl. Phys. Lett. 81,46524654 (2002). 13. V. G. Kozlov. V. Bulovic, P. E. Burrows and S . R. Forrest, Nature 389, 362-364 (1997). 14. G. F. Barlow and K. A. Shore, J. Mod. Opt. 47, 1921-1932 (2000). 15. S. Riechel, U. Lemmaer. J. Feldmann, T. Benstem, W. Kowalsky, U. Scherf, A. Gornbert and V. Wittwer, Appl. Phys. B 71, 897-900 (2000). 16. S. Riechel, U. Lemmer, J. Feldrnann, S. Berleb, A. G. Muckl, W. Brutting, A. Gombert and V. Wittwer, Opt. Lett. 26, 593-595 (2001). 17. N. Moll, R. F. Mahrt, C. Bauer, H. Giessen, B. Schnabel, E. B. Kley and U.Scherf. Appl. Phys. Lett. 80, 734-736 (2002). 18. S. F. Lim, W. Wang and S . J. Chua, Mater. Sci. Eng. B 85, 154-159 (2001). 19. K. Yarnashita, T. Mori and T. Mizutani, J. Phys. D 34. 740-742 (2001). 20. A. B. Chwang, M. A. Rothman, S . Y. Mao, R. H. Hewitt, M. S . Weaver, J. A. Silvernail, K. Rajan, M. Hack, J. J. Brown, X. Chu, L. Moro, T. Krajewski and N. Rutherford, Appl. Phys. Lett. 83,413-415 (2003). 21. M. D. J. Auch, 0. K. Soo, G. Ewald and C. Soo-Jin, Thin Solid Films 417, 47-50 (2002). 22. G. H. Kim, J. Oh. Y. Chu, Y. S. Yang. J. I. Lee, L. M. Do and T. Zyung, J. Kor. Phys. SOC.42, S376-S378 (2003). 23. Y. S. Jeong, B. Ratier, A. Moliton and L. Guyard, Synth. Met. 127, 189-193 (2002). 24. P. E. Burrows, G. L. Graff, M. E. Gross, P. M. Martin, M. Hall, E. Mast, C. C. Bonharn, W. D. Bennett, L. A. Michalski, M. S. Weaver, J. J. Brown, D. Fogarty and L. S. Sapochak, Proc. SPIE 4105,75-83 (2001). 25. P. E. Burrows, G. L. Graff, M. E. Gross, P. M. Martin, M. K. Shi, M. Hall, E. Mast, C. Bonharn, W. Bennett and M. B. Sullivan, Displays 22,65-69 (2001). 26. M. S. Weaver, L. A. Michalski, K. Rajan, M. A. Rothrnan, J. A. Silvernail, J. J. Brown, P. E. Burrows, G. L. Graff, M. E. Gross, P. M. Martin, M. Hall. E. Mast, C. Bonharn, W. Bennett and M. Zumhoff. Appl. Phys. Lett. 81, 2929-2931 (2002).
102
D. W. Matson et al.
27. J. D. Affinito, M. E. Gross, C. A. Coronado, E. N. Greenwell and P. M. Martin, Thin Solid Films 290-291,63-67 (1996). 28. G. L. Graff, R. E. Williford and P. E. Burrows, J. Appl. Phys. 96, 1840-1849 (2004). 29. G. Nisato, P. C. P. Bouten, P. J. Slikkerveer, W. D. Bennett, G. L. Graff, N. Rutherford and L. Wiese, Proc. 21'' Annual Asia Display, 81h International Display Workshop, Nagoya, 14351438 (2001). 30. G. Nisato, M. Kuilder, P. Bouten, L. Moro, 0. Philips and N. Rutherford, Society f o r Information Display, 2003 International Symposium, Digest of Technical Papers, Vol. XXXIV, 550-553 (2003).
Part 2 Metallized Plastics
Adhesion Aspecrs of Thin Films, Vol. 2, pp. 105-121 Ed. K.L. Mittal
0VSP 2005
Surface modification of polymers by ion-assisted reactions: An overview J. S. CHO, S . HAN, K. H. KIM, Y. G. HAN, J. H. LEE, C. S. LEE, J. W. SUNG, Y. W. BEAG and S. K. KOH" Research and Development Center, Plasma and Ion Beam Corporation, Shinnae Technotown No. 405, 485 Sangbong-Dong, Jungrang-Gu, Seoul 131-221, South Korea
Abstract-The research carried out on surface modification of polymers using the ion-assisted reaction (IAR) treatment is reviewed with outstanding results obtained regarding their hydrophilicity and adhesion. The IAR technique is an emerging technology in the field of surface modification of polymers which gives stable functional groups on surfaces and provides permanent hydrophilicity and strong adhesion of polymers. Low water contact angles, below 30°, and high surface energies, 60-70 mJlm2, are achieved on IAR-treated thermoplastic and thermosetting polymers such as polyethylene (PE), polypropylene (PP), poly(ethy1ene terephthalate) (PET), poly(viny1idene fluoride) (PVDF), polycarbonate (PC), poly(ether sulfone) (PES) and polyimide (PI). In this paper, the experimental results on IAR-treated polymers are presented and the changes in physical and chemical properties on the IAR-treated polymer surfaces are investigated by XPS, SEM and AFM. On the basis of these results, the interaction mechanisms among energetic ions, reactive gas molecules and polymer molecules involved in the IAR treatment are discussed. The improvement in adhesion between the IAR-treated polymers and other materials was explained in terms of the increased surface energy, as well as surface roughness of the polymers modified by the IAR treatment. Keywords: Surface modification; ion-assisted reaction; hydrophilicity; surface energy; adhesion
1. INTRODUCTION
Polymer materials with improved mechanical, optical and electrical properties have been widely used in a variety of industrial applications. However, their use is sometimes limited by the undesirable properties of the surface, as opposed to the very useful bulk characteristics, such as light weight, chemical inertness and high impact resistance. It is, therefore, necessary to modify the surface in a controlled manner to enhance wettability, printability, adhesion to other materials and compatibility (as in the production of blends using two immiscible polymers). The important feature of surface modification is that the surface properties of the treated material can be modified without altering their intrinsic bulk properties *To whom correspondence should be addressed. Tel.: (82-2) 3422-4100; Fax: (82-2) 3422-4105; e-mail:
[email protected]
106
J. S. Cho et al.
[l]. Several methods, including chemical, flame, corona and cold plasma treatments, have been devised for this purpose [ 2 , 31. After surface treatments, hydrophilic and/or polar groups can be formed on polymer surfaces, and these increase the polar component of surface energy of polymer materials. When polymer surfaces were modified by the above methods, however, undesirable effects, such as bond scission and carbonization, were also produced [4-61. Therefore, new surface modification methods are required to obtain polymer surfaces free of surface damage and having good wettability and high surface energy. In this paper, we review the experimental results obtained for various polymers modified by IAR treatment and investigate the changes in physical and chemical properties of IARtreated polymer surfaces. Based on these results, the adhesion enhancement between the modified polymers and other materials is explained in terms of wettability, surface energy and surface morphology. Adhesion enhancement mechanisms are proposed. Possible industrial applications of the IAR technique are discussed in the various fields of electronics, composites, biomaterials, etc. 2. EXPERIMENTAL
2. I . Sample preparation and IAR treatment Commercial polymer samples were cut into 10 x 10 mm2 sheets and were cleaned ultrasonically using methanol, isopropanol and triply distilled deionized water to remove the contamination on the surface. The samples were dried for 1 h in a drying oven. Contact angles on the untreated samples were measured and then the samples were loaded into the IAR system which is composed of an ion source, environmental gas supply, substrate holder and vacuum pumping system. The cold cathode ion source placed 50 cm below the samples was used to generate energetic ions, which were irradiated normal to the sample surfaces at room temperature. The range of ion beam energy was 500-1500 eV. The ion fluence was controlled by controlling the irradiation time at a fixed ion current density from 5 x 1014to 1 x 1017 ionskm’. The reactive gases were introduced near the sample surfaces during ion beam irradiation and the flow rate of reactive gases was varied from 0 to 10 ml/min by a mass flow controller. 2.2. Characterization
Contact angles were measured by the sessile drop method with a contact angle meter (ERMA, Tantec, USA), and each contact angle value represents an average value measured on five different samples prepared under the same experimental conditions. Test liquids used were typically distilled water and formamide (Junsei Chemical). From the contact angle data of water and formamide the surface energy was calculated. X-ray photoelectron spectroscopy (XPS) analysis was performed to determine the chemical bond environment for the modified polymer surfaces, using a Sur-
Surface modifcation of polymers by ion-assisted reactions
107
face Science Instruments 2803-S spectrometer. Atomic force microscopy (AFM) and field emission scanning electron microscopy (FE-SEM) were employed to measure the surface roughness and surface morphology, respectively. 3. RESULTS AND DISCUSSION
3.1. Wettability and surface energy
Figure 1 shows changes in water contact angle on PC as a function of ion fluence and oxygen gas flow rate. As shown in Fig. 1, the water contact angles on the PC surfaces irradiated without reactive gas are reduced from 78" to approximately 50", irrespective of the ion fluence, which are similar to those obtained by other methods such as corona discharge, plasma and ion implantation [4-61. The small reduction in water contact angle on the PC surfaces irradiated without reactive gas may be attributed to surface cleaning, surface roughness and remnant radical effect. In the case of PC surfaces modified in the oxygen gas environment, the
90 I ' t Ar+ion only
-+-
---o-
~ rion + + Oxygen gas (2 mumin) ~ rion + + Oxygen gas (4 mumin)
-4- Ar' ion + Oxygen gas (6 mumin)
F.l
3 30-
s
20
-
10 -
0 Lt'l * 0.0 io1*
I
I
m ' ' ' I
1Ol6
1017
Ion fluence (ions/cm'> Figure 1. Changes in water contact angle (in degree) on PC as a function of ion fluence at different flow rates of oxygen gas.
J. S. Cho et al.
108
changes in water contact angle are dependent on the ion fluence and the gas flow rate. The PC samples modified in oxygen gas environment have lower contact angles than those modified by Ar’ ion irradiation only, and have a minimum contact angle of 8” at optimum conditions (ion fluence = 1 x 10l6 ions/cm2, oxygen gas flow rate = 4 ml/min), which indicates a very hydrophilic and wettable surface. Therefore, it can be said that introduction of oxygen gas decreases the contact angle significantly and new functional groups are formed on the modified polymer surfaces through chemical reaction between oxygen and irradiated polymer surface [7]. It is well known that the wettability of solids and their adhesion to other materials are influenced by the surface energy of the solids. Many researchers have proposed a number of methods for determining the surface energy of solids [8]. Owens and Wendt [9] developed a method for determining the surface energy of solids, y s , by resolving the surface energy into contributions from dispersion and polar components. The dispersion, ysd, and polar, ysp, components of a solid were
90 80
P
-A-
y, : Surface energy
-0-
ySp:
Polar component
.,’-.--.
a,
0
$ rA
20
0
1015
10l6
1017
Ion fluence (ions/cm2) Figure 2. Surface energy and its dispersion and polar components for the IAR-treated PET as a function of ion fluence.
Surface modificatiori of polymers by ion-assisted reactions
109
calculated by contact angle measurements using two polar liquids and the total surface energy of the solid was obtained by the sum of two components,
Y,=
e
+ Y,”
(1)
Figure 2 shows the changes in surface energy for IAR-treated PET surfaces as a function of ion fluence. In the case of modified PET (Fig. 2), the dispersion component decreases from 32 to 18 mJ/m2, while the polar component is significantly increased from 13 to 51 mJ/m2, by ion irradiation with 8 ml/min flowing oxygen gas. It is apparent that the introduction of oxygen gas effectively increases the polar component of surface energy. From this result, the large decrease in water contact angle can be explained directly in terms of dominant increase in polar component of surface energy of the modified polymers.
3.2. Surface morphology Energetic ion bombardment causes sputtering of atoms and/or molecules on the polymer surface, as well as free radical formation, bond scission and chemical reactions. The ion bombardment is able to change the surface morphology of the substrate. It is well known that the wettability of a polymer surface is affected by the surface morphology, especially surface roughness. Wrobel et al. [IO] treated PET using plasmas initiated in various reactive gases: nitrogen, oxygen and ammonia. They explained that the increase in wettability was mainly due to the increase in surface roughness by the formation of micro-pores. Therefore, we performed AFM analysis to examine the change in surface morphology induced by energetic ion irradiation during IAR treatment. AFM images of irradiated PE surfaces are represented in Fig. 3. The rootmean-square (r.m.s.) roughness of PE surface increases to 7.6 nm at an ion dose of 1 x l O I 5 ions/cm2 compared to 5.6 nm for the untreated one. The r.m.s. roughness of PE surface increases to 7.1 nm as the ion dose increases from 1 x 10l6to 1 x 10’’ ions/cm2. This change in roughness for PE surface by ion bombardment in Fig. 3a cannot explain the noticeable reduction of contact angles sufficiently. Newly formed polar groups also play an important role in the wettability improvement. From the AFM results, it can be said that the surface damage in the IAR-treated polymers is negligible. According to Celina et al. [ 111, using a highintensity pulsed ion beam, a dramatic increase in surface area and roughness was achieved: the surfaces of PTFE, PC, PI and other materials turned into a microcellular foam. In thin film deposition where polymers are used as a substrate, this abrupt change in surface morphology and roughness of polymers is detrimental, as it can affect the resulting surface roughness of the deposited thin film. Therefore, the flatness of the IAR-treated polymer surface is advantageous in thin film deposition. The characteristics of various IAR-treated polymers are summarized in Table 1.
110 J. S. Cho et al.
Y
"E
.-
Surface modif cation of polymers by ion-assisted reactions
111
Table 1. Water contact angle, surface energy and surface roughness results for IAR-treated polymers Sample
Contact angle
("1
Surface energy (mJ/m2) Ysd
YSP
Y S
r.m.s. surface roughness (nm)
PE
Untreated IAR treated
95 28
30 18
4 51
34 69
5.6 7.1
PET
Untreated IAR treated
63 9
32 18
13 52
45 70
14.1 11.2
PC
Untreated IAR treated
78 8
15 20
12 48
27 68
1.4 2.6
PS
Untreated IAR treated
73 19
38 19
6 48
44 61
-
Untreated IAR treated
75 4
50 18
3 55
53 73
-
Untreated IAR treated
68 15
40 18
7
49
47 67
8.8 9.0
61
20
22
42
1.3
23
21
45
66
1.8
PI
PES
PMMA Untreated IAR treated
-
-
3.3. Chemical analysis Figure 4 shows the CISXPS spectra of untreated and modified PVDF surfaces. The CIScore level spectrum of the untreated PVDF includes -CH2- (286.2 eV) and -CF2- (290.8 eV) peaks, and shows a typical peak shape of PVDF. Compared to the untreated PVDF, the C,, spectra of PVDF modified at ion fluences of 5 x 10l4 ions/cm2 and 1 x 1015 ions/cm2 show that the peak intensity of the -CF2moiety drastically decreases, and new peaks related to oxygen and fluorine singly bonded to carbon appear between -CH2- and -CF2- peak positions. The -CH2peak is shifted to a lower binding energy of about 285 eV, which means the vicinity of -CH2- is changed from the most electronegative fluorine atoms to some other atoms. New bonds at binding energies > 286 eV, which indicates that -(C0)- and -(C=O)- are formed on the PVDF surface, and a remarkable reduction in -CF2- peak intensity are also seen. The CISspectra of PVDF modified by ion fluences of 1 x 10l6 ions/cm2, and 1 x l O I 7 ions/cm2 represent a sharp increase of doubly bonded carbon (=C=) and their shapes are skewed to lower binding energy. The carbonization of PVDF by high dose ion irradiation is similar to the previous results using highly energetic ion and/or heavy ion irradiation [12, 131. The change in the chemical structure of IAR treated PVDF surfaces has been explained in detail elsewhere [14].
112
J. S. Cho et al.
Binding energy (eV) Figure 4. CIScore level spectra of untreated and PVDF modified by the IAR treatment: (a) untreated, (b) 5 x l O I 4 ions/cm2, (c) 1 x 1015ions/cm2, (d) 1 x 10l6ions/cm2 and (e) 1 x lo’’ ions/cm2.
Figure 5 shows XPS spectra for (Fig. 5a) pristine PP, (Fig. 5b) IAR-treated PP and (Fig. 5c) water-washed PP after IAR treatment. IAR treatment was performed at an ion beam energy of 1000 eV with an ion fluence of 1 x 1016ions/cm2 and oxygen flow rate of 8 ml/min. C1, and 01,spectra hardly change after water washing, which means that hydrophilic groups formed by IAR treatment are not washed away. In the case of high-energy bombardment to form a hydrophilic polymer surface [ 15, 161, the concentration of newly formed functional groups decreased significantly after water washing due to the increment of solubility. The energetic ions result in the reduction of the chain length of the macromolecules and small chain length molecules easily dissolve in water. As the functional groups formed by IAR treatment do not dissolve in water, this means that the polymer surface treated by IAR treatment is not severely degraded.
Surface modification of polymers by ion-assisted reactions
113
294 292 290 288 286 284 282 538 536 534 532 530 528
Binding energy (eV) Figure 5. C,, core level spectra of (a) pristine PP, (b) IAR-treated PP and (c) water-washed PP after IAR treatment.
In order to explain the reaction mechanism induced by IAR treatment, a twostep model is suggested as shown in Fig. 6 [7]. The first step is the creation of unstable chains by the impact of energetic ions on the polymer surface that are sufficient to cause chain scission (a few eV). The second step is the formation of functional groups by interaction between the newly formed unstable chains and oxygen atoms.
J. S. Cho et al.
114
Reaction Gas
Ion irradiation
----
b
111111 -
Hydrophilic groups
~-~ -
Unstable Chains
1. Creation of unstable chains
2. Formation of hydrophilic groups
Figure 6. Schematic diagram for reaction mechanism of IAR treatment in terms of a two-step model.
50 p
i
50pm
Figure 7. SEM micrographs of HDPEhylon 66 blends; (a) untreated blend and (b) IAR-treated blend.
3.4. Applications In this section, the possible industrial applications of the IAR-treated materials having excellent wettability and adhesion characteristics are reported. Figure 7 shows the SEM micrographs of untreated and IAR-treated HighDensity Polyethylene (HDPE)/nylon 66 blends. The interfaces between HDPE and nylon 66 particles in untreated HDPE/nylon 66 blend can be identified clearly meaning that the HDPE and nylon 66 are immiscible. In the case of the IARtreated sample, however, the interfaces of particles disappear significantly, which means that HDPE and nylon 66 can be blended well. The fracture toughness of the untreated blend was found to be less than 1 J/m2, whereas that of the IARtreated sample increased to 200 J/m2 [17]. From this result, it can be said that new composite materials with advantages of each component can be made without using adhesives.
Surface modifcation of polymers by ion-assisted reactions
115
0 I
10'
1o2
,
1 1 1 1 1 1 1 1
,
, , , , , , , 1
io3
I
1o4
Frequency (Hz) Figure 8. Photographs after the adhesion test between PEDOTlPSS and PVDF film; (a) pristine PVDF, (b) IAR-treated PVDF and (c) sound pressure level (SPL) of a PVDF film speaker with the PEDOT/PSS electrode.
Figure 8a and 8b shows, respectively, the photographs after peeling off the Scotch@tape from the pristine and the IAR-treated PVDF films coated with the conducting polymer, poly(3,4-ethylene-dioxythiophene) (PEDOT) doped with poly(styrenesu1fonate) (PSS), and Fig. 8c represents the sound pressure level (SPL) of a film speaker made of PVDF films and PEDOT/PSS electrode. It has
J. S. Cho et al.
116
1.6 1.4 1.2
1.o 0.8 0.6
0.4
0.2 0.0 I
0
"
1015
1O l 6
I
1017
Ion fluence (ions/cm2) Figure 9. Peel strength of Cu films deposited on IAR-treated PI samples as a function of ion fluence.
been expected that PVDF can be used as a film speaker because of its piezoelectric property. Up to now, however, the practical usage of PVDF film as a speaker has been hampered because of poor adhesion between the PVDF film and the electrode. For the pristine PVDF sample with the PEDOT/PSS electrode, the electrode is completely detached from the PVDF surface by the Scotch tape. However, the electrode is not peeled off from the modified PVDF surface, as shown in Fig. 8b. A long-lasting flexible film speaker was fabricated and it lasted for more than 6 months. Figure 8c shows the sound pressure level (SPL) measured from the PVDF film speaker with PEDOT/PSS electrodes as reported in detail elsewhere [18]. The average SPL value with the highly conducting PEDOT/PSS electrode was 80 dB in the frequency range from 1 to 10 kHz. A Cu film on PI has been used widely in flexible circuit boards. In the case of evaporated Cu film on PI, tie-layers such as Ti, Cr and Ni are deposited on the PI before Cu deposition to improve the adhesion strength between the Cu film and the PI surface. Figure 9 shows the results of peel test for Cu films on the IAR-
Surface modification ifpolymers by ion-assisted reactions
35
117
I I
30 25
55 20
d
a
NE \
3
15
0 v
0
10
\
\
5
\e-
0
0
Ion fluence (iondcm') Figure 10. Oxygen transmission rate (OTR) of A10, film on IAR-treated PC samples as a function of ion fluence.
treated PI samples without depositing tie-layers. After the PI substrates were modified by IAR treatment, the seed Cu films with thickness of 500 nm and the thicker Cu films of 18 pm thickness were deposited by ion beam sputtering and electroless plating, respectively. As shown in Fig. 9, the peel strength of Cu layer on untreated PI is as small as 0.036 kg/cm, whereas those of Cu layers on IARtreated PI samples increase significantly and reach 1.4 kg/cm at an ion fluence of 1 x 10'' ions/cm2. From this result, a Cu film on PI without a tie-layer can be deposited successfully via the IAR treatment. Figure 10 shows the oxygen transmission rate (OTR) of A10, films with 30 nm thickness on untreated and IAR-treated PC samples, which are used as oxygen diffusion barrier in food packaging and plastic display panels. The oxygen transmission rate of A10, film on IAR-treated PC decreases significantly compared to that of A10, film on untreated PC. This result is due to the improvement of adhesion between the A10, film and the PC surface because of the decrease of interface diffusion path.
118
J. S. Cho et al.
Figure 11. Optical microscope images of PC12 cells grown in an IAR-treated Petri dish and commercial Petri dishes made by different manufacturers: (a) IAR-treated Petri dish, (b) commercial Petri dish 1, (c) commercial Petri dish 2 and (d) commercial Petri dish 3.
Figure 11 shows optical microscopic images of rat pheochromocytoma (PC12) cells grown on a IAR-treated Petri dish and commercial Petri dishes. As shown in Fig. 11, a distinct difference in cell growth is observed, which might be dependent on the surface hydrophilicity. The PS Petri dish modified by IAR shows an excellent culture ability of PC12 cells, whereas other commercial Petri dishes show a relatively low culture ability [19]. In cell culture, in order for cells to grow well, the cells should attach to the walls of cell-culture dishes. Therefore, the adhesion between cell-culture dishes and cells is an important factor. A good wettable surface of tubing materials is needed to improve the ability of a heat exchanger. Plain and low-finned tubes were treated by IAR to investigate the effects of hydrophilic surface treatment on heat transfer at the outside walls of copper tubes in an absorption chiller type heat exchanger. The experimental set-up and procedure for this study have been described in detail in Ref. [20]. Figure 12a shows that as the evaporation pressure increases, the heat transfer rate, 4, slightly decreases, resulting from the fact that the difference between
100,
1
0
80 -
-A-
Plain tubes (IAR treated)
-0-
Low-finned tubes (IAR treated)
500 a n
E
s-
60-
n
A
400
W
300
1s
n
24
6
u
0
A
40-
-
4 5
A Plain tubes (Untreated)
20
-
0' 0
A
Plain tubes (IAR treated)
0 Low-finned tubes (Untreated) 0 Low-finned tubes (IAR treated) I
200
I
10 20
I
30
I
I
-I
50
60
70
I
40
100
Psat (Torr)
-
0
10
20
30
40
50
60
70
P sat (Torr)
Figure 12. Expcrimental results of hcat transfer v e r ~ u the s saturation pressure: (a) heat transfer per unit surcace area and (b) the product of total heat transfer cocfficient, U , and heat transfer arca, A .
+ +.d
W
120
J. S. Cho et al.
the evaporator inlet temperature and the evaporation temperature decreases with the increase of chamber pressure. All heat-transfer rates obtained at several pressures for both hydrophilic-surface-treated plain tubes and low-finned tubes showed an increase compared to those of bare copper tubes. Figure 12b also shows that the effects of the surface treatment on the product, UA, of the total heat-transfer coefficient, U , and the heat-transfer area, A. The hydrophilic surface modification enhances the UA values by approximately 40% for a plain tube and 19-26% for a low-finned tube. While water on a hydrophobic surface, such as untreated tube, forms discrete sessile drops, the water on a hydrophilic surface drains as a fully wettable film. The water film is relatively very thin and the resultant thermal resistance to heat transfer is quite low. Therefore, enhancement of heat transfer efficiency through surface treatment can be explained by the fact that the thermal boundary-layer thickness has been reduced by the hydrophilic surface. 4. CONCLUSIONS
This paper deals with a new surface modification technique for polymers, the socalled ion-assisted reaction (IAR) to improve the surface properties of polymers, that has provided outstanding experimental results regarding wettability and adhesion of various polymers. The changes in water contact angle on modified polymers are explained in terms of the increased surface energy due to the formation of functional groups. From XPS analysis, it was found that more functional groups on the modified polymers were formed by ion irradiation in presence of oxygen gas than by ion irradiation alone, and the formation of functional groups is believed to be the result of chemical reactions between the irradiated polymer surface and oxygen gas. It is revealed by XPS study that the newly formed polar groups are [-(C-0)-1, [-(C=O)-] and [-(C=O)-0-1. The polar groups formed by IAR treatment are not washed away by water, signifying that the polymer surfaces treated by IAR treatment do not degrade severely. Excellent adhesion is achieved between other materials and IAR-treated polymer surfaces by the interaction of the polar groups with other materials. The results of the investigations reported here clearly demonstrate that the IAR treatment for surface modification should be beneficial in the fields of electronics and biomaterials, where the demand for enhanced surface wettability and adhesion is increasing. REFERENCES 1. R. L. Clough, Nucl. Instrum. Methods B185, 8 (2001). 2 . K. L. Mittal (Ed.), Polymer Surface Modification: Relevance to Adhesion, Vol. 2 , VSP, Utrecht (2000). 3. K. L. Mittal (Ed.), Polymer Surface Modification: Relevance to Adhesion, VSP, Utrecht (1996). 4. R. R. Lalauze, J. C. Le Thiesse, C. Pijolat and M. Soustelle, Solid State Zonics 12,453 (1987). 5 . W. Gopel, Prog. Surjiace Sci. 20, 9 (1985).
Surface modification of polymers by ion-assisted reactions
121
6. T. Susuki, T. Yamazki, H. Yoshioka and K. Hikichi, J. Mater. Sci. 23, 145 (1988). 7. J. S . Cho, W. K. Choi, H. J. Jung and S . K. Koh, J. Muter. Res. 12, 277 (1997). 8. K. L. Mittal (Ed.). Contact Angle, Wettabiliv and Adhesion, Vol. 2, VSP, Utrecht (2002). 9. D. K. Owens and R. C. Wendt, J. Appl. Polym. Sci. 13, 1741 (1969). 10. A. M. Wrobel, M. Kryszewski, W. Rakowski. M. Okoniewski and Z. Kubacki, Polymer 19, 908 (1978). 11. M. Celina, H. Kudoh, T. J. Renk, K. T. Gillen and R. L. Clough. Polym. Adv. Teclznol. 9, 38 (1998). 12. L. Torrisi, G. Ciavola, R. Percolla and F. Benyaich, Nucl. Znstrum. Methods B116, 473 (1996). 13. L. Torrisi, R. Percolla and F. Benyaich, Nucl. Znstrum. Methods B117, 387 (1996). 14. S. Han, S. K. Koh and K. H. Yoon, J. Elecrrochem. Soc. 146.4327 (1999). 15. B. M. Callen, M. L. Ridge, S. Lahooti and A. W. Neumann. J. Vac. Sci. Technol. A13. 2023 (1995). 16. Yu. I. Mitchenko, V. A. Frnin and A. S. Chegolya, J. Appl. Polym. Sci. 41.2561 (1990). 17. H. J. Kim, K. J. Lee, Y. S. Seo. S. K. Kwak and S. K. Koh, Macromolecules 34, 2546 (2001). 18. C. S. Lee, J. Y. Kim, D. E. Lee, J. Joo, S . Han. Y. W. Beag and S . K. Koh, J. Mater. Res. 18. 2904 (2003). 19. K. H. Kim. J. S. Cho, D. J. Choi and S . K. Koh, Nucl. Znstrum. Methods B175. 542 (2001). 20. H. Y. Kim and B. H. Kang, Appl. Therm. Eng. 23,449 (2003).
Adhesion Aspects of Thin Films. Vol. 2, pp. 123--144 Ed. K.L. Mittal
0VSP 2005
Contribution of chemical interactions between A1 atoms and different types of functional groups to the adhesion of Al-polymer systems R. MIX,* G. KUHN and J. FRIEDRICH Bundesanstalt fiir Materialforschung und -pri$ung, 12200 Berlin, Germany
Abstract-Monotype functionalizations with different types of functional groups at polypropylene and poly(tetrafluoroethy1ene) surfaces were achieved using the pulsed plasma-initiated homo- or copolymerization of functional groups-carrying monomers. The high degree of retained chemical structure and functional groups during the low-wattage pulsed-plasma polymerization was found to be a pre-requisite for producing well-defined adhesion-promoting plasma polymer layers as model surfaces with high concentrations of exclusively or predominantly one type of functional groups such as OH, NH2, or COOH. The maximum concentrations of functional groups were found to be 3 1 OH, 18 NH2 or 24 COOH groups/100 C atoms using allyl alcohol, allylamine or acrylic acid, respectively, as monomers. To vary the density of functional groups, a so-called plasma-initiated gasphase radical copolymerization with ethylene or styrene as “chain-extending’’ comonomer, or butadiene as “chemical cross-linker” was employed. Al-polymer systems were produced by depositing such monotype functional groups-carrying plasma polymers as adhesion-promoting interlayers onto PP or PTFE substrates followed by aluminium evaporation. The measured peel strength of aluminium deposits increased linearly with the concentration of functional groups. The ranking of the adhesion-promoting effect, H(CH2)
1. INTRODUCTION
The adhesion of metal layers deposited onto polymer surfaces is determined by the concentration and bond strength of chemical and physical interactions between the metal atoms and the functional (polar) groups at the polymer surfaces. Each type of functional group makes a specific contribution depending on its concentration to the interfacial adhesion and consequentially to the shear or peel strength of metal-polymer systemes (cf., Fig. 1). *To whom correspondence should be addressed. Tel.: (49-30) 8104-1632; Fax: (49-30) 8104-1637; e-mail:
[email protected]
R. Mix et al.
124
OH OH OH OH OH OH
OHCOOH 0 CHOOOHO
1 ( 1 1 1 1
non- specific functionalization
Me--Me--Me~-Me--Me--Me-I
,
I ,
, I
OH COOH
#
;
I
;
,
;
I
;
6 CHOOOHO
specific (monotype) functionalization
J
metallization
4/le--II/3e--Me--Me--~--~-, I
,
V
I
I
I I
I ,
I , ,
, I I
6 6 0 0 0 0
Figure 1. Principles of non-specific and specific (monotype) surface functionalizations.
To investigate the individual contribution of each type of metal-functional group (chemical) interaction it would be much easier if only one type of functional group was present at the polymer surface. Then, the interfacial adhesion is related to only one type of (chemical) interaction, depending on functional group concentration. However, the contributions due to London, Keesom and Debye interactions, hydrogen bonding and interdiffusion process to the macroscopic adhesion must be added [ 11. Four different functional groups, H(CH2), NH2, OH and COOH, on polypropylene (PP) and poly(tetrafluoroethy1ene) (PTFE) were tested for their adhesionpromoting properties towards aluminium top coatings. It is known from metalorganic chemistry that aliphatic groups do not form significant interactions with AI [2, 31. Pireaux [4] reported on interactions between AI atoms and aromatic ring of polystyrene as a precursor for Al-C formation. Bou et al. [5] also found A1-C formation when A1 was evaporated onto poly(ethy1ene terephthalate). Even unmodified polypropylene and polyethylene surfaces interact with evaporated AI atoms by charge transfer from A1 to carbon forming an A1-C bond [6, 71. The chemical interactions between A1 and primary amino groups are more controversial. Even at low temperature the amino group forms an A1+NH2 interaction (o-complex), which may re-arrange to Al-N [SI. Using Mg instead of A1 and N2 plasma for modification of PP, Nowak et al. [9] also observed the formation of Mg-N bonds. The improvement of A1 adhesion by introduction of N-containing groups was shown by Andre et al. [lo]. The OH group in poly(viny1 alcohol) (PVA) reacts with trimethylaluminium ((CH&Al) and also with metallic AI, forming A1-O-C complexes [ l l , 121, as also with Ag [ 131. Other authors interpret the electron transfer from A1 to different functional groups present at a polymer surface as acid-base interactions [14]. Friedrich and co-workers [15, 161 showed that A1 underwent a redox reaction with 0 functional groups present at 02-plasma-modified PP surfaces with formation of A1203 and reduction of 0 functional groups at the metal-polymer interface. This redox mechanism was confirmed by Silvain et al. [17] and Ding et al. [18], who found that A1 reacted with F of a fluoropolymer with formation of A1F3 at the interface.
Effect of AI-functional groups interaction on AI-polymer adhesion
125
A1 adheres very strongly to COOH-group-modified polymers [19, 201. The resulting chemical bond may be Al'C00- (salt-like) [ 131. Another mechanism prefers interaction of A1 with the carbonyl site first and and, more slowly, with the C-0 site [21] as it was also found with COOR groups [7, 22, 231. A third mechanism is the redox reaction of A1 and carboxylic groups, thus forming A1203[15]. Low concentrations (0.001-0.1 functional groups per polymer repeat unit) of appropriate reactive functional groups, introduced on the polymer surface, are found to increase the adhesion strength [24]. In some cases a very low concentration of functional groups can produce strong improvement in adhesion strength without, or negligible, altering the bulk properties of polymers. The type of functional group determines the specific resulting interactions to metals [25].Therefore, it can be concluded that the improved adhesion strengths result from formation of interfacial chemical bonds. The concentration of functional groups will influence the metal film formation, Le., cluster formation or homogeneous film formation, which had been neglected in the past and must be considered in the future (Fig. 2).
-OH OH OH OH OH OH
metallization
h$e---Me--~e---~e--~e--~e-,
,
,
I
t
0
6 6 0 6 0 0
ve--Me-- ve--Me-- Me--Me--
9, I
*
Me
.a
* I
0 Ihe--Me--Me--Me--M< 0
___)
metal film formation
island growth Figure 2. Schematic representation of adhesion in dependence on concentration of metal-functional groups anchoring points and possible metal film formation modes.
126
R. Mix et a1
In this work conventional polymers (PP and PTFE) were used as substrates. These polymers were coated with thin plasma polymers and copolymers (approx. 150 nm thick). The retention of chemical structure and functional groups during pulsed plasma polymerization was utilized for producing adhesion-promoting plasma polymer layers with high concentrations of exclusively or predominantly one type of functional group such as OH, NH2, or COOH. To vary the density of functional groups, a chemical (radical) copolymerization with ethylene as “chainextending” comonomer, or butadiene as “chemical crosslinker” was performed using the pulsed plasma. It was shown that in the plasma-produced homopolymers from allyl alcohol 95% of the theoretically expected OH groups were retained. Only small amounts of C=O “impurities” could be verified [26-281. The detection of primary amino groups in allylamine homopolymers is difficult because of incomplete derivatization reaction needed for quantification [29]. Moreover, the functionalization of surfaces with NH2 groups is superposed by 0 functionalization because of the extensive oxidation of the carbon atom in the a-position to the amino group when exposed to air [30]. Homopolymers from acrylic acid are stable on exposure to air; however, fragmentation of the acid group during the plasma process limits the yield of COOH groups to 80% of all detected 0 functionalities (and to 75% with respect to the original concentration of COOH in acrylic acid). Copolymerization was carried out to vary the density of monotype functional groups by inserting aliphatic sequences into the structure of plasma polymers. As in classic copolymerization, the main problem was to find compatible pairs of copolymers and to realize copolymerization in a broad range of mixing ratios of comonomers and to avoid the dominance of homopolymerization of individual comonomer [31, 321. It was pointed out that this process was a “pulsed-plasmainitiated chemical (radical) gas-phase copolymerization”, which strongly contrasted with the simple, so-called “plasma copolymerization”. This “old” process of plasma copolymerization also allows reaction of non-polymerizable (inert) “monomers” as demonstrated by Schuler et al. [33-351 and Yasuda and co-workers [36, 371. The most significant disadvantages of the “old” continuous-wave plasma copolymerization process are the nearly complete monomer molecule fragmentation to atoms and small fragments due to the amount of consumed energy per monomer during residence in the plasma zone, which is generally sufficient to break all chemical bonds in the monomer molecule [38-40]. The fragmentation is followed by random polyrecombination of fragments and atoms to an irregular, undefined structure. Additional defects in the deposited plasma polymer are produced by exposure of the growing copolymer layer to the plasma UV irradiation during the deposition process [41]. The “new” pulsed-plasma copolymerization process minimizes these disadvantages and strongly increases the fraction of chemical copolymers, which are produced in a pure chemical way during the “plasma-off’ period between two plasma pulses. However, as it was pointed out earlier, also with the “new” method irregularities were produced during the plasma pulses, but in much smaller quantities
Effect of AI-functional groups interactioii on AI-polymer adhesion
127
[321. Thus, it was shown that such pulsed-plasma-produced homo- and copolymers were well suited for their use as models with monotype functional groups of different types and variable density [26]. The deposition of adhesion-promoting plasma polymers with monotype functional groups onto PTFE required defluorination at polymer surfaces using H2 (reducing) plasma [42]. Without such pre-treatment the plasma polymer layers did not adhere to PTFE. It should be mentioned that the adhesion strength of evaporated A1 onto a non-modified PTFE surface is very low [43]. Usually the surface defluorination of PTFE is accomplished by reduction with alkaline metals [44-501. The application of a reducing hydrogen plasma pretreatment leads to defluorination and formation of HF, although numerous C radical sites are still present [51, 521. After finishing the pretreatment these radicals pick up oxygen from the air [53]. Post-plasma-formed oxygen-containing groups, especially peroxy groups and their degradation products, play an important role in the promotion of adhesion between thermally evaporated aluminium films and PTFE [54, 551. Another problem is the weak cohesive strength of PTFE. Therefore, a strengthening of the near-interface layer was advantageous by employing plasma-induced crosslinking (CASING) [56, 571. The degradation mechanism of PTFE at high doses during electron beam irradiation in presence of oxygen was described in Refs [58-611 and discussed in detail by Lunkwitz et al. [62]. These authors emphasised the role of COF (carboxylic acid fluoride) and COOH groups for improved wettability and adhesion properties. The perfluorocarboxylic acid groups are easily formed by hydrolysis of COF groups. Another way to produce wettability was the roughening of PTFE surfaces using CF4/02plasma [63]. 2. EXPERIMENTAL
2.1. Materials
Polypropylene (PP) foils of 100 pni (Goodfellow, UK) or 300 pm thickness (Ciba Geigy, Switzerland) were ultrasonically cleaned in diethylether bath for 15 min. Poly(tetrafluoroethy1ene) (PTFE) foil from DuPont (USA) had a thickness of about 1 111111. The monomers used were: acrylic acid, allyl alcohol and allylamine (all >99%) from Merck (Germany). All monomers were distilled before use. Ethylene and butadiene- 1,3 were supplied by Messer-Griesheim (Germany) and were used as received. 2.2. Deposition of adhesion-promoting plasma polymer layers
Deposition of plasma polymer layers was performed in a cylindrical plasma reactor of 50 dm3 volume. The design of the plasma reactor has been described in detail earlier [64]. The reactor was equipped with a pulsable radio-frequency (r.f., 13.56 MHz) generator with an automatic matching unit and an r.f. bar antenna
128
R. Mix et al.
Figure 3. Schematics of Al-PP and A1-PTFE systems.
(length 35 cm). The duty cycle of pulsing was adjusted to 0.1 and the power input was varied between 100 and 300 W. Mass flow controllers for gases and vapours, a heatable gashapour distribution in the chamber, and adjustment of pressure and flow by varying the speeds of the turbomolecular pump were used to control the pressure and the monomer flow. The gas flow was adjusted to 75 to 125 sccm and the pressure was kept at 26 Pa. The deposition rate was measured by a quartz microbalance.
2.3. Plasma pretreatment of polymers PP films were coated with plasma polymer layers of functional groups-carrying monomers without any plasma pretreatment. PTFE films were first exposed to H2 r.f. plasma (continuous-wave) for 1-1800 s at a pressure of 6 Pa and a power of 300 W, followed by deposition of adhesion-promoting plasma polymer layers (cf., Fig. 3).
2.4. Surface analysis XPS and IR analyses have been described in detail elsewhere [65]. Here, only some important facts are summarized. The XPS data acquisition was performed with a SAGE 150 Spectrometer (Specs, Germany) using a non-monochromatized MgK, or AlK,radiation with 12.5 kV and 250 W settings at a pressure of =lo-’ Pa in the analysis chamber. This instrument is equipped with a plasma reactor separated by a gate valve from the UHV system, where surface treatments can be carried out at a pressure of lO’-lO-’ Pa. XPS spectra were acquired in the constant analyser energy (CAE) mode at 90” take-off angle. Peak analysis was performed using the peak fit routine from Specs. FT-IR spectra were recorded with a NEXUS instrument (Nicolet, USA) using the ATR (Attenuated Total Reflectance) technique with a diamond or Ge cell (“Golden Gate”, Specac, UK). The IR signal is accumulated in the near-surface layer of the polymer film. The information depth depends on the material used as ATR crystal and amounts to about <1.5 ym using Ge and 2.5 pm using diamond.
Effect of AI-functional groups interaction on AI-polymer adhesion
129
2.5. Contact-angle measurements Contact-angle measurements were performed in the sessile-drop mode using water, formamide, ethylene glycol, benzyl alcohol and diiodomethane as test liquids. A Contact Angle Measuring System G2 including the appropriate software was used (Kriiss, Germany). The software employed for the calculation of surface energy is based on the approaches of Owens and Wendt, as well as Rabel and Kaelble [66-681. The surfaces of plasma polymer layers were inspected by Atomic Force Microscopy (AFM) to obtain information on their roughness. It was found that the layers were very flat for a thickness 1100 nm.
2.6. Metal deposition For measuring the AI-polymer peel strengths the plasma polymerization was performed using a plasma reactor equipped with sources for thermal evaporation (Ilmplasma 1200, Saskia, Germany). However, in case of PP the plasma-polymercoated PP samples were transferred to a separate electron beam metallizer (Auto 306, Edwards, UK). The thickness of deposited aluminium layers was adjusted to 150 nm using a quartz microbalance.
2.7. Peel strength measurements The metal-polymer systems produced are schematically shown in Fig. 3. The metal peeling technique for both AI-PP and AI-PTFE systems followed DuPont's preparation and peeling procedure. It has been described in detail elsewhere [69711. A 90" peel test was carried out at a peel speed of 25 mm/min for all Alpolymer systems. The standard deviation varied between 10 and 15%. Here, the thickness of evaporated aluminium layers was 150 nm and that of the adhesionpromoting pulsed plasma polymer layers was about 150 nm. After peeling of AI-polymer systems with adhesion-promoting plasma copolymer layers, the peeled surfaces were inspected with XPS to determine the locus of failure, i.e., whether the peel front propagated along the interface (interfacial failure) or within the material (cohesive failure). 3. RESULTS
3.1. Sugace free energy measurements The polar component of the surface energy of plasma copolymers was determined using the software described in Section 2.5. As expected, the polar component increased linearly or exponentially with growing concentration of OH groups in the plasma copolymers of allyl alcohol-ethylene or -butadiene (Fig. 4). Generally, the measured dependence confirms the absence of contributions of functional groups other than OH to the polar component. In contrast, low concentrations of OH or COOH groups produce only a very small polar component (Fig. 4). It is as-
R. Mix et al.
130
60
50 40 30 20
10 0 0
5
10
15
20
25
30
number of OH groups [per lOOC] Figure 4. Dependence of the polar component of surface energy on concentration of OH groups present in allyl alcohol-ethylene copolymers deposited employing 100 or 300 W power (duty cycle 0.1,f=103 Hz,p=25 Pa).
sumed that in this case the functional groups at the polymer surface are oriented towards the bulk and, thus, do not contribute to the polar component. The measured surface energy of the pure ethylene homopolymer (300 W) was 36 mJlm2, which is in the range of commercial polyethylene [l]. The polar component was also near zero, which qualifies the ethylene homopolymer as a pure chain-extending component in the copolymer and confirms the appropriateness of copolymers with ethylene sequences as a model surface with variable concentration of exclusively one type of functional group. The dispersion component was determined to be 35 d l m 2 which is comparable to that of polyethylene [l].Thus, the existing imperfections in the structure of the pulsed-plasma ethylene homopolymer (C=C double bonds, branched structures and other inhomogeneities) did not influence the dispersion component noticeably. For the surface energy of pulsed-plasma polymerized poly(ally1 alcohol) homopolymer a value of 5 1.6 d l m 2 was determined. Using butadiene as a comonomer in mixtures with allyl alcohol for both 100 W and 300 W powers a nearly linear increase in polar component with the concentration of OH groups was found (Fig. 5 ) . With ethylene and butadiene as comonomers the polar component increases much more with the concentration of OH groups in the case of 100 W. The butadiene homopolymer produced with 300 W has a surface energy of 38.5 mJ/m2 and possesses a higher
Effect of Al-functional groups interaction on Al-polymer adhesion
30
100 w
131
r.
I
25
-
20
-
I I I I II /
I
I I
0 0 °
I
0
I
15
300W
0
I / / 0
-
/+ I0
/0
4
10-
P@
I I
9 , /
- - %@J-
3-
- 0 -
.
,
/
,
*
5-
,----__--
o!
0
,.4-
_ _ r e
~
I
l
3
~
6
9
l
12
~
15
l
18
~
21
l
24
27
~
30
number of OH groups [per lOOC] Figure 5. Dependence of the polar component of surface energy on concentration of OH groups present in allyl alcohol-butadiene copolymers deposited employing 100 or 300 W power (duty cycle 0.1,f=103 Hz,p=25 Pa).
polar component (8.5 mJlm2) than that of the homopolymer formed at 100 W (3 mJlm2), thus confirming the assumption that more regular and defined structures are produced at “softer” (here, lower power input) plasma conditions. The dependence of measured polar components for the allyl alcohol-butadiene plasma copolymers produced at 100 and 300 W on concentration of OH groups showed a non-linear behaviour which is not well understood yet (cf., Fig. 5 ) . The polar component increases slowly at low OH concentrations but it increases very fast at higher OH concentration. The acrylic acid-butadiene copolymers show an exponential increase of surface energy in dependence on the concentration of COOH groups introduced (Fig. 6). Thus, especially at low concentrations of COOH, the polar component becomes smaller than expected. Also in this case, a linear correlation between the concentration of COOH groups and the polar component was assumed. However, in the case of low concentration of COOH groups these groups may be pointing away from the surface towards the bulk resulting in a lower polar component.
3.2. Peel-strength measurements of Al-plasma-produced homopolymer-PP systems Figure 1 shows of the differences between non-specific and specific (monotype) functionalizations of polymer surfaces for evaluating the contribution of each functional group to the adhesion (interaction) between metals and polymers.
l
~
132
R. Mix et al.
number of COOH groups [per lOOC] Figure 6. Dependence of the polar component of surface energy on concentration of COOH groups present in acrylic acid-butadiene copolymers deposited employing 100 W power (duty cycle 0.1, +lo3 Hz, p=25 Pa).
Such modified surfaces/interlayers are good models for studying the interaction between metal atoms and monotype functional groups in terms of metal-polymer peel strengths. Therefore, after evaporation of aluminium onto such model surfaces the peel strengths of the AI-plasma homopolymer-PP systems were measured. Also, it was confirmed that layers of pulsed-plasma-polymerized ethylene homopolymer did not promote any adhesion to A1 when applied in A1-PP systems as an adhesion-promoting interlayer. With NH2-groups-containing layers very weak peel forces were produced, in contrast to the use of OH and especially COOH-groups-containing interlayers (Fig. 7). Thus, the interactions between aluminium and the monotype functional groups of homopolymers depend strongly on the type of functional group in the order: COOH >> OH >> NH2 > H(CH2) [26].
3.3. Peel-strength measurements of Al-plasma copolymer-PP systems In the case of allyl alcohol-ethylene plasma copolymers, the maximum adhesion (650 N/m) was measured at 27 OH per 100 C atoms, which is significantly higher than the peel strength for the case of pure allyl alcohol (homo) polymer (80 N/m). Thus, the above ranking of the adhesion-promoting efficiency of different functional groups of homopolymers changes slightly if the peel strength of allyl alco-
Effect of AI-functional groups interaction on AI-polymer adhesion
133
700
600
E
500
& 5
400
w 0
-
300
c)
200
100 0 COOH
OH
NH2
H (CH2-CH2)
Figure 7. Bar diagram showing the adhesion-promoting effect of pulsed-plasma deposited homopolymer interlayers in AI-PP systems (dark gray), for copolymers (light gray).
hol-ethylene plasma copolymers is considered: COOH 2 OH B NH;? > H(CH2) (cf., Fig. 7). A nearly linear dependence of peel strength on the density of OH, NH2 and COOH functional groups was observed in the range of 0 to 27 OH, 0 to 15 NH2 and 0 to 10 COOH/100 C (cf., Fig. 8). Using poly(al1yl alcohol) the peel strength increases linearly with growing concentration of OH groups and the peel front propagates along the plasma polymerPP interface (interfacial failure). A plateau of maximum peel strength (650 N/m) at concentrations of 27-29 OH groups/100 C atoms was observed. At these concentrations partially or complete cohesive failures occurred on peeling. The pure allyl alcohol plasma polymer is tacky and weak and shows a low cohesive strength resulting in low peel strength and a pure cohesive failure within the allyl alcohol homopolymer layer. NH2 groups also showed a linear increase of adhesion strength but, as expected from the chemical point of view, on a low level (cf., Fig. S), resulting in a pure interfacial failure at the AI-plasma polymer interface and the lowest maximum peel strength to AI (cf., Fig. 7). COOH groups produce the highest peel strength to AI (cf., Fig. 8). In both cases of deposition of acrylic acid-ethylene or -butadiene pulsed-plasma copolymers (100 W, duty cycle 0.1) over a wide range of COOH concentrations (0-10 COOH/100 C) the AI peel strength depends linearly on the concentration of functional groups. Higher concentrations of COOH groups (>lo COOH/100 C) do not increase the peel strength further. This plateau in peel strength for higher concentrations of COOH groups is characterized by specimens with pure cohesive failures on peeling. The interpretation is that the interactions between A1 and COOH
R. Mix et a1
134 I
700 -
600 500 -
400 300 200 100 0I
I
I
I
I
I
I
I
I
0 5 10 15 20 25 30 concentration of OH, NH, and COOH groups [per 100 C] Figure 8. Dependence of AI peel strength in Al-plasma polymer-PP systems on the type and density of functional groups in plasma polymer interlayers.
groups become too strong; thus, the peel failure changes from purely interfacially to a cohesive failure within the polypropylene substrate [26]. This adhesion promotion by deposition of plasma polymer interlayers bearing monotype functional groups was also compared with 0 2 plasma pretreatment of polymer surfaces, which were also evaporated with A1 [26]. It should be remembered that this O2 plasma treatment produces a broad variety of different types of 0 functional groups at polymer surfaces. In this case, the maximum peel strength was measured after introduction of 20 0 atoms per 100 C atoms onto the surfaces of PE and PP. This result corresponds well also to the value of 15 0 per 100 C atoms for similar treatments and systems given by Wu [l]. These values of 15-20 0 functional groups also correspond well to the needed concentrations of functional groups at the surface of plasma polymer interlayers to establish maximum adhesion. As discussed before, in this case 10 COOH and 27 OH groups/100 C atoms were needed to produce maximum peel strength (each 600-650 N/m). It must be remembered that in the case of O2 plasma treatment OH, COOH and other 0 species are produced. The difference between these two types of adhesion promoting plasma modifications results in the lower maximum peel strength in the case of O2 plasma treatment (400-450 N/m) compared with that in the case of the deposition of plasma polymer interlayers (600-650 N/m). Here, it can be assumed that the O2 plasma modification causes polymer degradation at the surface, thus forming a weak boundary layer.
Effect of AI-functional groups interaction on Al-polymer adhesion PTFE as received
160-.
135
-.F
140-
.. 0.. 0
120100-
80 -
.10s
H,- plasma exposure time
[SI
Figure 9. XPS-measured 0 and F concentrations at PTFE surfaces 1’3.time of exposure to the hydrogen plasma (P=IOO W, p=6 Pa, cw plasma).
3.4. Plasma pretreatment of PTFE surfaces A few variants for producing well-adhered A1-PTFE systems were tested: (a) deposition of monotype functional groups-carrying plasma polymers as adhesion promoters onto virgin PTFE, (b) H2 plasma pretreatment of PTFE followed by deposition of an adhesion-promoting plasma polymer layer, (c) combined H2 and NH3 plasma pretreatment followed by deposition of an adhesion-promoting plasma polymer layer and (d) H2 plasma pretreatment of PTFE alone. Process a did not give any A1-PTFE bond strength because of poor adhesion of plasma polymers to the PTFE surface. Processes b and c provided adequate and strongly adhered A1-PTFE systems, whereas process d produced moderate to highly adhered A1-PTFE systems. Similar to the results described in Refs [72-741, defluorination in a continuouswave or pulsed (0.1 ms plasma “on”, 0.9 ms plasma “off’) hydrogen r.f. plasma leads to a minimum in the F/100 C ratio of 24 (35 [75]) and a maximum in the 0/100 C ratio of 19 (6.4 [75]) (Fig. 9). The introduction of oxygen is most likely due to post-plasma reactions of plasma-produced C radical sites at the PTFE surface by reaction with molecular oxygen when the samples were transferred from plasma reactor to XPS spectrometer with interim contact with the air [55, 761.
R. Mix et a1
136
20000 ....
18OOOj
16000
2
14000 12000
-10s
300
I
....... 30 s -. - .-, . 60 s -..- ..-. 120 s ......... 480 s ................ -. -. - -. -.
+-’
. . . . *-J unireaieu
298
i\
960 s 1920 s
296
294
292
290
288
286
284
282
binding energy [eV] Figure 10. CISpeaks of PTFE after varying exposure time to the hydrogen plasma (P=lOO W. p=6 Pa, cw plasma).
This post-plasma reaction is unavoidable, but in this case it may help to promote the adhesion between the deposited plasma polymer layer and PTFE. In Fig. 10, characteristic changes in the bimodal shape of the CISsignals can be observed by prolonging the time of exposure to the hydrogen plasma. It should be noted that the FA00 C ratio slightly increases and the O/C ratio (ex situ XPS measurements) significantly increases at treatment times longer than 120 s (Fig. 9).An explanation for the re-increase of the F/100 C ratio may be the increased temperature at longer treatment times and, therefore, enhanced hydrophobic recovery [77]. On the other hand, a simple abrasion test (wiping with a cotton cloth) and the solvent stability test (6 h rinsing in tetrahydrofuran, THF) of the defluorinated PTFE surface show that this near-surface layer becomes unstable when exposed to hydrogen plasma for more than 10 s. The respective XPS spectra became very similar to the spectrum of untreated PTFE. Therefore, for further plasmachemical processing and metal deposition a H2 plasma treatment time of 10 s was chosen. Figure 11 presents changes in relevant XPS peaks on applying H2/NH3 combined plasma. The idea of these experiments was to defluorinate the PTFE surface and additionally to introduce N functional groups as adhesion-promoting functional groups. The reduction of PTFE (defluorination) and introduction of H- and N-containing groups at the PTFE surface (3-5 N/100 C) was verified. Moreover,
Effect of AI-f~inctionalgroups interaction on Al-polymer adhesion
137
binding e n e r g y [eV] Figure 11. O,,, N1,, F,, and C I Speaks of PTFE after exposure to combined H2 (600 s) and NH3 plasma (60 s).
oxygen incorporation was also observed as discussed in the case of reduction with pure H2 plasma. Here, earlier results should be mentioned which showed that reduction of PTFE with evaporated Na metal or in a Na plasma, adopting the wet-chemical process for the vacuum experiment, led to the formation of carbide-like carbon species represented by a C1, subpeak at a binding energy (BE) of 283.5 eV [78]. The asymmetry of this C1, peak to higher binding energies suggests also the existence of C=C double bonds and 0-bonded C species. This interpretation was also given in Refs [1, 47, 79, 801. In contrast, hydrogen or ammonia plasma reduction of PTFE produces preferably a CH, component with a BE = 285 eV. The carbidelike species at BE = 283.5 eV could not be detected. Therefore, it was concluded that a complete degradation of the PTFE backbone did not take place as in the case of the alkaline metal reduction. It can be speculated that the hydrogen reduction at short treatment times (
R. Mix et a1
138
interfacial failure between A1 and PTFE 1 mixed failure between A1 and PTFE, AI and tape failure between tape and AI / I
$
4004
poly(acry1ic acid)
300-
' R
\
z
U
.C, 250-
,I
-1 .'. *
bl)
c Q)
3'" *
e
.
I . . . . . . . .\ <"""
- 1
i
Q)
poly(ally1 alcohol) 0
' \ *
150
, plasma I .
0 0 3 6 9 1 2
300
600
900
treatment time [SI Figure 12. A1 peel strength vs. time of plasma polymer interlayer.
Hzplasma pretreatment and type of adhesion-promoting
3.5. Peel strength measurements of AI-PTFE systems
3.5.1. Hydrogen plasma pre-treatment of PTFE The dependence of A1 peel strength on hydrogen-plasma-pretreated and then metallized PTFE is shown in Fig. 12. In comparison to the metallized untreated PTFE the peel strength is significantly increased. Only 10 s pretreatment is sufficient to increase the A1 peel strength to the maximum. However, very prolonged hydrogen plasma pretreatment (>lo00 s) results in lower peel strength. Moreover, another peel locus is observed. The failure was observed between the supporting tape
Effect of Al-functional groups interaction on Al-poljmer adhesion
139
(glued onto Al) and Al; thus, A1 could not be peeled off. The additional NH3 plasma modification, resulting in the introduction of N functional groups at the PTFE surface, did not increase the peel strength further.
3.5.2. Hydrogen plasma pretreatment of PTFE and deposition of plasma polymer layers The additional introduction of an adhesion-promoting pulsed plasma polymer interlayer onto the hydrogen plasma pretreated PTFE substrate improved the peel strength further to a range of 350-400 N/m, limited by the adhesion of the adhesive (supporting) tape onto the evaporated A1 layer. Because of this limitation of the peel test, the measured peel strength did not show a significant difference between OH and COOH groups at the PTFE interface (Fig. 12). However, the monotype-functionalized PTFE interface gave higher peel strength than the pure hydrogen-plasma-modified PTFE without plasma polymer interlayer. The locus of failure during peeling at maximum peel strength was identified by XPS by analyzing the peeled A1 and PTFE surfaces of H2 plasma-modified and pulsed-plasma allyl alcohol coated PTFE evaporated with A1 (200 nm). At maximum peel strength, the peeled PTFE surface showed an XPS CISsignal at a binding energy of 291 eV, characteristic of unmodified PTFE, and the peeled metal surface exhibited a signal near 285 eV that was assigned to CH components. Thus, the peel front must have propagated along the interface between unmodified PTFE and the H2 plasma-modified PTFE near-surface layer. 4. DISCUSSION
4.1. Contribution of chemical bonds to resulting adhesion strength The metal-polymer interaction and, therefore, the peel strength is based on the sum of discrete chemical bonds between aluminium and functional groups as described before. As expected, CH2-CH2 groups do not show any interaction with aluminium and, therefore, no peel strength could be determined. Using primary amino groups also very low peel strength was measured. In contrast, OH groups form strong chemical bonds to A1 (Al-alcoholates) and COOH groups form saltlike bonds and, thus, high peel strengths were measured. As regards the interactions between freshly evaporated A1 atoms and aliphatic chains in polyethylene and polypropylene, controversial information has been obtained. The proposed formation of A1-C bonds [5, 61 could not be confirmed [81, 821 or these did not influence the adhesion strength. Since aliphatic plasma polymers from ethylene or propylene have significant structural anomalies in comparison to the respective classical products [28], which should rather increase the adhesion strength than maintain it at a constant (low) level, so it is clear that the A1-CH2 interactions do not contribute to the adhesion of AI. Primary amino groups form Me-N bonds in contact with metals [8, 91. Using alkanethiol SAM model layers with NH2 end-groups and evaporating A1 the for-
140
R. Mix et al.
mation of A1-N bonds was observed [81]. However, the authors stated that impurities in the SAM layer (oxidation products, carbamate formation) contributed significantly to the A1-SAM interaction and were always superimposed on the original A1-NH2 interactions. The A1-NH2 interactions may also form Lewis acid-base interactions [83]. As shown here, the measured low peel strength of A1 to NH2-group-modified polymer surfaces indicates weak interactions in the AlNH, system. OH groups form A1-0-C bonds known as alcoholates [12]. The measured maximum adhesion strength confirms the existence of strong chemical bonds as shown in Figs 8 and 12. This was evidenced by the extensive consumption of evaporated AI in interfacial bond formation and redox reactions [15]. Metallic AI was only detected when thick A1 layers were evaporated (ea. 0.2 monoatomic A1 layer) [3]. All OH groups at the surface are consumed by reaction with AI atoms [81]. Generally speaking, A1 evaporated onto polymer surfaces always reacts with 0 present in C=O, C-0-C and OH groups. COOH groups strongly interact with metallic AI with the result that all carboxylic groups are consumed [81]. A1 was attached preferentially to the C=O site, thus forming A1..CLO interactions [84] but followed by formation of (weak) interactions to the C-0 site [21]. The end-product of the AI-COOH reaction is the formation of AI2O3.This reaction can be considered as a redox reaction [15, 801 because of the reduction of C-0 sites and the oxidation of Al. The other possibility is the formation of ionic, salt-like interactions. In interpreting the peel-strength dependence on concentration of functional groups the different slopes of the linear line sections, starting from the point of origin, reflect the strength of respective metal-polymer bonds (cf., Fig. 8). Thus, about 10 COOH and 27 OH/100 C are needed to reach 650 N/m peel strength (maximum peel strength before appearance of cohesive failure) whereas NH2 or CH2-CH2 groups allow, at best, weak interactions. These values correlate very well with the binding energies of A1 bonds with OH and COOH groups of organic compounds in AI-alcoholates or AI-salts. 4.2. Modelling the dependence of adhesion strength on concentration of functional groups at the polymer siirface The partially linear increase of adhesion strength with growing number of functional groups at the polymer surface corresponds very well with the results obtained with metal-polymer systems where functional groups are present in the polymer matrix [ 11. For example, systems of A1 with poly(viny1 chloride)-maleic acid copolymers were produced [85, 861. The copolymers were synthesized with variable concentrations of COOH groups. The dependence of lap shear strength of A1 on the concentration of functional groups (COOH) in the chemically synthesized copolymers shows the same type of peel curve as in the case of the Alplasma polymer-PP systems. Also in this case, joint strength increases rapidly with concentration of COOH and then levels off. The curve follows the Langmuir
Effect of AI-functional grcuips interaction on Ai-polyiner adhesion
141
adsorption isotherm, of = o f , m ~ q,(bc/l+bc), = where of,,is the fracture strength at complete surface coverage, 0 the fractional surface coverage, c the concentration of functional groups in the bulk and b is a constant [l]. The plateau in the peel strength may arise either from the saturation adsorption of the functional groups at the interface, or from cohesive failure of the adhesive or adherend. In our case, the cohesive failure of the PP substrate appears to be the mechanism. Considering the results of peel strength measurements, presented in Figs 6 and 7, the following relative ranking in adhesion improvement can be derived for the contribution (efficiency) of the respective AI-X interactions to the adhesion strength and, therefore, to the peel strength (N/m) of AI-PP system: COOH=65: OH=24: NH2=1; H(CH+O. In depositing A1 layers onto model surfaces with very low concentrations of (monotype) functional groups one problem remains and must be investigated in the future. This problem is concerned with the mode of metal film formation, i.e., whether homogeneous film formation or island growth (cf., Ref. [87]). This behaviour may have influence on resulting peel strength. However, the linearity of the peel strength vs. concentration of functional group dependence shows that there does not exist any significant influence of film formation process. As mentioned before, some deviations from the expected behaviour were observed if the concentration of functional groups was very low. 5. CONCLUSIONS
Adhesion-promoting plasma polymer interlayers, possessing a maximum of 3 1 OH or 18 NH2 or 24 COOH groups per 100 C atoms, were produced by applying low wattages and using the pulsed plasma technique. By carrying out a pulsedplasma-initiated chemically-dominated copolymerization, the density of monotype functional groups could be varied continuously between 0 and 31 OH, 0 and 18 NH2, and 0 and 24 COOH groups per 100 C atoms. Using such functional groups-carrying model layers as adhesion-promoting interlayers in A1-PP systems the peel strength of evaporated aluminium layers increased in the order: H(CH2)
142
R. Mix et al.
REFERENCES 1. S. Wu, Polymer Interface and Adhesion, Marcel Dekker, New York, NY (1982). 2. D. M. Brewis and D. Briggs, Polymer 22, 7 (1981). 3. G. L. Fisher, A. Hooper, R. L. Opila, D. R. Jung, D. L. Allara and N. Winograd, J. Electr. Spectr. Rel. Pkenom. 98-99, 139 (1999). 4. J. J. Pireaux, Syntk. Met. 67, 39 (1994). 5. M. Bou, J. M. Martin, Th. Le Mogne and L. Vovelle, Appl. Surf: Sci. 47, 149 (1991). 6. P. Marcus, C. Hinnen, D. Imbert and J. M. Siffre, Surf: Interf: Anal. 19, 127 (1992). 7. B. M. DeKoven and P. L. Hagans, Appl. Surf: Sci. 27, 199 (1986). 8. K. J. Klabunde, Ckemistly of Free Atoms and Particles. Academic Press, New York, NY (1980). 9. S . Nowak, M. Collaud, G. Dietler, P. Schmutz and L. Schlapbach, ,Sur$ Interf: Anal. 20, 416 (1993). 10. V. Andre, Y. de Puydt, F. Arefi, J. Amouroux, P. Bertrand and J. F. Silvain, ACS Symp. Ser. 440,424 (1990). 11. S . Akhter, X.-L. Zhou and J. M. White, Appl. Szu$ Sci. 37, 201 (1989). 12. A. Cotton and G. Wilkinson, Advanced Inorganic Chemistry. Wiley, Chichester (1966). 13. L. J. Gerenser, ACS Symp. Ser. 440, 434 (1990). 14. J. L. Droulas, Y. Jugnet and Tran Minh Duc, in: Metallized Plastics 3: Fundamental and Applied Aspects, K. L. Mittal (Ed.). pp. 123-140, Plenum Press, New York, NY (1992). 15. J. Friedrich, I. Loeschcke and J. Gahde, Acta Polym. 37,687 (1986). 16. J. Friedrich, in: Polymer-Solid Interfaces, J. J. Pireaux, P. Bertrand and J. L. Bredas (Eds.), pp. 443459. Institute of Physics Publishing, Bristol (1992). 17. J. F. Silvain, J. J. Ehrhardt and P. Lutgen, J. Adhesion Sci. Tecknol. 5, 501 (1991). 18. S.-J. Ding, Q.-Q. Zhang, D. W. Zhang, J.-T. Wang, Y.-D. Zhou and W. W. Lee, Appl. Sur$ Sei. 178, 140 (2001). 19. J. Schultz and L. Lavielle, Makromol. Ckem. Macromol. Symp. 23, 343 (1989). 20. J. D. Rancourt, J. B. Hollenhead and L. T. Taylor, J. Adhesion 40, 267 (1993). 21. Qu, T. Lee, M. Chtaib, J. J. Pireaux and R. Caudano, in: Polymer-Solid Interfaces, J. J. Pireaux, P. Bertrand and J. L. Bredas (Eds.), pp. 245-242. Institute of Physics Publishing, Bristol(l992). 22. A. K. Chakraborty, H. T. Davis and M. Tirrell, J. Polym. Sci. Part A Polym. Chem. 28, 3185 (1990). 23. Lj. Atanasoska, S. G. Anderson, H. M. Meyer 111, Z. Lin and J. H. Weaver. J. Vac. Sci. Tecknol. AS, 3325 (1987). 24. W. C. Wake, in: Recent Advances in Adhesion, L. H. Lee (Ed.), pp. 285-293. Gordon and Breach, New York, NY (1973). 25. C. Bischof and W. Possart, Adkusion, Akademie-Verlag, Berlin (1983). 26. J. Friedrich, G. Kuhn, R. Mix, A. Fritz and A. Schonhals, J. Adhesion Sci. Tecknol. 17, 1591 (2003). 27. J. Friedrich, W. Unger, A. Lippitz, I. Koprinarov, A. Ghode, Sh. Geng and G. Kuhn, Composite Intef. 10, 139-172 (2003). 28. J. Friedrich, R. Mix, G. Kuhn, I. Retzko, A. Schonhals and W. Unger, Composite Interf. 10, 173-223 (2003). 29. M. Muller and C. Oehr, Sur$ Coating. Tecknol. 116-119, 802 (1999). 30. J. Friedrich, G. Kuhn, R. Mix, I. Retzko, V. Gerstung, St. Weidner, R.-D. Schulze and W. Unger, in: Polyimides and Other High Temperature Polymers: Synthesis, Characterization and Applications, Vol. 2, K. L. Mittal (Ed.), pp. 359-387, VSP, Utrecht (2003). 3 1. H.-G. Elias, An Introduction to Polymer Science, VCH, Weinheim (1997). 32. R. Mix. G. Kuhn and J. Friedrich. Surf: Coating. Tecknol. 174-175, 81 1 (2003). 33. H. Schiiler and L. Reinebeck, Z. Natuforsck. 6a, 271 (1951). 34. H. Schuler and L. Reinebeck, Z. Naturj+orsck.7a, 285 (1952). 35. H. Schiiler and L. Reinebeck, Z. Naturforsch. 9a, 350 (1954).
Effect of ill-functional groups interaction on AI-polymer adhesion
143
36. H. Yasuda and C. E. Lamaze, J. Appl. Polym. Sci. 17, 1519 (1973). 37. H. Yasuda, H. C. Marsh, M. 0. Bumgarner and N. Morosoff, J. Appl. Polym. Sci. 19. 2845 (1975). 38. H. K. Yasuda, ACS Symp. Ser. 108, 37 (1979). 39. J. F. Friedrich. PhD thesis, Academy of Sciences of GDR, Berlin (1974). 40. J. Friedrich, J. Gahde, H. Frommelt and H. Wittrich. Z. Faserforsch. Textiltechn. / Z. Polymerenforsch. 27, 599 (1976). 41. J. Friedrich. I. Loeschcke, H. Frommelt, H.-D. Reiner, H. Zimmermann and P. Lutgen, Polym. Degrad. Stabil. 31, 97 (1991). 42. J. Friedrich, J. Gahde, H. Frommelt and H. Wittrich, Faserforsch. Textiltechiz. / Z. Polymereizforscli. 27. 604 (1976). 43. C.-A. Chang, Y.-K. Kim and A. G. Schrott, J. Vac. Sci. Technol. AS, 3304 (1990). 44. H. Brecht, F. Mayer and H. Binder, Angew. Makromol. Chem. 33, 89 (1973). 45. D. M. Brewis and I. Mathieson, Proceed. CZP’97, pp. 81-85, Societe Francaise du Vide, Paris (1997). 46. J. F. Friedrich, I. Koprinarov, R. Giebler, A. Lippitz and W. E. S. Unger, J. Adhesion 71, 297 (1999). 47. N. J. DeLollis and 0. Montoya, Adhesives Age 12, 36 (1969). 48. R. C. Snogren, Handbook of SLirfuce Preparation, Palmerton, New York, NY (1974). 49. C. A. Costello and T. J. McCarthy, Macromolecules 20. 2819 (1987). 50. M. S . Shoichet and T. J. McCarthy, Macromolecules 24, 982 (1991). 51. D. T. Clark and D. R. Hutton, J. Polym. Sci. PartA Polym. Chem. 25, 2643 (1987). 52. N. Inagaki, S. Tasaka and Y. W. Park, J. Adhesion Sci. Technol. 12, 1105 (1998). 53. M. C. Zhang, E. T. Kang, K. G. Neoh and K. L. Tan, J. Adhesion Sci. Technol. 13, 819 (1999). 54. E. E. Wirlitsch. N. A. Krotowa, N. N. Stefanowitsch, A. J. Wilensky, B. A. Radzig and T. A. Wladykina, Plast. Kautschuk 19, 17 (1972). 55. Y. Momose. Y. Tamura, M. Ogino and S . Okazaki, J. Vac. Sci. Technol. A10, 229 (1992). 56. R. H. Hansen and H. Schonhorn, J. Polym. Sci. B4, 203 (1966). 57. J. R. Hall, C. A. L. Westerdahl, M. J. Bodnar and D. W. Levi, J. Appl. Polym. Sci. 16, 1465 (1972). 58. W. Schnabel, in: Aspects of Degradation and Stabilization of Polymers, H. H. G. Jellinek (Ed.), Elsevier, Amsterdam (1978). 59. M. I. Bro, E. R. Lovejoy and G. R. McKay, J. Appl. Polym. Sci. 7,2121 (1963). 60. A. Ferse, K. Lunkwitz, H. Grimm, P. Dietrich. G. Engler, U. Gross, D. Prescher and J. Schulze, Acta Polym. 30. 348 (1979). 61. A. M o r a , E. Occhiello and F. Garbassi, Angew. Makromol. Chem. 180, 191 (1990). 62. K. Lunkwitz. W. Burger, U. Lappan, H.-J. Brink and A. Ferse, in: Polymer Surface Mod$ication: Relevance to Adhesion, K. L. Mittal (Ed.), p. 349, VSP. Utrecht (1996). 63. C. Haag and H. Suhr, Appl. Phys. A47. 199 (1988). 64. J. Friedrich, I. Retzko, G. Kuhn, W. Unger and A. Lippitz, in: Metallized Plastics 7: Fundamentals and Applied Aspects, K. L. Mittal (Ed.). pp. 117-142, VSP, Utrecht (2001). 65. I. Retzko, J. F. Friedrich, A, Lippitz and W. E. S. Unger, J. Elecrr. Spectr. Relat. Phenom. 121. 111 (2001). 66. D. K. Owens and R. C. Wendt, J. Appl. Polym. Sci. 13, 1741 (1969). 67. W. Rabel, Farbe Lack77,997-1009 (1971). 68. D. H. Kaelble, Physical Ckemistn of Adhesion, Wiley-Interscience, New York. NY (1970). 69. J. F. Silvain, J. J. Ehrhardt and P. Lutgen, J. Adhesion Sci. Technol. 5, 501 (1991). 70. J. Friedrich, P. Rohrer, W. Saur, Th. Gross, A. Lippitz and W. Unger. S u Coating. ~ Teclznol. 59, 371 (1993). 71. J. F. Friedrich, W. Unger. A. Lippitz, Th. Gross, P. Rohrer, W. Saur, J. Erdrnann and H.-V. Gorsler, in: Polymer Surface Modifcation: Relevance to Adhesion, K. L. Mittal (Ed.), pp. 49-72, VSP, Utrecht (1996).
144
R. Mix et al.
72. M. K. Shi, A. Selmani, L. Martinu, E. Sacher, M. R. Wertheimer and A. Yelon, in: Polymer Surface Modification: Relevance to Adhesion, K. L. Mittal (Ed.), p. 73, VSP, Utrecht (1996). 73. T. G. Vargo, J. A. Gardella, Jr., A. E. Meyer and R. E. Baier, J. Polym. Sci. Part A Polym. Chem. 29,555 (1991). 74. M. Charbonnier, M. Romand, M. Alami and T. M. Duc, in: Polymer Surface Modification: Relevance to Adhesion, Vol. 2, K. L. Mittal (Ed.), p. 3, VSP, Utrecht (2000). 75. T. Yagi and A. E. Pavlath, J. Appl. Polym. Sci. Appl. Polym. Symp. 38,215 (1984). 76. J. Friedrich, G. Kuhn, U. Schulz, K. Jansen, A. Bertus, S. Fischer and B. Moller, J. Adhesion Sei. Technol. 17, 1127 (2003). 77. T. Yasuda, T. Okuno, K. Yoshida and H. Yasuda, J. Polym. Sei. Part B Polym. Phys. 26, 1781 (1988). 78. J. F. Friedrich, W. E. S. Unger, A. Lippitz, R. Giebler, I. Koprinarov, St. Weidner and G. Kuhn, in: Polymer Surface Modification: Relevance to Adhesion, Vol. 2, K. L. Mittal (Ed.), p. 137, VSP, Utrecht (2000). 79. M. L. Miller, R. H. Postal, P. N. Sawyer, J. G. Martin and M. J. Kaplit, 1.Appl. Polym. Sci. 14, 257 (1970). 80. D. W. Dwight and W. M. Riggs, J. Colloid I n t e ~Sei. 47, 650 (1974). 8 1. A. Hooper and D. L. Allara, in: Metallized Plastics 5&6: Fundamentals and Applied Aspects, K. L. Mittal (Ed.), p. 203, VSP, Utrecht (1998). 82. A. Ringenbach, Y. Jugnet and T. M. Duc, in: Polymer Surface Modification: Relevance to Adhesion, K. L. Mittal (Ed.), p. 101, VSP, Utrecht (1996). 83. J. Kurdi, F. Arefi-Khonsari, M. Tatoulian and J. Amouroux, in: Metallized Plastics 5&6: Fundamental and Applied Aspects, K. L. Mittal (Ed.), p. 295, VSP, Utrecht (1998). 84. J. L. Droulas, “Chimie interfaciale des systemes metal-polymere”, PhD thesis, UniversitC Villeurbanne, France (1992). 85. A. D. McLaren and C . J. Seiler, J. Polym. Sei. 4, 63 (1949). 86. I. Lee and R. P. Wool, Macromolecules 33, 2680 (2000). 87. F. Faupel, R. Willecke and A. Thran, Mater. Sei. Eng. R22, 1 (1998).
Adhesion Aspects of Thin Films, Vol. 2, pp. 145-154 Ed. K.L. Mittal
0VSP 2005
Deposition of aluminum on three-dimensional polymeric substrates 0. KNOTEK, E. LUGSCHEIDER. K. BOBZIN, M. MAES and A. KRAMER" Materials Science Institute, RWTH Aachen Universiv, Augustinerbach 4-22, 0-52062 Aachen, Germany
Abstract-In the present investigation, 3-D polymer samples were coated with an aluminum layer. Aluminum is a well-known shielding material for electromagnetic compatibility (EMC) applications. The aim was to achieve a dense and uniformly deposited aluminum layer all over the 3-D substrates. The samples were coated by low-temperature magnetron sputter processes using either pulsed or DC power supplies. The main focus of this investigation was the uniformity of the coating over the substrate geometry and the adhesion between the coating and the substrate. In order to determine the uniformity of the coating, the thickness of the aluminum layer was evaluated by crater grinding tests. Further, scanning electron microscopy (SEM) analysis was used on the crosssectional areas of fractured samples. Coating adhesion was evaluated by scratch, Scotch tape and lattice cut tests. The results showed good adhesion between the aluminum layer and polymer substrates. Best results were achieved by reducing the deposition pressure. Adhesion was found to be independent of the choice of the power supply. Coating thickness, on the other hand, proved to be power supply dependent. Pulsed sputter processes gave higher deposition rates compared to DC sputter processes. Keywords: Pulsed sputtering; PVD-coated polymers; coating thickness distribution; coating adhesion.
1. INTRODUCTION
In recent years, an increasing number of additional electronics in cars like navigation systems or Electronic Stability Programmes (ESP) have been installed. With the integration of many electronic devices the question of protection against electromagnetic influences becomes important. An EMC-compatible component can be attained by a multifunctional polymer component [l, 21. The opportunity to use polymeric materials as construction materials offers the advantage of light weight constructions due to the low weight of plastics. Polymer components with EMC coatings offer an inexpensive, large-scale production capable alternative. *To whom correspondence should be addressed. Tel.: (49-241) 80-95578; Fax: (49-24 1) 80-92264; e-mail:
[email protected]
146
0. Knotek et al.
The big challenge is the joining of polymers to metals because of their different material characteristics. Typical coating methods are electroplating, hot pressing or physical vapor deposition (PVD) [3, 41. Especially PVD processes show great potential for coating polymers. Technological developments, like pulsed power supplies, reduce the process temperature to about 100°C. In this temperature range, PVD technology becomes more and more interesting for coating polymers. Particularly, thermoplastics often have melting points in the range 100-180°C. By developing new low-temperature sputtering processes, PVD becomes an attractive alternative method for metallization of plastics. In this paper, the results of aluminum coatings on polycarbonate substrates are discussed. The focus of the experiments was to study the influence of DC and pulsed magnetron sputtering processes on the coating thickness and the uniformity of the coating. In addition to the coating thickness distribution, coating adhesion was also investigated. Crater grinding tests [ 5 ]and SEM micrographs helped to evaluate the coating thickness at specified measurement points. A detailed model of the distribution of the coating thickness was ascertained on closer examination of all specified measurement points. A comparison of the coating thickness distributions which were achieved either by pulsed sputter deposition or by DC sputter deposition showed differences between the types of sputter deposition processes. Crater grinding tests gave a first impression of coating adhesion. The form and appearance of the crater margins revealed different coating adhesion levels. Further, scratch, Scotch tape and lattice-cut tests characterized qualitatively the coating adhesion. 2. EXPERIMENTAL
Samples for the experiments (Fig. 1) consisted of three polycarbonate parts: a triangle, a quadrant and a semicircle. The individual parts were fitted together by screws. This was done in order to manufacture three-dimensional samples which could be easily demounted into plane parts for further analysis. The semicircle was selected as base plate. The triangle and quadrant were fixed at the semicircle at defined positions at an angle of 90" to each other. The distribution of the coating thickness over these three-dimensional samples was the main focus of these experiments along with the adhesion between the coating and substrate material. Especially the differences due to different power supplies employed were examined. Both DC and pulsed magnetron sputtering processes were used for coating processes. The experimental set-up was the same for all deposition processes. Figure 2 shows a detailed view of this set-up. The semicircle was arranged perpendicular to the aluminum target. The distance between the top of the sample and the target was approximately 30 mm and between the bottom and the target 60 mm. Triangle and quadrant were positioned with an angle of 45" to the target. This arrangement should provide maximum information about the coating process
Aluininwn on polymer substrates
147
Figure 1. Samples used for coatings.
4
75 mm
+
Figure 2. Schematic view of the experimental set-up for deposition processes.
and the influences of the different power supplies on the distribution of the coating thickness. All coating processes were performed using a laboratory coating system Leybold Heraeus 2400. The pulsed power supply used was a Bipulsar 2000 (Eifeler Werkzeuge, Rastatt, Germany). Below, a short description of the coating procedure is presented. After fabricating the polycarbonate samples they were cleaned in an ultrasonic bath with alcohol and dried under nitrogen. The samples were positioned beneath the aluminum target. After evacuation, the coating process started with the sputtering of the target. Aluminum targets usually oxidise very fast. The sputtering of the target had the aim to remove the alumina layer grown on the target surface just before the real deposition process. An etching process was used with the aim to prepare the substrate surface for deposition. The coating deposition parameters are given in Table 1. Process pressure and power were chosen to be variables for pulsed sputter and DC sputter processes. In addition to these parameters, the pulse sequence was var-
148
0. Knotek et al.
Table 1. Coating parameters for DC and pulsed sputter depositions Pulsed sputter depositionsb
DC sputter depositions
No.
Pressure (Pa)
Power (W)
No.
DC 1 DC2
1 1
150 200
Pulsed 1 Pulsed 2
DC3 DC4 DC5"
1 0.5
0.5
(PSI
Pulsed 3 Pulsed4"
300 200 300
ton Itoff
Pressure (Pa)
20.5 119.5
1
20.5 I19.6
0.5
8.2131.8
1
20.5 / 19.8
1
"Etching process differed in that a mixture of argon and oxygen gases (100 : 20) was used instead of a pure argon atmosphere. bMean power for pulsed sputter deposition was 300 W.
0
measurement point
Figure 3. Specification of measurement points
ied while the frequency was fixed at 25 kHz. A rectangular single negative halfwave pulse was chosen, based on earlier experience with pulsed sputter deposition on polymer substrates [ 6 ] .The power was kept constant for pulsed sputtering. After the deposition, the samples were cooled in an inert environment. After coating, the samples were disassembled into pieces. The separate parts were evaluated with regard to coating thickness, distribution of the coating thickness and adhesion. Crater grinding tests and SEM analysis, which were used on cross-sectional areas of fractured samples, determined the coating thickness at specified measurement points (see Fig. 3). The distribution of the coating thickness over a polycarbonate part was determined from these results. The characterization of adhesion between the coating and the substrate was made by scratch, Scotch tape and lattice-cut tests. Also, the crater-grinding tests helped to specify the coating adhesion. The adhesion can qualitatively be evaluated by the appear-
Alurniizunz on polymer substrates
149
sample
tion zone ting ctuation
Figure 4. Crater grinding test. For a crater grinding test, the sample is fixed in the sample fixture. which is positioned at a tilt angle to the vertical. This ensures that the grinding ball has a point contact to the sample surface. The ball rotates while actuated by an engine shaft. This movement grinds a crater in the sample surface. This grinding procedure is supported by dropping a water-based emulsion, which contains diamond particles, on the rotating ball. In the ground crater one recognizes the coating and the substrate by different looking rings and circles, respectively, because the colours of the materials are not the same. One can draw conclusions from the form of the crater margin and from the form of the transition zones between different crater rings about the adhesion of different material layers. Margins and transition zones, which have a clear ring form, are signs of good adhesion. Bursts and delaminations in the margins or transition zones are signs of poor adhesion.
ance of the crater margins and of the transition zones between different materials (see Fig. 4). Sharp crater margins and clearly visible transition zones are signs of good adhesion of different coating layers. Burst margins and transitions zones are signs of bad adhesion. 3. RESULTS AND DISCUSSION
Metallization of polymers can only be successful if the coating material adheres to the polymer substrate. Metals and polymers have very different material characteristics, which is an important factor when joining these materials. For example, the difference between the thermal expansion coefficients of different materials groups is very high. Sputter processes for polymer metallization are developed with regard to the thermal expansion coefficients of the substrate and target material. In a sputter process the substrate is warm while it is coated. This means that the substrate is expanded in the deposition process. After deposition the coated substrate cools down and shrinks. Thus, residual stresses can be induced in the coating. The thermal expansion of the polycarbonate substrate is three times higher than that of aluminum coating. That is why lowering the deposition temperature for coating of thermoplastics is of vital importance. As an alternative to DC sputtering, pulsed sputtering is used. The advantage of pulsed sputtering is that the same power used for DC sputtering is applied in very short pulses by highly energetic particles, and this is the reason for a reduction of the temperature load of the substrates. There is only a short time available in which heat can be transferred to the substrate surface by highly energetic particles. Furthermore, the highly energetic particles offer the possibility to improve the adhesion of the coating system.
150
sample “DC5”
0. Knotek et al.
sample “DC2”
number = measurement point Figure 5. Comparison of crater grindings of two different samples. Here one can see photographs of crater grindings of the triangle parts of samples DC5 and DC2. These pictures are compared to each other with regard to adhesion. The form of the crater margin is a manifestation of good or poor adhesion. The crater grindings of sample DC5 show sharp circular margins, which means that the adhesion is good. At some crater grindings of sample DC2, especially numbers 3, 4 and 5, the form of the margin is not circular. One can see delaminations at the crater margin; this means poor adhesion.
Pulsed and DC sputter deposited coatings showed nearly similar adhesion behavior. All coated samples passed both Scotch tape and lattice cut tests. Even in scratch tests the critical load for all the samples was nearly the same. The values of critical load varied between 3 and 4 N. The critical load is defined as the load at which cracks appear in the scratches - according to the definition of Ollivier and Matthews [7]. Differences in adhesion were seen in crater grinding tests. The samples coated at reduced process pressure showed better results regardless of the type of sputter technique applied. In Fig. 5 , two DC sputtered samples of the same geometry (triangle) are compared. The left hand pictures show crater grinding tests on the sample where aluminum was deposited at a pressure of 0.5 Pa. Sharp transitions zones between the coating and substrate are observed in the craters which is a sign of good adhesion. On the right-hand side of Fig. 5 ground craters at measurement points of a sample coated at a pressure of 1 Pa are presented. Frayed-out margins (numbers 3, 4 and 5 ) are signs of low adhesion between the coating and substrate. The craters at measurement points 1, 2 and 6 have sharp margins. It should be noted that adhesion decreases with increasing distance from measurement point to the target. The craters of all samples were evaluated by arbitrary classification into adhesion classes from 1 to 6, where a class-1 adhesion means an optimal sharp margin of the crater. Comparing the coating thickness and the adhesion class for measurement points, it can be seen that no correlation between coating thickness and adhesion class was found. The distribution of the coating thickness over the sample geometry was also investigated. In the following, graphs of coating thickness distributions are presented (Figs 6-8). Each graph shows the coating thickness at the specified measurement points of triangle, quadrant or semicircle substrate geometry (see Fig. 3). On the left-hand side, Figs 6, 7 and 8 show the graphs for the coating thicknesses
Aluminurn on polymer substrates I
DC sputter processes
- 3
Pulsed sputter processes
-
E
p 3.5
f 2
8 2.5
2 2 5
~
i
15 1
::1.5
+ I
Pulsed 1 i -
Pulsed 2 iX
s
05 0 L__ 1
I
I
Pulsed 3
__ --
05
tDC5
2 3 4 5 measurement point
6
Pulsed 4
" ' 1
4 5 measurement point 2
6
3
Figure 6. Distribution of coating thickness of triangle substrate geometry.
Pulsed sputter processes
DC sputter processes
+
-
+DC1 DC2
Pulsed 1
~
DC3
I
Pulsed 2
*DC4 Pulsed 3
+De5 .-
1
2 3 4 5 measurement point
I
1
2
3 4 5 measurement Doint
6
*Pulsed 4
Figure 7. Distribution of coating thickness of quadrant substrate geometry.
I
1 1
DC sputrer processes
I
Pulsed sputter processes
~+-DC1 -c
DC2 81
PLsed 1
DC3
Pulsed 2 Pssed 3
-DC4 1
2
3 measurement
4
5
eDC5
. . .
1
2
- --
3 4 measurement point
Pulsed 4 5
Figure 8. Distribution of coating thickness for semicircle substrate geometry.
from the DC sputter processes and on the right-hand side the graphs of the results of the pulsed sputter processes. Thus, the coating thicknesses, achieved by DC and pulsed sputter processes, can be compared easily. The graphs for DC sputtering show that the deposition rate increases with increasing power. This is true for all observed geometries. There is an interesting point that the triangle and the quadrant show a dependence of the coating uniformity on the deposition pressure. Lower deposition pressure seems to give a higher uniformity of the coating thickness. For the semicircle geometry, this result could not be confirmed. A dependence of coating uniformity on deposition pressure was not recognized for this ge-
0. Knotek et a1
152
ometry. However, the coating thickness depends on the process pressure. Higher process pressure leads to higher coating thickness. If the coating thickness of the different pulsed sputter processes are compared it can be noted that the coating thicknesses achieved by the processes “Pulsed 1” and “Pulsed 2” are higher. This effect depends on the pulse sequence (Table 1). For this reason the deposition rate in the pulsed sputter process is dependent on the applied pulse sequence. Furthermore, it is interesting that the “Pulsed 4” process had the lowest coating thickness for all geometries. This can be attributed to the different etching processes (Table 1). In the etching process a mixture of argon and oxygen gases was used. The lower deposition rates are due to poisoning of the target surface. A similar etching process was used for “DC 5”. The coating thickness from this process was in the range from the other DC sputter processes. An explanation could be that the etching pressure was lower for this process. The low pressure resulted in a lower partial pressure of oxygen and thus, the poisoning effect of the target was negligible.
I
Pulsed I / DC3
-+Pulsed 31 DC3
-*
*
I
1
2
3 4 measurement point
5
Pulsed 2/ DC5
6
Figure 9. Comparison of the ratios of pulsed to DC deposition rates for quadrant substrate geometry.
I
3
9
P z e d I / DC3
m2
+ Pulsed 3/DC 4 -- 1
Pulsed 2/ DC5
1
2
3 measurement point
4
1
5
Figure 10. Comparison of the ratios of pulsed and DC deposition rates for semicircle substrate geometry.
Aluminum on polymer substrates
153
Comparisons of the ratios of pulsed and DC sputter deposition rates are shown in Figs 9 and 10. The ratios are for pulsed and DC sputter processes which had comparable deposition parameters. Figures 9 and 10 show that the pulsed sputter processes generally had higher deposition rates than those for DC-sputtered coatings. Especially, the measurement points, which were not placed in the direct line of sight, show higher coating thicknesses by pulsed deposition processes. Furthermore, an influence of the pulse sequence in pulsed sputter processes on the deposition rate can be seen. From these results it is clear that the aim to deposit aluminum uniformly along a complex substrate was not achieved. 4. CONCLUSIONS
A comparison of DC and pulsed sputter processes with regard to coating thickness distribution and coating adhesion for metallization of plastics was performed. As a substrate material polycarbonate was chosen. It is inexpensive and is used in wideranging applications. Metallization of plastics plays an important role for EMCcompatible components. Aluminum is well-known as shielding material for EMC applications. That is why polycarbonate was coated with aluminum. The material joining is a challenge caused by their different material characteristics. Therefore, a low-temperature sputter process was developed. This pulsed-vapor deposition process allows metal coatings on polymers without inducing too much residual stress in the coating. The deposition process was performed with either DC or pulsed power supply. The coated three-dimensional samples were characterized by crater grinding test and SEM analysis for the distribution of the coating thickness. It was shown that pulsed power supply gave higher deposition rates. Further, the distribution of coating thickness was more uniform by pulsed processes. The coating adhesion was evaluated by scratch, Scotch tape and lattice cut tests. The different tests showed no differences between the applied power supplies. Crater grinding tests helped to determine that deposition processes with lower deposition pressure achieved better adhesion results, regardless of the choice of power supply. A completely uniform aluminum coating was not obtained using any of the power supplies. Acknowledgements
The authors wish to acknowledge the financial support by the DFG (Deutsche Forderungsgemeinschaft - German Research Society) within the framework of the project Lu 232/74-1. REFERENCES 1. D. R. J. White, A Handbook on Electromagnetic Shielding Materials and Performance, Don White Consultants, Gainesville, VA (1980). 2. D. Gwinner, Interference Technology Engineer’s Master (ITEM), p. 138, R&B Enterprises, West Conchohocken, PA (1993).
154
0. Knotek et al.
3. K. L. Mittal (Ed.), Metallized Plastics 2: Fundamental and Applied Aspects, Plenum Press, New York, NY (1991). 4. R. Suchentrunk (Ed.), Kunststoff-Metallisierung - Handbuch fur Theorie und Praxis, Leuze Verlag, Saulgau. Germany (1991). 5. European Standard EN 1071-2, European Committee for Standardization. 6. E. Lugscheider. K. Bobzin, M. Maes and A. Kramer. Thin Solid Films 459, 313-3 17 (2004). 7. B. Ollivier and A. Matthews, J. Adhesion Sci. Technol. 8, 651-662 (1994).
Adhesion Aspects of Thin Films, Vol. 2, pp. 1555166 Ed. K.L. Mittal 0VSP 2005
Improvement of metal adhesion to silicone films: A ToF-SIMS study A. DELCORTE,'.=S. BEFAHY,' C. POLEUNIS,' M. TROOSTERS2 and P. BERTRAND'
'
Unite' de Physico-Chimie et de Physique des Mate'riaux, Universite' Catholique de Louvain, Croix du Sild 1, B-1348, Louvain-la-Neuve, Belgium 2Neurotech SA, Place des Peintres 7/003, B-1348 Louvain-la-Neuve, Belgium
Abstract-This report describes a fundamental study of strategies aimed at improving metal adhesion to poly(dimethylsi1oxane) (PDMS) films. The considered routes involve, in particular, the 2 keV Ar' ion bombardment of the PDMS film, which proves to be very efficient, as indicated by the tape test. Complementary treatments include thermal annealing of the PDMS film and its cleaning with selected solvents. At each step of the study, time-of-flight secondary ion mass spectrometry (ToF-SIMS) is used as a diagnostic tool for the assessment of the sample surfaces. Via a sample preparation procedure involving deposition of minute amounts of gold (20 nmol/cm2) on the surface to enhance ionization (Metal-Assisted SIMS or MetA-SIMS), ToF-SIMS reveals the presence of short PDMS chains on the surface of the films. The A r' ion bombardment/hexane-cleaning treatments of the films remove this oligomer overlayer and, thus, lead to a significant adhesion improvement of the titanium layer. Therefore, the ToF-SIMS results strongly support a scenario in which the adhesion of titanium to untreated films is limited by the presence of short, mobile oligomers on the PDMS surface.
Keywords: Adhesion; SIMS; metallization: poly(dimethylsi1oxane); PDMS.
1. INTRODUCTION
Silicone rubber is routinely used for biomedical applications and, in several cases, the polymer surface has to be covered with a thin layer of a biocompatible metal such as titanium. The adhesion of a metal film to a silicone surface, however, constitutes a challenging issue. Several strategies have been proposed over the years to enhance hydrophilicity or hydrophobicity of PDMS, and cell or coating adhesion to PDMS, including bombardment of the polymer surface by keV ions [1, 21, by laser beams [3,4] and exposure to plasmas [ l , 5-10]. The reasons why ion-bombardment and plasma^To whom correspondence should be addressed. Tel.: (32-10) 473-582; Fax: (32-10) 473 452: e-mail:
[email protected]
156
A . Delcorte et a1
treatment strategies provide a dramatic adhesion improvement for metals have been discussed in the literature. In general, it is believed that the initial lack of adhesion is caused by the presence of a weak boundary layer (WBL) at the surface of the polymer. For instance, in the case of PDMS, this WBL is expected to result from segregation of low-molecular-weight chains in the surface region. In this hypothesis, the role of the plasma or ion-bombardment pretreatment is to eliminate the WBL via cross-linking [l]. Some indirect evidence of a cross-linked surface after pretreatment has been provided by electron micrographs. In this paper, we show that the specific nature of the pristine and pre-treated surfaces can be elucidated using a novel and powerful chemical surface analysis procedure based on secondary ion mass spectrometry (SIMS). Recently, gold metallization has been introduced as a new sample preparation procedure for signal improvement in organic SIMS [ l l ] . It was shown in that article and subsequent contributions that not only the quasi-molecular ion intensities but also the yields of positive and negative fragment ions were significantly improved by the metallization procedure [ 12-14]. With respect to the traditional cationization method used in SIMS, consisting in the deposition of dilute analyte solutions on clean metal surfaces [ 15-17], the so-called Metal-Assisted SIMS (MetA-SIMS) procedure has the distinct advantage that it can be applied to all types of samples, irrespective of their nature and thickness. In particular, real-world, bulk organic samples, often difficult to analyze by SIMS because of electric charging effects, prove to be excellent candidates for MetA-SIMS. Using the MetA-SIMS procedure, it is demonstrated that the pristine PDMS surface is covered by an overlayer of low-molecular-weight oligomers. Cleaning the sample with hexane or bombarding it with 2 keV Ar' ions results in the disappearance of these oligomers. The adhesion improvement, however, is much higher with the latter treatment. It is our belief that Ar' ion-beam sputtering also creates reactive sites on the surface that eventually lead to a stronger binding of the evaporated metal. 2. MATERIALS AND METHODS
2.1. Material 2.1 . I . PDMSfilm The silicone film used in this study was MED-4750 from NuSil Technology (a division of Dow Corning). The elastomer is thermally cured using a Pt-catalyst to form a cross-linked network. 2.2. Methods 2.2.1. Film pretreatments The Ar'-ion-bombarded samples (1" series) were prepared using a 2 keV, defocused Ar' ion-beam. with a total ion fluence of 5 x l0l5 ionskm'. The hexane-
157
Improvement of metal adhesion to silicone films
Table 1. Results of adhesion tests performed on Cu-metallized PDMS films with or without pre-treatment (hexane cleaning/&+ ion irradiation)
Scotch@tape
Pristine PDMS
Hexane-cleaned PDMS
Ar'-ion-bombarded PDMS
-
+I-
+
+, deadhesion at the tape-metal
-.
interface; deadhesion at the metal-PDMS interface
cleaned samples (2ndseries) were first annealed at 373 K for 1 h before rinsing for 2 min in pure n-hexane (p.a. grade, from Vel). 2.2.2. Film metallization The metal evaporation (Ti and Cu for adhesion tests; Au for SIMS analyses) was carried out in an Edwards evaporator at an operating pressure of approx. mbar and a deposition rate of 0.1 nm/s (quartz crystal monitor). For the adhesion tests, an adhesion promoting layer of Ti of thickness 15 nm was followed by a 100 nm layer of Cu. 2.2.3. Adhesion tests After metallization, simple adhesion tests were performed for the three types of PDMS surfaces (pristine film, Ar'-ion-irradiated film and hexane-cleaned film). Two kinds of tapes were used for the tests: Scotch@and Tesa'. The results of the Scotch@tape tests are summarized in Table 1. 2.2.4. Pretreatment for analysis For MetA-SIMS analyses, the PDMS films were metallized by evaporating 20 nmol of Au per cm2 on their top surface. This amount was shown to be the optimum with respect to secondary ion yield enhancement in the case of gold metallized polystyrene oligomer films [ l 11. 2.2.5. Secondary ion mass spectrometry (SIMS) The secondary ion mass spectra were acquired in a PHI-EVANS Time-of-Flight SIMS (TRIFT 1) using a 15 keV Ga' beam (FEI 83-2 liquid metal ion source; approx. 550 pA DC current; 22 ns pulse width bunched down to approx. 1 ns; 5 kHz repetition rate for the mass range 0-5 kDa) [ 181. The experimental set-up has been described in detail elsewhere [19]. The ToF-SIMS spectra were obtained by collecting the secondary ion signals in the mass range 0 < m/z < 5000 for the 600 s bombardment of a 180 x 180 pm2 sample area, corresponding to a fluence of 3.5 x 10" ions/cm2. To further enhance the measured intensities, the secondary ions were post-accelerated by a high voltage (7 kV) in front of the detector.
158
A. Delcorte et al.
3. RESULTS
The aim of this study was to elucidate the modifications occurring at the surface of a PDMS film after pretreatments, namely 2 keV Ar' ion bombardment and rinsing with hexane. For this purpose, a new sample preparation procedure for SIMS analysis (MetA-SIMS) is proposed. In the results, we first discuss the effects of the MetA-SIMS procedure on the measured mass spectra of pristine PDMS samples. Afterwards, the influence of Ar' ion bombardment is described and analyzed. Finally, the results of the Ar' ion irradiation procedure are compared to the chemical cleaning of the PDMS surface in pure hexane. Before describing the details of the ToF-SIMS study, Table 1 summarizes the results obtained when performing the Tape test (Scotch@)for the three series of samples considered in this study (pristine PDMS film, Ar'-ion-irradiated film and hexane-cleaned film). The results are best for the Ar' ion irradiated film (deadhesion at the tape-metal interface). In contrast, the metal film is almost completely removed by the tape for the pristine films. The hexane-cleaned samples show an intermediate behavior, with regions where the metal sticks and regions where it is torn away by the tape. Nevertheless, these samples remain conductive. In the discussion (Section 4), the results of this macroscopic test are tentatively explained using the wealth of information provided by the SIMS analysis of the samples. 3.I , SIMS and MetA-SIMS mass spectra of a pristine PDMS film
The low mass range of the positive SIMS spectrum of a pristine PDMS film is shown in Fig. la. As expected, the characteristic peaks of siloxane polymers dominate the spectrum in this mass range (Si', m/z=28; SiC3H9', mlz=73; Si20CjH15t, m/z=147; Si3O3C5HI5',mlz=207; Si302C7H21',mlz=221, etc.) and the observed peak pattern mirrors the mass spectra published in the literature for PDMS [20, 211. After gold evaporation (Fig. lb), the fingerprint peak pattern of PDMS is almost unchanged, except for some intensity variation and the appearance of a strong Au' peak at mlz=197. In general, the intensities of the characteristic peaks are enhanced, especially for higher masses. For instance, the intensity of ions with a mass beyond 100 Da increases by an order of magnitude on average. Such large sensitivity enhancements have also been reported in previous articles [l 1, 221 for other types of organic samples and constitute the major interest of the method. The measured yields have been explained by the improved desorptionhonization induced by the presence of gold clusters in the top surface region of the films. The mass spectrum of Fig. l b does not show any sign of degradation induced by the gold metallization procedure. In contrast, the influence of the MetA-SIMS procedure on the high mass range of the positive mass spectrum is dramatic (Fig. 2). While there were virtually no high mass ions in the SIMS spectrum of the pristine film (Fig. 2a), the goldmetallized film exhibits a distribution of intense peaks between mlz=800 and m/z=3500, as seen in Fig. 2b. The peaks in the distribution are separated by
159
Improvement of metal adhesion to silicone film
0
I"
50
100
150
200
50
100
150
200
250
300
mlz
Figure 1. Fingerprint regions of the positive ion mass spectra of two PDMS films (logarithmic intensity scale). (a) Film 1. neat; (b) Film 2, after evaporation of 20 nmol/cm2 of gold (MetA-SIMS).
74 Da, which corresponds to the mass of the dimethylsiloxane repeat unit. The observed distribution can be safely attributed to an overlayer of PDMS oligomers at the surface of the film. The results obtained after hexane cleaning of the film (see Section 3.3) also confirm that the appearance of oligomers is not an artifact of the gold metallization procedure but rather a direct effect of the Aucationization (formation of a cation via recombination with a free Aut species) of the PDMS oligomers present at the sample surface. It should be noticed that an equivalent peak pattern is observed when short chains of PDMS are cast as a (sub)monolayer on a silver substrate [23]. 3.2. A r - ion -bornba rded P DMS $1 ins +
Positive mass spectra of Ar'-ion-bombarded PDMS regions covered with 20 nmol/cm2 of gold are presented in Figs 3 and 4. The high mass region shown in
160
A . Delcorte et al.
5
60:
10 v
S c
s-a
40 1
3
20:
+
O
0
8
“oool
m/z
-L 1000
2000
3000
4000
Figure 2. High-mass regions of the positive ion mass spectra of two PDMS films. (a) Film 1, neat; (b) Film 2 , after evaporation of 20 nmol/cm2 of gold. Note the distribution of Au-cationized PDMS oligomers.
Fig. 3a shows the disappearance of PDMS oligomers after the Ar’ ion bombardment procedure. It is not clear from Fig. 3a whether these oligomers have been fragmented or sputtered away during the high fluence bombardment or whether they might have cross-linked to form a more compact network, preventing single molecule emission. The mass spectrum of Fig. 3b, however, measured at the edge of the irradiated area and encompassing both bombarded and pristine regions, provides some more information. In addition to the regular distribution observed
Improvement of metal adhesion to silicone films
-
20
7
800
~
h
5
161
/z
600-
3
5 c 9
-2
400-
v)
4-
C
1
s3
g
I-
200
m/z+
O
1000
2000
3000
4000
Figure 3. High-mass ranges of the positive ion mass spectra of a PDMS film locally bombarded with 2 keV Art ions and metallized with 20 nmol/cm2 of gold. (a) Inside the Ar'-ion-irradiated area; (b) edge of the Ar+-ion-irradiated area. Distribution 1, initial oligomer distribution; distribution 2 , ion-beam-induced fragments.
on the pristine polymer, there is a more intense, low-mass distribution of PDMS chain segments in the range 800-1100 Da. This distribution of peaks is most probably the result of the fragmentation of larger oligomers induced by the partial Ar' ion bombardment in this region of the sample. The relatively high intensity of intermediate fragment peaks (between the major peaks of the distribution) also supports the idea of a quasi-random fragmentation of the polymer induced by physical sputtering. This result suggests that the picture of the events induced by the Ar' ion bombardment is probably more complex than just a straightforward cross-linking reaction [ 11.
162
106
A. Delcorte et a1
1
Si+
+ (n1~4
c
8103
([I
I-5102
101 100
50
0
100
200
150
250
Figure 4. Fingerprint region of the positive ion mass spectrum of a PDMS film bombarded with 2 keV Ar* ions and metallized with 20 nmol/cm2 of gold (inside the Art-ion-bombarded area).
The Ar' ion bombardment strongly affects the fragmentation region of the mass spectrum (Fig. 4). In comparison with the Au-metallized PDMS sample (Fig. lb), the irradiated sample displays a much simpler peak pattern. The characteristic peaks of PDMS have almost completely disappeared, indicating a profound transformation of the surface chemistry. The intensities of the SiC3H9' (mlz=73) and Si20C5H15' (m/z=147) peaks are more than three orders of magnitude lower. The major peaks remaining after irradiation are C', Si', Ga', Au' and a quite interesting series of ions beyond 200 Da. The most probable assessment for this group of peaks is: AuCHC (mlz=211) and AuCH3' (mlz=212), AuSi' (mlz=225), AuSiCH3' (mlz=240) and AuSi(CH3); (m/z=255). All these gold adducts were already present in the spectrum of the Au-metallized pristine sample. Their persistence in the spectrum of the bombarded sample shows that the irradiation step does not completely remove the specific molecular information of the PDMS repeat unit. 3.3. Comparison with hexane-cleaned surfaces A batch of pristine PDMS samples was annealed at 373 K for 1 h and then .-. . cieanea ror L m n in pure nexane. i n e amity or nexane to aissoive IOWmolecular-weight PDMS chains is indeed well known [21]. The positive mass spectrum of a Au-metallized, hexane-cleaned PDMS film exhibits the characteristic fragmentation pattern of PDMS (Fig. 5a, compare to Fig. lb), but shows no intact oligomers in the range 800<mlz<4000 (Fig. 5b). These spectra confirm two expectations. First, the high-mass distribution of peaks in Fig. 2b was indeed composed of oligomers, as opposed to long chain segments resulting from a bond breaking reaction, since they disappeared after a simple rinsing with the solvent. Second, the spectra indicate that hexane cleaning completely washes off the oli1 , -
A
.
.
m.
.
P
,
1
.
1
1
Improvement of metal adhesion to silicone filrns
50
W o
100
150
200
163
250
300
mlz
140,
F .- 120-
b)
Au,+
L3
3
E, 100m c?
80
-
u)
Au7+
c
C
2 60-
0
5 m 40 -
F
Au~+ I
20
n
rn/z-+ 1000
2000
3000
4000
Figure 5. Positive ion mass spectra of a PDMS film cleaned in hexane ( 2 min) and metallized with 20 nmol/cm2 of gold. (a) Fingerprint region. (b) high-mass region.
gomer overlayer without affecting the underlying film made of much longer PDMS chains. The high-mass spectrum also shows a distribution of gold ion clusters Au,: with 5
164
A. Delcorte et al.
4. DISCUSSION
Even though they do not provide a quantitative information, the simple tests performed with Scotch@and Tesa@tapes show that bombardment of the PDMS film with 2 keV Ar' ions induces a pronounced adhesion enhancement of the subsequently evaporated Ti and Cu layers (see Table 1). This observation is in line with the results reported in the literature [ 11. The same tests indicate that a simple hexane-cleaning provides a lower but still significant adhesion improvement. In previous articles [ 1, 241, the explanation for the adhesion improvement subsequent to ion bombardment or plasma treatment of PDMS surfaces was based on the elimination of the WBL through cross-linking. To our knowledge, no direct evidence of the process was shown. Using the new development in static SIMS analysis (MetA-SIMS), our results unambiguously prove a series of points: (i) there is an overlayer of oligomers on top of untreated commercial PDMS films; (ii) this layer disappears after 2 keV Ar' ion irradiation of the surface but also after a simple cleaning of the surface with hexane; (iii) unlike hexane cleaning, the Ar' ion bombardment strongly modifies the chemistry of the top surface layer (disappearance of all large characteristic fragments of PDMS, Fig. 4); and (iv) the bombardment-induced modification of the surface involves creation of shorter oligomer chains as an intermediate stage (Fig. 3b). From these observations, new hypotheses can be proposed. First, it is reasonable to think that the oligomer overlayer is partly or completely fragmented and sputtered away by the Ar' ion irradiation. This hypothesis is also supported by the common observation that PDMS oligomer contamination on other materials can usually be sputtered away by a very low primary ion fluence in SIMS analysis. Some of the oligomers might also cross-link, even though there is no direct proof of such a process. The drastic change in the fragmentation region of the mass spectrum after bombardment with Ar' ions, in particular, the disappearance of large characteristic fragments, could be an indication of cross-linking. Whatever its fate after Ar' ion irradiation, the oligomer overlayer is initially very thin and the cross-linking reaction should certainly affect the underlying high-mass chains, too. It should be noted, however, that the small Au-cationized fragments visible after pre-bombardment still contain CH3 residues, which were believed to be eliminated by the cross-linking reaction [ 11. In summary, our study demonstrates that the adhesion improvement observed after Ar' ion bombardment of PDMS is the result of a combination of factors involving both the removal of the so-called weak boundary layer made of oligomers and the severe chemical modification of the underlying, high-molecular-weight PDMS. In this respect, the observation that gold clusters appear in the positive mass spectrum of metallized, hexane-cleaned PDMS, but not in the mass spectrum of metallized, Ar'-ion-irradiated PDMS, suggests a stronger interaction between the metal overlayer and the irradiated PDMS. The chemistry modification induced by the Ar' ion bombardment probably creates a surface that is more reactive towards the incoming metal atoms. For instance, Si-OH, -C=O and -COOH
Improvement ojmetal adhesion to silicone films
165
groups might be introduced at the surface after reaction of the radicals formed under bombardment with the atmospheric O2 and H 2 0 molecules [2]. Such an increase of reactivity could explain the much higher adhesion of metals to the Art ion bombarded PDMS surfaces in comparison to hexane-cleaned films. 5. CONCLUSIONS
This work unambiguously shows that commercial PDMS films are covered with an overlayer of kDa-range oligomers which limits the adhesion of metals evaporated on their top surface. This layer disappears after 2 keV Ar’ ion irradiation, but also after a simple cleaning of the pristine films by hexane. However, the adhesion of titanium and copper layers on the surfaces is much higher when Ar’ ion bombardment pretreatment is used. Our results indicate that this improvement is related to the chemical modification of the top layers of the sample and their increased reactivity towards the evaporated metal atoms. From a methodological viewpoint, covering organic surfaces with minute amounts of gold induces a significant sensitivity enhancement for static SIMS analyses (MetA-SIMS). The positive fragment ion yields are, on average, enhanced by about one order of magnitude and the molecular ions become visible owing to an efficient ionization of the molecules via recombination with gold atoms. Such yield enhancements are promising for many problematic cases such as identification of kDa-range molecules on organic substrates and imaging SIMS applications where sensitivity usually constitutes a serious issue (e.g., polymer blends [25], polymers covered with additives [26], and cells and drughiornolecule arrays [27]). Acknowledgements
This work was supported by the Interuniversity Attraction Pole program on “Quantum sized effects in nanostructured materials” of the Belgian Federal State. The ToF-SIMS equipment was acquired with the support of the Rkgion Wallonne and FRFC-Loterie Nationale of Belgium. REFERENCES 1. P. Bodo and J. E. Sundgren, Thin Solid Films 136, 147 (1986). 2. C. Satriano, G. Marletta and E. Conte, Nucl. Instrum. Methods Phys. Res. B 148, 1079 (1999). 3. M. T. Khorasani, H. Mirzadeh and P. G. Sammes, Radiat. Phys. Chem. 47,881 (1996). 4. F. Abbasi, H. Mirzadeh and A. A. Katbab, Polym. Znt. 51, 882 (2002). 5. H. Hillborg and U. W. Gedde, Polymer 39, 1991 (1998). 6. H. Hillborg, J. F. Ankner, U. W. Gedde, G. D. Smith, H. K. Yasuda and K. Wikstrom, Polymer 41, 6851 (2000). 7. G. Bar, L. Delineau, A. Hafele and M.-H. Whangbo, Polymer 42, 3527 (2001). 8. B. Feddes, J. G. C. Wolke, A. M. Vredenberg and J. A. Jansen, Biomaterials 25, 633 (2004).
166
A. Delcorte et al.
9. H. Schmid, H. Wolf, R. Allenspach, H. Riel, S. Karg, B. Michel and E. Delamarche, Adv. Funct. Mater. 13, 145 (2003). 10. E. C. Rangel, W. C. A. Bento, M. E. Kayama, W. H. Schreiner and N. C. Cruz, Surf: Inte$ Anal. 35, 179 (2003). 11. A. Delcorte. N. MCdard and P. Bertrand, Anal. Chem. 74,4955 (2002). 12. D. Ruch, J.-F. Muller, H.-N. Migeon, C. Boes and R. Zimmer, 1.Mass Spectrom. 38, 50 (2003). 13. L. Adriaensen, F. Vangaever and R. Gijbels, paper presented at the XIVth International Conference on Secondary Ion Mass Spectrometry. San Diego, CA, September 2003. Appl. Su$ Sei. 231-232.256 (2004). 14. M. Inoue and A. Murase, paper presented at the XIVth International Conference on Secondary Ion Mass Spectrometry, San Diego, CA, September 2003. Appl. Su$ Sei. 231-232, 296 (2004). 15. H. Grade and R. G. Cooks, J. Am. Chem. Soc. 100,5615 (1978). 16. I. V. Bletsos, D. M. Hercules, D. van Leyen, B. Hagenhoff, E. Niehuis and A. Benninghoven, Anal. Chem. 63, 1953 (1991). 17. A. I. Gusev, B. K. Choi and D. M. Hercules, J. Mass Spectrom. 33, 480 (1998). 18. B. W. Schueler, Microsc. Microanal. Microsrruct. 3, 119 (1992). 19. A. Delcorte, X. Vanden Eynde, P. Bertrand and D. F. Reich, Int. J. Mass Spectrom. 189, 133 (1999). 20. J. C. Vickerman, D. Briggs and A. Henderson (Eds.), The Static SIMS Librarq, SurfaceSpectra, Manchester, UK (1997). 2 1. U. Oran, E. Unveren, T. Wirth and W. E. S. Unger, Appl. Su$ Sei. 227, 3 18 (2004). 22. A. Delcorte, J. Bour, F. Aubriet, J.-F. Muller and P. Bertrand, Anal. Chem. 75, 6875 (2003). 23. X. Dong, A. Gusev and D. M. Hercules, J. Am. Soc. Mass Spectrom. 9, 292 (1998). 24. H. Schonhom and R. H.Hansen, J. Appl. Polyn. Sei. 11, 1461 (1967). 25. B. Nysten, G. Verfaillie, E. Ferain, R. Legras, J.-B. Lhoest, C. Poleunis and P. Bertrand, Microsc. Microanal. Microstruct. 5, 373 (1994). 26. N. MCdard, A. Benninghoven. D. Rading. A. Licciardello, A. Auditore, Tran Minh Duc, H. Montigaud, F. Vernerey, C. Poleunis and P. Bertrand. Appl. Su$ Sei. 203-204, 571 (2003). 27. N. Winograd, Appl. Su$ Sei. 203-204, 13 (2003).
Adhesion Aspects of Thin Films, Vol. 2, pp. 167-176 Ed. K.L. Mittal 0VSP 2005
Mechanical stability of a Ti02 coating deposited on a polycarbonate substrate M. IGNAT,'3xS. GETIN,* B. LATELLA,3 C. BARBE3 and G. TRIAN13 'LTPCM, INP Grenoble, BP 75, Domaine Universitaive, 38402 Saint Martin d'Hdres, France 2CEA Grenoble, 17 rue des Martyrs, 38054 Grenoble, France 3ANST0 Materials Division, PMB 1, Menui, NSW 2234, Australia
Abstract-An analysis of the mechanical stability of a film-substrate system is presented. The film-substrate system consisted of a titanium oxide layer, deposited on a polycarbonate substrate. The analysis is based on in situ microtensile tests, which allow to follow the development of irreversible damage (cracking and de-adhesion) to the film when progressively pulled in tension. From this in situ degradation, the critical parameters corresponding to the initiation of irreversible damage mechanism are determined. Further, the observations and experimental determinations can be compared to analytical models, which describe how the stresses in a cracked film will be redistributed when pulling the substrate. If de-adhesion of cracked portions of the film is observed during an experiment, then, by considering the critical conditions for the first detachment and then buckling of the film, one can deduce intrinsic parameters related to the interface as, for example, its fracture energy and toughness.
Keywords: Cracking; de-adhesion: thin films; polycarbonate; titania: toughness: interfacial energy.
1. INTRODUCTION
The demands imposed on mechanical durability of film-substrate systems in many leading technologies (particularly microelectronics, photonics and biomaterials) are becoming more stringent and, thus, associated problems have to be understood and solved. Film-substrate systems are subjected to internal stresses, caused by thermoelastic mismatch, or to external mechanical stresses applied monotonically or cyclically. When reaching critical levels, these stresses may activate damage mechanisms such as cracking and de-adhesion of the film. Identifying these failures and understanding the critical conditions which cause them is essential, prior to any technological application of the system. The purpose of this work was to investigate the fundamental mechanical stability and adhesion of titania coatings used in optical devices. Titania was deposited "To whom correspondence should be addressed. Tel.: (33-4) 7682-6606; Fax: (33-4) 7682-6745; e-mail:
[email protected]
168
M. Ignat et al.
on a polymeric substrate by atomic layer deposition (ALD) to fabricate the model system. The cracking and debonding behaviour of the coating under external stresses was examined using in situ optical microscopy. Thus, the mechanical behaviour of the system was characterised from the results obtained with progressive application of an external monotonic tensile loading to the titania on polycarbonate system. This behaviour can be related to three main processes observed during loading: elongation of the system without any observed cracking, cracking of the titania layer and debonding of portions of the cracked film at a certain degree of crack density in the titania layer (see Fig. 1). For each of these processes, the associated recorded experimental parameters (applied force and corresponding displacement) allow calculation of corresponding strains and stresses, which cause the mentioned damage mechanisms, cracking of the film and de-adhesion, if observed.
Figure 1. Sequential micrographs taken during in situ tensile test. (a) The evolution of cracking is shown in successive micrographs; and (b) the de-adhesion and buckling in successive micrographs. The 100 Krn scale is indicated, the white arrows show defects taken as references to localize the zone.
Stability of a TiOz coating deposited on polycarbonate
169
In the following, first we shall describe briefly the experimental method used to characterise the system mentioned in situ. Then our experimental results will be discussed with regard to parameters related to the mechanical stability of the system, namely, the critical rupture stress and the fracture toughness of the film, and the interfacial energy associated with the observed de-adhesion. 2. EXPERIMENTAL
2.1. Sample preparation and micro-tensile device The film-substrate system consisted of a titania thin film deposited on a polycarbonate substrate. The titania films were deposited on the polycarbonate substrate by atomic layer deposition (ALD). This was performed using an AL-CVD reactor, with nitrogen as the carrier gas. The films were deposited at 373 K (lOO'C), and the reactants used were Tic& and HzO. The ALD deposition was performed up to 2000 pulses, each one increasing the film thickness by about one atomic layer. This procedure produced a 150-nm-thick film. To evaluate the elastic modulus of these ALD titania films, nanoindentation tests were performed, and the results are reported in Table 1. The nanoindentation procedure used for elastic modulus determinations corresponded to loading/unloading the film with a Berkovitch-type indenter [ 1-31. A polycarbonate substrate was chosen because of its good thermal stability (up to 473 K, 200°C) and ductility. The thickness of the substrates was 1 mm, and they were machined to obtain dogbone type, micro-tensile samples: their total length was 30 mm with a gauge length of 20 mm and width of 3 mm. For the materials used in this study, their mechanical properties as determined and then used in our calculations are reported in Table 1. The ALD titania films were directly deposited on dogbone type samples. The in situ tests were perTable 1. Mechanical and thermal parameters of TiOz on polycarbonate system Young's modulus (GW
Poisson's coefficient
Titanium dioxide
165" (thin film) 0.27 214 to 282
Polycarbonate (1 mm)
2.5 to 3.1
0.42
Flowhpture stress (MP4
Thermal expansion coefficient ( K-')
350* (tension) 10 103 (tension) 3500 (compression) 65 (tension) 67 110 (compression)
Residual thermal stress (MPa, calculated) 940*
14.3*
Note that the 350 MPa value for the rupture stress of the TiOz thin film in tension corresponds to the value deduced from the experiments. The values indicated by * represent our experimental determinations (nanoindentation and in situ tensile tests) and calculations (residual thermal stresses). All other values are taken from the Data Bank of Materials [ 1I].
170
M. Ignat et al.
formed with a micro-tensile device, which has been described elsewhere [4, 51. Such a device can be mounted directly on the stage of an optical microscope, as well as in the sample holder of a Scanning Electron Microscope. In both cases the in-plane rotation of the system holding the samples can be used to improve the observations. Besides, the micro-device allows a symmetrical movement in both extension (tension), as well as in contraction (compression). Thus, when pulling a sample symmetrically the operator can localize a zone and follow precisely the deformation without any drift of the image. In our case of straining a brittle film on a ductile substrate, the absence of any drift of the image allows to identify with precision the experimental parameters associated with the properties of the film, for example, its rupture (strength and toughness). Moreover, if debonding is observed, parameters related to the interface, such as the strain energy release rate, can also be determined by applying analytical models [6]. 3. RESULTS
3.1. In situ straining The in situ observations performed with an optical microscope when pulling our film-substrate systems in tension showed progressively the three stages mentioned previously (see Fig. 1): - During the loading of the substrate at a given strain, largely spaced cracks transversal to the tensile direction appear in the titania thin film. At this stage, on unloading the sample the elastic straining recovers and the cracks which appeared transversal to the tensile axis will close. Thus, on unloading the sample, no apparent damage is observed on the surface of the sample. However, it is important to point out, as the observation is continuous, that first a given force will be associated with the observed onset of cracks, and second, the critical rupture stress of the film can be deduced. - The second stage shows a progressive increase in the number of cracks developed at induced strains up to the linear elastic response of the system. Thus, the observed increase in the regularly distributed transversal cracks with strain reaches a characteristic inter-crack distance, which remains constant even though the elongation of the sample increases. This inter-crack distance is used in our calculations. - The third stage is related to de-adhesion of the film. As a matter of fact, careful observations show that de-adhesion seems to be triggered first at imperfections in the film. Then, in-between two consecutive cracks the film can buckle, with a well-defined dimension. When the last stage occurs, transversal cracking has reached saturation. This means that no new cracks will be generated and, as pointed out previously, a characteristic inter-crack distance will be observed.
Stability o f a TiOz coating deposited on polycarbonate
171
3.2. The critical parameters
From the in situ experiments and the reported forces and displacements associated with the observed cracking and de-adhesion, the rupture properties (fracture toughness, rupture energy) of the titania brittle film can be determined first. The starting point of the calculations involves the deduced stress associated with crack initiation and the elastic properties of the film and the substrate into the “shearlag” model. It should be noted here that the model mentioned was first developed to understand how the applied load was transferred between a matrix and its reinforcing phase in fiber-reinforced composite materials, at rupture of the fibers [7]. The shear-lag model assumes that when a high-modulus material breaks (the reinforcing phase in a composite, or the thin film attached to the substrate), adjacent to the cracks at the interface, the induced shear stresses cause slippage between the film and the substrate, or local plasticity in the substrate. Furthermore, by considering the experimental parameters associated with the cracked film and its corresponding critical cracking stress, intrinsic cracking parameters of the film can be determined. For example, for a simple system consisting of a brittle film on a ductile substrate, the fracture toughness of the film KIc is given by [8]:
In this relation, 0,is the critical stress deduced by observing the first crack, qSs is the yield shear stress of the substrate, F is a function which depends on the ratib of the film’s Young modulus Efto the substrate’s modulus E, and hf is the thickness of the thin film. By experimentally determinig the critical cracking stress of the film a, and knowing the modulus and other parameters in relation (l), the fracture toughness of the film can be calculated, Besides, by assuming that a uniaxial tensile stress state predominates in the film and from the deduced fracture toughness, the strain energy release rate of the film can be obtained by simply applying Griffith’s formalism. Indeed, the strain energy release rate GIc can be calculated from the ratio between the fracture toughness KIc and the Young modulus of the film Ef[9]:
By using the values of the parameters reported in Table 1, with relations (1) and (2), intrinsic rupture properties related to the film can be obtained, and then discussed with respect to bulk properties of the same type of material. Thus, the following analysis corresponds to the process of de-adhesion of the titania film. For this process the intrinsic parameters related to the interfacial behaviour, in particular, the energy to extend an interfacial crack can be deduced by considering the experimental observations:
M. Ignat et al.
172
In our system, the de-adhesion appears progressively. As a matter of fact, when the polycarbonate substrate reaches a critical level of strain, then we observe welldefined buckled areas in between two successive transversal cracks. The number of these characteristic buckled zones increases with increasing strain (Fig. 1). We must recall here that the buckling process depends principally on two factors: the critical dimensions of the detached area at the interface and the critical level of the compressive stress, produced by the Poisson’s contraction of the substrate which produces buckling. However, it must be considered that before the buckling, the strain which corresponds to film’s de-adhesion is difficult to deduce from the direct observations during the in situ experiment. As a matter of fact, only once buckling has started, de-adhesion of the film becomes detectable (Fig. 1). Consequently, the strain which corresponds to film de-adhesion before buckling remains a crucial parameter to be determined, because it is directly associated with the energy needed to extend an interfacial crack to produce de-adhesion. We deduce this parameter indirectly from considering the difference between the strain for initiation of regular buckling (experimental observation) and the theoretical strain relation corresponding to the condition to buckle a well-defined area of a debonded film. This critical strain &hb is given by [ 101: EP, =
Bnh: 2(1-v3 ’
(3)
where B is a constant which depends on the dimension of the buckled zone, vfis the Poisson coefficient of the film and hf is its thickness. Finally, the apparent interfacial cracking energy xntcan be calculated by considering the balance between the strain level prior to buckling E (deduced experimentally from the first observed buckled zone), and the theoretical strain level for buckling given by relation (3), which leads to [6]: y
=-hf int
Ef
2 1-v:
(4)
All the terms in this relation have been previously defined. 4. DISCUSSION
The following two main aspects should be considered: the critical values deduced from the above described relations and - the observed crack saturation with respect to the interfacial adhesion. The substrate and the film parameters determined and used in our calculations are reported in Table 1. In Table 2 we report exclusively the fracture toughness as well as the strain energy release rate (for the titania films and for the titanidpolycarbonate interface), also deduced from calculations. It should be men-
Stability of a TiOz coating deposited on polycarbonate
173
Table 2. Values deduced experimentally and typical values reported in the literature [l 11 for the strain energy release rate GIc( xc)and fracture toughness KIc for the thin film and for the interface
TiOz Interface
GIc (experimental, J . m-*)
Klc (experimental, MPa m0 5 ,
Glc (literature, J . m-*)
Klc (literature, MPa , m0 5 ,
1.5 5.9
0.5 0.5
22 Not found
2.5 8
tioned here that the starting point for the film parameters determination was the Young’s modulus from nanoindentation. This value is different from those which can be obtained from the literature [ l 11 corresponding to the bulk form: the deduced Young’s modulus for our films was lower (150 (GPa) < 250 (GPa)). At this stage, we may speculate that the generally observed microstructural differences between thin films and bulk materials often are related to the degree of crystallinity of the film. But this point needs verification by XR diffraction or Raman spectroscopy. A low degree of crystallinity, or a certain degree of amorphization in a thin film, often decreases the elastic modulus [ 121, As for the higher rupture energy deduced directly from the experiments, compared to that in bulk form, this can be explained by the combined effect arising from the large expansion coefficient, as well as elastic modulus mismatch between the substrate (soft and ductile) and the thin film (hard and brittle). Indeed, the combined effect of thermal expansion and modulus generates rather high thermoelastic compressive residual stresses, which will increase the level of tensile flow/rupture, about three times higher than the one suggested for the same material in the literature [ 111. With respect to the interface values, our determination of the energy associated with de-adhesion of the film is closer to values for oxide/metal interfaces, corresponding to adhesion energies deduced from measurements of wettability angles [13] and to values of oxides (SOz) on ductile metal substrates (Cu, Al) determined from the same type of experiments as reported here [ 14, 151. This may indicate that a certain degree of chemical continuity due to oxygen bonds subsists across the titania-polycarbonate interface. However, compared to the values of interfacial strain energy release rates in the literature, determined from other micromechanical type experiments (microhardness, bending, etc.), our energy value for cracking is lower. As a matter of fact, considering the values of interfacial energies reported in the literature (see, for instance, Refs [16, 171, reporting values higher than 100 J/m2), we can expect that the extended elastic response of the polycarbonate substrate excludes any effect caused by plastic deformation, increasing the level of strains used in calculations, and, thus, the deduced energies. Besides, if the system behaves elastically during the evolution of cracking and de-adhesion, we can determine the distribution of stresses in the cracked film at different levels of increasing strains. So by comparing with the experimental observations, this calculation will allow to discuss the degree of attachment of the film to the substrate from the inter-crack distance evolution.
M. Ignat et al.
174
-20
-1
0 Inter-crack distance [prn] (a1
0
+20
+I
Inter-crack distance [p m] IC1
Figure 2. Calculated stress distributions in defined cracked portions of the film. (a) up to 40 ,urn between two consecutive transversal cracks, (b) 6 pm, between two consecutive transversal cracks and (c) 2 pm, between two consecutive transversal cracks. For each diagram, the dotted line indicates the level of stress corresponding to the rupture stress of the film.
Stability of a TiOz coating deposited on polycai-boizate
175
The relation which describes the stress o:y (x) distributed along the in-plane longitudinal direction (x) in the vicinity of a single crack has already been established [ 181. Thus, when considering a portion of the film, limited by two consecutive cracks, the stress distribution becomes:
In this relation o: is the stress applied to the substrate, of the residual stress in the film, K the ratio of the Young’s modulus of the coating and substrate and L the inter-crack distance. The term u corresponds to the change in the variable for the calculation, step by step along an inter-crack distance (L). In Fig. 2 we report the results of calculations from relation ( 5 ) for three levels of stresses in the film corresponding to increasing strain levels. In each case, the critical crackmg stress is indicated (350 MPa). By taking into account the critical stress value and the stress-free crack surfaces, we note that theoretically for the titania on polycarbonate system, the limit distance between two consecutive cracks to maintain the cracking stress level of 350 MPa in the film is about 4 pm. As a matter of fact, this distance is less than the mean distance deduced from experimental observations 8 x m when reaching crack saturation. This discrepancy between the calculated and observed inter-crack distances indicates that a perfect transfer of displacement does not take place. Thus, we may speculate that the shear stresses at the interface may extend de-adhesion. This distance corresponds to the difference between calculated and observed inter-crack lengths at saturation. Similar results have been obtained in other hard brittle films deposited either by CVD, or by PVD on ductile substrates [19, 201. 5. CONCLUSIONS
In situ tensile experiments were performed on ALD Ti02 films deposited on a polycarbonate substrate. The following conclusions can be drawn from this study. The Young’s modulus of the TiOz coating determined by nanoindentation was lower than the values reported in the literature, which correspond to bulk material. The deduced critical cracking stress for the film allowed to determine film’s rupture properties, and parameters related to the interface. As for the rupturehlow stress in the film (higher than those reported in the literature for bulk material) it appears to be strongly dependent on the thermoelastic mismatch between the film and the substrate giving rise to high residual compressive stresses in the film. When calculating stress distribution from the observed inter-crack distances, a discrepancy exists between the calculated (perfect adhesion hypothesis) and
M. Ignat et a1
176
observed values. This discrepancy may point that interfacial shear extends debonding at the interface.
REFERENCES 1. E. R. Weppelmann, J. S. Field and M. V. Swain, J. Muter. Res. 8, 830 (1993). 2. M. V. Swain and J. Mencik, Thin Solid Films 253, 204 (1994). 3. G. M. Pharr and W. C. Oliver, MRS Bull. 17, 28 (1992). 4. M. Ignat, L. Debove and C. Josserond, Bull. Soc. Fr. Microsc. Electronique 14, 10 (1995). 5 . M. Ignat, Key Eng. Mater. 116-117,279 (1996). 6. M. Ignat, in: Surfuce Engineering Series, Vol. 2, p. 45, ASM International, Materials Park, OH (2001). 7. J. Aveston, G. A. Cooper and A. Kelly, The Properties of Composites, IPC Sci. Technol. Press, London (1971). 8. M. S. Hu and A. G. Evans, Acta Metall. 39, 1061 (1986). 9. A. A. Griffith, Phil. Trans. Roy. Soc. (London) A22, 163 (1920). 10. S. P. Timoshenko and J. M. Gere, Theory of Elastic Stability, 2"d edition, McGraw Hill, New York, NY (1961). 11. Cambridge Engineering Selector, Granta Design Limited, Version 3.1 (2000). 12. J. Mencik and M. V. Swain, Mater. Forum IMMAustrulasia 18, 277 (1994). 13. E. D. Hondros, Inst. Phys. Con5 Ser. 75, 121 (1986). 14. P. Scafidi and M. Ignat, Muter. Res. SOC. Symp. Proc. 309, 55 (1993). 15. M. Ignat, P. Scafidi, E. Duloisy and J. Dijon, Mater. Res. Soc. Symp. Proc. 338, 135 (1994). 16. A. G. Evans, B. J. Dalgleish, M. He and J. W. Hutchinson, Acta Metall. 37, 3249 (1989). 17. I. E. Reimanis, B. J. Dalgleish, M. Brahy, M. Ruhle and A. G. Evans, Acta Metall. 38, 2645 (1990). 18. S. M. Hu,J. Appl. Phys. 50,4611 (1979). 19. E. Harry, A. Rouzaud, M. Ignat and P. Juliet, Thin Solid Films 332, 195 (1998). 20. M. Ignat, T. Marieb, H. Fujimoto and P. A. Flinn, Thin Solid Films 353, 201 (1999).
Part 3 Adhesion Measurement
Adhesion Aspecrs of Thin Films. Vol. 2, pp. 170-193 Ed. K.L. Mittal 0VSP 2005
Advances in adhesion measurement good practice: Use of a certified reference material for evaluating the performance of scratch test instrumentation NIGEL M. JENNETT,’.*RIA JACOBS2 and JAN MENEVE2 ’National Physical Laboratory, Materials Peiformance, Hampton Rd, Teddington, Middlesex T W l l OLW, UK 2Mnterinls Technolog) Centre, Flemish Institute f o r Technological Research - VITO, Boeretnng 200, B-2400 Mol, Belgium
Abstract-The scratch test is widely used in industry but has, in the past, been widely distrusted because of apparent variability in results. This situation has been significantly improved as a result of the output of two European Commission (Standards, Measurement and Testing) funded projects, “FASTE” and “REMAST”. As a result. a European pre-standard. pr-ENV1071-3, now exists and a Certified Reference Material (CRM), BCR692, is available. This paper describes the results of the REMAST certification exercise and the improvements to scratch test good practice the CRM makes possible. In particular, the diagnostic power of the CRM is demonstrated by identifying previously unknown inter-laboratory variations in test results. Strategies are also described for using a CRM to monitor instrument performance and aid in the comparison of results between laboratories. for example in acceptance testing.
1. INTRODUCTION
The scratch test is routinely used for testing the strength of adhesion of a coating to its substrate. The most usual test (at least for thin hard coatings) consists of generating scratches by progressively loading a diamond stylus that is drawn across the surface of the coating-substrate system being tested [ 1-31. A progressive series of coating failures may be observed consecutively in the scratch track, at increasing critical load (L,) values. Although different coating-substrate systems fail in different ways [4], these failures are reproducible. As a consequence, the European Standards Committee CEN TC184 WG5 has established a European Standard prEN 1071-3 for the scratch test [ 5 ] ,which defines standard calibration, cleaning and test procedures.
%Towhom correspondence should be addressed. Tel.: (44-20) 8977-3222; Fax: (44-20) 8614-045 1; e-mail:
[email protected]
N.M. Jeiznett et al.
180
1.1. Uncertainties in the scratch test The development of the scratch test standard benefited from a research project, ‘FASTE’, funded by the European Commission [6]. ‘FASTE’ investigated the sources of uncertainty in the scratch test method. This included the effects of stylus shape, stylus and sample cleaning, calibration of the scratch test instruments, the test protocol used, and the reproducibility of operator judgement (in which is included the visibility of failure events). It was found that uncertainty in the Rockwell C stylus tip shape, either due to damage or to incorrect radius, was the dominant source of error in the scratch test method [7]. Low radii styli and rough or damaged styli, with sharp edges, initiated failure events at lower critical loads than large radii or worn indenters. Typical findings are shown in Fig. 1. It can be seen that the stylus effect is very significant when compared to the 95% confidence limits of the test result represented by the error bars on each column. In response to this finding, a project, ‘REMAST’, was funded by the European Commission to develop and certify a reference material that would act as a quality control tool for the testing and qualification of scratch test instruments [8]. This is the preferred solution for scratch test users, as validation of their instrument/stylus performance against a certified reference material only requires them to have and to use equipment that they already have. Direct measurement of stylus radius, for example, would require additional capital equipment and training expenditure, or an unacceptable delay while styli were sent to a third party for verification.
60
n
z
v
40
U
m
0 -
m
.-0
,E
20
0
0 A5
B5
E5
Stylus label Figure 1. Results from EC project FASTE showing the effect of stylus damage and increased radius on a scratch test critical load. Error bars represent the 95% confidence interval for the measurement.
Advances in adhesion measurement good practice
181
I .2. A cert$ed reference material The REMAST project consisted of two parts: a feasibility study to select a candidate coating and a production and certification exercise that generated over 1000 pieces of the Certified Reference Material (CRM reference number BCR-692) [9]. This project also devoted a considerable effort to improving the scratch stylus manufacturing process, because it was found that the commercially available Rockwell C scratch test styli rarely met the standard EN IS0 6508-2 [lo], which requires a tip radius of 200 k 10 pm in two orthogonal directions. Even during certification, the diagnostic power of the CRM was demonstrated by its identification of a hitherto unknown set of systematic offsets in the scratch test results obtained from different but traceably calibrated laboratories. As a result, an ‘NPL Measurement Good Practice guide’ [l 11 was developed to enable users to realize the full potential of the CRM and to assist them in distinguishing and understanding the different uncertainties that contribute to a scratch test result. This paper briefly describes the procedure that resulted in the certification of BCR692 and outlines the strategies promoted in the good practice guide. 2. EXPERIMENTAL
2. I . Candidate certified reference inaterials The REMAST project evaluated two candidate coatings, titanium nitride (TIN) and diamond-like carbon (DLC). DLC was selected as the reference material to be certified because, compared to the TIN coating, it showed a higher sensitivity to stylus tip shape variation, a lower data scatter and was found to cause less wear in the diamond styli. Three repeatable failure events were identified in the DLC coating (see Fig. 2) at critical load intervals in the range 5 N < L, < 45 N. These failures are defined as L,,, forward chevron cracks at the borders of the scratch track; Lc2,cracking as in LC1but accompanied by interfacial spallation; Lc3,gross interfacial shell-shaped spallation that extends across the whole scratch track. 2.2. Calibration of scratch test instruments
Four types of instruments were used in the certification of BCR-692. The instruments involved were: VTT Scratch tester, CSEM (now known as CSM Instruments) Revetest, CSM Instruments multipass Revetest scratch tester, and the Teer Coatings scratch tester. Each instrument was traceably calibrated according to procedures consistent with prEN 1071-3 for the following parameters: Sample planarity Applied load and loading rate The zero offset of the applied load Horizontal displacement and dj splacement rate The zero offset of the displacement
182
N.M. Jennett et al.
Figure 2. Failure events associated with L,,, Lc2and Lc3.These failures are defined as: LC1,forward chevron cracks at the borders of the scratch track; Lc2.cracking as in LClbut accompanied by interfacial spallation; and Lc3, gross interfacial shell-shaped spallation that extends across the whole scratch track.
In some cases additional optional calibrations were performed, including spring stiffness and horizontal stage compliance. Details of the certification design and homogeneity study are given in Appendix A. To summarise: Each laboratory made 40 scratches on each of two randomly selected samples using a loading rate of 100 N min-', a displacement rate of 10 mm min-', a starting load of 5 N and a finishing load of 45 N. Values for each of the three critical loads were measured and reported. This formed the core of the certification data set. The certification design also included a number of other quality control features, which allowed the individual offset biases expected from the variability of stylus radius and sample homogeneity to be separated. 3. RESULTS AND DISCUSSION
A part of the certification data for 15,' is plotted in Fig. 3. Figure 3 shows the certification data, including the control data from the pilot laboratory, obtained from 14 samples. As expected, it is immediately apparent from the pilot laboratory data that using a different stylus on the same sample of CRM changes the LC1value. However, a more surprising result is that, even after traceable calibration of all instruments in accordance with prEN 1071-3, there remain systematic differences in the results from different laboratories. This can be seen particularly in the case of sample 10 where the pilot laboratory records essentially the same result using both the laboratory stylus and the control stylus, but this value differs signifi-
Advances in adhesion measurement good practice
183
Lc, measurement data 0 I
E 16 U
3 14
t
12
oo
0°"
I "
10' I
0
5
Stylus offset
Data Set
I
10
I
15
Figure 3. L,, certification and pilot data from seven laboratories. Data are in groups of three sets of circles. Circle colour/position represents respectively: black(1eft) - certification data; Red(midd1e) pilot data ( 5 scratches) with lab stylus; Blue(right) - pilot data ( 5 scratches) with control stylus. Examples of stylus offset and laboratory offset are marked.
cantly from the result reported by the certifying laboratory using the same stylus and sample. Thus, the design of the certification campaign has ensured that this difference can be unambiguously attributed to a previously undiscovered laboratory offset. 3.1. Statistical regression analysis
The availability of the pilot laboratory data allowed further statistical analysis of the REMAST certification data to separate the independent systematic offsets due to stylus, specimen and to quantify the previously unknown laboratory offsets. The complete set of certification data and control data LclJkwas modelled (for each critical load) in terms of a reference value p , plus offsets from that value due to laboratory, stylus and sample (Ll,RJ and Sk, respectively) and a random measurement error eyk.Figure 4 is a schematic representation of the effect of the different offsets that act to bias the L, value of a scratch. The key concept is that, although each effect is represented by a distribution of possible biases, when a particular sample, stylus or laboratory is selected, a single value is selected from each of the individual distributions and the combination is a single offset. Mathematically, the model can be expressed as follows:
N. M. Jennett et al.
184
Lcijk= p + Li+ R j + S,
+ eVk.
If it is assumed that the expectation value (in this case the mean) of each distribution of offsets is zero, i.e.: E ( L l )= E ( R J )= E ( S , ) = E(eY,)= 0, var(elJ,)= c?,
(2)
and it is also assumed that the individual e,,k values are independent, with a variance 2 that is an unknown constant, then it is possible to determine estimates for the reference value, p , and the offsets, L,. In effect, each scratch test result provides a particular solution to the general model. In this case, it was important to have included the results of the pilot laboratory control scratches in the analysis, as this meant that each parameter had been independently varied. As a result, convergence on a single least squares fit to all the data was possible. Figure 5 shows a typical plot of the residuals to the fit and demonstrates that the solution obtained effectively explained the variations observed in the raw certification data. The mean of the regression solution agreed closely with the Certified Value (CV) obtained by taking the mean of the certification data. In addition, the combined sample variability, Sk,was consistent with the separate Analysis of Variance (ANOVA) homogeneity study. Therefore, a hybrid approach was taken to calculate a ‘verification range’ within which 95% of all scratches would fall if made by any instrument type using any qualifying stylus on any of the CRM pieces. The CV and the uncertainty of the material (UNC) were taken as the starting point. Thus, the homogeneity study ANOVA results were used to estimate the sampleto-sample and batch-to-batch variations; the stylus offset distribution was obVerification 95% limit 11111-1-11
Offset due to
7-+=? f Stylus offset
due to Sample + Lab. + Stylus choice I I
VR-
I III III III II III III II I
Verification 95% limit
Figure 4. Schematic representation of the combination of independent offsets to cause a repeatable overall offset to the critical load value measured using a particular combination of instrument, stylus and sample of CRM.
Advances in adhesion measurement good practice
185
tained from the radius sensitivity study undertaken by the pilot laboratory; and the laboratory offsets distribution was estimated by the statistical regression analysis described above. The verification range (VR) is a much larger range than the certified range (CV f UNC), but is the range against which an indirect validation of a scratch test instrument would be compared. This range is important to the user. At each extreme of the verification range, there is only a 2.5% chance that a result outside this range could be from a properly calibrated instrument with a valid stylus. The verification range is “two-tailed” and so, when using the 95% coverage factor, 2.5% of valid results will fall below the lower limit and, likewise, 2.5% of valid results will fall above the upper limit of the range. The two “tails” of 2.5% make up the 5 % of possible, valid results rejected by setting a 95% confidence interval. The actual value for the verification range depends, to some degree, on the number of scratches measured in an indirect validation exercise. It was decided that a pragmatic indirect verification requirement would be for a test laboratory to perform 5 scratches on a CRM sample and take the mean of the critical load values obtained. Therefore, a verification range (CV & VR(5)) was defined, being a combination of material effects, laboratory effects (u,) and stylus effects (u,,):
LCI residual data 0
0
0
5
Data set number 10
15
20
25
30
35
40
45
Figure 5. Residuals to the fit of the statistical regression model to the LC1certification and pilot data. The residuals are distributed about zero demonstrating that the stylus and laboratory offsets have been successfully modelled. Circle colour/position represents respectively: Black(1eft 14 groups) certification data; Red(midd1e 14 groups) - pilot data ( 5 scratches) with lab stylus; Blue(right 14 groups) - pilot data ( 5 scratches) with control stylus.
N. M. Jennett et al.
186
Table 1. Certified values (CV), uncertainties (UNC) and uncertainty components quoted in units of (N)
L, I Lc2 Lc3
Ubb
ubs
&s
sc
P
4 s
4
CV
UNC
VR(5)
0.40 0.44 0.8
0.49 0.49
0.55 0.53 1.30
1.72 2.31 1.78
18 18 14
1.05 2.26 1.34
0.8 1.11 2.11
13.6 17.0
k 1.9
27.9
f 3.6
f3.1 r 5.4 25.7
0.81
r2.0
is the standard deviation of the means of the p accepted data sets (1 data set consisting of the results from 40 scratches on a single certification sample), Ubb, ubs and uwsare uncertainty components (as standard errors) related to heterogeneity (between-batches Ubb, between-samples & and within-samples uWJ. us, is the uncertainty due to the range of allowable stylus radii and uI is the uncertainty related to the distribution of laboratory offsets. VR(5) is the verification range for each L,, within which the mean of 5 scratches will fall for 9 5 8 of calibrated machines using qualifying styli. s,
VR(5) = +- k
(S2,/p
+ U2bb + u2bs + u2w,/5+ u2,,+ u * ~ ) ” ~
(3)
where s, is the standard deviation of the means of the p accepted data sets (1 data set consisting of the results from 40 scratches made on a single certification sample), ubb, u b s and u,, are uncertainty components (as standard errors) related to heterogeneity (between-batches Ubb, between-samples ubs and within-samples uws) and k is a coverage factor to expand the range to include 95% of possible results. A full list of u values for each L, value is given in Table 1. The largest component of uncertainty for LC1and Lc2is the stylus-related term uSs.This confirms the high sensitivity of the scratch test to stylus effects. However, machine/laboratory effects, ul, also play a major role. It is interesting to note that the effect of stylus radius is less for Lc3. The certified reference material [9] is now available to support the correct evaluation of scratch test measurements. 3.2. Advances in measurement good practice A careful design enabled the certification exercise to separate the different effects that will cause an offset to the L, value expected from a laboratory test on a piece of BCR-692. Thus, proper experimental design can provide a route to significant improvements in instrument validation and scratch test measurement good practice. Simple strategies can be adopted by scratch test users to allow a more informed comparison of results from different machines. Diagnosis of the cause of a particular indirect validation failure is also possible. Each piece of BCR692 is supplied with a Good Practice Guide [l 11, which describes how to exploit the new diagnostic capabilities that the CRM makes available. The strategies described are summarised below. The most important use of the CRM is in the indirect validation of scratch test instruments. It is assumed that an indirect validation will consist of making a set of five scratches, measuring the L, values and calculating the average and standard deviation at each certified failure event. The standard deviation of each set of
Advances in adhesion measurement good practice
187
Table 2. Comparison table for a standard deviation of five L, values for each critical load Lc
L,, L,? LC7
(N)
Average standard deviation in certification (40 scratches)
Maximum expected standard deviation for 5 scratches
Minimum expected standard deviation for 5 scratches
0.55 0.53 1.30
0.8 1.2 2.2
2.1 3.2 6.9
0.2 0.3 0.5
urn5
The maximum and minimum values were estimated from the average standard deviation in the certification exercise by applying the F-statistic to take account of the reduction in sample size from 40 to 5 scratches. The certification within-sample uncertainty is given as a further comparison.
five L, values is a combination of real within-sample effects and the natural scatter of your instrument (repeatability). These two components cannot be separated. The standard deviation of any set of five scratches is very variable but there is a statistical range within which the results from a good instrument are likely to fall. The values in Table 2 represent the 95% confidence limits for the standard deviation of five scratches estimated using statistical tables for the F-statistic (from the x2 distribution) and the average standard deviation of the certification data. In 95% of instances, a correctly set up instrument should produce a standard deviation (of 5 scratches) that lies between the minimum and maximum values and probably near to the average value shown. A mean critical load value within the verification range (CV k VR(5)) implies that the instrument is working acceptably. If the mean L, value falls outside (CV k VR(5)), then this is not the case. The instrument has failed the verification test and the user should check first the stylus and, if necessary, the instrument calibration. Monitoring of instrument performance over time is also possible. This is done by plotting a control chart, using the first set of scratches on a piece of the CRM as a reference and comparing successive sets of results obtained by the same instrument and stylus on that particular CRM sample. If the verification range is marked on the chart as an outer limit (see Fig. 6), verification failure can be spotted instantly. However, the control chart can be used to provide much better quality control than is possible by comparing verification results solely with the verification range. Where there is no change in the CRM sample, stylus or instrument, future verification values should be distributed close to the first value and varying only by the normal amount of within-sample variation in combination with the normal instrument measurement variability. For example, L,, might be expected to vary by f 1.1 N (= 2 u , , , which is the within-sample uncertainty expanded to a 95% confidence interval by using a coverage factor of 2) about an average value that remains constant. If only the stylus is changing, then subsequent verification results will systematically drift. A drift of more than & 2.1 N (= 2u,J from the initial verification would indicate that the stylus has changed to be outside an acceptable calibration. Similarly, if only the instrument performance is changing, verification results
188
N.M. Jennett et al.
Lc [NI
15.6 14
12.4 Calibration or Stylus failure original ___________-________--------------------------------verification 10.5 N verification min.
Figure 6. Control chart for monitoring instrument performance over time. The control limits can be set by the user to achieve a level of quality control that is fit for purpose.
that have systematically drifted by more than t 1.6 N (= 2 4 ) from the original verification are also not allowed. Future standards should set guidelines for mandatory control limits, but in the meantime, the control limits can be set according to the level of reproducibility required by the user. Clearly, the best improvement in measurement good practice would be obtained by setting control chart limits to the 95% confidence interval of the smallest offset for each L, range. For L,,, this would be the instrument offset of & 1.6 N (as in Fig. 6). More sophisticated limits are also possible, e.g., limits that take into account the random within-sample variability that would be superimposed on any average drift. In a single indirect verification, it is not possible to determine why a verification failure occurred. However, it is possible for the advanced user to use the CRM to diagnose whether it is the stylus or the calibration that is causing instrument performance failure. A suggested strategy is to use an additional ‘master’ stylus. The ‘master’ stylus should be used solely for verification failure diagnosis so that the possibility of damage or wear is minimised. A control chart is maintained for each stylus, both using the same piece of CRM. The trend in the control charts may then be used to inform the user of the relative performances of instrument and stylus. If the trends of both master stylus and test stylus control charts change at the same time and in the same direction, it is likely that the instrument calibration is changing. Otherwise, it is most likely that the test stylus is at fault. This strategy is summarised in the diagnosis flowchart given in Fig. 7. This strategy requires a control chart history of the master and any test stylus on the same CRM sample. A simple strategy to maintain continuity of history is to use more than one CRM sample at a time. The combined strategy would then look something like that depicted in Fig. 8.
4dvances in adhesion measurement good practice
Verification failure or control chart anomaly requires diagnosis.
Is
Calibration problem indicated Recalibrate and repeat verification
OK see Table 2) Is mean Nithin the lerification range?
4
Replace test st!lus nith master st\lus and repeat tests
Plot result
control chart.
189
within the master control
Test stylus failure indicated. Replace test stilus. Repeat tests for all styli with no control chart history.
Figure 7. Flowchart to diagnose, by using a master stylus, the reason for verification failure.
1 cRM3
.
Figure 8. A strategy for continuous control chart history using multiple master and test styli. Diagnosis of calibration or stylus change requires a control chart for both the current master stylus and the test stylus on the same piece of CRM. The vertical lines link black dots that signify a test using the stylus indicated on each of the CRM pieces linked.
The black dots represent a test using the elements joined by the vertical lines. Since styli do wear away, the continual adoption of the previous master stylus as the new test stylus is one way of minimising exposure to the risk of biased results due to wear of the master stylus.
190
N. M. Jennett et al.
Estimate of difference in Lab, offsets
Figure 9. Schematic of comparison between two laboratories using the same stylus and CRM to identify the combined laboratory offset. No distribution is shown for the CRM as only a single sample is used.
The CRM can also be used to compare results from different instruments. This is an important issue when the scratch test is used for material or component acceptance testing. For the scratch test to be useful, it is essential to reduce the uncertainty in the comparison. If no special steps are taken, the variation between the results of two laboratories, even on the CRM, could be up to the entire range of VR(5). Even if the two laboratories test the same component or product sample, this potential difference makes agreement difficult. A meaningful comparison can only be achieved if the two laboratories understand the various offsets that can affect their results and act to remove or to characterise these offsets. The method proposed here revolves around estimating the differences between laboratories. It is not sufficient to exchange only test samples. CRM samples and, if possible, styli should be exchanged too. If the same piece of CRM and the same stylus are used by the two laboratories comparing results, the difference in results should reflect the combined laboratory offset (see Fig. 9) and the same difference should be obtained in both laboratories. If only the CRM is exchanged, then the difference between results will be a combination of laboratory offset and the difference in stylus offsets (see Fig. 10). At present, it is not possible to use knowledge of the combined laboratory offset to directly correct or normalise test results obtained on another material. Only the CRM is certified for within-sample reproducibility. The combined offset is, however, a useful guide to the differences between results that might be legitimate.
Advances in adherion measurement good practice
191
Estimate of combined inter-
Lab. 2 result +
I
I ..................................... Verification 95% limit
Figure 10. Schematic of comparison between two laboratories using the same CRM but different styli.
4. CONCLUSIONS
A certified reference material for the scratch test is now available to indirectly validate scratch test measurements. It exhibits three reproducible failure events at certified critical load ranges. In addition, a Measurement Good Practice Guide is available, which describes methods and strategies for using the reference material to compare results from different instruments. These tools enable the user to detect errors in the apparatus and the test, especially those caused by deviations of the stylus tip shape. The combination of these tools provides scratch test users with the opportunity to significantly improve current measurement practice in coating adhesion characterisation. Acknowledgements
The authors would like to acknowledge contributions from other National Physical Laboratory staff Dr. Peter Harris for performing the statistical regression analysis and Saffron Owen-Jones, who co-authored the NPL Good Practice Guide. Also acknowledged is Dr. Chris Ingelbrecht of the Inst. Reference Materials and Measurements, who performed the homogeneity study ANOVA and assisted with the official certification of BCR-692. We also acknowledge the partners and participating laboratories in the REMAST project, listed below, without which the certification of the CRM would not have been possible: Vlaamse Instelling voor Technologisch Onderzoek, Teer Coatings Ltd, National Physical Labo-
192
N. M. Jennett et al.
ratory, Laboratoire de Science et Genie des Surfaces - Ecole des Mines, Mossner GmbH - Hamodia, Inst. Pedro Nunes Univ. Coimbra, VTT Manufacturing Technology, Bundesanstalt fur Material Forschung und Priifung - BAM, Inst. for Reference Materials and Measurements, RSCE - Univ. Hull, Univ. de Poitiers and CSM Instruments. Moreover, we gratefully acknowledge the financial support of the European Commission (contract SMT4-CT98-2238) and the UK Government Department of Trade and Industry. REFERENCES 1. 0. S. Heavens, J. Phys. Radiat. 11, 355 (1950). 2. P. Benjamin and C. Weaver, Proc. R. Soc. London Ser. A 245, 163 (1960). 3. S . J. Bull, in: Adhesion Measurement of Films and Coatings, Vol. 2, K. L. Mittal (Ed.), pp. 107130, VSP, Utrecht (2001). 4. VITO, Atlas of Scratch Test Failure Modes (now contained within prEN 1071-3:2003). VITO, Mol (2003). 5. European Standard prEN107 1-3:2000:E, CEN Management Centre, Brussels (2000). 6. European Commission - Measurements and Testing Programme, Project Development and Validation of Test Methods f o r Thin Hard Coatings (FASTEJ,contract MAT1 -CT94/0045, completed 31/12/97. EC, Brussels (1997). 7 . J. Meneve, H. Ronkainen, P. Anderson, K. Vercammen, D. Camino, D. G. Teer, J. Von Stebut, M. G. Gee, N. M. Jennett, J. Banks, B. Bellaton, E. Matthaei-Schultz and H. Vetters, in: Adhesion Measurement of Films and Coatings, Vol. 2, K. L. Mittal (Ed.), pp, 79-106, VSP, Utrecht (200 1). 8. European Commission - Standards, Measurements and Testing Programme, Project “A Certified Reference Material f o r the Scratch Test - REMAST”, contract SMT4-CT98/2238, completed 31/12/2001. EC, Brussels (2001). 9. European Commission, Joint Research Centre, Certij?ed Reference Material for the Scratch Test BCR-692. Institute for Reference Materials and Measurements, European Commission, Directorate General, Joint Research Centre, Brussels (2002). Also accessible at http://www.innm.jrc.be 10. IOS, International standard I S 0 6508-2:1999, “Metallic Materials - Rockwell Hardness Test Part 2: Verification and Calibration of Testing Machines (scales A, B, C, D, E, F, G, H, K, N T).” International Organization for Standardisation, Geneva (1999). 11. N. M. Jennett and S . Owen-Jones, NPL Measurement Good Practice Guide No. 54 “The Scratch Test: Calibration, Validation and the Use of a Certified Reference Material“, NPL Materials Centre, Teddington (hardcopies may be obtained from http://www.npl.co.uWe-storel).
APPENDIX A
A.1. Certification design
Nine independent laboratories contributed data to the certification campaign, which also included homogeneity and stability testing. Each laboratory tested two randomly selected samples of the candidate CRM with a stylus taken from a set of certification styli that were specially manufactured and characterized to ensure that the requirements of EN IS0 6508-2 were met. Each laboratory performed 40 scratches on each of their samples and the critical loads were measured according to a certification procedure that was strictly enforced. A key component of the
Advances in adhesion measurement good practice
193
certification design was the use of a pilot laboratory. This laboratory not only contributed data to the certification using its own allocated stylus and samples, but it also used a control stylus to test all of the samples that had been tested by the other laboratories. Furthermore, it also used the stylus allocated to each laboratory to test the same samples allocated to that laboratory. Additional quality assurance was obtained by having the pilot laboratory re-measure the failure positions of a selection of each laboratory’s scratches. In a separate study, the pilot laboratory tested a single sample using 8 styli spanning the range of stylus radius allowed by EN IS0 6508-2, to establish a measure of the critical load sensitivity to sample radius. In this way, the effects due to the inevitable small variations in samples could be distinguished from the effects of stylus variation and calibration uncertainty. A.2. Calculation of certification value and uncertainty
From the large volume of scratch test data, a certified critical load range (CV ? UNC) was calculated for each of the three failure events (see Table 1). The certified value (CV) for each critical load was taken to be the average of all certification data, giving equal weight to each laboratory. This single value was assigned to all 1100 samples produced. The actual L, value obtained for any specific sample, however, will be distributed about the certified value as quantified by the expanded uncertainty, UNC. This uncertainty arises because the samples are unlikely to be absolutely perfectly homogeneous. There will inevitably be small differences between batches of samples, between samples in the same batch and possibly even within a single sample. On top of this, there will be an uncertainty in the estimate of CV that reflects the fact that there was a finite measurement uncertainty in each of the certification data sets and that not all of the 1100 samples were tested. The value of UNC was estimated by standard statistical methods. The batch-to-batch and sample-to-sample material heterogeneities were determined by a single laboratory using a single stylus in a standard ANOVA homogeneity study. The certification measurement uncertainty was taken to be the standard error of the mean values obtained by the certifying laboratories from the 18 samples tested. Thus, the uncertainty in the certified value is expressed as an expanded uncertainty, UNC, calculated as: UNC = f k . (s2,./p+ u2bb+u2bs+u2us)1’2,
(AI)
where s, is the standard deviation of the means of the p accepted data sets (1 data set consisting of the results of 40 scratches on a single certification sample), Ubb, ubs and u,, are uncertainty components (as standard errors) related to heterogeneity (between-batches ubb, between-samples ub, and within-sample uWs)and k is a coverage factor to expand the range to include 95% of possible results. In simple terms, the certification range (CV f UNC) represents the range in which 95% of L, values would fall using the hypothetical average stylus and instrument.
Adhesion Aspects of Thin Films, Vol. 2, pp. 195-201 Ed. K.L. Mittal 0VSP 2005
Film hardness effect on adhesion strength of Ti02 film on a glass substrate measured by the scratch test AKIRA KINBARA,13x EIJI KUSANO’ and HIDEHITO NANTO’ ‘Research Centerf o r Advanced Science and Technologj, The UniversiQ of Tokyo, 4-6-1 Komaba, Meguro-ku, Tokyo 153-8904, Japan ’Advanced Materials Science, R&D Center, Kanazawa Institute of Technology, 3-1 Yatsukaho, Matto-City, Ishikawa-Prej! 924-0838, Japan
Abstract-The effect of film hardness on the scratch adhesion test values of TiOz thin films on a glass substrate was investigated. In order to investigate the hardness effect on the value of scratch adhesion, multilayers of Ti/TiN were deposited on top of the Ti02 layer. The hardness of the multilayer has been found to vary with the thickness of Ti/TiN. We have used multilayers of various Ti/TiN thicknesses as overcoats to vary the hardness. The adhesion value determined by the scratch test showed a monotonous increase with the hardness of the multilayer overcoat. Our attention was focused on the indentation process as the indentation is assumed to be the main factor in detachment of the film from the substrate. An energetic approach is employed to interpret the hardness dependence of the critical load in the scratch adhesion test. Keywords: Thin film adhesion; scratch test; film hardness; Ti02 film.
1. INTRODUCTION
Mittal has listed 355 techniques for measuring adhesion of films and coatings [ 11. Since then, new techniques might have been added but the scratch test is still one of the most commonly used methods to evaluate the adhesion of thin films on solid substrates because of its easy operation, low price and wide applicability. However, the film detachment process in scratching is too complicated to be directly related to adhesion energy and other physically significant interface quantities. The ambiguity of the detachment process during scratching prevents a quantitative understanding of the scratching process [2]. Strictly speaking, adhesion is an interfacial phenomenon and the fundamental adhesion represents the interactions between the film and the substrate [ 11. However, scratching includes indentation and plowing processes. Both processes are not only affected by the interface property but also by both film thickness and film hardness. *To whom correspondence should be addressed. Tel.: (81-3) 5452-5483; Fax: (81-3) 5452-5484; e-mail:
[email protected]
196
A. Kinbara et a1
As the thickness dependence of the critical load to peel off the film by scratching has already been measured and discussed by Bull [2] and Davanloo et al. [3], in the present report, we focus our interest on the film hardness effect on the scratch test results. We measured the critical load to scratch Ti02 thin films of a fixed thickness on a glass substrate. In order to clarify the effect of the hardness of the sample, Ti/TiN multilayers with a fixed total thickness but with various layer numbers were deposited on top of the Ti02 thin film as overcoats of different hardnesses. The critical load for scratching was evaluated by a microtribometer (CSR-02, Rhesca, Tokyo, Japan) as a function of the Ti/TiN thickness and the effect of the film hardness on the critical load was discussed. 2. EXPERIMENTAL
2.1. Sample preparation Ti02 thin films were deposited onto a non-alkali (NA) glass substrate by dc reactive magnetron sputtering from a Ti target in an Ar/02 discharge gas mixture. The thickness of the Ti02 thin film was 50 nm. DC current was 0.4 A. Ar and O2 gas pressures were kept constant at 0.4 and 0.1 Pa, respectively, during the sample preparation. On top of the Ti02 film, a Ti thin film, 50-nm thick, was deposited as an adhesion promoting layer by Ar gas sputtering from a Ti target to assure a strong adhesion of the following multilayers. On the Ti thin film, Ti/TiN multilayers were deposited by reactive sputtering from the Ti target in an Ar-N2 discharge gas mixture. The flow rates of Ar and N2 gas were 4.8 and 1.0 sccm, respectively, during TiN deposition. For every multilayer, the thicknesses of Ti and TIN were the same. The thickness of the multilayer was fixed at 400 nm and, hence, the overall thickness of the coating sample was 500 nm. The hardness of the multilayer was varied by varying the layer number of the Ti/TiN. Thus, these multilayers acted as variable hardness overcoats. By varying the number of Ti/TiN layers in the multilayer (we denote the total thickness of the multilayer divided by the number of layers as the “Ti/TiN thickness” hereafter), the hardness of the sample was varied. The sample configuration is shown in Fig. 1. The hardness of multilayers has been investigated by a number of workers. For example, Ljungcrantz et al. [4] measured the hardness of a TiN/NbN multilayer grown by reactive magnetron sputtering. For this multilayer, the dependence of the hardness on the TiN/NbN thickness was not clear, although a slight dependence was observed. 2.2. Hardness measurement
Recent techniques for determining the hardness of samples have been reviewed by Oliver and Pharr [5]. The hardness of the multilayer samples was measured by a nano-indenter (ENT-1040: Elionix, Tokyo, Japan) using a Berkovich-type dia-
Film hardness effect on adhesion strength of TiOzfilm
197
Figure 1. Configuration of the test sample.
mond stylus in the present work. The details of the hardness measurement have been reported elsewhere [6]. The hardness of the multilayer samples was evaluated from the applied load on the stylus and the indentation depth. The loading and unloading curves for the applied load formed a hysteresis loop as has been usually observed in many other similar experiments. The hysteresis has been simulated using the finite element method [7]. If the hardness is determined from the indentation depth of the stylus during the loading process, the value obtained is called the “dynamic hardness”, denoted by HDin this paper, while, if the indentation depth is measured after removing the stylus, the hardness value obtained is called “plastic hardness” and is denoted by Hp. The HDincludes contributions from both Hp and the elastic part during the indentation process. This elastic part is termed “elastic hardness” and is denoted by HE. In the present study, we assume that plastic and elastic properties exist simultaneously in the Ti/TiN multilayer and the HD is related to them as follows:
+
( 1/HD)1’2 = ( 1/Hp)’/2 ( 1/HE)’j2
(1) Both HDand Hp are determined directly from the loading-unloading curve, and HE is calculated from this relation. The experimental results on H D , Hp and HEare shown in Fig. 2 as a function of Ti/TiN thickness. It was observed that both HD and H p increased with decreasing WTiN thickness but the HEreached a minimum at around 20 nm. Here, we consider the effect of applied energy during the indentation. The area under the loading curve corresponds to the total work done by the stylus [6]. The area under the unloading curve corresponds to the energy stored by elastic deformation, and the area surrounded by the loading-unloading curve corresponds to the energy dissipated during plastic deformation of the sample. Both elastic and
198
A. Kinbara et al.
plastic deformation energies are shown in Fig. 3 as a function of the Ti/TiN thickness. From Figs 2 and 3 , it becomes clear that energy dissipation in the plastic deformation is relatively small in hard samples.
70.0 60.0
4
50.0 h
?I
9 40.0
____
tDynamic
cn
{ 30.0
I
4-Plastic
U Elastic ______
I
20.0
10.0 0.0
0.0
5.0
10.0 15.0 20.0 Ti/TiN Thickness (nm)
25.0
30.0
Figure 2. Hardness of the multilayers as a function of Ti/TiN thickness. Total thickness of the film sample was 500 nm. H D ,dynamic hardness; H p , plastic hardness; HE,elastic hardness.
0.14 0.12 0.1 2 0.08 0.06 W
I
h
0.04 0.02 0 0.0
/ Plastic
10.0 20.0 Ti/TiN Thickness (nm)
30.0
Figure 3. Elastic and plastic energies of film samples during indentation. Load was 4.9 mN.
Film hardness effect on adhesion strength of TiOzfilm
199
3. RESULTS AND DISCUSSION
The critical load to peel off the multilayer film by scratching was measured using a micro-tribometer (CSR-02, Rhesca). Optical microscope observation showed that the film peeling always occurred at the interface between the TiOz film and the substrate. The results obtained are shown in Fig. 4 as a function of the Ti/TiN thickness. Combining Fig. 2 and Fig. 4, the relation between the hardness and the critical load is obtained. The dependencies of critical load on HD,Hp and HE are shown in Fig. 5a-c. It has been found that a harder film results in a larger value of the critical load. This means that hard films apparently behave as well-adhered films, irrespective of the interface properties. Recent study of stress distribution under a stylus by the finite element method shows that the stress and the plastic deformation are localized in a small area under the stylus [8]. If the film is hard, it is not easy to be deformed by the indentation of the stylus and the indentation depth becomes small. Hence, the film deformation is localized in a small indented area, and the stress or energy is not large enough to detach the film at the interface between the film and the substrate. Thus, a hard film behaves apparently as a strongly adhering film. Johnson, Kendall and Roberts have written an excellent account of the indentation process [9] and their analysis has been improved by Sridhar and Sivashanker [lo]. But these analyses are based on elastic deformation and these cannot directly be applied to our results for determining the interfacial energy because our results include the influence of plastic deformation. In Fig. 3, we observe that the plastic deformation energy dissipated in the film sample during indentation decreases with the decrease of the Ti/TiN thickness which is equivalent to an increase of the sample hardness. On the other hand, the
250.0
2 200.0 E
W
150.0
3 3 100.0 0 ..v
5
50.0 0.0
0.0
5.0
10.0
15.0
20.0
W T i N Thickness (nm) Figure 4. Adhesion (critical load) vs. Ti/TiN thickness.
25.0
30.0
A . Kinburu et al.
200
(4 250.0 200.0
T
2
E v
2
150.0
-2 100.0 1 0
.-
.$
0
50.0
0.0
0.0
z
250.0
-
200.0
-
4.0 6.0 8.0 Dynamic Hardness (GPa)
2.0
10.0
12.0
E v
; 150.0 1 0 -
~
2 100.0 .w .-
-
50.0
-
6
0.0
250.0
200.0 1 -
z
E
150.0
~
m
3 I
;:
100.0
-
50.0
-
I O
1
00
' 00
____~
1
100
50 ~
~
150 200 250 Elastic Hardness (GPa)
~
_
_
~
_
300
350
400
_-~ -_ _ _ ~
Figure 5. Adhesion (critical load) vs. (a) dynamic hardness, (b) plastic hardness, (c) elastic hardness.
Film hardness effect on adhesion strength of Ti02film
201
thickness dependence of the elastic energy is not clear. As the stored elastic energy is completely liberated on removing the stylus, so this energy is considered not to be used for scratching and thus only the dissipated (plastic deformation) energy is considered to be related to scratching. This energy is partly dissipated as heat during the plastic deformation of the film material but the rest is used for film peeling during scratching. In the usual scratching method, it seems to be impossible to separate the dissipated energy into the pure plastic deformation part and the work to peel-off the film. This makes it difficult to establish the scratching load as a physically meaningful quantity, although the test is still very useful from a practical point of view. 4. CONCLUSIONS
The scratching process during indentation has been discussed. The scratch test is a useful technique for an estimation of the adhesion strength of thin films. But the load required to peel off the film is strongly affected by the hardness of film. The results of the scratch test include the effects not only of the adhesion at the interface between the film and the substrate but also of the bulk hardness of the film. This test is a typical example of “practical adhesion” named by Mittal [l]. However, this test is still a useful technique for the adhesion evaluation and is a suitable indication of the durability of thin films. REFERENCES 1. K. L. Mittal, in Adhesion Measurement of Films and Coatings, K. L. Mittal (Ed.): pp. 1-13. VSP, Utrecht (1995). 2. S. J. Bull, in Adhesion Measurement of Films and Coatings, Vol. 2, K. L. Mittal (Ed.). pp. 107130, VSP. Utrecht (2001). 3. F. Davanloo, C. B. Collins and K. J. Koivusaari, in Adhesion Measurement of Films and Coatings, Vol. 2, K. L. Mittal (Ed.), pp. 141-157. VSP, Utrecht (2001). 4. H . Ljungcrantz, C. Engstrom, L. Hultman, M. Olsson, X. Chu, M. S. Wong and W. D. Sproul, J. Vac. Sci. Technol. A16, 3104-3113 (1998). 5 . W. C. Oliver and G. M. Pharr, J. Mater. Res. 7, 1564 (1992). 6. N. Kikuchi, M. Kitagawa, A. Sato, E. Kusano, H. Nanto and A. Kinbara, Surf: Coating. Technol. 126, 131-136 (2000). 7. M. Lichinchi, C. Lenardi, J. Haupt and R. Vitali, Thin Solid Films 312, 240-248 (1998). 8. C. F. Robertson and M. C. Fivel, J. Mater. Res. 14, 2251-2258 (1999). 9. K. L. Johnson, K. Kendall and A. D. Roberts, Proc. Roy. SOC.Lond. A324, 301-313 (1971). 10. I. Sridhar and S. Sivashanker, Surf: Coating. Technol. 167, 181-187 (2003).
Adhesion Aspects Ed. K.L. Mittal 0VSP 2005
of
Thin Films. Vol. 2. pp. 203-213
Two critical events observed on Cu films on glass substrate in the microscratch test SHIGERU BABA," YOSHIHIRO YAMAGUCHI, MASASHI OGAWA and TAKE0 NAKANO Department of Applied Phjsics, Seikei University, 3-3-1 Kichijoji-Kitumachi, Musushino-shi, T o b o 180-8633, Japan
Abstract-Copper films of 100-200 nm thickness were deposited on a Corning 7059 glass substrate. The adhesion was examined with a vibrating microscratch tester, employing diamond styli with tip radii R of 15 and 100 pm. In the friction signal measured during progressive scratch load, two types of film failures could be detected. The nature of the failures, especially their dependence on the film thickness and the stylus tip radius, was studied in detail. When a stylus of 100 pm radius was employed, the critical loads for both failures increased as the film thickness increased. However, when a stylus of 15 pm radius was employed, the critical load for the first failure decreased as the film thickness increased, while the critical load for the second failure increased with the film thickness. We speculated that the first failure corresponded to the failure at the surface of copper films. Therefore, we employed a nano-indentation hardness tester in order to examine the film surface hardness. The result showed that the film hardness decreased as the thickness increased. On the other hand, the second failure corresponded to the cohesive failure in the film, which was scraped off from the substrate surface. The observations demonstrate that the choice of tip radius and a careful inspection of the scratched trace are important in the scratch experiment.
Keywords: Scratch test; adhesion; copper film; stylus radius; glass substrate.
1. INTRODUCTION
A conventional scratch tester has been developed into a variety of microscratch testers for investigation of the tribological behavior of thin films and surfaces [1, 21. The vibrating microscratch tester was first proposed in 1986, and the performance was reported to be so sensitive that the adhesional failure of MgO films of less than 30 nm thickness could be detected in real time [ 3 ] . This microscratch tester has now been approved as one of the standard test machines for the adhesion of thin films on glass in the Japanese Industrial Standards [4]. The machine employs a hemispherical diamond stylus with a radius of curvature of 5-100 pm "To whom correspondence should be addressed. Tel./Fax: (81-422) 37-3780; e-mail:
[email protected]
204
S. Baba et al.
on the end of a cantilever beam. In the typical operation, the stylus is pressed against the film surface at a constant loading rate via the spring force of the cantilever, while the sample is driven horizontally. Along with the scratching, the pivot point of the cantilever is forced to oscillate perpendicular to the scratch direction with a small amplitude as in a reciprocating ball-on-disk tribometer [5]. Thus, the stylus exhibits a stick-slip motion due to the friction force [6]. That is, the stylus stops at the end of the oscillation when the force due to the spring constant of the cantilever is less than the friction force, and the stylus begins sliding when the spring force exceeds the friction force. The velocity difference between the stylus and the pivot point of the cantilever gives rise to an AC voltage signal by electromagnetic induction, just as a gramophone cartridge does. The absolute value of the AC voltage is proportional to the friction force [7]. The friction signal is plotted as a function of sliding time. The resultant figure of the friction force vs. stylus load (friction vs. load curve, for short) is a characteristic curve of the specimen, which is substantially the same as that obtained in the conventional scratch measurement [8]. As the critical load for film failure (Lc) varies with specimen materials, experimental parameters such as elastic compliance of the cantilever and the oscillation amplitude are set properly besides the stylus radius. If the oscillatory force is not applied, the machine operates as a conventional scratch tester. However, with the aid of oscillation, a much higher sensitivity for detection of film failure can be achieved, and the detection can be made more reproducible at the same time. Recently, copper has been considered as one of the key materials for interconnection in the ULSI technology because of its superior electrical and thermal properties. For the wiring formation by copper films, the importance of the mechanical behavior of copper under stress conditions has been reported [9]. We have been studying mechanical properties of copper films on glass with a vibrating microscratch tester. In the course of our study, we found that even harmonics of the AC signal voltage contained information about the surface failure and we could detect the surface failure clearly even for ductile materials [lo]. As for failure progression in copper films during scratching, we found another failure at a light load before the critical failure occurred. The characteristic behavior of the scratch failure of copper films is reported in this paper. 2. EXPERIMENTAL
Copper films were deposited by thermal evaporation onto a Corning 7059 glass substrate. The glass substrate (1.2 mm thick) was heated at 150°C for 15 min in a vacuum of about 2.0 x Pa and then cooled down to less than 50°C prior to film deposition. The vapor flux of copper was monitored with a quartz crystal microbalance and the deposition rate was kept constant at 0.18-0.26 n d s . Copper films of 100, 150 and 200 nm thickness were deposited. The scratch measurements were carried out using a microscratch tester CSR-02 (Rhesca, Tokyo, Ja-
Two critical events on Cufilms in microscrutch test
20s
pan). A diamond stylus of 15 or 100 pm radius was employed. The frequency and the peak-to-peak amplitude of the cartridge vibration were f = 30 Hz and A = 120 pm,-,. The stylus traveling speed v s was 10 p d s in the scratch direction, and the maximum sliding speed in the reciprocatory motion (= r f A ) was 11 m d s . 3. RESULTS
An optical micrograph of the scratch trace on the film surface is shown in Fig. 1. The stylus traveled on the surface in a serpentine manner from left to right with increasing stylus load. An illustration of the stylus motion is shown schematically on the left of the scratch trace. The advancing pitch of the stylus in the scratch direction per oscillation (= v,/f) was 0.3 pm in this experiment. It should be noted that the pitch length is greater than the film thickness and, hence, effects such as wear and fatigue due to the reciprocal motion of the stylus are not expected to be serious. The signal voltage which was proportional to the friction force was plotted versus stylus load, which is shown on the bottom. A small kink appeared in the curve at 7.9 mN. At this stage pile-up of film material was observed on the right edge of the scratch path (lower side edge in the photograph). As the stylus load increased
0
5
10
Stylus Load, L
15
n
20
[ mN ]
Figure 1. A typical scratch trace observed on the surface of a copper film after carrying out the vibrating microscratch test. The stylus with a radius of 100 km travels from left to right with the load increasing at a constant rate. The stylus load is shown along the scratch trace on the bottom. The measured friction force is also plotted. The critical load for film failure is indicated with a triangle mark. The stylus motion during scratching is illustrated schematically on the left.
S. Baba et al.
206
R= 15 pm
0.0
1.o
2.0
3.0
4.0
stylus Load , L [ mN ] Figure 2. A characteristic friction force vs. stylus load curve for copper films on the glass substrate, where a stylus of 15 p m radius was employed. A new type of film failure was observed at a light load LA, which is named “A-mode’’ failure in this paper. A typical failure observed at a high load LB is named “B-mode” failure.
further, an irregular fluctuation in the friction force occurred at about 18 mN. The signal irregularity was accompanied by a partial film delamination or scraping-off of the film material. The shape of the friction vs. load curve is a characteristic of the particular combination of film and substrate materials and their properties. The irregularity in profile in the friction vs. load curve for copper films on glass appeared almost in the same manner as above when a stylus of 15 pm radius was employed, but the characteristic profile could be identified more clearly as shown in Fig. 2. The first kink is observed at 1.8 mN. This irregularity which occurs at a light load is tentatively named the “A-mode’’ failure. The second irregularity which occurs at a high load is named the “B-mode” failure. According to the microscopic observation, the film was found to be very thin and looked transparent when it was illuminated from the back side of the glass beyond the point of failure B. The peeloff delamination does not take place for films of ductile metals, such as copper, unless the adhesion is very poor. Scratch measurements were repeated to study the statistical nature of the film failures, and samples of different thickness of copper were also examined. First, the distribution of the critical loads, where a stylus of 100 pm radius was employed, was studied. The correlation between the critical loads for failures A and B is shown in Fig. 3 for three film thicknesses. The scatter in the critical loads is
Two critical event,c on Cufilms in microscratch test R=l OOym
201
Cu film/Corning7059 glass
25 I
20
z
E
15
I
I1
Y
m
4
10
0
I
I
Figure 3. The distribution and the correlation of the critical loads for failure A and failure B for different film thicknesses, where a stylus of 100 pm radius was employed. Thicker films required greater loads to cause both failures.
5-10% among each group, but the critical load, on average, increases with the film thickness. The improvement of adhesion with film thickness has often been reported [ l l , 121. On the other hand, the results obtained employing a stylus of 15 ym radius are slightly different (Fig. 4). The critical load for failure B showed thickness dependence similar to that for 100 pm radius stylus, but the critical load for failure A is correlated inversely with the film thickness. Thicker films showed lower values of the critical load for failure A when they were scratched with a stylus of 15 pm radius.
4. DISCUSSION
In order to determine the critical loads more accurately, the friction coefficient was calculated by dividing the value of friction force with the stylus load. An example of the change in the friction coefficient which is calculated from the results in Fig. 2 is shown in Fig. 5. High values of the friction coefficient at the beginning of the scratch should be ignored as they are artifacts due to a background noise of the friction signal divided by a small value of the stylus load. Going beyond this region, we can see three regions where the friction coefficient does not
S. Baba et al.
208
R=l5pm
Cu film/Corning7059 glass
5
-z
4
E - 3
0 200nm 0 150nm A 100nm
1
0 ' 0.5
I
I
I
I
1.o
1.5
2.0
2.5
LA
[mN1
Figure 4. The distribution and the correlation of the critical loads for the failure A and the A u r e B for different film thicknesses, where a stylus of 15 pm radius was employed. Greater loads were required to cause the failure A for thinner films, while the failure B required greater loads for thicker films.
change greatly with the stylus load and two transition regions from one stable region to another. The first transition region corresponds to the failure A, and the second to the failure B. The curve of friction coefficient vs. load could be approximated by a combination of straight lines. Then by drawing straight lines as shown in Fig. 5 , the intersection points could be determined clearly. These kink points were indexed A1 , A2, B 1 and B2 as the beginning and end points of the respective failures. We applied this procedure to all the friction vs. load curves, and data sets of the critical load, friction and friction coefficient were obtained. The values of friction force were plotted against the corresponding stylus loads. Results are shown in Fig. 6 and Fig. 7 for the failures A and B, respectively. The figures show how these points A l , A2, B1 and B2 are distributed in the friction vs. load curve. First, the nature of the failure A is discussed. We can see in Fig. 6 that in the data points for points A1 and A2 the critical loads for thinner films are located slightly to the right. As for point A l , all data points seem to form a group, and the scatter in the friction force is less than that in the critical load. This fact suggests that failure A is triggered by the friction. In other words, for samples having a low coefficient of friction, a large stylus load is required to trigger failure A. As thick
Two critical events on Cufilms in microscratch test
0.0
209
1.o 2.0 3.0 Stylus Load, L [ m N ]
4.0
Figure 5. The friction coefficient is plotted as a function of the stylus load which was calculated from the friction-load curve in Fig. 2. The curve can be divided into three stable regions where the friction coefficient is almost constant and two transition regions between them. Intersection points A l , A2, B1 and B2 are determined by drawing straight lines around the linear regions.
R=15pm ~
Film thickness 0
200nm
Critical Load, LA [ mN ] Figure 6. The distribution of data points for points A1 and A2 for different film thicknesses, which are plotted in the friction vs. load curve employing a stylus of 15 pn radius.
S. Baba et al.
210
films have a tendency for a large friction coefficient, a light load is enough to trigger failure A in thick films. Another fact that the critical value of the friction force does not depend strongly on the film thickness implies that the failure A should be ascribed to a thickness-independent property of the filmhubstrate system. On the other hand, the data points for point A2 seem to form different lines for different thicknesses. It is again true that thinner films require more loads to complete failure A than thicker films do, because the friction coefficient is lower for thinner films. The distribution of the data points for failure B is shown in Fig. 7. Data points for point B 1 form straight lines which pass near the coordinate origin of the friction vs. load curve. Though the friction force at point B1 is slightly higher for thicker films, the load value does not depend clearly on the film thickness. On the other hand, data points for point B2 show a thickness dependence. Thick films have a tendency for high critical loads and high friction for point B2. By taking into account the microscopic observation that failure B corresponds to the scraping-off of copper films from glass substrate, it is speculated that the stylus touches the glass substrate at point B 1 and that the entire copper film is removed from the glass substrate at point B2. The removal of copper is not due to peeling-off, but is more due to wearing-off. In this sense, it can be said that either a cohesive failure in the copper film or failure between copper and glass takes place at point B2. The observation that the friction force at point B2 increased with film thickness
1800 3 1600@ a 1400' 12002? c, lO0Ol CU a> 8000 6001 $? 400L1: .o 200.22
I
I
S
h
i -
-
-
.-0
R =15pm Film thickness 0
<
-
I
A
200nm 150nm
I
100nm
I
I
I
Critical Load, LB [ m N ] Figure 7. The distribution of data points for points B1 and B2 for different film thicknesses observed in the friction VJ. load curve (corresponding to that of Fig. 6) employing a stylus of 15 pm radius.
Two critical events on Cufilms in microscratch test
21 1
agrees with the fact that the friction force on the coated surface consists of the plowing force and the shearing force at the interface, and the plowing component depends on the film thickness [ 131.
0.5
2 E
Y
U m 0
0.4
0.3 0.2 0.1 0.0 0
20
80
60
40
Indenter Depth [nm] Figure 8. A typical nanoindentation profile for a copper film (100 nm thick) on the glass substrate. Measurements were carried out with an Elionix ENT-1 lOOa tester, employing a Berkovich indenter.
v)
cn
Q)
500
C
400
CTJ I
300
.-0 E
200
2
e e*
L
>r 100
n
0 ’ 0
1
1
I
I
I
I
50
100
150
200
250
300
Film Thickness [nm] Figure 9. Thickness dependence of the dynamic hardness of the copper film on the glass substrate when a load of 0.5 mN was applied. To represent the hardness in Pa. the value should be multiplied by 10’.
S. Baba et al.
212
0.0
0.1
0.2
0.3
0.4
0.5
Indenter Load [mN] Figure 10. Dependence of the dynamic hardness on the indenter load, calculated from the load vs. indenter indenter depth curve in Fig. 8. To represent the hardness in Pa, the value should be multiplied by 10’.
The thickness dependence of the film surface hardness was investigated because the friction force is related to hardness. An ENT-1100a depth sensing indentation tester (Elionix, Tokyo, Japan) was employed for this copper/glass system. A typical example of the load-depth curve in the indentation process is shown in Fig. 8. The dynamic hardness (=Martens hardness [14]) at a load of 0.5 mN, which was calculated from the depth of the stylus penetrated from the original surface, had an increasing tendency with decreasing film thickness (Fig. 9). At the same time, the hardness was found to decrease with increasing indenter load. The dynamic hardness calculated at each point of the loading curve in Fig. 8 is shown in Fig. 10. It has been reported that the hardness of thin films is dependent in a complicated way on the material, thickness, hardness of substrate and indentation depth, which has been elucidated with a nano-indentation technique [ 15, 161. Furthermore, in the case of ductile metals, it has been reported that the abundance of crystallographic defects and impurities at the top surface of the film form a hard surface layer like a pie crust [17]. These facts suggest that the hard surface layer formed on the top surface of copper film and/or the effect of glass substrate was responsible for the failure A. It is speculated that a pointed stylus (Le., with low radius) thrusts into the copper film at the load A. When a stylus of 100 ym radius is employed, the contact area between the spherical stylus and film becomes wide and the maximum stress concentration occurs in a deep region beneath the surface [18], so that the failure begins to occur as a cohesive failure in the copper film. The cohesive strength is higher in thicker films. These observations indicate that the choice of a proper tip radius and a careful inspection of the scratch trace are very important in a scratch experiment.
Two critical events on Cu films in microscratch test
213
5. CONCLUSIONS
Copper films of thicknesses 100, 150 and 200 nm on a Corning 7059 glass substrate were examined with a vibrating microscratch tester. Two types of film failures were observed: one at a light load and the other at a heavy load. The nature of the failures was studied in detail in terms of the distribution of the critical load, friction force and friction coefficient and their thickness dependence. When a stylus with a radius of 100 pm was employed, the critical loads for both failures were found to increase with film thickness. When a stylus of 15 pm radius was employed, however, the critical load for the first failure was observed to increase with decreasing film thickness, while the second critical load increased with increasing thickness. The second failure was due to scraping-off of copper films from the glass substrate.
Acknowledgements The authors would like to thank Mao Komazawa of Elionix for nano-indentation experiments. REFERENCES 1. B. Bhushan, Handbook of Micro/Nunotribology, 2nd edn. CRC Press, Boca Raton, FL (1999). 2. K. L. Mittal (Ed.), Adhesion Measurement of Filnzs and Coatings, Vol. 2, VSP, Utrecht (2001). 3. S. Baba, A. Kikuchi and A. Kinbara, J . Vuc. Sci. Technol. A4,3015-3018 (1986). 4. JIS R3255-1997. 5 . R. Benzing, I. Goldblatt, V. Hopkins. W. Jamison, M. Mecklenburg and M. Peterson, Friction and Wear Devices, American Society of Lubrication Engineers. Park Ridge, IL (1976). 6. G. W. Stachowiak and G. W. Stachowiak, Engineering Tribology. 2nd edn., p. 469. Butterworth-Heinemann, London (2001). 7. S. Baba. A. Kikuchi and A. Kinbara, J. Vac. Sci. Technol. A5, 1860-1862 (1987). 8. J. Sekler, P. A. Steinmann and H. E. Hintermann, Surf: Coating. Technol. 36, 519-529 (1988). 9. K. N. Tu, J. Appl. Phys. 94,5451-5473 (2003). 10. S. Baba, T. Midorikawa and T. Nakano, Appl. S~irf:Sci. 144-145,344-349 (1999). 11. D. Sheeja, B. K. Tay, S. P. Lau, K. W. Leong and C. H. Lee, Znt. J. Modern Phys. B16,958-962 (2002). 12. A. A. Volinsky, N. R. Moody and W. W. Gerberich. Acta Mater. 50, 441-466 (2002). 13. J. Halling, Thin Solid Films 108. 103-1 12 (1983). 14. ISO/DIS 14577- 1.2. Metallic Materials - Instrumented indentation test for hardness and materials parameters. International Organization for Standardization (2001). 15. M. Iwasa, K. Tanaka, J. A. Barnard and R. C. Bradt, Mater. Res. Soc. Symp. Proc. 505, 199-204 (1998). 16. H. Ichimura, F. M. Rodriguez and A. Rodrigo, Surf: Coating. Technol. 127, 138-143 (2000). 17. T. Yoshida and T. Ohmura, J. Sui$ Finish. SOC.Jpn. 51, 262-265 (2000). 18. G. M. Hamilton and L. E. Goodman, J Appl. Mech. 33,371-376 (1966).
Adhesion Aspects of Thin Films, Vol. 2 , pp. 215-232 Ed. K.L. Mittal 0VSP 2005
Abrasion life and scratch durability of sputtered PTFE thin film SATORU IWAMORI," YUICHI NAGAYAMA, YOUSUKE YAMAGATA and YOSHINORI YAMADA Faculty of Engineering, Kanazawa University, 2-40-20, Kodatsuno, Kanazawa 920-8667, Japan
Abstract-Abrasion life and scratch durability of poly(tetrafluoroethy1ene) (PTFE) thin films sputtered onto metal substrates were evaluated using two different pin-on-disk type apparatuses. The scratch durability of the PTFE thin film deposited on the copper substrate was higher than of the film deposited on the nickel substrate. The abrasion life was found to be related to the hardness of the metal substrate. On the other hand, we found that the abrasion life for the mean contact pressure on the sputtered PTFE thin film on the nickel substrate was higher than that on the copper substrate. The abrasion life would also relate to the adhesion to the metal substrate. The adhesion strength between evaporated metal films and sputtered PTFE film was higher than that between evaporated metal thin films and the bulk PTFE. This difference was apparently due to differences in chemical bonding states between the sputtered PTFE thin film and the bulk PTFE. In order to increase the adhesion strength between the sputtered PTFE thin film and metal substrate, a poly(tetrafluoroethylene) (PTFE)-Cu mixed thin film was deposited as an interlayer by RF sputtering of PTFE-Cu mixed target. The sputtering rate of PTFE decreases only slightly with increasing pressure during sputtering, but that of Cu dramatically decreases. The electrical resistance of the PTFE-Cu thin film sputtered onto a glass substrate dramatically increases with increasing pressure when it is sputtered at a pressure beyond 5 mTorr. The PTFE-Cu thin film was introduced between the PTFE thin film and the Cu substrate (PTFE/PTFE-Cu/Cu substrate), and its tribological properties were evaluated. The abrasion life of the PTFE/PTFE-Cu/Cu substrate system improved compared to that of the PTFE thin film sputtered directly onto the Cu substrate. The PTFE-Cu layer acted as a functionally gradient material layer. Keywords: PTFE thin film; RF-sputtering; adhesion strength; tribological properties; PTFE-Cu interlayer.
1. INTRODUCTION
Sputtering is widely used in electrical and mechanical industries, because a sputtered thin film has a uniform structure and an excellent adhesion property to most substrates. Polymer thin films, such as polyimide (PI) and poly(tetrafluor0ethylene) (PTFE), have been sputtered since the 1970s and their tribological *To whom correspondence should be addressed. Tel./Fax: (8 1-76) 234-4950; e-mail:
[email protected],cic.jp
216
S. Zwainori et al.
properties have been characterized [ 1-71. As the tribological properties of the PTFE are known to be only slightly influenced by humidity, a PTFE thin film can be used as a solid lubricant. The scratch durability for the load on the pin and abrasion life are the most important properties for the solid lubricant. The abrasion life of the PTFE thin film was found to be shorter than that of the PI, because the PTFE thin film had poor adhesion property to the substrate [2]. In this paper, PTFE thin films were sputtered onto copper and nickel substrates, and their abrasion, scratch and adhesion properties were characterized. The abrasion life improved by the introduction of the PTFE-Cu mixed thin film as an interlayer between the PTFE thin film and the copper substrate. 2. EXPERIMENTAL
PTFE and PTFE-Cu mixed thin films were sputtered onto the substrates using a conventional RF sputtering apparatus. A TeflonTM(DuPont, USA) sheet (100 mm in diameter, 0.1 mm thick) was used as a sputtering target for depositing PTFE thin films. Five pieces of PTFE plates (25 mm x 25 mm square, 3 mm thick) were placed on the Cu plate (100 mm diameter, 5 mm thick) and used for the PTFE-Cu mixed target. The glass slide, copper plate and nickel plate were used as substrates for these sputtered thin films. The surfaces of these plates were cleaned by ultrasonic treatment in acetone and argon plasma treatment before sputtering. The thickness of the sputtered thin films was determined by measuring the height of the films sputtered onto a glass slide using a surface roughness measurement instrument (Surfcom 1400A-6: Tokyo Seimitsu, Japan). A pin-on-disk type abrasion test apparatus with a steel bearing ball (1 mm in diameter) as the slider was used for the evaluation of abrasion life. A pin-on-disk type scratch test apparatus allows to apply load continuously from 0.49 N to 9.8 N by appropriate compression of the spring with a timing belt and a motor. The load was increased at a rate of 0.4 N/s and the sliding speed of the sample was 20 d m i n . The abrasion life and scratch durability of these thin films on the substrates were evaluated by optical microscope and scanning electron microscope (SEM) observations. Deposition of metal thin films on PTFE thin films was carried out with a conventional vacuum evaporation apparatus, and the thickness of these metal films was 0.3 pm. In order to measure the adhesion strength between the evaporated metal thin films and the PTFE thin film, a stud (6 mm in diameter) was bonded to these metal thin films with an epoxy resin. The adhesion strength was determined by the pulling the stud using TendonTM(Touyou Sokuki, Japan). The pull speed was 5 rnm/min. The chemical bonding states of PTFE thin films and composition of the PTFECu thin films were determined by X-ray photoelectron spectroscopy (XPS). The electrical resistance of the PTFE-Cu thin films sputtered onto a glass slide was determined by measuring the resistivity with a four-point probe array [8].
Abrasion life and scratch dumbility of sputtered PTFE thin jtlm
211
3. RESULTS AND DISCUSSION
3.1. Abrasion life and scratch durability of the PTFE thin film
Figure l a and b shows, respectively, the SEM micrograph of the sputtered PTFE thin film onto the glass slide and the surface roughness profile of the film. The mean surface roughness (R,) of the PTFE thin film was 9 nm, which was almost at the same level as that for the glass substrate. The abrasion life of the PTFE thin film (thickness 0.9 pm) sputtered onto the nickel substrate was evaluated using the pin-on-disk type abrasion test apparatus. The load on the slider ball was 0.4 N. Figure 2a-c shows optical micrographs of the abrasion tracks after 5 , 10 and 15 revolutions, respectively. A slight abrasion track can be seen in Fig. 2a, and this track becomes larger with increasing number of revolutions. Finally, the PTFE thin film peels off from the substrate after 15 revolutions (Fig. 2c). The abrasion life of the PTFE thin film (thickness: 0.9 pm) sputtered onto the copper substrate was also evaluated using the same apparatus. The load on the slider ball was 0.4 N. Figure 3a-d shows optical micrographs of the abrasion tracks after 5 , 10, 20 and 25 revolutions, respectively. The PTFE thin film peeled off from the copper substrate after 25 revolutions [9]. The results on the abrasion life of the PTFE thin films sputtered onto nickel and copper substrates are presented in Fig. 4. The vertical axis shows the number of revolutions at which the PTFE thin film peeled off from the substrate. The abrasion life of the PTFE thin film sputtered onto the copper substrate was longer than that onto nickel substrate. In order to investigate the scratch durability, the PTFE thin films sputtered onto nickel and copper substrates were scratched with the apparatus described above. Figure 5a and 5b shows the optical micrographs of the scratch tracks in the PTFE thin films on nickel and copper substrates, respectively. These photographs also indicate that the PTFE thin film sputtered onto the
Figure 1. Scanning electron micrograph (a) and surface profile (b) of the sputtered PTFE thin film.
218
n W
m
n W
S. Zvvarnori et al.
0
i
Abrasion life and scratch durabilit); of sputtered PTFE thin film
219
(b)
-
-
200 p m
200 p m
-
-
200 pm
200 prn
Figure 3. Optical micrographs of the abrasion tracks in the PTFE thin film sputtered onto copper substrate after 5 (a), 10 (b), 20 (c) and 25 passes (d).
-.-2 50 0 3
9 40 2
!2
~~
~~
'
0
Ni substrate I
Cu substrate
20
0
0.2
0.4
0.6
0.8
Load (N) Figure 4. Relationship between abrasion life (in terms of numbers of revolutions) and load.
220
S. Iwamori et al.
h
e
Abrasion life and scratch durabilit): of sputtered PTFE thin film
22 1
(b)
200 p m
-
200 prn
200 p m
-
200 p m
Figure 6 . Scanning electron micrographs of the scratch tracks in the PTFE thin films sputtered onto copper substrate. (a) 1.5 N, (b) 2.0 N, (c) 4.0 N and (d) 6.0 N load.
copper substrate had a higher scratch durability than on the nickel substrate. Figures 6 and 7 show the SEM micrographs of the scratch tracks in PTFE thin films sputtered onto the copper and nickel substrates, respectively. The tracks after scratching with 1.0 N and 1.5 N loads were arc shaped (Fig. 6a and 6b). In addition, scratch debris was observed at the side of the tracks after scratching with 4.0 N and 6.0 N loads (Fig. 6c and 6d). The tracks in the PTFE thin film sputtered onto the nickel substrate can be observed after scratching with 1.0 N load, and crushed tracks can be observed after scratching with 1.5 N load (Fig. 7a and 7b). The PTFE thin film peeled off from the nickel substrate after scratching with 2.0 N load (Fig. 7c). In addition, the peeled width (200 pm) after scratching with 3.0 N load was three times the diameter of the ball (Fig. 7d). The PTFE thin films sputtered at various conditions were evaluated. Figure 8 shows the relationship between the durability of the PTFE thin films in terms of
222
S. Iwarnori et al.
(a)
(b)
I
1.4pm
I
14 pm
I
14 p m
I
33 p m
Figure 7. Scanning electron micrographs of the scratch tracks in the PTFE thin films sputtered onto nickel substrate. (a) 1.0 N, (b) 1.5 N, (c) 2.0 N and (d) 3.0 N load.
scratch load and the RF power during sputtering. The pressure during sputtering was 1.3 Pa. The scratch durability of PTFE thin films increased with the RF power up to 100 W, and then it remained constant. Figure 9 shows the scratch durability of the films in terms of load and the pressure during sputtering. The RF power during sputtering was 100 W. The scratch durability slightly decreased with increase of pressure. The scratch durability of the PTFE thin film sputtered onto the copper substrate, as well as the abrasion life, were higher than that onto the nickel substrate. In order to investigate the relationship between these peeled thin films and the hardness of the substrate, mean contact pressures between the PTFE thin films and the copper and nickel substrates were calculated. When the stress in the thin film due to the slider ball reaches a critical point, the film begins to be sheared by the stress. The PTFE thin films begin to peel from the substrate with increase of the shearing stress. The average shearing stress in the PTFE thin film was calculated as follows: the friction force was calculated from the friction coefficient and the load [9], and the average shearing stress was calculated from the friction force
223
Abrasion life and scratch durabiliQ of sputtered PTFE thinfilnz
4
1
Ni substrate
-I
-
100
150
I
3 n
z
:*
W
0
1
0
0
50
200
RF power (W) Figure 8. Scratch load vs. RF power during sputtering of PTFE thin films on nickel and copper substrates.
I 0 Ni substrate 1
T
I
3
‘I
n
z
I
T
:* W
0
J
1
0.1
I
IO
Pressure (Pa) Figure 9. Scratch load vs. pressure during sputtering of PTFE thin films on nickel and copper substrates.
224
S. Iwnmori et al.
and the contact area. The value of the average shearing stress in the PTFE thin film sputtered onto the copper substrate was 160 N/mm2 at 8.0 N load. On the other hand, the value for the PTFE thin film sputtered onto the nickel substrate was 350 N/mmz at 2.0 N load. These results indicate that the PTFE thin film sputtered onto the nickel substrate is peeled off from the substrate due to the large shearing stress, but the shearing stress in the PTFE thin film sputtered onto the copper substrate is small and does not bring about peeling. In order to estimate the contact area of the slider ball more correctly, the diameter of the contact area was calculated from the Hertz equation (1): S = [3/4 x WR{(1- vI2)/E1+ (1 - VZ~)/E*}]~’~,
(1)
where S: diameter of the contact area of the slider ball W: load R: diameter of the slider ball vl: Poisson’s ratio of the slider ball E l : Young’s modulus of the slider ball v2: Poisson’s ratio of the substrate E2 : Young’s modulus of the substrate The values of R, v1 and El are 0.5 mm, 0.30 and 20400 kg/mm2, respectively. The values of v2 and E2are 0.30 and 21 000 kg/mm2in the case of nickel substrate, and 0.34 and 11 000 kg/mm2 in the case of copper substrate. Based on the results of the calculation of S, mean contact pressures for the PTFE thin films sputtered on the copper substrate and the nickel substrate were calculated and are shown in Fig. 10. The mean contact pressure is calculated by dividing the friction force by the contact area. The abrasion life of the PTFE thin film sputtered on the nickel substrate was higher than that on the copper substrate. These results indicate that the sputtered PTFE thin film adheres to the nickel substrate more strongly than to the copper substrate. 3.2. Adhesion between evaporated metal thin films and sputtered PTFE thin film
In order to determine the effect of adhesion on the durability, the adhesion strength of metal thin films, such as gold, copper, nickel and aluminum, deposited on the sputtered PTFE thin films by vacuum evaporation was measured (Fig. 11). The nickel thin film adhered to the PTFE thin film most strongly of all the thin films. The adhesion strength between the copper thin film and the PTFE thin film was much smaller than that between the nickel thin film and the PTFE thin film. Bodo and Sundgren [lo] studied the adhesion of titanium, aluminum, chromium, nickel, copper, silver and gold thin films deposited onto high density polyethylene (HDPE) substrate by electron beam evaporation. The adhesion to the HDPE substrate was relatively high for nickel film, while it was low for aluminum, copper and gold films. The adhesion to the PTFE substrate was relatively low for these metal thin films, as shown in Fig. 11. The PTFE thin coatings were deposited by the spin-coating technique using polymer dispersions on the titanium, aluminum,
--.-
225
Abrasion life and scratch durability of sputtered PTFE thin film
zl
Ni substrate
- Cu substrate
1 ,
30
o 1 0
1
300
600
900
1200
1500
Mean contact pressure (N/mm2) Figure 10. Relationship between abrasion life (in terms of numbers of revolutions) and mean contact pressure.
4
I
H PTFE(sputtered)
0 CU
Ni
Figure 1 1. Adhesion strength of metal thin films prepared by vacuum evaporation to sputtered PTFE thin film and bulk PTFE.
chromium, copper and gold thin films prepared by electron beam evaporation. The adhesion of the PTFE spin-coated film was the highest for titanium film, followed by chromium film, and the least for copper film [ l l ] .This work reports on the adhesion for metal thin films on spin-coated PTFE film, and is different from our system (adhesion of metal thin films on the sputtered PTFE thin film). However, we think the adhesion of the metal thin films shows the same tendency in both cases. Although a metal-carbon bond (carbide bond) was observed at the in-
S. Iwamori et al.
226
terface between the metal film and the PTFE thin film in the titanium and chromium samples, it was not observed in the copper sample [ 111. In addition, it is known that copper hardly binds to carbon in the polymer structure with formation of a carbide bond. Although we have not confirmed the carbide bond at the interface between the nickel film and the PTFE thin film in this study, the carbide bond would relate with the adhesion between the nickel and copper thin films and the PTFE thin films. Figure 11 also shows the adhesion strength between metal thin films prepared by vacuum evaporation and the bulk PTFE. The adhesion strength was much lower than that between metal thin films and the sputtered PTFE thin film. Bodo and Sundgren [lo] reported that any metal that formed a stable carbide bond or organometallic complex with the substrate would adhere strongly to the polymer surface. The chemical bonding state of the sputtered PTFE thin film is different from that of the bulk PTFE [9, 121. Figure 12 shows the CISXPS spectrum of the sputtered PTFE thin film [12]. Although four peaks which represent -CF3, -CF2-, -CF- and -C- moieties are observed the surface of the sputtered PTFE thin film, only one peak, which represents the -CF2- moiety, can be observed on the surface of the bulk PTFE [9, 121. These results mean that almost all carbon atoms bind to two fluorine atoms in the bulk PTFE, but one carbon atom binds to three to zero fluorine atoms in the sputtered PTFE thin film. In addition, taking into consideration that the atomic composition of fluorine for carbon (F/C) was 1.3 to 1.7 [12], there would be double (-C=C-) or triple bonds (-C=C-) in the case of sputtered PTFE thin film. These double or triple bonds can react with high energy metal at-
295
292
289
286
Binding energy (eV) Figure 12. CISXPS spectrum of the sputtered PTFE thin film [ 113.
283
280
Abrasion life and scratch durability of sputtered PTFE thin film
227
oms and produce the carbide bond (-C-M). We think this is the reason why the adhesion strength of metal thin films onto the sputtered PTFE thin film was much higher than to the bulk PTFE.
3.3. PTFE-Cu mixed thin film prepared by sputtering In order to increase the adhesion strength between the PTFE sputtered thin film and copper substrate, a poly(tetraf1uoroethylene) (PTFE)-copper (Cu) mixed thin film was introduced between the PTFE thin film and the copper substrate by RF sputtering from a PTFE-Cu mixed target (Fig. 13b), and the abrasion life of this laminate was evaluated and compared to that without introduction of the mixed film layer (Fig. 13a). Before evaluation of the durability of the laminate, the PTFE-Cu mixed thin film was characterized. Figure 14 shows the sputter rates of the PTFE, Cu and PTFE-Cu mixed thin films at various pressures. The sputter rate of the PTFE thin film decreased only slightly with increase of pressure, that of the Cu thin film decreased dramatically. This result indicates that the atomic composition of the PTFE-Cu mixed thin film would be greatly affected by the pressure during sputtering. Table 1 shows the electrical resistance of the Cu thin film sputtered at various pressures. The resistance of the Cu thin film increased dramatically with increasing pressure. The Cu thin film deposited at high pressure would be oxidized during sputtering. Although the sputter rate of the PTFE-Cu mixed thin film was expected to be between that of the PTFE and the C h thin films, actually it was not. The Cu in the PTFE-Cu mixed thin film would react with the residual oxygen remained inside the chamber, and form copper oxides during sputtering when the PTFE-Cu mixed
400nm 2mm
200nm 200nm 2mm
Figure 13. Schematic diagrams of the abrasion life and scratch durability test samples without (a) and with (b) mixed thin film as an interlayer.
S. Iwamori et al.
228
0
10
20
30
40
50
60
Pressure (mTorr) Figure 14. Relationship between the sputter rate and pressure during sputtering for PTFE thin film (PTFE), the copper thin film (Cu) and PTFE-Cu mixed thin film (PTFE-Cu). Table 1. Relationship between electrical resistance and pressure during sputtering Pressure (mTorr) Resistance (SZ)
3 0.1
5 1.6
10 >io7
target is used. The sputter rate of copper decreased dramatically with increase of pressure. The sputter rate of metal oxide thin films prepared by reactive sputtering of the metal target is known to be much lower than that of metal thin films prepared by Ar sputtering of the metal target. This low sputter rate of metal oxide thin films is caused by the transformation from the metallic state to the oxide state [13, 141. In the case of the PTFE-Cu mixed thin film, although the oxygen concentration in the chamber was not measured, the Cu target would be oxidized and transformed from the metallic state to the oxide state when it is sputtered at 10 mTorr. This is the reason why the sputtering rate of the PTFE-Cu mixed thin film showed the unexpected value. Figure 15a and 15b shows, respectively, the optical micrographs of the PTFE thin film and the PTFE-Cu mixed thin film. A large number of particles, with diameter 10-30 pm, are seen only in the PTFE-Cu mixed thin film (Fig. 15b). We suppose that these are copper particles. The PTFE-Cu mixed thin films (323 nm and 410 nm thick) were sputtered at 3 mTorr onto the glass slide and their resistivity measured. The atomic compositions of these films are shown in Table 2. The copper content in the mixed film increased with thickness. Figure 16 shows the relationship between the resistivity and thickness of the PTFE-Cu mixed thin
Abrasion life and scratch durabiliQ of sputtered PTFE thin film
229
(b)
PTFE thin film
PTFE-Cu thin film
R,=9.1 nm
R,=26.7 nm
Thickness = 0.46 p m
Thickness = 0.40 p m
Figure 15. Optical micrographs of (a) the PTFE thin film and (b) PTFE-Cu mixed thin film prepared by sputtering. The R, values are (a) 9.1 nm and (b) 26.7 nm, respectively. R, is the arithmetic average surface roughness.
Table 2. Elemental composition of the PTFE-Cu mixed thin films PTFE-Cu mixed thin film thickness (nm)
Cu
F
0
(8)
(8)
(8)
m’c)
323 410
29 48
39 21
11 11
20 20
C
n
5 10
,
0
0.2
0.4
0.6
0.8
1
Film thickness (,um) Figure 16. Relationship between electrical resistivity and thickness of the PTFE-Cu mixed film.
S. Zwanzori et al.
230
a
b
C
d
Figure 17. Optical micrographs of the abrasion tracks in PTFE thin film sputtered onto Cu substrate (PTFEKu). (a) before durability test, (b) after 5-min durability test, (c) after 7-min durability test and (d) after 12-min durability test.
film. The resistivity decreased with increase of the film thickness. Biederman [7] reported that the metal and polymer targets contaminated each other during sputtering (cross-contamination effects of the target), which resulted in a low deposition rate and re-sputtering effects at the substrate. The differences in copper content and resistivity between these two films shown in Table 2 could be caused by the cross-contamination effect. 3.4. Evaluation of durability of the PTFE-Cu mixed thin film sputtered onto the copper substrate
Figure 17 shows the optical micrographs of the PTFE thin film sputtered on the copper substrate after the abrasion test with 0.49 N load using the apparatus described above. Figure 17a-d represents the optical micrographs before the test, and 5 min, 7 min and 12 min after the test, respectively. A slight abrasion track can be seen in Fig. 17b, and the film peeled off from the substrate after the test for 12 min (Fig. 17d). Figure 18 shows the optical micrographs of the abrasion tracks in PTFE/PTFECu mixed film on the copper substrate. Figure 18a-d represents optical micrographs before the test, and 5 min, 7 min and 15 min after the test, respectively. A slight abrasion track can be seen in Fig. 17d. These results indicate that the durability improved with the introduction of the PTFE-Cu mixed layer between the PTFE thin film and the copper substrate.
Abrasion life and scratch durability of sputtered PTFE thinfilm
a
b
C
d
23 1
Figure 18. Optical micrographs of the abrasion tracks in the PTFE/PTFE-Cu/Cu system. (a) before durability test, (b) after 5-min durability test, (c) after 7-min durability test and (d) after 15-min durability test.
The PTFE-Cu mixed layer would act as a functionally gradient material layer [15] and improve the adhesion strength between the PTFE thin film and the copper substrate. The abrasion life of the polyimide thin film sputtered onto the glass substrate was ten times higher than that of the PTFE thin film onto the glass substrate [2]. Although the abrasion life of the PTFE/PTFE-Cu/Cu substrate system shown in Fig. 18 improved two times compared to that of the PTFE/Cu substrate system, it is still lower than that of the polyimide thin film. 4. CONCLUSIONS
Abrasion life and scratch durability of poly(tetrafluoroethy1ene) (PTFE) thin films sputtered onto metal substrates were evaluated. 1. The abrasion life and scratch durability of the PTFE thin film sputtered onto the copper substrate were higher than of the film sputtered onto the nickel substrate. The abrasion life was found to be related to the hardness of the metal substrate. On the other hand, we found that the abrasion life for the mean contact pressure on the sputtered PTFE thin film on the nickel substrate was higher than that on the copper substrate. The abrasion life would also relate to the adhesion of the metal substrate. 2. The adhesion strength between evaporated metal films and sputtered thin PTFE film was higher than that between evaporated metal thin films and bulk
232
S. Iwamori et al.
PTFE. This result indicates differences in the chemical bonding states between the PTFE thin film and the bulk PTFE. 3. The PTFE-Cu mixed thin film was introduced as an interlayer between the sputtered PTFE thin film and the Cu substrate (PTFE/PTFE-Cu/Cu substrate), and its tribological properties were evaluated. The abrasion life of the PTFE/PTFE-Cu/Cu substrate system was improved compared to that of the PTFE thin film directly sputtered onto the Cu substrate. The PTFE-Cu mixed layer acted as a functionally gradient material layer. REFERENCES 1. M. Kitoh and Y. Honda, Thin Solid Filrns 271,92 (1995). 2. Y. Yamada, K. Tanaka and K. Saitoh, Surf: Coating. Technol. 43-44,618 (1990). 3. G. A. Hishmeh, T. L. Barr, A. Sklyarov and S. Hardcastle, J. Vac. Sci. Technol. A14, 1330 (1996). 4. H. Biederman, Vacuum 31, 285 (1981). 5. N. Marechal and Y. Pauleau, J. Vac. Sci. Technol. A10,477 (1992). 6. H. Biederman, M. Zeuner, J. Zalman, P. Bilkova, V. Stelmasuk and A. Boldyreva, Thin Solid Films 392,208 (2001). 7. H. Biederman, Vacuum 59,594 (2000). 8. Japanese Industrial Standard JIS K7194 (1994). 9. Y. Yamada, Y. Nagayama and K. Tanaka, J. Jpn. SOC. Tribol. 38, 817 (1993). 10. P. Bodo and J. E. Sundgren, Surf: Interf: Anal. 9,437 (1986). 11. C.-A.Chang, Y.-K. Kim and G. Schrott, J. Vac. Sci. Tecltnol. A8, 3304 (1990). 12. Y. Yamada, T. Kurobe, K. Yagawa and K. Ikeda, J. Mater. Sci. Lett. 18,415 (1999). 13. T. Hata, S. Sakano, Y. Masuda, K. Sasaki, Y. Haneda and K. Wasa, Vacuum 51, 583 (1998). 14. A. Belkind and J. Wolfe, Thin Solid Films 248, 163 (1994). 15. R. Fukuda, Mater. Sci. Technol. 35, 10 (1998).