Diamond Films Handbook

ISBN: 0-8247-9577-6 This book is printed on acid-free paper. Headquarters Marcel Dekker, Inc. 270 Madison Avenue, New York, NY 10016 tel: 212-696-9000; fax: 212-685-4540 Eastern Hemisphere Distribution Marcel Dekker AG Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland tel: 41-61-261-8482; fax: 41-61-261-8896 World Wide Web http://www.dekker.com The publisher offers discounts on this book when ordered in bulk quantities. For more information, write to Special Sales/Professional Marketing at the headquarters address above. Copyright 䉷 2002 by Marcel Dekker, Inc. All Rights Reserved. Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming, and recording, or by any information storage and retrieval system, without permission in writing from the publisher. Current printing (last digit): 10 9 8 7 6 5 4 3 2 1 PRINTED IN THE UNITED STATES OF AMERICA

Preface

It has been known for over two centuries that diamond is a particular form of carbon. Since we live in a world where carbon in general is omnipresent but carbon in the form of diamond is extremely rare, there has been longstanding interest in the development of ways to convert ordinary forms of carbon to diamond. A distinctive method of diamond synthesis at low pressures, in which diamond is metastable with respect to graphite, is often referred to as the ‘‘new diamond technology’’ to distinguish it from highpressure diamond synthesis, in which diamond is the stable form of carbon. Over two decades of intensive science and engineering have gone into developing and understanding low-pressure diamond deposition. Diamond Films Handbook is a comprehensive source of information for readers interested in the details of how low-pressure diamond synthesis is achieved as well as how the diamond may be used for practical applications. In low-pressure diamond synthesis, common sources of carbon, such as methane, are transformed by chemical vapor deposition to diamond at atmospheric or subatmospheric pressures and at temperatures that are common to materials processing. We are fortunate to include in this book a chapter on the history of this amazing achievement, written by John C. Angus, a pioneer in the field. An important practical result of this synthesis approach is the ability to deposit diamond in layers, or films, over substrates of dimensions not previously possible. Currently, the capability exists to synthesize layers of diamond over surface areas as great as several hundred square centimeters. A host of applications are based on this diamond deposition capability combined with the exceptional physical properties of diiii

iv

Preface

amond. The latter include extreme hardness, high thermal conductivity, low coefficient of friction, high degree of chemical inertness, low electron work function, low infrared and microwave absorption, and wide bandgap semiconductivity. In fact, there are competing methods of achieving low-pressure diamond deposition. The three that have received the most attention in the literature are the arc jet method, the hot-filament method, and the microwave plasma method. However, other methods also show significant promise, including optical assisted deposition and combustion synthesis. This book devotes a chapter to each of these methods, describing them in significant detail. Thus, the reader may compare apparatus requirements, processing conditions, and deposition results. To provide a better understanding of the various deposition methods, a chapter that provides the basic physics and chemistry of diamond film technology, as well as a chapter that describes diamond film quality characterization techniques, precede the deposition chapters. As for applications of diamond formed by low-pressure deposition, the scientific and engineering literature has addressed far too many potential applications for each to be considered in detail in one book. As is often the case in new technologies, the path of applications development has deviated significantly from early projections. Speaking broadly, the technical and marketing difficulties in developing applications-based products have been more difficult than initially envisioned. This book includes four chapters addressing applications that provide useful information on the types of issues generally important in product development. Two chapters discuss applications for which products based on low-pressure diamond deposition are available in a broad market sense: cutting tools and thermal management. Two additional chapters discuss applications for which low-pressure diamond products are primarily in the development stage: optical applications and electrical applications. We have been working jointly on diamond deposition since 1987. Our knowledge of microwave plasma source design and plasma-aided deposition led to an invitation from the Norton Company’s Diamond Film Division to participate in the field of diamond synthesis. Over the past 15 years, it has been remarkable to participate in and observe the evolution of the new diamond technology. Progress in the field, starting with initial efforts to deposit an exotic and useful material at low rates and in small areas to the current capability to deposit films with thickness up to 1 mm and over large areas, has been dramatic. For us it has been an exciting and invigorating experience, and interesting personal relationships have been established in the process that have contributed to many lasting memories.

Preface

v

Phenomenal advances in the field have resulted from methodical worldwide collaborations among those in industry, academia, and government laboratories. What is the future going to bring? People across the world are working on this technology, and we expect it will continue its rapid pace and development and that applications will appear in areas not yet envisioned. Thus, it is our hope that this book will provide those currently working in this area, and those contemplating entering the area, with a useful source of information, reference, and education. Diamond Films Handbook is intended to serve as a valuable source of information for readers interested in low-pressure deposition of diamond films and their applications. This is made possible by virtue of the chapter authors who have contributed their time and their unique expertise. We thank them for their contributions to this book, as well as to the field in general. We also express our gratitude to the publisher Marcel Dekker, Inc., and for the editing skills of Rita Lazazzaro and Ann Pulido, who made this book possible. Professor Asmussen thanks his wife Colleen for her years of patience, support, and tolerance of his cluttered offices. Professor Reinhard thanks his family, Sharon, Sarah, Daniel, and David for their long-standing support and much valued encouragement along the way. Both editors are grateful to the students with whom it has been their good fortune to work. Jes Asmussen D. K. Reinhard

Contents

Preface Contributors

iii ix

1.

Introduction Jes Asmussen and D. K. Reinhard

2.

A Short History of Diamond Synthesis John C. Angus

3.

Film Characterization Methods: Structure and Composition Ludwig Josef Balk and Ralf Heiderhoff

27

Deposition Chemistry: Deposition Pathways, Nucleation, and Growth David S. Dandy and Michael E. Coltrin

55

4.

5.

6.

Thermally Assisted (Hot-Filament) Deposition of Diamond Joseph E. Yehoda Plasma Torch Diamond Deposition Joachim V. R. Heberlein and Naoto Ohtake

1

17

119

141 vii

viii

Contents

7.

Microwave Plasma-Assisted Diamond Film Deposition Timothy A. Grotjohn and Jes Asmussen

211

8.

Combustion Synthesis of Diamond Colin A. Wolden

303

9.

Laser-Assisted and Optical Pumping Techniques for Diamond Synthesis Vish V. Subramaniam and Shashi M. Aithal

325

CVD Diamond Solutions for Machining and Other Mechanical Applications Brian L. Cline and James M. Olson

425

10.

11.

Diamond Heat Spreaders and Thermal Management Ajay P. Malshe and W. D. Brown

511

12.

Diamond Active Electronic Devices Hiromu Shiomi

579

13.

Diamond Film Optics D. K. Reinhard

607

Index

665

Contributors

Shashi M. Aithal* Center for Advanced Plasma Engineering and NonEquilibrium Thermodynamics Laboratories, Department of Mechanical Engineering, The Ohio State University, Columbus, Ohio John C. Angus Department of Chemical Engineering, Case Western Reserve University, Cleveland, Ohio Jes Asmussen Electrical and Computer Engineering Department, Michigan State University, East Lansing, Michigan Ludwig Josef Balk Department of Electronics, University of Wuppertal, Wuppertal, Germany W. D. Brown High Density Electronics Center and Department of Electrical Engineering, University of Arkansas, Fayetteville, Arkansas Brian L. Cline

Cline Innovations, LLC, Sterling, Massachusetts

Michael E. Coltrin Chemical Processing Sciences Department, Sandia National Laboratories, Albuquerque, New Mexico *Current affiliation: Novellus Systems, Inc., San Jose, California.

ix

x

Contributors

David S. Dandy Department of Chemical Engineering, Colorado State University, Fort Collins, Colorado Timothy A. Grotjohn Electrical and Computer Engineering Department, Michigan State University, East Lansing, Michigan Joachim V. R. Heberlein Department of Mechanical Engineering, University of Minnesota, Minneapolis, Minnesota Ralf Heiderhoff Department of Electronics, University of Wuppertal, Wuppertal, Germany Ajay P. Malshe High Density Electronics Center and Department of Mechanical Engineering, University of Arkansas, Fayetteville, Arkansas Naoto Ohtake Department of Mechanical Sciences and Engineering, Tokyo Institute of Technology, Tokyo, Japan James M. Olson Process Technology Development, Fairchild Semiconductor, South Portland, Maine D. K. Reinhard Electrical and Computer Engineering Department, Michigan State University, East Lansing, Michigan Hiromu Shiomi

Sumitomo Electric Industries Ltd., Itami, Japan

Vish V. Subramaniam Center for Advanced Plasma Engineering and NonEquilibrium Thermodynamics Laboratories, Department of Mechanical Engineering and Chemical Physics Program, The Ohio State University, Columbus, Ohio Colin A. Wolden Department of Chemical Engineering, Colorado School of Mines, Golden, Colorado Joseph E. Yehoda

Diamonex, Inc., Allentown, Pennsylvania

1 Introduction Jes Asmussen and D. K. Reinhard Michigan State University, East Lansing, Michigan

I.

DIAMOND FILM TECHNOLOGY—AN OVERVIEW

Advances in diamond science and technology represent a remarkable achievement in materials research and development. Diamond is, of course, not a new material. Many of the unique properties of naturally occurring diamond have been appreciated since time immemorial. Neither is the synthesis of diamond particularly new: reliable methods of forming diamond from carbon-containing compounds were reported in the middle of the twentieth century. What is remarkable is the current capability of synthesizing diamond structures and layers of dimensions not previously possible, with deposition areas up to several hundred square centimeters and direct deposition on a wide variety of substrates. This achievement has been made possible both by an improved scientific understanding of how diamond is formed and by a significant engineering development of chemical vapor deposition (CVD) systems designed specifically for the deposition of diamond. During recent years, it has become widely recognized that polycrystalline and homoepitaxial diamond can be deposited using a variety of CVD techniques. These new and relatively simple synthesis methods led to a worldwide research effort devoted to improved synthesis, to understanding the properties of the CVD films, and to applications of diamond. Figure 1 shows a simplified phase diagram of carbon. From this it may be seen that diamond is the minimum energy state, and therefore the stable form, of carbon at high pressures and temperatures, that is, at pressures of thousands of atmospheres and temperatures of thousands of degrees Celsius. Graphite, however, is the stable form of carbon under ordinary temperature 1

2

Asmussen and Reinhard

Figure 1 Phase diagram for carbon. This simplified phase diagram for carbon shows that at sufficiently high temperatures and pressures diamond is the stable, or minimum energy state, of carbon atoms. At lower temperatures and pressures, graphite is the stable form. Under ordinary conditions for temperature and pressure, near 1 atm and room temperature, diamond may be considered a metastable form of carbon. Even though it is not the minimum energy state, it does not spontaneously convert to graphite.

and pressure conditions. One method of synthesizing diamond is to subject graphite to conditions of approximately 55,000 atmospheres and temperatures of about 2000⬚C. The first group to report diamond synthesis under these conditions publicly was the General Electric group including Bundy, Hall, Strong, and Wentorf in 1955 [1]. Since then, a variety of technical

Introduction

3

approaches to achieve diamond growth under such conditions have been reported, and they are known collectively as high-temperature, high-pressure diamond synthesis. High-temperature, high pressure processes have been proved to be commercially successful in terms of producing diamond powder and grit useful in various industrial applications.* However, these processes are not amenable to direct deposition of diamond on substrates or to large-area deposition and are not the subject of this book. Instead, the focus of the book is on diamond formed under conditions of pressures generally less than an atmosphere and temperatures less than 1000⬚C. This approach to diamond synthesis is sometimes known as the new diamond technology. At ordinary temperatures and pressures, although diamond is not the minimum energy state of carbon, it is also not an unstable stage of carbon. In other words, once carbon atoms are in the diamond lattice spatial arrangement, the solid does not spontaneously convert to graphite under low-temperature, low-pressure conditions. Such nonminimum energy arrangements are known as metastable states. Formation of diamond from nascent carboncontaining species under metastable conditions is both thermodynamically allowed and readily achieved under proper deposition conditions. It is the lower temperatures and pressures associated with this method of diamond synthesis that have offered the capability of direct deposition of diamond on a variety of substrates and have opened the possibility of new applications of diamond. For many such applications, the diamond thickness need be only on the order of micrometers; hence the structures are referred to as diamond films. The applications interests in diamond are based on the exceptional properties of diamond including extreme hardness, high thermal conductivity, high electrical resistivity, low coefficient of friction, high degree of chemical inertness, low electron work function, large energy gap, low infrared absorption, and high breakdown voltage. In many cases, the properties of diamond are superlative. For example, it is reported to have the highest hardness of any material, the highest thermal conductivity, and the lowest compressibility. In other cases, the diamond material property is not necessarily the best, but diamond is still highly competitive with other materials. For example, the coefficient of friction is comparable to that of Teflon. In any case, diamond is properly considered to be a unique material because it exhibits an unusual constellation of highly attractive properties of interest for many applications.

*High-temperature, high-pressure methods have also produced gem quality diamond, apparently not under commercially viable conditions.

4

II.

Asmussen and Reinhard

METHODS OF DIAMOND FILM DEPOSITION UNDER METASTABLE CONDITIONS

A history of the development of diamond deposition methods under metastable conditions is provided by Angus in Chap. 2, and a detailed description of the key chemical and physics concepts is presented by Dandy and Coltrin in Chap. 4. The appropriate conditions for achieving low-pressure diamond synthesis can be obtained by a variety of methods. Generally, as shown in Fig. 2, these methods have certain things in common, including gaseous feedstock, a means of providing energy (thermal, electrical, combustion, or optical) that causes the gas components to dissociate, and a substrate on which the deposition takes place [2]. The gas feedstock includes at least one source of carbon-containing species. Most commonly, this is methane; however, a variety of other carbon-containing gas sources have been used. The energy provided to the system breaks down the input gas into a variety of subspecies such as, in the case of methane, CH3 and other molecules and atoms. It is the carboncontaining subspecies that are involved in donating a carbon atom to the growing diamond film on the substrate. In addition to the carbon-containing gas, most methods use hydrogen gas as a feedstock input. Often, for example, methane is diluted to between 0.5 and 5% in hydrogen. The energy provided to the system also breaks down the hydrogen gas, producing large amounts of atomic hydrogen. Atomic hydrogen plays an important role in that it helps maintain the diamond tetrahedral sp3 atomic configuration in the growing film as opposed to the sp2 configuration, which leads to graphite.

Figure 2 Generic deposition system for metastable synthesis of diamond. There are many types of diamond deposition systems that are capable of diamond synthesis under metastable conditions; however, they share generic features. All require a feedstock that contains carbon. The feedstock is usually gaseous and generally contains hydrogen as well as the carbon source. Energy is provided to the system in order to dissociate the molecules of the input gases. A variety of energy sources have been used, including thermal, electrical, optical, and combustion. A substrate is provided on which the deposition takes place.

Introduction

5

A variety of means have been developed to provide energy to the deposition system sufficient to break down the input gases. One of the most straightforward is the hot-filament method. In this method, the gases flow through a reactor that contains a wire or mesh filament apparatus typically made of tungsten, which is heated to a temperature sufficient to decompose significant fractions of the hydrogen and methane gases. Yehoda in Chap. 5 describes the hot-filament method in detail. Alternatively, a variety of approaches use electrical energy inputs to the deposition system in the form of dc, rf, or microwaves. Among the more successful of these are the arcjet method, described by Heberlein and Ohtake in Chap. 6, and microwave plasma–assisted CVD described by Grotjohn and Asmussen in Chap. 7. Combustion processes can also be used to provide the necessary species. Originally this approach used essentially a welding torch apparatus; however, the concept has also been applied to more sophisticated systems as described in Chap. 8 by Wolden. Finally, optical energy may be used, and Subramaniam and Aithal describe laser-assisted and optically pumped methods in Chap. 9. All of these methods have proved to be effective at depositing diamond under metastable conditions, although each has advantages and disadvantages as discussed in more detail in the cited chapters. The cost per carat of producing CVD diamond has decreased in recent years with improved system development. However, the capital cost associated with a production-scale deposition system is still a serious concern for all the various approaches. The associated economic considerations can present a barrier for the introduction of CVD diamond products. Therefore, deposition system development and improvement is a vital aspect of the future of CVD diamond applications.

III.

MATERIAL PROPERTIES OF DIAMOND

A.

General Properties

Pure diamond is composed only of the element carbon. The carbon atoms are arranged in the crystal structure known as the diamond lattice, in which each carbon atom has four nearest neighbors in the tetrahedral arrangement associated with sp3 chemical bonds. The nearest neighbor distance is 1.54 ˚ and the unit cell dimension is 3.567 A ˚ at 298 K. There are 8 atoms per A unit cell and 1.77 ⫻ 1023 atoms/cm3, which is the highest atomic density of any terrestrial material. The density at 298 K is 3.515 g/cm3. Diamond quantity is often expressed in carats, where one carat is equal to 200 mg, or about 57 mm3. The quality of diamond specimens varies considerably because both natural and synthetic diamond may contain impurities and defects. Natural diamond is classified into type I and type II, with the former

6

Asmussen and Reinhard

containing nitrogen as an impurity and the latter being essentially nitrogen free. Further information about diamond classification is contained in Chap. 13. As previously discussed, diamond is a metastable form of carbon at ordinary temperatures and pressures, whereas graphite is the stable phase. With sufficient perturbation, the metastable phase gives way to the stable phase. For example, if diamonds are heated in an inert atmosphere, the onset of graphitization is detected at 1800 K. The rate of graphitization increases such that at 2400 K a 0.1 carat diamond is completely reduced to graphite in less than 3 min [3]. Diamond is highly inert chemically, except for two situations. It is susceptible to oxidizing agents at high temperatures. For example, if diamond is heated in the presence of oxygen, oxidation begins at around 900 K. Also, diamond is subject to chemical attack by certain metals at high temperatures. These include carbide formers such as tungsten, tantalum, titanium, and zirconium as well as solvents for carbon such as iron, cobalt, manganese, nickel, chromium, and platinum. Thorough reviews of diamond properties are contained in the literature [4,5]. Several single-crystal diamond properties of particular interest for applications are briefly reviewed in the following sections. Later chapters cover many of these properties in more detail and relate them to deposition conditions and to particular applications. B.

Mechanical Properties

Diamond is the hardest known substance. Diamond also has the lowest compressibility, the highest elastic modulus, and the highest isotropic speed of sound (18,000 m/sec) of any known material. The degree of hardness is quantified in terms of both resistance to indentation and abrasion (or scratch) resistance. One indentation method uses a Vickers indenter with a square pyramidal diamond point. Another hardness test uses a Knoop indenter in which a pyramidal point with long and short axes in the ratio of 7:1 is used. Depending on the crystallographic orientation of the sample, Knoop indentation hardness values are reported to be in the range of 5700–10,400 kg/ mm2 where the values are determined on the basis of the indenter load and the dimensions of the indentation. As a comparison, the value for casehardened steel is approximately 400 kg/mm2. When applying the hardness concept in terms of scratch and abrasion resistance, the scratch hardness Mohs hardness value is 10, the largest of any material. Cline and Olson describe tests related to abrasion resistance of diamond under various conditions in detail in Chap. 10. In terms of compressibility, the ratio of tensile stress to linear strain, or Young’s modulus, is 1050 GPa, a value approximately five times higher

Introduction

7

than that of steel. Because of its brittle nature, diamond is not particularly strong; however, the cleavage strength of diamond is reported to be at least as great as the tensile strength of tungsten wire, approximately 400 kg/ mm2 [4]. C.

Thermal Properties

The thermal conductivity of pure diamond exceeds that of all solids in the range of 90 to 1200 K and is to an extent limited by isotropic purity because the naturally occurring 13C isotope acts as a phonon scatterer. At room temperature, the experimental value for the thermal conductivity of high-purity, natural single crystalline diamond is approximately 20 W/cm-K, which is about five times higher than the thermal conductivity of copper. Diamond’s thermal conductivity increases with decreasing temperature, reaching a maximum of 42 W/cm-K near 80 K, below which the thermal conductivity decreases. Impurities, such as nitrogen, reduce the thermal conductivity. Type I diamonds with 0.1% nitrogen have room temperature thermal conductivity values approximately 50% that of type II diamonds, which have a negligible nitrogen concentration. Isotropic purity increases the thermal conductivity. Synthetic diamond crystals grown with pure carbon-12 have thermal conductivities 50% higher than those of natural diamond for which the atomic weight is 12.01 because the material contains 1.1% carbon-13. Malshe and Brown discuss various practical issues associated with diamond’s thermal conductivity in Chap. 11. Diamond has the lowest specific heat of any solid in the temperature range between 0 and 800 K. At 300 K the value is 6.195 J/mol-K. The thermal expansion of diamond is also low. In the range of 300 to 1200 K, it increases from 0.8 ⫻ 10⫺6 to 4.8 ⫻ 10⫺6 with increasing temperature. D.

Electrical Properties

The electrical conductivity of diamond is even more sensitive to impurities than the thermal conductivity. The energy band structure of diamond exhibits an indirect energy gap with a value of 5.47 eV at 300 K. This is sufficiently large that at temperatures near room temperature the intrinsic carrier concentration is negligible and the material is an insulator with a dielectric constant of approximately 5.7. In practice, the resistivity of highly pure diamond is in the range of 1014 –1016 ⍀-cm. However, it is also possible to dope diamond substitutionally with boron. This introduces holes in the valence band such that the electrical conductivity drops dramatically with increasing boron concentration. Resistivities as low as 0.1 ⍀-cm can be achieved by boron doping. Consequently, diamond can also take the form

8

Asmussen and Reinhard

of a p-type semiconductor. The Hall mobility associated with holes in diamond is 1600 cm2/v-sec, a value that is considerably higher than the hole mobility of either silicon or gallium arsenide. The literature also contains reports of n-type doping of diamond; however, the consistent achievement of n-type diamond appears at this time to be highly problematic. Shiomi gives further information about diamond electronic properties in Chap. 12. E.

Optical Properties

It has been argued that pure diamond has the largest optical transmission bandwidth of any solid material. It extends from an ultraviolet (UV) cutoff of approximately 225 nm (corresponding to band-to-band excitation) through the far infrared and into the microwave range and beyond. There is slight absorption between 2.5 and 6 ␮m due to phonon excitation, with the absorption coefficient value peaking at about 12 cm⫺1. Throughout the entire visible region, pure diamond has essentially no optical absorption and is clear. Impurities can appreciably alter the optical transmission, however. As described by Reinhard in Chap. 13, impurities of various types can impart a wide range of colors to diamond and can even cause diamond to be opaque. Diamond has a large refractive index for a transparent material, approximately 2.4, and therefore a large reflection coefficient and a small angle for total internal reflection. Diamond is also photoconductive. There is a strong photoconductive peak at 225 nm due to excitation of electrons across the band gap in pure diamond, and in boron-doped diamond there are also peaks from 1.4 to 3.5 ␮m due to excitation of the deep-lying acceptor levels. F.

Polycrystalline Considerations

When CVD diamond is deposited on nondiamond substrates, the resulting film is polycrystalline. The question arises as to how similar the properties of polycrystalline CVD diamond are to those of single-crystal diamond. The answer is that for the most part, with some notable exceptions, it is possible to grow polycrystalline diamond with properties essentially identical to those associated with single-crystal diamond. One significant exception is that of fracture strength, which may be measured by loading diamond plates to the point of failure. Freestanding plates of CVD polycrystalline diamond are reported to have a fracture strength about an order of magnitude less than that of single-crystal diamond of the same thickness. The fracture paths in CVD diamond samples appear to be a mixture of transgranular and intergranular with no preferred direction, suggesting that the grain boundaries are not inherently weak [6]. The fracture origins appear to be at flaws [7]. A second significant difference between even the best polycrystalline dia-

Introduction

9

mond and single-crystal diamond is that of carrier mobility. Hall effect hole mobilities in p-type CVD diamond are an order of magnitude or more less than single-crystal values. Finally, polycrystalline CVD diamond samples do not show the well-defined cleavage planes of single-crystal diamond and so are not useful for traditional gem applications. Most other material properties of CVD polycrystalline diamond can be quite similar to those of single-crystal diamond—provided care is taken to achieve the desired effect. As a case in point, consider optical transmission. Raytheon, for example, has reported 350-␮m-thick CVD polycrystalline samples with optical transmission essentially identical to that of ideal diamond [6]. That is, except for the onset of band gap absorption at 225 nm and multiphonon absorption in the infrared, the sample is optically clear. However, the same cannot be said for most CVD diamond samples, which often show appreciable optical absorption and scattering, to the point of appearing to the eye to be gray or even black. The difference lies in deposition variables such as the chemical composition of the feed gases, system purity, and also, in this particular case, postprocessing. The sample surface was polished to reduce optical scattering effects. In terms of thermal conductivity, the best CVD diamond is reported to have the same value as single-crystal diamond. However, if CVD diamond is grown with more emphasis on growth rate than quality, such high values are not to be expected. Many nonoptimized CVD diamond samples have thermal conductivities that are one quarter to one half the value of single-crystal diamond. Even so, the thermal conductivity is still significantly higher than that of other high-thermal-conductivity electrical insulators such as aluminum nitride or beryllium oxide. In terms of mechanical properties, experimental studies of Young’s modulus for CVD polycrystalline diamond have resulted in values close to that of single-crystal diamond, that is, about 1000 GPa. In summary, for many (but not all) applications the properties of polycrystalline CVD diamond are sufficiently close to those of single-crystal diamond. However, not all CVD diamond is of the same quality. Therefore it is important to assess diamond quality. There are a variety of means of doing this, as reviewed by Balk and Heiderhoff in Chap. 3.

IV.

CVD DIAMOND APPLICATIONS

A.

Applications Overview

Diamond’s combination of attractive material properties makes it of interest for a large number of applications. Some of these are listed as follows. Cutting tool inserts Coated cutting and drilling tools

10

Asmussen and Reinhard

Rotating seals and wear surfaces Laser heat sinks Electronic packaging Electronic passivation layers Optical windows Optical coatings X-ray lithography Gyrotron windows Cold cathode electron emitters Photon emitters Acoustic speakers Active electronic devices Thermistors Pressure sensors Surface acoustic wave devices Nuclear particle detectors UV detectors Lasers Electrodes for electrochemistry Several of these applications are considered in detail in Chaps. 10, 11, 12, and 13, namely applications related to cutting tools, thermal management, electronics, and optics. The purpose of this section is to provide a brief overview of the broad spectrum of present and potential applications. B.

Cutting Tools

Among the early successful CVD polycrystalline diamond products have been those involved in cutting tools of various sorts [8]. One approach is to grow relatively thick layers of CVD diamond from which separate freestanding pieces are obtained. These pieces are then brazed onto a cutting tool. Alternatively, the diamond may be deposited directly onto the cutting tool. Both approaches have proved to be effective in cutting most nonferrous materials including aluminum, brass, bronze, ceramics, graphite, and glass fiber–reinforced structures. Diamond end mill, drill, and router products are available. A thorough treatment of this application may be found in Chap. 10. C.

Wear Surfaces

The abrasion resistance of diamond combined with the low coefficient of friction makes diamond of interest for wear surfaces such as bearings and

Introduction

11

seals. For this application, the diamond surface should be smooth. Generally, polycrystalline diamond film surfaces are not smooth in the as-grown state. Potential solutions are postprocessing, in the form of polishing; growth with highly oriented grains such that a relatively smooth surface results; and growth with extremely small crystallites, sometimes referred to as nanocrystalline films. This is a potentially significant market for which product development is still needed. D.

Thermal Management

The high thermal conductivity of diamond, combined in some cases with its chemical inertness and high electrical resistivity, makes it of interest for a variety of thermal management applications. Laser diamond heat sinks and other thermal management substrates formed from CVD polycrystalline diamond are examples of available products. Because diamond combines exceptionally high thermal conductivity with exceptionally low electrical conductivity, it is of considerable interest in electrical packaging applications. It provides efficient paths for heat flow without compromising the electrical isolation of individual components. This application area is covered in depth in Chap. 11. E.

Transmission Applications

Diamond is of interest for several applications in which it provides a window with high transmittance for various portions of the electromagnetic spectrum. In the x-ray portion of the spectrum, diamond is of interest for x-ray lithography masks. The low atomic number of diamond results in low x-ray absorption. Another example is in high-power gyrotrons such as are used in fusion research. This application requires the transmission of very large powers (megawatts) at microwave frequencies (170 GHz) as well as the ability to dissipate heat rapidly. The ability to transmit high powers in the optical portion of the spectrum is of interest to laser designers because the design of high-power lasers is power limited by damage limits to laser optics rather than limitations of the laser medium or pumping mechanisms. The scratch resistance and chemical inertness make diamond of interest as an optical coating material as well. Optical applications of diamond are covered in Chap. 13. F.

Electron and Photon Sources

Diamond exhibits a low, and sometimes negative, electron affinity. A negative electron affinity indicates that the conduction band minimum lies above

12

Asmussen and Reinhard

that of the vacuum level at the surface. Any electrons accumulating in the conduction band are thus readily available for emission into the vacuum. Therefore, diamond is of interest as an electron source and particularly as a cold-cathode electron emitter. Such field emitters are important for the development of field emission flat panel displays and vacuum microelectronic devices. Diamond has also been reported to be a photon source and, in fact, laser action has been reported based on optically active color centers [9]. Also, diamond-based diodes have demonstrated light emission upon forward bias. Further information about diamond-based photon sources is contained in Chap. 13.

G.

Acoustic Applications

High-quality loudspeakers coated with CVD diamond are marketed that show significantly improved performance at high frequencies. One version of the product consists of two layers of aluminum separated by a honeycomb structure. One of the aluminum layers is coated with diamond, giving the unit greater rigidity, which translates into improved high-frequency response. Surface acoustic wave (SAW) devices rely on interdigitated electrodes to launch elastic waves within a solid with frequencies in the MHz to GHz range. Such devices are used in the broadcast industry. Careful choice of the electrode periodicity and the material properties allows band-pass filters to be made. High-frequency operation requires very fine electrode spacing. Because acoustic waves traveling in diamond exhibit extremely high velocities, diamond offers the prospect of high-frequency (>2 GHz) operation without the need for technologically challenging electrode spacing [10].

H.

Detectors and Sensors

Diamond-based devices are also of interest for detecting a variety of radiation types as well as sensing various physical parameters such as temperature and pressure. For example, diamond thermistors have been proposed for temperature measurement in hostile environments such as chemical processing, gearbox oil, and cryogenics [11]. The piezoresistive effect of diamond has been used to sense pressure, and p-type CVD polycrystalline diamond is reported to have a large piezoresistive gauge factor [12]. Diamond is extremely radiation hard, with a 55-eV displacement energy for a carbon atom in the diamond lattice. It also acts as an ionizing radiation detector and is therefore of interest for radiation measurements where exposure to large doses is required. The large band gap of diamond make it of interest as a UV detector, based on photoconductivity, that is blind to visible light.

Introduction

I.

13

Diamond Electronics

Diamond diodes and diamond transistors have been fabricated on polycrystalline CVD diamond. In principle, diamond offers intriguing possibilities for such devices. Diamond has a large electron saturation velocity and a high breakdown field in comparison with both silicon and gallium arsenide. At electric fields above approximately 105 V/cm, the velocity of conduction band electrons in both silicon and gallium arsenide saturates at about 107 cm/sec. The saturation velocity in single-crystal diamond is over twice this high. Higher carrier velocities translate to faster switching speeds and higher operating frequencies. Also, the large band gap of diamond means that even at elevated temperatures, the intrinsic carrier concentration is negligible. In addition, the large thermal conductivity of diamond means heat generated by electronic devices fabricated in and on a diamond substrate would be rapidly dissipated. Together, these properties indicate a potential for diamond in high-speed, high-temperature, high-power, high-voltage applications. In fact, impressive results have been reported for polycrystalline diamond electronic devices with diodes operating at temperatures of 1000⬚C [13] and field effect transistors with a breakdown voltage above 200 V [14]. However the full potential of diamond electronics has not been realized because of the polycrystalline nature of the films. Therefore a significant challenge related to this area is the growth of single-crystal films on nondiamond substrates with significant areas. Chapter 12 discusses the area of active diamond electronics in detail. J.

Electrochemical Applications

The electrochemical behavior of boron-doped CVD diamond has proved to be of interest for electrochemistry applications. Compared with platinum electrodes, diamond electrodes give a much wider potential range over which no significant water decomposition occurs. Therefore, diamond electrodes are suitable substrates for reactions spanning a wide potential range in aqueous solutions [15]. They also have the advantage of chemical stability, even in highly aggressive environments.

V.

SUMMARY

The ability to deposit diamond directly over large areas on a variety of substrates has resulted in a variety of interesting current and potential applications. In most, but not all, cases, the quality of the diamond is sufficient for the intended application. However, market development has been slower

14

Asmussen and Reinhard

than many participants had expected. In many cases this is based on economic considerations. Customers for a new product are not easily convinced to commit to significant purchases unless the economic advantages are clear and significant. Therefore, continued reduction in the cost of CVD diamond production is important to the future of the field. In other cases, such as electronic and optical applications, additional product quality improvement is needed. Thus, improvements in deposition methodology and film quality also remain a critical area for future work.

REFERENCES 1. 2. 3. 4. 5. 6.

7.

8.

9. 10.

11.

12. 13.

FP Bundy, HT Hall, HM Strong, RH Wentorf. Nature 176:51–55, 1955. BV Derjaguin, DB Fedoseev. The synthesis of diamond at low pressure. Sci Am 233:102–109, 1975. Properties of Diamond. De Beers Industrial Diamond Division. RM Chrenko, HM Strong. Physical Properties of Diamond. General Electric Technical Information Series, Report No. 75CRD089, October 1975. ER Berman. Physical Properties of Diamond. Oxford: Clarendon Press, 1965. TJ Valentine, AJ Whitehead, RS Sussmann, CJH Wort, GA Scarsbrook. Mechanical properties of bulk polycrystalline CVD diamond. Diamond Relat Mater 3:1168–1172, 1994. DC Harris. Development of Chemical-Vapor Deposited Diamond for Infrared Optical Applications. Status Report and Summary of Properties. Naval Air Warfare Center Weapons Division, China Lake, CA, Report NAWCWPNS TP 8210, 1994. K Bigelow. Progress in the commercialization of CVD diamond. In: M Yoshikawa, M Murakaw, Y Tzeng, WA Yarbrough, eds. Second International Conference on the Applications of Diamond Films and Related Materials, MYU, Tokyo, 1993, pp 5–12. S Rand, L Deshazer. Visible color-center laser in diamond. Opt Lett 10:481– 483, 1985. MD Whitfield, B Audic, CM Flannery, LP Kehoe, GM Crean, C Johnston, PR Chalker, RB Jackman. Polycrystalline diamond films for acoustic wave devices. Diamond Relat Mater 7:533–539, 1998. BL Jones. Novel and electronic applications of diamond materials. In: M Yoshikawa, M Murakawa, Y Tzeng, WA Yarbrough, eds. Second International Conference on the Applications of Diamond Films and Related Materials, MYU, Tokyo, 1993, pp 21–28. M Aslam, I Taher, A Masood, MA Tamor, TJ Potter. Piezoresistivity in vapor deposited diamond films. Appl Phys Lett 60:2923–2925, 1992. A Vescan, I Daumiller, P Gluche, W Ebert, E Kohn. Very high temperature operation of diamond Schottky diode. IEEE Electron Device Lett 18:536–538, 1997.

Introduction 14.

15.

15

P Gluche, A Aleksov, A Vescan, W Ebert, E Kohn. Diamond surface-channel FET structure with 200 V breakdown voltage. IEEE Electron Device Lett 18: 547–549, 1997. H Martin, A Argoitia, JC Angus, AB Anderson, U Landau. Boron-doped diamond electrodes for electrochemical applications. In: A Feldman, Y Tzeng, WA Yarbrough, M Yoshikawa, M Murakawa, eds. Applications of Diamond films and Related Materials, 3rd International Conference, 1995, pp 91–94.

2 A Short History of Diamond Synthesis John C. Angus Case Western Reserve University, Cleveland, Ohio

I.

INTRODUCTION

In this chapter a brief discussion of the events leading to the successful synthesis of diamond at high pressures is given, followed by a somewhat more detailed description of the development of methods of growing diamond by chemical vapor deposition. An excellent source of information on many aspects of diamond has been given by Davies [1]. An important review is by Hazen [2]. Bundy [3], a member of the original General Electric diamond group, has presented an extremely informative account of the highpressure, high-temperature reactions of carbon to form diamond. DeVries [4] and Angus [5] have summarized developments in the low-pressure, chemical vapor deposition of diamond; Amato [6] has given an interesting popular account of the field.

II.

SCIENTIFIC UNDERSTANDING OF DIAMOND

Two fundamental facts about diamond had to be understood before serious attempts at diamond synthesis could be undertaken. The first was to prove that diamond was composed entirely of the element carbon. The second was to determine that diamond was denser than common forms of carbon, such as charcoal. These two empirical facts led to numerous attempts to convert less dense forms of carbon into diamond by compression at high pressures. Foremost among these early workers were Hannay, Moissan, and Parsons. It is now believed that none of these attempts was successful. 17

18

Angus

The very first scientific study of diamond appears to have been done by two Italian scientists, Averani and Targioni [7], at the Accademio del Cimento in Florence. They studied diamond combustion in 1694, but they did not trap or identify the product gases. In 1772, Lavosier showed that the combustion products obtained from diamond and charcoal were similar. In 1797, Tennant did further combustion experiments and concluded that diamond ‘‘consists entirely of charcoal, differing from the usual state of that substance only by its crystallized form.’’ Because diamond was known to be denser than charcoal, compression appeared to be a reasonable route to achieve diamond synthesis. It was not, however, until an understanding of chemical thermodynamics was achieved that high-pressure synthesis was approached from a truly scientific viewpoint. The first accurate determination of a point on the diamond-graphite equilibrium line was made by the American chemists Lewis and Randall [8]. They showed that, at room temperature, diamond and graphite are in equilibrium at approximately 10,000 atmospheres (⬃1 GPa). Subsequent calculations of the diamond-graphite equilibrium curve up to approximately 2000 K were performed by Simon [9], Liepunski [10], Rossini and Jessup [11], and others. A sketch of the phase diagram of carbon is shown in Figure 1. III.

HIGH-PRESSURE SYNTHESIS

The thermodynamic calculation of the diamond-graphite stability line meant that serious, scientifically based efforts to achieve diamond synthesis at high pressures could now be made. Professor Percy Bridgman at Harvard University was a pioneer in these efforts. However, the first high-pressure synthesis of diamond was achieved by Liljeblad and Lundblad at the Swedish firm Allemanna Svenska Elektrisia Aktiebolaget (ASEA). They inexplicably did not announce their success, and the first public announcement of successful diamond synthesis was made by the General Electric Corporation in 1955. Synthesis was achieved by dissolving graphite, or other nondiamond carbons, in molten transition metals and precipitating the carbon out as diamond. This was done in the diamond-stable region of the phase diagram at temperatures in the range 1700 to 2000 K and pressures from 7 to 10 GPa. The history of these developments has been ably documented by others [1–3]. IV.

DIAMOND SYNTHESIS AT METASTABLE CONDITIONS

A.

Earliest Work

Diamond synthesis at low pressures, where diamond is metastable with respect to graphite, received serious attention, even after the diamond-graphite

Short History of Diamond Synthesis

19

Figure 1 Sketch of carbon phase diagram. Regions of metastability of diamond and graphite are bounded by (dashed line) extensions of the melting curves of diamond and graphite, respectively. Approximate regions for high-pressure, high-temperature (HPHT) and chemical vapor deposition (CVD) synthesis of diamond are shown.

equilibrium line was established. The region of diamond metastability is most easily envisioned by extension of the diamond melting curve into the graphite-stable region (see Fig. 1). This extension defines a region where diamond will not transform into liquid carbon. In principle, diamond can be synthesized anywhere within this region of metastability if graphite formation can be suppressed. It is of great interest that only a small fraction of this region has been explored, even though many different approaches have been attempted. The following account of diamond synthesis in the region of diamond metastability is abstracted from Ref. 5. The earliest relevant report appears to be that of Meincke [12]. He reported a series of experiments with carbon arcs in which diamond was formed. Surprisingly, this work, which was reported in the open technical literature, received little attention, either at the time of publication or in recent years. It is also remarkable that Meincke sent some of his samples to William Eversole for identification [13]. William G. Eversole, at the Linde Division of the Union Carbide Corporation in the United States, presented the first documented report of dia-

20

Angus

mond growth at low pressures that was subsequently confirmed by others. Eversole first proposed growing diamond at low pressures in a series of corporate memoranda dated September 9, 1949, October 6, 1949, and October 20, 1949 [14]. In these memoranda he proposed that carbon monoxide be used as a source gas to precipitate diamond on a diamond seed crystal. Growth of diamond seed crystals was achieved in the period November 26, 1952 to January 7, 1953 [14]. Conclusive proof and repetition of the experiments took place from February 17, 1953 to October 15, 1953. This predates the successful diamond synthesis at high pressure by the General Electric Company, which was accomplished in December 1954 and announced publicly in 1955 [15]. It just predates February 15, 1953, indicated by Liander as the date for the first synthesis of diamond at ASEA in Sweden [16]. The record indicates that Eversole was the first to create new diamond by any method. It should be noted, however, that Eversole grew new diamond on preexisting diamond seeds; the General Electric and ASEA synthesis started with nondiamond carbons. The General Electric Company mounted a serious effort to grow diamond at low pressures from 1951 through 1956. This work did not result in successful diamond growth, despite the exploration of numerous methods. These experiments are summarized in an internal General Electric report [17]. The General Electric Group tried several methods for depositing diamond on diamond seed crystals, including deposition from solid iron-carbon alloys, deposition from a column of liquid metal whose carbon supersaturation was controlled by an imposed temperature gradient, electrodeposition from high-temperature carbon-containing electrolytes, and vapor deposition from carbon vapor and from CO, CH4, CCl4, and CBr4. During this era there was much doubt that diamond could be grown at all under metastable conditions. Neuhas [18], for example, used theoretical arguments to show that diamond would not nucleate under metastable conditions. It was accepted without question by many that diamond could be grown only at high pressures where it is the thermodynamically stable phase. A notable exception was Bridgman [19], who said We know from the thermodynamic potential that graphite is ordinarily the preferred form, but this does not enable us to say that the actual precipitate will be graphite and not diamond. As a matter of fact there are many known instances in which an element’s unstable form, corresponding to diamond, separates from a solidifying liquid or solution in preference to the stable form.

In other words, metastable phases can form from precursors with high free energy if the activation barriers to more stable phases are sufficiently high. As the precursors fall in energy they can be trapped in a metastable

Short History of Diamond Synthesis

21

configuration. Formation of a metastable phase depends on selecting conditions under which rates of competing processes to undesired products are low. The processes competing with diamond growth are spontaneous graphitization of the diamond surface and nucleation and growth of graphitic deposits. The critical role of hydrogen in permitting metastable diamond growth was recognized by some early workers. The low-energy electron diffraction (LEED) study at the Bell Laboratories by Lander and Morrison [20] is the most significant. It was shown that a (111) diamond surface saturated with hydrogen gave an unreconstructed (1 ⫻ 1) LEED pattern. The dangling bonds normal to the surface are terminated with hydrogen atoms, which maintain the bulk-terminated diamond lattice to the outermost surface layer of carbon atoms. When hydrogen is absent, the surface reconstructs into more complex structures. Lander and Morrison [20] also noted that there was a kinetic barrier to graphitization of a hydrogen-covered diamond surface and that there was significant mobility of carbon atoms (or vacancies) in the same temperature range. They pointed out that these conditions should permit epitaxial growth of diamond. They stated Epitaxial growth of diamond on diamond is possible despite the large instability with respect to graphite. It is simply necessary to inhibit the nucleation of graphite while adding carbon atoms to the surface. Presumably this restricts the process to temperatures below about 1400⬚C and to low addition rates. . . . Apparently diamond provides a spectacular example of the theorem that extension by epitaxial growth of an unstable phase is possible if the barriers to nucleation of all more stable phases are higher.

Several early U.S. patents also described the beneficial effects of hydrogen. Hibshman [21] synthesized diamond from CO with 5% H2 in the presence of a platinum catalyst. He speculated that the catalyst promoted the formation of atomic hydrogen. Vickery [22] used gas mixtures containing 95% H2 and 5% hydrocarbon to suppress the formation of graphitic carbons. Angus and Gardner [23] discussed the use of hydrogen in a 1972 patent. The detailed molecular mechanism responsible for diamond growth has been a subject of speculation since the earliest successful studies. Angus and his coworkers in 1968 [24] suggested that Thermodynamic computations indicate the partial pressure of free carbon atoms in the vapor phase is exceedingly low, so low, in fact, that free carbon atoms in the vapor probably are not a major source of carbon for diamond growth. A more likely mechanism involves reaction

22

Angus of CH4 or some other hydrocarbon fragment with the diamond surface resulting in the deposition of a carbon atom.

Based on growth rate data, which showed a first-order dependence on [CH4] and half-order with [C2H4], the Angus group proposed a three-step mechanism: formation of mobile surface species containing a single carbon atom, surface diffusion of the species, and subsequent addition to a vacant surface site. The surface sites were proposed to be ‘‘two-bonded’’ sites, that is, kink sites on 具110典 steps on {111} surfaces or sites on {100} surfaces [25]. Critical reviews published during this era accepted the conclusion that diamond could be grown at low pressures [26,27]. There was, however, considerable skepticism that the growth rates could ever be made high enough to support commercially viable technologies. B.

Work at Case Western Reserve University, Cleveland, Ohio

Angus concluded, after reviewing the earlier work and meeting with some of the principals, in particular, Eversole, Lander, and Hibshman, that chemical vapor deposition from hydrocarbons was the method most likely to succeed. To renew the field it would be necessary to obtain absolutely convincing experimental proof of low-pressure diamond growth. Support for this effort, as a ‘‘blue sky’’ project, was obtained from Richard F. Cornelissen of the Air Force Cambridge Research Laboratories. The first goal of reproducing Eversole’s work was achieved by the mid-1960s [24]. To further prove that new diamond was being grown, diborane was added to the gas phase. P-type, semiconducting blue diamond was obtained during diamond growth; no conductivity was achieved by annealing under diborane when growth was not taking place [28,29]. The role of hydrogen in diamond growth was a major focus of the research. This was strongly motivated by the work of Lander and Morrison, who showed that a hydrogen-rich environment should suppress the nucleation and growth of unsaturated, graphitic structures. Initial experimental results clearly showed that hydrogen played a critical role in the diamond growth process [25,30]. The presence of molecular hydrogen dramatically slowed down the spontaneous nucleation and codeposition of graphite. However, molecular hydrogen by itself was not enough to suppress graphite growth completely. Eventually, graphitic carbon nucleated on the surface and suppressed further diamond growth. It was then necessary to remove the graphitic deposits, using atomic hydrogen, and to repeat the sequence. Atomic hydrogen was first used at the suggestion of Nelson C. Gardner, who had studied the decomposition of H2 and desorption of H from

Short History of Diamond Synthesis

23

tungsten surfaces while a graduate student at Iowa State University. Hot tungsten filaments were used for generating atomic hydrogen, which was used to remove graphitic deposits from the diamond after growth. The atomic hydrogen also rationalized and prepared the diamond surface for the next growth cycle. However, atomic hydrogen was not used during the growth part of the cycle. Results from the experiments done at Case Western Reserve University were presented in Kiev in September 1971 at the ‘‘InterDiamond’’ meeting. The work appeared in the scientific literature only in Russian translation [28]. C.

Work at the Physical Chemistry Institute, Moscow, Russia

Research on low-pressure diamond synthesis started at the Physical Chemistry Institute in Moscow in 1956 [31]. The Soviet work prior to 1977 has been reviewed by Badzian and DeVries [4,32]. Their earliest report was an idea of Boris Spitsyn for deposition on diamond using CBr4 or CI4 at temperatures from 800 to 1000⬚C and pressures of about 3 ⫻ 10⫺6 mm of mercury [33]. This method was described in an unpublished author’s certificate filed July 10, 1956; the patent itself did not appear until 1980. The earliest published paper from the Soviet group appeared in 1968 and described growth of diamond filaments by a vapor-liquid-solid technique using molten iron as the liquid catalytic agent [34]. Subsequently, the Soviet group explored direct chemical vapor deposition from hydrocarbons and in 1969 at a meeting in Novosibirsk presented an independent confirmation of Eversole’s results [35]. Pure methane, at pressures from 0.1 to 0.3 torr and temperatures from 950 to 1050⬚C, was used for deposition on diamond seed crystals. Much higher rates were reported than those of Eversole. A major achievement of the Soviet group was the use of atomic hydrogen during growth. This gave much higher growth rates and also permitted nucleation of new diamond crystallites on nondiamond substrates. This development apparently came about as the result of a suggestion of Valentin Varnin (A. Badzian, personal communication, 1991). He had learned of the results from Case Western Reserve University on atomic hydrogen from Dmitri Fedoseev, who had attended the 1971 Kiev meeting where the work was presented. Varnin suggested using atomic hydrogen during the growth process; the first experiments were performed by Polchanskiya. The date of these experiments is not known, but the use of atomic hydrogen is not mentioned in a 1975 Scientific American article by Deryagin [36]. Prior to the use of atomic hydrogen, the Russian group used atomic oxygen to remove graphitic codeposits in a cyclic process [37].

24

Angus

Formation of diamond crystals on nondiamond substrates was first revealed by Deryagin and others in 1976 [38]. The crystals ‘‘were produced in a chemical transport reaction occurring in a closed system at a pressure below atmospheric and a substrate temperature of the order of 1000⬚C.’’ The use of atomic hydrogen was not mentioned. Fedoseev et al. [39] described the effect of molecular hydrogen on diamond nucleation, also in 1976. In a later paper Fedoseev et al. [40] described the effect expected from the use of atomic hydrogen during the growth step.

D.

Work at NIRIM, Tsukuba, Japan

Descriptions of methods for rapid growth of diamond at low pressures were first given by a group of Japanese researchers at the National Institute for Research in Inorganic Materials (NIRIM) in Tsukuba, Japan. The leader of the NIRIM program was Nobuo Setaka; key researchers included Yoichiro Sato, Seichiro Matsumoto, and Mutsukazu Kamo. A project on metastable diamond growth was initiated at NIRIM in 1974. Successful diamond synthesis was first achieved using a hot filament to activate CH4-H2 gas mixtures and reported in December 1981 at a fall meeting of the Carbon Society of Japan. Shortly thereafter, a remarkable series of papers was presented by the NIRIM group in which techniques for vapor growth of diamond at rates of several micrometers per hour were described [41–44]. The Japanese group revealed enough to permit confirmation of their results. The hot-filament and microwave-assisted deposition processes, still very popular, were developed at NIRIM. The current worldwide interest in the new diamond technology follows directly from the NIRIM effort.

V.

SUMMARY

Diamond synthesis at low pressure by chemical vapor deposition has been the result of a number of discoveries by research groups in the United States, the Soviet Union, and Japan. Extension of existing diamond seed crystals by chemical vapor deposition was first achieved in the early 1950s by William Eversole and predates diamond synthesis at high pressures. Understanding the role of molecular and atomic hydrogen in the nucleation and growth process led to higher growth rates, suppression of nucleation of graphitic carbons, and finally the de novo nucleation of diamond crystals on nondiamond substrates. One striking feature of the history was the reluctance by many to accept the possibility of diamond synthesis outside its range of thermodynamic stability.

Short History of Diamond Synthesis

25

ACKNOWLEDGMENTS Paul Mohr made available the report describing Eversole’s work at Union Carbide Corporation. Robert DeVries provided the author with the early internal General Electric report and with many delightful hours of conversation about early diamond work at General Electric and elsewhere. Andrzej Badzian, Valentin Varnin, Boris Spitsyn, and Dmitri Fedoseev provided insights into the relationship between the early American and Russian work. A modified version of this article appeared in Il Nuovo Cimento, The Proceedings of the 1996 Enrico Fermi Summer School on the Physics of Diamond. Permission to reprint it from the Societa Italiana di Fisica is gratefully acknowledged. REFERENCES 1. 2. 3. 4. 5.

6. 7. 8. 9. 10. 11. 12. 13. 14.

15. 16. 17. 18. 19. 20. 21. 22.

G Davies. Diamond. Bristol: Adam Hilger, 1984. RM Hazen. The New Alchemists: Breaking Through the Barriers of High Pressure. New York: Times Books, Random House, 1993. FP Bundy. Mater Res Soc Symp Proc 383:3–20, 1995. RC DeVries. Annu Rev Mater Sci 17:161, 1987. JC Angus. In: KE Spear, JP Dismukes, eds. Synthetic Diamond: Emerging CVD Science and Technology. New York: John Wiley & Sons, 1994, pp 21– 39. I Amato. Stuff: The Materials the World Is Made Of. New York: Basic Books, 1997. G Averani, CA Targioni. G Litt Ital 8:221, 1711. GN Lewis, M Randall. J Am Chem Soc 37:458, 1915. F Simon. Handb Phys 10:350, 1926. OI Liepunski. Usp Khim 8:1519, 1939. FD Rossini, RS Jessup. J Res Natl Bur Stand 21:491, 1938. H Meincke. Schweiz Arch Angew Wiss Tech 23:85, 1957. H Meincke. Gemmologist 26:46, 1957. AD Kiffer. Report, Tonowanda Laboratories, Linde Air Products Co, Synthesis of Diamond from Carbon Monoxide, June 6, 1956; WG Eversole, US Patents 3,060,187 and 3,060,188, 1962. FP Bundy, HT Hall, HM Strong, RH Wentorf. Nature 176:51, 1955. H Liander, E Lundblad. Ark Kemi 16:139, 1960. RA Oriani, WA Rocco. General Electric Research Laboratory Memo No MA 36, August 1957. A Neuhas, Angew Chem 66:525, 1954. PW Bridgman. Sci Am 193:42, 1955. JJ Lander, J Morrison. Surf Sci 4:241, 1966. HJ Hibshman. US Patent 3,371,996, March 5, 1968. EC Vickery. US Patent 3,714,334, January 30, 1973.

26

Angus

23. 24. 25. 26.

JC Angus, NC Gardner. US Patent 3,661,526, May 9, 1972. JC Angus, HA Will, WS Stanko. J Appl Phys 39:2915, 1968. SP Chauhan, JC Angus, NC Gardner. J Appl Phys 47:4746, 1976. JR Wedlake. In: JE Field, ed. The Properties of Diamond. London: Academic Press, 1979, pp 501–535. FP Bundy, HM Strong, RH Wentorf Jr. In: Chemistry and Physics of Carbon, Vol 10, Methods and Mechanisms of Synthetic Diamond Growth. New York, Marcel Dekker, 1973, pp 213–263. JC Angus, NC Gardner, DJ Poferl, SP Chauhan, TJ Dyble, P Sung. Sin Almazy 3:38, 1971; presented at the International Conference on Applications of Synthetic Diamonds in Industry, Kiev, 1971. DJ Poferl, NC Gardner, JC Angus. J Appl Phys 44:1418, 1973. SP Chauhan, JC Angus, NC Gardner. J Vac Sci Technol 11(1):423, 1974. BV Deryagin, DV Fedoseev. Russ Chem Rev 39:783, 1970. AR Badzian, RC DeVries. Mater Res Bull 23:385, 1988. BV Spitsyn, BV Deryagin. Author’s certificate dated July 10, 1956; USSR Patent 339,134, May 5, 1980. BV Deryagin, DV Fedoseev, BV Spitsyn, DV Lukyanovich, AV Lavrentev. J Cryst Growth 2:380, 1968. BV Deryagin, VA Ryabov, DV Fedoseev, BV Spitsyn, BM Lukyanovich, KS Uspenskaya. Second All-Union Symposium on Processes for Nucleation and Growth of Crystals and Films of Semiconducting Compounds, Novosibirsk, May 12–16, 1969. BV Deryagin, DV Fedoseev. Sci Am 233:102, 1975. BV Deryagin, DV Fedoseev. Proceedings of International Conference on Applications of Synthetic Diamonds in Industry, Kiev, September 14–18, 1971. Kiev: Naukova Dumka, 1974. BV Deryagin, BV Spitsyn, LL Builov, AA Klochkov, AE Gorodetski, AV Smolyaninov. Dokl Akad Nauk SSSR 231:333, 1976. DV Fedoseev, SP Vnukov, BV Deryagin. Zh Fiz Khim 50:2751, 1976. DV Fedoseev, KS Uspenskaya, VP Varnin, SP Vnukov. Izv Akad Nauk SSSR Ser Khim 6:1252, 1978. S Matsumoto, Y Sato, M Kamo, N Setaka. Jpn J Appl Phys Part 2 21:L183, 1982. S Matsumoto, Y Sato, M Tsutsumi, N Setaka. J Mater Sci 17:3106, 1982. M Kamo, Y Sato, S Matsumoto, S Setaka. J Cryst Growth 62:642, 1983. Y Matsui, S Matsumoto, N Setaka. J Mater Sci Lett 2:532, 1983.

27.

28.

29. 30. 31. 32. 33. 34. 35.

36. 37.

38. 39. 40. 41. 42. 43. 44.

3 Film Characterization Methods: Structure and Composition Ludwig Josef Balk and Ralf Heiderhoff University of Wuppertal, Wuppertal, Germany

I.

INTRODUCTION

Diamond thin films produced by chemical vapor deposition (CVD) are projected to have many future applications because of the unique properties of diamond [1]. The applications of diamond are often considered on the basis of the properties of natural diamond crystals. Natural diamond is known to be exceedingly hard and wear resistant, have high thermal conductivity and melting point, be optically transparent from the far infrared to the ultraviolet (5.5 eV), and have unique semiconducting properties. Common applications of diamond based on these properties are cutting tools, wear-resistant parts, medical coatings, heat sinks, acoustic response systems, inert optical coatings, and electronic and electro-optic devices. For this potential to be realized, the nucleation, growth, and resulting structure of diamond films need to be more thoroughly understood and controlled. Numerous techniques are useful in the characterization of diamond, too many to review in this chapter. In fact, even this review is limited to the techniques that have been used in the recent past to analyze CVD diamond. Carbon film deposition or growth has been carried out using different methods. Before the development of chemical vapor deposition methods, the films exhibited an sp2 bonding character, whereas the formation of sp2 bonding is suppressed or removed via an etching process at present. With respect to the critical areas of carbon phase identification, a definition was published in 1987 by Messier et al. [2] for the first time to distinguish diamond films 27

28

Balk and Heiderhoff

from other amorphous carbonaceous materials and hydrogen-rich diamondlike structures: Crystalline morphology can be observed by electron microscopy [scanning electron microscopy (SEM), and transmission electron microscopy (TEM)]. Monocrystallinity is observable by electron or x-ray diffraction. A small Raman peak is present at 1332 cm⫺1. This definition tolerates small impurities, such as nitrogen, hydrogen, silicon, or metals, that are also present in natural diamond. The two techniques that have become the standard methods of assessing the percentage of CVD diamond and the surface morphology are Raman scattering and SEM. The application of SEM is almost qualitative, but the Raman scattering results can be quantitative in some cases.

II.

RAMAN SPECTROSCOPY

Raman spectroscopy has emerged as the most frequently used tool to characterize diamond films prepared by various chemical vapor deposition methods because of the spectroscopic signature [3–13] (see Fig. 1). This technique provides well-known spectra for different carbonaceous compounds, such as graphite, glassy carbon, or amorphous diamond-like carbon, and is also used to determine the phase purity and crystalline perfection by analyzing the peak positions and peak width. The Raman spectra of different forms of sp2- and sp3-bonded carbon are illustrated in Figure 2. The spectra of diamond exhibit a sharp peak that is centered at a Raman shift from 1331 to 1335 cm⫺1 (at room temperature) with a bandwidth Du1/2 at half-intensity less than 2 cm⫺1, whereas graphite exhibits a broad band centered from 1565 to 1585 cm⫺1 (see Fig. 3). The peak frequency is known to downshift with temperature and to upshift when compressive internal stresses are present.

Figure 1

Schematical setup of a Raman experiment.

Film Characterization Methods

Figure 2

Raman spectra of different forms of sp 2- and sp 3-bonded carbon.

29

30

Balk and Heiderhoff

Figure 3

Raman-active normal modes in graphite.

The band broadening is related to internal stresses, thermoelastic stresses, pure thermal effects, and the density of defects in the crystal as well as the crystallite size. For microcrystalline graphite a strong feature appears at 1350 cm⫺1, and the 1580 cm⫺1 feature shifts to lower wave numbers with decreasing crystalline size. The feature at 1350 cm⫺1 has been identified as a vibrational mode that occurs at the boundary of the Brillouin zone.

III.

ELECTRON MICROSCOPY

In the early phase of the development of diamond technology, SEM micrographs captured the attention of researchers throughout the world. SEM is an excellent tool for the examination of the surface morphology of CVD diamond and of growth features such as steps and islands on the diamond surfaces (see Figs. 4 and 5). It has been recognized that CVD diamond films have a wide range of surface morphologies that are highly dependent upon the experimental conditions. These surfaces correlate with other techniques such as Raman spectroscopy. For example, if growth parameters were not optimized, the morphology could become sub-microcrystalline and then cauliflower-like. This correlated with a deterioration of the diamond peak and an increase in the graphitic peak in the Raman spectrum. A review of several reports on surface morphologies of diamond films that have related the morphological development to various deposition parameters is given by Zhu et al. [14]. At present, high-resolution transmission electron microscopes mainly work with acceleration voltages of 200 to 400 kV and can be combined with electron diffraction, x-ray microanalysis, and electron energy loss spectroscopy (EELS) modes (see Fig. 6). In the Bragg contrast mode, the objective diaphragm selects the primary beam for bright-field imaging and

Film Characterization Methods

Figure 4

31

SEM micrograph of a CVD diamond surface. (Courtesy of Zeiss.)

Figure 5 Unprocessed electron beam patterns from 100 nm2 areas of an uncoated polycrystalline diamond layer. (Courtesy of Zeiss.)

32

Figure 6

Balk and Heiderhoff

Schematical setup of a transmission electron microscope.

Bragg reflected beams for dark-field imaging (see Fig. 7). When using small electron probes, information about crystal symmetry, specimen thickness, and lattice distortion can be obtained by the convergent-beam electron diffraction methods (see Fig. 8). The possibility to use higher scattering angles in the diffraction mode makes it possible to record higher order Laue zone (HOLZ) diffraction patterns. Structural determination of the diamond crystalline morphology can be accomplished using transmission electron microscopy [15–19]. Further, TEM studies have shown that a large number of defects, which cannot be detected by other techniques, are present in the layers [20–24]. There are two methods of observing the thin films [25]. One

Film Characterization Methods

Figure 7

33

Bright-field image of a CVD diamond film. (From Ref. 18.)

method is the observation of a film that is thinned in the direction of a cross section or surface plane. This method can be used with any thick sample. But it takes a long time to prepare thin films. The other method is observing the side surface and does not require any thinning of the film or substrate. EELS and especially high-resolution electron energy loss spectroscopy (HREELS) obtained in the transmission mode in a TEM can also be utilized to examine the bonding character of a carbon film (see Fig. 9). Energy losses due to K-shell ionization and the fine structure due to interband transitions and plasmon losses that appears at higher energies can be used to identify different phases of carbon because the spectrum is strongly dependent on the type of bonding present. The fine features of these growth defects can be clearly observed and studied by high-resolution electron microscopy (HREM), which provides, in addition to detailed atomic resolution images, crystallographic information on the defects and their boundaries. Several studies obtained a view of the nature of stacking faults, twins and twinning configurations, as well as the interface between the substrate and the diamond film (see Fig. 10).

34

Balk and Heiderhoff

Figure 8

Diffraction pattern corresponding to Fig. 7. (From Ref. 18.)

Figure 9

EELS of bulk diamond and polycrystalline graphite.

Film Characterization Methods

35

Figure 10 HERM (high-resolution electron microscopy) of twins ending at a virtually amorphous inclusion. T, twin lamellae; I, inclusion; L, lattice image confusion. (From Ref. 19.)

IV.

DIFFRACTION

Low-energy electron diffraction (LEED) is well suited to study the structure of the outermost layers of diamond films. It is possible to combine LEED with Auger electron spectroscopy (AES) to establish the chemical surface situation if the depth of the scattered electrons is small and does not exceed one or two atomic layers. In order to carry out such studies, one must first obtain an unequivocally clean film surface. But it is very difficult to bombard the surface with, for example, argon ions that are sufficiently energetic to remove adsorbed or combined surface layers. The outermost surface layer can become graphitic [26–37] and cannot be annealed out, as a reconstruction of the diamond surface can occur if the impurities are removed by heat treatments [38–45] (see Figs. 11 and 12). LEED studies have shown that hydrogen plays an important role in the reconstruction of diamond surfaces [40,43]. Thermal desorption studies have indicated that hydrogen and oxygen are adsorbed on diamond surfaces [38,44]. Figure 12 illustrates the structural model for the (100) diamond surface: (a) 2 ⫻ 1 and (b) 1 ⫻ 2 reconstruction. Although AES provides valuable compositional information, it does not yield any information about trace impurities because of the sensitivity limitation of the technique. Conventional techniques such as secondary ion mass spectrometry (SIMS) or laser microscope mass analysis spectroscopy

36

Balk and Heiderhoff

Figure 11 Carbon 1s photoelectron spectra from (111) diamond. (a) Terminated surface polished and heated to 1100 K. (b) Clean surface heated to 1250 K. (c) Clean surface hydrogenated by atomic hydrogen. (From Ref. 32.)

(LAMMA) are much more sensitive than AES, and when electronic devices were realized, such chemical analysis became more important. Several alternative techniques such as Rutherford backscattering have been utilized for detecting hydrogen in diamond films. SIMS can detect hydrogen, but quantification is difficult because of the presence of background hydrogen in the vacuum system. Synchrotron radiation laboratories around the world are using x-ray diffraction (XRD) for the structural determination of diamond. These range from the nondestructive slicing of large diamonds by synchrotron x-ray sec-

Film Characterization Methods

37

Figure 12 Surface model for (a) 2 ⫻ 1 diamond (100) and (b) 1 ⫻ 1 diamond (100). (䡩) H atom; (●) C atom. (From Ref. 59.)

tion topography [46,47] to high-resolution studies of the spike diffuse reflections from platelet crystal topography (see Fig. 13). Mai et al. [48] pointed out the intrinsic suitability of diamond for xray beam conditioning, for example, due to its strong x-ray reflectivity and low x-ray absorption. A detailed comparison of natural and synthesized diamond for these applications was given by Lang [49]. A typical scattering geometry used for x-ray measurements is illustrated in Figure 14. In recent years, several authors have reported on the growth of textured diamond films. X-ray diffraction was used most often to determine the growth orientation and texture of these films [50–55] (see Figs. 15 and 16). The detailed analysis of the texture and surface morphology of polycrystalline diamond films utilized the fact that growth starts from randomly oriented nuclei on Si (100) and Si (111). The texture develops with increasing film thickness because of the competition of differently oriented grains (see Fig. 17). Twinning during the nucleation stage of the film appears to stabilize preferential growth along the (100) or (111) direction. The growth

38

Balk and Heiderhoff

Figure 13 (a) X-ray projection topography of type a natural diamond shaped as shown in b. (b) Perspective sketch of the part of the polished parallelepiped specimen that contains a large stacking fault. (From Ref. 49.)

parameter ␣ = V100 /(兹3V111), where V100 and V111 are the growth rates on {100} and {111} faces, can be varied by controlling the process parameters. In addition, other techniques such as extended x-ray absorption fine structure (EXAFS) or extended electron energy loss fine structure (EEELFS) are beginning to be utilized for bonding analysis. Several authors reported the x-ray absorption spectra of diamond films (see, for example, Ref. 56), especially to observe the Si-diamond interface [57–60].

V.

LUMINESCENCE

The luminescence from diamond has been studied for more than 50 years and is an established tool to investigate chemical and structural defects in

Film Characterization Methods

Figure 14 62.)

39

Typical scattering geometry used for x-ray measurements. (From Ref.

diamond films. The spectral range covered by the luminescence extends from 5.3 eV in the ultraviolet (‘‘edge emission’’ corresponding to the indirect energy gap [63]) to around 1.2 eV in the near infrared (‘‘H2’’ vibronic band [64]). Approximately 120 optical centers that have been documented for diamond are listed by Clark et al. [65]. The major spectra associated with edge emission, vibronic band, and ‘‘band A’’ emission are discussed in Ref. 66. Figure 18 illustrates a block diagram of an integral, spectral, and timeresolved CL experiment [69]. Cathodoluminescence has proved to be very useful in the characterization of the film because of the high spatial resolution and with respect to the wide energy gap of diamond (Eg = 5.5 eV) (see Fig. 19). The generation rate of a 50-kV beam at a current density of 0.01 A/cm2 is about 2 ⫻ 1023 electron-hole pairs per cm3 per second [67] and the penetration depth is comparable to that with GaAs. The electron beam generates simultaneously a wide variety of intrinsic and extrinsic processes. Cathodoluminescence spectroscopy is therefore less selective than photoluminescence spectroscopy, where the energies of the incident photons are less than the band gap. Diamond is an indirect semiconductor, and the edge emission may be observed only from films that are relatively free from defects. For indirect transitions a phonon must be created, with its energy បw being taken from the stored energy in the exciton. The energy emitted in the photon is given by hu = Eg ⫺ Ex ⫺ បw, where Ex is the exciton binding energy. Free excitons with the emission of transverse acoustic, transverse optic, and longitudinal optic phonons of wave vectors ⫾ k having values between 87 and 163 meV

40

Balk and Heiderhoff

Figure 15 SEM image and XRD pole figure of a highly oriented (100)-textured diamond film [exhibiting an 8⬚ full width at half-maximum of the (200) peaks] on a (100) Si substrate. (From Ref. 54.)

Film Characterization Methods

41

Figure 16 (a) Experimental XRD intensity map of the C 1S emission of a diamond (111) surface. (b) Calculation for a diamond cluster of 200 C atoms in 10 layers. (From Ref. 61.)

have been found. In contrast, the luminescence from a bound excition can proceed without creating phonons. Qualitatively, this reflects the localization of the hole on an acceptor center. The edge recombination radiation spectrum from a natural p-type semiconducting diamond is illustrated in Figure 20. Regarding donor-acceptor pair spectra, the donor captures an electron and the acceptor captures a hole when electron-hole pairs are generated by the electron beam. This mechanism was used for a long time to explain the broad band A luminescence that is observable in all types of natural, synthetic, and CVD diamond and extends from about 2 up to 3.5 eV. The energy is given by E(r) = Eg ⫺ EA ⫺ ED ⫹

e2 ea 2 ⫺ 4␲␧ r 4␲␧ r 6

where EA and ED are the acceptor and donor binding energies, ␧ is the static dielectric constant, r is the distance between the donor and acceptor, and a is an adjustable parameter in the polarization term [68]. Today, however, one believes that it is better to use the description band A than donor-acceptor pair recombination because at present there is no theory that describes all the phenomena that are observable, for example, in time-resolved and temperature-dependent measurements (see Fig. 21). Investigations have suggested that band A is due to dislocations or vacancy clusters acting as recombination centers for electron-hole pairs [71].

42

Balk and Heiderhoff

Figure 17 XRD patterns of one given diamond film for different film thicknesses. (From Ref. 62.)

Finally, one has to note that among band A and radiation damages, nitrogen is by far the most common impurity in diamond with many forms. Many sharp lines are correlated with the different nitrogen impurities. The 575-nm system, for example, is most intense in cathodoluminescence spectra in nitrogen-doped thermal CVD diamond. It turns out that the system is due to a single nitrogen atom and a vacancy [72]. Collins and Robertson [73] showed that the zero-phonon line responds relatively rapidly to random stress S and proposed that S can be estimated from the line width ⌬␭ using the expression ⌬␭ = S ⫻ 10 meV/GPA.

Film Characterization Methods

43

Figure 18 Block diagram of an integral, spectral, and time-resolved CL experiment. (From Ref. 69.)

VI.

THERMAL CONDUCTIVITY

It is well known that diamond crystals have the highest known thermal conductivity. The thermal conductivity ␬ of type IIa natural diamond at room temperature (␬ = 2200 W m⫺1 K⫺1) is a factor of 5 greater than that of pure

Figure 19

SEM and Cl micrographs of a CVD diamond film. (From Ref. 69.)

44

Balk and Heiderhoff

Figure 20 Cl spectrum of the edge recombination from natural p-type diamond. (From Ref. 63.)

Figure 21

Time-resolved Cl spectra of a natural diamond. (From Ref. 69.)

Film Characterization Methods

45

Table 1 Thermal Properties of Diamond and Other Materials at Room Temperature

Material Diamond (natural IIa) Copper AlN Si

Thermal conductivity (W m⫺1 K⫺1)

Thermal diffusivity (cm2 sec⫺1)

Linear thermal expansion (10⫺6 K⫺1)

2200 400 200 140

12 1.2 0.8 0.8

1 16 4 3

copper and over a factor of 10 higher than that of commonly used semiconductors (see Table 1). Many researchers have observed a large range of thermal conductivities in films and even in single crystals due to impurities or defects. Polycrystalline diamond films, because of their crystalline boundaries and graphitic contaminations, are not expected to have the same high thermal conductivity as single crystals. Today the highest values 2000 W m⫺1 K⫺1 to 2200 W m⫺1 K⫺1 were found for MWCVD films [70,74,75]. Anthony et al. [76] examined high-quality diamond crystals having different isotopic distributions of carbon. Experimentally, they found an increase of 50% in the diffusivity (thermal diffusivity ␣ related to the thermal conductivity ␬ by ␣ = ␬/c␳, where c is the thermal capacity and ␳ is the density of the material) of isotopically decreased diamond crystals (see Table 2), whereas they had expected only 5% for the isotopically purest specimen. Today there is no theory that is able to give a definite prediction of the measured enhancement in conductivity achieved by reducing the isotope concentration [77,78].

Table 2 Thermal Conductivity of High-Conductive Diamonds Depending on the Isotope Concentration Material 0.07% 13C diamond 0.5% 13C diamond 1.0% 13C diamond (natural abundance) Source: Ref. 76.

Thermal diffusivity (cm2 sec⫺1)

Thermal conductivity (W m⫺1 K⫺1)

18.5 14.5 12.4

3300 2600 2200

46

Balk and Heiderhoff

In the case of CVD diamond, the thermal conductivity is difficult to measure accurately because of the extremely high values. The difficulty is further increased by the anisotropic nature and the geometry of the samples. Even when comparing measurement techniques that measure bulk, in-plane thermal conductivity, results differ significantly. For instance, a sample that was measured at a large number of laboratories around the world was found to have a thermal conductivity of 1400 to 1550 W m⫺1 K⫺1 by the AC calorimetry method and the converging wave method and found to have a thermal conductivity of 2500 W m⫺1 K⫺1 according to a modified Angstrom method [79]. The measuring methods vary considerably in requirements for operating skills and sample preparation, and they can be divided into contact and noncontact, steady-state or dynamic methods. A review of the different methods is given by Fournier and Plamann [80]. A brief description of two standard methods is given as follows. A.

Conventional Method

In this method, the sample is shaped into a rectangular bar. Two thin-film heaters are evaporated directly onto the sample surface, one near each end (see Fig. 22). The heat flux through the sample is obtained by knowing the electrical power input to the heaters after all the revenues of heat loss have been quantified. Normally, the measurements are performed with the sample in vacuum, surrounded by a heated shield at a known temperature, the temperature of the thermal ground, to minimize unwanted heat transfer through radiation. A row of very fine thermocouples, so that they do not have a thermal gradient of their own, are then attached to the sample bar at a known distance from each other. The measured gradient dT/dx can be corrected for radiative loss and used to calculate ␬. An alternative is to use sinusoidal

Figure 22 Illustration of the two-heater heated-bar technique of measuring the thermal conductivity.

Film Characterization Methods

47

heat flows. Then the thermal waves would be detected and the change in either amplitude or phase of the waves used to determine the diffusivity. Further, it is possible to apply the heat pulse to one side of the sample from a flash lamp or laser and to monitor the temperature with an optical detector on the opposite side of the sample. For films that are transparent, an opaque layer must be evaporated onto the surface. B.

Mirage Effect Method

This thermo-optical method is based on the sensitive detection of thermal gradients in the gas layer adjacent to heated sample surfaces [81]. The experimental setup is shown in Figure 23. The light of a focused modulated laser beam is used as the source of heat and launches a thermal wave in the sample. Detection is through the mirage effect, which utilizes a second laser beam to traverse the heated air above the heated sample. The periodic deflection of the probe beam is monitored by the use of a position-sensitive or angular detector.

VII.

SCANNING PROBE TECHNIQUES

Some papers have shown the feasibility of the scanning tunneling [82,88] and scanning force techniques [89–92] to study the morphologies of CVD diamond films, especially nucleation and growth. Scanning tunneling microscopy (STM) and scanning force microscopy (SFM) routinely achieved atomic resolution on many substrates and are suited to the determination of the nanometer-scale structure of diamond films. STM is a useful characterization technique especially because it provides a direct method for observing surfaces and local electronic structures. STM has very high three-dimensional resolution and does not require special sample preparation, in contrast to, for example, TEM. The graphite form of

Figure 23

Experimental setup of the mirage effect method.

48

Balk and Heiderhoff

carbon has proved the standard for demonstration of atomic resolution in air and the technique has been used in the examination of other forms of bonded carbon. As the surface of CVD diamond is stable, STM images can be obtained even in air. It is generally recognized that this stability is due to the termination of freestanding bonds by hydrogen. It is very difficult to achieve high resolution because undoped diamond is normally an insulator and boron doping has been shown to change the surface morphology of diamond films. Therefore SFM, which has no such requirement, would seem to be a more generally applicable tool. Topography and microfriction can be determined simultaneously by SFM by measuring the normal and lateral deflections (torsion of the lever) of the cantilever by means of a laser beam. Since scanning probe microscopy was invented, many different applications have been developed to image various specimen features with high spatial resolution [93,94]. The efficiencies of STM-EBIC (EBIC: electron beam–induced current) and conventional SEM-EBIC measurements on diamond films were compared for the first time by Koschinski et al. [95]. EBIC investigations make it possible to characterize the electrical properties (carrier concentration, electrical potential) of the specimen under test on a microscopic scale [96] (see Fig. 24). Excess carriers are generated in a small volume of the sample

Figure 24

Schematical diagram of the STM-EBIC setup. (From Ref. 95.)

Film Characterization Methods

49

by the electron beam and induce a detectable current that depends on the structure of the sample, the position of the electron beam, and the electrical field inside the sample. This can be used to analyze inhomogeneities and local electrical properties of CVD diamond films (see Fig. 25). The determination of the electrical conductivity with respect to the localization of defects is a problem that is difficult to solve, especially when reaching a nanometer scale. Based on SFM, so-called scanning resistance microscopy (SRM) [97,98] and contact current measurements (CCMs) [99] are able to obtain two-dimensional profiles. In addition, CCM allows simultaneous measurement of the current signal and the topography. CCM determines the variation of the electronic properties of the sample by analyzing the current characteristics. For this purpose, a metallized SFM tip is scanned in contact mode over the sample surface and a potential is applied between the tip and a second fixed electrode. The current signal and the topography are obtained simultaneously and thereby can be easily correlated. It is useful to determine the thermal conductivity in the nanometer region by using an SThM (scanning thermal microscope) [100]. The probe consists of a filament that is resistively heated (see Fig. 26). Being the thermal element of the probe, this filament is part of a Wheastone bridge. High-thermal-conductivity areas of the sample cause the thermal feedback loop to increase the applied voltage to keep the temperature of the tip constant. The applied voltage, therefore, correlates directly with the thermal conductivity (see Fig. 27). A comparison of the determination of the electrical and thermal conductivity in the nanometer region is given by Maywald et al. [101].

Figure 25 Illustration of defects in natural diamond by the EBIC method: (a) SE and (b) EBIC images.

50

Balk and Heiderhoff

Figure 26

VIII.

Schematical drawing of the probe of an SThM. (From Ref. 95.)

SUMMARY

No other material system has been analyzed in such a way as diamond films around the world in recent years. Many applications of diamond related to the large spectrum of properties make it necessary to investigate the samples with many different characterization systems.

Figure 27 101.)

Topography and thermal images of an Si-diamond interface. (From Ref.

Film Characterization Methods

51

In this chapter we presented an overview of the analyzing systems as well as their problems with respect to these outstanding properties. Technological advances in film growth and characterization methods are occurring constantly and theories are beginning to evolve. In some areas such as heat sinks, x-ray windows, and wear-resistant coatings on cutting tools, diamond is already commercial available. However, CVD diamond growth technology is still far from realizing all of the further attractive potential applications. Some of those will certainly be available in the near future, whereas other will require continued research or may even prove impossible.

REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

JE Field. The Properties of Natural and Synthetic Diamond. London: Academic Press, 1992. R Messier, AR Badzian, T Badzian, KE Spear, P Bachmann, R Roy. Thin Solid Films 153:1, 1987. RO Dillon, JA Woolam, V Katkanant. Phys Rev B 29:3482, 1984. SK Sharma, HK Mao, PM Bell, JA Xu. J Raman Spectrosc 16:350, 1985. A Richter, HJ Scheibe, W Pompe, K-W Brzezinka, I Mu¨hling. J Noncryst Solids 88:131, 1986. RJ Nemanich, JT Glass, G Lucovsky, RE Schroder. J Vac Sci Technol A 6: 1783, 1988. DS Knight, WB White. J Mater Res 4:385, 1989. LH Robins, EN Farabaugh, A Feldman. J Mater Res 5:2456, 1990. RE Shroder, RJ Nemanich, JT Glass. Phys Rev B 41:3738, 1990. DW Kweon, J-Y Lee. J Appl Phys 69:8329, 1991. PV Huong. Mater Sci Eng B11:235, 1992. JE Graebner, JA Mucha, L Seibles, GW Kammlott. J Appl Phys 71:3143, 1992. PK Bachmann, DU Wiechert. Diamond Relat Mater 1:422, 1992. W Zhu, BR Stoner, BE Williams, JT Glass. Proc IEEE 79:621, 1991. Y Goto, K Kurihara, Y Sawamoto, T Kitakohji. Appl Phys Lett 60:172, 1992. PJ Fallon, LM Brown. Diamond Relat Mater 2:1004, 1993. H Sjo¨stro¨m, L Hultmann, JE Sundgren. Diamond Relat Mater 2:562, 1992. P Wurzinger, M Joksch, P Pongratz. Inst Phys Conf Ser 134:157, 1994. M Joksch, P Wurzinger, P Pongratz, R Haubner, B Lux. Diamond Relat Mater 3:681, 1994. BE Williams, JT Glass. J Mater Res 4:373, 1989. W Zhu, AR Badzian, R Messier. J Mater Res 4:659, 1989. JL Kaae, PK Gantzel, J Chin, WP West. J Mater Res 5:1480, 1990. J Narayan. J Mater Res 5:2414, 1990. K Suzuki, M Ichihara, S Takeuchi, N Ohtake, M Yoshikawa, K Hirabayashi, N Kurihara. Philos Mag 65:657, 1992.

52 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57.

Balk and Heiderhoff LD Marks, DJ Smith. Nature 303:316, 1989. S Evans, JM Thomas. Proc R Soc Lond A353:103, 1977. PG Lurie, JM Wilson. Surf Sci 65:453, 1977. TE Derry, RW Fearick, JP Sellschop. Nucl Instrum Methods 170:407, 1980. G Braunstein, R Kalish. Appl Phys Lett 38:416, 1981. TJ Whetton, AA Armstead, TA Grzybowski, AL Ruoff. J Vac Sci Technol A2: 477, 1984. JF Prins, TE Derry, JPF Sellschop. Phys Rev B 34:8870, 1986. S Evans. The Properties of Natural and Synthetic Diamond. London: Academic Press, 1992. R Kalish. Diamond Relat Mater 2:621, 1993. A Boudina, E Fitzer, G Wahl, H Estrom. Diamond Relat Mater 2:678, 1993. PK Bachmann, D Leers, DU Wiechert. Diamond Relat Mater 2:683, 1993. J Ullmann, A Weber, B Mainz, J Stiegler, T Schuhrke. Diamond Relat Mater 3:663, 1994. VS Varichenko, AM Zaitzev, MS Rusetskij, VF Stelmakh, K de Weldige, T Fries, K Wandelt, AJ Didyk, VA Laptev. Diamond Relat Mater 3:711, 1994. S Matsumoto, Y Sato, N Setaka. Carbon 19:323, 1981. WS Yang, J Sokolov, F Jona, PM Marcus. Solid State Commun 41:191, 1982. BB Pate, Surf Sci 165:83, 1986. TE Derry, L Smit, JF Van der Veen. Surf Sci 167:502, 1986. E Sowa, GD Kubiak, RH Stulen, MA Van Hove. J Vac Sci Technol A6:832, 1988. JJ Lander, J Morrison. Surf Sci 4:913, 1988. RE Thomas, RA Rudder, RJ Markunas. J Vac Sci Technol A10:2451, 1992. T Ando, T Aizawa, K Yamamoto, Y Sato, M Kamo. Diamond Relat Mater 3: 975, 1994. AR Lang. Diamond Relat Mater 2:106, 1993. M Moore, R Wagget, W Wierzchowski. Diamond Relat Mater 2:115, 1993. ZH Mai, S Mardix, AR Lang. J Appl Crystallogr 13:180, 1980. AR Lang, M Moore, JC Walmsley. The Properties of Natural and Synthetic Diamond. London: Academic Press, 1992. X Jiang, CP Klages, R Zachai, M Hartweg, PJ Ellis. J Mater Res 8:3438, 1993. A Badzian, T Badzian. Diamond Relat Mater 2:147, 1993. C Wild, P Koidl, W Mu¨ller-Sebert, H Walcher, R Kohl, N Herres, R Locher, R Samlenski, R Brenn. Diamond Relat Mater 2:158, 1993. M Schreck, R Hesssmer, S Geier, B Rauschenbach, B Stritzker. Diamond Relat Mater 3:510, 1994. BA Fox, BR Stoner, DM Malta, PJ Ellis. Diamond Relat Mater 3:382, 1994. C Wild, R Kohl, N Herres, W Mu¨ller-Sebert, P Koidl. Diamond Relat Mater 3:373, 1994. TW Capehart, TA Perry, CB Beetz, DN Belton, GB Fisher, CE Beall, BN Yates, JW Taylor. Appl Phys Lett 55:957, 1989. R Meilunas, MS Wang, KC Sheng, RPH Chang, RP Van Duyne. Appl Phys Lett 54:2204, 1989.

Film Characterization Methods 58. 59. 60. 61. 62. 63. 64. 65. 66.

67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84.

53

FX Lu, GS Jiang, BX Yang, J Chen, JJ Wang, Y Yuan. Diamond Relat Mater 2:575, 1993. T Ando, T Aizawa, K Yamomoto, Y Sato, M Kamo. Diamond Relat Mater 3:975, 1994. MC Polo, J Cifre, J Esteve. Diamond Relat Mater 3:492, 1994. OM Kuettel, J Osterwalder, L Schlapbach, R Agostino. Diamond Relat Mater 2:548, 1993. C Wild, N Herres, P Koidl. J Appl Phys 68:973, 1990. PJ Dean, EC Lightowlers, DR Wight. Phys Rev 140:A352, 1965. SC Lawson, AT Collins, S Satoh, H Kanda. Proceedings 2nd International Conference on New Diamond Science and Technology, 1991, p 709. CD Clark, AT Collins, GS Woods. The Properties of Natural and Synthetic Diamond. London: Academic Press, 1992. R Heiderhoff, LJ Balk. Presented at the Thirteenth Pfefferkorn Conference on Luminescence, Niagara Falls, Canada, May 13–18, 1994; submitted to Scanning Microsc Int, 1994. G Davies. The Properties of Natural and Synthetic Diamond. London: Academic Press, 1979. PJ Dean. Phys Rev A 139:588, 1965. R Heiderhoff, R Spitzl, M Maywald, V Raiko, J Engemann, LJ Balk. Proc SPIE 2151:59, 1994. R Heiderhoff, P Koschinski, M Maywald, LJ Balk, PK Bachmann. Diamond Relat Mater, in press. RJ Graham, F Shaapur, Y Kato, BR Stoner. Appl. Phys. Lett 65:292, 1994. AT Collins, SC Lawson. J Phys Condens Matter 1:6929, 1989. AT Collins, SH Robertson. J Mater Sci Lett 4:681, 1985. OW Ka¨ding, M Ro¨sler, R Zachai, HJ Fu¨ßer, E Matthias. Diamond Relat Mater 3:1178, 1994. PK Bachmann, H-J Hagemann, H Lade, D Leers, DU Wiechert, H Wilson, D Fournier, K Plamann. Diamond Relat Mater, in press. TR Anthony, WF Banholzer, JF Fleischer, L Wei, PK Kuo, RL Thomas, RW Pryor. Phys Rev B 42:1104, 1990. JR Olson, RO Pohl, JW Vandersande, A Zoltan, TR Anthony, WF Banholzer. Phys Rev B 47:14859, 1993. L Wei, PK Kuo, RL Thomas, TR Anthony, WF Banholzer. Phys Rev Lett 70: 3764, 1993. G Lu, A Feldmann, J Graebner, O Ka¨ding, PK Kuo, A Maesano, RP Miller, D Pickrell, S Preston, D Slutz. Diamond Relat Mater, in press. D Fournier, K Plamann. Diamond Relat Mater, in press. AC Boccara, D Fournier, J Baoz. Appl Phys Lett 36:130, 1980. KF Tuner, BR Stoner, L Bergmann, JT Glass, RJ Nemanich. J Appl Phys 69: 6400, 1991. MP Everson, MA Tamor. J Vac Sci Technol B9:1570, 1991. H-G Busmann, H Sprang, IV Hertel, W Zimmermann-Edling, H-J Gu¨ntherrodt. Appl Phys Lett 59, 1991.

54 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101.

Balk and Heiderhoff ME Welland, AW McKinnon, S O’Shea, GAJ Amaratunga. Diamond Relat Mater 1:529, 1992. H Sasaki, H Kawarada. Jpn J Appl Phys 32:L1771, 1993. VS Varichenko, AM Zaitsev, MS Rusetskii, VF Stelmakh, K de Weldige, T Fries, K Wandelt, AJ Didyk, VA Laptev. Diamond Relat Mater 3:711, 1994. L Vazques, O Sanchez, F Messeguer, JM Albella. Diamond Relat Mater 3: 715, 1994. LF Sutcu, MS Thompson, CJ Chu, RH Hauge, JL Margrave, MP D’Evelyn. Appl. Phys Lett 60:1685, 1992. LF Sutcu, CJ Chu, MS Thomson, RH Hauge, JL Margrave, MP D’Evelyn. J Appl Phys 71:5930, 1992. U Bo¨gli, A Blatter, A Ba¨chli, R Lu¨thi, E Meyer. Diamond Relat Mater 2:924, 1993. K Schiffmann, X Jiang. Appl Phys A 59:17, 1994. H-J Gu¨ntherodt, R Wiesendanger. Scanning Tunneling Microscopy 1–3. Berlin: Springer-Verlag, 1992. D Sarid. Scanning Force Microscopy. Oxford: Oxford University Press, 1991. P Koschinski, K Kaufmann, LJ Balk. Electron Microsc 2b:1121, 1994. HJ Leam. J Appl Phys 53:R51, 1982. Y Huang, CC Williams. J Vac Sci Technol B12:369, 1994. C Shafai, DJ Thomson, M Simard-Normandin. J Vac Sci Technol B12:378, 1994. LJ Balk, M Maywald. Mater Sci Eng B24:203, 1994. RB Dinwiddie, R Pylkki, PE West. Therm Conduct 22:668, 1994. M Maywald, RJ Pylkki, LJ Balk. J Scan Microsc 2:181, 1994.

4 Deposition Chemistry: Deposition Pathways, Nucleation, and Growth David S. Dandy Colorado State University, Fort Collins, Colorado

Michael E. Coltrin Sandia National Laboratories, Albuquerque, New Mexico

I.

INTRODUCTION

The unique properties of diamond, such as hardness and high thermal conductivity, make it an important new material in a wide range of applications, for example, protective coatings and thermal management. However, the high cost of material production has limited the commercial use of diamond thin films to a few, very specialized applications. In chemical vapor deposition (CVD) of diamond, the factors driving cost include low reagent utilization, low deposition rates, high energy consumption, large thermal management loads at the substrate, and capital equipment costs. For these reasons, there has been much research on the CVD diamond process, including both work on empirical process optimization and work on understanding the fundamental steps in the process. The latter, more basic research is done in the expectation that knowledge of the underlying chemistry, physics, transport, and materials issues will enable optimization (and cost reduction) through model-based process design. In this chapter a review is presented of the current understanding of diamond CVD growth mechanisms. Because diamond deposition depends on different chemical and transport processes occurring in the gas phase and on the surface, discussion of these mechanisms is separated into sections that focus on gas-phase processes, surface chemistry, and nucleation phe55

56

Dandy and Coltrin

nomena. Discussion has also been included on experimental and theoretical studies of diamond deposition to illustrate the insight gained from macroscopic measurements and predictions. Absent from this chapter is a discussion of alternative chemistries in diamond CVD. For example, the addition of oxygen in noncombustion systems to enhance growth rate and film quality and the use of fluorinated hydrocarbons to achieve deposition at lower substrate temperatures are both important areas of study, but these do not change the basic mechanisms controlling the growth of diamond by CVD.

II.

GAS-PHASE PROCESSES IN CVD DIAMOND

All diamond CVD processes have in common a highly energetic activation stage in the gas phase. This stage typically serves two purposes, to dissociate the hydrocarbon precursor molecule into fragments that react more readily at the deposition surface and to dissociate molecular hydrogen to create a superequilibrium concentration of gas-phase hydrogen atoms. Four commonly used diamond CVD reactors to achieve the activation are hot-filament reactors, microwave plasma reactors, dc arcjet reactors, and combustionsynthesis reactors. Although these deposition systems vary greatly in many engineering aspects, they have many important features in common, which is why each is able to produce high-quality diamond films. A large amount of energy, in the form of electrical or chemical free energy, is input to achieve dissociation of molecular hydrogen and the hydrocarbon feedstock. Moderately low pressures (in the range 10 torr to 1 atm) are used to prevent three-body recombination of H to form molecular hydrogen. High gas-phase temperatures, greater than 1700⬚C, are produced in the activation zone, and passive or active cooling is employed to maintain a substrate temperature in the neighborhood of 925⬚C. However, transport processes are quite different among the four reactor types. Hot-filament and microwave plasma reactors are diffusion dominated; typically, there is no thermal, velocity, or concentration boundary layer. Thus, one often finds linear gradients in temperature, velocity, or species concentration between the excitation region (hot filament or plasma ball) and the deposition surface in these reactors. Growth rates and other observables are a weak function of input flow rate (or velocity). On the other hand, arcjet and combustion CVD reactors are characterized by high velocities, for example, greater than 105 cm/sec, and are thus convection dominated. Thin boundary layers in temperature, velocity, and concentration are formed near the growth surface.

Deposition Chemistry

A.

57

Gas-Phase Chemical Kinetics

The gas-phase chemistry occurring in CVD diamond synthesis involves reactions of small hydrocarbons and their dissociation fragments. The chemical kinetics of C, H, and O species has been the subject of decades of experimental, theoretical, and modeling research in the combustion community [1]. Thus, the gas-phase chemistry aspects of diamond growth are probably better understood than for any other CVD system, with perhaps the exception of the CVD of Si from SiH4. Temperature- and pressure-dependent rate constants for a great number of the elementary gas-phase reactions have been measured experimentally and provide the direct input to CVD diamond kinetic models. However, in some cases rate constants for relevant reactions are not available. Sometimes use of a measured rate constant in a calculation, for example, gives a wrong prediction of flame speed or temperature. This may be because the measured rate constant itself is in error or may be due to other unknown sources of error in the simulation. The combustion research community has developed a number of reaction mechanism sets that have been tested and compared with experiment [2–8]. Usually some individual rate constants in a mechanism set have been adjusted such that the overall mechanism as a whole gives optimal predictions for certain types of flames. Hence, it can be dangerous to extract individual reactions from one mechanism and incorporate them into another without a thorough understanding of the limitations and validity of the particular numbers. Therefore, in searching for gas-phase reaction kinetics to use in a simulation of a CVD diamond experiment, it is best to consider published reaction sets ‘‘as a whole’’ and not mix and match information from different mechanism sets. The reaction mechanism of Miller and Bowman [4] was developed to describe nonsooting, oxygen-acetylene flames. It contains about 150 reactions of 51 hydrocarbons up to C2 species and is adequate for many diamond CVD applications in which the gas is rich in H atoms, and thus formation of higher hydrocarbon species is not very important [9]. A near-sooting mechanism was created by Miller and Melius [6] that contains approximately 220 reactions among 49 chemical species. The mechanism includes reactions and species leading to the formation of benzene; it is commonly assumed that benzene is a key precursor in the formation of polyaromatic hydrocarbons and soot. The Miller-Melius mechanism may be appropriate for modeling a combustion-synthesis diamond CVD experiment under fuelrich conditions; however, extremely fuel-rich conditions usually lead to diamond films of poor quality and as such are of less practical importance. A very large reaction mechanism (853 reactions among 190 chemical species)

58

Dandy and Coltrin

including hydrocarbons up to C8H18 was developed by Westbrook and Pitz [3,7]. Recently, a consortium of researchers from several different institutions has joined to produce validated combustion reaction mechanisms under the sponsorship of the Gas Research Institute. At the time of this writing, the latest version of the reaction set [8], GRI-Mech 2.11, contained 276 reactions of 49 species containing the elements C, H, O, and N. Wolden et al. [10] presented a brute-force sensitivity analysis of two different gas-phase reaction mechanisms, the 89-reaction mechanism of Frenklach and Wang [11] and the 24-reaction mechanism of Harris et al. [12,13], to develop a reduced gas-phase reaction mechanism. Wolden et al. used a zero-dimensional model to integrate the kinetics at a temperature of 2000 K and 20 torr for 0.1 sec; these particular conditions were chosen to mimic the conditions encountered by the gas mixture flowing past a hot filament [14,15]. Both of the starting reaction mechanisms reduced to the same set of nine (reversible) gas-phase reactions, which did a good job matching the time evolution of the C2H2, CH3, and H concentrations predicted by the longer reaction mechanisms. These reactions fall into three broad groups: (1) the three-body recombination of atomic H, (2) interconversion reactions between CH4 and CH3, and (3) methyl-methyl recombination reactions, followed by hydrogen-stripping reactions to form acetylene. A key role of hydrogen in the gas-phase chemistry of diamond CVD is to suppress formation of aromatic species [9]. Minimizing production of aromatics is assumed to inhibit formation and growth of graphitic phases on the deposition surface and thus improve diamond film quality.

B.

Reactions at a Hot Filament

Langmuir [16–18] first discovered the catalytic dissociation of hydrogen on a hot tungsten filament, and such a hot filament has proved to be an easily constructed, efficient H-atom source. Matsumoto et al. [19,20] first proposed use of hot filaments for diamond CVD. The majority of basic research has been done on hot-filament reactors because they can be assembled rather inexpensively on the laboratory scale, yet can produce good-quality diamond films. A thorough investigation of H2 dissociation at a hot filament was conducted by Jansen et al. [21], who used the difference in power consumption by the filament in a vacuum and in hydrogen as a measure of the hydrogen dissociation rate. The dissociation rate was shown to depend on the geometry of the heater element. Relatively high dissociation rates, normalized per heater area, were obtained for small-diameter wires, and it was argued that this indicates a nonequilibrium dissociation process. A mathematical analysis

Deposition Chemistry

59

of the H production rate was presented that depended upon parameters such as the diffusion coefficients, wire diameter, and dissociation rate constant. Meier et al. [22] used two-photon laser-induced fluorescence (LIF) to determine H-atom concentrations relative to a hot filament. In this experiment there was no substrate, and thus no diamond growth was occurring. They measured [H] as a function of total pressure, distance from the filament, filament temperature, filament diameter, filament material, and CH4 mole fraction. They found little effect on [H] due to addition of up to 5% CH4. However, in other studies, tungsten filaments have been found to carburize during diamond CVD upon high exposure to hydrocarbon feedstock [23]. Such carburization degrades the catalytic activity for hydrogen dissociation, and gas-phase H-atom concentrations have been found to drop by at least an order of magnitude when the methane concentration is increased between a factor of 3 and 10 [15,24–27]. As discussed in Sec. VI.A, Dandy and Coltrin [28] found that inclusion of filament poisoning effects was necessary to model the gas-phase chemistry in a hot-filament reactor because the H-atom concentration controls the kinetics to such a large degree. C.

Plasma Activation of the Gas Phase

Direct-current (dc) arcjet-assisted diamond CVD can attain very high growth rates [29–31] because of the high arcjet powers and material flow rates possible in these systems. Within the arcjet, electrical energy is converted to chemical free energy through dissociation of hydrogen. The degree of hydrogen dissociation is typically high, 20% or greater, but not known precisely. Detailed physical models of arcjet operation have not been included in diamond CVD models, although Dandy and Coltrin [32] gave a simple thermodynamic model of an arcjet that predicts the hydrogen dissociation fraction for a specified input power and operational temperature, as illustrated in Fig. 1. Temperatures at the exit of the arcjet can be in the range 1000–4000 K [33–35]. The combination of high temperature and large concentration of atomic H produces high reaction rates of the reactant gases in the free stream below the arcjet. Hydrocarbon feedstock is usually injected downstream from the exit of the arcjet. When methane is the reactant gas, the C1 species, i.e., CHx (x = 0–4) equilibrate very rapidly [36,37], and thus the concentration of methane is very low in the free stream. After equilibration (within a few cm of the exit of the arcjet) the C1 manifold of species remains essentially frozen in the convective free stream until the boundary layer is encountered near the substrate. In contrast, acetylene is much more stable than methane, and when acetylene is used as the reactant gas it undergoes significantly less gas-phase decomposition. Figure 2 shows the time evolution of species con-

60

Dandy and Coltrin

Figure 1 Predicted H-atom mole fraction exiting the arcjet and power required as a function of the arcjet operating temperature. (From Ref. 32.)

centrations from homogeneous chemical kinetics calculations [37] under conditions typical of the free stream beneath the exit of an arcjet, that is, 5000 K, 3 torr, 33% dissociated H2. One sees significant decomposition of methane within a few ␮sec in Fig. 2a, whereas there is much less decomposition when acetylene is the reactant in Fig. 2b. In a microwave plasma system, energy from the microwave electric field ionizes the gas, which is primarily H2 with small amounts of hydrocarbon. Energy is subsequently transferred from the electrons to vibrational levels in H2; then this vibrational energy serves to heat the gas through vibration-to-translation energy transfer to atomic H and vibration-to-vibration or vibration-to-translation energy transfer to other H2 molecules [38]. Ion chemistry is important in these systems as an electron-loss source. Electrons can dissociate H2 and eventually lead to a concentration of H⫹ ions, which can subsequently undergo charge exchange with hydrocarbon and

> Figure 2 Kinetics simulation of gas-phase chemistry under dc arcjet conditions for (a) 0.5% methane and (b) 0.5% acetylene as the hydrocarbon reactant. (From Ref. 37.)

Deposition Chemistry

61

62

Dandy and Coltrin

oxygen-containing species. Neutral chemistry in the gas phase is also initiated by electron-impact dissociation of H2, O2, and hydrocarbons [38].

III.

SURFACE CHEMISTRY IN CVD DIAMOND

A.

Calculation of Diamond Surface Structures and Energetics

Modern theoretical chemistry techniques are well suited to study the bonding and surface structures relevant to diamond CVD because the structures mainly consist of relatively light elements and thus few electrons, i.e., carbon and hydrogen, and because bonding is generally short range and covalent, so relatively small ‘‘slabs’’ can be used to model the lattice. A wide array of theoretical papers have appeared that report stabilities of proposed surface structures, relative bond strengths among competing adsorbate species, reaction barriers, and the like. Mehandru and Anderson [39] used atom-superposition and electrondelocalized molecular orbitals (ASED-MOs) to study adsorption energetics and bonding of CH3, CH2, CH, C2H, and C2H2 on clean and hydrogenated {111} diamond faces. On either the clean or hydrogenated surface, the adsorption bond dissociation energy (BDE) varied as C2H > CH ⬵ CH2 > CH3. The 1:1 (complete monolayer) coverage of CH3 was concluded to be very unstable due to steric repulsions (⫺2.3 eV BDE); the 1:1 coverages of CH2, CH, and C2H were all predicted to be stable. The favored bonding site for C2H2 on a clean {111} surface was di-␴ bridging. Bonding to the surface via a single ␴ bond was found to be very weak (1.2 eV) compared with the di-␴ bond (3.5 eV). Mehandru et al. [40] also used this technique to study hydrogen diffusion within the bulk diamond lattice. They found that a bond-centered (BC) site, that is, a C — H — C structure, was more stable than other interstitial sites (but still about 1.7 eV less stable than a gas-phase H atom plus the cluster). A minimum barrier to H migration (1.9 eV) was found for moving from one BC site to a neighboring one, moving in the {110} plane. They also examined the stability of from one to four H atoms bound to a lattice vacancy. Up to four H atoms could be bound in the vacancy, with the calculated BDE energies decreasing monotonically (5.3, 4.4, 3.6, 2.5 eV) as the number of H atoms increases. Alfonso et al. [41] used a combined density-functional/molecular dynamics simulation to calculate adsorption energies and minimum energy configurations for various hydrocarbons on flat {100}, flat {111}, and stepped {100} surfaces. For onefold adsorption sites, the bonding energies were ordered as follows: C2H > CH2 > CH3 > C2H2; stable twofold adsorp-

Deposition Chemistry

63

tion sites were found for C2H2 on terrace sites and near steps on the {100} surface. Bonding to {111} surfaces was weaker than to the other two surfaces studied. Alfonso et al. [42] studied the adsorption of various hydrocarbons on a diamond {100}-(2⫻1) surface. They used molecular dynamics and a dynamical quenching technique [43] to calculate stable adsorption binding configurations based on the many-body potential energy function of Brenner [44]. Bonding was studied on a number of different types of sites: at the end of a dimer (having removed an H from the dimer), bridging across the two carbons in a broken dimer (bridge sites), and bridging between the two ends of adjacent dimer pairs (troughs). For the flat {100}-(2⫻1) surface they calculated BDEs CH2 (4.67 eV) > C2H (4.34) > CH3 (4.18). Acetylene was bound considerably tighter at troughs (6.54 eV) than at bridge sites (3.62 eV), but the authors concluded that trough sites are much less abundant than bridge sites under steady-state conditions. The stability of hydrocarbon fragments near step edges was concluded to be very near the values calculated on flat terraces because of the strong covalent (and short-range) nature of the bonding. Angus and coworkers [45] studied nucleation of the diamond {111} plane upon the {0001} plane of graphite using energy minimization of a semiclassical Tersoff potential [46,47]. They found very favorable energetics for an interface matching three diamond {111} planes for every two graphite {0001} planes. A mechanism was proposed that starts with graphitic or aromatic structures, which are then hydrogenated to provide nucleation sites for diamond growth. Molecular mechanics MM2 [48] and MM3 [49] calculations were used by Harris and Goodwin [50] to calculate the free energy change in each step of their detailed reaction mechanism on the reconstructed {100}-(2⫻1):H surface, which in turn was used to calculate the reaction equilibrium constant and reverse rate constant. Yang and D’Evelyn [51] used MM3 to examine the relative stabilities of the {100}-(2⫻1) reconstructed surface with varying degrees of hydrogenation. The monohydride phase was found to be the most stable and was concluded to be the dominant surface phase under CVD conditions. MM3 was also used to examine the energetics of each step in a proposed growth mechanism on the {100}-(2⫻1) surface [52]. Huang and Frenklach [53] used MNDO [54] to calculate reaction energetics on the {100}-(2⫻1) reconstruction. They found a barrier of 37 kcal for H2 elimination from a surface dihydride to form the monohydride and that the dihydride is approximately 80 kcal/mol higher in energy than the monohydride plus H2(g). An alternative mechanism involving H-atom abstraction from the dihydride forming a dihydride surface radical, then H elimination and C — C bond formation, had a barrier of 49 kcal/mol, i.e., 12

64

Dandy and Coltrin

kcal/mol higher than the other channel. They calculated a barrier of 80 kcal/ mol for addition of CH3 to a surface dihydride radical; carbon atoms were found to add to a monohydride radical site with less than a 20 kcal barrier, and acetylene addition to a radical site proceeded with a barrier of 41 kcal/ mol. Brenner [44] developed a potential energy surface for diamond and studied adsorption of hydrocarbon fragments on the {111} surface. The potential uses many-body terms based on Tersoff bond-order expressions [55]. Alfonso and Ulloa [56] performed molecular dynamics calculations employing the Brenner [44] intermolecular potential to study the adsorption of methyl radicals on the diamond {100} surface. The adsorption probability was found to increase with CH3 incident kinetic energy and for normal incidence. Methyl radical did not react with a fully hydrogenated surface. Thus, a surface radical site was required for CH3 chemisorption. The same intermolecular potential was used by Garrison et al. [57], who discovered a low-energy reaction pathway for dimer-bond breaking and addition of CHx (x < 3) on the {100}-(2⫻1):H surface from analysis of their molecular dynamics simulations. B.

H-Abstraction/Radical Termination Reactions

Reactions of atomic hydrogen are a dominant factor in the surface chemistry of diamond CVD. They not only control the gas-phase chemistry, and thus the nature of the reactive species reaching the growth surface, but also determine the availability of reactive sites upon the surface. In addition, the recombination of hydrogen on the surface is an important source of heat in the system and must be accounted for in substrate thermal management. During deposition the diamond surface is usually hydrogen terminated. Growth is initiated by abstraction of an H from the surface, creating a reactive radical site H(g) ⫹ CH(s) ⇔ H2(g) ⫹ C*(s)

(1)

This reaction is accompanied by the radical-termination reaction H(g) ⫹ C*(s) ⇔ CH(s)

(2)

The net of these two reactions is heterogeneous recombination of atomic hydrogen to form H2. Although reactions (1) and (2) are written as reversible, for all practical purposes the reverse reaction rates are negligible. The abstraction reaction (1) is fast but does not occur with unit probability. The rate constants for gas-surface reactions in diamond CVD mechanisms are usually estimated by analogy with gas-phase reactions. This, in essence, assumes that the diamond surface behaves like a giant hydrocarbon

Deposition Chemistry

65

[58]. The gas-phase rate constant for abstraction of a tertiary hydrogen from an alkane is (in cm3 mol⫺1 sec⫺1) k1 = 1.3 ⫻ 1014 exp(⫺3674/T)

(3)

with the activation energy in K. The rate constant can be converted to a reaction probability ␥ using k=

␥ ⌫

冑

RT 2␲W

(4)

where ⌫ is the surface site density (taken here to be 3 ⫻ 10⫺9 mol cm⫺3) and W is the atomic weight of H; for reactions of H atoms,

␥ = 8.25 ⫻ 10⫺13

k

兹T

(5)

Thus, the rate constant in Eq. (3) corresponds to a reaction probability of 0.14 at a typical growth temperature of 1200 K. Kinetic modeling studies of diamond CVD have used values of ␥1 spanning the range 0.22 [59] down to 0.0037 [11]. Molecular dynamics calculations of Brenner et al. [60] yield an abstraction reaction probability of 0.04 at 1200 K. A reasonable analogue of reaction (2) would be the gas-phase reaction of H with the isopropyl radical, with a rate constant in the high-pressure limit of 2 ⫻ 1013 cm3 mol⫺1 sec⫺1 [2]. This rate constant corresponds to a probability of 0.48 at 1200 K. Previous models have used values of ␥2 ranging from essentially unity [59] to 0.016 [11]. Brenner et al. [60] calculated a value of 0.43 for ␥2 at 1200 K. Although there is quite a range in the probabilities that modeling studies have employed for reactions (1) and (2), there has been a good consensus among such models that the ratio ␥2/ ␥1 should be in the range of 3 to 4 [11,59,61]. This ratio was predicted to be approximately 11 via molecular dynamics [60]. A steady-state analysis [62] shows that the probabilities for reactions (1) and (2) can be combined to yield an effective probability, ␥H, for H-H recombination, 1 1 = ␥H 2

冉

冊

1 1 ⫹ ␥1 ␥2

(6)

Using the probabilities deduced from the gas-phase analogues in Eq. (6) yields an effective recombination probability ␥H = 0.22. Krasnoperov et al. [63] measured an effective H-atom recombination probability of 0.16; Harris and Weiner [64] measured a slightly smaller value, ␥H = 0.12, with about a factor of 2 uncertainty. Thus, we conclude that the rate constants (or equivalently) probabilities for H-atom abstraction and termination reactions de-

66

Dandy and Coltrin

rived from their gas-phase analogues are reasonable but perhaps are as much as a factor of 2 too large. C.

Growth Chemistry

The surface chemical reaction mechanism leading to diamond growth has been the subject of much research, debate, and controversy. It was widely assumed that the ‘‘correct’’ reaction mechanism could be found because diamond is a relatively simple material system: it is dominated by shortrange covalent bonds, chemically related to well-known hydrocarbon chemistry, and is amenable to theoretical or first-principles calculation of energetics and reaction rates. However, a universally accepted diamond growth reaction mechanism has still not yet been found. As an example, much controversy has reigned about the identification of the dominant precursor species leading to growth. In a hot-filament reactor environment, the most abundant gas-phase hydrocarbon species are C2H2 and CH3, and attention focused on these two candidates as the dominant growth species. Many conflicting experiments have concluded that one species or another could or could not account for diamond growth. It is unlikely that there is a single, simple diamond growth mechanism that applies to all deposition conditions. For example, the gas-phase environment present in a dc arcjet reactor is very different from that in a hotfilament reactor; there is no a priori reason to expect that the same growth mechanism will be operative under such different conditions. The first proposal of C2H2 as a growth species was by Frenklach and Spear [58]. A detailed kinetic reaction mechanism for diamond growth was proposed by Frenklach and Wang [11]. Their model was applied primarily to a hot-filament environment and included acetylene as the dominant growth species. However, shortcomings of the acetylene addition mechanisms were discovered [65]. Addition of acetylene go a monoradical surface site is thermodynamically unstable: the adduct formed has too short a lifetime to allow incorporation into the diamond lattice. Addition of acetylene to a biradical site on the {111} or {110} surface forms a thermodynamically stable molecule. However, the admolecule quickly desorbs after reaction with a gas-phase H atom, followed by a ␤-scission cleavage of its surface C — C bond [66]. Alternative reaction schemes for acetylene on the dimerized {110} surface have been proposed that do not suffer from the foregoing deficiencies [65,66]. Cappelli and Loh [67] conducted experiments in an arcjet reactor with C2H2 feed gas and reported high-quality diamond growth. Under these conditions, the concentration of C1 species was too low to account for the observed growth. They concluded that C2H2 was the dominant growth spe-

Deposition Chemistry

67

cies in this experiment. In a different growth environment, the opposite conclusion was drawn. Chu et al. [68,69] used a mixture of 13C-labeled methane and 12C acetylene for growth rate experiments in a hot-filament reactor. They measured homoepitaxial growth rates on different crystal faces of 0.4 ␮m/hr on the {100} face, 0.5 ␮m/hr on the {111} face, and 1.3 ␮m/ hr on the {110} face. The 13C fraction in the deposited film matched that of the gas-phase methyl radical (as deduced from mass spectrometry probe sampling) and differed significantly from the 13C fraction in the acetylene. From this observation they concluded that CH3 was the dominant growth species in the hot-filament experiment. A number of other experimental papers also concluded that C2H2 is not a viable growth species [69–73]. There has not been a well-accepted mechanism proposed for growth on {111} planes [74,75]. Steric repulsion appears to rule out complete coverage of the {111} plane by methyl radicals [39]. There is only moderate repulsion between just a pair of adjacent surface CH3 groups on the {111} plane [39,76,77], and such a configuration has been examined as a growth route. Atomic hydrogen abstraction from a single one of the adjacent methyls followed by bridging reactions is limited by a high activation barrier, approximately 50 kcal/mol [77]. Abstraction of an H from both adjacent methyls would be followed by rapid formation a C — C bond and thus a surface species such as — CH2 — CH2 — . However, subsequent abstraction of an H from this species will be followed quickly by ␤-hydride elimination of H and desorption of the — CH — —CH2 group [74]. Such a series of reactions has made drafting of a complete growth mechanism on the {111} and {110} planes difficult to date. Frenklach et al. [74] have proposed an atomistic model for step-flow growth upon a {100} facet between two {111} planes. The model is based on bridging methylene, CH2, as the mobile species. However, instead of migrating via surface diffusion, CH2 moves through a sequence of covalent bond breaking and formation. Butler and Woodin [78] propose a detailed reaction sequence for growth of diamond from CH3 (used as a ‘‘generic’’ growth species) on the {110} crystal face. In particular, their mechanism focuses upon reactions among chemisorbed species. Kinetic rate constants for the individual steps in the mechanism were not given, and thus it has not been further incorporated into numerical kinetic or reactor-scale models. Harris [59] proposed a widely used mechanism for growth on the unreconstructed {100} plane with CH3 as the growth species. Rate constants for the 12 reactions in the mechanism were mainly deduced from gas-phase analogues. Thermochemistry for each reaction, i.e., ⌬G, and thus the equilibrium constants and reverse rate constants, was estimated using the bicyclo[3.3.1] nonane molecule as a model of the diamond surface. Steps in this

68

Dandy and Coltrin

mechanism are illustrated in Fig. 3. Methyl radical (denoted by ‘‘M’’ in the figure) was assumed to adsorb on a surface radical site, followed by H abstraction from both the adsorbed methyl (‘‘M*’’) and an adjacent surface H and rapid formation of a C — C bond (‘‘biradical pairing’’). The Harris mechanism was used quite widely to simulate CVD diamond growth in a broad range of growth conditions, including hot filaments, oxygen-acetylene flame, dc arcjet, and rf torch reactors; see Ref. 50 for a summary. Steric crowding makes it very unlikely that the methyl radical can actually bond to the unreconstructed {100} surface [53], and so the original mechanism as proposed by Harris is probably not correct in all of its detail. In fact, it is unlikely that the unreconstructed {100} surface exists at all under growth conditions, due to steric repulsion between adjacent H atoms. The {100}-(2⫻1) reconstruction has been observed experimentally [79– 82], and the stability of this surface has been examined theoretically [51].

Figure 3 Steps in the methyl-addition growth mechanism of Harris [59], which uses the bicyclo [3.3.1] nonane molecule as a model of the unreconstructed {100} surface.

Deposition Chemistry

69

Harris and Goodwin [50] proposed a growth mechanism on the {100}(2⫻1):H (hydrogen-covered reconstruction) surface, illustrated in Fig. 4. The mechanism consists of two main parts: insertion into dimer bonds according to the route proposed by Garrison et al. [57], shown in Fig. 4b and c, and bridging across the trough between the ends of two dimers, as in Fig. 4f. The latter part of this mechanism basically follows the steps in Harris’ original methyl addition mechanism [59]. Harris and Goodwin give rate constants for each step in their mechanism and the corresponding thermochemistry, calculated via the molecular mechanics code MM3 [49]. Kinetic simulations showed that the trough portion of the mechanism was rate limiting because that portion required more H abstraction steps than the dimer

Figure 4 Steps in the mechanism proposed by Harris and Goodwin [50] for growth on the {100}-(2⫻1):H surface. In parts (a), (c), and (e), the letters A–D represent either an H or a surface radical site at which reaction can take place.

70

Dandy and Coltrin

addition portion. This mechanism was used by Harris and Weiner [83] in simulating the pressure and temperature dependence of diamond growth. There is experimental evidence that surface diffusion is operative on the diamond surface at high temperatures [84–87]. Theoretical analysis of surface diffusion of H atoms, hydrocarbon species, and radical sites found these processes to be very facile [88]. Surface diffusion will be enhanced at higher temperatures, potentially leading to smoother films. However, higher temperatures also lead to increased desorption from the surface and etching. The surface diffusion mechanism was also found to be unstable; the system was found to switch suddenly from a regime of smooth growth to formation of pits. D.

Reduced Reaction Mechanisms

The fine details of diamond growth mechanisms are very complex, still under debate, and dependent upon crystallographic plane and growth condition. However, many general aspects of diamond growth surface chemistry appear to be generic to the process and are widely agreed upon. As a result, a number of simplified models of diamond growth have been developed [32,62,67,78,89]. Although not intended to be mechanistically rigorous in detail, these models attempt to explain some of the overall behavior of the diamond CVD system, e.g., scaling of growth rate, surface coverage, defect formation, energy consumption, in terms of generic reactions taking place. Such models are typified by the discussion given by Butler and Woodin [78], whose simple analysis included reactions for H-atom abstraction and radical termination, incorporation of chemisorbed hydrocarbon species into the lattice, diamond growth, and parasitic growth of defects. This model is perhaps distinguished from some of the others in discussing the surface temperature dependence of these processes. One of the quantities predicted by the model is quality, which was defined as the ratio of the rate at which adsorbed carbons completed their lattice bonds before being overgrown to the total rate of carbon adsorption. Figure 5 illustrates the dependence of quality on substrate temperature for two different reactor types. Case (a) is representative of a dc arcjet system, for which the atomic hydrogen mole fraction at the surface was assumed to be 0.1 and the mole fractions of hydrocarbons were assumed to total 10⫺4. Case (b) represents conditions in a hot-filament reactor, with a near-substrate H mole fraction of 0.01 and CxHy mole fractions of 10⫺5. Goodwin [62] presented a simplified surface reaction mechanism that includes many of the important features present in more detailed kinetic schemes, including H-atom abstraction and termination reactions, addition of CH3 to a surface radical site, and a biradical pairing reaction to form a

Deposition Chemistry

71

Figure 5 Calculated quality (lattice incorporation/carbon addition) as a function of substrate temperature for conditions characteristic of (a) dc arcjet and (b) hot filament. (From the model of Butler and Woodin [78].)

surface C — C bond. The mechanism also included an alternative reaction path for formation of a buried defect, for example, hydrogen or sp 2 species trapped in the lattice. The simplified reaction scheme can be solved analytically, yielding expressions for deposition rate and lattice-defect fraction as a function of CH3 and H concentrations at the surface. The deposition rate expression containing two adjusted constants reproduces measured diamond growth rates from a wide range of experiments spanning growth rates from 0.1 to more than 7000 ␮m/hr and is G = 1.8 ⫻ 1011

[CH3][H] 5 ⫻ 10⫺9 ⫹ [H]

(7)

where G is the growth rate in ␮m/hr, and the concentrations are in mol/cm3. Assumptions in the analysis lead to the conclusion that the mole fraction of defects in the grown film, Xdef , will scale as Xdef ⬀

G [H]2

(8)

The form of Eq. (8) is especially important in trying to optimize diamond growth conditions. It illustrates the fundamental trade-off between achievable growth rate and film quality; i.e., defect density increases linearly with growth rate. Equation (8) predicts that growth rates can be increased quad-

72

Dandy and Coltrin

ratically with H-atom concentration for a fixed defect density. (This is achieved by simultaneously increasing the hydrocarbon fraction.) Goodwin points out that the scaling of defect fraction with H-atom concentration is uncertain and depends upon the form of the (unknown) defect formation reaction. He gives a more general scaling relationship for defect fraction as Xdef ⬀

G [H]n

(9)

Dandy and Coltrin [32] discussed diamond growth in terms of a simplified growth mechanism that extended the one given by Goodwin [62] to account for growth from atomic C atoms, which were quite abundant in the dc arcjet environment being modeled [36]. They also included a kinetic pathway for formation of lattice point defects due to overgrowth of surface hydrogen species and thus incorporation of hydrogen into the lattice as a defect. As shown in Fig. 6, the scaling of defect fraction was found to follow (very nearly) a linearly scaling with [H], that is,

Figure 6 Predicted growth rate as a function of gas-phase H mole fraction at the surface for a fixed diamond quality (defect concentration). Solid curve is for a defectformation reaction that is first order in surface reactant species; dashed curve is for a defect-formation reaction that is bimolecular in surface species. (From Ref. 32.)

Deposition Chemistry

Xdef ⬀

G [H]

73

(10)

when the defect formation reaction was first order with respect to surface species as reactants. When the reaction responsible for defect formation was assumed to be bimolecular with respect to surface species, a quadratic scaling as in Eq. (8) was obtained, which is also illustrated in Fig. 6. The difference in scaling laws for defect formation has important implications for scaling a CVD diamond reactor to higher arcjet power, for example. If defect formation decreases approximately linearly with [H], as in Eq. (10), higher power and thus increased H-atom production would allow a linear increase in attainable growth rate (holding diamond quality fixed). However, if defect formation scales as in Eq. (8), any increase in [H] is rewarded with a quadratic increase in the attainable growth rate (again, for fixed quality). Angus and Evans [89] present a simple reduced reaction set to describe many aspects of CVD diamond growth. Their model includes growth of diamond from graphitic nucleation sites [45], conversion of sp 2 sites to sp 3 hybridization, and gassification of hydrogenated surface species by atomic hydrogen. The reaction scheme can be solved analytically at steady state to predict growth rates of both diamond and sp 2 impurities as a function of the concentration of a growth precursor, [CHx], and atomic hydrogen, [H]. The model predicts that the impurity concentration will increase with [CHx] and thus with increasing growth rate, and it will decrease more than linearly with [H].

IV.

NUCLEATION PHENOMENA

It has become clear in research on diamond film growth that control of the fundamental phenomena associated with diamond nucleation and the early stages of growth is essential for applications in which material properties are sensitive to, or depend directly upon, the morphology of the polycrystalline film. Film anisotropy, grain size, and nano-, micro-, and macroscopic voids all strongly affect material properties such as thermal conductivity and optical transmissivity. Indeed, if diamond is ever to be used in semiconductor applications, it will be necessary to synthesize large-area, thick, single-crystal films. An increase in surface nucleation density may reduce morphological instabilities and surface roughness and, further, may improve the homogeneity of films and reduce formation of voids at the substrate or coating interface, leading to better diamond-substrate adhesion. Most early studies of the low-pressure CVD of polycrystalline diamond have focused on studying different deposition techniques and correlating

74

Dandy and Coltrin

these techniques with the material properties of the resulting films. These efforts have been successful in gaining a qualitative and semiquantitative understanding of how diamond is deposited from the gas phase. Another result of such studies has been an increasing awareness of the extreme variability in film morphology and crystallinity that could be produced with seemingly minor changes in the growth environment. Because of this observation, increasing attention has been paid to the nucleation and early growth stages of diamond; these nucleation studies have significantly contributed to understanding of diamond nucleation mechanisms in CVD. In this section possible mechanisms for diamond nucleation will be discussed, and the effect of surface conditions and the growth environment on epitaxy, morphology evolution, and texture will be considered.

A.

Nucleation Mechanisms

1.

Homogeneous Nucleation

The emphasis of most studies on nucleation and growth of diamond has been the heterogeneous formation of diamond particles and the crystallization and deposition of diamond films on substrate surfaces. Only limited work has been done to examine the possibility of achieving homogeneous nucleation in the gas phase at subatmospheric pressures. However, there is evidence that, at least in some cases, diamond may be nucleated homogeneously in the gas phase [90–93]. The primary reason that homogeneous nucleation of diamond has not been pursued more aggressively is that using a CVD process to produce diamond powder via homogeneous nucleation does not appear commercially competitive with other techniques because of the low nucleation rate; and conditions conducive to homogeneous nucleation are not favorable for film growth, so it may not even be practical to use the homogeneous nucleation phenomenon as a seeding technique. Classical nucleation theory indicates that homogeneous nucleation of diamond is possible [94], and a number of suggestions have been made regarding possible hydrocarbon cage molecules (such as adamantane) that may serve as gas-phase precursors for the diamond nuclei [95]. The adamantane molecule represents the smallest combination of carbon atoms that possesses the diamond unit structure, that is, three six-membered rings in a chair structure. However, chemical equilibrium calculations reveal that proposed precursors such as adamantane, tetracyclododecane, and hexacyclopentadecane are unstable at the temperatures and pressures typical of CVD diamond growth. Instead, Angus et al. [96,97] suggested that fully hydrogenated ring compounds structurally related to the boat-boat conformer of bicyclodecane are more plausible nucleation seeds due to their greater abun-

Deposition Chemistry

75

dance in CVD systems, thermodynamic stability, and higher reactivity with hydrocarbon radicals leading to atom addition. While dilute concentrations of 50–200 nm diamond crystals have been successfully produced via homogeneous nucleation using a number of reactant gas mixtures and reactor types [90,92,98], it has been found that higher nucleation densities may be achieved when impurities such as halogens, silane, or diborane are present in the system [99]. It has been suggested that, because of the higher rates of thermal dissociation of these impurities, the decomposition fragments quickly nucleate a relatively large number of small clusters. These clusters then serve as seeds for the subsequent addition of carbon. 2.

Heterogeneous Nucleation

It is possible to grow diamond homoepitaxially on a single-crystal diamond surface using methods such as hot-filament CVD [69,100] and microwave plasma–assisted CVD [101,102] using a variety of hydrocarbon sources. Diamond readily nucleates on cubic boron nitride (cBN) because the two materials have identical crystal structure, a lattice mismatch that is only 1.4%, and similar thermal expansivity [103,104]. However, in almost all other diamond CVD processes where diamond is grown on nondiamond substrates, the initial nucleation stage is extremely slow and nucleation density is very low unless some form of active substrate pretreatment takes place. Conventional growth of polycrystalline diamond generally consists of up to five distinguishable stages: (1) the incubation period, (2) three-dimensional nucleation of individual crystallites on the substrate surface, (3) termination of surface nucleation and subsequent three-dimensional growth of individual crystallites, (4) faceting and merging with neighboring crystallites, and (5) growth of the continuous film. These different stages are illustrated by the micrographs shown in Fig. 7. Before nucleation begins, the system undergoes an incubation period that may last from several minutes up to hours [13,105], depending upon substrate material, surface pretreatment, and deposition conditions. The lone nanocrystals formed during the nucleation stage often exhibit roughly spherical geometry. Nucleation density increases with time up to values that, again, depend upon the substrate and pretreatment method, after which surface nucleation ceases to occur at a measurable rate. The isolated crystallites grow and develop faceting due to the relatively high rate of surface carbon diffusion from the surrounding substrate surface. Once the crystals grow large enough to impinge upon one another, they form grain boundaries and then continue growing as a continuous film, as indicated by the highly textured morphology shown in Fig. 7f.

76

Dandy and Coltrin

Figure 7 Diamond growth on a nondiamond substrate occurs by sequential steps: (a) nucleation of individual crystallites; (b and c) termination of nucleation followed by growth of individual crystallites; (d) faceting and coalescence of crystallites; (e and f ) competition between growing crystallites and eventual overgrowth to form continuous film. (From Ref. 203.)

Under certain growth conditions the competitive growth process of the individual crystallites governs the subsequent growth of the continuous film. In these cases, the grains exhibit fastest growth in the direction perpendicular to the substrate, overshadowing any slower growing neighbors and forming a continuous film with a very distinct columnar structure [105]. This mode of film growth, where crystallites successfully overshadow their neighbors, has been observed in many vapor-deposited materials and is described as evolutionary selection [106]. Surface nucleation processes can be described with two quantities: the surface nucleation density, Nd (cm⫺2), and surface nucleation rate, Nr (cm⫺2 hr⫺1). The nucleation density is the number of nuclei grown per unit area, and the rate is the number grown per unit area per unit time. The density depends upon the number of activated nucleation sites available; thus, nucleation on the substrate will cease when all active sites have been occupied or when the discrete crystallites grow together and completely cover the surface. The nucleation density determines the ultimate thickness, average crystal size, homogeneity, substrate adhesion, texture, and roughness of the resulting film. Generally, the higher the nucleation density, the smoother the surface of the continuous film [107]. On an atomic scale, the surface nucleation process may include the following events: 1.

Atoms from the gas collide with the deposition surface and are adsorbed.

Deposition Chemistry

2.

3.

4.

5.

77

The adatoms may desorb back into the gas phase or diffuse along the surface; they may also diffuse into the substrate or form bonds with other surface species. As time increases, the concentration of surface adatoms also increases and clusters begin to form as the adatoms bond with one another. The clusters grow or decay in size, depending upon their thermodynamic stability, rate of atom addition from the gas phase, and surface diffusion from the surrounding substrate. Once the clusters reach a critical size, they become thermodynamically stable and will continue to grow as new atoms are adsorbed.

This model for nucleation is necessarily simplistic because it presumes an ideal substrate with no defects or existing impurities. In practical systems where pretreatment occurs, as discussed in Sec. IV.C, nucleation may take place on existing particles, at defects induced on the substrate surface, or at intermediate carbide layers of varying chemical composition. However, the general steps just outlined are adequate to describe the principal features of diamond nucleation. It was recognized fairly early in diamond CVD research that the surface nucleation process of diamond was a controlling factor in the formation and subsequent growth of continuous films [108,109]. Experiments were performed to nucleate diamond on nondiamond substrates, including singleand polycrystalline Au, Cu, Si, Mo, and W. From those studies, a number of general trends emerged regarding diamond nucleation on nondiamond substrates: 1.

2.

3.

4.

5.

Diamond nucleation rates on nondiamond substrates range from 103 to 108 cm⫺2 hr⫺1, depending upon the surface preparation method and deposition conditions. Nucleation of the diamond crystals is observed primarily on substrate defects (scratches, grain boundaries, dislocations), indicating that diamond nucleation is, indeed, occurring via a heterogeneous mechanism. Diamond nucleation rates are lower on single-crystal substrates than on polycrystalline substrates of the same material after identical surface pretreatment. Diamond nucleation rates on carbide-forming substrates (Si, Mo, W) are one to two orders of magnitude higher than on non–carbide-forming materials (Cu, Au). Nucleation rates on nondiamond substrates decrease as both substrate coverage and crystal size increase.

78

Dandy and Coltrin

A large number of experimental observations employing transmission electron microscopy, scanning electron microscopy, x-ray diffraction, x-ray photoelectron spectroscopy, low-energy electron diffraction, Rutherford backscattering spectrometry, electron energy loss spectroscopy, and Auger electron spectroscopy, to name but several techniques (for example, [110– 117]), validate the observations of Spitzyn et al. [108,109]. Together, the experimental studies of diamond nucleation and growth reveal that, in most cases, diamond does not actually nucleate directly on a nondiamond substrate surface but instead forms on an intermediate layer that develops at the interface of the substrate during the incubation period prior to actual diamond nucleation. The intermediate layer, formed due to chemical reactions between activated gas species and the substrate material, may consist of a combination of diamond-like amorphous carbon, metal carbides, or graphite, depending upon substrate pretreatment, the substrate material, and deposition conditions. It is generally agreed that the intermediate layer provides nucleation sites for diamond crystal growth and thus increases Nd on nondiamond substrates. Knowledge of the details of formation of the intermediate layers may provide a means of controlling the morphology and texture of the resulting diamond films. Diamond is often grown on materials that form refractory carbides, such as silicon, molybdenum, tantalum, and tungsten, in part because of the availability of these materials and their thermal properties (they can withstand the high temperatures and energy loads associated with diamond deposition) and in part because it was discovered that diamond would nucleate and grow on these materials with little or no pretreatment. It is now generally accepted that the reason for the ease of growth on these materials is the initial formation of a carbide layer that in turn acts as a medium for diamond nucleation. These intermediate layers are thin, often in the range of tens of nanometers, and do not strongly affect the material properties of the grown film. From observation and thermodynamic calculation [118,119], it has been suggested that diamond nucleation on Si must be preceded by the formation of a ␤-SiC buffer layer and that diamond actually nucleates on the surface of the carbide layer. This is supported by a large number of experiments in which particles or films of diamond were grown on silicon substrates. Systematic studies of diamond growth on other carbide-forming materials have also been carried out [120], and the same observation was made, namely that diamond nucleation was preceded by formation of a thin carbide layer. It has been postulated [121] that diamond growth on carbide-forming substrates is initiated by the dissolution of carbon into the substrate material, forming the stable carbide. Diamond nucleation on the carbide then occurs when the surface becomes saturated with adsorbed carbon. In this model, the diffusion of carbon into substrate and on the surface both play key roles

Deposition Chemistry

79

in the rate of diamond nucleation. It was found that qualitative features of the model agreed with experimental observation: materials with high carbon diffusivities had the longest incubation period because they required the most time to achieve a carbide layer thick enough for the surface to become saturated with carbon. However, enough experimental data exist to conclude that, although the formation of carbide interlayers is an important factor for diamond nucleation on carbide-forming substrates, it is probably not the sole mechanism by which nucleation occurs [122]. Intermediate layers of a-C:H or a-C were found to form on Si and Mo substrates, for example. Graphite interlayers have also been observed on carbide-forming substrates. What can be said is that, in general, formation of interlayers is a necessary step in the spontaneous nucleation processes of diamond on nondiamond substrates, but this alone is not sufficient for nucleation to occur. Two criteria must be satisfied for nonepitaxial surface nucleation when an interlayer is present: (1) carbon saturation of the substrate surface and (2) presence of high-energy surface sites (unsatisfied valences). Distinctly different interlayers may form on different substrate materials at different rates, and different reactant gas mixtures or different substrate temperatures may produce different intermediate layers on the same substrate. Low C:H ratios and/or high substrate temperature may favor the formation of carbides, while high C:H ratios and/ or low substrate temperature may lead to the formation of a-C or diamondlike carbon (DLC), or even direct nucleation of diamond on the bare substrate surface.

B.

Diamond Epitaxy and Morphology Evolution

1.

Epitaxy

Diamond films epitaxially deposited on both diamond and cBN single-crystal substrates demonstrate that, under typical low-pressure processing conditions, single-crystal diamond deposition is possible [109,123]. However, deposition of single-crystal diamond films on substrates other than diamond and cBN is desirable due to the difficulty in obtaining large-area single crystals of either of these two materials. Methods for obtaining large-area single crystals, or at least highly oriented films, must be developed for largescale electronic applications. Heteroepitaxial growth of diamond has been and will continue to be a major research objective. The primary difficulty associated with diamond epitaxy is the small number of materials available with suitable crystal structures and lattice constants. Some transition metals and ceramics, such as Ni, Cu, Fe, and cBN, constitute the few isostructural materials with sufficiently similar lat-

80

Dandy and Coltrin

tice constants (mismatch < 5%). Further, the extremely high surface energies of diamond, ranging from 5.3 to 9.2 J m⫺2 for the principal low-index planes, and the presence of interfacial misfit and strain all present major obstacles in forming oriented, two-dimensional diamond nuclei [124]. Early attempts to grow heteroepitaxial diamond on transition metals were unsuccessful, possibly due to the high solubility and mobility of carbon on these materials and the formation of intermediate layers, as discussed in Sec. IV.A. More recently, however, there have been a number of successes in obtaining heteroepitaxial growth on Ni, Cu, and cBN [104,124–126]. 2.

Oriented Growth

The crystal habit of diamond is, in general, determined by the relative growth velocities of the {100} and {111} planes, denoted as v100 /v111, and the appearance or disappearance of crystallographic planes in diamond films depends upon the growth velocities of the corresponding planes. The facets that appear on a crystal are those for which the normal growth velocity is the slowest. Based on the Wulff criterion for crystal habit [127], it is predicted that the most stable growth planes in diamond are the octahedral {111} planes, followed by cubic {100} planes and the {110} planes. At low substrate temperature, when v100 /v111 > 兹3, the crystal habit of diamond is octahedral, and at high substrate temperature, v100 /v111 ⱕ 兹3/3, it is cubic, and in between it is cubo-octahedral. Increasing the hydrocarbon concentration in the gas phase has the same effect as increasing the substrate temperature, specifically, that the morphology of polycrystalline diamond films evolves from octahedral, to cubo-octahedral, to cubic. Success in growing highly oriented, textured diamond on silicon represents a novel approach for obtaining near-single-crystal morphology over large areas [128,129]. In initial experiments no texture formed, in part because of the creation of a-SiC interlayers, which led to the growth of randomly oriented diamond particles. Subsequent attempts resulted in the growth of highly oriented, textured films on single-crystal Si{100} substrates. In the latter experiments, a two-step procedure was applied. First, diamond was nucleated using a CH4 /H2 gas mixture, and it is believed that in situ carburization occurred during this stage, converting the Si surface to an epitaxial ␤-SiC layer. Nucleation of diamond produced partially oriented nuclei. Then, during the growth stage CO was added to the CH4 /H2 mixture, resulting in textured growth. The ability to obtain highly oriented, textured films has led to the investigation of the dependence of film texture and morphology on deposition conditions [130]. Oriented films are categorized according to whether they are (1) strongly fiber textured, (2) epitaxially textured films grown on

Deposition Chemistry

81

{100} Si, or (3) homoepitaxial films grown on {100} diamond. The {100} faceted surfaces were selected because CVD diamond grown on {100} is known to contain significantly fewer structural defects than that grown on {111}. The data for the experiments of Wild et al. [130] were correlated by a growth parameter, ␣ = 兹3(v100 /v111), and it was concluded that the microstructure (texture and orientation) and morphology of diamond films could be controlled through manipulation of this parameter, which in turn can be controlled by altering the growth environment. For fiber-texture films, low CH4 concentrations or high substrate temperatures (corresponding to ␣ < 1.5) result in films with a pronounced 具110典 texture; at intermediate CH4 concentrations and substrate temperatures (1.5 ⱕ ␣ ⱕ 3), a transition of the fiber axis from 具110典 to 具100典 occurs; further increases in CH4 concentration or lower substrate temperature (␣ > 3) leads to a complete deterioration of the film morphology, such that the films lose their 具100典 texture, become very fine grained, and do not show any distinct faceting. The evolution of crystal shape as a function of ␣ is illustrated in Fig. 8. In that figure, the arrows indicate the direction of fastest growth. In the case of homoepitaxial or heteroepitaxially textured films of {100} orientation, the microstructure and morphology of the films are strongly affected by twin formation, which may lead to complete deterioration of the epitaxial orientation. Deposition conditions corresponding to values of the growth parameter ␣ ⬇ 2.5 tend to suppress twin formation, and the subsequent film growth improves epitaxial alignment of the crystals. In contrast, small values of the growth parameter (␣ ⬇ 1.5) usually indicate that the film orientation and morphology will deteriorate due to twinning. The fact that film texture and morphology can be correlated to this one parameter (␣) is a strong indication that these two properties depend most strongly on substrate temperature and the gas composition near the deposition surface. By investigating different combinations of Ts and reactant composition, it may therefore be possible to control film texture, surface morphology, and stability with respect to twin formation and, in particular, to grow 具100典 textured diamond films with {100} faceted surfaces.

Figure 8 Transition of crystal shape from cubic to octahedral as the growth parameter, ␣, is increased. The arrows indicate the direction of fastest growth.

82

3.

Dandy and Coltrin

Morphology Evolution

Regarding the size distribution and grain evolution during film growth, a number of experiments indicate that, for nucleation and growth on Si{100}, the crystal size and size spectrum both grow with increasing deposition time. During the early stages of growth, crystals are small and relatively uniform in size; as the films grow, the average size increases and, simultaneously, the size distribution broadens. The great disparity in crystal sizes and pronounced surface roughness in high-growth-rate systems led to a detailed set of experiments to study possible morphological instabilities during diamond growth [85]. Indeed, it was observed that in low-pressure, high-growth-rate systems (plasma arcjet and combustion), the lack of surface diffusion and reevaporation at diamond film surfaces during growth accelerated any stochastic interface perturbations, leading to fast growth of some crystals relative to others. The instabilities are then magnified during sustained growth as a result of competitive shadowing and reactant depletion by the taller crystals. This phenomenon is usually observed only in high-growth-rate systems and only for films thicker than 20 ␮m. The possible morphological instabilities inherent to diamond film growth not only result in extreme variations in the sizes of the crystallites but also lead to the incorporation of voids and noncrystalline phases in the films. As growth proceeds, smaller crystals are often found to be coated with DLC, which may be attributed to reactant starvation and lower temperature experienced by these smaller grains. Several possibilities exist for surmounting these instabilities in high-growth-rate systems. The first is to increase the nucleation density of diamond on the substrate, so that the inevitable onset of instability is delayed. The second is to grow the films in a cyclical manner, so that the growth phase is interrupted regularly with a renucleation phase. C.

Effects of Surface Conditions

It has long been recognized in film deposition that the substrate material and surface pretreatment both strongly affect the properties of the resulting film. As mentioned earlier, two key issues in diamond deposition are the surface nucleation density and rate, and a number of nucleation enhancement techniques have been developed to maximize both of these quantities. Nucleation density has been increased from less than 105 cm⫺2 on untreated substrates up to 1011 cm⫺2 on scratched or biased surfaces. The effects of surface conditions on nucleation processes have been investigated to provide guidelines for selection of optimal surface pretreatment methods. Methods of substrate surface pretreatment that have been tested include

Deposition Chemistry

83

(1) scratching with abrasives; (2) seeding with diamond grit and other ceramic powders; (3) electrical bias; (4) covering the substrate with graphite films or fibers; (5) coating the substrate with thin metal films, a-C, C70, cBN, SiC, WC, or hydrocarbon oil; (6) ion implantation; (7) pulsed-laser irradiation; (8) carburization; and (9) chemical etching. These different methods make it possible to control diamond nucleation density over many orders of magnitude. The reverse problem, that is, preventing diamond from growing on specific portions of a deposition substrate, is important for selective or patterned growth, a technique of particular importance in electronics applications. A number of surface pretreatment methods have been proposed for preventing diamond nucleation on foreign surfaces, including oxidation, sputtering, reactive ion etching, excimer laser irradiation, and surface melting with lasers. It is generally accepted that native oxides (SiO2) are the most effective at impeding diamond nucleation [131,132]. Chemical etching in HNO3, HF, and HCl; plasma etching in NF3 and Cl2; and sputtering in N2 have shown success in introducing surface roughness on selected regions of the surface but have not resulted in increased nucleation density and have, in fact, had the opposite effect. Scratching is the most widely used technique for enhancing diamond nucleation, although this pretreatment method is not easily applied to surfaces of complex shape and is not generally attempted when growing films that require extremely smooth surfaces. Often in this technique, the surface is scratched, abraded, or blasted with diamond particles or paste, but other abrasives are also used; these include borides, carbides, nitrides, silicides, oxides, and graphite. Diamond nucleation densities after scratching pretreatment typically range from 106 to 1010 cm⫺2. Scratching may also be accomplished using ultrasonic vibration with an abrasive paste suspended in methanol or acetone. It has been observed that an ultrasonically damaged surface has more uniformly distributed defects, and this leads to slightly higher nucleation densities (107 –1011 cm⫺2) and more reproducible effects [99,122]. A number of explanations have been put forth for the enhancement of nucleation density and rate caused by scratching. The first of these is the seeding effect: diamond, DLC, or other carbonaceous residues from the scratching process are left behind, either adhering to or imbedded in the substrate, and these nano- and microparticles may act as nucleation sites [133]. Because other abrasive materials yield a similar yet less pronounced effect, it is likely that this is not the only operative mechanism in diamond nucleation on scratched surfaces. A second possible mechanism for nucleation is that the highly disordered surface materials or microscopic crater edge sites on the polished surface create high-energy sites. Diamond nucle-

84

Dandy and Coltrin

ates preferentially at these sites because (1) there is a resultant minimization of interfacial energy from the formation of diamond nuclei on the sharp convex surfaces, (2) the microscopic craters and trenches formed by scratching may leave dangling bonds, and (3) carbon saturation will occur most rapidly at sharp edges. Topography patterning analyses lend credence to the second postulate [134]. It was observed that, although residual abrasive powder may enhance nucleation, the nucleation event is promoted by topological features alone, and the presence of residual abrasive is not a necessary condition for nucleation. A third possible mechanism for nucleation is that the scratching process produces nonvolatile graphitic particles through local pyrolysis; the graphitic particles are subsequently hydrogenated in the growth environment. Scratching and seeding cause substrate surface damage and contamination, as mentioned earlier. Therefore, these pretreatment methods are incompatible with many applications requiring extremely smooth, clean surfaces, such as diamond films for electronic devices, optical window materials, and smooth, wear-resistant coatings. Alternative pretreatment methods that yield high nucleation densities without substrate damage are of particular importance. Biasing pretreatment of substrates has been increasingly employed to enhance surface nucleation of diamond [129,135, 136]. Using a potential bias to obtain large nucleation densities on unscratched substrates provides an opportunity to control nucleation densities through variation of the applied voltage and current, while reducing surface damage. Most investigations of bias-enhanced nucleation have focused on mirror-polished Si substrates, and possible explanations for the nucleation enhancement have emerged. For negative biasing of an Si substrate, the role of biasing is to (1) increase the flux of carbon-containing cations (C⫹, ⫹ CH⫹ x , CxHy ) to the surface, expediting the local carbon saturation on the surface and leading to a thin layer of amorphous carbon on the SiC interlayer to form small clusters for diamond nucleation; (2) transfer more energy to the surface through ion bombardment, resulting in increased surface mobility of adsorbed species; (3) reduce and suppress surface oxide formation and remove native oxides; and (4) accelerate gas-phase reactions because of the increased ion-neutral collisions and higher energy of the sheath region, leading to higher concentrations of reactive hydrocarbon radical species. It is believed that nucleation enhancement on positively biased substrates is due to a high electron density, which in turn results in a high electron kinetic energy. Such a high-speed electron impingement process may increase the rate of decomposition of adsorbed hydrocarbons through hydrogen desorption.

Deposition Chemistry

D.

85

Effects of Deposition Conditions

Deposition conditions, such as substrate temperature, reactor pressure, and reactant composition, critically influence diamond nucleation rate and density, but conditions optimal for growth may not be ideal for nucleation. Experiments on bias-enhanced nucleation revealed that negative biasing greatly increases diamond nucleation, but poor-quality films are produced when biasing continues during growth [137]. Similarly, optimal values of temperature and pressure for growth may not be the same as those for nucleation. 1.

Substrate Temperature

Experiments carried out in different reactor systems present a consistent physical picture of the dependence of nucleation density on substrate temperature [138–140]. In these studies it was found that, for growth on silicon substrates, nucleation density falls off sharply for substrate temperatures below 820–850⬚C; for temperatures greater than 850⬚C, the nucleation density gradually decreases but remains on the order of 1010 cm⫺2 up to the highest temperatures studied (950⬚C). It is speculated that this overall dependence of nucleation density on substrate temperature is due to the change in adsorption state and surface diffusion length of growth precursors. At temperatures below approximately 900⬚C, the nucleation precursors are adsorbed primarily by physical adsorption (physisorption), whereas chemical adsorption (chemisorption) is dominant at higher temperatures. This change in the adsorption state causes an abrupt change in the diffusion length of the precursors as the temperature is increased above 900⬚C. As a result, the capture efficiency of precursors on the substrate surface and thus the nucleation rate and density dramatically increase when the temperature approaches 900⬚C. 2.

Reactor Pressure

Investigations indicate that there is an inverse relationship between pressure and nucleation density and rate, at least for diamond growth on silicon [111,141]. However, the effect does not appear to be nearly as strong as substrate temperature, resulting in a factor of 3–4 decrease when the pressure is increased from 2 to 50 torr. For silicon substrates, it is possible that the inverse dependence on pressure is due to a competition effect between ␤-SiC formation (which increases diamond nucleation density) and atomic hydrogen etching of material, which decreases the number of nucleation sites. Because of these observations it has been suggested that the nucleation phase be carried out at lower pressure (2–5 torr) and subsequent growth be carried out at higher pressure (20–40 torr).

86

Dandy and Coltrin

The effect of pressure has also been examined for nucleation on Mo substrates [142]. It was found that good nucleation density was achieved for high gas-phase hydrocarbon fractions at low pressure due to the promotion of carburization of the substrate surface and increased surface C concentration. Growth was then carried out at lower hydrocarbon concentrations and higher pressure to preserve diamond film quality. Films grown in this manner have been shown to be dense, homogeneous, and well crystallized. 3.

Reactant Composition

In nucleation on carbide-forming substrates, it has been observed that diamond nucleation density increases as the inlet CH4 mole fraction is increased. The gas composition influences not only the nucleation density but also the nucleation behavior and the resultant crystal morphology. At lower CH4 mole fractions (<0.5%), diamond nucleation may actually cease while a significant area of the substrate remains unnucleated, and the subsequent growth leads to good-quality, well-faceted, isolated diamond crystals. At higher CH4 concentrations (1–2%), isolated crystals may grow large enough to occlude one another before nucleation terminates [143]. A side effect of higher hydrocarbon concentrations is formation of appreciable nondiamond components. The addition of small amounts of oxygen has been generally found to reduce the incubation period and to increase nucleation density. For growth on silicon, it has been postulated that the role of oxygen is to participate in reactions leading to preferential formation of SiOx on the surface rather than SiC; the SiOx layer may impede Si diffusion from the bulk to the substrate surface, allowing the adsorbing carbon species to saturate the surface more quickly and then nucleate [144]. Further, the presence of oxygen allows lower substrate temperatures during nucleation due to the lower binding energy of OH to C (3.68 eV) compared with that of CH to C (4.63 eV); OH is more readily abstracted by atomic hydrogen, creating reactive sites for hydrocarbon adsorption. However, for growth on Ni and Pt substrates, oxygen may have an adverse effect. In experiments using these substrate materials it was observed that, although the addition of oxygen did not suppress growth of existing diamond and resulted in sharply faceted crystals, it degraded diamond nucleation by eliminating nucleation sites [114]. For high oxygen concentrations (O:H = 1), Ni and Pt surfaces are etched clean; oxygen must be decreased by a factor of 2 or more (O:H ⱕ 0.5) before nucleation sites last long enough for growth to occur. Once growth does occur, oxygen does not act as an inhibitor, and, in fact, the reason that large, well-faceted crystals form is the preferential etching of potential nucleation sites on the crystals, thereby suppressing secondary nucleation.

Deposition Chemistry

V.

87

IN SITU DIAGNOSTICS

The rate of growth of a diamond film and its uniformity, morphology, texture, and thermal and mechanical properties all critically depend upon the state of the gas phase adjacent to the deposition surface, that is, the composition, temperature, and flow field. The gas properties at the surface, in turn, depend upon the upstream conditions in the reactor. As discussed in Sec. VI, one- and two-dimensional physical models have been used, with varying degrees of success, to predict the flow field and temperature and species profiles within different types of diamond CVD reactors. Such models are a powerful tool for exploring issues related to diamond growth and may help provide links between operating conditions and film growth rate, morphology, and quality. However, the models cannot stand alone: to gain a true understanding of the mechanisms underlying diamond growth it is necessary to couple predictions with experimental measurement. In situ measurements of the gas phase and the deposition surface during diamond growth provide data that may be used both as input to the detailed models and as a means of validating such models. Diagnostic technique may be used, for example, to measure concentrations of postulated growth precursors such as CH3 or the species crucial for diamond growth, atomic hydrogen. Because the mechanisms describing the gas-phase chemistry are relatively well understood, it may only be necessary to measure several key species in order to have quantitative knowledge of the overall species distribution. A strong need also exists for on-line process monitoring in diamond CVD because this material is extremely expensive to synthesize, and many reliability issues remain to be addressed. Diagnostics well suited for fundamental studies are not necessarily appropriate as process monitors. Currently, most process measurement and control techniques in diamond reactors focus on average substrate temperature, reactor pressure, reactant flow rate, and power. These are important parameters, but they do not give enough explicit information to determine the state of the gas phase uniquely. For example, if laboratory measurements of gas-phase composition and temperature reveal a strong correlation between specific measurable quantities and film growth rate and/or properties, there may be an opportunity to exploit this information through the development of an on-line process sensor. Such a sensor could be tightly coupled to an overall process control strategy. Discussion of the diagnostic methods in this section is segregated according to the reactor type in which the measurement was made: hot-filament, plasma-assisted, and combustion. This section will not constitute a thorough review of diagnostic techniques employed in diamond CVD but is instead intended to indicate how measurements made by extraction, physical

88

Dandy and Coltrin

probes, and optical probes advance understanding of the growth process. A complete review of gas-phase diagnostic techniques in diamond CVD has been presented by Thorsheim and Butler [145].

A.

Hot-Filament Deposition

The hot-filament reactor is the most studied of the various diamond reactor types. Diagnostics in hot-filament systems have addressed three specific issues: (1) identification of diamond growth precursors, (2) determination of the role of other species in the growth process, and (3) quantification of the effect of operating conditions on gas-phase composition and spatial distribution. One of the first measurements of the gas-phase environment in diamond CVD was made by Kawato and Kondo [146] using gas chromatography (GC) to sample residual gas from a reactor. They observed detectable quantities of CH4, C2H4, C2H2, H2, and CO from a feed gas mixture of CH4, H2, O2, and Ar and found that most of the inlet CH4 was converted to C2H2. As more oxygen was added to the system, the amount of CO increased at the expense of C2H2. Because this was a sampling of the downstream stable gas species, the relationship between the measured concentrations and the conditions at the growth surface is not clear. Celii et al. [24] used infrared diode laser absorption spectroscopy for the in situ measurement of gas species in the region between the filament and the substrate. Using this method, C2H2, CH3, C2H4, and CH4 were detected. In contrast to the measurements of Kawato and Kondo, CH4 was found to be the predominant species (⬇8 ⫻ 1014 cm⫺3), with a concentration approximately four times greater than that of C2H2 and a factor of 10 greater than CH3. The fact that the measured CH4 concentration was over 40% of its inlet value indicates that a significant fraction of the inlet gas may completely bypass the filament, never undergoing thermal or heterogeneous dissociation. Using a combination of ex situ mass spectroscopy and x-ray photoelectron spectroscopy (XPS), Harris et al. [147] investigated diamond growth on platinum. By sampling at discrete points between the filament and the substrate, they observed that only CH4 showed a significant change with position, with its concentration growing by a factor of 5 at 30 mm from the filament. The concentrations of C2H2 and C2H4 remained flat in this region. Wu et al. [148] also used gas chromatography and quartz microprobe sampling through a temperature-controlled probe to measure gas-phase concentrations. They found that, while the CH4 concentration also increases with distance from the filament, the increase was a factor of 3 rather than the factor of 5 reported by Harris et al. [147]; the C2H2 mole fraction was seen

Deposition Chemistry

89

to decrease by a factor of 2. The discrepancies between the results may be explained by differences in deposition conditions and reactor geometry. Atomic hydrogen has also been measured in hot-filament reactors using laser-induced fluorescence (LIF) [22,149], and its concentration was inferred from crossed beam coherent anti-Stokes Raman spectroscopy (BOXCARS) measurements of H2 concentrations [150]. The LIF measurements indicated that, with no substrate present, H concentrations decreased by only a factor of 3 (relative to the value at the filament) at a distance of 30 mm from the filament. This result is expected because the low pressures used in the experiments (20 torr) greatly reduce the rate at which H recombines to form H2. The known high reactivity between H and heated surfaces (such as a substrate) would undoubtedly cause a more significant drop in H concentration with distance from the filament if a surface were present. The atomic hydrogen concentrations derived from the BOXCARS measurements [150] were an order or magnitude greater than those obtained from the LIF measurements. The differences between the two measurements were explained by filament temperature variation and other experimental factors. Molecular beam mass spectrometry (MBMS) measurements of predominant gas phase species were made within 0.1 mm of the substrate [15], and it was found that the absolute H concentrations were uniformly lower than those measured by either LIF or BOXCARS. As shown in Fig. 9, the MBMS measurements revealed that, as the inlet CH4 concentration was varied between 0.4 and 7%, a transition occurred in the predominant hydrocarbon species present near the substrate: C2H2 was dominant for CH4 feed fractions less than ⬇1%, while CH4 was dominant for feed fractions greater than this value. It was postulated that CH4 became the dominant species near the substrate because the filament became coated with carbonaceous material for high inlet carbon concentrations, and this poisoning resulted in lower atomic hydrogen concentrations and hence lower conversion of CH4. This was corroborated by a subsequent theoretical study using a stagnation flow model in which a primitive filament poisoning model was included [28], as described in Sec. VI.A. It was also observed in the MBMS study that, at the surface, CH3 closely tracked CH4, and its concentration was approximately 10% that of CH4, a result consistent with the infrared diode laser measurements of Celii et al. [24]. A possible diamond growth precursor, CH3, has a much stronger concentration dependence on distance from the substrate than do other measured stable hydrocarbon species such as CH4, C2H2, or C2H4, and it is expected that this larger spatial gradient is due to the higher reactivity of CH3. Resonance-enhanced multiphoton ionization (REMPI) was used to measure CH3 concentrations 4 mm above the substrate, in a system where the filamentto-substrate distance was varied from 7 to 16 mm [151]. The methyl radical

90

Dandy and Coltrin

Figure 9 Species mole fractions measured by MBMS sampling at the surface in a hot-filament reactor as a function of methane fraction in the feed gas. (From Ref. 15.)

concentration was a factor of 3 lower when the separation distance was 7 mm than when it was 16 mm. REMPI was used in a similar reactor system to measure concentrations of CH3, CH4, and C2H2 at different substrate temperatures [152]. By incorporating the concentration data with measured film growth rates, an apparent activation energy of 0.17 eV in CH3 concentration was determined, and no measurable substrate temperature dependences were observed for CH4 and C2H2. Absolute concentration profiles for CH3 as a function of distance from the filament were measured [153] using cavity ring-down spectroscopy (CRDS) [154]. A peak in CH3 concentration was detected several millimeters from the filament, indicating that this species is formed by gas-phase reactions rather than heterogeneous filament chemistry, as shown in Fig. 10. The CH3 concentration was also observed to increase linearly with susceptor temperature. Measurements of CH4 concentrations using CARS [155] and atomic hydrogen using REMPI [25] have shown that, near the filament, the up-

Deposition Chemistry

91

Figure 10 Spatial profiles of CH3 density and absorbance in a hot-filament reactor at two different susceptor temperatures. (From Ref. 153.)

stream and downstream spatial profiles are the same. In other words, the concentration profiles of CH4 and H are independent of whether the measurements are made upstream and downstream of the filament. The conclusion to be drawn from this result is that, indeed, mass transport in hotfilament reactors is completely dominated by diffusion, and convection plays a secondary role at best. Models of diamond growth in hot-filament reactors have confirmed that growth rate is independent of filament-substrate orientation [28,156]. Spatial temperature distributions have been measured using thermocouples [14], BOXCARS [150], and LIF [157]. These different experimental techniques consistently show a very sharp temperature drop from the filament to the surrounding gas. For example, LIF temperature measurements showed a drop from 2600 K on the filament to 1400 K 1 mm from the filament, in a reactor operating at 30 torr. The lower the operating pressure, the more pronounced is the apparent temperature discontinuity between the filament and adjacent gas. As shown in Fig. 11, BOXCARS measurements found that, for a 2870 K filament and 20 torr operating pressure, the tem-

92

Dandy and Coltrin

Figure 11 Gas temperature in a hot-filament reactor with 1% CH4 in H2, for filament power of 2.7, 2.0, and 1.5 kW. The open squares indicate the filament and substrate temperatures for 2.0 kW filament power. (From Ref. 150.)

perature dropped to 2600 K within 0.1 mm of the filament. This effect is well known as was first discussed by Langmuir [18], and it is observed in many thermal systems operating at low pressure; the discontinuity is observed within one or two mean free paths (75–125 ␮m at 20 torr) where continuum heat transfer theory is not applicable.

B.

Plasma-Assisted Deposition

A popular technique for determining the composition of plasmas is optical emission spectroscopy (OES). High-temperature plasma systems such as dc arcjets show elevated levels of CH and C2 when compared with systems such as microwave, and these emissions are useful in determining the presence of specific species and for use as a process monitor [158]. The use of OES as a sensor for process control in low-pressure dc arcjet [159] and microwave reactors [160] has been investigated. For the microwave system it was found that there was a linear relationship between C2, CH, and H(␤) emissions throughout the plasma. Molecular beam mass spectrometry has also been used to determine absolute species concentrations near the substrate in microwave [161] and

Deposition Chemistry

93

arcjet [162] plasma reactors. In the 20-torr microwave system, the effects of hydrocarbon source (CH4 and C2H2), hydrocarbon concentration, and substrate temperature on the gas-phase species concentration distribution near the substrate were studied. As illustrated by the data in Fig. 12, over the range of inlet carbon mole fractions considered, the composition near the substrate was independent of the specific source gas, indicating that the plasma was very efficient at scrambling the species during transport to the substrate. The composition was also found to be insensitive to substrate temperature, so that, although the film growth rate was temperature dependent, the gaseous species were governed by upstream processes that did not depend on the details of the heterogeneous kinetics. Regardless of the hydrocarbon reactant gas, when the mole fraction of carbon in the feed was less than 1%, CH4 was the dominant species detected near the substrate, whereas C2H2 was dominant for inlet mole fractions greater than 1%. Atomic hydrogen was found to be relatively insensitive to inlet carbon mole fraction, making up approximately 0.07% of the gas mixture. In the arcjet study [162], a 1% mixture of CH4 and H2 was passed through a 1 kW arc into a 200 torr reactor. It was found that a significant fraction of the inlet CH4 was converted to C2H2 and C2H4 and that the C2H2 concentration near the substrate grew approximately linearly with CH4 flow as the amount of CH4 in H2 was raised from 0.3 to 1.5%. Attempts were made to measure CH3 in that work, but it proved impossible to distinguish between actual CH3 and that produced as a fragmentation product. Actinometry has been used by a number of investigators to measure the relative ground state concentrations of species, most notably atomic hydrogen, using Ar as the actinometer [163–165]. It was observed by Mucha et al. [163] that, in CH4-H2-He-Ar mixtures, an increase in CH4 gave rise to a decrease in the H concentration, and the addition of oxygen resulted in an increase in H, possibly due to passivation of the reactor walls. Spatially resolved relative concentrations of atomic hydrogen were measured by Reeve et al. [165] in a dc arcjet reactor using a CH4-H2-Ar reactant mixture with no substrate present. It was observed that the H concentration dropped rapidly with distance from the plasma torch exit, reaching a relative value 30% of its initial maximum within 6 cm. The actinometry data, integrated along the line of slight, were found to be in excellent agreement with predictions of a one-dimensional model containing detailed pyrolysis chemistry. One drawback of actinometry, particularly in systems with appreciable electron densities, is the effect of the noble gas on plasma chemistry. Zhu et al. [166] carried out experiments in a microwave system with a CH4 /H2 plasma, and found that, upon addition of a noble gas, the emission intensity of various species changed, and the magnitude of the effect depended on the specific noble gas species introduced. It was determined that if a noble gas

94

Dandy and Coltrin

Deposition Chemistry

95

is to be used as an actinometer, its concentration should be kept below 1%. Ground state concentrations of atomic hydrogen have also been measured using third-harmonic generation [167,168]. The amount of atomic hydrogen present was seen to increase with plasma power, while it decreased as CH4 was added. Another nonlinear technique, resonant degenerate four-wave mixing, was used to measure CH and C2 concentrations and gas temperature within the boundary layer in a convection-dominated plasma reactor [169]. Measured CH profiles compared well with predictions obtained from a onedimensional stagnation-point flow model in which the experimentally determined temperature profile was used. Other work in plasma-assisted systems has been directed at determining concentrations of stable species such as C2H2 and CH4 using Fourier transform infrared spectroscopy (FTIR) in a microwave plasma reactor [170] and CARS in a diffuse rf plasma [155]. The quantitative differences in species’ concentrations measured in these two reactor types illustrate important distinctions in the gas-phase chemistry induced by the different energetic sources. In the microwave system, only C2H2 and CH4 were detected when the hydrocarbon constituted less than 1% of the inlet gas stream, whereas in the rf plasma no acetylene could be detected for comparable inlet conditions. Based on known detection limits it was deduced that, in the rf system, less than 20% of the inlet CH4 was converted to C2H2. As rf power was increased, the measured concentration of CH4 decreased, indicating that more gas-phase conversion was occurring. Efforts at measuring temperature in plasma systems have met with partial success. Early measurements using Langmuir probes in an arcjet system operating between 100 and 400 torr greatly overestimated the local temperature, and this was attributed to the disturbance of the plasma by the probe [171]; OES measurements in the same system gave temperatures that were 35–45% of the values measured by the probe. No substrate was present in that experiment, so it is difficult to extrapolate from the observed results to an actual deposition system. Measurements in a dc arcjet system at 60 torr using a floating double probe found that electron temperatures ranged from 2.3 eV at the exit of the plasma torch to 0.4 eV in the downstream (near the substrate) and peripheral regions of the plasma [172]. Electron densities in that system ranged from 6 ⫻ 1010 cm⫺3 near the plasma torch

< Figure 12 Mole fractions of species measured near the substrate in a microwave reactor as a function of the carbon mole fraction in (a) acetylene feed and (b) methane feed. Reactor pressure (20 torr), microwave power (850 W), and substrate temperature (825⬚C) were held constant. (From Ref. 161.)

96

Dandy and Coltrin

to 1 ⫻ 109 cm⫺3 in the downstream region, indicating that charged species concentrations were small compared with the 3 ⫻ 1017 cm⫺3 mixture number density. Determination of the gas kinetic temperature using OES of H2 and CN was carried out by Chu et al. [173]. LIF of CN was used to compare the rotational temperature obtained with OES with the gas kinetic temperature. It was found that the rotational temperature was an accurate measure of the gas kinetic temperature, even at pressures greater than 10 torr, and it was concluded from this that the rotational distribution is a better choice for temperature measurements than vibrational or electronic populations because of the more rapid thermal equilibration of the rotational population. One of the first detailed analyses of C2 and CH optical emission in a dc arcjet operating under diamond growth conditions with a CH4 /H2 mixture was carried out by Raiche and Jeffries [174]. Through examination of emission spectra, they concluded that the emission was likely due to chemiluminescent reactions; discrepancies in temperatures derived from emission spectra (5000 K) and those obtained from LIF measurements (2100 K) [175] were attributed to lack of thermodynamic equilibration of excited state populations. Rotational and vibrational excitation temperatures were determined in a dc arcjet reactor with a CH4 /H2 /Ar mixture, and a substrate located 1 in. below the exit of the plasma torch [176]. Although it was found that temperatures converged at the substrate and that rotational temperatures for C2 and vibrational temperatures for CH closely tracked the gas temperature, the vibrational temperatures for C2 and rotational temperatures for CH were anomalously high. Analysis of those results indicated that, in the plasma bulk, emission by CH was produced primarily by chemiluminescent reactions, while C2 was produced by electron-impact excitation. The general conclusion from that work was that OES of C2 provided an accurate means of obtaining spatially resolved gas kinetic temperatures. C.

Combustion Flame Deposition

Optical emission has been used in combustion synthesis to detect species such as CH, C2, H, and O in one system [177] and OH, CH, and C2 in another [178]. Yalamanchi and Harshavardhan observed that the primary combustion zone displayed the strongest emission from the carbon-bearing species, and the emission intensities of CH, H, and O dropped sharply downstream of this zone, while C2 emission remained fairly constant. Diamond growth was reported when CH emission was significantly stronger than C2 emission in the primary combustion zone. Hirose et al. [178] also examined emission from the center of the primary combustion zone and found that conditions optimal for optically transparent diamond correlated with an in-

Deposition Chemistry

97

crease in OH emission and a decrease in C2 emission. In subsequent work, emission from C2 at 510 nm was correlated with diamond growth in a C2H2/ O2 flame [179]; as can be seen in Fig. 13, C2 emission increased with the fuel-to-oxygen ratio, reaching a maximum when the ratio of the fuel flow rate to the oxygen flow rate was 1.1. This maximum in C2 emission corresponded to a maximum in the deposition rate. Subsequently, diamond films were grown using a MAPP/O2 mixture, and OES spectra from OH, CH, and C2 were measured [158]. MAPP gas in a mixture of methyl acetylene, propadiene, and liquefied petroleum gas. By examining the relative intensities of the CH and C2 emissions, it was found that when the MAPP and oxygen flow rates were manipulated such that the ratio of the C2 ⌬␯ = 0 transition to the CH A-X transition was between 1.2 and 1.6, well-faceted diamond films could be grown. Ground state species have also been measured using LIF and mass spectrometry [180] and GC [181]. The mass spectrometer probe sampled gas from the center of the primary combustion zone; the main gaseous spe-

Figure 13 Emission intensities of C2 and CH, and deposition rate as a function of stoichiometry in a C2H2 /O2 combustion flame reactor. (From Ref. 179.)

98

Dandy and Coltrin

cies detected in this region were CO and H2. Samples taken near the burner showed C2H2 and the carbon-bearing radicals to be in near equilibrium. As oxygen flow rate was increased, C2H2 concentration decreased near the combustion zone, while H2 remained nearly constant. The LIF experiments were performed with no substrate present and it was found that maximum OH fluorescence intensities were nearly independent of distance from the combustion zone. Flame temperature and the concentrations of CO, CO2, H2O, and OH have been measured during atmospheric-pressure diamond growth using FTIR [182]. Distinct differences in temperature and composition were observed depending on whether or not a substrate was present; the deposition flame was found to be wider and cooler than the free flame, with the gas temperature lowered by as much as 700 K with the substrate present. In both flames, CO was located primarily in the main combustion zone, while H2O, CO2, and OH were found mostly in the outer flame region. A large radial temperature gradient, 700 K/mm, was measured across the diamond growth zone. The steep temperature and composition gradients in many combustion torch systems, such as this one, point to the difficulties inherent in growing large-area, uniform diamond films using combustion synthesis. Obtaining the appropriate gas-phase composition with little or no radial variations points to the need for flat-flame burners and burner-stabilized flames [183–185].

VI.

REACTOR SCALE MODELING

Numerical and analytical solutions of the governing conservation equations for momentum (fluid flow), mass, and energy (temperature) have led to many insights concerning the mechanisms and conditions under which diamond may be grown by chemical vapor deposition. Modeling many types of diamond reactors has proved successful primarily because the principal hydrocarbon sources used are CH4 and C2H2, and as discussed in Sec. II, the pyrolysis and combustion mechanisms for these fuels are well understood for the stoichiometries used, and thermodynamic and thermophysical data for the various gaseous species are known. Also, as discussed in the following, the reactor configurations and operating conditions used in dc arcjet, rf plasma, combustion, and hot-filament systems lend themselves to geometric simplifications, which in turn make the numerical models tractable, even with very detailed homogeneous and heterogeneous chemistry. The onedimensional models of these reactors have proved successful in describing qualitative (species and temperature profile shapes, likely growth precursors) and quantitative (diamond growth rate, species concentrations) features in

Deposition Chemistry

99

systems where charged species (plasma) chemistry does not play a significant role. There is a need, however, for more detailed modeling whereby multidimensional effects are captured and ion-neutral chemistry is considered so that systems containing nonthermal plasmas, such as microwave reactors, may be adequately described. A significant factor complicating reactor models in the presence of the energy source. Whether it is the arcjet issuing from a dc plasma torch, the hot filament(s), or the microwave or rf plasmas, assumptions must invariably be made in constructing a physical model of the source. And, not surprisingly, the particular assumptions strongly affect the predicted temperature and species distributions within the reactor. Because each source type is handled differently, the discussion here is broken up according to reactor type: hot filament, plasma assisted, and combustion. Hot filament is the most studied diamond reactor type, both experimentally and theoretically, and it will be discussed first. A.

Hot-Filament Reactors

Most of the modeling efforts have focused on determining the temperature profile and, in particular, the species distribution in the region between the filament and substrate to identify likely diamond growth precursors and to develop and validate deposition mechanisms. The models have also been used to examine the role played by the filament in initiating gas-phase chemistry. Theoretical studies of hot-filament reactors may be roughly divided according to whether one- or two-dimensional models were used. The onedimensional models typically include detailed gas-phase and gas-surface chemistry and have been used to gain insight into the chemical kinetics governing diamond deposition and the relationship between mass transfer and chemistry. Two-dimensional models generally contain limited gas-phase chemistry and often do not contain deposition mechanisms at all. These models yield information on the details of mass, momentum, and energy transport in reactors. One of the first detailed studies of the chemistry in a hot-filament system was carried out by Goodwin and Gavillet [186]. It was assumed that the flow between the filament and substrate behaved as an ideal stagnation flow; the gas composition at the filament was determined using homogeneous kinetics and an experimental observation indicating that CH4 and C2H2 were present in roughly equal concentrations [14]. Figure 14 shows an example calculation of species concentration as a function of height above the substrate from this work. The availability of film growth rate data and the lack of an accepted deposition mechanism led Goodwin and Gavillet to examine the fluxes of individual hydrocarbon species to determine their

100

Dandy and Coltrin

Figure 14 Calculated mole fractions as a function of height above the surface in a hot-filament reactor (for all species greater than 0.1 ppm). (From Ref. 186.)

feasibility as growth precursors. An upper bound, mass transfer–limited growth rate of each hydrocarbon species was calculated, and it was calculated that only CH3, CH4, and C2H2 were present in sufficiently high concentrations at the surface to account for measured growth rates; the conclusion was that one of these three species must be responsible for deposition in the hot-filament environment. This result is consistent with the postulated deposition mechanisms that were being developed, specifically that one or both of CH3 and C2H2 were the species leading to film growth. Molecular beam mass spectroscopy data obtained by Hsu in a hotfilament reactor was the first quantitative measure of major species concentrations in a diamond deposition system [15,187]. By sampling through a pinhole in the center of the substrate, Hsu measured near-surface concentrations of CH4, C2H2, CH3, and H as a function of inlet CH4 concentration. (See Sec. V.A for more discussion of Hsu’s experiments.) A modeling study was subsequently carried out to examine Hsu’s system [28]. In that work, calculations were performed for H2-only and CH4 /H2 feed gases, using a one-dimensional stagnation flow model to describe the region between the

Deposition Chemistry

101

filaments and the substrate. For the hydrogen system it was found that, as filament temperature was varied between 1800 and 2600 K, good agreement with Hsu’s data [187] was obtained when (1) the H and H2 near the filament equilibrate around the filament temperature and (2) the heterogeneous abstraction and termination reactions (1) and (2) were included. A comparison between the experimental and numerical results is shown in Fig. 15. When the heterogeneous reactions at the surface were not accounted for, atomic hydrogen mole fractions were overpredicted by two orders of magnitude. This result demonstrates the importance of the gas-surface chemistry in determining the gas-phase composition near the surface, and it also demonstrates the significant mass transfer limitation in hot-filament reactors. That is, the overall kinetic rate of destruction of H by surface reactions is much faster than the rate at which H can be transported to the surface by diffusion. The ratio of these two time scales, diffusion to kinetic, is defined as the Damko¨hler number (Da); for the hot-filament system the effective Damko¨hler number for atomic hydrogen is on the order of 10–20, depending on conditions, confirming that the system is, indeed, mass transfer limited. Numerical results were also calculated for inlet CH4 concentrations between 0.4 and 7.2% [28]. To obtain good agreement with the experimental data [15] it was necessary to implement a simple filament poisoning model,

Figure 15 Experimental [187] and theoretical [28] results for near-substrate atomic hydrogen mole fraction in a hot-filament reactor containing only hydrogen.

102

Dandy and Coltrin

whereby it was assumed that heterogeneous H2 dissociation on the filament varied linearly from a maximum when no hydrocarbon was present to a minimum at 7.2% CH4. The calculated results from that study, together with the measured compositions, are shown in Fig. 16. As can be seen in Fig. 16a and b, it was determined that the predicted gas-phase composition did not depend strongly on whether or not a temperature discontinuity between the filament and surrounding gas [150] was used. However, when no filament poisoning was assumed, regardless of the temperature discontinuity, the calculated results for gas composition were in complete disagreement with Hsu’s data [15]. The conclusion from the results was that the hydrocarbon blocks reactive sites on the filament surface, and the extent to which sites are blocked depends on the amount of the hydrocarbon present. Hsu’s observation that, to maintain a constant filament temperature, current must decrease as inlet CH4 increases, confirms this supposition. Additional experiments and analyses were carried out to examine the effects of hydrocarbon source (CH4 versus C2H2) and substrate temperature (255 ⱕ Ts ⱕ 825⬚C) on gas composition near the substrate [27]. As with the previous study [28], the numerical model exhibited good qualitative and quantitative agreement with experiment for a variable CH4-H2 feed composition and a fixed substrate temperature. For low inlet methane fractions it was predicted that most of this species was converted to C2H2, but at higher inlet CH4 fractions significant conversion did not occur. Agreement between theory and experiment was not as good when the CH4 feed was replaced by C2H2. When the inlet carbon fraction was less than 1%, there was good quantitative agreement, suggesting that the model developed earlier [28] is applicable for C2H2 feed when there is little filament poisoning. However, the calculations significantly underpredicted (by as much as 60 times) nearsurface CH3 and CH4 concentrations at higher inlet C2H2 concentrations. The large amounts of CH4 and CH3 observed experimentally by McMaster et al. [27] suggest that facile decomposition of C2H2 occurred even when atomic hydrogen concentration was reduced by filament poisoning. The predictions of the model were relatively insensitive to the extent of the poisoning (and hence, the degree of H2 dissociation), and it is likely that heterogeneous decomposition of C2H2 on the filament is the explanation for the discrepancy between the model predictions and the observed data.

> Figure 16 Experimental [15] and theoretical [28] results for gas-phase species distribution at the substrate for a filament temperature of 2620 K and a temperature discontinuity at the filament of (a) 900 K and (b) 0 K.

Deposition Chemistry

103

104

Dandy and Coltrin

A discrepancy was also found regarding the dependence of near-surface H concentration on substrate temperature. At low inlet CH4 fraction (0.4%) and relatively high substrate temperature (825⬚C), there was good agreement between theory and experiment. But the one-dimensional calculations indicated that near-substrate H concentration should monotonically increase by almost a factor of 10 as Ts was lowered to 255⬚C, whereas the experimental data indicated only a modest rise in H. It was found that by including a simple radial diffusional loss term for H in the model to account for multidimensional effects, the predicted values for H at the substrate were brought into close quantitative agreement with experiment. The effect of substrate temperature on species distribution was also examined [188] for a different set of experimental data discussed in Sec. V.A [150]. As with the earlier study [27], it was found that the atomic hydrogen concentration was predicted to decrease drastically with increasing substrate temperature. In these calculations, for 600 ⱕ Ts ⱕ 1000 K, the mole fraction of CH3 at the surface rose by a factor of 5, whereas for substrate temperatures above 1000 K the CH3 mole fraction was only weakly dependent on Ts. By examining the predicted CH3 mole fractions at the surface with and without heterogeneous deposition chemistry, it was computed that for Ts ⱕ 950 K, there was a 14.2 kJ/mol activation energy for CH3 formation with surface chemistry and a 12.2 kJ/mol activation energy without surface chemistry. It was concluded that the exponential increase in CH3 mole fraction at the surface with increasing temperature was purely a gas-phase effect and could not be correlated with heterogeneous chemistry. Multidimensional, that is, two-dimensional and axisymmetric, calculations have been performed by a number of investigators to determine the role of heat and momentum transport in diamond growth and to relate predicted spatial nonuniformities to film growth rate and thickness variation. One of the first theoretical studies of heat transfer in a hot-filament diamond reactor was carried out by DebRoy et al. [156]. A vertical reactor was set up in that work and the filament was designed so that axisymmetric behavior could be assumed. It was found that, whether the filament was located above or below the substrate and whether the gases were introduced above or below the substrate, there was little variation in film growth rate. The model predicted the diffusion velocities of important species such as CH3 and H to be greater than any forced or natural convection velocities, demonstrating that concentration and thermal gradients are primarily responsible for species transport in hot-filament systems. This result has been confirmed by other investigators using one-dimensional models in which the inlet bulk velocity was varied by over an order of magnitude with no measurable effect on near-substrate composition or growth rate [28,188]. A subsequent study was carried out using the same axisymmetric reactor geometry to examine the

Deposition Chemistry

105

effect of atomic hydrogen on heat transfer [189]. It was determined that, for a given filament temperature, the substrate temperature in He was significantly lower than that in either pure H2 or a 1% CH4-H2 mixture and that small amounts of CH4 had no appreciable effect on the atomic hydrogen profile. In that work the predicted temperature profiles were in good agreement with measured values. Also, heterogeneous recombination of H was predicted to result in significant substrate heating due to its relatively high flux and the 4.5 eV exothermicity of the net recombination reaction to form H2. Because the sole energy source for the gas (and sometimes the substrate) typically consists of, at most, several discrete filaments, work has been done to study the effect of the filaments on substrate temperature uniformity. Wolden et al. [190] calculated temperature profiles in a system consisting of up to three parallel, cylindrical filaments oriented parallel to a silicon substrate in order to determine the relative contributions of radiation, conduction, and convection to substrate heating. It was found that predicted substrate temperature profiles were in reasonable agreement with measured values, and, in agreement with Tankala and DebRoy [189], it was determined that surface H recombination significantly affects absolute substrate temperature. Substrate temperatures were primarily determined by radiation flux, indicating that conduction and convection play secondary roles in the surface heat transfer process. Substrate temperature was predicted to be highest directly under the filaments, with appreciable gradients in the direction perpendicular to the filament axes. Through numerical experimentation it was found that a two-filament configuration could be designed that would yield relatively uniform substrate temperatures. A two-dimensional model containing a reduced gas-phase kinetic mechanism consisting of 10 species and 12 reactions has been developed [191]. Closed-form expressions for heterogeneous H production on the filament and for film growth rate were included in the model. Atomic hydrogen production on the filament was assumed to occur via reactions analogous to (1) and (2), and the relevant kinetic coefficients were adjusted to yield agreement with measured gas-phase H concentrations. At low pressure it was predicted that the H production rate on the filament was proportional to H2 concentration, and this result, coupled with tungsten filament data [192], yielded an H2 dissociation activation energy of 1.6–1.9 eV for filament temperatures between 2000 and 2500⬚C. At higher pressures, H concentrations appear to reach saturation. The two-dimensional reactor model was applied to three different experimental systems [27,149,150]. It was found that the temperature discontinuity at the filament had little effect on the results, which is consistent with earlier one-dimensional studies [28,186]. Also, total carbon near the filament was reduced relative to other regions of

106

Dandy and Coltrin

the reactor due to thermal diffusion effects. In general, it was concluded that the model adequately predicted the species’ concentrations and film growth rates in the different experimental systems. B.

Plasma-Assisted Reactors

Lack of quantitative data for temperature and composition in dc arcjet reactors has led investigators to use models to predict macroscopic observables such as film growth rate and to examine near-substrate gas composition for potential growth precursors. Using a modified mechanism with CH3 as the only growth precursor [59], Goodwin [193] modeled the growth environment of Raiche et al. [34] as a one-dimensional stagnation flow. It was determined that, although there were appreciable concentrations of atomic carbon near the substrate, the growth rate data of Raiche et al. could be adequately reproduced by the modified Harris mechanism. The uncertainty regarding the most probable growth precursor (CH3 versus C2H2) and the predicted high fraction of atomic carbon in dc arcjet systems (because of H supersaturation) led other investigators to propose a mechanism for the deposition of diamond and graphitic carbon in which all three species—CH3, C, and C2H2 —could participate [36]. The mechanism was incorporated into a stagnation flow model for a subatmospheric dc arcjet and used to investigate the dependence of growth rate and graphitic content on the degree of H2 dissociation and the CH4 concentration in the feed. It was predicted that CH3 was the predominant growth species when little of the H2 in the plasma torch underwent dissociation and that C became the dominant species at higher dissociation levels. The third growth species, C2H2, did not play a role in diamond growth under the conditions considered with less than 1% CH4 in the feed, but at higher input CH4 levels, the C and CH3 incorporation rates were only slightly higher than that of C2H2. In a subsequent study [194], the dependence of diamond growth rate on hydrocarbon injector location was modeled. As shown in Fig. 17, it was predicted that, for CH4 feed, growth rate could be increased by almost a factor of 2 by relocating the injector from a point near the plasma torch exit to just outside the substrate boundary layer. The growth rate increased as the injector was moved toward the substrate because this increased the near-substrate concentrations of C and CH3. An experimental study of the effect of injector location was carried out [165], and higher growth rate and improved film quality were observed when the hydrocarbon injector was located near the substrate. A two-dimensional model of the reacting gas flow, heat transfer, and electrodynamic phenomena in a subatmospheric dc arcjet has been developed [195,196]. Reaction rate coefficients for two irreversible reactions describing electron-assisted dissociation of H2 and CH4 were obtained in that

Deposition Chemistry

107

Figure 17 Predicted diamond growth rate as a function of hydrocarbon injector height above the substrate (Xinj) in a dc arcjet reactor for two different mole fractions of atomic hydrogen exiting the plasma torch. (From Ref. 194.)

work by solving the Boltzmann equation for the electron energy distribution function; the coefficients were then used in the gas-phase kinetic mechanism, which consisted of 15 species and 76 reactions. Instead of a detailed deposition mechanism, flux boundary conditions for H and CH3 were derived in closed form. The model predicted the temperature to peak along the discharge axis (the symmetry axis) and to decay smoothly in the radial direction to the electrodes. The model also predicted the deposition rate, as measured by CH3 flux, to vary significantly over the substrate surface. Interestingly, the flux nonuniformity was pressure dependent: at 150 torr the growth rate was minimum at the center and increased monotonically by over a factor of 2 at a radius of 1 cm. At 75 torr, however, the growth rate was maximum at the center, and it decreased by a factor of almost 3 at a radius of 0.8 cm

108

Dandy and Coltrin

before increasing again at distances greater than this. These results were attributed to the pressure dependence of the CH3 distribution in the reactor. Diamond deposition in an atmospheric-pressure rf plasma reactor has also been modeled [197,198]. A nonreacting (except for hydrogen chemistry), axisymmetric flow model coupled to two-dimensional electromagnetic field equations was used to calculate the velocity and temperature fields in the reactor. The temperature field was then used in a one-dimensional stagnation model to describe the processes occurring within the concentration boundary layer. The boundary layer thickness was chosen to lie at the 4000 K isotherm because chemical equilibrium was believed to exist at temperatures above this, that is, in the free stream. In the first study [197] a detailed surface mechanism was not used. Instead, it was assumed that H recombined at the surface to form H2, and C and C2 were assigned sticking probabilities of unity. In the second study [198] a detailed deposition mechanism was developed in which all CHx species (x = 0–3) were possible growth contributors. The predicted growth rate increased strongly as the boundary layer thickness decreased. For relatively thick boundary layers resulting from moderate jet velocities, growth was dominated by CH3, and for thin boundary layers growth was dominated by C. Because of the difficulties involved in modeling nonthermal plasma systems, fewer theoretical studies have been performed on microwave reactors than on dc or rf reactors. As a preliminary step in modeling the twodimensional (axisymmetric) behavior of microwave reactors, Hyman et al. [38] carried out detailed zero-dimensional calculations by solving the Boltzmann equation to determine the electron energy distribution. This equation was coupled to transient equations describing ion and neutral species chemistry. It was presumed that the simulation modeled the time-dependent reactor behavior at the center of the plasma ball; mass and energy transport effects were accounted for in this zero-dimensional model by including estimates for diffusion and conduction driving forces in the governing species and energy equations. For a reactor pressure of 40 torr and power deposition of 30 W/cm3, the model predicted a very rapid electron density buildup for the first several microseconds: it rose from 2 ⫻ 109 cm⫺3 at 0.01 ␮sec to 5 ⫻ 1011 cm⫺3 at 3 ␮sec, but increased by only another factor of 2 when the time increased to 2 msec. Once the gas reached its equilibrium temperature of ⬇3000 K after 2 msec, electron processes became unimportant relative to thermal processes in determining the chemical evolution of the gas. Atomic hydrogen concentration also reached equilibrium within 2 msec, constituting over 20% of the neutral gas mixture in a system that initially contained 98.75% H2, 1% CH4, and 0.25% O2. Most of the C1 species reached equilibrium values after approximately 4 msec: CO was the dominant C1 species, at approximately 0.3%, followed by C at 0.2%. Methyl

Deposition Chemistry

109

radical was present (at 0.01%) but at lower concentrations than both CH and CH2. Acetylene was predicted to be present in concentrations comparable to atomic carbon. The appreciable atomic hydrogen concentrations predicted to exist within the plasma ball are considerably higher than those found in laboratory microwave reactors [161] but may be characteristic of the conditions in the commercial ASTeX system being modeled. C.

Combustion Reactors

Analysis of combustion flame deposition has focused on the feasibility of using different hydrocarbon fuels, e.g., CH4, C2H2, C2H4, MAPP, and determination of the relationship between gas composition and film growth rate. Atmospheric deposition of diamond in premixed C2H2-O2-H2 and CH4-O2 strained flames has been modeled using a stagnation flow approximation [199]. Using a lightly sooting combustion mechanism [6] and a 35-reaction deposition mechanism, it was predicted that optimal growth conditions occurred when the burner-to-substrate distance was small, and the flame was lifted from the burner surface and stabilized on the deposition surface. The model was found to be in good agreement with experimental data available for the C2H2-O2-H2 flame [200]. It was also predicted that, because of the high gas velocities, temperatures in the C2H2 flame were higher than the theoretical adiabatic flame temperature; it was postulated that the reason for this behavior was the relatively long time required for C2H2 to dissociate to its equilibrium concentration. Concentrations of CH3 near the substrate were high enough in the C2H2 flame to account for observed growth rates. In contrast, temperatures in the CH4 /O2 flame were always lower than the adiabatic flame temperature, and while the gas velocity affected the flame location, it had little effect on peak temperature. In the CH4 flame, the fractions of CH3 and H near the substrate were approximately five times lower than those produced by the C2H2 flame. The bulk of the CH3 produced in the methane flame never reached the surface; instead, it was rapidly converted to C2H2 through bimolecular reactions. It was concluded that, even under optimal conditions, diamond growth rate was considerably lower in atmospheric methane flames than in acetylene flames. To increase film uniformity, combustion systems are often operated at relatively low pressure [183,184]. In a combined experimental and theoretical study of diamond growth, films were deposited at 35 torr to explore conditions that would simultaneously yield moderate growth rates, high film quality, uniformity, and large area coverage [201]. The gas was assumed to behave as an ideal stagnation flow, and the governing equations were coupled to a lightly sooting combustion mechanism [6] and a reduced deposition mechanism in which CH3 was the growth species and O and OH acted to

110

Dandy and Coltrin

abstract surface-bound hydrogen. The temperature profile predicted by the model was in close agreement with the profile obtained from OH rotational temperature data, and this was attributed in part to the large postflame reaction zone. Relative OH concentrations were obtained using LIF, and when these were normalized to the theoretical value at a single point (3 mm from the burner), the measured and predicted profiles were in good agreement. It was also found that, when normalizing the experimental OH concentrations, the presence or absence of heterogeneous chemistry in the model had no apparent effect. Because of this, the measured OH concentrations could not be used to evaluate the accuracy of the surface kinetic mechanism.

VII.

SUMMARY

This chapter has reviewed the state of understanding of the processes controlling the CVD of diamond. Gas-phase chemistry in this system is understood rather well, owing to much prior work in the combustion community. Much more contentious has been development of a mechanistic view of the elementary surface reactions leading to diamond growth. Indeed, our view is that a number of growth mechanisms and important growth precursor species can be operative depending upon details of the growth environment. Although there is still wide disagreement about the detailed, elementary steps in the deposition process, it appears that CVD diamond growth, and even defect formation, can be described quite well by reduced models. Such models, although not rigorous in detail, are useful in engineering reactorscaling studies and process optimization. Outstanding issues in our understanding of CVD diamond include elementary models for the differing growth rates observed on the different crystal faces of diamond. These different growth rates are key in controlling morphology and crystallite grain structure in polycrystalline diamond [202]. Another area still in need of better understanding is the initiation and control of diamond nucleation on nondiamond substrates. Although the nucleation of diamond has been widely characterized, fundamental and predictive models are still lacking.

ACKNOWLEDGMENTS It is a pleasure to acknowledge financial support of our work on CVD diamond by the Materials Science Program at DARPA, contract N00014-932002, and partial support has also been provided by Texas Instruments. The work performed at Sandia National Laboratories was also supported by the

Deposition Chemistry

111

United States Department of Energy under contract number DE-AC0494AL85000. We have benefited greatly from collaboration and discussions concerning this work with Drs. Richard Woodin, Robert Kee, Ellen Meeks, Wayne Weimer, Wen Hsu, Huimin Liu, James Butler, David Goodwin, and John Angus.

REFERENCES 1.

2. 3. 4. 5. 6. 7. 8.

9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

24. 25. 26.

J Warnatz, U Maas, RW Dibble. Combustion: Physical and Chemical Fundamentals, Modelling and Simulation, Experiments, Pollutant Formation. Berlin: Springer, 1996. J Warnatz. In: JWC Gardiner, ed. Combustion Chemistry. New York: Springer, 1984, p 197. CK Westbrook, WJ Pitz. Combust Sci Technol 37:117, 1984. JA Miller, CT Bowman. Prog Energy Combust Sci 15:287, 1989. DL Baulch, CJ Lobos, RA Cox, C Esser, P Frank, T Just, JA Kerr, MJ Rilling, J Troe, RW Walker, J Warnatz. J Phys Chem Ref Data 21:411, 1992. JA Miller, CF Melius. Combust Flame 91:21, 1992. A Burcat, WJ Pitz, CK Westbrook. Shock Tube Ignition of Octanes. UCRL102001, preprint, September 27, 1989. CT Bowman, RK Hanson, DF Davidson, WC Gardiner, V Lissianski, GP Smith, DM Golden, M Frenklach, M Goldenberg. GRI-Mech 2.11. http:// www.me.berkeley.edu/gri – mech/ M Frenklach. J Appl Phys 65:5142, 1989. C Wolden, KK Gleason, JB Howard. Combust Flame 96:75, 1994. M Frenklach, H Wang. Phys Rev B 43:1520, 1991. SJ Harris, AM Weiner. J Appl Phys 67:6520, 1990. SJ Harris, AM Weiner, TA Perry. J Appl Phys 70:1385, 1991. SJ Harris, AM Weiner, TA Perry. Appl Phys Lett 53:1605, 1988. WL Hsu. Appl Phys Lett 59:1427, 1991. I Langmuir. J Am Chem Soc 34:860, 1912. I Langmuir. J Am Chem Soc 34:1310, 1912. I Langmuir. J Am Chem Soc 37:417, 1915. S Matsumoto, Y Sato, M Tsatsumi, N Setaka. Jpn J Appl Phys 21:L183, 1982. S Matsumoto, Y Sato, M Tsatsumi, N Setaka. J Mater Sci 17:3106, 1982. F Jansen, I Chen, MA Machonkin. J Appl Phys 66:5749, 1989. U Meier, K Kohse-Hoinghaus, L Schafer, C Klages. Appl Opt 29:4993, 1990. TD Moustakas, JP Dismukes, L Ye, KR Walton, JT Tiedje. Polycrystalline diamond deposition from methane-hydrogen mixtures. Proceedings of Tenth International Conference on Chemical Vapor Deposition, Honolulu, 1987, p 1164. FG Celii, PE Pehrsson, H-T Wang, JE Butler. Appl Phys Lett 52:2043, 1988. FG Celii, JE Butler. Appl Phys Lett 54:1031, 1989. M Sommer, FW Smith. J Mater Res 5:2433, 1990.

112

Dandy and Coltrin

27. 28. 29. 30.

MC McMaster, WL Hsu, DS Dandy, ME Coltrin. J Appl Phys 76:7567, 1994. DS Dandy, ME Coltrin. J Appl Phys 76:3102, 1994. FM Cerio, WA Weimer. Rev Sci Instrum 63:2065, 1992. RL Woodin, LK Bigelow, GL Cann. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier Science Publishers, 1991, p 439. L Rongzhi, S Hailiang, Y Zhen, T Sen, Z Hesun. Transparent diamond film deposited by optimized DC arc plasma jet. Proceedings of Applications of Diamond Films and Related Materials, Auburn, AL, 1991, p 207. DS Dandy, ME Coltrin. J Mater Res 10:1993, 1995. N Koshino, K Kurihara, M Kawarada, K Sasaki. In: GH Johnson, AR Badzian, MW Geis, eds. Diamond and Diamond-Like Materials Synthesis. Pittsburgh: Materials Research Society, 1988, p 95. GA Raiche, GP Smith, JB Jeffries. Diagnostics of a diamond-depositing DCarc-jet plasma. Proceedings of New Diamond Science and Technology, Washington, DC, 1990, p 251. MH Loh, JG Liebeskind, MA Cappelli. Characterization of a supersonic hydrogen arcjet plasma thruster employed in diamond film synthesis. Proceedings of 29th Joint Propulsion Conference, Monterey, CA, 1993, p 2227. ME Coltrin, DS Dandy. J Appl Phys 74:5803, 1993. MH Loh, MA Cappelli. Study of precursor transport during diamond synthesis in a supersonic flow. Proceedings of Third International Symposium on Diamond Materials, Honolulu, 1993, p 17. E Hyman, K Tsang, A Drobot, B Lane, J Casey, R Post. J Vac Sci Technol A 12:1474, 1994. SP Mehandru, AB Anderson. J Mater Res 5:2286, 1990. SP Mehandru, AB Anderson, JC Angus. J Mater Res 7:689, 1992. DR Alfonso, SH Yang, DA Drabold. Phys Rev B 50:15369, 1994. DR Alfonso, SE Ulloa, DW Brenner. Phys Rev B 49:4948, 1994. OF Sankey, DJ Niklewsky, DA Drabold, JD Dow. Phys Rev B 41:12750, 1990. D Brenner. Phys Rev B 42:9458, 1990. JC Angus, Z Li, M Sunkara, C Lee, W Lambrecht, RL Li, B Segall. Diamond nucleation. Proceedings of Materials Research Society Symposium, Boston, 1993, p 128. J Tersoff. Phys Rev B 39:5566, 1989. WRL Lambrecht, CH Lee, B Segall, JC Angus, Z Li, M Sunkara. Nature 364:607, 1993. NL Allinger. J Am Chem Soc 99:8127, 1977. NL Allinger, YH Yuh, J Lii. J Am Chem Soc 111:8551, 1989. SJ Harris, DG Goodwin. J Phys Chem 97:23, 1993. YL Yang, MP D’Evelyn. J Vac Sci Technol A 10:978, 1992. M Zhu, RH Hauge, JL Margrave, MP D’Evelyn. Mechanism for diamond growth on flat and stepped diamond (100) surfaces. Proceedings of Third International Symposium on Diamond Materials, Honolulu, 1993, p 138. D Huang, M Frenklach. J Phys Chem 96:1868, 1992.

31.

32. 33.

34.

35.

36. 37.

38. 39. 40. 41. 42. 43. 44. 45.

46. 47. 48. 49. 50. 51. 52.

53.

Deposition Chemistry 54. 55. 56. 57. 58. 59. 60.

61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82.

83. 84. 85. 86. 87.

113

MJS Dewar, WJ Theil. J Am Chem Soc 99:4899, 1977. J Tersoff. Phys Rev B 37:6991, 1988. DR Alfonso, SE Ulloa. Phys Rev B 48:12235, 1993. BJ Garrison, EJ Dawnkaski, D Srivastava, DW Brenner. Science 255:835, 1992. M Frenklach, KE Spear. J Mater Res 3:133, 1988. SJ Harris. Appl Phys Lett 56:2298, 1990. DW Brenner, DH Robertson, RJ Carty, D Srivastava, BJ Garrison. In: JE Mark, ME Glicksman, SP Marsh, eds. Computational Methods in Materials Science. Pittsburgh: Materials Research Society, 1992, p 255. SJ Harris, DN Belton. Thin Solid Films 212:193, 1992. DG Goodwin. J Appl Phys 74:6888, 1993. LN Krasnoperov, IJ Kalinovski, H Chu, D Gutman. J Phys Chem 97:11787, 1993. SJ Harris, AM Weiner. J Appl Phys 74:1022, 1993. S Skokov, B Weiner, M Frenklach. J Phys Chem 98:8, 1994. S Skokov, B Weiner, M Frenklach. J Phys Chem 99:5616, 1995. MA Cappelli, MH Loh. Diamond Relat Mater 3:417, 1994. CJ Chu, MP D’Evelyn, RH Hauge, JL Margrave. J Appl Phys 70:1695, 1991. MP D’Evelyn, CJ Chu, RH Hauge, JL Margrave. J Appl Phys 71:1528, 1992. SJ Harris, AM Weiner. Thin Solid Films 212:201, 1992. LR Martin, MW Hill. J Mater Sci Lett 9:621, 1990. CE Johnson, WA Weimer, FM Cerio. J Mater Res 7:1427, 1992. WA Yarbrough, K Tankala, T DebRoy. J Mater Sci 7:379, 1992. S Skokov, B Weiner, M Frenklach. J Phys Chem 98:7073, 1994. M Frenklach, S Skokov, B Weiner. Nature 372:535, 1994. MR Pederson, KA Jackson, WE Pickett. Phys Rev B 44:3891, 1991. D Huang, M Frenklach. J Phys Chem 95:3692, 1991. JE Butler, RL Woodin. Philos Trans R Soc 342:209, 1993. T Tsuno, T Imai, Y Nishibayashi, K Hamada, N Fujimori. Jpn J Appl Phys 30:1063, 1991. H-G Busmann, W Zimmermann-Edling, H Sprang, H-J Guntherodt, IV Hertel. Diamond Relat Mater 1:979, 1992. LF Sutcu, CJ Chu, MS Thompson, RH Hauge, JL Margrave, MP D’Evelyn. J Appl Phys 71:5930, 1992. RE Thomas, RA Rudder, RJ Markunas. Thermal desorption from hydrogenated diamond (100) surfaces. Proceedings of Second International Symposium on Diamond Materials, Washington, DC, 1991, p 186. SJ Harris, AM Weiner. J Appl Phys 75:5026, 1994. C Wild, N Herres, P Koidl. J Appl Phys 68:973, 1990. KV Ravi. J Mater Res 7:384, 1992. WJP van Enckevort, G Janssen, W Vollenberg, JJ Schermer, LJ Giling, M Seal. Diamond Relat Mater 2:997, 1993. A Badzian, T Badzian. High temperature epitaxy of diamond. Proceedings of Third International Symposium on Diamond Materials, Honolulu, 1993, p 1060.

114

Dandy and Coltrin

88.

M Frenklach, S Skokov, B Weiner. On the role of surface diffusion in diamond growth. Proceedings of Fourth International Symposium on Diamond Materials, Reno, NV, 1995, p 1. JC Angus, EA Evans. Necleation and growth processes during the chemical vapor deposition of diamond. Proceedings of Materials Research Society Symposium, Boston, 1994, p 385. M Frenklach, R Kematick, D Huang, W Howard, KE Spear, AW Phelps, R Koba. J Appl Phys 66:395, 1989. M Frenklach, W Howard, D Huang, J Yuan, KE Spear, R Koba. Appl Phys Lett 59:546, 1991. W Howard, D Huang, J Yuan, M Frenklach, KE Spear, R Koba, AW Phelps. J Appl Phys 68:1247, 1990. S Mitura. J Cryst Growth 80:417, 1987. BV Derjaguin, DV Fedoseev. Sci Am 233:102, 1975. S Matsumoto, Y Matsui. J Mater Sci 18:1785, 1983. JC Angus, JA Mann, Y Wang, M Sunkara, PA Washlock, M Kuczmarski, Z Li. Progress Report, DARPA University Research Initiation Grant. Contract N00014-89J-3167, January 31, 1991. JC Angus, RW Hoffman, PH Schmidt. In: S Saito, O Fukunaga, M Yoshikawa, eds. Science and Technology of New Diamond. Tokyo: KTK Terra Scientific, 1990, p 9. PR Buerki, S Leutwyler. J Appl Phys 69:3739, 1991. G Popovici, MA Prelas. Phys Status Solidi A 132:233, 1992. E Kondoh, K Tanaka, T Ohta. J Appl Phys 74:2030, 1993. S Katsumata. Jpn J Appl Phys 31:3594, 1992. ZM Zhang, HM Cheng, SH Li, QY Cai, DL Ling, SJ Wang, ZW Hu, SS Jiang, CZ Ge, NB Ming. J Cryst Growth 132:200, 1993. WP Chai, YS Gu, M Li, H Mai, QZ Li, L Yuan, SJ Pang. J Cryst Growth 135:639, 1994. WA Yarbrough. J Vac Sci Technol A 9:1145, 1991. RA Bauer, NM Sbrockey, WE Brower Jr. J Mater Res 8:2858, 1993. A van der Drift. Philips Res Rep 22:267, 1967. BR Stoner, BE Williams, SD Wolter, K Nishimura, JT Glass. J Mater Res 7: 257, 1992. BV Spitzyn, BV Derjaguin. In: LL Bouilov, ed. Problems of Physics and Technology of Wide-Gap Semiconductors. Leningrad: Akad Nauk SSSR, 1979, p 22. BV Spitzyn, LL Bouilov, BV Derjaguin. J Cryst Growth 52:219, 1981. C Trevor, D Cherns, P Southworth. TEM studies of the nucleation of CVD diamond on silicon. Proceedings of Institute of Physics Electron Microscopy and Analysis Group Conference, Bristol, UK, 1991, p 275. D Kim, H Lee, J Lee. J Mater Res 28:6704, 1993. SM Kanetkar, AA Kulkarni, A Vaidya, RD Vispute, SB Ogale, ST Kshirsagar, SC Purandare. Appl Phys Lett 63:740, 1993. MT McClure, JA von Windheim, JT Glass, JT Prater. Diamond Relat Mater 3:239, 1994.

89.

90. 91. 92. 93. 94. 95. 96.

97.

98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108.

109. 110.

111. 112. 113.

Deposition Chemistry 114. 115. 116. 117.

118. 119.

120. 121.

122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133.

134. 135. 136. 137. 138. 139.

115

DN Belton, SJ Schmieg. Thin Solid Films 212:68, 1992. K Kobayashi, M Kumagai, S Karasawa, T Watanabe, F Togashi. J Cryst Growth 128:408, 1993. GA Hirata, L Cota-Araiza, M Avalos-Borja, MH Farias, O Contreras, Y Matsumoto. J Phys 5:305, 1993. BY Lin, CP Beetz, DW Brown, BA Lincoln. The influence of diamond surface perfection on the preferential nucleation of sp 2 carbon during methane pyrolysis. Proceedings of Evolution of Surface and Thin Film Microstructure Symposium, Pittsburgh, 1993, p 705. AR Badzian, T Badzian. Surf Coat Technol 36:283, 1988. JA Baglio, BC Farnsworth, S Hankin, C Sung, J Hefter, M Tabasky. In: RE Clausing, LL Horton, JC Angus, P Koidl, eds. Diamond and Diamond-Like Films and Coatings. New York: Plenum Press, 1991, p 635. PO Joffreau, R Haubner, B Lux. Int J Ref Hard Mater 7:186, 1988. B Lux, R Haubner. In: RE Clausing, LL Horton, JC Angus, P Koidl, eds. Diamond and Diamond-Like Films and Coatings. New York: Plenum Press, 1991, p 579. WA Yarbrough, R Messier. Science 247:688, 1990. M Komori, T Maki, T Kim, G Hou, Y Sakaguchi, K Sakuta, T Kobayashi. Appl Phys Lett 62:582, 1993. PC Yang, W Zhu, JT Glass. J Mater Res 8:1773, 1993. AR Badzian, T Badzian. Proceedings of Materials Research Society Symposium Proceedings, Boston, 1991, p 339. J Narayan, VP Godbole, CW White. Science 252:416, 1991. M Moore. Ind Diamond Rev 2:67, 1985. SD Wolter, BR Stoner, JT Glass, PJ Ellis, DS Buhaenko, CE Jenkins, P Southworth. Appl Phys Lett 62:1215, 1993. BR Stoner, SR Sahaida, JP Bade, P Southworth, PJ Ellis. J Mater Res 8:1334, 1993. C Wild, R Kohl, N Herres, W Mu¨ller-Sebert, P Koidl. Diamond Relat Mater 3:373, 1994. JS Ma, H Kawarada, T Yonehara, JI Suzuki, J Wei, Y Yokota, A Hiraki. J Cryst Growth 99:1206, 1990. JS Ma, H Kawarada, T Yonehara, JI Suzuki, J Wei, Y Yokota, A Hiaraki. Appl Phys Lett 55:1071, 1989. WA Yarbrough. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier Science Publishers, 1991, p 25. PA Dennig, H Shiomi, DA Stevenson, NM Johnson. Thin Solid Films 212: 63, 1992. A Sawabe, T Inuzuka. Thin Solid Films 137:189, 1986. M Katoh, M Aoki, H Kawarada. Jpn J Appl Phys 33:L194, 1994. BR Stoner, GHM Ma, SD Wolter, JT Glass. Phys Rev B 45:11067, 1992. Y Hayashi, W Drawl, R Messier. Jpn J Appl Phys 31:L193, 1992. S Haq, JA Savage, DL Tunnicliffe. In Y Tzeng, M Yoshikawa, M Murakawa,

116

140.

141. 142. 143. 144. 145.

146. 147. 148. 149. 150. 151. 152. 153. 154. 155. 156. 157.

158. 159.

160. 161. 162. 163. 164. 165.

Dandy and Coltrin A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier Science Publishers, 1991, p 405. JW Kim, YJ Baik, KY Eun. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier Science Publishers, 1991, p 399. JJ Chang, D Mantei, R Vuppuladhadium, HE Jackson. Appl Phys Lett 59: 1170, 1991. P Bou, L Vandenbulcke, R Herbin, F Hillion. J Mater Res 7:2151, 1992. RC Hyer, M Green, KK Mishra, SC Sharma. J Mater Sci Lett 10:515, 1991. SI Shah, MM Waite. Appl Phys Lett 61:3113, 1992. HR Thorsheim, JE Butler. In: KE Spear, JP Dismukes, eds. Synthetic Diamond: Emerging CVD Science and Technology. New York: John Wiley & Sons, 1994, p 193. T Kawato, K Kondo. Jpn J Appl Phys 26:1429, 1987. SJ Harris, DN Belton, AM Weiner. J Appl Phys 66:5353, 1989. CH Wu, MA Tamor, TJ Potter, EW Kaiser. J Appl Phys 68:4825, 1990. L Schafer, C-P Klages, U Meier, K Kohse-Hoinghaus. Appl Phys Lett 58: 571, 1991. K-H Chen, M-C Chuang, CM Penney, WF Banholzer. J Appl Phys 71:1485, 1992. FG Celii, JE Butler. In: R Messier, JE Butler, JT Glass, eds. New Diamond Science and Technology. Pittsburgh: Materials Research Society, 1991, p 201. EJ Corat, DG Goodwin. J Appl Phys 74:2021, 1993. EH Wahl, TG Owano, CH Kruger, P Zalicki, Y Ma, RN Zare. Diamond Relat Mater 5:373, 1996. P Zalicki, RN Zare. J Chem Phys 102:2708, 1995. SO Hay, WC Roman, MB Colket III. J Mater Res 5:2387, 1990. T DebRoy, K Tankala, WA Yarbrough, R Messier. J Appl Phys 68:2424, 1990. UE Meier, LE Hunziker, DR Crosley, JB Jeffries. Observation of OH radicals in a filament-assisted diamond growth environment. Proceedings of Second International Symposium on Diamond Materials, Washington, DC, 1991, p 202. K Yarina. Low pressure combustion CVD of diamond. MS thesis, Colorado State University, 1996. LS Plano, DA Stevenson, JR Carruthers. Characterization of DC plasmas for the control of diamond deposition. Proceedings of Second International Symposium on Diamond Materials, Washington, DC, 1991, p 290. A Inspektor, Y Liou, T McKenna, R Messier. Surf Coat Technol 39/40:211, 1989. MC McMaster, WL Hsu, ME Coltrin, DS Dandy, C Fox. Diamond Relat Mater 4:1000, 1995. KR Stalder, W Homsi. Appl Phys Lett 68:3710, 1996. JA Mucha, DL Flamm, DE Ibbotson. J Appl Phys 65:3448, 1989. Y Muranaka, H Yamashita, K Sato, H Miyadera. J Appl Phys 67:6247, 1990. SW Reeve, WA Weimer, DS Dandy. J Mater Res 11:694, 1996.

Deposition Chemistry 166. 167. 168.

169. 170. 171. 172. 173. 174. 175. 176. 177. 178. 179. 180. 181. 182.

183. 184. 185. 186. 187.

188. 189. 190. 191.

192. 193. 194. 195. 196.

117

W Zhu, A Inspektor, AR Badzian, T McKenna, R Messier. J Appl Phys 68: 1489, 1990. FG Celii, HR Thornsheim, JE Butler, LS Plano, JM Pinneo. J Appl Phys 68: 3814, 1990. HR Thorsheim, FG Celii, JE Butler, LS Plano, JM Pinneo. In: R Roy, R Messier, JE Butler, JT Glass, eds. New Diamond Science and Technology. Pittsburgh: Materials Research Society, 1991, p 207. DS Green, TG Owanao, S Williams, DG Goodwin, RN Zare, CH Kruger. Science 259:1726, 1993. T Mitomo, T Ohta, E Kondoh, K Ohtsuka, Y Habu. J Appl Phys 70:4532, 1991. KR Stalder, RL Sharpless. J Appl Phys 68:6187, 1990. SW Reeve, WA Weimer. J Vac Sci Technol A 12:3131, 1994. HN Chu, EA Den Hartog, AR Lefkow, J Jacobs, LW Anderson, MG Lagally, JE Lawler. Phys Rev A 44:3796, 1991. GA Raiche, JB Jeffries. Appl Opt 32:4629, 1993. GP Smith, JB Jeffries. Electrochem Soc Proc 91-8:194, 1991. SW Reeve, WA Weimer. J Vac Sci Technol A 13:359, 1995. RS Yalamanchi, KS Harshavardhan. J Appl Phys 68:5941, 1990. Y Hirose, S Amanuma, K Komaki. J Appl Phys 68:6401, 1990. K Komaki, Y Masaaki, I Yamamoto, Y Hirose. Jpn J Appl Phys 32:1814, 1993. Y Matsui, A Yuuki, M Sahara, Y Hirose. Jpn J Appl Phys 28:1718, 1989. Y Matsui, H Yabe, Y Hirose. Jpn J Appl Phys 29:1552, 1990. PW Morrison, JE Cosgrove, JE Markham, PR Solomon. In: R Roy, R Messier, JE Butler, JT Glass, eds. New Diamond Science and Technology. Pittsburgh: Materials Research Society, 1991, p 219. NG Glumac, DG Goodwin. Mater Lett 18:119, 1993. JS Kim, MA Cappelli. Appl Phys Lett 65:2786, 1994. KF McCarty, E Meeks, RJ Kee, AE Lutz. Appl Phys Lett 63:1498, 1993. DG Goodwin, GG Gavillet. J Appl Phys 68:6393, 1990. WL Hsu. Quantitative analysis of the gaseous composition during filamentassisted diamond growth. Proceedings of Second International Symposium on Diamond Materials, Washington, DC, 1991, p 217. B Ruf, F Behrendt, O Deutschmann, J Warnatz. J Appl Phys 79:7256, 1996. K Tankala, T DebRoy. J Appl Phys 72:712, 1992. C Wolden, S Mitra, KK Gleason. J Appl Phys 72:3750, 1992. YA Mankelevich, AT Rakhimov, NV Suetin. Two-dimensional model of a HFCVD reactor. Proceedings of Fourth International Symposium on Diamond Materials, Reno, NV, 1995, p 687. T Otsuka, M Ihara, H Komiyama. J Appl Phys 77:893, 1995. DG Goodwin. Appl Phys Lett 59:277, 1991. DS Dandy, ME Coltrin. Appl Phys Lett 66:391, 1995. YA Mankelevich, AT Rakhimov, NV Suetin. Diamond Relat Mater 4:1065, 1995. YA Mankelevich, AT Rakhimov, NV Suetin. Plasma Phys Rep 21:872, 1995.

118

Dandy and Coltrin

197. 198. 199. 200. 201. 202.

SL Girshick, BW Yu, C Li, HH Han. Diamond Relat Mater 2:1090, 1993. BW Yu, SL Girshick. J Appl Phys 75:3914, 1994. E Meeks, RJ Kee, DS Dandy, ME Coltrin. Combust Flame 92:144, 1993. M Murayama, S Kojima, K Uchida. J Appl Phys 69:7924, 1991. NG Glumac, DG Goodwin. Combust Flame 105:321, 1996. C Wild, P Koidl, W Mu¨ller-Sebert, H Walcher, R Kohl, N Herres, R Locher, R Samlenski, R Brenn. Diamond Relat Mater 2:158, 1993. M Yoshikawa. Diamond Films Technol 1:1, 1991.

203.

5 Thermally Assisted (Hot-Filament) Deposition of Diamond Joseph E. Yehoda Diamonex, Inc., Allentown, Pennsylvania

I.

INTRODUCTION

The earliest reports of CVD diamond growth by hot-filament chemical vapor deposition (HFCVD) were published in the early 1980s by Matsumoto et al. [1,2]. A simple W filament was used in a flow tube, the substrate being heated both by the filament and by an external furnace. Since that time, the number of papers dealing with HFCVD has increased manyfold. Figure 1 shows the number of yearly publications dealing with hotfilament, thermal filament, or HFCVD as compiled in the DIAmond Database [3]. It shows an incubation period of approximately 5 years followed by a very rapid increase as research groups within and outside Japan discovered HFCVD. In this chapter, I will give a brief description of HFCVD followed by a more detailed look at the practical aspects of this deposition technique, focusing on filament processes, substrate heating effects, filament contamination, and distribution of species. Other hybrid methods of HFCVD will be discussed along with issues of reactor scale-up. Other reviews dealing with HFCVD have already been published [4,5]. The growth chemistry will not be specifically addressed here. The reader is referred to Angus et al. [6], Butler and Woodin [7], and Goodwin and Butler [8] for a more detailed treatment. II.

DESCRIPTION

Hot-filament CVD can be described as a light bulb. In very much the same way a light bulb uses a refractory filament in a low-pressure environment, 119

120

Yehoda

Figure 1 Graph showing the number of publications dealing with hot-filament or thermal filament or HFCVD versus the year in which they were published. The data for 1999 are only through the first 6 months.

HFCVD also uses a filament, activated with either a dc or ac current, to cause the filament to glow (Fig. 2). At temperatures above approximately 1700–1800⬚C, the hydrogen in this low-pressure atmosphere undergoes a thermal dissociation reaction. It is this dissociation of molecular hydrogen into its atomic species that makes diamond growth possible. The atomic hydrogen (H 0) can then interact with the small percentage of hydrocarbon gas, frequently CH 4 , to form methyl radicals (CH3) via the reaction CH 4 ⫹ H 0 → CH3 ⫹ H2

(1)

This provides the most important active species for diamond growth [9–11]. The H 0 also has two other primary functions: (1) to etch nondiamond carbon from the growing film and (2) to surface terminate in H to prevent its collapse into a reconstructed graphitic surface, essentially shutting down the diamond growth completely.

Thermally Assisted Deposition of Diamond

Figure 2

III.

121

Schematic diagram of the HFCVD process showing the basic elements.

FILAMENTS

The primary role of the filaments in HFCVD is to provide a means of dissociating diatomic hydrogen (H2) into atomic hydrogen via the reaction H2 ⇔ H 0 ⫹ H 0 ⫺ 4.5 eV

(2)

From Eq. (2), it can be seen that the forward dissociation reaction is endothermic, absorbing energy as it occurs. Unlike conventional CVD processes that rely solely on thermal energy to initiate a reaction, for CVD diamond growth H 0 is a key ingredient. The filament also participates in decomposition of the hydrocarbon species but more indirectly through the production of H 0. The filaments also provide a secondary role, heating the substrate to the deposition temperature for diamond growth. Alternative means of heating or cooling can also be implemented to decouple the thermal effects of the filament on the substrate, but this cannot be completely decoupled because of the role of H 0 recombination [reverse of Eq. (2)]. The advantage of in-

122

Yehoda

dependently controlling the substrate temperature is to maintain optimum growth conditions. As will be seen in Sect. III.C, the heating of the substrate is primarily through H 0 recombination on the surface and secondarily through radiation from the incandescent filaments. Because Eq. (2) is endothermic, the energy requirements for dissociation should be visible during the CVD process, specifically on the energy supplied to the filaments. Jansen et al. [12] showed the effects of an unreactive gas such as He, compared with H2 and D2 , on the power dissipation of Re filaments. Compared with vacuum, there was only a small increase in power dissipated by the filaments, via convection and radiation, when He was used. When a gas such as H2 or D2 was introduced, the power dissipated in the filaments to maintain a constant filament temperature increased threefold. The result was the generation of H 0 and the energy expended, in dissociation of the diatomic hydrogen, was reflected in the increased power dissipation of the filaments; see Eq. (2). Thus the onset of power loss was associated with H 0 atom production. Sommer and Smith [13] investigated the effects of filament emissivity, filament resistance, and power consumption for W and Re filaments in CH 4-H2 and C2H2-H2 as gas chemistry systems. The filaments were observed to become ‘‘active’’ at a temperature ⱖ 1800⬚C. Evidence for this activity was the production of H 0 through thermal dissociation [Eq. (2)]. In other works [14,15], a similar effect of thermal dissociation of hydrogen on the power supplied to the filaments was observed. Some energy loss can happen at lower temperatures but is believed to be associated with increased vibrational and translational energy pathways for H2 rather than dissociation processes [14]. Again, comparison of power loss of Re filaments in vacuum, He, D2 , and H2 showed the onset of power loss associated with the endothermic hydrogen dissociation reaction [14], similar to that which was observed for W filaments [13]. A.

Filament Material

For a filament material to be a good candidate for HFCVD, it should have several important properties: (1) a high melting temperature of the metal or carbide, (2) resistance to attack by H 0, and (3) mechanical stability at the operating temperatures required for diamond growth. To date, the most successfully used filament materials have been W [13,16–18], Ta [17–19], and Re [12,14,18]. Other materials such as Pt [20] and C [21,22], have also been attempted but with only limited or no success. W and Ta are good carbide formers and also high-temperature materials. Carbon has low solubility in Re, so Re does not carburize, but its melting temperature is very high (3180⬚C). Table 1 gives the melting tem-

Thermally Assisted Deposition of Diamond

123

Table 1 Melting Point of Various Filament Materials and Their Carbides Filament material

Tmelting (⬚C) 3410 2860 ⬇2600 2996 ⬇3400 3880 3180

W W2C WC Ta Ta2C TaC Re

peratures of different filament materials and their carbides. Ta can be operated at higher temperatures than W by virtue of its higher temperature carbide, and it is also mechanically less prone to deformation than W [17]. Re has been shown to retain its shape better than other filament materials [18], giving a longer lifetime; however, its cost may be prohibitive. To a large extent, the choice of filament material has been shown to be independent of most deposition parameters [18]. Where it has been shown to make a difference, though, is in what methane concentration, [CH 4 ], and filament temperature result in a maximum growth rate. This has beens shown to depend on filament material [18]. Before the peak in growth rate is observed, the growth rates between similar filament materials appear similar [18,20].

B.

Filament Processes

Filaments such as Ta and W require a carburization step before they can be effectively used for diamond deposition. Because they are good carbide formers, and carbon incorporation is unavoidable, a carburization step helps stabilize the filaments before growth is started. This carburization step is normally performed for two reasons. The first is that because the lattice constants of the carbides are typically larger than that of the pure metal, the filaments will distort dimensionally, lengthening for straight filaments or twisting for spiral filaments. This can change the filament geometry, subsequently affecting the growth. The second reason for carburization is to stabilize the filament electrically. The subcarbides and carbides have different electrical resistivities than the pure metals. A consequence of this is a change in the power required to maintain the filaments at a constant temperature, Tfil . As an example, the resistivities of Ta (12.5 ␮ ⍀ cm), Ta2C (80

124

Yehoda

␮ ⍀ cm), and TaC (25 ␮ ⍀ cm) are seen to vary significantly from one another. Until the filament has reached equilibrium, because of the material changes caused by carburization along with the effects of the temperature coefficient of resistance, the power required for maintaining a constant filament temperature will change. Carburization is performed by heating the filaments in an atmosphere of hydrogen and a small percentage of carbon-containing gas, typically CH4. During this process the filament ‘‘removes’’ C from the ambient atmosphere of the reactor, forming a carbide from the surface inward. In the early stages of carburization, the filament can be composed of three distinct materials: a metal core surrounded by a subcarbide annulus that in turn is surrounded by a carbide ring. Figure 3 shows this schematically. The process can be tracked by analysis of the residual gases within the chamber, where the CH4 can be observed to decrease as the filament is heated.

Figure 3 Representation of a filament at successive stages of the carburization process. (a) The pure metal (M) filament; (b) partially carburized filament showing a metal core surrounded by a subcarbide annulus (M2C) followed by the metal carbide outer layer (MC); (c) a fully carburized filament composed of only the metal carbide.

Thermally Assisted Deposition of Diamond

125

A negative effect of this carbide formation is an increase in the brittleness of the filaments. This is particularly problematic with W and Ta and less so with Re due to its lack of carbide formation. This can shorten the lifetime of the filaments, especially if the filaments are cycled over large temperature ranges, e.g., room temperature to several thousand degrees Celsius. C.

Filament Effects on Substrate Heating

Just as the forward reaction in Eq. (2) requires energy, its reverse reaction will deposit energy during the recombination of H 0 into H2 . The effects of substrate heating during HFCVD have been studied, experimentally and through modeling [15,23–25]. Homogeneous chemical interactions and diffusional mixing govern the [H 0 ] in the gas phase [15]. Simple probes, constructed using thermocouples, have been utilized to detect H 0 recombination (heating) [15,22,24]. Measurements of the [H 0 ] and Tprobe were made as a function of filament-substrate distance. Studies of the Peclet numbers for mass transport and heat transfer showed that (1) diffusive rather than convective mass transport was dominant and (2) conductive heat transport dominates over convective transport [23]. Using mass transport as an example, the Peclet number is defined as the ratio of the convective mass transport to the diffusive transport: Pe =

convective uL = diffusive D

(3)

Values of Pe ⬇ 0.08 << 1 were determined, implying that diffusion transport dominates. The [H 0 ] values that were measured and modeled in pure H2 and 1% CH 4 in H2 agreed well with each other. In the case of CH 4 addition, the amount of [H 0 ] present was observed to be less than that in pure H2 . Homogeneous gas-phase reactions were ruled out while changes in the filament were suspected. Further verification of homogeneous gas-phase reactions not being significant was provided by the [H 0 ] profile shape not changing from pure H2 to additions of 1% CH 4 [24]. Power inputs into the filaments were observed to decrease with the methane additions as well as an indication of less [H 0 ] being generated [12]. In a comprehensive study of substrate heating effects, H2 , He, N2 , and Ar were used as ambients (without hydrocarbons) to characterize the substrate heating effects [25]. Temperature increases as high as ⬇250⬚C were observed in H2 over He or the other gases, H 0 recombination being the cause. Similar temperature increases have been measured and also ascribed to H2 recombination [22]. Pressure effects were also studied and showed an op-

126

Yehoda

timum range of 25–30 torr to operate within. At pressures higher than this, gas-phase recombination started to dominate over substrate surface recombination. For lower pressures, it was suggested that mixed flow regimes of viscous and molecular flow contributed to less efficient transport of H 0 to the substrate. For the geometry of the filament and substrate in this work, less importance was given to heat transport through radiation, but for more conventional geometry, radiation was believed to be more important. Heat flux in the reactor toward the substrate can be classified in order of importance: H 0 recombination > thermal radiation >> conduction > convection. Modeling of a three-filament system showed a 14% increase in Tsub when comparing the effects of H 0 recombination over radiation [26]. The same effect of substrate heating due to H 0 recombination has been observed in Diamonex’s 12-in. reactors. In this case, the substrate temperature in vacuum is approximately 500⬚C lower than that when H2 is introduced into the process chamber at typical filament temperatures for CVD diamond growth. D.

Filament Contamination

The main source of impurities in HFCVD films is incorporated hydrogen and metal from the filaments. Hydrogen can be incorporated from a fraction of a percent to several percent as an impurity [27–30]. This is not totally unexpected given that the typical growth ambient can be anywhere from 90 to 99% H2 . Metal contamination is normally lower, several tens of parts per million atomic (ppm-a) to hundreds of ppm-a. Using elastic recoil spectroscopy, differences in the [H] between the growth and substrate sides of thick (500 ␮m) diamond films were measured [30]. Both as-grown and polished samples were investigated. It was found that for the unpolished case, the substrate side had approximately 3.4 times more H incorporated (6.5% to 1.9%). When both the growth and substrate sides were polished, the substrate side was larger by only about 1.6 times, 4.2% for the substrate side and 2.6% for the growth side. The larger concentration on the substrate side was attributed to more nondiamond carbon that was believed to be where the H was bonding. One should remember that the substrate side, by virtue of nucleation and columnar growth, has a higher number of grain boundaries and associated defects. This can provide a greater number of sites where incorporated H can reside. The effect of the surface roughness on [H] was believed to be partially a geometrical effect of H being trapped by the roughness as it was ejected from the surface region toward the detector. Although this work used microwave CVD for the diamond deposition, the salient features of [H] in the diamond are believed to be appropriate to HFCVD as well.

Thermally Assisted Deposition of Diamond

127

The question of how much incorporated H comes from the H2 gas and how much is from the hydrocarbon gas was investigated using isotopic labeling studies utilizing H2 , CH 4 , deuteruim (D2), and deuterated methane (CD4) during growth [27]. Both microwave CVD and HFCVD were used, with little difference found in the level of incorporated H. Experiments suggested that the incorporated hydrogen did not originate from H attached to the hydrocarbon gas. However, the results pointed to another, as yet undefined, source of H contributing to incorporation. One of the consequences of the filaments being operated at high temperature is the evaporation of filament material during the diamond deposition process. With the exception of Re filaments, which do not form carbides, the other filament materials such as W and Ta evaporate less material as a carbide than as a pure metal; this is a consequence of higher melting temperatures and lower vapor pressures for the carbides. Impurities have been characterized by a number of methods; SIMS, RBS, EDX, and NAA. SIMS is one of the most popular and sensitive techniques. Matrix effects and the lack of reliable ‘‘standards,’’ however, usually complicate its quantitative interpretation. RBS and EDX lack sensitivity for anything but the most contaminated films. NAA is quantifiable, with the only stipulation that access to a source of neutrons is necessary. There is not an extensive literature on filament contamination. It was recognized and characterized early for Re filaments [12], where large amounts of Re (up to 2000 ppm-a) were detected via SIMS. The level of filament contamination was observed to be a strong function of filament temperature. Ta and W impurities have been detected in HFCVD films deposited on Si [31]. RBS showed W contamination from 10 to 80 ppm-a at 2000⬚C, whereas for an equivalent filament temperature for Ta, the contamination was over 1000 ppm-a. It was also found that the Ta contamination was significantly reduced if a precarburized filament was used. The lower evaporation rate of TaC compared with Ta was believed to be the cause. The distribution of metal contamination in the HFCVD diamond films has been observed to decrease as the film grows thicker [32]. The main explanation for this phenomenon is the early nucleation, coalescence, and growth of diamond. Because the first stages of growth involve a nucleation stage, the growth of diamond is slow compared with the evaporation rate of metal from the filaments. As the film coalesces to form a continuous film, the growth rate stabilizes and reaches equilibrium with the evaporation rate of filament material. The net result is a continuous contamination level throughout the rest of the film. More recent studies have used NAA to quantify the level of contamination in HFCVD diamond films [33]. It was found that the highest levels of metal contamination occurred with Re filaments, followed by Ta and then

128

Yehoda

W. This is shown in Table 2 for diamond deposited using these three different filament materials. Of the three filament materials, Re exhibits the highest level of contamination, by three to four orders of magnitude. Others [12] have observed the higher levels of Re contamination as well. As one would expect, higher filament temperatures resulted in higher levels of contamination via an increased evaporation rate: 0.279 ␮g cm⫺2 sec⫺1 for TaC at 2400⬚C compared with 0.00248 ␮g cm⫺2 sec⫺1 for TaC at 1950⬚C [33]. It is interesting to note that the calculated evaporation rate of WC, at 1950⬚C, was approximately a factor of 2 higher than that of TaC at the same filament temperature. However, its contamination level was substantially lower, 4.6 ppm-m (0.3 ppm-a) for W compared with 37 ppm-m (2.5 ppm-a) for Ta. The explanation for this difference, which has been observed by others, was the difference in how WC evaporates compared with TaC [33]. In WC, the C primarily evaporates, whereas for TaC, the Ta and C evaporate more or less equally. Gas chemistry effects on metal contamination have also been investigated to a limited extent. The obvious consequence for the addition of O2 , or other gases that allow operation at lower Tfil , is the lower evaporation rate that is present. Care needs to be exercised in using O2 in an HFCVD reactor, however. It has been shown that W and Ta filaments are stable against oxidation with added oxygen as long as a gas composition of C/(C ⫹ O) > 0.5 is maintained [34]. E.

Reactions with the Filaments

As long as the filament surface remains free of deposited C, which is excess carbon not involved in the carburization of the filaments, the dissociation of

Table 2 Metal Contamination, in Parts Per Million Atomic (ppm-a), in CVD Diamond Deposited Using the Three Different Filament Materials: Re, W, and Ta Contamination (ppm-a) Filament Re W Ta

Re

W

Ta

348 0.039 0.11

NA 0.30 0.17

<MDA 0.026 2.5

MDA for Re is 0.01 ppm-a. Source: Data courtesy of Dr. Robert Shaw, Oak Ridge National Laboratory.

Thermally Assisted Deposition of Diamond

129

H2 into atomic hydrogen proceeds without interruption. When the level of carbon in the gas phase becomes greater than a critical value, dependent on Tfil , deposition of carbon onto the filament surface occurs [13,19,35,36]. This deposition is commonly referred to as ‘‘poisoning’’ or ‘‘sooting.’’ Filament poisoning results in a reduced ability to transform H2 to H 0. The effects are reduced growth rate, morphology change, and reduced crystal size in the deposited film [36]. Sommer and Smith [37] thermodynamically modeled the filament as a substrate and defined a phase field where graphitic deposition occurred. This was done for the CH 4 /H2 and C2H2 /H2 chemistries using W and Re filaments [13]. Addition of reactive gases, such as oxygen, has been found to lower the Tfil where C deposition occurs. The same chemistries with the addition of O2 were also modeled [38]. It was observed that at sufficiently high filament temperatures, no deposited C phase occurred on the filaments. However, when the filament temperature was low and/or at higher [CH 4 ], the filament would become covered with a graphitic deposit. This graphitic deposit, commonly called ‘‘soot,’’ is really a microcrystalline C. This type of surface is no longer conducive to the dissociation of diatomic hydrogen, effectively shutting off the source. This has been found in attempts to use carbon filaments [21,22]. When the filament poisoning is not too advanced, cleaning at high Tfil and low or no added CH 4 can in most cases clean the filaments to such a degree that they can still be used. If the C layer is too thick, however, no amount of H 0 cleaning will lead to recovery. A hysterisis in the filament poisoning has been observed and characterized in terms of filament temperature, gas flow, and [CH 4 ] [36]. Figure 4 shows this for three different methane flows, where the filament temperature is observed to increase as the filament surface becomes covered with soot. Others have also seen this change in filament behavior as the [CH 4 ] was changed [13,19,38]. It has been observed that growth with poisoned filaments will result in a smaller grain size as well as lower growth rates. In most cases, the onset of changes in the filaments will be detected by a sudden apparent increase in the Tfil (Fig. 4), followed by a decrease in the substrate temperature. These two seemingly contradictory behaviors can easily be explained as follows. As the filaments start to build up a layer of carbon on their surface, the emissivity of the filaments is increased. This increase is perceived by an optical pyrometer, typically used to measure Tfil , as an increase in temperature. The subsequent decrease in Tsub is the result of the lower level of H 0 production, and as shown earlier in Sec. III.C, the H 0 recombination on the substrate is the primary mode of heating. Through careful monitoring of the decomposition conditions, the region of filament poisoning can be avoided.

130

Yehoda

Figure 4 Data showing the hysteresis in filament temperature as the methane fraction is varied for different total flows. This behavior is related to coverage of the filament by nondiamond carbon. (From Ref. 36.)

IV.

DETECTION AND DISTRIBUTION OF SPECIES

The detection of species during the HFCVD growth process has long been of interest. By probing the environment around the filaments and substrate, researchers have hoped to gain more insight into the growth process. A variety of probes have been used for detection of species at the filament and substrate. Infrared (IR) laser absorption has been used to detect species during HFCVD using 0.5% CH 4 in H2 [39]. Acetylene (C2H2) was found to be the most abundant species compared with CH 4 : 1.6 ⫻ 1015 cm⫺3 versus 2 ⫻ 1014 cm⫺3. The presence of CH3 was also detected with an estimated concentration of 5 ⫻ 1013 cm⫺3. Other species such as ethylene (C2H4) have measured concentrations of 6 ⫻ 1012 cm⫺3. The detection of C2H2 and CH3 could support either as a precursor to growth of CVD dia-

Thermally Assisted Deposition of Diamond

131

mond. Other hydrocarbon species were not detected by this method because of its sensitivity level. Methods that rely on sampling of the growth gases have been used. Mass spectrometry was used to sample the surface/near–surface region of a Si wafer during growth [40]. A chemical kinetics model was then used to assist in interpreting the measurements. Calculations were done as a function of distance for CH 4 in H2 . CH 4 and C2H2 were monitored. Because of the short lifetime of some species, they could not be detected directly by mass spectrometry. Indirect monitoring could be used, assuming that the presence of certain peaks was mainly due to them. For example, 2H = H2 2CH3 = C 2H 6 C 2H = C 2H2 This is one of the shortcomings of this type of technique. The gas temperature was also measured as a function of distance from the filaments. Using conventional thermocouples, the gas temperature was found to drop rapidly, to ⬇600 K, at a distance of 1 mm from the filaments. From their measurements and modeling, it was suggested that CH3 or C2H2 is primarily responsible for growth. Another optical method using lasers was used to detect the distribution of H 0 in HFCVD [41]. Resonance-enhanced multiphoton ionization, better known as REMPI, uses a focused laser to excite different levels in a species that are then detected. Profiles of H 0 versus Tfil showed that increasing filament temperatures resulted in higher levels of atomic hydrogen. Addition of CH 4 showed that the [H 0 ] decreased as the [CH 4 ] was increased. For the highest levels of [CH 4 ] this behavior was found to be reversed, the result of poisoned filaments. Although no proof of the growth species could be obtained, measurements of the [H 0 ] opposite the substrate showed behavior indicative of diffusional transport. Flow rate variations also supported this type of transport. Molecular beam mass spectrometry has also been used to measure growth species at the substrate surface [42]. Unlike conventional mass spectroscopy, this method was able to detect free radical as well as stable species. Using three parallel filaments, concentrations of H, H2 , CH3 , CH 4 , and C2H2 were measured. Measurements of H 0 at the substrate confirmed a ‘‘superequilibrium’’ state over what the normal H 0 equilibrium mole fraction would be for the substrate temperature. At [CH 4 ] ⱕ 1%, the [H 0 ] was stable, whereas [H 0 ] was found to decrease dramatically for [CH 4 ] > 1%, confirm-

132

Yehoda

ing what has been observed before [41]. This drop in [H 0 ] was more than an order of magnitude for an increase from 0.4 to 7.2% CH 4 . At the lower concentrations of methane, C2H2 was observed to be the dominant species. This was found to change as [CH 4 ] was increased; [C2H2] decreased with [CH 4 ] dominating and [CH3 ] steadily increasing, but at a slower rate as [CH 4 ] increased further. It was determined that there was a difference in the concentration of species if a single filament was used as opposed to an array. The reasoning was that a single filament allows more CH 4 to flow past without interacting with the filament. This is most obvious in the lower [C2H2 ] and [CH3 ] being detected. This also reconciled differences in species concentrations measured by others [41]. Coherent anti-Stokes Raman spectroscopy (CARS) has been used to measure both concentrations of species and temperature in rf-PACVD and HFCVD [43]. Upstream and downstream measurements were made of [CH4] from the filaments as a function of filament temperature. The lower concentration of CH 4 observed downstream was attributed to rarefaction of the gas by the filaments. Lifetimes of the H 0, before recombination, were found to be significantly longer (⬇0.5 sec) than those of other species that had lifetimes on the order of 100 ␮sec. This shows the dominance of H 0 as a species at the substrate. Two-photon laser-induced fluorescence (LIF) has been used to detect [H 0 ] profiles as a function of filament-to-substrate distance in HFCVD [44]. Measurements over a total pressure range of 1.5 to 100 mbar (1.1 to 75 torr) were made. Pressure effects were observed and found to give a saturation in [H 0 ] versus distance for pressures >10 mbar (75 torr). At a lower pressure, 1.5 mbar, the [H 0 ] was significantly lower, nearly 50% at the filament. Higher filament temperatures resulted in higher concentrations of H 0, falling off monotonically as a function of distance from the filament. The effects of filament diameter were explored for Ta filaments of 0.3, 1.0, and 2.0 mm diameter. At constant filament temperature and % CH 4 , the larger diameter filament resulted in a higher level of H 0 generation. This is believed to be an effect of available surface area; the larger surface area can support a higher number of dissociation events per unit area. As in previous studies [22,41], addition of CH 4 resulted in a reduction of [H 0 ]. The effects of excessive methane concentration were also observed here. The filament temperature was observed to increase by 80⬚C while the [H 0 ] decreased rapidly. The Tfil effect is not a true temperature change but a change in emissivity brought about by sooting of the filaments as explained before. The change in [H 0 ] was thought to be due to consumption in the gas phase. This supports later observations of the same effect [36].

Thermally Assisted Deposition of Diamond

133

Other materials besides Ta were investigated—W and Ir—but carburized Ta was reported to generate higher [H 0 ] than Ir. CARS has been used to determine [H2 ] and temperature distributions in HFCVD as a function of distance from the filaments [45]. Temperature distribution measurements of the H2 were made as a function of distance from the filaments for three different powers. It was observed that there was a substantial drop in temperature as one moved away from the filaments. Similarly, the gas temperature above the substrate was found to be over 500 K hotter than at the substrate surface. [H 0 ] values were calculated from the temperature distributions and the [H2 ] from CARS. Deficits in expected [H2 ] were attributed to [H 0 ]. [H 0 ] fractions as a function of distance from the filaments and filament power were shown. Mole fractions of 50 to 20% were calculated for distance at the filament to ⬇7 mm away (1% CH4 ). It was shown that the [H 0 ] is in superequilibrium except very near the filaments when compared with thermal equilibrium values, verifying earlier work [42]. There was also a substantial drop between the gas temperature and the filaments and substrate. Others [42] have used two-dimensional finite element modeling [46] to interpret HFCVD experiments. Gas-phase chemistry, filament catalysis, surface interaction for growth, and how these affect [H 0 ] generation-recombination were considered. Based on the model, gas-phase chemistry was determined not to be primarily responsible for the [H 0 ]. Only a 30% drop in [H 0 ] at the substrate, compared with the [H 0 ] at the filament, implied that gas-phase recombination is not dominant. From others [41], two activation energies of ⬇150 kJ/mole for Tfil > 2250 K and 330 kJ/mole for Tfil < 2100 K were found. The activation energy at the higher temperature was believed to be due to a chemical reaction. An expression for the H 0 production was developed. The saturation effect observed by others was predicted by this expression. In addition, the decrease in [H 0 ] when CH 4 was added was predicted. Fourier transform infrared (FTIR) analysis of the gas during growth has also been used in an attempt to study the gas species in an HF reactor with a rapidly rotating substrate [47]. Rapid substrate rotation was studied at various deposition pressures and found to affect only the Raman signal. At 200 torr the quality improved, but no effect was observed at 20 and 400 torr. At 20 torr the pumping action of rotation on the gas is marginal at best, but at higher pressures this improves. A competing effect of H 0 recombination opposed the uniformity of H 0 species over the substrate at the highest pressures. Absolute [CH3 ] values have been measured using cavity ring-down spectroscopy (CRDS) [48,49]. In CRDS a laser pulse is reflected back and forth between two mirrors. These form a cavity where the laser pulse is

134

Yehoda

circulated between the mirrors. Depending on the absorption of the medium within the cavity, the decay will be governed by this, and the decay time (ring-down time) can be determined. From this absorption the concentration can then be calculated by knowing the cross section for interaction. The [CH3 ] shows a maximum at ⬇2 mm from the filaments and is dependent on substrate temperatures. This maximum, seen in Fig. 5, is related to thermal diffusion effects, Sore´t diffusion. The higher Tsub results in higher [CH3 ] for the same Tfil . Detailed measurements near the substrate surface showed an increase in [CH3 ] by a factor of 2.8 for an increase in Tsub of ⬇2: 600 to 1200⬚C, a linear increase. An activation energy of 4.2 ⫾ 0.2 kcal mole⫺1 was found that agrees with other measurements [50]. Measurements of [CH3 ] at constant Tsub with Tfil from 2000 to 2550 K showed an increase followed by a saturation in [CH3 ] at ⬇2500 K. Others have also observed this [50,51].

Figure 5 Profiles of CH3 radical concentration, as a function of distance from a W filament, for two different substrate temperatures using CRDS. The gas flow is from left to right. Note the strong dependence of [CH3 ] on substrate temperature. The lines through the data represent modeled results. (From Ref. 49.)

Thermally Assisted Deposition of Diamond

135

This study used a single W filament. This could be the difference in some of the measurements; more CH 4 is able to move around the filament and not be affected by its presence. This behavior was observed in other work [42].

V.

HYBRID METHODS OF HFCVD

In addition to the conventional form of HFCVD, other hybrid methods have been developed. These have typically involved applying a bias to the substrate, either positive or negative, and generally dc. Radio frequency biasing has been reported [52], but there has not been much follow-on work in this area. Some of the earliest work on bias-assisted HFCVD was done using a positively biased substrate [53–61] and is referred to as electron-assisted CVD (EACVD). As the name implies, electrons from the hot filaments are drawn to the substrate when a positive bias is applied. The filaments are strong sources of thermonic electrons at the normal operating temperature for HFCVD. High growth rates, 3 to 7 ␮m/hr, and enhanced nucleation densities have been reported when using EACVD [53,54,58]. In the earliest work on EACVD, growth rates of 3–5 ␮m/hr were reported for bias voltages of 150 V [53,54]. Ten-fold increases in nucleation were also observed. Attempts to apply negative bias were not successful. It was speculated that the growth enhancement took place on the substrate surface and not in the gas phase because of the small probability of interaction between electrons and gas molecules. The increase in nucleation density was also cited as related more to surface processes than to gas-phase processes. The biased HFCVD process was further characterized electrically as a diode; the filaments are strong emitters of electrons, with electrons flowing easily toward the substrate, whereas the substrate is not an electron source [56]. Positive and negative biases were applied and the results studied. It was common, regardless of polarity, that nothing occurred until a plasma was formed and that higher levels of bias were necessary to achieve this. Different morphologies were observed depending on the bias conditions, and the reverse bias case resulted in nonuniform deposits. Using EACVD, thick diamond films of 500 ␮m have been grown at rates of 7 ␮m/hr [58]. Again, the presence of a plasma was found. It was suggested that greater H 0 generation occurred, and it was not e⫺ bombardment that was responsible for the growth rate. Raman and IR measurements showed characteristics of diamond films, with properties improving as the film thickness increased.

136

Yehoda

To further understand the mechanisms of EACVD, optical emission spectroscopy (OES) and IR absorption measurements were carried out during HFCVD with positive applied bias to evaluate the nucleation density [60]. OES measurements showed no increase in hydrocarbon species with applied bias and a very minor increase in H 0. This increase was not enough to account for the observed increase in nucleation density. The IR absorption intensity also showed no increase for growth species. Thus, it was concluded that the gas phase was not affected even though an increase in nucleation density was observed. Surface processes were postulated to be responsible. Negative biasing has also been applied to the substrate [62]. Nucleation enhancement was observed; however, it was nonuniform across the substrate, being higher around the edges. This effect was attributed to the substrate holder providing a more conductive path for charged species compared with the Si-wafer substrate. Subsequent modifications resulted in uniform nucleation. Nucleation densities were observed to increase from 104 to 109 cm⫺2. It was believed that the negative biasing results in positive ion bom⫹ bardment of the substrate by C ⫹, CH⫹ 3 , H , etc. species. Scanning electron microscopy (SEM) cross sections showed a disordered layer below the diamond film. Building upon this observation, selective area nucleation was achieved by using a grid above the substrate. The mesh grid acted as a ‘‘funnel,’’ directing the species responsible for the nucleation in a pattern. The grid mesh size was found to have a direct effect on the selective area nucleation density. In the case of EACVD, the bias is normally applied during the entire deposition. For a negative applied substrate bias, such as what is used for biased-enhanced nucleation (BEN), the bias is applied only during the initial stages of growth until sufficient nuclei are formed. At this point it is turned off, and normal HFCVD growth is continued. Spectroscopic ellipsometric study of the initial stages of biased-enhanced CVD identified four stages of the process [63]. These were carburization, etching, incubation, and nucleation. Observation revealed that during these four distinct processes, the plasma changed in a systematic way as well. Another unique hybrid method employing a hot filament has been used to deposit diamond. Good-quality CVD diamond was obtained, using a substrate that rotates sequentially between a hot filament and a graphite sputtering target [64]. The best diamond with this method was achieved under conditions of high [H 0 ]. Aside from the unique character of this method, combining a CVD and physical vapor deposition (PVD) technique, it attests to the importance of surface reactions of CVD diamond deposition and the importance of H 0.

Thermally Assisted Deposition of Diamond

VI.

137

SCALE-UP AND COMMERCIALIZATION OF HFCVD REACTORS

Hot-filament reactors lend themselves to scale-up. Unlike microwave reactors, hot-filament reactors have no fundamental restriction of wavelength to limit the usable area over which to deposit [12]. The important considerations are filament diameter and shape, filament array size, and power requirements. Although the scale-up is conceptually straightforward, dealing with problems of the stability of large filament arrays and the movement that occurs is a challenge. The issues of filament spacing and filament-to-substrate distance have been addressed [12,65]. Ideally, it is desirable to maintain the spacing between the filaments to approximately the filament-to-substrate distance [65]. This allows a uniform distribution of H 0 to impinge onto the substrate, resulting in more uniform deposits of diamond. In addition to Diamonex’s 12-in. HFCVD reactors [66], others have developed HFCVD reactors for deposition over large areas for application in thermal management [29,66] (Fig. 6) and tool coatings for planar as well as round tools [67–69].

Figure 6 CVD diamond used as a heat spreader in a multichip module substrate for a supercomputer thermal management application. Note the 20-mm-square unpolished diamond pieces in the foreground and the polished pieces on the board. (Courtesy of Dr. Paul J. Boudreaux, Laboratory for Physical Sciences, College Park, MD.)

138

VII.

Yehoda

CONCLUSION

Considerable knowledge has been gained since the early attempts at HFCVD. Large multizoned reactors have been developed for deposition over areas as large as 12 inches in diameter for such commercial applications as thermal management, tools, and protective coatings. Growth rates have also increased. With the application of active cooling during deposition, to carry away the heat of recombination and radiation, filaments operating near 2800⬚C are possible. Rates approaching ⬇19 ␮m/ hr have been reported [19] for single-filament systems and ⬇5 ␮m/hr for multifilament ones [70]. Unique configurations have allowed depositions onto spherical surfaces [71] as well. With continued scale-up to larger areas and increases in deposition uniformity and rates, the focus now shifts to improvements in other areas of diamond processing. For CVD diamond to become a commodity material in the thermal management arena and other areas, the whole processing stream requires simultaneous development to open commercial markets and carry us forward.

ACKNOWLEDGMENTS I would like to thank Dr. Bob Shaw for providing data on the ppm-a contamination, Dr. Daniel Morel for providing data to replot Fig. 4, Mr. Edward Wahl for Fig. 5 plus additional literature, and Dr. Paul Boudreaux for supplying Fig. 6. In addition, I would like to thank Diamonex, Incorporated for providing me with the time and encouragement to write this chapter and Dr. Fred Kimock for his suggestions while proofreading the manuscript.

REFERENCES 1. 2. 3. 4. 5. 6. 7.

S Matsumoto, Y Sato, M Kamo, N Setaka. Jpn J Appl Phys 21:L183, 1982. S Matsumoto, Y Sato, M Tsutsumi, N Setaka. J Mater Sci 17:3106, 1982. Data from DIAmond Database, Danish Technical Institute. The search criterion of hot filament or thermal filament or HFCVD was used. R Haubner, B Lux. Diamond Relat Mater 2:1277, 1993. C-P Klages, L Scha¨fer. In: B Dischler, C Wild, eds. Low Pressure Synthetic Diamond. Berlin: Springer-Verlag, 1998, p 85. JC Angus, A Argoitia, R Gat, Z Li, M Sunkara, L Wang, Y Wang. Philos Trans R Soc Lond A 342:195, 1993. JE Butler, RC Woodin. Philos Trans R Soc Lond A 342:209, 1993.

Thermally Assisted Deposition of Diamond 8.

9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

139

DG Goodwin, JE Butler. In: MA Prelas, G Popovici, LK Bigelow, eds. Handbook of Industrial Diamonds and Diamond Films. New York: Marcel Dekker, 1997, p 527. CJ Chu, MP D’Evelyn, RH Hauge, JL Margrave. J Mater Res 5:2405, 1990. E Kondoh, T Ohta, T Mitomo, K Ohtsuka. J Appl Phys 72:705, 1992. EJ Corat, DG Goodwin. J Appl Phys 74:2021, 1993. F Jansen, MA Machonkin, DE Kuhman. J Vac Sci Technol A8:3785, 1990. M Sommer, FW Smith. J Mater Res 5:2433, 1990. F Jansen, I Chen, MA Machonkin. J Appl Phys 66:5749, 1989. K Tankala, T DebRoy. J Appl Phys 72:712, 1992. TD Moustakas. Solid State Ionics 32/33:861, 1989. H Matsubara, T Sakuma. J Mater Sci 25:4472, 1990. S Okoli, R Haubner, B Lux. J Phys II 1:923, 1991. J Bru¨ckner, T Ma¨ntyla¨. Diamond Relat Mater 2:373, 1993. B Singh, Y Arie, AW Levine, OR Mesker. Appl Phys Lett 52:451, 1988. MC Mecray. Master’s thesis, The Pennsylvania State University, 1991. WA Yarbrough, K Tankala, M Mecray, T DebRoy. Appl Phys Lett 60:2068, 1992. T Debroy, K Tankala, WA Yarbrough, R Messier. J Appl Phys 68:2424, 1990. K Tankala, T DebRoy. Surf Coat Technol 62:349, 1993. R Gat, JC Angus. J Appl Phys 74:5981, 1993. K Tankala, T DebRoy, WA Yarbrough, CJ Robinson. Diamond Relat Mater 1: 1177, 1992. DC Ingram, JC Keay, C Tang, ML Lake, J-M Ting. Diamond Relat Mater 2: 1414, 1993. KM Rutledge, KK Gleason. Chem Vap Deposition 2:37, 1996. CP Schaffer, IC Chen, RL Sturdivant, AT Hunter, RG Wilson. Diamond Relat Mater 7:585, 1998. A Kimura, Y Nakatani, K Yamada, T Suzuki. Diamond Relat Mater 8:37, 1999. H-J Hinneberg, M Eck, K Schmidt. Diamond Relat Mater 1:810, 1992. M Griesser, G Stingeder, M Grasserbauer, H Baumann, F Link, P Wurzinger, H Lux, R Haubner, B Lux. Diamond Relat Mater 3:638, 1994. P Mehta Menon, A Edwards, CS Feigrle, RW Shaw, DW Coffey, L Heatherly, RE Clausing, L Robinson, DC Glasgow. Diamond Relat Mater 8:101, 1999. WD Cassidy III. Master’s thesis, Case Western Reserve University, 1995. DM Li, R Hernberg, T Ma¨ntyla¨. Diamond and Relat Mater 7:188, 1998. D Morel, W Ha¨nni. Diamond Relat Mater 7:826, 1998. M Sommer, FW Smith. Solid State Commun 69:775, 1989. M Sommer, FW Smith. J Vac Sci Technol A9:1134, 1991. FG Celii, PE Pehrsson, H-T Wang, JE Butler. Appl Phys Lett 52:2043, 1998. SJ Harris, AM Weiner. Appl Phys Lett 53:1605, 1998. FG Celii, JE Butler. Appl Phys Lett 54:1031, 1989. WL Hsu. Appl Phys Lett 59:1427, 1991. SO Hay, WC Roman, MB Colket. J Mater Res 5:2387, 1990. L Sha¨fer, C-P Klages, U Meier, KK Ho¨inghaus. Appl Phys Lett 58:571, 1991.

140 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58.

59. 60. 61. 62. 63. 64. 65.

66. 67. 68. 69.

70. 71.

Yehoda KH Chen, MC Chuang, CM Penney, WF Banholzer. J Appl Phys 71:1485, 1992. C Wolden, KK Gleason. Appl Phys Lett 62:2329, 1993. PW Morrison, O Taweechokesupsin, CS Kovach, B Roozbehani, JC Angus. Diamond Relat Mater 5:242, 1996. EH Wahl, TG Owano, CH Kruger, P Zalicki, Y Ma, RN Zare. Diamond Relat Mater 5:373, 1996. EH Wahl, TG Owano, CH Kruger, Y Ma, P Zalicki, RN Zare. Diamond Relat Mater 6:476, 1997. EJ Corant, EG Goodwin. J Appl Phys 74:2021, 1993. E Kondoh, T Ohta, T Mitomo, K Ohtsuka. J Appl Phys 72:705, 1992. T Komatsu, H Yamashita, Y Tamon, N Kikuchi. ISPC-8, Tokyo, 1987, 2487. A Swabe, T Innzuka. Appl Phys Lett 46:146, 1985. A Swabe, T Innzuka. Thin Solid Films 137:89, 1986. H Kaneko, M Kamada, R Kuwae, A Sawabe, T Inuzunka. Appl Surf Sci 33/ 34:546, 1988. K Kobayashi, S Karasawa, T Watanabe, F Togashi. J Cryst Growth 99:1211, 1990. YH Lee, PD Richaard, KJ Bachmann, JT Glass. Appl Phys Lett 56:620, 1990. H Li, M Mecray, W Yarbrough, XH Wang. In: R Messier et al., eds. New Diamond Science and Technologh (ICNDST-2). Proceedings of 2nd International Conference September 23–27, 1990, Washington, DC, MRS Conf Proc, 1991, p 461. G Popovici, CH Chao, MA Prelas, EJ Charlson, JM Meese. J Mater Res 10: 2011, 1995. J Cui, R Fang. Appl Phys Lett 69:3507, 1996. GS Fu, XH Wang, W Yu, L Han, LF Dong, XW Li. Appl Phys Lett 70:1965, 1997. W Zhu, FR Sivazlian, BR Stoner, JT Glass. J Mater Res 10:425, 1995. X Li, Y Hayashi, S Nishino. Diamond Relat Mater 6:126, 1998. MA Kelly, DS Olson, S Kapoor, SB Hagstrom. Appl Phys Lett 60:2502, 1992. L Scha¨fer, C-P Klages. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Diamond Films and Related Materials. Amsterdam: Elsevier Science Publications, 1991, p 453. Diamonex, Incorporated, 7331 William Ave., Allentown, PA 18106. Sp3, 505 East Evelyn Ave., Mountain View, CA 94041. CEME CON GmbH, Talbotstraße 21, D-52068, Aachen, Germany. T Leyendecker, O Lemmer, S Esser, M Frank. In: A Feldman, Y Tzeng, WA Yarbrough, M Yoshikawa, M Murakawa, eds. Applications of Diamond Films and Related Materials: 3rd International Conference, 1995, p 183. DM Li, T Ma¨ntyla¨, R Hernberg, J Levoska. Diamond Relat Mater 5:350, 1996. B Lunn, DA Wright, LY Zhang. Diamond Relat Mater 7:129, 1998.

6 Plasma Torch Diamond Deposition Joachim V. R. Heberlein University of Minnesota, Minneapolis, Minnesota

Naoto Ohtake Tokyo Institute of Technology, Tokyo, Japan

I.

INTRODUCTION

Deposition of diamond films using a thermal plasma has some process characteristics that are fundamentally different from those of all the other deposition methods. The thermal plasma generated either by an electric arc or by a radio frequency induction discharge (rfi), has extremely high energy densities (up to 108 J/m3) and temperatures (typically above 10,000 K), and this feature allows rapid dissociation of any substance or gas and high-rate generation of vapor-phase deposition precursors. As a consequence, high deposition rates are achieved and deposition at atmospheric pressure is possible. The advantages of thermal plasma chemical vapor deposition (TPCVD) have been recognized for the deposition of silicon carbide, silicon, and high-Tc superconductors [1], but only the promise of attaining high rate deposition of diamond using this process has led to strong further development efforts. Both thermal plasma generation methods, rfi and dc arc, were used to deposit diamond early on [2,3], but, at present, rf discharge reactors are more prominently used for process characterization, while the efforts in scaling the process to commercial size concentrate on dc arc reactors. For many applications of diamond coatings, economics is an overriding issue. Examples are diamond-coated tools or diamond heat spreaders in microelectronic applications. Economic studies of the scaling of the diamond deposition process have shown that the total deposition rate (in g/hr) is the most important factor in determining the cost of diamond-coated tools 141

142

Heberlein and Ohtake

[4], and for these applications, arcjet deposition is the most favorable coating method. Linear growth rates of up to 1 mm/hr and total deposition rates of more than 0.2 g/hr have been reported in small 10-kW reactors [5]. In addition, the carbon utilization efficiency (carbon in the diamond film over carbon introduced into the system) with more than 8% and the specific energy requirement with less than 43 kWhr/g [6] are more favorable than with any other method [7], leaving the cost of the higher gas flow rates the one operational variable that is higher than with other deposition methods. Because of the issue of economics, most research in arcjet diamond deposition has focused on developing an understanding of the factors influencing the deposition rate. However, following the directions of the investigations with the other diamond deposition methods, there is now a strong research effort addressing items such as control of film texture, control of the film-substrate interface properties to improve adhesion on substrates of interest, and deposition at lower temperatures. In this chapter, we first describe the concept of arcjet diamond deposition in detail and then review the operational characteristics of arcjet plasma generators. A review of the principles of the arcjet diamond deposition process follows using representative results of models and of diagnostic experiments, including descriptions of the influence that various operating parameters have on the film growth. This is followed by a review of the most significant developments of deposition reactors. Concluding, we present a brief analysis of some of the present research directions.

II.

DESCRIPTION OF CONCEPT

In plasma torch diamond deposition, the torch generates a thermal plasma consisting of a mixture of argon and hydrogen in varying proportions providing a source of atomic hydrogen and of heat for dissociating the deposition precursors. There exist numerous different configurations through which this task is achieved. Figure 1 shows a schematic of a typical system consisting of a dc arc between a rod type cathode and a cylindrical nozzletype anode. The plasma gas mixture enters the arcing region from the base of the cathode, is heated and expanded by the arc discharge, and leaves the anode nozzle at a high temperature and low density and with a high velocity. In the plasma jet, heat exchange with the surroundings lead to a cooling of the plasma, a reduction of the jet velocity, and a recombination of the ionized and dissociated species. Figure 2 shows some typical contours of temperatures and velocities in an argon-hydrogen plasma jet obtained with a plasma torch with a 6-mm-diameter nozzle and operated at a power level of 40 kW [8]. The temperature and velocity distributions can vary widely depending

Plasma Torch Diamond Deposition

Figure 1

143

Schematic of thermal plasma CVD process.

on the torch power, plasma gas composition, and chamber pressure. Arc powers can range from 100 W to more than 100 kW, and 40 to 90% of the electric power is transferred to the plasma, depending on the specific design and the operating conditions. Argon-hydrogen ratios can also vary widely from 10% hydrogen to 100%. Increasing the amounts of hydrogen usually leads to increased electrode erosion with possible film contamination and to lower density jets with larger boundary layer thicknesses. Deposition precursors are usually added downstream of the nozzle because most hydrocarbons can form volatile compounds with thermionic cathode materials, leading to strong erosion. Deposition of diamond requires the presence of atomic hydrogen and either atomic carbon or hydrocarbon radicals. Consequently, the substrate must be placed at a position where the hydrogen in the plasma and the deposition precursors are largely dissociated, i.e., where the temperature in the jet is above about 4000 K if equilibrium conditions exist in the jet. The deposition process depends on the rates of the recombination processes in the boundary layer. Ideally, one would like to have frozen chemistry in the boundary layer to preserve high concentrations of atomic hydrogen and of atomic carbon, and the thinner the boundary layer the more closely will this condition be approximated.

144

Heberlein and Ohtake

Figure 2 Temperature and velocity fields in an argon-hydrogen plasma jet at 1 atm. The torch nozzle exit is at z = 0. (From Ref. 8.)

The fluid dynamic boundary layer thickness ⭸ is a function of the free stream velocity u, and the gas density ␳ 0 and viscosity ␮ 0: ⭸ ⬀ (␮ 0 /␳ 0u)0.5

(1)

Besides the fluid dynamic boundary layer, one needs to consider the thermal boundary layer, and the concentration boundary layer. For the conditions of a high-temperature, high-velocity free jet impinging on a cooled substrate, the velocity and thermal boundary layers are often considered to coincide. However, for a mixture of gases with quite different densities and diffusion coefficients, such as argon and hydrogen, the concentration profile in the boundary layer may differ from the velocity profile. This fact is expressed by a value for the Schmidt number, Sc = ␮ 0 /( ␳ 0 D) (where D is the diffusion coefficient) that is not equal to one. Figure 3 illustrates the importance of the boundary layer thickness control. The distributions of the important chemical species in the boundary

Plasma Torch Diamond Deposition

145

Figure 3 Temperature (a) and mole fraction (b) profiles for selected species in the substrate boundary layer for two different free stream velocities of an argonhydrogen-methane jet resulting in two different boundary layer thicknesses. (From Ref. 9.)

146

Heberlein and Ohtake

layer of an argon-hydrogen-methane jet for two different free stream velocities and corresponding boundary layer thicknesses are shown [9]. The boundary layer edge has been assumed to coincide with the 4000 K isotherm, and equilibrium conditions are assumed at that location. It is apparent that the atomic hydrogen concentration at the substrate surface increases by an order of magnitude if the boundary layer thickness is decreased from 2 to 0.4 mm. Furthermore, the deposition mechanism can change during this decrease in boundary layer thickness from CH3 dominated to C dominated. This fact is illustrated in Fig. 4, where mole fractions of the important species for diamond growth at the substrate location are shown (Fig. 4a), as well as the growth rates for the different precursor species, both as a function of boundary layer thickness [9]. Besides controlling the boundary layer thickness by increasing the plasma velocity (i.e., increasing torch power or plasma gas flow rate or decreasing the chamber pressure), increasing the jet density can have similar effects (i.e., increasing the chamber pressure or changing the type of gas for the same velocity). One can have the paradoxical situation that reducing the amount of hydrogen in the jet and replacing it with the same volumetric flow rate of argon can actually lead to higher atomic hydrogen densities at the substrate surface due to stronger boundary layer compression [10]. Typical boundary layer thicknesses encountered in atmospheric pressure plasma jet diamond deposition range from approximately 0.5 to 2 mm. Lower pressures in the deposition chamber will increase the boundary layer thickness for the same jet velocity but also reduce the chemical reactions in the boundary layer. It is obvious that control of the boundary layer conditions is crucial for the deposition of diamond using thermal plasma chemical vapor disposition (TPCVD). However, a thin boundary layer leads to strong heating of the substrate by the hot plasma jet requiring effective substrate cooling to ensure the necessary control over the substrate surface temperature. While the high energy density of the thermal plasma allows very high deposition rates, it also brings some challenges for the process designer: there are usually strong radial nonuniformities associated with thermal plasma jets. These nonuniformities have the consequence that the high diamond growth rates exist over a small area, and strong radial variations of the film growth rate and the film morphology are observed. Also, the heat flux to the substrate is nonuniform, requiring special care in the substrate holder design to ensure a uniform substrate temperature. Various approaches have been pursued to increase the area of uniform deposition, such as reducing the pressure in the deposition chamber, which leads to a widening of the jet, spreading the arc jet magnetically, or moving the substrate. Most of these approaches require some sacrifice of the deposition rate. Also, re-

Plasma Torch Diamond Deposition

147

Figure 4 Mole fraction distributions in the boundary layer (a) and diamond growth rate (b) for different boundary layer thicknesses. (From Ref. 9.)

148

Heberlein and Ohtake

ducing the pressure carries the penalty of increased pump power with it. As can be seen in the discussion of the different plasma jet deposition reactors (Secs. IV. and V.), the principal design challenge is to preserve the advantages of this deposition method by transforming the high linear growth rates over a small area into uniform deposition over a large area while maintaining high overall deposition rates. The process characteristics of arcjet diamond deposition have several additional consequences: (1) A larger variety of precursors can be used; e.g., besides the commonly used hydrocarbons, various organic liquids can be used because the high temperatures will lead to rapid dissociation. (2) A variety of substrate materials can be used because the growth rate is frequently higher than the diffusion rate into the substrate material. However, this feature has somewhat limited benefits as the adhesion on many substrates is inadequate, and deposition on different substrates without good adhesion offers little advantage. (3) The high growth rates and high deposition precursor densities can result in highly stressed and sometimes very rough films and high secondary nucleation rates. Figures 5 and 6 illustrate these features, Fig. 5 showing diamond films obtained with several organic liquids used as deposition precursors and deposited at growth rates between 100 and 500 ␮m/hr [5]. Figure 6 shows some typical Raman spectra obtained from films deposited at high growth rates, indicating the relatively high purity of sp3 bonds [11]. Figure 7 shows a homoepitaxial film grown with methane at a growth rate of up to 220 ␮m/hr, depending on the crystal face [12]. The upper photograph shows the original diamond, and the lower one shows the crystal after 30 min of deposition. Although this homoepitaxial film is of excellent quality, the thickness that can be deposited remains limited due to the appearance of defects and secondary growth [13]. Although it is clear that the film characteristics can vary widely, the common features of these films are the high growth rates and the high purity of diamond as shown by the Raman spectra.

III.

PRINCIPLES OF PLASMA TORCH DESIGN

Plasma torch designs are adapted to the specific application, and they vary widely according to power, arc current, and energy density desired. For example, a cutting torch provides a transferred arc, highly constricted with a very high peak temperature (in excess of 15,000 K) for effective melting of the material, whereas a torch dissociating a gas and heating it to an average temperature of 4000 K for a chemical process will have a highly turbulent jet with a very wide temperature profile with a relatively low peak temperature. In the diamond deposition application, the torch has to disso-

Plasma Torch Diamond Deposition

149

Figure 5 SEM photographs of diamond films obtained with various liquid precursors: (a) acetone, (b) dichloromethane, (c) benzene, (d) carbon tetrachloride, (e) lacquer thinner, and (f) PCB compound. (From Ref. 5.)

ciate the hydrogen and provide sufficient energy for dissociation of the hydrocarbons. This review is limited to torches that are suitable for this purpose and operate in the power range between 1 and 100 kW. The arc discharge requires a cathode, an anode, and some kind of stabilization mechanism for the arc column. Furthermore, the arc discharge requires sufficiently high gas temperatures in the arc column to ensure adequate electrical conductivity (at least 10,000 K when no metal vapors are present). Under these conditions the electrons are heated by the applied electric field, rapidly achieve a Maxwellian distribution, and a sufficient number of high-energy electrons exist for ionizing collisions with atoms to

Plasma Torch Diamond Deposition

151

Figure 7 SEM photographs of a laser-cut diamond crystal (top) and the same crystal after 30 minutes in an arcjet reactor showing homoepitaxial growth. (From Ref. 13.)

< Figure 6 Typical Raman spectra obtained with diamond films from ethanol (a) and from acetone (b). (From Ref. 5; spectra courtesy of Dr. J. Butler.)

152

Heberlein and Ohtake

supply additional electrons and make up for those that have been lost through recombination or diffusion. Compared with other discharges, the arc discharge operates with the highest current values (1 A to more than 100 kA), the highest current densities (in the order of 105 A/cm2), and the lowest discharge voltage gradients (typically 5 to 40 V/cm). A.

Torch Cathode

The arc cathode has to provide the electrons to the arc column for transferring the current, and two dominant cathode types are used in plasma torches: the hot cathode, usually in the form of a rod or a button consisting of a refractory metal, and the cold cathode, usually consisting of a copper surface that is intensely cooled. The hot cathode provides electrons through thermionic emission; i.e., the temperature at the tip of the cathode is high enough that a sufficient number of electrons is emitted from the metal crystal lattice. The most frequently used cathode material is tungsten with the addition of a material such as thorium oxide that has a very low work function (around 2.6 eV compared with 4.5 eV for tungsten), thus increasing the number of electrons emitted at a given temperature. There is little movement of the arc attachment spot because the preferred attachment is the site of the highest electron emission, i.e., for a uniform material the site of the highest temperature. Typical spot temperatures are close to the melting point of tungsten (3680 K), and the dominant cooling mechanism of the spot is the emission of the electrons [14]. Because a substantial amount of heat is going into the cathode (in the order of 5% of the torch power), additional cooling of the cathode is usually provided (e.g., water cooling through internal cooling channels); however, if the cooling reduces the attachment spot size or temperature, increased erosion will result [15]. For operation with an inert gas such as argon, the thermionic cathode experiences little erosion at current levels up to 1000 A. Operation with hydrogen or an argon-hydrogen mixture will increase the erosion rate because of the higher heat fluxes associated with the more constricted hydrogen arc. Oxidizing, halogenated gases and hydrocarbons lead to strong cathode erosion because of the formation of volatile tungsten compounds. Thermionic cathodes are found in torches operating at power levels from 100 W to 100 kW. Cold cathodes are used in torches of intermediate power levels from 50 kW to 10 MW and for currents of up to 3 kA. The emission mechanism typically involves some evaporation and ionization of the cathode material, and the attachment spot is usually very unstable. Strong cooling of the cathode surface by a high-velocity water flow on the back surface results in a highly constricted arc attachment spot, and a forced motion of this spot will

Plasma Torch Diamond Deposition

153

distribute the erosion by evaporation over a larger area. The motion of the arc attachment is accomplished by controlling the gas flow over the cathode surface (e.g., swirl flow) and/or by applying magnetic fields to force a rotation of the arc attachment. The cathode material is usually copper or a copper alloy because of its high thermal conductivity. Evaporation of the cathode material leads necessarily to erosion and possibly contamination of a film, but these effects can be minimized for arc currents below 2000 A. B.

Arc Column

The arc column is inheritantly unstable because any disturbance of the current path through the plasma results in an increase in arc voltage and generation of magnetohydrodynamic (MHD) instabilities that can lead to arc extinction. Stable operation therefore requires a power supply with an opencircuit voltage significantly higher than the arc voltage and some design for a stabilization mechanism. Most torches use either convective stabilization, where a cold gas flow surrounding the arc forces the arc to remain in a preferred position of minimal heat loss, or wall stabilization, where proximity of a water-cooled wall to the arc forces it to the preferred position, or a combination of both. Convective stabilization has the consequence that an increase in arc current results in a widening of the discharge path or an increase in the plasma electrical conductivity or both, leading to a decrease in arc voltage (falling V-I characteristic) and requiring a current control in the power supply. The voltage gradient in the arc column is strongly dependent on the arc gas and the arc stabilizing design. An inert gas with a low thermal conductivity such as argon offers arcing with the largest arc diameter and the lowest voltage gradient, and the addition of hydrogen decreases the arc diameter and increases the voltage gradient dramatically due to the high thermal conductivity and specific heat of hydrogen. The consequence is that for the same torch diameter and the same arc current a thicker cold gas boundary layer exists if hydrogen is added to the plasma gas, leading to a strong temperature drop in the jet when the cold boundary layer mixes with the arc heated plasma. Higher arc currents and smaller nozzle diameters can avoid this temperature drop. C.

Anode

The function of the arc anode is to collect electrons from the arc plasma to allow current transfer to the power supply. However, the high current densities of the arc require intensive cooling of the anode. Heat fluxes between 5 and 20 kW/cm2 are encountered, and the total heat transfer can range from 5 to more than 50% of the torch power, depending on torch design. Ac-

154

Heberlein and Ohtake

cordingly, high cooling water flow rates are required, and the water velocities and water pressures at the back surface of the anode must be high enough to avoid film boiling. Usually, motion of the arc attachment is used to distribute the anode heating over a larger portion of the anode surface. The anode heat flux and the erosion are dependent on the constriction of the arc attachment, and a gas with high thermal conductivity and high specific heat such as hydrogen results in increased erosion. Also, a thick cold boundary layer between the arc and the cold anode surface forces the arc to constrict more, resulting in increased erosion. D.

Torch Design

Most torches with a thermionic rod cathode have the cathode tip placed at the converging entrance section of a cylindrical channel. The channel wall may be the anode surface, or it may be a section separating the anode from the cathode to assure longer arc lengths. Figure 8 shows the schematics of two different torch designs with thermionic cathodes. A torch designed for plasma spraying is shown in Fig. 8a; the torch is operated at atmospheric pressures at power levels up to 40 kW, and mixtures of argon and hydrogen can be used as plasma gas. The torch shown in Fig. 8b has been designed for space propulsion; hydrogen or hydrazine is heated by a low power (1 to 1.5 kW) arc, and the plasma is accelerated to high Mach numbers in the diverging section of the nozzle into a low pressure or vacuum ambient. Both types of torch design have been used for diamond deposition. The plasma gas is introduced at the base of the cathode with or without a swirl component, flowing toward the cathode tip, where part of it is heated by the electric current in the arc while the remainder forms the cold boundary layer along the channel wall. The heating of the plasma gas by the arc decreases its density by possibly two orders of magnitude, resulting in strong acceleration of the plasma flow. From a fluid dynamic point of view, the arc acts as a type of converging nozzle. The anode attachment on the inside wall of the channel is usually unstable resulting in fluctuations of the arc length with a frequency of a few kilohertz to a few tens of kilohertz, depending on torch design. For the application of diamond deposition, these fluctuations are of no major consequence because the characteristic processing times are on a larger time scale. The plasma exits the channel in a jet that has, in the case of an atmospheric pressure environment, velocity and temperature distributions similar to those shown in Fig. 2 [8]. A converging-diverging or a diverging nozzle is added at the exit of the plasma torch when supersonic flows are desired or when the deposition reactor is at a low pressure. In the case of a torch such as the one shown schematically in Fig. 8b, with a jet issuing into an environment at pressures less than 1 Torr, nozzle exit velocities of 13 km/sec at temperatures of around 5000 K can be achieved [16].

Plasma Torch Diamond Deposition

155

Figure 8 (a) Schematic of a plasma torch designed for atmospheric pressure plasma spraying: rod cathode at the left, anode nozzle with powder injection holes at right. (b) Schematic of plasma torch designed for space propulsion applications. (Part (a) courtesy of Praxair Surface Technologies; part (b) courtesy of NASA Lewis Research Center.)

Cold cathode torches typically position the cathode and anode cylinders coaxially side by side with the plasma gas entering through the gap between the electrodes with a swirl component (see Fig. 9). Usually, a solenoidal magnetic field is employed to act on the arc electrode attachments generating arc motion and reducing electrode erosion. The cathode is frequently used as the downstream electrode. The jets leaving such a torch usually have lower peak temperatures and velocities, and strong turbulence will make the profile more uniform. A nozzle is added when higher flow velocities are desired. However, for operation with high percentages of hydrogen as the plasma gas, use of thermionic cathodes is generally preferred because of easier control of electrode erosion [17].

156

Heberlein and Ohtake

Figure 9

Schematic of a swirl-stabilized cold cathode dc plasma torch.

Another design consideration besides power and desired velocity distribution is the arc heating efficiency of the torch, defined as the fraction of the electrical energy input, the product of current and arc voltage, I ⫻ V, which ends up as enthalpy flow in the plasma jet. This efficiency is determined by measuring the energy loss QL to the cooling water:

␩ = 1 ⫺ QL /(I ⫻ V)

(2)

Typical arc heating efficiencies range from 40 to 70% for torches with rodtype electrodes and from 70 to 90% for torches with cold electrodes. The efficiency increases with increasing gas flow rate and decreases with increasing arc current. Torch designs offering higher voltage operation have higher arc heating efficiencies as long as the length of the arcing channel does not exceed the length for establishing fully developed flow. IV.

CHARACTERIZATION OF THE PLASMA JET AND THE SUBSTRATE BOUNDARY LAYER

A.

Description of Atmospheric Pressure Plasma Jets

The plasma jet has been characterized for a number of configurations by using diagnostics and modeling. Also, because the plasma torches used for

Plasma Torch Diamond Deposition

157

diamond deposition have in general been developed for different applications such as plasma spraying or other materials processing tasks and for plasma propulsion, the characterization performed for these applications can be transferred to some extent to the diamond deposition process. However, one has to keep in mind that for different torch designs and operating conditions, the jet characteristics differ considerably and one has to be very careful when generalizations are attempted. Because of the strong nonlinearity of all plasma properties, the way the plasma gases and deposition precursors are heated and mixed can have a strong influence on the jet behavior. For example, mixing cold hydrogen with arc-heated argon will result in significant temperature drops in the jet and low degrees of hydrogen dissociation, and diamond deposition results obtained under these conditions cannot be compared with those obtained with arc-heated hydrogen. Some indication of the jet behavior can be found by calculating the jet power Pjet or the enthalpy flux after mixing of the reactants ˙ ⫻ have) Pjet = ␩ (I ⫻ V) = (m

(3)

where m ˙ is the total mass flow rate and have the average enthalpy of the mixture. Because in the temperature range between 2000 and 4000 K the equilibrium degree of dissociation changes drastically, a small change in the ratio of input power over mass flow rate can have a strong effect on the diamond deposition conditions. A value for have can be used to determine an equilibrium value for the degree of dissociation, and for arc-heated hydrogen this value will give a lower bound for the actual value, whereas for argonheated hydrogen, this value will present an upper bound. For the characterization of atmospheric pressure plasma jets, local thermal equilibrium (LTE) has been assumed in most cases. Also, time-averaged values are usually determined, although the fluctuations and instabilities may lead to somewhat distorted values [18]. Figure 2 gives such time-averaged LTE values of the temperature and velocity distributions in a plasma spray jet. Subsonic flow exists at the nozzle exit, and the plasma composition can be assumed to be close to chemical equilibrium at this location. Jets of this type have been well characterized concerning their time-averaged characteristics both experimentally and through models. For example, considering an argon jet issuing into an atmospheric pressure argon environment, it has been found that temperature and velocity distributions derived from spectroscopic and enthalpy probe measurements are reproduced by a standard low-Reynolds-number turbulence model for the jet and that they exhibit selfsimilarity with a Gaussian profile for radial distributions of temperature and axial velocity [19]. However, even for time-averaged conditions, chemical nonequilibrium exists.

158

Heberlein and Ohtake

There are usually three effects that lead to nonequilibrium conditions in the jet: (1) jet inhomogeneities caused by fluctuations and turbulence, (2) diffusion due to strong radial gradients, and (3) different relaxation times for the temperature equilibrium and chemical concentration equilibrium. Jet fluctuations are caused by arc anode attachment instabilities; i.e., a movement of the anode attachment downstream followed by a restrike of the arc upstream close to the cathode will result in variations of temperature, velocity, and composition at the nozzle exit. Figure 10 illustrates these fluctuations with a sequence of 100 nsec exposure time images of an atmospheric pressure arcjet. These variations will enhance the effect of large-scale turbulence and breakup of the laminar core of the jet because of the strong shear layers at the jet boundaries and the strong density differences between the plasma and the cold surrounding gas. The entrainment of the cold gas results in a decrease of the average temperature and velocity, but the rate of mixing of the cold gas with the hot plasma and the establishment of average conditions are relatively slow. This effect has been demonstrated experimentally by Pfender et al. [20], and a modeling description has been presented by Huang et al. [19]. A consequence is that one can find degrees of ionization and of dissociation that are far above the equilibrium values according to the average temperature. Figure 11 illustrates how the non-equilibrium conditions can affect temperature measurements in an argon-helium

Figure 10 High speed images (100 nsec exposure time) of an atmospheric pressure arcjet indicating jet instabilities.

Plasma Torch Diamond Deposition

159

Figure 11 Comparison of radial temperature profiles in an atmospheric pressure plasma jet obtained spectrometrically and using an enthalpy probe, indicating discrepancies of the results. (From Ref. 21.)

arcjet [21]. Temperatures derived from enthalpy probe measurements reflect the average energy content in the jet, and the spectral emission can be enhanced through nonequilibrium electron density distributions. Furthermore, the strong nonlinearity of the increase of the emission of radiation with temperature results in radiation measurements averaged over the line of sight that are higher than the associated average temperature. The effects of radial diffusion have also been well demonstrated experimentally and through modeling. Very high radial temperature and concentration gradients lead to radial diffusion of electrons, ions, and dissociated species out of the jet and molecular species into the jet, resulting in a higher concentration of molecular species in the central high-temperature portions of the jet and a widening of the distribution of electrons and dissociated species. Models for plasma jets with rigorous treatment of diffusion effects have been formulated [22,23] and combined with a kinetics formulation to determine the jet chemical composition [24]. The effect of lagging of the chemical reactions (recombination) compared with the drop in the gas temperatures results in conditions of a ‘‘super-

160

Heberlein and Ohtake

equilibrium’’ or higher densities of ionized and dissociated species than equilibrium relations would predict for the measured or calculated gas temperatures [25,26]. This includes significantly higher electron densities and atomic hydrogen densities compared with equilibrium conditions. The reason for this effect is that the temperature is determined by elastic collisions among all the plasma components, resulting in relatively short relaxation times once mixing has been achieved, while the electron-ion recombination reactions are largely determined by three-body collisions, which are considerably less frequent. This effect is stronger for supersonic jets where strong deviations from chemical equilibrium already exist at the nozzle exit, or for jets issuing into low-pressure environments [25,26]. Under these conditions, the nozzle design can have a significant influence on the transport of atomic hydrogen in the jet, and ideally, the nozzle design, torch power and plasma gas flow rate have to be optimized together [26]. On the other hand, the lower velocities in the jets of radio frequency torches (in the order of 10 to 30 m/sec) result in conditions much closer to chemical equilibrium. B.

Low Pressure Plasma Jets for Diamond Deposition

These arcjets have been to a large extent characterized with respect to diamond deposition, and the plasma temperatures and excited species distributions have been the primary focus of the investigations. Several measurements have been made in low-power (around 1 kW), laminar flow jets exhausted into a low-pressure chamber (around 5 to 200 torr) [27,28]. The methane has been introduced upstream of the arc and the low-current arc has been operated in a hydrogen-methane atmosphere. The small dimensions of the torch have resulted in very short residence times of the gas molecules in the plasma region (10 ␮sec), and the low arc power limited the enthalpy increase. Both emission spectroscopic and laser spectroscopic (LIF) measurements have been reported. Gas temperature values of approximately 5000 K at the axis approximately 5 to 10 mm from the nozzle exit have been derived from emission from C2 bands [27,28], and LIF measurements on CH bands resulted in axial temperature values of 2100 to 2500 K [28,29]. The authors explain that the temperatures derived from the CH band measurements are the true heavy species temperatures because of the relatively long lifetimes of the CH excited species and that the C2 band radiation is influenced by chemiluminescence effects. No atomic hydrogen lines and no atomic carbon lines are seen, and a chemical kinetics model shows that no dissociation equilibrium is obtained under these conditions. In a similar arcjet reactor with methane and hydrogen flow through the arc, but operating at somewhat higher power levels (1.6 kW), strong emission of atomic hydrogen and atomic carbon lines has been observed up

Plasma Torch Diamond Deposition

161

to 18 mm from the nozzle exit [30,31]. Farther downstream, emission from C2 and OH bands becomes dominant. A comparison with optical spectra from a microwave plasma shows the contrasting dominance of H2 and CH emission with the microwave plasma. A similarly strong emission of atomic hydrogen lines has been observed in a reactor operated with hydrogen at power levels up to 1.43 kW, with the hydrogen jet exhausted into a chamber having a background pressure of 0.35 torr [16,32]. Temperatures derived from LIF measurements on an atomic hydrogen line show values of approximately 4800 K at the axis near the nozzle exit dropping to 1800 K at an axial position 20 mm downstream, with some temperature recovery farther downstream. These temperatures are obtained with somewhat lower hydrogen flow rates compared with the measurements reported in Ref. 28, explaining the discrepancy. Velocity values derived from the Doppler shift of the LIF signals show peaks of about 13 km/sec near the nozzle exit, dropping to about 11.5 km/sec at 20 mm from the nozzle. A strong dependence of temperature and velocity on arc power is found: nozzle exit temperature values increase from about 1800 K at 0.8 kW to 4800 K at 1.43 kW. Laser-induced fluorescence measurements on CH radicals in an arcjet reactor operating with an argon-hydrogen mixture at a power level of 1.6 kW, with methane injected into the plasma flow inside the nozzle downstream of the arc, show temperature values ranging from 1800 K at the nozzle exit to about 2200 K at a position 20 mm downstream for the rotational temperatures and from 2700 K to about 2300 K for the vibrational temperatures [33]. It is not clear why the rotational temperatures should increase in the jet with increasing distance from the nozzle upstream of the recovery region in the boundary layer. These values for low-power arcjets are in strong contrast to measurements in medium-power plasma jets (larger than 5 kW), although less characterization has been performed in such reactors. The low-power arcjet measurements have been made some distance from the arc, and mixing between the hot plasma and the cold gas in the boundary layer has occurred to some extent. As pointed out in Sec. III., the low-current hydrogen arcs have small diameters and most of the gas is not heated by the arc. Mixing of the cold gas with the arc reduces the plasma temperatures in the jet. The atomic hydrogen mole fraction in these reactors can vary widely, depending on the fraction of the total hydrogen flow that is heated in the arc. In the mediumand high-power arc torches, the lower velocities and larger plasma volumes lead to a more uniform heating of the plasma gas and in general to higher temperatures at the nozzle exit (see Fig. 2). Ohtake et al. [34] present temperature measurements and temperature and flow field calculations for a 9kW reactor in which a supersonic jet exits a nozzle in front of the cathode,

162

Heberlein and Ohtake

and the anode attachment occurs to another jet from a torch located at a 90⬚ angle (see Sec. V. for details). In this reactor, peak temperatures of 9847 K have been measured at the axial location where the anode jet attachment occurs, i.e., right at the end of the arc, using line intensity ratios. The pressure has been 200 torr, and local thermodynamic equilibrium (LTE) has been assumed. This temperature drops to about 3550 K close to the substrate. The methane is introduced from the background gas in the reactor by entrainment. Obviously, there is much dissociated hydrogen at the boundary layer edge, verified by a strong atomic hydrogen spectrum. The C2 spectrum is very strong above the substrate, and a chemical kinetics model indicates C2H2 and CH2 as the major precursor species. There are few measurements characterizing the boundary layer conditions in front of the substrate because of the small spatial extent of this boundary layer (typically in the order of 1 mm or less). Most boundary layer characterization has therefore relied on modeling. Measurements in lowpower supersonic arcjets show the effect of a shock in front of the substrate leading to an initial temperature increase from approximately 2300 K to a peak value of approximately 3100 K at a position about 2 mm in front of the substrate [33]. C.

Diamond Deposition Reactor Models

Several models exist describing different arcjet reactor configurations and relating the diamond growth rates to the arcjet and boundary layer characteristics. The principal difficulties encountered are the formulation of an upstream boundary condition and the mixing of the reactants and the entrainment of the surrounding atmosphere. Because the entrainment is usually neglected, and the mixing is usually assumed to be instantaneous, the principal difference between the models is the upstream boundary condition, i.e., the temperature, velocity, and concentration profiles at the torch nozzle exit. A model for the arc in the anode nozzle would provide an upstream boundary; however, such models have so far had to rely on significant simplifications because the arc attachment to the anode wall with superimposed parallel flow is a three-dimensional time-varying phenomenon, and only two-dimensional steady-state descriptions are available at present for a limited set of conditions [35,36]. Results of a model by Ohtake et al. [34] give a good impression of the conditions in an arcjet deposition reactor. A 9-kW reactor for which experimental data have been obtained is modeled with a uniform jet inlet temperature of 9847 K (according to measurements), a uniform inlet velocity of 2366 m/sec, and a background pressure of 200 torr. The results are shown in Fig. 12, the streamlines (Fig. 12a), the axial velocity distribution (Fig.

Plasma Torch Diamond Deposition

163

Figure 12 Results of simulation of a hydrogen arcjet in a diamond deposition reactor: (a) streamlines, (b) axial velocity distribution, and (c) temperature fields. (From Ref. 34.)

164

Heberlein and Ohtake

12b), and the temperature distribution (Fig. 12c). The axial temperature drops to about 3400 K and the axial velocity to about 350 m/sec at a position 10 mm above the substrate (120 mm from the inlet). Chemical equilibrium is assumed in the jet up to a position 10 mm above the substrate. Results of a chemical kinetics calculation across the 10 mm show that C2H2 is the most likely diamond deposition precursor species. However, it is likely that superequilibrium exists in the jet with respect to hydrogen dissociation, and higher concentrations of atomic hydrogen would lead to an increase in C and CH3 in the boundary layer. Coltrin and Dandy [37] present an extensive investigation of the chemistry in an arcjet similar to the low-power arcjets just described. Again, a uniform reactor inlet temperature (i.e., temperature at the torch exit) of either 2500 or 3300 K and a uniform inlet velocity of 2000 m/sec are assumed, but the nonequilibrium dissociation is taken into account by considering various degrees of dissociation from 2.6 to 95%. The background pressure is 30 torr, and the CH4 is assumed perfectly mixed with the hydrogen jet shortly downstream of the inlet. The results show that the CH4 equilibrates rapidly with the hydrogen and that the mole fractions of the atomic hydrogen and the principal deposition precursor species are essentially constant up to the boundary layer edge. An increasing degree of hydrogen dissociation at the reactor inlet results in an increasing mole fraction of atomic carbon because of the fast hydrogen abstraction reactions from hydrocarbon molecules and in a decreasing mole fraction of CH3. Increasing the amount of methane injection results in an increase in C2H2. The effect of increasing the inlet gas temperature from 2500 to 3300 K while the degree of hydrogen dissociation is kept constant is a reduction of atomic carbon and an increase of CH3. These results demonstrate that virtually all kinds of deposition mechanisms can be encountered in arcjet diamond deposition, dependent on jet temperature, the degree of hydrogen dissociation, and the hydrogen-to-methane ratio, all of which can be controlled over a wide range by the reactor design and operation. The same authors demonstrate this by calculating the growth rate for different methane injection locations in the same reactor configuration [38]. Location of the methane injector just outside the boundary layer (at approximately 1 cm from the substrate) is found to result in the highest diamond growth rates because at this location the mixing of CH4 with atomic hydrogen results in rapid formation of CHx radicals serving as the main growth species, while insufficient residence times keep the formation of acetylene low. These calculations are for methane injection on the axis and instantaneous mixing with the hydrogen flow. In reality, one has to allow some time or distance for the mixing of the CH4 with the hydrogen.

Plasma Torch Diamond Deposition

165

Modeling results for a high-power reactor (98 kW) are presented by Kolman et al. [39]. An argon-hydrogen jet is considered with the temperature and mass fraction distribution at the reactor entrance taken from a plasma nozzle flow model for the same plasma torch operating conditions [26] (see Fig. 13). The hydrocarbon distribution at the reactor entrance has been assumed to be in equilibrium according to the temperature at the inlet. Two cases are considered, subsonic flow (Mach number = 1 at the inlet) and supersonic flow (Mach number = 2.92). The pressure at the reactor chamber exit is 35 torr. Figure 14 compares the results obtained for the subsonic jet (graphs on the left-hand side) with those of the supersonic jet. The axial distributions of temperature, velocity, pressure, and density from the nozzle exit (at the left side of each graph) to the substrate are displayed in the top graphs, and the mass fraction distributions of atomic hydrogen and various carbon-containing species are shown in the other two graphs. A gradual decrease of the atomic hydrogen concentration is observed, largely due to diffusion. The higher temperatures in the subsonic case lead to a dominance of atomic carbon in most of the jet expansion region. A strong temperature increase in the recovery zone in front of the substrate, particularly in the case of the supersonic jet, is obvious. Figure 15 shows the boundary layer profiles for the same quantities from a point 2 mm in front of the substrate. The drop in atomic hydrogen due to heterogeneous recombination on the substrate leads to an increase in acetylene concentration compared with atomic carbon as had been shown by Coltrin and Dandy [37]. It is interesting to notice that this drop is less pronounced in the case of the supersonic jet because of the strong temperature increase in the bowshock and because of the very thin boundary layer of 0.4 mm. Figure 16 shows the radial profiles of the gas-phase species at the substrate location and of the temperature, total heat flux, heat flux by conduction, and diamond growth rate on the substrate surface. The substrate heat flux is calculated considering the two major contributions to the energy transfer, heat conduction and surface recombination: qsub

冉冊 ␭ cp

with

eff

冉冊

␭ =⫺ cp =2

eff

dh ⫹ fs dz

冘 imax

(sihi)

i

␭/cpsub ␭/cpgas ␭/cpsub ⫹ ␭/cpgas

␭ = thermal conductivity, cp specific heat at constant pressure i = number of gas phase species si = destruction rate of species i on the substrate [kg/s m2]

(4)

166

Heberlein and Ohtake

Figure 13 Radial profiles of temperature, pressure, axial velocity, and atomic and molecular hydrogen mass fraction of an argon-hydrogen plasma jet at the nozzle exit for subsonic (top) and supersonic flow conditions. Profiles are calculated in Ref. 26 and used for jet calculation as presented here in Ref. 39.

Plasma Torch Diamond Deposition

167

Figure 14 Axial profiles of temperature, axial velocity, pressure, and density (top graphs), and of mass fractions of hydrogen, carbon, and several hydrocarbon species (middle and bottom graphs) for a subsonic (left) and a supersonic argon-hydrogen jet as calculated in Ref. 39.

Figure 15 Axial profiles across the substrate boundary layer of temperature, pressure and density (top graphs) and of mass fractions of hydrogen, carbon and selected hydrocarbons (bottom graphs) for a subsonic (left) and a supersonic argon-hydrogen jet as calculated in Ref. 39.

Figure 16 Radial profiles at the substrate surface of mass fractions of hydrogen, carbon and selected hydrocarbons (top graphs) and of temperature, total heat flux and heat flux by conduction, and total diamond growth rate g (bottom graphs) for a subsonic (left) and supersonic argon-hydrogen plasma jet as calculated in Ref. 39.

170

Heberlein and Ohtake

hi = specific enthalpy of gas species i fs = fraction of reaction heat transferred to substrate, assumed to be 0.6 in these calculations Clearly, the major contribution to the total heat flux is the hydrogen recombination energy released to the substrate. The model for the surface kinetics does not include formation of graphite. In this model laminar flow has been assumed, and turbulence will widen the jet and reduce the temperature and density gradients in the jet fringes.

D.

Scaling Relations and the Effect of Pressure

An approach for deriving scaling relations for arcjet diamond deposition reactors is presented by Goodwin [40,41]. Based on a simple boundary layer analysis and the film growth model by Harris [42], rules are derived for optimizing free stream velocity and operating pressure. This analysis shows that in most arcjet reactors the atomic hydrogen concentration is sufficiently high that it no longer strongly influences the growth rate, a result confirmed by detailed computations by Kolman et al. [43]. But the atomic hydrogen concentration does play an important role in reducing defect densities. Defects are assumed to result predominantly from sp2 bond formation. The atomic hydrogen concentration at the surface is in a first-order approximation proportional to [(up)ds]0.5, where u is the free stream velocity, p the pressure, and ds the substrate diameter. However, the free stream velocity is limited by the plasma torch design and operation, and the pressure has an optimum value because too high pressures strongly increase homogeneous recombination of atomic hydrogen. In a similar type of one-dimensional analysis using the heat and mass transfer analogy, Young [17] finds that the total mass deposition rate is proportional to the torch power. Increasing the substrate diameter requires an increase of the torch power proportional to ds1.5 to keep the same linear growth rate. This result is illustrated in Fig. 17, where deposition area and linear growth rate are shown as a function of torch power. Increased torch power allows an increase of the deposition area (see left figure), and this increase more than compensates for the decrease in linear growth rate of the film (right figure) if one considers the total deposited mass. Based on this analysis, Partlow et al. [4] find that strong cost benefits can be gained by scaling to a 100-kW reactor, with further scaling providing smaller gains. The predicted production costs of diamond films as a function of the output power are calculated, and it is concluded that it is necessary to use a plasma torch larger than 100 kW in a gas recycle mode to decrease the cost of toolgrade diamond to the $5/carat level.

Plasma Torch Diamond Deposition

171

Figure 17 Increase in deposition area with increasing power (left) and decrease in linear growth rate for the increase in deposition area. (From Ref. 4.)

A figure of merit for arcjet diamond deposition is presented by Bigelow et al. [7]. This figure is defined as the hydrogen concentration at the substrate divided by the torch power. An optimum pressure of 200 torr is found to maximize the hydrogen concentration at the substrate. At higher pressures homogeneous recombination losses will start to dominate. Such an optimum pressure has been found experimentally by Lu et al. [44] in a reactor where three arcjets are combined to cover a larger area (see Sec. V.). However, in this experiment the total mass flow rate of the plasma gases has been constant and not the velocity. The optimum pressure has been 270 torr for this reactor (see Fig. 18a). A similar optimum pressure for maximizing deposition rates has been reported (see Fig. 18b) by Hirata and Yoshikawa [45], although for their reactor the optimal pressure has been significantly lower (around 50 torr). However, these type of analyses do not represent the low-power arcjet deposition reactors at very low pressures with very high flow velocities. High deposition rates have been obtained under these conditions as well [32]. One may conclude that no one analysis is applicable for all arcjet reactors and that even for one reactor, the analysis applicable for the center of the substrate may not be correct for the outer parts of the substrate. Also, pressure variations affect not only the boundary layer. The arcjet becomes

172

Heberlein and Ohtake

Figure 18 (a) Linear growth rate dependence on deposition chamber pressure [44]. (b) Total mass deposition rate dependence on deposition chamber pressure. (From Ref. 45.)

Plasma Torch Diamond Deposition

173

less constricted and longer when the pressure is reduced from 1 atmosphere, and the heat as well as the mass is transferred to a larger portion of the substrate. Large area deposition will probably require reduced pressures.

V.

SPECIFIC PLASMA TORCH DEPOSITION REACTORS

There have been numerous efforts devoted to developing dc arcjet deposition reactors because of the promise of obtaining high deposition rates. In this section, several of these developments are described with the objective to present an overview of the different approaches to solve the problems associated with this method. Table 1 presents a summary of the specific developments described in this section. This table cannot be a complete representation of all arcjet diamond deposition developments, but it lists a representative sample.

A.

Common Features of Arc Jet Deposition

Diamond deposition using a dc arcjet was first reported by Kurihara et al. [2]. The plasma torch used simply consists of a tungsten cathode rod, a cylindrical anode made of pure copper and a dc power supply. They initiated a plasma jet consisting of Ar, H2 , and CH4 at 100–400 torr. The diameter of the anode nozzle was 2 mm and the distance between the cathode rod and the anode wall 1 mm. The substrate temperature was kept at 1000– 1500 K. They reported a growth rate of the diamond film of 180 ␮m/hr. Kurihara et al. [46] succeeded in forming functionally gradient films consisting of diamond/WC, diamond/Mo, etc. The WC or the Mo particles are melted by the plasma jet and sprayed onto a substrate, just as in a plasma spraying process. This plasma spraying process is combined with the diamond deposition process as shown in Fig. 19. The composition of the film changes from WC or Mo to diamond over a film thickness of 40 ␮m. The adhesive strength of the gradient composite film is better than 150 kgf/cm2. Loh and Cappelli [32] reported a supersonic plasma jet synthesis method that focused on reducing the thermal boundary layer effects in diamond deposition. They use two 35,000 L/min vacuum pumps in order to produce a process pressure as low as 10–100 Pa (0.07–0.7 torr). Figure 20 shows a schematic of the apparatus, a supersonic arc jet operating with hydrogen expanded to the low-pressure atmosphere. CH4 is introduced at the end of the expansion nozzle. The characteristics of this experiment are described in Sec. IV. The diamond film grown in this process on a molybdenum substrate at 820 ⫹ 30⬚C with a CH4 concentration in the jet of 0.3%

Table 1

List of Major Arcjet and rf Plasma Diamond Deposition Results

Type

Output (kW)

Gas (ᐍ/min at 20⬚C)

Pressure (torr)

Growth rate (␮m/h)

Fujitsu Tokyo Inst. Tech.

dc dc

1.8 9.0

100–400 200

80 930

Univ. of Minn Tokyo Inst. Tech.

dc dc

14.0 10.7

250 50

60 200

10 28.4

Univ. of Minn.

dc

19.3

100–760

1000

4

dc dc rf (4 MHz) rf (3.4 MHz) dc dc

1–3 1 60 40

CH4(0.01–0.2) ⫹ H2(5–20) CH4(0.08) ⫹ H2(1.5) ⫹ Ar(6.5) CH4(3.5–10) ⫹ H2(10–15) CH4(0.06) ⫹ H2(1.5) ⫹ Ar(6.3) (CH3)2CO(2 ml/min) ⫹ H2(18) ⫹ H2(1.7) ⫹ Ar(24) CH4(0.03–0.05) ⫹ H2(10) CH4(0.005) ⫹ H2(0.5) CH4(0.1–1.2) ⫹ H2(12) ⫹ Ar(60) CH4(0.8) ⫹ H2(20) ⫹ Ar(80)

NA NA

dc dc ⫹ rf dc dc

3.75 50 100 100

Institution

Stanford Chemnitz Univ. Tech. NIRIM Toyota Central Res. Ctr. Norton Texas Instruments and Olin Aerospace Kitachi Koki Univ. of Tokyo Westinghouse Univ. Sci. & Tech. Beijing NA: not announced.

Deposition area (cm2) 0.26 5

Remarks

Separate torch Triple torch 1 Cathode–3 anode Liquid source

Ref. 2 6 52,53 54 5

0.1 15 760

3 2 60

34 4 3.2

Supersonic plasma jet Hollow cathode Use induction coil

32 47 3

150

30

77

Use induction coil

48

CH4 ⫹ H2(⫹Ar?) NA

NA NA

NA NA

>80 320

Magnetic field enhanced Supersonic plasma jet

59 60

CH4(0.08) ⫹ H2(2) ⫹ Ar(20) CH4(0.3) ⫹ H2(15) NA CH4(0.04–1.2) ⫹ H2(8–12) ⫹ Ar(6–10)

44 250 NA NA

NA 1 4 40

3.2 168 730 97

On rotating substrate On rotating substrate Recycle mode Magnetic field, recycle mode

55 56 4 50

Plasma Torch Diamond Deposition

Figure 19 46.)

175

Schematic diagram of functionally gradient film deposition. (From Ref.

Figure 20 Schematic drawing of supersonic plasma jet CVD apparatus for diamond. (From Ref. 32.)

176

Heberlein and Ohtake

displays mostly {111} facets. The growth rate of the film has been 1–10 ␮m/hr, a very high value considering the low operating pressure. Stiegler et al. [47] report diamond synthesis with a hollow cathode arc jet CVD process. A schematic diagram of the apparatus is shown in Fig. 21. A hot refractory hollow cathode is made of tungsten, and the arc discharge occurs between the hollow cathode and a ring anode with possible discharge transfer to the substrate. The substrate temperature is kept at 973–1123 K by controlling the current to the substrate. They report that this apparatus produces high charge carrier densities (e.g., electron densities of 0.5–2.0 ⫻ 1011/cm3). The growth rate of the film has been 2 ␮m/hr over a deposition area of about 4 cm2. Matsumoto et al. [3] first reported rf induction thermal plasma diamond deposition. The arrangement of the deposition experiments is comparable to that shown in Fig. 22. They use a mixture of Ar and H2 (35 L/min Ar ⫹ 12 L/min H2), Ar (17 L/min), and Ar and CH4 (8 L/min Ar, 0.1–1.2 L/min

Figure 21 Schematic diagram of hollow cathode arcjet setup for diamond synthesis. (From Ref. 47.)

Plasma Torch Diamond Deposition

177

Figure 22 Schematic diagram of rf plasma CVD apparatus for diamond deposition. (From Ref. 48.)

CH4) as sheath, plasma, and carrier gases, respectively. The rf frequency is 4 MHz and the output power 60 kW. The growth rate of the diamond film has been approximately 60 ␮m/hr over a deposition area of 3 cm2. Kohzaki et al. [48] report large-area deposition of diamond films using an rf thermal plasma torch with an inner quartz sleeve of 65 mm diameter. The rf frequency is 3.4 MHz and the output power 40 kW. The process pressure is 150 torr. The growth rate of the film has been 30 ␮m/hr over a deposition area of 77 cm2 (12 in.2). They have obtained free-standing diamond films and have used them for tribological applications by brazing the free-standing films on steel materials.

178

B.

Heberlein and Ohtake

High-Rate Diamond Deposition

High-rate synthesis has always been one of the central objectives of arcjet diamond deposition process developments. A schematic of the arc jet apparatus featuring high-rate deposition reported by Ohtake and Yoshikawa is shown in Fig. 23 [6,49]. The apparatus mainly consists of a dc arc discharge torch, its associated power supply, and a reaction chamber. The plasma jet consisting of Ar and H2 is generated by a dc arc discharge. Typical operating conditions are 100 V and 90 A. CH4 is mixed into the plasma jet by en-

Figure 23 Schematic of separate-anode-type torch plasma jet diamond depositon reactor. (From Ref. 6.)

Plasma Torch Diamond Deposition

179

trainment enhanced by the vortex flow of the plasma gas. The plasma torch consists of separate cathode and anode assemblies that are placed at right angles to each other, and current flow is from a plasma jet exiting the anode assembly to the perpendicular plasma jet exiting the cathode assembly. The diameter of the cathode is 2 mm. To protect the cathode rod and the anode from erosion, only Ar is fed into the flow channel surrounding the electrodes. The cylindrical anode is made of oxygen-free copper, and the diameter of the anode nozzle is 3.5 mm. The distance between the cathode nozzle and the substrate has been varied from 100 to 150 mm. Figure 24 shows the appearance of the plasma jet during diamond deposition. The plasma jet impinges on the substrate where the diamond is deposited. Figure 25 shows a scanning electron microscope (SEM) picture of diamond deposited for a deposition time from 1 to 10 min. Figure 26 shows the variations of the nucleation density and average size of diamond crystallization with deposition time. The substrate has been lapped with 15␮m diamond paste. The substrate is then exposed for 1 min to an Ar ⫹ H2 plasma jet in order to remove the diamond dust particles remaining on the substrate surface. It can be observed that diamond has already deposited after 1 min; however the nucleation density increases with increasing deposition time for approximately 8 min. The deposited diamond forms a continuous film after 10 min. Figure 27 shows SEM photographs of the surfaces and the laser cut cross-sectional views of films that were deposited at different CH4 concentrations. The deposition time has been 1 hr for all samples. For a CH4

Figure 24

Appearance of plasma jet during diamond deposition.

180

Heberlein and Ohtake

Figure 25 (a–d) SEM photographs showing deposit at different deposition times. (From Ref. 49.)

concentration of 3%, a diamond film of 150 ␮m thickness was deposited within 1 hr, and the surface of the film was covered with triangular or rectangular automorphic diamond facets. When the CH4 concentration was increased to 5%, the maximum thickness of the film was about 700 ␮m and the diamond film had a large surface roughness. When the CH4 concentration was as high as 15%, cauliflower-like particles were observed on the film surface. The growth rate of the film decreases to about 100 ␮m/hr if we neglect the huge graphite particles grown on the film. Raman spectra of the films obtained at CH4 concentrations of (a) 3%, (b) 4%, (c) 8%, and (d)

Plasma Torch Diamond Deposition

181

Figure 26 Variations of nucleation density and average particle size with deposition time.

15% are shown in Fig. 28. The Raman spectra from the films deposited at CH4 concentrations of 3% and 4% have only one sharp peak at 1333 cm⫺1. When the CH4 concentration is increased to 15%, diamond, graphite, and amorphous carbon are detected in the Raman spectrum from the deposit. The most suitable deposition conditions for high-rate synthesis of diamond are reported to be a CH4 flow rate of 0.08 SLM, an H2 flow rate of 1.5 SLM, giving a CH4 concentration of 5.3%, and a substrate temperature of 950⬚C. Figure 29 shows the surface and the cross-sectional view of the diamond film obtained under these deposition conditions. The diamond film has large surface roughness, however, and many different diamond automorphic facets can be seen on the surface. The maximum thickness of the diamond film is 930 ␮m, indicating the very high growth rate of 930 ␮m/ hr obtained by this method.

182

Heberlein and Ohtake

Figure 27 (a–d) SEM photographs of surfaces and laser-cut cross sections of films deposited at different methane concentrations.

Plasma Torch Diamond Deposition

183

Figure 28 Raman spectra of diamond films deposited at CH4 /H2 concentrations of (a) 3%, (b) 4%, (c) 8%, and (d) 15%.

184

Heberlein and Ohtake

Figure 29 (a, b) SEM photographs of surface and cross-sectional view of highrate synthesized diamond film. (From Ref. 6.)

The relation between linear growth rate and CH4 concentration using this apparatus is shown in Fig. 30. Conversion efficiency (i.e., percentage of the carbon atoms in CH4 that formed diamond) is also plotted as a function of CH4 concentration. The linear growth rate increases with the CH4 concentration; however, the conversion efficiency has a maximum at at CH4 concentration of 4%. The decrease at higher CH4 concentrations may be caused by the deposition of nondiamond carbon. The gas flow rate also has an effect on the linear growth rate and the conversion efficiency (see Fig. 31). The CH4 concentration has been kept constant in these experiments. The growth rate saturates at H2 flows of 1.5 L/min, then decreases with further increasing H2 flow rate, mainly due to the decrease of the gas temperature. Figure 32 shows relations between the linear growth rate and the

Plasma Torch Diamond Deposition

185

Figure 30 Relation between linear growth rate and conversion coefficient and CH4 concentration (H2 flow rate, 1.5 lpm; Ar flow rate, 6.5 lpm; power, 9 kW; chamber pressure, 25 kPa; substrate temperature, 1000⬚C; deposition time, 1 hr).

186

Heberlein and Ohtake

Figure 31 Relation between linear growth rate and conversion coefficient and gas flow rate (Ar flow rate, 6.5 lpm; methane concentration, 4%; power, 9 kW; chamber pressure, 25 kPa; substrate temperature, 1000⬚C; deposition time, 1 hr).

conversion efficiency and the applied discharge power. Growth rate and conversion efficiency appear to be increasing more than linearly with torch power. Scale-up of deposition reactors to high power levels is generally accepted to reduce production costs. Following their scaling analysis, Young [17] and Partlow et al. [4] reported on the operation of a 100-kW system with a Westinghouse Marc 3 hot cathode swirl stabilized plasma torch. The plasma is expanded through a nozzle into the deposition chamber, which is

Plasma Torch Diamond Deposition

187

Figure 32 Relation between linear growth rate and conversion coefficient and output power (H2 flow rate, 1.5 lpm; Ar flow rate, 6.5 lpm; pressure, 25 kPa; substrate temperature, 1000⬚C; deposition time, 1 hr).

kept at a pressure between 10 and 50 torr. The gas is recycled after going through a heat exchanger. Lu et al. [50] also reported on a 100-kW dc arcjet installation for deposition of diamond films at costs as low as $4/carat level. They used a magnetic field–enhanced ‘‘W’’ type of rotating arc (see Fig. 33) that provides large area uniformity for not only composition but also temperature distribution in the jet in front of the substrate. The deposition area is reported to be 97 cm2. Gas recycling and coolant recycling are employed in this facility.

188

Heberlein and Ohtake

Figure 33 Schematic diagrams of (a) magnetic field–enhanced rotating plasma torch and (b) rotating arc. (From Ref. 50.)

C.

Large-Area and High-Quality Diamond Deposition

Enlargement of the deposition area has been mainly carried out by scale-up of the plasma torch itself as described in Sec. V.B. However, other ideas to enlarge the deposition area have also been reported.

Plasma Torch Diamond Deposition

189

Lu et al. [44,51,52] described a triple torch plasma reactor, which is shown in Fig. 34. The plasma generator consists of three identical dc plasma torches operating with an argon-hydrogen mixture at a total power level of about 36 kW. The three coalescing plasma jets impinge on a water-cooled substrate. A water-cooled gas feeding probe is located above the region where the three jets merge. This arrangement results in efficient mixing of the H2/CH4 reactant mixture introduced through this probe with the plasma. The three coalesced jets offer a larger plasma volume than most dc reactors and higher plasma flow velocities than the rf reactors. To improve temperature uniformity, the substrate holder has been shaped to allow nonuniform cooling of the backside of the substrate [53]. Although a uniform growth rate has been achieved in this way, the film morphology has varied with

Figure 34

Schematic of triple torch plasma reactor. (From Ref. 52.)

190

Heberlein and Ohtake

varying radius. However, providing additional methane to the outer portion of the coalesced jets through entrainment resulted in uniformity of film thickness and morphology over an area of approximately 35 mm diameter. Average linear growth rates of 100 ␮m/hr have been obtained [53]. Hirata and Yoshikawa [45,54] describe a unique plasma torch that uses three symmetrical anodes with one cathode as shown in Fig. 35. Each electrode can be moved along its axial direction by rotating a thread mounted at its end. The diameter of the plasma jet increases with increasing displacement y of the anodes from the center (see Fig. 36). According to Raman spectra taken from each radial portion of the film (Fig. 37), it can be seen that the quality of the diamond film is good all over the 60-mm-diameter substrate. Effects of process pressure on deposition rate (g/hr) and the surface profile are shown in Figs. 18b and 38, respectively. The maximum deposition rate (about 1 ct/hr) has been obtained at a process pressure of 6.7 kPa. The conversion efficiency of the carbon atoms in the source gas to diamond has been as high as 15% in this case. Although the surface uniformity has not been improved by this three-anode technique, the deposition area has increased by a factor of 5 compared with the deposition with the single-cathode single-anode torch. Noto and Mori [55] proposed enlarging the deposition area in terms of rotating the substrate during arc jet diamond deposition as shown in Fig. 39. The center of the substrate is shifted from the center axis of the plasma jet. The rotation speed of the substrate is kept at 15 rpm. The deposition area increases by a factor of 4 when the center of the plasma jet is shifted a quarter diameter from the center of the substrate. This idea demonstrates a convenient way to enlarge the deposition area, but Raman spectra from the diamond film (Fig. 40) indicate that the film contains small amounts of amorphous carbon, resulting from the deposit that grows in the peripheral part of the deposition region where the atomic hydrogen concentration and the substrate temperature are lower. Eguchi et al. [56] described diamond synthesis using a hybrid torch on a rotating substrate as shown in Fig. 41. The induction coupled plasma is coupled with a dc arcjet (‘‘hybrid plasma torch’’ [57]). The maximum substrate size has been 300 mm in diameter. The rotation speed has been 40 to 400 rpm. The temperature variation has been as small as 8 K per cycle, when the rotating speed of the substrate is 40 rpm. The diamond has grown over almost the entire surface of the rotating substrate. Yin et al. [58] reported diamond deposition on tungsten wires using the same rotating deposition system. Uniform diamond films approximately 1 ␮m thick are obtained on the wire with a diameter of 0.2 mm at a rotation speed of 450 rpm. Lu et al. [59] have applied a magnetic field to the arcjet in order to enlarge the plasma jet. A schematic illustration of the apparatus is shown in

Plasma Torch Diamond Deposition

191

Figure 35 (a, b) Schematic illustration of one-cathode, three-anode plasma torch apparatus for diamond deposition. (From Ref. 45.)

Fig. 42. The magnetic field is applied to the dc arc to force a rotation, and the rotating arc produces a plasma jet uniform over a larger area. An additional magnetic coil with an auxiliary anode is mounted above the substrate. This arrangement increases the atomic hydrogen concentration near the substrate and also contributes to improving the uniformity of the diamond film.

192

Figure 35

Heberlein and Ohtake

Continued

The deposition area has been larger than 78 cm2 (i.e., 10 cm in diameter), film thickness has been 500 ␮m, and the surface roughness has been approximately 5 ␮m after surface polishing. The grain size of the diamond has been as small as 20–50 ␮m for a film 500 ␮m thick. The diamond film prepared by this apparatus has a high optical transmission (t > 70%) and also provides high thermal conductivity (>2000 W/m/K). Diamond films produced with this apparatus have already been commercialized as heat sinks. McKenna et al. [60] present optical-quality high-growth-rate diamond deposition by supersonic arcjet CVD, focusing on generation of thick freestanding diamond films for optical components and infrared windows. The schematic of the deposition setup is shown in Fig. 43. They reported that they deposited an optical-quality diamond film 700 ␮m thick over the entire surface of an 8-in. (20-cm) substrate. A free-standing diamond dome was also produced by this method. The diameter of the dome was 2.5 in. and

Plasma Torch Diamond Deposition

193

Figure 36 Diameter of plasma jet versus displacement of anodes for different axial distances x between the cathode nozzle exit and the anode jet. (From Ref. 45.)

the thickness of the diamond film 700 ␮m. Process conditions were not given, but Raman spectroscopy and x-ray diffraction results for film indicated that the film had strong 具110典 orientation with very high film quality. D.

Liquid Source Diamond Deposition and Oxygen Addition Effect

Ohtake and Yoshikawa [61] reported the effect of oxygen addition in dc arcjet diamond deposition. They reported that the etch rate of graphite by the O2-Ar plasma jet was about 130 times larger than that by an H2-Ar plasma jet, and the rate of diamond etching by the O2-Ar plasma jet was about 55 times larger than that by the H2-Ar plasma jet. Consequently, the ratio of the etch rates of graphite and diamond obtained using the O2-Ar plasma jet was as high as 56, which is more selective etching of graphite in comparison with that by the H2-Ar plasma jet. The relationship between the oxygen concentration and the growth rate of diamond films is shown in Fig. 44. It can be seen that the growth rate decreases with increasing oxygen concentration. Figure 45 shows SEM pictures of diamond films deposited with oxygen concentrations of 0 and 33%. It can be seen that for a concentration of 33%, the crystals form films without pores, and the surface roughness is reduced. Figure 46 shows micro-Raman spectra of the diamond films that are deposited with oxygen concentrations of 0 and 33%. Both spectra have small peaks from amorphous carbon at 1500–1600 cm⫺1. The full

194

Heberlein and Ohtake

Figure 37 Raman spectra taken from different radial positions of the diamond film (Mo substrate 60 mm in diameter). (From Ref. 45.)

width at half-maximum (FWHM) of the peak from diamond deposited with oxygen is 7.5 cm⫺1. The value is about 0.67 times as large as that from the diamond film deposited without oxygen. Figure 47 shows the thickness profile of the diamond film. It can be seen that the thickness of the film is about 200 ␮m. The film appears white due to the light scattering from the surfaces and the grain boundaries in the film. A diamond film with 1.2 mm thickness has been prepared in a deposition time of 20 hr. The surface roughness of this film was about 150 ␮m. Diamond deposition using arcjets with liquid precursors was reported by Han et al. [5]. The schematic of the apparatus is shown in Fig. 48. The best results are obtained with acetone and ethanol as liquid precursors injected into the arcjet plasma. The liquids are injected using an atomizing probe located at the center of the substrate, opposing the plasma jet in a

Plasma Torch Diamond Deposition

Figure 38 Ref. 53.)

195

Effect of process pressure on surface profile of diamond film. (From

counterflow arrangement, or from the side with the atomizing probe 2 mm above the substrate [62]. High linear growth rates of more than 1 mm/hr have been achieved when the organic fluids are injected in the counterflow arrangement into an argon-hydrogen plasma jet. It was observed that using acetone resulted in smaller diamond particles than ethanol [62]. When the liquid is injected from the side of the plasma jet, high diamond growth rates and carbon conversion efficiencies are obtained only when the liquid is atomized close to the substrate and liquid droplets are actually entering the boundary layer. This situation has been described in a three-dimensional model that includes the fluid dynamics of the plasma jet; the atomization, transport, and evaporation of the liquid; the gas-phase chemical kinetics, and the surface chemical kinetics [43]. The very high growth rates observed experimentally can be explained by the model as a result of the high mass

196

Heberlein and Ohtake

Figure 39 55.)

Large area diamond deposition with a rotating substrate. (From Ref.

transport of deposition precursors in liquid form into the boundary layer, and the availability of oxygen and OH radicals to ensure film quality. E.

Effects of Substrate Bias on Arc Jet Diamond Deposition

Substrate biasing has been shown to be one effective way to accelerate the growth rate and to enlarge the deposition area of the diamond film. Substrate

Plasma Torch Diamond Deposition

197

Figure 40 Raman spectra of a diamond film with a rotating substrate at different radial positions (in mm). (From Ref. 55.)

bias effects in dc arcjet diamond deposition have been investigated by Matsumoto et al. [63] by using the setup shown in Fig. 49. The positive bias voltage is applied to the substrate or to the ring electrode near the substrate. The bias voltage has been ⫹100 to ⫹400 V, with a current flow of 0.5 to 5 A. The change in the appearance of the plasma jet with the bias is shown in Fig. 50. It is clear from Fig. 50b that a glowlike discharge extends from the tail of the plasma jet to the substrate, and the optical emission from the plasma jet is significantly enhanced especially near the substrate. The discharge changes to a transferred arc mode when bias is further increased, and because the transferred arc causes significant electron heating of the substrate, the substrate becomes overheated. Figure 51 shows the thickness profiles of the diamond films deposited with three different bias voltages. The growth rate of the diamond film increases by more than a factor of 2 when the bias current to the substrate reaches 2 A. In case of bias current of 3 A, the center region of the film is no longer diamond but graphite due to the large heat flux caused by the transferred arc initiated between the cathode and the substrate. These results show that the substrate biasing technique

198

Figure 41

Heberlein and Ohtake

Schematic of hybrid plasma torch. (From Ref. 56.)

can have an important effect on increasing the growth rate of diamond but that it will remain difficult to improve the uniformity of the diamond film. Baldwin et al. [64] also reported the effect of substrate biasing on the growth rate of diamond films. From emission spectroscopic measurements, it was determined that the bias leads to an increase in the concentration of atomic hydrogen near the substrate. The growth rate increased by a factor of 7 when the substrate was biased at 170 V and a current density of 4.9 A/cm2 was obtained. They measured the two-dimensional distribution of atomic hydrogen emission and concluded that the increased growth rate can

Plasma Torch Diamond Deposition

199

Figure 42 Schematic drawing of magnetic field-enhanced arcjet diamond deposition apparatus. (From Ref. 59.)

200

Heberlein and Ohtake

Figure 43

Schematic model of supersonic arcjet CVD setup. (From Ref. 60.)

be attributed to an increased flux of atomic hydrogen to the growth surface resulting from an elevated electron temperature that increases the dissociation near the surface.

VI.

CONCLUSIONS AND CURRENT RESEARCH ISSUES

Arcjet diamond deposition encompasses a large diversity of approaches, probably more than any other diamond deposition method. The different approaches vary in how the atomic hydrogen is generated, how the deposition precursors are generated, and how the atomic hydrogen and the dep-

Plasma Torch Diamond Deposition

201

Figure 44 Relationships between growth rate of diamond films and the ratio of oxygen flow rate to methane flow rate. (From Ref. 61.)

Figure 45 SEM photographs of diamond films deposited (a) without and (b) with oxygen addition (33% of methane flow). (From Ref. 61.)

202

Heberlein and Ohtake

Figure 46 Micro-Raman spectra of diamond films deposited at oxygen concentrations of (a) 0% and (b) 33%. (From Ref. 61.)

osition precursors are transported to the substrate. Because of this variety of approaches, it is difficult to arrive at general conclusions. The one common feature of all arcjet deposition methods is the relatively high growth rate of the films, and it appears that deposition reactors can be designed in which neither mass transport nor the gas-phase chemistry but rather the surface reactions are the limiting factors for the deposition rate. Furthermore, scaling

Figure 47

Thickness profile of diamond film deposited with oxygen.

204

Figure 48 Ref. 5.)

Heberlein and Ohtake

Schematic of counterflow liquid injection plasma reactor. (From

to large deposition areas and high overall deposition rates can be accomplished. Approaches have been demonstrated to solve the inherent weaknesses of the arcjet deposition method—the nonuniformity of the plasma jet leading to nonuniform distributions of film thickness, and the high heat fluxes to the substrate requiring substrate cooling—but more work on re-

Plasma Torch Diamond Deposition

205

Figure 49 Schematic drawing of substrate bias enhanced arcjet diamond deposition setup. (From Ref. 63.)

actor design for large-area high-rate deposition is anticipated. The major research issues associated with arcjet diamond deposition are essentially the same as those for all deposition methods; however, the special conditions of an impinging plasma jet may require different approaches for a solution. We see the areas that need to be further investigated: (1) the control of film morphology under high growth rate conditions, (2) the adhesion of

206

Heberlein and Ohtake

Figure 50 Change of appearance of plasma jet with substrate biasing (a) no bias and (b) 280V, 4 A. (From Ref. 63.)

Figure 51 Thickness profile of diamond film deposited with three different bias voltage. (From Ref. 63.)

Plasma Torch Diamond Deposition

207

the films on a variety of substrates, and (3) the deposition at low temperatures while maintaining a high growth rate. The control of film morphology requires mapping in the four-parameter space of carbon precursor concentration–atomic hydrogen concentration– total mass transport rate–substrate surface temperature, those regions associated with a certain texture of the growing film. It also requires control over secondary nucleation. Extensions of the maps of the ␣-parameter, which indicates the preferred direction of growth, to higher growth rate conditions and gas chromatography measurements giving information on the associated boundary layer chemistry can form the basis for this information [65]. Further experiments along these lines are necessary for arcjet deposition and for experiments with additional reactants such as oxygen and halogens. Such investigations will enhance the ability to tailor characteristics of highgrowth-rate diamond films for specific applications. Well-adhering diamond films have been obtained for some substrate materials, e.g., silicon or molybdenum; however, for many substrate materials of interest, adhesion is relatively poor. Being able to obtain well-adhering films will reduce the cost of diamond coatings for some applications and therefore allow wider utilization of the advantageous diamond characteristics. Arcjet diamond deposition makes the adhesion more difficult because the high growth rate is usually associated with an increased number of defects and with higher stresses. An understanding is needed of the relationship between defect and stress formation and the deposition conditions. Approaches such as the three-step deposition process in which an intermediate diamond-metal composite film is deposited between the substrate and the diamond film [66] may be a solution for some applications; however, the cost constraints will favor single-step processes for adhesion enhancement, possibly involving special substrate pretreatment such as deposition of a plasma sprayed intermediate layer of molybdenum [67]. Increasing the number of chemical reactants to include oxygen and halogens may be of interest for adhesion studies as well. A whole new world of applications for diamond films would open up if films could be deposited at high rates and at low temperatures. Film growth at low temperatures has involved use of oxygen or halogens, and rates have been very low. Arcjet deposition at temperatures of 550⬚C with deposition rates of 40 ␮m/hr [67] and even at 400⬚C with about 10 ␮m/hr growth rate have been achieved, but adhesion of these films has been rather poor. Approaches involving different surface chemistries may be necessary. Because arcjet deposition necessarily involves rather high heat fluxes, it may be possible to deposit diamond with this technology only on substrates that have a high thermal conductivity, but too few experimental data exist so far to reach any conclusion on this matter.

208

Heberlein and Ohtake

REFERENCES 1. 2. 3. 4.

5. 6. 7.

8. 9. 10. 11.

12. 13.

14. 15. 16. 17.

18. 19. 20.

21. 22.

J Heberlein, E Pfender. Thermal plasma chemical vapor deposition. Material Science Forum, pp 477–496, 1993. K Kurihara, K Sasaki, M Kawarada, N Koshino. Appl Phys Lett 52:6, 1988. S Matsumoto, M Hino, T Kobayashi. Appl Phys Lett 51:737, 1987. WD Partlow, J Schreurs, RM Young, I Martorell, SV Dighe, G Swartbeck, J Bunton. Low cost diamond production with large plasma torches. Proceedings of 3rd International Conference on Applications of Diamond Films and Related Materials, Gaithersburg, MD, 1995, pp 519–524. QY Han, J Heberlein, E Pfender. J Mater Synthesis Process 1:25–32, 1993. N Ohtake, M Yoshikawa. J Electrochem Soc 137:717, 1990. JCM Bigelow, NM Henderson, RL Woodin. In: R Messier, JJ Glass, eds. New Diamond Science and Technology. Pittsburgh: Materials Research Society, 1991, p 529. M Vardelle, A Vardelle, P Fauchais, MI Boulos. AIChE J 29:236–243, 1983. BW Yu, SL Girshick. J Appl Phys 75:3914–3923, 1994. SL Girshick, C Li, BW Yu, H Han. Plasma Chem Plasma Process 13:169– 188, 1993. QY Han, TW Or, ZP Lu, J Heberlein, E Pfender. High rate deposition of diamond using liquid organic precursors in atmospheric plasmas. Proceedings of the 2nd International Symposium on Diamond Materials. Vol. 91-98. Pennington, NJ: The Electrochemical Society, 1991, pp 115–122. KA Snail, CM Marks, ZP Lu, J Heberlein, E Pfender. Mater Lett 12:301–305, 1991. ZP Lu, K Snail, C Marks, J Heberlein, E Pfender. High rate homoepitaxial growth of diamond in thermal plasma. Proceedings of the 2nd International Symposium on Diamond Materials. Vol. 91-98. Washington, DC: The Electrochemical Society, 1991, pp 99–106. X Zhou, J Heberlein. Plasma Sources Sci Tech 3:564–574, 1994. X Zhou, J Heberlein. Plasma Chem Plasma Proc 16:229S–244S, 1996. JG Liebeskind, RK Hanson, MA Cappelli. Applied Optics 32:6117–6127, 1993. RM Young. Deposition area, growth rate, and arc power scaling laws for diamond film deposited by arc-heated gas flows. Proceedings of the 12th International Symposium on Plasma Chemistry. Vol. 4. Minneapolis, 1995, pp 2029–2033. JF Coudert, P Fauchais. High Temp Chem Process 3:443–453, 1992. PC Huang, J Heberlein, E Pfender. Plasma Chem Plasma Process 15:25–46, 1995. E Pfender, WLT Chen, R Spores. A new look at the thermal and gas dynamic characteristics of a plasma jet. Proceedings of the 3rd National Thermal Spray Conference. Long Beach, CA, 1990, pp 1–9. WLT Chen, J Heberlein, E Pfender. Plasma Chem Plasma Process 14:317– 332, 1994. JD Ramshaw, CH Chang. Plasma Chem Plasma Process 13:489, 1993.

Plasma Torch Diamond Deposition 23. 24. 25. 26. 27. 28. 29. 30. 31.

32. 33.

34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46.

47.

48.

49.

209

JD Ramshaw. J Non-Equilib Thermodyn 18:121, 1993. JD Ramshaw, CH Chang. J Comput Phys 116:359, 1995. MA Cappelli, TG Owano, OH Kruger. J Mater Res 5:2326, 1990. C George, G Candler, R Young, E Pfender, J Heberlein. Plasma Chem Plasma Process 16:43S–56S, 1996. KR Stalder, RL Sharpless. J Appl Phys 68:6187, 1990. GA Raiche, GP Smith, JB Jeffries. In: R Messier, JJ Glass, eds. New Diamond Science and Technology. Pittsburgh: Materials Research Society, 1990, p 251. GA Raiche, JB Jeffries. Applied Optics 32:4629–4635, 1993. K Kurihara, K Sasaki, M Kawarada. J Fujitsu Sci Tech 25:44, 1989. K Kurihara, K-I Sasaki, M Kawarada, N Koshino. Morphology of diamond films grown by DC plasma jet CVD. Materials Research Society Symposium Proceedings. Vol. 162. 1990, pp 115–118. MH Loh, MA Cappelli. Supersonic DC-arcjet synthesis of diamond. Proceedings of Diamond Films ’92, Heidelberg, Germany, 1993, pp 454–461. J Luque, W Juchmann, JB Jeffries. Absolute concentration measurements of CH radicals in a diamond depositing DC-arcjet reactor. Applied Optics 36: 3261–3270, 1997. N Ohtake, M Ikegami, M Yoshikawa. Diamond Films and Technology 2:1– 15, 1992. S Paik, PC Huang, J Heberlein, E Pfender. Plasma Chem Plasma Process 13: 379–397, 1993. C George, Ph.D. thesis, University of Minnesota, 1998. ME Coltrin, DS Dandy. J Appl Phys 74:5803–5820, 1993. DS Dandy, ME Coltrin. Appl Phys Lett 66:391–393, 1995. D Kolman, J Heberlein, E Pfender, R Young. Plasma Chem Plasma Process 16:57S–70S, 1996. DG Goodwin. J Appl Phys 74:6888–6894, 1993. DG Goodwin. J Appl Phys 74:6894–6906, 1993. SJ Harris. Appl Phys Lett 56:2298–2300, 1990. D Kolman, J Heberlein, E Pfender. Plasma Chem Plasma Process 18:73–90, 1998. ZP Lu, J Heberlein, E Pfender. Plasma Chem Plasma Process 12:55–69, 1992. A Hirata, M Yoshikawa. Dia Rel Mater 2:1402, 1993. K Kurihara, K Sasaki, M Kawarada, Y Goto. Formation of functionally gradient diamond films. Proceedings of the 1st International Conference Appl. Diamond Films and Related Materials, Auburn, Alabama. 1995, pp 461–466. J Stiegler, S Roth, O Stenzel, B Mainz, W Scharff. Plasma-assisted CVD of diamond films by hollow cathode arc discharge. Proceedings of Diamond Films ’92, Heidelberg, Germany. 1993, pp 413–416. M Kohzaki, H Higuchi, S Noda, K Uchida. Large-area diamond deposition and brazing of the diamond films on steel substrates for tribological applications. Proceedings of Diamond Films, ’92, Heidelberg, Germany. 1993, pp 612–616. N Ohtake, M Yoshikawa. Thin Solid Films 212:112, 1992.

210 50.

51. 52. 53. 54. 55. 56. 57. 58. 59.

60.

61. 62.

63. 64.

65. 66. 67.

Heberlein and Ohtake FX Lu, GF Zhong, YL Fu, JJ Wang, WZ Tang, GH Li, TL Lo, YG Zhang, JM Zhang, CH Pan, CX Tang, YP Lu. 100 kW DC arc plasma jet CVD system for low cost large area diamond film deposition. Proceedings of the 1st International Diamond Symposium, Seoul, Korea. 1996, pp 115–120. ZP Lu, L Stachowicz, P Kong, J Heberlein, E Pfender. Plasma Chem Plasma Process 11:387–394, 1991. ZP Lu, J Heberlein, E Pfender. Plasma Chem Plasma Process 12:35–52, 1992. QD Zhuang, H Guo, J Heberlein, E Pfender. Dia Rel Mater 3:319–324, 1994. A Hirata, M Yoshikawa. Dia Rel Mater 4:1363, 1995. N Noto, K Mori. J Jpn Soc Precision Engineering 58:1672, 1992. K Eguchi, S Yata, T Yoshida. Appl Phys Lett 64:58, 1994. T Yoshida, T Tani, H Nishimara, K Akashi. J Apply Phys 54:640, 1983. H Yin, K Eguchi, T Yoshida. J Appl Phys 78:3540, 1995. G Lu, KJ Gray, EF Borchelt, LK Bigelow. Free-standing white diamond for thermal and optical applications. Proceedings of Diamond Films ’92, Heidelberg, Germany. 1993, pp 1064–1068. T McKenna, JT Hoggins, JM Trombetta, P Klocek, C Vaughan. Optical-quality, high-growth-rate arcjet CVD diamond. Proceedings of SPIE. Vol. 2286. 1994, pp 160–173. N Ohtake, M Yoshikawa. Jpn J Appl Phys 32:2067, 1993. QD Zhuang, P Schwendinger, C Hammerstand, J Heberlein, E Pfender, R Young. Process characterization of CVD diamond from liquid precursors. Proceedings of the 12th International Symposium on Plasma Chemistry. Vol. 4. Minneapolis. 1995, pp 2303–2308. S Matsumoto, I Hosoya, T Chounan. Jpn J Appl Phys 29:2082, 1990. SK Baldwin, TG Owano, CH Kruger. Secondary discharge induced nonequilibrium chemistry in a CVD diamond DC arcjet reactor. Proceedings of the 12th International Symposium on Plasma Chemistry. Vol. 4. Minneapolis. 1995, pp 2011–2016. JW Lindsay, JM Larson, SL Girshick. Dia Rel Mater 6:481–485, 1997. C Tsai, JC Nelson, WW Gerberich, J Heberlein, E Pfender. Dia Rel Mater 2: 617–620, 1993. M Asmann, CFM Borges, J Heberlein, E Pfender. Thermal plasma chemical vapor deposition of diamond on steel. Proceedings of the 13 International Symposium on Plasma Chemistry. C.K. Wu, ed. Beijing: Peking University Press. 1997, pp 1206–1211.

7 Microwave Plasma-Assisted Diamond Film Deposition Timothy A. Grotjohn and Jes Asmussen Michigan State University, East Lansing, Michigan

I.

MICROWAVE PLASMA DISCHARGES

Microwave plasma-assisted diamond synthesis has been demonstrated over a wide operating pressure regime from a few millitorr to 1 atmosphere. The maximum film deposition rates vary from less than a tenth of a micrometer per hour at the low pressures of 10 mtorr to well over 10 ␮m/hr at operating pressures of 200 torr or greater. As the operating pressure is changed the behavior and properties of a typical microwave discharge vary significantly. At the low pressure of 10 mtorr to a few tens of torr the discharge usually fills the entire discharge chamber and is a relatively cold (gas temperature) nonequilibrium discharge where the electron temperatures are over 10,000– 20,000 K and the gas temperatures are less than 500–1500 K. As pressure is increased to 50 torr and above, the electron temperatures decrease and the gas temperatures increase. The discharge contracts and separates from the chamber walls as pressure is increased and takes on a ball-like or spherical shape that depends on the excitation electromagnetic fields and the geometry of the chamber. At higher pressures the gas temperatures usually are in excess of 1500 to 2000 K and the microwave discharge becomes a more spatially inhomogeneous discharge. More specifically, at higher pressures, the microwave discharge, similar to low-frequency and DC arcs, becomes a thermally inhomogeneous discharge. It has a hot central core with sharp thermal gradients existing between the discharge center and surrounding walls. Microwave energy is readily coupled into the electron gas in the hot discharge center because of 211

212

Grotjohn and Asmussen

its reduced neutral gas density, and neutral gas species are also easily ionized and excited in the hot discharge region. Major energy losses from the higher pressure discharges occur by heat conduction, convection, and radiation. The large temperature gradient causes heat conduction losses to become an important loss process. This loss mechanism includes the contribution to the heat conductivity by molecules, atoms, electrons, and ions as they diffuse to the walls. And also includes chemical reactions such as the transport of dissociation energy and ionization energy to the discharge/arc fringes by the free radicals and ions. Owing to the highpressure environment, electrons, ions, and free radicals recombine quickly outside the hot central core and thus convert their dissociation, ionization, and excitation energy into thermal energy. The result is a discharge with a radially varying gas temperature, ionization rate, and volume recombination rate. Gas temperature, ionization, dissociation, etc. are highest in the center of the discharge, while volume recombination and deexcitation of the different species increase radially away from the discharge center as the cooler, denser gas regions near the walls are approached. The central discharge core gas temperatures vary with gas type and pressure but typically are in excess of 2000 K, while temperatures external to the discharge are controlled by wall temperatures and the gas temperatures of the gas flowing around the discharge. This chapter reviews the studies that have been done on microwave plasma-assisted chemical vapor deposition (CVD) of diamond. As described earlier, this deposition occurs across a range of plasma discharge types extending from low-pressure nonequilibrium discharges to higher pressure, thermal-like discharges. This chapter is organized so that the general diamond deposition environment is first described; then a generic description of a typical plasma-assisted diamond deposition reactor is developed. Next, the details of several specific deposition systems are reviewed. Finally, details of the various microwave plasma-assisted diamond deposition processes and chemistries are examined. The approach used to organize the many diamond deposition results for microwave CVD systems is a control model approach where the deposition process is viewed in terms of input, internal, and output variables. The specifics of this approach are given in Sec. III.B. First though, Sec. II describes the general plasma chemistry environment that must be created to deposit diamond. II.

MICROWAVE PLASMA-ASSISTED CVD OF DIAMOND DEPOSITION ENVIRONMENT

The role of the microwave plasma-assisted machine in the deposition of diamond is to create the chemical/thermal environment needed for diamond

Microwave Plasma-Assisted Diamond Film Deposition

213

deposition. This environment for the standard hydrogen-methane deposition process has the following attributes. A.

Substrate Temperature

For diamond deposition the temperature of the deposition surface is generally in the range from 400 to 1100⬚C with the more typical values being 700–1000⬚C [1–3]. This range of substrate temperatures allows surface phenomena including various adsorption, desorption, and abstraction reactions to occur that lead to diamond growth. Too high a substrate temperature converts the diamond to graphite. Obtaining the correct substrate temperature often requires the design of substrate holders that incorporate either heating or cooling. B.

Atomic Hydrogen

The appropriate chemically reactive species need to be supplied to the growth surface. Two types of radicals are needed. First, atomic hydrogen is needed so that the surface is almost entirely covered with hydrogen to permit growth in the diamond phase of carbon and not the graphite phase. Second, appropriate carbon-containing growth species must be supplied to the growth surface. The ratio of the atomic hydrogen flux to the carbon growth species flux is typically 1000–10,000 for the deposition of high-quality diamond [4]. This need for the large flux of atomic hydrogen prevents the use of standard CVD where the substrate temperature provides the energy to dissociate the molecular species. At the typical substrate temperatures of 700– 1100⬚C the dissociation rate of hydrogen is very low. Hence, the hydrogen must be dissociated by a hotter thermal source located away from the substrate (e.g., a hot filament) or by a plasma discharge that dissociates molecular hydrogen by either electron impact dissociation or thermal dissociation due to a high gas temperature in the discharge. In the case of microwave CVD reactors the plasma discharge serves to dissociate the hydrogen via either electron impact dissociation or a combined electron impact dissociation and gas heating–produced thermal dissociation. Typically, the lower pressure discharges (few tens of torr) have higher gas temperatures (>2000 K) and thus thermal dissociation of hydrogen is the dominant dissociation process. C.

Carbon-Containing Radicals

The appropriate flux of carbon-containing growth species to the deposition surface is needed. The conditions of diamond deposition using microwave

214

Grotjohn and Asmussen

discharges are typically in the pressure range from 1 to 200 torr and gas discharge temperatures ranging from slightly above the substrate temperature (⬃1100–1500⬚C) to above 3000⬚C. Under these conditions in microwave discharges, the dominant carbon-containing radicals important to diamond growth are CH3 and C2H2 [3,4]. The methyl radical 1CH3 is generally acknowledged to be the most important species for diamond growth and is most often obtained from the methane (CH4) input feed gas. The CH4 chemical reactions in the typical diamond deposition environment (high gas temperature and significant atomic hydrogen concentrations) occur primarily as neutral-neutral chemistry rather than via electron collision processes, which are more important for the hydrogen dissociation. Almost always, the carbon chemistry reactions occur on a time scale much shorter than the residence time of the deposition gas in the discharge chamber for microwave CVD reactor systems. Hence, the carbon chemistry is typically assumed to have reached an equilibrium condition for the discharge gas temperature and atomic hydrogen concentration found in the discharge.

D.

Deposition Uniformity

Once a vapor-phase environment containing a substantial amount of atomic hydrogen and an appropriate number of carbon growth radicals is created adjacent to a temperature-controlled substrate, diamond growth can proceed. One of the major issues for the design of microwave plasma-assisted CVD diamond deposition machines is creating the appropriate and uniform vaporphase environment and temperature over the entire deposition surface. Several factors influence the uniformity of the deposition including the uniformity of the substrate temperature profile and the uniformity of the discharge located adjacent to the substrate. Substrate temperature profiles can be controlled or influenced through the appropriate design of the substrate holder cooling or heating. The uniformity of the discharge is an even more complex matter. The size, shape, and uniformity of the discharge are controlled by a number of factors including the discharge chamber shape and size, the discharge kinetics including recombination of charged species in the volume (via volume recombination) and/or on the surfaces (via surface recombination), the generation of radicals in the discharge, and the loss of these radicals in the discharge and on the surfaces surrounding the discharge. Many of the differences in the performance between different microwave diamond deposition CVD systems are concerned with the area or size of the deposition surface across which good uniformity is achieved.

Microwave Plasma-Assisted Diamond Film Deposition

III.

MICROWAVE PLASMA MACHINE DESCRIPTION, PROCESS VARIABLES, AND PERFORMANCE

A.

Microwave Plasma-Assisted Diamond CVD Machine Subsystems

215

A generic microwave plasma-assisted diamond CVD system is shown in Fig. 1. The subsystems include the gas flow rate control system, the vacuum pumping/pressure control system, the discharge chamber and substrate holder, the microwave power supply system, and the microwave energy coupling system, i.e., the microwave applicator subsystem. These subsystems work in unison, generally under computer control, to provide the deposition environment outlined in the previous section. A description of the basic operating purpose and design of each of these subsystems is provided in the following paragraphs. 1.

Gas Flow Rate Control Subsystem

The gas flow rate control system consists of first the gas cylinders containing the feed gases. Because of the sensitivity of the deposition process to impurities in the discharge plasma and the desire to minimize impurities in the deposited diamond, the feed gases should be of high and ultrahigh purity. The flow rate of each gas is controlled with a mass flow controller. The total flow rate ranges in the various systems and processes from tens to thousands of sccm. These controlled flow rate feed gases are then mixed together and delivered to the discharge chamber. 2.

Vacuum Pumping/Pressure Control Subsystem

The vacuum pumping/pressure control system as shown in Fig. 1 is used to pump the reacted gas from the deposition chamber in a continuous manner and to maintain the desired pressure in the deposition chamber. Most diamond deposition systems operate during deposition in the pressure range from a few torr to 200 torr, although values outside this range have certainly been reported. These common pressures of a few torr to 200 torr can be maintained at the flow rates used by using a mechanical roughing pump. The pressure in the deposition chamber is maintained by using an adjustable (throttle) valve located between the pump and the deposition chamber. The pressure is measured with a pressure gauge, most often a capacitance manometer. A control system automatically adjusts the throttle valve to achieve the desired pressure in the deposition chamber. Safety precautions included in the design of the pumping system are first a dilution of the hydrogendominated exhaust gas at the exit of the mechanical pump using a nitrogen

216

Grotjohn and Asmussen

Figure 1 Microwave plasma-assisted diamond CVD system showing gas flow control, vacuum pumping/pressure control, discharge chamber/substrate holder, microwave power supply, and applicator subsystems.

Microwave Plasma-Assisted Diamond Film Deposition

217

purge. The nitrogen dilution is used to reduce the hydrogen concentration to a nonflammable level at the exit of the mechanical pump. A second safety feature often included is a safety interlock system that prevents the pressure in the processing system from exceeding a preset level when hydrogen is present in the system. 3.

Discharge Chamber/Substrate Holder Subsystem

The next subsystem is the discharge chamber/substrate holder unit. The purpose of this system is to confine the plasma discharge and to position the substrate surface for optimum deposition. In the simplest case this consists of a microwave transparent tube that confines the discharge and controls the spatial variation of the processing gas flow. A platform within the tube holds the substrate. In realistic systems this subsystem generally requires some of the most careful detailed design in order to optimize the diamond deposition process. Several design variables of this subsystem work in an interacting manner to establish a plasma discharge shape and size that produces uniform plasma-assisted deposition across the substrate surface. The first set of design variables to be considered is the substrate holder that is designed to have a controlled, uniform, and repeatable substrate temperature. Such controlled temperature is acquired in some systems through active heating or cooling of the substrate holder. Typically, active heating of the holder is required when deposition is done at low pressures (less than 60 torr) or with lower microwave power levels. Conversely, systems that operate at higher pressures (greater than 80 torr) and high input microwave powers often require active cooling of the substrate holder. It is also found in some designs that substrate heating relies solely on heating from the plasma discharge itself with no additional active heating or cooling utilized. This type of substrate holder system is referred to here as a thermally floating system. It is important to note here that when operating in a thermally floating configuration the substrate temperature is determined by the pressure and microwave power. The second set of design variables in the discharge chamber/substrate holder subsystem comprises the size and shape of the discharge chamber and the location of the substrate holder. The discharge chamber is typically constructed of a fused silica bell jar or window and additional stainless steel walls. The fused silica portion of the discharge chamber provides the transparent material for microwave power coupling while providing a gastight chamber. The shape and size of the fused silica/stainless steel confinement region, along with the size and location of the substrate holder, strongly influence the discharge location and shape. Also important to the discharge size and location is the magnitude of the input microwave power and its heating profile. A typical example of how

218

Grotjohn and Asmussen

the experimental substrate temperature, Ts , and plasma volume, Vd , vary with the discharge pressure and input absorbed microwave power, Pt , for a microwave plasma-assisted CVD system is shown in Fig. 2 [5]. The solid lines shown in Fig. 2 have been experimentally measured for a system operating with a 7.5-cm-diameter substrate holder. In this figure the lower limit on the plasma volume, Vd(min), is determined by the discharge volume required to just cover the substrate holder. The upper limit of plasma volume, Vd(max), occurs when the plasma discharge begins to touch the discharge chamber wall, i.e., it just fills the discharge volume. Note that when the hot or bright part of the discharge comes in contact with the walls it can cause substantial wall heating. Thus this condition when Pt is greater than the Vd(max) boundary in Fig. 2 should be avoided. The discharge chamber size for this plot is 12.5 cm in diameter. The general trend in this plot is that as the pressure, p, increases the microwave power absorbed into the discharge

Figure 2 Diamond deposition reactor operating field map showing the substrate temperature in a thermally floating substrate system (no active cooling or heating of the substrate). This experimental plot is for a system with a 12.5-cm-diameter quartz dome and 7.5-cm-diameter silicon substrate described in Refs. 5 and 8.

Microwave Plasma-Assisted Diamond Film Deposition

219

must also increase to maintain a constant plasma volume. The other trend is that the temperature of the thermally floating substrate holder/substrate is determined primarily by the pressure and to a less extent by the input microwave power and flow rate. Plots such as the one given here in Fig. 2 that experimentally quantify the power-pressure region of operation can be developed for any microwave diamond CVD system. Different systems when measured this way can be expected to show similar trends but quantifiably different experimental operating curves. More specifically, these experimental Pt , p, Ts operating curves can be expected to be different for each of the microwave diamond CVD deposition systems described in Sec. IV. The last set of design variables of this subsystem is the location of the feed gas injection port(s), discharge gas flow directors, and exhaust gas pumping ports of the discharge chamber system. Some considerations include having radially uniform feed gas injection and radially uniform pumping of the discharge gas. Further, in some systems, especially those operated at higher pressures and flow rates, the injection ports and gas flow direction surfaces are designed to help improve the uniformity and shape of the discharge and ultimately the deposited film uniformity. At low pressures the discharges are more diffuse and detailed flow design considerations are not as crucial. An overall design issue for the discharge chamber/substrate holder subsystem is the selection of materials that are compatible with the high temperatures and reactive chemistry of the discharge. Common materials used for the substrate holder include molybdenum, stainless steel, and graphite. These materials to various degrees can withstand the substrate deposition temperatures of up to 1100⬚C. The walls of the discharge chamber are often constructed of fused silica and stainless steel. Where possible, water cooling of the walls is also included in the deposition discharge chamber design, especially for the stainless steel walls. 4.

Microwave Power Supply Subsystem

A typical microwave power supply subsystem is shown in Fig. 3. The microwave power is generated by a magnetron power supply most commonly operating at the frequency of 2.45 GHz. Another frequency used for the larger diameter and higher power systems is 915 MHz. The microwave circuit shown in Fig. 3 has the output of the magnetron power supply connected to a circulator that serves to isolate and protect the power supply from any power that may be reflected from the microwave cavity applicator. The reflected power is sent to an impedance-matched dummy load that is typically water cooled. The net power going to the plasma reactor is measured as the incident power, Pinc , minus the reflected power, Pref . The absorbed input

Figure 3

Typical microwave power supply and waveguide circuit.

Microwave Plasma-Assisted Diamond Film Deposition

221

power is then Pt = Pinc ⫺ Pref . Microwave power measurements are made using directional couplers inserted in the transmission line or waveguide between the circulator and the plasma reactor. The directional couplers sample a known fraction of the power flow in a given direction in the waveguide/ transmission line. This sampled power is then measured using power meters. Typical power supplies used in microwave plasma-assisted diamond CVD machines range from 100 W to 8 kW for the 2.45-GHz frequency regime and 6–60 kW for 915-MHz excitation. 5.

Microwave Applicator Subsystem

The last major subsystem is the microwave coupling system, i.e., the external microwave circuit and applicator coupling system. The components of this subsystem are a waveguide, cavity resonator, antenna, or plasma surface wave launcher that focuses or guides the microwave energy to create and maintain the plasma discharge. The variation of this coupling system is the fundamental difference between the many microwave plasma-assisted diamond deposition machine designs. Section IV of this chapter describes several such design variations. Also included in this subsystem are the tuning elements that match the impedance of the transmission line/waveguide to the reactor load. These tuning elements can include stub tuning adjustments, antenna/probe adjustments, resonant cavity size adjustments, and substrate position adjustments, to name a few. These adjustments not only control the impedance matching, which determines the amount of reflected power for a given input power, but also can influence the plasma discharge shape and size, which has a direct impact on the uniformity of the diamond deposition across the deposition surface. B.

Microwave Plasma Reactor Process Variables and Performance Variables

1.

The Multivariable Plasma Reactor

The performance of a microwave plasma diamond deposition reactor is a complex function of many experimental variables. Figure 4 displays a generic block diagram of a typical reactor where the nonlinear relationships between three groups of variables are schematically represented. The three basic groups are (1) input variables, U; (2) internal variables, X; and (3) output variables, Y. The input variables are defined as the variables that can be independently controlled by an experimental operator or reactor designer. As shown in Fig. 4 the input variables can be further subdivided into controllable input variables, U1 , such as operating pressure, methane concentration, gas flow rate, and deposition pressure; the reactor geometry variables,

222 Figure 4

Multivariate parameter space for microwave plasma-assisted diamond deposition.

Microwave Plasma-Assisted Diamond Film Deposition

223

U2 , i.e., the variables concerned with the reactor size and geometry, substrate holder configuration, etc.; and deposition process variables, U3 , i.e., substrate material, deposition time, seeding procedure, etc. The internal variables are defined as the internal plasma reactor states such as gas temperature, electromagnetic field strength, and plasma density. The output variables, Y, can be divided into two groups: those concerned with the reactor performance, Y1 , such as film growth rate and carbon conversion efficiency, and those concerned with the film properties, Y2 , such as film morphology, texture, and film quality. Figure 4, displays a number of example input, internal, and output variables that are important in the diamond deposition process. In order to begin to understand the microwave plasma-assisted diamond deposition process, not only must the relationships between the input variables and the output variables be understood but also the relationships between the reactor internal variables and both the input and output variables must be established. The output variables are a complex function of both the internal variable vector, i.e., Y = h(X), and input variables vector, Y = g(U ). A comprehensive theory that describes the relations between these variables does not exist and is only beginning to be developed. However, knowledge of these complex relations is essential for improved reactor design and ultimately for precise reactor control. In this chapter the important input and internal reactor variables are identified. These variables are then related to the reactor experimental outputs such as reactor performance, Y1 , and film properties, Y2 . This systematic approach enables improved understanding of the plasma reactor system through the identification of the cause-and-effect relationships between input and internal variables and then of the output reactor performance. Changes in reactor inputs such as operating pressure and gas chemistry, and in the reactor size and configuration are then related to reactor performance. 2.

Input Variables

Table 1 identifies a number of typical reactor variables. The independent, controllable input variables, U1 , are substrate temperature Ts , pressure p, incident microwave power Pinc , and feed gas composition and total gas flow rate ft . Reactor geometry variables, U2 , are associated with the applicator and discharge size. Here we assume that the applicator, the substrate holder, and the discharge chamber are all cylindrical. Thus the pertinent dimensions are the applicator radius RA and height LA and the quartz/fused silica discharge chamber radius R c and height L c . The chamber volume is then Vc = ␲R 2c L c . The substrate holder and the substrate itself have radii of Rh and Rs, respectively. These are some of the critical dimensions associated with the design optimization of the reactor and the reactor geometry variables asso-

224

Grotjohn and Asmussen

Table 1 Experimental Variables for Microwave Plasma-Assisted Diamond Deposition Reactors Input variables U

Controllable input variables, U1

Reactor geometry variables, U2

Deposition process variables, U3

Internal variables X

Output variables Y

Reactor performance, Y1 Film characteristics, Y2

Deposition pressure Incident microwave power Feed gas composition Substrate temperature Total flow rate Applicator size and configuration Substrate holder location and size Electromagnetic mode and cavity tuning Quartz dome geometry Substrate material and size Substrate seeding and nucleation procedure Deposition time Absorbed microwave power Plasma volume Absorbed power density Linear growth rate Total growth rate Carbon conversion efficiency Film uniformity Film structural quality Film morphology Film texture

ciated with the U2 variable set. Lastly, the deposition process variables need to be considered including substrate material and size, seeding or nucleation procedure, and deposition time. 3.

Internal Variables

Examples of experimentally measured internal variables, X, are (1) the power absorbed by the reactor, Pt ; (2) the plasma discharge volume, Vd ; and (3) the absorbed power density, < P> = Pt /Vd . The discharge volume Vd is given by ␲R 2dL d where the active discharge radius is R d and the height is L d . The power absorbed by the reactor, Pt , is defined as the difference between the incident microwave power, Pinc , and the reflected power, Pref , i.e., Pt = Pinc ⫺ Pref . The discharge is usually separated from the discharge chamber walls and thus Vd < Vc . The discharge volume Vd is often adjusted (see Fig. 2) by varying the power absorbed so that deposition uniformity is

Microwave Plasma-Assisted Diamond Film Deposition

225

achieved over the substrate surface. Other internal variables include neutral gas temperature, electron density and temperature, radical species concentrations, and gas residence time. 4.

Output Variables: Reactor Performance and Deposited Film Characteristics

A number of reactor process/performance figures of merit have been adopted to enable the evaluation of deposition process performance. These figures of merit, Y1 , are defined as follows: 1. 2. 3.

Linear growth rate (␮m/hr) is defined as the thickness increase of the diamond film (␮m) divided by deposition time (hr). Total growth rate (mg/hr) is defined as the weight increase of the substrate (mg) divided by deposition time (hr). Carbon conversion efficiency (CCE) is defined as the percentage of carbon atoms in the input gases that are converted into diamond. The equation for carbon conversion efficiency is CCE = Nfilm /Ninput where Nfilm is the number of carbon atoms in the deposited film and Ninput is the number of carbon atoms that entered the discharge chamber in the input gas flow.

The deposited film characteristics, Y2 , can include a wide range of properties including film uniformity, texture, morphology, thermal conductivity, electrical properties, and optical properties as well as others. The film characteristics that will be most closely tied to the microwave CVD machines in this chapter include film uniformity, film texture, and film morphology. The other film characteristics are examined in more depth in other chapters of this book. 5.

Quantifying the Input, Internal, and Output Variable Relationships

The sections that follow in this chapter quantify the relationships between the input variables and internal variables, X = f (U ), the input variables and output variables, Y = g(U ), and the internal variables and output variables, Y = h(X). In particular, Sec. IV establishes the range of reactor geometry variables, U2 , and some of the reactor performance relationships, Y1 = g(U2). Section 5 presents the output variables (both reactor performance and deposited film characteristics) as functions of the adjustable input parameters and the internal variables, i.e., Y = g(U1) and Y = h(X). Section VI describes the internal variables and their relationship to input variables based on experimental measurements X = f(U ), and lastly Sec. VII examines in more detail the internal variables based on modeling studies, X = f(U ).

226

Grotjohn and Asmussen

IV.

MICROWAVE PLASMA SOURCES FOR DIAMOND THIN-FILM DEPOSITION

A.

Basis for Comparison

Many different microwave plasma-assisted diamond deposition machines have been designed including tubular reactors, bell jar reactors, plasma jet reactors, ellipsoid reactors, plasma disk reactors, surface wave sustained reactors, and magneto microwave reactors. This section describes some representative reactors for the various types and gives a limited number of specifications for comparison of the reactors. The set of parameters selected for comparison includes microwave coupling structure, microwave frequency, discharge chamber size, substrate size, input microwave power range, and operating pressure range. The diamond deposition outputs (e.g., deposition rate) from each of the different reactor systems are only briefly described in this section. A detailed comparison of the diamond deposited by each system is not attempted because the different systems described here were designed on the basis of different objectives. Some of these design objectives have included designs focused on creating small inexpensive research systems, high-volume production systems, or low-substrate-temperature deposition systems. More detailed discussions of the actual diamond deposition outputs (e.g., deposition rate, film texture) will be undertaken in later sections of this chapter where the outputs will be studied as a function of input variables such as pressure, input chemistry, and flow rate.

B.

Selected Reactor Structures

1.

Tubular Reactor (NIRIM Reactor)

A schematic diagram of a typical tubular microwave reactor is shown in Fig. 5. The tubular microwave reactor was developed in the early 1980s by Kamo et al. [6] and has become the most common diamond film deposition reactor technology used for basic deposition research studies. As shown, the input gas, which is a mixture of hydrocarbon and hydrogen gas, is dissociated by the 2.45-GHz microwave energy coupled into the quartz tube inserted through a waveguide. A plunger is attached to the end of the waveguide to minimize the reflected power. The substrate is placed on a substrate holder located at the intersection of the tube and waveguide. Typical microwave power levels are 100 W to 1.5 kW and the pressure range is less than 80–100 torr. Typical quartz tube diameters are 4.5 cm or less and the substrate sizes are limited to 2–8 cm2. Tubular reactors are simple reactors to design and low in cost to construct, and as a result they have been used in many research investigations.

Microwave Plasma-Assisted Diamond Film Deposition

Figure 5

2.

227

Tubular microwave plasma-assisted CVD reactor. (From Ref. 6.)

Microwave Cavity Plasma Reactor

The microwave cavity plasma reactor (MCPR) [7,8] developed at Michigan State University and Wavemat/Norton in 1986–1995 is shown in Fig. 6. Two design variations are shown including an early version (1987) in Fig. 6a and a later version (1989) in Fig. 6b. The microwave discharge is produced inside a quartz dome which is located at one end of a microwave cavity. The cavity is formed by a cylindrical wall and a movable top short defining the top end of the cavity. The microwave energy is transmitted from a 2.45-GHz microwave generator through a rectangular waveguide to a transition into a coaxial waveguide that ends as a length adjustable excitation probe in the cylindrical cavity. A hemispherical plasma discharge is created and is placed in direct contact of the substrate by adjusting the length of the probe and the location of the sliding short, i.e., L p and L s . The diameter of the microwave cavity is 17.8 cm and the diameter of the discharge region is 9–12 cm. Various substrate holders and substrate sizes can be used ranging from 5 to 10 cm in diameter. The sliding short height L s is most commonly adjusted to a length of approximately 21 cm when a 2.45-GHz microwave excitation is used. This cavity height produces a resonant

228

Grotjohn and Asmussen

Figure 6 Microwave cavity plasma reactor: (a) early version and (b) later version. (From Refs. 7, 8.)

microwave mode that has been identified as the TM013 mode. Typical power levels are 500 to 5000 W and pressures from 5 to 140 torr. Care is taken in the design of the gas inlet structure and substrate holder structure so that the input gas is well mixed with the discharge and to get uniform gas injection and radially uniform discharge chamber exhaust gas pumping. The substrate holder can be designed with various heights. In addition, various substrate holders have been designed that are thermally floating, cooled, and heated. Figure 6a shows an early version of the reactor, which had a discharge diameter of 9 cm and a smaller substrate diameter. Figure 6b shows an improved reactor version designed for use with up to 10-cm-diameter substrates. Also, this more recent unit operates at higher pressures up to 180 torr and powers up to 6 kW. A larger 915-MHz version of the MCPR with a microwave applicator/cavity diameter of 45 cm was also designed, constructed, and operated (1994 at Norton Diamond Film). This larger design has a discharge diameter of 33 cm and substrate diameter

Microwave Plasma-Assisted Diamond Film Deposition

Figure 6

229

Continued

of up to 20 cm. It operates using 8–18 kW of microwave power generated by a 915-MHz power supply. 3.

ASTeX Bell Jar and High-Pressure Microwave Sources

The bell jar microwave plasma-assisted CVD reactor initially developed in 1987 [9,10] is shown in Fig. 7. The basic features of this reactor include a

230

Figure 7

Grotjohn and Asmussen

ASTeX bell jar microwave plasma-assisted CVD reactor. (From Ref. 9.)

silica bell jar that confines the plasma discharge. The initial designs used a 10-cm inner diameter bell jar and 7.5-cm diameter substrate holder. The microwave power generated by a magnetron supply is transmitted to the reactor using a rectangular waveguide. An antenna couples the energy from the waveguide into the reactor, which is cylindrical in shape. The position of the substrate could be moved and the substrate either was thermal floating or could be heated. The pressure range utilized was 40–70 torr and the power range for this unit was typically 1 kW. By 1992, ASTeX had improved the bell jar design and developed the ASTeX High Pressure Microwave Source (HPMS) as shown in Fig. 8 [10]. The former bell jar that confined the discharge and permitted the transmission of microwave energy was replaced by a silica microwave window. A plasma discharge was formed in the region between the silica window and the substrate holder. The position of the substrate could be adjusted to optimize the discharge-substrate interaction and the subsequent film uniformity

Microwave Plasma-Assisted Diamond Film Deposition

Figure 8

231

ASTeX high-pressure microwave source. (From Ref. 10.)

on the substrate. The substrate could be either heated or cooled as needed. The pressure range in which this system operated was from 10 torr to above 120 torr. The power levels used ranged from 1 to 5 kW with later systems able to operate at powers up to 8 kW. The lack of magnetron power supplies above 8 kW at 2.45 GHz prevented the extension of this 2.45-GHz design to even larger areas at the higher pressures. ASTeX designed a 915-MHz system that permitted the use of the higher power 30–60 kW magnetron power supplies [10]. The 915-MHz system had substrate sizes as large as 30 cm. This unit was reported as depositing diamond at a rate of 10 ␮m/hr over a 20-cm-diameter substrate. The mass deposition of diamond approached 1 g/hr. 4.

UC Berkeley Bell Jar Reactor

A schematic diagram of the UC Berkeley bell jar microwave plasma-assisted CVD system [11–14] is shown in Fig. 9. The bell jar is 10.2 cm in diameter

232

Grotjohn and Asmussen

Figure 9

Bell jar plasma-assisted CVD system. (From Refs. 11–14.)

with an annular gas feed system and gas exhaust system. The substrate utilized had a diameter of 2.5 cm and was supported by a quartz sample holder at the position near the plasma ball. This reactor was used extensively to study diamond deposition from solid-phase carbon sources. 5.

France-LIMHP Bell Jar Reactor

A schematic diagram of the LIMHP reactor studied by Gicquel and coworkers [15] is shown in Fig. 10. The reactor is composed of a quartz bell jar 10 cm in diameter that is surrounded by a 25-cm-diameter Faraday cage that acts as a resonant cavity for the microwave energy. The microwave power is coupled into the Faraday cage via a waveguide and an antenna. The substrate holder, which is 5 cm in diameter, can be moved up and down to obtain an optimum interaction between the plasma discharge and the

Microwave Plasma-Assisted Diamond Film Deposition

Figure 10 Ref. 15.)

233

LIMHP-France bell jar plasma-assisted diamond CVD system. (From

substrate leading to good deposition uniformity. The pressure range of this system is 20–120 torr and the input microwave power range is 600– 6000 W. 6.

Ellipsoidal Reactor

A schematic diagram of the ellipsoidal reactor developed by Funer et al. [16] is shown in Fig. 11. The microwave cavity is ellipsoid in shape with the microwave power input being located at the top. The power input from the waveguide to the cavity is done using a probe antenna. The design goal in this structure was to maximize the microwave electric field strength at the location where the plasma is desired, i.e just above the substrate. Using the ellipsoidal shape maximized the fields in this region by locating the end of the antenna at one focus of the ellipsoid and the desired plasma location at the other focus. This reactor configuration was constructed at both 2.45 GHz and 915 MHz. The 2.45-GHz system used 3–6 kW of power for operation in the pressure range from 35 to 150 torr. The substrate size for this system was 5–7.5 cm. The 915-MHz system was used to deposit diamond on 5- to 15-cm diameter substrates. The power for the 915-MHz system was 20–60 kW for the pressure range 35–150 torr.

234

Grotjohn and Asmussen

Figure 11

7.

Ellipsoid microwave plasma reactor. (From Ref. 16.)

Surface-Wave Microwave PACVD Reactor

A diagram of a surface-wave microwave plasma-assisted reactor [17,18] is shown in Fig. 12. The microwave energy is transmitted to the reactor through a waveguide to the waveguide surfatron. The discharge is confined in a fused silica vessel. The plasma is created and stabilized by adjusting the coaxial tuning stub and the waveguide tuning stub. The reactor operates at pressures of 1–60 torr with input power on the order of 1 kW. The substrate holder is heated and made with a molybdenum susceptor. The vertical position of the substrate holder is adjustable. The reactor size is a 2.5 cm inside diameter quartz tube and the substrate size is 2 cm in diameter. The surface-wave microwave plasma-assisted reactor was also scaled to 915 MHz [19]. The substrate holder diameter used in this configuration was 8.0 cm. 8.

Microwave Plasma Jet Torch System

Figure 13 shows a microwave jet torch reactor developed by Mitsuda and coworkers [20,21] starting in the late 1980s. The unit consists of a rectangular waveguide, a microwave transition unit, and a coaxial waveguide. The coaxial waveguide consisted of a 5.72-cm-diameter outer conductor and 2.0cm-diameter center conductor. Microwave energy was generated by a 2.45GHz generator with an output of 3.8–4.2 kW. The input gas mixture used consisted of hydrocarbon, hydrogen, and argon that flowed radially into the coaxial waveguide section and then passed through a nozzle (2.2 cm in diameter) and out of the reactor. A plasma jet was generated near the nozzle

Microwave Plasma-Assisted Diamond Film Deposition

Figure 12 18.)

235

Surface-wave microwave plasma-assisted CVD reactor. (From Refs. 17,

at atmospheric pressure and the resulting dissociated species arrive at the substrate. The substrate/deposition area was 2.5–5 cm2. This design, because of the high pressure, yielded linear diamond growth rates of 30 ␮m/hr. This is one of the highest diamond deposition growth rates reported using microwave plasma excitation. 9.

Magnetomicrowave and ECR Plasma CVD Reactor

A primary motivation for the deposition of diamond at low pressures is to achieve deposition at low substrate temperatures and across large areas. By lowering the pressure, gas heating is reduced and hence the substrate heating is reduced. In addition, at lower pressures the plasma discharge size is larger for a given input power and the plasma also diffuses more freely without volume recombination. At lower pressures microwave plasma sources are often operated with the addition of magnetic fields to improve microwave power coupling to the discharge. The addition of magnetic fields of sufficient strength allows collisionless heating of the electrons by the microwave fields to occur via the electron cyclotron resonance (ECR) mechanism. The use of magnetic fields in reactors with respect to diamond deposition was summarized by Bachmann [10]. He indicated that at pressures exceeding 7 torr the electron collision rate in the plasma is too high to permit substantial

236

Figure 13

Grotjohn and Asmussen

Microwave plasma torch CVD reactor. (From Refs. 20, 21.)

ECR heating. It was also noted that the magnetic fields hardly affect the deposition results at these higher pressures. Further, it was noted that at pressures of 75 mtorr and below, where ECR starts to become important, the diamond growth is significantly affected and it becomes difficult to grow well-defined polycrystalline diamond without getting a mixture of other graphitic–amorphous carbon phases. Well-defined polycrystalline diamond films were deposited by Mantei and coworkers [22–24] using a magnetomicrowave plasma CVD reactor operating at 1–3 torr as shown in Fig. 14. This reactor was built and investigated for the low-temperature deposition of diamond films. As shown, a magnetic field was established using a 15-cm-diameter, 9 cm high NdFeB permanent magnet with a magnetic flux density of 0.42 tesla at the magnet pole face. The discharge was operated with either CW microwave energy or pulsed microwave energy. The substrates used were 2.5 cm in diameter.

Microwave Plasma-Assisted Diamond Film Deposition

Figure 14

V.

237

Magneto-microwave plasma CVD reactor. (From Refs. 22–24.)

DIAMOND DEPOSITION PROCESSES/CHEMISTRIES IN MICROWAVE PLASMA-ASSISTED CVD REACTORS

This section describes the operation of microwave plasma-assisted diamond CVD reactors with respect to discharge chemistry. The focus in this section is on nongeometric input parameter variations. Whereas the previous section described reactors of various sizes, shapes, and designs, this section focuses on the general operation of microwave CVD reactors with respect to a variety of nongeometric input parameters including input gas composition, flow rate, and effect of impurities. The growth of diamond films can be divided into a two-step process beginning with substrate surface treatment or nucleation and followed by the growth process. Nucleation can be initiated through one of a number of processes including substrate surface scratching and bias-enhanced nucleation (BEN). The BEN process is usually performed in the same diamond

238

Grotjohn and Asmussen

deposition reactor as the diamond growth process. The BEN process consists of applying a bias voltage to the substrate to nucleate or start the diamond growth. This process is typically performed with a specific recipe that includes the bias voltage, bias time, and carbon percentage in the feed gas. The diamond deposition reactor operating conditions are then changed once the nucleation step is completed to facilitate diamond growth on the substrate. This section, as well as the rest of this chapter, focuses on the diamond growth process on a prenucleated substrate surface. Specifically, reactor design and operating conditions for bias-enhanced nucleation are not considered further in this chapter. The diamond growth process and results are discussed in this section with an emphasis on deposition done in microwave diamond CVD systems. In particular, the influence of gas chemistry, gas impurities, deposition pressure, and substrate temperature on the growth rate, quality, and texture of diamond deposited in microwave CVD systems is covered. Using the control formulation described in Sec. III.B, this section quantifies when possible the relationships between the controllable input variables, U1 , and the diamond deposition outputs, Y, for microwave diamond CVD systems.

A.

Simplified Surface Reaction/Growth Description for Microwave Plasma-Assisted CVD

The primary function of the microwave discharge is to create the radical species that participate in the growth of diamond. For example, diamond growth by using a hydrogen-methane mixture requires both atomic hydrogen and carbon-containing radicals. The microwave discharge serves to dissociate some of the molecular hydrogen into atomic hydrogen and to dissociate the methane molecules into CH3 and possible other radicals such as CH2 and CH. A general view of the deposition process [25] is one where atomic hydrogen is bonded to almost all of the surface. The primary mechanism that opens sites on the surface is abstraction of surface-terminating hydrogen by atomic hydrogen. The reaction and associated rate, Rabs⫺H , is given by CdH ⫹ H → C * d ⫹ H2 Rabs⫺H = k 1[H] where k 1 is the reaction rate constant and [H] is the atomic hydrogen concentration (flux) at the surface. The abstraction results in an open site on the diamond surface, which can be filled by either the adsorption of another atomic hydrogen or a carbon radical species (e.g., CH3). The reaction and associated rate, RadH , of hydrogen adsorption onto an open site, C * d , is

Microwave Plasma-Assisted Diamond Film Deposition

239

C d* ⫹ H → CdH RadH = k 2 [H] A growth species can also be adsorbed on to the surface with the reaction and associated rate for an open surface site given by C* d ⫹ CH3 → CdCH3 RadC = k 3 [CHx ] Once the growth species is on the surface, it can proceed along one of two primary paths. The first path is thermal desorption from the surface leading to an open site on the surface again. The reaction and associated rate for a given adsorbed site are CdCH3 → C d* ⫹ CH3 R des = k 4 The other pathway is the incorporation of the adsorbed carbon species into the diamond structure. The general event that begins this mechanism in growth models is an abstraction of a hydrogen from the adsorbed carbon species. CdCH3 ⫹ H → CdCH2 ⫹ H2 Rabs⫺CHx = k 5 [H] When the species adsorbed on the surface is CH2 , the carbon is believed to change to having two carbon-carbon bonds. Subsequent hydrogen abstraction and carbon incorporations complete the incorporation of the original carbon atom into the diamond crystal. The conditions that determine the rate of the five mechanisms listed above are surface temperature, gas-phase atomic hydrogen concentration at the surface, and gas-phase carbon radical concentration at the surface. In steady-state conditions, this set of reactions can be combined to give a growth rate G given by [25,26] G = k3

ns nd

冉

k1 k1 ⫹ k2

冊

[CH3][H] k4 ⫹ [H] k5

(1)

where ns is the surface site density (2.61 ⫻ 10⫺9 mol/cm2 on (100) surfaces) and nd is the molar density of diamond (0.2939 mol/cm3). During the deposition process, the surface is understood to be covered nearly 100% with atomic hydrogen as carbon radicals are being deposited in the diamond (sp3) phase. Low hydrogen surface coverage results in sp2like terminations on the diamond surface, which lead to defects and if the

240

Grotjohn and Asmussen

hydrogen coverage is low enough graphitic/amorphous film deposition. An empirical model [25] for the defects, Xdef , incorporated during diamond growth is Xdef ⬀

G [H]n

where n is an empirical fit that is often selected at n = 2. For n = 2 the simplest defect concentration model is given by Xdef ⬀

[CH4 ] [H2 ]

Hence, the higher quality (lowest defect density) diamond is deposited when the carbon species concentration in the input gas flow (e.g., methane concentration/hydrogen ratio) is low. Other work [4,27] has established the condition for achieving diamond growth. One necessary condition is to produce a flux of atomic hydrogen on to the surface that exceeds the carbon incorporation rate into the film by a factor of approximately 7000–11,000. Hence a key function of the plasma discharge in microwave CVD reactors is to provide a flux of atomic hydrogen that exceeds the carbon incorporation rate by a factor in the range of 104. To summarize, the many processes occurring on the diamond surface include adsorption of atomic hydrogen and carbon-containing species, hydrogen abstraction from the surface, and desorption of carbon radicals from the surface. The important variables for the diamond deposition process, including the microwave plasma-assisted CVD systems being considered in this chapter, are (1) substrate temperature, (2) atomic hydrogen flux, (3) carbon radical flux and composition, and (4) other ‘‘impurity’’ species fluxes. The rest of this chapter focuses on the discharge chemistry and diamond deposition results for microwave plasma-assisted CVD systems. Microwave plasma-assisted diamond CVD reactor performance as described earlier in Sec. III.B can be looked at either as the input variable settings determining the film properties (the approach used in this section, Sec. V, of this chapter) or as the input variable settings determining the reactor internal variables (e.g., [H], [CH3 ]) and then the internal variables determining the film properties (the approach used in Secs. VI and VII of this chapter). A somewhat reduced process variable model as shown in Fig. 15 is used as the basis for the discussion in the rest of this section. The input variables are input reactant gas composition, input gas flow rate, substrate temperature, input microwave power, and pressure. The output variables to be assessed include growth rate, diamond film quality, and film texture. The approach in this section is to give a general discussion of the trends for diamond growth in microwave CVD systems for a pressure range

Microwave Plasma-Assisted Diamond Film Deposition

Figure 15 process.

241

Reduced variable model of the microwave plasma-assisted CVD

from 1 to 200 torr for a variety of input gas chemistries. The key reactor input parameters and process quantities to achieve diamond deposition uniformly across a substrate surface are as follows: 1. 2. 3. 4.

5. 6.

Establish the desired input gas flow composition and rate. Establish and maintain a uniform and selected temperature across the deposition surface of the substrate. Establish the exit gas pumping rate to get the desired pressure. At higher pressures construct the input gas flow, discharge chamber, and exit gas flow to optimize the flow dynamics for uniform diamond deposition. Construct the reactor to prevent or minimize contamination of the plasma discharge from leaks and reactor materials via wall erosion. Input the microwave power to establish the plasma discharge. Optimize the input power magnitude, discharge chamber size and cooling, microwave power absorption profile, and substrate position to get uniform deposition across the substrate area.

The first generality in understanding the relationship of the deposition chemistry and output variables is the region of diamond growth with respect to input gas composition. The most commonly used input gas mixtures consist of hydrogen-, oxygen-, and carbon-containing molecules. Because the temperatures and activation of the gases in the plasma discharge are extensive, a primary determining factor for the film composition is the C, H, and O atomic molar composition of the input gas, regardless of the exact input

242

Grotjohn and Asmussen

species molecular composition. A useful summary of the input gas composition appropriate to grow diamond was developed by Bachman et al. in 1990 [1] and further refined in 1994 [28]. The plot in Fig. 16 shows such a ternary C-H-O diagram. The carbon deposition region exists very near and just above the H-CO line connecting 100% H with C-50% and O-50%, i.e., the crosshatched region shown in Fig. 16. The diamond growth region is very narrow to nonexistent on the right end of the H-CO line where [H] approaches zero and becomes wider on the left side of the H-CO line when [H] is larger. Once a specific input gas composition is selected in Fig. 16, the diamond deposition is still a function of a number of other variables. Several other variables are important including pressure, flow rate, and substrate temperature, as described in more detail in the following subsections.

B.

Methane-Hydrogen Deposition Discharge

It is useful first to summarize the general trends for diamond deposition using the simplest discharge chemistry consisting of methane and hydrogen. To do this, some background needs to be developed first. The diamond growth proceeds in a methane-hydrogen discharge as described earlier in this section primarily to supply the [H] flux to the substrate surface and carbon radical flux to the surface. The most important carbon flux in the microwave plasma considered here is the CH3 methyl radical. To deposit diamond, the [H] flux to the surface is quantified as being 7000 to 10,000 times larger than the carbon flux that is incorporated in the growing film [4,27]. When the [H] is larger than this factor, diamond is deposited on a diamond surface. Further, the larger the [H]/[CH3 ] ratio is at the growth surface, the less sp2-bonded carbon and defects are incorporated into the film and the higher is the diamond quality. The other general observation is that the growth rate is proportional to the [CH3 ] flux to the surface as indicated in Eq. (1). The general trends for the growth rate versus temperature, pressure, and methane percentage in microwave plasma-assisted diamond CVD systems are shown in Figs. 17, 18, and 19 respectively. The general trends are 1. 2.

3.

An increase in the growth rate as the pressure is increased [29] as shown in Fig. 17. An increase in the growth rate as the temperature increases up to a maximum, then a drop in the growth rate at higher temperatures [5,30] as shown in Fig. 18. An increase in the growth rate that is linear versus methane percentage at low methane percentages, that is sublinear versus methane concentration for moderate methane percentages, and that

Microwave Plasma-Assisted Diamond Film Deposition

243

Figure 16 C/H/O diagram of diamond CVD developed by Bachmann et al. (From Refs. 1, 28.)

saturates versus methane concentration at higher methane concentrations. These trends can be seen in the experimental data shown in Fig. 19. After the saturation value is reached, the growth rate may drop slightly as the methane concentration is increased further. Typically, in this high methane concentration region, nondiamond carbon or diamond with extensive secondary nucleation is predominantly deposited [5,26]. 1.

Growth Rate versus Pressure and Methane Percentage

One objective of this chapter is to quantify some of the experimental data obtained for microwave CVD systems across a wide parameter space. An empirical model for the growth rate G is G ⬃ R␥ where R is the methane fraction. In work by Harris and Weiner [31] the value of ␥ was pressure

244

Grotjohn and Asmussen

Figure 17 The total weight gain and absorbed microwave power density versus pressure. Note that the data in the 30–80 torr range include 7.5-cm and 10-cmdiameter substrates. The data above 80 torr are for 5-cm-diameter substrates. (From Ref. 8.)

dependent, varying from ␥ = 1 at 20 torr to ␥ = 0.5 at 200 torr. This growth rate model was developed in a hot-filament deposition system. A similar model can be developed for the microwave deposition system. To develop such a model for the microwave system we will consider data across a wide range of pressure and methane concentrations. Some data showing deposition rate versus methane percentage and pressure are shown in Fig. 20 for pressures of 18, 25, 53, and 135 torr [5,26,30,32]. The temperature used for these growth rates is at or near the measured or generally expected temperature range where the growth rate peaks. The methane concentration (#/unit volume) in the plasma discharge is a function of the discharge pressure, the discharge temperature, and the methane percentage in the input gas flow. In a microwave plasma source the discharge gas temperature increases as the pressure increases (see Sec. VI.B), assuming the discharge volume stays approximately constant. The neutral density in the discharge gas is proportional to (pressure)/(gas temperature). An approximate fit to the neutral gas

Microwave Plasma-Assisted Diamond Film Deposition

245

Figure 18 Diamond growth rate as a function of deposition temperature and methane concentration. (From Ref. 30.)

density versus pressure is given by density ⬃ (pressure)0.75 (see data in Sec. VI.B). Hence the density does not increase linearly with the pressure because of the gas heating that occurs as the pressure is increased. This is an empirical fit that models the discharge neutral density versus pressure. The full model for the growth rate is GTsmax = Cp0.75 R where C is a proportionality constant, p is the pressure, and R is the methane percentage. Figure 20 shows a plot of this model and experimental data for growth rates in microwave CVD system for deposition pressures from 18 to 135 torr. This model is for the region of methane concentrations where the growth rate varies linearly with R. The growth rate increases linearly versus the methane concentration for low methane concentrations, but increases beyond this linear region are characterized by the growth rate changing sublinearly versus methane increases as shown in Fig. 19. This can be understood by looking deeper at the carbon chemistry in the plasma discharge. The linear region at low methane concentrations is characterized by the growth species in the plasma

246

Grotjohn and Asmussen

Figure 19 Linear growth rate of diamond versus methane concentration. (From Refs. 5 (a) and 26 (b).)

Microwave Plasma-Assisted Diamond Film Deposition

247

Figure 20 Comparison of growth rate model with experimental data from Refs. 5, 26, 30, and 32. The growth rate model equation is (growth rate = Cp0.75R). This model is valid for low methane concentrations and for a range of substrate temperatures near the substrate temperature that gives a maximum deposition rate.

(CH3) increasing linearly with the methane concentration. Once the methane concentration becomes large enough, the CH3 concentration no longer increases linearly with the methane concentration as shown in Fig. 21 [26]. The gas-phase chemistry present in the discharge converts much of the CH4 to C2H2. This boundary where the CH3 percentage changes from a linear relationship versus methane concentration to a sublinear relationship varies as a function of pressure. A plot of the region of linearity versus pressure is shown in Fig. 22 based on experimental data [5,26,30,32]. The region below

Figure 21 CH3 and atomic hydrogen concentration versus input methane percentage as modeled by Farhat et al. [26] for a pressure of 7000 Pa and a plasma gas temperature of 2850 K.

248

Grotjohn and Asmussen

Figure 22 Regions of growth rate versus methane percentage as a function of methane percentage and pressure based on the experimental data of Refs. 5, 26, 30, and 32.

the dashed line exhibits a linear growth versus methane concentration. This plot uses data for substrate temperatures in the range where the maximum deposition rate occurs versus substrate temperature. 2.

Quality versus Pressure and Methane Percentage

The quality of diamond films can be defined in a number of ways including the number of incorporated defects and the Raman signal. Quality can also be defined in terms of the intended application for the diamond. Some measures of quality include the number of defects incorporated in the diamond crystals, the number or density/rate of secondary nucleation sites, the size of individual crystals in the polycrystalline film, and the nature of the grain boundaries. The general trend is that the larger the hydrogen flux is to the surface and the slower the growth rate (carbon incorporation into the growing film), the higher is the quality. In terms of the Bachmann diagram, the closer the plasma composition moves toward the C-O line, i.e., toward the no-growth region, the higher the quality of the deposited film. Because the definition of diamond film quality often depends on the intended application of the film, a simple definition of quality will be adopted here. This definition of quality has two regimes: the higher quality deposition regime, which has well-defined crystals, and the lower quality regime, which does not have well-defined crystals; i.e., it is extensively renucleated and often referred to as cauliflower-like films. Alternatively, a quality transition can be defined as the transition from a low secondary nucleation value, where well-defined crystals are formed, to a high secondary nucleation value, where cauliflower-like growth occurs. This region of

Microwave Plasma-Assisted Diamond Film Deposition

249

high-quality growth in a methane percentage and pressure parameter space is shown in Fig. 22 based on experimental data [5,26,30,32]. It is the region below the solid line in the figure. The cauliflower diamond growth region is above the solid line in Fig. 22. 3.

Growth Rate versus Substrate Temperature

The deposition rate of diamond from methane-hydrogen gas mixtures generally shows a maximum rate versus substrate temperature as seen earlier in Fig. 18. The temperature range of this peak is in the range 850–1150⬚C with the location of the maximum deposition rate being dependent on the pressure. Experimental results [5,26,30,32] have shown that the maximum growth rate temperature, Tmax , increases gradually with pressure as shown in Fig. 23. Some specific data for the growth rate versus substrate temperature that yields the maximum growth rate include work by Gicquel et al. [30] at 18 torr that gave peak values for substrate temperatures of 850– 950⬚C, Kondoh et al. [33] at 30 torr that gave peak values for substrate temperatures of 900–1000⬚C and Kuo [5,29] at 135 torr that gave peak values for substrate temperatures of 1050–1150⬚C. In even higher pressure, nonmicrowave CVD processes, the substrate temperature of maximum deposition rate at atmospheric pressure is 1200⬚C in work by Snail and Marks [34]. At temperatures less than Tmax , the growth rate dependence on substrate temperature is generally modeled with an Arrhenius expression using an associated activation energy Ea . The general growth rate expression is G ⬀ exp

冉 冊 ⫺Ea RT

Figure 23 Substrate temperature for maximum growth rate, Tmax , i.e., the shaded region, versus pressure from Refs. 5, 26, 30, and 32.

250

Grotjohn and Asmussen

where G is the growth rate, R is the gas constant, and T is the temperature in kelvin. In the temperature range 440–510⬚C, Ulczynski et al. [35] reported an activation energy of 14.6 kcal/mol in the pressure range 5–15 torr. Kondoh et al. [33] using a hot-filament system reported an activation energy of 23 kcal/mol. The data of Gicquel et al. [30] in the temperature range 700–850⬚C showed an activation energy in the range 11–16 kcal/mol. In fact, many results from growth rate measurements of polycrystalline diamond deposition show activation energies of 15–30 kcal/mol [4]. 4.

Growth Rate versus Flow Rate

A key parameter determining the importance of the flow rate is the residence time of the feed gas in the deposition chamber. The residence time, t, is given by t = V/F where V is the discharge chamber volume and F is the flow rate scaled to the pressure in the chamber. Example residence times in a typical discharge chamber range from 1 sec to a few minutes. For example, a 200 sccm (STP) flow is 7600 cc/min at 20 torr and 1000 cc/min at 150 torr. So if the discharge chamber volume is 1 L (1000 cm3) the gas residence time is about 8 sec at 20 torr and 60 sec at 150 torr. While the feed gas is within the plasma discharge region, it is undergoing chemical reactions and some of the carbon is being deposited as diamond on the substrate. A measure of how much of the carbon in the input feed gas is deposited on the diamond substrate is the carbon conversion efficiency. The carbon conversion efficiency increases as the flow rate is reduced as shown for some experimental data in Fig. 24 [29,36,37]. In terms of the effect of this carbon consumption (or sink) on the gasphase chemistry, lower flow rates lead to a gas-phase composition of the discharge that has a lower carbon content than present at higher flow rates. For example, if the input flow has a carbon input of 0.01 mole fraction and the carbon conversion efficiency is 25%, then the exit gas has a carbon concentration of 0.0075 mole fraction. Because of the long gas residence time in typical discharge chambers for diamond growth, the exit gas composition can be considered a close approximation of the carbon mole fraction in the chamber. Hence, to first order a lower flow rate is equivalent to a lower carbon concentration. This is particularly true when the carbon conversion efficiency is high. However, the possible anticipated higher quality diamond expected from lower carbon concentrations (i.e., lower flow rates/ higher carbon conversions) is often not observed because another effect of

> Figure 24

Carbon conversion efficiency versus flow rate. (From Refs. 36, 37.)

Microwave Plasma-Assisted Diamond Film Deposition

251

252

Grotjohn and Asmussen

lower flow rates and the associated longer gas residence times is that any leaks or chamber wall erosion will lead to larger impurity concentrations in the gas discharge. Such impurities can alter and sometimes degrade the deposited film process and properties. An example of the influence of flow rate is the work by Ralchenko et al. [37]. A microwave system using methane and hydrogen at 100 torr, 2.45 GHz and a substrate temperature of 720⬚C was utilized to grow diamond at flow rates ranging from 60 to 1000 sccm. They found the highest quality diamond films intended for millimeter wave windows were deposited at the higher flow rates. This result is suggestive of some impurity in the deposition process being reduced by the higher flow rate. 5.

Growth Rate versus Deposition Time

The growth rate as a function of the time is not constant for the entire duration of the film growth process. Initially the growth rate is slow as the film first nucleates and grows to cover the entire surface of the substrate. Next the growth rate increases as the film texture is formed and the crystals oriented with their fastest growth direction perpendicular to the substrate surface outgrow the other less favorable crystal orientations. Finally, once the film has become well textured, the growth rate becomes constant versus time. This phenomenon is indicated in Fig. 25a for growth at a pressure of 135 torr for 120 h [5]. The same effect was demonstrated in the work of Ralchenko et al. [37] at a pressure of 100 torr as shown in Fig. 25b. 6.

Dependence of Film Texture and Growth Parameter ␣ on Discharge Chemistry

As diamond grows, the shape of the crystals depends on the growth rate of the various surfaces of the crystals. A useful measure for understanding the relationship of growth on various surfaces is to study the growth along some of the primary directions in the diamond crystal structure. One finds that the growth rate along the three directions [100], [110], and [111] changes as a function of gas chemistry and substrate temperature [38]. The growth rate along the [110] direction is generally the fastest. As a result the {110} surfaces are rarely seen in the polycrystalline films grown in CVD reactors, including microwave reactors. This occurs because any (110) surfaces will grow out quickly, leaving other surfaces with slower growth rates defining the crystal. The growth rate in either the [100] direction or the [111] direction is next fastest depending on the growth conditions. Maeda et al. [39] did a study of the growth rate along these two directions in a microwave CVD

Microwave Plasma-Assisted Diamond Film Deposition

253

Figure 25 (a) Linear growth rate of as a function of deposition time [5]. (b) Growth rate versus film thickness. (From Ref. 37.)

254

Grotjohn and Asmussen

reactor. The study (Fig. 26) was done so that the growth rates along the [100] and [111] directions were measured as a function of methane concentration and substrate temperature. A useful parameter for understanding how the growth rates along the various directions determine the crystal shape is the growth parameter ␣ defined as

␣ = 兹3

V100 V111

where the growth rates along the [100] and [111] directions are given by V100 and V111 , respectively. When ␣ = 1 or less, the [111] direction grows quickly and the individual isolated crystals formed are cubes with (100) surfaces as shown in Fig. 27. Fig. 27 shows the diamond crystals that grow for values of ␣ ranging from ␣ = 1 to ␣ = 3 [40]. Also shown in Fig. 27 is the direction of fastest crystal growth for the various ␣ parameter values. The direction of fastest growth is indicated by the arrow by each cubooctahedral shape. For ␣ = 3 or larger, the [100] direction grows most quickly and the crystal shape formed has (111) surfaces. Conversely, for ␣ = 1 or less, the [111] direction grows most quickly and the crystal shape is formed by the more slowly growing (100) faces. In the case of ␣ = 1 or less, the longest direction in the crystal is the [111] direction. For ␣ = 3 or larger, the longest direction in the crystal is the [100] direction. At an intermediate value of ␣ = 1.5 the longest direction in the crystal is [110]. A plot of the ␣ parameter versus methane concentration and substrate temperature from the work of Maeda et al. [39] is shown in Fig. 28. The picture becomes more complicated when a polycrystalline film is grown because many crystals intersect and overgrowth occurs. The growth rates along the [100] and [111] directions determine the texture of the film and the surface morphology of the film. The texture of the film is the crystal direction that is longest in the individual crystals of the polycrystalline film. The texture in polycrystalline diamond films develops because of the crystals which have an orientation that has a direction of fastest growth perpendicular to the substrate surface. These crystals overgrow the crystals with other orientations. This is shown in Fig. 29, where a random orientation of small crystals is shown at the bottom of the film. As the film grows, the crystals that grow with their direction of fastest growth perpendicular to the surface become taller than their competitor crystals. Ultimately, these crystals with their fast growth direction normal to the substrate in Fig. 29 overgrow the competition, ending the possibility of further growth of the other crystals with less preferred orientations. A reliable measure of the ␣ parameter can be obtained using a texture analysis of the polycrystalline films from x-ray diffraction measurements.

Microwave Plasma-Assisted Diamond Film Deposition

255

Figure 26 Growth rate along the [100] direction (a) and along the [111] direction (b) versus substrate temperature for various methane concentrations. (From Ref. 39.)

256

Grotjohn and Asmussen

Figure 27 The cubo-octahedral shapes that result as the ␣ parameter is varied from 1.04 to 2.85. These are the shapes for isolated crystallites. (From Ref. 40.)

Measurements of the velocity of growth of diamond along various crystalline directions in microwave CVD systems have been made in work by Wild, Koidl, and their colleagues [41,42] as shown in Fig. 30 and by Maeda et al. [39] as shown in Figs. 26 and 28. The trend shown in these studies is that as the temperature increases the ␣ factor decreases. Also, these studies in general showed that as the methane percentage is increased, the ␣ factor increases. These figures display the general trends for ␣ only. The exact value of ␣ at a specific substrate temperature and methane percentage may change from reactor system to reactor system because different reactor geometries result in a different relationship between the input, internal, and output variables; i.e., the different reactor geometries discussed in Sec. IV have unique and different Pt , p, and Ts operating curves as discussed earlier in Sec. III.A.3. As will be noted in the following sections on the influence of various impurities, the addition of impurities also affects the growth along various crystalline directions and consequently the ␣ parameter value. The other property of polycrystalline diamond films strongly dependent on the growth rates in various directions is the surface morphology. The surface morphology can be defined as the crystal shapes, orientations, and

Microwave Plasma-Assisted Diamond Film Deposition

257

Figure 28 Growth rate ratio ([100]/[111] growth rate ratio) versus substrate temperature for various methane concentrations. (From Ref. 39.)

sizes as observed on the top surface of the growing polycrystalline film. Relating the direction of maximum growth rate and ␣ with the surface morphology of polycrystalline films must be done with great care. To appreciate this relationship, consider some randomly oriented seed crystals grown with ␣ = 1. As the individual crystals grow, they will have a cubic shape similar to the shape shown for ␣ = 1 in Fig. 27. Once the cubes start growing together, the surface will initially have the appearance of many cubes all of random orientation with respect to each other and the substrate surface. As the film grows to a larger thickness, the 具111典 texture forms because for ␣ = 1 the [111] direction grows fastest (see the arrow direction for the ␣ = 1 shape in Fig. 27). So as the film grows perpendicular to the substrate surface, the [111] direction will dominate and clear cube shapes will no longer be clearly seen. Rather, the corners of the cubes formed by overlapping and overgrowing crystals will be seen on the top surface. As a general statement, what the surface of films thick enough to have developed texture looks like can be roughly estimated as that seen by looking back down along the direction of the arrow for the crystals in Fig. 27. Commonly, films with square {100} facets are desired, and these can be formed in two basic ways. The first way is to operate with ␣ near 1. This will produce the square shapes during the initial stages of growth when the film thickness is only on the order of the initial spacing between the seed

258

Grotjohn and Asmussen

Figure 29 Sketch of polycrystalline diamond film growth from isolated, randomly oriented crystals shown at the bottom of the figure. The dashed lines indicate the grain boundaries. The solid lines show the surface of the film at selected times. (From Ref. 40.)

crystals or nucleation sites. Specifically, randomly oriented seed crystals will grow into small cube shapes with ␣ = 1. Continued growth under low secondary nucleation conditions will form a 具111典 texture because this direction grows the longest dimension (fastest). So observing square (100) facets when the films are thin may indicate values near ␣ = 1. The square facets in this case have orientations that are not well aligned with the substrate surface, i.e., faces parallel to the growth surface. Another way to form square facets is to grow a thick film with a 具100典 texture. This is done by using ␣ value slightly less than 3. As the film grows thick, the texture forms and the [100] directions are perpendicular to the substrate surface. The value of ␣ slightly less than 3 leads to the square tops on the crystals in the film. See, for example, the square tops on the crystal shapes for ␣ = 2.0 and ␣ = 2.25 in Fig. 27. These square facets are well aligned parallel to the surface of the growing film because of the strong texture of the film. This is the growth condition that leads to smooth coplanar (100) faceted films. So care must be used in interpreting the ␣ parameter value from just the examination of the surface morphology of polycrystalline films.

Microwave Plasma-Assisted Diamond Film Deposition

259

Figure 30 Growth parameter (␣) versus methane concentration for three substrate temperatures. An ␣ parameter of 1.5 gives a 具110典 texture and an ␣ parameter of 3.0 gives a 具100典 texture. (From Refs. 41, 42.)

Once the ␣ parameter has been mapped, two-step processes [43] have been developed that use an ␣ = 3 or near 3 to grow highly 具100典 textured films during the initial growth phase. This gives the film structure shown in Figs. 31a and 32b. Then for the second growth phase the growth conditions are changed so that the ␣ parameter is reduced in value. This increases the relative growth rate of the [111] direction relative to the [100] direction and the film growth fills in the valleys located between the (100) faces present in the initial growth phase. The resulting film has a top surface covered with (100) facets all reasonably parallel with the substrate surface as shown in Figs. 31b and 32b. Another factor that influences the surface morphology is the extent and influence of twinning that can occur. Wild and coworkers [42] indicated that the ␣ parameter also influences the formation and effect of twinning. They found that for ␣ > 2 the {100} facets are stable with respect to twinning.

260

Grotjohn and Asmussen

Figure 31 Schematic drawing of a two-step diamond deposition process. Step one grows the film so that is has a 具100典 texture, then step two grows the film so that smooth (100) faces dominate the top surface of the film. (From Ref. 43.)

Stable means that if a small twin occurs on a (100) face it will disappear as the film continues to grow. Conversely, an unstable condition is one where a small twin grows larger as the growth continues. The other stability condition noted in their work was that for ␣ < 1.5 the {111} facets are stable with respect to twins with an inclined twin plane. C.

Influence of Oxygen (Hydrogen-Methane-Oxygen Discharges)

The addition of oxygen to the input gas flow, either as molecular oxygen, O2 , or as another gas such as CO or CO2 , can be useful for diamond deposition. The use of some oxygen in the input gas (either O2 , CO, or CO2) generally yields films with larger grain sizes and less sp2 material. Also, the inclusion of oxygen has been shown to increase the deposition rate in some cases and the addition of oxygen allows deposition to occur at lower substrate temperatures.

Microwave Plasma-Assisted Diamond Film Deposition

261

Figure 32 SEM pictures of a polycrystalline diamond film after step one (a) and after step two (b and c) as described in the caption of Fig. 31. (From Ref. 43.)

262

Grotjohn and Asmussen

A basic requirement to achieve diamond deposition is that the C/O ratio must be near one (more specifically, equal to or slightly greater than one as shown earlier in the crosshatched region of Fig. 16). For oxygen-rich discharges no diamond growth occurs and for carbon-rich discharges (except those with large atomic hydrogen concentrations) nondiamond carbon is deposited. Any excess oxygen will etch the carbon from the surface, preventing growth, and any excessive carbon in large amounts is deposited as nondiamond carbon films. In general, a trade-off exists between higher growth rates (and the associated lower quality) that occur on having slightly more carbon versus lower growth rates (and the associated higher quality) that occur at slightly higher oxygen concentrations. Some additional understanding of diamond deposition with oxygencontaining discharges can be achieved by considering the gas-phase chemistry. The highly activated, high-temperature discharge environment of diamond deposition drives the mixture of carbon and oxygen species in the reactor toward primarily a CO molecular composition. If the feed gas mixture has substantial percentages of oxygen and carbon present, most of the oxygen and carbon are converted into CO. Four statements can be made regarding the influence of oxygen on the diamond deposition process: 1. 2.

In general, with the addition of oxygen, the substrate temperature can be dropped 100–150⬚C without a loss of growth rate [10]. The use of oxygen changes the growth rate, under some conditions to a higher rate. Figure 33 shows the growth rate versus percent oxygen (O2) added in the feed gas flow [44]. The deposition conditions used in this study by Weimer et al. [44] consisted of a tubular reactor operating at a 21 torr pressure, 170 W microwave power, and a 97 sccm H2 flow rate. The top portion of the figure shows the growth rate with a 2 sccm methane flow. It can be seen that as the oxygen flow is increased from 0 to 0.06% in the feed gas, the growth rate increases by a factor of 1.6. A similar phenomenon occurs for an acetylene feed gas as shown in the lower portion of Fig. 33. Also observed is that as the oxygen percentage increases further the growth rate drops. This presumably occurs because as the oxygen content in the discharge is increased more CO is generated, leaving less active carbon species available for deposition. Further, as the feed oxygen percentage increases, more oxygen radicals are available to etch the surface. In another study [10] of the influence of oxygen on the growth rate, the gas-phase chemistries near C/O/H = 1:1:1 gave a maximum deposition rate. This can be seen in Fig. 34, where the deposition rate peaks at the C/O/H = 1:1:1 condition [10].

Microwave Plasma-Assisted Diamond Film Deposition

263

Figure 33 Diamond growth rate versus oxygen percentage in the feed gas flow. The hydrogen gas flow rate was 97 sccm. (From Ref. 44.)

3.

The oxygen reduces sp2 incorporation in the film and also reduces secondary nucleation, which leads to larger crystals and altered surface morphology. This is believed to occur because of the increased etching effect of oxygen radicals on graphitic structure in the film [45]. It has also been noted by Harker and DeNatale [46] that the crystallite size increases as increasing oxygen content is

264

Grotjohn and Asmussen

Figure 34 Diamond growth rate versus C/H/O feed gas composition for two pressures of 70 and 140 torr. The feed gas composition of C/H/O = 1:1:1 gave the highest growth rate. (From Ref. 10.)

4.

added to the gas feed due to increased reactive etching rates during deposition that result in decreased crystal defect density. The introduction of oxygen into the feed gas influences the ␣ parameter (growth rates along the [100] and [111] directions), film texture, and film morphology. Figure 35 shows a plot of the parameter ␣ versus methane concentration at a fixed substrate temperature of 800⬚C [42]. The first of two observations to be made is that when oxygen is present in the discharge, variations in the methane concentration produce changes in the ␣ parameter in a manner similar to the no-oxygen case (i.e., ␣ increases as methane concentration increases). The second observation is that oxygen introduction with other conditions held constant produces shifts in the ␣ parameter and hence the texture and surface morphology. The change of ␣ (as indicated by film texture) with changes in the oxygen input flow was also demonstrated by Gu et al. [47]. In their study CO (0–10%) was added to a 100 sccm H2 and 1 sccm CH4 feed gas flow. The deposited films showed a large increase in the (100) XRD peak (more 具100典 texture) as the CO flow increased from 0 to 1%. Further, as the CO flow percentage in-

Microwave Plasma-Assisted Diamond Film Deposition

265

Figure 35 Growth parameter ␣ versus methane concentration for two different carbon monoxide concentrations of 17 and 67%. The substrate temperature was constant at 800⬚C. (From Ref. 42.)

creased to 4% and 10% the (100) XRD peak decreased, indicating a reduction in the 具100典 texture. A full mapping of the ␣ parameter dependence on oxygen content, hydrogen content, carbon content, and substrate temperature has not been done [40].

D.

Influence of Nitrogen Gas on Diamond Synthesis

In the mid-1990s it was demonstrated experimentally that the addition of small amounts of nitrogen (N2) gas to the conventional CH4-H2 diamond film synthesis gas has an important impact on the CVD synthesis of diamond films [48–53]. For example [49], high N concentrations result in nanocrystalline and black films containing graphite or amorphous carbon, and very high N concentrations, i.e. 20% N concentration can completely inhibit diamond growth. In contrast, the addition of only 20–100 ppm of nitrogen can have a beneficial effect. At a deposition pressure of 20–60 torr, an N

266

Grotjohn and Asmussen

concentration of only a few tens of ppm stabilizes the formation of smooth, 具100典 textured diamond films. In particular, Locher et al. [48] using a tubular reactor demonstrated that at intermediate methane concentrations (1–2%) at 37 torr, the addition of several tens of ppm of N2 stabilized the formation of smooth 具100典 textured diamond films. When no nitrogen was added with the CH4 percentage held equal to 1.5%, the film deposited had fine grains indicative of a high secondary nucleation rate or high twinning rate. When the nitrogen input level has set to 60 ppm, well-defined (100) faceted crystals grew with 具100典 texture. The introduction of nitrogen also increases the growth rate along the [100] direction which shifts the ␣ parameter to a higher value. This was also shown in the tubular reactor experiments of Jin and Moustakas [49]. The texture of the films shifts from 具110典 to 具100典 when tens of ppm to a few hundred ppm of nitrogen is added. This is further demonstrated in a homoepitaxial study by Wild et al. [50], who showed that the addition of 10–50 ppm of nitrogen increased the [100] growth rate much more than the [111] growth rate. Hence, the addition of small amounts of N2 to the deposition process can be used to stabilize the 具100典 growth surface. At higher deposition pressures and higher input power densities, the deposition rate was observed to increase by a factor of 3–5. Muller-Sebert et al. [51], employing a tubular reactor that operated at a pressure of 167 torr, a deposition temperature 830⬚C, and a methane concentration of 3%, demonstrated that the addition of only 25 ppm of nitrogen changed the growth rate from 2.6 to 13 ␮m/hr. Asmussen et al. and Ayres et al. [52], using a microwave cavity plasma reactor as shown in Fig. 6b, also studied the influence of nitrogen concentration from 5 to 1000 ppm at pressures of 38 and 50 torr. They noted a variation in growth rate of less than 10% over the range 5–1000 ppm. The substrate temperatures used in this study were T = 850⬚C at 38 torr and T = 910⬚C at 50 torr. Examples of how the nitrogen addition influenced the surface morphology are shown in Fig. 36. These experiments indicate that, as in the tubular reactor experiments, the addition of small amounts (200– 400 ppm) of nitrogen stabilizes the growth of high-quality 具100典 faceted films. However, the actual threshold nitrogen concentrations and the variation in these threshold concentrations versus the other independent experimental variables differ considerably from the tubular reactor performance, suggesting that reactor geometry has an important impact on the deposition process in the presence of nitrogen impurities. It is also interesting to note that even though nitrogen has a strong effect on the growth process, only a very small amount of the nitrogen is actually incorporated into the film [49,53]. Typical values for the incorpo-

Microwave Plasma-Assisted Diamond Film Deposition

267

Figure 36 SEM images of nine samples deposited with 200 sccm flow rate, 38 torr pressure, and 1% methane. Nitrogen gas-phase concentration and diamond Raman signal peak width are given at the top of each micrograph. (From Ref. 52.)

ration probability are 2 ⫻ 10⫺4 for the {100} faces and 7 ⫻ 10⫺4 for the {111} faces [53]. E.

Chlorine- and Fluorine-Based Deposition Chemistries

Diamond has been deposited with halogen gases and halogen-containing gases included in the feed gas flow. Halogen-containing gas mixtures including CF4 , Cl2 , CH3F, CH3Cl, C2H2Cl2 , C2H5Cl, C3F8, and others have been used to grow diamond [10,54]. Some of the advantages of using hal-

268

Grotjohn and Asmussen

ogens as noted in the literature include enhancement of growth rate, improvement of diamond quality, and reduction of growth temperature. The role of chlorine or fluorine in the deposition process is believed to be due to hydrogen abstraction from the surface. A model [55] for fluorine and chlorine is that the abstraction of the hydrogen from the surface is done more quickly by atomic fluorine and atomic chlorine at lower temperatures. This opens up sites more often for the attachment of a carbon species (such as CH3). The region of diamond growth for the C/H/Cl system in a microwave CVD system [10,56] occurs for a mixture of predominantly hydrogen with carbon composition percentages of 2–3% or less and chlorine composition percentages of 40% or less to get diamond. The region of diamond growth for the C/H/F gas composition was again for a mixture of predominantly hydrogen with C percentages of 2% or less and a fluorine composition percentage of 12% or less. The advantage of using halogen to lower the deposition temperature in hot-filament systems is well documented. Such an advantage is less clear in microwave CVD systems. Bachmann et al. [10,56] reported studies of C/H/F and C/H/Cl systems in which no distinct advantage in terms of growth rate or lower deposition temperature was found as compared with the C/H/O system. F.

Inert Gas/Methane/Hydrogen Deposition Chemistries

Two basic influences of adding a noble or inert gas to the input gas flow for diamond deposition have been identified. The first is an enhancement of the atomic hydrogen and the growth species CHx concentration in the plasma discharge by the addition of Xe. This work was done by Hosomi and coworkers [57]. They observed an increase in the deposition rate of diamond on adding 1% Xe to the feed gas composition of 1 part methane and 99 parts hydrogen. They used a microwave CVD system operating at 800 W 2.45 GHz, and 40 torr. The low excitation energy level of Xe resulted in a plasma discharge with a higher electron density. This then produced more atomic hydrogen and CH3 and a 50% increase in the deposition rate with the Xe addition. The second influence of inert gases occurs when large quantities of a noble gas are added to the input gas flow. In a series of studies [58,59] at Argonne National Laboratory, large percentages of argon in the input gas flow together with low hydrogen percentages have been used to deposit nanocrystalline diamond. Work by the Argonne group has shown that for Ar/H2 /CH4 (1% CH4) discharges the transition from a discharge containing no argon (hydrogen dominant) to one containing no hydrogen (argon dom-

Microwave Plasma-Assisted Diamond Film Deposition

269

inant) is accompanied by fundamental changes in the plasma chemistry, growth process, growth rate, and deposited film properties. However, throughout the entire variation of hydrogen-argon compositions with a small percentage of methane (⬃1%), diamond is grown. The plasma chemistry is observed to change from a plasma dominated by CH3 and H radicals at high hydrogen concentrations to a plasma having a high C2 concentration at high argon concentrations. Corresponding to the discharge chemistry changes, the diamond growth scenario for 70–99% hydrogen is dominated by the CH3 growth species, for the 20–70% hydrogen region the growth rate is dominated by the C2H2 species, and for the low hydrogen content region (less than 20%) the growth is dominated by the C2 species [59]. The C2 species is postulated as both a diamond growth species and a diamond nucleation species. The use of a predominantly argon discharge with small amounts of CH4 (⬃1%) and small amounts of hydrogen (0 to a few percent) permits the deposition of very smooth diamond films with randomly oriented 3- to 10-nm crystallites of diamond. Such diamond is referred to as nanocrystalline diamond.

VI.

INTERNAL VARIABLES

Several researchers have studied the internal variables in microwave plasmaassisted CVD diamond deposition reactors. These studies have quantified some of the relationships of input to internal variables. Some of the internal variable quantities measured have been species concentrations, gas temperature, electron density, electron temperature, and microwave power absorption density versus various input parameters. In this section these internal variable studies are reviewed with an emphasis on the studies that quantified the internal variable values. In particular, optical emission studies that provide only relative signal levels or relative concentration levels are not considered in this review. A.

Species Concentrations

Mitomo and coworkers [60] used in situ Fourier transform infrared (FTIR) spectroscopy to measure the molecular species concentrations in the microwave discharge region. In their study they used a number of different input gas mixtures including CH4 ⫹ H2 , C2H2 ⫹ H2 , C2H4 ⫹ H2 , CH3OH ⫹ H2, C2H5OH ⫹ H2 , CO ⫹ H2 , CCl4 ⫹ H2 , and CH4 ⫹ O2 ⫹ H2 for the discharge. The reactor was a tubular unit, shown in Fig. 5, operated at 2.45 GHz with a 4-cm-diameter tube, 30 torr pressure, and 300 W microwave power. The carbon-containing species observed using FTIR included CH4 , C2H2 , C2H4,

270

Grotjohn and Asmussen

and CO. The CH4 and C2H2 species were observed under all the conditions studied, while C2H4 was observed only when higher concentrations of carbon input gases were used. Lastly, CO was produced and measured when oxygen was present in one of the input molecular gas species. The species CH3OH, C2H5OH, and CCl4 were not seen in the discharge, indicating that plasma conditions were effective at dissociating these molecules. As shown in Fig. 37, when the input gas contained no oxygen the dominant species for low percentages of input carbon gas flow (i.e., 0.5% or less) was methane and the dominant species for percentages above 0.5% was acetylene. This was true regardless of the molecular form of the input gas. Another observation was that the addition of oxygen to the discharge reduced the concentrations of CH4 and C2H2 in the discharge as compared with the levels when no oxygen was present. However, the CO species concentration increased as oxygen was added. Hence, the discharge conditions in microwave CVD

Figure 37 Concentration of CH4 , C2H2 , and C2H4 in a microwave discharge region as measured by FTIR absorption spectroscopy versus percent methane in the feed gas at an operating pressure of 30 torr with no oxygen in the feed gas. (From Ref. 60.)

Microwave Plasma-Assisted Diamond Film Deposition

271

reactors favor the formation of CO; i.e., the addition of oxygen results in bonding of carbon with oxygen, forming CO. A mass spectrometric study by Weimer et al. [44] found similar results. They used a differential pumped quadrapole mass spectrometer to sample the exhaust gas from a microwave tubular reactor being operated for the deposition of diamond. The reactor had a 1.25-cm-diameter tube operating at a pressure of 21 torr, a microwave power of 170 W, and a frequency of 2.45 GHz. The input gas flow consisted of hydrogen, oxygen, and various carbon-containing molecules including CH4 , C2H2 , C2H4 , and C2H6 . The species detected in the exhaust gas included CH4 , C2H2 , CO, and H2O. Figure 38 shows plots of the various detected species as a function of the oxygen percentage in the input gas flow (0–3%). Once again, the dominant carbon species at low oxygen input was either CH4 or C2H2 . At oxygen flow rates of 1% or greater the dominant species was CO and the concentrations of methane and acetylene dropped as the oxygen flow increased. Using a molecular beam mass spectrometer that is capable of measuring radical and stable species in the discharge, Hsu [61] studied diamond deposition in an ASTeX microwave reactor (see Fig. 7). The experimental structure was constructed so that the sampling point of the discharge was a hole in the center of the substrate holder. The deposition conditions used during the mass spectrometer measurements included a substrate temperature of 1073 K, a pressure of 20 torr, and an input microwave power of 800 W. The species detected by the spectrometer included H2 , H, CH3 , CH4 , and C2H2 . The results, which are summarized in Fig. 39, show measured concentrations versus the methane input flow percentage from 0.2 to 3%. The two radicals measured were atomic hydrogen and the methyl radical, CH3 . The dominant carbon species at low methane input flow was methane, whereas at methane flow rates of 0.5% and greater the dominant species was acetylene. The measured hydrogen mole fraction was about 10⫺3. This, of course, is the measured value at the surface of the substrate holder where the discharge is sampled by the spectrometer. It was further estimated by Hsu that the H-atom mole fraction in the center of the discharge located above the substrate holder would be 8%. He also found that the hydrocarbon chemistry in the microwave system was largely unchanged from that found in thermal systems (i.e., hot-filament systems). This was due to the carbonhydrocarbon chemistry being driven by the gas temperature and the atomic hydrogen concentration. The primary effect of the plasma was suggested to be the dissociation of the hydrogen molecules to atomic hydrogen. Erickson et al. [62] measured in a unique microwave reactor configuration other radical species concentrations in a microwave discharge reactor. The measurement technique used was absorption spectroscopy. The species measured included CH3 and CH at a pressure of 20 torr in the discharge.

272

Grotjohn and Asmussen

Figure 38 Diamond deposition reactor exhaust gas composition versus oxygen percentage in the feed gas for input gas flows of 97 sccm hydrogen and (a) 2 sccm CH4 , (b) 1 sccm C2H6 , (c) 1 sccm C2H4 , and (d) 1 sccm C2H2 . (From Ref. 44.)

Microwave Plasma-Assisted Diamond Film Deposition

273

Figure 39 Mole fraction of chemical species measured in a microwave plasma CVD system versus variation in the methane fraction in the feed gas. Measurements done using a molecular beam mass spectrometer. (From Ref. 61.)

Based on the measured values of CH3 and CH at 20 torr, the atomic hydrogen mole fraction was estimated to be 0.8% or less. Wouters and coworkers [63] looked at hydrogen discharges created in a 1.6-cm-diameter tubular microwave reactor operating in the pressure range from 2 to 40 torr. They used the two-photon laser-induced fluorescence (LIF) method to measure the atomic H density when this discharge was operated in a pulsed mode with a 40% duty cycle at 100 Hz using 200 W of power. The results obtained for the absolute atomic hydrogen density and mole fraction are shown in Fig. 40. They reported gas temperatures in the discharge region were 1600 K at 40 torr and 1200 K at 2 torr as measured by H-atom Doppler broadening. The atomic hydrogen mole fraction can be seen to vary from 8% at 2 torr to 2% at 40 torr. Atomic hydrogen mole fraction in an LIMHP-France bell jar microwave CVD reactor (shown in Fig. 10) was measured by Gicquel and coworkers [26,64]. They used the method of actinometry, where the emission signals are properly calibrated so that absolute densities can be obtained. Actinometry is a technique that looks at the strength of optical emission

274

Figure 40 63.)

Grotjohn and Asmussen

Atomic hydrogen percentage and density versus pressure. (From Ref.

signals from the discharge. Specifically, the emission from atomic hydrogen and the emission from a gas added at a low concentration (single-atom gas such as argon) are both measured from the same discharge. If the electron excitation energy and cross section of excitation of the two emitting species are similar, then the density of the atomic hydrogen species can be determined because the density of the argon species is known from the input flow rate. Further, if the signal intensities are properly calibrated, the absolute densities can be determined. In this case the results obtained for the atomic hydrogen concentration are shown in Fig. 41 [26,64]. The system used for this measurement was a microwave resonant reactor with a 5-cmdiameter substrate and a 10-cm-diameter discharge region. The power density variation (6–30 W/cm3) was obtained by varying the pressure and the input power to obtain the same plasma volume as the pressure was increased from 18 to 100 torr. The power density and pressure are strongly coupled with pressure increases producing power density increases. As shown in Fig. 41a, the atomic hydrogen mole fraction increases from about 2 to 20% as the pressure increases. Figure 41b displays the variation of the atomic mole fraction of hydrogen as the methane concentration is increased from 0 to over 5%. Only a slight decrease is observed at the low power density of 9 W/cm3. At the higher power density of 22.5 W/cm3, no significant change in the atomic hydrogen concentration is observed as the methane concentration increases from less than 1% to 6%.

Microwave Plasma-Assisted Diamond Film Deposition

275

Figure 41 Atomic hydrogen mole fraction versus the microwave power density in part (a) and versus methane concentration in part (b). (From Ref. 26.)

B.

Neutral Gas Temperature

The neutral gas temperature of the discharge in microwave plasma-assisted diamond deposition reactors has been measured in a number of ways including Doppler broadening of optical emission signals, Doppler broadening of LIF signals, coherent anti-Stokes Raman spectroscopy (CARS), and rotational temperature of optically emitting molecules. One of the most systematic studies of the gas temperature in a diamond deposition reactor has been done by Gicquel and coworkers [65,66]. They used all four of the techniques just listed to measure the gas temperature in an LIMHP-France bell jar microwave discharge reactor (Fig. 10) operating at 2.45 GHz. This reactor had a discharge diameter of 10 cm and a substrate diameter of 5 cm.

276

Grotjohn and Asmussen

The results found by measuring the line-of-sight Doppler broadened emission of atomic hydrogen are shown in Fig. 42. The gas temperature ranges from 1600 K at the low power density (18 torr) to 3500 K at the highest power densities studied, which correspond to a pressure of 105 torr. Measurements performed using laser spectroscopy techniques are shown in Fig. 43. The two techniques utilized were CARS to measure molecular hydrogen and two-photon LIF to measure atomic hydrogen. The CARS technique measures the rotational temperature of molecular hydrogen. The LIF technique measures the Doppler broadening due to atomic hydrogen translational motion. Both of these techniques can be performed with good spatial resolution. For the data shown in Fig. 43 the location of the measurement was in the center of the discharge at a distance of 20 to 25 mm above the substrate. The effect of methane addition on the gas temperature was also studied as shown in Fig. 44. The results obtained by the line-of-sight optical

Figure 42 Atomic hydrogen translation temperature versus microwave power density. The temperature was measured using Doppler broadening of hydrogen OES signals. (Adapted from Refs. 65, 66.)

Microwave Plasma-Assisted Diamond Film Deposition

277

Figure 43 Gas temperature versus average microwave power density. Gas temperatures measured include atomic hydrogen temperature by two-photon LIF and hydrogen rotational temperature by CARS. (Adapted from Refs. 65, 66.)

emission measurements of the rotation spectrum of molecular hydrogen are shown in Fig. 45. The temperatures in this case range from 1300 K at 9 W/ cm3 to 1900 K at 22 W/cm3. These rotational temperature values are systematically lower than those found by the other techniques. Similar rotational temperatures versus pressure were also measured by Grotjohn et al. [67] for the MCPR reactor system of Fig. 6b. One variation of this reactor has a 5cm discharge chamber diameter and a 3-cm-diameter substrate holder. These results are shown versus pressure in Fig. 46. Some understanding of the difference in the temperature measurements can be obtained by considering the location and species being measured. The Doppler broadening measurement of atomic hydrogen is weighted toward the region that has the highest atomic hydrogen concentrations. That region is the center of the plasma, which has the highest gas temperature. Conversely, the molecular hydrogen, which is the species being observed

278

Grotjohn and Asmussen

Figure 44 Atomic hydrogen temperature (measured by LIF) and molecular hydrogen rotational temperature (measured by CARS) versus methane percentage in the feed gas. (Adapted from Refs. 65, 66.)

for the OES (optical emission spectroscopy) rotational temperature measurements, will have a higher density near the edges of the discharge where the gas temperature is lower. Hence, the line-of-sight rotational temperature measurements are weighted more toward the edges of the plasma discharge, which are expected to be cooler. C.

Electron Density and Temperature

The electron density in the discharges operated under conditions that deposit diamond have been measured. Grotjohn and coworkers [67] used a millimeter wave Fabry-Perot resonator operating at 30 GHz to measure the electron density in a 2.45-GHz microwave reactor operating in the pressure range from 10–60 torr. The results of the measurements are shown in Fig. 47 versus pressure. The density range is 1–5 ⫻ 1011 cm⫺3. In another investigation using a tubular reactor with a diameter of 2 cm, Tahara et al. [68] studied methane-hydrogen discharges in the pressure range from 0.1 to

Microwave Plasma-Assisted Diamond Film Deposition

279

Figure 45 Molecular hydrogen rotational temperature versus average microwave power density. Temperature was measured using light-of-sight OES. (Adapted from Refs. 65, 66.)

5 torr. The tube was placed through the center of a 2.45-GHz tunable resonant cavity. The methane flow rate in the input gas flow was varied from 1 to 10%. As shown in Fig. 48, the plasma density, which was measured with a Langmuir probe, stayed approximately constant at 1 ⫻ 1011 cm⫺3 across the pressure range studied. The microwave power utilized was 350 W. The last internal variable quantity is the electron temperature, which was measured in the pressure range 0.1 to 5 torr by Tahara et al. [68] using the system described in the preceding paragraph. The electron temperature values varied from slightly above 3.5 eV at the lower pressure of 0.1 torr to about 2eV at the higher pressure of 5 torr. D.

Microwave Power Density

Another important internal variable is the power density of the plasma discharge. Power density is defined by knowing the absorbed power of the

280

Grotjohn and Asmussen

Figure 46 Molecular hydrogen rotational temperature versus pressure for a 5-cmdiameter discharge system. (From Ref. 67.)

discharge and dividing by the volume of the discharge. The volume is typically estimated from the region of most intense light emission by the discharge. An example plot of the power density versus pressure measured for the MCPR reactor of Fig. 6b operated with a 3- to 4-cm diameter discharge and 3.2-cm-diameter substrate holder is shown in Fig. 49 [67]. Even though the pressure is a primary determining factor for the power density, other factors such as substrate size and total plasma volume are also determining factors. For example, the power density of the microwave system studied here does not match exactly the power density of other microwave diamond CVD systems with different diameters of substrate holders. In work by Kuo and Asmussen [8], the power density found for a microwave reactor with a 7.5-cm-diameter holder at 60 torr was 15 W/cm3 (see Fig. 17). In another system [2,30,65,66], the power density for a microwave system with a 5cm-diameter substrate holder operating at 60 torr was 22.5 W/cm3. So at the 60 torr pressure and 2.45 GHz excitation frequency, microwave systems with substrate holder diameters of 3.2, 5, and 7.5 cm have power densities of 32, 22.5, and 15 W/cm3, respectively. This relationship indicates that larger diameter plasmas, which are needed to cover larger substrates, operate with a lower power density than smaller diameter plasmas.

Microwave Plasma-Assisted Diamond Film Deposition

281

Figure 47 Electron density versus pressure for a microwave plasma-assisted diamond deposition discharge. (From Ref. 67.)

E.

Microwave Electric Field and Mode

The last quantity to be considered in this section is the measurement of the impressed microwave electric field strength and the spatial variation of the electromagnetic mode in diamond deposition reactors. The specific system considered here is the microwave cavity plasma reactor (MCPR) as shown earlier in Fig. 6b. The microwave electric field has been measured in the MCPR system by inserting a microcoaxial antenna probe as shown in Fig. 50 through holes drilled in a pattern along the applicator cavity walls, i.e., along the vertical or z direction and around the circumference. The microcoaxial antenna probe samples the electric field component normal to the wall, which in this case is the radial component. The antenna probe is inserted through the applicator cavity holes a small distance (⬃a few millimeters) so that only a small fraction of microwave power going into the applicator cavity is coupled into the sampling antenna probe (typically

282

Grotjohn and Asmussen

Figure 48 Electron temperature and plasma density versus pressure for a cavity resonance microwave hydrogen-methane discharge. (From Ref. 68.)

⬃10⫺6). This small fraction for the amount of power sampled by the antenna probe ensures that the probe does not perturb the field in the applicator. An example plot of the measured electric field strength versus cavity vertical position (z direction) is shown in Fig. 51 [69,70]. The symbols show the experimentally measured microwave electric field values in a discharge loaded applicator cavity. The dotted line is the expected strength of the radial field variation versus vertical position along the applicator cavity for the TM013 mode. A sketch of the ideal TM013 mode is shown in Fig. 52. Note in this figure that the electric field has three maximums along the longitudinal direction, which is also seen in the experimental data of Fig. 51. Although not shown in Fig. 51, when the microwave electric field is sampled around the circumference of the applicator cavity wall, the electric field strength is unchanged in agreement with the MPCR operating in a TM013 electromagnetic mode.

VII.

MICROWAVE DIAMOND CVD REACTOR MODELING

Microwave plasma-assisted CVD reactors for diamond deposition have been modeled in a number of investigations. The modeling of microwave CVD reactors needs to consider a number of factors including:

Microwave Plasma-Assisted Diamond Film Deposition

283

Figure 49 Microwave power density versus pressure for 5-cm-diameter hydrogenmethane discharge used for diamond deposition. (From Ref. 67.)

1. 2. 3.

Electron gas–initiated reactions in the discharges Neutral chemistry (including thermally activated) reactions in the discharges Microwave field excitation, propagation/resonance, and absorption including the spatial patterns of microwave fields and microwave power absorption

First, a few preliminary comments regarding H2 /CH4 discharge and microwave discharge operation in the pressure range from 1 to hundreds of torr are in order. The dissociation reactions, which are so important to diamond deposition (i.e., dissociation of H2 and CH4), can be driven by either electron dissociation or thermal dissociation. The general trend as shown in Fig. 53 for all molecular gases is that at low pressures, the electron temperature is high and the gas temperature is low so electron dissociation dominates. Conversely, at high pressures, the electron temperature is reduced

284

Grotjohn and Asmussen

Figure 50 Details of the microwave electric field measurement probe and a cutaway view of a microwave cavity plasma reactor.

Microwave Plasma-Assisted Diamond Film Deposition

285

Figure 51 Microwave electric field strength (radial direction) versus vertical position for a microwave cavity plasma reactor as shown in Fig. 50.

Figure 52

Sketch of TM013 electromagnetic resonant mode.

286

Grotjohn and Asmussen

Figure 53 Electron, ion, and neutral temperatures versus pressure for nonequilibrium (nonisothermal) and thermal-like plasmas. (From Ref. 10.)

and the gas temperature increases, resulting in increased thermal dissociation. The exact pressure range over which this transition occurs is dependent on many factors including molecular species type, other constituents in the plasma, and the plasma volume. One of the objectives in this section is to establish an understanding of this transition in the diamond deposition process. A.

Modeling Studies

Several researchers have studied diamond deposition in microwave plasmaassisted CVD systems. The key attributes that distinguish the various models are: 1. 2.

Spatial dimension of the solution [zero (one point), one, or two dimensional] Discharge chemistry model (hydrogen, hydrogen-methane, H-CO)

Microwave Plasma-Assisted Diamond Film Deposition

3.

287

Microwave power absorption profile (uniform, assumed profile, self-consistently solved profile)

Koemtzopoulos et al. [71] in 1993 reported on a model for microwave discharges that looked at hydrogen dissociation. The reaction rates were solved for by determining the electron energy distribution function (EEDF) using the Boltzmann equation. The model considered electron dissociation of the hydrogen as the primary dissociation mechanism. The model solved the electron density balance equation, the atomic hydrogen balance equation, and the power balance equation at pressures ranging from 15 to 100 torr. The geometry considered was a tubular reactor with the electric field and species concentrations assumed uniform throughout the plasma volume. The model showed that the electron distribution function is non-Maxwellian with the tail of the distribution function depleted by inelastic collisions. Further, it was demonstrated that even when 1% CH4 was added to the hydrogen discharge of EEDF did not change appreciably. This work also noted that the dissociation of methane by abstraction of atomic hydrogen (CH4 ⫹ H → CH3 ⫹ H2) is faster than electron dissociation at pressures as low as 20 torr with a gas temperature of 600 K and an atomic hydrogen fraction of 0.08. Hyman et al. [72] in 1994 developed a one-point numerical model of microwave plasma CVD diamond deposition reactors. The model solved the Boltzmann equation to determine the electron energy distribution function. A dominant energy loss mechanism for the electron gas was the vibrational excitation of the hydrogen molecules. The gas heating then occurred predominantly through vibrational-translational energy exchange. This model considered discharges consisting of carbon-hydrogen-oxygen mixtures. In the model the point of calculation was selected as the center of the plasma ball located above the substrate holder. Diffusion and heat conduction effects were accounted for by two boundary points fixed at the substrate condition and the wall/ambient condition. Their simulation example was given at 40 torr pressure, 30 W/cm3 microwave power density, and an input gas mixture of 98.75% H2 , 1% CH4 , and 0.25% O2 . Typical values of the plasma conditions simulated at time scales approximating steady state (after 4 msec of reaction time) are an electron temperature of 2.5 eV and a gas temperature of ⬃3000 K. In their model the initial gas mixture of H2 , CH4 , and O2 reacts to a mixture consisting of H2 , C2H2 , and CO (stable species) and H, C, CH3, and C2H (radical species). Hence the methane and oxygen molecules in the discharge are converted predominantly to C2H2 and CO molecules. Work by Tan and Grotjohn [69,73] on microwave plasma reactors for diamond deposition looked at the microwave fields and the absorption of the microwave energy by the discharge. They obtained full-wave two-di-

288

Grotjohn and Asmussen

mensional solutions of Maxwell’s equations to solve for the electromagnetic fields using the FDTD (finite-difference time-domain) method. The work simulated the resonant fields in a resonant cavity microwave reactor. Experimental measurements and simulations of the 2.45-GHz microwave fields showed close agreement. This model was limited to hydrogen discharges in the reactor. The reaction kinetics included electron dissociation and ionization processes of hydrogen, but thermal processes were either left untreated or treated in a very simple manner. The gas temperature in their model was set on the basis of experimental gas temperature measurements. This modeling work solved for the electromagnetic fields and the power absorption of the electron gas in the microwave reactor. Extensive modeling work on hydrogen and hydrogen-methane discharges has been done by Hassouni et al. [74–80] at LIMHP-CNRS in France. They developed one-dimensional discharge models for moderatepressure H2 and H2 /CH4 plasma discharges. They also developed a twodimensional model of H2 discharges using an assumed microwave power absorption profile. Their modeling work considered the electron energy, H2 vibrational energy, and total energy. The models use Boltzmann equation solutions to determine the EEDF, which is important for determining electron reaction rates because the EEDF is non-Maxwellian. The model also included gas heating in the discharge region. Extensive comparisons with good agreement are made of the simulated gas temperature and the experimentally measured gas temperature, which was obtained by CARS and Doppler broadening resolved OES. The hydrogen discharge considered in their modeling work included either 7 or 9 species and the H2 /CH4 model included an additional 12 neutral carbon-containing species and 9 ionic carbon species. The carbon chemistry for the pressure range considered (18– 100 torr) has been shown to be primarily governed by thermal chemistry with the gas temperature and the atomic hydrogen concentrations being the primary neutral carbon chemistry driving factors. The work by Tan and Grotjohn [69,73] on modeling the microwave fields in microwave diamond deposition reactors was combined with the work on hydrogen discharges by Hassouni et al. [74–80] to produce a twodimensional self-consistent microwave field and plasma discharge simulation for a moderate pressure (18–80 torr) hydrogen discharge reactor [81]. This model produced a self-consistent solution of the microwave fields in the plasma reactor, the absorption of the microwave energy by the discharge electron gas, and the hydrogen plasma discharge. The plasma model used was a three energy mode (gas, molecular vibration, and electron) with nine ⫹ ⫺ ⫺ species [H2 , H, H (n = 2), H (n = 3), H⫹, H⫹ 2 , H3 , H , and e ] model that accounted for the non-Maxwellian electron distribution function and had 35 reactions. Good agreement was obtained between the model results and the

Microwave Plasma-Assisted Diamond Film Deposition

289

experimental data for gas temperatures and atomic hydrogen concentrations. The model was also able to predict plasma discharge (ball) size, shape, and location in the discharge chamber with good accuracy. The specific reactor simulated, as shown in Fig. 10, had a 25-cm-diameter microwave resonant cavity structure operating at 2.45 GHz with a silica bell jar of 10 cm diameter. The substrate holder inside the bell jar had a diameter of 5 cm as shown in Fig. 10. Example two-dimensional plots of the gas temperature, atomic hydrogen mole fraction, absorbed microwave power density, electron temperature, and electron density are shown in Fig. 54. Numerical simulations of microwave plasma reactors for diamond CVD have also been done by Funer et al. [82,83]. In their study microwave fields are modeled in two dimensions (r and z) using a finite integration technique to solve Maxwell’s equations for the reactor structure. The plasma model used was a cold plasma dielectric function description of the discharge where the density of the discharge was calculated on the basis of the local electric field and a fitting parameter ␥ that is adjusted to have the modeling results agree with experimental observations. Their work looked at using the numerical model to design an optimum reactor. The various geometric dimensions of the reactor are described by parameters P1 , P2 , . . . , Pn . An optimization is done to determine the set of parameter values that maximize the ratio of the electric field (兩E兩2) within the plasma region to the electric field strength (兩E兩2) in the region surrounding the plasma region. The optimum parameter set was determined using the method of ‘‘direction of steepest gradient’’ to find the optimum solution. The result of this study was the elliptical reactor described earlier in Sec. IV, Fig. 11 [16]. B.

Internal Variables: Modeling Results

From the various modeling efforts, a number of important understandings regarding microwave-assisted diamond CVD reactors can be extracted. The first is that the chemistry of neutral carbon species (e.g., dissociation of CH4 to CH3) is essentially driven by gas temperature and H-atom concentration. Hassouni et al. [74] demonstrated using a one-dimensional H2 /CH4 plasma model that for the pressures they studied (18 torr and above) the carbon chemistry is driven by thermal and atomic hydrogen [H] processes and not by electron excitation or dissociation of carbon species. Another understanding that can be drawn from the various modeling results is the relative importance of thermal dissociation and electron collisional dissociation of hydrogen. First, Fig. 55 shows the atomic hydrogen concentration under equilibrium conditions due to just thermal dissociation at temperatures from 1600 to 2800 K (dashed line) for the two pressures of 20 and 100 torr. The equilibrium molar fraction for a pressure of 20 torr

290

Grotjohn and Asmussen

Figure 54 Spatial distribution (versus r and z) of discharge characteristics at 2500 Pa and 600 W. (a) Gas temperature (K), (b) atomic hydrogen model fraction, (c) absorbed microwave power density (W/cm3), (d) electron temperature (K), and (e) electron density (⫻1011 cm⫺3). (From Ref. 81.)

Microwave Plasma-Assisted Diamond Film Deposition

291

Figure 55 Hydrogen dissociation (molar percentage) versus gas temperature for hydrogen in equilibrium conditions (dashed line) and hydrogen in a plasma discharge used for diamond deposition (solid line).

goes from 0.038% at 1600 K to 1.1% at 2000 K to 41% at 2800 K. Data are also shown in Fig. 55 for a pressure of 100 torr. These quantities are taken from the work by Argoitia et al. [84]. In the two-dimensional simulations of a 2.45-GHz microwave plasma-assisted CVD reactor with a discharge size of approximately 5 cm diameter by Hassouni et al. [81] the simulation results yielded an average hydrogen molar fraction and gas temperature as a function of pressure as shown in Fig. 56. Note that as the pressure increases the gas temperature and the hydrogen molar fraction also both increase. If these gas temperature and hydrogen molar fraction values are now also plotted (Fig. 55), the solid line labeled 2.45 GHz is obtained. In Fig. 55, the lowest pressure of 18 torr shows that the hydrogen dissociation in the microwave CVD system is substantially greater that that expected from the gas temperature. Hence, at this lower pressure electrondriven dissociation of the discharge is an important process. At the higher pressures, the hydrogen molar fraction is approximately the same as or less than the thermal dissociation value based on the discharge gas temperature. Hence, at 40 torr and above the thermal dissociation of hydrogen becomes a dominant process. The modeled/measured value of the hydrogen dissociation is less than the equilibrium value because, in the reactor, the discharge chamber walls and cooler gas regions outside the active discharge volume provide sinks for atomic hydrogen via surface recombination and volume recombination. The influence of reactor scaling on the atomic hydrogen molar fraction is also interesting to consider. A larger diameter microwave diamond dep-

292

Grotjohn and Asmussen

Figure 56 (a) Plasma gas temperature versus pressure in a hydrogen-dominated diamond deposition discharge. (b) Atomic hydrogen percentage versus pressure in a hydrogen-dominated diamond deposition discharge.

osition reactor operating at 915 MHz with a substrate diameter of 15 cm was also simulated using the model in Ref. 81. The simulated data for the larger system are shown in Fig. 55 as a solid line labeled 915 MHz. It can be seen that at the higher gas temperatures of 2400 K and above the larger diameter 915 MHz system has an atomic hydrogen molar percentage that is closer to the thermal equilibrium values (dashed lines) than the smaller 5cm-diameter discharge 2.45-GHz system. Another important factor affecting the %H and the gas temperature at a selected pressure is the input microwave power. Table 2 shows the gas temperature versus input microwave power at a pressure of 18 torr in a

Microwave Plasma-Assisted Diamond Film Deposition

293

Table 2 Simulated Plasma Discharge Quantities as a Function of Input Microwave Power at a Pressure of 2500 Pa Microwave input power (W) Discharge parameters Vp (cm3) Tg peak value (K) Tg average value (K) Tg peak location (K) Te peak value (K) Te average value (K) Te peak location (cm) MWPD peak value (W/cm3) MWPD average value (W/cm3) MWPD peak location (cm) ne peak value (cm⫺3) ne average value (cm⫺3) ne peak location (cm) xH peak value (%) xH average value (%) xH peak location (cm)

300

500

700

800

900

19 1,730 1,500 1.3 19,840 16,030 0.3 18.1 8.7 0.3 3.3 ⫻ 1011 2.0 ⫻ 1011 0.9 0.65 0.4 1.1

33.6 1,920 1,620 1.5 19,920 16,000 0.3 19.2 8.5 0.3 3.4 ⫻ 1011 2.1 ⫻ 1011 1.1 0.81 0.52 1.5

49 2,020 1,640 1.7 19,650 15,930 0.3 18.4 8.0 0.3 3.3 ⫻ 1011 2.1 ⫻ 1011 1.3 0.9 0.55 1.5

61 2,080 1,710 1.9 19,570 15,730 0.3 17.5 7.4 0.3 3.3 ⫻ 1011 2.1 ⫻ 1011 1.3 0.93 0.60 1.7

74.5 2,065 1,640 1.9 19,560 15,320 0.3 17.2 6.4 0.3 3.1 ⫻ 1011 1.8 ⫻ 1011 1.1 0.9 0.52 1.7

Source: Ref. 81.

hydrogen discharge [81]. The primary factors affecting the gas temperature and percent hydrogen concentration once a particular pressure is selected are the plasma volume and the plasma boundary temperature (substrate temperature and wall temperatures). In general, the larger the plasma volume (as determined by the discharge chamber size and the input microwave power), the higher will be the gas temperature. The variations of other internal variables versus microwave input power are also given in Table 2. A companion table showing the variation of the internal variables versus pressure for hydrogen discharges of similar volumes is Table 3. Tables 2 and 3 provide a picture of the internal variable values and their variation with the pressure and microwave power (two input variables). One final utility of the models is in the design and operation of a reactor to get a uniform deposition across large areas. It has been noted by researchers that the uniformity of the deposition is controllable by the adjustment of the substrate position, microwave tuning circuit, pressure, and amplitude of the input microwave power. For example, work described by Bachmann [10] showed that films can be grown in the same reactor that vary from thickest in the center of a wafer area, to relatively uniform, to

294 Table 3

Grotjohn and Asmussen Simulated Plasma Discharge Quantities as a Function of Pressure/Power Pressure (hPa), microwave input power (W)

Discharge parameters 3

Vp (cm )* Tg peak value (K) Tg average value (K) Tg peak location (cm) Te peak value (K) Te average value (K) Te peak location (cm) MWPD peak value (W/cm3) MWPD average value (W/cm3) MWPDav experiment (W/cm3) MWPD peak location (cm) ne peak value (cm⫺3) ne average value (cm⫺3) ne peak location (cm) xH peak value (%) xH average value (%) xH peak location (cm)

25–600

52–1000

84–1500

110–2000

44.1 1,990 1,640 1.7 19,680 15,900 0.3 18.3 8.0 9.0 0.3 3.4 ⫻ 1011 2.1 ⫻ 1011 1.1 0.87 0.55 1.5

40.7 2,600 2,100 1.7 cm 17,220 14,600 0.3 40.1 16.0 15.0 0.3 5.1 ⫻ 1011 3.1 ⫻ 1011 1.1 1.19 0.78 1.5

39.8 3,100 2,540 1.7 cm 15,680 13,600 0.3 67.8 25.0 22.5 0.3 7.4 ⫻ 1011 4.3 ⫻ 1011 1.1 2.91 1.66 1.7

45 3,200 2,760 1.7 cm 13,990 12,400 0.3 79.1 30. 30. 0.3 9.1 ⫻ 1011 5.3 ⫻ 1011 1.1 6.98 3.8 1.9

*Vp, plasma volume. Source: Ref. 81.

thickest at the edge of the deposition area. Such a dependence of the uniformity was also seen in the discharge itself in Ref. 81, where simulations with the substrate holder at three difference vertical positions yielded the results of Fig. 57. Specifically, variation of the substrate height produced changes in the gas temperature and in the shape of the plasma discharge region above the substrate. Raising the substrate holder is shown in Fig. 57 to flatten the plasma so that it is more uniform across the diameter of the substrate.

VIII.

CONCLUSION

Microwave plasma-assisted CVD machines create and control the chemical environment and substrate conditions for diamond growth. The appropriate chemically active environment supplies neutral fluxes, such as CH3 and C2H2 , that provide the carbon for the growth process and also supplies a high-density atomic hydrogen flux so that deposition occurs in the diamond

Microwave Plasma-Assisted Diamond Film Deposition

295

Figure 57 Spatial distribution of gas temperature (K) at three substrate positions. (a) Lowered 1 cm, (b) nominal position, and (c) raised 1 cm. (From Ref. 81.)

296

Grotjohn and Asmussen

phase. The deposition machine must also control the substrate temperature to achieve the desired diamond film uniformity and properties. Since the early 1980s, when Kamo et al. [6] created the tubular reactor, substantial progress and major innovations have been made in the design and development of microwave plasma-assisted CVD machines. A wide range of machine types and sizes have been developed and have demonstrated successful diamond deposition. Using current technology and know-how, microwave plasma-assisted CVD is achieved over a wide experimental regime, including pressures ranging from 10 m torr to 1 atmosphere, microwave powers ranging from 100 W to 60 kW, and substrate diameters ranging from 2.5 to 30 cm. Today, microwave CVD systems can deposit diamond films uniformly across substrates as large as 30 cm in diameter and at rates of nearly 1 g/hr. By carefully adjusting the chemistry and controlling impurities, diamond films with the desired quality and morphology can also be synthesized. Considerable engineering has gone into the design of these systems, and reliable, automated operation is currently available for commercial applications. The approach utilized in this chapter was to view these diamond machines in terms of the relationships between the input variables, internal variables, and output variables. The relationships between these experimental variables are developed from both experimental studies reported over the past 15 years in the literature and modeling investigations of specific diamond deposition machines and their associated diamond deposition processes. Much has been learned from experimental and modeling investigations about the relationships between the input variables, such as pressure, gas flow rate and composition, and input microwave power, and the internal variables, i.e., chemistry, of the discharge. This understanding has come both from careful experimental measurements of the internal variables including gas temperatures, plasma density, radical densities, and microwave power densities and from careful modeling studies that can now solve for the plasma discharge kinetics and the microwave fields self-consistently in two dimensions. These studies, which were reviewed in this chapter, provide a very useful picture for the understanding of the engineering and scientific issues associated with the design of microwave plasma-assisted CVD machines and the diamond deposition process itself. Microwave plasma reactors are often used for fundamental scientific experimental investigations because microwave plasma-assisted machines operate without electrodes or filaments in contact with the discharge. Hence, electrode, filament, and other contamination issues can be minimized. For example, microwave plasma reactors have been extensively used to investigate the influence of impurities (even at levels as low as 10 ppm) on the deposition process and film properties. One result of these investigations has

Microwave Plasma-Assisted Diamond Film Deposition

297

been the design and development of very reliable, low-maintenance microwave plasma-assisted CVD machines that have very high vacuum system capabilities as well as the ability to control impurity gases to levels below a few ppm. Thus applications that require high-quality diamond synthesis, such as diamond window synthesis as well as synthesis of electronic quality films, are now usually carried out using microwave plasma reactors. However, even with the advances and successes over the past 15 years in the areas of machine design and process understanding, many interesting design and process understanding issues remain. Some of these include understanding the influence of various impurities or additives on the deposition process, e.g., the influence of small amounts of nitrogen, boron, and oxygen across the range of deposition conditions. Other issues center around the design of deposition machines that permit ever more uniform diamond deposition across large areas while minimizing unwanted contamination or impurities from the machine itself, i.e., impurities from wall erosion, substrate erosion, and fused silica bell jar or window erosion found at the high temperatures and in the highly active chemical environments of the deposition process. REFERENCES 1. 2.

3. 4.

5.

6. 7.

8.

PK Bachmann, D Leers, H Lydtin. Towards a general concept of diamond chemical vapour deposition. Diamond Relat Mat 1:1–12, 1991. W Zhu, BR Stoner, B Williams, JT Glass. Growth and characterization of diamond films on nondiamond substrates for electronic applications. Proc IEEE 79:621–646, 1991. JE Butler, H Windischmann. Developments in CVD-diamond synthesis during the past decade. MRS Bull 23:22–27, 1998. DG Goodwin, JE Butler. Theory of diamond chemical vapor deposition. In: MA Prelas, G Popovici, LK Bigelow, eds. Handbook of Industrial Diamonds and Diamond Films. New York: Marcel Dekker, 1998, pp 527–581. K-P Kuo. Microwave assisted plasma CVD of diamond films using thermallike plasma discharge. PhD dissertation, Michigan State University, 1997; and J. Zhany. Experimental Development of Microwave Cavity Plasma Reactors for Large Area and High Rate Diamond Film Deposition. PhD dissertation, Michigan State University, 1993. M Kamo, Y Sato, S Matsumoto, N Setaka. Diamond synthesis from gas phase in microwave plasma. J Cryst Growth 62:642–644, 1983. J Zhang, B Huang, DK Reinhard, J Asmussen. An investigation of electromagnetic field patterns during microwave plasma diamond thin film deposition. J Vac Sci Technol A 8:2124–2128, 1990. K-P Kuo, J Asmussen. An experimental study of high pressure synthesis of diamond films using a microwave plasma reactor. Diamond Relat Mater 6: 1097–1105, 1997.

298 9.

10.

11.

12.

13.

14.

15. 16. 17. 18.

19.

20. 21. 22.

23.

24. 25.

Grotjohn and Asmussen PK Bachmann, W Drawl, D Knight, R Weimer, RF Messier. In: A Badzian, M Geis, G Johnson, eds. Diamond and Diamond-like Materials. MRS Symposium Proceedings, EA-15, 1988, p 99. PK Bachmann. Microwave plasma chemical vapor deposition of diamond. In: MA Prelas, G Popovici, LK Bigelow, eds. Handbook of Industrial Diamonds and Diamond Films. New York: Marcel Dekker, 1998, pp 821–850. MC Salvadori, JW Ager III, IG Brown, KM Krishnan. Diamond synthesis by microwave plasma chemical vapor deposition using graphite as the carbon source. Appl Phys Let 59:2386–2388, 1991. MA Brewer, IG Brown, MR Dickinson, JE Galvin, RA MacGill, MC Salvadori. Simple, safe and economical microwave plasma-assisted chemical vapor deposition facility. Rev Sci Instrum 63:3389–3393, 1992. Z Feng, K Komvopoulos, IG Brown, DB Bogy. Effect of graphite carbon films on diamond nucleation by microwave plasma-enhanced chemical vapor deposition. J Appl Phys 74:2841–2849, 1993. Z Feng, MA Brewer, K Komvopoulos, IG Brown, DB Bogy. Diamond nucleation on unscratched silicon substrates coated with various non-diamond carbon films by microwave plasma-enhanced chemical vapor deposition. J Mater Res 10:165–174, 1995. A Gicquel, K Hassouni, S Farhat, Y Breton, CD Scott, M Lefebvre, M Prealat. Diamond Relat Mater 3:581–586, 1994. M Funer, C Wild, P Koidl. Novel microwave plasma reactor for diamond synthesis. Appl Phys Lett 72:1149–1151, 1998. CFM Borges, M Moisan, A Gicquel. A novel technique for diamond film deposition using surface-wave discharges. Diamond Relat Mater 4:149–154, 1995. CFM Borges, L StOnge, M Moisan, A Gicquel. Influence of process parameters on diamond film CVD in a surface-wave driven microwave plasma reactor. Thin Solid Films 274:3–17, 1996. S Schelz, C Campillo, M Moisan. Characterization of diamond films deposited with a 915 MHz scaled-up surface-wave-sustained plasma. Diamond Relat Mater 7:1675–1683, 1998. Y Mitsuda, T Yoshida, K Akashi. Development of a new microwave plasma torch and its application. Rev Sci Instrum 60:249–252, 1989. K Takeuchi, T Yoshida. The effect of oxygen on diamond synthesis in a microwave plasma jet. J Appl Phys 71:2636–2639, 1992. JJ Chang, TD Mantei, R Vuppuladhadium, HE Jackson. Effects of oxygen and pressure on diamond synthesis in a magneto-active microwave plasma. J Appl Phys 59:2918, 1992. Z Ring, TD Mantei, S Tlali, HE Jackson. Plasma synthesis of diamond at low temperatures with a pulse modulated magnetoactive discharge. Appl Phys Lett 66:3380–3382, 1995. Z Ring, T Mantei. Low-temperature diamond growth in a pulsed microwave plasma. J Vac Sci Technol A 13:1617–1618, 1995. DG Goodwin. Scaling laws for diamond chemical-vapor deposition. I. Diamond surface chemistry. J Appl Phys 74:6888–6894, 1993.

Microwave Plasma-Assisted Diamond Film Deposition 26.

27.

28.

29.

30.

31. 32.

33. 34. 35.

36. 37.

38.

39.

40.

299

S Farhat, C Findeling, F Silva, K Hassouni, A Gicquel. Role of the plasma composition at the surface on diamond growth. J Phys IV Fr 8(Pr-7):391–399, 1998. MC McMaster, WL Hsu, ME Coltrin, DS Dandy. Experimental measurements and numerical simulations of the gas composition in a hot-filament-assisted diamond chemical-vapor-deposition reactor. J Appl Phys 76:7567–7577, 1994. PK Bachmann, HJ Hagemann, H Lade, D Leers, F Picht, DU Wiechert. In: CH Carter, G Gildenblatt, S Nakamura, RJ Nemanich, eds. Diamond, SiC, and Nitride Wide Band Gas Semiconductors. MRS Symp Proc. Vol 339. Pittsburgh: Materials Research Society, 1994, p 267. K-P Kuo, J Asmussen. An experimental study of high pressure synthesis of diamond films using a microwave cavity plasma reactor. Diamond Relat Mater 6:1097–1105, 1997. A Gicquel, E Anger, MF Ravet, D Fabre, G Scatena, ZZ Wang. Diamond deposition in a bell-jar reactor: influence of the plasma and substrate parameters on the microstructure and growth rate. Diamond Relat Mater 2:417–424, 1993. SJ Harris, AM Weiner. Filament-assisted diamond growth kinetics. J Appl Phys 70:1385–1391, 1991. D Takeuchi, H Watanabe, S Yamanaka, H Okushi, K Kajimura. Defects in device grade homoepitaxial diamond thin films grown with ultra-low CH4 /H2 conditions by microwave-plasma chemical vapor deposition. Phys Status Solidi(a) 174:101–115, 1999. E Kondoh, T Ohta, T Mitomo, K Ohtsuka. Surface reaction kinetics of gasphase diamond growth. J Appl Phys 73:3041–3046, 1993. KA Snail, CM Marks. In situ diamond growth rate measurement using emission interferometry. Appl Phys Lett 60:3135–3137, 1992. MJ Ulczynski, DK Reinhard, M Prystajko, J Asmussen. Low temperature deposition of thin film diamond. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: Yoshikawa, 1994, pp 41–44. S Khatami. Controlled growth of microwave assisted plasma CVD of diamond. PhD dissertation, Michigan State University, 1997. V Ralchenko, I Sychov, I Vlasov, A Vlasov, V Konov, A Khomich, S Voronina. Quality of diamond wafers grown by microwave plasma CVD: effects of gas flow rate. Diamond Relat Mater 8:189–193, 1999. CJ Chu, RH Hauge, JL Margrave, MP D’Evelyn. Growth kinetics of (100), (110), and (111) homoepitaxial diamond films. Appl Phys Lett 61:1393–1395, 1992. H Maeda, K Ohtsubo, M Irie, N Ohya, K Kusakabe, S Morooka. Determination of diamond [100] and [111] growth rate and formation of highly oriented diamond film by microwave plasma-assisted chemical vapor deposition. J Mater Res 10:3115–3123, 1995. RE Clausing. Diamond morphology. In: MA Prelas, G Popovici, LK Bigelow, eds. Handbook of Industrial Diamonds and Diamond Films. New York: Marcel Dekker, 1998, pp 19–47.

300 41.

42.

43.

44.

45. 46. 47.

48. 49. 50.

51.

52.

53. 54.

55.

56.

Grotjohn and Asmussen C Wild, P Koidl, W Muller-Sebert, H Walcher, R Kohl, N Herres, R Locher. Chemical vapour deposition and characterization of smooth {100}-faceted diamond films. Diamond Relat Mater 2:158–168, 1993. C Wild, R Kohl, N Herres, W Muller-Sebert, P Koidl. Oriented CVD diamond films: twin formation, structure and morphology. Diamond Relat Mater 3:373– 381, 1994. X Jiang, K Schiffmann, C-P Klages, D Wittorf, CL Jia, K Urban, W Jager. Coalescence and overgrowth of diamond grains for improved heteroepitaxy on silicon (001). J Appl Phys 83:2511–2518, 1998. WA Weimer, RF Cario, CE Johnson. Examination of the chemistry involved in microwave plasma assisted chemical vapor deposition of diamonds. J Mater Res 6:2134–2144, 1991. DK Milne, PG Roberts, P John, MG Jubber, M Liehr, JIB Wilson. Diamond Relat Mater 4:395, 1995. AB Harker, JF DeNatale. J Mater Res 5:818, 1990. C Gu, Z Jin, C Wang, G Zou, Y Sakamoto, M Takaya. Growth of (100) orientation diamond film deposited by MWPCVD methods using gaseous mixtures of CH4 , CO and H2 . Diamond Relat Mater 7:765–768, 1998. R Locher, C Wild, N Herres, D Behr, P Koidl. Nitrogen stabilized 具100典 texture in chemical vapor deposited diamond films. Appl Phys Lett 65:34–36, 1994. S Jin, TD Moustakas. Effect of nitrogen on the growth of diamond films. Appl Phys Lett 65:403–405, 1994. C Wild, R Locher, P Koidl. Homoepitaxial growth of CVD diamond: effect of nitrogen contaminations on growth rates. Mat Res Soc Symp 416, pp. 75–80, 1996. W Muller-Sebert, E Wormer, F Fuchs, C Wild, P Koidl. Nitrogen induced increase of growth rate in chemical vapor deposition of diamond. Appl Phys Lett 68:759–760, 1996. J Asmussen, J Mossbrucker, S Khatami, WS Huang, B Wright, V Ayres. The effect of nitrogen on the growth, morphology, and crystalline quality of MPACVD diamond films. Diamond Relat Mater 8:220–225, 1999; VM Ayres, TR Bieler, MG Kanatzidis, J Spano, S Hagopian, H Balhareth, BF Wright, M Farhan, JA Majeed, D Spach, BL Wright, J Asmussen. The effect of nitrogen on competitive growth mechanisms of diamond thin films. Diamond Relat Mater 9:236–240, 2000. R Samlenski, C Haug, R Brenn, C Wild, R Locher, P Koidl. Appl Phys Lett 67:2798, 1995. Y-Z Wan, HS Shen, Z-M Zhang, X-C He. Ternary C-H-halogen phase diagram for CVD diamond. Mater Chem Phys 63:88–92, 2000 (also see references contained within). NJ Komplin, BJ Bai, CJ Chu, JL Margrave, RH Hauge. Proceedings of Third International Symposium on Diamond Materials, Honolulu, 1993. New York: Electrochemical Society, 1993, p 385. PK Bachmann, HJ Hagemann, H Lade, D Leers, F Picht, DU Wiechert. Proceedings of Advanced Materials ’94, International Symposium on Advanced Materials. Tsukuba, Japan: NIRIM, 1994, pp 115–120.

Microwave Plasma-Assisted Diamond Film Deposition 57.

58.

59. 60.

61.

62. 63.

64.

65.

66.

67.

68.

69.

70. 71.

72.

301

T Hosomi, T Maki, T Kobayashi, Y Yoshizako, M Taniguchi, M Sugiyo. Role of xenon additive in microwave plasma-assisted (H2 ⫹ CH4) chemical vapor deposition of diamond thin film. J Appl Phys 84:6059–6063, 1998. D Zhou, DM Gruen, LC Qin, TG McCauley, AR Krauss. Control of diamond film microstructure by Ar additions to CH4 /H2 microwave plasmas. J Appl Phys 84:1981–1989, 1998. D Gruen. Nucleation, growth, and microstructure of nanocrystalline diamond films. MRS Bull 23:32–35, 1998. T Mitomo, T Ohta, E Kondoh, K Ohtsuka. An investigation of product distributions in microwave plasma for diamond growth. J Appl Phys 70:4532–4539, 1991. WL Hsu. Gas-phase kinetics during microwave plasma-assisted diamond deposition: Is the hydrocarbon product distribution dictated by neutral-neutral interactions? J Appl Phys 72:3102–3109, 1992. CJ Erickson, WB Jameson, J Watts-Cain, JL Menningen, MA Childs, LW Anderson, JE Lawler. Plasma Sources Sci Technol 5:761–764, 1996. MJ Wouters, J Khachan, IS Falconer, BW James. Production and loss of H atoms in a microwave discharge in H2 . J Phys D Appl Phys 31:2004–2012, 1998. A Gicquel, M Chenevier, M Lefebvre. Spatially resolved spectroscopic analysis of the plasma. In: MA Prelas, G Popovici, LK Bigelow, eds. Handbook of Industrial Diamond and Diamond Films. New York: Marcel Dekker, 1998, pp 739–796. A Gicquel, M Chenevier, Y Breton, M Petiau, JP Booth, K Hassouoni. Ground state and excited state H-atom temperatures in a microwave plasma diamond deposition reactor. J Phys III Fr 6:1167–1180, 1996. A Gicquel, K Hassouni, Y Breton, M Chenevier, JC Cubertafon. Gas temperature measurements by laser spectroscopic techniques and by optical emission spectroscopy. Diamond Relat Mater 5:366–372, 1996. TA Grotjohn, J Asmussen, J Sivagnaname, D Story, AL Vikharev, A Gorbachev, A Kolysko. Millimeter wave Fabry-Perot resonator plasma diagnostic measurements of electron density in moderate pressure diamond deposition discharges. Diamond Relat Mater 9:322–327, 2000. H Tahara, K Minami, A Murai, T Yasui, T Yoshikawa. Diagnostic experiment and kinetic model analysis of microwave CH4 /H2 plasmas for deposition of diamondlike carbon films. Jpn J Appl Phys 34:1972–1979, 1995. W Tan, TA Grotjohn. Modeling the electromagnetic field and plasma discharge in a microwave plasma diamond deposition reactor. Diamond Relat Mater 4: 1145–1154, 1995. T Grotjohn. Modeling electromagnetic fields for the excitation of microwave discharges used for materials processing. J Phys IV Fr 8:61–79, 1998. CR Koemtzopoulos, DJ Economou, R Pollard. Hydrogen dissociation in a microwave discharge for diamond deposition. Diamond Relat Mater 2:25–35, 1993. E Hyman, K Tsang, A Drobot, B Lane, J Casey, R Post. One-point numerical

302

73. 74.

75.

76.

77.

78.

79.

80.

81.

82. 83.

84.

Grotjohn and Asmussen modeling of microwave plasma chemical vapor deposition diamond deposition reactors. J Vac Sci Technol A 12:1474–1479, 1994. W Tan, TA Grotjohn. Modeling the electromagnetic excitation of a microwave cavity plasma reactor. J Vac Sci Technol A-12:1216–1220, 1994. K Hassouni, O Leroy, S Farhat, A Gicquel. Modeling of H2 and H2 /CH4 moderate-pressure microwave plasma used for diamond deposition. Plasma Chem Plasma Proc 18:325–362, 1998. K Hassouni, CD Scott, S Farhat, A Gicquel, M Capitelli. Non-maxwellian effect on species and energy transport in moderate pressure H2 plasmas. Surf Coat Technol 97:391–403, 1997. K Hassouni, S Farhat, CD Scott, A Gicquel. Modeling species and energy transport in moderate pressure diamond deposition H2 plasmas. J Phys III Fr 6:1229–1243, 1996. M Capitelli, G Colonna, K Hassouni, A Gicquel. Electron energy distribution functions and rate and transport coefficients of H2 /H/CH4 reactive plasmas for diamond film deposition. Plasma Chem Plasma Process 16:153–171, 1996. A Gicquel, K Hassouni, S Farhat, Y Breton, CD Scott, M Lefebvre, M Pealat. Spectroscopic analysis and chemical kinetics modeling of a diamond deposition plasma reactor. Diamond Relat Mater 3:581–586, 1994. K Hassouni, CD Scott, S Farhat. Flow modeling for a plasma assisted diamond deposition reactor. In: MA Prelas, G Popovici, LK Bigelow, eds. Handbook of Industrial Diamond and Diamond Films. New York: Marcel Dekker, 1998, pp 653–672. K Hassouni, CD Scott, S Farhat. Modeling the diffusional transport of an H2 plasma obtained under diamond deposition discharge conditions. In: Handbook of Industrial Diamond and Diamond Films. New York: Marcel Dekker, 1998, pp 697–738. K Hassouni, TA Grotjohn, A Gicquel. Self-consistent microwave field and plasma discharge simulations for a moderate pressure hydrogen discharge reactor. J Appl Phys 86:134–151, 1999. M Funer, C Wild, P Koidl. Numerical simulations of microwave plasma reactors for diamond CVD. Surf Coat Technol 74–75:221–226, 1995. M Funer, C Wild, P Koidl. Simulation and development of optimized microwave plasma reactors for diamond deposition. Surf Coat Technol 116–119: 853–862, 1999. A Argoitia, CS Kovach, JC Angus. Hot-filament CVD method. In: Handbook of Industrial Diamond and Diamond Films. New York: Marcel Dekker, 1998, pp 797–819.

8 Combustion Synthesis of Diamond Colin A. Wolden Colorado School of Mines, Golden, Colorado

I.

INTRODUCTION

Combustion synthesis is one of the competing chemical vapor deposition (CVD) technologies for diamond film growth. It was first discovered in 1988 by Hirose and Kondo [1], who realized that acetylene flames produce copious amounts of atomic hydrogen and hydrocarbon radicals that are required for diamond growth. The discovery was subsequently confirmed at the Naval Research Laboratory [2] and has become the subject of intense study. It has been argued that combustion synthesis is the most flexible of the CVD alternatives because of its scalable nature, minimal utility requirements, and significantly reduced capital costs relative to plasma-aided processes [3]. In this chapter the development of combustion CVD will be reviewed from its inception using conventional welding torches to its present implementation with flat flame burners. The key experimental parameters and their impact on deposition are discussed in detail. The chemistry of combustion diamond CVD is discussed, highlighting studies that employed in situ diagnostics and detailed reactor modeling. Finally, combustion synthesis is contrasted with other CVD techniques in terms of both processing and economics.

II.

INVENTION: THE WELDING TORCH

A.

Description of System

Figure 1 shows the structure of an atmospheric oxyacetylene torch in its unperturbed state and in the configuration used for diamond deposition. 303

304

Wolden

Figure 1 Schematic diagrams of (a) the flame structure of an oxyacetylene torch and (b) the flame in a diamond deposition configuration.

There are three distinct regions in an acetylene flame: (1) the inner flame, (2) the acetylene feather, and (3) an outer diffusion flame. It is in the feather region that the substrate is placed for diamond growth. The most important parameter in combustion synthesis is the oxygen-to-acetylene ratio, defined here as R = O2 :C2H2 . At values of R near 1.0, a neutral flame is achieved, which is defined as the condition where the feather region just disappears because all the acetylene is consumed in the primary flame. The diamond growth regimes as a function of composition are outlined in Fig. 2. The highest quality diamond is obtained in slightly rich flames, R = 0.85–1.0 [4]. The value of R at which a neutral flame occurs is dependent on both burner design and total flow rate. Thus, for a particular experimental configuration the composition at the neutral flame condition may be shifted slightly, but the principles described in Fig. 2 still hold. The welding torch arrangement of Hirose has been used in numerous subsequent studies [5–13]. Polycrystalline diamond has been grown at rates up to 200 ␮m/hr [5] and thick (150 ␮m) homoepitaxial layers have been deposited [6]. Atmospheric torches have successfully produced large indi-

Combustion Synthesis of Diamond

305

Figure 2 Carbon deposition regions observed using atmospheric acetylene torches as a function of gas composition. (From Ref. 4.)

vidual crystals that approach 1 mm in diameter [7,8]. The drawback of the welding torch arrangement is that the deposited material is characterized by radial inhomogeneity [9] and often contains significant amounts of nondiamond carbon [10]. Because of these limitations, welding torches are being replaced by flat flame burners, which are more appropriate for large area deposition. Nevertheless, the wealth of data from experiments with atmospheric torches provided the foundation for the general understanding of combustion grown diamond that is described in the following. A more comprehensive summary of experiments performed using atmospheric torches can be found in an earlier review [11].

B.

General Considerations

1.

Chemistry

The composition of acetylene flames is much different from the traditional H2 /CH4 mixtures used in hot-filament and plasma CVD systems. However, the chemistry and the mechanism of diamond growth in combustion CVD

306

Wolden

are expected to be similar to those of other diamond growth systems. First, at R = 1.0 the acetylene flame is positioned in the direct center of the empirical growth regime diagram developed by Bachmann et al. [12] (see Chapter 4). The detailed chemistry of acetylene flames is discussed in detail in Sec. IV, but its essence may be distilled to the following two reactions: C2H2 ⫹ O2 → 2CO ⫹ H2 H2 ↔ H ⫹ H

⌬H⬚R = ⫺107 kcal/mol

(1)

Reversible: function (T, P)

(2)

First, nearly all of the oxygen and acetylene is rapidly consumed, producing carbon monoxide and hydrogen as described by the global reaction (1) [13]. The highly exothermic nature of this reaction results in flame temperatures >3000 K. At these temperatures a significant fraction of hydrogen molecules are converted into atomic hydrogen as expressed by global reaction (2) [14]. The degree of dissociation is strongly dependent on both temperature and pressure. Carbon monoxide is thermally stable and not expected to affect the growth processes under these conditions. A small fraction of the acetylene is converted into hydrocarbon radical species or remains unreacted. Thus, growth schemes discussed in Chapter 4 that suggest that diamond growth proceeds through the addition of hydrocarbon radicals in the presence of a superequilibrium concentration of atomic hydrogen may also be applied to combustion systems. Where filament and plasma systems employ electrical energy to create atomic hydrogen, combustion systems rely on the chemical energy stored in the acetylene molecule to generate heat and atomic hydrogen. 2.

Substrate Temperature

Substrate temperatures during the combustion CVD of diamond range from 950 to 1650 K. Because of these high temperatures, substrates have been limited to materials such as molybdenum, alumina, silicon, and diamond. Evaluating the true substrate temperature is difficult in all diamond CVD environments but especially so in combustion synthesis due to the extreme heat fluxes present. The two techniques most commonly used are optical pyrometry and thermocouples placed on the back side of the substrate. Pyrometry provides a good measure of the relative temperature of the substrate, but absolute values may be off by as much as 100 K due to uncertainties in emissivities. Care must be taken to position the pyrometer in order to avoid emission from the flame itself. Back-side thermocouples are affected by both the quality of the contact and the steep thermal gradients present. Both techniques work well for a single configuration, but caution is urged when comparing absolute temperatures obtained from different systems.

Combustion Synthesis of Diamond

307

Substrate temperature has a dramatic effect on two important properties: growth rate and morphology. Several authors [15–17] have observed that the growth rate increases exponentially with temperature as shown schematically in Fig. 3. The behavior shown by the curve in Fig. 3 is characteristic of combustion systems; however, its location may be shifted to the right or left depending on the operating conditions and the technique used to measure temperature. Weimer et al. [15] measured activation energies between 12 and 18 kcal/mol for diamond growth on individual crystal faces. Slightly higher values (16–23 kcal/mol) have been reported for polycrystalline films [16,17]. These values are in good accord with activation energies measured in hot-filament systems, indicating that the growth mechanisms may be similar [18,19]. As the substrate temperature increases, the growth rate reaches a maximum (Fig. 3). Increasing the substrate temperature beyond this point causes a rapid decline in both the quality and the growth rate [15–17]. In atmospheric torches the maximum growth rate occurs at substrate temperatures between 1450 and 1650 K [15,20]. In lowpressure (40–50 torr) flat flame burners the same growth rate behavior has been observed, but the maximum occurs at significantly reduced temperatures of 1050–1250 K [17,21]. Kim and Cappelli [21] suggested that the falloff at high temperature may be due to oxidation of diamond. An alter-

Figure 3 The effect of substrate temperature on growth rate observed in combustion CVD of diamond. (From Refs. 15–17, 20.)

308

Wolden

native possibility is that the growth precursor desorbs from the surface before having the opportunity to incorporate into the diamond lattice [22]. Substrate temperature also determines the diamond film morphology. The morphology of most films deposited by combustion CVD are typically dominated by {111} facets [20,21,23]. However, a number of researchers [17,20,23,24] have observed that at higher temperatures the morphology consists of {100} facets parallel to the substrate. As temperature increases, the growth habit changes from octahedron to cubic as shown in Fig. 4. The temperature where {100} faceting occurs was near the temperature where maximum growth rates were observed (Fig. 3) [17,20]. 3.

Reactant Composition

In atmospheric torches it has been observed that morphology also depends on reactant composition [25,26]. Decreasing the oxygen fraction has an effect similar to that of increasing substrate temperature as shown in Fig. 4. It was observed that the morphology changed from octahedron growth under oxygen-rich conditions to cubic growth as the acetylene content was increased [25,26]. The dependence on composition is in accord with observations of morphology variations with methane concentration in plasma reactors [27,28]. For flat flame burners this dependence on composition has not been reported, perhaps because the range of R where diamond is successfully formed is much smaller [17,21]. Of the two variables, morphology depends more strongly on substrate temperature than composition. C.

Modifications of the Welding Torch

In the welding torch operation shown in Fig. 1, diamond is often deposited in a small circle or in an annular ring. To increase the deposition area a

Figure 4 Schematic diagram showing changes in crystal growth habit as a function of both substrate temperature and reactant composition. (From Refs. 17, 20, 23–26.)

Combustion Synthesis of Diamond

309

number of modifications of the welding torch have been investigated, including turbulent flame operation, multinozzle burners, and enclosed flames. A few groups [29,30] have investigated operating the welding torch under turbulent flow conditions. The turbulent condition was achieved by increasing both the flow rate and the burner orifice. Despite some reports [31] of improved quality, it was generally concluded that the changes were not dramatic and not worth the costs associated with increased flow rates [29,30,32]. Some success was achieved with a multinozzle configuration was developed by Zhu and coworkers [33], although the morphology was not uniform over the deposition area. During atmospheric operation, diffusion of oxygen or impurities from the ambient may effect film quality. Attempts to minimize the effect by surrounding the torch with a coannular flow of inert gas have been studied [34,35]. Under these conditions the outer flame (Fig. 1) was suppressed, but no dramatic changes were observed in quality or growth rate. Another way to exclude the ambient is simply to enclose the torch in a vacuum chamber [36–39]. It was observed that the outer diffusion flame disappeared as there was no oxygen present and the deposition area was increased by 20% [36,37]. In addition, reducing the pressure to 300 torr increased the deposition area by 200%, but the growth rate decreased significantly. Morrison et al. [38] performed a systematic study of an enclosed flame at 700 torr and found that a low-temperature pretreatment step enhanced the nucleation density. Wang et al. [39] achieved some success using rastering techniques with an enclosed flame to coat larger areas.

III.

PRACTICAL IMPLEMENTATION: THE FLAT FLAME BURNER

A.

The Flat Flame Burner

1.

Burner Design

A number of groups [14,17,21,40,41–45] have addressed the issues of increasing the deposition area and improving uniformity through the use of flat flame reactors operated at both atmospheric and reduced pressure. Cooper and Yarbrough [40] first demonstrated diamond deposition in a commercially available burner at pressures between 25 and 40 torr. A schematic of a flat flame burner CVD for reduced pressure operation is shown in Fig. 5. Premixed gas enters a water-cooled burner and exits through a matrix that creates a radial uniform velocity profile with only an axial component. In practice this has been achieved by drilling 1-mm holes in a copper plate [21,41], and by using a honeycomb made of thin-walled silica tubes

310

Figure 5 pressures.

Wolden

Schematic of a flat flame burner CVD system operated at reduced

[17]. A flat, circular flame is formed stabilized below the burner surface as shown in Fig. 5. This is a very different structure than the conical flame that is associated with atmospheric torches (Fig. 1). The substrate is maintained at a certain temperature T and positioned at a distance L from the burner surface. This design was developed in the 1950s to measure flame velocities near extinction/ignition limits [46,47]. The burner’s characteristics of stability and geometrical simplicity that made it useful for that purpose are equally valuable for chemical vapor deposition. Atmospheric pressure flat flames are less stable and require more complicated burner design. Two examples that have been used include a coflow arrangement and a trumpet-bell design that are shown schematically in Fig. 6 [42]. Murayama and coworkers [43,44] used the coflow design with hydrogen as the external gas to stabilize their flame. Workers at Sandia have demonstrated the scalability of the trumpet-bell design [42,45]. Comparison of Figs. 5 and 6 shows another important difference between atmospheric and low-pressure operation. At low pressures the flame is stabilized on the burner surface, whereas atmospheric flames are substratestabilized flames located within a millimeter or two of the substrate. In lowpressure operation the substrate may be positioned at any distance from the burner, whereas in atmospheric systems the distance is constrained by the stability limits. In addition, in large-area atmospheric flames it was found that it was necessary to surround the substrate with equally spaced pillars to stabilize the flame [48].

Combustion Synthesis of Diamond

311

Figure 6 Two designs of atmospheric flat flame burners: (a) a coflow design and (b) a trumpet bell design. (From Ref. 48.)

2.

Radial Uniformity and Scalability

Regardless of operating pressure, the unique feature of this design is that films are produced that are uniform in both thickness and morphology. Figure 7 shows micrographs of surface morphology and a cross-sectional profile that demonstrate the radial uniformity achieved in flat flame burners. At low pressures, Glumac and Goodwin [41] demonstrated the scaleable nature of the system, producing uniform films over areas as large as 13 cm2. More recently, Hahn et al. [47] scaled up an atmospheric deposition system to 20 cm2. Another feature of the flat flame CVD reactor is that it is amenable to detailed kinetic modeling. Under proper operating conditions [49], the axisymmetric stagnation flow may be treated in one dimension using a similarity solution [50]. A number of groups have performed detailed kinetic modeling on these systems, which has provided insight into the deposition mechanisms. These studies are discussed in more detail in Sec. IV.

B.

Operating Conditions

1.

Reactant Composition

Low-pressure flat flame burners are operated with a value of R very near unity [17,21]. At atmospheric pressure, pure oxygen-acetylene flames are unstable. It has been found that the addition of hydrogen (18–25% by volume) while maintaining a C/O ratio ⬃1 was required to stabilize the flame and achieve diamond growth [42–45]. Because of the homogeneous nature

312

Wolden

Figure 7 Scanning electron micrographs (a–c) and cross-section profile (d) showing the radial uniformity of both morphology and growth rate achieved in flat flame burners. (From Ref. 17.)

of these flames, the composition range where diamond is deposited is much smaller than observed with the welding torch design [4]. Most researchers [42–45] report only a single composition that successfully produced highquality diamond. In low-pressure systems there is a small range where diamond is deposited [21,51]. The reason for the narrow range is shown in Fig. 8, where the equilibrium mole fractions of selected species are plotted as a function of composition. The three major species (CO, H2 , and H) are nearly constant across this range of R. For the minor species a very sharp transition is observed between R = 0.98 and R = 1.02. The concentration of carbon-containing species (C2H2 , CH3 , C) falls between four and seven orders of magnitude over this narrow regime. A concomitant increase in oxidation products (H2O, CO2 , O2) is also observed. Certainly, very good control of mass flow controllers is required for reproducible results using combustion synthesis. 2.

Operating Pressure

As already discussed, operating pressure has a major influence on burner design and flame stability. The pressure also influences a number of critical

Combustion Synthesis of Diamond

313

Figure 8 Equilibrium mole fractions of selected species in an oxyacetylene flame as a function of reactant composition R = O2 :C2H2 . Calculation performed at 40 torr with 5% argon.

issues including growth rate, temperature control, and deposition temperature. First, growth rates in atmospheric systems range from 25 to 40 ␮m/hr [44,45]. In comparison, growth rates in flat flame burners at ⬃45 torr have been reported between 4 and 5.5 ␮m/hr [17,21]. Although the gas density is about 15 times greater in atmospheric reactors, the growth rate is only a factor of ⬃7 times greater. More important with regard to economics, the carbon conversion efficiency (the fraction of acetylene atoms converted to diamond) is very similar in both cases. As discussed in Sec. II, growth temperatures in atmospheric systems are hundreds of degrees hotter than at low pressure. Substrate material selection is more limited for atmospheric flames, and these films are subject to the formation of large compressive stresses during cooling due to differing coefficients of thermal expansion [49]. Robust substrate temperature control is requisite for both controlling growth rate and achieving textured growth [17]. The proximity of the flame and substrate in atmospheric reactors re-

314

Wolden

Table 1 Comparison of Operating Flat Flames at Atmospheric and Reduced (40–50 torr) Pressures Aspect

Reduced pressure

Atmospheric pressure

Growth rates Carbon conversion efficiency Deposition temperatures Temperature control Flame stability Safety Vacuum chamber required Burner design

4–5.5 ␮m/hr Similar ⬃0.005% 400–1000⬚C Easy Yes Advantage Yes Simple

25–40 ␮m/hr Similar ⬃0.005% 700–1200⬚C Difficult No: requires H2 addition No Complex

Source: Refs. 17, 21, 42–45, 51, 52.

quires that energy must be removed at fluxes on the order of 500 W/cm2, requiring the use of elaborate cooling strategies [20,45,48]. In contrast, lowpressure reactors are nearly thermally neutral, and films have been deposited without any cooling present [41]. Simple control strategies can maintain the substrate temperature within a few degrees of a desired setpoint for hours of deposition [52]. Although atmospheric burner design is more complicated, an advantage is that no vacuum chamber is required. Table 1 summarizes some of the pros and cons of operating pressure in diamond deposition. C.

Applications

Combustion CVD was invented 7 years after the discovery of hot-filament and plasma techniques, and subsequently the majority of applied research has been performed in these conventional reactors. However, the studies that have been conducted demonstrate that combustion-grown diamond is comparable to material produced in other systems. Some of these examples are presented in this section with the intention of demonstrating the flexible nature of combustion CVD. 1.

Textured Film Growth

For a number of applications [53,54] films that demonstrate {100} texturing have several advantages. The combination of {100} growth conditions [27] and bias-enhanced nucleation techniques [55] has yielded highly oriented diamond films. Both the number and angle of grain boundaries are significantly reduced in these films relative to randomly oriented films [28,52]. Both the thermal [56] and the electronic [53] properties of diamond are

Combustion Synthesis of Diamond

315

adversely effected by the presence of grain boundaries. For example, the measured carrier mobility was three times greater in {100} oriented films than polycrystalline diamond [53]. In addition, {100} growth provides a means of planarizing the growth surface [52], enabling the use of conventional silicon technology to manufacture device structures without expensive grinding and polishing steps [54]. Figure 9 shows a scanning electron micrograph of a highly oriented {100} film that was produced by a two-step process. First, oriented nucleation was achieved using bias-enhanced nucleation techniques in a conventional microwave plasma reactor. The substrate was transferred to a low-pressure combustion reactor and grown for 3 hr [52]. In Fig. 9 one can see that a number of grains have coalesced, forming a very flat surface. 2.

Doping and Device Fabrication

Ravi and coworkers [57] demonstrated the high purity of combustion-grown diamond, producing films that had resistivities two orders of magnitude greater than natural diamond. The doping of diamond imparts semiconductor properties that are required for a number of applications [53]. Most of the limited work on doping of combustion-grown diamond has been carried out at the Naval Research Laboratory [58–61]. Glesener and coworkers [58]

Figure 9 Scanning electron micrograph of a highly oriented {100} film formed by a combination of bias-enhanced nucleation in a microwave reactor and textured growth in a low-pressure combustion CVD system. (From Ref. 52.)

316

Wolden

fabricated Schottky diodes by depositing boron-doped diamond on molybdenum. Further studies [59,60] indicated that high surface resistance was critical to form good junctions. The boron was added by bubbling part of the acetylene through a solution of diboric trioxide in acetone. Doverspike and coworkers [61] used an atomizer to entrain aerosol droplets that contained dopant species into the gas flow. Boric acid dissolved in methanol was used as a dopant solution. Recently, boron incorporation has been achieved without the use of a solvent. Trimethylborate is a high-vapor-pressure liquid that may be introduced by mass flow controllers at low pressure or with a bubbler for atmospheric operation [14]. 3.

Low-Temperature Deposition

Low-temperature processing is required for compatibility with a number of optoelectronic materials. In addition, low-temperature deposition minimizes the compressive stress that is formed when diamond is grown on material with differing coefficients of thermal expansion. Diamond deposition at rates of ⬃1 ␮m/hr on glass substrates was achieved at low temperature (750 K) in a low-pressure combustion reactor [51]. The low temperature was achieved by increasing the burner-substrate distance to 20 mm. Continuous, adherent films were deposited on Vycor and Pyrex glasses. The best films were optically transparent and no damage to the substrates was observed.

IV.

DEPOSITION CHEMISTRY

A.

Diagnostics

The chemistry of combustion diamond CVD has been primarily investigated using laser-induced fluorescence (LIF) and mass spectroscopy. The two techniques are complementary. Mass spectroscopy has been used to provide quantitative measurements of the major stable species, which are important for verifying kinetic models. In this technique a quartz microprobe samples gases at the diamond growth surface. Using mass spectroscopy, Wolden et al. [62] noted that the strong compositional dependence observed in a flat flame reactor was related to a sharp change in the acetylene concentration. On the other hand, LIF is a nonintrusive optical technique that has been used to detect radicals such as C2 , C2H, CH, and OH in the gas phase above the substrate. Researchers using plasma and hot-filament reactors have focused on C1 species such as methyl radical and atomic carbon as the dominant growth species [63,64]. However, evidence from combustion reactors suggests that C2 species are also very important. Three different groups [13,65,66] have noted a strong relation between C2 concentration profiles

Combustion Synthesis of Diamond

317

and diamond deposition rates. The role of CH is less clear, and OH appears to be of minor importance [65]. The work of Cappelli and Paul [65] suggested that C2H also plays an important role in the deposition process. B.

Reactor Modeling

An attractive feature of flat flame reactors is that they may be simulated using simple stagnation flow models with complex combustion kinetic mechanisms. Several workers have used this approach to model flat flames at diamond growth conditions [66–68]. Kim and Cappelli [67] argued that optimum diamond growth conditions result from a competition between hydrocarbon growth and oxidation. Simulations at Sandia indicated the importance of operating at high flame speeds to increase atomic hydrogen production [68]. A detailed kinetic mechanism containing 50 species and 218 reactions has been developed for the acetylene flame by Miller and Melius [69]. The model has been verified experimentally by Glumac and Goodwin [70] using mass spectroscopy and LIF at conditions slightly oxygen rich for diamond growth. Wolden and coworkers [14] performed a similar validation study at diamond growth conditions with a diamond substrate present. Figure 10 shows the profiles of selected species between the burner and substrate at diamond deposition conditions [14]. Acetylene and oxygen are rapidly converted to hydrogen and carbon monoxide as the gas exits the burner. At the high temperatures involved, a large amount of atomic hydrogen is produced. The concentration of H decreases near the surface due to recombination reactions with the diamond surface [71,72]. As noted in Sec. III, there is a very narrow composition window where diamond growth is achieved. Figure 11 shows the normalized concentration of selected species from model predictions plotted over a range of composition that spans the transition from graphite deposition to diamond deposition through etching [66]. The two species most often implicated in diamond growth, H and CH3 , hardly changed over this range. In contrast, the acetylene concentration decreased by 70%, which was the largest change observed for any species [66]. Although the mechanism is not fully understood, acetylene appears to be responsible for the sharp transition between diamond and graphite deposition.

C.

Alternative Fuels

Although acetylene is the most common fuel used for combustion synthesis, a number of alternatives have been examined including hydrogen [73], eth-

318

Wolden

Figure 10 Profiles of selected species between the burner (11 mm) and the substrate (0 mm) at diamond growth conditions: R = 1.00, P = 47 torr, Ts = 750⬚C, Uo = 600 cm/sec. (From Ref. 14.)

ylene [74], MAPP gas [75],* and methane [76]. The interest in other fuels is economic, due to the relatively high cost of acetylene. Diamond has been deposited successfully in each case, but the growth rates are significantly less than with acetylene. The superiority of acetylene is due to its high flame velocity and temperature. Table 2 compares these properties for a number of fuels. Examining Table 2, one finds that the flame temperature of acetylene is at least 350⬚C hotter than for the alternatives. All of these fuels produce hydrogen, but only atomic hydrogen is useful for diamond deposition. The importance of acetylene’s high flame temperature is shown in Fig. 12, which shows the dependence of hydrogen dissociation on both temperature and pressure. First, the temperature dependence is exponential. Thus, increasing the temperature from 2500 to 2900 K results in a fourfold increase in the H atom mole fraction. Besides growth rate considerations, modeling efforts

*MAPP gas is a commercially available mixture of about half liquefied petroleum gas (mostly propylene) and half C3H4 (a mixture of the two isomers methyl acetylene and propadiene).

Combustion Synthesis of Diamond

319

Figure 11 Variations of selected species as a function of reactant composition R = O2 :C2H2 over a range that spans the transition from graphite deposition to diamond-deposition to etching. (From Ref. 66.)

Table 2 Comparison of Adiabatic Flame Temperatures and Flame Speeds of Selected Fuels Fuel Acetylene, C2H2 Ethylene, C2H4 Hydrogen, H2 Propylene, C3H6 Methane, CH4

Adiabatic flame temperature (K)*

Flame speed (cm/sec)

2910 2565 2525 2505 2325

141.0 68.3 264.8 43.8 33.8

*Assumes base temperature = 25⬚C; oxidizer, air; equivalence ratio, 1.0.

320

Wolden

Figure 12 Equilibrium dissociation of atomic hydrogen as a function of temperature and pressure.

suggest that diamond quality increases quadratically with atomic hydrogen concentration [77]. Figure 12 also indicates that H-atom dissociation is enhanced at lower pressures in accordance with Le Chaˆtelier’s principle.

V.

COMPARISONS AND CONCLUSIONS

Although material deposited by combustion CVD has not been tested extensively, there is every indication that the quality is on par with that of diamond produced in hot-filament and plasma systems. Schottky diodes and high-energy particle detectors have been successfully fabricated using combustion-grown diamond. Issues of large-area deposition and uniformity have been addressed with flat flame burners. Growth rates are exceeded only by atmospheric plasma torches. Important features such as textured growth, boron doping, and low-temperature deposition have all been achieved in combustion reactors. Oriented nucleation, which has been achieved by substrate biasing in microwave and hot-filament reactors, has yet to be demonstrated in a combustion CVD system.

Combustion Synthesis of Diamond

321

Early reports concluded that combustion CVD was a much more expensive synthesis technique than plasma or hot-filament methods [78]. However, more recent analyses using data from a low-pressure flat flame burner indicates that the costs are very similar to those of microwave plasma CVD (M. Stonefield and C.A. Wolden, unpublished work). In comparison with plasma-aided processed, combustion CVD requires much less capital but has higher consumable costs because of the amounts of acetylene and oxygen required. In conclusion, the status of combustion synthesis of diamond using acetylene flames has been reviewed. The development of the technique from its inception using conventional welding torches to its present implementation using flat flame burners has been detailed. High growth rates and uniformity are characteristic of diamond deposition in flat flame burners. The design may be scaled to arbitrarily large areas. The quality of combustion diamond is comparable to that of material produced by other techniques. Economically, combustion CVD is competitive with plasma techniques. Combustion CVD requires less capital investment, but these savings are offset by high material costs. Substantial savings will be realized by improvements in growth rates and carbon utilization efficiency.

ACKNOWLEDGMENTS The author acknowledges the financial support of the National Research Council and Army Research Office during the writing of the manuscript. The author would like to thank Dr. John Prater of the Army Research Office for many helpful discussions and for stimulating my work in combustion CVD of diamond. Collaborations with the research groups of Professors Zlatko Sitar and Robert Davis in the Materials Science and Engineering Department at North Carolina State University were greatly appreciated.

REFERENCES 1. 2. 3. 4. 5. 6. 7.

Y Hirose, N Kondo. Program and Book of Abstracts, Japan Applied Physics 1988 Spring Meeting. Tokyo: Japanese Physical Society, 1988, p 434. LM Hanssen, WA Carrington, JE Butler, KA Snail. Mater Lett 7:289, 1988. KV Ravi. Diamond Relat Mater 4:243, 1995. Y Hirose, S Amanuma, K Komaki. J Appl Phys 68:6401, 1990. KA Snail, LM Hanssen. J Cryst Growth 112:651, 1991. G Janssen, WJP van Enckevort, JJD Schaminee, W Vollenberg, LJ Giling, M Seal. J Cryst Growth 104:752, 1990. T Abe, M Suemitsu, N Miyamoto, N Sato. J Appl Phys 73:971, 1993.

322 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42.

Wolden XH Wang, W Zhu, J von Windheim, JT Glass. J Cryst Growth 129:45, 1993. DB Oakes, JE Butler, KA Snail, WA Carrington, LM Hanssen. J Appl Phys 69:2602, 1991. LM Hanssen, KA Snail, WA Carrington, JE Butler, S Kellog, DB Oakes. Thin Solid Films 196:271, 1991. PW Morrsion Jr, JT Glass. In: G Davies, ed. Properties and Growth of Diamond. London: INSPEC, 1994, p 380. PK Bachmann, D Leers, H Lydtin. Diamond Relat Mater 1:1, 1991. Y Matsui, A Yukki, M Sahara, Y Hirose. Jpn J Appl Phys 28:1718, 1989. CA Wolden, RF Davis, Z Sitar, JT Prater. Diamond Relat Mater 7:133, 1998. RA Weimer, TP Thorpe, KA Snail. J Appl Phys 77:641, 1995. KA Snail, CM Marks. Appl Phys Lett 60:3135, 1992. CA Wolden, Z Sitar, RF Davis, JT Prater. Appl Phys Lett 69:2258, 1996. E Kondoh, T Ohta, T Mitomo, K Ohtsuka. Appl Phys Lett 59:488, 1991. C Chu, M D’Evelyn, R Hauge, J Margrave. J Appl Phys 70:1695, 1991. JJ Schermer, JEM Hogenkamp, GCJ Otter, G Janssen, WJP van Enckevort, LJ Giling. Diamond Relat Mater 2:1149, 1993. JS Kim, MA Cappelli. J Mater Res 10:149, 1995. CA Wolden, KK Gleason. Diamond Relat Mater 5:1503, 1996. GH Ma, Y Hirose, S Amanuma, M McClure, JT Prater, JT Glass. In: R Messier, JT Glass, JE Butler, R Roy, eds. New Diamond Science and Technology. Pittsburgh: Materials Research Society, 1991, p 587. KV Ravi, A Joshi. Appl Phys Lett 58:246, 1991. KV Ravi, CA Koch, HS Hu, A Joshi. J Mater Res 5:2356, 1990. K Hirabayashi, Y Hirose. Diamond Relat Mater 5:48, 1996. C Wild, R Kohl, N Herres, W Muller-Sebert, P Koidl. Diamond Relat Mat 3: 373, 1994. BR Stoner, SR Sahida, JP Bade, P Southworth, PJ Ellis. J Mater Res 8:1334, 1993. KA Snail, RG Vardiman, JP Estrera, JW Glesener, C Merzbacher, CJ Craigie, CM Marks, R Glosser, JA Freitas Jr. J Appl Phys 74:7561, 1993. JJ Schermer, LJ Giling, P Alers. J Appl Phys 78:2376, 1995. KA Snail, CL Vold, CM Marks, JA Freitas Jr. Diamond and Relat Mater 1: 180, 1992. JJ Scheimer, WA Elst, LJ Giling. Diamond and Relat Mater 4:1113, 1995. W Zhu, BH Tan, J Ahn, HS Tan. J Mater Res 30:2130, 1995. RC Aldredge, DG Goodwin. J Mater Res 9:80, 1994. T Abe, M Suemitsu, N Miyamoto. J Cryst Growth 143:206, 1994. M Murakawa, S Takeuchi. Surf Coat Technol 54/55:403, 1992. K Komaki, M Yanagisawa, I Yamamoto, Y Hirose. Jpn J Appl Phys 32:1814, 1993. PW Morrsion, A Somashekhar, JT Glass, JT Prater. J Appl Phys 78:4144, 1995. DY Wang, YH Song, JJ Wang, RY Cheng. Diamond Relat Mater 2:304, 1993. JA Cooper Jr, WA Yarbrough. Diamond Opt SPIE 1325:41, 1990. NG Glumac, DG Goodwin. Mater Lett 18:119, 1993. KF McCarty, E Meeks, RJ Kee, AE Lutz. Appl Phys Lett 63:1498, 1993.

Combustion Synthesis of Diamond 43. 44. 45. 46. 47. 48. 49. 50.

51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78.

323

M Murayama, S Kojima, K Uchida. J Appl Phys 69:7924, 1991. M Murayama, K Uchida. Combust Flame 91:239, 1992. DW Hahn, CF Edwards, KF McCarty, RJ Kee. Appl Phys Lett 68:2158, 1996. J Powling. Fuel 28:25, 1949. WT Bielder, HE Hoeschler. Jet Propul 25:1257, 1957. CF Edwards, DW Hahn. Proceedings of ILASS-95 (Institute for Liquid Atomization and Spray Systems), Troy, MI, 1995, p 34. GH Evans, R Grief. Numer Heat Transfer 14:373, 1988. RJ Kee, JA Miller, GH Evans, G Dixon-Lewis. In: Twenty-second Symposium (International) on Combustion. Pittsburgh: The Combustion Institute, 1988, p 1479. CA Wolden, Z Sitar, RF Davis, JT Prater. Diamond Relat Mater 6:1862, 1997. CA Wolden, SK Han, MT McClure, Z Sitar, JT Prater. Mater Lett 32:9,1997. BA Fox, BR Stoner, D Malta, P Ellis, RC Glass, FR Sivazlian. Diamond Relat Mater 3:382, 1994. SK Han, MT McClure, CA Wolden, Z Sitar. Mater Res Soc Symp Proc 423: 281, 1996. SD Wolter, BR Stoner, JT Glass, PJ Ellis, DS Buhaenko, CE Jenkins, P Southworth. Appl Phys Lett 62:1215, 1993. JE Graebner, S Jin, GW Kammlot, B Bacon, L Seibles, WF Banholzer. J Appl Phys 71:5353, 1992. KV Ravi, CA Koch, DS Olson, P Choong, JW Vandersande, LD Zoltan. Appl Phys Lett 64:2229, 1994. JW Glesener, AA Morrish, KA Snail. J Appl Phys 70:5144, 1991. JW Glesener, AA Morrish, KA Snail. Appl Phys Lett 61:429, 1992. JW Glesener, KA Snail, AA Morrish. Appl Phys Lett 62:181, 1993. K Doverspike, JE Butler, JA Freitas Jr. Diamond Relat Mater. 2:1078, 1993. CA Wolden, Z Sitar, RF Davis, JT Prater. J Mater Res 70:2733, 1997. SJ Harris, DG Goodwin. J Phys Chem 97:23, 1993. BW Yu, SL Girshick. J Appl Phys 75:3914, 1994. MA Cappelli, PH Paul. J Appl Phys 67:2596, 1990. RJH Klein-Douwel, JJL Spaanjaars, JJ ter Meulen. J Appl Phys 78:2086, 1995. JS Kim, MA Cappelli. J Appl Phys 72:5461, 1992. E Meeks, RJ Kee, DS Dandy, ME Coltrin. Combust Flame 92:144, 1993. JA Miller, CF Melius. Combust Flame 91:21, 1992. NG Glumac, DG Goodwin. Combust Flame 105:321, 1996. CA Wolden, S Mitra, KK Gleason. J Appl Phys 72:3750, 1992. SJ Harris, AM Weiner. J Appl Phys 74:1022, 1993. NG Glumac, DG Goodwin. Appl Phys Lett 60:2695, 1992. JS Kim, MA Cappelli. Appl Phys Lett 65:2786, 1994. SJ Harris, HS Shin, NG Glumac. Appl Phys Lett. 66:891, 1995. JS Kim, MA Cappelli. Appl Phys Lett 67:1081, 1995. DG Goodwin. J Appl Phys 74:6888, 1993. Final report for DARPA Contract Number N00014-93-C2044, 1997.

9 Laser-Assisted and Optical Pumping Techniques for Diamond Synthesis Vish V. Subramaniam and Shashi M. Aithal The Ohio State University, Columbus, Ohio

I.

INTRODUCTION

Diamond growth by chemical vapor deposition (CVD) processes requires activation of the gas phase adjacent to a substrate. The previous chapters of this book have discussed several methods of providing this activation needed for driving chemical reactions in the gas phase. Among these are hot-filament CVD (HFCVD), flame synthesis, and plasma CVD. All three methods rely upon thermal heating or near-equilibrium effects to induce reactions in the gas leading to diamond growth. Consequently, it is not possible to exercise selectivity from among the various competing chemical processes for diamond growth. Selectivity of desired chemical channels or lowering of growth temperatures, for instance, requires that attention be devoted to how the growth medium is energized and reacted. Lasers offer a unique means of activating the molecules in the growth medium. They are potentially capable of reducing growth temperatures as well as providing selectivity in a reactive process. In this chapter, the use of lasers in diamond synthesis is discussed. Use of lasers for diamond synthesis has not always been motivated with the goal of commercialization. Rather, lasers have been used to create environments that enable detailed study of the gas (or condensed) phase chemistry. These can also lead to better understanding of some of the reacting channels occurring in the other, more commercially viable processes 325

326

Subramaniam and Aithal

leading to process improvements. There are also other reasons that laserbased methods of diamond synthesis have been explored. These include the need to reduce substrate and gas (or liquid) temperatures during growth and to control and limit the codeposition of nondiamond carbon by selectively driving only the specific chemical channels that lead to diamond formation. In addition, lasers can be used to control diamond deposition patterns and enhance growth rates. Processes that meet these requirements are not necessarily the most economical as well. Hence, cost is not the only issue to consider when comparing laser-based methods with other better established CVD processes and General Electric’s high-pressure, high-temperature (HPHT) process. In order to appreciate the utility of lasers in CVD processes for diamond growth, it is necessary to understand how the energy added to a gas is redistributed within the various modes of its molecules. Gases comprise molecules and atoms and, in the case of plasmas, ions and electrons as well. These molecules, atoms, and ions in turn possess various internal modes of energy storage in addition to their translational (or what is called external) modes of motion. Figure 1 shows these various modes schematically. In addition to the three external degrees of freedom (i.e., translation in the three coordinate directions) that a particle possesses, molecules and molecular ions can store energy in rotation, vibration, and electronic excitation. Atoms and atomic ions can store energy only in electronic excitation in addition to their external modes. When energy is added to a gas in the

Figure 1 Schematic showing the different modes of energy storage for a diatomic molecule.

Laser-Assisted and Optical Pumping Techniques

327

form of simple heating, it first appears in the translational motion of the gas particles. Temperature is the common measure used to describe the average translational energy of the gas. The energy added by heating eventually makes its way into the other modes of molecular motion via collisions. Usually, exchange of translational energy between molecules takes place within a few collisions, and equilibration of energy between rotational and translational modes usually requires on the order of tens of collisions. In contrast, many collisions (on the order of thousands or many more) are required for the vibrational modes of molecules to equilibrate with rotational and translational modes [1–3]. Because the excitation of the vibrational modes is helpful in driving endothermic reactions, it can be seen that simple thermal heating is a circuitous and inefficient route to cause a gas to react. Thermal stimulation of a gas can be accomplished in a number of ways, including simple heating, electrical heating by partially ionizing it and then driving a current through the gas, or using a laser. It is usually inefficient and expensive to use a laser for thermal heating of a gas unless repetitive, very short duration pulse heating is required for a particular process. Lasers are also useful in instances where only the gas adjacent to a surface must be heated without heating the surface itself (see Fig. 2). In cases where the surface alone needs to be heated to drive a heterogeneous reaction, the gas must be transparent to the laser source whose energy is absorbed at the substrate-gas interface. This is often useful in the case where spot heating as opposed to bulk heating is required. Lasers drive chemical reactions by exciting specific transitions in a molecule. It may be pertinent to ask at this point whether or not a laser is necessary to excite a specific transition in a molecule. In other words, can one use an intense lamp (with an appropriate filter) to produce monochromatic (but incoherent) light to initiate reactions in gases? The answer of course is yes, and chemists were employing such techniques long before the invention of lasers to measure reaction rates [4]. However, inference of re-

Figure 2 Schematic showing the use of a laser to drive homogeneous gas-phase reactions by pyrolysis, photolysis, or laser excitation without heating of the surface.

328

Subramaniam and Aithal

action rates in complex chemically reacting systems is rendered difficult for two important reasons. First, incoherent monochromatic sources such as intense lamps are limited in intensity because they are diffuse, i.e., emit light in all directions. Usually, the higher the intensity (i.e., energy per photon multiplied by the number of photons per unit area, per unit time), the higher is the reaction yield, which makes it easier to measure changes in concentrations. Second, production of extremely short duration pulses (say of duration nanoseconds to hundreds of nanoseconds) of sufficient energy to drive fast reactions is nearly impossible with an incoherent, monochromatic source. Lasers, on the other hand, are monochromatic sources that can produce high pulse energies with extremely short durations and sustain these characteristics over long distances because they are coherent. We now digress a bit and examine this property of coherence in some detail, following the development in Ref. 5. Two sources are said to be coherent if they emit light or photons of a given frequency ␯ (or energy h␯, where h is Planck’s constant), with nearly the same amplitude, and in phase with each other. An important consequence of coherence is that a maximum intensity is observed at a point in space where the phase difference between two waves or photons reaching that point is an integral multiple of 2␲. Another way of stating this idea is that the maximum intensity will occur when the path difference1 between two waves or photons is an integral multiple of the wavelength ␭ (= c/␯, where c is the speed of light). The minimum intensity occurs when the phase difference between the photons is an odd multiple of ␲, i.e., (2m ⫹ 1)␲, where m is an integer (= 0, 1, 2, . . .). Coherent sources thus exhibit the important property of interference. The appearance of fringes (i.e., alternating segments of maximum intensity and minimum intensity) is due to interference. Because light is a form of electromagnetic radiation, it can be repre→ sented at any point in space as a time-varying electric field E(t) = → → E0e ⫺i␻te i␾(t), where E0 is the amplitude, ␻ = 2␲␯ is the circular frequency, i = 兹⫺1, t is time, and ␾→(t) is the time-dependent phase. Thus, when two → photons represented by E1 and E2 interact, the resulting intensity can be expressed as a time average: →

→

→

→

→

→

I = 具E⭈E*典 = 具(E1 ⫹ E2)⭈(E* 1 ⫹ E*)典 2 →

→

→

→

= 具兩E1兩2 ⫹ 兩E2兩2 ⫹ 2 Re(E1 ⭈E*)典 2 where the angle brackets denote the time average, defined as 1

If the path difference is x and the phase difference is ␦, the two are related by ␦ = 2␲x/␭, where ␭ is the wavelength.

Laser-Assisted and Optical Pumping Techniques

具 f 典 = lim → t0

⬁

1 t0

冕

329

t0

f(t) dt

0

The symbol Re denotes the real part of a complex number, and the superscript * denotes the complex conjugate of a complex number. Now for simplicity, if we assume that the two fields are polarized,2 then their vectorial nature can be ignored. The intensity then becomes I = I1 ⫹ I2 ⫹ 2 Re具E1E*典 2 where I1 = 具兩E1兩2典 and I2 = 具兩E2兩2典. Thus, if t is the time required for one photon to travel via a path A and if (t ⫹ ␶) is the time required for a second photon to travel via path B, then the intensity due to interference is ⫹ ␶)典 = 2 Re(⌫12(␶)). The function ⌫12(t) is called the mutual 2 Re具E1(t)E*(t 2 coherence function or correlation function of the two fields. Analogously, ⌫11(t) is called the autocorrelation function. By definition, ⌫11(t = 0) = I1 and ⌫22(t = 0) = I2. The intensity is also written as I = I1 ⫹ I2 ⫹ 2兹I1I2 Re{␥12(␶)} where ␥12(␶) is called the degree of partial coherence and is given by

␥12(␶) =

⌫12(␶) ⌫12(␶) = ⌫ (0)⌫ (0) 兹 11 兹I1I2 22

If 兩␥12兩 = 1, we have complete coherence. If 兩␥12兩 = 0, there is complete incoherence, and if 0 < 兩␥12兩 < 1, there is partial coherence. Now, suppose that the interference pattern varies between the two limits set by Imax and Imin: Imax = I1 ⫹ I2 ⫹ 2兹I1I2兩␥12兩 Imin = I1 ⫹ I2 ⫺ 2兹I1I2兩␥12兩 The fringe visibility is then defined as ᭙=

Imax ⫺ Imin 2兹I1I2兩␥12兩 = Imax ⫹ Imin I1 ⫹ I2

When I1 = I2, then ᭙ = 兩␥12兩. Thus, for complete coherence, ᭙ = 1 and the fringes have a maximum contrast of unity, whereas for complete incoherence, ᭙ = 0 and the fringe contrast is 0 (i.e., there are no distinguishable fringes). We will now attempt to relate the degree of partial coherence to the characteristics of a source. Suppose the source emits an electromagnetic 2

Simply defined, polarization is a chosen direction of the electric field vector in the plane normal to the direction of propagation of the electromagnetic wave.

330

Subramaniam and Aithal

wave whose field varies sinusoidally for a certain duration ␶0 and then changes phase abruptly. This can represent, for example, the case of an excited atom or molecule that emits radiation for a certain amount of time and then suffers a collision with another atom or molecule. This sequence of emission for a certain duration ␶0 followed by an abrupt change of phase is then repeated indefinitely. ␶0 is called the coherence time. The phase therefore may vary in a manner such as that shown in Fig. 3 for illustrative purposes. Suppose now that the light from this source represented by E(t) = E0e⫺i␻te i␾(t) is split into two beams and subsequently brought together to produce interference. Assuming that 兩E1兩 = 兩E2兩 = 兩E兩, the degree of self partial coherence is

␥(␶) =

具E(t)E*(t ⫹ ␶)典 具兩E兩2典

Substituting for E(t), we have

␥(␶) = 具e i␻␶e i [␾ (t)⫺␾ (t⫹␶)]典 = ei␻␶ lim → T

⬁

1 T

冕

T

e i [␾ (t)⫺␾ (t⫹␶)] dt

0

Figure 4 shows an illustrative plot of the phase difference ␾(t) ⫺ ␾(t ⫹ ␶) versus time. For the first time interval, 0 < t < ␶0, ␾(t) ⫺ ␾(t ⫹ ␶) = 0 when 0 < t < ␶0 ⫺ ␶. However, for ␶0 ⫺ ␶ < t < ␶0, the phase difference takes on some random value ⌬ between 0 and 2␲. Therefore, the degree of self partial coherence becomes

␥(␶) = e

i␻␶

= e i␻␶

1 ␶0

冕

␶0

e

i [␾ (t)⫺␾ (t⫹␶)]

dt = e

0

i␻␶

1 ␶0

再冕

␶0⫺␶

0

冕

␶0

dt ⫹

␶0⫺␶

冎

e i⌬ dt

1 {␶0 ⫺ ␶ ⫹ ␶e i⌬} ␶0

Each interval will have the same expression for ␥(␶), except that the value

Figure 3 Plot of phase ␾(t) versus time t for the example of a quasi-monochromatic source. (From Ref. 5.)

Laser-Assisted and Optical Pumping Techniques

331

Figure 4 Plot of phase difference, ␾(t) ⫺ ␾(t ⫹ ␶) versus time t for the discussion of coherent length and coherent time. (From Ref. 5.)

of ⌬ will be different. Because ⌬ is random, the time average of e (i⌬) is zero, and the average of ␥(␶) over all such intervals will be (1 ⫺ (␶/␶0)e i␻␶, whenever ␶ < ␶0. On the other hand, when ␶ > ␶0, the phase difference is always random so that its average over many intervals is zero. Hence, we have

␥(␶) =

or

再 再

兩␥(␶)兩 =

冉 冊

for ␶ < ␶0

0

for ␶ ⱖ ␶0

1⫺

1⫺ 0

␶ e i␻␶ ␶0

␶ ␶0

for ␶ < ␶0 for ␶ ⱖ ␶0

This function is shown schematically in Fig. 5. Since 兩␥兩 is equal to the fringe visibility ᭙ for two light waves of equal amplitude, it can be seen from Fig. 5 that ᭙ drops to zero when ␶ exceeds the coherence time ␶0. This means that the path difference between the two split light waves, say x, must not exceed the value of Lc = c␶0, where c is the speed of light. The quantity Lc = c␶0 is called the coherence length. Thus, for x > Lc, interference fringes will not be observed. Interference fringes can be observed only for x < Lc. In a radiating gas, the time between collisions of gas molecules is not constant but varies randomly from one collision to the next. However, an average time between collisions can be defined. Hence, ␶0 should be interpreted as an average value of the individual coherence times. The same holds true for the coherence length.

332

Subramaniam and Aithal

Figure 5 Plot of degree of self partial coherence, 兩␥ (␶)兩 versus ␶ for the discussion of coherent length and coherent time. (From Ref. 5.)

We are now in a position to understand the crucial differences between lasers and ordinary, intense, incoherent sources. The first point to realize is that no source of light is ever strictly monochromatic. There is always a spread of frequency about some mean frequency due to natural broadening.3 It can be shown using Fourier analysis that the width ⌬␻ of the frequency distribution is given by ⌬␻ = 2␲/␶0 or ⌬␯ = 1/␶0 [5]. Conversely, we can also say that a spectral line of width ⌬␯ has a corresponding coherence time of ␶0 = 1/⌬␯, with a coherence length of Lc = c␶0 = c/⌬␯. This condition can also be expressed in terms of wavelength as Lc = ␭2/⌬␭ because ⌬␯/␯ = 兩⌬␭兩 /␭. For ordinary light sources such as lamps or discharge tubes, the line widths in the visible portion of the spectrum (say around 500 nm) are on the order of an angstrom [5]. The corresponding coherence length is then on the order of 5000 wavelengths or about 2.5 mm. This means that with such light sources, interference fringes would disappear beyond a distance of about 2.5 mm. In contrast, the line width of a gas laser can be as narrow as 103 Hz or less,4 which corresponds to a coherence length of about 300 km! Thus, coherent sources are capable of producing interference effects over much larger distances than incoherent sources. The other essential difference between incoherent and coherent light sources is the property of directionality. Ordinary lamps and discharge tubes, which produce incoherent light, radiate in all directions. In contrast, light

3

Natural broadening refers to the finite line width (centered around a specific wavelength) observed for a spectral line due to the finite radiative lifetime of a given excited state of an atom or molecule. The width associated with natural broadening has its origins in Heisenberg’s uncertainty principle, which gives a relationship between uncertainties in time and energy. 4 The corresponding line width for ordinary, incoherent light sources is on the order of 1010 Hz.

Laser-Assisted and Optical Pumping Techniques

333

from a laser can come very close to an ideal plane wave whose divergence is primarily due to diffraction effects [6]. Therefore, the high directionality of the laser beam can be used to focus it to extremely small dimensions and thus attain high power densities or intensities. For the aforementioned reasons, laser-based methods have a great potential to drive chemical reactions. It is more efficient to stimulate internal modes of gas molecules (i.e., the more reactive modes such as vibrational or electronic excitation) directly rather than relying upon thermal heating. The former relies upon sustaining thermodynamic disequilibrium between internal and external modes of molecular motion. Excitation of the internal modes with minimal excitation of the translational modes also leads to some control over selectivity, as we shall see later in this chapter. In contrast, thermal heating of a gas, results in states that are in thermodynamic equilibrium or in local thermodynamic equilibrium (LTE).5 Lasers, depending on how they are used, are capable of producing gaseous environments that can be in LTE or far from it. Processes that use this departure from LTE to induce selective growth of diamond or produce films of desired properties or characteristics are discussed later in this chapter. Their ultimate utility in industrial applications, though, must be evaluated with regard to cost.6 The use of lasers for diamond synthesis (or synthesis of other materials for that matter) can be grouped into general categories: (1) rapid heating and pyrolysis, shown schematically in Figs. 2 and 6; (2) ablation or physical vapor deposition (PVD), shown schematically in Fig. 7; (3) photolysis, shown schematically in Figs. 2 and 6; (4) laser excitation or optical pumping, shown schematically in Figs. 2 and 6; and (5) hybrid methods. An obvious use of a laser is for intermittent, rapid heating. Laser pyrolysis is included in this grouping, although there are some important differences. A laser can be used to heat a gas by irradiating surfaces in contact with the gas. This takes advantage of the fact that solid surfaces tend to absorb radiant energy broad band (i.e., over a continuous range of wavelengths) whereas gases tend to absorb discrete lines (i.e., at specific wavelengths). Such an instance of a solid surface being heated is encountered 5

Strictly speaking, the state of thermodynamic equilibrium is defined as one that is homogeneous, isotropic, and devoid of any gradients. However, most practical engineering situations involve flows and gradients. Therefore, the concept of local thermodynamic equilibrium or LTE is typically introduced as an approximation to allow a flowing gas exposed to gradients to be treated as consisting of infinitesimal regions that are nearly homogeneous, isotropic, and devoid of gradients. 6 We shall address the issue of cost later, as each technique is described. We hasten to point out that the economics of diamond is largely artificial, and therefore we will avoid using units of dollars per carat to discuss cost. Rather, we evaluate processes based upon energy cost expressed here in units of joules per carat ( J/carat).

334

Subramaniam and Aithal

Figure 6 Schematic showing the use of a laser to drive gas-phase or heterogeneous reactions by pyrolysis, photolysis, or laser excitation while heating the surface locally.

when heterogeneous (i.e., gas-solid) reactions need to be driven. The laser wavelength must then be chosen so that the gas is transparent to this wavelength while the surface is not. A potential advantage over other methods of heating the surface is that the laser energy can be absorbed at the surface alone, with minimal heating of the bulk material. If the gas alone is to be heated, the laser energy must first be absorbed7 by the gas (at a specific wavelength or discrete wavelengths corresponding to vibrational or electronic transitions in the gas molecules). The energy initially absorbed in the internal modes is then rapidly redistributed to the external (translational) modes, thus raising the gas temperature and driving chemical reactions. This process, known as pyrolysis, can involve either specific constituents or all constituents of a gas mixture. Lasers can also be used to irradiate and vaporize a target material, with the vapor-phase products subsequently condensing on a surface placed near the target. We shall refer to this technique as laser ablation, and a schematic is shown in Fig. 7. In the literature, laser ablation is also known as laser physical vapor deposition or LPVD. This method has been used successfully to produce thin films of superconductors and other materials. The target or targets are usually pure materials or materials of fixed, known composition. Intense laser energy (usually in the ultraviolet) is focused on the target,

7

Laser energy can be absorbed by a gas molecule M either by resonant single-photon absorption or by multiphoton absorption. The former utilizes transitions between real energy levels in a molecule whereas the latter can also occur via virtual energy states: single-photon: h␯ ⫹ M(E1) → M(E2) where E2 > E1 multiphoton: nh␯ ⫹ M(E *) → M(E *) where E * 1 2 2 > E* 1 and E * 1 and E * 2 do not have to be real energy states belonging to the molecule M

Laser-Assisted and Optical Pumping Techniques

Figure 7

335

Schematic of laser physical vapor deposition or LPVD process.

causing portions of it to vaporize and in some cases ionize or react. The vapor then is made to condense on a substrate placed nearby (which may be heated or cooled), resulting in a thin film. In most cases, the laser also interacts with the vapor (which may or may not be ionized) emanating from the target. Thus, the laser ablation technique could be easily applied in conjunction with other techniques such as plasma CVD, hot-filament CVD, or flame synthesis, resulting in a hybrid method. The third category is photolysis, wherein multiphoton processes are used to break apart polyatomic molecules into smaller, radical fragments. Multiphoton absorption is a nonlinear process whereby two or more photons are absorbed by a molecule, resulting in dissociation of the polyatomic molecule into molecular or atomic fragments. The process can be written symbolically in the following form of a chemical reaction: Mn ⫹ p(h␯) → Ra ⫹ Sb

(1) ⫺34

where p photons of frequency ␯ are absorbed (h = 6.626 ⫻ 10 J-sec, is Planck’s constant) by a polyatomic molecule containing n (n = a ⫹ b) atoms. The result is that the molecule is taken to its dissociation limit and the resulting polyatomic fragments are R (containing a atoms) and S (containing b atoms). The two ways in which a multiple number of photons can be absorbed by a polyatomic molecule are shown schematically in Fig. 8. The photons do not necessarily have to access real energy levels within the molecule en route to the dissociation limit. The multiphoton process can access virtual energy states [4,7,8]. Further, the absorption process can be either sequential or simultaneous. An example of a naturally occurring multiphoton process is the dissociation of O2 in the earth’s upper atmosphere due to ultraviolet (UV) radiation from the sun: O2 ⫹ m(h␯) → O ⫹ O

(2)

336

Subramaniam and Aithal

Figure 8 Schematic showing multiphoton dissociation (MPD) process in a polyatomic molecule. MPD processes can take place through interaction of the molecule with many photons either sequentially or simultaneously and can involve either real energy levels or virtual energy levels.

where m is at least 2 for photons with a wavelength of 193 nm.8 The O atoms subsequently recombine with O2 to form ozone (O3). This technique of photolysis is routinely employed by chemists to measure reaction rates involving specific molecules or radicals [4,7,8]. There are laser-based techniques that take advantage of the coincidence between the wavelength of the laser light and energy difference for a particular transition (either electronic or vibrational) in a given molecule. Because such an application typically does not involve high powers but rather the right wavelength, these techniques are collectively referred to here as laser excitation or laser-excited CVD. A listing of different lasers presently available along with their operating wavelengths is given for reference in Table 1. A broader term that encompasses incoherent light (such as UV lamps) as well as coherent (laser) light is optical pumping. The terms laser excitation and optical pumping will be used interchangeably throughout this chapter, even when a laser is used to create the optical pumping. In optical pumping of a gas, the absorption is usually by a resonant single-photon process:

8

The particular wavelength of 193 nm is also the wavelength at which the ArF (argon-fluoride) excimer laser lases. The wavelength of 193 nm is in the range of absorption for O2 via the Schumann-Runge bands [4].

Laser-Assisted and Optical Pumping Techniques

337

Table 1 Available Laser Sources and Their Respective (Commonly Found) Wavelengths Laser type

Wavelength

ArCl excimer ArF excimer KrCl excimer KrF excimer Frequency-quadrupled Nd:YAG XeBr excimer XeCl excimer Nitrogen laser XeF excimer 7-Hydroxycoumarin dye 4-Methylumbelliferone dye Esculin dye Argon ion laser Argon ion laser Frequency-doubled Nd:YAG Puronin B dye Na-fluorescein dye 2,7-Dichlorofluorescein dye Rhodamine 6G dye Acridine red dye Rhodamine B dye Helium-neon laser Ruby (Al2O3 doped with Cr2O3) Nd:YAG (Y3Al5O12) HF chemical laser DF chemical laser CO laser CO2 laser Free electron laser

175 nm 193 nm 222nm 248 nm 266 nm 282 nm 308 nm 330 nm 351 nm 450–470 nm 450–470 nm 450–470 nm 488 nm 514.5 nm 532 nm Yellow 530–560 nm 530–560 nm 570–610 nm 600–630 nm 605–635 nm 632.8 nm 692.9 nm and 694.3 nm 1.064 ␮m (dominant line) 2.7 ␮m 3.8 ␮m Multiple lines from 4.7 to 5.8 ␮m 9.4 ␮m, 10.6 ␮m Visible to ultramicrowave (mm) wavelengths

Mn(E1) ⫹ h␯ → Mn(E2)

(3)

where h␯ = E2 ⫺ E1, and E1 and E2 are energy levels within the molecule Mn, comprising n atoms. This is followed by collisional processes that lead to energy transfer and subsequent chemical reaction. Lasers have been used successfully to synthesize diamond particles, films, and diamond-like films. Although lasers have also been used to enhance and control nucleation, reduce the surface roughness, and improve adhesion of the film to the substrate, the focus of this chapter is on synthesis.

338

Subramaniam and Aithal

It is not our intent to provide a general background on lasers or on their principles of operation. There exist several excellent books on the subject [9,10]. Laser operation and characteristics will be discussed only as and when specifically applicable to a particular process. The focus here is on the more nonstandard and novel applications of lasers for diamond growth, especially those that result in highly nonequilibrium environments. In other words, we explore the situations that can best be created by a laser and, in some cases, not by any other means. This chapter is organized as follows. Section II describes synthesis of diamond and diamond-like carbon reportedly resulting from rapid heating and/or cooling due to laser irradiation. Production of diamond-like carbon by ablation of a precursor target is described in Sec. III, followed by a discussion of use of laser CVD for diamond synthesis by photolysis of carbonaceous precursors in Sec. IV. Diamond synthesis by relatively low-power resonance absorption of laser energy (i.e., laser excitation) is discussed in Sec. V. A chapter summary is given in Sec. VI.

II.

RAPID HEATING AND COOLING

A.

Introduction

There are specific instances where laser beams provide an intense source of energy that results in heating of a target gas, solid, or liquid. Depending on the rate at which the incident energy is absorbed and redistributed per unit time, resulting temperatures and pressures can exceed thousands of kelvins and many hundreds of kilobars or more. Referring back to the (equilibrium) phase diagram for carbon (see Fig. 13), it can be seen that such conditions are conducive for conversion of nondiamond carbon to diamond. Even when the output power of a laser beam alone does not provide sufficient intensity or the required heating and compression, the beam can be focused to provide intense heating over a small region. The output beam from a laser is usually at least a few millimeters in diameter. When focused using a lens (of radius b < a) or focusing mirror, the ultimate radius of the beam of radius a can be estimated from physical optics to be [6,9] r0 = ␭ f/min{a, b}

(4)

where ␭ is the wavelength of laser light and f is the focal length of the lens or optical element used. In practice, however, the ultimate size of a focused laser beam is much larger than that given by Eq. (4) because of multimode emission from the laser and lens aberration. Now, if P is the power of the incident laser beam, then the intensity can be written as

Laser-Assisted and Optical Pumping Techniques

冕

339

a

I=

Ip(r)2␲ r dr

0

␲ r 20

When Ip is independent of r, we have I=

P ␲ r 20

(5)

We can also write Eq. (5) using Eq. (4) as I = Pa2/␲␭2 f 2

(6)

Usually, Ip depends on r in a Gaussian manner but Eqs. (5) and (6) can be easily rewritten for this case. Although Eq. (6) may not be quantitatively accurate for real laser beams for which Ip is a function of r, it does provide the correct trends. For instance, the intensity does scale inversely with the square of the laser wavelength and lens focal length and directly with the square of the initial (i.e., prefocusing) beam size. Thus, for a given laser output power, the intensity increases for shorter wavelength laser beams, smaller focal length lenses, and large initial beam diameters. B.

Gas-Phase Heating

An appropriate example of rapid gas-phase heating triggering reaction and ultimately leading to formation of diamond particles is the experiment of Buerki and Leutwyler [11]. This work combines both pyrolytic and photolytic decomposition of the gas phase. In this experiment, mixtures consisting of ethylene (C2H4), H2, and silane (SiH4) in compositions of 51.5–100%:0– 41.2%:0–7.2% were introduced vertically through a capillary nozzle into a reactor. These precursors were diluted in either N2 or Ar or both. The gas flow was then irradiated horizontally by a focused line-tunable continuouswave (cw) CO2 laser beam, 2–3 mm above the nozzle. Silane (a strong absorber of 10.6-␮m CO2 laser radiation) was used to elevate the gas temperature and to enhance pyrolysis or chemical decomposition of C2H4. Two processes that cause such gas mixtures to react are pyrolysis and photolysis. The process by which these small molecules decompose upon exposure to infrared radiation from the CO2 laser is known as infrared multiphoton dissociation or IRMPD [7]. Both SiH4 and C2H4 absorb infrared photons via multiphoton processes (of the type discussed in footnote 7 and shown in Fig. 8) triggered by the intense radiation field of the focused laser beam and are subsequently taken to their respective dissociation limits [7]. Inevitably, some of the laser energy is also transferred to the vibrational modes of these

340

Subramaniam and Aithal

molecules, resulting in rapid gas heating via vibration-rotation-translation (VRT) and vibration-translation (VT) collisional energy transfer. Thus, some of these molecules are also pyrolyzed and decomposed. The authors report that 1 g of particles black in appearance were produced from C2H4 /N2 /Ar/ H2, C2H4 /N2, C2H4 /Ar, and C2H4 /N2 /H2 mixtures. The particles (on the order of 5–15 nm in size) were characterized by bright-field transmission electron microscopy (TEM) imaging and electron diffraction. Based on this analysis, the authors report finding diamond for the C2H4 /N2 gas mixture and a graphite-diamond mixture in all the other cases. In addition, these authors report that a gas mixture of C2H4 /N2 /SiH4 /H2 yielded diamond particles ⬃120 nm in size in one instance. Although it is encouraging that they appear to be able to distinguish between diamond and graphite through bright-field TEM imaging and electron diffraction analysis, some caution is warranted. Disordered graphite is known to masquerade as diamond, especially in electron diffraction patterns [12,13]. The only way to establish deposits unequivocally as being diamond is through Raman spectroscopy. As an example, the Raman spectrum of a diamond film containing graphite and disordered graphite synthesized using an oxyacetylene flame is shown in Fig. 9. This particular film can be seen to contain diamond (1332 cm⫺1 line), graphite (as can be seen from the ‘‘G’’ peak at 1580 cm⫺1), and disordered graphite

Figure 9 Raman spectrum of a diamond film synthesized using an oxyacetylene flame, displaying peaks identifying diamond and disordered graphite. (From Refs. 15–17.)

Laser-Assisted and Optical Pumping Techniques

341

(as evidenced by the presence of the ‘‘D’’ peak at 1343 cm⫺1). Unfortunately, no Raman spectra are presented in Ref. 11. This is understandable given the small size of the particles and the small sample, both of which would yield an extremely small signal (if any) in Raman spectroscopy. In the absence of such evidence, as many additional methods as possible of characterization of the physical properties (such as hardness) and chemical properties (such as reactivity to hydrogen or oxygen when heated, or exposure to acids) are needed. Again, this is difficult when sample sizes are small. Gas-phase reactions with C2H4 as the carbonaceous precursor required about 10 W of power (the residence time was about 1 msec and the pressure was 1 atm) [11]. Based upon the values given in Ref. 11, a characteristic J/carat value (not counting capital cost) can be computed for this case using the optimistic assumption that all of the 1 g of deposit yield is diamond. Reference 11 reports a linear growth rate of 5.7 ␮m/sec, and assuming this is all diamond, we can compute an equivalent spherical volumetric growth rate of (4/3)␲(5.7 ⫻ 10⫺6)3 m3/sec or 7.8 ⫻ 10⫺16 m3/sec. Given the density of diamond (3515 kg/m3) and that 1 carat = 0.2 g or 2 ⫻ 10⫺4 kg, the growth rate in carats per second is 7.8 ⫻ 10⫺16 ⫻ 3515/2 ⫻ 10⫺4, or 1.4 ⫻ 10⫺8 carats/sec. The power required for this process according to Ref. 11 was 10 W or 10 J/sec, so that the energy cost in joules per carat is 10/1.4 ⫻ 10⫺8, or 7.1 ⫻ 108 J/carat. The value of 7.1 ⫻ 108 J/carat is enormous, and it is important to put it in perspective by comparing with the corresponding energy cost for diamond synthesis using an oxyacetylene flame. For flame synthesis, assuming a linear growth rate of 100 ␮m/hr on a 1 cm2 surface area, we obtain a typical value9 of 8 ⫻ 105 J/carat. Thus, we can see that the energy cost of this particular pyrolysis/photolysis process is three orders of magnitude higher and has no apparent gain in quality of the diamond or in selectivity. It must be further emphasized that laser efficiency and capital cost have not been considered here. The joule per carat value, however, does provide an index for comparing these different processes. A more recent example of extreme gas-phase heating to the point of ionization is the laser-plasma technique developed by Konov and Uglov [14]. In this work, a powerful (2.5 kW) cw CO2 laser beam was focused using an NaCl lens ( f = 7 cm) into a chamber containing a flowing mixture of Xe, H2, and CH4 (in proportions of 300:30:1) at 1 atm. The authors report growth of diamond films 1 cm2 in area at a rate of 30–50 ␮m/hr on water9

Assuming a growth rate of 100 ␮m/hr over a circular spot 1 cm2 in area, the volume growth rate is 2.8 ⫻ 10⫺12 m3/sec or 4.9 ⫻ 10⫺5 carats/sec. Taking 39 W as that power required to raise a total flow of 1.8 ⫻ 10⫺4 kg/sec of an acetylene-oxygen mixture from 25⬚C to the ignition temperature of 305⬚C, the cost is 8 ⫻ 105 J/carat. It must be pointed out that this value is typical of a laboratory-scale oxyacetylene flame.

342

Subramaniam and Aithal

cooled tungsten substrates, verified after growth using Raman spectroscopy. This growth rate is comparable to deposition using a laboratory-scale oxyacetylene flame, except that the power cost is higher by about two orders of magnitude. It must be pointed out, however, that this excessive power cost can be lowered if their process is optimized. This work is, however, very similar in nature to synthesis using the oxyacetylene flame except that conditions such as gas type and purity are more controllable. C.

Surface Heating

An alternative approach of using lasers to provide rapid heating has been demonstrated by several researchers [18,19]. In this instance, a stream of carbon black particles was exposed to CO2 and yttrium-aluminum-garnet (YAG) laser irradiation at various power densities and modes of operation (i.e., cw and pulsed). Conversion of carbon black particles to diamond is reported in Ref. 19. Again, no Raman spectra are given, but the authors claim that TEM imaging, x-ray, and electron diffraction data are able to distinguish diamond particles from disordered graphitic phases. Another report appears to confirm this finding based on TEM imaging, electron diffraction, x-ray photoelectron spectroscopy, and measurements of electrical resistivity [20]. In the latter work, a Q-switched Nd:YAG laser was used to irradiate a 200-nm-thick amorphous carbon (so called diamond-like carbon or DLC) film and to convert it to diamond. However, caution must be exercised about the credibility of these reported results given the lack of any Raman spectra. Further, an earlier study on pulsed Nd:YAG laser irradiation of amorphous carbon films actually reports conversion of the film to graphite at intensities of less than about 1 J/cm2 [21]. They confirm the structural change to graphite by Raman spectroscopy. Laser intensities of 4.5 J/cm2 were reportedly used in Ref. 20 and these intensities are not drastically different from those used in Ref. 21. It is also unlikely that the initial amorphous carbon films were very different. This comparison indicates that characterization methods other than Raman spectroscopy (such as TEM imaging and electron diffraction) may not be able to distinguish diamond from disordered graphite reliably. There has been an interesting attempt to use a laser to provide rapid heating of the surface layers of a single-crystal copper specimen, initially saturated with carbon. Because carbon is not soluble in copper, it can be driven out of the material with relative ease by heating. The technique described in Ref. 22 reportedly formed millimeter-size single-crystal diamond ˚ thick and were formed on the singlefilms. These films were about 500 A crystal copper substrate when irradiated with a laser beam. The presence of diamond was reportedly verified using Raman spectroscopy. This work was

Laser-Assisted and Optical Pumping Techniques

343

probably based upon the work of Ref. 23, which reported heteroepitaxial diamond growth by ion implantation of high-temperature carbon into a copper substrate. Unfortunately, neither the result reported in Ref. 22 nor that in Ref. 23 was reproduced in a subsequent attempt by Lee et al. [24]. Their approach essentially consisted of ion implantation of carbon at an ion energy of 50 keV and a flux of 1018 atoms/cm2 in a number of crystalline substrates [Cu(100), Cu(110), Cu(111), Ni(100), Ni(110), and single crystal Co]. The samples were then irradiated using separate laser wavelengths of 248 nm (KrF excimer laser with pulse width of 25 nsec), 308 nm (XeCl excimer laser with pulse width of 15 nsec), and 532 nm (frequency-doubled Nd:YAG laser with pulse width of 7 nsec) with power densities in the range of 1–6 J/cm2. The laser irradiation was carried out in vacuum and in the presence of nitrogen. No additional information was reported in their paper about whether or not the beam was focused or the spot size of the beams. Characterization by Raman spectroscopy, Auger spectroscopy, and TEM showed the resulting carbon films to be amorphous or composed of microcrystalline graphite. Lee et al. were therefore unable to reproduce the results of Refs. 22 and 23. Reference 24 appears to have reproduced all of the conditions reported in Ref. 22, but for the longest pulse width and perhaps the rise time of the laser pulse. In the experiments of Ref. 24, the longest pulse width was 25 nsec whereas the longest pulse width in Ref. 22 was 45 nsec. Because the pulse width affects the heat flux and total heat input into the surface, this could have affected the outcome of the experiments, with the rise time of the laser pulse being an important parameter. Figure 10 shows how two pulses can have different rise times while having the same total or integrated energy content. As we are dealing with pulse widths on the order of tens of nanoseconds, specific details such as rise time of the pulse and maximum pulse amplitude could be important in reproducing the experiments reported in Refs. 22–24. In other words, even if the heat addition over the duration of the laser pulse were the same in Refs. 22 and 24, the

Figure 10 Schematic showing two laser pulses having the same total or integrated energy while having different peak powers and rise times.

344

Subramaniam and Aithal

rate at which the heat is added may have been important. Unfortunately, pulse shapes are not reported in either Ref. 22 or Ref. 24. In addition, it would be interesting to repeat these experiments in the presence of atomic hydrogen, as this could help stabilize a diamond surface in the act of formation. It is instructive to explore the heating depth and evolving surface temperature distribution when a material is exposed to laser irradiation. Laser radiation can be reflected, transmitted, or absorbed at a surface and at depths below the surface. Of these, only the absorbed energy affects the temperature or average energy of the solid material. The absorbed radiation may be viewed as the result of interactions between the incident photons and electrons and photons and phonons10 in the material. In dielectric materials, phonon-phonon interactions are dominant whereas in metals photon-electron followed by electron-phonon interactions are important. The process of heat transport within the material therefore establishes itself (in the classical sense of heat conduction) only after a certain time ␶ has elapsed. This initial relaxation time, ␶, is the characteristic time required for photon-phonon or photon-electron and electron-phonon interactions (i.e., collisions) to exchange energy and establish local thermal equilibrium. In other words, the relaxation time is the time that must elapse before a local lattice temperature or local electron temperature can be defined. In metals, the characteristic times for phonon-phonon, photon-electron, and electron-phonon interactions are on the order of picoseconds, and in dielectrics the phononphonon relaxation time is on the order of nanoseconds to picoseconds. Therefore, if laser pulse widths are shorter than nanoseconds, the microscale or short-transient energy transfer must be considered, and for subnanosecond time scales it is likely that the temperature field is discontinuous [25]. During these short times, propagation of thermal energy within a material often follows a hyperbolic or wave equation unlike the usual parabolic equation describing the diffusion of heat.11 This means that a temperature distribution can propagate as a pulse or wave through the material, as shown in Fig. 11, with transient peak temperatures being much higher than normally predicted by the heat diffusion equation. Values of the quantities ␶q (relaxation time 10

Phonons are the quantum particles corresponding to the discrete or quantized vibrational states in the lattice of a solid. Typical lattice frequencies ␯ are on the order of 1013 Hz and energies are on the order of E = h␯, where h is Planck’s constant. Phonons propagate at the speed of sound, which for most solid materials at room temperature is on the order of 10 to 100 km/sec. 11 Although the hyperbolic form of the heat equation was first pointed out by Maxwell in the nineteenth century [26], it remained obscure until Cattaneo revived the idea in 1948 and again in 1958 [27].

Laser-Assisted and Optical Pumping Techniques

345

Figure 11 Plot showing the distribution of nondimensionalized temperature ␪(x, t) = (T(x, t) ⫺ T0)/T0 versus nondimensionalized distance ␦ = x/2兹␣␶q, at a nondimensional time ␤ = t/2␶q. The parameter B = ␶T/2␶q, where ␶T is the characteristic time for a meaningful temperature to be defined at a point or the phase lag of the temperature gradient, ␶q is the characteristic time for a meaningful temperature gradient to be defined at a point or the phase lag of the heat flux vector, ␣ is the thermal diffusivity, and T0 is the initial temperature at t = 0. The limit when B = 0 corresponds to what is known as a thermal wave, i.e., one in which a temperature pulse with an overshoot propagates as a wave through the material medium. (From Ref. 25.)

for the temperature gradient to be established) and ␶T (relaxation time for a temperature to be defined) inferred from measurements are 0.4348 and 70.833 psec, respectively, for metallic copper [25]. However, the corresponding values for oxides are larger, and it must be borne in mind that copper will have an oxide layer when exposed to air. Such differences can be important in explaining the different results obtained in Refs. 22–24. The effects of heating at short time scales (i.e., on the order of nsec or smaller) can be considerably different from behavior at long time scales (or times greater than a nanosecond), and this in turn can alter the diffusion of carbon from the bulk material to the surface.

346

Subramaniam and Aithal

Following the initial transient, the temperature response of a solid subjected to laser irradiation can be calculated for times on the order of tens of nanoseconds and greater. This is done by ignoring a part of the fast rising portion at the beginning of the laser pulse. In this instance, the heat flow in the material can be determined for most of the pulse duration of tens of nanoseconds using classical (Fourier-type) heat conduction. For a semi-infinite solid receiving a heat flux over a small, prescribed region on its surface, the transient temperature distribution is [28] T(x, t) ⫺ T(x, t = 0) = ⫺

q⬙x 0 erfc kT

1/2 2q⬙( 0 ␣Tt/␲) exp kT

冉 冊 x 2兹␣Tt

冉 冊 ⫺

x2 4␣Tt

(7)

where q⬙0 is the heat flux at the surface due to the absorbed laser energy, ␣T is the thermal diffusivity, kT is the thermal conductivity, x is the spatial coordinate, t is time, and erfc is the complementary error function. T(x, t = 0) is the temperature distribution following the fast transient portion of the laser pulse, after which time classical heat conduction occurs. The q⬙0 in Eq. (7) is a constant heat flux input into the surface. In reality, q⬙0 can be time dependent and varying even in the regime of classical heat conduction for the two pulse shapes shown in Fig. 10, so that Eq. (7) must be rederived in such an instance. In addition, the boundary condition (which yields the instantaneous value of q⬙) 0 must include a heat balance whereby heat is removed from the surface by radiation and convection. Hence, unrealistically high surface temperatures will be predicted by Eq. (7) if applied under conditions where radiative cooling of the surface is important. Although the situation described by Eq. (7) is not exactly that encountered in the experiments [22–24], it can nevertheless provide an indication of the temperature levels existing within the material during the period of laser irradiation. For illustrative purposes, temperature profiles given by Eq. (7) are plotted at various times in Fig. 12 for a laser intensity of 5 ⫻ 108 W/cm2, representative of the conditions existing in the experiments of Ref. 24. Because the rate of diffusion of carbon within the bulk Ni or Cu substrate material is temperature dependent, specifically varying as T 1/2, carbon initially near the surface would be driven out, aggregate, and condense to form a film. Although this would be expected to affect the growth rate of the carbon film at the surface, it is unclear what the effect of high local transient temperatures would be on the type of carbon aggregates thus produced. The nature of the resulting film could well be dependent on the rise time of the laser pulse.

Laser-Assisted and Optical Pumping Techniques

347

Figure 12 Temperature distributions calculated using Eq. (7) at various instants of time are shown here for an absorbed laser intensity of 5 ⫻ 108 W/cm2 for copper, representative of conditions in the experiments of Ref. 24.

Perhaps the most dramatic and as yet unsubstantiated claims have been the reports of diamond growth on a variety of substrates using CO2 as the carbonaceous precursor, with one or more laser sources (a CO2 laser, an excimer laser, and an Nd:YAG laser) [29–32]. It is important to point out that these claims were reported initially in the popular press, thereby circumventing the usual scientific peer review process. However, they are worth examining in view of what already exists in the published scientific literature and what is discussed in Secs. IV and V of this chapter. We will try to analyze this process, henceforth referred to as the QQC process (named after the industrial group that purportedly invented it). According to initial press releases, in the QQC process CO2 and N2 (as a shielding gas) are introduced in a plane stagnation flow over the substrate at an initial pressure of 1 atm. The process then consists of simultaneously irradiating a target surface to be coated, with simultaneous focused

348

Subramaniam and Aithal

beams from a CO2 laser, excimer laser (of undisclosed wavelength), and an Nd:YAG laser, which then reportedly produces diamond on a variety of substrates including stainless steel, cast iron, and a variety of metals and ceramics [29–32]. What is more remarkable is the claim of growth rates on the order of 1 ␮m/sec (specifically, a 45-␮m diamond coating is reportedly produced over a 1.6 cm2 area within 45 sec). Although few details are provided in Refs. 29–32, it is still possible to explore the feasibility of this approach. We now analyze the QQC process to explore whether or not the conditions might even be conceivably favorable for diamond growth. The excimer lasers and YAG lasers used in the QQC process are capable of producing extremely short pulses on the order of 10 nsec, so that the peak power can be on the order of 20 MW12 or more. This implies a peak intensity of 20 MW/mm2 or 2 GW/cm2 for a spot size or focal area of 1 mm2. An estimate of the gas temperature can then be made if we assume that the laser heating is balanced by radiative loss from the gas. This yields an approximate value of 13,000 K for the gas temperature of the resulting thermal plasma. Ionization and dissociation processes will tend to lower this estimated temperature so that realistic temperatures are probably in the range of thousands of kelvins. In the reported experiments, the target surface is exposed to a high intensity of infrared and ultraviolet energy. This must result in flash evaporation, in a manner similar to what occurs in laser shock processing or explosive welding [33–36]. The flash evaporation of the substrate material can result in an implosive (compressive) force at the material surface in the gigapascal range, depending on the pulse energy.13 This effect is put to good use in laser shock processing to improve the mechanical properties of metallic surfaces of dense or porous materials and in explosive welding [33–36]. The resulting gas mixture composed of vaporized substrate material, CO2, and N2 is rapidly heated to temperatures high enough to ionize the constituents. Such an ionized, electrically conducting gas is called a plasma. Accompanying this rapid heating is the production of extreme pressures. For a Gaussian-shaped laser pulse, assuming an absorption coefficient of 0.1 (typical for plasmas in this range of temperature and pressure), Ref. 35 gives the following relation between the maximum pressure in kbar and the absorbed intensity expressed in GW/cm2:

12

This is obtained assuming a peak energy of 200 mJ and a pulse width of 10 nsec. ˙ mass per unit time. This is typically milliA surface undergoing flash evaporation loses m ˙ on the order of 102 kg/ grams of material over the pulse duration of 10 nsec, yielding an m sec. Assuming a shock velocity on the order of the speed of sound at the temperature of 13,000 K, or O(103) m/sec, this yields a force of magnitude 105 N, or a pressure on the order of 1011 Pa (100 GPa, or 1000 kbar) on a spot 1 mm2 in area.

13

Laser-Assisted and Optical Pumping Techniques

Pmax = BI 1/2

349

(8)

where B is a constant with a value on the order of 10. It is interesting to note that this expression yields a maximum pressure that is independent of laser pulse duration and wavelength and one that is dependent only on the laser intensity at the focal point. Therefore, it is evident that the local processing conditions (i.e., in the gas and on the surface) achieved by such powerful lasers are on the order of GPa of pressure and temperatures on the order of thousands of kelvins. Shown in Fig. 13 and Fig. 14 are the phase diagrams for carbon in the absence of a catalyst and in the presence of catalytic metals such as nickel and iron, respectively. It can be seen that the pressures and temperatures shown in Figs. 13 and 14 match those expected in the QQC process. Thus, the QQC process appears capable of producing conditions locally comparable to those encountered in the well-known hightemperature, high-pressure (HTHP) process used for commercial diamond growth [37]. This places the QQC process within the realm of possibilities. However, the technique must be reproduced and independently verified before its claims such as growth rates of 1 ␮m/sec over an area of 1.6 cm2 can be substantiated. According to the analysis presented here, such growth rates may be possible on surfaces approximately 1 mm2 in area.

Figure 13 Ref. 38.)

Phase diagram for carbon without the presence of any catalyst. (From

350

Subramaniam and Aithal

Figure 14 Phase diagram for carbon in the presence of Ni catalyst (left) and Fe catalyst (right) at a pressure of 5.7 GPa. M, solid metal; L, liquid; G, graphite; D, diamond; C, carbide. (From Ref. 38.)

We can estimate the cost in terms of joules per carat for the QQC process, assuming that the growth rate of 1 ␮m/sec over a 1.6 cm2 area is indeed realizable. Using a pulse energy of 200 mJ/sec and volume growth rate of 1.6 ⫻ 10⫺10 m3/sec, we can calculate an energy cost of 71 J/carat, which is energetically much cheaper than diamond synthesis by one of the faster CVD processes, flame synthesis. D.

Heating of Surfaces Immersed in Liquids

The growth of some natural diamond (if not all of the larger crystals found in nature) is believed to occur in a state comprising solids immersed in

Laser-Assisted and Optical Pumping Techniques

351

liquids at extreme pressures and temperatures [38]. Nevertheless, synthesis of diamond by various deposition techniques has until recently largely focused on the gas phase alone. There have been a few attempts to grow diamond from a liquid phase or from solids submerged in the liquid phase [39–43]. Of these, Refs. 40–42 appear to be most intriguing in terms of potential application of lasers for diamond growth. In Cherian’s experiments [40–42], evidence for dissolution of diamond in a nickel crucible containing liquid NaOH at 1100 K and at atmospheric pressure is given, and more surprisingly, recrystallization of diamond back onto diamond surfaces upon ‘‘rapid’’ cool-down is also observed. As reported in Ref. 40, diamond crystallites consisting of chips and cleavages were dissolved in molten NaOH in a nickel crucible maintained at 1100 K in an oven. The crucible and its contents were then rapidly cooled to room temperature by removing them from the oven and placing in contact with a room-temperature metal surface or by cooling the crucible in room-temperature water. A cooling rate of ⬃1000⬚C/min was reportedly required for recrystallization to occur. Figures 15 and 16 show scanning electron microscopy (SEM) images of overgrowths that show the recrystallized diamond. Verification that the overgrowths are indeed diamond was carried out using several different characterization techniques, including microprobe Raman spectroscopy [40]. Reference [40] reports that careful experiments varying the quenching rate showed that the faster quenching rates yielded incomplete growth in the direction perpendicular to the substrate. Further, a metallic film of Ni was

Figures 15 and 16 SEM images of overgrowths that show the recrystallized diamond upon quenching an NaOH solution containing dissolved diamond particles in a nickel crucible. (From Ref. 41.)

352

Subramaniam and Aithal

found on the overgrowths of diamond, suggesting the role of an Ni-C complex as an intermediate facilitating the growth of diamond. Evidently, some of the carbon in the initial diamond seed material dissolved into the liquid NaOH, formed complexes with the dissolved Ni (from the walls of the Ni crucible), and upon cool-down recrystallized as diamond on remnant diamond seed surfaces. More insight into mechanisms could have been gained if isotopic labeling had been used. For instance, 13C diamond particles (used in the dissolution part) and 12C diamond seeds (used in the recrystallization part) may have been helpful in identifying the exact sources and targets of dissolution and recrystallization, respectively. Alternatively, a carbonaceous source other than diamond (graphite, for instance) could have been used in the dissolution phase of the experiment, while a diamond or silicon seed particle (thoroughly characterized prior to use) could have been introduced during the growth/recrystallization or cool-down phase of the experiment. More recently, Zhao et al. [43,44] have demonstrated that diamond can be grown on diamond seeds from (Ni)-C-H2O mixtures at a pressure of 1.4 kbar and temperature of 800⬚C.14 Experiments typically used mixtures of 3 wt% powdered nickel, 95 wt% glassy carbon, and 2 wt% 0.25 ␮m sized diamond in water (50–100 wt% of the glassy carbon) in a pressurized and heated vessel. Run durations were typically 50–100 hr, the shorter durations yielding diamond particles several microns in size. The presence of nickel was reportedly crucial in obtaining larger aggregates of particles with sizes ranging from tens of microns to 100 ␮m. The typical particle morphology observed in these experiments is shown in Fig. 17 and can be seen to resemble the morphology of diamond particles produced by General Electric’s commercialized high-pressure, high-temperature process. The use of lasers in such liquid-phase experiments may be awkward, although specific wavelengths do penetrate certain liquids. For example, the green line of the argon ion laser can be transmitted through water. Therefore, it may be possible to use lasers to transmit through a liquid and trigger heterogeneous reactions at a liquid-solid interface. However, this can be exploited only after growth mechanisms are clearly understood. Another 14

Laboratory experiments (see H. C. Noltimier and V. V. Subramaniam, Laboratory chemical vapor deposition [CVD] experiments applied to the geological formation and age of natural diamond, Poster Paper BTH-1 presented at Session 152 [Experimental Petrology] at the 1995 Geological Society of America Annual Meeting, New Orleans, November 9, 1995) also suggest the beneficial role of water vapor in accelerating growth of diamond by hot-filament CVD on minerals such as almandine and olivine, both found as impurities in natural diamond. These minerals can act as nuclei for diamond growth in C-H-O systems (and even at the low pressures of 30 torr typical in HFCVD). The authors also give compelling arguments indicating that most natural diamond may form during transit in the diamond diatremes instead of forming at depth, as commonly thought.

Laser-Assisted and Optical Pumping Techniques

353

Figure 17 Typical particle morphology observed in the experiments of Ref. 43. Note the resemblance to the morphology of diamond particles produced by General Electric’s commercialized high-pressure, high-temperature process.

means of possibly initiating diamond growth from a laser-excited liquid is discussed in Sec. V.C.

E.

Summary

Use of intense laser irradiation for pyrolysis (and subsequent reaction) of the gas phase and rapid heating of surfaces has been discussed in this section. Where rate of heating is necessary and crucial, lasers may be capable of providing a unique avenue for diamond growth. However, capital and maintenance costs remain prohibitive, especially in the energy cost per carat. Some reported techniques such as the QQC process, if reproduced, appear competitive with other existing techniques based upon reported growth rates and should be examined more carefully. Lasers have also been used to irradiate powders of carbon black and convert some of these into diamond. A relatively unexplored area is the use of lasers to drive heterogeneous reactions on surfaces immersed in liquids. Use of lasers to excite liquids themselves and drive liquid-phase reactions is discussed in Sec. V.

354

III.

Subramaniam and Aithal

ABLATION

The use of lasers to produce rapid heating is well established. When this rapid heating of solid surfaces leads to vaporization of the material at the surface, this is known as ablation. The resulting vaporized material often constitutes an activated gaseous phase that can be subsequently reacted and modified to yield a product or can condense onto another surface to produce a solid layer with different microstructural properties. This process, also known as laser physical vapor deposition, was shown schematically in Fig. 7. Intense laser energy (usually focused) is directed toward a target material, causing it to vaporize. The resulting vapor or plume is a gas comprising atoms and molecules from the target and is often ionized. This technique is often used to produce thin films on substrates by directing the electrically conducting plume toward a biased substrate. LPVD or ablation has been used for the deposition of thin films of different materials. Of particular relevance is the use of such an ablative technique for the deposition of diamond-like or amorphous carbon films [45–54]. In these experiments, graphite is used as a target (except in the case of Ref. 54, where Lucite targets were also used) and different highenergy lasers such as excimer lasers or Nd:YAG lasers or CO2 lasers are used. Intensities on the order of 108 W/cm2 are usually required. It is well established that LPVD is capable of synthesizing diamondlike carbon (DLC) or amorphous carbon films from ablated graphite targets. The term amorphous carbon refers to a material with properties distinct from those of graphite but lacking the long-range crystalline order characteristic of diamond. DLC films are known for their hardness approaching that of diamond. Many of the properties of amorphous carbon materials can be accounted for by the structural model of graphitic islands interlinked by small percentages of diamond-like sp 3 bonds [55]. Davanloo et al. [55] report the growth of ‘‘amorphic diamond’’ from a laser plasma source. These authors provide sufficient evidence for differences between their material and DLC (such as structural size on the order of 100–200 angstroms and band gap of 1.0 eV) to warrant distinguishing between their material and other established DLC properties. Such small crystallite dimensions usually pose great difficulty in acquiring Raman spectra so that this characterization technique is ineffective in unequivocally indicating the presence of diamond in the present case. Their technique is a variant of LPVD. An Nd:YAG laser (Q-switched repetition rate of 10 Hz, 15 nsec pulse width) was used to produce a peak intensity of 5 ⫻ 1011 W/cm2 at a graphite target, in conjunction with an electric discharge of 10 A maintained between the target and a rod electrode placed nearby. The special nature of this ‘‘amorphic diamond’’ film thus produced on a germanium window is evident in Fig.

Laser-Assisted and Optical Pumping Techniques

355

Figure 18 Transmittance curves for an amorphic diamond–coated Ge window (25 mm diameter) compared with a bare Ge window. The film was reportedly 0.3 ␮m thick with an index of refraction of 2.1. The authors point out the fact that absorption from CH bands at 2940 cm⫺1 is absent. (From Ref. 56.)

18. As can be seen from Fig. 18, the transmittance is higher for the amorphic diamond–coated Ge window compared with the bare Ge window and shows complete absence of the C — H absorption band at 3.4 ␮m. Reference 55 further reports a hardness value of 13 GPa, compared with the value of 77 GPa (at a temperature of 290 K) for diamond [37]. The amorphic diamond of Davanloo et al. was thus clearly subjected to numerous such tests, which enabled them to confirm that their films indeed exhibited diamond-like properties. Although LPVD and plasma-assisted LPVD have been successful in producing DLC and amorphic diamond films, there have only been a limited number of reports regarding growth of diamond using this technique [56,57]. Latsch and Hiraoka have shown that diamond can be produced by LPVD from targets of polymethyl methacrylate (PMMA) when a reactive mixture of oxygen and hydrogen (ratio of concentrations not reported) is present at pressures between 0.5 and 0.9 torr. The authors report the following critical parameters: target to substrate distance = 3–6 cm, substrate temperature = 600⬚C, ArF excimer laser (193 nm), fluence = 190 mJ/cm2, and substrate

356

Subramaniam and Aithal

bias = ⫺50 V [56]. The authors do not provide any information regarding any substrate pretreatment and report that ‘‘crystals started to appear after 10 minutes and thicker films were obtained for longer exposure times.’’ However, no SEM images are shown of the ‘‘films,’’ only of the isolated particles (see Fig. 19). Further, microprobe Raman spectra of the deposits (shown here in Fig. 20) indicate the presence of both ‘‘D’’ (at 1330 cm⫺1) and ‘‘G’’ (1597 cm⫺1) peaks. These peaks are usually indicative of graphite. Although the ‘‘D’’ peak usually occurs around 1343 cm⫺1, it can shift in some samples to the location where the diamond peak occurs. Therefore, caution must be exercised in interpreting Raman spectra especially when a broad peak near 1332 cm⫺1 and the ‘‘G’’ peak of graphite are simultaneously present. As an illustrative example, a Raman spectrum of a carbonaceous deposit produced from an oxyacetylene flame is shown in Fig. 21 [17]. The same film, exposed to atomic hydrogen (produced by a hot filament), removed the carbonaceous deposit, as can be seen from the Raman spectrum shown in Fig. 22 [17]. Given the uncertainty introduced by this fact, it is important to expose a film (or crystals) suspected to be diamond to H atoms at temperatures near 800⬚C in order to be sure that the deposit is indeed diamond. Subsequently, Polo et al [57] used ArF and KrF excimer laser ablation of graphite targets in hydrogen at 0.76 torr and reported growth of diamond particles 1–20 ␮m in size. The deposits were verified as diamond by Raman spectroscopy. The ablative method can be subject to problems of reproducibility. The composition of the target is often a critical parameter, as are the substrate temperature, position of the substrate relative to the target, substrate bias,

Figure 19

SEM image of isolated particles produced by LPVD. (From Ref. 56.)

Laser-Assisted and Optical Pumping Techniques

357

Figure 20 Microprobe Raman spectrum of the deposits from Ref. 56 showing the presence of both ‘‘D’’ (at 1330 cm⫺1) and ‘‘G’’ (1597 cm⫺1) peaks. This is indicative of disordered graphite rather than diamond.

Figure 21 Raman spectrum of a carbonaceous deposit produced from a flame. (From Ref. 17.)

358

Subramaniam and Aithal

Figure 22 The film whose Raman spectrum is shown in Fig. 21 was exposed to atomic hydrogen (produced by a hot filament) [17]. This figure shows the Raman spectrum after exposure to atomic hydrogen. Note the disappearance of the feature at 1334 cm⫺1, which clearly demonstrates that the original film was not diamond.

and laser pulse energy. Of these, variability in the composition of the target has the largest influence on the resulting film or deposits. In some cases, ablation of the target can be accompanied by gross ejection of particulate solid material from the target that is then incorporated in a growing film. Based upon the scant information presently available regarding growth rates in LPVD, it is not possible to estimate the energy cost per carat for this process. Further research in this area is warranted, especially with regard to use of different combinations of target material and reactive gases.

IV.

PHOTOLYSIS

The process of photolysis, or the use of energetic photons to fragment a polyatomic molecule, was introduced in Sec. I. Usually, such processes require the presence of intense laser irradiation to initiate interaction between a molecule and one or more photons. As explained in Sec. I, this multiphoton interaction between molecule and photons may be sequential (i.e., the molecule may sequentially absorb each photon) or simultaneous, ultimately lead-

Laser-Assisted and Optical Pumping Techniques

359

ing to fragmenting of the molecule into smaller molecular or atomic fragments. This generally falls within the area of photochemistry [4,7,8]. The wavelength range typically used in photochemistry varies from the vacuum UV (below 200 nm) to the microwave region (1–10 cm). This range is chosen because of the typical absorption characteristics of molecules. Most small (three or four atom) polyatomic molecules have bond dissociation energies on the order of 1–10 eV. This corresponds to photon wavelengths in the UV-visible range. Light in this wavelength range is usually absorbed via electronic transitions in molecules. Wavelengths in the IR usually result in vibrational excitation because molecular vibrational states differ by energies on the order of 0.1 eV or less. Radiation in the far IR to microwave regions (10 ␮m to 10 cm) usually corresponds to excitation of the rotational energies of molecules. The wavelength range applicable to photochemistry is further restricted by the availability of materials for optical elements such as windows, lenses, and mirrors. Typical wavelength ranges and other properties for common optical materials are listed in Table 2. Table 3 lists some useful conversion factors because different units are used for convenience. For instance, when dealing with UV or visible light, wavelength units in nanometers are typically used. When dealing with radiation in the near IR to the far IR, units of microns (␮m) are used for wavelength and wavenumbers or cm⫺1 are used for frequency. In the microwave regions, wavelength units of millimeters or centimeters are used. It is important to delineate the difference between reactions occurring in photochemistry, thermally induced reactions (i.e., those caused by heating), and reactions initiated by extremely short-wavelength radiation such as x-rays or gamma rays. In thermal reactions (such as pyrolysis), molecules are dissociated or reacted from their ground electronic manifolds by vibrational excitation usually fed by energy from the external modes (i.e., translation and rotation). In contrast, photochemical reactions typically involve electronically excited states. In reactions initiated by x-rays and gamma rays, the route is often via ionization. In other words, molecules are ionized (i.e., charged by stripping them of one or more electrons) and subsequently react with other ions or neutral molecules. The difference between thermally induced reactions and photolytic reactions is apparent in the following illustration provided in Ref. 4. Consider the thermal decomposition of ethane: C2H6 ⫹ M → CH3 ⫹ CH3 ⫹ M where M is any other collision partner. Here, translational energy is transferred from M to C2H6 by impact, resulting in an intermediate15 fol15

The intermediate (C2H6)† is typically vibrationally excited and is known as a transition state. In contrast, the intermediate in photolytic reactions is usually an electronically excited state.

Table 2

Typical Characteristics of Optical Materials

Material Al2O3 (sapphire) BaF2 CaF2 Ge LiF MgF2 KBr KCl KF KI Si SrF2 SrTiO3 TiO2 (rutile) Source: Ref. 59.

Transmission Water solubility Melting range (g/100 mL of Knoop Density point (␮m) Index of refraction H2O) hardness (g/cm3) (⬚C) 0.15–5.5 0.5–13 0.12–10 1.8–22 0.12–7 0.11–8 0.21–37 0.18–25 0.16–15 0.27–42 1.2–15 0.12–11 0.395–5.0 0.42–5.3

1.768 1.475 at 0.546 ␮m 1.424 at 2 ␮m 4.116 at 2 ␮m 1.379 at 2 ␮m 1.378 1.538 at 2 ␮m 1.475 at 2 ␮m 1.39 at 0.257 ␮m 1.631 at 2 ␮m 3.458 at 2 ␮m 1.439 at 0.546 ␮m 2.409 2.613–2.909

Insoluble

2200

3.97

2053

0.12 1.6 ⫻ 10⫺3 Insoluble 0.27 7.6 ⫻ 10⫺3 53.48 34.7 92.3 127.5 Insoluble 0.011 Insoluble Insoluble

82 158 780 102 415 7 9

4.89 3.18 5.323 2.635 3.15 2.75 1.984 2.48 3.12 2.33 4.24 5.175 4.26

1280 1360 937.4 842 1266 730 776 846 682 1410 1450 2080 1825

5 1150 130 595 700

Cleavage plane

Thermal expansion (⫻10⫺6 per ⬚C) 4.5

(111) (111) (100) Poor a or c (100) (100) (100) (100)

18.4 6.1 35 9.2–13.4 38.4 36.6 32 43 3

(111) 10.6 C-axis: 9.943

Laser-Assisted and Optical Pumping Techniques Table 3

361

Conversion Factors for Energy

1 eV

1 cm⫺1

␭ (angstroms)

1.6021 ⫻ 10⫺12 ergs/molecule 8065.73 cm⫺1 23.061 kcal/mole 96.487 kJ/mole 2.859 cal/mole 11.962 J/mole 1.9863 ⫻ 10⫺16 ergs/molecule 1.2398 ⫻ 10⫺4 eV/molecule 285915/␭ kcal/mole 12398/␭ eV/molecule

lowed by decomposition into CH3 molecules. In contrast, the primary photochemical channel for photolytic decomposition of ethane is C2H6 ⫹ h␯ → C2H4 ⫹ H2 proceeding via absorption of a photon by C2H6, leading to an electronically excited intermediate. As can be seen from this illustration, the products resulting from thermal reactions and photolytic reactions can be very different, showing how different products arise from different transition states. A photon absorbed by a molecule does not necessarily lead to a chemical change. The molecule certainly becomes excited upon absorbing the photon, but there are other loss processes that inhibit chemical change. For instance, the excited molecule could simply radiate the photon away by spontaneous emission. The molecule could also suffer a collision with another molecule while in this excited state. This may result in energy transfer to (into either internal or external modes of) the collision partner. Such processes are considered as losses because they do not result in dissociation of the target molecule or in chemical reaction, which is the primary goal of photolytic processes. In that sense, these loss mechanisms lead to inefficiencies in the photolytic process. The efficiency of a given photolytic process is qualitatively accounted for by the quantum yield, ␩q. It is defined as [60,61] ⌽q =

number of molecules undergoing the process number of photons absorbed by the system

(9)

In photochemistry, 1 mole (i.e., Avogadro’s number of 6.022 ⫻ 1023) of photons is called an Einstein. Typically, an overall reaction pathway may consist of several primary or elementary pathways and a quantum yield can be defined for each primary channel i:

362

Subramaniam and Aithal

⌽i =

number of excited molecules that proceed via elementary channel i number of photons absorbed by the system (10)

If moderate or low light intensities are involved, the probability that an excited molecule will absorb more than one photon during its short lifetime is small. For these conditions,

冘

⌽i = 1

(11)

i

where i represents the ith primary photochemical pathway. Equation (11) is referred to as the Stark-Einstein law and holds only for low or moderate light intensities. Further, though ⌽i ⱕ 1, the overall quantum yield for a product, ⌽oq, can be much larger than 1 as is the case for chain reactions. The earliest attempt to use a laser to photodissociate a carbonaceous precursor and produce diamond met with failure [13]. Although success was reported initially [12,13], subsequent characterization of the growth material showed the deposit to be heat-treated carbon black [13]. At the time, both CH4 (methane) and C2H2 (acetylene) were suspected of being growth species and Kitahama et al. used 193-nm radiation from an ArF excimer laser to photodissociate C2H2 diluted (from 1 to 30%) with hydrogen. The pressure was kept in the range of 25–30 torr, and substrate temperatures in the range 400–800⬚C were explored. After growth durations of 2 hr during which time C2H2 was photolyzed using 193-nm pulses (100 Hz repetition rate and 50 mJ/pulse), growth of particles (later shown to be nondiamond carbon) was observed along scratches induced by substrate pretreatment.16 The authors do not explain why higher hydrogen concentrations were not used in their mixtures, especially given the fact that the photochemistry of acetylene at 193 nm is well known [62]. In fact, the photolysis products of pure C2H2 have been reported to yield diacetylene, ethylene, hydrogen, and a polymer [62]. It is quite possible then that higher hydrogen concentrations or presence of oxygen in the precursor gases would have yielded diamond growth in these early experiments. The initial claims of diamond growth by photolysis of acetylene were subsequently retracted [13]. Despite the initial setback for photolysis, Goto et al. [63] succeeded in producing diamond from a gaseous mixture of hydrogen flowing at 1000 sccm and carbon tetrachloride (CCl4) flowing at 1–10 sccm and irradiated by a focused ArF excimer laser beam. The pressure ranged from 6 to 50 torr, the repetition rate was 100 Hz, the pulse energy was 80–200 mJ, and the average power was 5–12 W. The beam was focused using an F:300 16

Substrates were pretreated as is common with other growth techniques, in this case using no. 1800 diamond powder to enhance nucleation and accelerate growth.

Laser-Assisted and Optical Pumping Techniques

363

cylindrical lens, and the authors report that configurations where the laser irradiated the substrate and was parallel to the substrate, were separately explored. The latter arrangement is shown schematically in Fig. 23. The authors report the formation of an amorphous carbon film after deposition durations ranging from 30 min to 3 hr. However, when the hydrogen stream was predissociated using a microwave discharge or a resistively heated hot filament, the authors report growth of a 1- to 3 ␮m-thick diamond film on silicon substrates heated to 450⬚C. The films were characterized by Raman spectroscopy, which revealed a discernible peak at 1333 cm⫺1 confirming the presence of diamond. The crystal structure was further confirmed by electron diffraction. This was the first reported instance of diamond growth using a laser-based process and one of the first to demonstrate that substrate temperatures did not need to be as high as 800⬚C. The authors report on the importance of atomic hydrogen and methyl radicals for diamond growth. This idea has received much support in reviews of diamond growth mechanisms [64–66]. Following the success of Goto et al., Tyndall and Hacker [67] reported on photolysis of over 20 compounds of organic precursors using focused (cylindrical lens with a focal length of 100 mm) KrF excimer laser radiation. Laser fluences of 150 mJ/pulse/cm2 were reportedly used along with substrate temperatures between 25 and 400⬚C, while chamber pressures were maintained between 1 and 5 torr. The authors report that temperatures in excess of 400⬚C resulted in dramatically decreased growth rates. The organic precursors were admitted into the growth chamber by passing a carrier gas (reportedly argon, helium, or hydrogen) over the reservoir containing the organic precursor. Of the organic precursors examined, the authors report that aliphatic carboxylic acids such as CH3CO2H (acetic acid) and

Figure 23 Schematic of experimental arrangement used in Ref. 63 to produce diamond films from photolysis of CCl4/H2 mixtures.

364

Subramaniam and Aithal

H2C(CO2H)2 (malonic acid) yielded films containing diamond and disordered graphite after durations of 1.5 hr. The malonic acid had to be heated to 50– 70⬚C to provide sufficient vapor pressure to yield reasonable growth rates, but acetic acid could be used at room temperature. The authors do not specify which carrier gas or gases were used with these precursors. The Raman spectra of films obtained from these organic precursors display broad features at 1335 cm⫺1, which the authors claim to be diamond. However, this should be taken as insufficient verification, based upon the illustrative spectra shown in Figs. 21 and 22. As discussed earlier, this film shows the presence of a feature at 1332 cm⫺1, but this feature disappeared when the film was treated in atomic hydrogen (from a heated tungsten filament) after growth. Diamond films are usually quite resistant to etching by atomic hydrogen, and such disappearance of the well-known diamond feature in the Raman spectrum is not observed in true diamond films. On the other hand, graphitic films or films with disordered graphitic carbon show significant changes in their Raman spectra after postgrowth treatment in atomic hydrogen. Therefore, the spectra reported in Ref. 67 cannot be taken as irrefutable evidence of the presence of diamond. A postgrowth etch treatment of the film would have served as another means of verifying their claim. However, the authors of Ref. 67 do report on additional characterization of their deposits, such as electrical resistivity measurements and characterization by Auger spectroscopy. The electrical resistivity of these samples was greater than 109 ohm-cm (closer to the value for diamond, i.e., 1016 ohm-cm, than for graphite, i.e., 10⫺2 ohm-cm). Such additional characterization of the deposits is encouraging and bolsters the claims reported in Ref. 67. No SEM images of the deposits are shown, and therefore it is not clear whether the yield is a film covering a certain region of the substrate or consists of particles distributed over the substrate. In some instances, the choice of the carbonaceous precursor dictates the choice of growth technique. This is true in the case of growth of diamond from mixtures containing CO [68–70]. Use of CO as the carbonaceous precursor for diamond growth is of particular interest when isotopically pure diamond is desired. Diamond synthesized from a hydrocarbon source such as methane, but with a 13C content of 0.1% or less, is much more thermally conductive than diamond grown from CH4 with the natural abundance of 13 C of 1.1% [71]. Now, isotopically pure 12C and 13C are commercially produced as CO, and more processing steps are required to convert them to isotopically pure hydrocarbon sources. Consequently, 12CO will be cheaper than using 12CH4 in such an instance. Furthermore, because CO cannot be used in some conventional diamond growth systems such as HFCVD or oxyacetylene flame synthesis, growth options using CO as the precursor are

Laser-Assisted and Optical Pumping Techniques

365

limited. Consequently, plasma CVD or laser methods must be used when CO is the precursor. Growth of particles of diamond on seeded unheated monocrystalline silicon [(001) and (111)] substrates has since been reported from CO/H2 gas mixtures by photolysis of CO [68–70]. The photolysis of CO was accomplished using focused ArF excimer laser irradiation (the reader is referred to Ref. 70 for experimental details). Several orientations of the laser relative to the substrate were explored, as shown in Fig. 24, while the gas flow was directed perpendicular to the laser beam and toward the substrate. A CO flow of 0.7 sccm and H2 flow of 99.3 sccm were used, and the chamber pressure was maintained at 8 torr. Prior to growth, the substrates were thoroughly characterized using SEM imaging and energy-dispersive x-ray spectroscopic (EDS) analysis to ensure that no diamond particles remaining from the abrasive pretreatment used were present. Further, in order to distinguish seed particles from growth particles, the authors report using commercially available diamond grit (0–0.5 ␮m in size). The substrates were then thoroughly cleaned ultrasonically in acetone, rinsed with deionized water, and characterized before growth using SEM imaging and Raman spectroscopy. After run durations of 4 hr, irregularly shaped but faceted particles of size 5–10 ␮m of the type shown in Fig. 25 were observed. The morphology of these particles does not resemble that of diamond crystallites usually observed in other diamond processes (for example, see Fig. 28), suggesting that pure growth is involved here as opposed to the growth and etch processes simultaneously occurring in other CVD processes. Microprobe Raman spectra of these particles (see Fig. 26 showing a representative spectrum) exhibit a beautifully sharp diamond line at 1332 cm⫺1, with no discernible presence of graphitic or amorphous carbon. This work represents the first instance in which diamond growth is observed on unheated substrates by any known process, despite the slow growth rate. In some of their laser photolysis experiments, the H2 stream was dissociated using a tungsten filament resistively heated to about 2000⬚C [68–70]. As expected, the coverage of the particles is substantially increased, as shown in Fig. 27. The morphology of diamond particles produced by photolysis of CO in a hydrogen bath is significantly different from that of crystals normally grown by other CVD techniques. They are irregular in shape and resemble the 0- to 0.5-␮m diamond grit used for pretreatment of the substrates but are larger (5–10 ␮m in size). The irregularity in shape suggests the following. Because the resolution of SEM imaging is about 50 nm, it is likely that seed particles smaller than 50–100 nm went undetected. These small particles then would have acted as seeds around which diamond could grow, thus explaining the increase in size observed after growth. The typical cubooctahedral shapes (see Fig. 28) and other morphologies usually seen in other

366

Figure 24

Subramaniam and Aithal

Schematic of LCVD experiment. (From Ref. 70.)

Laser-Assisted and Optical Pumping Techniques

Figure 25 surface.

367

SEM images of diamond particles grown by LCVD on unheated

Figure 26 Typical Raman spectrum of diamond particles synthesized by LCVD on unheated surfaces.

368

Subramaniam and Aithal

Figure 27 SEM image of increased coverage obtained by LCVD, in the presence of a predissociated hydrogen stream.

Figure 28 Typical cubo-octahedral morphology of diamond particles synthesized by hot-filament CVD.

Laser-Assisted and Optical Pumping Techniques

369

CVD systems are a result of growth-etch cycles. In conventional CVD systems, the presence of (relatively) large amounts of atomic hydrogen accelerates diamond growth by passivating the growing surface while removing nondiamond carbon and partially etching diamond. An excellent illustration is provided by the interrupted temporal history of growth of diamond around a 100-␮m seed particle from a CO/H2 gas mixture in a microwave discharge reported in Ref. 72 and reproduced here in Fig. 29. As can be seen, the early stages of growth on the seed crystal give the appearance of an irregular shape for the particle. The morphologies usually observed in other conventional CVD systems begin to become apparent in this case of a CO/H2 plasma only after a long period of time (50 hr) has elapsed [72]. The irregular morphology observed in Refs. 68–70 can then be understood because their experiments spanned only 4 hr. The laser photolysis experiments of Refs. 68–70 provide an excellent example of the use of lasers to drive reactions selectively, specifically those

Figure 29 An excellent illustration is provided by the interrupted temporal history of growth of diamond around a 100-␮m seed particle from a CO/H2 gas mixture at 45 torr in a microwave discharge reported in Ref. 72. (a) The seed crystal, (b) after 10 hr, (c) after 20 hr, and (d) after 50 hr of growth.

370

Subramaniam and Aithal

that result in diamond formation while formation of the graphitic phase is suppressed. The choice of the ArF excimer laser as a source is dictated by the well-known multiphoton process that dissociates CO into C (31P) and ground state O(3P) atoms [73]. Once atomic carbon and oxygen are formed, they react as radicals with the surrounding molecular hydrogen resulting in the formation of desirable CHx species. It is important to note that if H2 were not present, copious amounts of C2, soot, and other nondiamond carbon would result [70]. This channel of C2 production from CO will be addressed further in Sec. V. More recently, it has been shown that optical pumping of CO/Ar gas mixtures with trace amounts of Fe or Ni catalyst results in growth of single-walled carbon nanotubes (see E. Ploenges et al. Carbon Nanotube Production in CO Laser Pumped Carbon Monoxide Plasmas. Paper AIAA2001-0651, presented at the 39th AIAA Aerospace Science Meeting and Exhibit, Reno, Nevada, January 8–11, 2001). In order to understand the laser-driven chemistry in the experiments reported in Refs. 65–67 better, we examine a simple chemical kinetics model of the gas-phase reactions expected to occur in the irradiated CO/H2 mixture. To enable a simple solution, the following assumptions are made. First, every CO molecule in the focal volume is assumed to have dissociated. Second, diffusion and convection are neglected in order to simplify the solution. From the flow rate of CO of about 1 sccm reported in Ref. 68, total pressure of 8 torr, and assuming a local gas temperature of 350 K, the initial number density of CO can be calculated to be ⬃2.2 ⫻ 1021 m⫺3. Similarly, the initial number density of molecular hydrogen based on its flow rate of about 99 sccm is 2.2 ⫻ 1023 m⫺3. Knowing the energy per photon,17 1.0 ⫻ 10⫺18 J and taking the energy per pulse as 50 mJ, the total number of photons in the laser-irradiated volume can be calculated. Assuming this volume to be about 1 mm3, the number density of photons is then approximately 4.8 ⫻ 1025 m⫺3. Because this is much bigger than the initial CO concentration of 2.2 ⫻ 1021 m⫺3 and assuming no more than three photons are required to dissociate a CO molecule, the first assumption is justified. Thus, the initial gas composition within the laser irradiated region can be taken to be nCO = 0 m⫺3, nH2 = 2.2 ⫻ 1023 m⫺3, and nC = nO = 2.2 ⫻ 1021 m⫺3. However, diffusion and flow will serve to replenish the laser-irradiated region with CO and H2 reactants. A more representative and realistic estimate of the gas composition can be obtained by scaling the initial concentrations of atomic carbon and oxygen with respect to the ratio of the irradiated volume to the

17

Because an ArF excimer laser was used in these experiments, the energy of a photon of wavelength 193 nm is (6.626 ⫻ 10⫺34 J-sec) ⫻ (3 ⫻ 108 m/sec)/(193 ⫻ 10⫺9 m) or 1.03 ⫻ 10⫺18 J.

Laser-Assisted and Optical Pumping Techniques

371

gas volume adjacent to the substrate. Taking the volume adjacent to the substrate to be about 1 cm3 and the irradiated volume as about 1 mm3, the initial composition for each successive interval is: nCO = 2.2 ⫻ 1021 m⫺3, nH2 = 2.2 ⫻ 1023 m⫺3, and nC = nO = 2.2 ⫻ 1018 m⫺3. With these initial values, the set of rate equations of the form dni = f (ni , nj ) dt

for all j

(12)

describe the evolution of concentration of species i, where ni and nj represent the number densities of the species listed in the reactions given in Table 4.

Table 4 Elementary Reactions and Rate Constants Evaluated at 300 K, Used to Simulate Gaseous Environment Existing Within the Laser-Irradiated Volume Adjacent to the Substrate in the Experiments Reported in Refs. 68–70 Reaction C ⫹ H2 → CH2 C ⫹ H2 → CH ⫹ H H2 ⫹ O → OH ⫹ H C ⫹ CO → C2O CH2 ⫹ H2 → CH3 ⫹ H CH2 ⫹ CO → C2H2 ⫹ O CH2 ⫹ CO → CH2CO H2 ⫹ C2O → CH2 ⫹ CO CH ⫹ H2 → CH3 CH ⫹ H2 → CH2 ⫹ H C2H ⫹ O → CH ⫹ CO CH ⫹ CO → HCO ⫹ C H ⫹ CO → HCO H ⫹ CO ⫹ H2 → CHO ⫹ H2 OH ⫹ CO → H ⫹ CO2 CH2 ⫹ CH2 → C2H2 ⫹ H2 CH3 ⫹ CH3 → C2H6 CH2 ⫹ CH3 → C2H4 ⫹ H CH2 ⫹ CH2 → CH3 ⫹ CH CH ⫹ H → C ⫹ H2 CH3 ⫹ H → CH2 ⫹ H2 C2H2 ⫹ O → CH2 ⫹ CO CO2 ⫹ O → CO ⫹ O2 CO2 ⫹ H → CO ⫹ OH CH3 ⫹ H → CH4

Rate constants 7.11 5.0 7.05 6.31 3.2 2.14 1.0 7.0 2.66 2.66 3.0 2.09 6.92 1.15 1.32 4.98 4.06 6.64 2.30 4.98 3.2 2.14 1.39 1.87 3.32

⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻

10⫺44 m3/sec 10⫺151 m3/sec 10⫺24 m3/sec 10⫺44 m3/sec 10⫺24 m3/sec 10⫺63 m3/sec 10⫺21 m3/sec 10⫺19 m3/sec 10⫺17 m3/sec 10⫺17 m3/sec 10⫺17 m3/sec 10⫺17 m3/sec 10⫺17 m3/sec 10⫺46 m6/sec 10⫺19 m3/sec 10⫺17 m3/sec 10⫺17 m3/sec 10⫺17 m3/sec 10⫺23 m3/sec 10⫺17 m3/sec 10⫺27 m3/sec 10⫺19 m3/sec 10⫺55 m3/sec 10⫺35 m3/sec 10⫺16 m3/sec

372

Subramaniam and Aithal

The reactions listed in Table 4 are not meant to be exclusive but represent only one set of reactions thought to be important for the conditions reported in Refs. 68–70. Equations (12) then represent a set of coupled nonlinear equations for ni that can be solved numerically using a method such as that described in Ref. 75. In these calculations, the ArF excimer laser pulses are assumed to remain off for 50 msec (corresponding to a 20-Hz repetition rate) and have pulse durations of 20 nsec. Each pulse interval is hereafter defined as the pulse duration of 20 nsec plus the laser-off time of 50 msec. Compositions of all species (except CO, H2, C, and O) are updated for each interval from values calculated at the end of the previous interval. The compositions after 2 sec or 40 such intervals thus calculated are shown in Figs. 30–32. It can be seen that the carbonaceous species present in order of abundance are HCO, C, CO, CH3, CH2CO, CH4, CH2, C2H6, C2H4, C2H2, and CH. Among these, the least credible is the unusually high amount of atomic carbon present. This must be regarded as a fictitious result because

Figure 30 Compositions of various species calculated using the reactions in Table 4 are shown here after 2 sec or 40 intervals.

Laser-Assisted and Optical Pumping Techniques

373

Figure 31 Compositions of various species calculated using the reactions in Table 4 are shown here after 2 sec or 40 intervals.

recombination of C to form either C2 or CO was not included in our reaction set listed in Table 4. Further, the absence of nondiamond carbon either on the SEM images or in the Raman spectra is conspicuous in the experiments reported in Refs. 68–70. This strongly suggests that C2 species must not have been present in abundance in the gas phase because these are known to produce graphitic carbon. Species capable of removing nondiamond carbon or atomic hydrogen or atomic oxygen from CHxOy complexes bound to the growing diamond surface are also present. In order of abundance, they are HCO, C, O, OH, H, CH3, CH2CO, CH2, and CH. Of these, only the high concentration of C should be ignored for the reasons already mentioned. Given the state of present understanding regarding mechanisms of diamond growth [64], we can attempt to make some assessment regarding the identities of possible precursors important in diamond growth by photolysis of CO.

374

Subramaniam and Aithal

Figure 32 Compositions of various species calculated using the reactions in Table 4 are shown here after 2 sec or 40 intervals.

Suppose the linear growth rate (i.e., growth rate normal to the substrate surface) is gr. Then, the number of carbon atoms added per unit time per unit area of the growing surface by the carbonaceous precursor, n˙ ⬙C , is given by n˙ C⬙ =

gr␳d mC

(13)

where gr is expressed in units of m/sec, ␳d is the mass density of diamond (3515 kg/m3), and mC is the atomic mass of carbon in kg (12 ⫻ 1.6605 ⫻ 10⫺27 kg). At the very least, for growth to occur, the flux of the carbonaceous precursor must exceed this quantity n˙ ⬙C for growth to continue. Thus, we have 1 ¯ ⱖ n⬙ ˙C nXC 4

(14)

where nX is the gas phase number density of the carbonaceous precursor,

Laser-Assisted and Optical Pumping Techniques

375

¯ = (8kBT/mX␲)1/2 is the random thermal speed. In the expression for and C ¯ kB is Boltzmann’s constant (1.3806 ⫻ 10⫺23 J/K), T is the gas temperature C, in the neighborhood of the growing surface, and mX is the molecular mass of the carbonaceous precursor. In using the expression for the random thermal flux in Eq. (14), it has been assumed that the velocity distribution function of species X is described by the Maxwell-Boltzmann distribution. Using Eqs. (13) and (14), the order of magnitude of the minimum concentration nX necessary to explain an observed growth of gr can be determined. For the specific conditions of the CO photolysis experiments reported in Refs. 68–70, let us take the growth rate to be 1 ␮m/hr, gas temperature T = 300 K, and mX to be on the order of 2 ⫻ 10⫺26 kg. For these values, Eq. (14) yields nX ⬇ 1018 m⫺3. Expected values of the gas composition for times on the order of an hour can be determined from calculated results such as those shown in Figs. 30–32 and compared with this minimum value of nX. Such an order of magnitude estimate shows that candidate growth species in diamond growth by photolysis of CO (excluding C and CH4) are HCO, CO, CH3, and CH2CO. Of these, CO can be eliminated as a likely growth precursor because it is an extremely tightly bound molecule and difficult to dissociate. CH3 emerges as a strong contender as growth species based upon what is known regarding growth mechanisms in other CVD systems [64]. However, the abundant presence of HCO and CH2CO cannot be overlooked, and these may well play a role in diamond growth by photolysis of CO. The other somewhat remote but possible candidate growth species is CH2, as its concentration is lower than that of CH3 only by about two orders of magnitude. Other species such as C2H6, C2H4, C2H2, and CH can be safely eliminated because they are present in substantially smaller amounts. Although the foregoing analysis yields some insight into the identity of the key carbonaceous precursor for diamond growth, it must be emphasized that it is inexact and can be viewed only as an order of magnitude estimate. Nevertheless, such analysis is useful for inferring trends and may suggest process improvements that lead to an increase in yield or growth rate. In this section, we have examined existing photolytic techniques for diamond synthesis. Of significance is the fact that growth of diamond particles has been realized under conditions where the substrate is not heated at all (see Refs. 68–70 on photolysis of CO), and growth of diamond films has been reported at temperatures as low as 450⬚C (see Ref. 63 on photolysis of CCl4). Given the strong interest in lowering growth temperatures [76], these represent significant advancements in the state of the art. It is also of significance to note that in the photolysis experiments involving CO there is evidence that selectivity has been achieved because nondiamond carbon was not codeposited. The absence of nondiamond carbon is conspicuous, and the microprobe Raman spectra show a particularly sharp diamond line

376

Subramaniam and Aithal

at 1332 cm⫺1. All these photolysis techniques, however, appear to require predissociation of hydrogen in order to increase growth rates. Of these, the photolytic decomposition of CCl4 in a predissociated hydrogen stream appears closest to commercial realization. An estimate of cost can be made based upon the results reported in Ref. 63. Using their reported growth rate of 1 ␮m/hr but over an area of 1 cm2 (the actual deposition area was not disclosed in Ref. 63) and an input average power of 10 W, the energy cost is approximately 2 ⫻ 107 J/carat for a yield of 4.9 ⫻ 10⫺7 carats/sec.

V.

OPTICAL PUMPING

A.

Introduction

Optical pumping refers to the excitation of the internal modes of atoms or molecules. Light, which is not necessarily coherent, may be used to excite rotational, vibrational, or electronic states of atoms and molecules by what is known as resonant absorption. In contrast to the relatively higher power, multiphoton processes introduced in the earlier sections of this chapter, here we focus on single-photon resonant absorption. The term resonant absorption refers to the fact that there is an excellent coincidence between the frequency of the exciting radiation or light and the energy gap (divided by Planck’s constant) associated with a particular rotational, vibrational, or electronic transition. Further, in contrast to multiphoton processes, optical pumping or resonant single-photon processes are characterized by much lower intensities. Coherence of the exciting radiation is therefore not strictly necessary, although they are needed in practice in order to exceed threshold intensities for absorption. Such threshold absorption levels exist because there are losses due to quenching by collisional or radiative processes that must be overcome. Internal modes of diatomic molecules were introduced briefly in Sec. I. As discussed earlier, these comprise rotational, vibrational, and electronic means of energy storage for molecules. At equilibrium, the average energy in these modes can be described by a single temperature, T. Further, the equipartition theorem18 from statistical thermodynamics tells us that each squared term in the Hamiltonian (corresponding to the degrees of freedom associated with each mode of molecular motion) contributes on average kBT/ 18

The equipartition theorem states that each squared term in the Hamiltonian contributes kT/2 to the average energy. Examples of squared terms are kinetic energy of a particle, which scales as velocity squared and potential energy in a linear spring, which scales as displacement squared. In Hamilton’s formulation of classical mechanics, velocity and displacement are examples of canonical independent variables.

Laser-Assisted and Optical Pumping Techniques

377

2 to the average energy of a molecule. For example, there are three degrees of freedom in translation for an atom or molecule. Each appears as a squared term in the kinetic energy arising from the sum of the squares of the three components of velocity (i.e., 12 mC 2x ⫹ 12 mC 2y ⫹ 12 mC 2z ). Hence, the contribution from the translational motion to the total energy of the particle is (3/2)kT. Similarly, for a diatomic particle, the contribution from the rotational modes is kT (corresponding to two degrees of freedom in rotation), and the contribution from the vibrational mode is kT (corresponding to the potential energy and kinetic energy associated with the single degree of freedom in vibration). The essential point here is that the energy in each of these modes is dependent on a single quantity, the gas temperature. When the gas is out of thermodynamic equilibrium, the energy in these modes can be different and is no longer identified by a single temperature. This is of relevance to optical pumping because the resulting situation is one where the energy in the vibrational and electronic modes cannot be described solely by the gas temperature, which characterizes the rotational and translational modes. Equilibration of the energies in the various modes of molecules occurs by collisions. As explained in Sec. I, rotation and translation equilibrate within a few collisions. However, for diatomic molecules such as CO, NO, and N2, it takes many collisions before the vibrational modes equilibrate with the external modes (i.e., rotation and translation). The rate at which energy is exchanged by the various modes via collisions can be described by a rate coefficient, analogously to chemical processes. For example, one can speak of a rate coefficient for vibration-vibration (VV) energy transfer, vibration-translation (VT) energy transfer, vibration-rotation (VR) energy transfer, vibration-electronic (VE) energy transfer, and so on. Thus, each species under such nonequilibrium conditions should be thought of as a specific molecule in a particular vibrational state of an electronic manifold. An example is provided by the energy level diagram of CO shown in Fig. 33, which shows the ground electronic state of CO (X1⌺g⫹) along with its vibrational levels (denoted as the many horizontal lines within the curve). Also shown in Fig. 33 are other electronic states along with their vibrational levels. The rotational levels are not shown for clarity, but they would appear as additional horizontal lines in between successive vibrational levels. For CO at 300 K, the characteristic time for VT transfer19 is expressed as ⬃10 sec-atm for the v = 1 level, while the characteristic time for VV energy transfer is ⬃0.1 msec-torr. At 1 atm, this yields a characteristic time of 10 19

This VT rate is for CO-CO collisions. The corresponding values for CO-He and CO-Ar VT relaxing collisions depleting v = 1 and populating v = 0 are 4 ⫻ 10⫺3 sec-atm and 103 secatm, respectively.

378

Figure 33

Subramaniam and Aithal

Energy level diagram for CO.

sec for VT relaxation for the v = 1 level while its characteristic time for VV exchange is 76 msec. This means that at any given pressure in the collisiondominated regime, vibrationally excited CO will tend to retain energy in vibration for relatively long durations if VT processes are ineffective. Chemical reactions proceed from vibrationally excited or electronically excited states because these promote the necessary breaking of chemical bonds. Optical pumping provides a means of producing vibrationally or electronically excited molecules. This technique has been pioneered by Rich and coworkers [2,77–79], Brechignac et al. [80], and Urban and coworkers [81,82] for studying various nonequilibrium molecular energy transfer processes with applications to electrical discharges, flames, and supersonic expansions. In this section, we discuss in detail the use of this novel technique for diamond synthesis. Although other laser sources and precursor molecules can be targeted using the technique presented in this section, we focus specifically on the technique known as laser-excited CVD or LECVD already reported in the

Laser-Assisted and Optical Pumping Techniques

379

literature [70,83,84]. In this method, a CO laser operating on several lines (i.e., transitions) from v = 12 → v = 11 through v = 1 → v = 0, is directed on CO gas molecules in a slowly flowing 95% CO/5% CH4 gas mixture at room temperature. The laser energy is absorbed by the target CO molecules initially in the ground electronic state (X1⌺⫹) and redistributed via anharmonic VV exchange collisions primarily to the internal modes of CH4 as well as to higher vibrational levels within CO. Diamond particles similar in morphology to those produced in Refs. 68–70 were observed on monocrystalline silicon wafers over substrate temperatures ranging from ⬃600⬚C down to room temperature. In order to understand this technique better, it is necessary to provide some detailed background on the optical pumping of CO, a topic on which there exists a considerable amount of literature. At room temperature, most CO molecules are in their ground electronic manifold (X1⌺⫹) with a majority of them occupying the ground vibrational level v = 0. The equilibrium distribution of populations versus vibrational level is described by the well-known Boltzmann distribution: nv = nv=0e⫺(Ev⫺Ev=0)/kT

(15)

where nv is the number density of CO molecules in vibrational level v with energy Ev = ប␻(v ⫹ 12) ⫺ ប␻␹e(v ⫹ 12)2, nv=0 is the number density of CO molecules in vibrational level v = 0 with energy Ev=0, and T is the translational mode temperature. In the expression for the vibrational energy Ev, ប = h/2␲, ␻ is the frequency of the molecule as an oscillator, v is the vibrational quantum number, and ␹e is the anharmonicity. Equation (15) states that at equilibrium at a temperature T, most molecules can be found in level v = 0, with the population in higher vibrational levels decreasing exponentially with v. Therefore, if radiation from a CO laser operating on vibrational transitions other than v = 1 → v = 0 is directed on gaseous CO at room temperature, no absorption will result and the gas will be transparent to the laser irradiation. It is evident that for cold CO gas to absorb the radiation, either the laser must have some output on the lowest transition v = 1 → v = 0 (i.e., shortest wavelength) or the gas must be heated (so that upper vibrational levels have sufficient population to absorb the longer wavelength laser transitions). Commercially available CO lasers typically do not have output on this shortest wavelength transition, and therefore such a CO laser must be specially built [70,83]. When laser output on the v = 1 → v = 0 transition is present, CO gas molecules can now absorb this laser energy and populate the v = 1 level. These molecules in level v = 1 now absorb the laser output from the v = 2 → v = 1 transition and populate the v = 2 level, and so on. In this manner, CO molecules can be optically pumped up to vibrational level v = 12, or whatever may be the vibrational level corre-

380

Subramaniam and Aithal

sponding to the longest wavelength laser transition. Collision-induced vibrational energy (VV) exchange processes then proceed according to CO(v) ⫹ CO(w) S CO(v ⫺ 1) ⫹ CO(w ⫹ 1),

wⱖv

(16)

This collisional vibrational energy exchange is favored due to the anharmonicity of molecules20 and serves to populate vibrational levels much higher than those initially populated by the laser irradiation. Levels as high as v = 40 in the ground electronic manifold of CO (corresponding to energies of about 10 eV) have been populated in this manner. This laser excitation followed by redistribution of the absorbed laser energy among the higher vibrational levels is known as anharmonic VV pumping. The resulting population distribution, henceforth referred to as the TRR distribution, is highly non-Boltzmann and was first derived by Treanor, Rich, and Rehm in 1968 [85]. For the limiting case of VV exchange collisions dominating VT exchange, the TRR distribution is [85]: nv = nv=0 exp

再

冎

(ប␻ v ⫺ 2ប␻␹ev) ប␻␹ev(v ⫺ 1) ⫺ kT kT1

(17)

where T1 is the ‘‘vibrational temperature’’ of level v = 1 defined by nv=1 = nv=0e⫺ប␻(1⫺2␹e)/kT1 or ប␻(1 ⫺ 2␹e) T1 = ⫺ nv=1 k ln nv=0

冉 冊

(18)

The form of Eq. (17) for the TRR distribution can also be derived from statistical thermodynamics in a manner analogous to that used to derive the Bose-Einstein or Fermi-Dirac distribution. This is done by maximizing the thermodynamic probability subject to the constraints that the total number of particles, total energy, and total number of vibrational quanta per molecule are fixed (i.e., conserved) [85]. Figure 34 shows a semilog plot of the normalized number density or population density as a function of vibrational quantum number for several cases. These are (1) the Boltzmann distribution as given by Eq. (15) but neglecting anharmonicity in the molecule, (2) the Boltzmann distribution 20

Anharmonicity refers to the fact that the vibrating bond between C and O atoms in the CO molecule exhibits departure from that of a harmonic oscillator. As a consequence, the gap between successive vibrational energy levels decreases as the vibrational quantum number (or vibrational energy) increases.

Laser-Assisted and Optical Pumping Techniques

381

Figure 34 Schematic showing a log-linear plot of normalized populations of molecules in vibrational level v versus vibrational quantum number (or vibrational energy) for three cases: (1) equilibrium, where the variation is given by the Boltzmann distribution [Eq. (15)] but without the effects of anharmonicity; (2) equilibrium, where the variation is given by the Boltzmann distribution [Eq. (15)] but including the effects of anharmonicity; and (3) nonequilibrium in which the average vibrational energy is different from the average translational energy.

including anharmonicity, and (3) the limiting TRR distribution as given by Eq. (17), with Tv ≠ T. It is important to point out that the concept of a vibrational temperature Tv that represents a meaningful average energy in vibration makes sense only when the variation of normalized number density versus vibrational quantum number is log linear. When departure from log linearity occurs, a single vibrational temperature cannot be defined for all vibrational levels. This is what happens in the case of the Boltzmann distribution including the effects of anharmonicity with Tv ≠ T or in the case of the ideal TRR distribution. The vibrational quantum number at which the number density attains a minimum for the TRR distribution can be determined by differentiating ln(nv /nv=0) from Eq. (17) with respect to v and setting it equal to zero. The result yields what is known as the Treanor minimum: vmin =

(1 ⫺ 2␹e) 2␹e

冉冊 T T1

⫹

1 2

(19)

A similar minimum exists for the Boltzmann distribution (i.e., equilibrium distribution) when the effect of molecular anharmonicity is included: (vmin)equilibrium =

(1 ⫺ ␹e) 2␹e

(20)

382

Subramaniam and Aithal

Values of frequencies and anharmonicities for some diatomic species are given in Table 5. As can be seen, the anharmonicity, ␹e, takes on values that are usually less than or equal to 0.01. This means that (vmin)equilibrium ⬇ 1/2␹e ⬇ 50. Such high vibrational levels are so close to the dissociation limit that the quadratic behavior of anharmonicity is never observed under equilibrium conditions. Thus, at equilibrium, the populations of vibrational levels are nearly log linear with vibrational quantum number below the dissociation limit, even when the effect of anharmonicity is included. In contrast, the TRR distribution yields vmin ⬇ (82)(1/5) ⬇ 16 for CO for T1 /T ⬇ 5. Thus, for the species listed in Table 5, the TRR distribution yields a minimum in the number density variation when the vibrational quantum number is on the order of 10 and the vibrational temperature of level v = 1 is a few times higher than the translational or gas temperature. In reality, the TRR distribution is modified by the competing effect of VT relaxation. VT energy transfer scales with vibrational quantum number and hence becomes important at the higher quantum numbers. The degree to which vibrational nonequilibrium persists in a gas therefore depends on the competing strengths of VV and VT or other loss processes such as spontaneous emission. In molecules such as CO and NO, spontaneous emission is slower than VT processes at pressures of interest. In homopolar diatomic molecules such as O2 and N2, there is no spontaneous emission because these molecules possess no natural or intrinsic dipole moment. We therefore illustrate the realistic VV up-pumping process by considering VV and VT processes alone, with the rotational modes in equilibrium with the translational modes at temperature T. Consider a diatomic molecule A2 undergoing the following single-quantum processes: VV energy transfer: k w⫺1,w v⫹1,v A2(v ⫹ 1) ⫹ A2(w ⫺ 1) → A2(v) ⫹ A2(w) ← k w,w⫺1 v,v⫹1

Table 5

(21)

Frequencies and Anharmonicities for Some Diatomic Molecules

Parameter ប␻ (cm⫺1) ប␻␹e (cm⫺1) ប␻/2ប␻␹e Source: Ref. 86.

CO

H2

N2

NO

O2

OH

2169.8 13.3 82

4401.2 121.3 18

2358 14.1 84

1904 14 68

1580 12 66

3735 83 22

Laser-Assisted and Optical Pumping Techniques

383

VT energy transfer: kv⫹1,v A2(v ⫹ 1) ⫹ M → ← A2(v) ⫹ M kv,v⫹1

(22)

where the VV and VT rates are given in Refs. 75 and 79. The evolution of the population of level v is then given by the following rate equations, including spontaneous emission as a loss mechanism [2]: dnv = VTv ⫹ VVv ⫹ SRDv dt

(23)

where the terms on the right-hand side represent net VV transfer, VT transfer, and spontaneous radiative decay into level v. Expressions for each of these terms, for CO, are given in Ref. 75. Process (21) is shown schematically in Fig. 35 for two general molecules A and B. Now, if E(v) is the energy of molecules in vibrational level v, then E(v ⫹ 1) ⫺ E(v) > E(w) ⫺ E(w ⫺ 1) for w ⱖ v because of the effect of anharmonicity. Thus, if we treat the right-hand side of process (21) as the ‘‘products’’ and the left-hand side as the ‘‘reactants,’’ we have ⌬E = E(v) ⫹ E(w) ⫺ E(v ⫹ 1) ⫺ E(w ⫺ 1) = [E(w) ⫺ E(w ⫺ 1)] ⫺ [E(v ⫹ 1) ⫺ E(v)] so that ⌬E ⱕ 0 for w ⱖ v. ⌬E is called the resonance defect. This shows that the forward process of (21) is favored at lower temperatures because it is exothermic in the sense that the energy difference ⌬E flows out of the vibrational modes into the external modes of rotation and translation for w ⱖ v. On the other hand, the reverse of process (21) is endothermic in the sense that energy must be provided from the external modes to sustain it. This can be expected to occur only at high temperatures. Energy can be retained in the vibrational modes, therefore, if an external source (such as a laser or optical pump) provides energy to the vibrational mode

Figure 35 Schematic showing intermolecular vibrational energy transfer, preferentially transferring vibrational energy from the molecule with the more widely spaced energy levels to the molecule with the more closely spaced energy levels [2]. When both collision partners are molecules of the same species, then such VV transfer is called intramolecular vibrational energy transfer. The difference between ⌬v and ⌬w is called the resonance defect.

384

Subramaniam and Aithal

to overcome loss processes such as process (22) or else process (22) is inhibited or minimized. Process (22) scales with vibrational quantum number v and temperature; i.e., the rate for vibrational deexcitation by VT transfer increases with increasing v and increasing T. Hence, as long as the energy corresponding to the resonance defect is removed either by having the gas flow or by having a suitable diluent such as argon, a low gas temperature can be maintained or VT losses can be lowered compared with VV exchange. Under these conditions, the solution of the rate equations (23) yields a fully VV pumped distribution and is indeed observed as shown in Fig. 36. The fully VV pumped distribution is observed in optically pumped gases as well as in common plasmas. It is, in fact, ubiquitous in plasmas containing tightly bound molecular species such as CO, NO, N2, or O2. In such plasmas, the mean electron energy is lowered because inelastic collisions between electrons and molecules dominate all other processes. Consequently, the electron (translational) energies in such discharges strongly resemble the fully VV pumped distribution shown in Fig. 36. The result is that a single electron temperature cannot be identified, and hence we must speak of electron energies and a corresponding electron energy distribution function or EEDF that is highly non-Boltzmann. These electron impact processes can be included in the rate equation formulation as given by Eq. (23) and are crucial in determining the chemical composition in a plasma. Process (16) is an example of intramolecular energy transfer, or transfer of vibrational energy between molecules of the same chemical species. Intermolecular energy transfer, or transfer of vibrational energy between molecules of different chemical species, also occurs. In both cases, energy is transferred preferentially from the molecules with widely spaced vibrational energy levels to those with more closely spaced energy levels. LECVD takes advantage of this point precisely. Further, this energy transfer is a result of molecular collisions and scales with pressure (or density) as long as relatively low external mode (i.e., translational and rotational) temperatures are maintained or as long as rates of population of vibrational levels by VV exchange collisions are favorable compared with the rates of depopulation of these levels by VT collisional processes. The effect of anharmonicity in a diatomic or small polyatomic molecule is such that any vibrational energy over and above the equilibrium value (corresponding to the translational temperature T ) is rapidly redistributed to higher vibrational states. Thus, molecular states at high vibrational levels can be populated, especially when the gas temperature is very low. The process, however, is not restricted to low gas temperatures and has been shown to occur at elevated gas temperatures on the order of 1500 K [87].

Laser-Assisted and Optical Pumping Techniques

385

Figure 36 Fully VV pumped distribution from Ref. 79 measured from an optically pumped mixture of 1% CO/Ar at a total pressure of 100 torr. Note that at low vibrational quantum numbers, the VV pumped distribution resembles the TRR distribution but instead of displaying a minimum exhibits a plateau known as the TRR plateau characterized by v ln(nv=0/nv) = constant. The distribution exhibits a sharp drop at the higher vibrational quantum numbers because the VT rates overwhelm the VV rates at high values of v.

B.

Optical Pumping of Gases (Laser-Excited CVD or LECVD) for Diamond Growth

Optical pumping of gases using a low-power laser can, under the right conditions, result in production of molecules with high vibrational energies. These molecules can then be made to react among themselves or transfer energy to other species to excite them to specific vibrational or electronic states and hence serve as huge heat baths for driving chemical reactions. A particularly dramatic demonstration of the power and usefulness of this tech-

386

Subramaniam and Aithal

nique is the early work of Kildal and Deutsch [88]. At the time, laser sources in the mid-IR such as the CO laser were not as well developed as they are today. Consequently, Kildal and Deutsch [88] went to elaborate lengths to obtain some coherent output at ⬃4.7 ␮m by frequency doubling the 9.4 ␮m laser line from a CO2 laser. The resulting laser output at 4.7 ␮m is coincident with the lowest vibrational transition, v = 0 → v = 1, in CO so that CO can be vibrationally excited to level v = 1 at room temperature. Once this state is populated, of course, other higher vibrational levels can be populated via anharmonic VV pumping: CO(v = 1) ⫹ CO(v = 1) → CO(v = 0) ⫹ CO(v = 2), followed by CO(v = 1) ⫹ CO(v = 2) → CO(v = 0) ⫹ CO(v = 3), and so on. Kildal and Deutsch used this approach to optically pump various gas mixtures containing CO, where CO is used as the energy ‘‘donor’’: CO/ C2H2, CO/C2H2 /H2, CO/CO2, CO/N2O, CO/CS2, CO/CS2 /H2 and CO/OCS [88]. Further, by placing mirrors on the chamber containing these gas mixtures, Kildal and Deutsch were able to obtain lasing from the ‘‘acceptor’’ molecules in which population inversions (partial or total) were sustained by energy supplied from the optically pumped and subsequently vibrationally excited CO molecules. The fact that such transfer lasers could be produced was testament to the rapidity of intermolecular vibrational energy transfer versus loss of this energy to translational modes. A laser source at 4.7 ␮m (i.e., laser output on the CO fundamental v = 1 → v = 0) would have made the optical pumping of CO much easier. Similar energy transfer and VV pumping have been previously reported in CH4, CD3H, CD4, CH2D2, and CH3F [89]. The CO laser was invented as early as 1965 by Patel [90], but theoretical understanding of its operating principle would await the development of the theory of anharmonic VV pumping developed by Treanor, Rich, and Rehm in 1968 [85]. This theoretical development subsequently led to more powerful supersonic and cryogenically cooled CO laser sources [77–79]. Today, the CO laser is the most efficient gas laser, displaying efficiencies near 50%. Steady development in the 1980s and 1990s led to compact cryogenically cooled CO laser sources capable of significant output on the lowest vibrational transition (v = 1 → v = 0). This has been important in enabling the optical pumping of CO at room temperature. We begin with a description of the optical pumping of CO diluted in argon. It is necessary to understand optical pumping in this simple mixture before proceeding with a description of the LECVD process that utilizes the optical pumping of other cold CO-containing gas mixtures. An excellent review can be found in Ref. 2. Consider a mixture of CO/Ar (typically a few percent CO with the balance being argon) flowing at 100 sccm, total pressure of about 50 torr, and temperature of 300 K, irradiated by a CO laser operating multiline with some of its output on the lowest transition (v

Laser-Assisted and Optical Pumping Techniques

387

= 1 → v = 0). Usually, lasing on the lowest transition is difficult but can be relatively easily produced using an intracavity Q switch [70,82]. This usually results in about 100 mW output on the lowest transition, with a total output of about 5 W on all lines. With improved designs [91], cw (continuous wave) output with several watts on the lowest transition can be obtained. The CO in the CO/Ar mixture absorbs this radiation on the v = 1 → v = 0 transition, which populates the v = 1 energy level above its equilibrium value. This in turn enables laser output on the v = 2 → v = 1 transition to be absorbed, populating the v = 2 level above its equilibrium value, and so on. In this manner, levels v = 9 down to v = 1 are populated by absorption of energy from the laser beam. A typical spectrum of the CO laser emission is shown in Fig. 37. It is important to highlight the fact that this absorption process is a resonant, single-photon process and therefore does not require high laser powers or pulse energies. This absorbed laser energy is then redistributed to the higher vibrational levels by anharmonic VV pumping. The result is a fully VV pumped distribution as shown in Fig. 36. Note that only one por-

Figure 37 Typical spectrum showing output from a CO laser. This laser is capable of operating at about 10 W cw while also lasing on the v = 1 → v = 0 transition (leftmost peak). Each successive set(s) of peaks represents higher vibrational bands from v = 2 → v = 1 up to v = 10→ v = 9. (From Ref. 91.)

388

Subramaniam and Aithal

tion of the TRR distribution is evident because VT rates scale or increase with vibrational quantum number and prevent the total inversion that is characteristic of the TRR distribution. Instead of a minimum, the real VV pumped distribution exhibits a near-plateau region over which vnv ⬇ constant, followed by a steep drop in the vibrational populations at the higher vibrational quantum numbers. The reason for both trends is that both VV and VT rates increase with v, but eventually VT rates exceed the VV rates that drive the up-pumping at the higher vibrational quantum numbers. A fully VV pumped vibrational population distribution in a flowing CO/Ar mixture results in population of vibrational states as high as v = 42, corresponding to energies of more than 8 eV [79]. This is significant because the dissociation energy of the CO molecule is about 11 eV. When a molecule such as CO is pumped to vibrational levels as high as v = 42, quite a bit of energy is engaged within the vibrational mode. This energy should be available to drive chemical reactions or for transfer to other modes of energy storage such as electronic excitation. Indeed, a host of energy transfer processes occur such as those that result in population of low-lying electronic states by VE (vibration-to-electronic) transfer [79,82]: CO(X1⌺⫹, v) ⫹ CO(X1⌺⫹, w) → CO(v⬘) ⫹ CO(a3⌸, w⬘), Ev ⫹ Ew > 6 eV ⫹

⫹

⫹

CO(X ⌺ , v) ⫹ CO(X ⌺ , w) → CO(X ⌺ , v ⬘) ⫹ CO(A ⌸, w⬘), Ev ⫹ Ew > 8 eV 1

⫹

1

1

CO(X ⌺ , v ⬇ 27) ⫹ M → CO(a ⌸, w⬘) ⫹ M) CO(X1⌺⫹, v ⬇ 40) ⫹ M → CO(A1⌸, w⬘) ⫹ M 1

3

(24)

1

(25) (26) (27)

Processes (26) and (27) represent direct collisional transfer to electronic states, one of which is the a3⌸ state, which is metastable with a radiative lifetime of 4.4–9.5 msec. Another electronic state, A1⌸, has a radiative lifetime of 10 nsec and therefore serves to limit the vibrational up-pump. It must be pointed out that around v = 42, 2[E(v = 42) ⫺ E(v = 40)] ⬇ [E(v = 1) ⫺ E(v = 0)], so that the following near-resonant process: CO(X1⌺⫹, v ⬇ 42) ⫹ CO(X1⌺⫹, v = 0) → CO(X1⌺⫹, v ⬇ 40) ⫹ CO(X1⌺⫹, v = 1) can occur with high rapidity, as does VT transfer (which scales with the vibrational quantum number v). While this is possible, the VE transfer via process (27) has been observed experimentally, and radiative emission from CO(A1⌸, w) → CO(X1⌺⫹, v) has been observed only after CO(X1⌺⫹, v ⬇ 42) is populated [79]. Thus, despite attaining energies near the dissociation limit, CO is not observed to dissociate because of the rapid VE transfer to

Laser-Assisted and Optical Pumping Techniques

389

the A1⌸ electronic state, followed by fast radiative decay to the ground electronic state or manifold [79]. Thus, infrared radiation goes into exciting the gas, while ultraviolet light comes out! In contrast, the a3⌸ by virtue of its relatively long lifetime suffers collisions and is either deexcited (collisionally quenched) or reacted.21 Processes (24) and (25) represent energy transfer by an associative mechanism referred to as vibrational pooling. Such processes have high probability of occurrence because they typically involve many combinations of molecular collision pairs with energies that add up to the required energy of a given excited electronic state. Such associative processes have been shown to be instrumental in producing ionization by vibrational pooling and may constitute a major source of ionization in CO-containing electrical discharges [92]: CO(X1⌺⫹, v) ⫹ CO(X1⌺⫹, w) →

再

CO⫹ ⫹ CO(X1⌺⫹, w⬘) ⫹ e⫺, ⫺ (CO)⫹ 2 ⫹ e

Ev ⫹ Ew ⱖ 13.6 eV

(28)

The lower pathway of process (28) resulting in the formation of the dimer ion is favored because it has a slightly lower ionization potential than CO. Regardless, the production of CO⫹ or (CO)⫹ 2 dimer ions usually leads to cluster ions of the form (CO⫹ ⭈ CO⭈ Cn) where n = (0, . . . , 11) [93]. In addition to the aforementioned energy transfer and associative ionization processes, chemical reactions such as the Boudouard reaction occur: {CO(X1⌺⫹, v)

or

CO(a3⌸, v⬘)}

⫹ {CO(X1⌺⫹, w) or CO(a3⌸, w⬘)} → C* ⫹ CO2, Ev ⫹ Ew > 6 eV

(29)

followed by: C ⫹ CO ⫹ M → C2O

(30)

C ⫹ C ⫹ M → C2 ⫹ M

(31)

The formation of C2 in the gas phase is generally thought to be undesirable for diamond formation. There is some work however, that suggests C2 could be a growth species for nanocrystalline diamond [94,95]. Past optical pumping experiments involving sufficient vibrational excitation of CO to drive the Boudouard reaction (29) leading to C2 formation via (31) have shown formation of graphitic soot [96]. Reaction (31) is one of the channels for Alternatively, there can be collision-induced E-V transfer from CO(a3⌸) into the CO(X1⌺⫹, v ⬇ 27) state, although it has not yet been observed experimentally by optical pumping.

21

390

Subramaniam and Aithal

formation of soot in electrical discharges containing substantial amounts of CO. In a plasma, free carbon is produced by electron impact dissociation of CO as well as by reactions (29) and (31) where the appropriate CO(a3⌸) or vibrationally excited CO(X1⌺⫹) is produced by electron impact processes. In contrast, formation of C2 in the gas phase in an optically pumped mixture first requires the formation of atomic carbon via reaction (29), which has an activation energy of about 6 eV. Therefore, it is clear that formation of CO in the a3⌸ state [whose lowest vibrational level is isoenergetic with the v ⬇ 27 state of CO(X1⌺⫹) with an energy on the order of 6 eV] or CO in vibrational levels beyond v ⬇ 13 in the X1⌺⫹ state must be suppressed. The requirement that CO in vibrational levels beyond v ⬇ 13 in the X1⌺⫹ state not be populated stems from the fact that molecules in these levels can pool their energies to form some molecules in v ⬇ 27 of the X1⌺⫹ state or in the low vibrational levels of the a3⌸ state. In general, light molecules or molecules with many rotational modes are very effective in VT relaxing vibrationally excited CO [2,3]. Molecular hydrogen and water are two examples of strong VT relaxers [2,75]. Hence, it is very difficult to produce reactive vibrationally excited CO molecules by optical pumping in mixtures predominantly comprising H2 or H2O, which is typical of growth mixtures used in diamond synthesis by CVD. Rebello et al. circumvented this problem by using a mixture containing mainly CO but with some CH4, which is commonly used as a carbonaceous precursor in CVD processes for diamond synthesis. Diamond growth was reported from a gaseous mixture of 95% CO and 5% CH4, where the CO was optically excited using a Q-switched CO laser operating with its shortest wavelength output on the v = 3 → v = 2 transition [70,83]. These experiments were the first to demonstrate that diamond growth was possible without the need to hyperdilute the gas phase with hydrogen or some other etchant. Such a mixture has never before been reported to yield diamond growth by any other CVD process (HFCVD, plasma CVD, or flame synthesis), and may therefore signal the existence of alternative routes to diamond synthesis. We review their experiment here in detail. A schematic of the LECVD or laser-excited CVD experiment is shown in Fig. 38. A CO laser Q-switched at 200 Hz with the lasing lines shown in Fig. 39 was used to optically pump a flowing mixture of 95 sccm of CO and 5 sccm of CH4 at a total pressure of 50 torr. Of the 2.3 W output (on all lines) from the Q-switched CO laser, approximately 1 W was reportedly absorbed by the CO in the gas mixture [70,83]. Single-crystal Si(100) wafers approximately 2 cm2 in size were used as substrates and heated to a temperature between 600 and 700⬚C. The substrates were supported on a heated holder within the reactor and oriented upside down [83]. This was done to

Laser-Assisted and Optical Pumping Techniques

Figure 38 Ref. 83.)

391

Schematic of the LECVD or laser-excited CVD experiment. (From

take advantage of the buoyancy-driven or free convective flow induced by substrate heating, which tends to drive the gas upward. Such a means of mounting was also used in experiments where the substrates were not heated [70]. The substrates were physically abraded using diamond particles (4- to 8-␮m sized particles in the earlier work where the substrate was heated [83] and with particles 0–0.5 ␮m in size in the later work where the substrates

392

Subramaniam and Aithal

Figure 39 Spectrum showing the lasing lines of the CO laser used in the LECVD experiments. This laser was Q switched at 200 Hz. (From Ref. 83.)

were not heated [70]) suspended in glycerol and subsequently cleaned in deionized water, acetone, and methanol, respectively. The authors reported that such a pretreatment was used to possibly reduce nucleation times and thereby enable short-duration experiments, as is typically done with other diamond growth techniques [70,83]. The substrates were then thoroughly characterized prior to growth using SEM imaging, and Raman spectroscopy (survey as well as with a microprobe). The substrates contained some residual abrasive particles only in a few isolated cases. However, these proved to be debris of the silicon substrate when analyzed using energy dispersive x-ray spectroscopic (EDS) analysis. Further, these debris particles did not show the characteristic diamond line at 1332 cm⫺1 in the corresponding Raman spectra. The authors carefully characterized the substrates before growth and reported on the formation of additional particles approximately 10 ␮m in size clustered together in groups that covered regions on the order of 50 or 60 ␮m2. Figure 40 shows SEM images of the substrate surface before and after growth. The fact that these growth particles were grouped together enabled Raman spectra to be obtained using a microprobe. A representative Raman spectrum and EDS spectrum are shown in Figs. 41 and 42, respectively. Of

Laser-Assisted and Optical Pumping Techniques

393

Figure 40 SEM images before and after growth in the LECVD experiment. (From Refs. 70, 83.)

significance is the fact that these experiments have been repeated under conditions where the substrate is not heated and similar diamond growth is observed [70]. Evidently, diamond growth by LECVD does not require the monocrystalline silicon substrate to be heated. During the LECVD growth process, there was no visible emission from the reactor chamber even when the substrate was heated because the resistively heated tungsten filament was hidden from view. However, there was radiative emission from the gas in the infrared. Figure 43 shows a low-resolution IR spectrum from the gas

394

Subramaniam and Aithal

Figure 41 Typical Raman spectrum of the diamond particles synthesized by LECVD. (From Ref. 70.)

phase for the case where the substrate was not heated, displaying the overtone emission (i.e., ⌬v = 2) from vibrationally excited CO and emission from the ␯3 mode of CH4. The ␯3 mode is the asymmetric stretch mode where a single hydrogen atom vibrates against the CH3 group in the methane. An energy level diagram for CH4 is shown in Fig. 44 along with the corresponding low-lying CO vibrational energy levels. In addition, faint emission was observed from the ␯4 or bending mode of CH4 [70]. The ␯4 mode emission is particularly difficult to detect because it overlaps with ambient blackbody radiative emission at 300 K. These spectra serve to verify that CO is vibrationally excited and transfers energy to CH4 to excite it vibrationally as well. The population distribution in CO(X1⌺⫹) can be inferred by comparing the experimentally measured radiative emission with synthetic spectra and is shown in Fig. 45 [70]. It can be seen that the CO in the CO/ CH4 mixture under conditions of interest to diamond growth exhibits vibrational nonequilibrium but is not fully VV pumped. Using the slope of the population distribution for the lower vibrational levels in Fig. 45, a vibrational temperature T␯ = 2200 K can be inferred.22 There are some interesting similarities and differences between these diamond crystals and those observed in other CVD systems. In the first 22

Note that here the translational temperature T ⬇ 300 K ≠ T␯ = 2200 K.

Laser-Assisted and Optical Pumping Techniques

395

Figure 42 Energy-dispersive x-ray spectroscopic (EDS) analysis spectrum of the diamond particles synthesized by LECVD. (From Ref. 70.)

place, the morphology of these particles is clearly irregular but still faceted. It resembles the morphology of the particles obtained by photolysis of CO in the CO/H2 mixtures in LCVD (see Sec. IV and Refs. 68 and 69). There is also a strong resemblance of the morphology of the growth particles to that of the high-pressure diamond grit used in the abrasion pretreatment, except the growth particles are larger. The well-defined crystal habits displayed in diamond synthesized by other CVD techniques are due to the simultaneous presence of growth and etch channels. If pure growth occurred without the presence of any etch mechanisms, morphologies such as those in Figs. 25 and 29 would be expected, as discussed earlier in Sec. IV (see Figs. 25 and 29, associated discussion, and Ref. 72). The Raman spectra reveal that the growth particles are indeed diamond as can be seen from the sharp 1332 cm⫺1 line in Fig. 41. The fact that no graphitic lines (at 1580 cm⫺1 and/or at 1343 cm⫺1) are observed in the Raman spectrum is indeed conspicuous. Codeposition of nondiamond (graphitic) carbon is usually ob-

396

Subramaniam and Aithal

Figure 43 Low-resolution IR spectrum from the gas phase, displaying the overtone emission (i.e., ⌬v = 2) from vibrationally excited CO and emission from the ␯3 mode of CH4. The ␯3 mode is the asymmetric stretch mode where a single hydrogen atom vibrates against the CH3 group in the methane. (From Ref. 70.)

served in diamond growth using other types of CVD processes and is clearly absent in LECVD. The EDS spectrum shows the presence of a small peak corresponding to oxygen. As this peak is observed only in diamond films grown using other CVD techniques in which oxygen or oxygen-bearing compounds are used in the growth gases [15–17,69], this suggests rather strongly that CO may play a role in diamond growth by LECVD. Such an oxygen peak is never observed in EDS spectra of diamond films exposed to ambient air. Therefore, it appears highly likely that oxygen is present on the diamond surface itself as a direct result of the specific LECVD growth environment and may in fact provide clues to growth mechanisms. Control experiments were also done to determine what the growth species might be [70,83]. Substrates having undergone the same abrasive pretreatment, and either heated to the same temperatures or unheated, were exposed to optically pumped 95% CO/5% H2 and 100% CO gaseous environments but showed no presence of diamond. These control experiments suggest that CH4 is the likely carbonaceous precursor. However, the CO

Laser-Assisted and Optical Pumping Techniques

397

Figure 44 Energy level diagram for CH4, along with the corresponding low-lying CO vibrational energy levels. (From Ref. 70.)

could have played a facilitating role as can be inferred from EDS analysis indicating the presence of oxygen on the diamond surface. Unfortunately, experiments involving 13CH4, which would have been more illuminating, were not done.23 As mentioned earlier, the same optical pumping technique has been applied to conditions where the substrate is not heated [70,84]. Earlier experiments [83] were conducted with a heated substrate and v = 3 → v = 2 23

The use of 13CO in the gas phase would have been problematic because the CO laser source would also have had to operate using the same isotope in order to take advantage of resonance absorption.

398

Subramaniam and Aithal

Figure 45 Population distribution in CO (X1⌺⫹) inferred by comparing the radiative emission with synthetic spectra. (From Ref. 70.)

as the shortest wavelength lasing transition from the CO laser. As discussed earlier in Secs. V.A and V.B, there is a negligible population of CO molecules in the v = 2 state at room temperature so that laser absorption to trigger the VV pumping could occur only at an elevated temperature. In the more recent work [70,84], however, the CO laser was operated with some output on the v = 1 → v = 0 transition so that CO molecules at room temperature could be vibrationally excited. Indeed, as in the LCVD or laser photolysis technique discussed in Sec IV, diamond growth was obtained on unheated surfaces [70,84]. It is of significance that two completely different laserbased techniques (LCVD and LECVD) involving two completely different gas mixtures (1% CO in 99% H2 for photolysis and 5% CH4 in 95% CO for LECVD) yielded diamond growth on unheated surfaces. This also suggests that growth mechanisms in these completely different experiments may share common features. We shall address the subject of mechanisms shortly. Optical pumping experiments have also been conducted in mixtures of 5% C2H2 in CO [70,84]. A low-resolution emission spectrum from this experiment is shown in Fig. 46 and displays the ␯3 band of acetylene and faint emission from possibly the combination bands ␯2 ⫹ ␯ 15, ␯2 ⫹ 2␯ 05 ⫺ ␯ 15, ␯2 ⫹ 2␯ 14 ⫹ ␯ 15 ⫺ ␯ 14, ␯3 ⫺ ␯ 14, and ␯1 ⫺ ␯ 15 [97]. The spectrum ascribed to the

Laser-Assisted and Optical Pumping Techniques

399

Figure 46 Low-resolution emission spectrum from optical pumping of CO/C2H2 mixture showing the ␯3 band of acetylene and faint emission from possibly the combinations bands ␯2 ⫹ ␯ 15, ␯2 ⫹ 2␯ 05 ⫺ ␯ 15, ␯2 ⫹ 2␯ 14 ⫹ ␯ 15 ⫺ ␯ 14, ␯3 ⫺ ␯ 14, and ␯1 ⫺ ␯ 15 [97]. (From Ref. 70.)

␯3 band may also contain some emission from the ␯2 ⫹ ␯ 14 ⫹ ␯ 15 combination band. The ␯3 band of acetylene arises from its asymmetric stretch mode wherein a hydrogen atom vibrates against the associated C2H. As in the case of the optically pumped CO/CH4 mixture, the vibrational populations of CO can be inferred from the overtone emission spectrum. This is shown in Fig. 47 for the CO/C2H2 mixture, and appears to indicate that levels up to about v = 9 or v = 10 are populated. However, no diamond growth was obtained from optically pumped CO/C2H2 mixtures despite the obvious energy transfer from CO to acetylene [70]. The LECVD experiments reported in Refs. 70, 83, and 84 were independently reproduced in the Laboratoire d’E´nerge´tique Moleculaire et Macroscopique, Combustion (E.M2.C.) at E´cole Centrale Paris in France [98]. These results confirm diamond growth both on unheated silicon substrates as well as on heated substrates [98]. Additional experiments in which high-resolution spectra were obtained are reported in Ref. 98 as well. Figure

400

Subramaniam and Aithal

Figure 47 Vibrational populations of CO inferred from the overtone emission spectrum are shown here for the CO/C2H2 mixture and indicate that levels up to about v = 9 or v = 10 are populated. (From Ref. 70.)

48 shows a high-resolution spectrum of the same bands of CO and CH4 that are displayed in Fig. 43 for the experiments reported in Refs. 70, 83, and 84. With 4 W of the incident 10.8 W cw laser power (on all lines) absorbed, Ref. 98 reported that emission measurements indicate that the CO alone (without addition of CH4) is pumped up to levels near v = 17, which corresponds to energies of about 4 eV. Upon addition of the CH4, emissions from CO molecules are observed only up to about v = 10, which corresponds to energies of about 2.5 eV. Note that in the latter case, formation of C2 in the gas phase is suppressed because reaction (29) requires an activation energy of about 6 eV. Reference 98 also presents detailed analyses using high-resolution emission and absorption spectroscopy. Careful analysis shows that the spectrum largely attributed to the ␯3 mode of CH4 actually contains other lines arising from the combination bands ␯3 ⫹ ␯4 → ␯4 and ␯3 ⫹ ␯2 → ␯2, also known as the ‘‘hot bands’’ of methane. From these spectra, a vibrational temperature T␯ of 1950 K was inferred [98]. Of significance is the fact that the detailed gas-phase spectroscopic studies of Ref. 98 indicate the presence only of CO and CH4 in both emission and absorption measurements. No evidence was found for any other chemical species or reaction by-products, at least within detection limits. Clearly, either re-

Laser-Assisted and Optical Pumping Techniques

401

Figure 48 High-resolution spectrum of the overtone bands of CO and ␯3 mode emission from CH4 from the E´cole Centrale Paris experiments. (From Ref. 98.)

action by-product concentrations were too small to be observed or reactions occurred on the growth surface. Diamond growth on unheated, seeded substrates has been reported using two completely different laser-based techniques (laser photolysis and LECVD) and using two very different gas mixtures (1% CO in H2 in the photolysis case and 5% CH4 in CO in the optical pumping case). These results beg the question of whether or not the two techniques share some commonality in growth mechanisms. At first glance, it appears they share

402

Subramaniam and Aithal

no common features. Laser photolysis uses a ‘‘conventional’’ gas mixture for diamond growth and therefore can be explained using present understanding of diamond growth by CVD processes [64], which identifies the CH3 radical as the growth precursor. Once CH3 groups are added to a growing diamond surface, the atomic hydrogen is removed by the atomic hydrogen present in the gas phase. In contrast, LECVD uses a gas mixture that normally would not yield diamond growth if used in other conventional non–laser-based CVD methods. There cannot be an appreciable amount of atomic hydrogen in the LECVD environment, or else the CO would not VV pump [2,75]. Because the particle growth rates for the two methods are comparable (on the order of 1 ␮m/hr) as are the resulting morphologies and Raman spectra, we expect the carbonaceous precursor to be present in concentrations on the order of 1018 m⫺3 following the same reasoning as in Sec. IV. It is extremely difficult to infer growth mechanisms or detailed channels based upon the experiments reported thus far [70,83,84,98]. However, a reasonable reaction scheme can be formulated if all the known evidence is viewed collectively. This evidence will be briefly summarized at the outset. First, control experiments indicate no diamond growth either in the absence of laser excitation or when CH4 is absent from the gas mixture. Further, no diamond growth was obtained from optically pumped CO/C2H2 mixtures. Second, elemental oxygen is detected in the EDS spectra of diamond particles grown using LECVD as well as laser photolysis. This suggests that CO somehow acts in concert with CH4 during growth. Third, there is clear evidence of energy transfer from CO to CH4 as can be seen from the IR emission from excited states of CH4. Fourth, the LECVD experiments conducted at E´cole Centrale Paris detected only CO and CH4 and no other species (within detection limits) in their emission and absorption studies [98]. Fifth, the morphology of the irregularly faceted particles grown by LECVD is identical to those grown by laser photolysis. Finally, the particle growth rates appear to be insensitive to whether or not the substrate is heated. Based upon these six essential points, we attempt to formulate a possible mechanism. There have been several important attempts to elucidate the surface mechanisms responsible for diamond growth [64,65,99–102]. These mechanisms are relevant for conventional CVD techniques for diamond growth in a hydrogen-rich environment and have shown promise in providing a mechanistic understanding of growth on (100) surfaces. Many of the proposed mechanisms focus on the incorporation of CHx species on the surface, specifically the CH3 molecule (i.e., methyl radical) [64,65,99–101]. We therefore concentrate here on the possibility of extending these ideas to understand the LECVD process. However, it must be borne in mind that

Laser-Assisted and Optical Pumping Techniques

403

these mechanisms have been postulated for a hydrogen-rich ambient and not for the conditions typical of LECVD. The first point that must be addressed is the production of methyl radicals either in the gas phase or on the surface of seed diamond particles. Given the relatively high vibrational temperature of CO (⬃2200 K) produced by laser excitation and subsequent VV pumping, it is reasonable to expect the translational temperature of CH4 to be between 300 and 2200 K. Under these conditions, production of CH3 can occur via the following channels: Vibrationally stimulated thermal dissociation: CH4 ⫹ M → CH3 ⫹ H ⫹ M Vibrational pooling:

CH* 4 ⫹ CH* 4 → CH3 ⫹ H ⫹ CH4

(32) (33)

CO(v) ⫹ CH* 4 → [CH3COH]* → [C2H4O]* → [CH3CHO]*

CH3 ⫹ HCO CH3CHO CH4 ⫹ CO

(34)

where the * denotes an excited species. The formation of [CH3COH]* can proceed via an insertion reaction whereby a CO molecule inserts itself into a vibrationally excited CH4. Such insertion reactions involving AlO inserted in CH4, have been previously reported [103]. The last step of process (34) has been observed in studies of decomposition of oxirane [104]. It is also interesting to note that CH3CHO is more stable than C2H4O by about 27 kcal/mol or ⬃1.17 eV/molecule. But, there is an activation energy of ⬃57 kcal/mol or ⬃2.47 eV/molecule for the isomerization C2H4O → CH3CHO. This activation energy can be provided via the vibrational excitation of the reactant CO and CH4 molecules by pooling because energies up to ⬃2 eV are available from CO molecules near v = 10 (see Fig. 45). A compelling argument can be made for processes (33) and (34) based upon energy transfer considerations. It can be seen from Fig. 44 that the ␯3 mode of CH4 cannot be populated by intermolecular energy transfer from CO(v = 1) because its energy is higher than the energy corresponding to the v = 1 → v = 0 transition.24 However, intermolecular energy transfer can occur to the ␯2 and ␯4 states of CH4 via CO(X1⌺⫹, v = 1) ⫹ CH4 → CO(X1⌺⫹, v = 0) ⫹ CH4(␯2 or ␯4)

(35)

It can thus be seen that the ␯3 mode can be populated by VV up-pumping 24

Two-quantum transfer via processes such as CO(v = 2) ⫹ CH4 → CO(v = 0) ⫹ CH4(␯3) are possible but much less probable than transfer of single vibrational quanta.

404

Subramaniam and Aithal

Table 6 Typical Characteristics of Low-Lying Energy Levels and VV Transfer in Methane at 2.5 Torr State

␯1 (symmetric stretch) ␯2 (symmetric bend) ␯4 (asymmetric bend) ␯3 (asymmetric stretch) 2␯2 2␯4 2␯3 2␯3 → ␯3 ␯2 ⫹ ␯4 ␯3 → ␯4 ␯3 → ␯2 ␯1 → ␯4 2␯4 → ␯2 ⫹ ␯4 ␯1 ⫹ ␯4 → ␯3

Characteristic time for VV transfer

Radiative lifetime

O(0.1–0.5 ␮sec) O(0.1–0.5 ␮sec) 1.1 ␮sec 1.6 ␮sec O(0.1–0.5 ␮sec) O(0.1–0.5 ␮sec) O(0.1–0.5 ␮sec) 0.92 ␮sec O(0.1–0.5 ␮sec) 0.72 ␮sec O(0.1–0.5 ␮sec) O(0.1–0.5 ␮sec) 1.6 ␮sec 0.22 ␮sec

>20 sec 13 sec 0.39 sec 37 msec — 0.20 sec — — 0.39 sec 25 sec 120 sec 8 sec — —

Source: Refs. 105–107.

within the methane initiated from either the ␯2 or ␯4 modes25 or by radiative decay from an excited state. In the latter case, an upper state could have been populated only by VV up-pumping within the methane. Once a specific excited state is populated, it can undergo one of several processes. It can radiate a photon (provided selection rules allow such a transition) and occupy a lower energy state or ground state. It can be collisionally deexcited to any lower state by a VT-type collision or collisionally excited to an upper state by intermolecular or intramolecular vibrational energy transfer. In CH4, the radiative lifetimes of the low-lying levels are longer than tens of milliseconds, as can be seen from Table 6 [105–107]. In contrast, VV transfer occurs within microseconds at pressures of interest to LECVD (i.e., PCH4 = 2.5 torr) [106]. Also, VT and VR collisional deactivation of these levels occurs on time scales of hundreds of microseconds or milliseconds at 2.5

25

The ␯2 and ␯4 levels of methane are isoenergetic with the v = 24 → v = 23 and v = 33 → v = 32 transitions in CO, respectively. Because such high vibrational levels are not significantly populated in CO (as can be seen from the overtone emission spectrum of CO), these methane energy levels must have been populated by off-resonance intermolecular energy transfer from v → v ⫺ 1 transitions in CO (with the v = 1 → v = 0 transition being the most likely one).

Laser-Assisted and Optical Pumping Techniques

405

torr [106]. Hence, if intermolecular energy transfer from CH4 back to CO does not depopulate the upper levels of CH4 appreciably, a compelling argument can be made for rapid VV exchange among CH4 molecules. We therefore expect CH4 molecules to be vibrationally hot. This is evident from the strong radiative emission from the ␯3 mode of CH4 despite the presence of rapid intramolecular VV transfer and self-absorption of radiation from the ␯3 mode by surrounding, cold CH4. It is therefore clear that vibrational energy from CO is rapidly transferred to CH4, thereby populating the ␯2 or ␯4 (or both) states. The presence of vibrationally excited CO and excited CH4 in the LECVD experiments has been inferred. These can lead to formation of CH3 radicals via processes (32) through (34). Similar reactions could occur between CO and CH4 molecules adsorbed on diamond seed surfaces as well, and there is not enough experimental evidence to argue for production of methyl radicals from the gas phase alone. Nevertheless, given such an initial source of production of CH3 radicals, the evidence gathered from the LECVD experiments collectively leads to two possible mechanisms. They are (1) growth from CH3 radicals with CO playing no part other than that of a spoiler and terminating growth and (2) growth by concerted action of CO and CH3 at the surface of the diamond seed particles. Let us consider these two possible mechanisms in some detail, especially from the perspective of experimental observations. For ease of discussion, we consider growth on the (100)-oriented surface. Once CH3 radicals are produced (via either gas-phase or surface processes), they can attach themselves to dangling bonds on the (100) surface as shown in Fig. 49. This can be written symbolically as: C•d ⫹ CH3 → Cd — CH3

(36)

• d

where C represents an activated site (i.e., dangling bond) on a seed diamond surface. Atomic hydrogen from the CH3 group can then be removed by successive attacks by other CH3 radicals: Cd — CH3 ⫹ CH3 → Cd — CH•2 ⫹ CH4(g)

(37)

In this manner, carbon can be successively added to and atomic hydrogen

Figure 49 Schematic showing attachment of CH3 radical on a dangling bond on the (100) surface of a diamond speed particle.

406

Subramaniam and Aithal

removed from the surface by CH3 radicals alone. Molecular dynamics calculations have shown that a high fraction of CH3 radicals (more than 0.5 at normal incidence) can be adsorbed on (100) diamond surfaces over kinetic energy ranges from 0.5 to 2 eV [108]. Other processes for abstraction of H atoms can also take place. One such process occurs by attack from vibrationally excited CO as suggested previously [109]: Cd — CH3 ⫹ CO(␯) → Cd — CH2 ⫹ HCO

(38)

The other can occur by the following process (after abstraction of an H atom from each of the CH3 groups): Cd — CH2• ⫹ Cd — CH2• → Cd — C2H4

(39)

In this scenario, the CO serves the sole purpose of exciting CH4 and producing CH3. It can also play the role of a spoiler by occupying a dangling bond and establishing a ketone-type structure. This structure is stable (unless attacked by a radical such as CH3) and can in principle inhibit growth by the CH3 mechanism just discussed. The ketone and other possible surface structures are shown in Fig. 50. The second possible mechanism requires the presence of both CO and CH4. Here, CO is added to the surface (the activation energy for adsorption supplied by vibrational excitation) so as to form a ketone-type structure (see Fig. 50a): —O C•d ⫹ CO → Cd — C —

(40)

The adsorbed CO is then attacked by a methyl radical, resulting in the conversion of the double bond between C and O to a single bond by ␤scission:

Figure 50 Schematic showing attachment of CO on the (100) surface of a diamond seed particle, forming a stable ketone-type structure (a) and an ether-type structure (b).

Laser-Assisted and Optical Pumping Techniques

— O ⫹ CH3 → Cd — C — O — CH3 Cd — C —

407

(41)

The resulting ether-like structure is shown in Fig. 51. After subsequent Hatom abstraction via (37), (38), or (39), an aldehyde-type structure can form momentarily, leading to ␤-scission of the underlying CO bond. This should result in evolution of formaldehyde (H2CO), leaving behind two dangling bonds (i.e., two active sites): Cd — C• — O — CH2• → Cd — C• ⫹ H2CO(g) The two dangling bonds on the carbon now allow another layer to be grown by repetition of the preceding processes. Both mechanisms serve to explain some of the key observations in the LECVD experiments. Both can account for the presence of small amounts of elemental oxygen (see the EDS spectrum shown in Fig. 42), and both give plausible descriptions of growth in an environment depleted of atomic hydrogen. Of the two, however, the first appears to explain additional observations reported in Ref. 98. The result of hydrogen abstraction by methyl radicals in the first mechanism would result in the evolution of CH4 from the surface, which would be indistinguishable from the methane already present in the gas phase. This would explain why only CO and CH4 were detected in the absorption measurements reported in Ref. 98. The question remains why CH3 radicals or perhaps HCO went undetected. A possible answer to this question is that CH3 concentration levels (according to our estimates, methyl radical concentrations on the order of nCH3 ⬇ 1012 cm⫺3 or 1018 m⫺3 are expected, or about 1 ppm) were well below detection limits. By this latter argument, the second mechanism cannot be ruled out because

Figure 51 Schematic showing the result of attack by the CH3 radical on a ketonetype CO structure occupying a dangling bond on the (100) surface of a diamond seed particle.

408

Subramaniam and Aithal

evolution of H2CO and HCO levels could have been below detection limits as well. Unfortunately, ultimate quantitative detection limits were not reported in Ref. 98 so that it is not possible to carry out further analysis of possible mechanisms. Given the current level of interest in synthesizing diamond at low substrate temperatures [76], both LCVD and LECVD appear to be significant advancements toward attaining this goal. Both laser-based techniques challenge the conventional myth that high substrate temperatures are necessary for diamond growth. Clearly, the environment of conventional diamond growth techniques is limited only by how energy is loaded into the gas mixture. Both techniques also offer some selectivity as can be seen by the conspicuous lack of codeposition of nondiamond carbon, prevalent in other CVD methods. LECVD also challenges our understanding of diamond growth mechanisms by displaying growth in an environment that is not rich in atomic hydrogen. Ultimately, that may be the reason why growth rates in the two methods are small—on the order of 1 ␮m/hr. For LECVD, as the absorbed power is on the order of 1 W, we may estimate the energy cost per carat as 1 J/sec ⫻ (1/10⫺12 cm3) ⫻ (1 hr ⫻ 3600 sec/hr) ⫻ (1/3.515) (cm3/g) ⫻ (0.2 g/carat) = 2 ⫻ 1014 J/carat. This value is many orders of magnitude higher than the corresponding value for diamond synthesis by an oxyacetylene flame. However, unlike LCVD and other methods of CVD diamond synthesis, the LECVD process scales with pressure and can be operated potentially at pressures on the order of 20 atm [77,83], so that the J/carat value can be reduced drastically. However, continuous films have yet to be produced by either LCVD or LECVD. The optical pumping technique shows tremendous promise for synthesis of other materials. For example, it has been shown that vibrationally excited nitrogen is important in the production of silicon nitride films [110]. Reference 110 reports that when silane was added to a flowing nitrogen discharge afterglow (where there is known to be copious amounts of vibrationally excited N2 molecules) and the reaction products made to impinge on a heated monocrystalline silicon substrate placed downstream, silicon nitride films were produced. In addition, reactions between the vibrationally excited N2 and silane in the afterglow led to formation of fine (micron-sized) particles of white silicon nitride powder [110]. Given the similarity in structure between CH4 and SiH4, it is likely that optically pumped CO/SiH4 mixtures could be used for deposition of silicon at low temperatures or for synthesis of silicon carbide. Indeed, Ref. 110 determined experimentally that about 15% of the vibrationally excited N2 quenching events resulted in excitation of the ␯3 mode of SiH4. Recently, a combination of optical pumping and application of a subbreakdown radio-frequency electric field has been shown to produce plasmas in CO/N2 gas mixtures seeded either with NO or

Laser-Assisted and Optical Pumping Techniques

409

O2 at total pressures ranging from 0.4 to 1.2 atm [111]. Such plasmas, although weakly ionized, contain substantial amounts of vibrationally excited CO and N2. The use of such plasmas for synthesis of novel materials remains to be explored. In this section, we have examined a laser excitation technique known as LECVD for diamond synthesis. Of significance is the fact that growth of diamond particles has been realized under conditions where the substrate is not heated at all [70,84] and without the need to hyperdilute the gas phase with hydrogen. In fact, diluents such as hydrogen poison the vibrational uppumping of CO and CH4, an effect that is central to the LECVD process. In the following section, we examine a potential means of extending the LECVD process to the condensed phase in order to increase growth rates. C.

Optical Pumping of Liquids

Energy transfer as well as the optical pumping of gases was discussed in Sec. V.B. Specifically, it was shown that small polyatomic molecules are very effective in retaining energy in their vibrational modes. For diatomic molecules, in the limit as VV exchange rates dominate VT and VR energy transfer rates, the population of various vibrational levels is described by the TRR distribution for anharmonic oscillators [see Eq. (17)]. As long as VV rates exceed VT, VR, and spontaneous emission rates, energy is effectively retained in the vibrational modes of molecules. It is important to point out that vibrational energy exchange among anharmonic oscillators is a collisional process. As long as the energy defect26 is carried away by a suitable cold bath, energy can be maintained for long times within the vibrational modes. This is irrespective of the pressure or density of the medium. In the gas phase, CO/Ar mixtures have been optically pumped in a manner so as to excite CO to very high vibrational states (near v = 40) at pressures of 20 atm [77]. Such strong vibrational excitation has also been observed in the liquid and solid phases. This is of significance if we are interested in driving chemical reactions, while maintaining some degree of control and selectivity over which reactions occur and to what extent. In this section, we describe extension of the LECVD technique described in Sec. V.A and V.B to the condensed phase. 26 Recall that the energy defect is defined as the energy difference between the total energy of the colliding molecules before the collision and that value after the collision. During anharmonic VV pumping, energy of magnitude equal to the energy defect is transferred from vibrational excitation to the external modes of translation and rotation. This results in eventual heating of the gas unless the gas is flowed or a suitable diluent is used to carry off this excess heat.

410

Subramaniam and Aithal

There exists a considerable literature on optical pumping of CO at elevated gas pressures as well as in the liquid and solid phases [77,112– 118]. These studies have been largely motivated by interest in fundamental energy transfer processes, isotope separation processes, and development of energy transfer lasers. There exists a relatively good coincidence between the laser lines emanating from a CO laser operating on the abundant isotope 12 16 C O and the isotopic liquid phase mixture of 88% 13C16O and 12% 13C18O. Several researchers have used this coincidence to excite CO vibrationally in the liquid phase and in the solid phase [112–118]. Remarkably, the condensed phase CO behaves as though it were a dense gas within the liquid argon matrix or on the solid single-crystal NaCl (100) surface [112–117]. Processes encountered in the gas phase (as discussed in Sec. V.A. and V.B) also occur in the condensed phases, and very high vibrational levels near v = 40 in CO have been observed [112]. Further, energy transfer from CO to small polyatomic molecules has also been reported [112,117,119,120], and not surprisingly, vibrationally excited CO and O2 have been reacted to form CO2 in the liquid phase at temperatures near 88 K. This is indeed a dramatic demonstration of the extreme disequilibrium between vibrational modes and external modes created by optical pumping. Although diamond growth from such optically pumped CO/CH4 liquid mixtures has never been demonstrated, the successful LECVD experiments described in Sec. V.B are strong indicators that diamond growth in optically pumped liquids could meet with success. Further, measured rate coefficients for energy transfer in the liquid and gas phases for a variety of mixtures of interest to diamond synthesis are quite encouraging. Table 7 gives measured values of rate coefficients for gas and liquid phases at a temperature of 80 K, as reported [119,120]. Remarkably, the variation in rate coefficients for vibrational energy transfer between gas and liquid states is less than 10% for the molecules listed in Table 7. Now, the rate of energy transfer depends on the product of the rate coefficient and number densities of the colliding partners. Therefore, the rate at which intermolecular vibrational energy transfer occurs increases by many orders of magnitude because the liquid phase is much denser than the gas phase. This has the potential to increase yield and growth rates. As many single-crystal materials are grown from melts in the liquid phase, optical pumping in the liquid phase may finally lead to epitaxial, single-crystal diamond films. The latter possibility is further bolstered by the fact that multilayers of ␣-form solid, single-crystal CO can be deposited on NaCl (100) substrates epitaxially because the lattice mismatch is less than 0.5% [116]. Further, the solid CO has been vibrationally excited up to levels as high as v = 23 [114], and as high as v = 30 at 22 K, by optical pumping using a CO laser [116]. In fact, this up-pumping is so rapid in the solid phase as to yield a population inversion (i.e., there are greater

Laser-Assisted and Optical Pumping Techniques

411

Table 7 Measured Rate Constants for Vibrational Energy Transfer in the Gas and Liquid Phases at 85 K System N2-12C16O N2-13C16O N2-13C18O 13 18 C O-O2 13 16 C O-O2 12 16 C O-O2 13 18 C O-CH4 13 16 C O-CH4 12 16 C O-CH4 N2-CH4 12 16 C O-CH4 13 16 C O-CH4 13 18 C O-CH4 12 16 C O-CF4 13 18 C O-CF4 12 16 C O-CD4 13 16 C O-CD4 13 18 C O-CD4 12 16 C O-CD3H 13 16 C O-CD3H 13 18 C O-CD3H

⌬E (cm⫺1) 187 234 286 487 540 587 509 562 609 796

kLiquid (cm3/sec) (3.2 ⫾ 0.2) (1.2 ⫾ 0.1) (3.9 ⫾ 0.1) (1.54 ⫾ 0.02) (7.0 ⫾ 0.5) (4.2 ⫾ 0.1) (4.03 ⫾ 0.09) (2.08 ⫾ 0.05) (1.33 ⫾ 0.05) (1.3 ⫾ 0.1) (1.33 ⫾ 0.05) (2.08 ⫾ 0.05) (4.03 ⫾ 0.09) (8.6 ⫾ 0.1) (7.50 ⫾ 0.09) (2.48 ⫾ 0.04) (2.28 ⫾ 0.08) (1.47 ⫾ 0.03) (1.15 ⫾ 0.02) (1.58 ⫾ 0.04) (1.26 ⫾ 0.02)

⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻

kGas (cm3/sec) 10⫺15 10⫺15 10⫺16 10⫺17 10⫺18 10⫺18 10⫺15 10⫺15 10⫺15 10⫺16 10⫺15 10⫺15 10⫺15 10⫺14 10⫺16 10⫺13 10⫺13 10⫺13 10⫺12 10⫺13 10⫺13

(3.6 ⫾ 0.2) (1.1 ⫾ 0.1) (3.79 ⫾ 0.08) (1.50 ⫾ 0.03) (7.0 ⫾ 0.9) (4.1 ⫾ 0.2) (4.17 ⫾ 0.08) (2.07 ⫾ 0.06) (1.27 ⫾ 0.04) (1.5 ⫾ 0.2) (1.27 ⫾ 0.04) (2.07 ⫾ 0.06) (4.17 ⫾ 0.08) (1.98 ⫾ 0.04) (1.0 ⫾ 0.1) (4.60 ⫾ 0.08) (4.3 ⫾ 0.1) (3.5 ⫾ 0.1) (1.32 ⫾ 0.03) (3.3 ⫾ 0.1) (2.38 ⫾ 0.05)

⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻ ⫻

kLiquid/kGas 10⫺15 10⫺15 10⫺16 10⫺17 10⫺18 10⫺18 10⫺15 10⫺15 10⫺15 10⫺16 10⫺15 10⫺15 10⫺15 10⫺13 10⫺15 10⫺13 10⫺13 10⫺13 10⫺12 10⫺13 10⫺13

0.97 1.09 1.03 1.03 1.00 1.02 0.97 1.00 1.05 0.87 1.05 1.00 0.97 0.43 0.75 0.54 0.53 0.42 0.87 0.48 0.53

⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.08 0.13 0.03 0.03 0.15 0.06 0.03 0.04 0.06 0.13 0.06 0.04 0.03 0.01 0.08 0.01 0.02 0.02 0.02 0.02 0.01

Source: Refs. 119, 120.

numbers of molecules in the higher vibrational states than in the v = 0 or ground vibrational state). We now describe an experiment that demonstrates some of the salient features of optical pumping of CO in the liquid phase [70]. A CO laser designed to operate cw (5 to 8 W on all lines) with some laser output on the v = 1 → v = 0 transition was used to pump CO in a liquid argon matrix. A schematic of the cell containing the cryogenic mixture is shown in Fig. 52. This cell is identical to that used in the experiments reported in Ref. 112, and details may be found therein. The output from the CO laser was focused into the center of the cryogenic cell containing a solution of CO in liquid argon, using a ZnSe lens (1 in. diameter, focal length of 9 in.). The liquid mixture was produced from an initially gaseous mixture of CO and argon, which was condensed upon being cooled by the liquid argon placed in a dewar immediately above the cell. A schematic of the mixing chamber

412

Subramaniam and Aithal

Figure 52 Schematic of the cell containing the cryogenic mixture of CO and argon. (From Ref. 70.)

is shown in Fig. 53. The CO/argon mixture is prepurified before allowing it to condense into the cell. The purification procedure consists of using the cold finger connected to the mixing chamber. The gases are condensed into the cold finger and, after some time, allowed to evaporate back into the mixing chamber while a small part of the cold finger is still immersed in liquid nitrogen. This process is repeated 15 times. Reference 70 reported pumping of CO/argon gas mixtures where the initial partial pressure of CO in the gas phase was varied between 70 mtorr and 30 torr, while maintaining a total pressure of 2 atm. The CO was an isotopic mixture comprising 88% 13C16O and 12% 13C18O. This isotopic mixture was selected because of its excellent coincidence in the liquid phase with CO laser lines when the CO laser was operated on the normally abundant 12C16O isotope. Following the aforementioned purification procedure, the CO/argon mixture was condensed into the cell and irradiated by the cw CO laser beam. Power measurements before the cell were in the range of 3.5 to 5 W. Using a side port, IR emission was monitored from the optically pumped liquid CO. A representative spectrum of the first overtone (⌬v = 2)

Laser-Assisted and Optical Pumping Techniques

413

Figure 53 Schematic showing the mixing chamber used in optical pumping of liquid CO/argon mixture. (From Ref. 70.)

transitions from the liquid CO for an initial gas-phase CO partial pressure of 9 torr, after background blackbody subtraction, is shown in Fig. 54. It is important to note that this is not a low-resolution spectrum. Rotational lines within each vibrational band can be either totally absent (as in the case of CH4) or substantially narrowed in the liquid state as compared with similar gas-phase spectra. It is immediately apparent from the emission spectrum that states at least as high as v = 38 (see the energy level diagram for CO

414

Subramaniam and Aithal

Figure 54 A representative spectrum of the first overtone (⌬v = 2) transitions from liquid CO in liquid argon, optically pumped using a CO laser. This spectrum was obtained for an initial gas-phase CO partial pressure of 9 torr and after background blackbody subtraction. (From Ref. 70.)

shown in Fig. 34) of the heavier isotope 13C18O are substantially populated. In addition, peaks corresponding to isotopes of carbon dioxide 13C16O16O, 13 16 18 C O O, and 13C18O18O are also observed. This is indicative of chemical reaction occurring among vibrationally excited CO molecules, possibly via channels (29). Production of CO2 has been observed previously but from vibrationally excited CO doped with O2 [112]. Assuming the Einstein A coefficients for spontaneous emission from these vibrational states are comparable between gas and liquid phases, relative populations can be inferred from the emission spectrum. This is shown in Fig. 55, displaying a fully VV pumped distribution in the heavier isotope. It is clear that many (if not all) of the processes observed in gas-phase optically pumped CO occur in the liquid phase as well. Anharmonic VV pumping resulting in highly vibrationally excited CO (levels as high as v = 38 or as energetic as 8 or 9 eV) is observed in both gas and liquid phases. Intermolecular energy transfer between CO and CH4 has been reported, although IR emission from highly excited methane has not yet been observed. Chemical reaction from vibrationally excited CO, and between CO and O2, yielding CO2 is observed in both the gas and liquid phases as well. Given this close similarity in the occurrence of key processes between gas and liquid phases and the success of the LECVD process for diamond synthesis, the likelihood that diamond can be synthesized from optically pumped liquid CO/CH4 /Ar mixture appears to be high. Should diamond synthesis occur in

Laser-Assisted and Optical Pumping Techniques

415

Figure 55 Relative populations of vibrationally excited CO in liquid argon inferred from the emission spectrum. This was obtained by assuming that the Einstein A coefficients for spontaneous emission from these vibrational states are comparable between gas and liquid phases. Note the fully VV pumped distribution in the heavier isotope. (From Ref. 70.)

such an optically pumped liquid mixture, it should be more rapid than in the gas phase because the rates of energy transfer scale as the square of the density. Finally, the production of flaw-free, cubic, solid CO as well as its extreme vibrational excitation by optical pumping resulting in population inversion bodes well for adsorbed CH4 species. Although growth of diamond films using such a technique has not been demonstrated, it is nevertheless worth exploring. Given the fact that many growth techniques for single crystals of other materials involve the liquid phase, this approach of optically pumping liquids appears promising for synthesis of single-crystal diamond. This is important before diamond can be considered as a candidate material for electronic applications.

VI.

SUMMARY

This chapter has discussed synthesis of diamond from gas and liquid phases using laser sources. The laser-based techniques include surface heating, py-

416

Subramaniam and Aithal

rolysis, ablation (LPVD), photolysis (LCVD), and optical pumping (LECVD). The laser-based techniques do not necessarily supplant existing CVD methods but offer the prospect of lowering growth temperatures, selectively depositing diamond while inhibiting the deposition of nondiamond carbon, and perhaps enhancing growth rates. In the case of rapid gas-phase heating such as pyrolysis, the laser simply serves as a source of thermal heating and is therefore an inefficient means of activating the gas. In this instance, it does not provide any advantages such as selectivity or enhanced quality of the deposit. Ablation and heating of surfaces can benefit from laser sources because reactions may need to be driven locally, without bulk heating of the growth medium, target, or substrate. In contrast, photolysis and optical pumping techniques use attributes of molecular structure to drive specific reaction pathways to form products. This is difficult or not even possible using conventional growth methods. Many of these laser-based processes have already demonstrated diamond growth, albeit at different rates. Although capital costs vary widely, a useful measure of the relative efficacy of each process is the joule-per-carat value. This measure considers the power required for the process as well as growth rate simultaneously. The pyrolysis and photolysis techniques exhibit growth rates on the order of microns per hour, as does the optical pumping technique at low gas pressures. However these exhibit drastically different energy costs in J/carat: O(108), O(107)–O(1014), and O(1014), respectively. These should be contrasted with typical values of O(108) J/carat for HFCVD27 and O(106) J/carat for synthesis using an oxyacetylene flame.28 The laser-plasma technique is equivalent to but less efficient energetically than the conventional oxyacetylene flame, exhibiting growth rates on the order of 10 to 100 ␮m per hour at a power level of 2.5 kW. Of all these laser processes, however, the well-publicized but as yet unverified QQC process reports growth rates on the order of millimeters per hour (i.e., microns per second) at a cost of about 71 J/carat. Clearly, if cost is the prevailing factor, the QQC process appears to be the best (provided its claims are substantiated), while some of the laser pyrolysis and laser photolysis methods are competitive with HFCVD and oxyacetylene flame synthesis. It would be premature to eliminate optical pumping based on a comparative measure such as joules per carat. Optical pumping, unlike the other methods, inherently scales directly with pressure. Therefore, for approximately the same power expenditure, the growth rate can be significantly increased sim-

27

This value is based on a power expenditure of 100 W and deposition over a 1 cm2 area at 1 ␮m/hr. 28 See footnote 9.

Laser-Assisted and Optical Pumping Techniques

417

ply by raising the pressure. This would reduce the J/carat value for this process from the unreasonable O(1014) and make it competitive with the others. Further enhancement in growth rate is possible in the liquid phase, in principle, although diamond growth in an optically pumped liquid is yet to be demonstrated. Moreover, this method has the added virtue of lower growth temperatures as well as process selectivity. Some of the more nascent but novel approaches such as LECVD, laser photolysis, and the QQC process have been critically examined. Some appear promising in terms of scaling to higher growth rates, lower growth temperatures, and providing selectivity. Conventional CVD and some of the laser-based (such as pyrolysis and laser-plasma) methods are useful for substrates that can withstand high temperatures and the presence of large amounts of atomic hydrogen or oxygen. The more novel of the laser-based methods show some promise of being able to circumvent these difficulties. It is particularly noteworthy that the optical pumping technique is capable of activating gases over a wide a range of pressures, as well as liquids. There are precedents for growth of single crystals of various materials from the liquid phase. Optical pumping lays the groundwork for development of a new class of processes that employ laser excitation to drive energy transfer and chemistry in cryogenic liquids and may someday play a role in the production of single-crystal diamond films.

ACKNOWLEDGMENTS Partial support from grants MSS-9157303 and MSS-9102076 from the National Science Foundation, The Ohio State University College of Engineering, The Center for Materials Research, The Ohio Board of Regents Investment Fund Program, and the Directorate of Defense Research and Engineering (DDR&E) under the Air Plasma Ramparts MURI program managed by the Air Force Office of Scientific Research is gratefully acknowledged. The authors thank Dr. Gregory Hall, Chemistry Department, Brookhaven National Laboratory for many helpful comments and suggestions.

REFERENCES 1. 2.

M Capitelli. Non-Equilibrium Vibrational Kinetics. Vol 39, Topics in Current Physics. New York: Springer-Verlag, 1986. JW Rich. Relaxation of molecules exchanging vibrational energy. In: EW MacDaniel, WL Nighan, eds. Applied Atomic Collision Physics. Vol 3: Gas Lasers. New York: Academic Press, 1982.

418

Subramaniam and Aithal

3.

JT Yardley. Introduction to Molecular Energy Transfer. New York: Academic Press, 1980. H Okabe. Photochemistry of Small Molecules. New York: John Wiley & Sons, 1978. GR Fowles. An Introduction to Modern Optics. 2nd ed. New York: Dover, 1989. K Thyagarajan, AK Ghatak. Lasers—Theory and Applications. New York: Plenum, 1981. JI Steinfeld, JS Francisco, WL Hase. Chemical Kinetics and Dynamics. Englewood Cliffs, NJ: Prentice Hall, 1989. JI Steinfeld. Molecules and Radiation—An Introduction to Modern Molecular Spectroscopy. 2nd ed. Cambridge: MIT Press, 1986. JT Verdeyen. Laser Electronics. 2nd ed. Englewood Cliffs, NJ: Prentice Hall, 1989. A Siegman. Lasers. Mill Valley: University Science Books, 1986. PR Buerki, S Leutwyler. Homogeneous nucleation of diamond powder by CO2-laser-driven gas-phase reactions. J Appl Phys 69:3739–3744, 1991. K Kitahama, K Hirata, H Nakamatsu, S Kawai, N Fujimori, T Imai, H Yoshino, A Doi. Synthesis of diamond by laser induced chemical vapor deposition. Appl Phys Lett 49:634–635, 1986. K Kitahama. Reinvestigation of the carbon films prepared by ArF excimer laser induced chemical vapor deposition. Appl Phys Lett 53:1812–1814, 1988. VI Konov, SA Uglov. Laser-plasma synthesis of diamond films. Quantum Electron (Kvantovaya Elektron) 28:281–282, 1998. EA Frey. Flame deposition of diamond on silicon. MS thesis, Ohio State University, 1992. EA Frey, A Tamhane, JHD Rebello, SA Dregia, VV Subramaniam. Morphological variations in flame-deposited diamond. J Mater Res 9:625–630, 1994. EA Frey, JHD Rebello, VV Subramaniam. Flame synthesis of diamond on silicon. Presented at the Fifth International Conference on Surface Modification Technologies, Birmingham, UK, September 1991; also see Proceedings of Surface Modification Technologies V, TS Sudarshan, JF Braza, eds. The Institute of Materials, 1992, pp 343–356. DV Fedoseev, VL Bukhovets, IG Varshavskaya, AV Lavrent’ev, and BV Derjaguin. Carbon 21:237–241, 1983. M Alam, T DebRoy, R Roy, E Breval. Diamond formation in air by the Fedoseev-Derjaguin laser process. Carbon 27:289–294, 1989. KS Harshavardhan, RS Yalamanchi, LK Rao. Formation of crystalline diamond from amorphous diamond-like carbon films by pulsed laser irradiation. Appl Phys Lett 55:351–353, 1989. S Prawer, R Kalish, M Adel. Pulsed laser treatment of diamond-like carbon films. Appl Phys Lett 48:1585–1587, 1986. J Narayan, VP Godbole, CW White. Science 252:416, 1991. JF Prins, HL Gaigher. In: R Messier, JT Glass, eds. New Diamond Science and Technology. Pittsburgh: Materials Research Society, 1991, p 561.

4. 5. 6. 7. 8. 9. 10. 11. 12.

13. 14. 15. 16. 17.

18. 19. 20.

21. 22. 23.

Laser-Assisted and Optical Pumping Techniques 24.

25. 26. 27.

28. 29. 30. 31. 32.

33. 34. 35. 36. 37. 38. 39.

40. 41.

42.

43. 44. 45.

419

S Tong Lee, S Chen, J Agostinelli, G Braunstein, LJ Huang, WM Lau. Laser processing of carbon-implanted Cu, Ni, and Co crystals: an attempt to grow diamond films. Appl Phys Lett 60:2213–2215, 1992. DY Tzou. Macro- to Microscale Heat Transfer: The Lagging Behavior. Washington, DC: Taylor & Francis, 1997. JC Maxwell. On the dynamical theory of gases. Philos Trans Lond 157:49– 88, 1867. C Cattaneo. Sur une forme de L’e´quation de la chaleur e´liminant le paradoxe d’une propagation instantane´e (A form of heat conduction equation which eliminates the paradox of instantaneous propagation). C R Acad Sci 247:431– 433, 1958: also see Sulla conduzione de calore. Atti Semin Mat Fis Univ Modena 3:3, 1948. FP Incropera, DP DeWitt. Fundamentals of Heat and Mass Transfer. 4th ed. New York: John Wiley & Sons, 1996. See Manu Eng, February 1995. See Chem Eng News, January 22, 1996, p 8. See Lucky lab accident yields new recipe for diamond coating. The New York Times Science Times, March 12, 1996, p B5. P Mistry, MC Turchan, S Liu, GO Granse, T Baurmann, MG Shara. A revolutionary diamond synthesis technique: the QQC materials deposition process. Innovations Mater Res 1(2):193–207, 1996. BP Fairand, BA Wilcox. J Appl Phys 43:3893, 1972. AH Clauer, BP Fairand, J Holbrook. J Appl Phys 50:1497, 1979. P Peyre, R Fabbrot. Laser shock processing: a review of the physics and applications. Opt Quantum Electron 27:1213–1229, 1995. B Dubrujeaud, M Jeandin. Cladding by laser shock processing. J Mater Sci Lett 13:773–775, 1994. G Davies, ed. Properties and Growth of Diamond. London: Institution of Electrical Engineers, 1994. YL Orlov. The Mineralogy of Diamond. New York: John Wiley & Sons, 1974. SB Ogale, AP Malshe, SM Kanetkar, ST Kshirsagar. Formation of diamond particulates by pulsed ruby laser irradiation of graphite immersed in benzene. Solid State Commun 84:371–373, 1992. KA Cherian. The dissolution and metastable autoepitaxial recrystallization of diamond. PhD dissertation, Sardar Patel University, India, 1983. KA Cherian. Nickel-assisted metastable diamond formation (surface recrystallization) in a dissolution medium at atmospheric pressure. Surf Coat Technol 47:127–143, 1991. KA Cherian. Metastable surface recrystallization of diamond from the liquid phase: substrate effects. In: Y Tzeng, ed. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier, 1991, pp 389–394. XZ Zhao, R Roy, KA Cherian, A Badzian. Hydrothermal growth of diamond in metal-C-H2O systems. Nature 385:513–515, 1997. RC DeVries. Diamonds from warm water. Nature 385:485, 1997. SS Wagal, EM Juengerman, CB Collins. Diamond-like carbon films prepared with a laser ion source. Appl Phys Lett 53:187–188, 1988.

420

Subramaniam and Aithal

46.

YA Bykovskii, PP Kozlenkov, IN Nikolaev, EV Charyshkin. Use of microsecond duration laser light to deposit diamond-like carbon films. Sov Tech Phys Lett 14:979–980, 1988. A Richter, I Muhling, M Klose. Properties of laser plasma deposited carbon films. J Noncryst Solids 107:128–131, 1988. T Sato, S Furuno, S Iguchi, M Hanabusa. Diamond-like carbon films prepared by pulsed laser evaporation. Appl Phys A45:355–360, 1988. J Krishnaswamy, A Rangan, A Srivatsa, J Narayan, Y Cong, R Collins, K Vedam. Low-temperature deposition of hard amorphous diamond-like films by laser evaporation. MRS Symp 129:219–224, 1989. J Krishnaswamy, A Rangan, J Narayan, K Vedam, CJ McHargue. Thin film deposition by a new laser ablation and plasma hybrid technique. Appl Phys Lett 54:2455–2457, 1989. RT Demers, DG Harris. Pulsed laser deposition of amorphous carbon films. Proc SPIE 1146:48–55, 1989. AP Malshe, SM Chaudhari, SM Kanetkar, SB Ogale, SV Rajarshi, ST Kshirsagar. Properties of carbon films deposited by pulsed laser vaporization from pyrolytic graphite. J Mater Res 4:1238–1242, 1989. RT Demers, DG Harris. KrF excimer laser deposition of diamond-like carbon films with germanium interlayers. Proc SPIE 1990. BR Mehta, EA Ogryzlo. Deposition of diamond-like films by laser ablation. Proceedings of the International Conference on Thin Films ICTF8/ICMC17, San Diego, 1990. F Davanloo, EM Juengerman, DR Jander, TJ Lee, CB Collins. Amorphic diamond films produced by a laser plasma source. J Appl Phys 67:2081– 2087, 1990. S Latsch, H Hiraoka. Diamond films by the excimer laser photoablation of polymers. J Mater 65–65, 1994. MC Polo, J Cifre, G Sanchez, R Aguiar, M Varela, J Esteve. Pulsed laser deposition of diamond from graphite targets. Appl Phys Lett 67:485–487, 1995. Precision Optics and Components for Lasers & Scientific Optical Systems. Janos Technology Inc, HCR #33, Townshend, VT. from Atomergic Chemetals Corporation, Farmingdale, New York. DC Harris, MD Bertolucci. Symmetry and Spectroscopy—An Introduction to Vibrational and Electronic Spectroscopy. New York: Dover, 1989. PV Hobbs. Basic Physical Chemistry for the Atmosphere Sciences. New York: Cambridge University Press, 1995. MP Irion, KL Kompa. UV laser photochemistry of acetylene at 193 nm. Appl Phys B27:183–186, 1982. Y Goto, T Yagi, H Nagai. Synthesis of diamond films by laser induced chemical vapor deposition. MRS 129:213–217, 1989. JE Butler, RL Woodin. Thin film diamond growth mechanisms. Philos Trans Soc Lond A342:209–224, 1993. SJ Harris. Mechanisms for diamond growth from methyl radicals. Appl Phys Lett 56:2298–2300, 1990.

47. 48. 49.

50.

51. 52.

53. 54.

55.

56. 57.

58. 59. 60. 61. 62. 63. 64. 65.

Laser-Assisted and Optical Pumping Techniques 66. 67. 68.

69. 70. 71.

72.

73.

74. 75. 76. 77. 78. 79. 80. 81. 82. 83.

84.

85.

421

JC Angus, A Argoitia, R Gat, Z Li, M Sunkara, L Wang, Y Wang. Chemical vapor deposition of diamond. Philos Trans Soc R Lond A342:195–208, 1993. GW Tyndall, NP Hacker. KrF* laser induced chemical vapor deposition of diamond. MRS Symp 1990. JHD Rebello, VV Subramaniam, TS Sudarshan. Diamond growth by laserdriven reactions in a CO/hydrogen mixture. Appl Phys Lett 62:899–901, 1993. JHD Rebello, VV Subramaniam. Diamond synthesis at low temperatures. J Mater 46:60–63, 1994. JHD Rebello. Synthesis of diamond by selective laser excitation of gaseous precursors. PhD dissertation, Ohio State University, 1995. TR Anthony, WF Banholzer, JF Fleischer, L Wei, PK Kuo, RL Thomas, RW Pryor. Thermal diffusivity of isotopically enriched 12C diamond. Phys Rev B 2:1104–1111, 1990. H Itoh, T Nakamura, H Iwahara, H Sakamoto. Morphology and growth process of single crystalline diamond grown by microwave plasma CVD in the CO-H2 system. MRS International Conference Proceedings on New Diamond Science and Technology, 1991, pp 479–484. G Meijer, AM Wodtke, H Voges, H Schluter, P Andresen. State-selective detection of CO using a tunable ArF excimer laser. J Chem Phys 89:2588– 2589, 1988. W Hack, W Langel. Formation of C2(e3⌸g , C1⌸g) in the photolysis of CO ( X1兺g⫹) at 193 nm with an ArF laser. Chem Phys Lett 81:387–390, 1981. V Babu, VV Subramaniam. Numerical solutions to nozzle flows with vibrational non-equilibrium. J Thermophys Heat Transfer 9:227–232, 1995. Y Muranaka, H Yamashita, H Miyadera. Worldwide status of low-temperature growth of diamond. Diamond Relat Mater 3:313–318, 1994. JW Rich, RC Bergman. C2 and CN formation by optical pumping of CO/Ar and CO/N2 /Ar mixtures at room temperature. Chem Phys 44:53, 1979. RC Bergman, GF Homicz, JW Rich, GL Wolk. Chem Phys 78:1281, 1983. RL DeLeon, JW Rich. Vibrational energy exchange rates in carbon monoxide. Chem Phys 107:283–292, 1986. P Brechignac, JP Martin, G Taieb. IEEE J Quantum Electron QE-10:797, 1974. H Dunnwald, E Siegal, W Urban, JW Rich, GF Homicz, MJ Williams. Chem Phys 94:195, 1985. W Urban, JX Lin, VV Subramaniam, M Havenith, JW Rich. Treanor pumping of CO initiated by a CO laser. Chem Phys 130:389–399, 1989. JHD Rebello, DL Straub, VV Subramaniam. Diamond growth from a CO/ methane mixture by laser excitation of CO: laser-excited chemical vapor deposition. J Appl Phys 72:1133–1136, 1992. VV Subramaniam, JHD Rebello. Laser-based synthesis of diamond at low temperatures. Presented at the 187th Meeting of the Electrochemical Society, Reno, NV May 21–26, 1995. CE Treanor, JW Rich, RG Rehm. Vibrational relaxation of anharmonic oscillators with exchange dominated collisions. J Chem Phys 48:1789, 1968.

422

Subramaniam and Aithal

86.

BM Smirnov. Physics of Weakly Ionized Gases. Moscow: Mir Publishers, 1981. C Flament, T George, KA Meister, JC Tufts, JW Rich, VV Subramaniam, JP Martin, B Piar, MY Perrin. Nonequilibrium vibrational kinetics of carbon monoxide at high translational mode temperatures. Chem Phys 163:241, 1992. H Kildal, TF Deutsch. Optically pumped infrared V-V transfer lasers. Appl Phys Lett 27:500–502, 1975. DS Anex, GE Ewing. Transfer and storage of vibrational energy in liquids: collisional up-pumping of carbon monoxide in liquid argon. J Phys Chem 90: 1604–1610, 1986. CKN Patel. CW laser on the vibrational-rotational transitions of CO. Appl Phys Lett 7:246, 1965. Provided courtesy of KA Essenhigh, Molecular Energy Transfer Laboratory, Non-Equilibrium Thermodynamics Laboratories, Department of Mechanical Engineering, Ohio State University, August 7, 1998 (unpublished). IV Adamovich, S Saupe, MJ Grassi, O Schulz, S Macheret, JW Rich. Vibrationally stimulated ionization of carbon monoxide in optical pumping experiments. Chem Phys 173:491–504, 1993. Y Kaufman, P Avivi, F Dothan, H Keren, J Malinowitz. Ion clusters in HeCO and Ar-CO glow discharges. J Chem Phys 72:2606–2611, 1980; also E Plonjes, P Palm, unpublished results from optical pumping experiments at Ohio State University, 1998. DM Gruen. Nucleation, growth, and microstructure of nanocrystalline diamond films. MRS Bull 32–35, 1998. PC Redfern, DA Horner, LA Curtiss, DM Gruen. J Phys Chem 100:11654, 1996. DL Straub, Laser-excited chemical vapor deposition (LECVD) of carbon films. MS thesis, Ohio State University, 1991. EE Bell, HH Nielsen. The infra-red spectrum of acetylene. J Chem Phys 18: 1382–1394, 1950. F Chaouchi-Gueoual. Excitation vibrationnelle par transferts collisionnels CO ( X1兺⫹, v)-CH4 . Production de diamant. DSc thesis, University of Paris, Orsay, 1996. SJ Harris, DG Goodwin. J Phys Chem 97:23, 1993. S Skokov, B Weiner, M Frenklach. Elementary reaction mechanisms for growth of diamond (100) surfaces from methyl radicals. J Phys Chem 98: 7073–7078, 1994. S Skokov, B Weiner, M Frenklach. Molecular-dynamics study of oxygenated (100) diamond surfaces. Phys Rev B 49:11374–11382, 1994. S Skokov, B Weiner, M Frenklach. Elementary reaction mechanism of diamond growth from acetylene. J Phys Chem 98:8–11, 1994. DP Belyung, A Fontijn. Gas phase reactions between hydrocarbons and metal oxides: the AlO ⫹ CH4 reaction from 590 to 1380 K. J Phys Chem 97:3456, 1993. RR Baldwin, A Keen, RW Walker. Studies of the decomposition of oxirane

87.

88. 89.

90. 91.

92.

93.

94. 95. 96. 97. 98.

99. 100.

101. 102. 103.

104.

Laser-Assisted and Optical Pumping Techniques

105. 106. 107. 108. 109. 110.

111.

112.

113. 114.

115. 116. 117.

118.

119.

120.

423

and of its addition to slowly reacting mixtures of hydrogen and oxygen at 480⬚C. J Chem Soc Faraday Trans 1 80:435–456, 1984. JT Yardley, CB Moore. Vibrational energy transfer in methane. J Chem Phys 49:1111–1125, 1968. P Hess, AH Kung, CB Moore. Vibration → vibration energy transfer in methane. J Chem Phys 72:5525–5531, 1980. P Hess, CB Moore. Vibrational energy transfer in methane and methane-raregas mixtures. J Chem Phys 65:2339–2344, 1976. DR Alfonso, SE Ulloa. Molecular-dynamics simulations of methyl-radical deposition on diamond (100) surfaces. Phys Rev B 48:12235–12239, 1993. JA Mucha, DL Flamm, DE Ibbotson. On the role of oxygen and hydrogen in diamond forming discharges. J Appl Phys 65:3448–3452, 1989. LG Piper. The importance of vibrationally excited nitrogen in silicon nitride deposition systems. Report WL-TR-93-2097, Aero Propulsion and Power Directorate, Wright Laboratory, Wright-Patterson AFB, August 1993. E Ploenjes, P Palm, WC Lee, WR Lempert, IV Adamovich. Unconditionally stable high-pressure nonequilibrium molecular plasma. Paper AIAA-20002627 presented at the 31st AIAA Plasmadynamics and Lasers Conference, Denver, June 19–22, 2000. WF Calaway, G Ewing. Vibrational relaxation of small molecules in the liquid phase: liquid nitrogen doped with O2 , CO, and CH4 . J Chem Phys 63:2842, 1975. H Dubost, R Charneau. Laser studies of vibrational energy transfer and relaxation of CO trapped in solid neon and argon. Chem Phys 12:407, 1976. N Legay-Sommaire, F Legay. Observation of strong vibrational population inversion by CO laser excitation of pure solid carbon monoxide. IEEE J Quantum Electron QE-16:308, 1980. HC Chang, HH Richardson, GE Ewing. Epitaxial growth of CO on NaCl (100) studied by infrared spectroscopy. J Chem Phys 89:7561–7568, 1988. HC Chang, GE Ewing. Vibrational energy transfer and population inversion in CO overlayers on NaCl (100). J Chem Phys 94:7635–7641, 1990. E Weitz, G Flynn. Vibrational energy flow in the ground electronic states of polyatomic molecules. In: J Jortner, RD Levine, SA Rice, eds. Advances in Chemical Physics. Vol XLVII: Photoselective Chemistry (Part 2). 1981, pp 185–235. R Disselkamp, GE Ewing. High vibrational states of carbon monoxide in liquid argon: overtone intensity enhancement and reactions with oxygen. J Phys Chem 93:6334–6339, 1989. MR Buckingham, HT Williams, RS Pennington, CJSM Simpson, MM Maricq. A comparison of rate constants for vibrational relaxation in liquid argon and in the gas phase at the same temperature. Chem Phys 98:179–195, 1985. HT Williams, MH Purvis, CJSM Simpson. Non-resonant VV transfer in liquid air and in the gas phase at 85 K. Chem Phys 115:7–14, 1987.

10 CVD Diamond Solutions for Machining and Other Mechanical Applications Brian L. Cline Cline Innovations, LLC, Sterling, Massachusetts

James M. Olson Fairchild Semiconductor, South Portland, Maine

I. A.

INTRODUCTION Growth Engine for Advanced Cutting Tool Materials

The machining industry is always seeking a means of improving productivity and machined part quality. Over the years, this relentless demand has resulted in countless iterations of productivity improvements. Today, the machine tool industry produces automated high-speed machining centers that have the ability to machine complex shapes quickly with high accuracy and minimal operator assistance. These dramatic machine tool improvements create an interrelated demand for improved cutting tool materials that enhance productivity by extending tool life under increasingly aggressive machining conditions. Improvements to cutting tool materials are also needed for enhanced cutting tool performance in machining many advanced workpiece materials. For example, the aerospace and automotive industries are utilizing increasing quantities of lightweight materials such as high-silicon aluminum alloys, fiber-reinforced composites, and metal matrix composites (MMCs), which offer unique options for component design. Unfortunately, many of these enabling materials are very abrasive and difficult—if not impossible—to machine using conventional cutting tools [1–3]. 425

426

Cline and Olson

This chapter contains an overview of how and why chemical vapor deposition (CVD) diamond technology is beginning to assume a key role in the materials evolution for advanced cutting tool and specialty wear part applications. Section I continues with a history of cutting tool materials development including high-speed steel, cemented carbides, hard coatings, and synthetic diamond materials. Section II overviews the issues and challenges associated with CVD diamond tool design and manufacture. This section also includes discussion of CVD diamond properties and application specifics that can affect the performance of CVD diamond cutting tools as well as common approaches to CVD diamond tool design. Section III is focused on microstructural features of CVD diamond that affect the performance of diamond coatings as well as freestanding diamond in mechanical applications. Section IV contains cutting performance information for various CVD diamond tool types used in a wide range of materials. The chapter closes with an overview of other types of mechanical component applications in Sec. V followed by a summary in Sec. VI. B.

Evolution of Conventional Cutting Tool Materials

1.

Carbon Tool Steels and Nonferrous Cast Alloys

Early industrial tooling materials included carbon tool steels—iron-based materials with a wear-resistant ‘‘case’’ surrounding a tougher FeC alloy. Although these materials display high fracture resistance, their low hardness and low wear resistance limit their wear life in industrial machining of metals and highly abrasive materials. The introduction of high-speed steels (containing varying amounts of W, Cr, V, Mo, and Co) led to the rapid replacement of carbon tool steels in metalcutting applications due to marked improvements in wear and fracture resistance. The added wear resistance and high-temperature hardness properties (‘‘hot hardness’’) of nonferrous cast alloys (containing varying amounts of Cr, W, C, and Co) allowed further improvements in cutting capabilities for selected metalcutting operations. 2.

Cemented Carbides Including Tungsten Carbide

The subsequent development of cemented carbides in the late 1920s provided an immediate advantage over traditional high-speed steels and cast alloys [4]. The combination of properties offered by cemented carbide cutting tools is typically attained by sintering large volumes of ceramic grains with 3 to 15% (by volume) of metallic binder and various sintering aids. Cobalt-sintered tungsten carbide combines wear-resistant, hard tungsten carbide (WC) ceramic particles with the fracture resistance of a ductile, metallic, Co-based binder phase. Many ‘‘tungsten carbide’’ grade formulations are

CVD Diamond Solutions for Machining

427

readily available for a wide range of applications in the cutting tool and wear part industry. The metal sintering approach to engineering cutting tool materials bypasses the challenges of introducing the high hardness and wear resistance of otherwise brittle ceramic materials into cutting tools. Although WC-Co cutting tools have lower fracture resistance than high-speed steel, their mechanical properties are suitable for a very wide range of heavy-duty applications. The higher hardness and improved hot hardness characteristics of cemented carbides allow operation at cutting speeds three to four times higher than with high-speed steel. C.

The Historic Use of Diamond and Ceramic Cutting Tool Materials

1.

Single-Crystal Diamond

For decades, natural and high-pressure, high-temperature (HPHT) singlecrystal diamonds have been used for specialty cutting and wear applications. Single-crystal diamond tools are known to have unmatched edge quality and create excellent surface finish due to their precision, continuous diamond edges. Applications include woodworking, contact lens machining, jewelry manufacture, and nonferrous metal finishing. The cost and difficulty of fabricating tools of complex geometries limit the usage of single-crystal tips to simple shapes relative to other tooling materials. However, the hardness, thermal stability, and corrosion resistance of single-crystal diamond tooling continue to provide added value in many demanding industrial applications. Single-crystal tools are also limited by their low fracture resistance, anisotropic wear behavior, and morphological shape constraints relative to other forms of diamond. 2.

Polycrystalline Diamond Compacts

In the early 1970s, technology was developed that allowed the production of metal-sintered polycrystalline diamond (PCD) compacts in large, twodimensional flat plates sintered directly to a cemented carbide backing. Similar in construction to cemented carbide, PCD diamond tooling has significant advantages related to the high wear resistance of diamond grains combined with the fracture resistance afforded by the metallic binders used in sintering. Figure 1 contains scanning electron microscope (SEM) images that compare the two-phase microstructures of a cemented carbide cutting tool and a PCD cutting tool. The added hardness and wear resistance of the diamond grains in PCD relative to the tungsten carbide grains in WC-Co offer distinct advantages for PCD-tipped tooling. PCD tools are normally fabricated by radio-frequency or torch brazing flat, tungsten carbide-backed,

428

Cline and Olson

CVD Diamond Solutions for Machining

429

polycrystalline tool tips directly onto a cemented carbide or high-speed steel tool body. After brazing in air, PCD tools are typically finish ground or cut to final shape by electric discharge machining (EDM). The performance advantages of PCD over tungsten carbide tools clearly outweigh the added cost of brazed PCD tooling in many nonferrous metal and abrasive nonmetallic machining applications. PCD is commonly used for machining aluminum, copper, brass, bronze, and cemented carbide in applications including turning, boring, profiling, grooving, milling, and holemaking [5]. 3.

Advanced Ceramics

Ceramic cutting tools are produced from a wide range of polycrystalline ceramic materials including aluminum oxide, cubic boron nitride (‘‘PCBN’’), and silicon nitride [6–9]. The hardness, chemical stability, and high wear resistance of these advanced ceramic cutting tools have demonstrated costperformance advantages over conventional cutting tool materials in applications ranging from abrasive and corrosive polymers to superalloy machining. PCBN-tipped tools are rapidly becoming the favored cutting tool choice in many high-throughput ferrous applications. Although the wear resistance, high-temperature corrosion resistance, and hot hardness of ceramic cutting tools are impressive, the application range of ceramic tooling is often limited by the lower fracture resistance of ceramic materials relative to other cutting tool materials. Special cutting edge preparation, whisker reinforcement, or other approaches are commonly used to improve the performance and broaden the application range of ceramic cutting tools. D.

Development of Coating Technologies for Cutting Tools

In the 1970s, thin-film deposition technology began to play a key role in furthering performance of commercial cutting tools [10]. Coated carbide and

< Figure 1 Structural similarities between cemented tungsten carbide and PCD diamond: SEM photomicrographs showing (a) the fractured cross section of a 6% CoWC cemented carbide at 5000⫻ magnification, (b) a polished cross section of a 6% Co-WC cemented carbide at 2000⫻ magnification, and (c) a 50⫻ image showing the construction of a PCD cobalt-sintered diamond cutting tool. Because both cemented carbides and PCD are two-phase materials composed of hard ceramic grains (WC or diamond, respectively) sintered in metal, they are commercially available in various compositions and grain sizes tailored for a wide range of applications. (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc.)

430

Cline and Olson

coated high-speed steel tools utilize hard coatings such as titanium nitride, titanium carbide, and aluminum oxide to enhance cutting performance. Most coatings are produced using CVD or physical vapor deposition (PVD) techniques. This ‘‘layer-composite’’ approach to cutting tool design couples the high-temperature wear resistance, corrosion resistance, and hardness advantages of advanced ceramic materials with the superior fracture resistance of the underlying tool body. Today, hard coating technology has evolved to the development of multilayered coatings in which individual coating layers offer unique performance-enhancing properties beyond single-layer structures [10,11]. Coating technologies for tungsten carbide and high-speed steel resulted in significant improvements in machining productivity and serve as the predecessors of CVD diamond-coated cutting tools. E.

Commercialization of CVD Diamond Technology for Cutting Tools

The relatively recent development of CVD diamond technology [12–14] has enabled further advances of diamond as a strategic material in the cutting tool industry. Like many new materials, CVD diamond has had its technical and commercial challenges. Today, however, the industry has achieved a track record of success and continues to develop processes that ensure consistency, reliability, and performance [1,2,15–23]. Presently, CVD diamond cutting tools are commercially available in two primary product forms: ‘‘thick-film’’ freestanding CVD diamond cutting tool tips and ‘‘thin-film’’ CVD diamond coatings. 1.

Thick-Film CVD Diamond Cutting Tools

Thick-film CVD diamond is vapor-deposited diamond manufactured in the form of solid, freestanding wafers of pure, binder-less diamond with typical thickness ranging from 150 to over 1000 ␮m. For most cutting tool uses, CVD diamond wafers are laser cut into hundreds of mirror-finish polished CVD diamond tips that are subsequently attached to cutting tools in a manner similar to cobalt-sintered polycrystalline diamond (PCD). The development of successful brazing techniques for thick-film diamond in the early 1990s bypassed the adhesion and substrate selection challenges of thin-film CVD diamond tools [8,23]. Figure 2a contains an image of two large-diameter, as-deposited, freestanding diamond wafers in addition to a collection of mirror-finished freestanding diamond materials. Figure 2b shows examples of finish-ground thick-film CVD diamond tooling (also known as diamond sheet tools [11,23]) adjacent to examples of aluminum workpiece materials. Thick-film CVD diamond cutting tools are engineered to compete

CVD Diamond Solutions for Machining

431

Figure 2 Thick-film CVD diamond is vapor-deposited diamond that can be manufactured in the form of large-diameter, freestanding wafers of pure, binder-less diamond. The larger wafer shown in the background of image (a) is approximately 175 mm in diameter. For most cutting tool uses, CVD diamond wafers are laser cut into hundreds of mirror-finish polished CVD diamond tips [(a), foreground], which are subsequently attached to cutting tools (b) in a similar manner to cobalt-sintered polycrystalline diamond (PCD). Freestanding CVD diamond cutting tools are used for a wide range of applications including aluminum automotive component machining (b). (From Ref. 1.)

432

Cline and Olson

with conventional PCD tools and are useful in a wide range of applications. As with PCD, thick-film CVD diamond tools are normally made with oversized tips that can be reground multiple times by diamond tool fabricators. Approaches to engineering thick-film CVD diamond-tipped tools are discussed further in Sec. II.D.2. 2.

Thin-Film CVD Diamond-Coated Cutting Tools

Thin-film diamond is CVD diamond that is deposited as a continuous, conformal coating directly onto a suitable substrate. The ability to deposit polycrystalline diamond coatings directly onto cutting tools enables the properties of diamond to be captured in shapes previously not available from nature or the diamond industry. In addition, the complete coating can eliminate the depth of cut limitations of thick-film CVD tools and PCD-tipped tools. Commercial diamond coatings generally fall in the thickness range of 5 to 35 ␮m. The diamond-coated tungsten carbide end mill depicted in Fig. 3 shows the complex shape, high-aspect-ratio coating capability presently available from the CVD diamond industry. Approaches to engineering thin-film CVD diamond-coated tools are discussed in Sec. II.D.1. Thin-film adhesion issues, challenges, and solutions are the topic of Sec. III.B.

Figure 3 Thin-film CVD diamond is vapor-deposited diamond deposited directly onto a component such as a cemented carbide cutting tool. This image contains a diamond-coated tungsten carbide rotary tool showing the complex shape coating capability now available in the CVD diamond industry. The end mill shown is 20 mm in diameter and 160 mm in overall length with a coated flute length of 110 mm. (From Ref. 15.)

CVD Diamond Solutions for Machining

II.

433

ISSUES AND CHALLENGES IN CVD DIAMOND CUTTING TOOL DESIGN AND MANUFACTURE

Capturing the unique properties of diamond in the design of CVD diamond cutting tools requires a knowledge of the application range and a fundamental understanding of the processes available to manufacture the tool. A broad-scope assessment of the complex properties-process-application interaction is critical because the performance advantage of a CVD diamond tool —if any—will depend not only upon the properties of the diamond itself but also on whether the chosen tool design and manufacturing process effectively capture the desired properties for the application of interest. A.

Critical Properties of CVD Diamond for Mechanical Applications

The key properties of CVD diamond for mechanical applications are not limited to the often-referenced hardness of diamond. The true challenge in CVD diamond tool manufacturing is capturing this potential within a functional tool that delivers reproducible performance. Table 1 provides a comparison of selected mechanical and thermal properties for PCD, single-crystal diamond, cemented carbide, and silicon nitride. Selected properties are summarized further in the remainder of this section. 1.

Hardness and Young’s Modulus

Diamond’s unmatched hardness and elastic modulus contribute to excellent wear resistance. Unlike metal-bonded PCD compacts and cemented carbides, diamond-like carbon (DLC), and many other hard coatings, CVD diamond maintains its hardness and wear resistance at elevated cutting temperatures. Diamond’s hot hardness is due in part to the high-temperature structural stability of its dense, polycrystalline construction coupled with the lack of metal binders that soften at elevated temperatures. 2.

Coefficient of Friction

CVD diamond’s low friction coefficient contributes to lower cutting forces, lower spindle power consumption, and less frictional heating in many cutting applications. The reduced cutting temperature that results from diamond’s high lubricity is viewed as a major advantage in applications where coolant elimination and higher speed operation are being considered. This property is not included in Table 1 because measured values of the coefficient of friction vary significantly depending on environment, surface roughness, and measurement technique. The coefficient of friction of polished CVD dia-

434

Table 1

Comparison of Thermal and Mechanical Properties for CVD Diamond and Other Hard Materials

Property Indentation hardness (GPa)

Young’s modulus (GPa) Tensile strength (MPa)

CVD diamond

Natural diamond

85–100 Technique unspecified Commercial data Ref. 90

50–100 Technique unspecified Commercial data Ref. 90

78.5 ⫾ 18.6 Vickers indentation Commercial data Ref. 91

56–102 Knoop indentation Normal load of 1–500 kg Ref. 92 1000–1100 Commercial data Ref. 90 1050–3000 Orientation dependent Measurement technique unspecified Typical commercial data Ref. 90

1000–1100 Commercial data Ref. 91 Growth face in tension: 500–800 (0.5 mm thick) Three-point bend geometry Estimated from Fig. 11 of Ref. 21 Nucleation face in tension: 800–1300 (0.5 mm thick) Three-point bend geometry Estimated from Fig. 12 of Ref. 21

PCD 50 Technique unspecified Commercial data Ref. 90

WC-Co cemented carbide 7.8–23.5 Vickers indentation Typical range Ref. 93

Si3N4 ceramic 7.8–17.7 Vickers indentation Range based on commercial data from two suppliers

21.5 Knoop indentation Ref. 92

776 Commercial data Ref. 90 1260 Measurement technique unspecified Commercial data Ref. 90

400–700 Ref. 93 700–1500 Typical range Ref. 93

120–330 Typical data Ref. 94 586–843 Range observed for three commercial grades Four-point bend with crosshead speed of 0.004 cm/s NIST evaluated data Ref. 95

Compressive strength (MPa)

16,000 Commercial data Ref. 90

7600 Commercial data Ref. 90

3000–7000 Typical range Ref. 93

>4000 Commercial data ASTM C773 Ref. 97

1.3 Commercial data Ref. 90

8680 average 16,530 maximum Natural octahedron containing no visible flaws or inclusions 10⫺6 m2 estimated load area Ref. 96 2.9 Commercial data Ref. 90

Transverse rupture strength (GPa)

1.2 Commercial data Ref. 90

0.6–3.0 Typical range Ref. 93

5.3–7.0 Commercial data Ref. 91 500–2200 Commercial data Ref. 90

3.4 Commercial data Ref. 90 600–2200 Commercial data Ref. 90

8.81 Commercial data Ref. 90 560 Commercial data Ref. 90

500–1100 Commercial data Ref. 90

600–1100 Commercial data Ref. 90

200 Commercial data Ref. 90

5–30 Typical range Ref. 93 20–120 Typical range for manufacturers quoted values Ref. 93 —

0.76–0.85 Commercial data Ref. 4, p. D116, Kennametal, Inc. 1.5–8.5 Typical values Ref. 94 7–43 Typical data Combined range for material types Ref. 94 —

Fracture toughness [MPa(m)1/2] Thermal conductivity at room temperature (W/m ⭈ K) Thermal conductivity at 200⬚C (W/m ⭈ K)

435

436 Table 1

Continued

Property Thermal expansion coefficient (ppm/K in 100–250⬚C range) Thermal expansion coefficient (ppm/K at 500⬚C) Thermal expansion coefficient (ppm/K at 1000⬚C) Mean thermal expansion coefficient (ppm/K in denoted range)

CVD diamond

Natural diamond

PCD

WC-Co cemented carbide

Si3N4 ceramic

1.2 Commercial data Ref. 90

1.2 Commercial data Ref. 90

4.2 Commercial data Ref. 90

—

—

3.8 Commercial data Ref. 90

3.8 Commercial data Ref. 90

—

—

—

4.5 Commercial data Ref. 90

4.5 Commercial data Ref. 90

6.3 Commercial data Ref. 90

—

—

—

—

—

4.5–8.5 Typical range for 0–800⬚C temperature range Ref. 93

2.8–3.7 Typical data for 20–1000⬚C range Ref. 94

CVD Diamond Solutions for Machining

437

mond is generally superior to that of cemented carbides [24] and advanced ceramics [25] and is similar to values attained for single-crystal diamond (e.g., 0.05 [24]). The high lubricity is often compared with that of fluoropolymers.* 3.

Thermal Conductivity

High-quality diamond has the highest thermal conductivity of any material. As Sec. III.A will explain, the use of the highest crystalline quality CVD diamond is not necessarily the best overall solution for mechanical applications of diamond. Regardless, it has been demonstrated that freestanding diamond that has been engineered for mechanical applications can offer thermal conductivity over five times that of tungsten carbide and over 50% greater than that of PCD.† Although diamond’s high lubricity helps to avoid frictional heating effects, CVD diamond’s excellent thermal conductivity relative to other cutting tool materials acts as an additional safeguard in maintaining lower cutting temperatures by dissipating heat generated in the cutting zone. 4.

Thermal Expansion Coefficient

CVD diamond and single-crystal diamond have essentially identical thermal expansion characteristics, as Table 1 delineates. Diamond has a very low thermal expansion coefficient relative to most other cutting tool materials such as cemented carbide and particularly high-speed steel, especially in the 100 to 250⬚C range. This property can be beneficial in controlling the positional accuracy of a cutting tool relative to a workpiece during precision machining. Because of the high-temperature nature of commercial CVD diamond coating and brazing processes, attachment of CVD diamond to cutting tool materials such as cemented carbide requires strict attention to the thermal expansion mismatch. The residual thermal stresses that result from high-temperature attachment necessitate good interfacial adhesion for both braze fabrication of thick-film CVD diamond and direct thin-film coating. Section III.B contains a detailed discussion of diamond adhesion issues for thin-film cutting tools.

*According to E.I. du Pont de Nemours and Company, the coefficient of friction of Teflon is generally in the range of 0.05 to 0.20, depending on the load, sliding speed, and particular Teflon coating used. †Comment is based on Norton Diamond Film’s ‘‘PolyTurn’’ and ‘‘DiamaPak’’ cutting tool products, which were retracted from the market upon business closure.

438

5.

Cline and Olson

Fracture Strength

As Table 1 depicts, the compressive strength of CVD diamond is generally superior to that of single-crystal diamond and PCD. Unfortunately, the pure polycrystalline structure results in a material that is generally inferior in tensile strength. Even so, the tensile strength of freestanding CVD diamond is sufficient for a wide range of industrial applications, especially when incorporated into tool designs that accommodate the unavoidable fact that diamond is a brittle, ceramic-like material. The measured values of the Weibull modulus [21] indicate that the breadth of the strength distribution for bulk CVD diamond is similar to that for advanced ceramics such as silicon nitride. The observed values ensure that CVD diamond can yield the reliable, reproducible tensile strength necessary in many mechanical applications. 6.

Fracture Toughness

Fracture toughness is a property of a material that is frequently defined as resistance to crack propagation. High fracture resistance is necessary for a cutting tool material to survive the impact and fatigue that can result from aggressive cutting operations such as milling or interrupted cutting or when machining workpiece materials containing material inhomogeneities such as hard phases, fibers, or particles. As shown in Table 1, the fracture toughness values of bulk, freestanding CVD diamond are comparable to those of advanced ceramic materials such as silicon nitride. Furthermore, because the microstructure of polycrystalline CVD diamond can provide a crack deflection mechanism [26–28], proper microstructural engineering can yield fracture toughness values exceeding those of single-crystal diamond and approaching values common to cobalt-sintered PCD. Section III.A includes approaches to optimizing the bulk properties of CVD diamond including fracture toughness. 7.

High-Temperature Phase Stability

Black, graphite-like coatings have been observed on diamond surfaces exposed to atomic oxygen at temperatures as low as 625⬚C [24]. However, diamond oxidation occurs at very low rates in air below 800⬚C and generally does not limit the application range of CVD diamond tooling. CVD diamond is thermally stable to temperatures of at least 1200⬚C in vacuum. By contrast, PCD is particularly susceptible to thermal degradation at temperatures above 700⬚C [29]. The cobalt binder phase can catalyze graphite formation at the sintered contacts between the diamond grains. This high-temperature phase instability, which can be induced by excessive cutting temperatures or im-

CVD Diamond Solutions for Machining

439

proper brazing or grinding procedures, is known to result in a general degradation of the mechanical properties of the PCD cutting tool tip. 8.

Chemical Corrosion Resistance

The chemical reactivity of carbon limits the use of diamond tooling in ferrous metal and superalloy machining. All forms of diamond tooling are at risk of corrosion or graphitization induced by machining iron-, nickel-, and cobalt-containing metals at elevated cutting temperatures [30]. Molten manganese, chromium, and the platinum group of metals are also true solvents for all forms of carbon including diamond [24]. Other metals capable of reacting with diamond include the carbide-forming transition metals titanium, tantalum, tungsten, and zirconium. Even so, diamond’s superlative combination of properties still maintains it as the preferred form of tooling for selected applications such as bimetallic machining of aluminum–cast iron deck faces for automotive engines [5]. With the exception of the metallic interactions just noted and hightemperature oxidizing environments, pure forms of diamond are extremely chemically inert. Unfortunately, the corrosion resistance of CVD diamond is a property that is often overlooked. For example, the metallic constituents of PCD and cemented carbides limit tool lifetime in corrosive polymer machining. Although the corrosion resistance of ceramic cutting tools can offer measurable improvements over traditional tooling, the advent of CVD diamond technology now allows the production of tooling with complex geometry that can offer both corrosion resistance and unmatched abrasive wear resistance. B.

The Critical Combination of Abrasive Wear Resistance and Fracture Toughness

In order to meet new machining requirements for difficult-to-machine materials, improve productivity, and hold high workpiece tolerances, new cutting tool materials must have both suitable fracture toughness and high abrasive wear resistance. The importance of fracture toughness, a measure of the resistance to crack propagation, was discussed in Sec. II.A.6. However, the specific abrasive wear resistance of a CVD diamond tool is very application specific and, therefore, was not discussed in Sec. II.A. Abrasive wear resistance is the result of a complex interaction between the properties and characteristics of the diamond tool, the workpiece material being machined, and the nature of the specific application. Improved abrasive wear resistance allows machinists to maintain tight tolerances with minimal tool or machine adjustments and reduces machine downtime associated with rapid tool wear and frequent tool changes.

440

Cline and Olson

Figure 4 depicts schematically the trade-off between fracture toughness and abrasive wear resistance in cutting tool materials. For example, advanced ceramic cutting tool materials, known to be very hard and resistant to abrasive wear, tend to be lower in fracture toughness. Metallic cutting tools such as high-speed steel, known to be high in fracture toughness, tend to have relatively low resistance to abrasive wear. Although cemented carbide does not have the superior toughness of high-speed steel, it combines an abrasive wear resistance approaching that of ceramic materials and fracture toughness that is sufficient for a wide range of heavy-duty industrial machining needs. In most applications, the exceptional fracture toughness of high-speed steel is not necessary or worth the inherent trade-off in abrasive wear resistance. Figure 4 also shows that ceramic-like CVD and PVD coatings can markedly enhance the abrasive wear resistance of both tungsten carbide and high-speed steel tooling while the critical fracture toughness of the underlying tool substrate is maintained. From the success of hard coatings in the

Figure 4 Abrasive wear resistance and fracture toughness trade-offs for common cutting tool materials.

CVD Diamond Solutions for Machining

441

cutting tool market, it is clear that the added hardness and wear resistance of diamond-coated tungsten carbide offer a well-balanced fracture toughness-wear resistance combination. The unique combination expected for CVD diamond-coated cutting tools is depicted at the top center of the figure. Because fracture toughness is a bulk mechanical property and the underlying substrate of a diamond-coated carbide tool is tougher than bulk CVD diamond, the diamond-coated carbide concept offers the potential to exceed the fracture toughness of bulk, freestanding CVD diamond. However, as discussed in Sec. III.B of this chapter, the composite properties of a diamondcoated carbide are captured in the tool design only if the adhesion strength exceeds the forces of the application. The abrasive wear resistance of the cobalt-sintered diamond in PCD can offer as much as 100 times greater tool life over cemented carbide [5]. By contrast, as Fig. 4 shows, PCD-tipped cutting tools are usually inferior in fracture toughness to cemented carbides. Fortunately, the fracture toughness of PCD is sufficient to avoid cutting edge chipping in most rigid machining situations. In fact, as the operating speeds and feeds move beyond the material limits of traditional cutting tool materials in aggressive highproductivity operations, the long life and predictable finish advantages of diamond tooling are often amplified.

C.

Application Characteristics That Affect CVD Diamond Cutting Tool Performance

Every cutting tool application is unique, and industrial cutting tools need to be designed for the worst-case scenario within the targeted range of applications. There are many factors that can directly or indirectly affect cutting tool performance. The feasible application range of a CVD diamond cutting tool design may not be clear until a tool is thoroughly tested in the field. As with any cutting tool, success in a specific operation on a specific machine tool does not ensure the same results with the same tool on the same type of machine at a different location. Because of the complexity of machining processes, this section is provided as an overview of the application characteristics that should be factored into the design and manufacturing processes used to produce CVD diamond cutting tools. 1.

Impact of Machine, Fixture, and Operating Conditions on Performance

Factors such as spindle balancing and rigidity, workpiece vibration, or even poor tool handling practices can dictate whether or not a tool will offer

442

Cline and Olson

a performance advantage in a specific application. Clearly, the type of machining operation (e.g., turning, boring, milling, drilling) and machine operating parameters (e.g., cutting speed, cutting feed rate, depth of cut) have a major influence on the nature of the forces on the CVD diamond cutting edge and the diamond-tool interface. For instance, machining operations where the cutting edge is not in constant contact with the workpiece (e.g., interrupted turning, milling, or end milling) generally require more robust design than continuous cutting operations (e.g., turning, boring, drilling). The impact-like forces generated in interrupted operations require good material strength and fracture toughness to minimize the damage of a CVD diamond tip or diamond coating during use. Other important application details include the coolant type (e.g., dry, mist, flood), chip-forming behavior, and efficiency of chip evacuation from the cutting area. Because CVD diamond (like other forms of diamond) is capable of operating at high cutting speeds and feed rates, machine tools with limited ranges of speed and feed can significantly limit the performance potential of diamond. By contrast, the wear resistance of many traditional cutting tool materials can degrade rapidly when machining processes are put into ‘‘fast forward’’ at high speeds and feeds. As a demonstration, Cline [31] and Tanikella et al. [32] have reported the results of end mill machining tests that were performed on SiC-containing ‘‘bisque ceramic’’ that was heat treated (‘‘presintered’’) at temperatures greater than 1200⬚C. The workpiece material and machining conditions were selected in an effort to accelerate the diamond wear mechanism common to abrasive applications such as ‘‘green ceramic’’ (unfired polymer-ceramic compacts) and graphite machining. The results of the machining tests showed the dramatic performance advantage of thin-film CVD diamond-coated end mills at the selected machining conditions. Figure 5a and b show that the uncoated tungsten carbide and TiNcoated tungsten carbide tools failed within 16 cm of machining (two 7.6cm-long passes). The failure was highlighted by excessive abrasive wear, which was enhanced by frictional heating of the tools. The PVD TiN coating offered no clear advantages under these machining conditions. Optical pyrometry of the carbide and TiN-coated carbide tools indicated temperatures exceeding 900⬚C. By contrast, a diamond tool that had already run 80 previous passes was barely warm to the touch after eight sequential passes at the same machining conditions. Freeze-frame images of the 88th pass are shown in Fig. 5c. Both the TiN-coated and uncoated carbide tooling showed a strong sensitivity to density variations in the presintered ceramic workpiece. In the case of one uncoated carbide tool, the excessive wear and overheating reduced the 9.5 mm cutting diameter of a new tool by over 2.5 mm within less than 23 cm of machining. More details of this case study are summarized in Sec. IV.B.2.c.

CVD Diamond Solutions for Machining

443

Figure 5 Effect of high cutting speeds on various cutting tool materials: sequential freeze-frame images showing presintered ceramic machining using (a) an uncoated tungsten carbide end mill, (b) a TiN-coated carbide end mill, and (c) a CVD diamond-coated end mill. The uncoated and TiN-coated tungsten carbide tools failed within 16 cm of machining due to excessive wear and frictional heating. The diamond-coated tool operated at cutting temperatures at least 500–1000⬚C lower than the uncoated equivalent and was still worthy of continued testing after more than 670 cm of machining. (From Ref. 31.)

444

Cline and Olson

Figure 5

2.

Continued

Impact of Workpiece Material Characteristics on Performance

The relative performance advantage of one tool over another is dependent on the workpiece material and the requirements for finished part quality. The characteristics of the workpiece such as abrasiveness, corrosiveness, and microstructural homogeneity can significantly affect the utility of a diamond tool in machining operations. Although less aggressive applications such as graphite machining will not always differentiate between tools of varying integrity, applications that generate higher, impact-like cutting forces can rapidly differentiate between tools that offer predictable abrasive coating wear and tools that fail unpredictably due to film delamination. The characteristics of diamond tooling are of particular interest in machining composite materials. The unique advantages offered by composites are generally the result of combining two or more dissimilar metallic and/ or nonmetallic materials in an effort to yield combined properties that exceed those of the individual constituents. For instance, glass fiber–reinforced plastics (GFRPs) combine the stiffness of the reinforcing fibers with the lowcost, lightweight characteristics of an otherwise pliable polymer matrix. Other useful composites, such as high-silicon aluminum alloys, metal matrix

CVD Diamond Solutions for Machining

445

composites (MMCs), and structural aerospace composites, offer advantages ranging from weight reduction to wear resistance and fatigue resistance. By design, the most unique and enabling composites tend to be composed of materials that have diverse properties and, therefore, vastly different machining behavior. The combined results are workpiece materials that can be abrasive, corrosive, and generally difficult to machine with conventional cutting tool materials [2,3,23,33,34]. Table 2 lists some commonly used metallic and nonmetallic composite materials along with their associated machining challenges. Many early efforts to fabricate components from these and other composites using conventional tooling proved to be cost prohibitive due in part to high machining costs. As the industrial demand for innovative composite materials continues to increase, so does the interrelated demand for cutting tools that can endure the challenges of machining these progressive materials. Many of the case studies discussed in Sec. IV are related to this increasingly important class of materials.

D.

Design and Manufacturing Options and Constraints for CVD Diamond

The design and manufacture of diamond-coated cutting tools varies significantly from thick-film, brazed-on CVD-diamond cutting tools. The last step in manufacturing a thin-film, diamond-coated tool is typically the CVD coating process, whereas freestanding CVD diamond wafer fabrication is the first of many steps in the fabrication of a thick-film, brazed-on CVD diamond tool. Many of the process-related advantages of the thin-film CVD diamond coating process are similar to advantages observed with other hard coatings used for cutting tools (see Sec. I.D). On the other hand, the processrelated advantages of the fabrication process for thick-film CVD diamond are more typical of PCD diamond tooling (see Sec. I.C.2). 1.

Approaches to Engineering Thin-Film CVD Diamond-Coated Tools

Today, CVD diamond processes enable direct coating of high-aspect-ratio, complex shapes ranging from specially molded, chip-breaking inserts to the end mill shown in Fig. 3. However, the manufacture of thin-film, CVD diamond-coated cutting tools does not come without unique technical challenges. Most noteworthy are the challenges associated with adhesion of diamond deposits to tungsten carbide—the preferred substrate for diamond

Table 2

Examples of Difficult-to-Machine Composite Materials Example

Examples of use

Machining issues

Related CVD diamond performance case studies

Composite: A390 (18% Si in Al) Matrix: Aluminum Reinforcement: Silicon particles Composite: Duralcan (30% SiC in Al) Matrix: Aluminum Reinforcement: Silicon carbide ceramic particles Composite: 25% cobalt tungsten carbide Matrix: Cobalt-based alloy Reinforcement: sintered tungsten carbide grains Composite: Carbonepoxy Matrix: Epoxy polymer Reinforcement: Highdensity carbon fibers Composite: G10 Matrix: Polymer Reinforcement: Glass fibers

Reduced-weight, wear resistant, temperatureresistant pistons

Although aluminum is generally free-machining, hard silicon particles are extremely abrasive

Thick-film inserts: IV.A.1.a Thick-film rotary tools: IV.A.2 Thin-film inserts: IV.B.1.a, IV.B.1.b

Brake rotors Lightweight structures

Although aluminum is generally free-machining, hard SiC ceramic particles are extremely abrasive

Thick-film inserts: IV.A.1.c Thin-film rotary tools: IV.B.2.c

High-fracture-toughness wear parts

Tungsten carbide sintering results in strong bonding or contiguity between WC grains. The WC grains are very hard. At elevated temperatures, the metal binder can react with the carbon in diamond. Carbon fibers are extremely abrasive; grinding is often preferred over cutting due to excessive tool wear

Thick-film inserts: IV.A.1.d

Glass fibers induce abrasive tool wear; polymer can cause corrosive tool wear

Thin-film rotary tools: IV.B.2.c

Material type Hypereutectic or ‘‘high-silicon’’ aluminum

Metal matrix composites (MMCs)

Cemented tungsten carbide

Structural ‘‘aerospace composites’’

Fiberglassreinforced polymers (FRPs)

Source: Modified from Ref. 2.

Stiff, lightweight support structures for commercial aircraft Strong, lightweight sporting goods Lightweight, insulative circuit boards Low-cost structural composites

Thin-film rotary tools: IV.B.2.c

CVD Diamond Solutions for Machining

447

coating commercialization. Many early efforts to introduce diamond-coated carbide cutting tools into the aluminum machining market failed or were significantly delayed because of lack of adhesion treatment technologies that could reproducibly yield WC-Co surfaces worthy of CVD diamond processing and subsequent use. Many early development efforts in the CVD diamond cutting tool industry involved the use of advanced ceramic materials as interim substrate solutions because the chemical compatibility and thermal expansion mismatch were more favorable for diamond coating. A variety of substrate materials including SiAlON, binder-less tungsten carbide and, more commonly, silicon nitride and silicon carbide were employed [23]. Although diamond-coated ceramic tools do have utility in selected applications, the inferior fracture toughness of diamond-coated ceramic inserts and end mills generally limited their use to nonmetallic machining operations such as those involving graphite, polymers, green ceramics, and nonmetallic composite materials. a. Interfacial Integrity. Figure 6 contrasts the failure modes of a ‘‘bad’’ tool (Fig. 6a) that has failed by unpredictable, catastrophic film delamination related to poor interfacial adhesion and a ‘‘good’’ tool (Fig. 6b) that has failed by predictable abrasive wear. The relative utility of these two tools is clear. If the diamond adhesion is not suitable for the application, the desired coating properties simply are not captured in the tool design. As discussed in Sec. III.B in detail, proper diamond-coated WC-Co cutting tool design requires strict attention to both the coating properties and the substrate surface properties. b. Morphology and Surface Roughness Characteristics. Today, most thin-film CVD diamond cutting tools are made from well-faceted CVD diamond deposits because of the higher abrasive wear resistance typically offered by coarse-grained diamond. However, relatively smooth, more lubricious, microcrystalline diamond coatings are commercially available for applications where workpiece surface finish is more critical. Most diamondcoated tools are presently marketed in an as-deposited state with no postdeposition finishing. Depending on the interfacial roughness introduced by adhesion treatment, coating thickness, and the morphological features of the coating, the surface roughness and edge blunting of some diamond-coated tools can limit their use to intermediate finishing and roughing applications. The morphological features of CVD diamond are discussed in more detail in Sec. III.A. c. Postdeposition Finishing Options. Postdeposition finishing techniques can be applied to modify the generally rough, matte finish of diamond-coated products. Diamond-coated ceramic tools with mirror-finished

448

Cline and Olson

Figure 6 Contrasting failure modes of CVD diamond coatings: SEM images contrasting the failure mode of two different diamond-coated tool designs after aluminum metalcutting tests. (a) A tool that failed catastrophically due to insufficient coating adhesion. A metallic buildup of aluminum is evident in the delaminated zone on the tool tip where the adhesion-treated carbide was exposed. (b) A tool that offered a suitable level of adhesion for the application and therefore maintained the slow, predictable wear characteristics desired in a diamond-coated component. As reviewed in Sec. III.B, the diamond/substrate interface characteristics have a significant influence on the nature of tool failure. (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc.)

CVD Diamond Solutions for Machining

449

rake faces similar to PCD tools have been produced in the industry.* Oles et al. [23] have reported the success of a CVD diamond buffing technique in reducing insert surface roughness for the flank face of a diamond-coated carbide insert. Turning tests in A390 (18% Si-Al) showed that the buffed flank surfaces substantially improved the workpiece surface finish. However, the turned component finish of the buffed tools was still inferior to the finish generated by finish-ground PCD tools tested in the same study. Kanda et al. [17] have reported progress in grinding rotary tools to enhance cutting edge characteristics and resulting workpiece surface finish. Although the general tendency of the industry is toward smoother surfaces that mimic the finish ground characteristics of PCD and cemented carbide tools, Oles et al. [23] have also reported a ‘‘microscopic chip-breaking’’ advantage of diamond-coated cemented carbide inserts. Milling and turning studies of low-silicon and high-silicon alloys showed the advantages of diamond facets on the chip-forming rake surface in creating a ‘‘highly stressed chip that is more readily broken.’’ Metallic chips in the shape of small ‘‘sixes’’ and ‘‘nines’’ were reported to be more desirable than continuous stringers or coils that are common to smooth rake-face tools such as PCD. As CVD diamond technology advances, it is likely that the use of polishing, grinding, or other finishing techniques for CVD diamond-coated tooling will become more prevalent. Increases in CVD diamond cutting tool demand will provide justification for tailoring film properties such as thickness and surface roughness to specific applications. Such justification is critical to the presence of CVD diamond in the market because it will undoubtedly lead to an expansion in application range for diamond-coated products. 2.

Approaches to Engineering Thick-Film CVD Diamond Tools

a. Design and Manufacture of Thick-Film Tips. The primary technical challenge in the manufacture of freestanding thick-film CVD diamond cutting tool tips is attaining a throughput of freestanding diamond material having the dimensional and mechanical characteristics needed for downstream processes and ultimate use [35]. This is commonly approached through the manufacture of large-area, crack-free, CVD diamond discs of the necessary thickness with minimal thickness variation and minimal bow. Attaining CVD diamond wafers of this quality is exceptionally difficult due to the unique mechanical properties of diamond and the tendency for these *Mirror finish observed on diamond-coated silicon carbide cutting tools once produced by DeBeers.

450

Cline and Olson

properties to vary measurably with changes in deposition conditions such as temperature and gas chemistry. Clearly, factors such as CVD diamond reactor type, the associated diamond deposition area, deposition rate, and vacuum process control capabilities are critical to success in freestanding diamond production. Although this issue has been discussed in various forums [35], many of the details of the industrial methods used to generate large freestanding diamond discs are presently treated as trade secrets and will not be discussed in this text. Regardless of the challenges, pure CVD diamond wafers for mechanical applications have been commercially manufactured in the thickness range of 150 to over 1000 ␮m with diameters up to 175 mm (see Fig. 2a).* By comparison, commercial high-temperature, highpressure PCD manufacture is presently limited to disc sizes of about 74 mm. Cutting CVD diamond components from freestanding CVD diamond wafers is presently done using laser technology as opposed to the electric discharge machining (EDM) ‘‘wire cutting’’ common to the PCD industry. Pure CVD diamond wafers contain no metallic binders and, therefore, do not afford the conductive properties needed in an EDM workpiece. The general acceptance of EDM wire techniques for diamond tool fabrication coupled with the present lack of the necessary laser technology at cutting tool fabricator sites is a significant barrier to commercialization of thick-film diamond tooling. b. Attachment Techniques for CVD Diamond Tips. Most freestanding, thick-film CVD diamond used for cutting tool applications is approximately 0.5-mm-thick polished diamond. The pure diamond can be attached to the tool body using a high-temperature inert vacuum braze (normally >850⬚C). The use of a high-temperature braze ensures good chemical bonding and excellent braze strength. Commercially available forms of freestanding CVD diamond cutting tool tips are generally made from well-faceted deposits that may have crystal sizes exceeding 50 ␮m. The surface roughness of the thick-film, freestanding diamond is usually removed by diamond finishing techniques in an effort to mimic the desirable characteristics of PCD. In most commercial thick-film CVD cutting tool applications, the relatively smooth nucleation surface is used as the cutting surface and the rough, faceted face becomes part of a vacuum-brazed interface typically bonded to tungsten carbide. A flank-face view of a finish-ground CVD diamond cutting tool tip is shown in Fig. 7. Commercially available thickfilm cutting tool tips offer superior wear resistance, which is typical of coarse-grained thin-film diamond deposits, coupled with mirror finish (commonly <0.5 ␮m R a) polished rake faces and finish-ground flank faces at*Comment refers to Norton Diamond Film’s thick-film product range prior to business closure.

CVD Diamond Solutions for Machining

451

Figure 7 Flank-face electron microscope view of a freestanding CVD diamond cutting tool after use in a machining operation. The bright white region along the cutting edge contains the abrasive wear scar. As the 100⫻ image depicts, the faceted coarse-grained ‘‘growth’’ face of the CVD diamond tip is incorporated into the vacuum braze layer of the tool (see lower horizontal feature). The polished, fine-grained, ‘‘nucleation’’ face of the diamond is used as the rake face of the tool (see top, dark region above the U-shaped cutting edge). To the naked eye, the rake face of this tool actually has a black mirror finish. (From Ref. 1.)

tained by industrial finishing and grinding techniques, respectively. This unique combination enables thick-film cutting tool tips to generate superior workpiece finish relative to CVD diamond coatings and most other alternative cutting tool technologies. c. CVD Diamond Grinding Challenges. Good cutting edge grinding techniques can yield CVD diamond edge quality similar, if not superior, to that attainable by most or all PCD grades. However, CVD diamond’s erosion resistance often requires more fabrication time and cost to ensure good edge quality. According to a representative of the diamond fabrication industry (T. Drury, J&M Diamond Tool, personal communication, 2000) [1] it takes 10 to 20% longer to grind a CVD diamond tool with an edge quality comparable to that of PCD tooling. Diamond superabrasive wheels are normally used to grind CVD diamond. As with PCD fabrication, toolmakers have the option of using vitrified bonded, resin-bonded, and metal-bonded diamond wheels. Aggressive grinding of CVD diamond can induce flaws that sacrifice the mechanical behavior and cutting action of the cutting edge. When proper

452

Cline and Olson

grinding techniques are employed, the strong nonmetallic bonding of the CVD diamond crystallites minimizes grain pullout on the cutting edge of thick-film CVD diamond tips. Most toolmakers claim that reduction in grinding wheel feed rates can yield CVD diamond edge finishes superior to what PCD can offer. Once the CVD diamond edge quality is attained through proper fabrication techniques, the binder-less construction of CVD diamond offers a continuous polycrystalline cutting edge that lacks the soft binder phase inherently present in metal-sintered PCD and cemented carbides.

III.

OPTIMIZING MECHANICAL PERFORMANCE OF CVD DIAMOND CUTTING TOOLS

Because abrasive wear and mechanical failure mechanisms control the wear life of cutting tools in most applications, optimization of the mechanical properties of CVD diamond cutting tools is essential to maximizing performance. Fortunately, CVD diamond process parameters can be adjusted and monitored in an effort to modify and reproduce desirable material properties and physical characteristics of the deposits. This section contains discussions of the issues and challenges of attaining good mechanical performance of CVD diamond tools. Section III.A contains approaches to engineering surface structure and the bulk mechanical properties of both thin-film and thickfilm CVD diamond deposits. Section III.B will be focused on the fundamentals of the thin-film diamond adhesion and includes specific approaches that have been taken to overcome adhesion to tungsten carbide cutting tools. A.

Optimizing the Surface and Bulk Structure of CVD Diamond Deposits

1.

CVD Diamond Morphology and Defect Structure

One of the most recognizable physical characteristics of CVD diamond deposits grown under different process conditions is the microstructure or morphology of the individual diamond crystals. As shown in Fig. 8, diamond film morphology may range from coarse-grained (diameter >1 ␮m), faceted crystallines to fine-grained (diameter <0.1 ␮m), ‘‘microcrystalline’’ structures. The properties of the diamond film and associated performance can vary widely depending on the selected morphology and application details. Because of the organic nature of CVD diamond synthesis from carbonhydrogen gas mixtures, diamond deposits produced under typical CVD diamond growth conditions may contain small fractions of nondiamond or non-sp3 bonded carbon. Fine-grained deposits (typically grown with higher

CVD Diamond Solutions for Machining

453

Figure 8 Examples of CVD diamond morphology. Depending on the needs of the application, the morphology of CVD diamond deposits can range from coarse (diameter >1 ␮m) faceted crystallites to fine-grained (diameter <0.1 ␮m) or ‘‘microcrystalline’’ structures. The SEM micrograph in (a) shows the free surface of a coarse-grained diamond film. Note the misorientation of grains relative to each other and the high density of twin defects observed at the surface. The SEM micrograph in (b) shows the free surface of a microcrystalline diamond film. The average grain size for this film is on the order of 100–200 nm. (From Ref. 28.)

454

Cline and Olson

carbon content gas mixtures) generally contain higher amounts of non-sp3 carbon than coarse-grained deposits. In addition to nondiamond carbon defect structure, lattice defects such as vacancies, dislocations, and/or twin planes may also be incorporated into the films. Microporosity in the form of fissures and voids has also been reported in thick-film diamond [35]. High-resolution electron microscopy has been used to resolve the fine structure of lattice defects in films and results have been correlated with growth conditions and mechanisms [36,37]. The resulting influence of defects on the material properties has been determined to be a function of the defect type and density [35,38–42]. 2.

Structural Analysis of CVD Diamond by Raman Spectroscopy

Raman spectroscopy has been demonstrated to be one of the most versatile techniques used to characterize the structural properties of synthetic diamond. Graphite, sp2-bonded amorphous carbon, diamond, and other forms of carbons have strong and easily identifiable Raman spectra. In diamond materials, the primary diamond peak exhibits changes in width that have been related to the degree of structural order, and small shifts in the wavenumber position of the primary diamond peak have been correlated with the stress state in the diamond. Nondiamond carbon has a distinctly different Raman signature that is more intense than the diamond spectrum, allowing the detection of very small amounts of ‘‘graphitic’’ material in CVD diamond films [43]. Many early CVD diamond references refer to the relative amount of non-sp3 carbon versus sp3 carbon in a deposit as a measure of CVD diamond ‘‘quality,’’ where the quality was loosely defined by the percentage of nondiamond carbon and the density of defects incorporated in the deposit [44]. This definition assumes that the ideal ‘‘highest quality’’ Raman characteristic for polycrystalline CVD diamond is one that mimics the perfect singlecrystal diamond—minimal defect structure with a sharp, narrow signature Raman peak at 1332 cm⫺1. In applications such as freestanding optics, this definition may apply. However, in most predominantly mechanical applications, the best performing material often has intergranular and intragranular defect structures that yield Raman spectra quite different from those of a pristine single-crystal. That is, the use of the highest crystalline quality CVD diamond is not necessarily the best overall solution for mechanical applications of diamond. The Raman spectrum of the coarse-grained (‘‘higher quality’’) diamond film depicted in Fig. 8a is presented in Fig. 9. The Raman results show a sharp primary diamond band located at about 1332 cm⫺1 and a low-intensity,

CVD Diamond Solutions for Machining

455

Figure 9 Raman spectra for a coarse-grained diamond coating: Raman spectrum of coarse-grained diamond film used for thin-film cutting tool applications. Note the sharp primary diamond band located at about 1332 ⌬cm⫺1. This primary diamond band displays a full width at half of the maximum intensity (FWHM) of approximately 12 ⌬cm⫺1. The individual peaks upon which the experimental data are superimposed were found to represent the best fit solution to the spectrum. The broad band located at approximately 1556 ⌬cm⫺1 is attributed to a small amount of sp2bonded carbon content in the coating. (From Ref. 28.)

broad band located at approximately 1556 cm⫺1. Because the Raman scattering efficiency of the sp2-bonded material is approximately 50 times greater than that of sp3-bonded carbon [43,45] and the intensity of the broad band is quite low compared with the first-order diamond band, the sp2 carbon content is actually quite low in this sample. The corresponding microstructure of this film displays a high density of microtwinning and a high degree of misorientation of the free-surface crystal planes (see Fig. 8a). The Raman spectrum of the microcrystalline (‘‘lower quality’’) diamond film depicted in Fig. 8b is shown in Fig. 10. In addition to the primary diamond band at

456

Cline and Olson

Figure 10 Raman spectra for a microcrystalline diamond coating: Raman spectrum of a microcrystalline diamond film used for thin-film cutting tools or sliding wear applications. In addition to the primary diamond band located at about 1332 ⌬cm⫺1, high-intensity bands at approximately 1357 ⌬cm⫺1 and 1580 ⌬cm⫺1 indicate a higher component of sp2-bonded carbon material in this sample compared with the spectrum presented in Fig. 9. The low-intensity band near 1180 ⌬cm⫺1 has been attributed to a breakdown in Raman scattering rules observed for very small grained crystals. (From Ref. 28.)

1332 cm⫺1, the Raman results for this sample display several broad bands at approximately 1357 cm⫺1 and 1580 cm⫺1 that are attributed to disordered sp2-hybridized carbon bonds. The low-intensity band near 1180 cm⫺1 has been attributed to a breakdown in Raman scattering rules observed for very small grained crystals [46,47]. The average size of the diamond grains in this film is estimated to be in the range of 0.10–0.20 ␮m. The results of characterization of CVD diamond by many researchers appear to indicate that sp2-bonded carbon generally resides at the boundaries of the individual crystallites in the CVD diamond films grown by most techniques [48,49]. In much the same way that impurities can influence the grain size of other

CVD Diamond Solutions for Machining

457

polycrystalline materials, it is likely that significant amounts of nondiamond carbon actually limit the crystalline periodicity and, therefore, the grain size in this type of deposit (see Fig. 8b). 3.

Effect of CVD Diamond Structure on Mechanical Performance

In order to investigate the relationship between diamond characteristics and wear resistance, Olson and Dawes [28] characterized the microstructure, phase purity, and defect structure for diamond films grown under a variety of process conditions using scanning electron microscopy (SEM), x-ray diffraction (XRD), and Raman spectroscopy. The material characteristics were then correlated with ‘‘performance’’ in a metalcutting application. The results of this examination suggest that the density and type of defect have a very strong influence on the wear mode and wear rate of the film. The mechanism and rate of abrasive wear of cutting tools depend upon free-surface characteristics such as chemical and phase composition and bulk properties such as strength and toughness [50]. As discussed in Sec. III.A.1, the phase composition and microstructure of the film surface may vary significantly as a function of processing conditions. In most cases, attempts to characterize the tribological consequence of such free-surface microstructural variations in CVD diamond films have been unsuccessful due to limitations in adhesion strength [14]. Interfacial reactions between the cutting tool surface and workpiece during machining can dramatically affect wear rate; therefore, the potential reactivity between diamond and the workpiece needs to be considered. In highly abrasive applications (chemical wear excluded), diamond wear proceeds by brittle fracture or cleavage [50]. Materials with high hardness are usually brittle and diamond is a notable example. Fracture analysis of CVD diamond films grown under a range of conditions reveals that the fracture mode may vary from nearly 100% intergranular to nearly 100% intragranular [26–28]. In materials that fracture through an intergranular mode, wear proceeds by crack propagation between individual crystallites. This fracture mode is typical for CVD diamond films with weak intergranular bonding. In materials that fracture through an intragranular mode, wear proceeds by cleavage through the grains of diamond. This fracture mode is typical for CVD diamond films with strong intergranular bonding. The general relationship between grain size and strength is well established for ceramics and other brittle materials [51]. However, evidence suggests that the operating mechanism that limits grain size in CVD diamond films has a more significant influence on the wear mechanism than the grain size itself. Because non-sp3-bonded material often resides preferentially at

458

Cline and Olson

the grain boundaries of the crystals that make up the polycrystalline material [48,49], this chemical defect can limit the grain boundary strength and may ultimately alter the wear mechanism. Well-faceted, highly oriented films that are characterized by weak grain boundaries also display very poor resistance to wear. A fractographic comparison of weak and strong grain boundaries is shown in Fig. 11. In materials that fracture through an intragranular mode, wear may be controlled through the introduction of planar defects that can act as microscopic crack deflection sites [26–28]. Step formation restrains the propagation of brittle cracks by absorbing extra energy in the vicinity of the connecting riser between adjacent crack planes. The increase in crack resistance energy associated with step formation depends on the density and height of the steps [51]. Experimental evidence supports the hypothesis that planar defects and strong, randomly oriented grain boundaries in CVD diamond films may serve as microscopic crack deflection sites that redirect cracks and absorb crack energy [26–28]. Experimental evidence also supports that CVD diamond films characterized by intergranular fracture propagation (Fig. 11b) have wear resistance inferior to that of CVD diamond deposits characterized by intragranular fracture (Fig. 11a) in most applications [28]. B.

Optimizing Adhesion of CVD Diamond Coatings to Cutting Tools

CVD diamond processes enable cutting tool designers to capture the unique properties of thin-film diamond while exploiting the fracture toughness and other bulk properties of the underlying substrate base material. In order to transform this design concept into functional tooling, the strength of adhesion of the diamond film to the underlying substrate must allow the thin film and substrate to operate as a ‘‘composite’’ system. Because diamond is a low thermal expansion material and commercial CVD diamond thin-film

> Figure 11 Comparison of intergranular and intragranular fracture in CVD diamond. SEM micrographs of the fractured cross section of thin, CVD diamond films grown on WC-Co substrates show the dramatic variations in fracture characteristics that can result from simple changes in CVD diamond process conditions. A highly oriented film seen as the dark layer in the top of the 2000⫻ image in (a) displays weak grain boundaries and a fracture mode that is intergranular. A randomly oriented film seen as the dark layer making up most of the 4000⫻ image in (b) displays strong grain boundaries and a fracture mode that is nearly 100% intragranular. (From Ref. 74.)

CVD Diamond Solutions for Machining

459

460

Cline and Olson

processes operate at high temperatures, diamond coating requires special attention to substrate chemistry and general process control that is not as critical for most other coatings. Diamond coating challenges are amplified by the fact that the primary market demand is for diamond coatings on existing grades of tungsten carbide that contain not only cobalt but also various grain growth inhibitors and metallic impurities. The high-temperature reactivity of diamond with the cobalt-based binder phase of tungsten carbide cutting tools generally prevents the direct, high-temperature deposition of CVD diamond directly onto ‘‘off-the-shelf’’ tungsten carbide grades. Even after high-temperature chemical compatibility of the interface is ensured using adhesion treatment processes, the thermal expansion mismatch between the film and substrate often requires attention as it can be the source of very large residual stresses. Ignoring any one of these factors can lead to poor interfacial integrity and may result in delamination of the film or coating during use. Because overcoming diamond adhesion challenges is part of the critical path to successfully manufacturing a thin-film CVD diamond cutting tool, this section is devoted specifically to adhesion issues. Section III.B.1 is an overview of the specific adhesion challenges in coating cutting tools with a focus on commercial grades of tungsten carbide. Section III.B.2 includes various approaches that have been used to overcome adhesion barriers in an effort to capture the enabling characteristics of diamond-coated tungsten carbide. Section III.B.3 contains a case study on the development of a specific adhesion treatment intended for the manufacture of diamond-coated metalcutting inserts. 1.

Diamond-Substrate Characteristics Necessary for Adhesion

Adhesion strength may be loosely defined as the force or energy required to separate two objects. It may be thought of as the energy necessary to break the bonds at an interface, thus driving the extension of an interfacial crack. Extension of such a crack is referred to as interfacial crack propagation. It is interfacial crack propagation that leads to failure of an interface and ultimately yields visually apparent signs of component failure referred to as ‘‘delamination,’’ ‘‘spalling,’’ or ‘‘flaking.’’ Using this definition, interfacial failure is defined not solely by a fracture resistance parameter (‘‘interfacial fracture toughness’’) uniquely related to the bonding across the interface but by the ‘‘adhesion strength,’’ which is determined by the combined influences of interfacial fracture toughness, strength controlling defects, and residual stresses [52]. The interfacial fracture toughness is related to the type and density of interfacial defects (cracks, voids, weak interfacial

CVD Diamond Solutions for Machining

461

phases, etc.), the density of chemical bonds between the diamond film and substrate at the interface, and the presence of an interfacial crack deflection mechanism [52–54]. Therefore, maximizing adhesion strength for CVD diamond-coated cutting tools requires minimizing specific interfacial defects such as voids, cracks, or weak interfacial phases; maximizing chemical bond density; and providing a crack deflection mechanism that increases the energy necessary to drive an interfacial crack. In order to minimize the interfacial defects and maximize the chemical bond density in diamond-coated WC-Co, it is essential to maximize diamond nucleation density, ensure that the interface is chemically stable, and suppress the formation of ‘‘soft’’ nondiamond phases at the interface. As will be discussed in Sec. III.B.1.d, toughness-enhancing crack deflection characteristics can be engineered into the diamond/WC-Co interface. It should be emphasized that, in practical applications, the ‘‘adhesion strength’’ (in quotation marks to emphasize the potential difference in failure mechanism) is a measure of the utility of the tool in the application and its resistance to decohesion during use. That is, a quantitative assessment of adhesion strength is meaningless unless the conditions mimic the specific failure mode observed in the application. Using this alternative definition, the failure mode commonly described as delamination, spalling, or flaking is failure that is located at or near the interface. This distinction encompasses crack propagation within the nucleation layer of the diamond film as well as failure within the free-surface grains of the substrate such as the adhesion-treated region of a diamond-coated WC-Co component. Although this definition is of much more practical utility, it necessitates the use of semiquantitative failure analysis to verify the specific mechanism and location of the ‘‘interfacial’’ failure. a. Nucleation Density of Diamond. Effective nucleation of CVD diamond onto foreign substrate materials is notably more difficult than nucleation of other hard coatings produced by CVD or PVD processes. Unlike other hard coating processes, ‘‘diamond seeding’’ processes are often employed to enhance the nucleation density of diamond [55]. Heteroepitaxially grown diamond films begin as submicrometer-sized nuclei formed on the top of a nondiamond substrate. The physical characteristics of the nuclei are dependent on the deposition conditions and the composition of the underlying substrate. As schematically shown in Fig. 12, subsequent growth normally occurs by addition of atoms to already formed nuclei rather than by continued formation of new nuclei. If the individual nuclei are separated by a ‘‘large distance,’’ interfacial voids may form when the individual nuclei coalesce and form a continuous layer. By maximizing the nucleation density, the inclusion of internucleic voids at the interface may be minimized. Be-

462

Cline and Olson

Figure 12 CVD diamond nucleation and growth. On many nondiamond surfaces, growth of CVD diamond films occurs by a Volmer-Weber or ‘‘island growth’’ type of nucleation/growth mechanism. In this process, a finite number of nuclei are formed during a nucleation stage (a). Following this stage, adatoms are incorporated into the individual nucleation sites until the islands have grown to some maximum diameter at which their boundaries intersect (b). If the density of diamond nuclei is low and individual nuclei are separated by large distances relative to their size, interfacial voids will form when the nucleation sites have grown to a size at which their diameters coalesce (b). The size of these interfacial or internucleic voids may be minimized (assuming the same nucleation and growth mechanism) by increasing the nucleation density as shown in (c). (From Ref. 74.)

cause these voids may act as crack nucleation sites, they can lower the critical stress necessary for crack propagation at the interface. Increasing the nucleation density also reduces the time necessary for the nuclei to coalesce into a continuous film. Many methods of optimizing the nucleation density for a range of substrate materials have been presented in the CVD diamond synthesis technical literature [55–60] and the details are not discussed in this chapter. b. Interfacial Chemical Bonding. The chemical composition of some substrate materials can impair or negate the ability to form strong bonds between the film and substrate, especially with the high-temperature nature of CVD diamond processes. Physical or mechanical adhesion alone is simply not sufficient to ensure performance. Chemical bonding is necessary to counter the interfacial stresses induced by the thermal expansion mismatch and the applied load of cutting tool operations. Several approaches

CVD Diamond Solutions for Machining

463

may be taken to optimize the chemical bonding between the film and substrate. In general, the most effective means of maximizing bond density is by minimizing the mismatch between coefficients of thermal expansion and lattice spacings between the film and substrate. A more detailed consideration of this topic may be found in Allen [61]. For the specific case of CVD diamond films on carbide-forming substrates (Ta, Cr, Ti, W, etc.), a carbide ‘‘interphase’’ may be formed that can relax the lattice mismatch between the film and substrate [62,63]. However, in the case in which amorphous carbon or other nondiamond material is used to enhance the nucleation density, these nondiamond layers may form a weak interfacial layer suppressing the strength of chemical diamond-substrate bonds. Therefore, deposition process conditions must be carefully controlled in order to suppress the formation of a weak interfacial nondiamond phase or, if this phase cannot be avoided, to minimize its thickness. c. Substrate Chemical Composition. Because cobalt catalyzes the formation of nondiamond carbon (sp, sp2-bonded carbon) in the diamondCVD growth environment, it is important that the binder phase be absent from the substrate surface during nucleation [64,65] and that diffusion of the binder phase from the bulk to the free surface is suppressed [66]. It is believed that the presence of Co at the growth surface or at a region of already formed diamond can perturb the phase stability of the deposit. Under typical CVD growth conditions, this may result in a phase transformation from diamond to graphite or other nondiamond carbon phases. The solubility of carbon in cobalt under the elevated temperatures of CVD diamond deposition conditions may drive the dissolution of the carbon at the interface into the binder phase [67]. For this reason, the binder phase composition and the proximity of the binder phase to the diamond film may also affect the phase stability of the interface when subsequently exposed to the hightemperature diamond deposition conditions. d. Interfacial Microstructure. Consider a film-substrate system as illustrated in Fig. 13, which is characterized by an ‘‘atomically smooth’’ interfacial microstructure. Because adhesion strength is related to the density of chemical bonds that are present at the interface, the only resistance to interfacial crack propagation in this case is provided by the strength of these chemical bonds. If we consider the interface to contain residual stress and other weakening characteristics such as internucleic voids, the applied force required to drive an interfacial crack is further reduced. Now consider the film-substrate system as illustrated in Fig. 14, which is characterized by a rough interfacial microstructure. Recalling the previous definition that adhesion strength is a measure of the force or energy required to drive a crack along the interface between the film and substrate, this

464

Cline and Olson

Figure 13 Schematic representation of a smooth substrate-film interface system. Interfacial crack extension or propagation is resisted only by chemical bonding between the film and substrate. If an additional stress is applied to a film that has a preexisting crack at the interface (a), this crack will extend along the interface resulting in delamination (b). The adhesion strength of films that display this type of interface is limited by the strength and density of the chemical bonds, residual stresses, and the type and density of interfacial defects such as cracks or voids. (From Ref. 74.)

energy needed is significantly increased for a system that incorporates a nonporous, ‘‘rough’’ interface. This is due to several factors. Generally, the diamond-WC interface (assuming that no Co or Co interaction is present) can be considered a brittle interface. In brittle materials, crack propagation occurs very rapidly due to high crack velocities. Studies of single-crystal diamond have shown that the crack velocities exceed 5000 m/sec [68]. Fortunately, the same crack deflection phenomenon described for the bulk diamond in Sec. III.A.3 is also applicable to the diamond-substrate interface. For a crack to extend along the interface, it must follow a tortuous path as it is redirected along the rough interface as illustrated in Fig. 14. Because of the high crack velocity, the lowest energy crack path is most likely a linear path into the film or into the substrate. Extension of the crack into the film or into the substrate may pin the crack tip and arrest the interfacial crack, effectively toughening the interface. In addition, because the interfacial surface area between the film and substrate is much greater for rough interfaces, the increased density of chemical bonds further enhances adhesion strength.

CVD Diamond Solutions for Machining

465

Figure 14 Schematic representation of a rough substrate-film interface system. In addition to the increase in surface area available for chemical bonding, this type of interface may provide a tough, crack deflection mechanism. In brittle interfaces characterized by little plasticity, crack extension is redirected from the interface (a) and may be arrested by the film or substrate (b). Thus, the increase in chemical bond density in addition to the toughening mechanism results in an increase in resistance to interfacial crack propagation and delamination. (From Ref. 74.)

2.

Selected Approaches to Solving the Tungsten Carbide Adhesion Problem

a. Wet-Chemical Etching Processes. Wet-chemical etching techniques have been commonly used as an adhesion treatment approach targeted at removing the metallic binder phase from the surface of WC-Co substrates. For example, Kikuchi et al. [69] proposed the formation of an ‘‘etch layer’’ extending between 0.1 to 1.0 ␮m into a WC-Co substrate with a 1–4 wt% Co content. Etching methods successfully allow the nucleation and growth of diamond films on the etched surface without the preferential deposition of graphite. However, most wet-chemical techniques require stringent process control to ensure utility of the diamond-coated part. The binder phase, which is located at the interstices of the hard, WC grains, provides the toughness characteristic of ductility to the WC-Co material. It also serves to ‘‘cement’’ the hard, somewhat contiguous, WC ‘‘skeletal structure’’ together to form a composite material with excellent mechanical properties. Chemical removal of the binder phase to a depth greater than the dimension of the free-surface WC grains essentially removes the cement and generally results in the formation of an embrittled layer at the surface of the WC-Co

466

Cline and Olson

part. In the presence of residual thermal stresses or applied stresses encountered during use, failure of the interface by WC grain ‘‘decohesion’’ or by crack extension in this embrittled area may result in delamination of the film as illustrated in Fig. 15. Diamond-coated tungsten carbide components that have failed by this mode characteristically have WC grain pullout from the WC-Co substrate onto the back side (interface/nucleation side) of the delaminated diamond film. An example of this is shown in Fig. 16. By contrast, the removal of the binder phase to a depth less than the dimension of the free-surface WC grains may allow too much chemical interaction between the diamond and binder phase at the high temperatures typical of traditional CVD diamond synthesis. This interaction results in dissolution of the diamond and diffusion of the carbon into the metallic binder phase as well as diffusion of the binder phase into the diamond as illustrated in Fig. 17. This effectively reduces the interfacial contact area

Figure 15 Overetched WC-Co surfaces yield WC-Co/diamond interfaces that are susceptible to mechanical failure. (a) A schematic representation of the cross section of a WC-Co substrate. The goal of most wet-chemical etch processes is to remove the binder phase from a region near the free surface of the WC-Co surface. If the etch depth extends further into the substrate than the top layer of the WC grains (b), the toughness of the interface will be severely degraded. Following diamond deposition on this surface (c), the diamond-coated part will cool from deposition temperature. Because of the mismatch in the coefficient of thermal expansion between the film and substrate, the interface will be in tension due to residual thermal stresses. In the presence of further applied stresses, crack extension along the interface between the near-surface WC grains and the unetched region of the substrate (d) can lead to failure of this region. This is sometimes referred to as grain pullout as the interface side of the diamond film contains WC grains that were pulled from the substrate (e). (From Ref. 74.)

CVD Diamond Solutions for Machining

467

Figure 16 SEM micrograph of the interface side of a diamond coating that delaminated due to the failure mode shown in Fig. 15. The bright contrast, faceted grains were verified to be WC grains using energy dispersive spectroscopy (EDS). Detection of these WC grains on the interface side of the diamond film verifies that failure of the interface region occurred by crack extension in the region of the substrate in which the binder phase was removed. (From Norton Diamond Film/ Saint Gobain Industrial Ceramics, Inc.)

between the film and substrate and sacrifices the mechanical integrity of the diamond-coated component. This failure mode has also been characterized by SEM and energy dispersive spectroscopy (EDS) as presented in Fig. 18. Metal machining operations induce cutting forces that often exceed the level of adhesion offered by etching processes. Extensive cutting tests and interfacial analysis of diamond coatings on pre-etched WC-Co surfaces indicate* that etched interfaces of common WC-Co grades offer an insufficient level of adhesion for reproducible performance in applications such as aluminum machining. The adhesion limitations of etched WC-Co are especially evident under more aggressive machining conditions that induce impact forces on the interface. Notable examples include milling, interrupted turning, and the machining of workpiece materials with structural inhomogeneities (e.g., high-silicon aluminum, MMCs). Other testing has indicated that good etching process control can allow production of diamond films on *Comment is based on studies performed within and monitored by Norton Diamond Film prior to business closure.

468

Cline and Olson

Figure 17 Underetched WC-Co surfaces yield WC-Co/diamond interfaces that are susceptible to chemical degradation during coating. (a) Schematic representation of the cross section of a WC-Co substrate. (b) The goal of most wet-chemical etch processes is to remove the binder phase from a region near the free surface of the WC-Co surface. (c and d) If the etch depth does not extend sufficiently far into the substrate, the binder phase and diamond film may interact under the high temperatures typical of most CVD diamond growth conditions. The Co can induce a phase transformation from diamond to graphite. The carbon then diffuses into the binder phase as the cobalt diffuses into the film. The resulting interfacial porosity results in poor bonding between the film and substate. (e) Depending on the magnitude of the damage, delamination of the film can occur because of residual thermal stresses or the applied stresses of the application. (From Ref. 74.)

> Figure 18 Structural and chemical analysis of cobalt-induced interfacial degradation. At the high temperatures present in typical CVD growth environments, diffusion of the binder phase into the film and subsequent diffusion of carbon from the film into the binder phase may occur. Given time, the diamond-substrate interface may be dissolved, thereby reducing the interfacial contact area and adhesion strength. The SEM image of a film that failed by this mode (a) was taken of the interface side of a diamond film following delamination. The pores displayed in the image were created in the diamond film by binder phase diffusion from the bulk into the film during deposition. Within the interior of the voids, small spheres of Co have solidified upon cooling down from deposition temperature. The corresponding energy dispersive spectrum (b) identifies the composition of the small, bright spheres as primarily cobalt. Co-diamond interactions may also be viewed in cross section as observed in the composite SEM image of a diamond WC-Co interface shown in (c).

CVD Diamond Solutions for Machining

469

This SEM image was taken of the interface region between a diamond film and WC-Co substrate in which the interface between the diamond and substrate has resulted in the diffusion of carbon from the diamond into the binder phase of the substrate. Upon cooling down from deposition, the carbon in the binder phase has precipitated out in the form of graphite (shown as the ‘‘fuzzy’’ dark contrast material in the central region of the image). (From Ref. 74.)

470

Figure 18

Cline and Olson

Continued

common grades of etched WC-Co in applications that generate lower cutting forces such as nonmetallic machining. b. Diffusion Barrier Approaches. Recognizing that a physical barrier or so-called diffusion barrier to diamond-binder interaction may improve adhesion by preventing interaction between the binder phase and the diamond film, other researchers [70–72] have developed techniques for the

CVD Diamond Solutions for Machining

471

formation of interlayers between the CVD diamond film and WC-Co substrate. Proper selection of such a layer could also reduce residual stresses between the diamond film and the underlying substrate by choosing an interlayer material with a coefficient of thermal expansion that falls between the film and underlying substrate. Destructive analysis of commercial CVD diamond products produced by various industrial vendors indicates that interlayer technology is used; however, the specific intent of the interlayer design was not evident and observed failures were typically at the interlayer.* Testing indicated that the specific approaches studied in commercial samples failed to yield sufficient interfacial toughness to resist delamination in abrasive aluminum machining; however, the interfacial integrity of the same products may have been suitable for other less aggressive operations. Academic research on interlayers reported by Kupp [71] revealed a suppression of chemical interaction but generally poor cutting test performance based on industrial expectations. Although diffusion barrier approaches can add measurable cost and complexity to adhesion treatment, it is quite possible that unique interlayer solutions will evolve as the CVD diamond industry advances. c. Mechanical Roughening Techniques. Other researchers have recognized that rough interfaces potentially provide an increase in the adhesion strength of CVD diamond films or WC-Co materials. This has led to methods that include mechanical roughening by abrasion, wet-chemical etching of the WC phase [73], and laser ablation or blasting with abrasive grit. Unfortunately, most of these techniques form an interface that has sustained significant damage and actually reduce the interfacial toughness rather than improve it. d. High-Temperature Heat Treatments. Saijo and coworkers [73] discuss a process for treatment of WC-Co with a binder phase of 4 wt% or less using a decarburizing gas composed of oxygen and hydrogen at a temperature between 500 and 1200⬚C. Although the decarburization of the freesurface WC grains promotes chemical bonding between the diamond film and substrate through recarburization during CVD diamond deposition, the method produces a free surface in which the WC grains are smaller than in the bulk. According to Olson and Windischmann [74], this process therefore does not provide the crack deflection or interfacial toughening mechanism believed to be essential in highly abrasive applications and may create an embrittled surface layer due to decreased grain contiguity.

*Comment is based on studies performed within Norton Diamond Film prior to business closure.

472

Cline and Olson

e. Present Status and Future Needs for Adhesion Treatments. The often undisclosed, proprietary adhesion treatments used by diamond coating vendors have been successful in producing suitable adhesion for selected applications but have still not resulted in diamond-coated WC-Co products that can be used for the full range of machining applications envisioned for CVD diamond a decade ago. For instance, diamond-coated tungsten carbide products offer measurable performance advantages in machining selected nonmetallic materials; however, adhesion issues have clearly limited the success in machining common metallic workpiece materials such as aluminum. It now seems apparent that the nature and extent of adhesion challenges for diamond-coated tungsten carbide cutting tools and wear parts create the need for a range of adhesion treatment solutions that are selected based on the specific demands of the application. For purposes of instruction, Sec. III.B.3 contains a case study involving adhesion treatment technology developed specifically for metalcutting insert manufacture. Section IV.B.1.a contains related metal cutting results. 3.

Case Study: Development of an Adhesion Treatment for Metalcutting Inserts

In order to maximize adhesion strength of diamond coatings of WC-Co substrate materials, it is essential to optimize many key features of the diamond-substrate interface. Unless all of these characteristics are in place, optimal adhesion strength cannot be realized for the most abrasive applications. In an approach reported by Olson and Windischmann [74], the essential characteristics of the diamond-substrate interface are produced that, acting together, optimize the adhesion strength of the diamond film to the underlying substrate. These characteristics fall in three general categories: (1) chemical composition, (2) phase composition, and (3) microstructural composition. The remainder of this section contains a detailed discussion of the specific attributes of this adhesion treatment for WC-Co cutting tool inserts. The chemical composition of the interface is altered to avoid deleterious binder phase–diamond reactions that can reduce the chemical bonding of the diamond film to the substrate and also induce a phase transformation of the diamond film to graphite. However, unlike techniques that remove the binder phase to some depth below the exposed WC-Co substrate surface, binder phase removal is done in a way which limits removal to an area that is directly exposed to the CVD growth species or ‘‘free surface.’’ In addition, the composition of the WC phase at the free surface is altered to maximize the density of direct chemical bonding between the diamond film and substrate. Unlike chemical etching methods which decarburize the WC grains

CVD Diamond Solutions for Machining

473

by chemically ‘‘attacking’’ them, this is performed in a manner intended to maintain the mechanical properties of the substrate and interface. Finally, the microstructural composition of the interface is altered to minimize crack nucleation sites due to interfacial voids and provide a toughening, crack deflection mechanism that resists interfacial crack propagation. This feature essentially arrests or deflects cracks that may nucleate at the interface and impedes the propagation of these cracks, thereby suppressing delamination. The process utilizes commercially available WC-Co cutting tool inserts with a specific grain size and composition. The typical composition of the starting grade is 94.0 wt% WC, 6.0 wt% Co with an average grain size of the WC phase of about 1–2 ␮m (Sandvik H13A, Carboloy 883, and others). Because the substrate in this case is in the form of a commercially available cutting tool insert, the free surface of the material is usually finish ground to final dimensional tolerances. Therefore, the free-surface usually displays the impression of residual grind lines and surface damage shown in Fig. 19a. This surface contains fragments of WC grains that have been fractured during the grinding step and a binder phase, which is composed mostly of Co, smeared across the surface. In the current approach, the free-surface binder phase is vaporized using a proprietary technique. Because of the nature of the process, the freesurface WC grain size is increased while the stoichiometry of the WC grains is shifted to a slightly C-deficient ratio. This leaves the free surface or ‘‘shell’’ of individual WC grains at the free surface of the substrate with a slightly decarburized ‘‘case.’’ Figure 19b shows the marked change in surface structure relative to the as-received WC-Co surface in Fig. 19a. Elemental analysis of the adhesion treated surface (Fig. 19b) by EDS indicates a surface that is essentially free of metallic binder phases (Fig. 19a). The combination of the simultaneous vaporization of the binder phase, which consequently increases the surface free energy of the WC grains, and the slight decarburization of the WC grains increases the dangling bond density of the free-surface grains. Enlargement of these free-surface WC grains is induced simultaneously with the binder phase vaporization and shift in stoichiometry. Thus, all three components of the required interface may be evolved: (1) the binder phase is removed through vaporization, (2) the chemical bonding between the diamond film and substrate may be maximized through a recarburization/diamond film growth step, and (3) the enlarged WC grains coupled with a high nucleation density provide the crack deflection enhancing microstructure that toughens the interface. All of these features are created using a one-step process developed to produce these characteristics in a way that minimizes the trade-offs of other important physical features of the carbide substrate.

474

Cline and Olson

Figure 19 (a) SEM micrograph showing the finish-ground surface of a 6% CoWC cutting tool. Most adhesion technologies are targeted at removal of the metal binder that is smeared on the surface during grinding. (b) Using the surface modification technique described in Sec. III.B.3, the binder phase can be selectively removed from commercially available grades of as-ground WC-Co cutting tool inserts. As the image depicts, the WC grains are restructured to create a pristine WC surface ready for diamond coating. EDS spectra (not shown) indicate that an essentially cobalt-free substrate surface can be produced by this technique. (From Ref. 74.)

CVD Diamond Solutions for Machining

475

Following the evolution of the surface microstructure and chemical composition of the substrate as just described, the surface is coated with a diamond film [74]. Diamond growth was carried out in both high-rate, microwave plasma CVD and DC arcjet CVD systems under high deposition rate conditions that suppress bulk-to-interface diffusion of the binder phase. The early stages of the CVD diamond process were tailored to optimize the nucleation density without formation of a nondiamond interphase. The characteristics of the diamond film have been developed to optimize the wear properties for abrasive environments as described in Sec. III.A and previously by Olson and Dawes [28]. The interface structure of a diamond-coated WC-Co cutting tool insert prepared by this technique is contrasted with that of a diamond-coated insert prepared by chemical etching in Fig. 20. Section IV.B.1.a contains the aluminum machining performance of inserts produced using this specific interfacial engineering approach.

IV.

CVD DIAMOND CUTTING TOOL PERFORMANCE

As discussed in Secs. II and III, there are many factors that can affect the performance of a specific CVD diamond cutting tool design in a chosen application. This section contains application case studies of both thin-film and thick-film CVD diamond cutting tools. The case studies are intended to clearly show that the relative performance advantage of CVD diamond tooling over conventional tooling varies widely depending on the specifics of the machining application and the tool design concepts employed. The case studies include discussion of the most critical application characteristics as well as the cutting tool wear mechanisms that dictate tool life and machining efficiency. A.

Freestanding Thick-Film CVD Diamond Cutting Tool Performance

1.

Thick-Film CVD Diamond Insert Performance

a. ‘‘High-Silicon’’ Aluminum Machining with Thick-Film CVD Diamond. As noted in Table 2 (Sec. II.C.2), silicon is commonly added to aluminum compositions to modify alloy properties. The Si-Al alloy has a eutectic near 12.2% Si-Al. Higher amounts of silicon in the composition result in the precipitation of hard, abrasive silicon particles in the workpiece that enhance the wear resistance of the aluminum surface. For instance, A390 (18% Si-Al) hypereutectic aluminum, which is commonly used in automotive pistons, contains approximately 6% hard silicon particles that vary in

476

Cline and Olson

Figure 20 Interfacial comparison of two adhesion treatment approaches. Suitable diamond deposition onto substrates produced using the surface modification technique described in Sec. III.B.3 yields an interface that is free of large voids and of the chemical interfacial degradation observed in Co-diamond interactions. (a) Crosssectional SEM image of the diamond film–substrate interface. (b) For comparison, a cross-sectional SEM image of a diamond film on an etched WC-Co substrate. Note the porosity of the binder-depleted zone (embrittled layer) and the low interfacial roughness of the interface shown in (b) versus (a). (From Ref. 74.)

CVD Diamond Solutions for Machining

477

size and distribution depending on the aluminum processing technique. According to Whitacre, a piston industry representative [75], ‘‘diamond tooling is an absolute requirement to successfully machine hypereutectic [high-silicon] alloys.’’ Because the volume of silicon particles increases with increasing silicon content, the machinability of low-silicon hypoeutectic aluminum alloys is significantly better than that of high-silicon hypereutectic aluminum alloys. Although PCD tooling is widely used for hypereutectic alloy machining in the piston industry, some companies have been benefiting from the enhanced tool life offered by brazed-on, thick-film CVD diamond tips since the early 1990s. CVD diamond properties captured in a properly brazed and finish-ground thick-film CVD diamond tool tip can yield extended tool life in hypereutectic aluminum alloys while producing better and more consistent surface finish than PCD. For a piston machining operation, Hay [8] reported that a CVD diamond tool offered three times the tool life of PCD in a finish profile turning operation. Using a cutting speed of 700 m/min, water–soluble oil coolant, and a depth of cut of 0.15 mm, A390 hypereutectic aluminum pistons were machined in a two-pass operation with the first pass using a feed rate of 0.30 mm/revolution and the second pass using a feed rate of 0.15 mm/rev. The 25-␮m grain size PCD tools yielded approximately 10,000 pistons/tool before finish specification was lost. By comparison, the properties of CVD diamond enabled the facility to produce an average of 30,000 pistons/tool in the same operation. The 100⫻ SEM image in Fig. 7 is that of a wear scar found on a CVD diamond piston turning tool after machining A390 aluminum pistons. This tool has worn to a point that requires regrinding of the flank face of the tool in order to resharpen the cutting edge. Depending on the extent of edge damage before each resharpening, tools in this application are reground more than five times before disposal. During its useful lifetime, a single CVD diamond tip is capable of machining over 100,000 high-silicon aluminum pistons. Wear of the CVD diamond can be seen along the cutting edge where the finish-ground flank face of the tool meets the polished rake face. The higher magnification 500⫻ image in Fig. 21 shows that the wear scar has some striations that may be the result of anisotropic wear of the diamond crystallites or, more likely, silicon particle impact damage. Hard particle damage to the polished rake face would tend to evolve into striations along the flank face wear surface, which is in constant contact with the rotating piston. Although the tool tip is visually reflective with no significant buildup of aluminum, the dark region of contrast at the top of the 100⫻ image in Fig. 7 is likely to be a submicrometer aluminum layer resulting from chip flow. Also noteworthy is the apparent waviness of the rake face in the bright region on the left side of the 100⫻ image. This feature is

478

Cline and Olson

Figure 21 Flank-face SEM image of a freestanding CVD diamond cutting tool after use in turning high-silicon aluminum pistons. This image is a 500⫻ close-up of the 100⫻ image shown in Fig. 7. A single CVD diamond tip is capable of machining over 100,000 high-silicon aluminum pistons. (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc.)

probably induced by the erosive nature of the Si-Al chips that are formed in this region of the tool. b. ‘‘Low-Silicon’’ Aluminum Machining with Thick-Film CVD Diamond. Although hypoeutectic low-silicon aluminum compositions have good machinability and can generally be machined with nondiamond tooling materials such as tungsten carbide, diamond-tipped tooling is commonly used in mass production environments, which benefit from diamond’s ability to endure many hours of stable, high-speed operation. Drury has reported (T. Drury, personal communication, 2000) [1] performance in an aluminumfinishing operation in which CVD diamond-tipped tooling not only outlasted PCD by 30% but also maintained the required surface finish throughout the tool life. Hay [8] has reported the results of a finish-boring application where hard, anodized 6061 aluminum turbine cases were machined with both CVD and PCD diamond-tipped tools. Using a cutting speed of over 457 m/min, a depth of cut of 0.013 mm, and a feed rate of 0.05 mm/rev, CVD diamond outperformed fine-grain PCD (5 ␮m grain size) under stringent surface finish specifications. The CVD diamond tool reportedly increased tool life from 106 parts per tool up to 345 parts per tool and surface finish was more consistent than that of the PCD tooling. This is one of many examples where

CVD Diamond Solutions for Machining

479

thick-film CVD diamond not only extends wear life but also produces and maintains superior surface finish relative to PCD tools. c. Metal Matrix Composite Machining with Thick-Film CVD Diamond. As noted in Table 2 (Sec. II.C.2), metal matrix composites (MMCs) are typically reinforced with over 20% ceramic particles or fibers, making them extremely difficult or impossible to machine with conventional tooling such as cemented carbide. One of the present uses of MMCs is for automotive brake rotors where aluminum matrix composites are employed. MMCs are also of interest for improved high-speed rail braking systems and components for the electronics industry. CVD diamond tooling has shown promise as an alternative to PCD tooling or, in some cases, grinding or water-jet cutting MMCs. In tests performed by an Austrian research group [76] using a 30% SiC reinforced aluminum-matrix composite, it was determined that larger grain size PCD significantly outperformed smaller grain size PCD. In the same tests, CVD diamond-tipped tools outperformed all forms of PCD tested in finishing the MMC material. As an example, after 10,000 m of finishing, the CVD diamond tool had a wear scar less than 25 ␮m in width, whereas 25- and 40-␮m grain size PCD had worn to over 120 and 230 ␮m, respectively. The superior wear resistance offered by CVD diamond is attributed to its binder-free, pure diamond construction, which reduces the tendency for diamond grain ‘‘pullout’’ when eroded by the SiC particles. Because CVD diamond can be manufactured with high diamond-to-diamond grain boundary strength, the grain size of the diamond crystallites is not nearly as critical for CVD as it is for cobalt-sintered PCD. However, as discussed in Sec. III.A, the grain boundary strength of CVD diamond deposits is likely to be very critical in abrasive applications such as MMC machining. The high erosion resistance of thick-film CVD diamond in machining MMCs and high-silicon aluminum materials is probably due to the combination of diamond’s hardness, fracture toughness, lubricity, and other key properties. High-silicon aluminum or MMC machining requires very high erosion resistance to hard particles. In general, the metal matrix of both MMCs (including high-silicon aluminum) is free-machining aluminum; therefore, tool wear is dictated by the size and distribution of the hard particulate reinforcement. Table 3 contains erosion resistance data for CVD diamond and other materials known for their hardness or wear resistance. In these tests, CVD diamond clearly showed superior erosion resistance compared with alternative tooling materials with a relative volume loss four times lower than for PCD and 120 times lower than for tungsten carbide. The performance of thin-film tooling in erosive workpiece materials requires not only erosion-resistant diamond but also interfacial impact resistance be-

480

Cline and Olson

Table 3

Relative Erosion Resistance of CVD Diamond and Other Materials

Material

Relative volume loss

CVD diamond (low pressure, vapor deposited) PCD (high pressure, cobalt sintered) Cemented tungsten carbide (6% Co-WC sintered) Aluminum oxide ceramic (99.5% Al2O3) Silicon carbide ceramic Silicon nitride ceramic Erosion test parameters: 2% SiC particles in water (30 grit; >0.5 mm average particle size) 91.4 m/sec abrasive flow rate 45⬚ angle of impingement

1 4 120 220 360 920

Source: Norton Diamond Film, SGIC, Inc. internal test data.

tween the tooling material and the coating. The performance of diamondcoated inserts in high-silicon aluminum machining is discussed in Sec. IV.B.1.a. The performance of diamond-coated drills in MMCs is discussed in Sec. IV.B.2.c. d. Sintered Tungsten Carbide Boring with Thick-Film CVD Diamond. Tungsten carbide not only offers unique solutions as a cutting tool material as discussed in Secs. I.B.2 and II.B, but also is commonly used in tough, wear-resistant wear parts. In selected applications, tungsten carbide is precision machined using PCD, PCBN, and, more recently, CVD diamond. In a boring application, a 25% Co-WC cylinder with a 20 mm inner diameter and 40 mm length was machined using a speed of 30 m/min, feed of 0.05 mm/rotation, and depth of cut of 0.12 mm. Flood coolant was used. The results are summarized in Fig. 22. The CVD diamond thick-film tool machined eight parts before failure. PCD machined only five parts per tool and PCBN was only able to machine one part. Another carbide boring application demonstrated the ability of CVD diamond to produce a similar finish to single-crystal diamond tooling. The ability to machine dry, with no coolant, has also been demonstrated with CVD diamond thick-film tooling. 2.

Thick-Film CVD Diamond Rotary Tool Performance

The performance advantages of thick-film CVD diamond relative to PCD are similar to that observed for thick-film CVD diamond inserts. For example, in a 19% Si-Al wrist pin bore reaming operation, a PCD reamer was replaced with a thick-film CVD diamond blade reamer insert. Reamer life

CVD Diamond Solutions for Machining

481

Figure 22 Performance comparison of CVD diamond, PCD, and PCBN in sintered tungsten carbide boring (25% Co-WC). (Data provided by a customer of Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc. and Ceratonia.)

was extended by three to four times using a cutting speed of 628 m/min, speed of 51 mm/rev, with a depth of cut of 0.076 mm per side. Even though the potential performance advantage are significant, the present use of thickfilm CVD diamond in rotary tooling including reamers, end mills, routers, and drills is much less extensive than the usage in the simpler geometries of inserts. One of the major barriers for thick-film CVD rotary tools is the inability to EDM ‘‘wire-cut’’ CVD in a manner similar to that with PCD. Many of these challenges will ultimately be overcome as methods of fabricating with flat, thick-film CVD shapes evolve or alternative, adherent solutions are developed with thin-film diamond. Since diamond-coated tungsten carbide rotary tools are the present focus of the industry, performance advantages of thin-film diamond rotary tool designs are discussed extensively in Sec. IV.B.2. B.

Thin-Film CVD Diamond Cutting Tool Performance

1.

Thin-Film CVD Diamond Insert Performance

Depending on the application, diamond-coated inserts can show a cost-performance advantage relative to uncoated carbide, other hard coatings, and

482

Cline and Olson

PCD. Although there are cases where single corners of a CVD diamond insert will outperform a single PCD tip, the indexability (and disposable design) of diamond-coated inserts is generally a key advantage relative to regrindable PCD tips. CVD diamond-coated inserts generally cannot produce the superior surface finish that is expected from fabricated PCD tips. In some cases, diamond-coated insert design limits the use of diamond-coated tools to roughing and intermediate finishing processes. In such instances, fabricated, thick-film CVD diamond tips are a viable alternative for long wear life and improved surface finish over PCD. a. Case Study: Performance of Diamond-Coated Inserts. To evaluate fully the level of performance of the CVD diamond-coated WC-Co cutting tool inserts manufactured using the process described in Sec. III.B.3, inserts were tested in a range of industrial applications under a variety of conditions involving workpiece materials ranging from graphite to hypereutectic aluminum. The tests were carried out in machining laboratory settings as well as various industrial settings by independent evaluators, and the results were compared with those for tools traditionally used in the specific application. HIGH-SILICON ALUMINUM TURNING. In A390 high-silicon aluminum turning tests, the tools were demonstrated to last significantly longer than uncoated carbide tools or PCD-tipped tools. Using a cutting speed of 680 m/min, feed rate of 0.2 mm/rev, and depth of cut of 1.0 mm, diamondcoated C2 grade WC-Co TPG 321 style inserts were tested against uncoated carbide, coarse grain PCD (25 ␮m grain size), and thick-film CVD diamond of the same tool geometry. Both the PCD and thick-film CVD tools are regrindable by design; the diamond-coated and uncoated carbide tools are disposable. Tools were tested to a predetermined failure criterion defined by 0.375 mm of flank wear. Figure 23a shows the results of the tool life testing. Thick-film CVD diamond (500 ␮m diamond brazed-on and finish ground) produced the best per tip tool life, followed by thin-film CVD diamond (25– 35 ␮m coating), PCD, and uncoated C2 grade, 6% Co-WC carbide. Because diamond-coated tools are indexable and a TPG 321 tool is triangular, the per tool lifetime was actually highest for the disposable CVD diamond thinfilm tools. The thin-film tool’s rough interface (created by the adhesion treatment discussed in Sec. III.B.3) coupled with the faceted diamond morphology critical to high wear resistance in the A390 workpiece limits the ability of this tool design in finishing selected workpiece materials. As Fig. 23b summarizes, the finish-ground cutting edges on both PCD and thickfilm CVD diamond yielded surface roughness values near 0.4 ␮m R a, whereas, the diamond-coated tools produced values near 0.75 ␮m R a. Using this and other case studies, the adhesion treatment approach discussed in Sec. III.B.3 proved to offer suitable adhesion to endure various

CVD Diamond Solutions for Machining

483

Figure 23 Comparison of PCD, CVD diamond thick-film, CVD diamond thinfilm, and uncoated WC-Co tooling in turning high-silicon aluminum. Tool lifetime is compared in (a). The surface finish of the Mahle A390 (⬃18% Si-Al) workpiece is shown in (b). (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc.)

484

Cline and Olson

abrasive wear and impact-oriented applications. Unfortunately, the efforts to engineer the necessary properties into the diamond-coated WC-Co ‘‘composite’’ required a trade-off in surface finish of the workpiece that generally limits the use of this particular tooling design concept to intermediate finishing and roughing. Figure 6 (Sec. II.D.1.a) depicts this surface roughness versus adhesion trade-off. Figure 6b shows a relatively rough, diamondcoated tool produced by this method. A uniform wear pattern can be observed at the cutting corner. Figure 6a shows a relatively smooth, but delaminated tool produced by a wet chemical etching adhesion treatment. As Fig. 23 depicts, the thick-film, brazed-on diamond tip not only demonstrated superior life relative to coarse-grained PCD but also delivered comparable finish. b. Performance of Other Thin-Film Diamond Tool Designs HIGH-SILICON ALUMINUM TURNING. A CVD diamond machining study by Oles et al. [23] reported turning test results for two hypereutectic alloys (Reynolds A390 and Mahle 138) that contained nominally 18% silicon. Machining conditions were as follows: 762 m/min, 0.127 mm/rev, 0.635 mm depth of cut, 15⬚ lead angle, and flood coolant. The tested tools were SPGN120308-style inserts with a diamond-coating thickness of approximately 30 ␮m. Regardless of the alloy type, diamond-coated and PCD tools lasted approximately two orders of magnitude longer than the uncoated carbide tools, which reached the failure criterion of the testing within seconds. This exceptionally rapid wear was induced by the hard silicon particles and clarifies why diamond tooling is generally accepted as the only suitable solution for high-productivity machining of high-silicon aluminum alloys [75]. The study also showed the strong influence of the workpiece microstructure on the wear rate and general performance of CVD diamond–coated tools relative to PCD. In Mahle 138 alloy, the steady-state wear rates on the coating average 0.005 mm/min, whereas the wear rate averaged 0.012 mm/min when machining Reynolds A390. The higher wear rates observed in the Reynolds A390 were attributed to the higher level of impact damage induced by the larger, less homogeneous silicon particulates. The microstructure of the A390 alloy contained a nonuniform distribution of relatively larger, irregularshaped silicon particles (⬃0.09 mm size, 100 particles/cm2), whereas the 138 alloy contained a uniform distribution of smaller, spheroidal particles (⬃0.02 mm size, 160 particles/cm2). LOW-SILICON ALUMINUM TURNING. Oles et al. [23] also reported turning tests using CVD diamond-coated tools in hypoeutectic 383.2 aluminum (11% silicon) using the same machining conditions summarized before for the hypereutectic alloy study. The steady-state diamond wear rate

CVD Diamond Solutions for Machining

485

was more than five times lower than the wear rates reported for the highsilicon alloys in the previous section. The wear rate for uncoated carbide was 0.011 mm/min compared with 0.001 mm/min estimated for the diamond-coated tooling. The authors noted that the much lower wear rates of the cemented carbide tooling explain the general acceptance of uncoated and TiN-coated carbides as a solution for machining hypoeutectic, low-silicon aluminum alloys. CARBON-FILLED PHENOLIC COMPOSITE TURNING. Hay [8] has reported the performance of diamond-coated silicon nitride ceramic tools in a 30 vol% carbon-filled phenolic resin machining application. Precision parts were turned using a cutting speed of 122 m/min, feed rate of 0.003 ipr, and a depth of cut of 0.635 mm. To counter the corrosive effects of the workpiece, aluminum oxide ceramic tools were used to attain approximately 50 parts per cutting corner and a marked cost-lifetime advantage over metalbonded PCD and tungsten carbide tooling. The introduction of diamondcoated silicon nitride tools into the same operation now offers up to 650 parts per corner due to an unmatched combination of corrosive and abrasive wear resistance. 2.

Thin-Film CVD Diamond Rotary Tool Performance

a. General Status of Nonmetal Machining with Diamond-Coated Rotary Tools. Some of the most impressive tool life improvements with CVD diamond have been reported in applications using diamond-coated rotary tools. In cases where the level of adhesion is suitable for the application, the lifetime advantage of a diamond-coated tungsten carbide tool is often dramatic. Because of the geometric complexity of most rotary tools, PCD (or thick-film diamond) tooling is generally not a cost-effective option; therefore, tungsten carbide and hard coatings on carbide are commonly used in many high-throughput environments. Lifetime advantages of diamondcoated tools relative to other hard coating or uncoated carbide generally fall in the range of 10 to 100 times for machining nonmetallic materials including structural composites, abrasive or corrosive polymers, green ceramics, graphite, and hard carbon. Selected case studies of nonmetallic machining applications are summarized in Sec. IV.B.2.c. b. General Status of Metal Machining with Diamond-Coated Rotary Tools. Although similarly impressive performance advantages have been observed in metal machining applications, not all vendors recommend their diamond-coated rotary tool products for metalcutting at the time of publication. Although much of the potential for diamond-coated tooling is projected to be automotive and aluminum machining, the present use of diamond-coated rotary tools in metalcutting applications is not nearly as

486

Cline and Olson

widespread as for nonmetals. Engineering diamond-coated tools to sustain the shear forces of metal machining has been a major barrier to expansion of their application range. There have been cases where commercial claims of metalcutting capability have been retracted from the marketplace because of reproducibility issues in the field. Even with these challenges, rotary tool users who machine nonferrous metals such as aluminum alloys, copper, brass, bronze, and magnesium should closely monitor the expanding application range of CVD diamond-coated tooling as commercial adhesion technology and application know-how continue to evolve. The following section contains a summary of a case study on MMC drilling. c. Diamond-Coated Rotary Tool Case Studies GRAPHITE AND HARD-CARBON MATCHING. The advantages of CVD diamond-coated rotary tools in enhancing graphite-machining operations are presently aggressively marketed by many vendors and are becoming well known in the machining industry [18–20,77]. Ten to 20 times tool lifetime advantages are commonly observed in small job shops and large machine shops involved in EDM sinker-die electrode and other graphite component production. Success has been achieved with both outdated milling machines and high-speed machining centers. High-speed machines amplify the advantages of CVD diamond as other hard coatings and uncoated carbide both fall short at operating speeds in excess of 900 m/min. When forms of carbon harder than graphite are machined, the relative performance numbers are often even more dramatic than the graphite machining results. In an antimony-impregnated hard-carbon (carbon-graphite) end milling application (T. Uemura, personal communication of field test data, 1999), a diamond-coated end mill with a faceted 10 to 15-␮m coating was used as a replacement for a commercially available diamond-like carbon (DLC) coating. Using a 9.5 mm diameter, four-flute, square end mill, notches were machined into a gas seal ring. DLC-coated end mills were able to produce approximately 120 parts per tool using the following conditions: 1000 rpm (30 m/min), 38 mm/min feed rate, and a depth of cut of 5.3 mm. Not only did CVD diamond demonstrate superior performance at the same conditions, but also the properties of the coating enabled operation under more aggressive conditions: 3000 rpm (a threefold increase), 1016 mm/min feed rate (a 26.7-fold increase), and a depth of cut of 5.3 mm. Under these conditions, the CVD diamond-coated end mill produced approximately 11,000 parts and the customer realized tool life increases of over 90 times relative to the DLC-coated tool. More important, the machining process productivity was dramatically affected by the increases in speed and feed enabled by the diamond-coated tool as per part cycle times were decreased from 20 to 3 minutes per seal ring.

CVD Diamond Solutions for Machining

487

Although the present costs of DLC tooling are markedly lower than those of CVD diamond coatings, the cost-performance advantage of pure, crystalline, CVD diamond is clear in many applications such as the preceding hard-carbon case study. Most DLC coatings are limited in higher temperature cutting applications (often induced by higher speed operations) because of phase instability at temperatures as low as 250–350⬚C. The level of phase instability has been determined to be a function of the hydrogen content in DLC materials with pure amorphous carbon having the best stability of materials in the continuum of amorphous hydrogenated carbon (aH:C) materials. Although Horsfall [78] has reported properties of noncrystalline carbon coatings that have low hydrogen content and high-temperature stability above 500⬚C, the continuum of diamond-like carbon materials are limited in performance relative to CVD diamond because of their relative thermal instability. Furthermore, due to compressive stresses in DLC coatings, they are often limited to a coating thickness of approximately 2 ␮m, over five times thinner than many commercially available diamond-coated tools. In addition to per tool cost advantages, DLC coatings have an advantage over diamond coatings in edge sharpness (a function of coating thickness), coating surface roughness (typical of amorphous coatings), and much lower sensitivity to substrate selection (due to low-temperature coating process). PRESINTERED CERAMIC MACHINING. In the presintered ceramic machining test discussed earlier in Sec. II.C.1, the performance of 9.5 mm diameter, four-flute diamond-coated end mills was compared with that of uncoated tungsten carbide and PVD TiN-coated carbide tools of the same geometry. The spindle speed was 10,000 rpm, with a feed rate of 508 mm/ min (20 ipm) corresponding to a cutting speed near 300 m/min (1000 sfm) and a chip load of approximately 0.012 mm. The presintered ceramic workpiece was machined with no coolant using alternating, 7.6 cm long climb (up-cut) and conventional (down-cut) passes with a 4.8 mm radial depth of cut and a 12.7 mm axial depth of cut. Cutting forces on the workpiece were measured using a three-axis dynamometer and a computerized data collection system. Testing details have been reported elsewhere by Tanikella et al. [32] and Cline [31]. Figure 5 visually depicted the difference in cutting temperatures between diamond and the alternative tools in this test. This dramatic result has been attributed to the superior lubricity (low coefficient of friction) of the diamond coating when in contact with the ceramic workpiece material. According to Quinto et al. [79,80], the hardness of PVD TiN-coated carbide at 1000⬚C is approximately 25% of its room temperature hardness. This ‘‘hot hardness’’ limitation explains the lack of any performance advantage afforded by the PVD TiN coating relative to uncoated carbide under these

488

Cline and Olson

cutting conditions. The reduced friction and wear were clearly reflected in the force and power data summarized in Table 4. In the first 7.6 cm long machining pass, the spindle power consumption for the TiN-coated carbide and uncoated carbide tools was over 10 times higher than the power consumption for the diamond tool after 80 passes. The lower cutting forces seen with CVD diamond-coated tools offer the potential for reduced yield losses in green or presintered ceramic machining. The low cutting temperature seen with the CVD diamond tool is viewed not only as an indication of the tremendous potential for CVD diamond tooling in high-speed machining of ceramics but also as the promise of CVD diamond coatings for machining many other difficult-to-machine abrasive materials. CARBON-EPOXY COMPOSITE MACHINING. Carbon-epoxy composites are of interest for lightweight, structural applications ranging from the aerospace industry to sporting goods. Because the hard, stiff carbon fibers in carbon-epoxy induce excessive wear in traditional cutting tool materials, some manufacturers have resorted to using alternative shaping techniques such as water jet cutting or computer numerical control (CNC) grinding with diamond grit tools. Other manufacturers are beginning to benefit from the advantages of CVD diamond-coated rotary tools in this difficult-to-machine material. Tanikella et al. [32] and Cline [2] have reported the results of carbonepoxy machining tests to determine the advantage of diamond-coated rotary tools over uncoated tungsten carbide two-flute, 9.5 mm diameter routers.

Table 4 Cutting Forces and Spindle Power Measured During Dry Machining of Presintered Ceramic Pass number

Fx (newtons)

Fy (newtons)

Fz (newtons)

Power (watts)

Typical measurements for uncoated or TiN-coated tungsten carbide end mills First pass (climb) Second pass (conventional)

200–250 300–400

100–150 500–700

20–50 80–100

800 1250–1400

Typical measurements for diamond-coated tungsten carbide end mills Average for pass 1 through 8 Average for pass 81 through 88 Source: Ref. 31.

40

40

15–25

50

60

60

25

50

CVD Diamond Solutions for Machining

489

Using a spindle speed of 20,000 rpm with a table speed of 2540 mm/min, alternating climb milling and conventional milling passes were performed on a 17.8 cm long, rigidly fixtured panel. The selected cutting conditions were in accordance with the interests of a large commercial aircraft manufacturer. As shown in Fig. 24a, severe wear was observed on the carbide tool in less than 36 cm of cutting. Figure 24b shows that cutting forces for the uncoated carbide tool increased dramatically during the first (climb) machining pass. After 20 passes, the CVD diamond-coated tool was still sharp and running predictably as depicted in Fig. 24c and d. Cutting noise and cutting odor typical of uncoated carbide tools were dramatically reduced with the use of CVD diamond-coated routers. Figure 25 shows the machined cross section of the carbon-epoxy composite after 40 passes using a diamond-coated router. The machined surface is burr free and cut cleanly enough to enable observation of the layered structure of the composite. It is interesting to note that the spacings between the composite layers in Fig. 25 match the spacings between the worn notches in the uncoated carbide tool shown in Fig. 24a. CVD diamond-coated tooling clearly has the potential to minimize costly deburring operations in composites fabrication. Even with flood coolant, the hardness and frictional heating generated by the carbon fibers clearly exceeded the performance limits of the uncoated carbide cutting edge. In another carbon-epoxy machining study [81], production testing of drills, drill countersinks, and two-flute helical end mills in carbon fiber composites showed that CVD diamond-coated cutting tools last up to 50 times longer than tungsten carbide and more than twice as long as PCD tools. This same reference reported that CVD diamond-coated tools were projected to increase the lifetime of machine tools by as much as 50%, significant cost savings to a machine shop. FIBER-REINFORCED POLYMER MACHINING. Matching tests were performed in a glass fiber-reinforced polymer (GFRP) using CVD diamondcoated tungsten carbide and uncoated carbide. The specific GFRP studied contained nominally 65% glass fibers and is, therefore, extremely difficult to machine. SEM micrographs of the composite cross section shown in Fig. 26 depict the extremely high density of fibers having a nominal diameter of 10 ␮m. Using 6.4 mm diameter, four-flute, square-end end mills, the GFRP was machined with no coolant at 12,000 rpm (239 m/min), a feed rate of 2438 mm/min, and an infeed of 3.2 mm. The selected machining conditions create a feed per tooth or ‘‘chip load’’ of 0.051 mm per cutting edge. Table 5 contains end mill diameter measurements for the CVD diamond-coated tool relative to an uncoated carbide tool. The dimensional stability of the CVD diamond-coated tool is dramatically better than that of the uncoated carbide tool. According to laser diameter measurements, the

490

Cline and Olson

Figure 24 Comparison of uncoated and CVD diamond-coated tungsten carbide router performance in carbon-epoxy machining. (a) After only two cutting passes less than 36 cm, optical inspection of the uncoated carbide tool showed severe tool wear. (b) The rapid degradation of the uncoated cutting edge was further exemplified in the spindle power data for the first pass (made in the climb direction). (c) In contrast, the CVD diamond-coated carbide tool exhibited relatively insignificant amounts of wear after 40 cutting passes (711 cm.). (d) The operational stability of the CVD diamond-coated tool is shown in the spindle power data for the 17th through 20th passes. Note: 1 division = 0.5 mm in (a) and (c). (From Ref. 2.)

CVD Diamond Solutions for Machining

Figure 24

491

Continued

diameter of the uncoated carbide was reduced by over 40 ␮m after only 62.9 cm3 of material removed, whereas the diamond-coated end mill lost less than 2 ␮m of diameter after removing 20 times more workpiece volume (1265.5 cm3) under the same cutting conditions. As Fig. 27 shows, structural analysis of the wear scar verifies that the CVD diamond coating shows only

492

Cline and Olson

Figure 25 Oblique view of the cross section of a carbon-epoxy panel that was machined in the 40th cutting pass with a CVD diamond-coated helical router. The layer spacings of the composite material seen in this image match the spacing between the worn notches in the uncoated carbide tool shown in Fig. 24a. (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc.)

superficial wear of the faceted diamond peaks at this point of testing. In contrast, the uncoated tool has significant amounts of wear after machining less than 5% of the workpiece volume (62.9 cm3). The workpiece finish produced by the CVD diamond-coated tool was superior and did not deteriorate during the testing. The test was stopped when the uncoated carbide tool reached a point at which it had clearly worn out (62.9 cm3 of material removed). Table 6 contains data for the power consumption by the machine tool spindle observed during the testing. The data confirm that the dimensional stability of the CVD diamond-coated cutting tool results in a consistent power draw on the spindle, whereas, the power draw on the spindle had increased by 50–70% for the uncoated carbide tool. This excess energy is primarily dissipated as frictional energy during cutting, which tends to overheat both the cutting tool and the polymer-based workpiece. The machinist noted that the uncoated carbide tools were ‘‘too hot to touch’’ from the onset of the test and that they started to glow ‘‘red hot’’ after about 47 cm3 of machining. The CVD diamond-coated tools ran ‘‘cool to the touch’’ throughout the entire test, presumably because of the high lubricity of the diamond when in contact with the high density of glass fibers in the workpiece.

CVD Diamond Solutions for Machining

493

Figure 26 SEM images of the cross section of a glass fiber-reinforced polymer (GFRP) component that was machined with a CVD diamond-coated end mill. The image in (a) depicts the high fiber volume percentage characteristic of the most abrasive GFRP materials. The image in (b) shows that the ⬃10 ␮m diameter fibers are pulverized during the machining action of the diamond tool and clarifies the need for abrasion-resistant cutting tool material in machining this particular workpiece. (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc. Test results and photographs courtesy of Walter Swann.)

Table 5

Laser Measurements of End Mill Diameter from GFRP Composite Machining Study

Tool type Uncoated WC-Co CVD diamond–coated WC-Co

Volume of material removed (cm3) 62.9 1262.5 (20 times more)

Pretest tool diameter measurements (in.)

Posttest tool diameter measurements (in.)

Change in tool diameter due to tool wear

␾1 ␾2 ␾1 ␾2

␾1 ␾2 ␾1 ␾2

␾1 ␾2 ␾1 ␾2

= = = =

0.24859 0.24862 0.24963 0.24970

= = = =

0.24692 0.24694 0.24960 0.24965

= = = =

⫺0.00167 ⫺0.00168 ⫺0.00003 ⫺0.00005

Source: Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc. Internal test data provided courtesy of Walter T. Swann.

in. in. in. in.

(⫺42 ␮m) (⫺43 ␮m) (⫺0.76 ␮m) (⫺1.27 ␮m)

CVD Diamond Solutions for Machining

495

Figure 27 Performance of CVD diamond-coated carbide routers in GFRP machining. The lifetime advantage of CVD diamond-coated WC-Co end mills relative to uncoated WC-Co was estimated to exceed 20 times based on laboratory testing. Low-magnification optical images of a diamond-coated WC-Co end mill (a) before testing and (b) after 1262.5 cm3 of GFRP material was machined indicated that no measurable wear was present. Low-magnification optical images of the uncoated WC-Co end mill (c) before testing and (d) after only 62.9 cm3 of GFRP material was machined indicated that extensive wear was present on the cutting edge. SEM images at 100⫻ magnification of the cutting edge taken after testing verified that essentially no wear was observed on the CVD diamond-coated end mill after 1262.5 cm3 of machining (e), whereas the uncoated tungsten carbide tool had suffered extensive abrasive wear after machining 20 times less (62.9 cm3) workpiece material (f ). (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc. Test results and photographs courtesy of Walter Swann.)

496

Cline and Olson

Figure 27

Continued

Consequently, a pungent burning odor of the polymer was noted only for the uncoated tools. The photographs in Fig. 26 are actually the machined surface created with a diamond-coated tool. Clearly, these conditions require the cutting edge to pulverize the hard, abrasive glass fibers during the machining process. The coating hardness and interfacial impact resistance are both very critical in ensuring predictable, abrasive wear with the CVD diamond-coated tool. POLYCARBONATE MACHINING. In a polycarbonate milling application, a small job shop replaced a tungsten carbide rotary tool with a 12.7 mm diameter CVD diamond-coated end mill (coating thickness of 10–15 ␮m). Using an outdated milling machine, the user was able to machine over 3000 polycarbonate parts using the CVD diamond-coated end mill. By com-

Table 6

Spindle Power Consumption During GFRP Composite End Milling Spindle power consumption (watts) End of testing

Start of testing Machining direction → Uncoated WC-Co CVD diamond-coated WC-Co

Climb cutting

Conventional cutting

Climb cutting

Conventional cutting

80 75

80 80

120 80

135 80

Source: Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc. Internal test data provided courtesy of Walter T. Swann.

CVD Diamond Solutions for Machining

497

parison, an uncoated tungsten carbide end mill typically produced 30–35 parts per tool. The machine-limited cutting conditions were 3200 rpm with a feed rate up to 280 mm/min. The 100 times lifetime advantage of the CVD diamond tool relative to WC-Co is attributed to the corrosion resistance of the pure diamond composition relative to the cobalt-sintered tungsten carbide (discussed in Sec. II.A.8). If the binder phase of WC-Co (or PCD) is eroded, the wear mechanism is likely to shift from a slow, abrasive wear of the hard particle phase (WC or diamond, respectively) to a rapid, corrosioninduced disintegration (grain pullout) of the particulate composite at the cutting edge. MMC DRILLING. In magnesium-matrix composite drilling tests [82], CVD diamond coatings were shown to offer superior performance relative to uncoated tungsten carbide and TiAlN-coated tungsten carbide. TiAlN and CVD diamond coatings were being tested in an effort to find a lower total cost solution than PCD. The workpiece material compositions included particle-reinforced Mg (12 vol% SiC), fiber-reinforced Mg (20 vol% Al2O3), and hybrid-reinforced Mg (15 vol% SiC particles ⫹ 5 vol% Al2O3 fibers). The performance advantage of the CVD diamond-coated drills was most dramatic in the SiC-containing composites, where excessive wear was observed in TiAlN-coated and uncoated carbide. Although a dramatically improved performance was observed for the TiAlN coating in the Al2O3 fiberreinforced MMC, CVD diamond performed well in all composites tested. The satisfactory performance of TiAlN in the Al2O3 fiber-reinforced MMC, but not the SiC-containing composites, demonstrates the criticality of selecting the appropriate cutting tool material for each specific application. The superior performance of CVD diamond was attributed to several factors including diamond’s higher hardness relative to both SiC and Al2O3 as well as the higher level of adhesion of the diamond coatings relative to TiAlN coatings. Diamond’s unmatched hardness reduces the sensitivity to changes in the hardness of the MMC ceramic reinforcement phase. EVOLVING PROMISE FOR FERROUS MACHINING. Although many researchers believe that diamond tools are unable to machine ferrous materials because of the chemical incompatibility discussed in Sec. II.A.8, Kanda et al. [17] reported progress in the use of CVD diamond-coated rotary tools in machining ferrous materials. In tests using annealed gray cast iron (FC250; JIS G 5501), a diamond-coated, two-flute, ball-nose end mill (4 mm diameter, 8 mm flute length, and 80 mm total length) was tested. Tools were operated in both an up-cut (climb) and down-cut (conventional) machining direction using compressed air as a coolant under the following cutting conditions: 24,000 rpm, 2 m/min feed rate, with an axial depth of cut of 0.5 mm. Under these conditions, the wear of the CVD diamond-coated tool was less than 10% of that of an uncoated tool. In another test [17], the same

498

Cline and Olson

researchers reported the relationship between the wear rate of the CVD diamond coating and the cutting conditions in carbon steel (S50C; JIS G 4051) with a hardness of 200 HB. Using a ball-nose end mill (2 mm diameter, 4 mm flute length, and 60 mm total length), the carbon steel was machined for a total cutting length of 4.0 m using various operating conditions: 8000 rpm and 16,000 rpm with an axial depth of cut of 1.5 mm and a radial infeed of 0.1 mm. Tests were performed using both compressed air and water-soluble emulsion. The work of Kanda et al. [17] indicates that the selection of a proper tool design and operating conditions can suppress the chemical corrosion of the diamond cutting edge, thus enabling superior performance over traditional tooling materials.

V.

OTHER MECHANICAL APPLICATIONS OF CVD DIAMOND

Both the unique characteristics of CVD diamond and the flexibility of CVD diamond processes allow the use of CVD diamond in mechanical applications beyond cutting tools. For decades, natural and synthetic (HPHT) single-crystal diamond has been used not only for specialty cutting applications as noted in Sec. I.C.1 but also for unique wear applications such as wire dies, engraving tools, and dressing tools for grinding wheels. The advent of thin-film and thick-film CVD diamond offers an alternative type of diamond for these applications, and the ability to coat or laser-cut large complex shapes of CVD diamond material opens up opportunities for the use of diamond in new applications where diamond materials were previously unavailable or cost prohibitive. This section is intended to provide examples of unique wear applications in which diamond coatings and freestanding diamond offer unique cost-performance advantages. Because the specific requirements for wear part applications vary widely, this section is organized in unique wear part case studies including both thin-film and thick-film product concepts. The selected case studies include diamond-coated mechanical seals (Sec. V.A), freestanding CVD diamond dressing tools (Sec. V.B), and freestanding CVD diamond wire dies (Sec. V.C). A.

CVD Diamond-Coated Mechanical Seals

Cemented tungsten carbide materials such as WC-Ni are commonly used in a wide range of mechanical seal applications. Advanced ceramic materials such as aluminum oxide, silicon carbide, and silicon nitride are used in seal applications that can afford the trade-off in fracture toughness and strength relative to cemented carbide for the benefits offered by the unique properties

CVD Diamond Solutions for Machining

499

of ceramic materials. Over the last decade, there have been ongoing efforts to develop CVD diamond coating on cemented tungsten carbide as well as nonoxide ceramic seals for selected applications. Because many of these efforts have been commercial in nature and performed under confidential terms within the industry, very little performance data has been released into the CVD diamond industry. Cline [25] has reported a friction study of diamond-coated mechanical seals that were engineered for a commercial aqueous pumping application. Although ceramic seal faces are often mated to a softer, lubricious, hardcarbon face in many applications, the study was focused on comparing the benefits of diamond-on-diamond versus SiC-on-SiC seal face combinations. This study was focused on unfinished, as-deposited, diamond-coated seals produced using four different CVD process conditions. The sintered ␣-SiC material was a commercially available grade commonly used in the seal industry.* All seal pairs were tested in a friction tester using light seal face pressure, a mean surface speed of 306 m/min, and a starting temperature of 25⬚C. Both lubricated (deionized water) and unlubricated (dry) operating conditions were investigated. It is interesting to note that the surface speed used is nearly the same as that used in the presintered ceramic machining test summarized in Sec. IV.B.2.c. The results of the friction study are summarized in Table 7. Because of the proprietary nature of the results, the friction data were averaged for the four unique, but similar, diamond coating types. The table shows the diamond-on-diamond seal face combination was clearly superior to the SiCon-SiC. The diamond coatings decreased the mean steady-state friction coefficient during both lubricated and dry operation. The friction coefficient measurements of the diamond-coated faces indicate that CVD diamond coating can be tailored to produce friction coefficient values comparable to those of well-known lubricious materials such as fluoropolymers.† Because of the added lubricity of the diamond-coated faces, the temperature rise observed during the worst-case scenario of dry operation was reduced by over 100⬚C to a temperature range that minimizes the risk of damage to other components within the aqueous pump. Abusive rig testing conducted after the friction testing resulted in minimal diamond erosion compared with that observed for the silicon carbide face combination. Extensive application testing over a period of years indicated that the unique combination of properties

*Carborundum’s Hexoloy SA sintered ␣-SiC material. † According to E. I. du Pont de Nemours and Company, the coefficient of friction of Teflon is generally in the range of 0.05 to 0.20, depending on the load, sliding speed, and particular Teflon coating used.

Table 7

Dynamic Friction Test Results of CVD Diamond-Coated Mechanical Seals

Seal face lubrication

Mean steady-state friction coefficient (ring-on-ring)

SiC-on-SiC assembly

Lubricated, deionized water

0.08–0.15

SiC-on-SiC assembly

Unlubricated, dry

0.20–0.30

Diamond-on-diamond assembly Diamond-on-diamond assembly

Lubricated, deionized water

0.03–0.04

Unlubricated, dry

0.05–0.13

Seal face material combination

Seal face temperature rise 6–8⬚C above bulk fluid temperature after 58 min of operation >350⬚C above ambient temperature after 5 min of dry operation 1⬚C above bulk fluid temperature after 58 min of operation 200–250⬚C above ambient temperature 14 min of dry operation

Source: Test data provided by a customer of Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc.

CVD Diamond Solutions for Machining

501

afforded by the CVD diamond coatings can eliminate several failure modes common to the silicon carbide seal faces commonly used in the application. B.

CVD Diamond Dressing Tools

Dressing tools are generally used to remove dull ceramic or superabrasive grains from grinding wheel surfaces in an effort to suppress grinding temperatures, improve workpiece finish, and improve workpiece tolerances. They are also used for ‘‘truing’’ or shaping grinding wheels. Historically, most dressing and truing has been performed with stationary dressing products that are fed into a rotating wheel. These products are often made using single-crystal or PCD diamond. Today, many high-productivity grinding operations employ rotary CNC dressers, or shaped dressing rolls, which can be used to dress one side of a grinding wheel continuously as the wheel is simultaneously grinding a workpiece on its opposite side. CNC rotary dressers may contain over 100 similar pieces of diamond that are hand-set over the outer diameter of the dresser and can be engineered to create unique forms in a grinding wheel. Examples of stationary and rotary diamond dressing tool designs are shown in Fig. 28. CVD diamond in a form similar to that shown in Fig. 29 offers several unique advantages to manufacturers of dressing tools as well as dressing tool end users. CVD diamond has been shown to offer performance enhancement for stationary dressing tools as well as rotary dressing tools in-

Figure 28 Examples of rotary (left) and stationary (right) diamond dressing tools used for shaping and truing grinding wheels. (From Ref. 1.)

502

Cline and Olson

Figure 29 SEM images of a laser-cut piece of freestanding diamond typical of that used in a dressing tool. The 100⫻ image in (a) depicts an oblique view of the coarse-grained diamond facets common to as-deposited, freestanding CVD diamond. The 122⫻ image in (b) shows the relatively smooth nucleation face of the diamond (top center). The striations seen at the lower left of both (a) and (b) are the result of the laser-cutting process. The fractured cross section seen at the right end of (a) and front, center of (b) depicts the dense, polycrystalline structure of the diamond deposit. Note that the fractured cross section does not have the weak grain boundary characteristics (intergranular fracture) previously shown in Fig. 11a. (From Norton Diamond Film/Saint Gobain Industrial Ceramics, Inc. Images courtesy of Walter Swann; part (a) from Ref. 1.)

CVD Diamond Solutions for Machining

503

cluding CNC profiling dressers and plunge form dressers. The commercial availability of synthetic, thick-film CVD diamond of good mechanical integrity creates a virtually unlimited supply of uniform laser-cut pieces that can offer superior performance relative to single-crystal diamond. Otherwise cost-prohibitive or unavailable, high-aspect-ratio pieces of diamond (such as 8 mm ⫻ 1 mm ⫻ 1 mm thick) can be cost-effectively incorporated into CVD diamond dressing tool designs allowing longer life and better dressing efficiency. Because of the precise shape of the laser-cut CVD diamond, the diamond can be strategically oriented to achieve 100% diamond consumption during the lifetime of the dresser. The availability of large-area CVD diamond helps to minimize or even eliminate design challenges associated with the crystal habit constraints and shape variability inherent in natural diamond. This enables small, intricate geometric forms such as sharp corners to be designed into dressing tools. The bulk structure of the CVD diamond can be engineered to enhance fracture toughness and strength for purposes of optimizing dressing tool life and dimensional stability beyond other forms of diamond (see Sec. III.A). This allows plunge form dressers to be produced with very small convex radii (e.g., 0.05 mm). C.

CVD Diamond Wire Dies

The production of metal wires is normally achieved by drawing metals such as copper, tungsten, or stainless steel through precision-made dies composed of hard materials such as diamond or tungsten carbide. Single-crystal diamond wire dies are much more wear resistant than tungsten carbide dies, but they have relatively poor chip resistance and are extremely difficult to fabricate. Furthermore, the intrinsic anisotrophy of a single crystal can result in nonuniform wear or cleavage-induced catastrophic failure under the extreme pressures of the wire drawing process. According to Wilks and Wilks [83], ‘‘the best choice of [crystallographic] direction is not too obvious because as the wire passes through the die its circumference is abrading the diamond on a whole 360⬚ range of planes, and the rates of wear on these planes will be somewhat different. Hence, the originally circular hole will not only grow larger but will loose its shape. However, 具110典 directions offer the advantage that the wire is abrading the sides of the hole with {001} and {011} orientations in abrasion resistant directions.’’ Over the years, efforts have been made to overcome the anisotropic trade-offs of single-crystal diamond wire dies by using polycrystalline diamond compacts [84,85], but these approaches were limited by the two-phase construction of a diamond composite and a secondary material of inferior mechanical properties. With the advent of CVD diamond processes, the ability to produce pure, dense layers of diamond created new

504

Cline and Olson

alternatives for the wire drawing industry. Patents by Anthony et al. [86– 89] report the advantages of CVD diamond that has been microstructurally engineered for enhanced performance in wire drawing. Included in the claims of these patents is a 具110典 preferred, columnar orientation of the microstructure within the freestanding CVD diamond die. The polycrystalline diamond component of the wire die is oriented in a manner crystallographically similar to that described by Wilks and Wilks [83] for singlecrystal wire dies where the {001} and {011} orientations are present in abrasion-resistant directions. Although specific performance advantages are not presented here, the potential utility of CVD diamond synthesized in an oriented manner is quite clear in this application. D.

Other Wear Part Opportunities

Although cutting tool performance information is frequently reported in trade journals, trade magazines, and marketing literature, it is not feasible to predict the true state of the technology for the full range of mechanical applications where CVD diamond offers unique solutions. This is due mostly to the nature of commercial competition in the market. The introduction of commercial products for many wear part applications is likely to accelerate as CVD diamond production costs are reduced and the industry matures. Many wear part applications such as those described here require attention to detail specific to the application. Although much of the activity in the CVD diamond marketplace presently appears to be focused on larger, less fragmented markets, it is likely that the use of CVD diamond wear products will evolve and or be exposed in time. This trend will be driven by a necessary link between (1) a more thorough understanding of the structureproperty-performance relationships of CVD diamond design concepts and (2) commercial processes capable of reproducibly fine-tuning the necessary characteristics of the specific application.

VI.

SUMMARY

For decades, diamond has played a key role in industrial productivity. The 1950s saw the introduction of synthetic diamond grit. In the 1970s, metalsintered polycrystalline diamond (PCD) gained a foothold, offering unique flat shapes previously not available. Today, the evolution of diamond for mechanical applications continues with the emergence of CVD diamond as a versatile industrial material for cutting tools and other wear components. In machining applications, CVD diamond’s superlative combination of properties offers not only the promise of longer tool life under standard operating

CVD Diamond Solutions for Machining

505

conditions but also opportunities for cost-effective high-speed and reducedcoolant operation.

A.

Thick-Film CVD Diamond

The inherent purity of CVD diamond deposition processes and industrial advances in freestanding CVD diamond manufacturing provide outstanding machining solutions for many difficult-to-machine materials. In most applications, freestanding, thick-film, CVD diamond tips are used as an alternative to cobalt-sintered PCD or single-crystal diamond. The ability to ‘‘tailor’’ the fracture toughness, strength, and other critical properties of freestanding CVD diamond is paramount to the ongoing development and commercialization of high-performance diamond cutting tool and wear products. Cutting tool application results indicate that the bulk mechanical properties of thickfilm CVD diamond can meet the needs of applications ranging from sintered tungsten carbide turning to high-silicon aluminum milling. The more recently discovered utility of thick-film CVD diamond for enhancing rotary and stationary dressing tool performance is a clear indication that the mechanical integrity of commercial grades of CVD diamond is suitable for very aggressive operations. The application range for thick-film diamond will continue to expand as diamond cutting and dressing tool fabricators gain a greater practical understanding of the unique properties of CVD diamond relative to PCD, single-crystal diamond, and other superhard materials.

B.

Thin-Film CVD Diamond

As practical methods continue to be developed for harnessing CVD diamond’s outstanding properties in diamond-coated products, the application range of thin-film CVD diamond continues to expand. Now that CVD diamond reactor development is reaching a level that supports commercialization and mass production, the primary technical challenge in the manufacture of diamond-coated carbide cutting tools and wear parts is ensuring suitable, reproducible diamond adhesion. A broadening portfolio of industrial adhesion technologies has led to the expanded use of diamond-coated cemented carbide tools in a wide range of metals and especially nonmetals. Depending on the vendor, the carbide grades used for diamond coating may be stringently controlled to ensure coating quality, adhesion, and long cutting life. As adhesion treatments become more robust, the ‘‘hidden promise’’ of CVD diamond-coated cemented carbide will finally be universally realized in the marketplace.

506

Cline and Olson

ACKNOWLEDGMENTS Both authors are previous employees of the Norton Diamond Film business unit of Saint Gobain Industrial Ceramics, Inc. in Northboro, MA. The authors would like to acknowledge those who directly and indirectly contributed to this chapter including Saint Gobain coworkers, past customers of Norton Diamond Film, as well as academic mentors at Penn State University, Worcester Polytechnic Institute, and Bates College. Erica Olson and Heather Cline are acknowledged for their patience and support of the respective authors during their academic and industrial endeavors in the CVD diamond field. Portions of this work were funded by a NIST/DOC ATP program entitled Accelerated Commercialization of Diamond-Coated Round Tools and Wear Parts, an R&D joint venture between Kennametal, Inc. and Norton Diamond Film. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

B Cline. Cutting Tool Eng 52:76, 2000. BL Cline. Metalwork Equipment News March:68, 1999. RB Aronson. Manuf Eng January:52, 1999. KJA Brookes. World Directory and Handbook of Hardmetals and Hard Materials. 5th ed. Hertfordshire, UK: International Carbide Data, 1992. PL Smith. Am Mach December:74, 1999. J Schneider. Manuf Eng January:66, 1999. PM Stephan. Ceram Bull 71:1623, 1992. RA Hay. The new diamond technology and its application in cutting tools. In: ED Whitney, ed. Ceramic Cutting Tools. Park Ridge, NJ: Noyes, 1994, p 305. DB Arnold, FJ Momper. Manuf Eng November:62, 1997. DT Quinto. Int J Refractory Metals Hard Mater 14:7, 1996. WR Pfouts. Manuf Eng 125:98, 2000. KE Spear. J Am Ceram Soc 72:171, 1989. WA Yarbrough, R Messier. Science 247:688, 1990. JP Dismukes, KE Spear, eds. Synthetic Diamond: Emerging CVD Science and Technology. New York: John Wiley & Sons, 1994. BL Cline. Metalworking Equipment News January:24, 1999. K Kanda, S Takehana, S Yoshida, R Watanabe, S Takano, H Ando, F Shimakura. Surf Coat Technol 73:115, 1995. SY Kazutaka Kanda, K Yoneshima, H Ando. Diamond Films Technol 8:93, 1998. D Myers. Moldmaking Technol October:35, 1999. D Myers. Moldmaking Technol November:42, 1999. D Myers. Moldmaking Technol December:33, 1999. RS Sussmann, JR Brandon, SE Coe, CSJ Pickles, CG Sweeney, A Wasenczuk, CJH Wort, CN Dodge. Finer Points 10:7, 1999.

CVD Diamond Solutions for Machining 22. 23. 24. 25. 26.

27. 28. 29.

30.

31. 32.

33. 34. 35. 36. 37. 38. 39. 40. 41.

42.

43. 44. 45. 46.

507

A Johnson. Cutting Tool Eng February:34, 1996. EJ Oles, A Inspektor, CE Bauer. Diamond Relat Mater 5:617, 1996. MN Yoder. In: RF Davis, ed. Diamond Films and Coatings: Development, Properties, and Applications. Park Ridge, NJ: Noyes Publications, 1993, p 1. BL Cline. Mechanical applications for CVD diamond. Presented at Diamond 99 Conference, Prague, Czech Republic, 1999. HA Hoff, AA Morrish, WA Carrington, JE Butler, BB Rath. In: JT Glass, R Messier, N Fujimori, eds. Diamond, Silicon Carbide and Related Wide Bandgap Semiconductors. Mater Res Soc Symp Proc 162:279–284, 1990. HA Hoff, AA Morrish, JE Butler, BB Rath. J Mater Res 5:2572, 1990. JM Olson, MJ Dawes. J Mater Res 11:1765, 1996. N Johnson. CVD diamond properties and products for material removal, wear and thermal management applications. Proceedings for Diamond Films 1990, Crans Montana, Switzerland, 1990. J Wilks, EM Wilks. Abrasion and wear of diamond. In: JE Field, ed. The Properties of Diamond. 2nd ed. London: Harcourt Brace Jovanovich, 1990, p 351. BL Cline. Ceram Ind 149:51, 1999. BV Tanikella, W Swann, B Cline. Machining of composite materials and bisque ceramics using CVD diamond coated WC-Co rotary tools. Presented at 100th Annual Meeting of the American Ceramic Society as part of a Symposium on Innovative Processing and Synthesis of Ceramics, Glasses and Composites, Poster Session SXIP-032-98, 1998. R Komanduri. Mach Sci Technol 1:113, 1997. MM Schwartz. SAMPE J 32:18, 1996. KJ Gray, H Windischmann. Diamond Relat Mater 8:903, 1999. D Shechtman, JL Hutchinson, LH Robins, EN Farabaugh, A Feldman. J Mater Res 8:473, 1993. BE Williams, HS Kong, JT Glass. J Mater Res 5:801, 1990. VS Vavilov, EA Konorova, EB Stepanova, EM Trukhan. Sov Phys Semicond 13:635, 1979. JF Prins. Appl Phys Lett 41:959, 1982. VI Nepsha. In: MA Prelas, G Popovici, LK Bigelow, eds. The Handbook of Industrial Diamond and Diamond Films. New York: Marcel Dekker, 1998. AM Zaitsev. In: MA Prelas, G Popovici, LK Bigelow, eds. The Handbook of Industrial Diamond and Diamond Films. New York: Marcel Dekker, 1998, pp 227–376. AG Gontar. In: MA Prelas, G Popovici, LK Bigelow, eds. The Handbook of Industrial Diamond and Diamond Films. New York: Marcel Dekker, 1998, pp 377–412. DS Knight, WB White. J Mater Res 4:385, 1989. PW May. Endeavor Mag 19:101, 1995, accessed via http://www.chm.bris.ac. uk/pt/diamond/end.htm. P Bou, L Vandenbulcke. J Electrochem Soc 138:2991, 1991. RJ Nemanich, SA Solin. Solid State Commun 23:417, 1977.

508 47.

48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75.

76. 77.

Cline and Olson W Yarbrough, R Roy. In: GG Johnson, AR Badzian, MW Geis, eds. Diamond and Diamond-Like Materials Synthesis. Extended Abstracts No 15. Pittsburgh: Materials Research Society, 1988. RE Shroder, RJ Nemanich, JT Glass. Phys Rev B 41:3738, 1990. AR Badzian, T Badzian, D Pickrell. In: Diamond Optics. SPIE 969. 1988, p 14. D Tabor. In: JE Field, ed. The Properties of Diamond. 2nd ed. London: Harcourt Brace Jovanovich, 1990, p 325. B Lawn. Fracture of Brittle Solids. 2nd ed. Cambridge: Cambridge University Press, 1993. DB Marshal, AG Evans. J Appl Phys 56:2632, 1984. AG Evans, JW Hutchinson. Int J Solids Struct 20:455, 1984. AG Evans, BJ Dalgleish, M He, JW Hutchinson. Acta Metall 37:3249, 1989. JJ Dubray, CG Pantano, M Meloncelli, E Bertran. J Vac Sci Technol A 9:3012, 1991. J Robertson. Diamond Relat Mater 4:549, 1995. H Liu, DS Dandy. Diamond Relat Mater 4:1173, 1995. Y Chakk, R Brener, A Hoffman. Appl Phys Lett 66:2819, 1995. R Polini, D Gazzoli, E Molinari, V Sessa, ML Terranova, P Ascarelli, S Fontana. Diamond Relat Mater 1:205, 1992. JW Kim, YJ Baik, KY Eun. Diamond Relat Mater 1:200, 1992. KW Allen. J Adhes 21:261, 1987. PO Joffreau, R Haubner, B Lux. R&HM 1988, p 186. R Haubner, A Lindlbauer, B Lux. Diamond Relat Mater 2:1505, 1993. M Murakawa, S Takeuchi, H Miyazawa, Y Hirose. Surf Coat Technol 36:303, 1988. M Murakawa, S Takeuchi, Y Hirose. Surf Coat Technol 39/40:235, 1989. BS Park, YJ Baik, KR Lee, KY Eun, DH Kim. Diamond Relat Mater 2:910, 1993. S Kubelka, R Haubner, B Lux, R Steiner, G Stingeder, M Grasserbauer. Diamond Films Technol 5:105, 1995. JE Field. The Properties of Diamond. 2nd ed. London: Harcourt Brace Jovanovich, 1990, p 281. N Kikuchi, T Komatsu, H Yamasita, H Yoshimura. US Patent 4,731,296, 1988. JM Pinneo. European Patent Application 93308250.5, 1993. ER Kupp, PhD thesis, Pennsylvania State University, 1995. DS Park, PJ Diconza. European Patent Application 93307379.3, 1993. K Saijo, M Yagi, K Shibuki, T Sadahiro. US Patent 5,100,703, 1992. JM Olson, H Windischmann. Diamond Films Technol 8:105, 1998. JP Whitacre. Hypereutectic aluminum: a piston material for modern high specific output gasoline engines. SAE Trans Vol 96, Sec 3, Passenger Car Meeting and Exposition on October 19, 1987 in Dearborn, MI, Society of Automotive Engineers, Warrendale, PA, 1987. F Feuchtenschlager. Austrian Research Centers, data presented at commercial seminar in Wurzburg, Germany, 1998. J Sebzda Sr. EDM Today 2:24, 1999.

CVD Diamond Solutions for Machining 78. 79. 80. 81. 82.

83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97.

509

R Horsfall. Cutting Tool Eng August:42, 1997. DT Quinto, AT Santhanam, PC Jindal. Mater Sci Eng A105/106:443, 1988. DT Quinto. J Vac Sci Technol A 6:2149, 1988. Diamond coating cutting. Tool Prod 65:20, 1998. K Weinert, D Biermann, M Liedschulte. Drilling of magnesium-matrix composites. In: Composites in Manufacturing. Newsletter of the Composites Manufacturing Association of the Society of Manufacturing Engineers 15:1, 1999. J Wilks, E Wilks. Properties and Application of Diamond. Oxford: ButterworthHeinemann, 1991, pp 505–507. LC Carrison, RG Wetzel. US Patent 4,016,736, 1977. RH Wentorf Jr, WA Rocco. US Patent 3,831,428, 1974. TR Anthony, BE Williams. US Patent 5,377,522, 1995. TR Anthony, BE Williams. US Patent 5,363,687, 1994. TR Anthony, BE Williams. US Patent 5,361,621, 1994. TR Anthony, BE Williams, KM McNamara. US Patent 5,465,603, 1995. Introduction to CVD Diamond. Internet information from the DeBeers Industrial Diamond website (http://www.debid.co.uk/), September 2000. Based on information provided in April 1999 product announcements for DeBeers Industrial Diamond Diafilm PC, RF, and TM products. Properties of Diamond. DeBeers Industrial Diamond Division, Technical brochure. KJA Brookes. World Directory and Handbook of Hardmetals and Hard Materials. 5th ed. Hertfordshire, UK: International Carbide Data, 1992, p 81. S Hampshire. In: SJ Schneider Jr, ed. Engineered Materials Handbook. Vol 4. Materials Park, OH: ASM International, 1991, p 812. GA Graves, NL Hecht. Effects of environment on the mechanical behavior of high-performance ceramics. Report UDR-TR-94-136, 1995. EH Hull, GT Malloy. J Eng Ind (Trans ASME) November:373, 1966. Silicon nitride compressive strength data from Enceratec, Inc via http://www. enceratec.com, September 2000.

11 Diamond Heat Spreaders and Thermal Management Ajay P. Malshe and W. D. Brown University of Arkansas, Fayetteville, Arkansas

I.

INTRODUCTION

Almost from the instant that active electronic devices were invented, a drive to increase the operating speed of electronic systems containing them was initiated. At the same time, a parallel effort focused on packing more electronics into a smaller and smaller volume. These trends continue unabated today. However, it is becoming more and more difficult to make progress in these areas using conventional single-chip packaging and present-day thermal management technologies. Multichip module (MCM) packaging, in which integrated circuits (ICs) are mounted extremely close together on a single substrate in order to increase the operating speed of an electronic system, only contributes to the thermal management challenge. As a consequence of these continuing trends in electronic systems, there are three main difficulties that must be addressed: how to provide for (1) the required power at the proper locations in the system, (2) sufficient input/output (I/O) connections to and from the system, and (3) adequate heat removal from the system. A.

Need for Thermal Management in Electronic Packaging

In IC technology, increasing functionality and operating speed of a chip are accomplished by reducing the minimum feature size, thereby shrinking device size and interconnection (i.e., wire) width. The reduced device dimen511

512

Malshe and Brown

sions dictate a reduction in the operating voltage(s) in order that material stress levels not be exceeded, thereby preserving the reliability of the IC. A reduction in interconnect width is accompanied by an increase in interconnect resistance, which then must be alleviated by reducing the interconnect length. On an IC, reduced interconnect length is a direct consequence of reducing device size. The final result of these changes is that an IC contains more active devices in a given area operating at higher speeds. This, in turn, means that the power consumption/dissipation in the given area also increases as illustrated in Fig. 1. Thus, a higher functionality IC with its accompanying higher power dissipation presents a thermal management challenge. As the drive to pack more and more of these ICs into smaller and smaller volumes without adversely affecting the operating speed continues, thermal management of electronic systems will present formidable challenges. The multichip module, noted previously, is a technology presently being used to pack ICs into a smaller volume by mounting them on a single substrate. The substrate provides both mechanical support for and interconnections between the ICs. By using array I/O on the MCM substrate, both the large I/O requirement and power insertion problems are relieved. However, thermal management becomes an even bigger problem because, as the power dissipated in a given volume increases, conventional approaches to dealing with the resulting heat quickly become inadequate. For example, the size of conventional heat sinks would have to increase to handle higher

Figure 1

Trends in clock speed and power consumption. (From Ref. 95.)

Diamond Heat Spreaders and Thermal Management

513

power chips. This, in turn, would lead to a requirement for higher cooling air velocity and flow. However, in order to maintain or even reduce the size of a given electronic system, larger conventional heat sinks are not acceptable because of the increased volume and weight that they present. On the other hand, if the excess heat is not dissipated in an effective and timely fashion, there is the danger of degrading the system, which, at the very least, would adversely affect the long-term reliability of the system or, at worst, would prove fatal. Because of the increasing demands on packaging of electronic systems and the accompanying thermal management challenges, in recent years much attention has been directed at the development, processing, and characterization of new electronic materials and thermal management technology. Thus, the requirements and role of heat spreaders (also referred to as thermal spreaders) and heat sinks and the various materials and schemes used for thermal management of electronic packages are topics of a significant amount of research and development. B.

Thermal Heat Spreaders and Heat Sinks

A thermal heat spreader material serves as an interface between a heatdissipating device, such as an IC, and a heat sink. The primary function of the heat spreader is to collect heat from the source (e.g., IC) and quickly spread it to a larger device (e.g., heat sink) in order to eliminate heat from an electronic system. Thus, in terms of the roles they play in removing heat from an electronic system, the function of a heat spreader is different from that of a heat sink. In other words, the role of the heat spreader is to extract the heat rapidly from the source and spread it over a large area; it is the heat sink that actually dissipates the heat. Because of their different roles, a heat spreader needs to have a high thermal conductivity and excellent electrical insulating properties and should be capable of withstanding the high temperatures generated by the heat source device. To ensure intimate contact between the heat source and the heat spreader, the surface of the heat spreader in contact with the heat source should be extremely flat, of the order of or exceeding 0.1 ␮m/mm, and the material should be stress free. Stress generally causes surface curvature, which is undesirable. Because of the requirement for an extremely flat mating surface, inexpensive processes for surface finishing, such as polishing or planarization, should be available and procedures for dicing or cutting should be simple and clean. Also, the definition of complex metallization patterns using photolithography should be manufacturing transparent. Apart from these features, the cost of the heat spreader must be sufficiently low as to make it economically viable.

514

Malshe and Brown

The physical dimensions of a heat spreader required for a specific application can be estimated by modeling. Usually, the thermal properties of the material used for the heat spreader, the configuration of the system generating the heat, and the amount of heat generated by the system are the input parameters to such a simulation program. The results from one such program involving the estimation of isotherms in half-cylindrical heat spreaders of different thicknesses are shown in Fig. 2 [1]. Single-crystal, natural diamonds have excellent properties for heat spreader applications. They have a very high thermal conductivity of over 20 W/cm K and electrical resistivities in excess of 1013 ⍀-cm. However, natural diamonds have a number of limitations that restrict their use in electronic applications. They are scarce, making them expensive, and from a more technical point of view, their composition is likely to vary from sample to sample. Furthermore, their physical dimensions severely limit the area to which they can be applied and, hence, their useful applications. To overcome these limitations, diamond substrates can be synthesized by chemical vapor deposition (CVD). Although CVD diamond deposition rates are reasonably high, the costs associated with its growth and postdeposition processing are still too high to allow it to be used routinely as a thermal management substrate in either MCMs or single-chip packages. However, given that diamond will probably be a viable MCM substrate in the future, a primary technical consideration is the minimum diamond substrate thickness required for efficient thermal management. As seen in Fig. 3, the efficiency of diamond in spreading heat decreases with increasing thickness for an edgecooled package geometry. This fact has important technical as well as economic implications when determining the optimum thickness of a substrate. Currently, approximately 1000-␮m-thick substrates are considered acceptable for MCM applications. However, according to some simulation results [2], thinner diamond substrates (20 to 30% thinner) can be used to achieve similar thermal management performance. This finding suggests that the CVD growth time and correspondingly the cost of diamond can be significantly reduced. Thus, CVD diamond, by far the technically best material for heat spreader applications, has the potential to become a viable contender economically. C.

Materials and Schemes for Thermal Management of Electronic Packages

Of the several conventional MCM substrates used for thermal management, silver (thermal conductivity, k = 4.28 W/cm K), copper (k = 3.96 W/cm K), beryllium oxide (k = 2.23 W/cm K), and aluminum nitride (k = 2.30 W/cm K) are the only materials having thermal conductivities close to that of

Diamond Heat Spreaders and Thermal Management

515

Figure 2 Isotherms in half-cylindrical heat spreaders of different thicknesses. (From Ref. 1.)

516

Figure 3

Malshe and Brown

Thermal management efficiencies of some materials. (From Ref. 2.)

diamond. Figure 3 compares the thermal management efficiencies of these materials and CVD diamond, with k values as high as 16.0 W/cm K, although it can vary over a wide range of values depending on the growth conditions and it is anisotropic with the higher value in the direction parallel to the growth direction. It shows that even under extreme power dissipation conditions, the temperature rise at the center of a diamond substrate of optimum thickness in an edge-cooled geometry is much less than that of conventional materials. Moreover, the temperature rise for diamond is well within the acceptable limits for all practical applications (see Fig. 4). Many different schemes for thermal management of electronic packages are possible, depending on the particular application requirements. These schemes can be broadly classified into two categories: cooling of the package from the back and cooling from the edges. Although diamond has the highest thermal conductivity of any material, it has a relatively low heat capacity. Therefore, to use diamond to its best advantage, it should be used as a heat spreader by incorporating it as an intermediate layer between a heat source and a heat sink. Thus, as noted previously, the primary function of the diamond would be to spread the heat quickly from a smaller area (the heat source) to a larger area (the heat sink) where the heat can be disposed of more effectively and efficiently (see Fig. 5). In an application as a heatspreading substrate, the heat source (e.g., an IC) would be mounted on one

Diamond Heat Spreaders and Thermal Management

517

Figure 4 Temperature distribution of plot of diamond substrate—simulation results. (From Ref. 2.)

Figure 5 Ref. 4.)

Placement of diamond film between the source and sink of heat. (From

518

Malshe and Brown

flat surface of the diamond with the heat sink mounted on the other side. In this configuration, the CVD diamond would quickly pull the heat from the IC, spread it over a large area (the diamond heat spreader), and transfer it to the heat sink, which is more efficient than diamond at dissipating heat. The fact that diamond has an excellent thermal conductivity but relatively low heat capacity reinforces the need to determine the optimum thickness for a particular application, thereby minimizing the economic impact of using a diamond heat spreader. Maximum performance benefits of diamond heat spreaders occur when the lowest junction operating temperature of the device generating the heat is achieved for a given configuration. On the other hand, minimum cost is realized by making the diamond substrate as small as possible because the cost of CVD diamond correlates directly with its volume. It has been observed [3] that the optimum diamond heat spreader should have a thickness that is 0.5 to 1.0 times the radius of the heat source and a radius three times that of the heat source (see Fig. 5). It has also been found that [3] when a CVD diamond–coated substrate is used as a heat spreader, the coating is effective only when the heat source radius is comparable to the diamond thickness. Unfortunately, the layer of solder used to attach the heat source (e.g., IC) to the diamond coating can present a major thermal impedance, thereby reducing the effectiveness of the diamond in reducing the temperature of IC. Another feasible approach to packaging electronic devices, such as solid-state lasers, is by stacking them vertically to effect a two-dimensional array instead of the conventional approach where devices are spread out in the horizontal direction over a single heat spreader creating a simple onedimensional array. The only efficient method for removing heat from the two-dimensional configuration is horizontally through heat spreaders placed between the devices. The thermal management problem is then one of heat flow in one dimension along a horizontal bar. Similarly, three-dimensional MCMs, which offer the greatest potential for speed performance enhancement of electronic systems, can be realized by taking advantage of the excellent thermal properties of CVD diamond. However, they also present the greatest challenges to the package designer and substrate fabricator. D.

Diamond Heat Spreaders

Effective, economical thermal management is essential for satisfactory operation of both single-chip and multichip electronic packages. Diamond has an excellent thermal conductivity (k = 20.00 W/cm K for natural diamond and in the range of 7.00–16.00 W/cm K for thermal management quality CVD diamond), a low coefficient of thermal expansion (⬇10⫺6), high elec-

Diamond Heat Spreaders and Thermal Management

519

trical resistivity (>1013 ⍀-cm), and high mechanical strength (Young’s modulus ⬇ 1011 N m⫺2), making it an ideal choice for two-dimensional as well as three-dimensional MCM packaging. Thus, diamond is clearly a qualitatively superior material, compared with its competitor materials such as silicon, alumina, and beryllium oxide, for thermal management applications. In fact, it has been shown to perform well as a substrate for specific devices, such as high-power semiconductor laser diodes, and has been used as the substrate in prototype MCMs. The only obstacles to more widespread use of diamond substrates in such applications have been its limited availability in large sizes, which are stress free and of uniform thickness, and cost. However, advances in CVD diamond deposition technology make it possible to fabricate fairly large area (10 cm2) diamond substrates in the laboratory at a reasonable cost, thereby increasing its market considerably (see Fig. 6). The superiority of diamond as a thermal management material arises primarily from its high thermal conductivity combined with its electrical insulation properties. The purest natural diamonds (type IIA) have a thermal conductivity of about 20 W/cm K at room temperature. Those containing nitrogen as an impurity (type I) have a lower thermal conductivity. CVD diamond shows a variation in conductivity from near that of copper (4 W/ cm K) to about 16 W/cm K. Variations in the thermal conductivity are also observed when measured along and across the plane of the substrate. This variation arises from variations in crystalline size, chemistry, and crystal orientation. Although the thermal conductivity is slightly more in the direc-

Figure 6

Large-area (10 cm2) diamond substrates. (Courtesy of Norton Diamond.)

520

Malshe and Brown

tion perpendicular to the substrate surface, the in-plane thermal conductivity is still excellent in high-quality material (see Table 1), making diamond an excellent choice for thermal management applications. Furthermore, the low coefficient of thermal expansion is well matched to that of silicon, ensuring that stress is minimal when used with silicon ICs. Diamond is also a lowloss dielectric material, which is an important property for substrates used in high-speed digital circuitry. CVD diamond films or substrates are already competitive in price for applications requiring ‘‘small’’ pieces. For example, a small-area diamond submodule could be used as a drop-in module in a large electronic module. However, there are a number of applications where thin, large-area CVD diamond substrates could be used. Numerical calculations and simulations have shown that depending on the application, there is an optimum thickness for the CVD diamond film that will efficiently distribute the heat. This makes it possible to realize some devices and configurations that have not been possible to date. This includes multiple heat sources on a single substrate. Each heat source may be quite small, and provided the separation of the heat sources is sufficiently large, the required thickness of the substrate will be determined by the size (and power dissipation) of the individual heat-

Table 1

Measured Properties of CVD Diamond

Thermal conductivity (at 25⬚C) In-plane Perpendicular Optical transmission ␭ = 10 ␮m ␭ = 2 ␮m Young’s modulus Shear modulus Bulk modulus Poisson’s ratio Thermal expansion (RT 400⬚C) Density Specific heat at RT Hydrogen content Dielectric constant Loss tangent Electrical resistivity Flexural strength Chemical resistance Source: Ref. 4.

13.50 to 16.10 W/cm K 21.70 W/cm K >70% 63% 1180 GPa 514 GPa 558 GPa 0.148 2.0–2.6 ⫻ 10⫺6/K 3.51 g/cm3 0.54 J/g K 20 ppm 5.6 <0.001 1013 ⍀ cm 1000 MPa All acids and alkalis

Diamond Heat Spreaders and Thermal Management

521

generating devices. Some prime examples of this type of assembly are laser diode arrays and high-power amplifiers. There is a further possibility of stacking high-power laser diode arrays to give a two-dimensional pattern to realize a high-efficiency, high-intensity light source, which could be used for pumping solid-state lasers. Another potential application for CVD diamond involves substrates that utilize microchannel cooling. In present-day microchannel cooling, channels are etched into the back surface of a silicon substrate. Water or some other coolant is then circulated through the microchannels. The material between the heat-generating device and the coolant channels does not serve as a heat spreader but simply serves as a heat conductor, so the lower its thermal resistance, the better. Thus, if the silicon were replaced by CVD diamond of the same thickness and with similar microchannels, the thermal resistance of the substrate would be reduced by the ratio of the thermal conductivities of silicon and diamond. CVD diamond has also been utilized as the substrate material in microwave, as well as radio-frequency circuit packaging. Because microwave systems generally require the dissipation of high power from small monolithic microwave integrated circuits, the thermal impedance through a conventional alumina substrate is often too high. To circumvent this problem, large precision thermal vias are formed by drilling holes through the substrate and filling them with a metal. The dies are then attached to the substrates directly over the thermal vias. However, the creation of thermal vias in a substrate adds to both the processing complexity and cost. Another approach to packaging of such systems would utilize diamond as the substrate. The packaging of a microwave system using diamond instead of alumina as the substrate has been attempted [4]. A frequency divider circuit for use in an avionics automated test system for the F-16 aircraft was built on a CVD diamond substrate [4]. The superior heat-spreading efficiency of a CVD diamond substrate enables denser device packaging. As a result of using a CVD diamond substrate, the number of components used was reduced and the assembly process was simplified. Thus, diamond proved to be more economical to use than alumina because it allowed fewer and simpler assembly steps and it resulted in improved system reliability. In order to achieve higher system operating speeds, MCMs are now utilizing three-dimensional (3-D) packaging to reduce interconnect distances between substrates. The 3-D approach then requires the use of vertical interconnects between the substrates. This requires metallized signal vias through the substrates and fuzz buttons between the substrates in order to pass signals in the vertical direction. Because a 3-D package contains a large number of high-speed ICs in a relatively small volume of space, it presents a formidable thermal management problem. However, CVD diamond is an

522

Malshe and Brown

ideal thermal management material for this package configuration because it will provide both the required mechanical support for the ICs and a lateral path through which the heat generated by the ICs can be removed to a heat sink at the edges of the ‘‘cube.’’ If the diamond edges are cooled effectively, the high thermal conductivity of the diamond will maintain the dies within their rated operating temperatures [5]. Thus, CVD diamond offers viable and reasonably cost-effective solutions to the challenge of ever increasing speed and power requirements of the single-chip and multichip IC packages. However, it remains a material searching for applications that take advantage of its superior thermal as well as outstanding electrical, mechanical, and chemical properties. E.

Diamond Synthesis

Understanding and controlling the heteroepitaxial growth of diamond has been a major challenge and goal of researchers ever since CVD diamond deposition was first demonstrated. One of the difficulties associated with this effort is the small number of substrate materials with suitable crystal structures and lattice constants. The crystal structure, grain size, and thermal properties of diamond are very sensitive to the chemical vapor deposition conditions. The substrate-diamond interfacial properties are affected by the substrate surface chemistry, atomic arrangement, and deposition conditions such as substrate temperature and flow rates of the gases. 1.

Introduction to Diamond Deposition

Diamond films can be deposited from a variety of carbon-containing gases mixed with other reactive and inert gases. Methane, acetylene, carbon monoxide, carbon dioxide, and organic compounds including methyl alcohol, ethyl alcohol, isopropyl alcohol, acetone, and several ethers have been used for diamond deposition. Methyl radicals and acetylene are the predominant carbon-containing precursors detected near the growth surface. The methyl radical has been the most frequently suggested growth species, based at first on the fact that the diamond lattice is composed of sp3-bonded carbon and, later, on isotopic labeling experiments reported by Chu et al. [6]. However, a comprehensive study is needed to understand fully the competitive molecular processes and to determine which species are responsible for diamond growth. It has been shown [7] that the successful growth of diamond by CVD methods requires a combination of carbon, hydrogen, and oxygen in appropriate ratios. However, variations in gas-phase composition are not sufficient to explain the large differences in growth rate. In general, diamond depo-

Diamond Heat Spreaders and Thermal Management

523

sition has been reported between the temperatures 500 and 1200⬚C. In a few cases, it has been deposited below 500⬚C using C-H-O gas mixtures [8–11]. The substrate temperature should always remain below 1400⬚C to prevent graphitization of the growing film [11]. The gas phase is always rapidly quenched from very high to relatively moderate substrate temperatures in almost all of the deposition techniques. The substrate temperature also affects the crystal orientation at the growth surface. At low temperatures and pressures, (111) crystal phases dominate, while at higher pressures and temperatures, polyhedral crystals with (111) and (100) phases dominate [12]. Badzian [13] reported that the (111) phase dominates at lower temperatures (<900⬚C) and the (100) phase dominates at higher temperatures. Crystallographic orientation of the film-substrate has a measurable effect on the thermal conductivity and, hence, on heat-spreading efficiency. The reactor pressure affects the diamond growth rate because it determines the recombination length of the active species and, hence, its lifetime and the drift distance of atomic hydrogen [14]. The growth rate in a lowpressure plasma is generally small. 2.

Nucleation of Diamond Growth

Nucleation during the diamond growth affects the density of the film, the voids at the substrate interface and cavities within the substrate. Eventually, these voids are detrimental to the efficiency of thermal spreading capabilities (Fig. 7). There are a number of reports describing the possible mechanism(s) leading to the synthesis of diamond [15–19]. The low nucleation density of diamond on silicon, a commonly used substrate, may arise from the high surface energy of diamond, the large lattice mismatch of diamond and silicon, the low sticking probability, or low surface mobility of growth species [16]. The process of film nucleation is usually divided into several stages such as the adsorption of arriving particles, surface diffusion of adsorbed atoms, and formation of nuclei and their subsequent growth and coalescence [17,18]. For diamond growth, continued efforts by the scientific community have been devoted to increasing the nucleation density in order to reduce the surface roughness of the deposited films as well as to increase the concentration of the diamond phase in the deposited material. Increased nucleation densities also reduce the formation of voids at the substrate–diamond film interface, leading to better adhesion of the diamond film to the substrate [19]. Surface pretreatment plays a major role in the diamond nucleation process. The degree of diamond nucleation is strongly dependent on the substrate material and its surface condition as well as on the method of deposition. Consequently, several substrate materials and nucleation tech-

524

Malshe and Brown

Figure 7 Schematic and a typical SEM micrograph of a microcavity in the substrate. (From Ref. 2.)

niques have been developed to enhance growth of diamond on nondiamond substrates. Abrasion of the substrate surface using hard materials such as diamond powder is a widely studied method for enhancing diamond nucleation [20–26]. Ultrasonic or mechanical processes are generally used to effect the abrasion. This method yields nucleation densities in the range 107 to 109 cm⫺2. Different mechanisms, both morphological and chemical, could be responsible for the nucleation. Generally, these mechanisms create randomly oriented nucleation sites, which results in the growth of randomly oriented diamond crystallites. Chakk et al. [27] used a mixture of metal and diamond particles to abrade the substrate ultrasonically and enhance the nucleation density. Powders of metals such as tungsten, tantalum, molybdenum, titanium, aluminum, iron, nickel, and silicon are used and result in a nucleation density enhancement of about two orders of magnitude. The enhanced nucleation is found to be associated with a physicochemical modification of the substrate surface attained through a cooperative effect of both the metal and diamond particles during the ultrasonic abrasion process.

Diamond Heat Spreaders and Thermal Management

525

The application of a negative bias to the substrate during deposition is another method for enhancing diamond nucleation [28–31] and could eliminate the abrasion techniques used at the present time. The negative bias collects carbon ions at higher rates onto the silicon surface and increases their bond strength with the surface silicon owing to ion mixing. This enhances the formation of carbon clusters, overcoming reevaporation and diffusion. It is generally believed that energetic ion bombardment also plays a critical role in bias-enhanced nucleation. Ion bombardment could increase the mobility of surface species and help them find nucleation sites by destroying the weak sp2 structure and increasing the sp3 structure, leading to the formation of diamond nuclei precursors. Enhancement of diamond nucleation was observed when a silicon substrate was coated with a catalytic material such as iron [32]. The catalyst plays an important role in the diffusion of carbon atoms into the substrate and the creation of more diamond nucleation sites. A few minutes of incubation time was required for the growth of nuclei, which increased with increasing thickness of the iron film. In addition, carbon fibers [33], fullerene clusters [34], graphitic carbon [35], and amorphous carbon [36] have been used to enhance the nucleation density. A novel technique referred to as electrostatic seeding has been reported [37]. The technique utilizes electrostatically charged, nanometer-size diamond particles to spray coat a substrate on which the diamond is to be deposited. The spraying process allows a wide range of particles to be deposited and, when the spraying is optimum, results in films that are smooth, dense, and relatively microcavity free (see Fig. 7). In addition to being a relatively inexpensive process to implement, this approach to providing nucleation sites for diamond growth can be used to nucleate substrates of any shape. Diamond growth conditions such as type and concentration of CVD gases, source of gas activation, plasma temperature, applied substrate bias, uniformity of the activated gas species, field or plasma uniformity, geometry of the substrate, surface chemistry of the substrate, and substrate temperature during deposition as well as variations in any of these conditions during deposition greatly influence the crystal size, chemistry (i.e., purity), porosity, and crystal orientation of the resulting diamond as well as the residual stress and defects in the film. These factors also control the in-plane and perpendicular thermal conductivities, thereby determining the thermal management efficiency of the diamond. This is obviously of extreme importance for diamond that is to be used as a heat spreader. With this in mind, a limited discussion of techniques known to be employed by some commercial CVD diamond vendors to deposit thermal management quality diamond films and substrates will be presented in a later section of this chapter.

526

Malshe and Brown

Finally, it is important to use the proper substrate or mandrill when depositing a thick diamond film that is intended to be freestanding so that it is released from the substrate or mandrill with minimal effort and damage. If the substrate or mandrill is not properly chosen, there is the potential for contamination of the diamond by the substrate or mandrill and/or mechanical damage due to a large CTE mismatch. In general, vendors use a proprietary substrate or mandrill, but the general philosophy is to use substrates that present a ‘‘moderate’’ CTE mismatch with diamond. This allows relatively shockless release of the diamond from the substrate or mandrill upon cooling from the deposition temperature to room temperature. The selection of non– carbide-forming materials for the substrate or mandrill greatly reduces the potential for chemical contamination of the diamond and problems with removal of the diamond from the substrate or mandrill.

F.

Growth Techniques

The synthesis of diamond at low pressures by CVD techniques is of interest both for its crystal growth science and as a process with significant commercial potential. The CVD approach is the only realistic alternative to highpressure, high-temperature synthesis methods. Although diamond films have been synthesized in the laboratory by a number of methods, the discussions presented in this section, based on published literature, will be brief and limited to three of the most commonly used techniques employed by commercial vendors of thermal management quality diamond. These are the DC arcjet discharge, hot filament, and microwave plasma-enhanced CVD techniques. 1.

Diamond Deposition Methods

The growth of diamond films at higher rates was made possible by the development of thermal and plasma-enhanced chemical vapor deposition methods in which a hydrocarbon gas, diluted with excess hydrogen, is used as the carbon precursor. The various CVD methods, although slightly different, share a number of characteristics. Activation of the carbon-containing gas (a hydrocarbon, CO, CO2, or an alcohol) is necessary in all processes. A concentration of species sufficient to etch graphite and/or to suppress graphite precursors is required. Atomic hydrogen is generally considered to be a good agent for this purpose, but atomic and molecular oxygen, fluorine, OH, and molecular hydrogen have also been used. The substrate surface must support the nucleation and growth of diamond from the vapor phase and should not include any catalysts that promote the formation of graphite.

Diamond Heat Spreaders and Thermal Management

527

Carbon-bearing species must be transported from the gas phase to the substrate surface either thermally or by electric fields in the plasma. a. DC Arcjet Discharge. The direct current (DC) arcjet discharge technique can be characterized as a very high growth rate process. Thick, freestanding diamond substrates can be routinely synthesized using this method. A DC arcjet discharge reactor for diamond deposition consists of a gas injection nozzle, composed of a rod cathode, typically made of tungsten, concentric with a tube anode. Gases are allowed to flow between the cathode and anode and are sprayed out from an orifice in the anode. A high-temperature discharge jet is created and sustained by a DC voltage across the electrodes. The substrate is located downstream from the jet stream on a water-cooled substrate stage. Carbon precursor and graphite etchant gases could be introduced at different locations depending on the desired activation temperature. Although this reactor is widely used because of its high growth rates, the film can suffer from high compressive stresses, microvoids, and high surface roughness. b. Hot-Filament CVD. Thermally activated CVD, or hot-filament CVD, is one of the earliest developed approaches to low-pressure synthesis of diamond. A filament made of a refractory metal, normally tungsten, is heated to high temperatures, around 2300⬚C, by resistance heating and is used to activate the hydrocarbon-hydrogen gas mixture. The filament, which is located a few millimeters above the substrate, also provides heating for the substrate. The hydrocarbon-hydrogen gas mixture is allowed to flow across the hot filament, where it is activated. Diamonex Inc. (Allentown, PA) uses this type of reactor to produce CVD diamond substrates commercially. Materials that have been used for the filament include tungsten, tantalum, and rhenium. Source gases reported in the literature include methane, propane, ethane, and oxygen-containing hydrocarbons such as acetone and ethanol. A DC bias has also been applied between the filament and the substrate to enhance electron or ion bombardment of the substrate. Variations of this basic reactor have been reported. For example, a tantalum tube at 2200⬚C has been used to flow the gas mixture onto a silicon substrate. A hot-filament CVD reactor is relatively inexpensive and easy to construct. Several filaments (i.e., a filament array) are used to perform depositions on large-area substrates. The important parameters in this process are the filament temperature, the position of the substrate with respect to the filament, and the gas flow dynamics. This technique, however, may have some disadvantages including contamination of the diamond film by the

528

Malshe and Brown

filament, erosion and sagging of the filament, and a relatively slow growth rate. It is also necessary to supply constant power throughout a deposition using a proper power controller. It is important to note that substrate temperature uniformity is difficult to achieve when using multiple filaments. c. Microwave Plasma-Enhanced CVD. Microwave plasma-enhanced chemical vapor deposition (MPECVD) is one of the most widely used diamond deposition techniques. A magnetron is generally used to generate microwave energy at 2.45 GHz and a waveguide assembly is used to couple the energy to a resonant cavity. Several variations of microwave reactors, such as a tubular reactor (commercially manufactured by ASTeX Inc., Woburn, MA) and a bell jar reactor (commercially manufactured by Wavemat, Inc., MA), have been used to deposit diamond. An advantage of a microwave plasma reactor over other reactors is that it is an electrodeless process, and hence there is no contamination from the electrode material. The microwave plasma excitation of hydrogen generates superequilibrium concentrations of atomic hydrogen. The collisions of electrons with gas atoms and molecules generate high ionization fractions. In a tubular microwave reactor, the microwave energy is coupled to a fused silica tube through a set of waveguides, tuners, and power monitors. The substrate is located on a platen inside the fused silica tube. The platen may be a susceptor that can be heated by coupling of the microwaves. The substrate susceptor may be independently heated using resistive heating. The use of a water-cooled substrate stage is also a common practice. The substrate temperature is normally monitored using an optical pyrometer. ASTeX Inc. currently uses a high-power, microwave plasma system to deposit freestanding diamond substrates. In a bell jar type of microwave reactor, the microwave energy is coupled to a cylindrical cavity through a waveguide and an antenna. The process chamber is isolated by a fused silica bell jar. The substrate is usually mounted on a thin graphite plate. The combination of microwave coupling and plasma bombardment heats the substrate to more than 800⬚C. The direct coupling of microwave energy with some electrically conductive substrates, such as silicon, may also cause some heating. Independent heating or cooling of the substrate can be used to ensure good control of the substrate temperature. d. Low-Temperature Deposition. Microwave plasma-assisted CVD of diamond has also been found to be effective in the temperature range 440 to 520⬚C [38]. The ability to deposit at low substrate temperatures allows the use of diamond films on integrated circuit substrates as well as on borosilicate glasses. The temperature dependence shows [38] that the process is consistent with a model that assumes the presence of radical surface site

Diamond Heat Spreaders and Thermal Management

529

pairs rather than single radical sites [39]. A schematic of the configuration is shown in Fig. 8. 2.

Specifications of Some Commercially Available CVD Diamond Films and Substrates

Specifications of CVD diamond films and substrates supplied commercially by various manufacturers are listed in Table 2. G.

Other Diamond-Based Thermal Management Solutions

One of the primary objectives of diamond film synthesis is to provide a technically and economically viable solution to thermal management of electronic systems. For a multichip module that utilizes diamond as a substrate material, the diamond serves as the thermal link between the electronic device junctions and a heat sink. It is very important, then, to have a good coefficient of thermal expansion (CTE) match between the electronic devices and the diamond and between the diamond and the heat sink. This is also a critical requirement for any other material used for thermal management. Consequently, the synthesis of thermal management materials using carbondiamond, copper-diamond, or other metal matrix–based composites must exhibit not only excellent thermal management properties but also an acceptable CTE. 1.

Synthesis of Diamond-Based Composites

Because any composite material composed of diamond and a second material essentially consists of a diamond particle matrix embedded in the second material, it is important to consider the various modes of thermal conduction in materials. For example, in metals, conduction electrons are

Figure 8

Schematic of low-temperature deposition of diamond. (From Ref. 39.)

Table 2

Supplier Specifications for CVD Diamond Supplier

Property Thermal conductivity Electrical resistivity Coefficient of thermal expansion Dielectric constant Loss tangent Dielectric strength Density Chemical resistance Oxidation behavior Shear strength Poisson’s ratio Young’s modulus Tensile strength Hardness Compressive strength a b

No data available at the time of writing. Data not available at the time of writing.

Diamonex (HFCVD) >1300 W/m⬚C >1011 ⍀ cm ⬃2 ⫻ 10⫺6/⬚C (25–200⬚C) 5.7 <0.0005 at 15 GHz 100–300⫹ V/␮m b

Inert to acids, alkalis, and solvents Resistant to 600⬚C ⬃108 N/m2 b b b b b

Astex (microwave plasma CVD) a

Diamond Heat Spreaders and Thermal Management

531

responsible for thermal conduction. However, in nonmetals such as diamond, phonons are responsible for thermal conductivity. Although the carriers of thermal energy are different in metals and in diamond, the fact that the presence of discontinuities and imperfections degrades thermal conductivity is common to both of them. Consequently, composite materials must be carefully synthesized so that they perform well in thermal management applications. Furthermore, although the bulk thermal conductivity of a material is size dependent, it does change as the physical dimensions of the material begin to approach the mean free path of the heat transport mechanism: electrons for metals and phonons for nonmetals. Consequently, the minimum diamond powder size that should be used in a diamond composite material has been estimated to be at least 10 times the phonon mean free path in diamond, or approximately 1 ␮m [40]. 2.

Commercially Available or Reported Composite Materials

The excellent thermal properties of diamond have led to its use in combination with various other materials to form diamond-based composite materials. Such materials find applications where pure diamond heat spreaders with their associated high costs are not necessary. This section discusses some commercially available diamond-based composite materials and some that have been discussed in the literature. a. Dymalloy. Sun Microsystems and Lawrence Livermore National Laboratory have jointly developed an alloy that they call Dymalloy [40]. It is a copper-diamond composite material reported to have a thermal conductivity of 4.2 W/cm K, which is better than that of copper, and an adjustable coefficient of thermal expansion that is nominally 5.5 ppm/⬚C at 25⬚C, a value compatible with both silicon and gallium arsenide. In order to overcome the thermal conductivity limitations of a diamond composite material, the diamond powder used is treated in a special way to make it suitable for composites. For example, this is illustrated by the fabrication process used by Davidson et al. [40] in the synthesis of the Dymalloy copper-diamond alloy. The diamond powder is first coated with ˚ of W-Rh using a physical vapor deposition (PVD) technique. To 100 A ensure uniform coating of the irregular diamond grains, the powder is tumbled in the coating system. This is accomplished by placing the powder in a small metal pan that is mechanically vibrated in the vertical direction by a piezoceramic transducer. The idea behind this process is to deposit a carbide-forming material on the surface of the diamond. However, all carbideforming materials oxidize rapidly in air, and bonding to oxide surfaces is ˚ difficult. This problem is circumvented by in situ deposition of a 1000 A layer of a brazable material such as pure copper onto the W-Rh coated

532

Malshe and Brown

particles. This form of diamond powder can then be packed into a form and vacuum infiltrated to form a composite. The major advantage of the powder technique is the ability to compress the powder into any desired shape. Also, the coefficient of thermal expansion can be tailored to any specification by controlling the amount of diamond powder in the composite. The surface of the composite will have a surface texture due to the embedded diamond particles. If a smoother surface is required, it can be plated up and polished. b. Polycrystalline Diamond Fibers for Metal-Matrix Composites. While the alloy mentioned in the previous section has some attractive features, Jyn-Ming Ting and Max L. Lake (Applied Sciences Inc., Cedarville, OH) have suggested the use of polycrystalline diamond fibers to form metalmatrix composites [41]. They suggest that their technique is simpler and, because it is adaptable to continuous processing, it may have commercial significance as a thermal management composite. The process for synthesizing polycrystalline diamond fiber (PDF) begins with the formation of vapor-grown carbon fiber (VGCF). VGCF is produced in three stages through the pyrolysis of hydrocarbon gases in the presence of transition metal particles. The first stage is the reduction of the catalyst, which is supported on a substrate in a hydrogen atmosphere. In the next stage, the gas flow is changed to a mixture of methane and hydrogen in a linearly increasing temperature sweep to 1000⬚C. The fibers are nucleated and elongated as methane decomposes on the catalyst. The catalytic particle migrates down the reactor in the direction of gas flow. In the final stage, the gas mixture is enriched with methane, allowing thickening of the fiber through the CVD of carbon on the fiber surface. These fibers can be grown to the desired diameter and size by controlling the deposition parameters. The VGCF fibers are then selected for suitable diameter and are preconditioned before being coated with polycrystalline diamond. Preconditioning consists of a treatment in an ultrasonic bath using diamond dust in an aqueous solution. The fibers are then mounted in a microwave plasmaenhanced chemical vapor deposition (MPECVD) reactor operating at 2.45 GHz and exposed to the plasma for a specific period of time. A gas mixture of CH4 and H2 is used and the temperature of the substrate, on which the graphite fibers are mounted, is maintained at approximately 1000⬚C. Figure 9 shows the cross section of a polycrystalline diamond fiber produced using this process. This technique offers the option of varying the diameter of the precursor graphite fiber and the thickness of the diamond coating, thereby tailoring its physical properties. For example, when a 1-␮m-thick diamond

Diamond Heat Spreaders and Thermal Management

Figure 9

533

A cross section of polycrystalline diamond fiber. (From Ref. 41.)

coating is grown on an 8-␮m-diameter carbon fiber (resulting in about 36% volume diamond), the fiber is predominantly graphite and has anisotropic properties. However, if a 9.5-␮m-thick diamond coating is applied to a 0.5␮m carbon fiber yielding a diamond volume of about 90%, the resulting fiber is predominantly polycrystalline diamond with isotropic properties [41]. c. Other Diamond-Based Thermal Management Composite Materials. As described previously, diamond-based composite materials are attractive because of the possibility of tailoring their thermophysical properties. They are used in a variety of applications, including heat exchangers in hypersonic vehicles operating in the temperature range 520–930⬚C [43], cooling fins in space power systems, and, of course, in electronic packaging. Diamond-based material can be fabricated in a variety of forms, such as thin films, composites, or tubes with hollow cores. Diamond-based fibers can be made by chemically depositing diamond onto metallic wires or nonmetallic fibers, and, as long as the core diameter is small compared with the fiber diameter, the fiber properties depend primarily on the diamond properties. Hollow cores can lead to a reduction in the composite density and may provide channels for gases or liquids for convectional cooling. In applications which demand that the thermal management material be strong and light, yet have excellent heat conducting properties, the options could include hollow core, diamond-based composites such that the core diameter is limited to under about 200 ␮m to ensure sufficient mechanical strength.

534

Malshe and Brown

High-thermal-conductivity graphite-copper composites with diamond coatings have been synthesized for thermal management packaging applications by MER Corporation [42]. The fabrication process employs a hollow cathode sputtering process to deposit a bonding layer, followed by copper, onto spread graphite fibers, which are then consolidated into composites of the desired conductivity and thermal expansion. However, graphite-copper composites are not acceptable for applications requiring electrical isolation. For such applications, the composite is coated with an electrically insulating diamond film that also serves as a heat spreader. It is important to note that significant effort is being expended in thermal simulation of electronic packages that utilize diamond-based composite materials [42,43]. H.

Thermal Conductivity

This section has been broadly divided into three subsections. The first addresses the need for measuring thermal conductivity, the second briefly describes the terminology used, and the third describes some of the techniques used commercially to measure thermal conductivity. 1.

Need for Thermal Conductivity Measurements

Diamond films or substrates exhibit a wide range of thermal conductivities depending on the synthesis conditions used to deposit them. For example, films can be grown with a variety of grain sizes. Larger grain sizes may promote the formation of large voids between adjacent grains in the films, which affects the thermal conductivity. On the other hand, films with smaller grain sizes have smaller voids, but the surface area, or the grain boundary, increases, which results in a degradation of effective thermal conductivity because nondiamond carbon is found in the grain boundaries. Each such interface acts as a resistance for heat transfer. Thus, deposition conditions must be optimized to yield thermal management quality diamond films. In particular, deposition conditions should be well controlled across the substrate, particularly while growing large-area depositions, to ensure uniform thermal properties across the substrates. As noted previously, diamond films of very high thermal conductivity can be grown using a variety of CVD techniques. Unfortunately, CVD rates are generally slow, making diamond films relatively expensive. Altering the deposition parameters in order to increase the deposition rate results in films with a high void density and a smaller thermal conductivity. Thus, from an economic point of view, it is important that the user of diamond not specify a thermal conductivity higher than what is actually needed for a specific application. Furthermore, the fact that a wide range of thermal conductivities

Diamond Heat Spreaders and Thermal Management

535

is available points out the need for simple, fast, and inexpensive methods for measuring the thermal conductivity of diamond films. 2.

Thermal Conductivity Measurement Terminology

Heat conduction in materials, including diamond, occurs primarily via phonons. Unfortunately, phonons are scattered by discontinuities in materials, such as lattice defects, voids, and interfaces, which limit heat conduction. The diamond crystal size itself can limit thermal conductivity. Thus, the measurement of thermal conductivity of diamond and other thermal management materials requires an understanding of the heat conduction process. Thermal conductivity (␬) governs the steady-state temperature distribution of a system while thermal diffusivity (D) governs the transient response. They are related by

␬ = ␳CD

(1)

where ␳ is the mass density of the material and C is its heat capacity per unit mass. The general approach to the measurement of thermal properties of materials is to measure ␬, use the ␬ data as the input for numerical modeling, and obtain ␬ values by carefully choosing the sample geometry and thermal shielding. Minimizing heat loss by convection and radiation helps to improve the reliability of the data. 3.

Thermal Conductivity Measurement Techniques

All methods used to measure thermal conductivity employ either the static method or the modulated heating method (also called pulsed heating). The analysis of results obtained using either method depends on the solution of the homogeneous heat diffusion equation. Because, in most applications, the heat flow is either one-dimensional or three-dimensional, it is convenient to use the cylindrical coordinate system. For a modulated heat source, the homogeneous heat equation takes the form D1 = {(⭸2T/⭸r2) ⫹ (1/r)(⭸T/⭸r)} ⫹ D2⭸2T/⭸z2 ⫹ i␻T ⫺ hT = 0 ⫺i␻T

(2)

where the heat source is assumed to have the form qe . In the 1-D case, the radial derivative terms are, naturally, zero. Most of the treatments that follow assume an isotropic medium, that is, the properties are direction independent. The terms that appear in Eq. (2) have the following meanings: D1: thermal diffusivity in the radial direction (in-plane diffusivity) D2: thermal diffusivity in the z direction (bulk diffusivity) ␻: modular angular frequency

536

Malshe and Brown

T: complex Fourier component of temperature at ␻ h: coefficient of heat loss by radiation a. A˚ngstrom’s Method. A schematic of the measurement setup is shown in Fig. 10. A bar is heated at one end at a frequency ␻ and the temperature is measured along the bar. This then is a one-dimensional problem. Further, if loss to the surroundings due to radiation is minimal, then the temperature [ln(Tmag)], where Tmag is the magnitude of temperature T, and ␸ (phase angle of Tmag) are linear in z, the distance from the heating source. The slope of the linear curve is (⫺␮⫺1). Thus, by measuring the temperature Tmag or ␸ at two or more points along the bar, ␮ and hence D can be calculated. If, however, the specimen is thin, there could be significant losses to the surroundings, especially at low ␻. In this case, although both [ln(Tmag)] and ␸ remain linear in z, their slopes have different values. It can be shown that the product of the two slopes is ␮⫺2, which is independent of h. Thus, D can be determined uniquely. From the knowledge of D, it is then possible to compute h. b. 3␻ Method. The designation 3␻ method arises from the way the thermal signal is measured. A schematic diagram of the measurement setup is as shown in Fig. 11. A narrow strip of metal and metal pads for electrical contacts are deposited onto the surface of the test specimen. An ac current I at ␻ = ␻0 , is passed through the strip, which is assumed to have negligible thickness. By measuring the voltage, V(␻0), one can calculate the wire resistance, R, which is a measure of the specimen temperature because of the temperature dependence of the wire resistivity. The strip also acts as a line source of heat due to the power dissipated, P = I 2R, which produces a

˚ ngstrom’s method for measuring thermal difFigure 10 Schematic diagram of A fusivity. (From Ref. 94.)

Diamond Heat Spreaders and Thermal Management

537

Figure 11 Schematic diagram of 3␻ technique for measuring thermal conductivity. (From Ref. 94.)

cylindrical thermal wave at ␻ = 2␻0 . Thus, the strip serves a dual purpose. The temperature in the specimen is T = (P/L)K0(r/␮ ⫹ ir/␮)/(␲␬)

(3)

where P/L is the power generated per unit length of wire and K0 is the zeroth-order modified Bessel function. It is also assumed that the strip width ws << ␮. In the limit, r (the radial distance from the wire) << ␮, and we have T = (P/L) ln(2␻0)/(2␲␬)

(4)

where all terms independent of frequency have been ignored. The slope of T versus ln(2␻0) can be used to calculate ␬. The ac component of R will depend on 2␻0 . The measured voltage at 3␻0 provides a measure of R and, hence, T. The dependence of R on temperature T can be calibrated by performing measurements of V(␻0) as a function of temperature. Thus, the specimen temperature and the thermal signal can be determined simultaneously. c. Photothermal Radiometry. This technique consists of heating a specimen with a modulated or pulsed heat source and measuring the thermal signal with an infrared (IR) detector. This method has been widely used to measure the thermal diffusivity (D) of diamond. The experimental arrangement is shown in Fig. 12. The experiment can be conducted in several ways. In the first method, a modulated laser beam of radius r0 >> w and r0 >> ␮ heats the top surface of the specimen. The large beam size allows 1-D anal-

538

Malshe and Brown

Figure 12 Schematic diagram of the photothermal radiometry method for the measurement of ␬. (From Ref. 94.)

ysis. A CaF2 lens focuses the IR radiation emitted by the center of the heated area onto an InSb detector. The thermal signal is measured as a function of ␻. In another approach, the detector scans the thermal signal as a function of r, which is the distance from the modulating heating beam. In this case, the thermal signal primarily depends on the parallel component of the diffusion coefficient, i.e., on D㛳. Usually, a nonlinear fitting procedure is required to obtain the thermal diffusivity from thermal radiometry measurements. Obviously, the 3-D analysis is more complex than the 1-D procedure. d. Photothermal Deflection (Mirage Effect). A schematic diagram of the test setup for this measurement is shown in Fig. 13. The specimen is heated by a modulated laser beam that is focused normally onto the surface. The heating causes a thermal wave to propagate in both the specimen and the air surrounding the specimen. The temperature gradient in the air close to the specimen causes a change in the refractive index of the air in the immediate vicinity of the specimen. A probe laser beam, which is made to scan the heated surface of the specimen, is deflected by the index gradient in the air. The deviation of the beam depends on the distance of closest approach x between the probe beam and the heating beam. The deviation angle is projected into two angles ␪n and ␪t, which represent the component of the deflection normal to the specimen and the component parallel to the specimen, respectively. These deflections are measured with a quadrant detector. By measuring ␪n and ␪t as a function of x for different values of ␻ and fitting the data to theoretical functions, the thermal diffusivity can be obtained.

Diamond Heat Spreaders and Thermal Management

539

Figure 13 Schematic diagram for the photothermal deflection method (the Mirage effect). (From Ref. 94.)

e. Round-Robin Measurements. The National Institute of Standards and Technology (NIST) brought people together from U.S. industries for the purpose of discussing the need to standardize the procedures for thermally characterizing CVD diamond. To address the issue of measuring the thermal conductivity of CVD diamond accurately, a working group on standardization of thermal conductivity measurement of CVD diamond was established [44]. It was agreed that standardizing the measurement of thermal conductivity (␬) of CVD diamond was important. However, because a standardized measurement procedure did not exist at that time, the group decided to conduct a round-robin thermal conductivity measurement experiment in which several laboratories agreed to perform measurements of ␬ on a given set of specimens [43]. A considerable variation in measured values of ␬ resulted from this effort, independent of whether the measurements were made at the same laboratory using different techniques or at different laboratories. The variations have been attributed to various causes [45]. In some instances, a single laboratory obtained different values at several locations on the same specimen, indicating substantial inhomogeneity over the surface of the specimen. Some measurement techniques, such as the photothermal deflection and transient thermal gradient methods, have depth selectivity so that parameters measured on one surface differ from those measured on another surface because of differences in growth conditions at the two surfaces. Because diamond synthesis progresses through columnar growth, the density of voids in a sample is also a factor that affects thermal conductivity. Consequently, if the probed depth varies from one technique to another, then

540

Malshe and Brown

the measured parameter values affecting thermal conductivity (␬) will be averaged over that depth. As the void density is expected to vary with depth, it is natural that the measured ␬ values would also vary if the measurement technique has depth sensitivity.

II.

POSTDEPOSITION PROCESSING OF DIAMOND HEAT SPREADERS

As noted previously, depending on the deposition parameters, CVD diamond films of different orientations, sizes, and thicknesses can be grown using a variety of techniques. However, the particular parameters used to grow diamond films dictate, to some degree, their thermal characteristics. In fact, diamond films are usually also nonuniform in thickness and grain size. Furthermore, they are composed of randomly oriented crystallites and exhibit intercrystal stresses unless grown in a very controlled manner [46]. A.

Postdeposition Processing

The nonuniform thickness of CVD diamond films, along with their surface roughness, can severely limit their thermal management efficiency due to poor contact between a heat source and the diamond and between the diamond and a heat sink. Photolithographic definition can also be limited by these film parameters. Furthermore, metallization adhesion can be adversely affected. To overcome these potentially serious problems, CVD diamond films must be subjected to postdeposition processing. The exact postdeposition processing steps required depend on the specific application of the diamond. For thermal spreader applications, it is essential that both surfaces of the diamond are flat and clean. B.

Cutting and Drilling

The typical freestanding diamond substrate produced commercially using state-of-the-art CVD techniques has a collar at its outer edge resulting from diamond growth extending down the side of the mandrill on which the diamond is grown [47]. Also, the outer portion of an as-grown substrate generally has a nonuniform grain size/chemistry if it is a large substrate. Consequently, the first required postdeposition processing is the removal of the collar so that the resulting substrate is planar. Furthermore, there is a need to cut the as-grown substrate into various sizes and shapes, as dictated by specific thermal management applications.

Diamond Heat Spreaders and Thermal Management

541

Cutting of polycrystalline CVD diamond poses some challenges. The hardness of diamond, along with its chemical inertness, mechanical brittleness, and polycrystalline nature, presents serious challenges for precise cutting using conventional methods. The problems are compounded if the diamond film to be cut is thin and is deposited on another substrate that has different thermal properties. Typically, CVD diamond is cut using a pulsed neodymium yttrium-aluminum-garnet (Nd-YAG) laser, assisted by a reactive gas jet. The wavelength, power, and repetition rate of the laser, as well as the chemical quality and thickness of the diamond, affect the definition of the cut, the cutting-induced burr, contamination, melting, and microcracking of the material. In fact, the definition of a cut corner is a function of the laser wavelength. Laser cutting speed can be enhanced by having the right photochemistry at the diamond surface and by introducing the proper gas(es), such as a combination of a reactive gas, like oxygen, and an inert gas such as argon. Although a typical application of diamond in thermal management has the IC mounted directly on one of its sides, the same and/or other side might be used for high-density interconnection. Also, the edges of the substrate may be used to extract heat, depending on the cooling scheme employed. Furthermore, drilling of vias, which are filled with metal to provide electrical interconnection from one substrate surface to the other, may be necessary for some thermal management diamond substrates and requires that the diamond chemistry of the via surface be retained following drilling. These vias are typically 100–150 ␮m in diameter, although other sizes are possible, depending on the particular application. Depending on the size and design of a diamond-based MCM, the number of vias required could vary from a few hundred to a few thousand on one substrate. The quality of a via is determined by its edge angle, wall contamination, and the definition of the top surface hole wall junction. Consequently, the drilling of vias in diamond films using conventional techniques has not had much success. Although drilling of diamond using lasers is not new, very little literature is available on the subject because of its potential commercial value. At the present time, an Nd-YAG laser operating at 1.06 or 53.2 ␮m is used in the pulsed mode for drilling via holes in CVD diamond [47]. However, laser drilling of thousands of vias in a large-area, freestanding CVD diamond substrate can cause bowing of the substrate during drilling, thus making the process difficult. However, successful laser via drilling has been accomplished by drilling from both sides of the substrate [48]. In another novel process, fine laser drilling of diamond has been demonstrated by laser treatment of diamond in a reactive liquid medium [49]. It was shown that liquid chemistry plays an important role in laser machining.

542

C.

Malshe and Brown

Polishing and Planarization

State-of-the-art physical and chemical methods of polishing and planarization of CVD diamond can be broadly classified as (1) nonreactive contact methods, (2) reactive contact methods, and (3) noncontact methods. Each method offers some advantages but also has some disadvantages. 1.

Need for Polishing

The rough surface of diamond substrates is an innate limitation of state-ofthe-art diamond growth techniques [50]. Consequently, the surface must be polished before the substrate will perform effectively in thermal management or other applications. Polishing is typically achieved by microchipping or dissolution or both. Hence, the factors to be considered when selecting a polishing method for a given application include the as-deposited surface quality, the polish rate (and hence the total time required to attain a specified surface smoothness), the minimum achievable roughness, surface contamination, as well as the equipment and operating costs. Different methods used for polishing the diamond film are described next. 2.

Polishing Techniques

For CVD diamond, the term polishing is typically used to describe material removal using any process. However, the total polishing process actually consists of two very different operations: lapping (coarse material removal) and polishing (fine material removal). In the discussions that follow, unless specified otherwise, a distinction is not made between the two operations. a. Mechanical Polishing. The mechanical polishing of diamond proceeds by grinding the samples on a flat metal wheel or scaife, which is usually made of cast iron. An abrasive powder, consisting of material harder than that to be polished, and a binder, often olive oil, are spread on or impregnated into the wheel. For polishing diamond, diamond powder is used as the abrasive material. The cast iron wheel is rotated at a high speed (>2500 rpm). A load is applied to force the surface of the diamond against the rotating wheel [51]. A coarse powder (>1 ␮m) is used in the initial stage of polishing. This process is called lapping. Figure 14 shows an example of a mechanically lapped diamond surface. The primary advantages of postdeposition processing of diamond using a process such as mechanical polishing are that the process is simple and scalable. Also, there is no requirement for substrate heating or for reactive gases. Furthermore, the cost of the equipment is low. Moreover, there are virtually no limitations on the size of diamond samples that can be polished. Finally, the chemical quality of the film is not drastically changed after the

Diamond Heat Spreaders and Thermal Management

Figure 14

543

Example of a mechanically polished diamond surface. (From Ref. 95.)

polishing process. There are, however, some limitations that should be weighed against the advantages. The polishing time can be quite long, up to several days, and the applied pressure on the substrates can cause microcracking. The polishing rates depend on the quality of the diamond and orientation. b. Thermochemical Polishing. This technique is based on the atomic dissolution of carbon into a hot metal. The technique was introduced by Grodzinski in 1953 [52], and Yoshikawa has used it in more recent times [53]. Yoshikawa’s setup is schematically depicted in Fig. 15. CVD diamond substrates are polished using a hot plate made of iron [53], nickel [54], manganese [55], or molybdenum [54]. The polishing rate depends on the diffusion of carbon atoms from the diamond surface into the hot metal plate. In general, it has been found that of these materials, iron yields the highest polishing rate [56]. However, as polishing progresses, the carbon content in the iron increases, which decreases the polishing rate. The plate temperature is required to be maintained in the range of 700 to 950⬚C. At lower temperatures, the polishing rate is low because of insufficient chemical reaction. At higher temperatures, some voids have been observed in the diamond surface being polished. Their number has been found to increase with temperature [54]. They are formed because of selective etching of diamond in the presence of oxygen. Thus, the polishing temperature must be optimized

544

Figure 15 Ref. 95.)

Malshe and Brown

Schematic diagram of a thermochemical polishing apparatus. (From

Diamond Heat Spreaders and Thermal Management

545

and the optimum temperature depends on the metal from which the plate is made. Thermochemical polishing is affected by the ambient. In vacuum, using iron, an etch rate of about 7 ␮m/hr is the maximum possible. However, the finish is coarse. A smooth finish is possible in a hydrogen atmosphere, but the polishing rate is only about 0.5 ␮m/hr. Consequently, it is common practice to start the polishing in vacuum to take advantage of the higher material removal rate and then finish the polishing in a hydrogen atmosphere to achieve a smooth finish. Figure 16 shows an example of a thermochemically polished diamond surface. Researchers at AT&T Bell Labs have used different metals, such as manganese, cerium, and, lanthanum instead of iron [57,58]. Under optimized conditions, etch rates as high as 50 ␮m/hr have been realized using these materials. The cost of equipment required for thermochemical polishing is slightly higher than that for mechanical polishing. A vacuum chamber may be required. The operating cost depends primarily on the cost of the metal and the heating medium. The limitation on substrate size is similar to that for mechanical polishing. Particles from the metal plate that remain on the diamond surface are a major source of contamination. Also, due to the higher processing temperatures at the diamond-metal interface, the possibility of forming a thin nondiamond layer, such as graphite or diamond-like carbon, on the surface of the diamond exists. Finally, the processing time is of the order of several hours and the technique may not be suitable for thin films. c. Chemical-Assisted Mechanical Polishing and Planarization (CAMPP). Chemical-assisted mechanical polishing and planarization has

Figure 16

Thermochemically polished diamond surface. (From Ref. 95.)

546

Malshe and Brown

been shown to be a promising technique for fine polishing, not lapping, of diamond substrates [59]. CAMPP uses mechanical polishing in conjunction with chemicals to enhance the removal rate. Diamond under load comes in contact with an alumina base plate in the presence of oxidizing chemicals such as KOH or KNO. The diamond removal rate is directly proportional to the speed of the polishing wheel. Hard materials, such as diamond powder, may be added to the chemicals to enhance the polishing rate. The polishing rate using this technique is higher than that for mechanical polishing. Surface contamination of the substrate is less than that of either mechanical or thermochemical polishing. The technique, in principle, is also capable of polishing nonplanar surfaces. The polishing rate is a function of diamond chemistry. Figure 17 shows an example of a CAMPPpolished diamond surface. The required equipment and operating costs are only slightly higher than those for mechanical polishing. d. Laser Polishing. Contact polishing methods, such as abrasive polishing, apply stress to the diamond surface, which makes them impractical for polishing very thin (typically a few micrometers thick) diamond films or substrates. Moreover, most of the contact techniques are unsuitable for contoured surfaces or for selectively polishing small areas on larger surfaces. It is for such polishing needs that noncontact polishing techniques, such as laser polishing, prove beneficial.

Figure 17

Example of CAMP polished diamond surface. (From Ref. 95.)

Diamond Heat Spreaders and Thermal Management

547

Laser polishing typically uses an Nd-YAG Q-switched pulsed laser [60], an excimer laser [61], or a combination of the two [62] to effect rapid polishing of diamond. The polishing is accomplished by irradiating the diamond surface with a pulsed laser operating at a repetition rate between 1 and 100 Hz. The polishing mechanism involves transient thermal oxidation and/or evaporative ablation of diamond [63]. A schematic diagram of a laser polishing system is shown in Fig. 18, and Fig. 19 shows the surface morphology of a diamond film following laser polishing. Laser polishing is well suited for coarse material removal over large areas [62]. Polishing proceeds by traversing the diamond film under a stationary laser beam. Traverse step size should be a fraction of the average grain size to ensure homogeneous irradiation of the diamond surface and hence a consistent surface topology after polishing. Depending on the laser beam wavelength and power, the laser pulses will heat the diamond film to a depth of a few submicrometers to a few micrometers. Depending on the polishing parameters, an approximately 150-nm-thick layer of nondiamond carbon may remain on the sample surface after polishing [64]. The polishing rate is limited by the grain size, ambient conditions, laser spot size, and laser power. It has been shown [65] that a high peak power (>1 GW/cm2) YAG laser in an oxygen ambient leads to a polishing rate of about 100 nm/min. The fastest coarse diamond polishing rate using a laser was demonstrated using multiple lasers on a freestanding thermal management substrate [62].

Figure 18

Schematic of laser polishing of diamond thin films. (From Ref. 95.)

548

Malshe and Brown

Figure 19 Surface morphology of a diamond film after being laser polished. (From Ref. 95.)

The time required to polish a diamond substrate depends on the size of the sample, the wavelength of the laser, the laser beam size, ambient conditions, and the scanning capability of the stage/laser. The surface finish attainable is limited by the quality of the different diamond facets and grain boundaries, which have different evaporation rates in response to the same laser radiation [66]. Moreover, if the polishing technique is applied to thinfilm diamond, the laser can generate shock waves, resulting in microcracks that can adversely affect the adhesion of the diamond to the substrate [67]. Laser polishing is probably the best method for rapid, coarse polishing of diamond. The technique does, however, have limitations for fine polishing. Although laser equipment costs are high, the technique is still economical because of its high polishing rates. Furthermore, as noted previously, laser polishing is attractive because it can be used on contoured surfaces. e. Ion Beam Polishing. Similar to laser polishing, ion beam processing is a noncontact technique that uses a focused ion beam. The polishing mechanism proceeds by sputtering of carbon atoms by the ion beam from the diamond surface reactively and/or nonreactively depending on the

Diamond Heat Spreaders and Thermal Management

549

ion being used. A schematic of an ion beam polishing system is shown in Fig. 20. The diamond polishing rate depends on the orientation of the sample with respect to the angle of incidence of the bombarding ion beam. Diamond samples are typically heated to a temperature of about 700⬚C and bombarded with argon or oxygen ions. At normal incidence, the yield for 500 eV O2 ions is about seven times larger than the yield for 500 eV Ar ions. Oxygen ions oxidize the carbon, which possibly accounts for the higher rate of material removal. An etching rate of 20 nm/sec has been reported [68] in an oxygen atmosphere. The equipment required for ion beam polishing is very expensive. Although the sputtering rate depends on the energy of the ion beam, higher

Figure 20

Schematic of ion beam polishing setup. (From Ref. 95.)

550

Malshe and Brown

beam energies have a higher potential for producing surface damage and may lead to surface graphitization if the beam energy is very high. However, although the technique is expensive, it is clean and does not leave any chemical contamination on the diamond. The sample size is limited only by the size of the ion beam chamber. f. Reactive Ion Etching. Reactive ion etching (RIE) of diamond films has been reported [69] using O2 and H2 gases. The etching was performed in a DC plasma at an rf power of 200–300 W and at a pressure of ˚ /min have been 65–80 mtorr. Diamond etch rates on the order of 300–400 A achieved using 400 eV oxygen ions. Hydrogen ions are known to produce lower etch rates than oxygen ions. However, using oxygen, RIE etches at a rate about three times higher than ion beam polishing. RIE can produce surface contamination due to plasma heating and may cause anisotropic etching of the diamond. The cost of the equipment is relatively high and the specimen size is limited by the active plasma area. 3.

Planarization

As noted previously, in state-of-the-art CVD diamond synthesis, depending on the growth temperature and pressure conditions, favored crystal orientations dominate the competitive growth process. As a result, the grown films are columnar and have a polycrystalline structure and a rough surface morphology, as seen in Fig. 21. They also contain a number of microactivities, which are a consequence of their polycrystalline and columnar structure. These microcavities manifest themselves as surface pits when diamond is polished. Surface roughness and surface pits are undesirable features, predominantly because of the photolithography and via processing limitations they pose during fabrication of a thermal management substrate (as shown in Fig. 7 earlier). Generally, the surface roughness increases with increasing the film thickness. For 500- to 900-␮m-thick diamond substrates, the large surface roughness (tens of ␮m or larger) limits the thermal management efficiency as well as the photolithographic resolution. Chemical and/or mechanical polishing is used to reduce the surface roughness, but polishing does not eliminate the surface pits as seen in Fig. 22, which shows a scanning electron micrograph (SEM) that compares an unpolished diamond surface with a chemical-assisted mechanical polished and planarized (CAMPP) CVD diamond substrate surface. As microcavities are a natural result of state-of-the-art diamond deposition techniques, surface polishing will not eliminate them because, as polishing proceeds, some surface pits are polished away, but others are created as microcavities in the bulk approach the surface.

Diamond Heat Spreaders and Thermal Management

551

Figure 21 Atomic force microscopy (AFM) image of an unpolished diamond substrate. (From Ref. 95.)

In a novel approach to achieving planarization of CVD diamond, Malshe et al. [70,71] developed a technique that involves filling the surface pits with a material, such as a polyimide, spin-on glass, metal paste, or electroplated metal, that is compatible with microelectronic processing technology. It was shown that commercially available polyimides are especially attractive for effectively planarizing a diamond surface. The ability of a material to planarize a diamond surface is determined by its viscosity, solid content, and flow and shrinkage during cure. If desired, the filler layer can then be polished back to expose the diamond surface, although this polishing step has been shown to be unnecessary from a thermal performance standpoint [71]. The materials used as a filler material for planarization were polyimide PI-2721, PI-2610, and PI-2611 from DuPont [72]. The first layer was PI-2610, which has a low viscosity and, hence, is effective in filling the microcavities. By using a low-viscosity material such as PI-2610, the potential difficulty of air being trapped inside the microcavities is resolved,

552

Malshe and Brown

Figure 22 Scanning electron micrograph (SEM) comparing an unpolished diamond surface with a chemical assisted mechanical polished and planarized (CAMPP) CVD diamond substrate. (From Ref. 95.)

as the low-viscosity material can easily flow into the high-aspect-ratio microcavities and fill them up. This layer is partially cured and the diamond substrate is then treated with a coat of PI-2611. PI-2611 is more viscous than PI-2610 and completely planarizes the diamond substrate. Curing the PI-2611, layer along with PI-2610, completes the planarizing process. The adhesion strength of polyimide on diamond is reported [50] to be close to 4000 psi. The polyimide films exhibited excellent adhesion of CVD diamond even after subjecting the planarized diamond samples to thermal shock testing, as specified in MIL-STD-883. The reduction in adhesion strength was only about 33%, so the adhesion strength remained well above acceptable limits for packaging applications. The surface roughness of polyimide planarized diamond is quite adequate for photolithographic processes. Moreover, it was shown by simulation studies [71] that a polyimide thickness of about 5 ␮m does not significantly degrade the thermal performance of the diamond substrate.

Diamond Heat Spreaders and Thermal Management

D.

553

Metallization

Metals are used in the packaging of ICs to provide electrical interconnection paths for signals, power, and ground. The most popular metals used in packaging are gold, silver, copper, and aluminum. Copper readily oxides and cannot be deposited onto many polymers because it reacts with them. Aluminum is also easily oxidized, although the oxidation process is self-limiting. Gold, on the other hand, is a very attractive metal for applications as a primary conductor as it is one of the best electrical conductors known, has good thermal conductivity, is essentially inert in most ambients, is very bondable, and provides a low contact resistance. However, due to its high cost, gold is normally used only as a top layer metal. 1.

Requirements

Metallization of CVD diamond is essential for bonding ICs to the diamond and providing electrical interconnects between the ICs. Bonding of ICs to diamond using metal is necessary for electronic packaging that is intended to take advantage of the excellent thermal and electrical properties of diamond. A metal contact on diamond is generally required to be both strongly adherent and stable at high temperature. To obtain strong adhesion, unless the metal is a carbide former, it is necessary first to deposit a layer of metal (which acts as the adhesion layer) that chemically reacts with diamond. It is the stability and chemistry of this layer that essentially determine the hightemperature thermal stability of the metallization system. The interface layer generally must satisfy three requirements: (1) interdiffusion of the interface metal into the primary conductor and the substrate should be low, (2) the interface metal should be thermodynamically stable with respect to the primary conductor and the substrate, and (3) the interface metal should adhere well to both the primary conductor and the substrate and should be of uniform thickness. No single metal meets all these requirements for all substrates. Depending on the physical and chemical mechanisms involved, diffusion barriers may be classified into three categories: passive barriers, sacrificial barriers, and stuffed barriers. In a passive barrier, the interface metal is chemically inert with respect to the substrate and the primary conductor. In a sacrificial barrier, the interface barrier reacts with the substrate and/or the primary conductor in such a way that it will not be consumed by the reaction until the lifetime of the product has expired. For a stuffed barrier, the grain boundaries, which provide the easiest diffusion paths, are plugged. The top metal layer of a multilayer metallization system must provide (1) good electrical conductivity, (2) corrosion resistance, (3) strong adherence to dielectrics, and (4) compatibility with the assembly process. Gold is

554

Malshe and Brown

an excellent material for top layer metallization because of its thermodynamic stability as well as its compatibility with die attachment and wire bonding processes. Two schemes have been used for metallization of diamond substrates. In one of the approaches, the adhesion layer is chromium (Cr) or titanium (Ti), with the overlayer metal being either gold (Au) or a gold-tin (Au-Sn) alloy. This system will fail at high temperature because Cr and Ti are soluble in Au and, therefore, diffuse into the gold, resulting in degradation of the diamond-metal bond and loss of adhesion [72,73]. In another system, the adhesion layer and the overlayer are the same material, but they are separated by a diffusion barrier. The diffusion barrier layer is typically an amorphous film of Ti ⫹ Si ⫹ N that is reactively sputter deposited onto the adhesion layer before depositing the overlayer. The best results have been achieved using Ti for the adhesion layer and overlayer with a barrier layer of Ti ⫹ Si ⫹ N [74]. 2.

The Metallization Process

Metallization for electronic applications is challenging because it demands optimization and control of the process for consistency and repeatability. Furthermore, thermal management applications require thick metallization, which has to be economic. One of the most reliable and economic techniques for realizing reliable and consistent metallization is electroplating [75]. However, extra effort is required when electroplating an insulating material such as CVD diamond because insulators cannot be used as an electrode in an electroplating bath. Hence, an electrically conducting seed layer must first be deposited onto an insulator before it can be electroplated. Deposition of a seed layer is generally accomplished by sputtering or by thermal evaporation. Although a seed layer is required, electroplating is one of the best, most reliable, and most cost-effective methods for metallizing CVD diamond. Wang et al. [76] discuss a method for metallizing plated through holes (PTHs) using electrolytic copper. To ensure uniform plating inside PTHs that are 13 mils or smaller in diameter, they use a very high throwing power. This is realized by using a low copper concentration and a high acid concentration in the plating solution. The throwing power can be further enhanced by using a low current density during the electroplating. In this way, it is reported that through holes as small as 6 mils in diameter can be plated [76]. Baker [77] described plating baths for the alkaline gold sulfite system in which arsenic was used as an additive to produce electroplates with hardness depending on the amount of oxalic acid added to the electrolyte. Larger amounts of oxalic acid produced lower hardness in the electrodeposit.

Diamond Heat Spreaders and Thermal Management

3.

555

Thick Film Metallization

Dupont Electronics Materials (Research Triangle Park, NC) and Crystalline Materials Corporation (San Ramon, CA) have reported on their development of thick-film metallization for CVD diamond, specifically for MCM applications [94]. Thin-film metallization has traditionally been the approach to providing metallization on CVD diamond. This newly developed thick-film metallization system, which uses a unique inorganic binder, is designed for nontraditional substrates such as diamond [92]. 4.

Via Filling

The use of at least one side of a diamond substrate for direct bonding of ICs, the same and/or other side for high-density interconnections, and the edges for heat extraction is a realistic configuration of a single high-density substrate used in a 3-D multichip module. However, this type substrate structure requires the use of vias to provide paths for electrical interconnection between the two sides. The material used to fill the vias obviously needs to be a good conductor, should have a coefficient of thermal expansion close to that of diamond, and should also have excellent adhesion to diamond. Tungsten-copper composite via-fill material has a reasonably good conductivity and a coefficient of thermal expansion more closely matched to diamond than most other conductors [48]. Raytheon Company (Sudbury, MA) and Micro Substrates Corporation (Tempe, AZ) have reported the results of their combined efforts to realize good via-filled diamond substrates [50]. In order to minimize the diamond substrate polishing effort, the smoother, mandrel side of the diamond substrate was used to form the metal interconnects by photolithography. A via-filled diamond substrate is realized in the following sequence: (1) CVD diamond is synthesized on suitable mandrels, (2) the diamond is separated from the mandrel, (3) the diamond is cut and polished, (4) vias are drilled into the substrate, (5) an adhesion layer is deposited in the vias because of poor adhesion between the via wall and conventional via-fill material [50], and (6) the via-fill material is deposited. The via-filled diamond substrate is then processed further to reestablish the resistivity of the diamond substrate. The following sections provided more detail on the process steps required to form filled vias. a. Laser Drilling. An Nd-YAG laser, operating at the fundamental wavelength (1.06 ␮m and 532 nm), is found to be suitable for drilling holes in diamond, a few hundreds of micrometers thick. The process is attractive because it is clean and good-quality circular holes can be drilled. For drilling holes in thicker CVD diamond substrates, a process that uses laser core drilling from both sides of the substrate results in clean, circular holes.

556

Malshe and Brown

A CO2 laser, which is often used for high-rate drilling of alumina substrates, is not very effective for drilling holes in CVD diamond substrates due to the very low energy absorption of the CO2 wavelength by diamond. However, better results are obtained with CO2 lasers in the presence of oxygen. b. Adherent Deposits. Adhesion of the via-fill material to diamond is generally so poor that via plugs can be pressed out of their holes very easily. To improve the adhesion, an adherent layer of titanium-tungsten (TiW) can be deposited between the diamond and the via-fill material. Other materials, some proprietary, have also been used as adhesion promoter materials. The adhesion material is sputtered into the vias. Limiting the adhesion promoter layer to the via wall is important because it eliminates the need to repolish the diamond surface after deposition of the via-fill material. An etch-back process, involving a simple via-fill masking process that does not use any photomasks but uses a high-viscosity filling material, has been developed [48]. The material fills the via to the surface of the opposite side. A two-step fill, etch, and strip sequence provides good results. Completion of the etch can be determined by observing conductivity using microprobing techniques. c. Metallizing Vias. For small vias with a large aspect ratio, lowpressure CVD (LPCVD) techniques [79] are used to fill vias ranging from 0.001 to 0.010 in. in diameter in 0.010- to 0.030-in.-thick CVD diamond (aspect ratios in the range 1–30) [80]. For large holes with low aspect ratios, a hybrid sputter-plating process is used. The critical issues for a via-filling process include adhesion and electrical conductivity. Of these, an adhesion problem can be handled by using adhesion-promoting deposits as described previously. For small holes with high aspect ratios, penetration into the via and porosity of the deposit are critical. Tungsten is generally chosen for via filling using LPCVD because it is a good carbide former and is the best electrical conductor among the transition elements (which are known to be very good carbide formers and hence are all potential candidates for metallization of CVD diamond). Copper and aluminum are also excellent choices for via filling but require Ti-W as an adhesion material. Heat treating a thick (>0.5 ␮m) tungsten film on diamond to 1000⬚C results in severe adhesion degradation. The tungsten is almost completely delaminated leaving no trace of tungsten (W) or any of its carbide (W2C or WC) on the surface. In ˚ ) of tungsten, when heat treated at 960⬚C for contrast, a thin layer (⬃200 A 30–60 min, maintains its excellent adhesion to diamond. Following the deposition and proper heat treatment of a thin tungsten film, the diamond substrate can be cooled and thick layer of tungsten can be applied.

Diamond Heat Spreaders and Thermal Management

557

d. Resistivity Restoration. This step is used to remove via-fill material that may have adhered to the substrate surface during the via-fill process. It is interesting to note that, although the diamond surfaces may exhibit increased electrical conductivity following a via-fill process, the fill materials are not responsible for the increase. The reason for the increased conductivity is the presence of graphite, which is formed when the diamond is elevated in temperature during the various via-filling process steps. Restoration of the diamond resistivity is achieved by placing the substrate in an oxygen plasma. The plasma oxidizes the graphitized diamond layers, thereby restoring the resistivity. However, care needs to be taken to ensure that the metallization in the filled vias and other locations on the diamond surfaces are not adversely affected. It should be noted that neither argon sputtering nor plasma etching is capable of restoring the preprocessing conductivity of the diamond.

E.

Die Attachment—Face-Up and Flip-Chip

In MCM technology, reliable die attachment implies that there is excellent electrical communication between the ICs and the substrate on which they are mounted. Two die attach technologies are presently being employed to attach ICs to thermal management diamond substrates. One is referred to as flip-chip die attachment and the other is the conventional face-up attachment using wire-bonding attachment for electrical connection between the ICs and the substrate. These two approaches are depicted schematically in Fig. 23.

Figure 23 Schematic of flip-chip and face-up (wire bonded) die attachment configurations. (From Ref. 82.)

558

Malshe and Brown

Flip-chip die attachment, which was originally developed by IBM in the 1960s and referred to as Controlled Collapse Chip Connect (C4) [81], is now finding more widespread use due to its distinct advantages of selfalignment, high assembly yield, and good reliability in small coefficient of thermal expansion (CTE) packages. The configuration also offers very high I/O capabilities and a minimum I/O signal path length for excellent electrical performance compared with wire bonding. However, flip-chip attachment is more expensive and it also suffers from problems such as conduction heat dissipation in substrate-cooled packages and structural fatigue failure in applications with large CTE differences. Although the flip-chip approach places the ICs closer to the substrate, the solder bumps present an overall bottleneck to heat flow because they do not present a continuous interface layer. Thus, the choice of a die attach technology must be guided by the package design required for a given application. 1.

Thermal Modeling Procedure

Three-dimensional finite element analysis (FEA) models have been developed that estimate the temperature buildup of a device. The inputs to such a model are material parameters of the major components of the electronic package system (i.e., die, substrate, and attachment layer). Brown et al. [82] included the solder bumps of a flip-chip when performing a flip-chip analysis because the localized conduction associated with each bump can result in hot spots in the die. Further, the solder bumps were constructed using rectangular elements but were given a thermal conductivity value calculated to account for the C4’s typical truncated spherical shape. The heat load was assumed to result from heat dissipated by ICs operating uniformly over the surface of the substrate. Cooling was modeled as a constant temperature plane on the bottom or the edge of the substrate, two commonly used configurations to connect a diamond heat spreader to a heat sink. Parameters of interest included the die material properties, die dimensions, solder material properties, solder bump diameter and shape, solder bump array pitch, underfill material properties, substrate dimensions, and substrate material properties. As a point of reference, a baseline case was established with the model parameters having values as listed in Table 3. The maximum temperature differentials between the die junction temperature (Tj) and the substrate cooling temperature (Tc) as well as the heat flow paths were noted from the finite element solutions to compare package performance as a function of the selected parameters. a. Die Material. The choice of die material (Si or GaAs) does not have a large effect on either the face-up or flip-chip package thermal performance. In the face-up, bottom-cooled model, the normal resistance

Diamond Heat Spreaders and Thermal Management Table 3

559

Base Case Model Parameters

Die material Die size Die thickness Die power Attachment material Face-up attach thickness Flip-chip bump height Solder bump conductivity Underfill conductivity Solder bump pitch Substrate materials

Substrate size Substrate thickness Cooling plane temperature

GaAs (45 W/m K) 10 mm ⫻ 10 mm 404.6 ␮m 40 W (uniform) 95 Pb/Sn solder (35 W/m K) 25.4 ␮m 101.6 ␮m 44 W/m K (150 ␮m diameter) 2 W/m K (high k filler) 357 ␮m Diamond (1200 W/m K) Cu (400 W/m K) Si (148 W/m K) 15 mm ⫻ 15 mm 404.6 ␮m 50⬚C

Source: Ref. 82.

through the die is kept small by the short conduction path and the edgecooled package resistance is dominated by the substrate. Although heat is not forced through the die in a flip-chip package, it is known that heat would flow through the higher thermal conductivity solder bumps to the substrate rather than through the underfill. The largest temperature difference was observed for an edge-cooled substrate, which resulted in the Tj of a silicon die being 13⬚C cooler than that of a GaAs die. b. Die Width. Simulation studies utilizing dies of different widths showed that while a bottom-cooled package saw a constant thermal conduction path and a proportional increase in heating and cooling loads, the edge-cooled package experienced an increasing thermal conduction path and only a linear increase in cooling area. Thus, the bottom-cooled package performance remained almost constant over a range of die widths, whereas the edge-cooled package experienced significant increases in Tj ⫺ Tc , which became more pronounced as substrate conductivity decreased. Figure 24 shows the flip-chip model results, which are very similar to the results for the face-up bonded case. If Tj = 100⬚C is defined as the die temperature at which device failure occurs, then a diamond substrate would be the only edge-cooled package that would survive over the full range of die widths considered.

560

Malshe and Brown

Figure 24 Maximum package temperature differential versus die width for flipchip die attachment. E-C, edge-cooled; B-C, bottom cooled. (From Ref. 82.)

c. Die Thickness. The effect of die thickness variation on the overall package performance is negligible when the conduction paths are small. Increasing the die thickness does increase the normal die resistance, but it simultaneously reduces the lateral resistance. Thus, in the face-up bonding case, the rise in the temperatures (Tj ⫺ Tc) is found to increase with increasing die thickness. In the flip-chip configuration, because it is the lateral resistance that is critical, cooler junction temperatures are achieved with increasing the die thickness. d. Die Power. Over the heat dissipation range of 20 to 80 W, which represents power dissipation levels in present-day and near-future highspeed, high-power electronics, all packages exhibit a linear increase in (Tj ⫺ Tc). Two approaches are commonly used for the cooling schemes: the bottom-cooled and the edge-cooled approach. Despite the temperature rise, a bottom-cooled substrate package remains at safe levels due to the short thermal paths. The edge-cooled diamond substrate is found to handle about 400% increase in power dissipation as compared with a Si substrate [84]. e. Flip-Chip and Face-Up Die Attachment Comparison. Flip-chip die attachment has a thermal advantage over face-up die attachment because of the presence of two paths for heat conduction: through the substrate and through the top of the die. Simulation results [83] show that, as the heat sink resistance decreases, the (Tj ⫺ Tc) for an edge-cooled substrate decreases toward the values for a bottom-cooled substrate. A die heat sink is

Diamond Heat Spreaders and Thermal Management

561

essential for adequate cooling of an IC if a package uses a high-thermalresistance substrate without bottom cooling. However, using a diamond substrate eliminates the necessity for a die heat sink because the junction temperature would be maintained within safe operating limits. In other words, diamond is an extremely attractive substrate material for electronic packaging applications for thermal management of high-power-density ICs packaged into very small spaces.

III.

APPLICATIONS AND EXAMPLES OF VARIOUS PACKAGES

This section describes some of the passive electronic applications of CVD diamond as a thermal spreader reported in the published literature. A significant amount of progress has been made in the utilization of diamond in electronic packaging over the past few years, but challenges still remain. A.

Background

The U.S. Defense Advanced Research Projects Agency (DARPA) should be recognized for its continued interest in and financial support of research and development of synthetic diamond for electronic thermal management. There is no doubt that it would have been difficult, if not impossible, for many researchers to pursue diamond thermal management research at any level were it not for DARPA support. The following sections address the applications of diamond to electronic packaging. B.

Optoelectronics

Diamond finds applications in numerous optoelectronic applications. Conventional light-emitting diodes (LEDs) have a very high thermal resistance and a low light output. This makes them impractical for high-flux applications such as automotive tail lights. Work has been carried out at the Optoelectronics Division of the Hewlett Packard Company (San Jose, CA) in which LEDs were soldered to a diamond heat spreader mounted on a copper slug. The result is an enhanced ability to drive a high-light-flux LED, a larger LED chip at higher powers [90], resulting in a fivefold increase in the total light output per element. Table 4 compares the characteristics of conventional and high-light-flux LEDs. The enhanced performance achieved by HP may allow LEDs to compete on a cost per lumen basis with incandescent light sources in many applications.

562

Malshe and Brown

Table 4 LEDs

Comparison of Performance of Conventional and High-Light-Flux

Device Conventional LED High-light-flux LED

Thermal resistance (⬚C/W)

Light output (lumen)

Maximum drive power (W/cm2)

LED chip area (cm2)

300

1.8

250

0.00064

>316

0.00250

4

10

Source: Ref. 90.

From a manufacturing standpoint, processing precut, and generally small, diamond heat spreaders such as those used with LEDs proves to be difficult and expensive. A possible alternative is the use of relatively large diamond wafers after they are partially etched to define the edges of specific heat spreader sizes. Such wafers could then be broken into individual heat spreaders. This approach is reported [84] to have nearly 100% yield, is cost effective, and is convenient for postdeposition processing. Furthermore, gold-plated diamond heat spreaders have been successfully attached to silver-plated copper slugs using a thermocompression bonding technique [84]. This eliminates the need for an adhesive layer, which only adds to the thermal resistance of an electronic package. Diode lasers have already become an attractive choice for laser applications because of their low cost and high efficiency. Consequently, efforts continue to be directed at making diode laser arrays on a commercially viable scale that are reliable and cost effective. However, laser arrays result in an even greater need for efficient thermal management materials and cooling technologies. Not surprisingly, CVD diamond substrates are presently being used for thermal management of laser diode arrays. Sumitomo Electric USA Inc. has developed a composite diamond heat spreader material for laser diode applications [91]. C.

Three-Dimensional High Density Packaging

The multichip module approach to achieving very high packaging and interconnect densities in electronic packaging is of particular interest for digital system engineers. However, as the density of high-performance ICs in an MCM package increases, the power dissipation also increases, making thermal management a challenging task. Assuming that the cost of diamond can be reduced, it offers a promising solution to many difficult electronic

Diamond Heat Spreaders and Thermal Management

563

packaging challenges, as can be seen from the comparison of the properties of electronic packaging materials in Table 5 [84]. Figure 25 illustrates the ability of diamond to quickly transport heat away from a heat source, such as an IC. 1.

Diamond for 3-D MCM Applications

There are several advantages in going to three-dimensional electronic packaging, but the most important one is a significant reduction in signal path length between contiguous 2-D multichip modules, thereby increasing the operating speed of a system. In 3-D MCM packaging, use is made of the fact that IC chips are thin (typically 0.6 mm or less), permitting very high volume density packaging. Figure 26 illustrates that in a 2-D MCM there can be only eight nearest neighbor ICs for a given IC. However, in a 3-D MCM, the same IC would have 116 nearest neighbors within the same distance. Assuming that the MCMs are 1.5 cm square and continuing the example further, if the interconnect delay is 150 psec/in. and the VLSI chips used have 10,000 gates, then the 3-D geometry places over a million gates within a worst-case delay time of about 650 psec. If these ICs were spread out horizontally, the worst-case delay would be in excess of 1.7 nsec, which

Table 5

Comparison of Electronic Packaging Material Properties

Material Diamond (natural) Diamond (CVD) Beryllia, BeO AlN Alumina, 99% GaAs Silicon Kovar (FeCoNi) Molybdenum Aluminum Copper Silver Diamond-epoxy Silver-epoxy Polyimide Source: Ref. 84.

Thermal conductivity (W/m⬚C)

Electric insulator? (Y/N)

Thermal expansion (ppm/⬚C)

2000 700–1700 220 70–230 29 45 149 17 146 237 396 427 8.7 5.8 0.2

Y Y Y Y Y Semi N N N N N N Y N Y

0.8–1.0 1.0–1.5 6.4 3.3 6.3 5.9 2.6 5.9 5.1 23.8 16.8 19.6 120 120 >50

564

Malshe and Brown

Figure 25 Ref. 84.)

Diamond capabilities in reducing ‘‘hot spot’’ temperature rise. (From

Figure 26 Ref. 84.)

Comparison of 2-D and 3-D MCM signal interconnect lengths. (From

Diamond Heat Spreaders and Thermal Management

565

would represent a significant performance compromise for system clock rates in the 50–100 MHz range and virtually preclude operation at rates above 250–350 MHz. With 3-D packaging, this is extended to nearly 1 GHz. Diamond is attractive for 3-D MCM applications, consisting of vertically stacked 2-D multiple chip modules, primarily because of its extremely high thermal conductivity, a requirement for any substrate utilized in 3-D packaging of high-power ICs. Furthermore, technologies exist that will allow electrical interconnections to be made in the vertical direction (i.e., between the 2-D multichip modules). As shown in Fig. 27, conductive fuzz buttons can be used in spacer boards placed between the 2-D multichip modules to provide vertical interconnection. Such fuzz buttons are typically 0.020 in. in diameter and extend beyond both sides of the board by about 40 mils, thereby making solid contact with pads on the substrates above and below when a mating force is applied. Such an arrangement facilitates rework of defective modules. 2.

Edge Cooling

One possible approach to dealing with very high power density and the associated heat dissipation is to extend the diamond substrate and make it a part of the heat exchanger itself. As an example, if an active 100 mm ⫻ 100 mm MCM substrate is extended on the heat sink ends by 5 mm or so,

Figure 27

Cross section of 3-D diamond MCM assembly. (From Ref. 84.)

566

Malshe and Brown

then it is possible to perforate these extended ends with slots of 0.1 mm width and 2.5 mm length on a pitch of 0.25 mm to create a total surface area of about 50 cm2. This would reduce the power density from 250 W/ cm2 to about 50 W/cm2. This would make thermal management more effective and maintain the interface temperature at an acceptable level (see Fig. 28). The edge-cooling approach has been attempted experimentally in a joint effort by Sandia National Laboratories (Albuquerque, NM) and Allied Signal Corp. (Kansas City, MO) [87]. They have built and tested a single substrate intended for a 3-D edge-cooled MCM. Their results indicate that it is possible to achieve a thermal resistance as low as 0.22⬚C/W for a single substrate. The challenge is to achieve the same kind of thermal impedance for 3-D stacks of these substrates. D.

Power Packages

In recent years, there has been a trend toward higher power ICs while, at the same time, reducing their physical size. Typical power densities are now

Figure 28 84.)

Edge cooling technique utilizing extended cooling slots. (From Ref.

Diamond Heat Spreaders and Thermal Management

567

on the order of 200 W/cm2 and rising. Power electronics poses one of the most difficult problems for miniaturization, but it should be possible even considering the challenges of thermal management by taking advantage of multichip module technology called multichip power modules (MCPMs). Although the components themselves are relatively small, often requiring only a few electrical connections, the package size is often determined by thermal loads and the desired thermal performance of the complete system. Diamond is an ideal thermal management material for such applications. However, if a layer of diamond is placed between a heat-generating device and a copper heat sink, the thermal shear stress acting at the diamond-copper interface far exceeds the shear strength of diamond. To overcome this problem, a 200-␮m-thick aluminum (Al, 99.995%) layer has been placed between the diamond and copper and also between the diamond and device, as shown in Fig. 29 [88]. Pure Al is very soft compared with copper and exhibits a plastic stress-strain behavior. The aluminum layer is thus able to absorb the thermal stress in the module. Figure 30 shows the temperature difference between the coolant and device surface for a module with aluminum oxide and diamond [89], clearly depicting the improvement in thermal management by using CVD diamond substrates. Diamond substrates are being used by Sandia National Laboratories (Livermore, CA; Albuquerque, NM) for MCM packaging of power electronic subsystems of flight computers [89]. The flight simulator computer is built around a radiation-hardened microprocessor, nonvolatile memory, and analog driver circuits and has multiple responsibilities including (1) providing a digital communication interface to the airplane or missile, (2) performing signal processing as required, (3) systems status monitoring, and (4) actuation signals to other systems (see Fig. 31). The actuation signals are large-power, short-duration signals for activation of mechanical and pyro-

Figure 29 Schematic cross section of a power module with diamond layer for thermal management. (From Ref. 86.)

568

Malshe and Brown

Figure 30 Photograph of working diamond MCM (dark substrate) and alumina MCM in copper tungsten secondary package. (From Ref. 87.)

technical devices. Two typical activation signals are a 150 W, 12 msec pulse and a 32 W, 20 Hz square wave of 3 sec duration. As the total activation time is brief, thermal performance is dominated by the first level of packaging around the active and passive devices. Temperature measurements made as a comparative study using diamond and aluminum nitride (AlN) as the thermal management materials revealed that in order to achieve the same temperature difference, the surface area of AlN required was about four times larger than that of CVD diamond substrates [89,90]. The effectiveness

Figure 31

Schematic of flight computer block diagram. (From Ref. 87.)

Diamond Heat Spreaders and Thermal Management

569

of CVD diamond in thermal management applications involving a reduction of volume is thus obvious. The MCM version of the power electronics subsystem occupies a little more than half the volume of a conventionally packaged, through-hole PCB system. However, the shrinking of the power electronic subsystem does require the excellent thermal management capabilities of CVD diamond. In another approach, Vandersande et al. reported using thermoelectric coolers and diamond films for temperature control of power electronic circuits [93]. The idea is to mount the highest power components directly onto thin-film thermoelectric elements, which maintains the temperature of the components from a few degrees to a few tens of degrees below that of the diamond substrate. To maximize the efficiency of the thermoelectric cooler, diamond films acting as thermal lenses are used to spread the heat from the small power device to the larger coolers. However, thermoelectric coolers are not generally useful for temperature control of high-density electronic circuits. Passive cooling with high-thermal-conductivity CVD diamond films is a much more realistic method for cooling of electronic circuits above room temperature.

E.

Plastic Packages

Commercial, monolithic, microwave GaAs ICs (MMICs) are usually packaged in injection-molded plastic packages. When packaged in plastic, these devices are restricted to power dissipation level of 1–2 W. If the power dissipation exceeds these levels, more expensive ceramic packages are used. By integrating a diamond substrate into the plastic package, the thermal dissipation capabilities can be increased to about 15 W at about half the cost of a ceramic package. Thus, a combination of the superior material properties of CVD diamond and the economics of plastic packaging offers a potential solution for relatively inexpensive packaging of microwave ICs. Norton Diamond Film’s NorCool (NorCool is a trade mark of Norton Diamond Film) package combines the excellent thermal properties of CVD diamond with the low cost of plastic packages. The NorCool package replaces the die paddle of a typical copper lead frame with a diamond substrate. The diamond substrate is physically attached to all the leads, thereby providing tremendous improvement in thermal performance, yet keeping the leads electrically isolated from each other [92]. The major problem associated with the attachment of GaAs MMICs to diamond substrates could be the coefficient of thermal expansion mismatch, as the CTEs of GaAs and diamond, respectively, are 5.7 ⫻ ppm/⬚C and 1.4 ppm/⬚C.

570

F.

Malshe and Brown

Concluding Remarks

Recent success in the synthesis, processing, and application of CVD diamond has demonstrated its potential as one of the most promising thermal management candidates for challenging electronic applications. Significant efforts have resulted in optimization and commercialization of synthesis and processing technologies. Emphasis should continue to make CVD diamond synthesis and processing more cost effective as well as manufacturing transparent. Diamond films and substrates will continue to find more and versatile applications for the ever growing thermal management needs for miniaturization, increased operating speed, and cost reduction in electronic packaging.

G.

List of U.S. Patents Relevant to Diamond Heat Spreaders 5,472,370 Method of planarizing polycrystalline diamonds, planarized polycrystalline diamonds and products made therefrom 5,663,595 Diamond heat sink comprising synthetic diamond film 5,660,894 Process for depositing diamond by chemical vapor deposition 5,653,952 Process for synthesizing diamond using combustion method 5,650,639 Integrated circuit with diamond insulator 5,648,148 Thermal management of electronic components using synthetic diamond 5,647,964 Diamond film synthesizing apparatus and method thereof using direct current glow discharge plasma-enhanced chemical vapor deposition 5,645,617 Composite polycrystalline diamond compact with improved impact and thermal stability 5,642,779 Heat sink and a process for the production of the same 5,633,088 Diamond film and solid particle composite structure and methods for fabricating same 5,631,046 Method of metallizing a diamond substrate without using a refractory metal 5,628,824 High-growth-rate homoepitaxial diamond film deposition at high temperatures by microwave plasma-assisted chemical vapor deposition 5,624,719 Process for synthesizing diamond in a vapor phase 5,620,512 Diamond film growth from fullerene precursors 5,614,258 Process of diamond growth from C.sub.70

Diamond Heat Spreaders and Thermal Management

571

5,609,955 Diamond film and mixed diamond and nondiamond particle composite compositions 5,609,683 Apparatus for making industrial diamond 5,607,723 Method for making continuous thin diamond film 5,607,560 Diamond crystal forming method 5,567,986 Heat sink 5,560,779 Apparatus for synthesizing diamond films utilizing an arc plasma 5,545,030 Diamond film and solid nondiamond particle composite compositions 5,539,176 Method and apparatus of synthesizing diamond in vapor phase 5,538,765 DC plasma jet CVD method for producing diamond 5,535,816 Heat sink 5,531,184 Method for producing synthetic diamond thin film, the thin film and device using it 5,523,160 Highly oriented diamond film 5,523,121 Smooth surface CVD diamond films and method for producing same 5,518,759 High-growth-rate plasma diamond deposition process and method of controlling same 5,516,554 Cyclic hot-filament CVD of diamond 5,514,242 Method of forming a heat-sinked electronic component 5,510,157 Method of producing diamond of controlled quality 5,508,230 Method for making a semiconductor device with diamond heat dissipation layer 5,507,987 Method of making a freestanding diamond film with reduced bowing 5,506,452 Power semiconductor component with pressure contact 5,505,158 Apparatus and method for achieving growth-etch deposition of diamond using a chopped oxygen-acetylene flame 5,504,303 Laser finishing and measurement of diamond surface roughness 5,500,157 Method of shaping polycrystalline diamond 5,500,077 Method of polishing/flattening diamond 5,499,601 Method for vapor-phase synthesis of diamond 5,496,596 Methods for producing diamond materials with enhanced heat conductivity 5,495,126 Polycrystalline diamond heat sink having major surfaces electrically insulated from each other 5,492,770 Method and apparatus for vapor deposition of diamond film

572

Malshe and Brown

5,491,002 Multilayer CVD diamond films 5,490,002 Multilayer CVD diamond films 5,490,963 Preparation of thin freestanding diamond films 5,488,350 Diamond film structures and methods related to same 5,488,232 Oriented diamond film structures on nondiamond substrates 5,487,945 Diamond films on nondiamond substrates 5,485,804 Enhanced chemical vapor deposition of diamond and related materials 5,480,686 Process and apparatus for chemical vapor deposition of diamond films using water-based plasma discharges 5,479,875 Formation of highly oriented diamond film 5,479,874 CVD diamond production using preheating

ACKNOWLEDGMENTS The authors would like to acknowledge Defense Advanced Research Projects Agency (DARPA) for their financial support. Discussions with Dr. Gordon, Dr. Naseem, and Dr. Schaper at High Density Electronics Center (HiDEC), University of Arkansas, Fayetteville, have been very helpful at many stages of research and are sincerely appreciated. Finally, the authors appreciate the efforts of Dr. Railkar of Materials and Manufacturing Research laboratory (MRL) at the University of Arkansas, Fayetteville in compiling this chapter.

REFERENCES 1. 2. 3.

4. 5.

6.

M Seal. Passive electronic applications. Diamond Relat Mater 1:1075–1081, 1992. AP Malshe, S Jamil, MH Gordon, HA Naseem, WD Brown, LW Schaper. Diamonds for MCMs. Adv Packag September/October:50, 1995. EF Borchett, G Lu. Applications of diamond made by chemical vapor deposition for semiconductor laser submounts. Proceedings of SPIE OE/LASE, Los Angeles, January 1993. TJ Moravec, A Partha. Adv Packag October:8, 1993. TJ Moravec, RC Eden, DA Schaefer. The use of diamond substrates for implementing 3-D MCM’s. Proceedings of 1993 International Conference and Exhibit and MCM’s, 1993, pp 85–91. CJ Chu, MP D’Evelyn, RH Hauge, JL Margrave. Mechanism of diamond growth by chemical vapor deposition on diamond (100), (111) and (110) surfaces: carbon-13 studies. J Appl Phys 70:1695, 1991.

Diamond Heat Spreaders and Thermal Management 7. 8. 9. 10.

11.

12. 13. 14.

15.

16. 17.

18. 19. 20.

21.

22.

23. 24.

573

PK Bachmann, D Leers, H Lydtin. Towards a general concept of diamond chemical vapor deposition. Diamond Relat Mater 1:1, 1991. A Inspektor, Y Liou, T McKenna, R Meisser. Plasma CVD diamond deposition in C-H-O systems. Surf Coat Technol 39/40:211, 1989. Y Liou, R Weimer, D Knight, R Messier. Effect of oxygen in diamond deposition at low substrate temperatures. Appl Phys Lett 56:437, 1990. RA Rudder, JB Posthill, GC Hudson, DP Malta, RE Thomas, RJ Markunas, TP Humphreys, RJ Nemanich. Chemical vapor deposition of diamond films using water:alcohol:organic-acid solutions. Mater Res Soc Proc 242:23, 1992. PK Bachmann. General aspects of CVD growth of diamond and their correlation. In: G Davies, ed. Properties and Growth of Diamond. Data Reviews Series No 9. London: INSPEC, 1994, p 364. S Matsumuto, Y Sato, M Tsatsumi, N Sekata. Growth of diamond particles from methane hydrogen gas. J Mater Sci 17:3106, 1982. AR Badzian. Adv X-Ray Anal 31:113, 1988. TR Anthony. Methods of diamond making. In: RE Clausing, LL Horton, JC Agnus, P Koidl, eds. Diamond and Diamond-like Films and Coatings. NATO Advanced Study Institute on Diamond and Diamond-like Films. New York: Plenum, 1991, p 555. M Tomellini, R Polini, V Sessa. A model kinetics for nucleation at a solid surface with application to diamond deposition from a gas phase. J Appl Phys 70:7573, 1991. J Robertson, J Gerber, S Sattel, M Weiler, K Jung, H Ehrhardt. Mechanisms of bias-enhanced nucleation of diamond on Si. Appl Phys Lett 66:3287, 1995. M Tsuda, M Nakajima, S Oikawa. The importance of the positively charged surface for the epitaxial growth of diamonds at low pressure. Jpn J Appl Phys 26:L527, 1989. B Lewis, JC Anderson. Nucleation and Growth of Thin Films. New York: Academic Press, 1978. S Soderberg, A Gerendas, M Sjostrand. Factors influencing the adhesion of diamond coatings on cutting tools. Vacuum 41:1317, 1990. J Rankin, RE Boekenhauer, R Csencsits, Y Shigesato, MW Jacobson, BW Sheldon. Nucleation and growth during the chemical vapor deposition of diamond on SiO2 substrates. J Mater Res 9:2164, 1994. RA Bauer, NM Sbrockey, WE Brower Jr. Quantitative nucleation and growth studies of PACVD diamond film formation on (100) silicon. J Mater Res 8: 2858, 1993. M Silviera, M Becucci, E Castelluci, FP Mattiot, V Barbarossa, R Tomaciello, F Galluzzi. Non-diamond carbon phases in plasma-assisted deposition of crystalline diamond films: a Raman study. Diamond Relat Mat 2:1257, 1993. T Takarada, H Takezawa, N Nakagawa, K Kato. Diamond synthesis on substrates treated by collisions with particles. Diamond Relat Mater 2:323, 1993. EJ Bienk, SS Eskildsen. The effect of surface preparation on the nucleation of diamond on silicon. Diamond Relat Mater 2:432, 1993.

574 25. 26.

27.

28. 29.

30.

31.

32. 33.

34. 35. 36. 37.

38. 39.

40. 41.

42.

Malshe and Brown H Maeda, S Ikari, S Masuda, K Kusakabe, S Morooka. Effect of surface pretreatment on diamond deposition. Diamond Relat Mater 2:758, 1993. R Steiner, G Stingeder, M Grasserbauer, R Haubner, B Lux. Investigation of surface preparation for diamond deposition on molybdenum substrates by secondary ion mass spectroscopy. Diamond Relat Mater 2:958, 1993. Y Chakk, R Berner, A Hoffman. Enhancement of diamond nucleation by ultrasonic substrate abrasion with a mixture of metal and diamond particles. Appl Phys Lett 66:2819, 1995. S Yugo, T Kania, T Kimura, T Muto. Generation of diamond nuclei by electric field in plasma chemical vapor deposition. Appl Phys Lett 58:1036, 1991. X Jiang, K Schiffmann, A Wetphal, CP Klages. Atomic force microscopic study of heteroepitaxial diamond nucleation on 具100典 silicon. Appl Phys Lett 63: 1203, 1993. X Jiang, CP Klages, R Zachai, M Hartweg, HJ Fusser. Deposition and characterization of diamond epitaxial thin films on silicon substrates. Appl Phys Lett 62:3438, 1993. SD Wolter, BR Stoner, JT Glass, PJ Ellis, DS Buhaenko, CE Jenkins, P Southworth. Textured growth of diamond on silicon via in-situ carburization and bias enhanced nucleation. Appl Phys Lett 62:1215, 1993. K Kobayashi, N Mutsukura, Y Machi. Catalyst effect for diamond nucleation in the low pressure process. Mater Manuf Processes 7:395, 1992. Y Nakamura, K Tamaki, Y Watanabe, S Hirayama. Diamond nucleation by carbon fibers on unscratched substrate by hot-filament chemical vapor deposition. J Mater Res 9:1619, 1994. RJ Meilunas, RPH Chang, S Liu, MM Kappes. Nucleation of diamond films on surfaces using diamond clusters. Appl Phys Lett 59:3461, 1991. JJ Dubray, CG Pantano, WA Yarbrough. Graphite as a substrate for diamond growth. J Appl Phys 72:3136, 1992. PN Barnes, RLC Wu. Nucleation enhancement of diamond with amorphous films. Appl Phys Lett 62:37, 1993. AP Malshe, A Khanolkar, WD Brown, SN Yedave, HA Naseem. Electrostatic seeding (ESS): a novel approach to seeding substrates for CVD diamond growth. Proceedings of International Meeting of the Electrochemical Society, Paris, August 31–September 5, 1997. J-M Ting, ML Lake. Polycrystalline diamond fibers for metal-matrix composites. JOM March:23, 1994. MJ Ulczynski, DK Reinhard, M Prystajko, J Asmussen. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, p 41. HL Davidson, JA Kerns, D Makowiecki, NJ Colella. ISHM Advanced Technology Workshop, Beaver Creek, CO, 1996. PG Patridge, G Lu, P May, JW Steeds. Potential high-strength high thermal conductivity metal matrix composites based on diamond fibers. Diamond Relat Mater 4:848–851, 1995. CH Stoessel, C Pan, JC Withers, D Wallace, RO Loutfy. High thermal conductivity graphite copper composites with diamond coatings for thermal

Diamond Heat Spreaders and Thermal Management

43. 44. 45.

46. 47.

48. 49.

50. 51. 52. 53. 54. 55. 56. 57. 58. 59.

60. 61. 62. 63. 64. 65.

575

management packaging applications. Mater Res Soc Symp Proc 390:147, 1995. L Yan, A Haji-Sheikh, JV Beck. Thermal characteristics of two-layered bodies with embedded thin film heat source. J Electron Packaging 115:276, 1993. A Feldman, C Beetz, P Klocek, G Lu. J Res Natl Inst Stand Technol 98:375, 1993. A Feldmane. Round robin thermal conductivity measurements on CVD diamond. In: A Feldman, Y Tzeng, WA Yarborough, M Yoshikawa, M Murakawa, eds. Applications of Diamond Films and Related Materials: Third International Conference, 1995, p 627. LY Andrew, HC Ong, LM Stewart, RPH Chang. J Mater Res 12:1796, 1997. AP Malshe, HA Naseem, WD Brown, LW Schaper. Recent advances in diamond based multichip modules. In: A Feldman, Y Tzeng, WA Yarborough, M Yoshikawa, M Murakawa, eds. Applications of Diamond Films and Related Materials: Third International Conference, 1995, p 611. WE Wesolowski, RJ DeKenipp, MPR Panicker. Proceedings of International Conference on Multichip Modules—ICEMCM’95, 1995, p 146. S Jamil, MH Gordon, GJ Salamo, HA Naseem, WD Brown, AP Malshe. Proceedings of Applied Diamond Conference 1995, Gaithersburg, MD, August 1995. DG Bhat, DG Johnson, AP Malshe, HA Naseem, WD Brown, LW Schaper, CH Shen. Diamond Relat Mater 4:921, 1995. B Watermeyer. Diamond Cutting. 2nd ed. Hohannesburg: Centour, 1982. P Grodzinski. Diamond Technology. London: NAG Press, 1953. M Yoshikawa. SPIE Diamond Opt III 210, 1990. H Tokura, M Yoshikawa. Applications of Diamond Films and Related Materials: Third International Conference, 1995, p 253. WS Lee, YJ Baik, KY Eyn. Diamond Relat Mater 4:989, 1995. C-F Yang, H Okura, M Yoshikawa. In: S Saito, O Fukunaga, M Yoshikawa, eds. Science and Technology of New Diamond. 1990, p 135. S Jin, JE Graebner, TH Tiefel, GW Kammlott, GJ Zydzik. Diamond Relat Mater 1:949, 1992. S Jin, W Zhu, JE Graebner. Presented at Applications of Diamond Films and Related Materials: Third International Conference, 1995, p 209. GS Raju. Chemical assisted mechanical polishing and planarization of CVD diamond substrate for MCM applications. MSEE thesis, University of Arkansas Library, Fayetteville, 1994. S Tezuka, M Yoshikawa. New Diamond 17:36, 1990. M Rochechild, C Arnone, DJ Ehrlich. J Vac Sci Technol B4:310, 1986. AM Ozkan, AP Malshe, WD Brown. Diamond Relat Mater 6:1789, 1997. T Chien, C Cutshaw, C Tanger, Y Tzeng. Applications of Diamond and Related Materials: Third International Conference, 1995, p 257. U Bogli, A Blatter, SM Pimenov, AA Smolin, VI Konov. Diamond Relat Mater 1:782, 1992. KV Ravi, VG Zarifis. Laser polishing of Diamond, Proceedings of the Third

576

66. 67. 68. 69. 70. 71.

72. 73. 74. 75. 76. 77. 78.

79. 80. 81. 82.

83. 84. 85. 86. 87.

88. 89.

Malshe and Brown International Symposium on Diamond Materials: The Electrochemical Society, Honolulu, 1993. P Tosin, W Luthy, HP Weber. Applications of Diamond Films and Related Materials: Third International Conference, 1995, p 271. AP Malshe, BS Park, WD Brown. ICMC-TF’97, San Diego, CA, 1997. DF Grogan, T Zhao, BG Bovard, HA MacLeod. Appl Opt 31:1483, 1992. GS Sandhu, WK Chu. Appl Phys Lett 55:437, 1989. AP Malshe, WD Brown, HA Naseem, LW Schaper. US Patent 5,472,370, December 5, 1995. MR Pawar, WD Brown, AP Malshe, HA Naseem, SS Ang, RK Ulrich. Planarization of large area CVD-diamond films using polyimide. Int J Microcircuits Electron Packag 19:316, 1996. WD Brown, HA Naseem, AP Malshe, JH Glezen, WD Hinshaw. Mater Res Soc Symp Proc 391:59, 1995. C Iacovangelo, E Jerabek. Int J Microcircuits Electron Packag 17, 1994. L Gilbert. NASA Tech Briefs June:63, 1997. J Glezen. Gold electroplating of CVD diamond. MS thesis, University of Arkansas, Fayetteville, 1995. CT Wang, D Dinella, SS Wagner. Trends in high density plating. Plat Surf Finish 15(4):75, 1993. KD Baker. US Patent 3,776,822, 1973. LM Napolitano Jr, E Meeks, MR Daily, D Miller, DP Norwood, DW Peterson, CA Reber, JE Robles, W Worobey, D Johnson. Int J Microcircuits Electron Packag 16:271, 1993. CD Iacovangelo, EC Jerabek, RH Wilson. GE Patent Appl RD23427, accepted 1993. CD Iacovangelo, PJ DiConza, EC Jerabek, KP Zarnoch. Mater Res Soc Symp Proc 337:401, 1994. LF Miller. IBM J Res Dev 13:239, 1969. MD Brown, AP Malshe, MH Gordon, RK Ulrich, WF Schmidt, WD Brown. ISPS’97, sponsored by International Microelectronics and Packaging Society, San Diego, December 2–5, 1997. MD Brown, AP Malshe, MH Gordon, RK Ulrich, WF Schmidt, WD Brown. Paper submitted to 3M. DA Schaefer, RC Eden, TJ Moravec. Proceedings of the Technical Program, Anaheim, CA, Vol 3, February 1993, p 1177. DW Peterson, JN Sweet, DD Andaleon, RF Renzi, DR Johnson. ISHM Third International Conference on MCM’s, 1994. B Fiegl, M Hibler, W Kiffe, F Koch, R Kuhnert, R Messer, H Schwarzbauer. Hybrid Circuits 35:15, 1994. LM Napolitano Jr, MR Daily, E Meeks, D Miller, DP Norwood, DW Peterson, CA Reber, JE Robles, W Worobey. ICEMM Proceedings—International Conference and Exhibition on Multichip Modules, Denver, April 1993, p 92. JW Vandersande, R Ewell, J-P Fleurial, HB Lyon. Applications of Diamond films and Related Materials: Third International Conference, 1995, p 623. M Gomes-Cassers, PM Fahis. IEEE MTT-S Digest, 1996, p 227.

Diamond Heat Spreaders and Thermal Management 90.

91. 92.

93. 94.

95.

577

JA Carey, D Collins, BP Loh. Abstract from the IMAPS Advanced Technology Workshop on Thermal Spreaders and Heat Sinks, Beaver Creek, CO, April 1997. M Nagai, K Tanabe, Y Kumazawa, N Fujimori. Presented at the ISHM-97 Conference, Philadelphia, 1997. J Crumpton, R Koba, K Valenta, B Speck, C Roach, D Kuty, R Keusseyan. Abstract submitted to the IMAPS Advanced Technology Workshop on Thermal Spreaders and Heat Sinks, Beaver Creek, CO, April 1997. D Tilton. Electron Packaging Prod August:71, 1995. A Feldman, HPR Frederikse. Proceedings of Second International Conference on the Applications of Diamond Films and Related Materials, Omiya Saitama, Japan, August 25–27, 1993, p 261. AP Malshe, BS Park, WD Brown, A Muyshondt, HA Naseem, Diamond Relat Mater 8:1198, 1999.

12 Diamond Active Electronic Devices Hiromu Shiomi Sumitomo Electric Industries Ltd., Itami, Japan

I.

INTRODUCTION

Diamond is a promising material for electronic and optical devices under severe conditions because it has a large electron energy gap. Diamond is also an attractive material for high-frequency and high-power devices due to its high breakdown field, high electron and hole mobilities, and low dielectric constant. Figures of merit for materials are useful to compare diamond and other semiconductors and to consider the advantages of different materials for certain types of electronic applications. The author shows three figures of merit, Johnson’s figure of merit, Keyes’ figure of merit, and Baliga’s figure of merit, and discusses the performance of diamond devices. Johnson’s figure of merit [1] is usually used to compare material properties for power microwave applications. It is given by JFM =

冉 冊 EMvs 2␲

2

(1)

where vs is the scattering limited saturated velocity of carriers in the semiconducting material and EM is the peak electric field strength at breakdown. Diamond’s high electron saturation velocity (vs = 2.8 ⫻ 107 cm/sec at room temperature) [2] combined with a breakdown field EM in excess of 107 V/ cm [3] results in a high Johnson figure of merit, JFMdiamond/JFMSi = 2601. Keyes’ figure [4] of merit is an indication of the thermal limitation on a material’s high-frequency electrical performance. It is given by KFM = ␭

冉 冊 cvs 4␲␧r

1/2

(2) 579

580

Shiomi

where c is the velocity of light, ␧r is the dielectric constant, and ␭ is the thermal conductivity. Diamond’s high electron saturation velocity and high thermal conductivity make Keyes’ figure of merit high, for example, KFMdiamond/KFMSi = 32. Baliga et al. [5,6] proposed a practical figure of merit for power devices. Although higher values of JFM and KFM are desirable for a good power semiconductor switch, they are not sufficient criteria. From a practical point of view, the drift region resistance is an important parameter that needs to be minimized for increasing the power-handling capability of a semiconductor device. Baliga’s figure of merit is based on the parameter ␴A, which can be interpreted as the on-state drift region conductance per unit area of an abrupt one-sided p-n junction, ␴A was given by Shenai et al. [5],

␴A =

q␮NB WM

(3)

where ␮ is the low-field carrier mobility, NB is the drift region doping density, WM is the depletion layer depth at breakdown, and q is the electronic charge. This equation was later corrected for semiconductors with deep donors or acceptors by Collins [7],

␴A =

q␮nc WM

(4)

where nc is the concentration of charge carriers. Assuming an abrupt one-sided p-n junction on uniformly doped material, the drift region doping density can be related to avalanche breakdown as NB =

␧sE2M 2qVB

(5)

where ␧s is the permitivity of the semiconductor and EM is the peak electric field strength at breakdown. The depletion layer depth at breakdown, WM, is 2VB/EM. For the calculation of ␴A in Eq. (4), nc = NB for Si and GaAs at or above room temperature. However, this is not the case for diamond. The ionization energy of boron acceptors in diamond is 0.37 eV [8], and only about 0.2% of the acceptors are ionized at room temperature in a typical natural semiconducting diamond [9]. The following expression must be used to determine the hole concentration p: p(p ⫹ ND) = (NA ⫺ ND ⫺ p)

冉

冊

2␲m*kT h2

3/2

exp

冉 冊 ⫺EA kT

(6)

where NA is the boron concentration, ND is the donor concentration, and m* is the hole density-of-states effective mass [9].

Diamond Active Electronic Devices

581

The author calculated the temperature dependence of ␴A as already described for a one-sided junction made of diamond, Si, GaAs, and 6H-SiC, as shown in Fig. 1, using the parameters after Collins [7]. Collins used an activation energy of 0.37 eV and a low value for the mobility of p-type diamond, 290 cm2 V⫺1 sec⫺1, which was the typical value for homoepitaxial diamond films. For these parameters, ␴A is larger for diamond than for GaAs or Si at temperatures greater than about 180⬚C because of the deep level of the acceptor. However, the actual carrier concentrations are expected to be more because the energy level EA decreases as the doping concentration increases [10–12]. In the temperature range investigated [10], the activation energy changed from 0.2 to 0.37 eV due to the different acceptor concentrations. After modification of the growth process, mobilities between 1000 and 1430 cm2 V⫺1 sec⫺1 were achieved in the improved chemical vapor deposition (CVD) films, which had B concentrations between ⬃5 ⫻ 1016 and 2.2 ⫻ 1018 cm⫺3 [13]. If the mobility is 1300 cm2 V⫺1 sec⫺1 with an activation energy of 0.37 eV, ␴A is the larger for diamond than for GaAs or

Figure 1 Conductance/cm2 calculated for one-sided diodes made from diamond, SiC, GaAs, and Si.

582

Shiomi

Si at temperatures greater than about 110⬚C. When the author uses an activation energy of 0.22 eV [10,14] and a mobility of 1300 cm2 V⫺1 sec⫺1, ␴A is larger for diamond than for GaAs or Si in the entire temperature range and for 6H-SiC at temperatures greater than about 70⬚C. It is concluded that diamond-based devices offer significant advantages in high-power applications. In the case of a high-frequency system, Trew et al. [15] examined the frequency performance of GaAs, silicon carbide, and diamond electronic devices, as shown in Fig. 2. Because frequency performance scales directly with saturation velocity, diamond devices are expected to operate at higher frequencies than GaAs or silicon carbide devices. Diamond MESFETs are capable of producing over 200 W of X-band power as compared with about 8 W for GaAs and about 60 W for silicon carbide. Trew’s group [16] further analyzed the output performance of diamond MESFETs and showed its

Figure 2 RF power performance versus frequency for diamond, SiC, and GaAs MESFETs. (From Ref. 15.)

Diamond Active Electronic Devices

583

temperature dependence. Power performance of the diamond MESFETs exceeds that of GaAs MESFETs when operated at temperatures higher than 177⬚C. The author speculates that the temperature limit should be lower, following a discussion similar to the earlier discussion of on-state drift region conductance.

II.

DIAMOND TRANSISTORS

A.

Introduction

Table 1 shows the pioneering works on diamond transistors. The author categorizes the history of the development of diamond transistors into three periods: 1982–1988 for preliminary bulk transistors, 1989–1992 for preliminary field-effect transistors, and 1993– for improved FETs and circuits.

Table 1

Historical Development of Diamond Transistors

Year

Topic

1982 1987

Bipolar transistor on natural diamond Point-contact transistor on HPHT diamond Permeable base transistor on HPHT diamond First MESFET on CVD diamond films First MESFET on CVD diamond films MiSFET with CVD intrinsic diamond MOSFET with SiO2 MESFET by rapid thermal processing MOSFET by the selective growth method MOSFET with ion implanted channel MiSFET on polycrystalline diamond films Gate-recessed MOSFET MOSFET on polycrystalline diamond films First diamond circuit MESFET using surface conductive layer Self aligned gate-recessed MESFET MOSFET on improved diamond films Pulse-doped (␦-doped) MESFET Normally-off diamond MESFET Transconductance exceeding 10 mS/mm

1988 1989 1990

1991

1992

1993 1994

1995 1997 1999

Researchers

Ref.

F. Prins W. Geis et al.

17 19

W. Geis et al.

20

Shiomi et al. Sh. Gildenblat et al. Shiomi et al. A. Grot et al. Tsai et al. Sh. Gildenblat et al.

22 25 26 27 30 28

R. Zeisse et al. Nishimura et al.

32 34

A. Grot et al. J. Tessmer et al.

29 37

Zeidler et al. Kawarada et al. Shiomi et al. L. Dreifus et al. Shiomi et al. J. Looi et al. Tsugawawa et al.

39 41 44 45 47 48 49

584

B.

Shiomi

1982–1988, Preliminary Bulk Transistors

The first active diamond electronic device was demonstrated in 1981 when Prins [17] reported bipolar transistor action in ion-implanted natural diamond. The device was fabricated by ion bombarding C⫹ ions onto the surface of a natural p-type diamond in two regions separated by 3.2 ␮m, as shown in Fig. 3a. The bombarded regions served as the n-type regions, although they were actually radiation-damaged regions. The common emitter current gain of the device is less than one, although the output I-V characteristics do show complete current saturation, as shown in Fig. 3b. In 1987, Tzeng et al. [18] reported a similar npn bipolar junction transistor, fabricated by ion implanting arsenic in a p-type natural diamond. The current gain obtained in the device was about 0.8. A point-contact transistor, operational at 510⬚C, was reported by Geis et al. [19]. This is the first report of diamond transistors that have power gain. Point-contact transistors were fabricated on a synthetic boron-doped diamond produced by a high-pressure, high-temperature (HPHT) process. From room temperature capacitance-voltage measurements of Schottky diodes on the diamond, it was determined that the net ionized boron acceptor concentration was on the order of 1016 cm⫺3. The base, collector, and emitter contacts consisted of tungsten probes that formed Schottky or ohmic contacts on the diamond surfaces, as shown in Fig. 4a. The collector current as a function of base-collector voltage is shown in Fig. 4b. When operated at room temperature, the small-signal current gain (the ratio of the collector to emitter currents) varied from 2 to 25 and the calculated small-signal power gain varied from 6 to 35, depending on the transistor and the operating conditions. At a high temperature, 510⬚C, the current gain typically varied from 1.6 to 0.5 and the power gain varied from 4.5 to 1.3, depending on the devices. In the following year, Geis et al. [20] reported the fabrication of a permeable base transistor. An ion-beam etching technique was used to fabricate the vertical structure of this device [21]. A schematic diagram of the device structure and I-V characteristics are shown in Fig. 5. A transconductance of 30 ␮S/mm was obtained, which was limited by a parasitic resistance (⬃104 ⍀ cm) in the substrate of the device, not an inherent material property nor the structure.

C.

1989–1992, Preliminary Field-Effect Transistors

In 1989, the first diamond MESFETs were fabricated independently by two groups on boron-doped homoepitaxial diamond films grown by the plasmaassisted chemical vapor deposition method [22–25]. Shiomi et al. [22] ep-

Diamond Active Electronic Devices

585

Figure 3 First active diamond electronic device, an npn bipolar junction transistor in natural p-type diamond. (a) Schematic illustration of the masking of the base area during carbon ion implantation. (b) I-V characteristics. (From Ref. 17.)

586

Shiomi

Figure 4 Diamond point-contact transistor. (a) Schematic drawing of the transistor. The ohmic contact was formed either by tungsten probes or by e-beam evaporated tungsten on the (111) facets of the diamond. (b) I-V characteristics. (From Ref. 19.)

itaxially grew diamond films on an HPHT diamond (100) surface in the selected condition of high CH4 concentration with B2H6 as dopant gas. Ti was used for the source and drain ohmic contacts and Al was used for the gate Schottky contacts. The gate length and width were large, 140 ␮m and 1.8 mm, respectively, because the MESFET patterns were fabricated directly with a shadow mask without using a photolithography technique. The device structure and the drain current versus drain voltage characteristics are shown in Fig. 6. The gate voltage was varied from ⫺1.0 to 5.0 V when the leakage current was less than 1 ␮A. The gate voltage controlled the drain current

Diamond Active Electronic Devices

587

Figure 5 Diamond permeable base transistor. (a) Device structure. To fabricate this device, ion beam-assisted etching was used to produce the grating in the diamond. (b) I-V characteristics. (From Ref. 20.)

588

Shiomi

Figure 6 First diamond MESFET. (a) Device structure. (b) I-V characteristics. (From Ref. 22.)

although the ohmic property of the source to drain was not optimized without annealing due to the fabrication process. Gildenblat et al. [25] fabricated a MESFET with a circular Al gate, as shown in the inset of Fig. 7. The leakage current remained under 50 nA at reverse biases up to 16 V. This structure shows some degree of gate control over the drain current ID. As can be seen in Fig. 7, a 6 V reverse gate bias results in a reduction of the drain current. Following the first fabrication of MESFETs, the two groups previously mentioned made attempts to improve the device performance by reducing the gate leakage with the incorporation of a gate insulator. A field-effect transistor with an insulated gate is generally called an insulated-gate field-

Diamond Active Electronic Devices

589

Figure 7 First diamond MESFET; device layout in the inset and I-V characteristics. (From Ref. 25.)

effect transistor (IGFET) or a metal-insulator-semiconductor field-effect transistor (MISFET). Shiomi et al. [26] inserted a 0.23-␮m-thick intrinsic diamond layer between the gate metal and boron-doped diamond layer. To specify this structure in Table 1, the author calls it a metal–intrinsic semiconductor–semiconductor (MiS) structure. The device structure and drain current versus drain voltage characteristics are shown in Fig. 8. The gate voltage was varied from 0 to 20 V in 5 V steps and the drain voltage was applied up to ⫺50 V. The drain current tends to saturate as the drain voltage increases. The intrinsic diamond layer withstood 70 V between gate and drain. The maximum transconductance was 2.0 ␮S/mm. Grot et al. [27,28] used 0.10-␮m-thick SiO2 as a gate insulator. It was formed by sputtering in an Ar (50%)–O2 (50%) plasma at 350⬚C. To specify this device structure in Table 1, the author calls it a metal-oxide-semiconductor (MOS) structure. During the fabrication of these devices, they used a selective growth process, where boron-doped homoepitaxial diamond was grown in the opening of sputtered SiO2. The hole mobility of selectively grown films varied between 210 and 290 cm2/V sec for hole concentrations between 1.0 ⫻ 1014 and 6.9 ⫻ 1014 cm⫺3. The device structure is shown in Fig. 9a. An Au (1.25 ␮m)/Ti (50 nm) metallization was deposited on top of the selectively grown homoepitaxial diamond island. The transistor characteristics at 300⬚C are shown in Fig. 9b. High-temperature operation of the device is made possible by incorporation of an insulating gate. The gate leakage current remained under 10 pA even during operation at 300⬚C. To

590

Shiomi

Figure 8 First diamond MiSFET with intrinsic diamond layer between gate metal and boron-doped diamond layer. (a) Device structure. (b) I-V characteristics, illustrating the tendency of drain current saturation at high positive gate voltage bias. (From Ref. 26.)

improve the transistor characteristics, Grot et al. [29] introduced the electron cyclotron resonance (ECR) plasma etching process and fabricated mesaisolated recessed gate field-effect transistors that operated at temperatures up to 350⬚C. The maximum transconductance was 87 ␮S/mm at 200⬚C, although they did not exhibit complete saturation or pinch-off of the channel. In 1990, Tsai et al. [30,31] reported on the fabrication of a diamond MESFET with the channel formed by solid-state diffusion. The device structure and the I-V characteristics are shown in Fig. 10. Boron was diffused into type IIa diamonds and simultaneously electrically activated by a rapid thermal annealing technique using a cubic boron nitride planar diffusion source in an argon atmosphere. The annealing took place at ⬃1370⬚C for 20 sec. Boron concentrations as high as 3.5 ⫻ 1019 cm⫺3 were detected at ˚ in the diamond substrate using secondary ion mass speca depth of 500 A

Diamond Active Electronic Devices

591

Figure 9 First diamond MOSFET with SiO2 as a gate insulator. (a) Device structure. (b) I-V characteristics at 300⬚C. (From Ref. 27.)

trometry (SIMS). Electrical contacts of Ti/Au were made on diamond via evaporation, and subsequent ohmic annealing was carried out for 30 min at 800⬚C. Finally another, unannealed Ti/Au bilayer 0.15 mm in length and 1 mm in width was deposited to form the Schottky gate. The MESFET was observed to pinch off at high positive gate bias, and reverse-bias leakage was below 10⫺12 A at 5 V. In addition to the boron doping processes already mentioned—the epitaxial growth process and the solid-state diffusion process—Zeisse et al. [32] introduced the ion-implantation process and fabricated MOSFETs. The device structure and the I-V characteristics are shown in Fig. 11. A p-type conducting layer was formed in a substrate of semi-insulating natural dia-

592

Shiomi

Figure 10 Diamond MESFET with ultrashallow boron-doped layer formed by rapid-thermal-processing (RTP) solid-state diffusion using cubic boron nitride as the dopant source. (a) Device structure with an ultrashallow p-doped channel less than ˚ at a boron concentration of about 3 ⫻ 1019 cm⫺3. (b) I-V characteristics, 500 A illustrating pinch-off behavior at high positive gate bias. (From Ref. 31.)

mond (type IIa) by boron implantation. Boron ions were implanted at 80 K using a multiple implant scheme (25 keV, 1.5 ⫻ 1014 B⫹/cm2; 50 keV, 2.1 ⫻ 1014 B⫹/cm2; and 100 keV, 3.0 ⫻ 1014 B⫹/cm2) intended to provide a uniform p-type layer 210 nm thick. After implantation, the diamond was annealed at 1263 K in nitrogen to remove the implantation damage and

> Figure 11 Diamond MOSFET fabricated by ion implanting boron into natural insulating type IIa diamond. (a) Plan (top) and cross section (bottom) views of an ion-implanted MOSFET. The thickness of the implanted layer is approximately 210 nm. The circular geometry does not require a mesa etch to contain the current flow from source to drain. (b) I-V characteristics at room temperature. The transconductance approaches 9.5 ␮S toward saturation for a change in gate voltage from 0 V to 2 V. (From Ref. 32.)

Diamond Active Electronic Devices

593

594

Shiomi

activate the implanted boron. The gate insulator, consisting of an SiO2 film approximately 100 nm thick, was deposited by indirect plasma-enhanced chemical vapor deposition [33] at a temperature of 300⬚C. Saturation and pinch-off were both observed at room temperature. The transconductance was 3.9 ␮S/mm and the output conductance was 60 nS/mm. This was the first reported use of ion implantation to fabricate a field-effect device successfully in diamond. The fabrication of active devices on polycrystalline diamond films is attractive because of the availability of large-area inexpensive substrate material. Nishimura et al. [34–36] first reported field-effect transistor action in polycrystalline diamond films. The device structure and the I-V characteristics are shown in Fig. 12. The maximum transconductance was 5 ␮S/mm. The device did not show saturation or pinch-off. Tessmer et al. [37,38] improved the polycrystalline diamond FETs and obtained a larger estimated maximum transconductance of 121 ␮S/mm, although the device did not show the pinch-off condition.

D.

1993– , Improved FETs and Circuits

Zeidler et al. [39] fabricated a simple demonstration circuit with a voltage gain of two using two MOSFETs. Natural type IIa diamonds were implanted with boron at 77 K, using the multiple scheme, low-temperature boron ion implantation procedure to produce an approximately uniformly doped p-type layer about 200 nm thick. This layer was used to fabricate MOSFETs operating in the depletion mode, with good uniformity from device to device. Two of the devices made the first diamond electric circuit, as shown in Fig. 13a. As indicated by the labels ‘‘DOT’’ and ‘‘RING’’ in Fig. 13a, the transistors were connected in such a way as to place a parasitic resistor across the power supply rather than across the output. The circuit was tested at room temperature under direct-current conditions. A gain (Vout/Vin) of approximately two was measured over a wide range of VDD, as seen in Fig. 13b. Kawarada et al. [40,41] fabricated MESFETs using a surface conductive layer of hydrogen-terminated undoped homoepitaxial diamond films. Hydrogen-terminated films display p-type conduction [10,42,43]. These FETs exhibited enhancement-mode effects and a high value of transconductance of 200 ␮S/mm (Fig. 14), although these operations should depend on the environment, which controls the surface conductive layer. Shiomi et al. [44] introduced a reactive ion etching process and fabricated MESFETs with a self-aligned recessed structure. The device structure and the I-V characteristics are shown in Fig. 15. Saturation and pinch-off characteristics were both observed in diamond MESFETs for the first time.

Diamond Active Electronic Devices

595

Figure 12 Polycrystalline diamond transistor. (a) Top and cross-sectional views of the MiSFET. (b) I-V characteristics of the MiSFET at room temperature. The inset shows the I-V characteristics between source and gate. (From Ref. 36.)

596

Shiomi

Figure 13 Diamond circuit. (a) Circuit configuration with parasitic resistances shown in dashed lines. (b) Voltage gain as a function of applied bias. (From Ref. 39.)

Diamond Active Electronic Devices

597

Figure 14 Diamond MESFET using a surface conductive layer of undoped homoepitaxial film. (a) Cross section of the MESFET. (b) I-V characteristics of the MESFET with a gate length of 10 ␮m. The transconductance obtained is 200 ␮S/mm. (From Ref. 41.)

598

Shiomi

Figure 15 Gate-recessed structure MESFET. (a) Cross section of MESFET. (b) IV characteristics of the MESFET with a gate length of 4 ␮m. (From Ref. 44.)

Dreifus et al. [13,45,46] demonstrated diamond MOSFETs that exhibited complete pinch-off of the active channel and a large transconductance on the order of a mS/mm at the elevated temperature. In this research, highquality homoepitaxial films of semiconducting diamond were deposited on IIa diamond substrates. The room temperature hole mobility exceeded 1400 cm2/V sec. In addition, compensation was measured as 3 ⫻ 1015 cm⫺3, which was similar to donor levels in the best natural type IIb single crystal diamonds. These films were used to fabricate MOSFETs. Pinch-off and saturation of the channel current were observed at temperatures up to 500⬚C. Figure 16 shows the I-V characteristics at 400⬚C. Diamond field-effect transistors were configured in analog-amplifier and a two-input NOR digital logic circuit. Successful logic operation has been observed up to 400⬚C. Shiomi et al. [47] have fabricated MESFETs utilizing a boron pulsedoped layer as a conducting channel. The layered structure of the MESFET, as shown in the inset of Fig. 17a, consists of an undoped 150-nm-thick buffer layer grown first, followed by a boron pulse-doped layer of 11 nm (the full

Diamond Active Electronic Devices

599

Figure 16 Diamond MOSFET with an improved epitaxial layer; I-V characteristics at 400⬚C. Gate-to-source voltage varies from 0 to 14 V in 2 V increments. (From Ref. 13.)

width of the half-maximum of the boron profile in Fig. 17a) and a 75-nmthick undoped diamond cap layer. These layers were grown by controlling the flow of the dopant gas during the microwave plasma-assisted chemical vapor deposition. The doping levels in the undoped layers and the boron pulse-doped layer were found to be <1017 and >1019 cm⫺3, respectively, as shown in Fig. 7a. The I-V characteristics in Fig. 7b showed that this fieldeffect transistor with a gate length of 4 ␮m and a gate width of 39 ␮m exhibited an extrinsic transconductance of 116 ␮S/mm with both pinch-off characteristics and current saturation. In 1997, Looi et al. reported the first normally off enhancement mode MESFET structures formed with a near-surface p-type hydrogenated diamond layer [48]. Such devices play an enabling role in circuit design. The transistors were formed on free-standing polycrystalline diamond grown by microwave plasma-enhanced CVD. Thermally evaporated gold was used for the ohmic contacts and aluminum was used for the gate metal. With no gate bias, the drain-to-source current was less than 1 nA, corresponding to an off-state. By applying negative gate voltage, an on-state was established. With a gate voltage of ⫺1.8 V, the drain-to-source current magnitude was 4.7 ␮A, with the gate current remaining less than 10 nA. Room temperature transconductance values were 0.14 mS/mm. On single crystalline diamond, a hydrogenated p-type layer can be utilized to form FET’s with even higher performance figures of merit. Tsugawa et al. formed both MESFET and MOSFET devices on homoepitaxial CVD diamond deposited on a synthetic Ib diamond (100) surface [49]. By

600

Shiomi

Figure 17 Pulse-doped diamond p-channel MESFET. (a) Depth profile of boron concentration measured by SIMS and cross-sectional drawing of a pulse-doped MESFET in the inset. (b) I-V characteristics. (From Ref. 47.)

appropriate selection of gate metal, both normally-off (enhancement mode) and normally-on (depletion mode) devices were achieved. In this case, lead was used as the gate metal for an enhancement mode MESFET and copper was used as the gate metal for a depletion mode MESFET. Transconductance values exceeded 10 mS/mm for both enhancement mode and depletion mode transistors.

Diamond Active Electronic Devices

III.

601

SUMMARY AND DIRECTION OF FUTURE WORK

A simple analysis based on the measured peak electric field strength at avalanche breakdown is presented to evaluate the performance of various semiconducting materials for high-power electronics. The calculated results indicate that diamond is superior material for fabricating power semiconductor devices. The high saturation velocity of diamond is shown to enable diamond to handle high power at high frequency. Moreover, the highest thermal conductivity and large energy gap of diamond offer an insulating layer which can be conveniently used for conducting the heat away from the active region of devices. The development of transistors is a benchmark for a new semiconductor material. It has been made clear by many researchers in the last several years that diamond is not just a simple gemstone but also a semiconducting material. However, researchers have now also demonstrated the feasibility of diamond active devices. For practical application, the author thinks that we need major breakthroughs in the following three techniques; the growth technique, the doping technique, and the microfabrication technique. In terms of the growth technique, the high-speed growth technique or the heteroepitaxial growth technique should be developed in the quest to obtain large single-crystal diamond wafers or films. For example, high-quality 6H-SiC, which is supplied as a wafer by the improved sublimation method [50] and is grown as a film by step-control epitaxy [51], has accelerated the development of SiC high-power devices [52–55]. High-quality GaN has been heteroepitaxially grown on sapphire [56,57], and soon after the brightest blue light–emitting diode (LED) was successfully fabricated [58]. These examples show that the production of high-quality wafers or films is the main breakthrough for newcomer semiconductors. In terms of the doping technique, the concentration and distribution of dopant should be well controlled. And n-type diamond with low resistivity should be obtained for a wide range of electrical applications. The recent progress in improving CVD diamond may enable the growth of n-type diamond. In terms of the microfabrication technique, the processing tools for diamond devices must be further developed to the point of fabricating reproducible devices. Recent progress shows that microfabrication techniques established for other semiconductors can be utilized for diamond. At present, it is not clear that diamond active devices have performance competitive with that of conventional materials. However, not a year passes without significant progress in the material and device quality of diamond transistors. In the next few years, high-power and high-frequency diamond devices may surpass the performance of devices fabricated in conventional

602

Shiomi

semiconductors and SiC. And the mentioned breakthroughs will open the way to the commercial use of diamond active devices. REFERENCES 1. 2. 3.

4. 5. 6. 7. 8. 9. 10. 11. 12.

13.

14. 15.

16.

17.

EO Johnson. Physical limitations on frequency and power parameters of transistors. RCA Rev 26:163, 1963. DK Ferry. High field transport in wide-bandgap semiconductors. Phys Rev B 12:2361, 1975. AV Bogdanov, IM Vikulin, TV Bogdanova. Investigation of microplasma breakdown at a contact between a metal and a semiconducting diamond. Sov Phys Semicond 16:720, 1982. RW Keyes. Physical limits in digital electronics. Proc IEEE 63:740, 1975. K Shenai, RS Scott, BJ Baliga. Optimum semiconductor for high-power electronics. IEEE Trans Electron Devices ED-36:1811, 1989. BJ Baliga. Power semiconductor device figure of merit for high-frequency applications. IEEE Electron Device Lett 10:455, 1989. AT Collins. Diamond electronic devices—can they outperform silicon or GaAs? Mater Sci Eng B11:257, 1992. AT Collins, WS Williams. The nature of the acceptor centre in semiconducting diamond. J Phys C 4:1789, 1970. JE Field. The Properties of Diamonds. New York: Academic Press, 1979. H Shiomi, Y Nishibayashi, N Fujimore. Characterization of boron-doped diamond epitaxial films. Jpn J Appl Phys 30:1363, 1991. JW Glesener. Hole capture in boron-doped diamond. Appl Phys Let 64:217, 1994. M Werner, O Dorsch, HU Baerwind, E Obermeier, L Haase, W Seifert A. Ringhandt, C Johnston, S Romani, H Bishop, and PR Chalker. Charge transport in heavily B-doped polycrystalline diamond films. Appl Phys Lett 64:595, 1994. JT Glass, DL Dreifus, RE Fauber, BA Fox, ML Hartsell, RB Henard, JS Holmes, D Malta, LS Plano, AJ Tessmer, GJ Tessmer, HA Wynands. High quality semiconducting diamond films and devices. Fourth International Conference on New Diamond Science and Technology, MYU, Kobe, Japan, 1994, p 335. N Fujimori, H Nakahata, T Imai. Properties of boron-doped epitaxial diamond films. Jpn J Appl Phys 29:824, 1990. RJ Trew, J Yan, PM Mock. The potential of diamond and silicon carbide electronic devices for microwave and millimeter-wave power applications. Proc IEEE, 79:598, 1991. MW Shin, RJ Trew, GL Bilbro. High temperature dc and RF performance of p-type diamond MESFET: comparison with n-type GaAs MESFET. IEEE Electron Device Lett 15:292, 1994. JF Prins. Bipolar transistor action in ion implanted diamond. Appl Phys Lett 41:950, 1982.

Diamond Active Electronic Devices 18.

19.

20. 21. 22. 23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

603

Y Tzeng, TH Lin, JL Davidson, LS Lan. Fabrication and high temperature characteristics of diamond electronic devices. Proceedings of Seventh Biennial University/Government/Industry Microelectronics Symposium, IEEE, Rochester, NY, 1987, p 187. MW Geis, DD Rathman, DJ Ehrlich, RA Murphy, WT Lindley. High-temperature point-contact transistors and Schottky diodes formed on synthetic borondoped diamond. IEEE Electron Device Lett EDL-8:341, 1987. MW Geis, NN Efremow, DD Rathman. Summary abstract: device applications of diamonds. J Vac Sci Technol A 6:1953, 1988. NN Efremow, MW Geis, DC Flanders, GA Lincoln, NP Economou. Ion-beam– assisted etching of diamond. J Vac Sci Technol B 3:416, 1985. H Shiomi, Y Nishibayashi, N Fujimore. Field-effect transistors using borondoped diamond epitaxial films. Jpn J Appl Phys 28:L2153, 1989. H Shiomi, Y Nishibayashi, N Fujimore. Diamond transistors. Fiftieth Fall Meeting of The Japan Society of Applied Physics and Related Society, Japan, 28a-E-10, 1989, p 384. N Fujimori, T Imai, H Nakahata, H Shiomi, Y Nishibayashi. Epitaxial growth of diamond and diamond devices. Proceedings of Fall 1989 MRS Meeting, Boston, Vol 162, 1989, p 23. GS Gildenblat, SA Grot, CW Hatfield, CR Wronski, AR Badzian, T Badzian, R Messier. Electrical properties of homoepitaxial diamond films. Proceedings of Fall 1989 MRS Meeting, Boston, Vol 162, 1989, p 297. H Shiomi, Y Nishibayashi, N Fujimori. Characterization of boron-doped diamond epitaxial films and applications for high-voltage Schottky diodes and MESFETs. Second International Conference of New Diamond Science and Technology, Material Research Society, Washington, DC, 1990, p 975. SA Grot, CW Hatfield, GS Gildenblat, AR Badzian, T Badzian. Semiconductor device development using selectively grown thin-film diamond. Second International Conference of New Diamond Science and Technology, Washington, DC, 1990, p 949. GS Gildenblat, SA Grot, AR Badzian. High-temperature thin-film diamond field-effect transistor fabricated using a selective growth method. IEEE Electron Device Lett EDL-12:37, 1991. SA Grot, GS Gildenblat, AR Badzian. Diamond thin-film recessed gate fieldeffect transistors fabricated by electron cyclotron resonance plasma etching. IEEE Electron Device Lett EDL-13:462, 1992. W Tsai, M Delfino, LY Ching, G Reynolds, D Hodul, CB Cooper III. Boron doping of diamond via solid state diffusion. Second International Conference of New Diamond Science and Technology, Material Research Society, Washington, DC, 1990, p 937. W Tsai, M Delfino, D Hodul, M Riaziat, LY Ching, G Reynolds, CB Cooper III. Diamond MESFET using ultrashallow RTP boron doping. IEEE Electron Device Lett EDL-12:157, 1991. CR Zeisse, CA Hewett, R Nguyen, JR Zeidler, RG Wilson. An ion-implanted diamond metal-insulator-semiconductor field-effect transistor. IEEE Electron Device Lett EDL-12:602, 1991.

604 33. 34.

35.

36.

37. 38.

39.

40.

41. 42. 43.

44.

45.

46.

47.

Shiomi LG Meiners. Indirect plasma deposition of silicon dioxide. J Vac Sci Technol 21:655, 1982. K Nishimura, R Kato, S Miyauchi, K Kobashi. Metal semiconductor field effect transistor with polycrystalline diamond films. Japan New Diamond Forum, Fifth Diamond Symposium, Tsukuba, Japan, 1991, p 34. K Nishimura, H Koyama K, Miyata, K Suzuki, K Kobashi. Fabrication of field effect transistors made of polycrystalline diamond films. Third International Symposium on Diamond Materials at the ECS Spring Meeting, Hawaii, 1993. K Nishimura, K Kumagai, R Nakamura, K Kobashi. Metal/intrinsic semiconductor/semiconductor field effect transistor fabricated from polycrystalline diamond films. J Appl Phys 76:8142, 1994. AJ Tessmer, K Das, DL Dreifus. Polycrystalline diamond field-effect transistors. Diamond Relat Mater 1:89, 1992. AJ Tessmer, LS Plano, DL Dreifus. High-temperature operation of polycrystalline diamond field-effect transistors. IEEE Electron Device Lett EDL-14:66, 1993. JR Zeidler, CA Hewett, R Nguyen, CR Zeisse, RG Wilson. A diamond driveractive load pair fabricated by ion implantation. Diamond Relat Mater 2:1341, 1993. N Nakamura, M Aoki, M Itoh, N Jin, H Kawarada. Fabrication of MESFET utilizing hydrogen-terminated homoepitaxial diamond films. Fourth International Conference on New Diamond Science and Technology, MYU, Kobe, Japan, 1994, p 729. H Kawarada, M Aoki, M Ito. Enhancement mode metal-semiconductor field effect transistors using homoepitaxial diamonds. Appl Phys Lett 65:1563, 1994. MI Landstrass, KV Ravi. Resistivity of chemical vapor deposited diamond films. Appl Phys Lett 55:975, 1989. H Nakahata, T Imai, N Fujimore. Change of resistance of diamond surface by reaction with hydrogen and oxygen. Second International Symposium on Diamond Materials, The Electrochemical Society, Washington, DC, Vol 91, 1991, p 487. H Shiomi, Y Nishibayashi, N Toda, S Shikata, N Fujimori. Fabrication of a gate-recessed-structure MESFET on diamond films. Fourth International Conference on New Diamond Science and Technology, MYU, Kobe, Japan, 1994, p 661. DL Dreifus, AJ Tessmer, JS Holmes, C-T Kao, DM Malta, LS Plano, BR Stoner. Diamond field-effect transistors. MRS Spring Meeting, San Francisco, 1994. BA Fox, ML Hartsell, DM Malta, HA Wynands, C Kao, LS Plano, GJ Tessmer, RB Henard, JS Holmes, AJ Tessmer, DL Dreifus. Diamond devices and electrical properties. Diamond Relat Mater 4:622, 1995. H Shiomi, Y Nishibayashi, N Toda, S Shikata. Pulse-doped diamond p-channel metal semiconductor field-effect transistor. IEEE Electron Device Lett EDL16:36, 1995.

Diamond Active Electronic Devices 48.

49.

50.

51. 52. 53. 54.

55.

56. 57. 58.

605

HJ Looi, LYS Pang, Y Wang, MD Whitfield, RB Jackman. High-performance metal-semiconductor field effect transistors from thin-film polycrystalline diamond. Diamond Relat Mater 7:565–568, 1998. K Tsugawa, K Kitatani, H Noda, A Hokazono, K Hirose, M Tajima, H Kawarada. High-peformance diamond surface channel field effect transistors and their operation mechanism. Diamond Relat Mater 8:927–933, 1999. RF Davis, JW Palmour, JA Edmond. Epitaxial thin film growth, characterization and device development in monocrystalline ␣- and ␤-silicon carbide. Diamond Relat Mater 1:109, 1992. H Matsunami. Progress in SiC epitaxy-present and future. Fifth SiC and Related Materials Conference. Washington, DC: IOP Publishing, 1993, p 45. LG Matus, JA Powell. High voltage 6H-SiC p-n junction diodes. Appl Phys Lett 59:1770, 1991. M Bhatnagar, PK McLarty, BJ Baliga. Silicon carbide high-voltage (400V) Schottky barrier diodes. IEEE Electron Device Lett EDL-13:501, 1992. T Kimoto, T Urushidani, S Kobayashi, H Matsunami. High-voltage (>1 kV) SiC Schottky barrier diodes with low on-resistances. IEEE Electron Device Lett EDL-14:548, 1993. CE Weitzel, JW Palmour, CH Carter Jr, KJ Nordquist. 4H-SiC MESFET with 2.8 W/mm power density at 1.8 GHz. IEEE Electron Device Lett EDL-15:406, 1994. S Nakamura. GaN growth using GaN buffer layer. Jpn J Appl Phys 30:L1705, 1991. S Nakamura, M Senoh, T Mukai. Highly p-typed Mg-doped GaN films grown with GaN buffer layers. Jpn J Appl Phys 30:L1708, 1991. S Nakamura, M Senoh, T Mukai. High-power InGaN/GaN double-heterostructure violet light emitting diodes. Appl Phys Lett 62:2390, 1993.

13 Diamond Film Optics D. K. Reinhard Michigan State University, East Lansing, Michigan

I.

INTRODUCTION

The optical properties are among the most interesting of the physical properties of diamond. These include the widest spectral optical transmission range of all known solid materials [1] and, for a transparent material, an unusually high index of refraction. Such attributes, combined with extreme hardness, high thermal conductivity and chemical resistance, make diamond a material of obvious interest for optical applications. However, optical applications of diamond have traditionally been very limited because of the notoriously high cost of quality diamond, even for samples of modest dimensions. In spite of this, early experiments brought convincing evidence of the potential of diamond for optical uses. A case in point is the diamond window used as the optical port for an infrared radiometer experiment on the Pioneer Venus probe launched in 1978. This 18.2 mm diameter by 2.8 mm thick diamond window survived earth atmosphere on launch, the cold and vacuum of space, and the Venus atmosphere (largely carbon dioxide with various acids, including sulfuric acid, at temperatures of 800 K and pressures of 90 atm) [1]. Other important demonstrations included the use of single-crystal diamond as high-power laser windows. By virtue of low optical absorption combined with high thermal conductivity, the windows successfully passed CO2 laser beams with continuous wave (CW) power densities in excess of 1 MW/cm2 [2]. Such experiments demonstrated unique optical capabilities for diamond; remaining as a significant barrier for more widespread use was the prohibitive economic cost. 607

608

Reinhard

The development of chemical vapor deposition (CVD) methods to synthesize diamond dramatically altered the issue of material availability. CVD technology has demonstrated the ability to create diamond structures of sizes that would be prohibitively expensive or simply not available in the traditional single-crystal form. For example, CVD has been used to form freestanding diamond structures in a variety of shapes including domes, tubes, and plates [3] as well as to form thin-film diamond coatings on a variety of substrates with diameters up to 20 cm and larger. Consequently, optical applications not previously feasible are now within consideration. Such applications include freestanding optical elements as well as coatings of lenses and other optical elements and devices. Nevertheless, there are several nontrivial issues associated with the application of CVD diamond for optical purposes. The intrinsic optical properties of diamond are based upon high-purity, single-crystalline samples. CVD samples, however, are generally polycrystalline and, depending on the deposition processes, may or may not contain significant defects and impurities. An additional concern with CVD material is mechanical strength and, of particular issue for thin-film applications, the integrity of a diamond coating in terms of film stress and adhesion to the substrate. This chapter is concerned primarily with the optical properties associated with polycrystalline CVD diamond and with related optical application issues of both thick films and thin films. The optical properties of singlecrystal diamond are first reviewed briefly in order to provide a frame of reference. Next, optical properties of freestanding polycrystalline films are considered, followed by a discussion of diamond film–coated substrates. Finally, electro-optic applications of diamond are briefly considered, including photodetection, photoconductive switching, and photon emission. The scope of the chapter is primarily concerned with the ultraviolet, visible, and infrared, with only passing reference to other portions of the electromagnetic spectrum. Regarding the illustrations and data presentation in this chapter, it is noted that the spectrally dependent optical properties of materials are variously presented as a function of wavelength, ␭, or as a function of wavenumber (that is, ␭⫺1, or inverse wavelength), or as a function of photon energy, E, which is related to the free-space wavelength as E = hc/␭, where h is Planck’s constant and c is the velocity of light in free space.* All three methods of data presentation are in common use in the literature and are used in this chapter.

*If the energy is expressed in electron volts and the free-space wavelength in micrometers, then E = 1.24/␭.

Diamond Film Optics

II.

OPTICAL PROPERTIES OF SINGLE-CRYSTAL DIAMOND

A.

Reflection and Absorption

609

Optical phenomena associated with single-crystal diamond have been a topic of study for hundreds of years. Such studies have been motivated in significant part by the fact that diamond gem commerce is based on the optical properties of naturally occurring single-crystal diamond as obtained from the earth. These properties have been extensively reviewed, for example, by Davies [4] and by Clark [5]. They are discussed here briefly as a prerequisite to the discussion of polycrystalline CVD diamond. Consider an optical beam of intensity I0 in air normally incident on the surface of a diamond sample that is characterized by an optical absorption coefficient ␣. The optical intensity at some depth d in the sample, neglecting reflections from the back of the sample, is I(d) = I0(1 ⫺ R)exp(⫺␣d)

(1)

where R is the reflection coefficient at the front surface given by R=

冉

na ⫺ nd na ⫹ nd

冊

2

(2)

na and nd being the refractive indices of air and diamond, respectively. The refractive index of diamond is given by the Sellmeier equation [6], n2d ⫺ 1 =

4.3356␭2 0.3306␭2 2 ⫹ 2 ␭ ⫺ (0.1060) ␭ ⫺ (0.1750)2 2

(3)

where ␭ is the wavelength expressed in ␮m. Figure 1 illustrates the wavelength dependence of the refractive index. There is clearly considerable dispersion in the visible portion of the spectrum, with nd ranging from near 2.47 in the violet to 2.40 in the red, a property that is also enhanced by the cut of diamond gems—light is refracted into a spectrum of colors when it passes out into the air and leads to the ‘‘fire’’ of diamond. Diamond, in fact, shows the largest color dispersion of the gems. The large refractive index of diamond causes a large normal incidence reflection value, approximately 17%, and also results in a small critical angle for total reflection, approximately 24 degrees.* Equation (3) is valid from the infrared to the onset of band-to-band absorption in the ultraviolet. The room-temperature diamond electronic band *This is taken advantage of by gem cutters. Facets in the so-called brilliant cut are arranged such that it is not possible for light to fall on the lower facets at lessor angles, so no light is refracted out the back.

610

Reinhard

Figure 1 Wavelength dependence of the refractive index according to the Sellmeier equation.

gap, or separation between the valance-band maximum and conduction-band minimum [5], is 5.47 eV (␭ ⬇ 0.23 ␮m), which effectively establishes an upper limit on photon energy for transmission.* Because the energy gap in diamond is indirect, photon excitation of an electron from the valence band to the conduction band must be accompanied by phonon absorption or emission in order to conserve momentum. For an indirect gap semiconductor with energy gap EG the absorption coefficient corresponding to transitions involving absorption of a photon of energy E and absorption of a phonon of energy EP is of the general form [7]

␣a =

A(E ⫺ EG ⫹ EP)2 exp(Ep/kT) ⫺ 1

(4)

for E > EG ⫺ EP and zero for E < EG ⫺ EP and where A is a slowly varying *According to Clark [5], the energy gap decreases at a rate of 5.4 ⫻ 10⫺5 eV/K.

Diamond Film Optics

611

function of photon energy. For the emission of a phonon the absorption coefficient is

␣e =

A(E ⫺ EG ⫺ EP)2 1 ⫺ exp(⫺Ep /kT)

(5)

for E > EG ⫹ EP and zero for E < EG ⫹ EP. When the photon energy is sufficient to cause phonon emission, ␣e dominates ␣a. The actual situation is somewhat more complex in that more than one type of phonon must be considered* and excitons can also play a role. Clear from the general theory, however, is the fact that for photon energies greater than the energy gap, the absorption increases sharply. Experimental results for diamond show that the absorption coefficient increases from approximately 15 cm⫺1 at 5.5 eV to 5000 cm⫺1 at 5.96 eV [4,5]. In addition to the fundamental absorption described by Eqs. (4) and (5), many semiconductors are observed experimentally to exhibit an exponential increase in absorption with energy near the band gap. The absorption coefficient associated with this observation is modeled empirically by Urbach’s rule according to the expression [8]

␣U = ␣0 exp

冋

册

␴(T)(E ⫺ E0) kT

(6)

where T is the temperature in kelvins and k is Boltzmann’s constant. Experimental observation of such absorption has been reported for high-purity single-crystal diamond. Thomas and Tropf [9] reported Urbach tail parameters at room temperature to be ␣0 = 4.23 ⫻ 1011 cm⫺1 and ␴ = 0.585 with E0 being the value of the direct energy gap in diamond, 6.5 eV. Figure 2 illustrates the wavelength dependence of the Urbach edge absorption coefficient which contributes an exponential absorption edge tail below the band gap. In addition to the Urbach tail, a temperature-independent weak absorption tail is experimentally observed that extends farther into the band gap, from approximately 0.23 to 0.35 ␮m. The magnitude of the weak tail varies somewhat from sample to sample [5] but on a representative high-purity single-crystal sample has been reported to be modeled as [9]

␣w = 0.28 exp(0.45E)

(7)

where E is in eV and the absorption coefficient is in cm⫺1. The weak tail absorption is primarily an ultraviolet phenomenon—the value of ␣w becomes on the order of 1 cm⫺1 by the blue portion of the visible spectrum. *Phonon energies at high-symmetry points in the Brillouin zone range in energy from 70 to 165 meV [4].

612

Reinhard

Figure 2 Wavelength dependence of the absorption coefficient near the band edge as calculated by an Urbach edge model.

For longer wavelengths, throughout the visible and into the infrared, where the absorption coefficient as modeled in Eqs. (4)–(7) becomes zero or vanishingly small, high-purity diamond should exhibit essentially no optical absorption until excitation of vibration modes of C — C bonds occurs in the range of approximately 2 to 7 ␮m. Diamond in the absence of significant defects or impurities has no infrared-active single-phonon absorption because diamond is a symmetric covalent material with no dipole moment to interact with the electric field of light. However, even in pure diamond, combinations of phonons can create weak dipole moments and two-phonon, three-phonon, and four-phonon infrared absorption peaks have been observed [9]. This multiphonon absorption is not strong; the absorption coefficient associated with the largest peak is about 12 cm⫺1. At higher wavelengths, the intrinsic absorption becomes negligible again.

Diamond Film Optics

613

In summary, pure diamond has an optical window spanning the ultraviolet, visible, infrared, millimeter, and microwave regions of the electromagnetic spectrum, with slight absorption in the infrared due to phonon absorption and an ultraviolet cutoff due to band-to-band optical excitation. These properties are referred to as the intrinsic optical properties of diamond, that is, those associated with carbon atoms arranged on a perfect diamond lattice. B.

Impurities and Diamond Types

Few natural single-crystal diamonds approach the intrinsic perfection. Impurities may give rise to additional electronic absorption levels and to vibration-induced absorption not found in pure diamond. Also, defects and impurities break the lattice symmetry, allowing one-phonon infrared absorption [10], which is forbidden in pure diamond. In fact, the effects of impurities on the optical properties of diamond are primarily responsible for the classification of natural diamond into four main types: Ia, Ib, IIa, and IIb [11]. Type I diamonds contain significant nitrogen impurities. Types Ia and Ib differ in that for Ib diamonds the nitrogen is present in substitutional sites and contributes to an electron paramagnetic resonance (EPR) signal. Type Ia is not EPR active, and at least in some cases the nitrogen forms aggregates or platelets within the crystal. Substitutional nitrogen gives rise to absorption in the higher energy portion of the visible spectrum, which causes the diamond to have a yellow appearance. When the nitrogen is not EPR active, the effect on color is less. For example, a type Ib diamond with 50 ppm atomic nitrogen has a pronounced yellow-gold color, whereas a type Ia diamond with 50 ppm atomic nitrogen may have only a slight tinge of color [11]. However, type Ia diamonds can have nitrogen concentrations as high as 2500 ppm atomic, and such crystals do show pronounced absorption in the visible portion of the spectrum. Type II diamond differs from type I in that type II does not contain appreciable nitrogen. Type IIa is ‘‘pure’’ diamond although purity is a matter of degree, varying from stone to stone. It has been suggested that a good specification for dividing between types I and II would be at the nitrogen level of 1 to 5 ppm atomic [11]. Type IIb is also nitrogen free but contains boron. Substitutional boron gives rise to strong infrared absorption that has a tail extending into the lower energy portion of the visible spectrum. As a result, a boron concentration of only 10 ppm produces a deep blue color in the crystal. Figure 3 illustrates comparative transmission plots for a colorless type IIa diamond, a pale yellow type Ia diamond, a gold-yellow type Ib diamond, and a blue boron-doped type IIb diamond. The type Ia in this figure is a natural stone; the others are single crystals grown synthetically

614

Reinhard

Figure 3 The single-crystal transmission spectra of (A) a type IIa colorless diamond, (B) a blue type IIb diamond, (C) a pale yellow type Ia diamond, and (D) a golden yellow type Ib diamond [11].

by the high pressure General Electric process [11]. Of note for electronic purposes, substitutional boron acts as an acceptor, introducing holes in the valence band such that type IIb diamond is semiconducting. All four types of diamond occur naturally, with the greatest preponderance being type Ia. In one study of natural diamond, 98% were type Ia, 0.1% were type Ib, and the remainder were type II.

Diamond Film Optics

Figure 3

615

Continued

Because nitrogen is so commonly found as an impurity, most natural diamonds exhibit various shades of yellow and brown. Colorless diamonds are comparatively rare. The largest by far colorless diamond discovered to date is the Cullinan diamond found in South Africa in 1905 and weighing 3106 carats as a rough stone. It was subsequently cut into smaller stones, the largest being 530 carats. Large type IIb diamonds are quite rare, the largest being the Hope diamond, which is approximately 45 carats. Although there are several much larger diamonds, this stone, which originated in India,

616

Reinhard

is famous for its color, a deep sapphire blue.* Also rare, but occasionally found, are diamonds of other colors including orange, pink, mauve, green, red, and black. These also result from impurities or defects. For example, manganese is associated with a broad absorption at 0.55 ␮m that produces a pink or mauve color. Color may also be induced synthetically by creating optical centers using radiation-induced damage. Lattice damage caused by electron, neutron, and gamma ray irradiation produces an absorption band centered near 2 eV in the visible and also an ultraviolet band extending from approximately 3 eV to the absorption edge. The net effect of irradiation in a type IIa diamond is a change in appearance from colorless to green, bluegreen, or blue, depending on the irradiation [11]. Further information on color centers in diamond may be found in Chapter 3 of this book and in Refs. 4 and 5 of this chapter.

III.

OPTICAL PROPERTIES OF POLYCRYSTALLINE CVD DIAMOND

A.

Scattering Effects in CVD Diamond

The very best CVD polycrystalline diamond optical samples do in fact exhibit many optical properties quite comparable to those of type IIa singlecrystal diamond [12]. For example, in a review report by Harris [13] that summarizes the results of a multigroup effort funded by the U.S. Navy to produce optical quality CVD diamond, clear (from the band gap cutoff in the ultraviolet to the phonon absorption in the infrared) windows with thicknesses of 0.3 to 1.0 mm and diameters† up to 25 mm were described. However, much CVD diamond does not have this level of optical quality. Freestanding samples in the as-grown state are often opaque and show a color ranging from a gray to a black appearance. One factor that limits the optical performance of polycrystalline diamond films is scattering at the rough growth surface of the polycrystalline films. First, however, consider the case where scattering effects are neglected. For wavelengths below the band gap and not in the portion of the infrared where absorption occurs, a freestanding, optically smooth diamond sample should show the transmission spectra of a transparent plate in air. For such *One of several diamonds of historical interest, this gem was acquired by the royal family in France in the 17th century, disappeared in the turmoil of the French Revolution in the 18th century, appeared again in England in the 19th century, and in the 20th century became part of the Smithsonian collection in Washington D.C. † In subsequent years, the achievable diameter for high-quality optical samples of CVD diamond increased significantly.

Diamond Film Optics

617

cases, assuming that the incident light is a transverse electromagnetic (TEM) wave (which is valid except for cases of rapidly diverging or converging light), the optical power transmission for normal incidence on a diamond plate with front and back air interfaces is given by the expression [14] T=

(1 ⫺ r2ad)(1 ⫺ r2da) 2 1 ⫹ r2adrda ⫹ 2radrda cos(2␦)

(8)

where the Fresnel coefficients are rad =

na ⫺ n d na ⫹ n d

(9)

rda =

nd ⫺ n a na ⫹ n d

(10)

and

at the top and bottom interface, respectively, and

␦=

2␲ ndt ␭

(11)

is the change in phase of the beam on traversing the diamond film of thickness t. The constructive and destructive interference resulting from multiple reflections at the two interfaces results in a transmission oscillating between 100% and approximately 50% with peaks separated in wavenumber by integer multiples of (ndt/2)⫺1. This is illustrated in Fig. 4A, in which T is plotted as a function of wavenumber using Eq. (8). Also shown for reference is T plotted as a function of ␭ in Fig. 4B. Such oscillatory behavior can be observed experimentally throughout the spectrum if scattering is eliminated by polishing the diamond surface to optical smoothness. It is also observed on nonpolished samples if the optical wavelength is much larger than the surface roughness [15,16]. For example, Fig. 5 shows the transmission spectrum for a freestanding 76.2-␮m-thick film with grain size in the range 2– 8 ␮m. The measured transmission [16] is shown as a function of wavenumber, from 20 to 350 cm⫺1, corresponding to wavelengths from 500 to approximately 29 ␮m. In the longer wavelength portion of the transmission spectrum, the transmission behavior approximates the ideal case of Fig. 4. However, for shorter wavelengths, the transmission drops off because of scattering at the rough surface. Scattering at a rough surface affects both transmitted and reflected light. For light entering a rough diamond surface from air, the roughness of the surface causes phase differences in transmitted light. This introduces a correction factor for power transmission which, following Filinski [17], may be written as

618

Reinhard

Figure 4 Ideal transmission for a 75-␮m-thick freestanding, optically smooth diamond plate. (A) The regularly spaced peaks when plotted as a function of wavenumber; (B) the same information plotted as a function of wavelength.

Diamond Film Optics

619

Figure 5 Measured transmission for a 76.2-␮m-thick film with grain size in the range of 2 to 8 ␮m. (From Ref. 16.)

ST = exp

冋冉

冊册

2␲␴ (nd ⫺ na ⫺ ␭

2

(12)

Here ␭ is the free-space light wavelength and ␴ is the root mean square (rms) of the surface roughness, assuming a Gaussian distribution of surface height about the mean and ␴ small compared with the wavelength [18]. Additional models have been developed for scattering of shorter wavelengths from rough surfaces [19]. The rough surface also affects the interference term due to multiple reflections within the diamond film. The internal power reflection at a rough surface is multiplied by a correction term [17] SR = exp

冋 冉 冊册 4␲␴ nd ⫺ ␭

2

(13)

Again, the reduction of intensity is caused by the phase fluctuations appearing in the reflected beam. Generally on CVD diamond films, the top surface, or growth surface, is much rougher than the bottom surface. Considering only the top surface to be rough and assuming that the electric field reflection coefficient is corrected by a factor (SR)1/2, then the expression for T is modified to account for the rough surface to become TR, where [20] TR =

2 (1 ⫺ r2ad)(1 ⫺ rda ) ST 2 2 1 ⫹ SRradrda ⫹ 2radrda兹SR cos(2␦)

(14)

Figure 6 shows the theoretical transmission for a 75-␮m-thick freestanding plate with an rms surface roughness of 2 ␮m on the top surface. Comparing Fig. 6 with Fig. 4, significant loss in transmission is observed in both the

620

Reinhard

Figure 6 Theoretical transmission for a 75-␮m-thick freestanding diamond film that has an rms surface roughness of 2 ␮m on one side and is optically smooth on the other side. Results may be compared with the theoretical transmission for a smooth diamond film in Fig. 4 and the experimental results of Fig. 5.

experimental and theoretical cases as the wavelength decreases. For ␭ in the visible portion of the spectrum, the transmission would be quite small in both cases. Partly as a result of scattering, as-grown polycrystalline CVD diamond samples often have both reflectance and transmittance values less than 1%, far different from the high values associated with gem quality single-crystal diamonds. The resulting optical appearance has caused more than one diamond researcher to be faced with incredulity when presenting CVD samples to the public for inspection. The general conclusion to be drawn from the literature regarding optical transmission through, and reflection from, polycrystalline diamond is that both are often limited by the surface roughness of the growth surface [10,21–24]. In the far infrared, where the wavelength is large compared with surface roughness, the transmittance approximates that of a transparent plate in air. At smaller wavelengths the transmittance drops off due to scattering at the rough surface. The smoother the surface, the shorter the wavelength at which appreciable transmittance can occur. Consequently, surface roughness can be a major impediment for applications. Methods for reducing

Diamond Film Optics

621

this problem by growing smooth surfaces and by postprocessing are addressed in Secs. IV and V of this chapter. B.

Impurity-Induced Absorption in CVD Diamond

CVD diamond films often contain chemical impurities that decrease optical transparency in portions of the optical spectrum. Some of these impurities are also found in natural diamond and play optical roles similar to those in single-crystal diamond. Examples of these are nitrogen (due, for example, to air contamination in the CVD vacuum system or in the feed gases) and boron (intentionally introduced as a p-type dopant). Nitrogen introduces states in the energy gap reported variously to be between approximately 1.6 and 1.9 eV below the conduction band. These states cause absorption at photon energies below the intrinsic energy gap of diamond, in the ultraviolet and blue portion of the spectrum, such that a yellow coloration results. Boron introduces states in the gap at approximately 0.4 eV above the valance band. These states cause absorption that peaks at approximately 2.5 ␮m wavelength but tails to approximately 0.5 ␮m in the visible, causing a blue coloration [10]. However, CVD diamond often has additional impurities or defects not commonly associated with natural single-crystal diamond. If the CVD system contains metallic electrodes or filaments, metallic contaminants in the film can result. Likewise, if the CVD system involves a discharge contained within silica walls, both silicon and oxygen may be incorporated in the film. Feed gases are also sources of impurities. For example, hydrogen is commonly incorporated in feed gases, both directly and as a component of the carbon-containing molecule. As a result, hydrogen is incorporated in the growing film. When the CVD chemistry involves oxygen-containing species, the CVD diamond film also contains oxygen. In addition to chemical impurities, diamond films contain structural defects. In particular, depending on growth conditions, non-sp3 carbon bonds may be present, corresponding to nondiamond carbon, giving rise to additional allowed energy levels and correspondingly additional optical transitions. Other structural defects include crystallographic defects (twins, stacking faults, and dislocations) as well as grain boundary regions between crystallites. A variety of analytical methods have been used to assess defect and impurity concentrations in diamond films including nuclear magnetic resonance (NMR) [25], x-ray photo-electron spectroscopy (XPS) and x-ray diffraction [26], electron paramagnetic resonance (EPR) [27], photothermal deflection spectroscopy (PDS) [28], transmission electron spectroscopy (TEM) [29,30], cathodoluminescence [31,32], Raman spectroscopy [33,34], Auger electron spectroscopy, secondary ion mass spectrometry (SIMS), and electron energy loss spectroscopy (EELS) [35]. An example of an impurity anal-

622

Reinhard

ysis of CVD diamond by McNamara et al. [26] with regard to hydrogen, oxygen and nitrogen is shown in Table 1. Considering the nitrogen content, it is noted that amounts in this table range from 20 to 100 ppm, which, following the discussion in Sec. II, corresponds to type I diamond. Additional information concerning the characterization of impurities and defects in diamond films may be found in Chap. 3 of this book. Various impurities cause absorption beyond that of the intrinsic optical absorption of diamond by allowing either additional photon-induced electronic transitions or additional photon-induced lattice vibrations. In the case of hydrogen it is principally the latter, such that the effect of carbon-hydrogen bonds in diamond is to cause additional absorption in the infra-red. This added absorption is particularly pronounced in the CH stretch region (2750– 3050 cm⫺1 wavenumbers or approximately 3.5 ␮m wavelength); however, additional absorption due to hydrogen incorporation is also observed in other parts of the infrared spectrum. In the study reported by Harris [13], which included diamond samples grown by various methods including hot-filament, microwave plasma, and dc arcjet CVD processes, hydrogen contents ranged from 16 ppm to fractions of an atomic percent depending on the growth method and gas feed composition. At the higher hydrogen concentrations, C — H stretching absorption bands dominated the infrared absorption spectra with a strong absorption peak near 3.5 ␮m. At the lower range of hydrogen content in the film, the effect of hydrogen on absorption is negligible. Figure 7 illustrates infrared absorption spectra in hydrogen-contaminated polycrystalline diamond [11]. Not only is additional optical absorption

Table 1 Example of Impurity Content Analysis of Diamond Films Deposited by Microwave Plasma-Assisted CVDa Sample A B C D E F G H J

Hydrogen (%)

Nitrogen (%)

Oxygen (%)

0.085 0.180 0.027 0.045 0.021 0.043 0.020 0.015 0.092

0.0048 0.0047 0.0020 0.0035 0.0018 0.0088 0.0024 0.0039 0.0101

0.0091 0.0075 0.0053 0.0157 0.0034 0.0832 0.0132 0.0032 0.1970

a Results are expressed in atomic percent. Source: Ref. 26.

Diamond Film Optics

623

Figure 7 Absorption spectra showing the effect of hydrogen impurities in polycrystalline diamond. The impurities activate one-phonon absorption and also cause absorption due to C — H bonds [11].

present due to C — Hx stretching modes, but also the presence of hydrogen impurities activates one-phonon absorption, which is forbidden in pure diamond. Consequently, samples with strong C — Hx absorption also exhibit strong one-phonon absorption. This single-phonon absorption is also illustrated in Fig. 7 in the form of absorption in the range 8–10 ␮m. Most materials transparent in the region 8–10 ␮m are soft, making diamond an attractive alternative for infrared applications. To the extent, therefore, that hydrogen is incorporated into CVD diamond, this issue can be a performance-limiting concern for CVD diamond. Based on the results shown in Fig. 7, if the hydrogen content is reduced to less than 0.02% atomic, such absorption is negligible even for thick films. For thinner films the effect on percent transmission will, of course, be correspondingly less. However, experimental observation of infrared absorption due to hydrogen impurities

624

Reinhard

has also been reported on thin freestanding membranes [15,36]. For example, Johnson and Weimer [15] removed the silicon substrate from diamond films by etching to study the transmission of a 3-␮m-thick film and observed small but detectable impurity absorption peaks at 2925–2830 cm⫺1 and 3325 cm⫺1, which were attributed to C — Hx groups. Significantly larger peaks were observed on a 13-␮m-thick film. Fortunately, if appropriate means are taken in choosing deposition parameters, the hydrogen content of CVD diamond can be reduced to a point at which the effect on absorption is negligible. Unfortunately, this generally involves the use of relatively low methane concentrations, which results in a trade-off between optical quality and growth rate. The other elements in Table 1, oxygen and nitrogen, also play major roles in the optical properties of CVD diamond. The effect of nitrogen with regard to causing increased absorption in the near UV and visible portion of the spectrum has been previously noted. EPR data suggest that most nitrogen in CVD diamond is present as point-defect substitutional impurities rather than aggregates [27], which in terms of traditional diamond classification corresponds to type Ib rather than type Ia. In terms of oxygen, this element is sometimes added to the gas flow directly as O2, or in the form of CO or CO2, in order to reduce the graphitic content of the diamond film and thereby improve optical transmission in the visible [37]. However, both oxygen and nitrogen also cause increased absorption in the infrared. IR spectroscopy has identified absorption due to a variety of oxygen- and nitrogen-containing groups, including O — CH3 and N — CH3 [26] which introduce absorption in the 8–12 ␮m wavelength region. Therefore, adding oxygen in order to improve visible transmission may be accompanied by a trade-off in terms of decreased transmission in the far infrared. Also noted is that even small leaks in CVD diamond systems can lead to observable oxygen and nitrogen incorporation. In addition to hydrogen, oxygen, and nitrogen, optical effects due to other chemical impurities have been reported. Lithium, of interest as an ntype dopant, has been reported to cause increased absorption in the vicinity of 1.5 eV [28]. The incorporation of silicon has been reported to be associated with absorption at 1.68 eV [31]. C.

Defect-Induced Absorption in CVD Diamond

Structural defects other than chemical impurities are also strongly correlated with optical properties of CVD diamond. Principal among these is the effect of graphitic carbon or sp2 bonds. An early and instructive study in this regard was reported by Collins et al. [31] in which absorption and cathodoluminescence spectra of polycrystalline diamond were investigated for different

Diamond Film Optics

625

growth conditions. The diamond films were deposited on silicon using microwave plasma-assisted CVD systems and methane-hydrogen feed gas mixtures with methane concentrations varying from 0.3 to 3.0%. In order to avoid scattering effects on the optical absorption measurements, the rough growth surface of the diamond was mechanically polished. The central region of the silicon wafer was then etched away, leaving a supported diamond film that was smooth on both sides. In addition to the CVD films, isolated CVD crystals were investigated by growing individual crystals about 100 ␮m across on sparsely seeded substrates. These individual crystals were detached from the substrate and mounted on optical sample holders. As shown in Fig. 8, the results of Collins et al. [31] showed that absorption in the films was a strong function of methane concentration and of a type not attributable simply to C — H bonds. All samples exhibited an almost monotonically increasing absorption with increasing photon energy. The band gap absorption edge is clearly visible in Fig. 8 for the 0.3% methane film; however, sub–band gap absorption is appreciable with ␣

Figure 8 Absorption spectra of CVD diamond films grown with varying methane concentrations as noted, as well as the absorption spectrum of an isolated CVD crystal. (From Ref. 31.)

626

Reinhard

reaching approximately 500 cm⫺1 prior to the onset of band-to-band generation. The color of this film was described as pale yellow, consistent with appreciable absorption in the higher energy portion of the visible. For the higher methane concentrations, the film samples were too strongly absorbing to extend the measurements into the ultraviolet region. The 1.5% methane film shows an absorption coefficient of over 2000 cm⫺1 in the blue portion of visible spectrum and exhibited a dark brown color to the eye. Based on Raman analysis, evidence suggested that the absorption evident in Fig. 8 was due to nondiamond carbon in the form of sp2 bonds. The importance of nondiamond carbon bonds in causing appreciable sub–band gap absorption was further confirmed and quantified in later studies. In an investigation by Rosa et al. [38] of microwave plasma-assisted CVD diamond, a set of diamond films were deposited on silicon under variable growth conditions such that films ranged from white polycrystalline films to brown nanocrystalline films. The value of ␣ at 2.41 eV ranged from a low of 3 cm⫺1 for the least absorbing films to as high as 500 cm⫺1. Films were also characterized by a Raman quality factor defined as

␤=

ID ID ⫹ IN

(15)

where ID and IN are the integrated intensities of the diamond and nondiamond Raman signals, respectively. From ␤, a volume fraction of the nondiamond component was calculated and correlated with the absorption coefficient at 2.41 eV with an approximately one-to-one correlation. When the absorption coefficient is nonnegligible, the expression for transmission is modified as shown in Eq. (16), which accounts for both scattering due to a rough surface and absorption within the film to arrive at a transmission, TRA, given by [22,39] TRA =

2 (1 ⫺ r2ad)(1 ⫺ rda )exp(⫺␣t) 1 ⫹ S r r exp(⫺2␣t) ⫹ 2radrda兹SR exp(⫺␣t)cos(2␦) 2 2 R ad da

(16)

The importance of absorption clearly depends on the diamond thickness. For an absorption coefficient of 500 cm⫺1 and a 1-␮m-thick film, the transmission loss resulting from absorption is approximately 5%, which may be negligibly small for many applications. However, for a 500-␮m-thick freestanding diamond sample, the transmission loss would be 99.3%. Because the optical properties of CVD diamond samples may depend on both wavelength-dependent scattering and various wavelength-dependent absorption mechanisms, separation of optical constants is nontrivial. The procedure involves best-fit solutions to the optical equations using floating values of the free parameters [22,39–41].

Diamond Film Optics

IV.

FREESTANDING DIAMOND AND DIAMOND DIAPHRAGMS

A.

Application and Growth Considerations

627

One class of optical applications of diamond involves the use of either freestanding diamond or diamond diaphragms as optical elements. This section discusses a variety of application-related issues associated with such structures. Various applications for diamond windows have been reviewed by Seal and van Enckevort [1], most of which involve operation in difficult environments. Diamond windows have been used in spectroscopic instrumentation for various spacecraft including the Nimbus and aforementioned Pioneer series as well as the Galileo Jupiter probe. On earth, diamond windows are used in the chemical industry for spectroscopic quality control of various caustic, high-temperature processes and also in laboratories. Because of diamond’s resistance to scratching and ease of cleaning, diamond components are used in instrumentation by the food industry for contact analysis by optical means. In the field of medicine, diamond optical elements have found use as miniature endoscope windows. The transmissive properties of diamond have also been employed for medical use in diamond blades (which have superb cutting edges) in that laser energy can be piped to the cutting edge, where it is used to cauterize blood vessels. In some cases, laser energy at the blade edge also seems to facilitate the cutting of certain tissues. Freestanding diamond has also been used for high power light transmission windows [42,43]. The most common optical material, silica-based glass, is not well suited for high-power windows because its thermal conductivity is so low, typically on the order of 0.01 W/cm deg. The thermal conductivity of diamond is over a thousand times higher and the optical absorption coefficient of diamond can be as low as or lower than that of glass. Consequently, the optical power transmission of diamond is considered to be at least 1000 times higher than that of glass. This is relevant to the designers of high-power lasers, where limits on power are set by the damage limits to laser optics rather than limitations of the laser medium or pumping mechanisms. It is also relevant to high-power gyrotrons in the mm wave portion of the spectrum. A major limitation to optical diamond applications has historically been size limitation imposed by cost. Whereas the cost of a 0.5-mm-diameter window of single-crystal diamond is measured in tens of dollars, increasing the diameter by an order of magnitude to 5 mm causes the price to be measured in thousands of dollars. To increase the diameter another order of magnitude to 50 mm is essentially to leave the realm of availability of singlecrystal diamond. It is for this reason that CVD capabilities offer the potential to greatly increase the applicability of diamond to optics. In fact, by the

628

Reinhard

Figure 9 Measured transmission for a polished, 350-␮m-thick film (solid line) and a 275-␮m-thick type IIa crystal (dashed line). The data are smoothed so as not to show the interference oscillations. Evident in both cases are the intrinsic multiphonon absorption in the infrared between 4 and 7 ␮m and the onset of band gap absorption in the UV [13].

early 1990s, the ability to grow freestanding diamond of significant size with transmission characteristics throughout the entire optical range comparable to intrinsic diamond had been demonstrated. Figure 9 shows transmission results for a 350-␮m-thick CVD diamond sample, fabricated at Raytheon in 1992 [13]. Data are plotted in direct comparison with transmission results for a 275-␮m-thick type IIa single crystal, and essentially identical properties are observed.* Evident in both samples is the onset of band gap absorption at 0.23 mm and the multiphonon absorption inherent to the diamond lattice in the infrared. Absent are any significant effects due to defects and impurities as discussed in Sec. III of this chapter. Within 2 years, such *The data in Fig. 9 are smoothed so that interference effects are not seen. Also, the CVD sample was polished to optical smoothness after growth.

Diamond Film Optics

629

visually transparent samples were available from the same company at 60 mm diameter [44] using microwave plasma-assisted CVD. Gray et al. [45,46] at Norton Co. reported 100-mm diameter CVD ‘‘white diamond’’ wafers, 1 mm thick, with infrared transmission and Raman characteristics comparable to those of type IIa diamond in approximately the same time frame, 1992. In fact a variety of CVD methods have been used to deposit freestanding or supported diaphragm diamond for optical applications or investigations, including hot filament [47], arc plasma jet [45,46,49], and microwave plasmas [16,50–52]. Regardless of deposition method, an important practical consideration for freestanding diamond elements is that of the available deposition rate. Unsupported diamond layers intended for optical applications must be of sufficient thickness to provide necessary robustness. Generally such films are hundreds to several hundreds of micrometers thick. High deposition rates are therefore desired for reasonable production times. However, high deposition rates do not necessarily yield the best quality diamond. In the previously mentioned U.S. Navy–funded study [13], it was found for all production methods used in the investigation (hot filament, arcjet, and microwave) that maximum growth rates were limited to 1 to 5 ␮m/hr for optical quality diamond. Higher growth rates were obtainable by increasing the methane concentration, but this approach to high deposition rates yielded lower optical quality. Therefore, one challenge for freestanding CVD optical diamond is to achieve high deposition rates while maintaining little or no sp2-bonded carbon and a low concentration of C — H bonds, thereby maintaining the low absorption coefficient necessary for high transmissivity in thick films. Issues specifically related to increasing growth rates are addressed in other chapters in this book that are related to deposition methodologies. B.

Diamond Surface Considerations

Another important consideration for applications is the diamond surface condition. As previously noted, the rough growth surface of a polycrystalline diamond film presents an important performance limitation due to optical scattering. Generally, the surface roughness of diamond films increases as the thickness increases. In the course of film growth, growth competition between the individual crystallites results in more and more crystallites being buried and a decreasing number of surviving crystallites at the surface with the size of the crystallites increasing as growth continues. Consequently the growth surface of an as-grown CVD diamond sample with a thickness of 500 ␮m may show well-faceted crystallites with average sizes of 50 ␮m and larger [45]. Such a surface can easily have an average surface roughness

630

Reinhard

Ra value of several micrometers. For such samples, except for the far infrared, optical transmission losses due to scattering are appreciable and it is necessary to polish the diamond surface in order to achieve good transmission in the visible and ultraviolet. Because diamond is the hardest known material, polishing a rough diamond surface is a nontrivial task. However, a variety of methods have been developed. Drawing from experience in the gem industry, a traditional method is to rub the rough diamond surface against a cast iron surface. One version of this approach is to rotate an iron wheel (called a scaife) at high speed (2000 rpm) and press the diamond against the rotating surface.* Significant heat is developed by the friction between it and the diamond. Another approach is to heat the iron to approximately 350⬚C and use slow rotation [50]. The polishing in either case is believed to be due to a chemical reaction between the diamond carbon and the iron to form iron carbide, which is subsequently oxidized in air, due to the high temperature, and is easily removed by mechanical polishing. This method can take several weeks to achieve an optically smooth surface. If polishing is too aggressive, the diamond piece can be shattered. Alternatively, the surface can be polished with high-speed lapping using diamond grit as the abrasive [53]. Surface smoothing and diamond removal have also been achieved by using diffusional reactions with hot metal plates. In this approach, the diamond workpiece is placed in static contact with iron, manganese, cerium, lanthanum, or La-Ni eutectics at 600–900⬚C in an argon atmosphere, under which conditions diamond dissolves in the metal [54–58]. For example, as reported by Jin et al. [55], a freestanding CVD diamond piece (1 cm ⫻ 0.5 cm ⫻ 250 ␮m thick) was sandwiched between metal sheets with compressive stress of 3000 psi for iron and 200 psi for manganese for 48 hr at 920⬚C. The rough growth facets were reported to be mostly removed and the diamond thickness was reduced by approximately 50 ␮m. Using rare earth metals, removal rates were more than five times higher because of the large liquid solubility of carbon at high temperatures. Noncontact methods including ion beams, plasmas and laser treatment have also been investigated for improving the diamond surface [59–64]. This approach can be facilitated by first applying a sacrificial planarizing layer to the diamond surface and then using a material removal process that removes the diamond and planarizing layer at comparable rates. For example, Bovard et al. [62] have described the use of photoresist/titanium-silica as a planarizing layer followed by reactive ion beam etching with 500 eV

*Iron is the most common scaife material but not the only possibility. Nickel, for example, can also be used.

Diamond Film Optics

631

oxygen ions to achieve optical quality smoothness. Starting with a 5-cmdiameter CVD sample with a 1-␮m peak-to-valley surface roughness, the reactive ion beam polishing procedure reduced the peak-to-valley distance to 55 nm and resulted in an rms surface roughness of 5.5 nm [62]. In some cases freestanding diamond is grown to a thickness greater than that needed for the final product so that postgrowth diamond removal can be used to correct for nonuniform thickness or sample bowing. Noncontact methods, such as electron-cyclotron-resonance (ECR) oxygen plasmas, have been shown to be much faster than conventional scaife or lapping methods for bulk removal of diamond from workpieces [65]. For example, Chakraborty et al. [66] have reported the use of ECR plasmas to etch 10cm-diameter CVD freestanding diamond, 1 mm thick, with O2/SF6/Ar plasmas. Diamond removal rates of 7 ␮m/hr with ⫾10% uniformity across the substrates were reported. The etch mechanism in such cases is primarily highly anisotropic ion-assisted reactive etching by diamond. To first order, the etch rate increases linearly with increasing ion energy. Therefore, high etch rates are facilitated by substrate bias. The role of SF6 is to avoid the formation of a residual black film on the diamond surface [66,67]. Such etching erodes the rough crystal facets, but it does not produce an optically smooth finish, which requires further processing. Although it is possible to postprocess CVD diamond surfaces to optically smooth surfaces by various polishing means, it is clearly of advantage to grow the films with smooth surfaces such that the need for polishing is greatly reduced or eliminated. One approach, described by Wild et al. [51,68,69], is to grow textured diamond films on silicon substrates with a preferred orientation of directional growth such that a relatively flat growth surface is covered with {100} faces results. After growing a thick diamond film, the silicon substrate is removed, leaving a freestanding diamond sample. By this growth method, 150-␮m-thick films were deposited with surface roughness values of 150 nm. This Ra value is still an order of magnitude higher than that needed for lossless transmission throughout the entire visible spectrum, but it is substantially less than for polycrystalline films grown by conventional means. Such orientation is achieved by adjusting growth chemistry such that the ratio of growth rates on the {100} and {111} faces is slightly below 兹3. In this case, cubo-octahedral nuclei with small {100} faces develop. The direction of fastest growth is perpendicular to these faces, and as the film grows thicker the sizes of these {100} faces increases. Consequently, growth techniques that result in preferred growth orientations are useful for optical applications. Table 2 summarizes a specific method as described by Maeda et al. [70] for growing diamond that is textured in the 具100典 direction and results in highly oriented {100} faces on silicon substrates in a microwave plasma. This approach uses a three-step

632

Reinhard

Table 2 Three-Step Microwave Plasma CVD Conditions for Growing Highly Oriented {100} Diamond Films on Silicon Parameters CH4 (sccm) H2 (sccm) Bias voltage (V) Treatment time (min) Microwave power (W) Substrate temperature (⬚C) Pressure (kPa)

Carburization 2 98 0 15–120 570 830 2.65

Bias treatment 2 98 ⫺150 1–15 830 830 2.65

Growth 0.5–2 98–100 0 >120 460–670 800–900 4

Source: Ref. 70.

growth sequence in which the first step carburizes the surface, the second step adds an electrical bias to the substrate which creates a textured growth surface, and the third step involves the growth of the bulk of the film. Film texture has also been correlated with hydrogen incorporation and therefore with infrared absorption. Haq et al. [71] investigated infrared absorption for various film textures and observed that the lowest C — H absorption was associated with 具110典-oriented films, followed by 具100典 and then 具111典. Single-phonon absorption was also found to be highest in the case of 具111典 textured films and lowest for 具110典 and 具100典 with no significant difference between the latter two. An alternative approach used to achieve smooth surfaces that works for relative thin films is to deposit very fine grain films [69,72,73]. If the nucleation density is very high, on the order of 1010 cm⫺2 or higher, and the film thickness is kept sufficiently small that large grains do not have a chance to form, then quite small surface roughness values can be achieved. Yang et al. [74] reported nucleation densities as high as 1011 cm⫺2 and a mean surface roughness of 30 nm for 1-␮m-thick films. The role of fine grain films is discussed further in this chapter in the context of substrate coatings in Sec. V.B. C.

Mechanical Considerations for Plates and Diaphragms

For several optical applications of freestanding diamond, mechanical strength is an important figure of merit as well as optical transparency. Whereas it is possible to make freestanding diamond pieces of significant size that essentially have the optical transmission of intrinsic diamond, the

Diamond Film Optics

633

same cannot be said of mechanical strength. The mechanical strength of optical quality 0.5- to 1-mm-thick CVD diamond has been reported to be significantly less than that of natural diamond [53]. Optical quality diamond has been reported to grow in a highly stressed state that can lead to macroscopic and microscopic cracks [13].* It has been suggested that the incorporation of voids and defects arises as a result of random in-plane grain orientations coupled with the overgrowth of twin boundaries and that this is relatively independent of reactor design [75]. Such cracks and defects play a critical role in mechanical strength because the strength of polycrystalline CVD diamond is governed by flaws. One way to measure mechanical strength is by means of the ring-onring test, in which one side of a diamond plate is supported on a metal ring and the other side is contacted by a metallic load ring of smaller radius. The load magnitude is gradually increased until the point of fracture. In one study of ring-on-ring load testing of CVD diamond plates produced by a variety of deposition methods, measured strengths were reported to be an order of magnitude less than would have been expected for single-crystal diamond [13]. The fracture origins in the CVD diamond plates corresponded to relatively large flaws, on the order of 100–300 ␮m in size for 800-␮m-thick disks. In a study by Valentine et al. [76], the bursting strength of diamond plates with thicknesses in the range of 179–319 ␮m and diameters in the range of 14–25 mm were measured by applying gas pressure to diamond disks supported at the disk periphery. In this case, the CVD diamond strength was reported to be about half that of natural diamond. Fine polishing of the surface seems to have no effect on strength [13]. For some optically related applications, it is possible to consider the use of diamond membranes, or diaphragms. For these structures the mechanical strength of CVD diamond does not necessarily differ significantly from that of single-crystal diamond, perhaps because the thinner films cannot have flaws as large as those found in the thicker structures [53]. Diamond diaphragms subject to loads can, however, show considerable deflections, resulting in distorted optical surfaces. The load deflection behavior of pointloaded CVD diamond thin-film diaphragms was investigated by Glime et al. [77]; circular diaphragms consisting of diamond film 1.5 ␮m thick and 2– 4 mm in diameter adhering to a silicon rim were prepared by mechanically grinding and chemically etching the single-crystal substrate. A nanoidentor was used to provide load versus deflection data. The maximum deflections

*It should be noted that natural diamonds are also occasionally found that are in highly stressed states. There are reports of large stones of considerable potential value spontaneously shattering when brought to the surface of the mine because of temperature changes.

634

Reinhard

were observed to be up to five times the film thickness before failure. When the central deflection of a plate exceeds about one half the plate thickness, classical plate theory relating load and deflection to the mechanical material properties is no longer valid and nonlinear models must be used [78]. The relationship between applied central point load, P, deflection at the center of the diaphragm, w, and diamond thickness, t, is reported to be of the form 4␲Et4 P= 3(1 ⫺ ␯2)r2

冋

w ⫹A t

冉 冊册 w t

3

(17)

where r is the radius, E is Young’s modulus, ␯ is Poisson’s ratio, and A is a constant equal to about 0.43 for a point-loaded circular plate loaded in the nonlinear region and diamond diaphragm deflection data agrees well with this equation [77,79]. Diamond diaphragm structures often contain internal stresses. Berry et al. [80] have used vibrating membrane methods to measure internal stresses in CVD diamond deposited on silicon using microwave plasma deposition. Central regions of the substrates were back etched to leave the diamond film as a taut membrane, with net tensile stress, supported circularly at the edges. Film stresses ranging from 9.4 to 139 MPa were measured, depending on growth temperature. Flowers et al. [81] found the burst pressure for diamond diaphragms to increase linearly as a function of a parameter R, where R is the square root of the film thickness divided by the diaphragm diameter. In some cases, compressive internal stresses in diamond diaphragms have resulted in film buckling, causing wrinkling and nonplanar surfaces [77]. The issue of internal stresses in diamond films is discussed farther in the context of film coatings in Sec. V.C. Mechanical strength is also of interest for shapes more complex than simple plates. Of particular interest for certain airborne applications has been the fabrication of freestanding CVD diamond domes such as could be used to protect nose cone infrared sensing instrumentation where aerodynamic considerations are crucial [53,82]. Near net shape domes have been reported by depositing on nonplanar growth surfaces. Potential applications of such window structures involve speeds as high as mach 4.* Consequently, significant differential pressures and complex strain patterns must be considered [83]. Another mechanical strength–related figure of merit particularly relevant for airborne applications is the erosion resistance to liquid impact. In

*The limit on speed is imposed by temperature considerations. At mach 6, stagnation temperatures are such that diamond may convert to graphite.

Diamond Film Optics

635

a study by Jilbert et al. [84], the rain erosion resistance of diamond was investigated by using high-velocity jets, 0.8 mm in diameter, designed to simulate the effects of spherical raindrop impact. Using 1-mm-thick samples, the threshold velocity for catastrophic failure was evaluated for both type IIa natural single crystals and CVD diamond. For the type IIa samples, the threshold velocity for catastrophic failure was 515 m/sec and for CVD diamond the threshold velocity for one sample was 325 m/sec and for another was 375 m/sec. For sand erosion, the CVD diamond was again found to be considerably weaker than its natural counterpart, as illustrated in Fig. 10. Although CVD diamond is inferior to single-crystal diamond in regard to mechanical strength, it should be noted that competing infrared-transmitting windows all suffer to some degree from resistance to erosion and other issues related to mechanical strength. A bursting strength that is half that of single-crystal diamond is still approximately an order of magnitude higher than those of other materials transparent at an IR wavelength of 10.6 ␮m, such as GaAs, Ge, and ZnS [76]. Also, figures of merit for a given application generally involve multiple material parameters that can further favor diamond. For example, CVD diamonds offer a useful combination of optical transmission and resistance to erosion and thermal shock. Harris [53] noted that even with a relatively low mechanical strength, the extremely high ther-

Figure 10 The exposure time required to form cracks in natural and CVD diamond upon exposure to sand erosion with a flux of 10.5 kg m⫺2 sec⫺1. (From Ref. 84.)

636

Reinhard

mal conductivity of CVD diamond coupled with its low thermal expansion provides a thermal shock resistance that is two orders of magnitude greater than that of other optical window materials. Many other aspects of CVD diamond windows compare well with single-crystal diamond. It is noted that the hardness and modulus of CVD diamond are found equivalent to those of type IIa natural diamond [53]. Also, high optical quality and high thermal conductivity go hand in hand for CVD diamond. In fact, optical absorption measurements have been proposed as an inexpensive technique for estimating the thermal conductivity of CVD diamond [85]. Optical quality CVD diamond thermal conductivity is reported to have room temperature conductivities of approximately 2000 W/m deg. Combined with low absorption over most of the optical spectrum and a low thermal expansion coefficient, 1 ppm/K at room temperature, the same as for type IIa diamond, CVD diamond is capable of providing the attributes required for high-power-transmitting windows.

D.

Effects of Oxidizing Environments

Finally, whether diamond is in the form of single-crystal or CVD polycrystalline, unprotected diamond is not suitable for use at high temperatures in oxidizing environments. Diamond maintains its unique transmissivity at high temperatures in nonoxidizing atmospheres and the refractive index has been measured up to at least 1200⬚C [86]. However, it is oxidized in air at temperatures above approximately 700⬚C. The resilience of polycrystalline CVD diamond varies considerably with film quality. Bachmann et al. [87] noted that some samples survive 20 min in oxygen at 700⬚C while others disappear completely within only 5 min [87]. Raman and SEM analysis indicated that defective diamond and nondiamond carbon composites are preferentially etched. Nitrogen-doped films and films grown in deposition systems with air leaks showed poor resistance to oxygen etching because nitrogen induces the decoration of diamond grains with highly defective or nondiamond carbon. The operating temperature range of diamond in oxidizing environments can be increased by applying protective coatings. An added advantage of such coatings would be to act as an antireflection layer. Moran et al. [88] have considered a variety of materials for high-temperature coatings including selected oxides, hydrides, carbides, fluorides, and nitrides of yttrium, hafnium, cerium, aluminum, silicon, and boron. In nonoxidizing environments, an upper limit for diamond applications would be 1600⬚C, at which diamond transforms to graphite, which is the stable phase of carbon at atmospheric pressure [13].

Diamond Film Optics

V.

DIAMOND-COATED OPTICAL SUBSTRATES

A.

Optical Considerations

637

The attractive surface properties of diamond—extreme hardness and chemical resistance—can be transferred to optical substrates that lack these features by means of thin-film diamond coatings. This section considers the optical properties of a system consisting of a diamond film on a transparent substrate. Consider, for example, a diamond film of thickness t on a glass substrate as shown in Fig. 11. For the ideal case of zero absorption in the diamond and of smooth surfaces such that there is no scattering, the transmission is limited only by specular transmission at the interfaces. Considering light again in the form of a TEM electromagnetic wave incident on the diamond, the light in the glass substrate is determined by a modification of Eq. (8) as TG =

2 (1 ⫺ r2ad)(1 ⫺ rdg ) 2 2 1 ⫹ radrdg ⫹ 2radrdg cos(2␦)

(18)

where the coefficients rad and ␦ are given in Eqs. (9) and (11) and rdg =

nd ⫺ n g nd ⫹ n g

(19)

Equation (18) does not account for reflection at the glass-air interface or for multiple reflections in the glass. The observed transmission for the total structure, defined as TT, is related to TG as

Figure 11

Considerations for the ideal transmission of a diamond film on glass.

638

Reinhard 2 2 TT = TG(1 ⫺ rga )[1 ⫹ rga (1 ⫺ TG)] ⫹ higher orders

(20)

where rga =

ng ⫺ n a ng ⫹ n a

(21)

The higher order terms typically contribute on the order of 0.1% and may be neglected for most purposes. Equation (20) is based on the assumption that the light reflected within the glass substrate is not sufficiently monochromatic to produce interference effects over the width of the glass substrate. Consider, for example, the case where the glass is 1 mm thick and the incident light is from a monochromator with a bandwidth of 10 nm. For these conditions the wavelength separation of maximum in transmission due to multiple reflections within the glass is much smaller than the bandwidth of the monochromator—the light behaves incoherently as far as multiple reflections in the glass are concerned. Figure 12 shows the ideal transmission resulting from a 1-␮m-thick diamond film on glass using Eq. (20) and refractive index values of 1.54 and unity for glass and air, respectively, where Eq. (3) is used to generate refractive index values for diamond. Note that the glass, whose refractive

Figure 12 Calculated transmission for a diamond film on glass with zero scattering and zero absorption.

Diamond Film Optics

639

index is between those of diamond and air, acts as an antireflection layer, increasing the average power transmission. For a diamond plate with air interfaces on both sides, the average power transmission is given by the average of Eq. (8), which is approximately 67% for a diamond refractive index of 2.4. For a diamond-glass combination with air interfaces on top and bottom, the average power transmission is given by the average of Eq. (20), which is approximately 75%. For semiconductor infrared detector substrates, a diamond overlayer acts as an antireflection coating. For example, for a semiconductor with a refractive index of 3.6, a diamond coating would increase the fraction of light entering the semiconductor from 68 to 79%. As has been previously noted, a rough top surface will cause scattering losses due to phase differences in both transmitted light and internally reflected light. Following the treatment of Sec. III.A, the light entering the glass substrate under conditions of a rough diamond surface, defined here as TGR, is given by a modification of Eq. (14) such that TGR =

2 (1 ⫺ r2ad)(1 ⫺ rdg )ST 2 2 1 ⫹ SRradrdg ⫹ 2radrdg兹SR cos(2␦)

(22)

and the expression for transmission through the entire structure, accounting for reflection loss at the glass-air interface, defined here as TTR is given by TTR = TGR(1 ⫺ r2ga)(1 ⫹ r2ga(1 ⫺ TGR)) ⫹ higher orders

(23)

Figure 13 illustrates the calculated transmission using Eq. (23) for rms surface roughness of 40, 25, and 10 nm throughout the same visible/nearinfrared spectral region used to illustrate the ideal (no scatter) case of Fig. 12. For the case of 40-nm surface roughness, the effect on scattering in the visible range is nonnegligible, with the transmission value dropping to less than 40% in the blue portion of the spectrum. However, for the case of 10nm surface roughness the transmission compares very closely to the results of Fig. 12 for the ideal film without scatter. B.

Diamond-Coated Substrate Examples

One approach to applying diamond to an optical substrate is to form the two materials separately and then bond them together. For example, Partlow et al. [89] have described an ‘‘optical brazing’’ process that results in the thermal bonding of diamond to infrared-transmitting windows of ZnSe and ZnS. In this method, diamond is deposited on silicon by microwave plasmaassisted CVD. A chalcogenide bonding glass, whose refractive index is closely matched to diamond, is used as an interlayer between the diamond and, for example, ZnSe. A sandwich is made of the II-VI compound, the chalcogenide, and the diamond-coated substrate, with the diamond facing

640

Reinhard

Figure 13 Calculated transmission for a diamond film on glass for rms surface roughness values of 10, 20, and 40 nm.

the chalcogenide. The sandwich is then warm pressed, during which most of the chalcogenide is squeezed out, leaving a thin film that bonds the rough diamond surface to the ZnSe. Finally, the silicon is etched away, leaving the smooth side of the diamond film exposed and the rough side of the diamond film immersed in the bonding glass. As another example, Fuentes et al. [90] have described a method to bond diamond films anodically to glass substrates. Diamond was found not to bond anodically directly to glass rings, ˚ thick, was sputtered on the surface so an interfacial layer of SiC, 1000 A of diamond films, which had been previously deposited on silicon substrates. The Si-diamond substrate was then placed with the diamond against glass, and a combination of heat and electrical bias was used to accomplish the bonding. Subsequently, the silicon was removed by etching, leaving diamond membranes attached to glass rings. Also, a method for attaching diamond to infrared-transmitting glass using a commercial optical glue has been reported by Ran et al. [91]. The majority of studies of diamond-coated optical substrates, however, have involved deposition of the diamond onto the substrate in a CVD growth chamber. The rest of this section is concerned with examples that use this direct growth approach. CVD methods have been used to deposit diamond films on a variety of substrates of optical interest including quartz [92,93], fused silica [94–

Diamond Film Optics

641

96] sapphire [93,96,97], soda-lime glass and lead glass [98], borosilicate glasses [99–101], germanium [96], MgO [98], and ZnS [97]. Silicon, one of the most common substrates used for CVD diamond deposition, is also of interest for certain optical applications including detectors. For some optical substrates, CVD of diamond is problematic because of (1) low temperature requirements, (2) coefficient of thermal expansion (CTE) mismatches with diamond, or (3) the substrate being damaged by the deposition plasma. All three of these concerns are present, for example, with ZnS and ZnSe. These materials would be severely etched by the atomic hydrogen present in typical diamond deposition plasmas. Also, chemical considerations require that the substrate temperature during deposition be less than 650⬚C. Finally, the CTE of these substrates is several times higher than that of diamond. In spite of these concerns, successful diamond deposition on these substrates can be accomplished by first applying a protective interlayer, as described by Costello et al. [97] which protects the substrate from the deposition plasma [102]. In contrast, silicon exhibits none of these concerns to a serious degree. A silicon substrate is compatible with direct contact with deposition plasmas, temperatures over 1000⬚C can be tolerated, and although the CTE of silicon is not identical to that of diamond, the differences are such that good adhesion can be obtained over a wide range of deposition conditions. Most optical substrates fall somewhere between the extremes of ZnS and Si, with one or more of the preceding three concerns being of issue. An optical substrate of considerable interest for diamond coating is silicate-based glass. Such glass, arguably the most common optical material, lacks the resistance to erosion and abrasion exhibited by diamond. Consequently, direct CVD of diamond on common commercial glass is of interest for enhancing resilience to hostile environments. Issues associated with such deposition depend on the glass type. Table 3 shows the composition of several common glasses, including soda-lime, borosilicate, and lead glass,

Table 3 Comparison of Commercial Silicate Glasses Showing Composition by Weight %, Strain Temperature Ts, and Room-Temperature Coefficient of Thermal Expansion Glass type

SiO2

B2O3

Al2O3

CaO

MgO

Fused silica Soda lime Borosilicate Alkali lead

99.9 72.6 81.0 77.0

0.8 13.0

1.7 2.0

4.6

3.6

Source: Refs. 103 and 104.

1.0

PbO

Na2O

8.0

15.2 4.0 9.0

K2O

Ts (⬚C)

CTE (10⫺7/⬚C)

5.0

956 473 510 390

5.5 92 33 93

642

Reinhard

as well as the CTE in each case* [103,104]. Soda-lime glass is commonly used for windows, containers, and lamps; borosilicate glass is used for cookware, labware, and headlamps; and alkali lead glass is used for lamp tubing and sealing applications. Of principal interest is the CTE, which ranges from a value somewhat below that of diamond for the case of fused silica to values an order of magnitude higher than diamond for the case of soda-lime and lead glasses. Also varying greatly with glass composition is the effect of temperature on the mechanical properties, with the glass changing gradually from an elastic solid at room temperature to a viscous liquid at high temperatures. A temperature of relevance to deposition applications is the strain temperature, which corresponds to a viscosity of 1014.5 poise. Generally speaking, internal strain cannot be introduced into the glass at temperatures below this value. In order to minimize internal strain in the diamond film on the cool-down after growth, the deposition temperature should be less than the strain temperature of the glass. Considering the compositions in Table 3, the strain temperatures vary from 956⬚C for fused silica to less than 500⬚C for the soda-lime and lead glasses. In any case, for silicate-based glasses the principal matrix seen by the diamond film is that of SiO2. The interface between CVD diamond films and SiO2 and between CVD diamond films and Si has been studied by Pickrell et al. [95] by analyzing the interface side of flakes of diamond film removed from the substrate. In the case of conventional silicon substrates, several groups have reported that an essentially continuous SiC layer forms during the initial growth [95,105,106]. However, in the case of SiO2 substrates, silicon carbide was observed only in discrete, particulate-like areas. Both Si and SiO2 will react with C to form SiC at elevated temperatures. However, in the reaction of C with SiO2 a gas, CO, is evolved. The silicon carbide can also react with the SiO2 substrate to form more gases, as [95] 2SiO2 ⫹ SiC = 3SiO ⫹ CO

(24)

where both products are gases. The extent of reaction between arriving carbon and silica to form silicon carbide will be limited by the diffusion of gaseous species away from the interface, a process that will slow down as the diamond film grows, eventually reaching a local equilibrium. As suggested by Pickrell et al. [95], the crystals of silicon carbide can act as nucleation sites for diamond. However, gas evolution may impede the growth of diamond parallel to the substrate, which produces a noncontinuous, par-

*In terms of commercial codes, the glasses in Table 3 correspond to the following Corning numbers: fused silica (Corning 7940), soda lime (Corning 0080), borosilicate (Corning 7740, also called ‘‘娃Pyrex’’), alkali lead (Corning 0010).

Diamond Film Optics

643

ticulate type of intermediate SiC layer at the interface. Other potential nucleation sites on silica surfaces include residual diamond crystals from seeding procedures, surface scratches, and possibly a ternary Si-O-C phase [107]. In spite of the evidence for a patchy SiC interface, block-on-ring tribotests have indicated that diamond films adhere well to fused silica substrates [94]. Diamond deposition on fused silica can be accomplished using the normal range of CVD diamond deposition temperatures, typically between approximately 700 and 900⬚C. However, for the more common glasses, it has been necessary to develop lower temperature deposition systems. In the case of microwave plasma-assisted CVD, for example, lower substrate temperatures have been accomplished by a variety of methods including [92,94,99,100,108–118]: decreasing the microwave input power and lowering the pressure such that the size of the plasma increases, thereby decreasing the plasma power density; cooled substrates; pulsed power systems; and magnetoactive plasmas, including electron-cyclotron-resonance (ECR) plasmas. Although most low-temperature deposition reports have involved microwave plasma CVD systems, other deposition systems have been adapted to low-temperature deposition, including hot-filament systems [119], combination ECR/filament systems [120], and rf plasma systems [121]. Input gas mixtures that have been optimized for high-temperature deposition are not generally optimum for low-temperature deposition. In many, but not all, of these cited cases, oxygen-containing gases were added to hydrogen-methane mixtures, including CO, CO2, and O2. In some cases, H2 has been eliminated from the input gases [114], and in other cases alternate sources of carbon, including fullerenes, have been investigated [122]. A hindrance to low-temperature deposition is the fact that abstraction of hydrogen from growth sites is a temperature-activated process, generally involving hydrogen atoms. As a result, growth rates at low temperatures exhibit an experimental activation energy of approximately 0.6 eV [99], consistent with a model by Harris and Weiner [123] in which radical surface site pairs are involved in the incorporation of new carbon species on the surface. Consequently, low-temperature (less than 500⬚C) CVD diamond deposition that results in uniform coverage of areas of significant size show rather low growth rates, on the order of 0.1 ␮m/hr. Also of interest as feed gases for low-temperature deposition are halogens [124]. The dissociation of molecular fluorine into fluorine radicals occurs at rather low temperatures and fluorine may be beneficial in abstracting hydrogen from the growth surface at low temperatures. As would be expected from the discussion in Secs. III.A and V.A, high transparency of diamond-coated substrates requires a smooth diamond surface. Two examples of diamond-coated borosilicate glass are shown in Fig. 14. Both samples in this figure consist of polycrystalline CVD diamond

644

Reinhard

Figure 14 Photograph of two diamond-coated borosilicate glass samples in the as-grown state. The sample on the left is a large-grain film with appreciable scatter and low transmission in the visible. The sample on the right is a highly transparent, small-grain diamond film [100].

deposited on 5-cm-diameter glass substrates by microwave plasma-assisted deposition and both samples are in the as-grown, nonpolished state [99]. The difference in visible transmission between the two samples is striking. Light transmission through the sample on the right-hand side of the photograph allows good visibility of the optical bench. Holes in the bench that are approximately a meter distant from the sample are clearly noted. In contrast, the sample on the left has sufficient scattering that the holes in the optical bench cannot be seen. Both samples are of comparable thickness—1.4 ␮m for the sample on the right and 1.7 ␮m for the sample on the left. However, they differ in terms of the initial seeding method and therefore the final grain size and surface roughness. The sample on the left is representative of films with grain sizes on the order of a micrometer and with surface roughness rms values on the order of 100 nm. As a result of the surface roughness, there is strong scattering of visible light and the sample displays a frosted glass appearance. Such samples appear translucent rather than transparent;

Diamond Film Optics

645

objects are clearly visible only if they are directly adjacent to the diamond film. The sample on the right is representative of finer grain films where the average grain size is substantially less than a micrometer and the surface roughness is measured in tens of nanometers. For the particular sample illustrated, the surface roughness was experimentally measured to be in the range of 20 to 30 nm. Although the grain size is small, the Raman spectrum, shown in Fig. 15, shows the characteristic diamond peak at 1332 cm⫺1 and does not indicate appreciable nondiamond carbon content. Figure 16 shows the measured transmission through the two samples shown in Fig. 14, with a 1-cm-diameter detector located 2 cm behind the sample and a 0.4-cm-diameter beam incident on the sample. For the largegrain film, the measured transmission is less than 10% throughout the visible. For the small-grain film, the transmission falls from about 70% in the red portion of the visible spectrum to approximately 40% in the blue. Comparing these results with the calculated transmissions in Fig. 14, the fall-off in experimental transmission with decreasing wavelength is consistent with a surface roughness value of approximately 35 nm, slightly larger than the surface roughness as measured by a stylus. The calculated transmissions of Fig. 14 predict that a surface roughness of 10 nm or less should result in high transmission throughout the entire visible range. Such transmission is, in fact, achievable for as-grown polycrystalline CVD diamond films on glass, as shown in Fig. 17 for a 0.43-␮m-thick diamond film deposited on Corning 7059 borosilicate glass [125].*

C.

Stress and Adhesion Considerations

Film adhesion can be an important issue for CVD diamond films on optical substrates because of stress in the films. This is particularly true for substrates whose thermal expansion coefficients are poorly matched to that of diamond. For example, Raman analysis of diamond films on soda-lime glass have shown a shift in the diamond peak from 1332 to 1340 cm⫺1 [98] with the shift to higher wavenumber indicating that compressive film stress is present. Large stresses can cause sample bowing and facilitate film delamination from the substrate. In fact, the phenomenon has been proposed as a means of forming freestanding diamond films, based on a study in which diamond deposited by hot-filament CVD onto Corning 7059 glass was observed to separate spontaneously from the substrate [101]. For most appli-

*Corning 7059 is a low sodium content borosilicate glass, otherwise similar to 娃Pyrex. It is used in electronic applications, including flat panel displays.

646

Reinhard

Figure 15 Raman spectrum for the transparent diamond film shown on the righthand side of Fig. 14.

cations, however, it is desired that the film remain intact to the optical substrate. The dependence of stress in diamond films has been studied by Windischmann et al. [126] as a function of gas composition and deposition temperature. If diamond is deposited on a substrate whose CTE is higher than that of diamond, a compressive contribution to the total stress in the film is developed as the film-substrate combination is cooled from the deposition temperature to room temperature. In addition to the thermally induced stress caused during cool-down, diamond films have an intrinsic tensile stress that depends on methane concentration, among other things. Its origin is explained in terms of a grain boundary relaxation model due to a fine-grained porous microstructure [126]. Consequently, diamond films may be under either tensile stress or compressive stress, depending on the substrate and on deposition conditions. If thermally induced compressive stress and intrinsic tensile stresses cancel, the resulting films have very low net stress. For numerical calculation purposes, the thermal compressive stress may be estimated by breaking the temperature into i intervals and calculating a stress for each interval according to [126]

␴i =

冉 冊 E 1⫺␯

(␣F ⫺ ␣S)⌬Ti

(25)

where E/(1 ⫺ ␯) is the biaxial modulus and (␣F ⫺ ␣S) is the average difference in CTE values for the film, ␣F, and the substrate, ␣S, at the midpoint of the ith temperature interval, ⌬Ti. The total thermal stress, ␴th, is calculated by summing the ␴i. An estimate of the internal tensile stress, ␴in, is given by Windischmann et al. as

Diamond Film Optics

647

Figure 16 Measured transmission in the visible and near infrared for the two samples shown in 14. (a) The large-grain film and (b) the small-grain film [125].

648

Reinhard

Figure 17 Measured optical transmission for a 0.43-␮m-thick diamond film on borosilicate glass (Corning 7059) that shows high transmission throughout the visible and near infrared [125].

␴in =

冉 冊冉 冊 E 1⫺␯

␦ d

(26)

where d is the grain size of the polycrystalline film and ␦ is estimated to be 0.077 nm for diamond [126]. The total stress, ␴tot, is equal to the difference between ␴th and ␴in and either compressive or tensile, depending on the sign. For silicon, the thermal stress is always compressive. Although the CTE of silicon drops below that of diamond at high temperatures, the difference is not enough to counteract the larger CTE of silicon at low temperatures. For diamond films on silicon with grain sizes of the order of 0.1 ␮m and larger the combined stress is compressive, based on both calculations and experiments that measured the bowing of diamond diaphragms produced by back etching the silicon. In this regime, as the grain size increases, other things being equal, the net strain would be expected to increase. This is consistent with a report by Gamlen et al. [127], who investigated diamond film adhesion on silicon by measuring delamination after Vickers indentation and found the delamination diameter to increase with increasing grain size. For very fine grain diamond films on silicon, however, the net stress is generally tensile.

Diamond Film Optics

649

Figure 18 shows the CTE values for diamond and Corning 7059 glass as a function of temperature. The glass CTE value is fairly constant until near the strain point of 593⬚C, at which it increases sharply. Consider, as an example, a diamond film on 7059 glass cooling down from 475 to 25⬚C. Taking the biaxial modulus for diamond to be 1345 GPa [128], the thermally induced strain may be estimated from Eq. (25) to be approximately 1.7 GPa compressive, using a ⌬Ti value of 100⬚C. Consequently, from Eq. (26), films with grain sizes larger than approximately 70 nm would be in a state of compressive stress and films with smaller grain sizes would be in a state of tensile stress. Grain size depends most directly on seeding density; however, for a given seeding density, grain size generally increases with increasing film thickness. Equations (25) and (26) are useful for determining trends. Neglected, however, are effects such as structural relaxation in the substrate [129] and compressive stress due to residual hydrogen and sp2 carbon [126]. In experimental studies of diamond on glass, some films have shown good longterm adhesion with, for example, diamond films on both Corning 7059 and 7040 substrates that pass the tape test for adhesion several years after deposition and after repeated thermal cycling to 200⬚C. Other films, however, have shown spontaneous delamination after times ranging from several minutes to several days after removal from the deposition chamber. Adhesion

Figure 18 The coefficient of thermal expansion coefficients for diamond and Corning 7059 glass.

650

Reinhard

Table 4 Film Thickness Dependence for Diamond Film Adhesion to Glassa Thickness (␮m)

Adhered?

0.7 1.0 1.7 2.5

Yes Yes Yes No

a

Substrates are 5-cm-diameter Corning 7040 (娃Pyrex) glass.

trends for translucent diamond films on borosilicate glass such as pictured on the left side of Fig. 14 are shown in Tables 4 and 5 with adhesion becoming more problematic as both substrate temperatures increase and as film thickness increases. These trends are consistent with general expectations from Eqs. (25) and (26), given the CTE values of the glass substrates. Long-term adhesion of such films on 7059 glass was reported when diamond film thicknesses were less than 3 ␮m and deposition temperatures were less than 480⬚C [100]. VI.

PHOTON DETECTION AND EMISSION

This section briefly considers diamond’s role in electro-optics. Both radiation-induced electrical conductivity within diamond and photon emission from diamond have been used extensively for diagnostics of diamond electronic states. The diagnostic use of such phenomena is not within the scope

Table 5 Substrate Temperature Dependence and Thickness Dependence of Diamond Adhesion to Glassa Pressure (torr)

Temperature (⬚C)

Thickness (␮m)

Adhered?

15 13 11 9

505–510 480–483 445–455 440

2.3 2.1 1.3 1.0

No No Yes Yes

a Substrates are 5-cm diameter Corning 7059 glass. The substrate temperature was measured by optical pyrometry.

Diamond Film Optics

651

of this chapter but is treated in Chap. 3. The focus here is on applicationrelated issues. To the extent that photon absorption in diamond involves electronic transitions, the electric properties of diamond are altered. Photons with energies greater than the band gap, that is,with energies greater than 5.47 eV, will with some probability generate free carriers in diamond that cause a photoconductive effect. Electronic transitions involving states in the gap can also contribute to photoconductivity in diamond for photons of lessor energy. Likewise, electronic transitions between states, whether band to band or involving gap states, can with some probability involve photon emission. Consequently, both photon detection and photon emission applications are of interest. The use of single-crystal diamond for photon detection has a long but not particularly extensive history with early work on diamond photoconductivity being reported in 1923 [130]. However, as is the case for many other applications, the advent of CVD diamond provided an impetus for renewed interest. The wide band gap of diamond makes it of particular interest for UV detection [131–135]. An ultraviolet detector should preferably have no sensitivity to visible light, a large sensitivity to UV light, and a small dark current. A useful figure of merit is the quantum efficiency, QE, taken as the total number of electrons collected at the contacts divided by the number of incident photons, or as QE =

Iphh␯ Poptq

(27)

where Iph is the photo current, h␯ is the photon energy, and Popt is the optical power. The simplest UV detector structure is a two-terminal device consisting of diamond between two metal electrodes, an example of the so-called metal-semiconductor-metal (MSM) structure. The quantum efficiency for an MSM device is related to diamond material properties and the electrical bias according to [136] QE =

␩F (␮␶) L

(28)

where F is the electric field applied between the metal contacts, L is the separation between the contacts, ␮ is the carrier mobility, ␶ is the carrier lifetime, and ␩ is the fraction of incident power absorbed in the diamond,

␩ = (1 ⫺ R)(1 ⫺ e⫺␣t)

(29)

In an early study of CVD diamond MSM UV detectors by Binari et al. [133], a comparison was made between the performance of a natural diamond type IIa detector and that of a CVD diamond detector. In both

652

Reinhard

cases, the MSM structure consisted of two evaporated aluminum stripes separated by 350 ␮m on the surface of the diamond. Both devices showed an acceptable dark current, approximately 10⫺12 A, and both showed a substantially larger UV response than visible response. The quantum efficiency of the type IIa detector at a UV wavelength of 200 nm was approximately three orders of magnitude higher than the quantum efficiency in the blue portion of the visible spectrum (400 nm) and the quantum efficiency of the CVD diamond detector was approximately two orders of magnitude higher at 200 nm than at 400 nm. However, the two detectors differed significantly in the UV 200-nm quantum efficiency, which, when biased at 100 V, was 39% for the type IIa diamond and 0.09% for the CVD diamond. As is evident from Eq. (28), the quantum efficiency of an MSM photodetector is strongly dependent on the ␮␶ product of the sensor material. Polycrystalline materials in general tend to exhibit lower mobilities and lower lifetimes than their single-crystal counterparts and this is also the case for diamond. Therefore, a materials challenge for CVD diamond photodetectors is the achievement of higher ␮␶ products [137]. For a given bias voltage, the QE will also improve with decreasing electrode spacing. Using MSM structures with 20-␮m inter-electrode spacing, L, Chan et al. [134] have fabricated devices with five orders of magnitude difference in sensitivity between deep UV light and visible light. Another materials-related issue for CVD diamond UV detectors is persistent photoconductivity, which results in a memory effect that can have significant time constants, on the order of 1000 sec or longer. The origin of this effect is that electrons excited by the UV irradiation are trapped in deep traps that have long release times. Detectors can be reset, or bleached, by heating them sufficiently to release the trapped carriers. This phenomenon can be put to use, as proposed by Gonon et al. [138] for dosimetry, by measuring thermally simulated currents after high-energy photon absorption. However, for typical UV detection applications a persistent photoconductivity is undesirable. Chan et al. [134] reported a postgrowth methane-air treatment that reduces the duration of the persistent photoconductivity from in excess of 500 sec to less than 5 sec. Diamond electronic device structures other than the MSM, for example, Schottky barrier diamond photodiodes, are also applicable for photon detection [139,140]. A related application is photoconductive switching. Single-crystal type IIb diamond has been evaluated as a detector for several pulsed laser sources by Young et al. [141] and shown to have subnanosecond response times. Subnanosecond, kilovolt switching of microjoule optical pulses has been reported in insulating single-crystal diamond [142]. In the range of 0.22– 0.6 ␮m, defect and impurity photoabsorption provides the switching mechanism. As noted by Angus et al. [131], CVD diamond may have significant

Diamond Film Optics

Figure 19 Ref. 143.)

653

Typical waveform of a CVD diamond photoconductive switch. (From

advantages over natural diamond because of reproducible levels of vacancy and nitrogen defect complexes as well as increased simplicity of doping. Using CVD diamond, Aikawa et al. [143] demonstrated subnanosecond detection of light pulses from a KrF laser as shown in Fig. 19. With regard to electro-optic applications that involve photon emission, band-to-band recombination events in diamond would not be efficient sources of photons because diamond has an indirect band gap. As discussed by Yoder [144], however, another approach may be considered for diamond in that impurities may be intentionally added in order to add states within the large forbidden energy gap so as to provide optical transition energies. In fact, visible-light lasting has been demonstrated in single-crystal diamond by external pumping of electrons into such color centers [145]. Also, diamond based p-n diodes and Schottky diodes have demonstrated electroluminescence (EL) upon forward bias [146–148]. In these cases, the EL and cathodoluminescence spectrum coincide and reflect the impurity and defect content of the diamond. Fahrner et al. [149] used nitrogen implantation into natural type IIa single-crystal diamond and showed a modification of the EL spectrum.

VII.

SUMMARY

The unique intrinsic optical properties of diamond are realizable in polycrystalline CVD diamond in both thick- and thin-film form, so the material is of interest for optical applications involving photon transmission, photon

654

Reinhard

detection, and photon emission. With regard to transmission, special care is needed to avoid scattering caused by rough surfaces. Smooth surfaces are achieved either by polishing, by fine grains, or by oriented growth. Also necessary for high transmission are growth conditions that contribute neither appreciable impurities nor appreciable nondiamond carbon [150]. For thick films containing large grains, the mechanical strength is significantly less than for single-crystal diamond because of the large flaw sizes inherent in such structures. However, the strength is still greater than that of many other contending materials. The attractive surface attributes of diamond, including hardness and chemical resistance, may be added to more common optical substrates by thin-film diamond coatings [151]. For many substrates, stress in such films is controllable by balancing compressive thermal stress and tensile internal stress. As an indirect band gap material, photon emission heavily involves impurity and defect states. Such impurities can be intentionally added to provide color centers for the enabling of optical transitions. The large band gap allows ultraviolet photon sensing with essentially visibleblind response. Viability of CVD diamond for optical applications is tied to (1) deposition technology and the ability to achieve simultaneously useful growth rates and the necessary optical properties, (2) surface roughness values, (3) mechanical strength considerations, and (4) compatibility between CVD diamond and optical substrates.

REFERENCES 1. 2. 3. 4.

5. 6. 7. 8. 9. 10.

M Seal, WJP van Enckevort. Applications of diamond in optics. SPIE Diamond Opt 969:144–152, 1988. R Seitz. The end of the beginning. SPIE Diamond Opt 969:124–137, 1988. T Obata, S Morimoto. Free-standing diamond films—plates, tubes, and curved diaphragms. SPIE Diamond Opt 1146:208–216, 1989. G Davies. The optical properties of diamond. In: PL Walker Jr, PA Thrower, eds. Chemistry and Physics of Carbon. Vol 13. New York: Marcel Dekker, 1977, pp 2–143. CD Clark. Optical properties of natural diamond. In: R Berman, ed. Physical Properties of Diamond. Oxford: Clarendon Press, 1965, chap 11. ¨ ber Brechungsindizes und Absorptionkonstanten des Diamanten F Peter. U zwischen 644 und 226 m␮. Z Phys 15:358–368, 1923. RA Smith. Semiconductors. 2nd ed. Cambridge: Cambridge University Press, 1978. F Urbach. Phys Rev 92:1324, 1953. M Thomas, WJ Tropf. Optical properties of diamond. Johns Hopkins APL Tech Dig 14:16–22, 1993. A Feldman, LH Robins. Critical assessment of optical properties of CVD diamond films. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds.

Diamond Film Optics

11. 12.

13.

14. 15. 16.

17. 18. 19. 20.

21. 22.

23.

24. 25.

26.

27.

655

Applications of Diamond Films and Related Materials. Amsterdam: Elsevier Science, 1991, pp 181–188. RM Chrenko, HM Strong. Physical Properties of Diamond. General Electric Technical Information Series, Report 75CRD089, October 1975. MP D’Evelyn, EB Stokes, PJ Codella, BE Williams. Properties of CVD diamond for optical applications. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 547–550. DC Harris. Development of Chemical-Vapor-Deposited Diamond for Infrared Optical Applications. Status Report and Summary of Properties, Naval Air Warfare Center Weapons Division, China Lake, CA, Report NAWCWPNS TP 8210, 1994. OS Heavens. Optical Properties of Thin Solid Films. New York: Academic Press, 1955. CE Johnson, WA Weimer. Optical properties of microwave-plasma-assisted CVD diamond film. Diamond Opt II, SPIE 1146:188–191, 1989. AJ Gatesman, RH Giles, J Waldman, LP Bourget, R Post. Optical properties of polycrystalline diamond in the far-infrared. SPIE Diamond Opt 1325:170– 177, 1990. I Filinski. The effects of sample imperfections on optical spectra. Phys Status Solidi (b) 49:577–588, 1972. H Davies. Proc Inst Elec Engr 101:209, 1954. HE Bennet, JO Porteus. Relation between surface roughness and specular reflectance at normal incidence. J Opt Soc Am 51:123–129, 1961. XX Bi, PC Eklund, JG Zhang, AM Rao, TA Perry, CP Beetz Jr. Optical transmission and reflection of freestanding filament assisted CVD diamond films. SPIE Diamond Opt 1146:192–200, 1989. T Feng. Correlation of optical properties and surface roughness of polycrystalline diamond films. SPIE Diamond Opt 1146:159–165, 1989. LH Robins, EN Farabaugh, A Feldman. Determination of the optical constants of thin chemical-vapor deposited diamond windows from 0.5 to 6.5 eV. SPIE Diamond Opt 1534:105–115, 1991. C-F Chen, S-H Chen, T-M Hong, J-S Leu. Optical characterization of diamond synthesized using CH4-CO2 mixtures without supplying hydrogen gas. Thin Solid Films 253:162–167, 1994. H-W Ko, SE Hsu, SJ Yang, MS Tsai, YH Lee. Characterization of diamond films for optical coatings. Diamond Relat Mater 2:694–698, 1992. KM McNamara, KK Gleason, CJ Robinson. Quantitative correlation of infrared absorption with nuclear magnetic resonance measurements of hydrogen content in diamond films. J Vac Sci Technol A 10:3143–3148, 1992. KM McNamara, BE Scruggs, KK Gleason. Effects of impurities on the IR absorption of chemically vapor deposited diamond. Thin Solid Films 253: 157–161, 1994. M Hoinkis, ER Weber, MI Landstrass, MA Plano, S Han, DR Kania. Para-

656

28. 29. 30. 31.

32.

33. 34. 35.

36.

37.

38.

39. 40.

41.

42. 43. 44.

Reinhard magnetic nitrogen in chemical vapor deposition diamond thin films. Appl Phys Lett 59:1870–1871, 1991. M Nesladek, M Vanecek, LM Stals. Defect-induced optical absorption in CVD diamond films. Phys Status Solidi (a) 154:283–303, 1996. BE Williams, HS Kong, JT Glass. Electron microscopy of vapor phase deposited diamond. J Mater Res 5:801–810, 1990. W Zhu, AR Badzian, R Messier. Structural imperfections in CVD diamond films. J Mater Res 4:659–663, 1989. AT Collins, M Kamo, Y Sato. A spectroscopic study of optical centers in diamond grown by microwave-assisted chemical vapor deposition. J Mater Res 5:2507–2514, 1990. R Heiderhoff, R Spitzl, M Maywald, V Raiko, LJ Balk, J Engemann. Characterization of MPCVD-diamond films grown on porous silicon. Diamond Film Semiconductors, SPIE Proceedings, Vol 2151, 1995. SA Solin, AK Ramdas. Raman spectrum of diamond. Phys Rev B 1:1687– 1698, 1970. DS Knight, WB White. Characterization of diamond films by Raman spectroscopy. J Mater Res 4:385–393, 1989. BE Williams, JT Glass. Characterization of diamond thin films: diamond phase identification, surface morphology, and defect structures. J Mater Res 4:373–384, 1989. P John, DK Milne, IC Drummond, JIB Wilson, MG Jubber, JA Savage. Attenuated total reflectance infrared absorption in CVD diamond films. pp 209– 217, SPIE Diamond Opt V 1759:209–217, 1992. C Zheng, J Ran, Z Yuan. Decrease in surface roughness of diamond film synthesized by MWPCVD. In: M Yoshikawa, M Murakawa, Y Tzeng, WA Yarbrough, eds. 2nd International Conference on the Applications of Diamond Films and Related Materials. Tokyo: MYU, 1993, pp 745–750. J Rosa, M Vanecek, M Nesladek, LM Stals, V Vorlicek. Quantitative study of Raman scattering and defect optical absorption in CVD diamond films. Diamond Relat Mater 6:704–707, 1997. X Ying, Y Shen, H Xue, L Jianhai, Z Xing, J Xu, F Zhang. Determination of optical constants of thin films. Diamond Opt V SPIE 1759:218–223, 1992. O Stenzel, R Petrich, S Roth, B Mainz, W Scharff. Determination of the optical constants of fine grained diamond layers on silicon substrates using curve-fitting procedures. Diamond Relat Mater 2:704–707, 1992. R Petrich, O Stenzel. Spectrophotometric characterization of rough diamond layers on silicon. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 387–390. JR Seitz. Laser systems with diamond optical elements. US Patent 3,895,313, July 15, 1975. S Singer. The potential of diamond as a very high average power transmitting optical material. Diamond Opt SPIE 969:168–177, 1988. DC Harris. Properties of diamond for window and dome applications. SPIE 2286:218–228, 1994.

Diamond Film Optics 45. 46.

47.

48.

49.

50. 51.

52.

53.

54.

55.

56.

57. 58.

657

KJ Gray. Electromagnetic window properties of CVD diamond. SPIE Diamond Opt V 1759:203–208, 1992. G Lu, KJ Gray, EF Borchelt, LK Bigelow, JE Graebner. Free-standing white diamond for thermal and optical applications. Diamond Relat Mater 2:1064– 1068, 1992. Q Lichang, P Xin, Y Peichun, X Zhenwu, L Erkai, H Li. Preparation and characterization of optically high quality diamond film. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier, 1991, pp 213–216. L Rongzhi, S Hailiang, Y Zhen, T Sen, Z Hesun. Transparent diamond film deposited by optimized dc arc plasma jet. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier, 1991, pp 207–212. A Hirata, M Yoshikawa. Optical properties of diamond plates synthesized in arc discharge plasma jet chemical vapor deposition. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 379–382. XH Wang, L Pilione, W Zhu, W Yarbrough, W Drawl, R Messier. Infrared optical properties of CVD diamond films. J Mater Res 5:2345–2350, 1990. C Wild, W Mu¨ller-Sebert, T Eckermann, P Koidl. Polycrystalline diamond films for optical applications. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier, 1991, pp 197–205. RLC Wu, AK Rai, A Garscadden, P Kee, HD Desai, K Miyoshi. Synthesis and characterization of fine grain diamond films. J Appl Phys 72:110–116, 1992. DC Harris. Diamond optics: status for infrared applications. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 539–546. S Jin, JE Graebner, GW Kammlott, TH Tiefel, SG Kosinski, LH Chen, RA Fastnacht. Massive thinning of diamond films by a diffusion process. Appl Phys Lett 60:1948–1950, 1990. S Jin, LH Chen, JE Graebner. Thinning, shaping and patterning of CVD diamond by diffusional reactions. In: M Yoshikawa, M Murakawa, Y Tzeng, WA Yarbrough, eds. 2nd International Conference on the Application of Diamond Films and Related Materials. Tokyo: MYU, 1993, pp 189–196. S Jin, JE Graebner, M McCormack, TH Tiefel, A Katz, WC DautremontSmith. Shaping of diamond films by etching with molten rare-earth metals. Nature 362:882, 1992. M Yoshikawa. Development and performance of a diamond film polishing apparatus with hot metals. Diamond Opt III SPIE 1325:210–221, 1990. SK Choi, SK Jung, YH Jin, YS Park, SH Kim, JW Lee. Microstructure control and surface polishing of CVD diamond thin film for optical application. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds.

658

59. 60.

61. 62. 63. 64.

65. 66.

67. 68.

69.

70.

71. 72.

73.

Reinhard Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 383–386. A Hirata, H Tokura, M Yoshikawa. Smoothing of chemically vapor deposited diamond films by ion beam irradiation. Thin Solid Films 212:43, 1992. A Boudina, E Fitzer, G Wahl, H Esrom. Improvement in IR properties of chemically vapour-deposited diamond films by smoothening with KrF excimer radiation. Diamond Relat Mater 2:678–682, 1992. Z Tianji, DF Grogan, BG Bovard, HA Mcleod. Diamond polishing with argon and oxygen ion beams. Diamond Opt III SPIE 1325:142–151, 1990. BG Bovard, T Zhao, HA Macleod. Oxygen-ion beam polishing of a 5-cm diameter diamond film. Appl Opt 31:2366–2369, 1992. DF Grogan, T Zhao, BG Bovard, HA Mcleod. Planarizing technique for ionbeam polishing of diamond films. Appl Opt 31:1483–1487, 1992. KV Ravi, VG Zarifis. Laser polishing of diamond. Proceedings 3rd International Symposium on Diamond Materials, Electrochemical Society, Pennington, NH, 1993, p 861. SJ Pearton, A Katz, F Ren, JR Lothian. ECR plasma etching of chemical vapour deposited diamond thin films. Electron Lett 28:822–924, 1992. RN Chakraborty, DK Reinhard, PD Goldman. Etching of diamond wafers with electron-cyclotron-resonance plasmas. Electrochem Soc Extended Abstr 95-1:396–397, 1995. O Dorsch, M Werner, E Obermeier. Dry etching of undoped and boron doped polycrystalline diamond films. Diamond Related Mater 4:456–459, 1995. C Wild, N Herres, R Locher, W Mu¨ller-Sebert, P Koidl. Control of twin formation: a prerequisite for the growth of thick oriented diamond films. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 149–152. P Koidl, C Wild, R Locher, R Kohl. Smooth diamond films for optical applications. In: M Yoshikawa, M Murakawa, Y Tzeng, WA Yarbrough, eds. 2nd International Conf. on the Applications of Diamond Films and Related Materials. Tokyo: MYU, 1993, pp 87–94. H Maeda, M Irie, K Kusakabe, S Marooka. Nucleation of diamond by chemical vapor deposition in bias-enhanced microwave plasma. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 153–156. S Haq, DL Tunnicliffe, JA Savage. Optical characterization of textured microwave CVD diamond. Diamond Relat Mater 3:593–597, 1994. RLC Wu, AK Rai, A Garscadden, P Kee, HD Desai, K Miyoshi. Synthesis and characterization of fine grain diamond films. J Appl Phys 72:110–116, 1992. AV Khomich, PI Perov, VI Polyakov, IG Teremetskaya, VP Varnin, VI Konov, SM Pimenov, ED Obraztsova. Polycrystalline CVD diamond films with optical properties close to those of type II diamonds. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond

Diamond Film Optics

74.

75.

76.

77.

78. 79. 80. 81.

82.

83. 84.

85.

86.

659

Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 589–592. GS Yang, M Aslam, KP Kuo, DK Reinhard, J Asmussen. Effect of ultrahigh nucleation density on diamond growth at different growth rates and temperatures. J Vac Sci Technol B 13:1030–1036, 1995. GD Barber, WA Yarbrough, K Gray. Microstructural evolution and flaws in CVD diamond. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 551–556. TJ Valentine, AJ Whitehead, RS Sussmann, CJH Wort, GA Scarsbrook. Mechanical property measurements of bulk polycrystalline CVD diamond. Diamond Relat Mater 3:1168–1172, 1994. WH Glime, ED Case, DK Reinhard, J Mattavi. Deflection behavior of point loaded CVD diamond thin film diaphragms. Proceedings of the American Society for Composites 8th Technical Conference on Composite Materials. Lancaster, PA: Technomic Publications, 1993, pp 921–930. S Timoshenko, S Woinowsky-Krieger. The Theory of Plates and Shells. 2nd ed. New York: McGraw Hill, 1959. B Banerjee. Large deflection of a circular plate under a concentrated load— a new approach. J Ind Math Soc 33:57, 1983. BS Berry, WC Pritchet, JJ Cuomo, CR Guarnieri, SJ Whitehair. Internal stress and elasticity of synthetic diamond films. Appl Phys Lett 57:302–303, 1990. LED Flowers, ED Case, RM Brown, DK Reinhard. Burst pressure of diamond thin film diaphragms as a function of film diameter and thickness. Proceedings of the American Society for Composites 8th Technical Conference on Composite Materials, Technomic Publications, Lancaster, PA, 1993, pp 911–920. CJH Wort, JR Brandon, BSC Dorn, JA Savage, RS Sussmann, AJ Whitehead. Properties of CVD diamond domes. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 569–572. CA Klein. Diamond windows for IR applications in adverse environments. Diamond Relat Mater 2:1024–1031, 1992. GH Jilbert, CSJ Pickles, EJ Coad, JE Field. Diamond: an erosion resistant aerospace material. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 561– 564. JE Graebner. Inexpensive technique for estimating the thermal conductivity of CVD diamond from the optical absorption. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 557–560. RE Rawles, MP D’Evelyn. Optical properties of diamond at elevated temperatures. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshi-

660

87. 88.

89.

90.

91.

92.

93.

94. 95.

96. 97.

98.

99.

100.

Reinhard kawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 565–568. PK Bachmann, D Leers, DU Wiechert. Post-depositional diamond etching. Diamond Relat Mater 2:683–693, 1993. MB Moran, KA Klemm, LF Johnson. Antireflection/antioxidation coatings and polishing results for PECVD diamond optics. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. Amsterdam: Elsevier Science, 1991, pp 233–240. WD Partlow, RE Witkowski, JP McHugh. CVD diamond coatings for the infrared by optical brazing. In: Y Tzeng, M Yoshikawa, M Murakawa, A Feldman, eds. Applications of Diamond Films and Related Materials. 1991, pp 163–168. R Fuentes, LM Trolio, JE Butler. Anodic bonding of diamond to glass. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 581–584. JG Ran, CQ Zheng, XS Li, P Fan. Study on infrared glass window adhering to diamond film. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 577–580. Y Muranaka, H Yamashita, H Miyadera. Low temperaure (⬃ 400⬚C) growth of polycrystalline diamond films in the microwave plasma of CO/H2 and CO/ H2/Ar systems. J Vac Sci Technol A 9:76–84, 1991. A Joseph, Y Tzeng. Diamond coated quartz and sapphire optical windows. In: M Yoshikawa, M Murakawa, Y Tzeng, WA Yarbrough, eds. Second International Conference on the Applications of Diamond Films and Related Materials. Tokyo: MYU, 1993, pp 429–432. TP Ong, RPH Chang. Low-temperature deposition of diamond films for optical coatings. Appl Phys Lett 55:2063–2065, 1989. DJ Pickrell, W Zhu, AR Badzian, RE Newnham, R Messier. Near-interface characterization of diamond films on silica and silicon. J Mater Res 6:1264– 1277, 1991. AB Harker. Properties of low temperature plasma CVD diamond films. Diamond Opt II SPIE 1146:152–158, 1989. MC Costello, DA Tossell, DM Reece, CJ Brierley, JA Savage. Diamond protective coatings for optical components. Diamond Relat Mater 3:1137–1141, 1994. Y Liou, A Inspektor, D Knight, R Weimer, D Pickrell, AR Badzian, R Messier. Low temperature diamond deposition on glass. Diamond Opt II SPIE 1146: 12–20, 1989. MJ Ulczynski, DK Reinhard, M Prystajko, J Asmussen. Low temperature deposition of thin film diamond. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 41–44. M Ulczynski, DK Reinhard, M Prystajko, J Asmussen. Thin film diamond coatings on glass. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M

Diamond Film Optics

101. 102.

103. 104. 105. 106. 107. 108.

109. 110.

111.

112.

113.

114.

115.

116.

117.

661

Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 573–575. KE Nariman, DB Chase, HC Foley. Growth of free-standing diamond films on glass. Chem Mater 3:391–394, 1991. T Hartnett, R Miller, D Montanari, C Willingham, R Tustison. Intermediate layers for the deposition of polycrystalline diamond films. J Vac Sci Technol A 8:2129–2136, 1990. RH Doremus. Glass Science. New York: John Wiley & Sons, 1973. CTL Babcock. Silicate Glass Technology Methods. New York: John Wiley & Sons, 1977. DN Belton, SJ Harris, SJ Schmieg, AM Weiner, TA Perry. Appl Phys Lett 54:416, 1988. R Meilunas, MS Wong, KC Sheng, RPH Chang. Appl Phys Lett 54:2204, 1989. DJ Pickrell, PhD thesis, Pennsylvania State University, University Park, 1990. WL Hsu, DM Tung, EA Fuchs, KF McCarty, A Joshi, R Nimmagadda. Lowtemperature diamond growth in a microwave discharge. Appl Phys Lett 55: 2739–2741, 1989. Y Liou, R Weimer, D Knight, R Messier. Effect of oxygen in diamond deposition at low substrate temperatures. Appl Phys Lett 56:437–439, 1990. R Ramesham, T Roppel, C Ellis, DA Jaworske, W Baugh. Selective and low temperature synthesis of polycrystalline diamond. J Mater Res 6:1278–1286, 1991. H Kawarada, KS Mar, A Hiraki. Large area chemical vapor deposition of diamond particles and films using magneto-microwave plasma. Jpn J Appl Phys 26:L1032–L1034, 1987. S Jin, TD Moustakas. Growth of diamond thin films by ECR plasma-assisted CVD at low pressures and temperatures. Diamond Relat Mater 2:1355–1359, 1993. CR Eddy Jr, DL Youchison, BD Sartwell. Comparison of (CH4-O2)-H2 and CO-H2 plasmas for low temperature diamond film deposition by electron cyclotron resonance plasma-assisted chemical vapor deposition. Diamond Relat Mater 3:105–111, 1993. C-F Chen, S-H Chen, H-W Ko, SE Hsu. Low temperature growth of diamond films by microwave plasma chemical vapor deposition using CH4 ⫹ O2 gas mixtures. Diamond Relat Mater 3:443–447, 1994. TD Mantei, JJ Chang. Diamond deposition in a permanent magnet microwave electron cyclotron resonance discharge. J Vac Sci Technol A 10:1423–1425, 1992. Z Ring, TD Mantei, S Tlali, HW Jackson. Plasma synthesis of diamond at low temperatures with a pulse modulated magnetoactive discharge. Appl Phys Lett 66:3380–3382, 1995. T Yara, H Makita, A Hatta, T Ito, A Hiraki. Low temperature fabrication of diamond films with nanocrystal seeding. Jpn J Appl Phys 34:L312–L315, 1995.

662

Reinhard

118.

A Hatta, K Kadota, Y Mori, T Ito, T Sasaki, A Hiraki, S Okada. Pulse modulated electron cyclotron resonance plasma for chemical vapor deposition of diamond films. Appl Phys Lett 66:1602–1604, 1995. M Ihara, H Maeno, K Miyamoto, H Komiyama. Diamond deposition on silicon surfaces heated to as low as 135⬚C. Appl Phys Lett 59:1473–1475, 1991. W Tsai, GJ Reynolds, S Hikido, CB Cooper III. Electron cyclotron resonance plasma-enhanced filament-assisted diamond growth. Appl Phys Lett 60:1444– 1446, 1992. I Watanabe, H Matsumoto. Low temperature synthesis of diamond films by rf plasma CVD. Jpn Soc Promot Sci 6:197–202, 1993. R Csencsits, CD Zuiker, DM Gruen, AR Krauss. Grain boundaries and grain size distributions in nanocrystalline diamond films derived from fullerene precursors. Solid State Phenom 51–52:261–270, 1996. SJ Harris, AM Weiner. Pressure and temperature effects on the kinetics and quality of diamond films. J Appl Phys 75:5026–5032, 1994. DE Patterson, CJ Chu, BJ Bai, ZL Xiao, NJ Komplin, RH Hauge, JL Margrave. Diamond Relat Mater 1:768, 1992. MJ Ulczynski, B Wright, DK Reinhard. Diamond coated glass substrates. Dia Rel Mat 7:1639–1646, 1998. H Windischmann, GF Epps, Y Cong, RW Collins. Intrinsic stress in diamond films prepared by microwave plasma CVD. J Appl Phys 69:2231–2237, 1991. CA Gamlen, ED Case, DK Reinhard, B Huang. Adhesion of polycrystalline diamond thin films on single-crystal substrates. Appl Phys Lett 59:2529–2531, 1991. JE Field. The Properties of Diamond. London: Academic Press, 1979. SE Rosenberg, PY Wong, IN Miaoulis. Analysis of depositions stress during thin-film growth on a relaxing substrate. J Appl Phys 77:6273–6277, 1995. B Gudden, R Pohl. Z Phys 17:331, 1923. JC Angus, CC Hayman, RW Hoffman. Diamond and ‘diamond-like’ phases grown at low pressure: growth, properties and optical applications. Diamond Opt SPIE 969:2–13, 1988. DR Kania, MI Landstrass, MA Plano, LS Pan, S Han. Diamond radiation detectors. Diamond Relat Mater 2:1012–1019, 1992. SC Binari, M Marchywka, DA Koolbeck, HB Dietrich, D Moses. Diamond metal-semiconductor-metal ultraviolet photodetectors. Diamond Relat Mater 1:1020–1023, 1992. SSM Chan, RD McKeag, MD Whitfield, RB Jackman. UV photodetectors from thin film diamond. Phys Status Solidi (a) 154:445–454, 1996. VV Tokiy, VI Timchenko, VA Soroka, N Tokiy, BV Spitsyn, LL Bouilov. Diamond film metal-semiconductor-metal photodetector. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 145–148. MA Karim. Electro-Optical Devices and Systems. Boston: PWS-Kent, 1990.

119.

120.

121. 122.

123. 124. 125. 126. 127.

128. 129. 130. 131.

132. 133.

134. 135.

136.

Diamond Film Optics 137.

138.

139.

140.

141.

142. 143.

144. 145. 146. 147. 148. 149.

150.

151.

663

LS Pan, DR Kania, S Han, JW Ager III, M Landstrass, OL Landen, P Pianetta. Electric transport properties of undoped CVD diamond. Science 255:830– 833, 1992. P Gonon, S Prawer, Y Boiko, D Jamieson. Electrical conduction in polycrystalline diamond and the effects of UV irradiation. Diamond Relat Mater 6: 860–864, 1997. M Marchywka, JF Hochedez, MW Geis, DG Socker, D Moses, RT Goldberg. Ultraviolet photoresponse characteristics of diamond diodes. Appl Opt 30: 5011–5013, 1991. G Popovici, MA Prelas, S Khasawinah, T Sung, VI Polyakov, PI Perov, NM Rossukanyi, AI Rukovishnikov, AV Khomich, AI Krikunov, IG Termetskaya, VP Varnin. Diamond photovoltaics: characterization of CVD diamond film based heterostructures for light to electricity conversion. In: A Feldman, WA Yarbrough, M Murakawa, Y Tzeng, M Yoshikawa, eds. Applications of Diamond Films and Related Materials: Third International Conference. NIST Special Publication 885, 1995, pp 99–102. JF Young, LA Vermuelen, DJ Moss, HM van Driel. Subnanosecond timeresolved photoconductive response of semiconducting diamond. Appl Phys Lett 42:434–436, 1983. PS Panchhi, HM van Driel. Picosecond optoelectronic switching in diamond. IEEE J Quantum Electron QE-22:101–107, 1986. Y Aikawa, K Baba, N Shohata. Characteristic of photoconductive switch with diamond thin film. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 679–682. MN Yoder. Artifact diamond. Its allure and significance. SPIE Diamond Opt 969:106–113, 1988. S Rand, L Deshazer. Visible color-centered laser in diamond. Opt Lett 10: 481–483, 1985. JF Prins. Mater Sci Rep 7:271, 1992. N Fujimori, Y Nishibayashi, H Shiomi. Jpn J Appl Phys L30:1728, 1991. Y Taniguchi, K Hirabayashi, K Ikoma, N Kurihara, M Matsushima. Jpn J Appl Phys 28:1848, 1989. WR Fahrner, B Burchard, AM Zaitsev, AA Melnikov, AV Denisenko, VS Varichenko. Optoelectronic Devices Based on Diamond. In: S Saito, N Fujimori, O Fukunaga, M Kamo, K Kobashi, M Yoshikawa, eds. Advances in New Diamond Science and Technology. Tokyo: MYU, 1994, pp 737–740. AM Zaitsev. Optical properties. In: MA Prelas, G Popovici, LK Bigelow, eds. Handbook of Industrial Diamonds and Diamond Films. New York: Marcel Dekker, Inc. 1998. DK Reinhard, JR Kleindienst. Diamond films as optical coatings for silicon. New Diamond Frontiers Carbon Tech 10:13–23, 2000.

Index

Ablation, laser, 333–335, 354–358 Abrasive wear, 439, 440 Absorption, optical band-to-band, 609 defect induced, 624–626 due to C — H bonds, 622, 632 impurity related, 613–616, 621–624 infrared, 612, 622–624, 632 intrinsic, 613 multiphoton, 336 resonant, 376 ultraviolet, 609 Urbach tail, 611 weak tail, 611 Accademio del Cimento, 18 Acceptors (see Boron) Acetylene, 66, 67, 109, 164, 165, 214, 247, 269–272, 288, 296, 304 Acoustic applications, 12 Actinometry, 93, 273 Activation energy, 250 Adhesion diamond to substrate, 437, 458–475, 645–650 metal to diamond, 556 treatments, 476

AES (see Auger electron spectroscopy) AFM, 551 Allemanna Svenska Elektrisia Aktiebolaget, 18, 20 Alloys high-silicon aluminum, 425, 444, 475, 482, 484 low-silicon aluminum, 478, 484 ␣ parameter, 38, 81, 207, 252–260, 264, 265, 266 Alumina, 519 Aluminum (see also Alloys) as electric contact, 586, 599 as via fill, 556 Aluminum nitride, 568 Amorphic diamond, 354, 355 Amorphous carbon, 342, 354, 462 Angstrom’s method, 536 Anharmonic pumping, 380 Annealing, 590, 59 Antireflection diamond, as, 634 layers, on diamond, 636 Applications of diamond acoustic, 12 cold cathode emitter, 12 665

666 [Applications of diamond] cutting tools, 10, 137, 141 (see also Tools, cutting) detectors, 12, 651, 652 diodes, 13, 580, 652 electro chemical, 13 electrodes, 10 electron and photon sources, 11 electronic active devices, 10, 579–602 passivation layers, 10 microelectronics, 14 packaging, 10, 561–566 gyrotron windows, 10, 11, 627 heat sinks, 10 lasers, 10, 11, 607, 627 nuclear particle dectors, 10 optical coatings, 10, 11, 637–648 optical windows, 10, 11, 627 sensors, 10, 12 speakers, 10 surface acoustic wave, 10, 12 thermal management, 11, 137, 141, 513–522 thermistors, 1, 10 transistors, 13, 579–602 transmission (see also Transmission, optical), 11 ultraviolet dectors, 10, 12, 651, 652 vacuum microelectronics, 12 wear surfaces, 10, 439–441 x-ray masks, 11 Applied Science and Technology (see ASTek, Inc.) Applied Sciences Inc., 532 Arcjet CVD (see also D-c arcjet CVD), 5, 56, 59, 66, 93, 72, 141–210, 527 Arrhenius, 249 Arsenic as dopant, 584 ASTeX Inc., 229 Atomic density of diamond, 5 Atomic Force Microscopy (see AFM) AT&T Bell Laboratories, 545

Index Auger spectroscopy, 35, 78, 343, 364, 621 Bachmann diagram, 242, 243, 248 Balagia’s figure of merit, 579, 580 Band gap (see also Energy gap) direct gap, 611 indirect gap, 610 Bell Laboratories, 21 Beryllium oxide, 519 Bias enhanced nucleation, 237, 238, 524 Binder, metallic, 426, 460, 470, 474 Bipolar transistor, 585 Blackbody radiation subtraction of, 413 Boltzmann’s constant, 374 Boltzmann distribution, 380 Boron activation energy of, 582 as acceptor, 580, 586, 589–592, concentrations of, 590 implantation of, 592 in Hope diamond, 615 in type IIb diamond, 614 ionization energy, 580 optical effects of, 613, 621 Boudouard reaction, 389 Boundary layer, 156–159 Brazing, 430, 450 Breakdown, electric field, 579 Bulk modulus, of diamond, 520 12

C (see Carbon-12) C (see Carbon-13) C — H stretch modes, 622 C2, 96, 97, 160, 161, 162, 269, 389 C2H2 (see Acetylene) Carat volume, 5 weight, 5 Carbon, non-diamond amorphous, 454, 525 atomic, 106, 108, 143, 164 carbon black, 353, 362 fullerene, 525 13

Index [Carbon, non-diamond] glassy, 352 graphitic, 525 phase diagram, 2, 19, 338, 350 Carbon-12, 352, 364, 412 Carbon-13, 352, 364, 412 Carbon conversion efficiency, 142, 184– 187, 223, 225, 250, 251 Carbon tool steels, 426 Carbonaceous precursors, 341, 364 Carrier concentration, 581 mobility, 9, 581 CARS (see Coherent anti-Stokes Raman spectroscopy) Case Western Reserve University, 22, 23 Cast iron milling, with diamond, 497 polishing diamond with, 630 Catalysts, 350, 525, 532 Cathodeluminescence, 34, 653 Cemented carbides, 426 Ceramic bisque, 442 green, 442, 447 CH3 (see Methyl) CH4 (see Methane) Chemical resistance, of diamond, 439, 520, 530 Chemical vapor deposition (see also Arcjet, Combustion, Laser, Hot filament, Microwave), 19, 22 high-pressure, 18 laser excited, 5, 378 low-pressure, 73 thermal plasma, 141 Chlorine, 267, 268 C/H/O diagram, 242, 243 Circuits, diamond, 594–598 Cleavage strength, of diamond, 7 Cleavage, through grains, 457 CO as carbon source, 270, 288, 364, 365 energy levels of molecules, 378 laser spectrum, 387

667 [CO] laser transitions, 378, 379 optical pumping of, 386 single crystal, 410 vibrational mode, 388 Cobalt, 426, 460, 463 as crystalline substrate, 343 Coherence length, 331 Coherence time, 330 Coherent anti-Stokes Raman spectroscopy, 89, 91, 95, 132, 133, 275, 276, 278, 288 Coherent light, 328–332 Color, of diamond blue, from boron, 613 brown, from nitrogen, 615 brown, from sp2 bonds, 626 green, from irradiation, 616 pink, from manganese, 616 yellow, from nitrogen, 613–615 Color centers, 653 Combustion CVD, 5, 56, 68, 96–98, 109, 303–323 Competitive growth, 76 Composites carbon-epoxy, 488 graphite-copper, 534 metal-matrix, 425, 444, 479, 532, 533 Compressibility of diamond, 6 Conduction band, 610 Conductivity electrical of diamond (see also Electronics, diamond), 7, 49 thermal (see Thermal conductivity) Contacts (see also Schottky) ohmic, 584, 591, 599 Contamination by filament, 621 in feedgas, 621 Conversion factors, energy, 361 Copper as crystalline substrate, 343 as electrical contact, 600 as heat sink, 567 as via fill, 556 Correlation function, 329

668 Cost, of diamond, 341 Crack propagation, 464 velocity, 464 CRDS (see Ring-down spectroscopy) Crystalline Materials Corp, 555 CTE (coefficient of thermal expansion), 529 of diamond, 7, 436, 437, 518, 530, 649 of glass, 641, 649 Cubic boron nitride, 429 Cubo-octohedral diamond, 368 Cullinan diamond, 615 Cutting, of diamond, 450, 540, 541 Cutting tools (see Tools) CVD (see Chemical vapor deposition) DARPA, 561 D-c arcjet anode, 149, 152 cathode, 149, 152, 153 CVD, 59, 68, 70, 92, 98, 106 instabilities, 158 nozzle, 154–156, 160 DeBeers, 449 Defects, in diamond absorption spectra effects, 624–626 interfacial, 464 Degree of mutual coherence, 329 Delamination, 460, 468, 645, 648–650 Density, of diamond, 5, 520, 530 Detectors ultraviolet, 651, 652 Diamond classification Type Ia, 613–615 Type Ib, 613–615 Type IIa, 592, 613–615 Type IIb, 613–615 Diamond grit, 365 Diamond like carbon, 338, 342, 486 Diamond properties band-gap, 13, 610, 611 boron doped (see also Boron), 13 cleavage strength, 7 compressibility, 6

Index [Diamond properties] density, 5, 520, 530 electrical conductivity, 7, 49, 581 energy gap, 39, 579, 610–613 Hall mobility, 8 hardness, 6, 433, 434, 530 optical transmission (see also Transmission, optical), 8, 9 refractive index, 8, 609 resistivity, 7, 514, 519, 520, 530 specific heat, 7, 520 thermal conductivity (see also Thermal conductivity), 7, 9, 13, 43, 45, 46, 49 thermal diffusivity, 45 Diamond seeds, 352 Diamond synthesis (see Hot-filament CVD, Arcjet CVD, Combustion CVD, Microwave CVD, rf plasma CVD, Laser-assisted CVD) Diamonex, 137 Diaphragms, diamond, 627–629, 631– 636 Diatomic molecule energy storage, 326, 376 single-quantum processes, 382 Dielectric, diamond breakdown (see Electrical breakdown) constant, 520, 530, 579 loss tangent, 520, 530 strength, 530 Diffusion of boron, 590 of carbon, 345 Dipole moment, 382 DLC (see Diamond like carbon) Domes, diamond, 608 Doner, 584 Dopant Electronics Materials Research, 555 Doping, of diamond, 586, 590, 598–600 Drilling, of diamond, 540, 541, 555, 556 Drilling tools (see Tools, rotary) Dymalloy, 531–531

Index EEDF (see Electron energy distribution function) EELS (see Electron loss spectroscopy) Effective mass, 580 Electric breakdown field, 530 Electro-chemical, 13 Electroluminescence, 653 Electron energy distribution function (EEDF), 256–288 Electron density, 278, 293, 294 Electronics, diamond, 10, 141, 579– 602 Electron loss spectroscopy (see Film characterization) Electron-photon interactions, 344 Electro-optics (see Optoelectronics) Emission spectrum, 399 End mills (see Tools) Energy, conversion factors, 361 Energy cost (see Joules per carat) Energy dispersion spectrum, 392, 395, 402 Energy distribution Boltzmann, 380 Treanor-Rich-Rehm, 380 Energy gap, 39, 579, 610–613 Energy transfer (see also Vibrations) Epitaxial heteroepitaxial, 79, 80, 461 homoepitaxial, 1, 75, 148, 581, 589, 594 Erosion rain resistance, 634, 635 sand resistance, 635 Etching, of diamond ECR plasma, 631 ion beam, 584 mesa, 592 reactive ion, 550, 594 Ethylene, 339, 362 Excimer (see Lasers) Excitation electronic, 326 rotation, 326 vibration, 326

669 Failure modes of tools, 448 Ferrous, machining by diamond, 497 FET (see Field effect transistor) Field effect transistor, 582–599 Filament carburization, 124 contamination, 125 materials rhenium, 122, 128 tantalum, 122, 128 tungsten, 58, 119, 122, 128 poisoning, 101, 102, 129 sooting, 129 Film characterization, 21–54 Auger electron spectroscopy, 35, 78, 343, 364, 621 cathodoluminescence, 34 electron loss spectroscopy, 30, 33, 34, 78 high-resolution electronmicroscopy, 33 laser microscope mass analysis spectroscopy, 35 low-energy electron diffraction (LEED), 21, 35 Raman spectroscopy (see also Raman), 28, 148, 150 Rutherford backscattering, 36, 78, 127 scanning electron microscopy (see also SEM), 28, 30, 40, 78, 180, 182, 184, 201 scanning force microscopy, 47–49 scanning probe measurements, 47, 48 scanning thermal microscope, 49, 50 scanning tunneling microscopy, 47, 48 secondary ion mass spectrometry, 35, 36, 37 thermal conductivity measurements (see also Thermal conductivity), 46 conventional method, 47 thermo-optical method, 47 transmission electron microscopy, 28, 30, 32, 33, 78, 340, 342

670 [Film characterization] X-ray diffraction, 28, 36, 37, 40, 41, 78, 264 Finite element analysis, 288, 558 Flame temperature, 98 Flip-chip, 557–560 Flourine, 267–268, 526 Fourier analysis, 332 Fourier transform infrared analysis, 95, 98, 132, 269 Fracture, modes of, 457 Fracture toughness, of diamond, 435, 438–441 Freestanding diamond, 177, 627–629 Friction, coefficient for diamond, 433, 434, 499 FTIR (see Fourier transform infrared analysis) Fullerene, 525 Gallium arsenide, 569 Gas chromatography, 88 Gem diamond, 609 General Electric, 2, 17, 18, 20, 326, 352 Germanium, as substrate, 354, 355 Glass (see also Substrates) alkali lead, 641, 642 borosilicate, 641–650 TM Pyrex, 650 soda lime, 641, 642 Gold as electric contact, 590, 591, 599 Grain boundaries, in diamond, 458, 553 Grains cleavage through, 457 in PCD, 427 Graphite disordered, 340 machining, 486 microcrystalline, 343 transformation of diamond to, 636 Growth, of diamond on diamond seed, 405–409

Index Growth rate, 71, 72, 141, 142, 147, 148, 173, 174, 176, 181, 185, 186, 190, 201, 207, 211, 225, 242–247, 249–252, 260–265, 341 calculation of, 373, 374 Gyrotron (see Windows, diamond) Hall mobility, of diamond, 8 Hard coatings, 440 Hardness, of diamond, 6, 433, 434, 530 Harvard University, 18 Heat capacity of diamond, 7, 516 Heat exchangers, 533 Heat sink, 10, 512, 516, 518, 529, 540, 565 Heat spreader, 513–522 Hewlett Packard, 561 High-pressure, high temperature diamond, 3, 18, 326, 349, 352, 586 High-resolution electron microscopy, 33 Hole concentration, 580 mobility, 581 Homoepitaxy, 1, 67, 75, 79, 80, 148, 266 Hope diamond, 615 Hot-filament CVD, 5, 24, 56, 58, 66, 67, 68, 70, 75, 88, 89, 98, 99–105, 119– 140, 368, 527, 528 HREM (see High-resolution electron microscopy) Hydrogen abstraction, 64, 65, 67, 69, 238, 239, 240, 268 atomic, 59, 60, 64, 65, 71, 72, 87, 89, 93, 95, 100, 101, 104, 105, 108, 142, 143, 146, 160, 161, 170, 213, 214, 238, 239, 240, 242, 247, 262, 273, 274, 276, 356, 363

Index [Hydrogen] dissociation, 58, 102, 106, 121, 128, 143, 149, 157, 213, 238, 291, 292 in polycrystalline diamond, 622 molecular flow rate, 370 optical effects, 621 p-type diamond, 594, 599 recombination, 105, 125 surface passivation, 369 IBM, 558 Impurities absorption spectra effects, 621–624 in microwave CVD diamond, 622 Incoherent light, 328–332 Infrared absorption (see Optical absorption) Infrared diode laser absorption spectroscopy, 88 Infrared laser absorption, 130 Infrared spectrum gas phase, 396 Insulated gate, 588 Interface (see also Adhesion, Microstructure) chemical bonding, 462 degradation, 476 liquid-solid, 352 Interference fringes, 331 Ion implantation, 343, 585, 591, 592 Ionization energy of boron acceptors, 580 of molecules, 359 Iron, as catalyst, 350 Johnson’s figure of merit, 579, 580 Joules per carat, 341, 350, 376, 408, 416 Kennametal, Inc., 506 Key’s figure of merit, 579 Knoop hardness of several optical materials, 360 Knoop indenter, 6

671 Langmuir probe, 279 Laser chemical vapor deposition, 366– 369 Laser driven reactions, 327, 370 Laser excited chemical vapor deposition (see also Optical pumping) CO lasing lines used in, 392 experimental setup, 391 Laser-induced fluorescence, 89, 91, 96, 97, 110, 132, 160, 161, 273, 275, 276, 278 Laser microscope mass analysis spectroscopy, 35 Laser physical vapor deposition (see Ablation) Laser-plasma technique, 341 Laser pumping, 336, 337 Laser pyrolysis, 333 Lasers argon ion, 352 CO (see also CO), 379 CO2, 348 excimer, 348, 371, 547 for diamond synthesis, 325–417 high power, 607, 627 Nd-YAG, 342, 354, 547, 555 polishing with, 546–550 Q-switched, 342, 354, 386, 547 types of, 337 YAG, 348 Lawrence Livermore National Laboratory, 531 Lead as electrical contact, 600 LEED (see Film characterization) LIF (see Laser-induced fluorescence) Liquid immersion of substrates, 350– 353 Liquid-solid interface, 352 Loss tangent (see Dielectric) Low-energy electron diffraction (see Film characterization) Low-temperature, deposition, 408, 528, 529, 643 Lubricity, of diamond, 437

672 Mandrill (see also Substrate), 526 Manganese optical effects, 616 Mass spectroscopy, 88, 97, 131 Maxwell-Boltzmann distribution, 374 MCM (see Multichip module) Membranes (see Diaphragms) MESFET, 588 Metallization, of diamond, 553–557 Metal matrix (see Composites) Methane, 362 bending mode of, 394 emission from, 401 energy levels, 397, 404 hot bands, 400 stretch mode of, 393 Methyl radicals, 62, 64, 67, 70, 71, 87, 89, 93, 95, 100, 104, 106, 107, 109, 120, 130, 131, 134, 136, 146, 164, 214, 238, 242, 247, 268, 269, 271, 273, 287, 288, 296, 363, 407 Michigan State University, 227 Microcrystalline diamond, 456 Microcrystalline graphite, 343 Microporosity, in diamond, 454 Microstructure, of diamond, 444 interfacial, 463 Microsubstrates, Corp., 555 Microwave (see also Packaging) plasma enhanced assisted CVD, 5, 24, 56, 75, 92, 95, 99, 108, 126, 211–301, 528 power density, 279, 280, 292, 293 Microwave applicators, 215, 221 Microwave reactor types ASTeX, 229–231, 271 Berkeley, 231, 232 cavity, 227–229, 266, 277, 278, 280 France bell jar, 232, 233, 273 ellipsoidal, 233, 234 jet torch, 234, 235 magneto/ECR, 235–237 surface-wave, 234 tubular, 226, 227, 266, 269, 271, 273, 286

Index Mirage Effect (see Photothermal deflection) Mobility of holes in diamond, 581, 582 Modeling numerical, 285–296 thermal, of packaging, 558–561 Molecular beam mass spectroscopy, 88, 92, 100, 131, 271 Morphology, of diamond, 80–82, 86, 207, 223, 365, 452, 453 MOSFET, 591 MPD (see Multiphoton dissociation) Multichip module, 511, 512, 514, 521, 557, 562–569 Multiphoton absorption, 335 Mutual coherence function, 329 Nanocrystalline diamond, 268, 389 Nanoindentor, 633 National Institute for Research in Inorganic Materials (see NIRIM) National Institute of Science and Technology (NIST), 539 Navy, U.S., 629 Nickel as catalyst, 350, 352 as crucible, 351 as single crystalline substrate, 343 NIRIM, 24, 226 Nitrogen aggregates of, 613 effect on synthesis, 265 effect on thermal conductivity, 519 ESR active, 613 in polycrystalline diamond, 622 optical effects, 613, 621 NorCool, 569 Norton Diamond Film, 227, 448, 506, 569, 629 n-type diamond, 601 Nuclear particle dectors, 10 Nucleation, 37, 55, 73–79, 82–84, 110, 136 by abrasion, 83, 524 by carbon fibers, 525

Index [Nucleation] by fullerenes, 525 by metallic powder, 524 by substrate bias, 83–85, 136, 237, 238, 524 carbide layer, 78 density, 73, 74, 78, 82, 84, 86, 461, 523, 524 homogeneous, 74 incubation period, 75, 76 rates, 76, 77, 82 three-dimensional, 75, 76 ultrasonic, 524 OES (see Optical emission spectroscopy) Ohmic (see Contacts) Optical (see Absorption, Reflection, Substrates, Transmission) Optical absorption (see Transmission) Optical coatings, 10, 21, 355, 637–648 Optical emission spectroscopy, 92, 95, 96, 97, 136, 276, 278, 288 Optical materials, properties, 360 Optical pumping, 336, 337, 376–415 for diamond growth, 385–409 of liquids, 409–415 Optical reflection (see Reflection) Optical transmission (see Transmission) Optics, diamond, 607–654 Optoelectronic (see also Packaging) uses of diamond, 650–653 Overtone spectrum, 414 Oxidation of diamond, 438, 530, 636 Oxyacetylene, 340, 342 Oxygen growth, 260–265 optical effects, 624 Oxygen, in polycrystalline diamond, 622 Ozone, 336 Packaging (see also MCM) microwave, 521, 569 optoelectronic, 561, 562

673 [Packaging] radiofrequency, 521 three-dimensional, 521, 562–566 PCD tools, 427, 428 Permeable base transistor, 587 Phase diagram (see Carbon) Phase stability, 438 Phonon absorption multiphoton, 612, 623 single photon, 623 Phonon-phonon interactions, 344 Photochemistry, 359 Photoconductivity in diamond, 651–653 switching, 652, 653 Photodissociation, 362 Photolysis, 327, 339, 358–376 Photon detection in diamond, 650–652 emission from diamond, 653 infrared dissociation, 339 multiphoton dissociation, 336, 358 single photon absorption, 376 Photon-electron interactions, 344 Photon-phonon interactions, 344 Photothermal deflection, 538 radiometry, 537, 538 Physical Chemistry Institute, 23 Planarization, of diamond, 550–552 Plasma plume, 335 p-n junction, 580 Poisson’s ratio, 520, 530, 634 Polarization, 329 Polishing, of diamond chemical assisted, 545–546 for optics, 629, 630 ion beam, 549 laser, 546–550 mechanical, 542,543 thermochemical, 543–545 Polycarbonate, 496 Post deposition finishing, 447, 449 Precursors, for diamond, 364, 373–375 p-type diamond, 590, 594

674 Pyrolysis, 327, 339, 353 Pyrometry, 650 QQC process, 347–349, 353, 417 Quantum efficiency, 651 Quantum number, vibrational, 382, 396 Quantum yield, 362 Quench rate, 351 Radiometry (see Photothermal) Raman diamond peak, 340 D peak, 340, 341, 357 G peak, 340, 357 micro, 351, 357 quality factor, 626 spectra, 28, 180, 181, 190, 193, 203, 248, 340, 357, 358, 367, 394, 455, 456, 646 spectroscopy, 28, 148, 150, 343, 454 Rate constants for energy transfer, 411 of elementary reactions, 371 Raytheon, 555, 628 RBS (see Rutherford backscattering) Reactor scale-up, 98, 99, 119, 137, 141, 170–173 Recrystallized diamond, 351 Reflection, optical, 609 Refractive index of diamond, 8, 609 of several optical materials, 360 Sellmeier equation, 609 Relaxation time, 344, 345 REMPI (see Resonance-enhanced multi photon ionization) Resistivity, electrical of diamond, 7, 514, 519, 520, 530 Resonance defect, 383 Resonance-enhanced multiphoton ionization, 89, 90 Rf plasma CVD, 68, 98, 108, 141, 160, 171, 176, 198 Rhenium, 122, 128 Ring-down spectroscopy, 133 Rutherford backscattering, 36, 78, 127

Index Saint Gobain Industrial Ceramics, 448, 506 Sandia National Laboratories, 567 Saturation velocity, 580 SAW device (see Surface acoustic wave device) Scanning electron microscopy (see SEM) Scanning force microscopy (see SFM) Scanning tunneling microscopy (see STM) Scattering optical effects, 616–620, 626, 629, 639, 640 Schmidt number, 144 Schottky contacts, 584, 586 photodiodes, diamond, 652 Seals, diamond coated, 498 test results, 500 Secondary ion mass spectroscopy (see SIMS) Seeding (see Nucleation) Sellmeier equation (see Refractive index) SEM, 28, 30, 40, 78, 180, 182, 184, 201, 261, 267, 351, 356, 367, 393, 428, 448, 451, 453, 458, 467, 469, 470, 474, 476, 478, 493, 552 Semiconductor diamond as, 580, 598 Sensors, 12 SFM, 47, 48, 49 Shear modulus, diamond, 520 Silane, 339 Silica (see Substrates) Silicon (see Substrates) optical effects, 624 Silicon carbide diamond interface, 642 SIMS, 35, 36, 37, 591 Sooting, 109, 129 sp2 bonds, 4, 27, 28, 29, 71, 73, 170, 239, 242, 260, 263, 454 optical effects, 624, 625 sp3 bonds, 4, 28, 29, 73, 148, 239, 452

Index Spalling, 460 Specific energy, 141 Specific heat, of diamond, 7, 520 Stagnation flow, 347 Stainless steel, as substrate, 348 Stark-Einstein Law, 362 STM, 49, 50 Strain temperature of glass, 641 Strength of diamond burst strength, 520 compressive strength, 435 flexural strength, 633 mechanical strength, 519, 633–636 rupture strength, transverse, 435 shear strength, 530 tensile strength, 434, 530 Stress due to CTE mismatch, 437, 646–649 in films, 645–650 Substrate bias, 196–200 Substrate holder, 217–219 Substrate temperature, 85, 213, 214, 217, 218, 223, 240, 249, 250, 262 Substrates boron nitride, 75 diamond, 518–522 germanium, 355 glass, 641–650 optical, 637–650 silica, 641, 643 silicon, 78, 641, 642 tungsten carbide, 432 ZnS, 639 ZnSe, 639 Sun Microsystems, 531 Surface acoustic wave, 10, 12 Surface roughness, 447, 550, 629, 632 optical effects of (see Scattering) Tantalum, 122, 123, 128 TEM, 28, 30, 32, 33, 78 bright field, 340 of diamond, 340, 342 of graphite, 340

675 Temperature distribution, 345–347 gas, 275–278, 291, 292–294, 348, 385 gradient, 345 substrate, 85, 213, 214, 217, 218, 219, 223, 240 vibrational, 380 Tersoff potential, 63, 64 Textured growth, 37, 80, 81, 252–260, 631, 632 Thermal conductivity measurement methods, 535 of aluminum nitride, 514 of beryllium oxide, 514 of copper, 514 of diamond, 7, 9, 13, 43, 45, 46, 49, 435, 437, 514, 520, 530, 580, 636 of silver, 514 Thermal diffusivity, 45 Thermal expansion (see CTE) Thermal management, 11, 137, 141 Thermionic, 152, 155 Thermistors, 10 Thick film cutting tools, 430 Thin film cutting tools, 432 Thorium oxide, 152 3␻ method, 536, 537 Titanium as electrical contact, 589, 590 as filament material, 122, 128 Tools, cutting applications, 10, 137, 141, 441– 444 coated, thin film, 432 conventional, 426, 427 design of, 445–452 diamond adhesion to (see Adhesion) inserts, 475–478, 481–484 PCD, 427, 428 performance of, 441–445, 452–458, 475–498 thick film, 430–432 tungsten carbide, 426, 427 Tools, dressing, 501–503 Tools, rotary, 480, 481, 485–498

676 Transistors, diamond, 13 bipolar, 583 frequency, 582 history of, 583 IGFET, 589 MESFET, 582 MISFET, 589 MOSFET, 594–599 permeable base, 586, 587 point-contact, 584 power, 582 Translational energy, 327, 359 Transmission electron microscopy (see TEM) Transmission, optical diamond, 8, 9, 11, 520, 608 diamond coated substrates, 637–648 germanium windows, 355 of polycrystalline diamond, 616–621, 628 Treanor minimum, 381 Tungsten, as cathode material, 152 as electrical contact, 584 as filament, 58, 119, 122, 125, 527 as via fill, 556 melting point, 123 Tungsten carbide (see Tools) Turning (see Tools) Type I, 517 Type Ia diamond color of, 613 nitrogen in, 613 transmission spectra, 615 Type Ib diamond color of, 613 nitrogen in, 613 transmission spectra of, 614 Type II, 5, 7, 43 Type IIa diamond Cullinan diamond, 615 nitrogen in, 613 Type IIb diamond boron in, 613 Hope diamond, 615 transmission spectra of, 14

Index Ultraviolet (UV) (see also Absorption) diamond detector, 10, 12, 651 vacuum, 359 Union Carbide Corp., 19 Vacuum microelectronics, 12 Valence band, 610 Velocity, distribution, 374 Via, filling of, 555 Vibrational pooling, 389 Vibration electronic energy transfer, 377, 388 Vibration-rotation energy transfer, 404 Vibration-rotation-translation energy transfer, 340 Vibration-translation energy transfer, 340, 377, 382–384, 404 Vibration-vibration energy transfer, 377, 382–384, 404 Vicker’s indentation, 6, 648 Wavelength, 359, 608 Wavenumber, 359, 608 Wear resistance, of diamond, 439–441, 479, 504 Wear surfaces, 10 Windows, diamond endoscope, 627 gyrotron, 10, 11, 627 optical, 10, 11 Wire dies, 503–504 X-band, 582 XPS (see X-ray photoelectron spectroscopy) X-ray initiated reactions, 359 X-ray diffraction (XRD), 28, 36, 37, 40, 41, 78, 264, 457 X-ray masks, 11 X-ray photo electron spectroscopy, 88 X-RD (see X-ray diffraction) Young’s Modulus, 6, 9, 433, 434, 520, 530, 634 biaxial, 646 ZnS (see Substrates) ZnSe (see Substrates)