ADHESION ASPECTS OF

THIN FILMS

Volume 1

Editor: K.L. Mittal

VSP BV P.O. Box 346 3700 AH Zeist The Netherlands

Tel: +31 30 692 5790 Fax: +31 30 693 2081

vsppub@compuserve.com www.vsppub .com

0 VSP BV 2001

First published in 2001

ISBN 90-6764-338-6

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner.

Printed in The Netherlands by Ridderprint bv, Ridderkerk.

Adhesion Aspects of Thin Films, Vol. 1, pp. vii-viii Ed. K. L. Mittal 0 VSP 2001

Preface

This book chronicles the proceedings of the First International Symposium on Adhesion Aspects of Thin Films held under the auspices of MST Conferences in Newark, New Jersey, October 28-29, 1999. Apropos, the second symposium on this topic is planned to be held also in Newark, New Jersey, November 5 - 7 , 2001.

Films and coatings are used for a variety of purposes - decorative, protective, functional, etc. - in a host of applications. Irrespective of the intended function or application of a film or a coating, their adequate adhesion to the underlying substrates is of cardinal importance. Concomitantly, the need to understand the factors controlling adhesion and to tailor adhesion to a desired level is quite patent. Apropos, what is the difference between a film and a coating? The answer is: A film is a thin coating and a coating is a thick film. It is merely a matter of thickness.

Adhesion of films and coatings is dictated by a legion of factors, e.g., substrate condition or pretreatment, mode of deposition and deposition parameters and their thickness, and mechanical properties. Stresses in films and coatings are of paramount importance in dictating their adhesion behavior, and such stresses depend, inter alia, on both the deposition technique and parameters, as well as the nature and properties of films and coatings.

Even a cursory look at the literature will evince that there has been tremendous interest and brisk activity in understanding and tailoring adhesion of films and coatings, so we decided this symposium was both opportune and needed. This symposium was planned with the following objectives in mind: (i) to bring together the community interested in this topic, (ii) to provide a forum for discussion of latest developments, (iii) to provide an opportunity for cross-pollination of ideas; and (iv) to identify areas which offered good promise and warranted vigorous pursuit. The technical program for the symposium contained a total of 25 papers and the presenters hailed from academia, industry and other laboratories. Many different subtopics within the broad topic of adhesion aspects of thin films were discussed. There were lively and enlightening (not exothermic) discussions throughout the symposium, both formally in the auditorium as well as in corridors. Most of the presentations dealt with films, but a few papers dealing with coatings, i.e. relatively thick films, were also presented.

Now coming to this book which contains a total of 16 papers, as some papers are not included for a variety of reasons. It must be recorded here that all manuscripts were rigorously peer reviewed, suitably revised (some twice or thrice) and properly

viii Preface

edited before inclusion in this book. So this book is not merely a compilation of unreviewed and unedited papers, rather it represents the highest standard of a publication. The topics covered include: mechanisms, origin, evolution and measurement of stresses in thin films; surface stress effects on the intrinsic stress; various factors affecting stresses in thin films; delamination of coatings caused by residual stress; effects of surface treatments on the adhesion of metallic films; adhesion of CVD diamond to carbide cutting inserts; effect of carbon contaminant on adhesion of aluminum films; effect of interlayers on adhesion of ceramic coatings; effect of residual stress on adhesion and wear resistance of hard coatings; tribological properties of ceramic films; oxide layers as barrier coatings on a plastic substrate; adhesion aspects of organic coatings to metals; and adhesion of thin plasma polymerized fluorocarbon films.

I certainly hope this book providing a commentary on the current state of knowledge of adhesion of thin films will be useful to anyone interested in thin films and, hopefully, will provide ideas on how to improve or tailor adhesion of a film or a coating for a given situation.

Acknowledgements

First, my thanks are extended to Dr. Robert H. Lacombe for helping in many ways in organizing this symposium. The unsung heroes (reviewers) must be thanked for their time and efforts in providing valuable comments which are a desideratum to maintain the highest standard of a publication. My sincere thanks go to the authors who supplied the raw material (manuscripts) without which this book had not seen the light of day. Finally, I would like to express my appreciation to the staff of VSP for giving this book a body form.

K. L. Mittal

CONTENTS

Preface

Stresses in thin films J. BQttiger, J. Chevallier, P. KringhQj and K. 0. Schweitz

Fundamental aspects of residual stress evolution in thin metal films during energetic particle deposition A. Misra and M. Nastasi

Surface stress effects on the intrinsic stress in thin films R. C. Cammarata

vii

1

17

31

Influence of the ion bombardment on the stress in thin films produced by ion beam and RF sputtering techniques S. Scaglione, F: Sarto, A. Rizzo, M. Alvisi and M. A. Tagliente 35

Delamination of thin hard coatings induced by combined residual stress and topography U. Wiklund, S. Hogmark and J. Gunnars 51

Effects of surface treatments on the adhesion of metallic films to ceramic substrates A. J. Pedraza 67

The state-of-the-art in adhesion of CVD diamond to carbide cutting inserts M. A. Taher, W E Schmidt, A. I? Malshe, E. J. Oles and A. Inspektor 79

Adhesion improvement of diamond films to silicon nitride substrate for cutting tools H. Itoh, R. Sasai, M. Kamiya, S . 3 . Lee, K. Kuroda and 7: Tsutsumoto 141

Quantifying the effect of carbon on the practical adhesion of aluminum films to sapphire substrates J. A. Schneider, S. E. Guthrie, W M. Clip and N. R. Moody 159

The effect of a titanium-based interlayer on the adhesion of ceramic coatings M. T Vieira, S. Roque and A. S. Ramos 171

Effect of annealing on residual stress, strength, adhesion and wear resistance of thin, hard coatings on low alloy steel T Z. Kattamis and C. G. Fountzoulas 181

vi Contents

Study of adhesion and tribological properties of some ceramic films J. Takadoum and B. Cretin

Properties of oxide coatings deposited on a plastic substrate by a successive pulsed plasma anodisation process B. M. Henry, A. G. Erlat, C. R. M. Grovenor, G. A. D. Briggs and R. R Howson

Practical adhesion of organic coatings to metals: The role of the interphase and its residual stresses J. Bouchet, A. A. Roche, E. Jacquelin and G. W Scherer

Epoxy-diamine adhesives on metals: The interphase formation and characterization S. Bentadjine, A. A. Roche and J. Bouchet

Improving the adhesion of plasma polymerized thin fluorocarbon films on silicon using (CHF3 + SF6) radio-frequency discharge pretreatments k! Ianev and N. Schwesinger

195

207

217

239

26 1

Adhesion Aspects of Thin Films, Vol. 1, pp. 1-16 Ed. K. L. Mittal 0 VSP 2001

Stresses in thin films

J. B0TTIGER*, J. CHEVALLIER, P. KRINGH0J and K. 0. SCHWEITZ Institute of Physics and Astronomy, University of Aarhus, DK-8000 Aarhus C, Denmark

Abstract-The mechanical and other physical properties of thin metallic films are strongly influenced by stresses which may approach the plastic-deformation limit. Large stresses may cause adhesion failure, and, due to stress relaxation, the properties of the thin films may alter during their service life. Depending on the deposition process, thin metallic films can be deposited with either tensile or compressive stresses. Typically, evaporated thin films are in tension, mainly caused by grain boundaries of the thin films. Compressive stresses may arise when, during the deposition process, the growing film is bombarded with energetic particles in the energy range from tens to hundreds of electronvolts. This is due to “atomic peening”, Le. atoms close to the surface are recoil-implanted into spaces smaller than their atomic volume. In addition to the above-mentioned growth stress, stresses also originate from interfaces and the film surface, as well as due to differences in thermal expansion coefficients of the film and the substrate if the temperature is changed after deposition. By deposition of thin films using magnetrons, the magnitude and sign of the stress, tensile or compressive, can be altered through control of the deposition parameters. In multilayers with bilayer lengths in the nm range, the contribution to the stress from interfaces may dominate, and the film stress thus becomes highly dependent on the bilayer length. Various stress-relaxation mechanisms exist, whereby - at moderate temperatures (500-700 K) and times (tens of minutes) - significant changes in stress are observed.

Keywords: Thin films; intrinsic and thermal stresses; measurements of stress; the origin of stress; stress control: stress relaxation.

1. INTRODUCTION

Today, thin films are an indispensable part of modern technology. They play a key role in microelectronics, for example in Very-Large-Scale-Integration systems such as the Central Processing Units and dynamic memories in computers. Thin films are also key components in optical devices such as mirrors, including X-ray mirrors, filters, solar cells, and in various sensors. Also, the application of hard coatings for tribological protection of tools and critical machine parts in the machine industry is increasing.

‘To whom correspondence should be addressed. Tel.: 45 89 42 36 89; Fax: 45 86 12 07 40; E-mail: bottiger@ifa.au.dk

2 J. BQttiger et al.

During deposition with various techniques, stresses (force/cross-sectional area) develop in the thin films. Depending on the deposition parameters, large stresses which come close to or even exceed the limit for plastic deformations may develop. This may result in peeling off or cracking of the film. Stresses also affect the physical properties such as, for example, mechanical hardness and magnetic properties. Therefore, it is of paramount importance to control the level and sign, tensile or compressive, of the film stress during deposition and understand the relevant stress-relaxation processes to make certain that no stress changes occur during the service life of the film with the specific environment it experiences. To maximise the adhesion, the absolute level of film stress should be minimised. Sometimes, however, it can be necessary to build significant stresses into the films to obtain the desired physical properties, whereby the adhesion strength diminishes. Also, in tribological applications, if e.g. the coating will experience large tensile stresses, it can be beneficial to build-in compressive stresses in the coating.

In the sixties and seventies, the stress formation in thin films was studied extensively, and the reader is referred to reviews by Hoffman [l], Buckel [2] and Campbell [ 3 ] . A recent, extensive review has been written by Koch [4]. In a review by Nix [5] on mechanical properties of thin films, stress and stress relaxation in thin films are also discussed. The early experimental studies of stress formation were hampered by poor vacuum systems which made it difficult to reproduce results obtained in various laboratories. With the introduction of UHV systems, it became possible, for example, to separate the contribution of impurities and structural defects to stress.

In the present paper, stresses in thin films are reviewed. In Section 2, a general discussion of stresses in films on substrates is given, followed by descriptions of experimental methods to measure thin film stresses. The origin of intrinsic stress due to defects, surfaces and interfaces is outlined in Section 3 together with a description of thermal (extrinsic) stress, which arises during change of temperature from the deposition temperature to room temperature because of differences in thermal expansion coefficients of film and substrate. In Section 4, the control of stress during deposition by evaporation and magnetron sputtering is discussed. An example of a measurement of interface stress is given in the next section, and finally stress relaxation is discussed.

2. STRESSES IN THIN FILMS ON SUBSTRATES

Consider a thin film on a substrate, see Fig. la. The film is free of stress, with a thickness much smaller than that of the substrate (the thin film approximation) and lateral dimensions much larger than the total thickness of film and substrate. Imagine the film being removed from the substrate without dimensional changes and formation of stress. Thereafter, in the self-supporting state, the film experiences a volume change through the introduction of defects, a phase transformation or a change in temperature, see Fig. lb. For the present discussion, assume that the volume change is uniform corresponding to a dilatational strain e D . For a pure

Stresses in thinfilms 3

Figure 1. Illustration of introduction of compressive stress in a thin film by a volume change and the associated substrate bending, based on Ref. [ 5 ] .

dilatation strain, the principal strain components are then E: = E: = ET = e D / 3 (the z-axis being perpendicular to the film surface).

Consider now the film being reattached to the substrate. As the lateral dimensions of the film and substrate are no longer the same, a biaxial stress (elastic isotropy is assumed in the plane of the film) is imposed on the film to elastically deform it back to the same dimensions as those of the substrate, see Fig. IC. The strain developed is E = cx = E ) = -eo/3. Using Hooke’s law, we obtain a biaxial stress in the reattached film:

IY = ox = U ) = ME. (1)

M is the biaxial modulus given by M = E/(1 - u ) . E is the Young modulus and u is the Poisson ratio. Notice that a volume expansion of the self-supporting thin film (as shown in Fig. 1 where e D is positive) results in compressive stresses, Le. negative IY values, while a volume shrinkage results in tensile stresses (positive IY values). As long as the forces applied to deform the self-supporting film are still present, the stress in the reattached thin film does not change. When they are taken away, the normal tractions from the edges are removed and shear forces on the film/substrate interface near the edges are produced. These shear stresses supply the forces required to maintain the biaxial stress in the thin film and result in bending of the filmhubstrate composite. Tensile film stresses bend the substrate concavely upwards, compressive film stresses bend the substrate convexly outwards (Fig. 1).

In this case, the film/ substrate composite may be rectangularly shaped, and the substrate should be smooth (polished) and planar with a well-known in-plane isotropic Young’s modulus. Silicon substrates are well-suited for this type of stress measurements. The radii of curvature of the substrate before and after deposition of a thin film, q,d and ?+ad, respectively, are measured ex situ, for example by use of a

The film stress can be obtained from this bending [6].

4 J. Bgrtiger et al.

profile-measurement device, and the stress, CT, is calculated by use of the Stoney equation [7 ] :

E, is the in-plane Young’s modulus of the substrate, us its Poisson ratio, d, and df the thicknesses of the substrate and film, respectively. Note that only the elastic properties of the substrate enter into the Stoney formula. This is related to the thin film approximation, i.e. the film thickness is much smaller than the thickness of the substrate.

When multiple thin films are deposited sequentially onto a very much thicker substrate, the resulting stress is the weighted average of the stresses of the individual layers.

It is sometimes assumed that the deposition of a highly stressed cap layer on top of another film, which is attached to a thick substrate, causes large stresses to arise in this film. This is wrong. Although the stress in the cap layer is compensated by stresses in the underlying layers, these stresses are usually extremely small due to the much thicker substrate.

Frequently, measurements of residual stress are carried out using the X-ray diffraction sin2 @ method [8]. Assuming a biaxial stress state in a polycrystalline sample, lattice constants d, are obtained from measurements of planar distances of a specific family of planes as a function of @ which is the angle between the planes and the film surface. The dependence of d, on $ is given by

d,j = - + ’ ado sin2 @ + do E (3)

where do is the stress-free lattice parameter, CT is the biaxial stress, and E and LJ

are the Young modulus and the Poisson ratio of the film, respectively. Based on equation (3), it is seen that the biaxial stress can be obtained from the slope in a d, versus sin2 @ plot.

3. THE ORIGIN OF STRESS

In general, growth processes proceed far from thermodynamic equilibrium and are mainly kinetically controlled. This results in many different types of defects which create stresses. Below, the most important contributions to stress formation are discussed. More details can be found in [4].

Some polycrystalline films contain a large amount of low-angle grain boundaries separating the grains. Numerous experimental investigations have shown that on average these grain boundaries produce a tensile stress in the films. This is explained by considering the atomic forces acting between the boundary atoms of neighbouring grains. On average, the distances between these atoms are larger than the equilibrium atomic distances, and thus the forces are attractive, i.e. neighbouring grains are strained in tension.

Stresses in thinjlms 5

Recry stallisation may occur when, during or after deposition, the self-diffusion is sufficiently high. During recry stallisation, defects are annealed out and the average grain size of polycrystalline films increases. This leads to a densification of the film and therefore a tensile contribution to the stress is expected [9]. Also, during solid state reactions and/ or interdiffusion, volume changes may occur which introduce stress.

It is well-established that small isolated particles (grains), due to their surface tension, are in a state of compressive stress [lo], related to an increased hydrostatic pressure, P , inside the particle which is given by

2Y p = --, R (4)

where y and R are the surface tension and the particle radius, respectively. From this equation it is seen that the compressive stress decreases with increasing particle radius. Below, the compressive stress in small grains due to the surface tension will be denoted the capillarity effect.

Various impurities built into the films during growth influence the stress. As an example, oxygen may be incorporated into Cr and Fe films during deposition [ l l , 121, probably at interstid sites or in grain boundaries, since large compressive stresses were formed.

When the bulk lattice constants of the film and substrate in epitaxial growth are different, misfit stresses arise because the film during the initial growth has the same lateral lattice spacings as the substrate. At a critical film thickness, the stress is released by formation of misfit dislocations. For further information on misfit stress, the reader is referred to the literature, see e.g. Ref. [13].

The interface (surface) stress, g, represents the work required to elastically deform a unit area of interface (surface) by a unit strain. Interface (surface) tension, y , is the free energy connected to create a unit area of new interface (surface). The connection between g and y [ 141, assuming in-plane elastic isotropy, is

dY g = y + - .

dE ( 5 )

The derivative in the equation accounts for the change in y upon straining the surface and may assume positive or negative values. An introduction to interface (surface) stress and tension has been written by Cammarata [14].

The surface stress has been calculated by first-principle calculations, see, e.g., Ref. [15]. It is more the rule than the exception that single-crystal surfaces are subject to surface stress which sometimes results in surface reconstructions. So far, very few absolute measurements of surface stress are available.

As described below, interface stresses have been measured in a few systems, see, e.g., Refs [ 16- 191. First-principle calculations of interface stress have also been carried out but theoretical and experimental interface stress values do not agree. A compilation of experimental and theoretical papers on interface stress is given in Ref. [19].

6 J. BQttiger et al.

In contrast to the above-discussed thin film stresses, which are named intrinsic (or growth stresses when related to defects), thermal stresses are denoted extrinsic. They arise due to the difference in thermal expansion coefficients of film (af) and substrate (a,). When changing the temperature from the deposition temperature (Tdep) to the ambient temperature (Tamb), thermal stress D is produced, given by

(af - as)(Tdep - Tamb). (6 ) Ef

1 - Uf r J = -

Ef and uf are the Young modulus and the Poisson ratio of the film, respectively.

4. CONTROL OF STRESS DURING DEPOSITION

In some applications of thin films, a specific stress is required to obtain the desired physical properties or to increase the ability of the film to withstand specific external forces (tribology). In the present section, it is described how the film stress can be tailored to specific applications through control of the deposition parameters. The section is divided into two parts: the first describing stress control during deposition with techniques where the atomic species arriving at the growing film have thermal energies; and the second part dealing with depositions where atomic species with energies up to several hundreds of electronvolt bombard the growing film. The topic of the present section is described by way of examples, and a review is not attempted.

4.1. Control of stress during thermal evaporation

Thurner and Abermann [20] have investigated the intrinsic stress in Cr films. The Cr was deposited by thermal evaporation on a bending beam. By measuring the bending during deposition and using the Stoney equation, the force per unit width due to the stress was obtained as a function of film thickness. By dividing this force with the film thickness, the average film stress was obtained. The results are shown in Fig. 2. For three different deposition temperatures, 300 K, 473 K and 573 K, the force per unit width is shown as a function of film thickness. The room temperature film had columnar structure, and, as expected with such a structure, the corresponding force- thickness curve shows a nearly constant tensile stress in the growing film (a constant slope). Going to higher substrate temperatures, the columnar structure became less pronounced due to more equiaxed grains, resulting in fewer low-angle grain boundaries. This is reflected in the curves in the figure. For the highest deposition temperature, the stress level is dramatically reduced and the stress is changed from being tensile to being compressive. According to Abermann et al. [21], considering the 573 K curve, the initial tensile forces in the discontinuous film are due to formation of grain boundaries and recrystallisation, while, when the film becomes continuous, capillarity effects are dominating.

In summary, in the above example of stress in films deposited by thermal evaporation, the substrate temperature, which determines the thermal energy and

Stresses in thin films 7

I I I 1 C.102-384

150

THICKNESS (nm)

Figure 2. Film force per unit width in Cr films deposited at various temperatures versus thickness. Solid line: 300 K; dashed line: 473 K; dotted line: 573 K. The data are taken from Ref. [20].

thus the mobility of the diffusing atoms, is the key deposition parameter. At low temperatures, high tensile stresses are formed, while at higher temperatures a low- level compressive stress is observed.

4.2. Control of stress during deposition with simultaneous energetic bombardment of the growing Jilm

When bombarding a growing film with energetic atomic species with energies ranging from a few to hundreds of electronvolt, compressive stresses may arise through a process of atomic peening [22, 231. The energetic atomic species recoil- implant atoms, which are close to the surface, further in the growing film into spaces smaller than the usual atomic volume. Hereby the film expands outwards from the surface and, in the plane of the film, compressive stresses are formed.

Davis [24] has developed a simple theoretical model to account for the compres- sive stress which builds up during film growth due to the simultaneous bombard- ment with energetic atomic species. In the model, the compressive stress is due to atomic peening, and a component of stress relaxation comes from collision cascades formed at higher incident-particle energies. From the model, the compressive stress u can be calculated:

E‘/2 (7)

Y ~ = k -

1 - u R / J + 0.016/3(E/Eo)5/3’

where k is a constant, Y is in-plane Young’s modulus, u is Poisson’s ratio, R is the rate of film atoms deposited per unit area, J is the flux of energetic incident ions,

8 J. BIttiger et al.

p is a materials-dependent parameter of the order of unity, E is the energy of the bombarding ions and Eo is an excitation energy of the order of 10 eV.

Davis' model does not take the fact into account that, without irradiation, tensile stresses arise, and it also disregards any possible relaxation which is not connected with the collision cascades. To correct for this, Schweitz et al. [25] have added a constant to equation (7), whereby the following expression is obtained:

Here, b, k l and d (originating from the constant added to equation (7)) are constants to be determined from experiments.

Schweitz et al. [25] have also experimentally studied the stress in Ni films deposited by e-gun evaporations with simultaneous irradiation with monoenergetic Ar+ ions in the energy range 60 to 800 eV (Ion Beam Assisted Deposition). Evaporation rates in the range from 0.5 to 2.0 A/ s were used, and the R / J ratio was varied from 0.4 to 6.4. The stress was obtained from the bending of the substrate. The 168 data points - the measured compressive stresses as a function of E and R / J - were fitted to equation (8). In Fig. 3, the stress as a function of Ar+ energy is shown for the experimental data with R / J = 1.20 f 0.12 (closed circles) and for R / J = 5.8 f 0.58 (open circles) together with the curves calculated from equation (8) with the constants obtained by fitting. Note that the compressive stresses are plotted as positive values in this figure (since this was the way it was plotted in the original figure). The solid lines are for the R / J values of 1.08 and 1.32, corresponding to the closed symbols, and the dashed lines for R / J values of 5.22 and 6.38, corresponding to the open symbols. Except for the highest energy of 800 eV, good agreement between the model predictions and the experimental data is observed. The deviations at 800 eV are probably due to formation of strong in-plane texture at this energy [25].

It is seen from Fig. 3 that the stress is highly dependent on the energy of the bombarding ions and less sensitive to the R / J ratio. The compressive stress increases with increasing ion energy due to the atomic peening, reaching a maximum and then even decreases as the collision cascades start to relax the stress at higher energies.

These results clearly show that when depositing films with techniques involving energetic bombardment during growth, the stress can be controlled, mainly, by adjusting the bombardment energy through suitable control of the deposition parameters. As is the case with vapour deposition, the substrate temperature also plays an important role. Due to defect annealing, higher substrate temperatures lower the level of intrinsic stress.

To illustrate how to control the stress when depositing films by magnetron sputtering, experimental studies [26] of the stress in amorphous Ta50Cr50, deposited by magnetron sputtering, are presented.

The amorphous films were produced at room temperature by magnetron co- sputtering with an Ar pressure varied between 0.4 and 2.0 Pa, the chamber base

Stresses in thin films 9

Ar ION ENERGY (eV)

Figure 3. The stress in Ni films as a function of bombardment energy for the experimental data with R/ J = 1.20 i 0.12 (closed circles) and for R / J = 5.80 f 0.58 (open circles). Solid lines: model calculations (equation (8) and the constants as obtained by fitting) for R / J values 1.08 and 1.32. Dashed lines: model calculations for R/ J values of 5.22 and 6.38. Note that the compressive stress is plotted as positive values. The data taken from Ref. [25].

pressure being 2 x low5 Pa. Two magnetrons were placed approximately 13 cm from the substrates. The deposition rates varied between 2 and 4 A/s , with final thicknesses of the order of 1 pm. The film stress was obtained from measurements of the substrate curvature.

In Fig. 4, the results of measurements of the stress in amorphous Ta50Cr50 film as a function of the Ar pressure are shown. The Ta magnetron consumed 800 W d.c. power and the Cr magnetron 600 W r.f. power. No bias was applied to the substrate. When the films corresponding to the closed-circle data points were deposited, the substrates were rotated, in contrast to the open-circle data points where the substrates were kept fixed. Traces of crystalline material were detected by X-rays in the films which had been rotated. However, the two sets of data points, open and closed circles, showed the same trends. As also observed with magnetron sputtered crystalline materials [27], varying the pressure causes the intrinsic stress to change from a highly compressive state, -2 GPa, to a highly tensile one, +1 GPa.

The energetic particles originating from the magnetrons, i.e. Ar+ ions being neu- tralised and reflected at the magnetron surface and sputtered particles, experience only a few collisions on their way to the substrate at low pressures. Therefore, the bombardment at low pressures is more energetic, resulting, due to the peening ef- fect, in large compressive stress values. In contrast, at higher pressures the peening effect is less important and the stress turns tensile.

Figure 5 shows the results of measurements of the stress as a function of r.f. power (bias voltage) applied to the substrate, which increases the energy of the bombarding

10

o 0-

n: I- o

2

-1 -

I i i i

€

D

L

0 -2 I I I I

0 0.5 1 .o 1.5 2.0 2.5

Ar PRESSURE (Pa)

Figure 4. Stress in amorphous Ta5oCr5o-films produced by magnetron sputtering as a function of the Ar pressure, no bias applied. The data taken from Ref. [26].

1

O t

I- E o -ll -1

"0 50 100 150 200 250 300 350

RFPOWER (W)

Figure 5. Stress in amorphous TasoCr50 films produced by magnetron sputtering as a function of r.f. power (bias voltage). The Ar pressure was 1.5 Pa. The data taken from Ref. [26].

particles. The Ar pressure was 1.5 Pa and the magnetrons were powered as before. The substrate power range from 50 to 300 W corresponds to average d.c. voltages ranging from -240 to -640 V. Without a bias voltage, the stress was tensile, whereas a strong peening effect (large compressive stresses) was observed with an applied bias voltage. A maximum in compressive stress is observed in the figure at a

Stresses in thinjilrns 11

power value of 150 W, and at higher power values, the relaxation caused by higher- energy collision cascades starts to play a significant role. This can be explained by the Davis model [24].

5. MEASUREMENTS OF INTERFACE STRESS

In some cases, a significant part of the total film stress may originate from the surface and/or interfaces. However, as mentioned above, very few data are available on surface and interface stresses. To illustrate the importance of interface stress when considering the total film stress, measurements of the interface stress in Ag/Ni multilayers are presented.

Schweitz et al. [28] have studied Ag/Ni multilayers which were produced by magnetron sputtering. The structural characterisation was carried out by X-ray diffraction (XRD) and transmission electron microscopy (TEM), and the interface stress was determined by the method developed by Ruud et al. [16].

The mechanical contributions to the energy of multilayer/ substrate composites consist of the strain energy of the individual elemental layers and the substrate (arising from defects and bending) and a significant contribution from the interfaces for small bilayer lengths. Ruud et al. [16] assumed that the energy of the film- substrate interface and the free film surface was negligible compared to the total energy of the interfaces and that the interface stress was isotropic and independent of strain. They minimised the mechanical energy with respect to the radius of curvature of the film/substrate composite which resulted in the equation

2 h

a - (a) = - f (9)

where a is the total film stress, (a) is the average stress in the individual layers, h is the bilayer length and f is the interface stress. It is seen that a plot of a (obtained by measuring the radius of curvature and using the Stoney equation) minus (a) (obtained by measuring in-plane lattice spacings by Grazing Incidence XRD) versus the reciprocal bilayer length, l / h , should yield a straight line with a slope of twice the interface stress.

In Fig. 6, data from the studies by Schweitz et al. [28] are presented. The total stress (closed circles) and the average stress in the individual layers (open circles) are shown as a function of the reciprocal bilayer length. Both the total stress and the average stress in the individual layers increase as the reciprocal bilayer length increases. Note that the total film stress is small although a significant growth stress, (a), is observed. This shows that, in the present case, the interface and growth stresses are of the same magnitude but of different signs.

In Fig. 7, the difference a - (a) is shown as a function of the reciprocal bilayer length, l / h . The closed circles are the measured values and the solid line is a linear fit to the data of the multilayers with bilayer lengths between 5 nm and 15 nm. The open circles represent the multilayers with the smallest modulation length of 1.7 nm

12 J. Bmttiger et al.

2 I I I I I I C.102.381

F

Q P

P 0

i -1 I I I I I I I I 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

X’ (nm-’)

Figure 6. The total stress (closed circles) and the average stress in individual layers in Ag/Ni multilayers versus the reciprocal bilayer length.

Figure 7. The difference between the measured total stress and the average stress in the individual layers in Ag/Ni multilayers versus the reciprocal bilayer repeat length. The closed circles are the measured values and the solid line is a linear fit to data from multilayers with bilayer lengths between 5 nm and 15 nm. The open circles represent the multilayer with the smallest modulation length of 1.7 nm and the pure Ag and Ni films.

and the pure Ag and Ni films. The slope of Fig. 7 gives the compressive interface stress f = -2.24 =t 0.21 J/m2.

6. STRESS RELAXATION

As mentioned above, stress relaxation is detrimental to the performance of many thin-film devices. Therefore, for specific thin films, the stress-relaxation mecha- nism(s) has to be identified and quantified by mathematical equations. This makes it possible to predict the maximum changes in stress, which the individual film expe- riences during its service life when exposed to, for example, elevated temperatures.

Stresses in thinJilms

1 1 1 1 1 c-102.381 -600

13

-1000 I I I I I I I 300 350 400 450 500 550 600

TEMPERATURE [K]

Figure 8. Stress in an amorphous Ta54C1-46 film with initial stress of -770 MPa as a function of temperature. The sample is ramped up in temperature at the rate of 5 Wmin. At 400 K, 440 K, 480 K, 520 K, and 560 K, the ramping is stopped for half an hour so that the stress relaxation can be measured at constant temperature. After relaxation at 560 K, at the rate of 5 Wmin the temperature is decreased to 40 K, kept there for 1 min, raised to 560 K again, and then brought back to room temperature.

As an example, this section presents studies of stress relaxation in amorphous Ta54C1-46 film by inhomogeneous flow, in which the strain is localised in shear bands. In addition, stress-relaxation mechanisms in crystalline films are briefly discussed.

In the stress-relaxation studies by Andersen et al. [26], three different amorphous Ta54Cr46 films with as-deposited compressive stresses of -770 MPa, -655 MPa and -260 MPa were prepared by magnetron sputtering. Their stress-relaxation behaviour was measured by use of a cantilever-beam technique, the stress being obtained by the substrate bending. In Fig. 8, as an example of the measurements carried out during stress relaxation, the stress in a film with the as-deposited stress -770 MPa is shown as a function of temperature. The sample is ramped up in temperature at the rate of 5 K/min. At 400 K, 440 K, 480 K, 520 K and 560 K, the ramping is stopped for half an hour, so the stress relaxation can be measured versus time at constant temperature. The slope of the stress curve mainly arises due to the difference in thermal expansion coefficients between the film and the substrate, introducing further compressive stress in the film during heating. From the figure it appears that the significant stress relaxation occurs at the last three hold temperatures, represented by the vertical lines. The data corresponding to the vertical line at 520 K are shown in Fig. 9 as the relative change in stress, Aa/oo, versus time. (The data cannot be discerned from the solid line which is a fit to the data, see below.) Here, 00 is the start stress at the particular relaxation temperature. Curves similar to the one of Fig. 9 have been obtained for all three samples at the relaxation temperatures 480, 520 and 560 K.

14

TIME (sec.)

Figure 9. The relative change in stress, Aa/oo, versus the time at 520 K. The data, which cannot be discerned from the solid line, correspond to the ones of the vertical line in Fig. 8 at 520 K. The solid curve is the fit obtained by use of equation (10).

The curves have been fitted with the expression:

A 0 - = alog(1 + b(t - to)), 00

where l / b is the relaxation time and to determines the start of the relaxation process. The equation was developed to describe the creep due to dislocation movement in work-hardened crystalline materials [29], and it was, therefore, not clear beforehand that it could be valid for amorphous materials. However, an equivalent expression was derived for amorphous films, considering inhomogeneous flow [30] in which the strain is localised in shear bands [31, 321.

In Fig. 9, the solid line is the fitted curve (due to the excellent fit, it is difficult to discern experimental data from the fitted curve). For all the experimental data, a good agreement between fitted curves and experimental points was found. Thus, for the various relaxation temperatures and initial stresses, the values of a , b and to are obtained. Extrapolations and the use of equation (10) enable the stress relaxation in the amorphous Ta54Cr46 films for a chosen temperature and initial stress to be predicted. The thermodynamic parameters of activation free energy A F and activation volume 52 of the model of inhomogeneous flow can be calculated from the values of a and b and the following expressions [30]:

b = --ioexp($ E52 + e). kT kT

Stresses in thinj lms 15

E is Young’s modulus and EO is a constant. The relaxation of compressive stresses in arc-evaporated TiC,N1 -, films during

thermal annealing has been investigated as a function of composition and initial stress level [33]. The relaxation was attributed to defect annealing, including recry stallisat ion.

Various stress-relaxation mechanisms in metallic-alloy thin films have also been identified. For example, Gardner et al. [34] observed stress relaxation in an aluminium alloy due to grain growth. Dislocation glide mechanisms have been considered, among others by Flinn and coworkers, who managed to describe the shape of stress-temperature curves for aluminium [35] and copper [36] using a model based on thermally activated dislocation glide. Besides dislocation-glide mechanisms, Thouless et al. [37] and Vinci et al. [38] applied stress-relaxation models which also included grain-boundary diffusion.

7. SUMMARY

In contrast to free-standing films, films attached to a substrate may be stressed in tension or compression. The film stress can be obtained from the bending of the film/substrate composite by measuring the radius of curvature and using the Stoney equation. Frequently, measurements of stress are also carried out using the X-ray diffraction sin2 @ method. In general, growth processes proceed far from thermodynamic equilibrium and are mainly kinetically controlled. This results in many different types of defects which create stresses. The most important of these defects have been described. Extrinsic stress of thermal origin arises when the deposition temperature is different from room temperature and the thermal expansion coefficients of the substrate and the film are different. Through control of the deposition parameters - substrate temperature and the energy and flux of energetic particles bombarding the growing film - the stress can be tailored to applications which require specific tensile or compressive stresses. Interface stresses play a role in thin films and multilayers with small bilayer lengths. An example of the measurements of interface stress has been presented. Finally, the importance of controlling stress relaxation has been emphasised and an example of stress relaxation in amorphous Ta54Cr46 film has been given.

REFERENCES

1. R. W. Hoffman, in: Physics of Thin Films, G. Hass and R. E. Thun (Eds), Vol. 3, p. 211.

2. W. Buckel, J. Vac. Sci. Technol. 6, 606 (1969). 3. D. S. Campell, in: Handbook of Thin Film Technology, L. I. Maissel and R. Glang (Eds). p. 123.

4. R. Koch, J. Phys.: Condens. Matter 6,9519 (1994). 5 . W. D. Nix, Metall. Trans. 20A, 2217 (1989). 6. M. Ohring, The Materials Science of Thin Films, Chapter 9. Academic Press, San Diego (1992).

Academic Press, New York (1966).

McGraw-Hill, New York (1970).

16 J. B@ttiger et al.

7. G. G. Stoney, Proc. R. SOC. A32, 172 (1909). 8. I. C. Noyan and J. B. Cohen, Residual Stress. Springer-Verlag, New York (1987). 9. R. Koch and R. Abermann, Thin Solid Films 140,217 (1986).

10. J. C. Heyraud and J. J. Metois, Sulf: Sci. 100, 59 (1980). 11. G. Thurner and R. Abermann, Vacuum 41, 1300 (1990). 12. D. Winau, R. Koch and K. H. Rieder, Appl. Phys. Lett. 59, 1072 (1991). 13. J. H. Van der Merwe and W. A. Jesser, J. Appl. Phys. 64,4968 (1988). 14. R. C. Cammarata, Prog. Surf: Sci. 46, 1 (1994). 15. R. J. Needs, Phys. Rev. Lett. 58, 53 (1987). 16. J. A. Ruud, A. Witvrouw and F. Spaepen, J. Appl. Phys. 74, 2517 (1993). 17. G. Gladyszewski, S. Labat, P. Gergaud and 0. Thomas, Thin Solid Films 319, 78 (1998). 18. S. Labat, 0. Thomas, P. Gergaud, A. Charai, C. Alfonso, L. Barrallier, B. Gilles and A. Marty,

19. K. 0. Schweitz, Ph.D. thesis, University of Aarhus (1999). 20. G. Thurner and R. Abermann, Thin Solid Films 192, 277 (1990). 21. R. Abermann, R. Koch and R. Kramer, Thin Solid Films 58, 365 (1979). 22. E M. d’Heurle and J. M. Harper, Thin Solid Films 171, 81 (1989). 23. H. Windischmann, J. Vac. Sci. Technol. A9, 2431 (1991). 24. C. A. Davis, Thin Solid Films 226, 30 (1993). 25. K. 0. Schweitz, J. Amdt, J. Battiger and J. Chevallier, Nucl. Instrum. Meth. Phys. Res. B127/128,

26. P. Andersen, M. Moske, K. Dyrbye and J. Battiger, Thin Solid Films 340, 205 (1999). 27. C. Hudson and R. E. Somekh, Mater. Res. SOC. Symp. Proc. 239, 145 (1992). 28. K. 0. Schweitz, H. Geisler, J. Chevallier, J. Bgttiger and R. Feidenhans’l, Mater. Res. SOC. Symp.

29. J. Weertman and J. R. Weertman, in: Physical Metallurgy, 3rd edn, R. W. Cahn and P. Haasen

30. R. J. DiMelfi, Scr Metall. 21,421 (1987). 31. E Spaepen, Acta Metall. 25,407 (1977). 32. A. I. Taub, Acta Metall. 28, 633 (1980). 33. L. Karlson, L. Hultman, M. P. Johansson, A. Horling and G. Ramanath (to be published). 34. D. S. Gardner, T. L. Michalka, P. A. Flynn. T. W. Barbee, Jr., K. C. Saraswat and J. D. Meindl,

35. P. A. Flinn, D. S. Gardner and W. D. Nix, IEEE Trans. Electron Devices 34, 689 (1987). 36. P. A. Flinn, J. Mater Res. 6 , 1498 (1991). 37. M. D. Thouless, J. Gupta and J. M. E. Harper, J. Matel: Res. 8 (8), 1845 (1993). 38. R. P. Vinci, E. M. Zielinski and J. C. Bravman, Thin Solid Films 262, 142 (1995).

J. P h y ~ . IV6, C7-135 (1996).

809 (1997).

Proc. 505, 559 (1998).

(Eds), p. 1309. North-Holland, Amsterdam (1983).

in: Proc. 2nd Int. IEEE VLSIMultilevel Interconnection Confi, p. 102 (1985).

Adhesion Aspects of Thin Films, Vol. I , pp. 17-29 Ed. K. L. Mittal 0 VSP 2001

Fundamental aspects of residual stress evolution in thin metal films during energetic particle deposition

A. MISRA * and M. NASTASI Los Alamos National Laboratoiy, Materials Science and Technology Division, Los Alamos, NM 87545

Abstract-We have studied the stress evolution in thin metal films on Si substrates as a function of Ar pressure during sputtering, substrate bias and post-deposition ion irradiation. With increasing bombardment, the tensile stress is observed to increase to a maximum, and then transitions to compressive stress that also reaches a maximum. We show that the maximum tensile strain may be estimated from the grain size and the interatomic potential. We further show that the compressive stress maximum is related to the saturation in point defect concentration, with smaller contribution from entrapped Ar. We use dislocation-based models to estimate the film yield strength and show that the maxima in both tensile and compressive residual stresses are set by the film yield strength. Compressive yield strength is higher as compared to tensile strength due to hardening from point defects.

Keywords: Residual stress; sputtering; Cr films; point defects; film yield strength.

1. INTRODUCTION

Physical vapor deposited (PVD) coatings are used in a wide range of applications such as microelectronic, integrated optoelectronic and microelectromechanical systems (MEMS) devices, and tribological systems. It is well known that residual stresses in PVD thin films may have deleterious effects such as film cracking (tensile stress) or delamination from the substrate (compressive stress) [l-31. On a less catastrophic scale, the residual stresses may adversely affect the electrical, magnetic or optical properties of the films. With continuing trends in reduction of the microelectronic device dimensions, the need to enhance the reliability and performance of thin metal films such as in interconnect lines is even higher. Crucial to this effect is a fundamental understanding of the stress evolution in thin films and methods to tailor residual stress are very significant for the reliability of devices

*To whom all correspondence should be addressed; Phone: (505) 667-9860; Fax: (505) 665-2992; E-mail: amisra@lanl.gov

18 A . Misra and M. Nastasi

that involve PVD films. The purpose of this article is to present insights in the physical mechanisms of tensile and compressive stress evolution in sputtered films. We focus only on the intrinsic (Le., growth related) stresses as observed in ambient temperature deposited films, and ignore the extrinsic stresses that result from a mismatch in co-efficients of thermal expansion of the film and the substrate when the film is cooled to room temperature after elevated temperature deposition. Furthermore, we only consider the nanocrystalline, continuous films deposited at low temperatures (e.g., homologous temperature of - 0. 14Tm for Cr deposited at room temperature, where T , is the melting point of Cr) where the effects of thermally activated diffusive processes may be assumed to be insignificant. In our experiments described in this article, the film thickness is kept constant and stress evolution is studied as a function increasing energy and/or flux of particle bombardment either during or after deposition. Specifically, three experiments will be described:

(i) Effect of Arpressure. The two primary sources of energetic species irradiating the growing film are (a) sputtered metal particles which are ejected with average energies of a few to a few tens of eV and (b) reflected Ar neutrals which are back scattered from the target with typical energies in the tens to a few hundred eV range that scale with the target to the Ar ion mass ratio. These particles ejected from the target lose energy by collisions with Ar in the plasma and hence, the higher the Ar pressure in the chamber, the lower the final energy of the ejected particles on reaching the substrate.

(ii) Effect of negative substrate bias voltage. In addition to the two sources of energetic particles mentioned above, applying a small negative potential to the anode (normally at ground potential) results in the acceleration of a fraction of Ar ions to the anode (substrate).

(iii) Post-deposition ion-irradition. Sputter deposition at very high Ar pressures and no substrate bias is similar to ambient temperature evaporation (non- energetic deposition). These films were subjected to irradiation with high- energy Ar ions in the 110-300 keV range after deposition. The stress evolution as a function of ion dose was then followed in situ during irradiation.

The common trends from these experiments are interpreted using models based on defect evolution in the bombarded films.

2. EXPERIMENTAL PROCEDURES

Cr films were deposited by dc magnetron sputtering on {loo} Si wafers using 300 W power to 10 cm diameter targets. The target-to-substrate distance was -10 cm. In the first kind of experiment, Ar pressure was varied from 1 to 7.5 mTorr and no substrate bias was used. In the second set of experiments, negative substrate bias up to 500 V was used at a fixed Ar gas pressure of 5 mTorr. In the third experiment, films sputtered at 5 mTorr pressure were irradiated after deposition with Ar ions in the energy of 110-300 keV to a total dose of 5 x loL5 ions/cm2. All these

Fundamental aspects of residual stress evolution in thin metal films 19

experiments were performed on films with 150 nm nominal thickness. The second type of experiment was also performed on 1 p m thick films.

The residual stresses for the first and second experiments were calculated using the Stoney equation [4], with the substrate curvature measured before and after deposition by a laser deflection apparatus [5] . The residual stresses in the third experiment were estimated by monitoring the deflection at the end of a cantilever beam sample during ion irradiation, following an approach similar to that used by Van Sambeek and Averback [6]. Transmission electron microscopy (TEM) was performed on a Philips CM30 microscope operating at 300 kV.

3. RESULTS

3.1. Stress evolution with Arpressure

The evolution of tensile stresses in sputtered 150 nm nominal thickness Cr films as a function of Ar gas pressure in the 1-7.5 mTorr range is shown in Fig. 1. With decreasing Ar pressure, the biaxial tensile residual stress increases to a maximum of - 1.7 GPa and then rapidly decreases. At the lowest Ar pressure of - 1 mTorr used, the stress was still tensile.

3.2. Stress evolution with negative substrate bins

With increasing substrate bias in the 150 nm thick film (Fig. 2), we observe an increase in tensile stress to a maximum of - 1.6 GPa, followed by a complete relaxation of tensile stress, and finally the rapid build up of compressive stress to a saturation value of -2.1 GPa. The stress evolution with substrate bias for 1 p m thick Cr films is shown in Fig. 3. The Ar pressure was kept constant at 2.5 mTorr. Note from Fig. 1 that 2.5 mTorr is to the left of the peak in tensile stress and hence, no initial increase in the tensile stress is observed with increasing bias at 2.5 mTorr deposition (Fig, 3 ) . However, with increasing bias at 5 mTorr deposition (which is

2 I I I I I I I I I , I I I . I I I I I , I I I I I , , I - 1 - 4 , -

I 0 t . . , , I . . . ' ' ' ' ' ' I . ' " I . . . . I . , , . I . . , . 1

1 2 3 4 5 6 7 8 Ar pressure (mTorr)

Figure 1. Dependence of tensile residual stress in 150 nm thick Cr films on the Ar pressure during sputtering.

20 A. Misra and M . Nastasi

1 - A -

6 v1 v1

g - 1 : c n - - 2 1 I i

- 3 - - - 5 4 1 n 8 3 " a a m ' l n n c ' l ' n a L a ' a c c

-600 -500 -400 -300 -200 -100 0 Substrate Bias (V)

Figure 2. Evolution of residual stress in 150 nm thick Cr films with negative substrate bias.

- 2 . 5 ~ " ' " " " " " " " ' " " " ~ -500 -400 -300 -200 -100 0

Substrate Bias (V)

Figure 3. Evolution of residual stress in 1 p m thick Cr films with negative substrate bias. Note the relaxation of compressive stress at bias higher (Le., more negative) than 300 V.

to the right of the peak in Fig. l), initial increase in tensile stress to a maximum is observed (Fig. 2).

3.3. Stress evolution with post-deposition ion irradiation

The evolution of stress in 150 nm thick Cr films, deposited at 5 mTorr Ar pressure without any substrate bias, with post-deposition ion irradiation is shown in Fig. 4. The irradiation was done with 110 keV Ar ions that penetrate about 80 nm of film thickness. With constant ion energy and increasing dose, the stress evolution is similar to that shown in Figs 1 and 2, Le., tensile stress initially increases, then decreases to zero, and then compressive stress builds up. No saturation of compressive stress was observed for the dose range considered here. In previous experiments [7], 300 keV Ar ions were used where the depth of ion penetration was greater than the film thickness. Both experiments gave identical results indicating that factors such as modification of interface stress and the non-uniform stress due

21 Fundamental aspects of residual stress evolution in thin metaljilms

Ion Dose (x 10' /cm2)

Figure 4. Evolution of residual stress, measured in situ with dose, in 150 nm thick Cr films irradiated with 110 keV Ar ions. Films were deposited at 5 mTorr Ar pressure with no substrate bias prior to deposition.

to irradiation of only a fraction of the film thickness have little influence on the observed stress evolution.

3.4. TEM observations

The evolution of stress was correlated with the microstructures of the films through TEM observations. The microstructures were investigated for four processing conditions: (a) films deposited at high Ar pressures without any substrate bias (Fig. Sa); (b) films deposited under conditions that produced the maximum tensile stress (Fig. Sb); films deposited under conditions that produced almost no stress (Fig. 5c), and (d) films deposited under conditions that produced the maximum compressive stress (Fig. 5d). Typically, nanocrystalline columnar microstructures are observed as shown in the cross-sectional TEM micrograph in Fig. Sa from a 150 nm thick Cr film deposited at 7.5 mTorr Ar pressure without substrate bias. The intercolumnar regions exhibited Fresnel fringes in through-focus imaging: bright fringes for under focused images, dark fringes for over focused images and no contrast for exact focus images. Through-focus images have been presented elsewhere [7]. These Fresnel fringes indicate voided inter-columnar regions that may only be a few atomic layers thick [8]. A complete series of cross-sectional TEM images as a function of Ar pressure have been presented elsewhere [9]. As the bombardment energy is increased, the inter-columnar voids tend to shrink and at the tensile stress maximum, no Fresnel fringes were observed (Fig. 5b). At near-zero stress, clear grains are observed with no evidence of open volume or radiation-induced defects in the film (Fig. Sc). At compressive stress maximum, clear evidence of radiation-damage-type defect structures is observed (Fig. Sd), although the details of the radiation-induced point defects cannot be analyzed from these TEM images. However, by comparing to Fig. Sc, it is obvious that the major fraction of the grains is covered with a high density of radiation damage that may include vacancies (either isolated or clustered), self-interstitials (either

22 A. Misra and M . Nastasi

Figure 5. TEM images showing (a) cross-sectional view of a 150 nm thick Cr film deposited at 7.5 mTorr Ar pressure without substrate bias, (b) plan view of a 150 nm thick Cr film deposited at S mTorr Ar pressure and -50 V bias, (c) plan view of a 1 p m thick Cr film deposited at -SO V bias at 2.5 mTorr Ar pressure, (d) plan view of a 1 p m thick Cr film deposited at -200 V bias at 2.5 mTorr Ar pressure: (e) schematic illustration of the stress evolution with particle bombardment, points a, b, c and d roughly correlate to the TEM images (a), (b), (c) and (d) with the stress evolution.

isolated or clustered), Frenkel defects and entrapped Ar. The correlation between the above TEM images and stress evolution trends is shown schematically in Fig. 5e. Finally, it should be emphasized that the microstructure evolution with bombardment correlates well with the stress evolution irrespective of the film thickness. For a given stress, the primary difference in the microstructure of the 150 nm and 1 p m thick Cr films was in the grain size: -16 nm in the former and -50 nm in the latter.

Fundamental aspects of residual stress evolution in thin metaljilms 23

+

Figure 5. (Continued).

24 A. Misra and M. Nastasi

4. DISCUSSION

The three different experiments performed reveal the same common trend in intrinsic stress evolution with increasing energetic particle bombardment: tensile stress increases to a maximum, then transitions to compressive stress which builds to a maximum. In this section, we discuss the physical mechanisms that lead to this stress evolution.

4.1. Tensile stress

The initial increase of the tensile stress to a maximum corresponds to closing of the nm-scale inter-columnar porosity as inferred from the TEM images shown in Fig. 5. Recently, Nix and Clemens [lo] have shown that the maximum tensile stress in PVD films may be estimated by balancing the sum of the elastic strain energy and the grain boundary energy with the energy of the two free surfaces that get eliminated in the island coalescence. This upper bound estimate may be as much as an order of magnitude higher than the experimentally measured peak stresses [ 111. Here we attempt to estimate the maximum tensile residual stress using the following expression from the grain boundary relaxation model originally proposed by Hoffman [12]:

E Amax tensile - Omax (1 - u ) d ’

where c r ~ ~ ~ l l e is the maximum tensile stress, E is Young’s modulus, u is Poisson’s ratio, A is the relaxation distance and d is the grain size. According to this model, the attractive interatomic forces between the adjoining islands lead to tensile elastic straining of the film. The elastic strain continues to increase until the islands coalesce completely to form a grain boundary. In other words, significant tensile stress may be generated even before the two free surfaces are replaced by a grain boundary. Using equation ( I ) , the tensile stress can be calculated if the elastic displacement A can be estimated. However, a major limitation of the Hoffman model [12] is that there is no direct way to know A. This in fact was a motivating factor in the development of the Nix-Clemens model to allow estimation of tensile stresses without knowing A. We show that the interatomic potentials may be used to estimate Amax and obtain oE$~ values that compare well with the experimental values.

The well known interatomic potential energy (E)-interatomic distance ( a ) rela- tionship is shown in Fig. 6a. Here the potential energy ( E ) is plotted against the ratio of a to the equilibrium interatomic distance (ao). As an approximation, we have used the universal binding-energy relation developed by Banerjea and Smith [ 131. The differentiation of E with respect to a gives the interatomic force and the differenti- ation of E with respect to volume ( V ) gives the interatomic mean stress (negative of hydrostatic pressure). Thus, the E vs. ala0 relation (Fig. 6a) may be used to obtain the dependence of interatomic mean stress on a/ao, shown in Fig. 6b. In Fig. 6b, the positive numbers indicate tensile stress and negative numbers indicate

Fundamental aspects of residual stress evolution in thin metal films 25

0

-1

- - 2 5 h !? 2 -3 W

- 4 I substrate I insignificant interatomic forces between columns

(b) (dl a > - l.lao

-5 1 .5 2 alao 0.5

-+I Ir- A 20

u o w 01 01

E-20 ;j

substrate

-40 I

maximum attractive interatomic -60 " " " " . I " " " " " ~ " " " "

0.5 alao 1.5 2 forces between columns

Figure 6. (a) Interatomic potential energy vs. ala0 where a is the interatomic distance and a0 is the equilibrium value of a, and (b) interatomic mean stress vs. ala0 calculated using the universal binding energy relation [13]; (c) and (d) schematic illustrations of the development of tensile stresses in films as the distance between adjoining columns, a, decreases.

compressive stress. Note the similarity between the evolution of tensile stress with decreasing interatomic distance (Fig. 6b) and the stress evolution with increasing bombardment (Fig. 5e). We correlate the interatomic force- distance relation with the film microstructure through the approximation that the islands (or columns) in the film are separated by a distance a . As shown in Fig. 6c, the adjoining islands will not experience any significant attractive interatomic forces if the gap, a , be- tween the islands is large. In fact, tensile stresses (trend shown in Fig. l ) have been observed to decrease almost to zero as Ar pressure is increased to -20 mTorr for tungsten films 1141. From Fig. 6b, it follows that when a > -1.7~0, the stress tends to zero. Further, attractive interatomic forces reach a maximum when a FZ 1 . 1 2 ~ 0 (Fig. 6d). Hence, Amax = 1 . 7 ~ 0 - 1.12~0 = 0 . 5 8 ~ ~ .

For Cr, a0 = 0.249 nm and u = 0.21 1151, grain size d = 21 f 2 nm from Fig. 5b, and E = 230 GPa as determined by us on these films by nanoindentation 1161. Substituting these values in equation (l), we get = 2 f 0.15 GPa. This is in good agreement with the experimentally measured maximum tensile stress of - 1.6 GPa (Figs 1 and 2). There are two possible reasons for the discrepancy between the experiment and the model: (i) the approximate interatomic potential

26 A. Misra and M . Nastasi

Figure 7. Schematic of the Orowan mechanism for dislocation glide in thin films (only one columnar grain is shown for simplicity). The process involves dislocation bowing between the obstacles (film-substrate interface or grain boundary) and creating interface dislocations as the bowed segment advances on the slip plane. In the films studied here, the grain size (d ) is significantly smaller than the film thickness ( h ) and hence d is taken as the obstacle spacing in the Orowan model.

use for Cr, and (ii) stresses measured are ex situ, i.e., samples are taken out of the vacuum chamber for stress measurement and any oxygen adsorption on the open columnar boundaries may reduce the tensile stress [15]. Finally, we note that the maximum tensile residual stress may not exceed the biaxial yield strength of the film. An estimate of the films yield strength is described next.

In thin films with the microstructural length scales (in-plane grain size and film thickness) on the order of a few tens of nanometers, long dislocation pile-ups can- not be supported and hence, Hall-Petch type models for strengthening due to grain refinement break down [18]. The plastic flow in these fine-scale structures is ac- complished by the motion of single dislocations by bowing between the boundaries (or interfaces) and creating interface dislocations, as shown schematically in Fig. 7 [19, 201. The yield strength as determined by this Orowan bowing stress is given as [ 201 :

oorowan = M Gb d-' in( g), 4x(l - U )

where M is the Taylor factor ( ~ 2 . 7 5 for bcc metals), G is the shear modulus (95 GPa for Cr films [16]) and b is Burgers vector (e2.5 A for Cr). In general, grain size and film thickness are additive when applying the Orowan model to estimate the yield strength of thin films [21], but in our case since the film thickness is significantly greater than the grain size, we have only considered the grain size. Substituting these values we obtain aorowan = 1.4 & 0.1 GPa, for d = 21 f 2 nm, in good agreement with the maximum tensile stress measured. This indicates that the biaxial yield strength limits the experimentally measured residual stress.

4.2. Compressive stress

Tensile stress evolution was interpreted in terms of the attractive interatomic forces that develop when excess volume is eliminated from the growing films. On the other hand, compressive stress develops due to creation of excess volume through

Fundamental aspects of residual stress evolution in thin metal films 27

point defects. Recent molecular dynamics simulations show that during energetic particle deposition of metal films both vacancy [22] and stable interstitial formation [23] are observed. The self-interstitials and incorporated Ar will tend to build up compressive stress, while vacancies may have a small but opposing effect. We, therefore, model the compressive stress in terms of the Frenkel defect misfit volume and concentration. The compressive stress due to defects, a;"', is given as:

where ~ F D and f~~ are the atomic concentrations of Frenkel defects and entrapped Ar respectively, S2 is Cr atomic volume, AS~FD is the volume change associated with a Frenkel defect and A a A r is the volume change associated with an Ar atom entrapped in Cr. f~~ has been measured by RBS as -0.005 [16]. Typically, the saturation concentration of ~ F D in metals is on the order of 0.01 f 0.005 [24], (ARFD/S~> z 1 for Cr [25] and ( A S ~ A ~ / S ~ ) z 1.75 [16]. The biaxial modulus, E/(1 - u ) , for Cr films with large compressive stress was ~ 3 0 0 GPa, as measured by nanoindendation [ 161. Substituting these values in equation (3) we estimate 02"' = 1.875 k 0.5 GPa. The relatively large error range is due to the uncertainty in the saturation value of fFD. However, the estimate compares well with the experimentally measured compressive stress maximum of z2.1 GPa.

Finally, we note that the compressive saturation stress is higher than the maximum tensile stress (Fig. 2). If the maxima in residual stress are defined by the film yield strength, this implies that the yield strength of films processed under conditions of high bombardment is higher. The model for yield strength shown in Fig. 7 defines the yield strength through the Orowan bowing stress with grain boundaries as the obstacles. However, if the grains contain radiation-induced point defects (Frenkel pairs or interstitial clusters), these will pin the dislocations and further impede the expansion of Orowan loops in the grains. Hence, the additional hardening due to point defects has to be considered. Hardening due to interstitial solutes is more rapid as compared to substitutional solutes since the strain field around interstitial solutes is usually elliptical while substitutional solutes have spherically symmetric strain field [26]. This rapid hardening (Aadefects) due to interstitial solutes is described by the following model [26]:

where a is an empirical parameter, c is the solute concentration and AE is the difference between the longitudinal and transverse strains in the elliptical strain field of interstitials. The A a c( AE dependence implies a diffuse interaction between solutes and dislocations, whereas a localized interaction is more applicable for impenetrable solute clusters leading to a A a c( A E ~ / ~ type dependence [26]. From the parameters used in equation (3) , we estimate the solute concentration as, c = ~ F D + f~~ = 0.015 f 0.005. As = 0.40 and a = 20 for typical interstitial hardening in bcc metals such as Fe at room temperature [26]. Using these values,

28 A . Misra and M. Nastasi

we obtain A o d e f e c t s from equation (4) as 0.64 f 0.1 GPa. The new yield strength of the film (at high bias deposition), crJ , is given as:

( 5 ) eb = cyIeld + A g d e f e c t s .

With o$ld = 1.56 f 0.1 GPa (using grain size of 19 nm for films with maximum compressive stress) and Aodefec ts = 0.64f0.1 GPa, it follows from equation ( 5 ) that

= 2.2 f 0.2 GPa. This is in good agreement with the experimentally measured saturation in compressive stress.

For post-deposition ion irradiation at high energies and to high doses, the self- interstitials may form clusters or even collapse into interstitial loops [27]. These loops preferentially align normal to the tensile stress direction [28] to relax the tensile stress and continued growth in the size of the loops will eventually lead to the development of compressive stress [29, 301. The compressive stress buildup in these high energy ion-irradiated films has been shown to be proportional to (dpa)*I3 where dpa is displacements per atom [29, 301.

The relaxation of compressive stress with increasing bias (Fig. 3 ) has also been observed in other investigation [31]. Since we have shown that the maximum in compressive stress is correlated with point defect build up and saturation, it is likely that plastic flow in these materials is accomplished, in part, by irradiation enhanced creep [29].

5. SUMMARY

The findings of this investigation are summarized as follows: (i) Metal films deposited at low homologous temperatures in the absence of

any significant energetic particle bombardment exhibit tensile stresses and nanoscale columnar porosity. With increasing bombardment, the tensile stress initially increases to a maximum as the gaps between the columns decrease and the attractive interatomic forces between the adjoining columns increase in magnitude.

(ii) Further increases in bombardment result in a complete relaxation of tensile stress, build up of compressive stress to a maximum followed by a gradual relaxation. The universal nature of this trend was revealed in three different experiments: decreasing Ar pressure in sputtering, increasing substrate bias and post-deposition ion-irradiation.

(iii) Compressive stress evolution may be interpreted using a model based on bombardment-induced point defects (e.g., self-interstitials and entrapped Ar) that effectively add free volume in the plane of the film.

(iv) Maximum values of tensile and compressive stresses are limited by the film yield strength. Hardening due to point defects makes the compressive yield strength higher than the tensile yield strength.

Fundamental aspects of residual stress evolution in thin metaljilms 29

Acknowledgements

This research was sponsored by the Department of Energy, Office of Basic Energy Sciences.

REFERENCES

1. H. Windischmann, Crit. Rev. Solid State Mater. Sci. 17, 547 (1992). 2. M. F. Doerner and W. D. Nix, Crit. Rev. Solid State Mater. Sci. 14, 25 (1988). 3. J. A. Thornton and D. W. Hoffman, Thin Solid Films 171, 5 (1989). 4. G. G. Stoney, Proc. Roy. SOC. (London) A82, 172 (1909). 5. C. A. Volkert, J. Appl. Phys. 70, 3521 (1991). 6. A. I. Van Sambeek and R. S. Averback, Muter. Res. SOC. Symp. Proc. 396, 137 (1996). 7. A. Misra, S. Fayeulle, H. Kung, T. E. Mitchell and M. Nastasi, Appl. Phys. Lett. 73, 891 (1998). 8. E. S. Machlin, Materials Science in Microelectronics - The Relationships between Thin Film

9. A. Misra and M. Nastasi, J. Matel: Res. 14, 4466 (1999). Processing and Structure, Vol. 1, pp. 157- 184. GIRO Press, New York (1995).

10. W. D. Nix and B. M. Clemens, J. Mater. Res. 14, 3467 (1999). 11. A. Misra, H. Kung, T. E. Mitchell and M. Nastasi, J. Muter. Res. 15, 756 (2000). 12. R. W. Hoffman, Thin Solid Films 34, 185 (1976). 13. A. Banerjea and J. R. Smith, Phys. Rev. B 37, 6632 (1988). 14. M. Itoh, M. Hori and S. Nadahara, J. Vac. Sci. Technol. B 9, 149 (1991). 15. C. J. Smithells (Ed.), Metals Reference Book, 5th edn. Butterworths, London (1976). 16. A. Misra and M. Nastasi, Appl. Phys. Lett. 75, 3123 (1999). 17. A. Misra and M. Nastasi, J. Vac. Sci. Tecknol. A 18, 2517 (2000). 18. B. M. Clemens, H. Kung and S. A. Barnett, MRS Bulletin 24, 20 (February 1999). 19. J. D. Embury and J. P. Hirth, Acta Metall. 42, 2051 (1994). 20. W. D. Nix, Scripta Mater. 39, 545 (1998). 21. C. V. Thompson, J. Muter. Res. 8, 237 (1993). 22. X. W. Zhou, R. A. Johnson and H. N. G. Wadley, Acta Mater. 45,4441 (1997). 23. C. M. Gilmore and J. A. Sprague, NanoStructured Mater. 9, 643 (1997). 24. M. Nastasi, J. W. Mayer and J. K. Hirvonen, Ion-Solid Interucrions - Fundamentals and

Applications, Cambridge Solid State Science Series, p. 343. Cambridge University Press, Cambridge (1996).

25. P. Ehrhart, Atomic Defects in Metals, Landolt-Bornstein, New Series, Group 111, Vol. 25. Springer, New York (1991).

26. R. L. Fleischer, in: The Strengthening ofMetals, D. Peckner (Ed.), pp. 93-162. Reinhold Press, New York (1964).

27. D. J. Bacon, A. F. Calder and F. Cao, Radiation Effects and Defects in Solids 141, 283 (1997). 28. A. B. Lidiard and R. Perrin, Phil. Mag. 14, 49 (1973). 29. A. Misra, S. Fayeulle, H. Kung, T. E. Mitchell and M. Nastasi, Nucl. Instrum. Methods B. 148,

30. A. Jain, S. Loganathan and U. Jain, Nucl. Instrum. Methods B 127/128, 43 (1997). 31. B. Window, F. Sharples and N. Savvides, J. Vac. Sci. Technol. A 6, 2333 (1988).

211 (1999).

Adhesion Aspects of Thin Films, Vol. 1, pp. 31-34 Ed. K. L. Mittal 0 VSP 2001

Surface stress effects on the intrinsic stress in thin films

R. C. CAMMARATA * Department of Materials Science and Engineering and Department of Mechanical Engineering Johns Hopkins University, Baltimore, M D 21218, USA

Abstract-During the early stage of non-epitaxial growth of metallic thin films, it is often observed that the discontinuous film is deposited in a state of compressive stress. This behavior can be understood in terms of surface stress effects. Surface stress is a thermodynamic parameter characteristic of every solid surface and is associated with the reversible work per unit area to elastically deform the surface. It is shown how surface stress effects lead to a compressive intrinsic stress during the initial island stage of thin film growth once the islands become firmly attached to the substrate.

Keywords: Intrinsic stress; surface stress; island growth.

1. INTRODUCTION

Virtually all thin films grown on a substrate are in a state of internal stress [ 1 - 81. In the case of non-epitaxial films, the types of stresses observed in thin films can be classified as either thermal stresses generated by a temperature change when the film and substrate have different coefficients of thermal expansion, or intrinsic stresses resulting from processes that would change the dimensions of the film were it not attached to the substrate. It is noted that such stresses can be generated only if there is a strong interaction between film and substrate, so that there is an obvious and important connection between adhesion and internal stress in thin films.

The intrinsic stress behavior during non-epitaxial metal film growth has recently been reviewed by Nix and Clemens [9], who based much of their discussion on the extensive experimental work of Abermann and coworkers [ 10- 141. Compressive stresses commonly develop during the early stage of metallic film growth prior to the film becoming continuous. During island coalescence, the intrinsic stress generally displays a sudden jump that results in the intrinsic stress becoming tensile. Several theories have been proposed to explain the development of the tensile intrinsic stresses [l-91; in particular, it is believed that in many cases the tensile stress

*E-mail: rcc @ jhu.edu

32 R. C. Cammarata

can be associated with structural relaxations at the grain boundaries formed during island impingement [1, 2, 5, 10, 111. It has been suggested that the compressive stress that develops before coalescence is the result of a dependence of the lattice spacing on crystal size [15]. There appears to be some confusion in the literature as to the precise formulation of this effect, its relative importance, and even the sign of the intrinsic stress generated. In this paper, the development of the early stage compressive stress is discussed in terms of surface stress effects.

2. SURFACE STRESS

The free energy y of a solid surface represents the reversible work per unit area required to create a new surface by processes such as cleavage or creep. The surface stress is associated with the reversible work per unit area needed to elastically deform a pre-existing surface [16]. Letting denote the surface elastic strain tensor (i, j = 1, 2), the surface stress tensor f i J = yai j + a y / a ~ ~ ] , where ai j is the Kronecker delta. For a surface possessing three-fold or higher rotational symmetry, the surface stress is isotropic, and can be rewritten in scalar form as

(1)

The physical origin of the surface stress can be understood in the following way. The nature of the bonding of the atoms on a surface is different from the bonding between atoms in the interior (e.g., atoms at the surface have fewer nearest neighbors than atoms in the interior). Because of the difference in the nature of the bonding, the surface atoms would have an equilibrium interatomic distance different from that of the interior atoms if the surface atoms were not constrained to remain structurally coherent with the underlying lattice. As a result, the surface can be considered as exerting a pressure, often referred to as the Laplace pressure, on a finite-size solid that results in a volume elastic strain. In the case of low index metal surfaces, experimental measurements and theoretical calculations [ 161 indicate that the surface stress is positive and of order 1 to 5 N/m.

f = y + a y / a s .

3. LAPLACE PRESSURE AND INTRINSIC STRESS GENERATION

The Laplace pressure P exerted on an isotropic, finite-size solid owing to the surface stress can be obtained by setting the virtual work acting on the solid, PdV, resulting in a variation in the volume dV, equal to the virtual work performed by the surface stress, f dA, resulting in a variation in the surface area dA:

P = f(dA/dV). (2) It is noted that the Laplace pressure of a solid depends on f and not on y as is often assumed. Consider an island formed during the early stage of film growth that is modeled as a disk. Let this disk be referred to a Cartesian coordinate system where the z-axis is perpendicular to the plane of the depositing film. For

Surface stress effects on the intrinsic stress in thinfilms 33

simplicity, the only surface areas that will be considered are those parallel to the x-y plane. In addition, it will be assumed that the strain state of the island can be taken as approximately uniform. Let f denote the surface stress associated with the top surface, and let g denote the surface stress associated with the base of the disk (the film-substrate interface). It should be noted that it is possible to define two independent surface stresses for a solid-solid interface [16]; for the present purposes, g denotes the surface stress associated with the interfacial work produced when the film and substrate are elastically deformed by the same amount parallel to the film-substrate interface. Using equation (2), the in-plane Laplace pressure P for the island is

where t is the island thickness. This Laplace pressure imposes a biaxial volume stress a,, = aJ, = - P . The in-plane radial strain E induced by these volume stresses can be expressed using Hooke’s law as E = S;larx + S i p , , , where Si, and Si, are elastic compliances referred to the previously defined Cartesian coordinate system. Using equation (3), the equilibrium in-plane strain in the bulk of the island, as a function of thickness, is

where Y = l/(S{l + Si2) is the in-plane biaxial elastic modulus. In order for an intrinsic stress to be developed, a growing island that initially

could migrate across the surface must become firmly attached to the substrate. Let the thickness when this attachment first occurs be denoted as to. This thickness will be determined in part by the degree of adhesion between the island and the substrate, as well as by various kinetic processes occurring during deposition. The island of thickness to will be in a stress-free state with its equilibrium in-plane lattice spacing equal to n [ l + &(to)] , where a is the bulk lattice spacing and &(to) is the strain given by equation (4), evaluated at the thickness to. During further deposition, if the island were not constrained by the substrate, the equilibrium strain relative to the bulk state, E @ ) , would be given by equation (4), and the equilibrium strain relative to that for the island of thickness to would be As = ~ ( t ) - &(to). Since the island of thickness to is constrained by the substrate such that it could not elastically deform in the x - y plane, the substrate must impose an in-plane biaxial stress to oppose the latent strain A& as the thickness increases. As aresult, an intrinsic stress a = -Y A& is generated in the island due to surface stress effects that can be expressed as [ 161

When t >> to, the thickness-dependent intrinsic stress contribution will approach a limiting value of

For an order of magnitude calculation, the sum of the surface stresses can be taken as f + g = 2 nm, resulting in a0 = -2/t0 GPa, where to is in units of nanometers.

34 R. C. Cammarata

If the crystallites first become firmly clamped to the substrate at a thickness of about 2 nm, then the limiting value for the intrinsic stress generated will be of order -1 GPa. This example illustrates how surface stress effects can explain the compressive stresses often observed during the island growth of non-epitaxial metal films. (Surface stress effects may also explain the origin of compressive stresses often observed in fully grown, continuous metallic and dielectric films [17].) A more complete analysis would include the contribution of the surface stress of the curved surface of an island that, at coalescence, becomes transformed into a grain boundary. This contribution leads to interesting effects that are discussed elsewhere [ 171.

Acknowledgements

The author gratefully acknowledges support from the National Science Foundation administered through the Materials Research Science and Engineering Center at Johns Hopkins University.

REFERENCES

1. D. S. Campbell, in: Handbook ofThin Film Technology, L. I. Maissel and R. Glang (Eds), p. 123.

2. K. Kinosita, Thin Solid Films 12, 17 (1972). 3. R. W. Hoffman, in: Physics of Non-Metallic Thin Films, NATO Advanced Study Institute

4. K. Kinosita, Thin Solid Films 50, 205 (1978). 5. M. F. Doerner and W. D. Nix, CRC Crit. Rev. Solid State Mater. Sci. 14, 225 (1988). 6. M. Ohring, The Materials Science ofThin Films, Chap. 9. Academic, Boston (1992). 7. R. Koch, J. Phys. Condens. Mater. 6, 9519 (1994). 8. F. A. Doljack and R. W. Hoffman, Thin Solid Films 12,71 (1972). 9. W. D. Nix and B. M. Clemens, J. Mater. Res. 14, 3471 (1999).

McGraw-Hill, New York (1970).

Series B, Vol. 14, C. H. S. Dupuy and A. Cachard (Eds), p. 273. Plenum, New York (1976).

10. R. Abermann, R. Kramer and J. Maser, Thin Solid Films 52,215 (1978). 11. R. Koch and R. Abermann, Thin Solid Films 129,71 (1985). 12. R. Abermann, Thin Solid Films 186, 233 (1990). 13. R. Abermann, Vacuum 41, 1279 (1990). 14. R. Abermann, Mater. Res. SOC. Symp. Proc. 239,25 (1992). 15. M. Laugier, Vacuum 31, 155 (1981). 16. R. C. Cammarata, Prog. Surf: Sci. 46, 1 (1994). 17. R. C. Cammarata, T. M. Trimble and D. J. Srolovitz, J. Matel: Res. (in press).

Adhesion Aspects of Thin Films, Vol. 1, pp, 35-50 Ed. K. L. Mittal @ VSP 2001

Influence of the ion bombardment on the stress in thin films produced by ion beam and RF sputtering techniques

SALVATORE SCAGLIONE I , * , FRANCESCA SARTO I ,

ANTONELLA RIZZO 2 , MARC0 ALVISI and MARIA ANTONIETTA TAGLIENTE

ENEA, Dipartimento Innovazione, CR Casaccia, Via Anguillarese 301, 00060 Rome, Italj Pastis CNRSM, S.S.7 Appia km. 712, 72100 Brindisi, Italy

Abstract-In general, optical thin films, produced by physical vapor deposition methods involving ions (ion beam sputtering, RF sputtering) are denser than films grown by the conventional thermal evaporation methods. Nevertheless, the high level of the film stress and its effect on film-substrate adhesion is the main disadvantage that limits the use of the ion techniques to deposit optical coatings. A knowledge of the relationship between deposition parameters and the intrinsic stress should be helpful in pursuit of coating mechanical stability enhancement.

In this paper, the intrinsic stress in optical thin films is shown to be related to the deposition parameters.

Ion beam sputtered hafnium oxide with additional ion pre-treatment as protective layer for plastic optics and the possibility of improving the film/substrate adhesion by reducing the total stress is reported. The optical and structural properties of zinc selenide thin film, deposited by RF sputtering technique, are investigated considering the momentum transfer from neutral particles to the growing film as the main parameter governing the film properties. We propose that the momentum transferred to the film during deposition is the main factor governing the optical, structural and mechanical properties of the films.

Keywords: Thin film; ion beam assistance; RF sputtering; Hf02; ZnSe; stress

1. INTRODUCTION

The ion bombardment during thin film growth by ion beam sputtering at low energy appears to be a promising technique to improve the adhesion quality of films. When an energetic ion impinges on the surface of a material, solid or liquid, a momentum exchange process takes place: the bombarded atoms receive sufficient energy to

*To whom correspondence should be addressed. Tel.: 0039 06 30486322; Fax: 0039 06 30486364; E-mail: salvatore.scaglione@casaccia.enea.it

36 S. Scaglione et al.

recoil, break the bonds and begin a collision cascade process in which both the incident ions and the recoiling atoms lose fraction of their energies.

This energy is available to promote several effects [ 11: (i) surface cleaning, i.e., removal of organic contamination from the surface,

(ii) etching of material from the surface, which can remove a weak boundary layer, (iii) cross-linking or breaking of near-surface molecules, which can cohesively

(iv) modification of the surface-chemical structure, due to ion mixing and/or

All these processes, alone or in synergistic combination, enhance the film/ sub- strate adhesion.

It is well known that stresses are present in thin films even without the application of external load. Many factors contribute to the residual stress in thin films. Among these is the thermal stress due to mismatch between the expansion coefficients of the film and the substrate. The total residual stress at a film/substrate interface is usually measured and correlated with different factors. The residual internal stress in the film can strongly affect the filmlsubstrate adhesion [2-41. Usually films under compressive stress are harder than those under tensile. On the other hand, the ion beam assisted deposition technique has been used to control residual stress by a suitable choice of the deposition parameters such as, current density, mass and energy of assisting ions and growth rate [ 5 ] .

The film microstructure dictates mechanical stability and optical properties of films. To improve the mechanical and optical properties of the films, the microstruc- ture must be modified in order to obtain dense thin films with a low intrinsic stress, high adhesion between the film and the substrate as well as between individual lay- ers, and low optical absorption.

The first stage in the growth of a thin film can be represented as the condensation of vapor phase material on the substrate. During the coalescence process, in most cases, not all film volume is filled and a pronounced columnar structure with voids located both along the boundaries and in the interior of each column occurs. Ion beam bombardment of the growing film can increase the mobility of the condensing atoms, thus avoiding the columnar structure formation. An enhancement of the adhesion and a reduction of the intrinsic stress for ion assisted deposited films has been reported [6, 71. The role of ion beam assistance during film growth has been extensively studied to obtain more compact, dense, adherent and stress controlled films. Most of the effort has been directed to find the correlation between the film properties and the parameters of the assisting ion beam (energy, ion mass, and current density). The importance of the momentum transfer parameter has been recognized in a few cases only. Furthermore, the resulting changes in optical and mechanical properties have been primarily studied and correlated to the momentum transfer parameter. Kester and Messier [8] have focused their attention on structural changes besides changes in optical and mechanical properties and have concluded a more precise interpretation of the momentum transfer parameter effect.

strengthen the surface layer,

chemical reaction with the incident ions.

Influence of the ion bombardment on the stress in thin,fi/ms 3 1

Among the high index materials in the UV spectral range hafnium oxide (HfOz) is a very promising material for laser optical coatings, due to its relatively high damage threshold and good thermal and mechanical stabilities. The microstructure of the deposited hafnia films (HfOz), in particular the orientation of the crystallites, was found to be strongly dependent on the momentum transfer parameter values

Other limitations appear when using hafnia as a high refractive index coating in ophthalmic applications, especially when plastic substrates are used. In fact, one of the main factors responsible for adhesion failure of coatings is the stress developing at the coating/ substrate interface when the coating is cooled to room temperature. In particular, the thermal stress plays an important role in adhesion failure of plastic substrates coated with oxide films, because the expansion coefficient of polymers is about an order of magnitude greater than that of hard oxides. Therefore, the bombardment of the substrate with energetic ions prior to deposition of the coating reduces stress and thus improves the coating/substrate adhesion [ 101. Ion beam etching of the substrate before film deposition activates the substrate surface, and results in better film/substrate adhesion [I 1 - 131.

Zinc selenide (ZnSe) thin films prepared by physical vapor deposition (PVD) techniques on single-crystal substrates are widely used for a variety of technological applications, for example in optoelectronic devices such as blue light emitting diodes, solar cells, and dielectric mirrors [ 141. Many important properties of interest for such applications are determined by the microstructure of the films developed during the deposition process. In fact, both the grain size and texture of the ZnSe film can affect the wear behavior of the coating. The residual stress is also very important in the mechanical stability of thin film devices. In almost all PVD processes the film is left in a strained state at the end of the deposition and in some circumstances cracking or peeling of the coating can occur. This kind of failure usually occurs at some critical thickness, when the internal stress reaches a value which overcomes even strong chemical bonds at the interface. The study of the reasons for the appearance of stresses and the methods of reducing them is a critical problem in the optical coating manufacturing.

The stress field depends on both the materials and conditions existing during the deposition: pressure, composition of the residual gases, substrate temperature and deposition rate. Both tensile and compressive stresses were observed [15] in the case of sputtered thin films. The magnitude and the sign of the stress depend on the parameters which determine the momentum transferred to the growing film by the energetic particles bombarding it during the deposition such as ions (for ion assisted films) or reflected neutral species (for films sputtered without ion assistance) [16]. This dependence has been verified in many works dealing with ion-beam assisted sputtering [17]. On the other hand, due to the difficulty in measuring the parameters dictating the momentum transfer (type, energy and number of bombarding species) in radio frequency (RF) sputtering, only a few investigations have been dedicated to the study of the properties of RF sputtered thin films as a function of the momentum transfer parameter [ 161.

( P I 191.

38 S. Scaglione et al.

In this paper the influence of the particles bombardment on the optical and me- chanical properties of the growing hafnia and zinc selenide thin films is discussed. The deposition parameters are expressed in terms of the momentum transfer from the bombarding species to the deposited atoms. A correlation between the momen- tum transfer and the residual stress in the films is reported in order to improve the mechanical stability (reduced residual stress) of optical materials used for techno- logical applications, i.e. hafnia film on a plastic substrate and thick zinc selenide layer for infrared optical coating (A. = 10.6 pm). In addition to the mechanical sta- bility, an optical coating should be very dense in order to prevent the variation of the optical properties induced from the absorption of moisture from the environment.

2. HfOz LAYER

2.1. Ion beam assisted evaporation

The momentum transferred from the bombarding ions to the growing film influences the density (refractive index) of the film [17] itself. In the ion beam assisted deposition (IBAD) the densification mechanism can be explained in terms of the collision cascade model [18], where the atoms recoiling from the surface are implanted beneath the surface itself. In previous investigations, Muller [ 19, 201 had considered that the implantation of recoiling atoms increased the density of the region just below the surface and, as a consequence, a depletion zone formed at the film surface. In the IBAD process, the depleted region can be filled by properly adjusting the energy and the ion-to-atom arrival ratio (see Fig. 1). As such, the density of the ion assisted deposited thin film depends not only on the energy of the bombarding ions, but also on the ion-to-atom arrival ratio.

The momentum transfer per arriving atom, i.e. momentum transfer parameter P , can be defined as

Where nion and natom are the ion and atom numbers, respectively, determined by measuring the growth rate and the ion current density at the substrate, m and E are the mass and the energy of ions, respectively, and y = (4 . m . M ) / ( ( m + M ) * ) where M is the mass of deposited atoms.

Figure 2 shows the variation of the refractive index of a set of Hf02 samples deposited by varying the momentum transfer parameter P of the assisting Xe ions. The ion beam energy (E) , the ion current density (4 and the growth rate (Rate) were the parameters that were varied to obtain different values of P (Table 1). By increasing the momentum transfer parameter the refractive index increases too [21]. The change in the refractive index with the momentum transfer parameter shows the densification effect of Xe ion assistance on the hafnia films [22].

The sharp increase in the refractive index for P 3 200 (atomic mass unit.eV)’/’ or (amu.eV)’/2 could be justified by the preferential sputtering of oxygen atoms during the bombardment process. In fact, as the oxygen atoms are lost the percentage of

Influence of the ion bombardment on the stress in thinfilms

I I I I I 2.14- 2.12-

2.10- 2.08 -

-

....'; h = 500 nm

$ 2.06- - LI 2.04-

;5 2.02-

.- - > .- - 5 2.00: -

39

Figure 1. Layout of the ion beam assisted process.

1.96 1.98 v 1.94 4

0 50 100 150 200 250

momentum transfer parameter, P (amu.eV)"*

Figure 2. Refractive index of HfO2 thin film at E, = 500 nm plotted vs. the value of the momentum transfer parameter P .

the metallic component (hafnium in this case) increases as well as the refractive index. This effect is negligible for momentum transfer parameter lower than 200 (amu . eV)''', where the increase of the refractive index is due to the increase of the film density. The behavior of the refractive index (A = 3 10 nm) with the momentum transfer parameter shown in Fig. 3 confirms this.

The X-ray diffraction patterns of the samples grown at different momentum transfer parameters present different features, as observed in Fig. 4 for IBAD deposited hafnia thin films. The samples can be divided in three different groups.

40 S. Scaglione et al.

Table 1. Deposition parameters of e-gun Xe ion assisted deposited hafnia thin film. E b and I b are the energy and the ion current, respectively, of Xe ion beam. Rate is the growth rate measured without the ion bombardment, J is the ion current density at the substrate and P is the momentum transfer parameter calculated by equation (1)

Sample # Eb (ev) I b (mA) Rate (nm/s) J (pA/cm2) P (amu.eV)1/2

2.14-

2.12-

2.10-

2.08 -

0 1 2 3 4 5 6 7 8 9

- 150 200 200 200 200 200 200 200 300

- 70 70 63 78 84 63 84 63 70

0.8 0.7 0.5 0.5 0.5 0.5 0.3 0.3 0.2 0.2

- 15 17.5 20 22 26 20 26 20 20

0 37 71 81 89

105 134 174 20 1 247

X

c 4 .-

A h=310nm 2.16

/ /

+ , , , , , . , , , , , I 0 50 100 150 200 250

momentum transfer parameter, P (amu.eV)’”

Figure 3. Refractive index of HfO2 thin film at i = 310 nm for different momentum transfer parameters, P .

For P values ranging between 37 and 81 (amu . eV)’/2 the deposited films show the presence of a fully crystalline structure. The whole diffraction pattern could be indexed by the Hf02 monoclinic form [23]. For higher values of P ( P = 89 and P = 105 (amu . eV)’/2) the patterns present a broad halo with some small and narrow peaks superimposed. Such diffraction patterns are typical of nearly amorphous structure with some crystallites embedded in the amorphous matrix (4% for the sample deposited at P = 105 (amu . eV)’/*). At the highest P values ( P from 134 to 250 (amu . eV)’l2) the film structure is again fully crystalline, but the presence of a single intense peak points out the presence of a strong preferred orientation of the micro-crystallites. The angular position of this peak clearly

Influence of the ion bombardment on the stress in thin films 41

20 40 60 80 100 2 8 (degrees)

Figure 4. X-ray diffraction patterns of the hafnia films deposited with different P values plotted vs. the diffraction angle. Broken lines indicate the diffraction planes.

excludes a transition from the monoclinic to the cubic or tetragonal Hf02 phase. On comparing the X-ray diffraction patterns of the low P value samples with those of a standard hafnia powder the peak positions are similar. The occurrence of a preferred orientation is accompanied by a peak shift towards lower scattering angles as shown in Fig. 5. This shift indicates a cell expansion along the direction perpendicular to the substrate plane, which is caused by an in-plane compressive stress state. This result is in agreement with the behavior of the residual stress with the momentum transfer parameter shown in Fig. 6 and calculated by the Stoney equation:

E S2 o =

6(1 - U ) t

where Rb and R, are the mean radii of curvature of the substrate before and after the deposition, respectively, t is the coating thickness, 6 is the substrate thickness and E and u are Young’s modulus and Poisson’s ratio of the substrate, respectively.

2.2. Ion beam sputtering

A similar mechanism can be envisioned for ion beam sputtered films, even without ion assistance. The sputtered atoms arrive at the substrate surface with an energy (a few tenths of an eV) high enough to cause recoiling of the surface atoms into the film [24], thus filling the voids and pores of the deposited coating. A higher growth rate corresponds to a higher flux of energetic atoms bombarding the growing film. A theoretical support for this interpretation can be given by TRansport of Ions in Matter (TRIM) simulations. Figure 7 shows a comparison of the distributions of recoiling atoms produced by 300 eV Xef ions with those caused by 44 eV Hf and 35 eV 0 beams. The distributions are obtained from TRIM in units of number of recoiling atoms per incident particle. The calculation for Xe ion bombardment at 300 eV simulates the effect of the ion assistance. These distributions have been

42 S. Scaglione et al.

26 28 30 32 20 degrees

Figure 5. (-111) X-ray diffraction peak enlarged for an hafnia sample with low P (tensile stress) and one with high P (compressive stress).

w

0 50 100 150 200 250 300

momentum transfer parameter, P (amu.eV)"*

Figure 6. Residual stress in e-gun Xe ion assisted deposited hafnia thin films for different P values. Positive values represent tensile stress, negative values compressive stress.

multiplied by the number of ions/cm2 per second corresponding to the current density value (10 pA/cm2) used in the experiment by Sarto et al. [25]. The energy values of 44 eV chosen for Hf and 35 eV for 0 are the average values with which Hf and 0 atoms are sputtered from the Hf02 target, in the direction normal to surface, when an energetic (1500 eV) Ar+ beam hits the target at an angle of incidence of 45". The calculated distributions have been multiplied by the number of atoms/cm2 per second arriving at the growing film surface. The number used was determined using the measured sputtering growth rate of 0.1039 nm/s. The results of Fig. 7

Injueizce of the ion bombardment on the stress in thin films 43

-A- Hf recoils due to Xe -A- 0 recoils due to Xe -.- Hf recoils due to Hf and 0 -0- 0 recoils due to Hf and 0

0.0 0.5 1 .o 1.5 2.0 2.5 3.0 depth from surface (nrn)

Figure 7. Distribution of recoils produced by the bombardment of a Hf02 target by the following species: Xe ion beam with an energy of 300 eV and a density of 6.25 x l O I 3 ions/cm2 s at an angle of incidence of 45' (filled triangle for Hf recoils and open triangle for 0 recoils); a mixed beam of Hf atoms, with an energy of 44 eV and a density of 2.84 x l O I 4 atoms/cm2 s, and 0 atoms, with an energy of 35 eV and density of 5.68 x l0l4 atoms/cm2 s (filled squares for Hf recoils and open squares for 0 recoils).

clearly show that the effect of sputtered atoms bombardment is comparable to that of the direct ion bombardment, especially in the region close to the surface.

A trend can be observed in the packing density values with the growth rate. The packing densities were calculated from the refractive index values by the linear interpolation method by Kinosita and Nishibori reported by H. K. Pulker [26]. The extinction coefficient values were quite similar for the different samples, and low enough (- 6 x lop3 at 550 nm) to be reasonably assume that the films were not substoichiometric. Consequently, the refractive index values were mainly influenced by the packing density 127, 281 rather than by the metal atom concentration. For both ion-assisted and no ion-assisted sputtered samples, the packing density increases by increasing the sputtering growth rate, and it tends to saturate at the higher growth rate values, as seen in Fig. 8.

The ion beam sputtering seems to be a promising technique for thin film deposition on plastics. First of all, it allows the substrate temperature to be kept low enough to minimize thermal stress at the film/substrate interface. Both thermal and intrinsic stresses are summed up to give the total residual stress in the coating. In the case of hafnia thin films deposited on polycarbonate substrate, the thermal stress is compressive, because the expansion coefficient of polymers is greater than that of hard oxides. Thus, the compressive thermal stress could, in principle, be balanced if the film had an intrinsic stress of opposite sign, i.e. tensile in this case. Films with high packing density usually show compressive stress. On the other hand, dense films are usually harder than porous ones, and maintain their optical properties when exposed to a humid environment.

The effect of the growth rate on the residual stress is shown in Fig. 9. The samples were deposited on polycarbonate substrate with and without ion assistance. The film with no ion assistance shows an increase in the compressive residual stress level as

44

0'96:

.- ? 0.92- c

S. Scaglione et al.

A A-A 0

W

1

Figure 8. Packing density values versus sputtering growth rate of Hf02 films deposited on polycarbonate substrates without ion assistance (squares): with 150 eV (circles) and 300 eV (triangles) Xe+ ion assistance.

-2.0 I ' I ~ I ~ I ~ I , , ~

0.04 0.05 0.06 0.07 0.08 0.09 0.10 0

sputtering growth rate (nm/s) 1

Figure 9. Residual stress versus sputtering growth rate for Hf02 films deposited on polycarbonate substrates without (squares) and with Xe+ ion assistance at 150 eV (circles) and 300 eV (triangles) beam energies.

the growth rate increases. This result could be the consequence of the intrinsic characteristic of the ion beam sputtering process, where the sputtered atoms arrive at the substrate with energy of a few tenths of an electron volt.

As concerning the samples grown with ion assistance, the effect of the bombard- ing ions on the residual stress seems to be more complex than a simple densification mechanism. Indeed, densification phenomenon occurs, but some other mechanisms have to be involved in order to account for the stress behavior. As reported by some of the authors [29], the ion assisted deposited films of HfOz on a glass substrate change their stress from compressive to tensile as the momentum transfer parameter decreases (Fig. 6). The limit of zero momentum transfer parameter (no ion assis- tance) showed a tensile stress. On the contrary, in the case of ion beam sputtering,

Injuence of the ion bombardment on the stress in thiizjilms 45

the stress is compressive for no ion-assisted coatings. This can be due to the fact that the energy of the depositing atoms is higher in the case of sputtering than that of evaporation, thus producing a higher packing density. From these considerations, it is quite evident that in the case of hafnia thin films deposited by ion beam sput- tering and bombarded during growth, the ion assistance produces relaxation of the compressive stress (Fig. 9). This fact suggests the possibility of growing Hf02 films with zero residual stress but high packing density on polymeric substrates.

3. ZnSe LAYER

3.1. Momentum transfer parameter in RF sputtering

The systematic investigations on the intrinsic residual stress in magnetron sputtered metal films were performed by Hoffmann and Thornton [30] to find out the dependence of the stress reversal from tensile to compressive on the operational parameters such as the ratio between the target atomic mass and the sputtering gas mass, cathode shape, orientation and distance of the substrate with respect to the target, or plasma power density. The mechanism proposed by Hoffmann and Thornton in order to explain the compressive stresses in magnetron sputtered coatings involves the target elastic rebound of energetic neutral particles which bombard the growing film (Fig. 10). The ions, accelerated in the cathode sheath at energies approaching the discharge voltage, tend to be neutralized just prior to their impact on the cathode surface. A fraction of these incident species is reflected from the cathode. These energetic species, largely neutral, escape from the cathode and a portion of them reaches the substrate and impacts with energy decreased by the collision with the working gas. If the arrival energy is above the appropriate threshold values, the film atoms recoil beneath the surface inducing a compressive strain. According to this model, the energy and the flux of the energetic

Figure 10. Layout of the RF sputtering process.

46 S. Scaglione et a1

particles bombarding the growing film determine the nature and the magnitude of the stress. Since Hoffmann and Thornton studies, a number of researchers have shown that the results obtained for magnetron sputtered thin films are generic to any PVD process involving super-thermal particles striking the film [31]. In PVD processes the residual stress and microstructure properties of thin films are strongly affected by bombardment with energetic particles during the growth process. This is experimentally verified especially for ion-assisted deposition processes, where determination of the energy and the flux of the bombarding ionic species is feasible and reliable. For plasma based sputtering processes the stress variation is generally studied as a function of a process parameter such as pressure or substrate bias, since the energy and the flux of the particles striking the substrate are difficult to quantify. According to these findings, equation (1) (normalized to the incoming atom flux (a)) can be rearranged as [32]:

J P = --JzM,1/E,

U (3)

where J (atom/cm2 + s) is the energetic particles flux per unit area, E is the energy of the bombarding particles of mass Mi and y is the energy transfer factor which is given by [33]:

(4) 4Mi Mat

(Mi + at>^ ’ Y =

where Ma, is the mass of the atoms being bombarded by the energetic particles. On the other hand, in the case of RF sputtering processes the deposition parame-

ters (cathode voltage, gas pressure and radio frequency power) are not independent of each other. Accordingly, a knowledge of the exact values of J and E in equa- tion (3) is quite complicated. Nevertheless, an estimation of the dependence of P on the measurable parameters can be done. The energetic particles flux J can be estimated by considering the maximum flux Jm,, which, assuming the electron den- sity to be zero within the cathode and the particles in the sheath to be ions, is given by the Child law [34]:

where V is the voltage across the sheath, I is the sheath width, 60 the vacuum dielectric constant, e the charge of electron and Mi the ion mass (Fig. 10).

Both experimental and theoretical investigations [35] show that the reflection coefficient R (the ratio between the neutrals reflected by the target and the neutrals impinging on the target) depends weakly on the reduced energy and strongly on the mass ratio. In this case R = 0.20 is determined using the data reported in Ref. [36]. The energy of the reflected particles is assumed to be a maximum at the energy value eV( l - y ) . The reflected particles reach the growing film through the plasma in which the atoms move at a comparatively negligible velocity. The mean path length is inversely proportional to the chamber pressure and the momentum transfer cross section for atom-atom collision. The flux of neutral atoms reaching the substrate

Influence of the ion bombardment on the stress in thinjilms

2.48 l ~ l ~ l ~ l ~ l ~ , ~ l

2.46

Q, 2.38- > p 2.36- 0 2 2.34- 5 L 2.321

2.30-

41

I

3" / k10.6 pm

/f

Figure 11. Refractive index at A = 10.6 p m of RF sputtered ZnSe films deposited at different values of 8.

decreases exponentially with target-to-substrate distance d . The fraction of particles that can travel along a distance x without a collision will asymptotically approach zero as x increases, rather than suddenly drop to zero at the mean path length A. Due to these assumptions, the momentum transfer parameter P should be proportional to /3 = 0.2 x V 2 exp(-d/A). All the properties of the investigated samples are reported as function of j3, which will be expressed in arbitrary units. As shown in Fig. 11, the refractive index n at a wavelength of 10.6 p m of RF sputtered ZnSe thin films is significantly affected by the momentum transfer parameter. For /3 values lower than 29 a.u., n is lower than the bulk value and it increases with /3; whereas for /3 > 29 a.u., n is almost a constant close to that of the ZnSe bulk (2.45). Since the packing density of the films is generally related to the refractive index, the samples deposited at j3 < 29 a.u. exhibit poor compactness. The value of /3 2 29 a.u. was determined from Fig. 11 using linear extrapolations.

The X-ray diffraction patterns of ZnSe films exhibit many peaks and the reflection at high angles can be monitored to observe the evolution of the residual strain with the parameter j3. The larger amplifications of the variations in the d-spacings at high angles justify this choice. The X-ray diffraction patterns depicted in Fig. 12 show the 4 3 3 spacing for a set of ZnSe samples deposited by RF sputtering technique at different /3 values. The 4 3 3 spacing was calculated from the (333) reflection maximum position using Bragg's law; whereas the strain component cZz = ( 4 3 3 - do)/do was calculated by taking for do the spacing d333 of a c-ZnSe powder which should be stress free. In Fig. 13, the E,, strain component (changed in sign in order to have the same sign as the in-plane stress component [37]) is shown for different values of the parameter j3. Tensile in-plane stress (positive -czz values) were found at small values of /3 and an inversion of the sign was observed at /3 2' 29. For /3 > 29 the stress becomes compressive (negative -czz values). For /3 ranging from 55 to 90, the strain value was not measured because of the coating delamination, probably due to very high compressive stress.

48 S. Scaglione et al.

20 40 60 80 100 1 LO 28 (degrees)

Figure 12. X-ray diffraction patterns of RF sputtered ZnSe films grown at different values of ,6.

0.002 0.0°41 i\ wz 0.000 i l i

-0.004 compressive strain

-0.006! I I I . I I I I I . I I I I 0 20 40 60 80 100 120

P(a. u .) Figure 13. Plot of the the figure, the rectangle indicates the range of ,63 values at which the film delamination occurs.

strain component of RF sputtered ZnSe thin films as a function of /3. In

4. SUMMARY

This work was mainly concerned with the study of ZnSe and Hf02 coatings deposited using different physical vapour deposition techniques. All the films were either ion beam, electron beam or RF sputtered. The effect of ion beam assistance during film deposition was investigated. We used a combination of optical spectroscopy, residual stress measurements, and X-ray diffraction patterns to characterize the samples. Our results indicate that the momentum transferred to thin growing film influences the microstructural evolution of thin film, its optical and structural properties, as well as the level of intrinsic stress and hence film- to-substrate adhesion. The momentum transfer parameter P was used for ion

Influence of the ion bombardment on the stress in thinfilms 49

beam and electron beam sputtered films, whereas a parameter /3 proportional to the momentum transferred was used for RF sputtered films.

The hafnium oxide layer deposited by IAD shows that an almost linear correlation exists between P and the refractive index and as well as between P and the sign of the residual stress. The tensile to compressive stress transition in Hf02 thin film produced by e-gun evaporation and xenon ion beam assistance is not an abrupt transition from a random phase to an oriented one. The random oriented crystalline structure is progressively amorphized by increasing the energy delivered to the growing film. Additional energy allows the formation of an oriented phase with compressive stress and high density.

For the same hafnia films. increasing the growth rate produces densification effect similar to that caused by ion assistance. This effect could be tentatively explained by envisioning that both Hf and 0 atoms arriving at the substrate surface from the target have enough energy to produce recoils into the depositing film. The residual stress in no ion-assisted film samples has been found to be always compressive and strictly related to the film density. In the case of ion assisted samples the residual stress changes from tensile to compressive, while the packing density remains quite high. This fact indicates that some relaxation mechanism or structural modification is involved during ion assistance, in addition to simple densification phenomenon.

On the other hand, the ZnSe thin film, produced by RF sputtering technique, grew as the ZnSe cubic phase. At low ,8 values, the coatings exhibited a well-defined [ 11 11 texture, a refractive index typical of a poorly compacted microstructure and were found in a in-plane tensile stress state. At ,8 2 29, we found a sign inversion in the strain component and a less oriented c-ZnSe phase. For /3 > 29, the coatings were in-plane compressive stress state, exhibited [ 11 11 texture and refractive index close to the bulk value.

REFERENCES

1. M. R. Wertheimer, L. Martinu and E. M. Liston in: Handbook of Thin Film Process Technology,

2. U. A. Handge, I. M. Sokolov and A. Blumen, Phys. Rev. B 59, 8541 (1999). 3. J. Y. Robic, H. Leplan, Y. Pauleau and B. Rafin, Thin Solid Films 290-291, 34 (1996). 4. S. Kumar, P. N. Dixit, D. Sarangi and R. Bhattacharyya, J. Appl. Phys. 85, 3866 (1999). 5. S. J. Bull, A. M. Jones and A. R. McCabe, Surf: Coat. Technol. 54/55, 173 (1992). 6. P. J. Martin and R. P. Netterfield, Thin Solid Films 199, 351-358 (1991). 7. S. Scaglione, D. Flori and G. Emiliani, Appl. Sur jke Sci. 43. 224-227 (1989). 8. D. J. Kester and R. Messier, J. Appl. Phys. 72, 504 (1992). 9. M. Alvisi, S. Scaglione, S. Martelli, A. Rizzo and L. Vasanelli, Thin Solid Films 354, 19-23

D. A. Glocker and S. I. Shah (Eds). IOP, Philadelphia (1996).

(1999). 10. H. Dong and T. Bell, Su$ Coat. Technol. 111, 29 (1999). 11. H. C. Ong, R. P. H. Chang, N. Baker and W. C. Oliver, Surf: Coat. Technol. 89, 38 (1997). 12. E Sarto, M. Alvisi, E. Melissano, A. Rizzo, S. Scaglione and L. Vasanelli, Thin Solid Films 346,

13. D. Rats. V. Hajek and L. Martinu, Thin Solid Films 340, 33 (1999). 14. H. R. Dobler, Appl. Opt. 28, 2698 (1989).

196 (1999).

50 S. Scaglione et al.

15. H. E. Reedy and G. L. Herrit, Proc. SPIE 1020, 180 (1988). 16. R. K. Waits, in: Thin Film Processes, J. L. Vossen and W. Kern (Eds), p. 131. Academic Press,

17. L. Caneve, S. Scaglione, D. Flori, M. C. Cesile and S. Martelli, J. Adhesion Sci. Technol. 10,

18. J. D. Targove and H. A. MacLeod, Appl. Optics 27, 3779 (1988). 19. K. H. Muller, Appl. Phys. A 40, 209 (1986). 20. K. H. Muller, J. Appl. Phys. 59, 2803 (1986). 21. S. Scaglione, D. Flori and L. Caneve. Proc. SPIE 2253, 1243 (1994). 22. P. J. Martin, R. P. Netterfield and W. G. Sainty, J. Appl. P h p . 55, 235 (1984). 23. Joint Committee on Powder Diffraction Standards, Card 43- 1017. 24. P. Sigmund, in: Sputtering by Particle Bombardment I . R. Behrisch (Ed.), p. 9. Springer-Verlag,

25. F. Sarto, L. H. AbuHassan, M. Alvisi, L. Caneve and S. Scaglione, submitted to Thin Solid Films. 26. H. K. Pulker, Coatings on Glass. Elsevier, Amsterdam (1984). 27. L. Caneve, A. Rizzo, F. Saxto and S. Scaglione, in: Future Directions in Thin Film Science and

Technology, J. M. Marshall, N. Kirov, A. Vavrek and J. M. Maud (Eds), pp. 192- 199. World Scientific Publishing, Singapore (1996).

28. P. J. Martin and R. P. Netterfield, in: Handbook ofIon Beam Processing Technology, J. J. Cuomo, S. M. Rossnagel and H. R. Kaufman (Eds). p. 373. Noyes Publications. Peak Ridge, New Jersey (1989).

Orlando, Florida (1978).

1333 (1996).

Berlin (198 1).

29. F. Sarto, M. Alvisi and S. Scaglione, Proc. SPIE 3836, 35 (1999). 30. D. W. Hoffmann and J. A. Thornton, J. Vac. Sci. Technol. 20, 355 (1982). 31. H. Windischmann, J. Vac. Sci. Technol. A 9, 2431 (1991). 32. P. B. Mirkarimi, K. F. McCarty and D. L. Medlin, Mater: Sci. Eng. R21,47 (1997). 33. P. Sigmund, Phys. Rev. 184, 383 (1969). 34. H. R. Kaufman, J. Vac. Sci. Technol. 15, 272 (1978). 35. J. Bottiger, J. A. Davies, P. Sigmund and K. B. Winterbon, Radiation EfSects 11, 69 (1971). 36. E. Franke, H. Neumann, M. Zeuner, W. Frank and F. Bigl, Surf: Coat. Technol. 97,90 (1997). 37. J. J. Cuomo, J. M. Harper, C. R. Guarnieri, D. S. Yee, L. J. Attanasio, J. Angilello, C. T. Wu and

R. H. Hammond, J. Vac. Sci. Technol. 20, 349 (1982).

Adhesion Aspects of Thin Films, Vol. 1, pp. 5 1 -65 Ed. K. L. Mittal 62 VSP 2001

Delamination of thin hard coatings induced by combined residual stress and topography

URBAN WIKLUND '.*, STURE HOGMARK' and JENS GUNNARS * ' Department of Materials Science, Uppsala Universiy, SE-751 21 Uppsala, Sweden

Det Norske Veritas, PO. Box 49306, SE-100 29 Stockholm, Sweden

Abstract-Most of today's thin ceramic coatings generated by PVD or CVD techniques are in a residual stress state. This state has a number of origins and will be of either tensile or compressive nature. On topographic substrates which are not perfectly smooth, flat and infinitely large, tensile stresses are always induced across the coating-substrate interface. These topography induced "lift off" stresses, and the risk of delamination they represent, have been predicted and investigated using finite element calculations. Surface features such as edges and grooves in the substrate, interface roughness, as well as coating termination at cracks or pores in the coating have been studied.

The general experience that highly stressed PVD coatings have to be sufficiently thin so as not to spontaneously detach from the substrate has been both experimentally and numerically verified. For instance, if the thickness of a coating surrounding an edge is increased, interfacial "lift off" stresses comparable in magnitude to the residual stress will develop and may cause coating detachment.

It is also shown that high stress develops across the interface of a coating on a flat but rough substrate. The stress level is independent of the coating thickness once the thickness exceeds about three times the amplitude of the interface roughness. It is also shown that the stress across the interface is very sensitive to the local inclination of the interface profile relative to the macroscopically flat interface plane. At such positions the interfacial stress scales approximately linearly with the maximum inclination of the surface profile.

Various highly stressed ceramic coatings deposited on surface irregularities such as substrate corners and grinding grooves are used to experimentally confirm the predictions and to demonstrate hazardous combinations of stress and topography.

Kejwords: Coating: delamination: surface topography; surface roughness: non-planar interface; residual stress: interface stress.

1. INTRODUCTION

Coatings are used to improve a vast number of properties of surfaces. These properties may be mechanical, thermal, chemical, optical or magnetic. Almost

*To whom correspondence should be addressed. Telephone: +46 18 4713092; Fax: +46 18 4713572; E-mail: urban.wiklund@angstrom.uu.se

52 U. Wikluad et a1

any coating is exposed to residual stresses which normally are a consequence of the manufacturing process, e.g. due to the growth mechanism and a thermal mismatch between the substrate and coating material. Significant stresses may also build up when a coated component is cooled from the deposition temperature or possibly heated to its service temperature. For ceramic coatings on metal substrates the residual stresses in commercial coatings made by physical vapour deposition (PVD) are in the range of 1-3 GPa [ 13. Depending on the material combination, the stresses in chemically vapour deposited (CVD) coatings may be either of compressive or tensile nature [2].

Delamination and fracture of coatings are often encountered in mechanical appli- cations and on components exposed to thermal cycling. Furthermore, the interface between the substrate and the coating is usually the weakest part of a coating com- posite. Provided the substrate surface has been cleaned from contamination, delam- ination and cracking are strongly promoted by residual stresses in the coating and this is further worsened by any irregularities in the substrate surface.

In theory, the hardness and wear resistance of most tribological PVD coatings would be improved if they were manufactured with high compressive residual stresses. This would also be the case in practice provided the stress in the coating is uniformly distributed and the surface of the substrate is perfectly flat and smooth and infinitely large. However, no engineering component can display this geometry and, as a consequence, severe stresses may develop at critical positions.

Most tribological coatings are applied to relatively rough surfaces and it is a common experience that decohesion and detachment of engineering ceramic coatings is associated with high levels of residual stress. Coating detachment may occur already during the deposition process when the coating reaches a critical thickness. If surviving the growth phase, spallation may be caused by stresses induced during the cooling phase. In addition, coated components are usually exposed to some kind of external loading which imposes additional stresses which can trigger detachment. Due to the limited thicknesses of most PVD and CVD coatings these phenomena are not easily quantified experimentally.

Finite element calculations have been used in this work to investigate these phe- nomena and elucidate the stress concentrations which evolve due to the interaction between the residual stress in the coating and the substrate geometry and topogra- phy. The calculated results are also correlated to selected experiments and real case situations for a number of different coating systems.

2. COATING SYSTEMS

Ceramic coatings are widely used to enhance the tribological performance of mechanical components. Metal cutting tools is an application area where coatings have become established since the late seventies. Tic, TIN and CrN are three frequently used coatings but also other coatings, such as diamond and TiB2 have emerged recently. These five coatings together with relevant substrate materials (tool steel, high speed steel (HSS), cemented carbide (CC) and austenitic stainless

Delamination of thin hard coatings 53

Table 1. Coating composites used as demonstrators and their typical properties

Coating/ substrate Ratio of Young’s moduli

~~

CrN on tool steel TiN on HSS TiN on CC T ic on CC Diamond on CC TiB2 on stainless steel

3501220 = 1.6 5001220 = 2.3 500/620 = 0.8 4501620 = 0.7

10001620 = 1.6 6401210 = 3.0

Deposition process

PVD PVD PVD CVD CVD PVD

Characteristic residual stress (GPa)

Coating thickness (pm)

-1 -4 -4 +0.2 -2 - 10

4 4 4 4

10 2

steel (SS)) represent systems under both compressive and tensile residual stresses of various levels. Typical properties of the coating systems used as demonstrators in the models and experiments to follow are given in Table 1.

3. MODELS

The residual stress in a homogeneous coating does not induce any shear or normal stresses across the interface provided the substrate is perfectly smooth, flat and of infinite extent, see Fig. la . In this ideal case, a biaxial stress state evolves in the coating where only normal stress components parallel to the interface are present. We define this stress as the characteristic residual stress a*. The level of a * is characteristic for the specific material combination, manufacturing process and service temperature [3], cf. Table 1.

However, the topography of engineering components does not generally meet the prerequisites above. Most substrates display some kind of surface irregularities such as ridges (Fig. lb) and grooves (Fig. IC) and nearly all have edges (Fig. Id). In addition, the manufacturing often introduces flaws such as pores (Fig. le). In Fig. 1 interfacial stresses induced by a homogeneous compressive residual stress a* in the coating are indicated by arrows.

To predict coating detachment, the shear stress T , , ~ and the normal stress a,, across the interface, see Fig. 2, are examined. Selection of these two stresses is motivated by the observation that the interface is generally weaker than either the coating or the substrate. Thus, tnt and a,, across the interface will be more important than the maximum principal stress a1 which is normally used to predict initiation of cohesive failure. Additional support for focusing on these stresses comes from the fact that, in the case of coating systems with residual stress, the highest 01 usually coincides with the maximum of the interfacial stress on.

The four cases of Figs lb - l e are analysed by considering three fundamental geometries as defined by Fig. 2 [4, 51. The stress state is analysed by finite element calculations. Plane deformation conditions and homogeneous isotropic elastic materials are assumed. The elastic properties are defined by Young’s modulus, E , , and Poisson’s ratio I/, (i = c. s), where the subscript refer to the coating and the substrate, respectively. The coating has a thickness, r,, which is very small

54 U. Wiklund et al.

(b) (C) (4 (e)

Figure 1. (a) Illustration of the ideal case of an infinite coating with a uniform compressive stress applied to a perfectly flat and smooth substrate. (b-e) Ridges, grooves, and corners of the substrate and coating pores, respectively, are features in real coating composites where compressive residual stress will induce tensile and shear stresses at the interface.

(b)

Figure 2. Models and definitions used in the numerical calculations. (a) A periodic rough interface. (b) A coating on a curved 90" substrate edge. (c) A terminated coating (coating edge).

compared to the lateral dimensions of the composite. In the models, residual stress corresponding to o* is imposed by a thermal strain. Finite element analyses are performed using the code Adina [6]. Element meshes are designed to resolve the stress state at the interface [4, 51.

The rough interface of Fig. 2a represents grooves and ridges commonly found in the surfaces of engineering components. The interface is idealised as a periodic profile described by circular arcs defined by the radii R, and R, and the angle y . R, represents the size of the grooves and R, the size of the ridges. The ratio R,/R, describes the asymmetry of the interface and y represents the maximum local inclination of the interface relative to the interface average plane. When the radii are small compared to the coating thickness, the model represents the important case of small amplitude surface roughness which is almost invariably present on engineering surfaces.

Delamination of thin hard coatings 5 5

With the assumption of a periodical interface and using symmetry and periodicity boundary conditions, the calculations can be restricted to a periodic unit cell, see Fig. 2a. Stress distributions for non-planar interfaces are thoroughly analysed in [SI. Discussions on the influence of substrate roughness and curved interfaces are found in [7-101.

Figure 2b shows the model used for analysing the case of a coating deposited on a 90" substrate edge with radius R. Coatings deposited on such geometries are commonly found to be very sensitive to detachment.

Once delamination has started, the geometry of Fig. 2c is representative for the remaining part of the coating. Such a free coating edge, which can also be used to model the situation at a pore in the coating, has been studied in 14, 11, 121.

4. NUMERICAL RESULTS

The stress fields for the geometries in Fig. 2 are analysed for a wide range of coating thicknesses and profile radii for the materials in Table 1. Poisson's ratios used are u, = 0.2 and us = 0.3, except for the CC substrate where us = 0.2 was used. As mentioned above, the maximum of the interfacial stress a,, and the principal stress a1 normally coincide.

4. I . Rough interfaces

The calculated levels of the maximum interfacial stress a, along a non-planar interface of a flat substrate are shown in Fig. 3. A symmetric interface profile ( R , = R, = R ) and y = 45" are assumed. The stress level is of the same order of magnitude as a*, unless R / t is very large. The maximum tensile stress an is generated at the summit of the profiles in coatings with compressive residual stress, e.g. TiN on HSS. In coatings with tensile residual stress, represented by T i c on CC, maximum tensile stress a,, develops in the bottom of the valleys of the profiles.

When the interface roughness is small compared to the coating thickness, the maximum normal stress across the interface can be approximated by 151

The inclination angle y is given in degrees. This approximation is valid for R < OSt,, moderate stiffness differences and y less than 45". It is also shown in [ 5 ] that for interface radii larger than the coating thickness ( R =- 3,) the maximum a,, is independent of y and approximately given by

For each coating system in Fig. 3, a radius of curvature can be found for which the interfacial stress reaches its highest value. For smaller radii, the stress levels off at about 75% of the highest value. The constant level is given by equation (1).

56 U. Wiklund et al.

rnax on [GPal

0.1

0.6 max q, [GPa] 0.5

0.4

0.3

0.2

0.1

0

Tic I CC a' = 0 2 GPa

1, = 20 pm -

CrN I Tool steel a. = -I GPa

t 1, = 20 pm

f, = 20 pm 1.0 -

0.2 - u' =-4GPa

1 IO 100

2.0

1.6

1 2

0.8

0.4

0

1.2 max on [GPa] 1.0

0.8

0.6

0.4

0.2

i I diamond I CC u* = -2.0 GPa

2.4 max on [GPa] 2.0

1.6

1.2

0.8

0.4

10

I I 10 100 10 100

(e) R [ P ~ I (0 R [ ~ m l

Figure 3. Maximum tensile stress un across the interface of a rough substrate vs. radius of curvature of the interface profile. The stress levels are shown for different coating thicknesses for six different coating systems assuming a periodical interface with R, = R, = R and y = 45'.

For larger radii, the stress level is reduced monotonously towards zero according to equation (2).

Figure 4 shows the distribution of the maximum principal stress 01 in the calculation domain, cf These results represent an extremely uneven interface profile and a coating thickness set to half the radius of the interface profile. The arrows show the direction of 01 at the central point of each arrow. For compressive residual stress, see Fig. 4a representing TiB2 on SS, the maximum value of 01 is found at the summit of the profile. The direction of 01 implies that an interfacial crack is likely to nucleate at this point. In case the residual stress is

Fig. 2a.

Delamination of thin hard coatings 51

Figure 4. Distributions of the principal stress a1 in a periodic unit of a coated substrate. The maximum inclination angle is 45" and the profile radius equal to 2t,. The patterns represent (a) a high compressive residual stress (TiB2/SS) and (b) a small tensile residual stress (TiC/CC).

58 U. Wiklund et al.

Figure 5. Interfacial stresses for a coating with compressive residual stress (diamond on CC) deposited around a substrate edge. (a) Distribution of interface stresses ( R = tc = 10 pm). (b) Maximum stresses at the edge as a function of R/t,.

tensile, e.g. TIC on CC in Fig. 4b, the maximum value of C T ~ is found at the free surface in the bottom of the profile, parallel to the interface. The arrows parallel to the interface and positioned at the interface show the stresses in the coating. This suggests that for this case, cracks are likely to initiate perpendicular to the interface. However, the normal stress across the interface, on, also has its maximum in the bottom of the valley and the residual stress may be relieved by interfacial cracks if the strength of the interface, in this case, is less than one fourth the cohesive strength of the coating material.

4.2. Substrate corner

The distributions of residual normal interfacial stress on and shear stress tnnt around a coated substrate edge are symmetric and asymmetric, respectively. The peak value of C T ~ is found at the tip of the edge, see Fig. 5a which shows the case of a 10 p m diamond coating on a CC substrate edge with 10 p m radius. Note that the maximum tnt occurs at the transitions from curved to planar interface, cJ: Fig. 2b. Increasing the radius for a specific coating thickness has the effect of reducing the maximum interfacial stress as shown in Fig. 5b.

Figure 5 is representative of coatings with compressive residual stress. A tensile residual stress would reverse the curves.

4.3. Coating edge

Figure 6 presents the stress distribution at a terminated coating (coating edge), cJ Fig. 2c, with compressive residual stress CT*. The stresses shown are the normal stress across the interface oz, the shear stress along the interface tXz, and the normal stress in the substrate parallel to the interface axs). All stress components are found to become singular at the edge (x = 0). An increase in coating thickness has the effect of distributing the stress over a larger area as the high stresses are

Delamination of thin hard coatings 59

[GPaI 0.2

0

-0 2

-0.4

-0 6 0 10 20 30 40 50

[GPaI 0.8

0.4

0

-0.4

-0.8

0 IO 20 30 40 50

Figure 6. Distribution of interfacial stresses at a terminated coating (diamond on CC). (a) Normal stress across the interface oZ and (b) interfacial shear stress txi and normal stress CT$) parallel to the interface in the substrate.

approximately confined to the region within 5t, from the edge. For large distances from the edge az, a~:’) and txc all approach zero (and a~r approaches a*).

As in the previous case, all stresses are reversed if a tensile residual stress is considered instead of the compressive one. In particular, for this case the high positive a, at the corner will promote initiation of coating delamination.

5. EXPERIMENTAL RESULTS AND DISCUSSION

5.1. Rough interfaces

All engineering components have surfaces that are rough on some scale. Applying a residually stressed coating will induce interfacial “lift off” stresses, on, as demonstrated by Fig. 3. If the coating thickness is much larger than the roughness of the interface (tc > 2 R ) an interfacial stress of the order of 15 to 30% of o* is achieved, depending on y . For a sufficiently smooth substrate (or sufficiently thick coating) the stress level is independent of the coating thickness. Decreasing the maximum inclination y , e.g. by polishing, will further reduce the stress.

TiB2 coatings with different thicknesses were deposited on stainless steel to demonstrate the effects of high residual stress in combination with substrate topog- raphy. The substrates had been given a controlled topography of well-defined ridges by grooving and polishing. For this highly stressed system, spontaneous spallation occurred at the point of minimum convex radius, in this case approximately 10 pm, for coatings 2 p m and thicker, see Fig 7. Fracture was initiated at, or close to, the interface, cf. Fig. 4a. In crosssection, Fig. 7b, the crack can be seen to initially fol- low the interface and then deflect away from it. This deflection from the interface suggests that the interface is quite strong. As expected from Fig. 4a, no cracks were visible in the valleys of the interface profile.

60 U. Wikh id et al.

Figure 7. Illustrative damages in a TiB2 coating with high compressive residual stress (10 GPa) deposited on a grooved stainless steel substrate. (a) Spallation along the edge of a groove. Note especially the exposed substrate. (b) Crosssection showing that the initiation of the cracks is close to the point of minimum convex curvature of the interface.

Delarniriation of thin hard coatings 61

Figure 8. (a) Extensive damage to a TiN coating on a ground HSS substrate induced by scratching. (b) Cracking and spallation of a TiN coating associated with the turning texture on the side of a HSS punch tool.

62 U. Wiklund et al.

When a coated component is placed in service, external loading is likely to be superimposed on the residual stress. This may lead to nucleation of extensive delamination on a coated component with otherwise subcritical stresses, see Fig. 8. Note how the scratch in Fig. 8a initiates delamination at a site of high roughness. Delamination then spreads spontaneously far away from the scratch only driven by the residual stress. Similar craclung and coating delamination is seen at the punch tool in Fig. 8b.

5.2. Substrate corner

During the first attempts to deposit diamond on CC metal cutting inserts, coating delamination at the edge of the tool was frequently observed. There are mainly two contributing reasons for this delamination. The first one is the weakened substrate material near the surface. This is a consequence of the removal of the cobalt binder phase, which is performed by etching. Such a removal is a prerequisite for efficient diamond nucleation and good adhesion. The second contributing factor is the thermal mismatch between the coating and the substrate, i. e. high compressive residual stresses are incorporated in the coating. Together, these facts promote nucleation of interfacial cracks at the cutting edge, primarily due to the normal stress, a,,, see Figs 5a and 9. Spontaneous coating detachment was often observed for sharp edges and thick coatings, which is in agreement with Fig. 5b.

The most common way to reduce the risk of delamination for this geometry is to restrict the coating thickness to small values in comparison to the edge radius or, if possible, to increase the edge radius.

Figure 9. Detachment of a diamond coating on a CC metal cutting tool, initiated at the edge by the combination of high compressive residual stress and a sharp edge geometry.

Delamination of thin hard coatings 63

(b) Figure 10. (a) A coating with high compressive residual stress (PVD TiB2 deposited on SS) containing a pin-hole down to the substrate. (b) Cross section of the scratch-like coating damage in (a) revealing extensive substrate yield.

64 U. Wiklund et al.

5.3. Coating edge

If a part of the coating is removed, e.g. due to mechanisms described in Sections 5.1 and 5.2, the new situation can be described by Fig. 2c. However, the geometry in Fig. 2c can also represent different types of coating defects such as pin-holes or terminated cracks perpendicular to the interface.

Stress concentrations at the free edge of a pin-hole in a coating with high compressive residual stress may initiate extensive cracking, see Figs 6 and 10. The crosssection shown in Fig. 10b reveals that the stress in the TiBz coating is high enough to cause extensive yielding of the substrate material. This results in a peculiar type of substrate yielding and coating cracking. The compressive stress is locally relieved by shear in the substrate material and simultaneous fragmentation of the coating. Once initiated at the pin-hole, this damage spreads spontaneously over the surface.

6. CONCLUSIONS

With the aim to predict coating failure by delamination we have used the finite element technique to calculate stresses at the interface in residually stressed coatings. The analysis provides models for interface stresses induced by interface roughness and substrate geometry. Using model experiments and observations from real case situations, the numerical results were qualitatively confirmed.

Important conclusions for coating manufacturers as well as end-users include:

0 Normal stress across the interface of a magnitude comparable to that of the resid- ual stress o* may be induced at corners, edges and topographical irregularities of the substrate surface.

0 For a coating system with a rough interface, the stresses reach a maximum when the coating thickness is approximately equal to the amplitude of the interface roughness profile. A thinner coating is subject to lower stress.

0 The model shows that the local inclination of the interface profile is an important measure of the surface roughness. When the amplitude of the interface roughness is small compared to the coating thickness, the stress across the interface is primarily determined by this parameter.

0 In a coating with compressive residual stress interfacial cracks are likely to nucleate at the site of minimum convex radius of the rough interface.

0 In a coating with tensile residual stress interfacial cracks are likely to nucleate at the minimum concave radius of the rough interface. If the interface is strong, cohesive cracks may form perpendicular to the interface.

0 For a given coating thickness a substrate roughness radius exists below which the interfacial stress is constant.

Delamination of thin hard coatings 65

Acknowledgements

The financial support from the Swedish Research Council for Engineering Sciences (TFR) and the Swedish Institute (SI) is gratefully acknowledged by the authors.

REFERENCES

1. K. Holmberg and A. Matthews, in: Coatings Tribology, Tribology Series 28, D. Dawson (Ed.).

2. L. Chollet, H. Boving and H. E. Hintermann, J. Mater: Energy Sys. 6, 293 (1985). 3. J. Gunnars and U. Wiklund, Mater: S a . Eng. A (accepted). 4. J. Gunnars and A. Alahelisten, Surface Coatings Technol. 80. 303 (1996). 5. J. Gunnars, “On fracture of layered materials,” PhD Thesis no. 1997:37, Lulei University of

6. ADINA Theory and Modelling Guide. Adina R&D Inc., Watertown. MA (1992). 7. A. G. Evans and J. W. Hutchinson, Acta Metall. 37. 909 (1989). 8. H. Gao, Int. J. Solids Struct. 28, 703 (1991). 9. S. D. Akbarov and A. N. Guz’, Appl. Mech. Rev. 45, 17 (1992).

Elsevier (1994).

Technology, Sweden, ISSN: 1402-1544, paper VI (1997).

10. S. J. Lee. in: ASME Winter Annual Meeting, MD-Vol. 44, Ceramic Coatings. New Orleans

11. D. B. Bogy, J. Appl. Mech. 38, 377 (1971). 12. D. Munz and Y. Y. Yang, J. Appl. Mech. 59, 857 (1992).

(1993).

Adhesion Aspects of Thin Films, Vol. 1. pp. 61-11 Ed. K. L. Mittal 0 VSP 2001

Effects of surface treatments on the adhesion of metallic films to ceramic substrates

A. J. PEDRAZA* Department of Materials Science and Engineering, The Universi? of Tennessee, Knoxville, TN 37996-2200, USA

Abstract-Pulsed laser irradiation before or after film deposition strongly enhances the bonding between metallic films and ceramic substrates. Low-energy ion bombardment of the substrate prior to film deposition also promotes strong bonding with ceramic substrates. Irradiation by laser and ions produces in many instances metastable structures. In consistency with previous results, it is reported that laser irradiation of sapphire substrates prior to film deposition strongly enhances the copper film-sapphire bonding. The effect of the oxygen content in the irradiation atmosphere on metal/oxide adhesion bonding is analyzed with particular reference to alumina and sapphire substrates. The presence of oxygen during irradiation promotes strong bonding because it produces an intermediate compound. However, an oxygen-containing atmosphere is not a requisite for the formation of a strong bond between a metallic film and an oxide substrate. Irradiation in a reducing atmosphere in the case of laser, or at low pressure during ion bombardment, also promotes a very strong bond. The oxygen- depleted surfaces produced by these treatments have metastable structures with high reactivity that also form intermediate compounds. In all cases studied here, the presence of an intermediate or an interfacial compound appears to be the main factor responsible for the adhesion enhancement.

Keywords: Surface modification; excimer laser; adhesion: metallic films; ceramic substrates.

1. INTRODUCTION

A high degree of adhesion between a metal and a ceramic may be secured by appropriate engineering of the interface connecting them. Interfacial manipulation aimed at promoting or improving adhesion bonding can be conducted at any of the following three processing stages: 1. prior to film deposition, denoted as “surface modification” (Fig. 1 a), 2. after film deposition, also called “interface stitching” (Fig. lb), or 3. during deposition, that can be named “interface construction”.

*Phone: (865) 974-7809: Fax: (865) 974-41 15; E-mail: apedraza@utk.edu

68 A. J. Pedrazn

Substrate Im

i Film Modified Layer

Substrate 0

i Film Interfaid Compound I

Substrate (iii)

Figure la. Schematics illustrating adhesion enhancement by laser-induced surface modification of the substrate. (i) Surface modification by laser irradiation, (ii) film deposition and (iii) formation of an interfacial compound.

Film ~

Substrate

I I (ii> Substrate I

Figure lb. Schematics of laser-induced interface stitching (i) film deposition and (ii) laser irradiation.

Adhesion of metallic films to ceramic substrates 69

Bombardment using photon or ion beams are two means employed to achieve these effects, plus thermal annealing in some instances [ l ] . For interface stitching, used for fairly thin films, the ions or photons must possess enough energy to reach the interface between the film and the substrate (Fig. lb). When lamps or other low-fluence sources of photons are used, the thickness of the film must be in the nanometer range [2]. The ionizing radiation must reach the interface and promote reactions between atoms at both sides of the interface. When high fluence photon sources, e.g. lasers, are used the electromagnetic radiation generated can be absorbed before reaching the interface, generating heat. In this instance modification of the interface takes place by thermal activation [3]. The penetration depth of ions in materials is a function of the ion energy, the ion species, and the nature of the material. Calculations of ion ranges and displacement damage can be readily made using the TRIM computer code [4].

In this paper, the use of pulsed-laser irradiation for both interface stitching and surface modifications aimed at improving the adhesion of metallic films to ceramic substrates is discussed. The generation of a damaged near-surface region by low- energy ion bombardment and its effect on metal/ ceramic bonding is also analyzed. The mechanisms of adhesion enhancement for dissimilar materials are analyzed and basic general principles are highlighted. A practical example on how these guidelines can help to improve bonding is given at the end of the paper for the case of hydroxyapatite films (a bioceramic material) deposited on germanium single crystals (a semiconductor material).

2. ADHESION ENHANCEMENT BY PULSED-LASER IRRADIATION

Very rapid heating of the near-surface region occurs when a material is exposed to nanosecond pulsed laser irradiation. Part of the incoming UV radiation is reflected at the surface and part is absorbed in the material. In metals, the reflectivity decreases from the visible wavelength to the ultraviolet (UV) range, where 10 to 40% of the light is reflected; most of the absorption takes place in the first tens of nanometers where very rapidly (- s) the absorbed energy is dissipated into heat. The very rapid heating produces thermal gradients in the film of -10' K/cm and, at high laser energy densities, intense evaporation takes place generating shock waves. Single crystal insulators, on the other hand, are mostly transparent although at very high light intensities they become opaque [5-71. In general, microstructural defects such as dislocations, external surfaces, grain and particle boundaries, microcrack and microvoid surfaces as well as impurities and point defects can strongly enhance light absorption. F centers are generated during laser irradiation of aluminum oxide [8, 91. Point defects generated during laser irradiation continuously increase the absorption of UV light until the ceramic material is melted. In this paper, we shall discuss the effects of a laser having a wavelength of 308-nm with a 42-ns full width at half-maximum pulse duration (FHWM).

70 A. J. Pedraza

2.1. Interface stitching

The adhesion of copper and nickel films to sapphire and of copper and gold films to silica was strongly enhanced by laser irradiation [ 10- 151. Metallic films 10-75 nm- thick were initially deposited on the ceramic substrates and laser irradiated. The procedure followed for enhancing bonding of metallic films to sapphire is different from that followed to enhance the bonding to silica. In the case of sapphire, very thin films 10 nm in thickness were initially deposited and laser irradiated at a fluence of 0.50 J/cm2. Thicker films were built by a sequence of depositions followed by laser irradiations. Following each laser irradiation, the specimen was exposed to air and transferred to a deposition chamber for the next step. The surface changes due to air exposure are irrelevant because the laser thermally affects approximately a depth of 1 pm, melting every time the total film thickness. After the first irradiation, because both the copper film and a very thin layer of sapphire are melted, a compound is formed at the metal/ ceramic interface. When the laser irradiation was performed in argon-4% hydrogen the compound was found to be sapphire-like. The selected area diffraction patterns of the compound show that all the a-sapphire reflections are present, including two that should not be present due to the general conditions of extinction for their space group. Structure factor calculations show that substituting copper for some of the aluminum accounts for the observed extra reflections. Also there is a slight change in the lattice parameter. When the irradiation is performed in air the diffraction pattern indicates the formation of a trirutile-like structure in the interface.

Adhesion testing was done using a scratch tester that had a balanced lever arm with a diamond stylus of 0.2 mm radius and a load platform above the stylus. During the test, the stage with the specimen had a translational motion at a constant speed of 27 pm/s. A strain gauge rigidly attached to the load platform measured the applied tangential forces. The scratch tests done on a laser treated film showed no de-adhesion from the substrate after applying a load of 400 g while an as-deposited film showed clear signs of removal at the root of the scratch at a normal load of only 200 g. The shear stresses where no de-adhesion was detected were 0.65 GPa and 0.21 GPa for the laser-treated and for the as-deposited films, respectively. The optical micrograph of Fig. 2 shows a scratch made with a 100 g normal load on a copper film deposited on sapphire, the right portion of which was irradiated (lighter region). The scratch goes from the non-irradiated portion (darker region) to the irradiated part (lighter region). Ripples indicating de-adhesion can be seen in the unirradiated region. This effect was further illustrated by the profiles of the groove taken in the non-irradiated region (left) and the irradiated part (right).

A very different mechanism is operational in the case of silicon dioxide. Under UV-laser irradiation, silica behaves very differently from aluminum oxide. Copper or gold films, 20 nm-thick, do not react with silica when they are irradiated with UV laser light at 0.5 J/cm2. Instead, a very peculiar process takes place, viz. the film is melted, breaks down into small spheres and is encapsulated inside the silica. Cross-sectional transmission electron microscopy (XTEM) showed that these films penetrated into the silica 10 to 15 nm. XTEM showed that an 80 nm-thick gold

Adhesion of metnllicjlms to ceramic substrates 71

Figure 2. Optical micrograph of a scratch made with 100 g normal load in a 1 pm-thick copper film produced by sequential film deposition followed by laser irradiation. Ripples can be seen in the unirradiated portion (left side). No ripples or evidence of peeling appears on the irradiated side (right).

film deposited on silica also clustered into small particles when irradiated in air at 1.5 J/cm2, but these particles were only partly driven into the silica. In this case, the silica did not coat the particles entirely, as was the case for the previous irradiation. In the case of the partially coated particles, a new gold film deposited on top of the irradiated system will make contact with the portions of the emerging particles produced from the initial layer. The partly exposed particles serve as mechanical anchoring for the new film deposited after laser irradiation. As the laser energy is increased there is also ablation of silica and film taking place. The result is the removal or weakening of the anchoring film. The adhesion strength, measured on copper and gold films deposited on silicon dioxide following a two-stage process, is shown in Fig. 3. The two stages were as follows: in the first step 75 nm-thick metallic films were deposited on silica substrates and irradiated at several laser fluences. After the irradiation another 75 nm-thick film was deposited on each specimen and pull tests were performed. For both, Cu and Au films, a maximum adhesion strength is observed at intermediate fluences in agreement with the TEM observations.

In summary, it was shown that the mechanisms of laser stitching depended on whether a reaction between the metallic film and the substrate had taken place. If a reaction occurs, the adhesion enhancement takes place by chemical modification of the metal/ceramic interface; if a reaction does not occur, the adhesion enhancement can take place by mechanical anchoring.

2.2. Surface modijicatioiz

When a ceramic such as alumina is laser irradiated the significant amount of heat that evolves in the near surface can eventually produce melting of a thin layer. For laser fluences higher than 1 J/cm2 a 0.3-0.5 pm-thick layer is melted in a few tens of nanosecond. Non-equilibrium phases can be formed due to the very rapid

72 A. J. Pedraza

Laser Fluence (J/cm2)

Figure 3. Adhesion strength of copper and gold films deposited on Si02. The films were prepared by first depositing 75 nm-thick films, next the couples were irradiated at laser fluences from 0.5 J/cm2 to 3 J/cm2. After irradiation another 7 5 nm-thick metallic film was deposited on the specimens.

solidification that follows the laser irradiation. Typically, only a few hundreds of nanoseconds are required for the liquid to solidify. In the sapphire, a metastable phase - y-alumina - was formed as a result of rapid solidification [16]. This y - alumina layer grew epitaxially from the unmelted sapphire substrate with the orien- tation relationship (001)y//(0001)a and [ 1 lO]y//[ 11 l ]a . Also resolidified sapphire with a polytype structure was found in similarly laser-irradiated samples [9]. An amorphous layer -0.2 pm-thick formed in the near-surface region of alumina after laser irradiation at an energy density of 1 - 1.3 J/cm2. When the irradiation was per- formed at an energy density of 1.6 J/cm2 or higher the melted region solidified with a cellular microstructure with the same orientation as the matrix. In some instances, pockets of amorphous material could be seen in samples irradiated at 3 J/cm2.

We have performed laser irradiation of alumina substrates at several laser fluences followed by deposition of copper, gold and nickel films [17-221. Deposition can be done many hours after the laser treatment. It was found that a very strong increase in adhesion was obtained if a low temperature annealing was performed after film deposition. In Figs 4a and 4b the adhesion strength, as measured by pull test, is plotted as a function of the laser fluence. The alumina was irradiated in an Ar-4% H2 atmosphere (Fig. 4a) and in air (Fig. 4b). Also in Fig. 4b, the values of the adhesion strength of gold film deposited on alumina irradiated in oxygen atmosphere are plotted. All the laser processing was performed at a pressure of 1 atm. The maximum adhesion strength that can be measured with this test is limited by the strength of the epoxy/metal interface. We have found that this limit is around 70 to 72 MPa. These experimental results show that the adhesion strength is a function of the atmosphere: oxidizing atmospheres tend to promote stronger bonding than reducing atmospheres. However, the bonding promoted by

Adhesion of metallic$lms to ceramic substrates 73

8ol

2 I:j/ 20

10 2

0 0 0 1 2 3 4 5

Laser Fluence (J/cm2)

Figure 4a. Adhesion strength of metallic films sputter-deposited on alumina substrates laser irradiated in Ar-4% H2, as a function of laser fluence.

80 7

70 a : h 0

f

.- (I) 2 20

10

0 0 1 2 3 4 5

Laser Fluence (J/cm2)

Gold- * Oxygen

+ Copper + Nickel

Figure 4b. Adhesion strength of metallic films sputter-deposited on alumina substrates laser irradiated in air, as a function of laser fluence, and adhesion strength of gold films sputter-deposited on alumina substrates laser irradiated in oxygen, as a function of laser fluence.

laser irradiation in reducing atmospheres can be very strong as is the case for copper and nickel films. The maximum strength achieved is a function of the chemical nature of the metal, as indicated by quite different adhesion strengths of sputter- deposited metals on unirradiated alumina substrates: 0.1 MPa for Au, 13 MPa for Cu, and 32 MPa for Ni. Finally the bonding also depends on the type and extent of laser-generated disorder/ damage, since for gold significant bonding enhancement is obtained only if pulsed-laser melting of the alumina has occurred.

14 A . J. Pedraza

A careful study by Auger electron spectroscopy (AES) showed that the very high adhesion strength of copper films deposited on laser irradiated A1203 was due to the formation of a transitional region. In the case of laser irradiation in air, the transitional region consists of a Cu-A1 double oxide which is the most probable cause of strong bonding. In the case of a gold film deposited on alumina laser- irradiated under oxygen, AES revealed that the Auger peaks of gold from the film and of oxygen and aluminum from the substrates shifted by - 1.5 eV, 1.6 eV, and 1.4 eV, respectively. On the other hand, no shift of the Auger peaks was found at the interface when the laser-irradiation was performed in argon-4% H2.

Adhesion enhancement between copper films and aluminum nitride is obtained as well, when aluminum nitride is irradiated prior to film deposition. The bonding behavior of copper on aluminum nitride is complicated by the fact that several phases can appear in the near-surface region after irradiation. First, aluminum oxide is present on the as-received substrates; this oxide layer decreases in thickness after irradiation. In addition, during irradiation aluminum nitride decomposes into metallic aluminum and gaseous nitrogen that is desorbed. Laser irradiation prior to deposition increases the adhesion strength of the copper film to A1N from 25 MPa to 40 MPa, whether the irradiations are performed in air or in Ar-4% H2. This adhesion enhancement takes place without any annealing after irradiation and at a laser fluence of only 0.5 J/cm2. As the laser fluence increases the adhesion strength strongly decreases reaching a value of 18 MPa for substrates irradiated in air at 2 J/cm2. The decrease in adhesion strength is even more abrupt for A1N substrates irradiated in Ar-4% H2. This decrease in adhesion strength correlates with the decomposition of A1N to form islands of metallic aluminum on the surface because the amount of metallic aluminum increases with the laser fluence.

Pulsed laser irradiation of silica prior to deposition does not improve the bonding of copper or gold films. Laser irradiation produces significant ablation of silica but it fails to produce surface modifications conducive to bonding. It is also the failure of the silica to chemically react with copper or gold films that precludes a chemical bonding during laser stitching. Conversely, other metals such as Ti do react with silica during irradiation and it is this reaction that precludes their laser encapsulation.

It is illustrative to compare interface stitching with surface modification by laser irradiation as techniques for enhancing the adhesion of metal/ceramic systems. It is clear from the pull test data that the bonding of metallic films to aluminum oxide is stronger if surface modification procedures are employed instead of laser stitching. However for the case of silica only interface stitching produces a significant increase in adhesion bonding with copper or gold films.

Finally, the thermal stability of gold films deposited on laser irradiated aluminum oxide was studied by measuring the increase in electrical resistance of the gold films after an annealing for 45 minutes at 550°C. Prior to the 550°C annealing, the 20 nm- thick gold films were deposited on irradiated aluminum oxide substrates. Surface modification was done at several fluences, under oxidizing or reducing atmospheres. The electrical resistance measures the integrity of the film after annealing. Consis-

Adhesion of metallicjlms to ceramic substrates 15

tent with the results obtained from the adhesion strength measurements, it was found that the film deposited on a substrate irradiated at 1 J/cm2 in oxygen had the highest thermal stability.

A practical example on how surface modification can enhance the adhesion of dissimilar materials is the case of films of hydroxyapatite, a bioceramic material, deposited on germanium single crystals. It was required that 100 nm-thick hydroxy- apatite films should remain attached to the Ge element upon immersion in a liquid medium containing a microbial community (saliva) for several hours. Hydroxy- apatite films sputter-deposited on as-received optically polished germanium sub- strates detached after 30 seconds of immersion in water. The detachment starts at surface imperfections and scratches. The water penetrates through these imperfec- tions, percolating between the film and substrate and very rapidly lifting the entire film. The problem is related to the presence of a native oxide layer that is water sol- uble. An intermediate silicon layer was deposited onto the as-received substrate to serve as an adhesion-promoting agent, however the problem persisted with the film adhesion showing signs of degradation after only a few minutes. A ready solution came by removing the existing native oxide layer by laser irradiation before deposit- ing the silicon layer and the hydroxyapatite. Under this condition, film stability was tested in artificial saliva for over three hours, showing no signs of delamination.

3. ADHESION ENHANCEMENT BY SURFACE MODIFICATION USING LOW-ENERGY IONS

Pre-sputtering to clean surfaces is routinely done before film deposition using low energy ions to induce desorption of contaminants. In this section we are going to discuss another type of process induced by low-energy ion bombardment. The bonding of copper or gold films to aluminum oxide can be strongly enhanced by bombarding the substrate surface, before deposition, with Ar+ ions in the range of 0.5 to 10 keV. Two studies, one performed using X-ray photoelectron spectroscopy and the other AES, show that an interfacial compound is formed at the metal/ceramic interface [23, 241. Both studies support the idea originally put forward by Baglin [25] that one of the effects of the low-energy ion bombardment of the substrate was to generate a surface environment conducive to the formation of an interfacial compound. The presence of the interfacial compound was only detected in the metakeramic couples that exhibited the stronger bonding.

In a separate set of experiments it was shown that the adhesion strength of gold films had a ten-fold increase by bombarding the substrate prior to deposition. For the AES study a film 10 nm-thick was deposited on sapphire that was bombarded with 7 keV Arf ions. The gold fildsapphire couple was sputter etched until the film was removed only at the center of the 200 p m x 200 p m crater. The metal-ceramic interface was studied taking advantage of the variable film thickness in the vicinity of the crater. The advantage of this technique is that an interface buried only a few monolayers deep can be analyzed, avoiding significant damage. A chemical shift of 1.1 eV toward lower energies was detected in the Au NVV Auger peak together

76 A . J. Pedraza

with a shift of 0.5 eV in the 0 KLL peak toward higher energies. The aluminum peak at the interface and up to - 1.5 nm showed that the aluminum environment was significantly different from that expected in aluminum oxide. Three important changes are detected in the A1 KLL peak in the interfacial region: the shift of the AI KLL peak toward metallic aluminum, the peak broadening, and the fine structure of the peak which is different from both metallic aluminum and aluminum in an environment of A1203.

These studies show that after ion bombardment the very near surface of alumina is left in a defective state with significantly less oxygen than the bulk A1203 and, probably, with an almost metallic aluminum monolayer. Thus, it is expected that the bombarded surface has a significantly higher surface free energy than the original surface. These studies further suggest that sputtered gold atoms with an incoming energy of 10 to 15 eV striking the modified surface promote rearrangements in the near-surface region. These rearrangements produce an interfacial compound generating a chemical bonding between the gold film and the bombarded substrate.

4. CONCLUSIONS

1. Laser-promoted interface stitching enhances the bonding of copper and gold films to A1203. Two adhesion enhancement mechanisms have been identified: chemical bonding (formation of an interfacial compound) and mechanical an- choring (partial film encapsulation).

2. Laser-induced surface modification of ceramic substrates promotes strong bond- ing of sputter-deposited metallic thin films. Strong bonding was experimentally found to occur always, and only, when an interfacial compound was formed.

3. Laser-induced adhesion enhancement by surface modification of a given ceramic substrate depends upon (1) the chemical nature of the metal, (2) the type and extent of surface modification, and (3) the irradiation atmosphere.

4. Not all of the surface modifications lead to adhesion enhancement, as found for A1N substrate with the formation of metallic aluminum at its laser-irradiated surface.

5 . If the causes of de-bonding are known, a practical surface modification treatment can be designed for improving the adhesion of other dissimilar materials as well. Hydroxyapatite films deposited on optically polished Ge single crystals are removed when immersed in water, the most likely cause being the presence of a water-soluble native oxide at the surface. A successful treatment to remove the native oxide prior to deposition is irradiating the substrate in a reducing atmosphere, plus using an intermediate adhesion-promoting film (silicon).

6. Ion-induced surface modification strongly enhances the bonding of Cu and Au films to A1203 and sapphire substrates. Strong bonding is attributed to the formation of an interfacial compound during film deposition.

Adhesion of metallic films to ceramic substrates 77

REFERENCES

1. A. J. Pedraza, in: Adhesion Promotion Techniques; Technological Applications. K. L. Mittal and

2. C. J. Sofield, C. J. Woods. C. Wild. J. C. Riviere and L. S. Welch. Mater: Res. SOC. Symp. Proc.

3. A. J. Pedraza and M. J. Godbole. Metall. Trans. A23, 1095 (1992). 4. J. P. Ziegler, J. P. Biersack and U. Littmark, The Stopping and Range of Ions in Solids, Vol. 1 ,

5. N. Bloembergen, IEEEJ. Quantum Electronics QE-10, 375 (1974). 6. X. A. Shen. S. C. Jones and P. Braunlich, Plzys. Rev. Lett. 62, 271 1 (1989). I. A. J. Pedraza, Nucl. Instruin. Meth. Phjs. Res. B 141, 709 (1998). 8. G. A. Shafeev, Appl. Phys. A 55, 3879 (1992). 9. G. A. Shafeev, Ad)! Mater: for Opt. and Elect. 2, 183 (1993).

A. Pizzi (Eds), pp. 245-288. Marcel Dekker. New York (1999).

25. 197 (1984).

p. 115. Pergamon Press, New York (1985).

10. J. Pedraza, M. J. Godbole, E. A. Kenik, D. H. Lowndes and J. R. Thompson. J. Vac. Sci.

11. A. J. Pedraza. M. J. Godbole, D. H. Lowndes and J. R. Thompson. J. Mater: Sci. 24. 115 (1 989). 12. M. J. Godbole, A. J. Pedraza, D. H. Lowndes and E. A. Kenik, J. Mater: Res. 4, 1202 (1989). 13. J. Godbole, A. J. Pedraza, D. H. Lowndes and J. R. Thompson, J , Mate?: Res. 7, 1004 (1992). 14. A. J. Pedraza, in: Laser Materials Processing III, J. Mazumder and K. N. Mukherjee (Eds),

15. A. J. Pedraza and M. J. Godbole, Metall. Trans. A 23, 1095 (1992). 16. S. Cao, A. J. Pedraza, D. H. Lowndes and L. F. Allard, Appl. Pkjs. Lett. 65. 2940 (1994). 17. A. J. Pedraza. R. A. Kumar and D. J. Lowndes, Appl. Phys. Lett. 66, 1065 (1995). 18. A. J. Pedraza. M. J. DeSilva, R. A. Kumar and D. H. Lowndes, J. Appl. Phys. 77,5176 (1995). 19. A. J . Pedraza. S. Cao, L. F. Allard and D. H. Lowndes, Muter: Res. SOC. Symp. Proc. 357. 53

20. J. W. Park. A. J. Pedraza, D. H. Lowndes and W. R. Allen, J. Mater: Res. 12, 3174 (1997). 21, S. Cao, A. J. Pedraza, D. H. Lowndes and L. F. Allard. J. Mater: Res. 12, 1747 (1997). 22. J. W. Park, A. J. Pedraza and D. H. Lowndes, J. Mater. Sci. 34, 1933 (1999). 23. J. E. E. Baglin, A. G. Schrott, R. D. Thompson, K. N. Tu and A. Segmuller, Nucl. Instrum. Meth.

24. J. W. Park, A. J. Pedraza and W. R. Allen, Appl. Sur5 Sci. 103, 39 (1996). 25. J. E. E. Baglin, in: Ion Bean7 Modijication of Insulators, P. Mazzoldi and J. W. Arnold (Eds).

Techno/. A 6, 1763 (1988).

p. 183. TMS, Warrendale, PA (1989).

(1995).

Phys. Res. B 19/20,782 (1987).

Ch. 15. Elsevier (1986).

Adhesion Aspects of Thin Films, Vol. 1. pp. 19-139 Ed. K. L. Mittal 0 VSP 2001

The state-of-the-art in adhesion of CVD diamond to carbide cutting inserts

MAHMOUD A. TAHER I , * , WILLIAM E SCHMIDT ', AJAY P. MALSHE ', E. J. OLES and A. INSPEKTOR2 ' Materials and Manifacturing research laborator): (MRL), Department of Mechanical Engineering,

' Kennametal Inc., Corporate Technology Center; Latrobe, PA 15650-0231, USA University of Arkansas, Fayetteville, AR 72701, USA

Abstract-Since the discovery of diamond thin film synthesis using chemical vapor deposition (CVD), a wide range of applications have benefited from this technology. In the machining industry, cemented carbide tool inserts were coated with thin CVD diamond films to extend their wear life beyond their nominal machining performance. Premature failure of CVD diamond coatings to carbide inserts caused by a combination of mechanical, chemical and thermal factors remains a challenging difficulty. Procedures such as chemical etching with an acidic solution, mechanical scratching with diamond grit, and the use of intermediate layers of various materials, among other novel techniques, were shown to enhance nucleation of diamond growth and improve adhesion. To evaluate the effect of such techniques, some comparative adhesion evaluation procedures have been developed. Whereas useful studies addressed several of the issues, the field remains open to additional key investigations that will enlighten the full understanding of this topic.

In this paper, we discuss the state-of-the-art of the adhesion of CVD diamond coatings to cemented carbide tool inserts. Through an in-depth review, a comprehensive discussion on the factors that affect adhesion and a survey of the recent advancements in novel techniques to enhance adhesion of CVD diamond coatings is laid out. In addition, a summary of the available procedures to evaluate the adhesion of CVD diamond coatings to carbide inserts is presented.

Keyvvords: CVD diamond; carbide inserts; adhesion; adhesion testing; adhesion enhancement.

1. INTRODUCTION

The rapidly expanding markets of aerospace and automotive industries caused a significant increase in the machining volume of advanced alloys known for their highly abrasive properties. Examples of such alloys are: hypo and hypereutectic

*To whom correspondence should be addressed. Current address: Caterpillar Inc., Technical Center, Bldg. E-854, P.O. 1875, Peoria, IL 61656-1875. Telephone: (309) 578 3463; Fax: (309) 578 2953; E-mail: TAHER-MAHMOUD-A @ CAT. COM

80 M. A. Tuher et al.

Figure 1. Comparison of major properties of various materials used in machining. The values shown are approximate and for comparison only. Properties of different materials can overlap depending on the method of measurement or the conditions in which a tool services.

aluminum- silicon alloys, metal-matrix composites, and magnesium alloys. To meet this large production volume, advanced cutting tools were in great demand. Diamond is a good candidate material that can be applied to large volume- highly abrasive cutting processes. Figure 1 qualitatively compares some important properties of diamond with those of other widely used cutting materials. Diamond has a low friction coefficient, good chemical inertness, and the highest hardness of any material. The synthesis of thin film diamond coatings by chemical vapor deposition (CVD) led to its application to the widely used (WC-Co) carbide inserts [1-381. The ultimate goal of using CVD diamond coatings is to extend the performance of carbide tools beyond their conventional wear life. This can be readily justified by observing the comparative properties of these two materials as given in Table 1. However, using a hard coating does not guarantee the aimed enhancement. Other considerations that are necessary for an advanced tool include the reliability of the tool and its consistent performance.

Premature adhesion failure of CVD diamond coatings is the major remaining chal- lenge preventing the useful utilization of thin diamond films as wear resistant coat- ings for carbide inserts. Many efforts have been reported in the published literature addressing this difficulty and describing methods used to improve adhesion. On this track, valuable research was conducted to study the chemical and mechanical aspects of this difficulty. To fully comprehend the issue of adhesion of CVD dia- mond coatings to carbide inserts, and to envision the future of this technology, it is important at this time to interconnect all the aspects of this topic and summarize the current status. The aim of this paper, therefore, is to provide scientists and en-

Adhesion of CVD diamond to carbide cutting inserts 81

Table 1. Properties of WC-Co and CVD diamond

Property WC-6 w t 8 CO Ref. CVD diamond Ref.

Coefficient of thermal expansion 4.3 (200°C) ( m l m K) 5.9 (1000°C)

Thermal conductivity (W/m K) Density (g/cm3) Modulus of elasticity (GPa) Tensile strength (GPa) Compressive strength (GPa) Fracture toughness MPa Hardness Coefficient of friction

100- 121 15 6 14- 648 1.45-1.52 5.17-5.93 13 92 HRA (15 GPa) 0.2 pLs on steel

~ 7 1 0.8-1.2 (25°C) ~ 9 1 1.2-1.21 (100-350°C) 3.84 (6OOOC) 4.45 (75OOC)

[871 1000 (90-127°C) ~ 9 1 [871 3.5 ~ 9 1 [871 700- 1079 [891 [871 0.5-2.2 WI ~371 30 [891 [871 6 1891 [87] ([MI) 75-111 GPa [891 [881 0.08 /&j (low speeds) [88]

0.1 /Id (high speeds) Natural diamond on sphere of steel ball bearing

gineers working in the tooling industry a complete and in-depth discussion of the state-of-the-art in the adhesion of CVD diamond coatings to WC-Co cutting inserts.

The topic of adhesion is addressed through five steps. The first step is to identify and define the exact meaning of adhesion to the particular case of diamond coatings to carbide inserts. The second step is to understand and discuss the issues and factors that cause poor adhesion. Based on understanding of these issues, the third step then becomes to describe the novel methods reported in the literature and how these methods contribute to improved adhesion. To estimate the effectiveness of particular procedures in improving adhesion, experimental tests used for adhesion evaluation are discussed. Finally, all the remaining avenues that still require further investigation are summarized.

2. UNDERSTANDING ADHESION

Before understanding the characteristics of adhesion of CVD diamond films to cemented carbide inserts one has to first define the term “adhesion”. Adhesion has been defined in many ways. According to ASTM, adhesion is defined as “the state in which TWO surjiaces are held together by interfacial forces which may consist of valence forces or interlocking forces or both”. Mittal [39] has categorized adhesion into two classes. The two classes of adhesion emerge from the kind of force or work acting on a certain coating-substrate system. The force or work can be either that of attachment or detachment. If a system is expressed in terms of the attachment force or work then adhesion in these cases could be defined as “Fundamental Adhesion” which is simply a summation of all intermolecular or interatomic interactions.

82 M. A. Taller et al.

Therefore, to truly quantify the Fundamental adhesion in a certain system, all ionic, covalent, and van der Waals forces have to be calculated. Measuring such forces in practical situations is difficult because this would require beforehand a complete knowledge of the exact chemical and atomic status at the interface. Such a status is always determined by presumption and by experimental methods that have limited capability and accuracy. Therefore, the role of Fundamental Adhesion is difficult to evaluate since it is an intrinsic property that cannot be explicitly quantified, in addition it may or may not be related to the exact behavior of a system when put into service. The second class of adhesion refers to systems where the work or force of detachment is measured by the application of an external load capable of causing failure to the system under investigation. This class is important, and is referred to as “Practical Adhesion” since it determines the critical state of adhesion at which a system will fail under service. Therefore, practical adhesion is of most interest when studying the performance of the CVD diamond-coated cemented carbide inserts.

Another well-known definition of theoretical adhesion is termed the thermody- namic work of adhesion W,. This kind of work is inter-related to the surface ener- gies y1 and y2 of any two surfaces in contact, and the energy of the interface y I 2 , and is defined as follows [40]:

The theoretical strength of the interface is another quantity that characterizes the bonding across the interface and is another interpretation of Fundamental Adhesion. Thouless [40] outlines the relationship between the thermodynamic work of adhesion and an applied peel stress in a manner similar to a classical tensile test. If a load is applied to a system of adhering materials where the load and displacement are monitored, a plot similar to that in Fig. 2 can be generated. The area under such curve is the thermodynamic work of adhesion W,, and the peak stress a,,, gives the theoretical strength of the interface. However, as will be discussed in the coming sections, the theoretical strength is never achievable due to the presence of cracks that cause failure at stress levels much lower than the theoretical maximum.

In another similar interpretation of ASTM [41], Fundamental adhesion was related to practical adhesion by the following equation:

Fundamental adhesion = Practical adhesion + Internal stress.

This equation, however, is a qualitative representation of the relationship between practical and Fundamental adhesion. In order for this equation to have a quantitative interpretation, the conditions required for stress superposition have to be met. For example, practical adhesion should have the same units and orientation as the internal stress, and this normally is not the case in practical situations.

The various definitions of adhesion caused a great diversity in defining the criteria for its evaluation. As a result, reported studies discussing adhesion appear to be scattered rather than cumulative. Therefore, as the discussion continues, an attempt

(2)

Adhesion of CVD diamond to carbide cutting inserts

CT t t t

( 4 1 1 1 0

83

6

Figure 2. (a) Uniformly separating an interface by an applied stress, a , would give (b) an idealized stress-displacement (a vs. 6 ) plot yielding both the thermodynamic work of adhesion, WA, and the theoretical strength of the interface, a,,, (adapted from Thouless [40]).

will be made to define adhesion for CVD diamond coatings to carbide cutting inserts based on the issues that affect it.

3. ISSUES RELATED TO THE ADHESION OF CVD DIAMOND COATINGS TO CARBIDE CUTTING INSERTS

The state of adhesion of CVD diamond to carbide inserts is generally influenced by mechanical, chemical, and thermal issues. Failure mechanisms, machining stresses, residual stresses, elastic stored energy, mechanical interlocking, and the interface toughness are some of the mechanical issues that are of concern. On the other hand, the effect of the cobalt binder, and surface reactivity are additional chemical issues that are important to examine and know. At any time, two or more of the above- mentioned factors occur simultaneously and, therefore, cannot be separated. In addition, some of these factors can be controlled to enhance adhesion while others cannot. Understanding these issues in the case of CVD diamond coated carbide inserts is vital before addressing means to improve adhesion.

3. I . Mechanical issues

3.1.1. Failure of diamond coatings. Generally, when no external loads are applied to a coating its mechanical failure was classified into two classes depending

84 M. A. Tnlzer et al.

on the region in which failure occurred. If failure occurs within the bulk of the substrate or the coating it is termed cohesive failure. However, debonding of the coating due to a complete separation of the coating from the substrate with minimal apparent damage of either surface is classified interfacial failure [39, 421. Until a recent time, little information has been available on the exact type of failure of CVD diamond coatings on carbide inserts. However, in a study by Perry et al. [42], it was shown that the failure of CVD diamond coatings deposited on W substrates occurred discretely at the nucleation plane of the diamond film. These W substrates were treated with the minimal requirements for diamond nucleation by abrading the surface with fine diamond powder in an ultrasonic bath. Auger scans detected the formation of several carbide compounds of W at the backside of the detached diamond films. However, little traces of W were found on the detached diamond coating indicating that the failure was interfacial with respect to the original interfaces, but could be considered cohesive with respect to the carbide layer that formed. Soderberg et al. [3] reported that a weak interface caused by etching of the cobalt binder resulted in the failure of the WC grains together with the diamond coating. In other studies [43], visual inspection revealed that the detachment of diamond coatings deposited on carbide tool inserts usually

Figure 3. Two modes of failure for diamond coatings on cemented carbide tools. (a) cohesive failure and (b) interfacial failure.

Adhesion of CVD dinmorid to carbide cutting inserts 85

(b) Figure 3. (Continued).

occurred with no apparent cohesive failure in the substrate. The contradiction in such reports indicates that failure of diamond coatings depends on the conditions of pretreatment, the cause of failure and the definition of the interfacial region. Therefore, a reasonable conclusion to draw is that both modes of failure occur simultaneously. The challenging question is to determine which mode occurs first and hence leads to the occurrence of the other mode. Figure 3 shows examples of interfacial detachment and cohesive failure that occurred in diamond coatings right after the deposition process.

3.1.2. Machining stresses. For attaining a satisfactory adhesion of CVD dia- mond coatings to carbide tool inserts, it is initially important to obtain information about the values and orientations of mechanical stresses that a coating- substrate system will experience during machining. However, due to other factors that in- clude the type of material being machined, and the cutting conditions, the collection of such information would require enormous databases associated with numerous experiments. In addition, the knowledge of such stresses solely cannot establish the baseline adhesion strength required for a CVD diamond coating deposited on a carbide insert. Other residual stresses are present in the coating that can have either a beneficial or detrimental effect on adhesion and, therefore, have to be accounted

86 M. A. Tuhev et al.

for as well. Despite the importance of collecting machining data, only a few stud- ies have specifically addressed this issue. From elementary orthogonal machining mechanics, the mean steady state normal, and shear stresses can be calculated as follows [44]:

ornr = 2Fvl WLc, (3)

rrnf = 2pFp/A, (6)

where: om, and emf: the maximum normal stresses on the rake and flank faces, respectively; rmr and rmf: the maximum shear stresses on the rake and flank faces, respectively; 8: chip flow direction; I,: tool-chip contact length; 1; : tool- chip contact length during sticking friction only; Fv: main cutting force (parallel to the work velocity); Ff: feed cutting force; Fp: radial cutting force; p: coefficient of friction of the flank face; W: chip width; A: flank wear area.

The geometry of the cutting process is shown in Fig. 4. Calculating machining stresses requires knowing the values of radial forces and the area of flank wear zone. These data can be collected from either experimental measurements or finite element modeling (FEM) of the machining process. In studying the effects of the post-deposition polishing of CVD diamond films, Bhat et al. [22] measured the reduction of radial cutting forces due to mechanical polishing of diamond coated inserts. The study revealed that for an as-deposited diamond coated insert the radial force increased steadily from 43 to 115 N with the cutting time increasing from 1 to 6 minutes, respectively. Although no flank wear data were reported in this study, it can be assumed that the flank wear area during machining was 200 x 1000 p m which is reasonable when compared to several studies under similar cutting conditions [24]. Thus, the calculated maximum normal stress in the flank region reaches a value of 1150 MPa. Lin and Liu [45] constructed an orthogonal FEM cutting model to investigate the distribution of machining stresses within the workpiece and the cutting forces in the tool for two different tool materials, a diamond cutting tool and a carbide cutting tool. Although the study focused more on the role of the thermal conductivity of each of the later tools on the amount of heat generated at the cutting interface, some useful information about the forces generated was supplied. The effective (Von Mises) stress at a point on the rake surface was reported to be equal to 11 10 MPa. Despite the direct role that

Adhesion of CVD diamond to carbide cutting inserts 87

Figure 4. Geometry of the cutting action during orthogonal cutting. Note that Fv is normal to the plane drawn (adapted from Younis [44]).

machining stresses have in determining the adhesion necessary to sustain acceptable machining performance, it is a factor that cannot be controlled for enhancing adhesion. However, using a proper set of machining parameters one can reduce these stresses.

3.1.3. Pre-service stress (residual stress). It is widely known that residual stresses that exist in CVD diamond-coated inserts consist of two independent stresses: thermal stress (ath), and intrinsic stress (alnt) related according to [46-481:

atotal = a t h -k P i n t . (7)

The thermal stress is caused by the mismatch of the coefficient of thermal expansion between CVD diamond and WC-Co. This stress generates after the deposition process is complete and the diamond- substrate sample is left to cool down to room temperature. Thermal stress can be calculated according to the following equation:

(8) a t h = Ef(af - a , )AT/( l - Uf).

where E f is Young’s modulus of the diamond film, a, and cyf are the carbide substrate and diamond film coefficients of thermal expansion (CTE), respectively. AT is the difference between the deposition temperature and the room temperature, and uf is Poisson’s ratio for the diamond film. The average value of thermal stress calculated for CVD diamond coatings on carbide inserts in several reports

88 M. A. Taher et al.

was -3.7 GPa. Intrinsic stresses, on the other hand, cannot be expressed in a straightforward manner. There are several factors that contribute to intrinsic residual stresses such as, impurities, grain boundaries, dislocations, voids, and lattice mismatch. Generally, the total residual stress (qtota~) can be measured by the utilization of several analytical techniques. The two common techniques used for measuring stresses in CVD diamond coated inserts are the d - sin2 3 method using X-ray diffraction [23, 48, 491, and Raman spectroscopy [51-53]. These techniques measure the total residual stress from which the calculated thermal stress (qh) can be deducted giving the value of intrinsic stress. However, each of the mentioned techniques has its own limitations and benefits. An in-depth comparison of these techniques was reported by Windischmann and Gary [47], and Mohrbacher e ta l . [53].

There have been some studies that reported the value of the total residual stress in diamond coatings on carbide inserts [24, 28, 50-541. In general, the total residual stress was found to lie in the range of -300 MPa to -3.0 GPa. Due to the nature of residual stress measurement techniques, the thermal and total stresses were always assumed to be biaxial and parallel to the substrate-coating interface to make it easier to quantify. This assumption is reasonable as long as the thickness of the coating remains small [55] compared to the thickness of the substrate. However, finite element modeling conducted by Gunnars and Alahelisten 1561, and Taher et al. 1571 revealed that other peel and shear stresses existed especially at the carbide insert tip. Figure 5 shows the distribution of thermal residual stresses across the length of the cutting edge of a CVD diamond-coated carbide model. With the existence of such stresses a more precise understanding of how they affect adhesion is needed especially when these residual stresses are combined with the stresses that generate at the cutting edge during machining. Although extensive research has been conducted on studying the residual stress due to diamond CVD deposition, the fundamental question of the exact role of these stresses on the adhesion strength has not been completely answered. There is a common belief that the presence of residual stress is detrimental to the adhesion of CVD diamond coatings. However, this statement does not reflect the true effect of residual stresses. Depending on the orientation and values of the residual stresses their role can be detrimental as well as beneficial at the same time. Again, the mode of failure of the coating, whether cohesive or interfacial, plays a role in determining the effect residual stress has on adhesion. One way in which residual stress can be detrimental is that it can cause brittle cohesive failure in the diamond coating when stress exceeds the strength of diamond. In this case, the combined use of FEM to determine principal stresses and a brittle failure theory such as the Columb-Mohr theory becomes necessary. Also, residual stress increases the strain energy stored by the coating. The increase in stored elastic energy places the coating in a state of conditional mechanical stability that can lead to catastrophic failure when external loads such as machining forces are applied to the diamond-carbide system. On the other hand, residual stresses can become beneficial in controlling crack propagation from the surface to the bulk of the coating as will be discussed later.

Adhesion of CVD diamond to carbide cutting inserts 89

A Direction for which thermal residual stresses in part (h) are plotted X

(4 1000 500

0

A -500 0 g -1000 ; -1500 v) E -2000

-2500 -3000

-3500

-4000 0 0.2 0.4 0.6 0.8 1

X/L (k6mm)

(b! Figure 5. Results from a finite element model showing the distribution of thermal residual stresses on a diamond coated cemented carbide tool ( [57] ) . (a) Geometry of the FEM model. (b) Distribution of thermal stresses along the rake face.

90 M. A. Tuher et al.

A high residual stress should not necessarily indicate poor adhesion when considering the interfacial failure of diamond coatings. On the contrary, it indicates that the adhesion at the interface is strong enough such that when cooling occurs after the CVD process, the substrate becomes restrained from retrieving its original dimensions inducing the high thermal compressive strain in the coating. Therefore, for identical deposition conditions a lower residual stress in the coating should signal the lack of proper adhesion, because this would mean that the substrate has restored most of its original pre-deposition dimensions.

Unlike machining stresses, residual stresses can be controlled, to a certain extent, by varying the deposition conditions such as the hydrocarbon gas percent, the deposition temperature and the reactor pressure. The role of deposition conditions on residual stresses in diamond coatings deposited on various materials such as Si, Sic , Mo, Cu, Ni, Fe-Ni alloy, W, and Ti has been numerously studied [50, 51, 58-71]. Fewer reports have been found regarding the same study on diamond coatings deposited onto WC-Co inserts specifically [24, 521. Kuo et al. [52] showed that the compressive residual stress in the diamond coating increased from -1.5 GPa to -2.9 GPa with increasing the deposition time from 1 to 8 hours then decreases beyond 8 hours to a value of -1.75 GPa. Also, the compressive residual stress in the coating was found to increase from -2.1 GPa to -3.4 GPa with increasing the methane to hydrogen ratio (CH4 : H2) from 1 to 1.5 ~01%. A similar trend was reported by Taher et al. [24] in which the compressive residual stress in the diamond coating was found to increase from -270 MPa to - 1.3 GPa by increasing the methane concentration in the reactor from 1 to 5 ~01%. The addition of intermediate layers of materials having a CTE closer to that of diamond can reduce thermal stresses as well [18,21,27, 32-34,721. Table 2 lists some important reported mechanical stresses in the CVD diamond/ WC-Co system. Information

Table 2. Some important reported mechanical characteristics of the CVD diamond cemented carbide (WC-Co) system

~ ~~

Mechanical characteristic Reported values Remarks

Total residual stress (atotal) -300 to -3 GPa 0 Normal stress parallel to interface 0 Averaged from several references 0 Other peel and shear stresses exist at

0 Normal stress parallel to interface 0 Averaged from several references 0 Normal stress parallel to interface 0 Averaged from several references 0 Assumed for a region 200 p m x 1000 p m

cutting tip (FEM) Thermal stress (athermal) -3.7 GPa

Intrinsic stress (crintrinsic)

Maximum normal stress 1150 MPa in flank region in area Maximum normal stress 1110MPa 0 From FEM of orthogonal machining in rake region

3.4 to 0.7 GPa

Adhesion of CVD diamond to carbide cutting inserts 91

about the superposition of residual stresses and machining stresses is valuable and has not been reported yet.

3.1.4. Crack initiation. The failure of any brittle material such as CVD diamond occurs in three steps. First crack initiation, followed by crack propagation, then complete failure. There are several origins of crack initiation in the CVD diamond/ carbide system that can be intrinsic or mechanically induced. Intrinsic cracks occur due to the nature of CVD diamond deposition where nucleation is followed by three-dimensional growth with little secondary growth causing the formation of voids at the interface. These voids that increase in number if pre- deposition nucleation is low are the main source of intrinsic crack initiation [ 3 ] . Mechanically induced cracks are initiated by the gradual release of the stored elastic energy generated by high residual stresses [73]. A preferred avenue for this crack formation process would be through the weak byproduct defects of the CVD process such as impurities of hydrogenated diamond carbon and sp2 bonded carbon. These non-diamond species have a lower mechanical strength than diamond and, therefore, they become the preferred avenues for crack formation and propagation. One way to limit crack initiation sites and improve the adhesion of CVD diamond coatings is to insure proper seeding of the substrate prior to diamond deposition to enhance the nucleation density. To minimize the role of other defects such as defective grain boundaries and impurities in crack initiation, optimizing the hydrocarbon gas percentage and controlling the growth rate becomes necessary.

3.1.5. Crack propagation by the release of the elastic strain energy. As men- tioned earlier, the theoretical strength of an interface is a quantity that cannot be truly analyzed. However, a more important property of a bonded system is the mechanical toughness of the interface G, also known as the critical strain energy release rate [74]. The G, is quantified experimentally by determining the load re- quired to propagate a sharp crack, then using this load and the geometry data G, can be computed [40]. When a load is applied to a system in which cracks have initiated, the mechanical strain energy G in the system is altered. If the applied loads cause the strain energy of the system to exceed G, of the interface, then G be- comes the driving force for crack propagation that may cause delamination. In this case, all the stored elastic energy is instantaneously transformed to kinetic energy and failure occurs. Although no data were found concerning G, for CVD diamond coatings on carbide inserts, one study addressed this issue by calculating the strain energy release rate of diamond films on titanium. Chandra er al. [74] combined the thermal residual stress calculated at the temperature at which a diamond coating de- bonded from a Ti substrate during cooling with the intrinsic stress calculated from a similar coating on a W substrate. Using this total stress value at de-bonding, G, was calculated. Therefore, enhancing adhesion can be achieved by increasing the inter- face toughness and thus increasing G,. A study addressed solely at quantitatively calculating G, of CVD diamond on carbide inserts will be of most significance in

92 M. A. Tiher et al.

predicting the permissible external loading that a diamond-coated cutting tool can withstand without premature de-bonding.

3.1.6. Wear mechanisms. The failure of CVD diamond-coated inserts during machining can be in the form of flaking (interfacial failure) or abrasive wear (gradual cohesive failure) [ 2 2 ] . Ideally, a test of superb adhesion is when the diamond coating fully deteriorates by wear rather than flaking. Flaking will occur primarily due to poor adhesion between the diamond coating and the carbide substrate [6]. Therefore, flaking is clearly undesirable because the benefit of using a diamond coating is lost, except for the chip breaking assistance of Faceted diamond crystals at the rake surface [29, 751. If the adhesion strength of the CVD diamond coating is sufficient to withstand the machining stresses, then the abrasive action between the workpiece material and the diamond coating becomes the primary failure mechanism.

Unless the CVD diamond coating is polished, a two-step wear mechanism is ex- pected to occur. The first step is caused by the initial high surface roughness of the CVD diamond coating in which crack initiation occurs at the surface. The mecha- nism that describes such behavior was proposed by Gunnars and Alahelisten [%I. They described a three-zone wear model as shown in Fig. 6. In this model, the role of residual stresses becomes significant in controlling crack propagation from the surface to the interface that could lead to interface failure (flaking). As outlined earlier, the high total compressive residual stress present in CVD diamond coatings on carbide inserts was assumed to be biaxial and oriented parallel to the interface. Wear starts to occur at the surface, which, because of geometry, allows stress to relax. A crack is more likely to initiate at protruding grains in zone I and propa- gate preferentially along the (1 1 1) easy cleavage planes of diamond. The geometry at deeper depths, however, prevents the compressive residual stress from relaxing. Therefore, as the crack propagates deeper in the coating, it encounters higher com- pressive stresses that cause the cracks to redirect their paths deviating from cleavage planes to a direction parallel to the interface in region 11. The high compressive stress now causes cracks to propagate fast parallel to the interface resulting in a smooth surface in region 111. Due to the smoother surface, fewer asperities will be present and it becomes harder to nucleate cracks.

Figure 6. The three zones wear mechanism proposed by Gunnars and Alahelisten [56] .

Adhesion of CVD diamond to carbide cutting inserts 93

In addition the high residual stress now present at the surface will make it more difficult for cracks to open. Even if a crack does nucleate in this region, the high compressive stress will quickly redirect its path to the free surface. It should be noted that the peel stress is zero at the free surface and increases with the increase in depth where it becomes largest at the interface. Therefore, peel stress will promote propagation of a crack originating at the interface due to voids or other defects explained earlier. This wear mechanism concludes that compressive normal residual stresses that are parallel to the interface reduce crack propagation from the surface to the interface, while tensile peel residual stresses cause interfacial cracks to propagate parallel to the interface. It also shows how residual stresses, if controlled, can become beneficial in increasing the life of a diamond coating.

Since now the free surface has become smoother, the second step in the wear mechanism becomes dominant. In this step proposed by Malshe et al. [75] wear oc- curs due to abrasion resulting from chemical interactions at the diamond- workpiece interface. As the localized temperature rises due to the absence of cutting fluids at the cutting interface, certain surface diamond crystals undergo several phase trans- formations to non-diamond structures that exhibit poor mechanical properties. Fol- lowing this stage, micro-chipping of the non-diamond phases occurs and flank wear progresses. The fracture toughness and cohesion of diamond crystals in CVD dia- mond depends on the chemical quality of the CVD diamond film. CVD diamond is a polycrystalline structure in which individual crystals are grouped together with less than perfect atomic bonds along the grain boundaries. If the chemical qual- ity of the CVD coating is poor, there will be large quantities of amorphous carbon or hydrogenated diamond present that will deteriorate the cohesion of the diamond crystals. This weak cohesion of diamond crystals will adversely affect the rate of abrasive wear. Fortunately, the chemical quality of CVD diamond used for tooling applications includes minimal amounts of impurities [22, 361 making the role of crystal cohesion less important.

Gunnars and Alahelisten [56] mentioned that as the cutting surface became smoother it would be extremely hard to initiate and propagate cracks into the coating and consequently the wear rate becomes very low. In this sense, the first step in the wear mechanism described by the three-zone model can be eliminated by subjecting the CVD diamond coating to a post-deposition polishing process. As a result, the requirement for a high compressive residual stress would be less important, therefore benefiting the adhesion of the coating. Bhat et al. [22], showed that by post-deposition polishing, lower radial cutting forces were experienced at the cutting edge as well, which demonstrates another benefit from having a smooth surface. This reduction in the cutting forces lowers the mechanical load on the coating, thus minimizing the risk of adhesion failure.

3.1.7. Mechanical interlocking. One of the significant factors that should affect the adhesion of CVD diamond to WC-Co inserts is the mechanical interlocking that occurs at the coating-carbide interface. Fundamentally, mechanical interlocking is based on a mechanism by which initial diamond nuclei grow inside surface cavities

94 M. A. Tuher et al.

Mode I

wc Mode I1

u c o @$@ CVDdiamond

Figure 7. The two modes of mechanical interlocking. Mode (I): resistance to peeling is offered by combined friction and obstruction forces. Mode (11): resistance to peeling is offered by friction forces.

of the substrate. Upon completion of growth of a single diamond crystal, it would be partially embedded in a group of substrate grains as shown in Fig. 7 . According to the orientation of how the diamond crystal will grow, the interlocking action can occur with two contact modes. One mode in which resistance to peeling is generated by obstruction from the substrate grains (I), and, therefore, the magnitude of the detachment force will depend on the strength of the substrate. The second mode (11) is when the resistance to peeling is the friction force that develops between the diamond crystal and the substrate grains. Apparently, the second mode of interlocking is weaker than the first. If neither mode occurs, then adhesion will presumably be sustained by chemical bonding and attraction forces only, the significance of which has not been thoroughly investigated. Generally, it is believed that both modes of interlocking occur simultaneously in most cases. However, there has been no quantitative study to address this issue.

If adhesion is to be improved, it is necessary to implement conditions in which mode I of interlocking would be dominant such as proper seeding of the substrate prior to diamond coating. In addition, optimizing the substrate surface roughness helps in creating surface cavities that can trap the seeded diamond particles. Surface roughness of the substrate plays another important role in adhesion optimization. It is known that crack propagation is accelerated by a smooth surface because less energy is required to move the crack. However, a rough surface would obstruct the movement of a crack by the dispersion of energy at the asperities [76]. For a crack to continue its growth, excess input energy would be required. Alam et al. [76] conducted a study to determine the role of the surface roughness of W substrates prior to deposition on the adhesion of the diamond coating. Adhesion was found to improve initially up to a substrate surface roughness of 110 nm, but then dropped again when the substrate surface roughness approached 180 nm. The initial improvement was explained by the increase of surface area and mechanical interlocking between the coating and the substrate. However, as the substrate

Adhesion of CVD diamond to carbide cutting inserts 95

surface roughness increased, the concentration of coating defects and voids resulted in a decrease in adhesion. A study focusing on the effects of surface roughness of WC-Co substrates on the mechanical interlocking is necessary and would be expected to yield valuable information.

3.2. Chemical issues

3.2.1. The role of the cobalt binder. One of the early conclusions explaining poor adhesion of CVD diamond coatings to the WC-Co system is the existence of cobalt as a binder material [14, 77-79]. By observing the phase diagram shown in Fig. 8 of the Co-C system, it is apparent that at the deposition temperatures of CVD diamond (800-1000°C) C is completely soluble in Co up to 0.9 wt% C. Therefore, in the initial stages of WC-Co exposure to a CVD plasma (rich in hydrocarbon radicals) there is a rapid diffusion of carbon species from the surface of the substrate into the bulk until the carbon solubility in Co is exceeded. The catalytic effect of Co being a transition metal with a partially filled 3d (Co has an electronic configuration of 1 s2 2s2 2p6 3s2 3 p 6 3d74s2) shell preferentially promotes the formation of a layer of graphite. When the substrate becomes completely saturated with C, graphite deposits at the surface before any diamond nucleation could occur. The formation of this graphite layer has been shown to occur in previous studies [8, 15, 17, 771 and resulted in poor adhesion. Prolong exposure to the high deposition temperatures could also lead to cobalt out-diffusion [16, 441,

Atomic Percent Carbon 0 2 4 6 8 10 12 14 16 18

1800 , , , , , I ' m , , , I , , 7 ' ) , , m i " ' , , , , , , , , ' , , , , ,

- 422°C €

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Weight Percent Carbon

Figure 8. The phase diagram of the Co-C system [95].

96 M. A. Taker et al.

(a> Figure 9. SEM pictures of the detrimental effects of cobalt. (a) Cobalt out-diffusion, (h) cobalt damage to full grown diamond crystals.

which attacks fully-grown CVD diamond crystals as shown in Fig. 9. Currently, all researchers agree that suppressing the activity of Co from the substrate surface of cemented carbide inserts is essential in promoting diamond growth. The methods by which Co is suppressed will be discussed in the following sections.

3.3. Sunirnuty

To conclude the understanding of the term adhesion for diamond-coated carbide inserts, and based on the previous discussions, adhesion will be categorized as follows.

Practical Adhesion of diamond coatings to carbide tool inserts is the measure of force or work required to cause interfacial removal of the diamond coating from the carbide substrate. Such force or work may be intrinsically needed to: overcome the interfacial mechanical interlocking force and/or the chemical bonding: initiate interfacial cracks: or increase the elastic stored strain energy caused by residual stresses beyond the critical value of the interface. However; such force or work may not contribute to initial cohesive damage of the diamond coating.

Adhesion of CVD diamond to carbide cutting inserts 97

(b) Figure 9. (Continued).

4. TECHNIQUES TO IMPROVE ADHESION

After discussing the various issues that affect CVD diamond adhesion to WC-Co inserts, it now becomes easier to decide on methods to improve adhesion by simply addressing the difficulties with appropriate solutions. Enhancement of the adhesion strength of CVD diamond can be achieved through the following avenues: 1. Halting the detrimental effect of cobalt by selective surface removal or by

2. Maximizing fundamental adhesion by increasing the interfacial interactions

3. Increasing mechanical interlocking by optimizing the substrate surface rough-

4. Controlling crack propagation by optimizing the residual stress, and increasing

The four avenues listed above are behind most of the research conducted so far aimed at improving the adhesion strength of CVD diamond coatings to carbide tool inserts. Some of the pre-deposition treatments and the results obtained by their application that have been reported in the literature are discussed in the following

suppressior,.

through chemical bonds.

ness and thus improving nucleation.

the interface toughness.

98 M. A. Tuher et al.

sections. An attempt will be made to point out the reasons for implementing a certain method from a chemical and/or a mechanical point of view, whenever applicable. Tables 3 and 4 contain a comprehensive list of most of the novel procedures used for enhancing the adhesion of CVD diamond coatings specifically on cemented carbide tool inserts.

4.1. Cobalt etching

The most popular approach to suppress the cobalt effect is to remove cobalt from the interface by etching the substrate surface with a chemical solution for a specific amount of time. The most common etching agents used are listed in Table 5. Although in certain cases the use of the chemical solution was to increase the surface roughness of the substrate, the major goal was to selectively eliminate cobalt from the surface and subsurface layers of the substrate.

The etching time is critical since excessive cobalt removal can lead to a weak brittle interface by depleting the binder phase in the WC matrix. The samples together with the etching solution are usually placed in an ultrasonic bath for 5 - 15 minutes. This was shown to be effective in removing cobalt from the surface [19]. Huang et al. [8] examined the effect of Co% on adhesion. They reported that the size of cracks generated from indentation testing was reduced from 450 p m for samples with 25% Co, to 250 p m for samples with 6% Co. In addition etching caused a reduction in the size of cracks from 350 p m to 150 pm.

A quantitative comparison of the effectiveness of several etching agents was reported by Deuerler et al. [26]. A shown in Fig. 10, the most effective agent that generated the least amount of Co at the substrate surface was a 1 : 3 HN03 : HC1 solution.

To examine the effect of etching time on the adhesion strength of the coating, Fan et al. [ 181 compared the critical load for radial crack formation during indentation of various diamond-coated substrates treated for different etching times. It was shown that the optimum etching time using a 10% HN03 solution was 10 minutes beyond which no improvement in adhesion was observed. Most of the other studies conducted by researches showed similar improved adhesion of CVD diamond coatings as a result of cobalt surface etching and, therefore, this step should be implemented in every substrate preparation.

4.2. The application of surface rubbing, grinding, sand blasting and water peening for sugace roughness optimization

Increasing the substrate surface roughness prior to diamond deposition was under- taken to enhance adhesion since an increase in the surface area of contact, better nucleation and mechanical interlocking were expected to occur. The procedure re- ported varied depending on the scratching medium, grit size, time and method of application. The simplest of all methods is to rub the surface with a diamond paste, followed by ultrasonic cleaning. Saijo et al. [2], Saito et al. [9] and Tsai et al. [ 111, for example, applied the later procedure. All these studies reported enhanced adhe- sion as a result of the rubbing action. More complex methods used an ultrasonic bath

Table 3. List or research works conducted to enhance thc adhesion of CVD diamond coatings

Treatment method

Substrate CVD deposition method

Thickness of coating

Adhesion tcsting method (see Table 7)

Results Ref.

1 WC-6-27 wt% Co HFCVD NIS

5

IS0 K10 grade SPGNl20308

IS0 K10

WC-6 wt% Co

WC-6 wt% Co

6 w c - C o

7 w c - C o

MWPCVD

MWPCVD

HFCVD

d.c. jet CVD

20 LLm

20 p m

NIS

NIS

Indentation and machining

Indentation and machining Indentation and machining Nonc

Indentation

r.f. CVD 2- 15 pin None

r.f. CVD 30 jLm Nonc

Lower Co content exhibited better nucleation and adhesion than higher Co content from machining and indentation test Improved adhesion as observed from indentation and machining tests Improved adhesion as observed from indentation and machining tests The deposition temperature, and the duration for which the sample was kept in diamond suspension afkctcd nucleation density Improvcd adhesion with no detectable damagc acter using

6

i 2

200 N Rockwell-C indenter for Ag interlayers < 6 p m in thickness

I O iLm thick diamond coatings Adequate adhesion up to 30 p m thick diamond coatings

Adequate adhesion up to 1111

[ 111

Table 3. (Continued)

Treatment Substrate method

CVD Thickncss deposition of coating method

Adhesion Results testing method (see Table 7)

Ref.

9

10

SPGN 120308 wc-5.5 wt% Co wc-Co

1 1 IS0 K I O WC-6 wt% Co

8 WC-6 ~ 1 % CO HFCVD 6.5-30 p m Indentation, Adhesion failure by flaking [201 hand was reduced by increasing grinding the surface roughness of

the substrate

than untreated substrates HI’CVD I 1-24 p m Indentation Qualitatively, better adhesion 1161

HI’CVD N/S Indentation, Improved adhesion as 1181 grinding indicated from indentation

and wear testing Machining Improved adhesion by I261

applying a combination of mechanical and chemical pre-treatments to the substrate as shown by the machining performance of the coated cutting tools Qualitatively improved adhesion comparcd with untreated substratcs

13 Indcntation Good adhesion with and machining performance machining comparable with that of

brazed PCD cutting tools

12 IS0 grade K68 w c - 6 w t 8 Co SNGN & SPGN w c - c o

d.c. jef CVD

NIS

HFCVD N/S

N/S 25 - 30 b m

Nonc

128,291

Table 3. (Continued)

Thickness Adhesion Treatment Substrate CVD mcthod deposition of coating testing method

method (see Table 7)

Results Ref.

14 w e - 6 wt% Co HFCVD N/S Scratch

15 we-IO wt% co

16 w e - 5 wt% Co

DCPACVD 5-7 p m Indentation

N / S Indentation MWPCVD

17 w e - c o DCCVD N / S Indentation

18 we-2 wt%; Co HFCVD 5 p m Erosion blasting

w e - 8 wt% e o

Insufficient adhesion with a 5.6 pm thick TiN intermcdiate layer. Low adhesion with an a-C intermediate layer. Best adhcsion with a 9.6 p m thick Sic and a 6.8 pm thick SilN4 interrnediatc layers Highly adherent coatings when tested

by Vickers indentation Improved adhesion using two separate treatments when compared to untreated sample as indicated from indentation Best adhesion achieved with substrates that had the ideal convex surface feature when prepared by sandblasting Best adhesion achieved with sample having 2 wt% Co, etched and water peened at a prcssure of SO MPa

1271

[33, 341

1321

I351

1381

~

HFCVD: hot filament CVD. MWPCVD: microwave plasma CVD. DCCVD: direct current CVD. DCPACVD: direct currcnt plasma assisted CVD. r.f. CVD: radio frequency CVD. N/S: data not specilied by the authors.

Table 4. Description of reported state-of-the-art procedures to enhance CVD diamond adhesion to WC-Co cutting tool inserts

Method Criteria of adhesion Cobalt Etching Roughness Novelty procedure Ref. No. enhancement etching solution optimization

(Table 5)

I To suppress cobalt, increase Yes 1 Yes surface roughness

2 To suppress cobalt and Yes NIS Ycs increase interlace toughness

3 To suppress cobalt and NIS NIS NIS increase interface toughness

4 To suppress cobalt, increase Yes I Yes nucleation density

1. Chemically etched in a I : I HNOl : HzO solution 181 for 3 minutes

2. Polished with 2-4 p m diamond paste lor IO minutes 3. Plasma etched for dccarburization at 930°C for I h (only 5

2 for 2.5 h -i

2 5

samples with 12 wt% Co) b

s m 4. Scratched with grade 600 S i c powders in a ball mill

Adding a few percents of a metal M belonging to thc group 4a or Sa of the periodic table to the WC-Co system. Prior to deposition the substrates were heat treated in inert gas Adding a few percents o f a metal M belonging to thc group 4a or Sa of the periodic table to the WC-Co system by using ion-plating method. Prior to deposition the substrates were heat treated in inert gas

2. Ultrasonic treatment in a diamond suspension

[9]

[IO]

1. Chemically etched in I : I H N 0 3 : H 2 0 solution with a I141

concentration of25 x 1 0 ~ ~ g/mI for 20 minutes

Table 4. (Continued)

Method Criteria of adhesion Cobalt Etching Roughness Novelty procedure No. en hanccmcnt etching solution optimization

(Table 5)

Ref.

5 Suppress cobalt, create Yes I Yes a stress relief multilayer

To create anchoring sites Yes for improved mechanical interlocking

increase interface toughness To suppress cobalt and Ycs

Ycs

No

I . The substrates were first ground using a diamond 1121 grinding wheel

for 30 s 2. Chemically etched in 10 vol% solution of llNO3 at 50°C

3. Rubbed with a diamond paste of grain size 1-3 p m 4. A stress relief multilayer was deposited using magnetron

sputtering at a substrate temperature of 200°C that consisted of: a 0.04 pm thick Nb (or W) layer;

bsorbing Ag layer whose thickness varied between 2-25 pm; and a top layer or 0.04 p m Nb (or W)

were heated to a tcmperature above the melting temperature of Ag to obtain a good wetting between Ag and diamond

5. At the last stage of CVD diamond growth, the samples

Etching of cobalt followed by rubbing with 0.1-0.25 pm siLed diamond powder and embedding them into voids between WC grains, then cleaning the excess powder I . Initial deposition of diamond crystals without forming

2. Electroplating o f a Ni binder to till the voids between

3. Diamond growth continued on the Ni-diamond composite

11 1 I

I l l 1 a continuous coating

diamond crystals

R 5. 3 s R

I

0 W

e 0 P

Table 4. (Continued)

Method Criteria of adhcsion No. enhancement

Cobalt Etching Roughness ctching solution optimization

(Table 5 )

8 To suppress cobalt by removal, increase surface roughness by lapping and sintering

Yes 8 Yes

9

I O

To suppress cobalt by forming No NIA N 0

non-catalytic Co compounds (borides and silicides)

To suppress cobalt by removal, to rcduce residual stress by interlayer deposition, and improve nucleation by decarburization

Yes 1 Yes

Novelty proccdurc Ref.

- Four different procedures: 1201 I . Lapping 2. Lapping and removing cobalt in a I : 3 H N 0 3 : J 1 2 0

3. Sintering and removing cobalt in a I : 3 HN03 : 1 1 2 0

4. Sintering and decarbonizing in H2 + 1 % 0 2

Samples were ultrasonically cleaned in acetone. The

(grain size, approx. 100 pm) were placed i n quartz ampules that were sealed under vacuum (about 13.3 Pa) The substrates w’ere treated for 3 to 24 h at a temperature of 700°C

2. Mechanically polished and etched in dilute nitric

3. Decarburized in H2 atmosphcre for 5- IS

4. The samples suspcndcd in a diamond suspcnsion

solution for I0 minutes

solution for 10 minutes ;s ? 3 z atmosphere at 860°C for 30 minutes

1161 suhstratc together with boron and silicon powders P

EL

I . Samples cleaned i n acetone and methanol 1181

acid for I , 5 , I O minutes

minutes

with a diamond particle size of 0.3 pm

Table 4. (Continued)

Method Criteria of adhesion Cobalt Etching Roughness Novelty procedure Ref. NO. enhancement etching solution optimization

(Table 5)

I I To suppress cobalt, improve Yes nucleation, increase surface roughness, and generate chemical bonds

2 ,5 ,6 Yes

b a 2.

5. A discontinuous layer of CVD diamond deposited 6. 0.5- 1 p m thick ceramic coatings of TiN, TIC, WC, Sic, 2-

% 7. A second diamond layer deposited on each

of the ceramic coatings 8 Mechanical: 1261 S’ A: as ground 2 P: polished R U: ultrasonic nucleation enhancement with isopropanol and diamond powder (< 1 pm)

and Si3N4 deposited by laser CVD on the discontinuous diamond layer

3

0

a

s

13: R Chemical:

EH: etching with HNO3 (3

E enhancement 00

EQU: etching with HN03 + 3HC1+ ultrasonic nucleation enhancement EM: etching with Murakami’s solution EMU: etching with Murakami’s solution

r)

EHU: etching with H N 0 3 + ultrasonic nucleation

EQ: etching with HN03 + 3HCl (aqua regia)

CJ. s

2. m 6

+ ultrasonic nucleation EMH: etching with Murakami’s solution + etching with HNO3

L

0 VI

+- 0 m

Table 4. (Continued)

Method Criteria of adhesion Cobalt Etching Roughness Novelty procedure No. enhancement etching solution optimization

(Table 5)

Intermediate layer: ACP: polished + amorphous carbon (a-C) by arc-ion plating ACU: untreated + a-C by arc-ion plating SICP: polished + S i c by magnetron sputtering SKU: untreated + S i c by magnetron sputtering WP: polished + W by magnctron sputtering CUP: polished + Cu by magnetron sputtering MOP: polished + Mo by magnetron sputtering NBP: polished + Nb by magnetron sputtering TIP: polished + Ti by magnctron sputtering TAP: polished + Ta by magnetron sputtering PTP: polished + Pt by magnetron sputtering WCP: polished + WC-Tic by magnetron sputtering

I . Lapping with 1 p m diamond paste for IO minutes 12 To suppress cobalt by etching, Yes I , 2 ,7 Yes Several routes:

relieve residual stress by intermediate layer, and increase surface rouehness

2. Etching Co with HN03/H20 for I O minutes 3. Etching Co with HN03/HF/H20 for I O minutes

Ref.

- - by scratching 4. Etching with Murakami’s reagent for 30 minutes

+ scratching with 1 p m diamond paste for I O minutes

Table 4. (Continued)

Method Criteria of adhesion Cobalt Etching Roughness Novelty procedure Ref. No. enhancement etching solution optimization

(Table 5)

13

14

To suppress cobalt by No evaporation, increase surface roughness by heat treating

To suppress cobalt diffusion No by intermediate layers, and increase surface roughness by scratching

N I A No

NIA Yes

5. Route B + H2 plasma treatment at 800"C, P = 933 Pa

6. Route C + H2 plasma treatment to decarburize the

7. Route D + HN03 treatment for 15 minutes 8. Magnetron sputtering deposition of 0.1 p m tungsten

Heat treating the WC-Co for a period of time at a certain temperature and in a protective atmosphere that produced WC grain growth at the surface, increases surface roughness (> 0.6 p m Ra) without forming substrate porosity, and eliminates surface cobalt by evaporation I . Si3N4 and S i c intermediate layers with a thickness

(7 Torr), for 1 h, to decarburize the substrate

substrate

intermediate layer

of 6.8-9.6 p m were deposited on WC-Co substrates by hot-wall and cold-wall plasma assisted CVD (PACVD)

of 5.8 p m were deposited on WC-Co substrates by hot-wall PACVD using a pulsed dc glow discharge

carbon with a thickness of ( 1 0- 100 nm) were deposited on WC-Co substrates by a laser-arc process

2. TiN and Tic intermediate layers with a thickness

3. a-C intermediate layers containing 70% sp3 bonded

128,291 3 3 0 3 R

2. m CI, 3

Table 4. (Continued)

Method Criteria of adhesion Cobalt Etching Roughness Novelty procedure No. enhancement etching solution optimization

(Table 5 )

Ref.

15 To suppress cobalt by using a TiCN interlayer, enhance nucleation by diamond seeding, increase surface roughness, and prevent forming the brittle g-phase of WC

No N I A Yes

4. Each set of samples precoatcd with the intermediate layer, except those coated with a&, were dipped into a mixture of 50 p m diamond powder in n-hexane under ultrasonic irradiation

1. The substrate was sand-blasted and then ultrasonically cleaned in alcohol

2. An interlayer was deposited by arc-evaporation PVD at 800°C by introducing Ti and N ions combined with a substrate biasing voltage of 1 kV for 30 minutes

3. Bias decreased to 100-300 V accompanied by gradual decrease in the amount of N ions

4. The final composite was a TiCN initial layer followed by a layer containing gradually decreased amounts of C and N

particles followed by laser ablation formed an adherent Tic layer prior to diamond CVD coating

5. Depositing a seeding layer of nano-sized diamond

16 To suppress cobalt by Yes 9 Yes I . The substrates were polished with diamond

2. Then, the substrates were etched using 1 : I removal, reduce residual paste stress by multilayer deposition, increase chemical HF: HzO solution

133, 341

is ?

Table 4. (Continued)

Method Criteria of adhesion Cobalt Etching Roughness Novelty procedure Ref. No. enhancement etching solution optimization

(Table 5 )

17

bonding using compatible materials such as Si and Ti

To suppress cobalt, optimize Yes surface roughness by sand blasting, and improve nucleation

I O Yes

3. Then treated with two separate methods: A. Diamond particles with 20-9596 substratc

coverage were first deposited, followed by coating a 2.8 p m thick Ti coating using DC sputtering, and coating a 300 nm thick Si coating using an E-gun

B. Diamond particles with 20-9592 substrate coverage were first deposited, followed by coating a 2.8 p m thick Ti coating using DC sputtering, and coating a 300 nm thick Si coating using an E-gun. Then the samples were heat treated at 950°C for I h in a reducing atmosphere (N2 + 396H2) to transform amorphous Si to polycrystalline Si and to from Ti and Si carbides

then rinsed in deionized water

each group was ground and sandblasted with #I20 or #50O alumina sands, at a pressure of 2.5-3 N/cm2 for 30 minutes

i n an ultrasonic bath of ethanol, rinsed i n deionized water, and finally cleaned with HF acid and rinsed i n deionized water

I . Substrates were degreased in acetone and ethanol 1351

2. The substrates were divided into four groups where

3. After sandblasting, the substrates were suspended

F s 0

a F

c CL 0

Table 4. (Continued)

Method Criteria of adhesion Cobalt No. enhancement etching

18 To suppress cobalt by selective Yes peening and etching, increase surface roughness by water peening

Etching Roughness Novelty procedure Ref. solution optimization ;s (Fable 5) ?

N/S Yes 1 . Substrates with 2 wt% Co were treated using ~381 f ~

2

4 a high pressure jet of water peening P, r 2. The peening pressure and duration were optimized

to achieve best adhesion

N/S: data was not specified by the authors N/A: not applicable.

Next Page

Adhesion of CVD diamond to carbide cutting inserts 111

Table 5. List of chemical etching solutions used for cobalt removal

Etchant Etching solution No.

Reference

1 2 3 4 5 6 I 8 9

10 11 12 13 14

1 : 1 HNO3 : H20' [ 5 , 8, 24, 251 Murakami's solution: 10% KOH + 10% K3Fe(CN)6 + 80% H20b [26] 1 : 1 HNO3 : CH3COOH [111 H2 SO4 - H2 0 2 P O I HNO3 [261 1 : 3 HN03 : HC1 [261 HNO3/H20/HF [251 1 : 3 HN03 : H20 [201

HF [351 H2S04 solution [911 HC1 [921 9 : 1 H202 : H2SO4 [931 1 : 33 FeC13 : H 2 0 [931

1 : 1 HF:H20 [321

' Effective in removing Co from the surface up to depths from 5- 10 km. Etches the WC-Co composite and thus has less efficiency in selective Co removal.

A P U EHU EQ EQU EM EMU EMH Pretreatment [method]

Figure 10. The Co/W proportion at the substrate surface with different surface preparations. For a description of the preparation methods check Table 4, row 11 (adapted from Deuerler et al. [26]).

with a suspension of diamond powder in a solvent such as isopropanol to increase the surface roughness of the substrate [26, 141. Mehlmann et al. [14] conducted quantitative measurements to relate the nucleation density of CVD diamond parti- cles in the initial stages of deposition to the time of the ultrasonic treatment and the concentration of the diamond powder in the treatment solution. They showed that

Previous Page

112 M. A. Taher et al.

if the same amount of diamond suspension was kept in the ultrasonic field to pre- treat several samples, the nucleation density sharply decreased during the first 200 minutes, and then decreased slowly. If a fresh suspension was used for each pre- treatment, the nucleation density was shown to increase from 7 x lo7 particles/cm2 to 5 x lo8 particles/cm* with the increase of suspension particle concentration from 25 x lop5 g/ml to 25 x lop3 g/ml in the ultrasonic bath. In addition, for the same concentration of diamond powder in the ultrasonic bath, the nucleation density was found to increase as much as 1.5 x lo8 particles/cm2 with the increase in treat- ment time from 1 minute to 20 minutes. Beyond this time, the nucleation density remained unchanged.

Recently, blasting was employed to optimize the surface roughness of the sub- strate [35, 381. By using sand blasting, and surface energy calculations, Zhang and Zhou [35] reported that there existed an optimized substrate shape that eased the for- mation of diamond nuclei. Their analytical calculations revealed that the number of atoms required to form a critical nucleus was the smallest for a convex-shaped sub- strate. By controlling the sand blasting process parameters, such a substrate shape was achieved and exhibited the best adhesion characteristics. The parameters re- quired to reach such a shape included blasting of #500 Alumina sands at a pressure of 0.25 MPa for 30 minutes in a water carrier. Another similar process was reported by Tonshoff et al. [38]. However, in this case, water peening was used with no suspension of abrasive particles. A highly pressurized water jet caused local plas- tic deformation on the substrate surface and resulted in increased substrate residual stress. In addition, the peening process increased the substrate surface roughness and removed the smeared Co-binder phase from the surface. Water peening was conducted at different pressures ranging from 25 to 100 MPa and different peening times from 1 to 20 seconds with a nozzle diameter 1.5 mm. The optimum peening pressure and time were found to be 50 MPa and 40 seconds respectively. The re- sults showed that adhesion of diamond coatings improved with the increase in the substrate surface roughness due to the formation of better interlocking.

As observed, there appears to be a general trend of adhesion enhancement by optimizing the surface roughness of the substrate prior to diamond deposition. Sand blasting and water peening showed promising results that indicated that they could be the methods of choice for tailoring the surface roughness in the future.

4.3. Pre-diamond deposition heat treatment

Heat treatment of the carbide substrate was reported to improve adhesion by a few researchers [2, 18, 291. The mechanism by which adhesion was improved differs, however, in all cases some sort of surface reconstruction occurs. Saijo et al. [2] used heat treatment in a H2-02 plasma to decarburize the WC substrate prior to diamond coating. The decarburization resulted in the formation of a layer of metallic W. During the CVD diamond deposition process of the substrate, which was seeded by scratching of diamond particles, metallic W recarburized into finer WC grains trapping the seeded diamond particles. The trapped seeds offered anchoring sites for mechanical interlocking. Indentation showed a drastic improvement in adhesion

Adhesion of CVD diamond to carbide cutting inserts 113

when the heat-treated samples were compared to untreated ones. However, no data were reported about the exact temperature and duration of the heat treatment process. Yet in another study by Fan et al. [18] the same procedure was conducted in a H2 atmosphere and similar results were reported. The duration of the decarburization was in this case 15 minutes but again the heating temperature was not indicated.

Oles et al. [28] reported the application of a heat treatment process that was shown to greatly improve the adhesion of CVD diamond coatings to carbide inserts. The procedure involved heat-treating the WC-Co substrate for an undisclosed time and at a certain temperature in a protective atmosphere. As a result, Co was completely eliminated from the surface by evaporation. As Co evaporated, W and C precipitated on the existing WC grains causing growth at the surface and an increase in the substrate surface roughness ( R , > 0.6 pm). Diamond films deposited on these heat- treated substrates exhibited excellent machining characteristics when compared to untreated substrates. The authors mentioned that the selection of the heat-treating conditions had to be tailored according to the composition of the cemented carbide insert.

4.4. The application of interlayer materials

The application of interlayers of various materials as a substrate pretreatment method was shown to be an effective way to improve the adhesion of CVD diamond coatings to carbide inserts. The improvement in adhesion by using these layers is believed to be due to one or more of the following factors: suppressing cobalt, relieving the residual thermal stress, improving nucleation, increasing fracture toughness of interface, or increasing the chemical bonding. To obtain these enhancements an intermediate layer material was usually selected. The selection was based on one or more of the following properties: an intermediate value of CTE between WC-Co and CVD diamond, a low Co diffusion coefficient to minimize the thickness of the layer material, a high affinity to forming carbides, and a high fracture toughness if possible. Practically no material can meet all these requirements and this led to the diversity in selecting a suitable material. During the past decade, numerous researchers have reported improvement in adhesion using various interlayer materials such as: the metallic materials Cu, Mo, Nb, Ni, Pt, Ta, Ti, and W; the ceramics Si3N4, SIC, TIC, TIN, and WC; and the non-metallic materials B, and Si. Yet, despite this diversity, little information was provided about the reason for selecting a certain material. For comparison, the relevant physical and mechanical properties of the materials used as interlayers are listed in Table 6. In the following discussion some examples of the use of interlayers will be outlined.

4.4.1. Metallic interlayers. Interlayers of metallic elements have been shown to improve adhesion of CVD diamond coatings [9, 10, 12, 261. These layers were usually deposited with a thickness in the 0.5 pm- 1 p m range. The exact method used for depositing the intermediate layers differed, however, ion implantation, laser assisted CVD, and evaporation were among the techniques used to deposit such

Table 6. Propcrties of materials and chemical compounds used as interlayer materials

1 nterlayer Crystal Coefficient of Diffusion coefficient” (D) Solubility of C (wt%) based Carbide material struclure thermal expansion of Co in the material on phase diagram at T = 900°C

(m/m “C) [Ref.] listed at T = 900°C (cm2 h x 1 0 ) [Ref.]

and standard pressure

B RHOMB 8.3 (20-750°C) 1871 .? 0% 1951 B4C %

M 0 BCC 5.2-5.75 (20- 1000°C) [88] rn 0.14% [95] MoC ? Nb BCC 7.19-7.88 (20-1000°C) [88] 0.423 0.1% 1881 NbC, 2

Nb2C,Nb3C, $- Nb4C 2

Ni CPH & FCC 13.3- 16.3 (20-900°C) [88] 2.12 0.6% 1951 Yes (NiRC) !-

c u FCC 17.1 -20.3 (20-1000°C) 1881 1063 0.01% 1951 None

e,

metastable Pt FCC 9.1-10.2 (20-1000°C) 1881 1 .os ‘7 None Si D 2.8 (20- l00OC) 1881 ? 0% 1951 Sic Si3N4 ORTHO 2.9 (25-800°C) 1941 ? - -

Sic FCC (ZnS, 4.5 (25-800°C) [94] ? - -

sphaleri te) Ta BCC 6.6 (500°C) 1941 ,! 0% [95] TaC, Ta2C Ti CPH 8.8-9.9 (20-800°C) 1941 247.83 0.4% [95] TIC

Table 6. (Con ti nued)

lnterlayer Crystal Coefficient of Diffusion coefficient* ( D ) Solubility of C (wt%) based Carbide material structure thermal expansion of Co in the material on phase diagram at T = 900°C

(mlm "C) [Ref.] listed at T = 900°C (cm2 h x lo9) [Ref.]

and standard pressure

Tic FCC (NaCI) 6.9 (25-800°C) 1941 ? ~

~

Ti N FCC (NaCI) 9.0 (0-900°C) 1181 ? ~

-

W BCC 4.5-4.6 (20- 100°C) 1881 '? 0% 1951 WC, W2C WC HEX 4.7 (20-800°C) 1881 '7 - -

BCC: body centered cubic. CPH: close-packed hexagonal. D: diamond (two interpenetrating FCC lattices). FCC: face centered cubic. HEX: hcxagonal. ORTHO: orthorhombic. RHOMB: rhombohedral. *DiWussion coefficient calculated using the Arrhcnius relation 11 = A exp(-Q/RT), where A and Q were extracted from refercnce 1881

b R T 3. s

r)

E a. 4

116 M. A . Taher et al.

metallic layers. More detailed description of the procedures is given in Tables 3 and 4.

Saito et al. [9] deposited CVD diamond coatings on a cemented carbide tool that contained a few percents of a metal M belonging to group IVa or Va of the periodic table (for example Ti, Nb and Ta). Metals of these groups are known to form stable carbides thus offering good chemical attraction through possible C -C bonds. Prior to diamond deposition, the carbide tools to be coated were heat-treated in an inert gas. The surface morphology was compared to that of a carbide insert with no metal M but subjected to the same heat treatment process. This observation revealed a drastic change in the shape of the grains at the surface as shown in Fig. 1 1. For the M treated carbide tool, the surface of the insert was extremely rough with numerous voids that should offer excellent anchoring sites for diamond nucleation and growth. The carbide tool with no metal M showed a typical morphology with WC grains depleted of Co due to etching and exhibited many fine pores. The effect of enhancing the surface roughness can be seen in Fig. 12 where the grown diamond film was strongly anchored to a solid solution layer of M carbide and WC with no Co. The heat treatment process was believed to provide a mean of forming a solid solution through diffusion. The specific parameters used for this process were not laid out, however, the results indicated improved adhesion by indentation and machining.

Another use of metals from group IVa, Va or VIa of the periodic table was reported by the same group [lo]. In this case, however, the metal was introduced as an intermediate layer deposited by ion implantation. The thickness of the intermediate layer ranged from 0.1 to 3 pm. A heat treatment process followed the intermediate layer deposition to also form a Co-free solid solution of the metal carbide and WC, and to increase the surface roughness of the substrate. The formation of the metal carbide was confirmed by X-ray diffraction. They concluded that deposition of the metallic layer alone did not improve adhesion unless followed up with the heat treatment process. It can be speculated that without the heat treatment, the diamond coating would have been deposited directly on a smooth metallic layer that may not have had good adhesion with the substrate. Heat treatment, therefore, assured the formation of a solid solution by inter-diffusion and an increase in the solid solution grain size thus increasing the surface roughness. Since the metals used for these two studies were not identified, comparisons based on their physical properties such as Co migration and thermal stresses generated cannot be made.

Cappelli et al. [25], applied a 0.1 p m thick W layer by physical vapor deposition as an intermediate layer prior to diamond coating. Although no adhesion testing was conducted, W seemed to function well as an interlayer because it has a matching structure and a good diffusion barrier for Co migration. No data were found regarding the diffusion of Co in W and, therefore, its effectiveness as a barrier layer can only be justified by a similar metal in the same group of the periodic table namely Mo. Co shows a very low diffusion coefficient in Mo as given in Table 6. In one study by Kawadra et al. [SO] a 100 p m thick Mo layer acted as a successful diffusion barrier against Co diffusion.

Adhesion of CVD diamond to carbide cutting inserts 117

(a> Figure 11. SEM images of the surfaces of the heat-treated inserts according to Saito et al. [9]. (a) WC- Co insert, (b) WC-Co-M insert.

A more extended study was reported by Deuerler et al. [26] comparing the use of W, Cu, Mo, Nb, Ti, Ta and Pt as intermediate layers. For comparison with other non-metallic layers, the use of amorphous carbon (a-C), S i c and WC-10% TiN was examined as well. The study focused more on the effectiveness of these metals on enhancing the nucleation density rather than adhesion. However, a good nucleation is a necessary precursor to enhanced adhesion. The metallic interlayers were deposited using d.c. magnetron sputtering. All substrates were initially subjected to plasma etching using Ar and then mechanically polished by diamond powder with

118 M. A. Tuher et al.

(b) Figure 11. (Continued).

a grain size of 1 pm. The highest nucleation density was reported for the composite interlayer of WC-10% TIC followed by Pt, Ta, Ti, Mo, Nb, W with Cu exhibiting the worst nucleation density. This is in contradiction with the nucleation density reported on the same materials in bulk form [80, 421. The best adhesion obtained was with the use of the a-C layer rather than layers of the other metallic materials. a-C films (also known as diamond-like carbon) provide an increased number of chemical bonds with CVD diamond due to the chemical compatibility of DLC and CVD diamond.

Adhesion of CVD diamond to carbide cutting iiiserts 119

Figure 12. SEM image of the cross section of a diamond-coated WC-Co-M insert indicating the excellent mechanical interlocking at the interface [9].

A different application of metallic interlayers was reported by Tsai et al. [ 111. According to this study, a three-step process was conducted to coat WC-Co samples with CVD diamond coatings. In this case, diamond crystals were initially deposited using CVD to form a discontinuous coating. After that a layer of Ni was electroplated to fill the voids between the diamond crystals. Finally a continuous film of CVD diamond was grown on top of the initial crystals. The results indicated that well adherent diamond films were successfully deposited with a thickness up to 30 pm. The improvement in adhesion was attributed to two factors. The first

120 M. A. Tuher et al.

factor was the formation of a diffuse interface between the plated Ni layer and WC enhancing the chemical bonding. The second factor was the infiltration of the Ni binder into the gaps between the WC grains thus recovering the loss of strength due to Co depletion through etching. In addition, the authors suggested that Ni acted as a thermal stress relaxing interlayer. However, Ni has a higher CTE than WC-Co and diamond that would contradict this particular conclusion since higher thermal stresses would be expected to generate at the interface.

The concept of a stress relieve multilayer (SRM) was applied by Nesladek et al. [12]. The SRM consisted of a 0.04 p m thick sputtered layer of Nb (or W), a 2-25 p m thick stress absorbing layer of Ag, and a 0.04 p m thick top layer of Nb (or W). In the final stages of the CVD diamond coating process, the samples were heated to a temperature above the melting point of Ag to relieve the thermal stress build-up. This process was effective in improving the adhesion when tested by indentation. The choice of Ag was based on its popular application in metal joining cases where high interfacial stresses are expected. In addition Ag offers good wetting effect with diamond especially if alloyed with small amounts of refractory metal. An optimum value for the Ag thickness was found to be in the range of 2-6 pm. It could be presumed that an increase in thickness of the Ag layer increased the residual stress to a level much higher than the amount of relief that could be offered.

Care should be taken when selecting a metal for use as a barrier interlayer. Some metals can form more than one carbide compound some of which are mechanically brittle. For example, in the WC-Co system, an extremely brittle q-phase ( C O ~ W ~ C ) is known to form if there is a slight deficiency in the amount of carbon present. The brittleness of the q-phase can result in a weak interface and, therefore, poor adhesion.

4.4.2. Non-metallic interlayers. In the previous section, it was shown that metals were used as interlayer materials because of their strong mechanical properties. In addition metals have the potential of forming carbides at the interface thus enhancing the chemical bonding. Yet some non-metals can be good interlayer materials as well because of their capability to form carbides and a compatible crystal structure with diamond. Crystalline Si, for example, is an excellent candidate that could be used because it has the exact crystal structure as diamond, and in addition, it has a low CTE. However, Si is known to have weak mechanical strength and the deposition of crystalline Si is rather challenging. To correct for these two drawbacks, Lin et al. [32] designed two procedures by which Si could be used as an interlayer material. In the first procedure, after etching and polishing the carbide substrate, a discontinuous CVD diamond layer was deposited which only included scattered diamond particles. The samples were then coated with a Ti interlayer by DC sputtering to offer the mechanical strength necessary to compensate for silicon’s weak strength. Amorphous Si was then deposited using an electron gun. After that, the samples were coated again with a continuous layer of CVD diamond. This procedure showed optimum adhesion when tested by indentation for a Ti interlayer

Adhesion of CVD diamond to carbide cutting inserts 121

thickness of 2.8 pm, a Si thickness of 0.3 pm, and a 50% diamond coverage in the initial step. In the second procedure, no initial diamond deposition was performed, however, the same interlayers of Ti and Si were deposited on the substrate directly. The difference this time was to heat-treat the interlayer-coated samples in a reducing atmosphere of (N2 + 3%H2) at 950°C for 1 hour. As a result, the formation of T ic and crystalline Si was observed by X-ray diffraction. The CVD diamond coating that was then deposited exhibited better adhesion when compared to the first procedure. This study indicated the usefulness of acquiring the beneficial properties of a metal together with a non-metal to enhance the diamond coating adhesion.

Another application of non-metals was reported by Kubelka et al. [16]. The goal of their study was to suppress Co by the use of Si and B treatments instead of traditional chemical etching. Si and B layers were applied to carbide samples by embedding the substrates in silicon (or boron) 100 p m sized powder inside quartz ampoules under vacuum for 4-24 hours at a temperature of 700°C. Secondary ion mass spectrometry analysis (SIMS) revealed changes in the Co binder phase composition during CVD diamond deposition. Improved adhesion was reported by examining the indentation behavior of the coatings on Si and B treated and untreated samples. The adhesion enhancement was explained by the possible formation of stable boride and silicide compounds that decrease the Co vapor pressure and thus reduce the negative effect of metallic Co. However, a diamond coating with a thickness larger than 20 p m was shown to flake off regardless of the application of Si or B pre-deposition treatment.

4.4.3. Ceramics and other compounds. Ceramic materials have been used ex- tensively by several research groups as interlayer materials for enhancing diamond film adhesion to carbide inserts. One detailed study was conducted by Fan et al. [18] in which the effectiveness of TIC, TiN, WC, Sic, Si3N4 was examined. These materials were deposited in a multilayered fashion. Initially the samples were me- chanically polished with diamond paste, and then etched for 10 minutes to remove cobalt from the surface. A decarburization process explained in a previous section was then conducted. Prior to the application of the mentioned interlayer materials, an initial discontinuous CVD diamond coating was deposited having an average grain size of 0.5-1 pm. Then by using laser physical vapor deposition (LPVD) at a substrate temperature of 600°C, 0.5- 1 p m thick layers of Tic, TIN, WC, Sic and Si3N4 were deposited on the initially grown discontinuous layer. A final CVD diamond layer was deposited on each of the treated samples. To compare the adhe- sion of these samples, indentation and a specially designed grinding process were used. The results indicated that good adhesion was achieved by the use of the Tic multilayered structure where the critical indentation load that caused cracking in the coating was 45 kg (441 N). However, the other multilayered structures exhib- ited inadequate adhesion for different reasons. The enhanced adhesion of the TIC multilayered structure was attributed to the good adhesion of T ic to WC as well as to diamond due to compatible chemical bonding and structure. In addition, the difference in the CTE of Tic, WC and diamond is not large, yielding lower thermal

122 M. A. Taher et al.

stresses. The critical indentation loads associated with TiN, WC, S i c and Si3N4 were 35 kg (343 N), 20 kg (196 N), 20 kg (196 N) and 20 kg (196 N) respectively. The weaker adhesion provided by the TiN multilayered structure was explained by the lack of sufficient C-C bonding and a larger CTE mismatch between WC, TiN and diamond. For the WC interlayer, the poor adhesion was due to the formation of the brittle carbide of W, namely W2C. The low temperature of the LPVD process caused the deposition of S i c and Si3N4 in an amorphous structure rather than a crys- talline structure like Tic, TiN and WC. Therefore, the multilayered structures using S i c and Si3N4 demonstrated a weak interface and hence poor adhesion. Overall, in this study the use of all the multilayered structures resulted in better adhesion of the coating when compared to untreated carbide samples. The improvement was observed because these multilayered deposits improved the fracture toughness by reducing the stress intensity and accommodation of thermal strains.

Endler et al. [27] used 5-10 p m thick interlayers of Tic, TiN, S i c and Si3N4 as well to improve the adhesion of diamond coatings to carbide inserts. In this case, however, the multilayered structure described by Fan et al. [18] was not used. The interlayers were deposited directly on the substrate using plasma assisted chemical vapor deposition (PACVD) at 900°C. With this process, Si3N4 was deposited in a crystalline structure rather than an amorphous structure. As a result, the best adhesion of the CVD diamond coating was obtained with an interlayer of Si3N4. This was due to the lower CTE of Si3N4 as compared to TIC and TiN.

The studies described above indicate that the use of interlayers can be highly beneficial in improving the adhesion of diamond coatings. If these layers are to be applied commercially, then a set of systematic experiments should be conducted to optimize the thickness of the coating, the deposition temperature of diamond, and the heat treatment time and temperature if needed prior to coating. In addition, an awareness of the carbides and intermetallic compounds that could form under the optimized conditions is required to avoid secondary defects. One reminder is that metals and other compounds deposited by physical vapor deposition or sputtering are known to behave differently from their bulk state. In addition, the conditions of compounds formation or interaction are different from the thermodynamical equilibrium conditions at which phase diagrams are normally generated. Therefore, the use of phase diagrams of bulk systems should be used with caution.

5. ADVANCES IN ADHESION MEASUREMENTS

Quantitative measurement of the adhesion strength of diamond coatings is a matter of considerable consideration. Since adhesion has been associated with the strength of an interface [81], the purpose of any practical quantitative adhesion test method is to apply controlled loads capable of producing critical stresses at the interface to initiate interfacial failure rather than cohesive failure. In particular, for interfacial failure to occur, large peel and shear stresses are needed at the interface but with values that would not cause initial cohesive failure. One the most difficult tasks is to truly quantify adhesion in its fundamental meaning for diamond coatings. The

Adhesion of CVD diamond to carbide cutting inserts 123

difficulty arises from the nature of the physical properties of diamond. Any genuine adhesion test would certainly involve the direct contact of the diamond coating with another foreign material. The low chemical interaction of diamond with adhesives makes it difficult, for example, to apply the pull test because non-consistent adhesion would exist at the pull stud-diamond interface. The superior hardness of diamond makes it challenging to use any other material for the well-known scratch test. Due to these difficulties, and regardless of the extensive research conducted in the past decade, there has been no reliable quantitative adhesion testing procedure to evaluate the interfacial adhesion of diamond coatings to WC-Co cutting inserts. All the methods that have been developed so far are fundamentally different from each other and only compare the apparent practical adhesion qualitatively. In fact, the use of the word apparent is necessary in this context because what is being measured in not the practical adhesion defined earlier. The few widely reported adhesion testing procedures used for diamond-coated carbide inserts are the indentation method and physical machining. Other methods include sand blasting [37, 381 and the scraper test [I]. These methods will be discussed in more detail in the following sections. However, it is worth mentioning here that other testing procedures applied to diamond coatings on other substrate materials and geometries could also be used for carbide substrates. One method was a non-destructive evaluation procedure that used an ultrasonic angle beam technique [82] to compare the adhesion of several Si3N4 diamond-coated substrates. Another method was a non-contact compression test [83] by which large shear stresses were generated at the interface between a CVD diamond coating and a Si3N4 substrate. A uniaxial compression force was applied on the Si3N4 substrate. The size of the substrate in this case was larger than the coating. The results from this procedure indicated that the compressive load needed to debond the coating decreased with the increase of the CVD deposition temperature. A quantitative pull test was reported by Ramesham et al. [84] in which a Sebastian Five multipurpose tester was used to perform adhesion measurements. Pull studs were attached using epoxy to CVD diamond that was selectively grown on Si, Si3N4, Si20, A1203, Mo, and BN. Although the adhesion strength values were reported, all samples failed at the epoxy-diamond interface rather than the diamond-substrate interface. A similar test was conducted by Alam et al. [76] on W substrates in which some samples exhibited a mixed failure at the diamond-W interface as well as the epoxy-diamond interface. These two studies, although not applied to carbide tools, clearly indicate that the pull test, for example, cannot be a reliable adhesion test method.

The difficulty of designing a reliable adhesion test procedure for diamond coated- carbide cutting inserts is currently one of the major challenges in this technology. A test that can report reliable data will have to consider many factors and provide answers to several questions related to adhesion testing as indicated by Brown [81] as follows: (1) Whether temperature of the test is a true representation of the temperature at

which the coated carbide tool will service. As an indication of the service tem- peratures of diamond-coated tools Muller-Hummel and Lahres [ 851 reported

124 M. A. Taher et al.

quantitative temperature measurements at the diamond- chip interface during dry machining of a AlZnMgCu1.5-malleable alloy and a TiA16V4-malleable alloy. Their measurements indicated that the temperature could reach high val- ues ranging from 800°C for the AlZnMgCu1.5-malleable alloy to more than 1300°C for the TiA16V4-malleable alloy. Clearly at these temperatures, me- chanical and chemical coating- substrate interactions would be expected to oc- cur and have to be accounted for during any adhesion test.

(2) Some of the methods used for testing adhesion apply stress or energy unevenly to the coating- substrate interface causing an unacceptable measure. One result of this is that the stress or energy density at the locus of failure initiation is not really known. Therefore, the calculated value for the adhesion from the data obtained, using the conventional formulas for this case, may have an unacceptable, and often unrecognized, error associated with it.

(3) Are the samples truly representative of the coating-substrate system that will be in the practical application?

(4) How does the unmeasured residual stress within the coating affect the adhesion? These questions pose a big challenge for designing an adequate adhesion test.

However, they offer valuable guidelines for engineers and scientists attempting to propose innovative adhesion testing procedures. It is expected that within the coming few years, novel-testing procedures will emerge.

In the meantime, important adhesion test methods that were reported are discussed in the following sections. The reader should keep in mind that most of the described methods cannot truly quantify practical adhesion but rather compare the quality of adhesion with improved substrate pretreatment methods. A list of all adhesion evaluation methods found in the literature is given in Table 7.

5.1. Indentation

The indentation method has been implemented by several researchers [2, 5 , 8-10, 12, 18, 20, 24, 33 - 351 to evaluate qualitatively the adhesion of diamond coatings to carbide tool inserts. The most widely indentation procedure applied is the Rockwell indentation method using a 0.2 mm tip radius diamond spheroconical indenter. This process involves forcing the indenter into the surface of the coating under two specified forces, called preliminary and total test forces [86]. There are two general classifications of the Rockwell test: the Rockwell hardness test and the Rockwell superficial hardness test. In the Rockwell hardness test the preliminary test force is 10 kg (98 N). Total test forces are 60 kg (589 N) designated as Rockwell A, 100 kg (981 N) designated as Rockwell D and 150 kg (1471 N) designated as Rockwell C. In the Rockwell superficial hardness test the preliminary test force is 3 kg (29 N) and total test forces are 15 kg (147 N), 30 kg (294 N), and 45 kg (441 N). In using indentation as an adhesion evaluation technique the criterion used in most reports was the comparative size of the detached area that resulted from the indentation process. In some cases, mircohardness indentation using a pyramid Vickers indenter was reported instead. This technique, however, uses loads in the range of 1 to 1000 g

Table 7. List of reported adhesion test procedures applied to diamond-coated carbide inserts

Adhesion Adhesion Theory and procedure Advantages Disadvantages test method technique

1 Indentation Rockwell hardness, and Vickers microhardness indentation are applied to the sample and the size of the damaged area i s compared

2 Machining Physical machining using AI-Si alloys and other workpiece materials as detailed in Table 9

3 Scratch Scratch with an apparatus having a continuously

4 Hand Hand grinding using No. 180 SIC sand paper

5

increasing load and automatic recording

grinding Grinding Using a stainless steel block, the sample is rubbed

against a glass plate with diamond paste of grain size 5 wm for a specific amount of time. The area of retained coating is the criterion for adhesion comparison A highly pressured jet of particles is impinged on the sample. The time required for coating detachment (erosion) i s the adhesion criterion Scrapping of the coating using a sharp blade and recording the load. Using this load, the work of adhesion is calculated

7 Scraping

6 Blasting

Easy, fast

Most relevant to true service conditions

Easy, Fast

Easy, fast

Relatively easy, can offer good comparison if sever flaking is avoided

Selective targeting

Quantitative

Qualitative, costly, method of failure uncertain Qualitative, costly and categorizes wear rather than adhesion Qualitative, questionable reliability Qualitative, questionable reliability Qualitative

Evaluates abrasion resistance rather than adhesion Causes cohesive Failure, data scatter is large

126 M. A. Tuher et al.

250

0

-250 - -500 ‘ -750 z. a -1000

-1250

-1500

-1 750

-2000

-2250

6”

- 0 - Sy (X-ray Total)

+ Sy combined (Total X-ray + Indentaion)

0 0.02 0.04 0.06 0.08

Radius from point of application of indenter (mm) 0.1

Figure 13. The distribution of combined peel indentation stress and peel residual stress at the diamond-carbide interface of a diamond-coated carbide tool insert. Residual peel stress measured by X-ray diffraction and indentation peel stress determined by contact finite element modeling [57].

(0.01 - 10 N). Guseva et al. [33] and Konyashin et al. [34] used higher loads up to 100-200 N which far exceeds the limits of the standard Vickers indentation. The numerous applications of these two indentation methods are due to the ease of the process. A single test would take a few seconds to perform and the sample can be readily examined under a scanning electron microscope.

However, indentation has several disadvantages. The increased hardness and high surface roughness of the CVD diamond coating cause severe damage to the expensive diamond Rockwell and Vickers indenters. Also, an examination of the stresses generated during indentation revealed that this process did not give a true indication of adhesion. In a study by Taher et al. [57], contact finite element modeling was used to determine the distribution of stresses that originated from indentation. At the onset of cohesive failure in the diamond coating, the peel and shear stresses generated at the interface were recorded. Using superposition, the indentation peel and shear stresses at the interface were combined with the total residual peel and shear stresses as shown in Figs 13 and 14. The stresses were plotted starting from the point of application of the indenter on the coating surface and for the layer of elements present at the interface. As observed, a normal tensile peel stress (a)) required to classify adhesion did not occur with the use of the indentation test. Despite the occurrence of a high tangential shear (a,,) stress, it is concentrically distributed around the indentation point of application, and thus its effect will only emerge if brittle fracture occurs first in the coating material.

The interesting result of combining the total residual peel stress (al) with that of indentation was the increase observed in the normal compressive stress due to indentation from zero to a positive tensile value in the region away from the indenter. Yet it should be noted that regardless of how large the value of the peel stress became, it would have never exceeded the value of the peel stress already

Adhesion of CVD diamond to carbide cutting inserts 127

800

700

600

500

400

300

200

100

0

-e- Sxy (X-ray total ) h

0 n z v) rn 2 z

-100 1 0 0.02 0.04 0.06 0.08 0.1

Radius from point of application of indenter (mm)

Figure 14. The distribution of combined shear indentation stress and shear residual stress at the diamond-carbide interface of diamond-coated carbide tool insert. Residual shear stress measured by X-ray diffraction and indentation shear stress determined by contact finite element modeling [57].

available due to the total residual stress. This conclusion would always be true even if there were uncertainties in the measurement of the total residual stress by X-ray diffraction. It was concluded that if a CVD diamond coating deposited onto a cemented carbide insert survived the total residual stresses acted upon it due to thermal and intrinsic factors, interfacial failure would not occur if a high compressive stress such as that of indentation was applied to the coating. Failure will occur only if these high compressive stresses initiate cohesive failure in the diamond coating surface layer thus generating cracks that can travel to the interface as described earlier. Therefore, any adhesion test that subjects a diamond coating to a pure compressive stress such as indentation will not yield useful information regarding adhesion. Table 8 lists a summary of indentation methods used for qualitative adhesion testing of diamond-coated tools.

5.2. Machining

The ultimate test for a diamond-coated carbide tool insert is to examine its cutting performance by the use of controlled machining experiments. Since the early work conducted on the application of CVD diamond to carbide inserts, results from cutting experiments have been frequently reported. The criterion evaluating the cutting performance was the amount of wear experienced at the cutting edge in the flank area. The wear data were collected under a certain set of machining conditions and generally using a face turning operation on a lathe machine. The machining parameters included the cutting speed u , the feed rate f, the depth of cutting d , type of coolant, and the material being machined. The vast diversity in the conditions and workpiece materials applied during tests by various authors make a comparison of literature data difficult. For example, if the differences in cutting speeds are too

Table 8. Summary of indentation methods used for qualitative adhesion testing of diamond-coated carbide tools

Indentation Type of indenter Load range Criterion method

Remarks References

Brale Diamond cone indenter with 0.2 mm radius

Superficial Diamond cone Rockwell indenter with 0.2 mm

radius

Superficial Diamond cone Rockwell indenter with 0.2 mm

radius Rockwell A Diamond cone

indenter with 0.2 mm radius

Rockwell D Diamond cone indenter with 0.2 mm radius

15-45 kg (147-441 N)

15-45 kg (147-441 N)

15-45 kg (147-441 N)

60 kg (589 N)

100 kg (981 N)

The critical load for radial crack observation

The slope of the plot showing the lateral crack diameter vs. the indentation load Area of damaged region

Area of damaged region

Area of damaged region

0 The only semi-quantitative use of indentation

0 The damaged area was not concentric

0 The lateral crack diameter ranged from 100-450 p m

0 The radius of damaged area -100 pm

0 Reported as a Rockwell C 0 The damaged area was not

concentric 0 The radius of damaged

area: 150-400 pm 0 The damaged area was not

concentric 0 Brittle and semi-ductilc

failures were observed 0 The radius of damaged

area: 350-650 p m

18,321

% b

1121

2 e, -

12,5, I 0,20,351

1241

4

Table 8. (Continued)

Indentation Type of indenter method

ci 8 Load range Criterion Remarks References

Rockwell C Diamond conc indenter with 0.2 mm radius

Vi ckers Diamond pyramid with a 136" face angle

LL 3

150kg(I471 N) Arca of damaged region 0 The damaged area was not 191 concentric &

0 The radius of damaged 3 0

area: 250-500 wm

R ~ 2 0 0 - 2 5 0 p m E

diamond indentation 09

200-400 N Area of damaged region 0 Diagonal length was 133, 341 $ 1̂

"' 0 Good Vickers due to smooth

Table 9. List of rcported cutting experiments and parameters used for diamond-coated carbide inserts

Substrate Deposit Cutting conditions Perforrnancc

Composition L M Work mat. V .f d t tc FIA Vol.lt Vol. WID Ref.

we-4 wt% eo w e - 6 wt% co WC-6 wt% CO w e - ? wt% co we-? wt% co w e - ? wt% eo w e - ? wl% eo w e - ? WtYO c o w c - ? wt%1 eo wc-5.5 wt% co we-5.5 wt% co we-5.5 WtYO co wc-I 2 wt.% eo we- 17 wt% eo WC-6.7 ~ 1 % CO

4-5 8 8

12-15 '?

MW HF HF F&T >

AI- 18% Si A1-17% Si AI-l I% Si A1-18% Si A1-20% Si AI- 10% Si

AI-I8% Si AI- 18 wt% Si AI-16 wt% Si AI-16 wt% Si AIL16 wt% Si AI-I2 wt% Si AI-17 wt% Si AIL1 1 wt% Si- 3.8 wt% c u AIL18 wt% Si AIL17 wt% Si- 4.5 wt% c u A1-7-9 wt% Si- T6 AI-7-9 wt% Si- T6

'?

200 500 250 350 200

1000 800 300 350 330 500 500 250 605 250 350 200 762

0. I 0.1 0. I 0.08 0.1 0.1 0.3 0. I 0.077 2.5 2.5 2.5 0. I 0.1 0.1

0.5 1 .o I .5 0.7 0.5 0.5 0.3 1 .5 0.2 0.13 0.13 0.13 0.2 0.5 0.06

5 l o x 103 50 x 103 37 x 103 19 x 103 l o x 101 SO 103 72 x I O 3 45 x 101 5.39 x 103 107 103

s X 1 0 3

I62 x 10' 162 x 107

30.2 x 1 0 3 1.8 x 1 0 3

0.05 x 106 ~

-

200 90

9 > 340 2 > I O >> 10 > 30 >> 40 >> 2.5

~

~

2.0 x 106 4.5 x 106 ~

0.405 x IO6 > 1.83 x 106 0.22 x 106

>> 1.62 x 106

1.21 x 106

>> 4.5 x 106

> 1.62 x IO6

0.15 x IO6

7 7

? '?

? MW HF HF HF HF

? HF

7

-

0.9 1 0.8 ~

-

F

F + A F + A F + A

~ 10 7

12 25

? ?

')

~

2.5

l o x 103 61 x 103

2.5 x 106 1.22 x 106

we-? wt% eo w c - ? wt%1 co

20 7

MW MW

0. I 0.127

0.5 0.635

250 20

5 ~

F + A A

we-? wt% c o 6 HCDC 1850 2500 2150

0.25- 0.8 0.5

0.5-3 (I U F (I U D 1311

h h h h w e - ? wt% co 6 HCDC 0.8 1 .5

F + A w 1311

Table 9. (Continued)

Substrate Deposit Cutting conditions Performance

Composition L M Work mat. u s d t t , FIA Vol./t Vol. W / D Ref.

WC-? wt%Co 6 HCDC AI-l 1 wt% Si

WC-‘! wt% Co 6 HCDC Cu WC-5wt%Co 15 HF AIL19 wt% Si WC-6wt%Co 30 MW AIL18 wt% Si-

WC-6 Wt%J CO 30 MW AI- I8 wt% Si WC-2wt%Co 5 HF Mg-

6.5 vol.% Zn- 12 vol.% S i c

5 wt% Cu

< < b 2100- 0.35- 0.5- ‘ ( F w 1311 2600 0.65 1.0 2 300 0.15 1 d F + A d D 1311 2. 200 0.1 I > I 2 5 A 20 x 10’ 0.24 x IO6 D 1241 % 762 0.127 0.635 10.4 0.12 A 61 x 103 0.63 x IO6 W [291 3

f 762 0.127 0.635 43.6 0.14 A 61 x 10’ 2.7 x IO6 W 1291 c,

100 0.2 0.2 175 IOe F 4 x 103 0.7 x I06 D [3X] $ F R

0

Life of the diamond coated tool was described as thc machining of 140 truck wheels vs. 7 for uncoated carbide. 2 R ,. E

’’ Life of the diamond coated tool was described as thc machining of 2250 wheels vs. 690 for uncoated carbide. ‘Life of the diamond coated tool was described as the machining of 800 wheels vs. 200 for uncoated carbide. ’Life of the diamond coated tool was described as the finishing o f 4 armatures vs. 2 for PCD. L: diamond layer thickness (pm). 1): cutting speed (m/min). 2.

=. 3

00

m M: diamond deposition method. 1: feed rate (mm/rev). Y 2

F&T hot tilanient & combustion flame. HF: hot tilamcnt CVD. HCDC: high current direct current CVD. DC: direct current CVD. MW: microwave CVD. F/A: flaking/abrasion. WID: wetldry.

d: depth of cut (mm). t : cutting time ~ tool life criterion: 0.2 mm flank wear width (min). tc: cutting time for uncoated carbide to reach 0.2 mm flank wear (min). Vol./t: chip volume per time = IJ x f x d (mm’/min) also known as the Material Removal Rate (MMK). Vol.: total chip volume = u x j’ x d x t (mm’).

132 M. A . Tuher et al.

large, different wear mechanisms occur making meaningful conjectures practically impossible. However, a list is given in Table 9 that summarizes most of the results reported in the published literature for machining experiments conducted using diamond-coated carbide inserts. The criterion for life is the time it takes the tool to reach a flank wear of 200-254 pm. Note that in some cases that criterion was not reached within the time of the experiment and is indicated as such. Also, whenever appropriate, a comparative value of the time it takes an uncoated carbide tool to reach the same wear criterion is reported. One important observation that should be noted is that all these cutting experiments are not truly continuous. To record the progression of flank wear, the tools have to be removed from the machine and examined at specified time intervals. After that, the tool is reset and the machining process is resumed which gives time for both the tool and workpiece to cool down.

As improvements in substrate pretreatment advanced, by one or more of the methods mentioned earlier, the mode of failure of the diamond coating during machining showed a change from pure flaking to pure abrasion. In earlier studies by Saito et al. [2], Soderberg et al. [3] and Oakes et al. [5], the primary wear mechanism appeared to be flakmg due to inadequate adhesion. Wear failure by pure abrasion was reported by Inspektor et al. [17], Oles et al. [29] and Malshe et al. [75], which indicated far more improved adhesion. In contrast to early studies, the performance of CVD diamond-coated tools even surpassed that of brazed polycrystalline diamond (PCD) tools, as shown in Fig. 15. This ascertains the effectiveness of some pretreatment methods implemented such as heat treatment of the substrates [28, 291. The advantages of using machining, as an adhesion evaluation method, is that it offers direct information regarding the

43.6

CVD thin film (30pm thick)

v =762 mlmin, f = 0.127 mmlrev d= 0.635 mm, 15' Lead Angle, Flood Coolant, SPGN 120308

k Reynolds A390 Mahle 138

Figure 15. Machining performance in orthogonal cutting of hypereutectic (x 18% Si) aluminum alloys (adapted from Oles and Cackowski [29]).

Adhesion of CVD diamond to carbide cutting inserts 133

wear characteristics of a certain tool when placed in true manufacturing conditions. The comparative cutting performance characterized does not, however, quantify adhesion but rather quantifies wear life. Physical machining is a costly test because it requires thorough preparation, skilled personnel and expensive workpiece materials.

5.3. Erosion blasting

A new method used to evaluate the adhesion characteristics of diamond coatings on carbide inserts was introduced by Schneider et al. [37]. In this method, a sandblast erosion test was conducted on the coating-carbide system. The criterion for adhesion evaluation was the time required to cause detachment of the diamond coating in the blast spot area near the cutting edge. The critical time was termed the critical blasting time by Schneider et al. [37]. The blasting test was conducted using 90- 125 p m sized alumina particles (A1203) with a dynamic blasting pressure of 0.4 MPa and a nozzle diameter of 0.8 mm. Since the size of these particles was ten to twenty times larger than the coating thickness, fatigue was expected to occur within the substrate as well. The results of applying the blasting test indicated that the relative qualitative adhesion depended on the substrate pretreatment method and diamond coating thickness. The critical blasting time under the listed conditions ranged from 2 seconds to 130 seconds.

The blasting test is a simple procedure that can yield useful comparative adhesion results. The small size of the nozzle makes it versatile since it can be applied to different regions of the coating. The drawback of this testing procedure is that the detachment of the coating can be caused due to other factors not related to adhesion. For example, the diamond quality determines the hardness and hence the erosion resistance of the coating. As a result, what is being measured in the blast test can be a representation of the resistance to erosion rather than the adhesion strength.

Tonshoff et al. [38] applied the erosion test as well to determine the adhesion of diamond-coated carbide inserts. In this case, S i c particles with an average grain size d = 75 p m were blasted onto the coating with a pressure of 0.5 MPa and a mass flow rate of 10 g/min. The critical blasting time under these conditions varied from 15 to 90 seconds. In their work they clearly stated that cohesive failure of the diamond coating was the criterion of adhesion evaluation. This criterion contradicts the adhesion definition given earlier.

5.4. The scraper test

The scraper test was an early adhesion testing procedure reported by Murakawa and Takeuchi [l]. It involved the use of a sharp scraping blade that moved along the interface between the diamond coating and the substrate as shown in Fig. 16. The scraping motion started on the sample where no coating existed and then was continued along the interface for a specified distance. The scraping force was recorded and plotted against the distance that the blade moved. A typical force diagram related to this test is shown in Fig. 17. Note that the load defined as Pt,

134 M. A. Taher et al.

Substrate holder

Scraping distance (I)

Scraping load (P) Scraping speed = 5.7 mdmin

A yf B=1.2mm Blade width

(A- arrow view)

Figure 16. Schematic illustration for the scraping tester developed by Murakawa and Takeuchi [l].

Cemented carbide substrate h

Measurement range ( I , = 1 mm)

(a) 1, = 1 mm

Figure 17. Illustration of the scraping test on a diamond-coated carbide tool insert. (a) Relation between coating film and scraping area: S-S, start line of scraping operation; T-T, end line of scraping operation. (b) Example of force diagram: I , scraping distance. (Adapted from Murakawa and Takeuchi [ 11 .)

corresponded to the frictional resistance from the substrate during the initial motion over the coatingless region. Once contact with the coating occurred, the load P increased until it reached a state of steady scraping and then leveled to a constant value. Murakawa and Takeuchi [ 11 defined the scraping force F as a force or work

Adhesion of CVD diamond to carbide cutting inserts 135

per unit area, F = AP11/Bl1 = A P / B , where A P ( = P - Pb) was the net load required to scrape off the coating.

An important advantage of this test was that it offered a true quantitative measure of the adhesion as defined in this paper to a certain extent. However, the data collected showed a noticeable variation in the values of the scraping force that limits the use of this test to a restricted range of applications. For example, it cannot be used to compare the adhesion of samples with a varying coating thickness and quality. Upon testing a set of cemented carbide tools coated with 7- 11 p m thick diamond films the scraping force showed a tendency to increase with the increase in film thickness. This trend was believed to be due to the increase in the tear resistance along the stripping edge lines in the coating. Another questionable feature of this test is whether the scraping action is supplying the driving force for true interfacial failure or rather for cohesive brittle failure.

Despite some of the limitations of the prototype scraper test, it offers the most quantitative approach as compared to other state-of-the-art adhesion evaluation techniques. A modified setup of the same experiment that minimizes the uncer- tainties was not reported.

6. CONCLUSIONS AND FUTURE WORK

The issues that affect the adhesion of CVD diamond coatings to carbide inserts have been discussed. Mechanical issues such as residual stress, interface toughness, crack initiation and propagation and mechanical interlocking were shown to greatly affect adhesion. For example, residual stresses were shown to have both detrimental and beneficial effects. To control residual stresses proper selection of deposition conditions would be required. The interface toughness can be presumably improved by the use of interlayer materials. Also, the mechanical interloclung can be optimized by optimizing the roughness of the substrate surface.

The important chemical issue that affects adhesion was found to be the role of the Co binder in promoting non-diamond growth at the interface. This can be addressed by proper chemical etching of the substrate prior to diamond deposition or the use of diffusion barrier interlayers.

The adhesion of CVD diamond to carbide tools has been improved during recent years by the application of numerous novel methods. For example, heat treatment, the use of multilayered intermediate materials, surface grinding and blasting were successful adhesion enhancement methods reported.

To evaluate adhesion, testing procedures such as indentation, machining, erosion blasting and the scraper test were used. It was shown that these tests were effective in comparing the adhesion of diamond coatings, however they did not offer a quantitative evaluation.

It appears that the adhesion difficulty is being properly addressed, and that successful achievements have been reached. However, some key studies are still required in order to fully understand certain issues.

136 M. A. Taher et al.

It is suggested that a study should be addressed at evaluating the machining stresses that generate at the cutting tip of diamond-coated tools. Without the knowledge of these stresses, a definite adhesion criterion cannot be established. The machining stresses should also be combined with the residual stresses that exist prior to machining.

Important information about the exact adhesion mechanism of CVD diamond coatings to carbide tool inserts is also missing. Although it is known that mechanical interlocking and chemical bonding both occur, it is not clear which has the dominant contribution. Therefore, a systematic study in that area is needed.

Finally, no complete understanding of adhesion can be achieved without a true quantitative method to test it. The design of a consistent, quantitative adhesion testing procedure is highly necessary because with its use, all the remaining key issues described above can be properly addressed.

REFERENCES

1. M. Murakawa and S. Takeuchi, Thin Solid Films 181,443-450 (1989). 2. K. Saijo, M. Yagi, K. Shibuki and S. Takatsu, Surface Coatings Technol. 43/44, 30-40 (1990). 3. S. Soderberg, A. Gerendas and M. Sjostrand, Vacuum 41, 1317- 1321 (1990). 4. M. Murakawa and S. Takeuchi, Surface Coatings Technol. 49,359-365 (1991). 5. J. Oakes, X. X. Pan, R. Haubner and B. Lux, Surface Coatings Technol. 47, 600-607 (1991). 6. B. Lux, R. Haubner and P. Renard, Diamond Rel. Mater. 1, 1035-1047 (1992). 7. P. M. Stephan, R. A. Hay and C. D. Dean, Diamond Rel. Mater. 1,710-716 (1992). 8. T. H. Huang, C. T. Kuo, C. S. Chang, C. T. Kao and H. Y. Wen, Diamond Rel. Muter. 1, 594

9. Y. Saito, T. Isozaki, A. Masuda, K. Fukumoto, M. Chosa, T. Ito, C. E. Bauer, A. Inspektor and

10. T. Isozaki, Y. Saito, A. Masuda, K. Fukumoto, M. Chosa, T. Ito, E. J. Oles, A. Inspektor and

11. C. Tsai, J. C. Nelson, W. W. Gerberich, J. Herberlein and E. Pfender, Diamond Rel. Mater. 2,

12. M. Nesladek, J. Spinnewyn, C. Asinari, R. Lebout and R. Lorent, Diamond Rel. Mater. 2,

13. T. H. Huang, C. T. Kuo, T. S. Lin and C. S. Chang, Diamond Rel. Muter. 2, 928-932 (1993). 14. A. K. Mehlmann, A. Fayer, S. F. Dirnfeld, Y. Avigal, R. Porath and A. Kochman, Diamond Rel.

15. F.-M. Pan, J.-L. Chen, T. Chou, T.-S. Lin and L. Chang, J. Vuc. Sci. Technol. A 12, 1519-1522

16. S. Kubelka, R. Haubner, B. Lux, R. Steiner, G. Stingeder and M. Grasserbauer, Diamond Rel.

17. A. Inspektor, C. E. Bauer and E. J. Oles, Surface Coatings Technol. 68/69, 359-368 (1994). 18. W. D. Fan, X. Chen, K. Jagannadham and J. Narayan, J. Mater: Res. 9, 2850-2867 (1994). 19. H. Itoh, T. Nakamura, H. Iwahara and H. Sakamoto, J. Mater: Sci. 29, 1404- 1410 (1994). 20. P. X. Ling and G. Z . Ping, Thin Solid Films 239, 47-50 (1994). 21. W. D. Fan, K. Jagannadham and J. Narayan, Mater: Res. Soc. Symp. Proc. 363, 163-168 (1995). 22. D. G. Bhat, D. G. Johnson, A. P. Malshe, H. Nasem, W. D. Brown, L. W. Schaper and C.-H. Shen,

23. P. Scardi, S. Veneri, M. Leoni, R. Polini and E. Traversa, Thin Solid Films 290-291, 136-142

(1992).

E. J. Oles, Diamond Rel. Mater. 2. 1391 (1993).

C. E. Bauer, Diamond Rel. Mater. 2, 1156- 1159 (1993).

617-620 (1993).

98-104 (1993).

Mater. 2, 317-322 (1993).

(1994).

Mater. 3, 1360-1369 (1994).

Diamond Rel. Mater. 4,921-929 (1995).

(1996).

Adhesion of CVD diamond to carbide cutting inserts 137

24. M. A. Taher, W. F. Schmidt, W. D. Brown, S. Nasrazadani, H. A. Naseem and A. P. Malshe,

25. E. Cappelli, F. Pinzari, P. Ascarelli and G. Righini, Diamond Rel. Mater. 5,292-298 (1996). 26. F. Deuerler, H. van den Berg, R. Tabersky, A. Freundlieb, M. Pies and V. Buck, Diamond Rel.

27. I. Endler, A. Leonhardt, H.-J. Scheibe and R. Born, Diamond Rel. Mater. 5, 299-303 (1996). 28. E. J. Oles, A. Inspektor and C. E. Bauer, Diamond Rel. Mater. 5, 617-624 (1996). 29. E. J. Oles and V. J. Cackowski, in: Proceedings of the 1st International Diamond Symposium,

30. V. Weihnacht, W. D. Fan. K. Janannadham. J. Narayan and C.-T. Liu, J. Mater Res. 11,

31. J. Karner, M. Pedrazzini, I. Reineck, Me. E. Sjostrand and E. Bergmann, Mater. Sci. Eng. A 209,

32. C. R. Lin, C. T. Kuo and R. M. Chang, Thin Solid Films 308-309, 273-278 (1997). 33. M. B. Guseva, V. G. Babaev, V. V. Khvostov, G. M. Lopez Lundena, A. Yu. Bregadze,

34. I. Y. Konyashin, M. B. Guseva, V. G. Babaev, V. V. Khvostov, G. M. Lopez and A. E. Alexenko,

35. B. Zhang and L. Zhou, Thin Solid Films 307, 21-28 (1997). 36. M. A. Taher, W. F. Schmidt, W. D. Brown, S. Nasrazadani, H. A. Naseem and A. P. Malshe,

J. Mater Sci. 33, 173-182 (1998). 37. D. Schneider, B. Schultrich, P. Burck, H.-J. Scheibe, G. Jorgensen, M. Lahres and J. Karner,

Diamond Rel. Mater. 7, 589-596 (1998). 38. H. K. Tonshoff, A. Mohlfeld, C. Gey and J. Winkler, Surface Coatings Technol. 116, 440-446

(1999). 39. (a) K. L. Mittal, in: Adhesion Measurement of Thin films, Thick films, and Bulk Coatings,

K. L. Mittal (Ed.), STP 640, pp. 5-17. ASTM, Philadelphia (1978). (b) K. L. Mittal, in: Adhesion Measurement of Films and Coatings, K. L. Mittal (Ed.), pp. 1-13. VSP, Zeist, The Netherlands (1995).

Surface Coatings Technol. 86-87, 678-685 (1996).

Mater. 5, 1478- 1489 (1996).

D. S. Lim and H. H. Koo (Eds), pp. 91- 108. The Consortium, Seoul, Korea (1996).

2220-2230 (1996).

405-413 (1996).

I. Y. Konyashin and A. E. Alexenko, Diamond Rel. Mater. 6, 89-94 (1997).

Thin Solid Films 300, 18-24 (1997).

40. M. D. Thouless, IBM J. Res. Dev. 38, 367-377 (1994). 41. M. Murakawa and S. Takeuchi, J. Adhesion Sci. Technol. 9, 695-710 (1995). 42. S. S. Perry, S. P. McGinnis and G. A. Somorjai, J. Adhesion Sci. Technol. 9,711-724 (1995). 43. M. A. Taher, Master Thesis, University of Arkansas, Fayetteville, AR, USA (1995). 44. M. A. Younis, Trans. ASME, J. Eng. for Industry 114, 377-385 (1992). 45. Z. C. Lin and C. C. Liu, Intl. J. Machine Tool Manuf. 36,73-88 (1996). 46. H. Windischmann, G. F. Epps, Y. Cong and R. W. Collins, J. Appl. Phys. 69, 2231-2237 (1991). 47. H. Windishmann and K. J. Gray, Diamond Rel. Mater. 4, 837-842 (1995). 48. S. Chatterjee, A. G. Edwards and C. S. Feigerle, J. Mater Sci. 32, 3355-3360 (1997). 49. P. R. Chalker, A. M. Jones, C. Johnston and I. M. Buckley-Golder, Surface Coatings Technol.

50. V. G. Ralchenko, E. D. Obraztsova, K. G. Korotoushenko, A. A. Smolin, S. M. Pimenov and V. G. Pereverzev, Mater. Res. SOC. Symp. Proc. 383, 153- 158 (1995).

51. V. G. Ralchenko, E. D. Obraztsova, K. G. Korotoushenko, S. M. Pimenov, A. A. Smolin, V. G. Pereverzev, Y. V. Lakhotin and E. N. Loubnin, in: Applications of Diamond and Related Materials: Third International Conference, A. Feldman, Y. Tzeng, W. A. Yarbrough. M. Yoshikawa and M. Murakwa (Eds), pp. 635-638. National Institute of Standards and Technology (NIST), Washington DC, USA (1995).

47, 365-374 (1991).

52. C. T. Kuo, C. R. Lin and H. M. Lien, Thin Solid Films 290-291, 254-259 (1996). 53. H. Mohrbacher, K. Van Acker, B. Blanpain, P. Van Houtte and J.-P. Celis, J. Mater. Res. 11,

1776- 1782 (1996).

138 M. A. Taher et al.

54. M. A. Taher, J. L. Shultz, S. Nasrazadani, H. A. Naseem, W. D. Brown and A. P. Malshe, in: Proceedings of the Fourth International Symposium on Diamond Materials. Vol. 95-4, pp. 614-623. The Electrochemical Society, Pennington, NJ (1995).

55. B. D. Cullity, Elements of X-ray Dzffraction, 2nd edition. Addison-Wesley (1978). 56. J. Gunnars and A. Alahelisten, Surface Coatings Technol. 80, 303-312 (1996). 57. M. A. Taher, W. F. Schmidt and Ajay P. Malshe, in: Proceedings of 1998 ASMEIMED Inter-

national Mechanical Engineering Congress and Exposition. American Society of Mechanical Engineers, New York (1998).

58. M. Yoshikawa, G. Katagiri, H. Ishida, A. Ishitani, M. Ono and K. Matsumura, Appl. Phys. Lett. 55. 2608 (1989).

59. B. S. Berry, W. C. Prichet, J. J. Cuomo, C. R. Guarnieri and S. J. Whitehair, Appl. Phys. Lett. 57,

60. N. S. Van Damme, D. C. Nagle and S. R. Winzer, Appl. Phys. Lett. 58, 2919-2920 (1991). 61. H. Guo and M. Alam, Thin Solid Films 212, 173- 179 (1992). 62. J. A. Baglio, B. C. Farnsworth, S. Hankin, G. Hamill and D. O’Neil, Thin Solid Films 212,

63. P. K. Bachmann, H. D. Bausen, H. Lade, D. Leers, D. U. Wiechert, N. Herres, R. Khol and

64. A. B. Harker, D. G. Howitt, S. J. Chen, J. E Flintoff and M. R. James, Proc. SPIE 2286,254-261

65. W. Zhu, R. C. McCune, J. E. devries, M. A. Tamor and K. Y. Simon Ng, Diamond Rel. Mater.

66. D. Rats, L. Bimbault, L. Vandenbulcke, R. Herbin and K. F. Badawi, J. Appl. Phys. 78

67. P. Scardi, S. Veneri, M. Leoni, R. Polini and E. Traversa, Thin Solid Films 290-291, 136- 142

68. S. K. Choi, D. Y. Jung and H. M. Choi, J. Vac. Sci. Technol. A 14, 165- 169 (1996). 69. W. L. Wang, M. C. Polo, G. Sanchez, J. Cifre and J. Esteve, J. Appl. Phys. 80, 1846-1850

70. E. Liu, B. Blanpai, J. P. Celis, J. R. Roos, G. Alvarez-Verven and Th. Priem, Surface Coatings

71. Y. Nakamura, S. Sakagami, Y. Amamoto and Y. Watanabe, Thin Solid Films 308-309, 249-253

72. W. D. Fan, K. Jagannadham and J. Naryan. Matel: Res. Soc. Symp. Proc. 356, 847-852 (1995). 73. R. E. Reed-Hill and R. Abbaschian, Physical Metallurgy Principles, 3rd edn. PWS-Kent, Boston

74. L. Chandra, M. Chhowalla, G. A. J. Amaratunga and T. W. Clyne, Diamond Rel. Mater. 5,

75. A. P. Malshe, M. A. Taher, A. Muyhshondt, W. F. Schmidt, H. Mohammed and H. Mohammed.

76. M. Alam, F. He, D. E. Peebles, J. A. Ohlhausen and D. R. Tallant, J. Adhesion Sci. Technol. 9,

77. M. Murakawa and S. Takeuchi, H. Miyazawa and Y. Hirose, Surface Coatings Technol. 36, 303

78. A. K. Mehlmann, S. F. Dimfeld and Y. Avigal, Diamond Rel. Mater. 1, 600 (1992). 79. B. S. Park, Y.-J. Baik, K.-R. Lee, K. Y. Eun and D. H. Kim, Diamond Rel. Mater. 2,910 (1993). 80. M. Kawadra, K. Kurihara and K. Sasaki, Diamond Rel. Mater. 2, 1083 (1993). 81. S. D. Brown, J. Adhesion Sci. Technol. 8, 687-711 (1994). 82. M. Saka, S. Sat0 and H. AbC, NDT&Elntl. 30, 305-311 (1997). 83. R. Rozbicki, V. L. Rabinovich and V. K. Sarin, J. Adhesion Sci. Technol. 9, 737-751 (1995). 84. R. Ramesham, T. Roppel, R. W. Johnson and J. M. Chang, Thin Solid Films 212,96- 103 (1992). 85. P. Muller-Hummel and M. Lahres, Diamond Rel. Mater. 4, 1216- 1221 (1995).

302-303 (1990).

180-185 (1992).

P. Koidl, Diamond Rel. Mater. 3, 1308- 1314 (1994).

(1994).

4, 220-233 (1995).

4994-5001 (1995).

(1996).

(1996).

Technol. 80, 264-270 (1996).

(1997).

( 19 9 2).

674-681 (1996).

in: Proceedings of NAMRAC Conference, p. 254. Society of Manufacturing Engineers (1998).

653-679 (1995).

(1988).

Adhesion of CVD diamond to carbide cutting inserts 139

86. Standard Test Methods for Rockwell Hardness and Rockwell Superjicial Hardness of Metallic

87. M. Bauccio (Ed.), ASMMetals Reference Book, 3rd edn. ASM International, Metals Park, Ohio,

88. E. A. Brand (Ed.), Smithell Metals Reference Book, 6th edition. Butterworth & Co., London, UK

89. M. A. Prelas, G. Popovici and L. K. Bigelow (Eds), Handbook of Industrial Diamond and

90. M. G. Peters and R. H. Cummings, European Patent 0519587 A1 (1992). 91. K. Shibuki, M. Yagi, K. Saijo and S. Takatsu, Surface Coatings Technol. 36, 295 (1988). 92. Y. Saito, K. Sato, S. Matuda and H. Koinuma, J. Mater. Sci. 26, 2938 (1991). 93. K. J. Grannen, F. Xiong and R. P. H. Chang, Surface Coatings Technol. 57, 155 (1993). 94. ASM Engineered Materials Reference Book. ASM International, Metals Park, Ohio, USA (1989). 95. H. Baker (Ed.), ASM Handbook, Vol. 3, Alloy Phase Diagrams, Second printing. ASM

96. R. F. Davis, Diamond Films and Coatings. Noyes Publications, Park Ridge, NJ (1993). 97. T. Yashiki, T. Nakamura, N. Fujimori and T. Nakai, Surface Coatings Technol. 52,81-86 (1992).

Materials, ASTM Publication E l 8-93 (1993).

USA (1993).

(1983).

Diamond Films. Marcel Dekker, New York (1998).

International, Metals Park, Ohio. USA (1997).

Adhesion Aspects ofThihin Films, Vol. I , pp. 141-158 Ed. K. L. Mittal 0 VSP 2001

Adhesion improvement of diamond films to silicon nitride substrate for cutting tools

HIDEAKI ITOH ',*, RYO SASAI ', MASAHIRO KAMIYA ', SUNG-SO0 LEE ', KOTARO KURODA and TAKAHIRO TSUTSUMOTO

Research Center for Advanced Waste and Emission Management, Nagoya University, Chikusa-ku, Nagoya, 464-8603, Japan Department of Materials Science and Engineering, Graduate School of Engineering, Nagoya Universiw, Chikusa-ku, Nagoya, 464-8603, Japan Hiroshima Prefectural Western Industrial Research Institute, 2-10-1, Agaminami, Kure, Hiroshima 737-0004, Japan

Abstract-A high adhesion of the diamond film to the substrate is required for application of diamond coated cutting tools. A procedure to grow an adherent diamond film onto a pretreated sintered silicon nitride substrate was developed by a two-stage microwave plasma CVD technique using the CO and H2 reactant mixture. The substrate was pretreated in a hot and strong acid solution and then microflawed ultrasonically with diamond grains suspended in ethanol. An anchored deposition of diamond into the micropores of the acid-treated substrate resulted in an excellent adhesion between the diamond film and the silicon nitride substrate. The insertion of diamond-like carbon (DLC) intermediate layer was also attempted to further improve the adhesion strength. The graded texture and structural variation of C-C bonding from the substrate side to the diamond film were clearly characterized by SEM/XMA and TEM analyses or micro-Raman spectroscopy. Adhesion was found to increase considerably in the presence of a DLC interlayer, which was confirmed by a compression topple test. X-ray stress measurements suggested a relaxation of large residual stress in the interfacial region between the diamond film and the substrate. The actual cutting test by milling the work material, A1-20 wt% Si alloy, revealed a significantly long tool life by using a thick diamond coated specimen.

Keywords: Diamond film; adhesion improvement; silicon nitride; cutting tool.

1. INTRODUCTION

Adhesion improvement of CVD diamond films to a substrate material is still one of the crucial problems to be solved especially for application of diamond coatings to cutting tools [ 1 - 101. Moreover, the adhesion of diamond films to a cemented

"To whom correspondence should be addressed. Tel.: +81-52-789-5854; Fax: +81-52-789-5853; E-mail: hitoh@rescwe.nagoya-u.ac.jp

142 H. Itoh et al.

carbide substrate has not been enough to provide a wide range of commercial cutting tools with reliable performance and long life [1, 2, 4-7, 101. A major cause of low adhesion strength has been believed to be a large thermal stress induced during the cooling process of the heated specimen after the CVD treatment. The bond strength between the diamond film and the substrate material is too weak to endure the residual stress and the mechanical shock from the work material during the cutting process.

So, a procedure to prepare highly adherent diamond films to pretreated silicon nitride substrates was developed by the present authors [ 1 1 - 171. The formation of a graded interfacial microstructure including the anchored diamond deposition was found to be effective in relaxing the large residual stress between the diamond film and the substrate. The insertion of a diamond-like carbon (DLC) interlayer further increased the adhesion strength of the diamond film. Further study has been carried out to clarify the mechanism of adhesion improvement. In the present paper, our earlier work is briefly reviewed, and the recent experimental results on the interfacial TEM observation and the adhesion evaluation are described as well as the discussion of adhesion improvement mechanism using X-ray residual stress measurements.

2. EXPERIMENTAL

Figure 1 shows the procedure for the preparation of a diamond coated specimen. The diamond film was grown by the microwave plasma enhanced CVD using a reactant gas mixture of CO and H2 [9, 181. Commercially available silicon nitride (p-Si3N4) sintered compact for cutting tools was used as substrate. A triangular chip with edge side length of 16 mm and thickness of 3.0 mm was mainly used for cutting test, and the chamfer angle and chamfer length were 30 degrees and 0.15 mm, respectively. The substrate was pretreated in a strong acid, 60% HN03 and 47% HF (1 : 1) at 25-40°C for 30-60 min for preferential etching of sintering additives (sialons or others). Then the acid-treated substrate was subjected to the first microflawing treatment [ 181 in diamond suspended ethanol in an ultrasonic filed (200 W, 30 min) for increasing the nucleation sites. The DLC interlayer was formed by r.f. plasma CVD using a CH4-H2 reactant gas [19]. After the second microflawing treatment on the DLC surface, the diamond was grown onto the DLC coated substrate by a two-stage CVD process [11, 121, which consisted of the first CVD of fine diamond grains (microwave power: 750 W, total pressure: 2 kPa and CO concentration: 10 ~01%) and the second higher rate CVD of a thick diamond film (microwave power: 550 W, total pressure: 4 kPa and CO concentration: 25 ~01%).

The deposited material and the substrate were identified by X-ray diffraction (XRD) and micro-Raman spectroscopy. The surfaces and cross sections of the specimens were observed by a scanning electron microscope (SEM) and X-ray microanalyzer (XMA). The microstructure in the interfacial region between the diamond film and the substrate was also observed by using a high resolution transmission electron microscope (TEM). The TEM specimens were fabricated by

Adhesion improvement of diamondjlms to silicon nitride 143

sintered compact

Acid treatment

1 Microflawing ( I ) 1 - Microflawing (11)

by microwave plasma CVD Two-stage diamond coating

r-----I Diamond coated specimen

Figure 1. Procedure for preparation of diamond coated specimen

a focused ion beam technique [17]. The practical adhesion [20-221 of the diamond film was determined by an edge compression test or a compression topple test of the coated specimens [23]. The tool life and the cutting performance of diamond coated specimens were evaluated by a milling test using A1-20 wt% Si as the work material under the conditions of cutting speed 393 m/min, depth of cut 0.25 mm, feed rate 0.1 mm/rev and spindle revolution 2500 rpm [12, 131.

The X-ray residual stress measurement was performed using chromium as the target material. The measured diffraction line was 220 at the diffraction angle 28 = 130.42 degree.

3. RESULTS AND DISCUSSION

3.1. Effects of the pretreatment and CVD conditions on diamond coating properties

After the pretreatment of silicon nitride substrate, many small pits were formed by depletion of an amorphous binding phase such as sialon and yttria. Even the needle shaped crystals of B-Si3N4 were detached from the substrate in the case of pretreatment at 40°C for 60 min due to an excess corrosion of the binders. The number and depth of such pits or micropores were found to increase with increasing etching temperature and duration. The average depth of the pits was estimated to be in the range of 5-10 p m [12].

Microflawing pretreatment was carried out after the acid treatment. The nucle- ation sites for diamond deposition were considerably increased in number [ 181 by the collision of fine diamond grains into the inner parts of the micropores formed

144 H. Itoh et al.

80

n E

2 60

E i? 20

0

I ' I ' I

I 1 I

10 20 30

Reaction time (h)

Figure 2. Variation of diamond film thickness with total CVD reaction time.

by the acid treatment. A part of the fine diamond grains used for microflawing treatment remained in the micropores, which had a positive effect on increasing the nucleation density.

A two-stage CVD was performed to control the microstructure and deposition rate of the diamond film after the above pretreatment of the substrate [ 1 11. In the first stage CVD, the diamond was found to deposit from the inside of the micropores of the pretreated substrate with a simultaneous deposition on the surface of p-Si3N4 needle crystals, then filling the pores with tiny diamond grains. At the reaction time of 60 min, the whole area of the surface was apparently occupied by diamond grains. Finally, a dense and uniform thin film which was similar to that grown on an unetched substrate was formed in 90 min.

A thickening of the diamond film was attempted by the second stage CVD. Figure 2 shows the relationship between the diamond film thickness and the total CVD reaction time (the first stage is for 1 h and the rest is the second stage). The film thickness at the edge part of the substrate was greater than that in the central part with increasing the CVD treatment time. It is also noted that the growth rate of diamond film increases with an increase in the treatment time. This is caused by the higher columnar growth rate of preferentially oriented diamond crystallites.

Figure 3 shows the SEM micrographs of cross sections of a diamond coated spec- imen, which was prepared by the acid treatment at 25°C for 30 min and the total CVD reaction time of 10 h. Etch pits after the removal of diamond anchors from the substrate can be seen on the fractured relief surface, as shown in Fig. 3a. An adherent and tight diamond film with thickness of 12 p m was formed at the edge corner of the cutting insert, as shown in Fig. 3b.

Adhesion iniprovement of diamond films io silicoii nitride 145

Figure 3. Scanning electron micrographs of cross sections of the diamond coated specimen. Acid treatment: 25"C, 30 min. CVD reaction time: the first stage for 1 h and the second stage for 9 h. (a) Fractured surface of the relief side, (b) polished cross section of the edge part.

146 H. Itoh et al.

3.2. Microstructure of the inte flacial region between the diamondJilm and the substrate

Figure 4 shows the XRD patterns of diamond coated specimens with different CVD treatment times. Diffraction lines for 111 and 220 of the diamond were identified with their intensities increasing with increasing treatment time. Other diffraction lines were attributed to those for silicon nitride sintered substrate. A crystalline phase such as silicon carbide was not identified. Preferred orientation to 220 direction of diamond increased with increasing treatment time, which was apparently related to the development of a columnar structure as seen in Fig. 3b.

Figure 5 shows the micro-Raman spectra at various positions on the cross section of the diamond coated specimen prepared with CVD treatment time of 30 h. In the spectra of positions 0 and @ within the diamond film, one can see the typical diamond peaks at 1333 cm-' and broad peaks corresponding to a small amount of amorphous carbon around 1500 cm-'. Analogous peaks can be confirmed at position 0 near the interface in the substrate side, which verifies the anchored deposition of diamond into the micropores. Only a weak diamond peak can be observed at position 8, which is located at about 10 p m inside from the interface. At position @ about 30 p m inside from the substrate, a broad peak of DLC, which shifts slightly to a higher wavenumber, is seen, as well as the silicon nitride peaks in the wavenumber range lower than 1000 cm-' (the spectrum of silicon nitride at far inside of the substrate is shown in a). It is noted that no peak corresponding to silicon carbide is observed in any of the spectra. Therefore, the results suggest

20 30 40 50 60 70 80

2 8 , degree (Cu K,) Figure 4. X-ray diffraction patterns of diamond coated specimens with CVD treatment times. (a) Si3N4 sintered substrate untreated, (b) 2 h, (c) 10 h, (d) 20 h, (e ) 30 h.

Adhesion improvement of diamoad~7lms io silicon nitride 147

I I I I I I I I I I

1600 1400 1200 1000 80C Raman shift (cm-’)

Figure 5. Micro-Raman spectra at the positions specified on the cross section of a diamond coated specimen in the upper SEM photograph. CVD treatment time: 30 h. @ spectrum of the silicon nitride substrate in a bulk region.

148 H. Itolz et al.

that carbon atoms of the anchored diamond in the etched micropores diffuse into the substrate side during a long duration of the second stage CVD, when strong C-C bonds as contained in the DLC are formed in the bulk region up to 30 p m depth within the substrate.

Figure 6 shows cross-sectional TEM micrographs in both bright and dark fields in the interfacial region between the silicon nitride and the diamond. The contact between the diamond and the B-Si3Nd needle-shaped crystals (the diffraction pattern is shown on the right hand below) seems to be excellent with no void seen around the P-Si3N4 crystals. Small micropores were observed in some places, however, where particles of DLC were identified as amorphous carbon by electron energy loss spectroscopy (EELS) [17]. The TEM images of the diamond crystal in the boundary region are shown in Fig. 7. Many planar defects can be seen as a streak pattern in the diamond crystal, which suggests that diamond growth is accompanied by the formation of micro-twins and stacking faults during the CVD treatment.

3.3. Tool life and cutting performance of diamond coated specimens

Figure 8 shows the tool life plotted against the CVD treatment time, where the tool life is defined here as the number of passes in the cutting amount unit (2.4 cm3) of the work material (Al-20 wt% Si alloy) under milling conditions described in Section 2. The plots show the number of passes when the diamond film peeled from the substrate or the flank wear of 0.15 mm was attained without peeling. It was found that diamond film peeled off abruptly halfway during the milling test on the specimens prepared for the CVD treatment time up to 20 h. The adhesion between the diamond film and the substrate was too weak to mill the A1-20 wt% Si work material. As regards effects of the acid treatment, a longer acid pretreatment in a hotter solution seems to be more effective for increasing the adhesion. On the other hand, neither peeling nor chipping occurred after the milling test in cases A and B. These facts show that a more adherent diamond film can be formed on a silicon nitride substrate by a more aggressive acid pretreatment and a longer CVD treatment time. A considerable increase in the tool life after treatment times of 30 and 50 h verifies normal abrasive behavior of the diamond film.

Figure 9 shows the cutting performance of diamond coated specimens prepared at various treatment times. Flank wear is plotted as a function of the cutting amount of the work material. The cutting amount is found to increase with increasing CVD treatment time up to 50 h.

3.4. Effects of the DLC interlayer on the adhesion strength of the diamondjlm to the substrate

The effects of inserting a DLC layer between the diamond film and the pretreated substrate were investigated. The DLC deposited in the beginning into the micro- pores of the acid-treated and microflawed specimen and a smooth DLC layer was grown for 2 h, where the optimum growth conditions: CH4 concentration 20 vol%, total gas flow rate 50 ml/min and the system pressure 33 Pa were employed. This

Adhesion inzproi9ement of dinmondjlms io silicon nitride 149

Figure 6. TEM images of the interfacial region between silicon initride and diamond with a diffraction pattern of Si3N4 crystal. The picture on the right hand below shows the diffraction pattern for the cross section of a silicon nitride needle-shaped crystal.

150 H. Itoh et al.

Figure 7. TEM images of the diamond crystal in the CVD coated film.

Adhesion improvement of dinmondjlms to silicon nitride

, I I I I I

151

E

Treatment time (h)

Figure 8. Tool life measured by the milling test as a function of CVD treatment time. The number shown in the plot means the thickness (pm) of diamond film. Acid treatment conditions: (A) 25OC, 30 min; (W) 40°C. 30 min, (0) 40°C, 60 min. 1 pass = the cutting amount of 2.4 cm3.

film was identified by Raman spectroscopy to be a typical DLC film which had a Raman peak at 1580 cm-' and a shoulder peak in the wavenumber range of 1350- 1400 cm-'.

A two-stage CVD of diamond was carried out onto the microflawed surface of the DLC interlayer. Fine diamond grains were seen to nucleate during the first stage CVD for 1 h. The second microflawing treatment was essential for nucleation sites enhancement, without which larger and only a few diamond grains were deposited

152 H. Itoh et al.

No fracture at 200 N 0

150 - 3 0 w

t” & 100 -0 w

0

10 20 30 40 50

Thickness (,u m) 0

Figure 10. Edge compression test results on diamond coated specimens prepared at different CVD treatment times. Solid symbol: without DLC interlayer. open symbol: with DLC interlayer. (m) 10 h, (A, A). 20 h. (0) 30 h with acid pretreatment; (0) 20 h without acid pretreatment.

on the DLC film. Diamond films as thick as 30-80 p m were grown by prolonging the second stage CVD treatment time up to 30 h.

The cross-sectional micro-Raman spectra in the interfacial region revealed a slightly shifted and weak diamond peak, referred to as “quasi-diamond”, at 1331 - 1332 cm-’ on the substrate side about 10 p m deep from the interface [16]. The DLC was identified at 20 p m inside from the interface. This micro-Raman analysis verified that a graded C-C bonding structure from the DLC ( s p 2 + sp3 bonding) to diamond ( sp3 bonding) film existed in the deep surface region of the substrate.

The edge compression test was carried out to evaluate the adhesion of diamond films, which were coated on a square chip with an edge side length of 10 mm and thickness of 3.0 mm. The edge part of the chip was compressed with a counter edge part of an orthogonally arranged bar of cemented carbide. Figure 10 shows the edge fracture loads for the diamond coated specimens with different coating conditions and thicknesses. The CVD reaction times are shown in parentheses near the plots. No fracture occurred even at loads of 200 N in both cases with and without DLC interlayer when the CVD treatment time exceeded 20 h. It is surprising that the 15 p m diamond coated specimen with a DLC interlayer has a high adhesion even without acid pretreatment.

The compression topple test for adhesion evaluation was performed using the diamond coated specimens obtained under different pretreatment and deposition conditions [20]. Compression stress was applied to the two sides of a rectangular prism of diamond coated specimen, i.e., the diamond film was compressed in parallel to the film surface direction. A small topple bar with 4 mm diameter and

Adhesion improvement of diamondjlms io silicon nitride 153

Thickness (t), p m 500 100 50 30 20 10 5

16001 I I I I I I 1 I

-* 1200 b v v) 1000

F v)

+ u) 800 C 0 v) 600 v)

400

200

.- g!

E

0.1 0.2 0.3 0.4 0.5

t-ln , ( p m)-ln

Figure 11. Compression topple test results on diamond coated specimens prepared under different conditions. (+) Diamond/DLC/acid-pretreated silicon nitride: (0) diamond/acid-pretreated silicon nitride: (0) diarnond/DLC/non-pretreated; and (A) diamond/non-pretreated silicon nitride.

100 mm length was attached to the film surface with an epoxy adhesive. A weight of 10 g was hung at the other edge. This topple bar plays a role in monitoring the peeling point of diamond film. The compression stress ( B ) at peeling was measured and the fractured surface was examined by SEM or XMA.

Figure 11 shows the compression topple test results on the specimens coated with various diamond film thicknesses ( t ) . Total CVD treatment times are shown near the symbols (+, 0, 0) for the substrates with or without acid pretreatment or DLC coating. The plots marked with A show the data obtained for comparison on the microflawed Si3N3 sintered substrate without an acid pretreatment as well as a DLC coating, when the diamond was coated by the hot filament CVD method using H2-5 vol% CH4 reactant gas. It is apparent from the linear relationship between B vs. reciprocal square root thickness (I/&) that the compression stress to cause peeling decreases with increasing film thickness. Higher adhesion strength was demonstrated when using the substrate pretreated in a strong acid. A remarkably high value of B = 1235 MPa was attained for the acid-treated specimen which was covered with a DLC interlayer and coated with 25 p m thick diamond film.

Figure 12 shows the typical SEM photographs of the fractured surfaces of the specimens after the compression test. Diamond films with thicknesses of 15 pm, 30 p m and 50 p m were deposited on the pretreated substrate for CVD treatment times of (a) 10 h, (b) 20 h and (c) 30 h, respectively. A large difference in surface microstructure is recognized among the specimens (a), (b) and (c). The surfaces of specimens obtained for CVD reaction times of 10 h or 20 h suggested that peeling

154 H. Itoh et al.

(a) 10 h

(b) 20 h

Figure 12. SEM photographs of the specimen surfaces appearing after the fracture by the compression topple test. Total diamond coating time: (a) 10 h, (b) 20 h. (c) 30 h, (d) 30 h (larger view with a low magnification).

occurred at the interface between the silicon nitride substrate and the intermediate DLC layer which was formed spontaneously during the second stage CVD. In contrast, peeling seemed to occur between the DLC layer (including quasi-diamond) and the diamond film in the case of CVD reaction time of 30 h, as shown in Fig. 12c. However, the magnified SEM picture of this specimen revealed two different areas, as shown in Fig. 12d, where one area (D) was analogous to Fig. 12c and the other area (S) was similar to Fig. 12b. The carbon map obtained by XMA revealed a greater amount of carbon distribution in the area (D), where fine-grained diamond particles adhered to the substrate side. This is the evidence that the adhesion strength between the substrate (B-Si3N4) and the DLC layer increased by a long-term CVD treatment in the second stage.

Adhesion improvement of diaiizondjilrns to silicon nitride 155

(c) 30 h

(d) 30 h (a larger view) - 50 p m Figure 12. (Continued).

Figure 13 shows the micro-Raman spectra of the fractured surfaces of the same specimens as shown in Fig. 12. In the spectra obtained for the specimens (a and b) CVD coated for 10-20 h, Raman peaks for the DLC can be observed, which suggests that a small amount of DLC still remains on the substrate side. On the other hand, quasi-diamond peak can be identified along with the DLC peaks, as shown by arrow 4 in the D spectrum of the diamond coated specimen for 30 h. This suggests that such a graded C-C bonding structure from the DLC to diamond via quasi-diamond contributes to the increased adhesion strength, where a large residual internal stress would have been relaxed in the interfacial region.

3.5. Internal residual stress distribution in the diamond coated specimens

Figure 14 shows the result of the X-ray residual stress measurements on the diamond coated specimens with CVD treatment times of 10 h, 20 h and 30 h. It is confirmed

156 H. Itoh et al.

1700 1600 1500 1400 1300 1200 1100

Raman shift, cm-I Figure 13. Micro-Raman spectra in the areas (a) and (b) in Fig. 12, and in areas (S) and (D) in Fig. 12d.

from the diagram that the 28 -sin2 C$ relationship is non-linear, and the intermediate range of sin2 C$ value with a gentle slope expands to the lower value side of sin2 4 with increasing CVD reaction time. Considering a deep X-ray penetration depth in the diamond, the information from the lower angle side (lower value of sin2 4) reflects the stress behavior in the interfacial region between the diamond and the substrate, while the information from the higher angle side (higher value of sin2 @) reflects the stress behavior on the free-surface of the diamond film [24, 251. Consequently, one can summarize qualitatively from the sign of the slope that a compression stress acts in the interfacial region and a tensile stress acts on the free surface. The intermediate zone of stress relaxation expands with increasing CVD treatment time. Such an internal stress distribution must arise by a large thermal stress in the interfacial region and a large intrinsic stress in CVD diamond film.

4. CONCLUSIONS

Adherent diamond coatings onto a pretreated sintered silicon nitride substrate were developed by a two-stage microwave plasma CVD technique using the CO and H2 reactant system. The pretreatment of the substrate in a hot and strong acid solution and subsequent micoflawing treatment with diamond grains resulted in an anchored deposition of the diamond into the micropores of the acid-treated substrate. An excellent adhesion between the diamond film and the silicon nitride substrate was attained by the two-stage CVD of diamond, i,e. a fine-grained diamond deposition in the first stage, and higher growth rate of diamond film in the second stage. The

Adhesion improvement of diamond films fo silicon nitride

130.90

M e, -0

m N

- 130.80

130.70 000 0.10 020 0.30 040 0.50 0.60 0.70 0.80

sir+$

M 0

W

a cv

3 1 .OO

30.90

. C .

130.80 ! I I

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 sin211

M e, '0

m N

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80

157

Figure 14. 20 -sin2 4 diagrams obtained by X-ray residual stress measurements on specimens with different CVD treatment times. Other treatment conditions are the same as those for specimens (c), (d) and (e) in Fig. 4.

insertion of a diamond-like carbon (DLC) intermediate layer was found to enhance the adhesion strength further. The graded texture and structural variation of the C-C bonding from the substrate side to the diamond film were confirmed by SEM/XMA and TEM analyses or micro-Raman spectroscopy. 'The high adhesion strength was verified by the edge compression test, the compression topple test or the milling test on the diamond coated specimens. The actual cutting test by milling the Al- 20 wt% Si alloy work material revealed a significantly long tool life when using a thick diamond coated specimen. The X-ray residual stress measurements suggested that the thermal compression stress in the interfacial region was relaxed by an intrinsic tensile stress on the free-surface of the diamond film.

158 H. Itoh et al.

REFERENCES

1. S. Soderberg, A. Gerendas and M. Sjostrand, Vacuum 41, 1317-1321 (1990). 2. R. Polini, G. Marcheselli and E. Traversa, J. Amer. Ceram. Soc. 77, 2043-2048 (1994). 3. B. Lux and R. Haubner, in: Thin Film Diamond, A. Lettington and J. W. Steeds (Eds).

4. M. Neslidek, K. Vandierendonck, C. Quaeyhaegens, M. Kerkhofs and L. M. Stals, Thin Solid

5. R. Polini, E. Traversa, A. Marucci, G. Mattei and G. Marcheselli, J. Electrochem. Soc. 144,

6. S. Chatterjee, A. G. Edwards and C. S. Feigerle, J. Mater. Sci. 32, 3355-3360 (1997). 7. N. Dilawar, R. Kapil, Brahamprakash, V. D. Vankar, D. K. Avasthi, D. Kabiraj and G. K. Mehta,

8. B. Lux and R. Haubner, in: Low Pressure Sjnthetic Diamond, B. Dischler and C. Wild (Eds),

9. R. Polini, S. Mattei, A. Marucci and E. Traversa, 1. Ceram. Soc. Japan 106, 1167-1171 (1998). 10. R. Polini, F. L. Normand, G. Marcheselli and E. Traversa, J. Amel: Ceram. Soc. 82, 1429- 1435

11. H. Itoh, S. Shimura, H. Iwahara and H. Sakamoto, J. Ceram. Soc. Japan 104. 1137- 1142 ( 1996). 12. H. Itoh, S. Shimura, K. Sugiyama, H. Iwahara and H. Sakamoto, J. Amer. Ceram. Soc. 80,

13. H. Itoh, K. Sugiyama, H. Iwahara and H. Sakamoto, J. Japan Soc. Powder Powder Metallurgj

14. H. Itoh, K. Sugiyama, H. Iwahara, Sung-Soo Lee and H. Iwahara, J. Surface Finishing Soc.

15. H. Itoh, New Diamond 14, 24-29 (1998). 16. H. Itoh, S. S. Lee, K. Sugiyama and H. Iwahara, Surjace Coatings Technol. 112, 199-203

(1999). 17. T. Seguchi, M. Takahashi, K. Kuroda, H. Saka, K. Sugiyama, I. Arai and H. Itoh, in: Proc.

Special Symp. Advanced Materials, T. Imura, H. Fujita, T. Ichinokawa and H. Kawazoe (Eds), pp. 295-298. CO-OP Printing Bureau of Nagoya University (1998).

pp. 127-141. Chapman &Hall, London (1994).

Films 270, 184-188 (1995).

1371 - 1375 (1997).

Thin Solid Films 323, 163- 169 (1998).

pp. 223-242. Springer, Berlin (1998).

(1999).

189-196 (1997).

43, 1455- 1460 (1996).

Japan 47. 1042-1047 (1996).

18. H. Itoh, T. Osaki, H. Iwahara and H. Sakamoto, J. Mater. Sci. 26, 3763-3768 (1991). 19. Sung-Soo Lee and 0. Takai, in: Proc. 9th S p y . Plasma Sci. f o r Materials, pp. 27-31.

20. K. L. Mittal, in: Adhesion Measurement of Thin Films, Thick Films and Bulk Coatings,

21. K. L. Mittal, Electrocomponent Sci. Technol. 3, 21-42 (1976). 22. K. L. Mittal, in: Adhesion Measurement of Films and Coatings, K. L. Mittal (Ed.), pp. 1-13.

VSP, Utrecht, The Netherlands (1995). 23. T. Tsutsumoto, A. Nakao and H. Matsubara, in: Advances in New Diamond Sci. and Technol.,

S. Saito, N. Fujimori, 0. Fukunaga, M. Kamo, K. Kobayashi and M. Yoshikawa (Eds),

University of Tokyo, Tokyo (1996).

K. L. Mittal (Ed.), STP No. 640, pp. 5- 17. ASTM, Philadelphia (1978).

pp. 763-766. MYU K.K., Tokyo (1994). 24. S. K. Choi, D. Y. Jung and H. M. Choi, J. Vac. Sci. Technol. 14, 165-169 (1996). 25. H. Mohrbacher, K. Van Acker, B. Blanpain, P. Van Houtte and J.-P. Celis, J. Muter. Res. 11,

1776- 1782 (1996).

Adlzesiorz Aspects of Thirz Films, Vol. 1, pp. 159- 170 Ed. K. L. Mittal c VSP 2001

Quantifying the effect of carbon on the practical adhesion of aluminum films to sapphire substrates

J. A. SCHNEIDER*, S . E. GUTHRIE, W. M. CLIFT and N. R. MOODY Sandia National Laboratories, Livermore, CA 94551, USA

Abstract-The adhesion of aluminum films onto sapphire substrates in the presence of controlled carbon contaminants was investigated. In this study, nanoindentation techniques were used to quantitatively assess the practical adhesion of the film-substrate systems. This technique induces controlled delamination blisters of the film from the substrate, even in ductile films if constrained by a highly stressed overlayer. The geometry of these blisters is modeled as interfacial cracks. utilizing existing linear elastic fracture mechanics models, and is characterized by a critical strain energy release rate.

Specimens with 10 nm of carbon at the interface displayed a weak adhesion characterized by a strain energy release rate of 1-0.8 J /m2 with debonding occurring within the carbon layer. As the carbon thickness was decreased to 1 nm, there was no discernible change in the practical adhesion from that of a clean surface as characterized by a strain energy release rate of 5.1-6.1 J /m2. Formation of a carbide, observed in Auger analysis, suggests there may be a critical amount of carbon associated with weak interfaces.

Keywords: Interfacial fracture; indentation; thin films; adhesion; nanomechanics.

1. INTRODUCTION

A common concern to those working with thin films or coatings is how well these adhere to the substrate or material of interest. In the electronics area, materials of interest are commonly metal films on hard, insulating substrates. How well these films adhere, both initially and over time, depends on the presence of contaminants which can weaken the adhesion, along with the presence of residual stresses which can combine to overcome the interfacial strength. Of interest is how to measure and interpret the effect of contaminants on this adhesion or interfacial strength. Quantification of this adhesion strength is necessary to evaluate film processing or process changes and allow comparison between different film systems.

*To whom correspondence should be addressed. Dept. of Mechanical Engineering, MSU, Mis- sissippi State, MS 39762, Phone: (662) 625-3260; Fax: (662) 325-7223; E-mail: schneider @me.msstate.edu

160 J. A . Schneider et al.

Adhesion can be defined as either the thermodynamic work of adhesion, W,, or the practical work of adhesion, WAd. The thermodynamic work of adhesion is that required to create two new surfaces and is a function of the surface and interface energy. Since there are no direct methods for measuring the thermodynamic work of adhesion for a solid-solid system [l], the practical work of adhesion [2-41 is that which is generally measured. Methods for evaluating the practical adhesion of as-deposited thin films to substrates generally involve removal of the film and fall into two categories. Semi-quantitative tests, such as peel or bend tests, provide a comparative basis for adhesion, but they cannot be directly related to material mechanisms of adhesion [5, 61. More quantitative methods have been developed which utilize linear elastic fracture mechanics concepts to allow assessment of the practical adhesion [7- 111. These methods use various techniques to initiate and propagate a delamination event between the film and the substrate, such as the blister technique [ 12, 131 or indentation techniques [ 14- 161. Nanoindentation test typically uses a device which measures the continuous load and displacement of a diamond indenter tip as it is pushed into the as-deposited film. Indentation stresses, along with residual film stresses, influence the interfacial cracking response by inducing buckling of the film above the crack and by providing an additional crack driving force once buckling occurs. The resulting controlled film delamination events are modeled as bimaterial cracks and practical film adhesion is characterized by the strain energy released per unit increase in delaminated area.

The formulas used in this study are those derived by Marshall, Evans, and Hutchinson [16, 171 and Hutchinson and Suo [lo] for continous indentation test. However, application of these techniques to ductile or strongly adhering films has been limited due to the difficulty in developing sufficient elastic strain energy in the film to initiate delamination [ 181. The use of highly stressed overlayers deposited on the film of interest has been reported to reduce this limitation [6, 19-24]. In this study, nanoindentation techniques applicable to testing of as-deposited films of ductile A1 on hard sapphire substrates are evaluated.

2. MATERIALS AND PROCEDURE

The effect of contaminants on adhesion was investigated by sputtering carbon of various thicknesses onto the (0001) sapphire substrate prior to deposition of AI films. A graphite target was used as the carbon source. Prior to deposition, the sapphire substrate surface was subjected to an HC1 etch, de-ionized water rinse, N2

drying, and mounted in an analysis chamber with a base pressure of 2 x lo-' Torr. The substrate was heated to 250°C and held for 1 h to drive off moisture. The evolving species were verified as water vapor by an in-situ quadrupole mass spectrometer. Surface cleanliness was verified by in-situ Auger analysis before and after all processes. After characterization, the sample was moved under vacuum directly into an ancillary chamber for film deposition. Film thickness was determined in-situ by a quartz crystal deposition monitor and, subsequently, verified using independent stylus measurements.

Quantibing the effect of cnrbon on the practical adhesion 161

A1 films 500 nm thick were deposited by physical vapor deposition (PVD) onto smooth single crystal sapphire substrates at a background pressure of 5 x lop9 Torr. The substrates were not actively heated, but the temperature rose during the deposition from an ambient 24°C to 34"C, due to radiant heating from the source. Ta overlayers, 1 p m thick, were deposited by DC magnetron sputtering onto the surface of the 500 nm thick A1 films using low pressure ( 5 mTorr) argon gas.

A Rigaku X-ray diffractometer (XRD) with a thin film detection system and Cu K , radiation was used to characterize the films. Out-of-plane strain, indicated by shifts in the 28 peak position, was used to calculate the in-plane residual stress, assuming isotropic elasticity [25].

The hardness and elastic moduli of the A1 and Ta films on sapphire were calculated for a series of indents. A continuous recording of the indentation loads and displacement data was made using a Nanoindenter II@ system. A triangular (Berkovich) pyramid-shape diamond indenter tip, with approximately 320 nm tip radius, was driven to a maxiumum depth of 500 nm at a constant loading rate of 30 p N / s . The Berkovich indenter tip was previously calibrated by indentations into a material of known modulus to correlate the contact area as a function of the indenter penetration depth. The use of this area function and the recorded depth of the identer allowed calculation of the elastic moduli and hardness properties using the methods of Oliver and Pharr [26].

For indentation induced delamination of the film from the substrate, the Nanoin- denter was configured with a conical diamond indenter, with a nominal 1 p m tip radius and a 90" included angle, that was driven into the film at a constant loading rate of 300 pN/s . From blistering events, the fracture energy or work of practical adhesion of the film to the substrate can be calculated using the elastic approach of Marshall, Evans, and Hutchinson for indentation induced delamination [ 10, 16, 171. The geometries of the delamination blisters were measured from optical micro- graphs using Nomarski contrast. Scanning force microscopy (SFM) was used to measure the indent profile.

The practical work of adhesion ( WAd) is calculated from the strain energy release rate (G) for fracture of thin films. A crack induced at the interface between the film and substrate advances when the strain energy release rate is equal to the crack growth resistance (r,) as summarized in equation ( I ) [ 101. For this analysis, TI is considered to be an indication of the practical adhesion strength, Le.,

The analysis of the nanoindentation driven delamination utilizes the combined bi- layer properties of both the Ta and AI films for elastic modulus, E , Poisson's ratio, u , stress state, a, thickness, t , and delamination radius, a, to calculate the strain energy release rates [lo]. For indentation of the film, the portion released (G) when the crack advances at the interface is given in equation (2) [lo, 16, 171.

t q ( 1 - u2) ctai(1 - u ) ct(t71 - a&1 - u ) - - + E E

G = 2E 3

162 J. A. Schaeider et al.

where c = [ l + 0.902(1 - v)]-'; CTR is the residual stress contributed by both films; a1 is the indentation stress determined from volume-conserving accommodation of the indentation volume, VI, given by equation ( 3 ) [lo];

and oB is the critical buckling stress for a clamped circular plate given by equa- tion (4) [lo];

where k is equal to 14.7 for a unconstrained plate, shown in Fig. la, which typically delaminates during indenter unloading. Figure l b illustrates a plate pinned at the center, which delaminates and buckles as the indentation is loaded into the film, with k = 42.7. A third mode of buckling has also been reported and is illustrated in Fig. IC [27]. At present this type of blister would be modeled as a double or annular buckle.

For the general case of a thin bi-layer film, the appropriate expressions for the stresses, oR, aI, and aB, have been determined using standard solution methods for pure bending of non-homogeneous members and the application of thin plate mechanics to describe the curvature and stress distributions in multilayers after release from the substrate as a function of the residual stress. This derivation has

Figure 1. After indentation, the released film may display different modes of delamination. Single buckle (a), double or annular buckle (b), or bending buckle (c).

Quantihing the effect of carbon on the pnzctical adhesion 163

been published elsewhere for indentation delamination of a multilayer thin film system from a substrate [20, 211.

3. RESULTS AND DISCUSSION

The XRD analysis of the Ta films deposited directly onto sapphire (Fig. 2a) and over the A1 films (Fig. 2b) indicates similar texturing which corresponds to the B-Ta phase [28, 291. Spontaneous blisters were observed in the Ta films soon after deposition, and were used to calculate an upper bound for the residual stress in the film. Using equation (4) to determine the critical buckling stress for a clamped plate of 40 mm nominal diameter [ 101, a residual stress level of 1.8 GPa in the film would be required to drive the delamination induced buckling. Since this is an elastic solution, any plastic deformation would relieve some level of residual stress and result in an overprediction of stress. For a lower bound estimate of typical residual stress levels, we use the data from the review of Thorton and Hoffman [30] in which similar argon gas pressures (5 mTorr) were used to sputter 200 nm thick Ta films with a resulting compressive residual stress level of 0.4 GPa. Using the 26' peak position shift of the Ta film on sapphire, a biaxial in-plane compressive stress of 0.8 to 1 .O GPa was calculated for the films in this study.

A1 films on sapphire all display a (111) texture, both with and without the Ta overlayer, as shown, respectively, in the XRD diffractograms in Figs 2b and 2c. The 26' peak position shift of the A1 films on sapphire was used to calculate a biaxial, in-plane tensile stress in the range of 0.06 to 0.07 GPa. Stress levels in the A1 films are significantly less than those in the Ta films.

The elastic modulus and hardness values for the A1 and Ta films were calculated using indentation data from the top 10% of the film to avoid the influence of the substrate. Indentations in the B-Ta film on sapphire indicate a hardness of 16.6 f 1.8 GPa and an elastic modulus of 266.7 :& 23.4 GPa. This is in close agreement with the published hardness of 16-18 GPa for (002) oriented P-Ta films [29]. Textured (1 11) A1 films on sapphire indicate a hardness of 0 .7f0 .14 GPa and an elastic modulus of 73.8 f 11.3 GPa. The hardness of the A1 film is slightly higher than that of bulk (1 11) single crystal Al, which is not uncommon in thin film layers [31].

Initial indentation testing with a conical indenter in the as-deposited A1 films was unsuccessful in driving delamination blistering. Sputtered Ta films were then deposited at low temperatures over the A1 films, and the indentation tests repeated. Delamination blisters induced from these tests are shown in Fig. 3a for the sample with 10 nm carbon at the interface and in Fig. 3b for the sample with the clean interface. Since a load of 250 p N was required to induce blistering in the sample with the clean interface, this load was kept constant in subsequent tests.

Modified calculations, based on the methods of Marshall, Evans, and Hutchinson [16, 171, which take into account the bi-layer film properties [20, 211, were applied to the indentation induced delamination blister geometry. The calculated fracture energies to drive delamination in the films with 10 nm of carbon at the interface were

164 J. A. Schneider et al.

1

0.8

0.6

F .I v1

g0.4 C c U

0.2

n 20 30 40 50 60 70 80 90 100

(a> 1

0.8

0.6

F 2

.C

8 0.4 E -

0.2

0

2 8

20 30 40 50 60 70 80 90 100

(b) 2 0

Figure 2. XRD analysis of A1 and Ta films deposited on sapphire substrates. Predominant (002) orientation is observed in 1000 nm thick sputtered B-Ta film deposited directly on sapphire substrate (a) and over a 500 nm thick PVD A1 film on sapphire substrate (b). Deposition of the Ta film on the A1 film does not change the (1 11) texture observed in the 500 nm thick A1 films on sapphire without the Ta film overlayer (c).

in the range of 0.8 to 1.3 J/m2. Although the fracture energy from the indentation tests is slightly higher than the van der Waals type of force (0.05 to 0.7 J/m2) [32-351, the low fracture energy indicates a weak type of bond.

A significantly higher fracture energy of 5.1 to 6.1 J/m2 was calculated for the smaller diameter delamination blisters observed in the films with the clean interface indicating a stronger bond. Similar results have been published for 500 nm thick Ta2N overlayers on 178 nm thick A1 films in which a fracture energy of 7 J/m2 was reported [23].

The removal of the film from the substrate with carbon at the interface provided the opportunity to inspect the delaminated surfaces of both the sapphire substrate and the A1 film. An Auger depth profile using Ar+ sputtering of the sapphire

Quanribing the effect of carbon on the practical adhesion 165

1

0.8

0.6

h U

m .- 0.4 * I C

0.2

0

(c) Figure 2. (Continued).

20 30 40 50 60 70 80 90 100

2 e

Figure 3. Nanoindentation tests were used to induce blisters in the samples with Ta/Al films on sapphire under a 250 mN load. Larger diameter blisters are observed in the light micrograph images in the sample with 10-20 nm of carbon at the interface (a) rhan in the sample with the “clean” interface (b).

substrate, shown in Fig. 4a, indicates a residual carbon thickness of 9.5 nm before the A1 spectrum from the sapphire substrate is observed. Figure 4b shows an Auger depth profile of the interface side of the A1 film. The Auger peak shape analysis, using a linear least squares fit, indicates an initial layer of 0.5 nm of carbon which transitions to a carbide before exposing the underlying Al.

166 J. A. Schneider et al.

100

80

$ 60

u *# 2 40

20

0

0 2 4 6 8 10

Depth in nm

60

40

20

0 0 2 4 6 8 10

Depth in nm

(b) Figure 4. Separation of the film from the substrate allowed Auger depth profile of the delaminated regions. A 9.5 nm residual carbon layer remains on the delaminated sapphire substrate surface (a). On the A1 film delaminated surface (b), a 0.5 nm layer of carbon layer remains, indicating the fracture occurred in the carbon layer. Carbide formation is noted between the carbon and Al.

Quant ib iq the effect of carbon on the pr,ictical adhesion 167

Next, the carbon thickness was decreased to 1 nm and the indentation tests repeated. No differences were observed in the initial, optical observations of the indentation induced delaminations, which would indicate similar bond strengths. Using scanning force microscopy (SFM) in the indented region, Fig. 5 shows a comparison of the indent profiles. The clean interface, shown in Fig. 5a, displays an annular type blister in which the film remains attached to the substrate under the indenter tip, suggesting the crack propagated during loading. With 1 nm of carbon present, the blister in Fig. 5b represents the bending type. The primary difference with 1 nm C is that the film under the indenter tip has detached from the substrate, although permanent deformation has occurred. The detachment of the film would indicate a slightly weaker adhesion, although using the fracture

(a> Figure 5. Scanning force microscope (SFM) images of indents into Ta/Al/sapphire systems with clean interface (a) and I nm of carbon at the interface (b).

168 J. A. Schneider et al.

(b) Figure 5. (Continued).

mechanics approach, similar values for the fracture energy are calculated; although the failure of the film under the indenter tip suggests a possible embrittlement of the ductile A1 film in the presence of contaminants.

The evidence of fracture within the carbon layer supports similar trends reported in whisker pull-out tests in which a carbon layer formation has been reported to decrease the fracture resistance [36]. Weak bonding has been reported for a range of interfacial carbon layer thicknesses. Brennan [37, 381 has cited an extensive fiber pull-out occurring in composites with a 10-40 nm layer of carbon present, attributed to debonding within the carbon layers. The effect on shear strength was recently reported by Dehm et al. [39] in which a decrease in the adhesion of copper to sapphire was observed by incorporating up to 110 nm thick carbon layers. Our current investigation shows that the debonding within a carbon layer occurs in much thinner layers than previously reported. The presence of a carbide suggests that

Quantifiing the effect of carbon on the practical adhesion 169

a reaction may be occurring between the A1 and the carbon, indicating a critical carbon thickness for transition from a weakly bonded to a strongly bonded A1 film on sapphire substrate.

4. SUMMARY

In this study, indentation induced delamination was used to determine the effects of varying amounts of carbon on the adhesion strength of A1 films to sapphire sub- strates. This method, which can be applied to as-deposited films, utilizes elastic fracture mechanics based solutions for measuring of practical adhesion in terms of the energy dissipation required for delamination. The use of highly stressed over- layers appears to increase the driving force for indentation induced delamination of highly ductile films, which otherwise are unable to support sufficient elastic strain energy to promote interfacial delamination.

Specimens with 10 nm of carbon at the interface displayed a weak adhesion characterized by a strain energy release rate of 1-0.8 J /m2. No discernible change in the practical adhesion between a clean substrate and one with 1 nm of carbon at the interface was measured as characterized by a strain energy release rate of 5.1-6.1 J/m2. Auger depth profiles indicated that the fracture occurred within the 10 nm thick carbon layer, thinner than previously reported for weak interfaces. The presence of a carbide adjacent to the A1 film surface suggests a possible transition from a weakly bonded to a strongly bonded A1 film on sapphire substrates. The changes in the mode of delamination blister may provide additional information on the effect of contaminants at the interface.

Acknowledgements

This work was supported by the U.S. Department of Energy under Contract #DE-AC04-94AL85000. This paper is written in memory of Steve Guthrie, auf Wiedersehen mein Freund.

REFERENCES

1. L. Vitos. A. V. Ruban, H. L. Shiver and J. Kollar, Surface Sci. 411, 186 (1998). 2. K. L. Mittal. in: Adhesion Measurement of Thin Films, Thick Films and Bulk Coatings,

3. K. L. Mittal, Electrocomponent Sci. Technol. 3, 21 (1976). 4. K. L. Mittal, in: Adhesion Measurement of Films and Coatings, K. L. Mittal (Ed.), pp. 1- 13.

5. M. Ohring, The Materials Science of Thin Films. Academic Press, New York (1992). 6. A. Bagchi and A. G. Evans, Thin Solid Films 286, 203 (1996). 7 . Z. Suo and J. W. Hutchinson, Int. J. Fracture 43, 1 (1990). 8. J. R. Rice, J. Appl. Mech. 55, 98 (1988). 9. A. G. Evans, M. Ruhle, B. J. Dalgleish and P. G. Charalambides, Matel: Sci. Eng. A126, 53

K. L. Mittal (Ed.), STP No. 640, pp. 5- 17. ASTM, Philadelphia (1978).

VSP, Utrecht, The Netherlands (1995).

( 1990).

170 J. A. Schneider et al.

IO. J. W. Hutchinson and Z. Suo, in: Advances in Applied Mechanics, J. W. Hutchinson and T. Y. Hu

11. A. G. Evans, M. D. Drory and M. S. Hu, J. Mater: Res. 3, 1043 (1988). 12. H. Chai, Int. Fracture 46, 237 (1990). 13. H. M. Jensen, Engr: FractureMech. 40,475 (1991). 14. L. G. Rosenfeld, J. E. Ritter, T. J. Lardner and M. R. Lin, J. Appl. Phys. 67, 3291 (1990). 15. C. Rossington, A. G. Evans, D. B. Marshall and B. T. Khuri-Yakub, J. Appl. Phys. 56, 2639

16. D. B. Marshall and A. G. Evans, J. Appl. Phys. 56,2632 (1984). 17. A. G. Evans and J. W. Hutchinson, Int. J. Solids Structures 20,455 (1984). 18. M. R. Turner and A. G. Evans, Acta Mater. 44, 863 (1996). 19. A. Bagchi, G. E. Lucas, Z. Suo and A. G. Evans, J. Mater. Res. 9, 1734 (1994). 20. M. D. Kriese, N. R. Moody and W. W. Gerberich, Mater: Res. Soc. Symp. Proc. 473, 39 (1997). 21. M. D. Kriese, N. R. Moody and W. W. Gerberich, J. Mater: Res. 14, 3007 (1999). 22. M. D. Kriese, N. R. Moody and W. W. Gerberich, J. Mater: Res. 14, 3019 (1999). 23. D. E Bahr, J. W. Hoehn, N. R. Moody and W. W. Gerberich, Acta Mater. 45, 5 I63 ( 1997). 24. M. D. Kriese, D. A. Boismier, N. R. Moody and W. W. Gerberich, Eng. Fracture Mech. 61, 1

25. D. B. Cullity, Elements of X-ray Diffruction. Addison-Wesley, Reading, MA (1967). 26. W. C. Oliver and G. M. Pharr, J. Mater: Res. 7, 1564 (1992). 27. M. V. Swain and J. Mencik, Thin Solid Films 253, 204 (1994). 28. M. Oda, A. Ozawa, S. Ohlu and H. Yoshihara, Jpn. J. Appl. Phys. 11, 2616 (1990). 29. R. Saha and J. A. Barnard, J. Crystal Growth 174, 495 (1997). 30. J. A. Thorton and D. W. Hoffman, Thin Solid Films 171, 5 (1989). 31. M. E Doerner, D. S. Gardner and W. D. Nix, J. Mater Res. 1, 845 (1986). 32. K. L. Chopra, Thin Film Phenomena. McGraw-Hill, New York (1969). 33. M. T. Laugier, Thin Solid Films 117, 243 (1984). 34. D. C. Agrawal and R. Raj, Mater: Sci. Eng. A126, 125 (1990). 35. P. Benjamin and C. Weaver, Proc. Roy. Soc. London A252,418 (1959). 36. R. J. Kerans, R. S. Hay and N. J. Pagano, Ceramic Bull. 68, 429 (1989). 37. J. J. Brennan, in: Tailoring Multiphase and Composite Ceramics, R. T. Tressler, G. L. Messing,

38. J. J. Brennan, ONR Tech. Report R87-917546-4 (1987). 39. G. Dehm, R. Raj and M. Ruhle, Mater: Sci. Forum 207-209, 597 (1996).

(Eds), p. 63. Academic Press, New York (1992).

(1984).

(1998).

C. G. Pantano and R. E. Newnham (Eds), p. 549. Plenum, New York (1986).

Adhesion Aspecrs ofThin Films, Vol. I , pp. 171-180 Ed. K. L. Mittal 0 VSP 2001

The effect of a titanium-based interlayer on the adhesion of ceramic coatings

M. T. VIEIRA *, S. ROQUE and A. S. RAMOS Depto. Enga. Mecrinica da Faculdade de CiCncias da Universidude de Coimbra, Polo II Pinhal de Marrocos, 3030 Coimbra, Portugal

Abstract-The present work was carried out to study the effect of a titanium-based interlayer on the adhesion of sputtered nitride coatings. Tungsten nitride coatings were deposited onto high-speed steel substrates after deposition of a 0.8 p m interlayer for improving the adhesion. Pure titanium and titanium alloys interlayers with different aluminium concentrations were used. The critical adhesion loads in the scratch test were significantly lower with no interlayer than if titanium with aluminium was used as an interlayer. The best practical adhesion was obtained when an interlayer with 28 at.% of aluminium was used.

Keywords: Practical adhesion; scratch test; interlayer; titanium-. aluminium; tungsten nitride.

1. INTRODUCTION

Ceramic coatings are frequently used in the mechanical field, namely in tool applications. Nevertheless, it is known that ceramic-to-metal adhesion can be poor.

Regardless of the purpose of the coating, a fundamental prerequisite is a good adhesion between the coating and the substrate that could result from:

A low interfacial energy which induces a good structural matching between the substrate and the film. The decrease of the stress gradient between the coating and the substrate resulting from the different thermal expansion coefficients of the coating (ceramic) and the substrate (metal alloy) in order to absorb the shear strain without initiation of fracture.

And/or a good chemical bonding with formation of ductile compounds at the interface.

*To whom correspondence should be addressed. Telephone: 351-239790765; Fax: 351-239790701; E-mail: teresa.vieira@mail.dem.uc.pt

112 M. i? Vieira et al.

The adhesion depends on the properties of both substrate and film, but also on the interface itself. Bonding between interface atoms may be attributable to a variety of interactions including van der Waals forces and the formation of metallic, ionic or covalent bonding configurations. The adhesion generally requires the formation of strong chemical bonds at the interface. Strong and stable interatomic bonding, preferentially metallic, linking the substrate species to the coating remains the crucial underlying source of thin film adhesion. In systems that have no bulk affinity such bonding can be promoted by tailoring the interface.

Nowadays, it is usual to improve the adhesion between ceramic coatings and steel substrates, especially TIN coatings, by introducing a titanium interlayer that promotes a decrease of the compressive stresses between the ceramic film and the metallic substrate [ 1,2]. However, Helmersson et al. [ 11, concluded that the increase of the substrate temperature above 400 "C during presputtering and sputtering could reduce significantly the adhesion due to the diffusion of carbon of the substrate to the TIN film with the formation of Tic at the interface. This carbide has a more covalent character than TIN, which is responsible for a ductility decrease. Nevertheless, these authors did not take into consideration the reactivity of titanium with substrate surface oxides, only highlighting the role of FeO in the increase of structural matching between the substrate and the TiN coating. In the work of Tang et al. [3], the influence of Ti and Cr interlayers with different thicknesses on friction and wear behaviours of Tic coatings was studied. Among the coatings with an interlayer, those with 50 and 500 nm Cr or 50 nm Ti exhibited less delamination than those with a 500 nm Ti interlayer and their hardness and elastic moduli were significantly higher, thus enhancing their wear and friction behaviours. However, the high chemical reactivity of titanium, that presupposes a good adhesion, does not always give rise to an adhesion improvement [4, 51. The titanium reactive interlayer only improves adhesion, if stable and non-fragile reaction products are formed at the interface. The titanium reacts with the oxygen from the substrate surface forming a family of oxides ranging from Ti0 to TiOz, depending on its activity [6]. Among these oxides, the T i0 is the most stable [7] and promotes a strong and ductile bonding with the steel substrate due to its more metallic character. To obtain the desired oxide, titanium with an element of dilution could replace the titanium interlayer. The composition of this interlayer should be carefully chosen because the titanium activity will dictate the reaction products formed. The aluminium is an interesting element for dilution because it forms solid solutions or stable intermetallic compounds with Ti [8], leading to a decrease of titanium activity and/or formation of Ti3AlC compounds with the carbon from the substrate (Ti3AlC compounds have a more metallic character than TIC). Both these factors contribute to adhesion improvement. Lii et al. [9] reported that TiAl was very promising for use as an interlayer to enhance the adhesion of TiAlN coatings to high-speed steel substrates. Single TiAl coatings adhere very well to steel substrates and the adhesion between TiAl and TiAlN was guaranteed by the nitrogen diffusion. However, the influence of the TiAl interlayer was not compared to a pure Ti one which according to the above explanation might also lead to the same results.

Titanium-based interlayer and adhesion ojceramic coatings 173

The aim of the present work was to study the influence of the titanium interlayer dilution on the adhesion of tungsten nitride coatings to high-speed steel substrates. Interlayers with different amounts of aluminium (element of dilution) were pro- duced in order to obtain different titanium activities making it possible to show its influence on the adhesion of the bilayer coatings to the substrate.

2. EXPERIMENTAL DETAILS

2.1. Deposition technique

Polished high-speed steel substrates (AIS1 M2) were coated with tungsten nitride by reactive d.c. magnetron sputtering from a tungsten target. To improve the adhesion a titanium- aluminium interlayer was introduced between the substrate and the ceramic coating. This interlayer was produced from a titanium target with a variable number of small aluminium foils. Before deposition, the sputtering chamber was evacuated by a turbomolecular pump down to a final pressure of lop4 Pa. Afterwards, the substrate surfaces were heated and ion cleaned by an ion gun at a pressure of 1.5 x lo-' Pa. During the deposition, the substrates were at a negative bias of 70 V. To deposit the bilayer coatings (interlayer + ceramic film) two cathodes, one for the tungsten target and another for the titanium target, were used. Both cathodes had independent power sources. The deposition times were selected in order to obtain an interlayer with 0.8 p m and a ceramic coating with 3.2 p m thickness, thus the total thickness being 4 pm. Sputtering was performed in pure argon during the production of the interlayer, and in an argodnitrogen mixture with a N2/Ar partial pressure ratio of 1.5 to obtain the ceramic film using the reactive mode. Coatings with no interlayer, and with an interlayer with different amounts of aluminium were produced by the sputtering technique using the deposition parameters summarised in Table 1.

2.2. Characterisation techniques

The properties of the interlayer and/or ceramic coating were evaluated using the following equipments: 0 Perthometer CSD rugosimeter for the determination of coating thicknesses; 0 Cameca SXSO electron microprobe to obtain the chemical composition of the

0 Philips PW 3040/00 X-pert diffractometer with a Co K , radiation for the

0 JEOL T330 scanning electron microscope (SEM) to examine the morphology of

0 Fisherscope H 100 ultramicrohardness tester equipped with a Vickers indenter to

coatings;

evaluation of film structure;

transverse fracture surfaces of the coated samples;

determine the hardness.

174 M. 7: Vieiru et al.

Among the different techniques for adhesion measurement of coatings the scratch test is probably the most popular as it is quick, reproducible and easy to per- form [lo]. However, what is actually measured is the so-called practical adhesion defined as the “force or work required to remove or detach a film or coating from the substrate irrespective of the locus of failure” [ l l ] . The practical adhesion is function of fundamental or intrinsic adhesion but also of many other factors such as the stresses in the coating, the thickness and mechanical properties of the coating, the mechanical properties of the substrate, the mode of failure and the technique used for adhesion measurement [ 1 1, 121. A European standard [ 131 on the determi- nation of practical adhesion by scratch test is being developed in order to establish a standard procedure. In this study the scratch tests were performed with a CSEM- Revetest fitted with an acoustic emission detector. Ten scratches were made on each sample and for all of them the load that gave rise to the first cohesion failure, Lcl , and the load that was responsible for the first adhesion failure, Lc2 were deter- mined. The tests were performed according to the following conditions: a 200 p m diamond tip was used with a scratching speed of 10 mms-’ and a loading rate of 100 N min-’ . The load, applied progressively, varied from 0 N to 80 N. The obser- vation of the indentation channels by optical microscopy allowed the determination of the critical loads, Lcl and Lc2.

3. RESULTS AND DISCUSSION

3. I . Preliminary studies

To obtain the desired thickness of the interlayer and hard coatings it was neces- sary to first establish the deposition rate of the titanium-aluminium layer and of the tungsten nitride. After selecting a nitrogedargon ratio of 1.5 for the tung- sten nitride coatings, a deposition rate of 0.85 nm s-l was obtained using the con- ditions presented in Table 1. The thickness measurements of titanium and all titanium- aluminium coatings were used to obtain an average deposition rate, in spite of its slight increase with the percentage of aluminium in the Ti target. This

Table 1. Deposition parameters used on the production of coatings by sputtering

Number of Power 1 Power 2 Deposition Deposition aluminium (kW) (kW) pressure 1 (Pa) pressure 2 (Pa) micro foils

- 2 - 4 x 10-1 -

0 2 0.5 4 x 10-1 3 x 10-1 2 2 0.5 4 x 10-1 3 x 10-1 4 2 0.5 4 x 10-1 3 x 10-1 8 2 0.5 4 x 10-1 3 x 10-1

12 2 0.5 4 x 10-1 3 x 10-1

Titanium-based interlayer and adhesion of ceramic coatings 175

may be attributed to the higher ejection rate of aluminium compared to that of ti- tanium [14]. An average deposition rate of 0.4 nms-' was obtained for the films that would constitute the interlayers. As for the dilution of titanium, single Ti-A1 coatings were produced using a Ti target with different numbers of A1 foils superim- posed. The compositions obtained by selecting four different numbers of aluminium foils are presented in Table 2 . The reproducibility of the results was guaranteed by the analysis of three different coatings for each condition.

The structure identification of the ceramic coatings indicated the presence of the W2Nl+x fcc phase [15]. As for the titanium coatings a hcp a-Ti phase was formed. The structure of the titanium-aluminium coating (Fig. 1) consisted of a-Ti solid solution, where the aluminium seems to substitute for titanium [ 151.

The hardness of the titanium- aluminium sputtered coatings with different A1 contents (10 < H < 14 GPa) was higher than that of the titanium coating (7 GPa). The hardness of the Ti-A1 interlayer is between that of the substrate (8.6 GPa) and that of the ceramic coating (24 GPa), thus enhancing the matching of the mechanical properties of substrate and ceramic coating, which is favourable to adhesion improvement.

Table 2. Composition of the titanium-aluminium coatings on high-speed steel as a function of the number of aluminium microfoils

Number of Titanium Aluminium aluminium (at.%) (at.%) microfoils

2 4 8

12

93.0 81.8 71.9 60.0

7.0 18.2 28.1 40.0

176 M. T Weka et al.

3.2. Ceramic/interlayer coatings

In order to compare with the practical adhesion results of the bilayer coatings, a single tungsten nitride coating with the same 4 p m thickness was produced. The thickness should be the same because an apparent increase of the interface strength is observed with the increase of coating thickness [16, 171. Thicker layers may require a greater surface stress to achieve the same shear stress at the interface, resulting in higher critical loads.

An observation of the coatings cross section (Fig. 2) allows to identify their morphology and to distinguish the interlayer and the ceramic coating. According to Thornton’s model, the morphology of the ceramic coating is type T with a low density and the interlayer is much denser [18]. The cross-sectional images also make possible to confirm the thickness values measured using the rugosimeter.

Concerning the practical adhesion, there are different kinds of adhesion failures defined in the literature [19, 201. In the case of the bilayer coatings the failures occur at the interlayer/substrate interface, as the X-ray maps of Fig. 3 confirm by the substrate exposure on the scratch tracks. Before failing at the interface with the substrate the coatings exhibited conformal cracking which is a cohesion failure. The critical load, L,1, was around 22 N for coatings both with and without interlayer. The maintenance of the cohesion critical load was expected since the tungsten nitride top coating was the same.

Photomicrographs of the end part of the indentation scratch (higher loads) are shown in Fig. 4. As expected, the single ceramic coating behaved like a brittle coating with a typical adhesion failure -buckling, Fig. 4a [ 181. Among the bilayer coatings different types of adhesion failures were observed. When the interlayer consisted of pure titanium, which is a ductile metal, Fig. 4b, large coating portions were removed from the inside of the scratch track (rubbing out). When the stylus

-f ceramic I coating

3 interlayer

Figure 2. SEM cross section image of a titanium interlayer + tungsten nitride coating on high-speed steel.

Titanium-based interlayer and adlzesion oj ceramic coatings 111

Middle part of the channel End part of the channel

(b)

Figure 3. Iron X-ray maps of indentation channels in bilayer coatings. (a) Ti interlayer on high-speed steel. (b) Ti-A1 interlayer on high-speed steel.

advances, the titanium interlayer undergoes plastic deformation, which gives rise to an extensive cracking. The small pieces of the coating are then easily detached from the substrate, preferentially inside the track. Taking this into account, and also from the observation of Fig. 3a, a low critical load for adhesion failure is expected for the tungsten nitride/pure Ti coatings. If instead of pure titanium a less ductile Ti-A1 interlayer is used this behaviour is not observed. Chipping on the tracks edges occurred for all the bilayer coatings with a less ductile Ti-A1 solid solution as interlayer, Fig. 4c.

The critical adhesion loads were measured by optical microscopic observations of the scratch tracks. Figure 5 represents the average and standard deviation of critical adhesion failure loads for the various samples studied. Taking into consideration the standard deviation and considering as well the nature of the scratch test, it is possible to conclude that the scatter associated with the ten tests performed for each sample is in general quite low. The single ceramic coating has a critical adhesion failure load (31 N) close to the tungsten nitride/titanium bilayer coating (32 N). In Fig. 5 it is

178 M. 7: Vieiru et al.

Figure 4. Adhesion failures exhibited by the coatings deposited onto high-speed steel. (a) Tungsten nitride single coating (buckling). (b) Tungsten nitride coating + titanium interlayer (rubbing out). (c) Tungsten nitride coating + titanium- aluminium interlayer (chipping).

Titanium-based interlayer and adhesion of ceramic coatings 179

0 10 20 30 40 50

Ai at % in the interlayer

Figure 5. Critical adhesion failure load (Lc2) of the bilayer coatings on high-speed steel as a function of the interlayer aluminium content.

clear that in the presence of a pure titanium interlayer the resulting critical adhesion failure load is significantly lower than when titanium-aluminium was used as interlayer. On one hand, it was expected that the adhesion of a single ceramic coating would be lower than in presence of a reactive interlayer. On the other hand, the extensive plastic deformation of the Ti ductile interlayer could be responsible for a poor adhesion, as already mentioned. The rubbing out of the coating might be less pronounced if a thinner Ti interlayer is used. In fact, various authors concluded that an optimum thickness should be determined for ductile interlayers under ceramic coatings [ 3 , 4, 91. Moreover, with a pure titanium interlayer the excessive reaction could make the interface brittle. The titanium- aluminium interlayer could promote the formation of chemical bonds with the substrate by reacting with the oxygen from the metallic substrate surface and, consequently, the adhesion could be enhanced (A1 from 7 to 40 at.%). The dilution of the titanium with the aluminium, by reducing the titanium activity, prevents an excessive reaction and consequently the formation of an undesirable Ti02 thin layer at the interface. The degree of dilution does not seem to substantially influence the practical adhesion. Nevertheless, the higher critical adhesion failure load was achieved when the interlayer had 72 at.% Ti and 28 at.% Al. Other contents of aluminium in the titanium solid solutions result in a lower critical adhesion failure load. So, probably, the above composition is responsible for an optimum titanium activity promoting the formation of T i 0 and avoiding an excess of reaction. The aluminium concentration could also influence the stresses developed in the titanium structure that can be too high when an optimum A1 concentration is exceeded.

The practical adhesion improvement can be achieved through the interfacial chemical bonding enhancement, but the mechanical properties also play an im- portant role in the adhesion mechanism [2, 211. In the present case the practi- cal adhesion improvement could also result from the mechanical properties of the titanium-aluminium interlayer. In fact, the Ti-A1 interlayers may work as a better interface between the steel substrate and the ceramic coating. The hardness and ductility gradient across the coated system could promote a better adhesion to the substrate.

180 M. T Vieira et al.

4. CONCLUSION

The following conclusions can be drawn from this study: 0 An interlayer consisting of a solid solution of aluminium in titanium is suitable to

improve the practical adhesion between a tungsten nitride coating and an alloyed steel.

0 The optimum practical adhesion is obtained when the chemical composition of the interlayer is 72 at.% Ti and 28 at.% Al.

0 The study of the role of titanium-aluminium solid solution interlayers suggests that this interlayer could improve the adhesion of other coated metallic substrates; but further studies are necessary in order to distinguish the contribution of the chemical reactivity and mechanical properties to adhesion failures.

REFERENCES

1. U. Helmersson, B. 0. Johansson, J.-E. Sundgren, H. T. G. Hentzell and P. Billgren, J. Vac. Sei.

2. D. S. Rickerby, S. J. Bull, T. Robertson and A. Hendry, Surface Coatings Technol. 41, 63 (1990). 3. J. Tang, L. Feng and S. Zabinski, Surface Coatings Technol. 99,242 (1998). 4. J. Valli, U. Makela, A. Matthews and V. Murawa, J. Vue. Sci. Technol. A3, 241 1 (1985). 5. M. T. Vieira and A. S. Ramos, J. Mater. Proc. Technol. 92/93, 156 (1999). 6. M. G. Nicholas, Joining ofCeramics. Chapman and Hall, London, 1990. 7. J. H. Westbrook and R. L. Fleischer, Intermetallic Compounds, Vol. 1. John Wiley, New York

8. U. R. Kattner, J. C. Lin and Y. A. Chang, Metull. Trans. 23A, 2081 (1992). 9. D.-F. Lii, J. L. Huang and M. Lin, Surface Coatings Technol. 99, 197 (1998).

Teclznol. A3, 308 (1985).

(1995).

10. K. L. Mittal, Electrocomponent Sci. Technol. 3, 21 (1976). 11. K. L. Mittal, in: Adhesion Measurement of Films and Coatings, K. L. Mittal (Ed.), pp. 1-13.

12. K. L. Mittal, in: Adhesion Measurement of Thin Films, Thick Films and Bulk Coatings,

13. European Prestandard, PrEN 1071-3 “Determination of adhesion and other mechanical failure

14. J. C. Pivin, J. Mater. Sci. 18, 1267 (1983). 15. International Center for Diffraction Data, Newtown Square, PA, USA. 16. A. J. Perry, Thin Solid Films 107, 167 (1983). 17. J. H. Je, E. Gyarmati and A. Naoumidis, Thin Solid Films 136, 57 (1986). 18. J. A. Thornon, J. Vac. Sci. Technol. 11, 666 (1974). 19. P. J. Burnett and D. S. Rickerby, Thin Solid Films 154.403 (1987). 20. J. Palmers, M. Van Stappen, J. D’Haen, M. D’Olieslaeger, L. M. Stals, G. Uhlig, M. Foller and

21. N. Laidani, A. Miotello, L. Guzman, S. Tuccio and L. Calliari, J. App. P h y . 76, 285 (1994).

VSP, Utrecht, The Netherlands (1995).

K. L. Mittal (Ed.), STP No. 640, pp. 5- 17. ASTM, Philadelphia (1978).

modes by a scratch test” (1999).

E. Haberling, Surface Coatings Technol. 74-75, 162 (1995).

Adhesion Aspects of Thin Films, Vol. 1, pp. 181-193 Ed. K. L. Mittal 0 VSP 2001

Effect of annealing on residual stress, strength, adhesion and wear resistance of thin, hard coatings on low alloy steel

THE0 Z. KATTAMIS '.* and COSTAS G. FOUNTZOULAS Department of Metallurgj and Materials Engineering, Universioi of Connecticut, Storrs,

Army Research Laboratory, APG, M D 21005-5096 CT 06269-3136

Abstract-The effect of residual stresses and their evolution during annealing were investigated on steel specimens coated with thin, hard coatings. Amorphous hydrogenated silicon carbide and amorphous silicon-containing diamond-like carbon (Si-DLC) thin coatings were deposited on AIS1 4340 low alloy steel specimens. The respective deposition methods used were plasma-enhanced chemical vapor deposition and Ar' ion beam-assisted physical vapor deposition of tetraphenyl- tetramethyl-trisiloxane (704 Dow Corning diffusion pump oil). During annealing in an argon atmosphere the residual stress attributed to hydrogen entrapment during deposition gradually changed from compressive to tensile due to loss of hydrogen. The rate of stress increase decreased with increasing annealing time. The cohesion and adhesion failure loads and the abrasive wear resistance decreased with increasing annealing time, as did the friction coefficient between the coating and a diamond stylus.

Keywords: Silicon-containing diamond-like carbon coatings; S i c coatings; alloy steel substrates; residual stresses; cohesion failure load; adhesion failure load; tribological properties.

1. INTRODUCTION

Transient or residual stresses in thin coatings may originate from a mismatch in thermal expansion coefficients between the substrate and the coating [ 1-41, as well as from a lattice mismatch between them, and from growth defects within the coating and grain impingement [ 5 ] . They may also originate from the entrapment of hydrogen within the coating, as was observed in amorphous S i c films obtained by plasma-enhanced chemical vapor deposition (PECVD) on carbon fibers and single-crystal silicon wafers [6]. In these films the entrapped hydrogen led to the establishment of biaxial residual compressive stresses. After annealing, the decrease

*To whom all correspondence should be addressed. Telephone: (860) 4864718; Fax: (860) 4864745; E-mail: tkattami @mail.ims.uconn

182 T. Z. Kattamis and C. G. Fountzoulas

in hydrogen content led to a significant reduction in coating thickness and a gradual change of biaxial compressive stresses into tensile.

Residual stresses are expected to affect coating microhardness. A coating under residual compressive stress resists penetration of the indenter and hence exhibits a measured apparent microhardness which is higher than its intrinsic value. Under residual tensile stresses the opposite trend was observed [7]. Residual stresses are also expected to affect adhesion to the substrate, as well as wear resistance. For example, Jindal et al. [SI established that for TiN-coated WC-Co cemented carbide substrates an increase in residual compressive stress led to a reduction in measured adhesion failure load. Assuming that for a given coating thickness a certain critical compressive stress is required for buckling or delamination, the applied stress that causes delamination will be reduced in presence of a higher compressive residual stress in the plane of the coating. Buckling and delamination by excessive compressive stresses were reported elsewhere [9, 101. The beneficial effect of residual compressive stresses on wear resistance was also shown [ I 1, 121.

Si-DLC, as well as S i c coatings were deposited on AISI 4340 low alloy steel specimens by two different methods and their properties were investigated. The work reported here is basically focused on the effect of annealing on residual stresses within the coating, and on coating strength, adhesion to the substrate and wear resistance.

2. EXPERIMENTAL PROCEDURE

2. I . Specimen preparation

2.1.1. Si-DLC-coated specimens. AISI 4340 low alloy steel specimens (20 mm in diameter and 5 mm thick) were polished down to 0.05 p m alumina and coated on one side with Si-DLC, using the ion beam assisted deposition (IBAD) process. These specimens were used to evaluate the strength, adhesion to the substrate and tribological properties of as-deposited coatings. The actual composition of 4340 steel was: 0.41%C, 0.71%Mn, 0.18%Si, O.S2%Cr, 1.75%Ni, 0.26%Mo, balance Fe. A Zymet 100 non-mass analyzed ion implanter was used for the preparation of the DLC coatings. As previously described [13, 141 DLC coatings formed upon energetic ion bombardment of a vapor-deposited precursor material: tetraphenyl- tetramethyl-trisiloxane (704 Dow Coming diffusion pump oil). The substrates were initially cleaned in methanol and acetone and subsequently sputter-cleaned with a 40 kV Ar+ ion beam (40 pAcm-* for 10 min). A water-cooled sample stage maintained the temperature of the substrates during deposition close to room temperature. The base pressure was 2 x lop6 Torr and the deposition was carried out at 3 x Torr. The precursor material was deposited using a 3 mm diameter nozzle connected to a heated oil reservoir. The oil temperature was maintained at about 140°C as monitored by a thermocouple. The substrate was inclined at 45" with respect to both the horizontal ion beam and the vertically aligned oil bath nozzle. A shutter was placed above the oil container to start and stop the oil vapor flow. The

Effect of annealing on residual stress, strength, adhesion and wear resistance 183

growing coating surface was continuously bombarded by a slowly oscillating (1 Hz) ribbon (2 cm x 20 cm) of Ar+ ion beam at 40 keV. The bombardment time used was 4 h. The peak ion current density in the beam was varied from 55 p A cm-* to 140 p A cm-*, as read by a Faraday cup.

For evaluating residual stresses within the coatings a second series of amorphous Si-DLC-one-side-coated AISI 4340 steel foils (20 mm in diameter and about 200 p m thick) were also processed under identical conditions. The coated specimens were placed in Vycor glass tubes which were evacuated and back-filled with Ar. The annealing in the Ar atmosphere was conducted at 200°C for times up to 30 min.

2.1.2. Sic-coated specimens. A first series of amorphous SiC-one-side-coated AISI 4340 low alloy steel foils (20 mm in diameter and about 200 p m thick) were prepared by PECVD, using the equipment built and described by Landis [15]. These specimens were used for residual stress evaluation. A glow discharge was generated within the deposition chamber (450 mm in diameter and 225 mm high). The deposition conditions were 50 mTorr Ar pressure, and a gas mixture consisting of SiH4 at a flow rate of 4.2 sccm, CH4 at a flow rate of 6.7 sccm, and Ar at a flow rate of 10 sccm, with 100 W supplied to the upper (target) electrode, while the lower (substrate) electrode was grounded. The spacing of the two 150 mm diameter electrodes could be changed from 25 mm to 125 mm. The accelerating potential could be varied between 25 and 110 V. Ion species within the plasma were analyzed by plasma ion mass spectrometry (PIMS), or plasma ion energy spectroscopy (PIES). The specimens were prepared with an ion bombardment energy of about 15 eV, as determined by PIES. A second series of specimens (wafers 20 mm in diameter and 5 mm thick) were prepared in a similar fashion and were used to evaluate cohesion, adhesion to the substrate and tribological behavior of the coatings. Similar coatings were found to be amorphous and to contain an excess amount of trapped hydrogen which is responsible for the generation of residual compressive stresses [15]. To study the effect of gradual hydrogen elimination on the nature and magnitude of residual stresses, several thin specimens from the first series were encapsulated in fused silica tubes in a vacuum of about 5 mTorr and annealed at 650°C for 5, 10, 15, 20, 25 and 30 min.

2.2. Measurement of nanohardness

A mechanical properties microprobe (Nanoindenter, Nan0 Instruments, Inc., Knox- ville, TN, USA) was used to determine both nanohardness and Young’s modulus of specimens. From the force-displacement curve plotted during the nanoindentation test, hardness was calculated at any point during the loading part of the cycle which includes both elastic and plastic contributions [ 16- 181. However, Young’s modulus could only be calculated during the unloading part of the cycle which contains no contributions from plastic deformation. The contact area was calibrated vs. depth by producing very shallow indents and imaging replica films using TEM. The hardness of the coating was taken as being equal to the force required to bring about plastic

184 7: Z. Kutrarnis and C. G. Fountzoulas

penetration, divided by the projected indenter/ specimen contact area. The elastic modulus was determined by linearly fitting the intial few points during unloading and dividing by the indenter contact area [ 161. For both types of films a total elastic plus plastic indentation depth of about 100 nm and a constant indentation loading rate of 100 mN/s were used. For as-deposited Si-DLC coatings the 100 nm depth was attained with a load of about 9 mN, whereas for Sic , with a load of about 2 mN. The ratio of this indentation depth to film thickness does not exceed the critical value of 1/10 at and below which the contribution of the substrate is negligible.

2.3. Measurement of residual stresses

The residual stress, a,, in the amorphous DLC and S i c coatings on 20 mm disk- shaped steel foils was evaluated by measuring the curvature of the foils after deposition on only one side, using Stoney’s expression [19]:

OC = E,t,”/6(1 - u,)t,R,

where R is the spherical radius of curvature of the substrate steel disk-shaped foil, E, and v, are its Young’s modulus and Poisson’s ratio, respectively, and ts and tc are the thicknesses of the substrate and coating, respectively. This expression is valid for a coating which is substantially thinner than the substrate. The coating surface will assume a convex shape, if the residual stress within the coating is compressive and a concave shape, if it is tensile. The curvature of the coated specimen was measured with a Dektak I1 profilometer. If d is the trace half-length and h is the vertical displacement, the radius of curvature may be approximated by R = d2/2h . The coating thickness was measured metallographically and by scanning electron microscopy (SEM) on random transverse cross sections of the coated specimen. Results were cross-checked by measuring the depth of the substrate from the coating surface by interference optical microscopy at random locations where the substrate was exposed following surface scratching. Five traces were made on each specimen and five radii of curvature were determined. The average value was adopted as the radius for that specimen.

2.4. Evaluation of the coating cohesion, adhesion and suvface friction coeficient

The cohesion failure load, Lc , and adhesion failure load of the coating L A were evaluated by automatic scratch testing with a CSEM-Revetest, as described elsewhere [20]. During testing the loading rate was dL/dt = 100 Nmin-’ and the table translation speed d x / d t = 10 mmmin-’, yielding a loading gradient dL/dx = 10 Nmm-’. The acoustic emission (AE) signal intensity, the frictional force Ft and the friction coefficient p* were plotted vs. normal load F, between 0 and 70 N. The friction coefficient was also plotted vs. time at constant loads of 5, 10 and 15 N. Five scratches were performed per specimen.

Effect of annealing on residual stress, strength, adhesion and wear resistance 185

2.5. Evaluation of the abrasive wear resistance

A pin-on-disk apparatus was used to evaluate the abrasive wear resistance of the coatings, following the procedure described previously [21]. The pins used (8 mm x 8 mm) were carefully sectioned with a diamond saw from the various coated specimens. The disk was covered with a cloth which was uniformly coated with 1 p m average particle size diamond paste along a 25 mm wide band that included the pin trajectory. An average normal pressure PN = 0.581 MPa was applied to the specimen holder. The tests were performed in air. The diamond paste was periodically soaked with polishing fluid. The rotational velocities were set with a stroboscope at 50, 100 and 200 rev min-' and corresponded to tangential velocities of 0.22, 0.434 and 0.87 m s-l, respectively. Each test was interrupted often to measure the coating thickness reduction. Measurements of abrasive wear resistance are reported as specific wear rate W, (mm3 N-' m-l ), w here W, = A ~ / P N L , A t (nm) is the thickness reduction measured metallographically on polished transverse sections and L (m) is sliding distance. In this expression, L = 2nrNt , where N (rev min-') is the rotational velocity, r = 0.0415 m is the radius of the pin trajectory and t (min) is time. Four measurements were made for each as-coated or annealed specimen.

3. RESULTS AND DISCUSSION

3.1. Coating thickness and composition

Observations made on the prepared Si-DLC-coated 4340 steel specimens in the as-coated condition were similar to those reported previously [14]. Thus, the as- deposited coating thickness was fairly uniform and its average value increased with increasing beam current density. The average coating composition was measured by Rutherford backscattering spectrometry (RBS) for all elements except hydrogen, which was determined by forward elastic recoil analysis. The average stoichiometry of the as-deposited coating was C67Si9O6HI5Ar3. Following a 30 min annealing at 200°C in argon the stoichiometry changed to C75Si1008H3Ard. There were obviously a substantial loss of hydrogen, small increases in silicon, oxygen and argon contents and a significant increase in carbon. Coating thickness varied between 1.85 and 4.31 p m and was reduced by about 7 pct after a 30 min annealing.

It was established by Landis [15] that S i c coatings deposited on silicon wafers under similar conditions as those that prevailed in the present investigation consisted of amorphous hydrogenated silicon carbide (a-SiC:H). It is, therefore, assumed the S ic coatings deposited and studied in the present investigation have the same composition. Sic coating thickness varied between 0.99 and 1.23 p m and was reduced by about 9 pct after a 30 min annealing.

3.2. Residual stresses

In amorphous hydrogenated silicon carbide (a-SiC:H) and Si-DLC coatings the entrapped hydrogen is most likely in solid solution and leads to positive misfit in

186 7: Z. Kattamis and C. G. Fountzoulas

-600 ' I 0 10 20 30 40

ANNEALING TIME, mln

Figure low alloy steel specimens, annealed at 200°C and 650°C, respectively.

Residual stress in coating vs. annealing time. Si-DLC-coated and Sic-coated AIS 43

the plane of the coating and to biaxial compressive stresses [6]. During annealing the loss of hydrogen causes substantial shrinkage and reduction in thickness, and results in the development of biaxial residual tensile stresses [15]. Landis [15] had observed that in Sic-coated silicon wafers, at coating thicknesses in excess of 0.75 pm, tensile stresses supplied the strain energy required for spontaneous delamination and separation of the entire coating from the substrate. In the present investigation coating thickness exceeded 0.75 pm, yet no such delamination was observed.

The results of residual stress calculation using equation (1) are plotted vs. annealing time in Fig. 1 for both types of coatings. As an example, for Si- DLC-coated steel specimen in the as-coated condition d = 11.02 x lop3 m, h = 11.3 x m, hence R = 5.36 m, t, = 200x lop6 m, tc = 0.92 x lop6 m, and for AIS1 4340 ( E , = 211 x lo3 MPa and us = 0.287) the calculated compressive (-) residual stress a, is 400.08 MPa. In this calculation the actual coating thickness which changes with annealing time was used. For both coatings the coating stress gradually changed during annealing from compressive (-) to tensile (+). The rate of increase of the residual stress decreases with increasing annealing time. Devitrification which is operative during annealing has been neglected. The partial or total transformation of glass into crystalline material would be expected to cause a reduction in specific volume and relaxation of compressive stress, allowing the transition from residual compression to tension to occur after a shorter annealing time, hence shifting the curve of Fig. 1 to the left. Figure 2 exhibits the decrease in Hloo, the nanohardness corresponding to an indentation depth of about 100 nm, with increasing annealing time. The increase in hardness under residual compressive stress conditions was reported by LaFontaine et al. [7]. It is clear that both coatings

-Si-DLC-coated AIS1 4340 Low Alloy Steel --+-SIC-coated AIS1 4340 Low Alloy Steel

ANNEALING TIME, min

Figure 2. Coating nanohardness vs. annealing time. Si-DLC-coated and Sic-coated AIS1 4340 low alloy steel specimens.

behave in a similar fashion during annealing. The nanohardness of Si-DLC is for all annealing times higher than that of Sic.

3.3. Coating cohesion and adhesion failure loads

The cohesion and adhesion failure loads for both coatings on 4340 steel are plotted in Fig. 3 vs. annealing time. Whereas the residual stresses were measured on coated thin 4340 steel foils, the measurement of cohesion and adhesion failure loads was carried out on coated thicker specimens. It was assumed that under the same processing and annealing conditions the residual stresses in coatings of the same thickness deposited on substrates of different thicknesses are the same. Let the length 1 of a coating area of unit width and thickness tc shrink by AI after a given annealing treatment. Assuming that the substrate is rigid and will not shrink, the residual stress a, within the coating can be approximated by:

= (AZ/O(&/l - u,>, (2)

where E, and u, are Young’s modulus and Poisson’s ratio of the coating, respec- tively. Thus, a, appears to be independent of the substrate thickness. On the con- trary, in the absence of any externally applied stress, the average residual stress within the substrate, as, can be expressed by:

0 s = ( & / t s > a c (3)

and therefore depends on the relative thickness of the coating with respect to the substrate. Figure 3 shows that for both coating systems the cohesion failure load, Lc , decreases with increasing annealing time. Contributing factors can be the decrease in residual compressive stress which ultimately changes to tensile, a change in chemical composition with loss of hydrogen, as well as devitrification

188 7: Z. Kattamis and C. G. Foiintzoulas

20 8 I I I I

- - v 4 J J

o Si-DLC-coated AIS1 4340 Low Ailoy Steel 1 8 1 : n Sic-coated AIS1 4340 Low Ailoy Steel

16

1

i

1 0 1 0 20 3 0 4 0

0 '

ANNEALING TIME, rnin

Figure 3. Cohesion and adhesion failure loads vs. annealing time in coating. Si-DLC-coated and Sic-coated AIS1 4340 low alloy steel specimens.

followed by grain growth. Tensile cracking in the scratch occurred at the trailing edge of the stylus, as exemplified by Fig. 4 for a Si-DLC-coated specimen with a residual stress of -50.2 MPa (microcracks A) and a Sic-coated specimen with a residual stress of -81.2 MPa (microcracks B). At the edges of the scratch track the tensile cracks deviate from the circular shape, appearing at an angle of about 45" with respect to the scratching direction. Secondary tensile cracks appear to be responsible for localized chipping, which is sporadically observed and becomes more frequent with increasing applied load. The imposed frictional stress a at a given location is proportional to the applied normal load. This stress is tensile behind the advancing stylus and compressive ahead of it. Let OT and OB be the resistances of the coating to tensile cracking and buckling, respectively, both of which are expected to change with annealing, and oC be the residual stress. In the present case no buckling failure has been observed; hence, UT < DB. Tensile cracking at a given location will occur when

0~ + 0 > aT. (4)

Behind the advancing stylus, a > 0. If a, < 0 (compressive stress) a higher cr will be tolerated prior to cracking; hence measured Lc will be higher. With increasing annealing time, a, decreases in absolute value and eventually becomes positive (tensile). A lower imposed frictional stress or applied normal load will then be sufficient to cause cracking.

Similar considerations apply to crack propagation, and hence to the adhesion failure load. This is the trend exhibited in Fig. 3 for both types of coatings. In the as-coated condition the adhesion failure load of Si-DLC is about 14 N and is reduced to about 10 N with a 30 min annealing. For the S i c coating the respective adhesion failure loads are about 4.2 N and 3.1 N. They are significantly lower than for Si-DLC. These load measurements are fairly precise, because the AE signal

Effect of annealing on residual stress, strength, adhesion and 1vear resistance 189

Figure 4. Photomicrographs of scratches on (a) Si-DLC-coated and (b) Sic-coated AIS1 4340 low alloy steel specimens corresponding to residual stresses of -50.2 MPa and -81.2 MPa. respectively. Tensile cracks are shown at locations A and B.

190 7: Z. Kattamis and C. G. Fountzoulas

is much stronger than the noise generated by the friction of the diamond stylus against the coating surface, hence crack initiation and the onset of delamination can be unambiguously identified. Again, the decrease in adhesion failure load with annealing time can be attributed to changes in stress state, microstructure and chemical composition.

3.4. Friction coeficient

The coefficient measured with the automatic scratch tester is in fact a scratching coefficient between the diamond stylus and the coating surface and may be defined as p* = pa + pp, where pa is the adhesion component and pp the ploughing component [22]. The instrument used does not allow direct, accurate measurement of p* below a normal load of about 2.5 N. To circumvent this obstacle, a previously reported procedure [22] was followed. The coefficient p* was measured vs. time at constant loads of 5, 10 and 15 N. Average values of p* per scratch were plotted vs. normal load and the resulting curves were extrapolated to a load of 1 N. The calculated value, p* (1 N) was arbitrarily taken as equal to the actual friction coefficient and plotted in Fig. 5 vs. annealing time. These values of p* were measured between a diamond stylus and the Si-DLC and S i c coatings at sliding velocities of 1.66 x ms-l, respectively. For Si-DLC they are somewhat higher than those reported in the literature, which were measured between a steel ball and Si-DLC coatings at sliding velocities of 0.04-0.5 ms-' [23] and 0.2 ms-' [24]. The higher p* values measured here could possibly be attributed to the lower sliding velocity used and the chemical similarity of the two counterbodies [25]. Figure 5 shows that the average friction coefficient of S i c decreases with increasing annealing time. This is difficult to explain because the possible formation of silica, due to residual oxygen within the

ms-' and 1.80 x

5 u r! Y Y W 0 0 Z

I-

K U

0 o

Si-DLC-coated AIS1 4340 Low Alloy Steel 0 Sic-coated AIS1 4340 Low Alloy Steel

0 10 20 30 40

ANNEALING TIME, min

Figure 5. Coating average friction coefficient vs. annealing time. Si-DLC-coated and Sic-coated AIS1 4340 low alloy steel specimens.

Effect of annealing on residual stress, strength, adhesion and wear resistance 191

annealing tubes would have led to an increase in friction coefficient [26]. The same figure also shows a small decrease in average friction coefficient of DLC, presumably due to its transformation into graphite which is more common at higher temperatures. The decrease in friction coefficient with increasing annealing time implies a parallel decrease in frictional force corresponding to a given applied normal load. The increase in measured L A with decreasing friction coefficient, when all other conditions remain unchanged, has been discussed elsewhere [27,28]. It can be concluded that Lc and L A measured at increasing annealing times are larger than those which could be measured, if the friction coefficient remained unchanged and equal to that of the as-coated specimen.

3.5. Abrasive wear resistance

The dependence of specific wear rate of the two types of coatings on annealing time is illustrated in Fig. 6. Each point represents an average of three tests of equal duration. For as-coated Si-DLC in the specimens the wear rate is about 4 x mm3 N-' m-l f or specimens annealed for 30 min.

The same trend is observed for Sic, with the specific wear rate increasing from about 8 x lop7 mm3 N-' m-' i n the as-coated condition to about 6 x

mm3 N-' m-l after a 30 min annealing. Again, the increase in specific wear rate with annealing time can be attributed to changes in stress state, microstructure and chemical composition. The specific wear rate is plotted in Fig. 7 vs. adhesion failure load for both coatings and shows an improvement in abrasive wear resistance with enhanced adhesion. The beneficial effect of a residual compressive stress on adhesion and wear resistance was reported elsewhere [l 1, 29, 301. Lin et al. [ 111 concluded that a residual compressive stress in Fe2B surface layer on boronized

mm3N-'m-' a nd increases up to about 1 x

l0-'l -Si-DLC-coated AIS1 4340 Low Alloy Steel --SIC-coated AIS1 4340 Low Alloy Steel

0 1 0 2 0 30 4 0

ANNEALING TIME, min

Figure 6. Specific wear rate vs. annealing time. Si-DLC-coated and Sic-coated AIS1 4340 low alloy steel specimens.

192 7: Z. Kattamis and C. G. Fountzoulas

i

3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 10.8' " " " " " " " " " " " "

LA, ADHESION FAILURE LOAD, N

Figure 7. Specific wear rate vs. adhesion failure load. Si-DLC-coated and Sic-coated AIS1 4340 low alloy steel specimens.

medium carbon steel could improve the wear resistance under low loads, as is the case in the present investigation. On the contrary, high residual compressive stresses may promote buckling and eventual delamination [3 11, thus reducing the wear resistance. The detrimental effect of residual tensile stresses on wear properties was reported also by Zum Gahr [32].

4. CONCLUSIONS

During annealing, the residual stress in amorphous, hydrogenated Si-DLC and S i c coatings on 4340 low alloy steel gradually changes from compressive to tensile. The rate of increase in residual stress decreases with increasing annealing time. Both cohesion and adhesion failure loads of annealed specimens, as well as their abrasive wear resistance decrease with increasing annealing time. The friction coefficient between the coating surface and a diamond stylus decreases with increasing annealing time.

REFERENCES

1. T. Yamamoto and K. Kamachi, J. Jpn. Inst. Metals 49, 120 (1985). 2. A. J. Perry, Thin Solid Films 146, 165 (1987). 3. D. S. Rickerby, J. Vac. Sci. Technol. A6, 2809 (1986). 4. D. S. Rickerby, B. A. Bellamy and A. M. Jones, Surf: Eng. 3, 138 (1987). 5. D. S. Rickerby, A. M. Jones and A. J. Perry, Surf: Coat. Technol. 36,631 (1988). 6. A. S. Argon, V. Gupta, H. S. Landis and J. A. Cornie, Mater: Sci. Eng. A107,41 (1989). 7. W. R. LaFontaine, B. Yost and C.-Y. Li, J. Mater: Res. 5,776 (1990). 8. P. C. Jindal, D. T. Quinto and G. J. Wolfe, Thin Solid Films 154, 361 (1987).

Effect of annealing on residual stress, strength, adhesion and wear resistance 193

9. D. B. Marshall and A. G. Evans, J. Appl. Phys. 56, 2633 (1984). 10. D. B. Marshall and A. G. Evans, J. Appl. Phys. 56,2639 (1984). 11. Z. Lin, Z. Wang and X. Sun, Wear 138,285 (1990). 12. S. Janowski, J. Senatorski and W. Szyrle, Eurotrib 85, €'roc. 4th European Tribologj Congr.,

13. C. G. Fountzoulas, J. D. Demaree. W. E. Kosik, W. Franzen, W. Croft and J. K. Hirvonen, Mater:

14. T. Z. Kattamis, C:. G. Fountzoulas. J. K. Hirvonen and C. V. Cooper, J. Adhesion Sei. Technol.

15. H. S . Landis, Ph.D. Thesis, Massachusetts Institute of Technology (1988). 16. C. P. Beetz, Jr., C. V. Cooper and T. A. Perry, J. Mater: Res. 5, 2555 (1990). 17. E. H. A. Dekempeneer, R. Jacobs, J. Smeets, J. Meneve, L. Eersels, B. Blanpain. J. Roos and

18. M. F. Doerner and W. D. Nix, J. Mater: Res. 1. 601 (1986). 19. G. C. Stoney, Proc. Roy. Soc. London, Ser. A 82, 172 (1909). 20. T. Z. Kattamis, K. J. Bhansali. M. Levy. R. Adler and S. Kamalingham, Muter. Sei. Eng. A161.

21. T. Z. Kattamis and T. Suganuma, Mater: Sci. Eng. A128, 241 (1990). 22. S. Skolianos and T. Z. Kattamis, Mater. Sei. Eng. A163. 107 (1993). 23. T. Hioki, Y. Ito, A. Ito, S. Sibi and J. Kawamoto, Surf: Coat. Technol. 46, 233 (1991). 24. K. Oguri and T. Arai, J. Matel: Res. 7, 1313 (1992). 25. I. V. Kragelskii, Friction and Wear, pp. 154, 155, 180, 184. Butterworths (1965). 26. K. Miyoshi, Surt Coat. Technol. 36, 487 (1988). 27. P. A. Steinmann. Y. Tardi and H. E. Hintermann, Thin Solid Films 154, 333 (1987). 28. S. J. Bull, D. S. Rickerby, A. Matthews, A. Leyland, A. R. Pace and J. Valli, Su$ Coat. Technol.

29. Th. Roth, E. Broszeit and K. H. Kloos, Surf: Coat. Technol. 36, 801 (1988). 30. Th. Roth, K. H. Kloos and E. Broszeit, Thin Solid Films 153, 123 (1987). 31. A. G. Evans, Mater. Sci. Eng. 71, 3 (1985). 32. K. H. Zum Gahr, Z. Metallkd. 69. 643 (1978).

Vol. IV, p. 4. Elsevier, Amsterdam (1985).

Res. Soc., Symp. Proc. 279. 645 (1993).

10,939 (1996).

D. J. Oostra. Thin Solid Films 217, 56 (1992).

105 (1993).

36, 503 (1988).

Adhesion Aspects of Thin Films, VoZ. I , pp. 195-205 Ed. K. L. Mittal e VSP 2001

Study of adhesion and tribological properties of some ceramic films

J. TAKADOUM ‘3’ and B. CRETIN ’ Laboratoire de Microanalyse des Sugaces, ENSMM, 26 chemin de l’kpitaphe 25030 Besangon, France Laboratoire de Physique et Me‘trologie des Oscillateurs, 32 avenue de l’observatoire 25044 Besangon Cedex, France

Abstract-Adhesion and tribological properties of titanium nitride (TiN), titanium carbonitride (TiCN) and diamond-like carbon (DLC) films, prepared by vapour deposition techniques, were studied. The coatings were deposited on steel substrates. The results show that hardness of the substrate affects greatly both adhesion and wear resistance.

Some characterisations of the films using the Scanning Thermoelastic Microscopy (SThEM) are also presented. We showed that this technique was a new method for thin film adhesion characterisation.

Keywords: Thin films; adhesion; tribology; scanning thermoelastic microscopy.

1. INTRODUCTION

Ceramic thin hard coatings produced by vapour deposition techniques are widely used to improve tribological properties of tools and machine parts. Among these materials, nitride and carbide films are most frequently used due their high bond strength to the substrate and their excellent resistance to wear, erosion and heat. The use of diamond-like carbon (DLC) films in tribological applications is relatively recent. These films combine high hardness (2000- 10 000 HV), low friction coefficient (0.05-0.25), and chemical inertness, which makes them very good candidates for tribological applications [ 1 -31,

Microstructure, texture, stoichiometry and consequently mechanical properties of vapour deposited films depend greatly on the technique of deposition (chemical vapour deposition, reactive ion plating, reactive magnetron sputtering, . . .) and for a given technique they are affected by the experimental conditions (nature of the gas

*To whom correspondence should be addressed. Phone: 33-3-81402857; Fax: 33-3-81402852; E-mail: Jamal.Takadoum@ens2m.fr

196 J. Takadoum and B. Cretin

used, working pressure, substrate temperature and bias voltage, target-to-specimen distance . . .) [4-71.

In the present paper we report some results concerning the characterisation of adhesion and tribological properties of TiN and TiCN obtained by reactive ion plating and DLC prepared by Physical Enhanced Chemical Vapor deposition (PECVD). We will present in the last section some results obtained using the scanning thermoelastic microscopy which is a relatively unknown technique for thin film characterization.

2. EXPERIMENTAL PROCEDURE

Two different steels were used as substrates: X85WMoCrV6542 (hardness = 880 HV) and 35CrMo4 (hardness = 250 HV). They were polished mechanically to obtain a roughness R, = 0.02 pm.

TIN and TiCN films were deposited by reactive ion plating while the DLC films were prepared using a radio-frequency plasma system [8]. The films were 3 to 4 p m thick.

Friction tests were conducted in air at room temperature using a ball-on-disk tribometer. An alumina ball with a diameter of 5 mm was used. The length of the wear track was 15 mm and the total sliding distance was 8 m. The applied load was 25 N and the sliding speed was 2 mm/s.

Adhesion was investigated with the scratch test method. The scratches were made at a loading rate of 100 Nmin-' and a diamond tip velocity of 10 mmmin-'. Scratch tests were conducted three times and the results were averaged. Internal stresses in the coatings were measured using the curvature method. Specimen size of 20 mm x 10 mm x 2 mm were cut from the coated substrate. They were polished mechanically with a S i c paper to a lmm thickness and then thinned

coating

Electochemical

Figure 1. measurements.

The different operations carried out for thinning the substrate before internal stress

Adhesion and tribological properties of some ceramic films 197

electrochemically to remove residual stresses introduced by mechanical polishing. The final thickness of the substrate was about 500 pm. The radius of curvature of the specimen was calculated from surface height measurements using a profilometer (Fig. 1).

3. RESULTS AND DISCUSSION

Figure 2 shows depth profiles obtained by glow discharge optical spectrometry (GDOS) on the three films. The interface shows an interdiffusion layer of about 1 Fm. Prior to DLC deposition an S i c interlayer about 500 nm thick was deposited to improve adhesion. Figure 2c shows the presence of this layer at the interface. Mechanical properties of the films were investigated using the nanoindentation technique. Hardness and Young’s modulus of the films were deduced from loading/unloading curves (Fig. 3). The values obtained are presented in Table 1. They are comparable to those reported in other studies [ l , 7, 91.

Internal stresses measured with the curvature method showed that in all cases the stresses were compressive. The results are presented in Table 2 where they are compared to values obtained in other studies.

The values of the critical loads (L,) deduced from the scratch tests are shown in Table 3. It appears clearly that TiCN presents the best adhesion to the substrate. In addition, in all cases L , values are lower for the softer substrate (35CrMo4). During the scratch test the soft substrate deforms plastically leading to high stresses in the matter pushed at the border of the scratch and consequently leads to film detachment (Figs 4 and 5) . Similar behavior was observed in the case of friction tests. Figure 6 shows an SEM photograph of the track wear after sliding against an alumina ball. Numerous cracks parallel to the sliding direction can be seen, which indicates that fracture (fatigue wear) occurs at the surface of the material. In addition the figure shows that film detachment takes place at the border of the wear track. On the contrary, the film deposited on the hard substrate (X85WMoCrV6542) did not suffer any measurable wear due to the improved load support afforded by the substrate. Table 4 shows the wear coefficients measured for TIN, TiCN and DLC films. A quite similar values of wear coefficient were obtained for TiCN and DLC films whereas in the case of TIN a higher wear coefficient was measured. This is due to two effects. The first is the fact that DLC and TiCN are harder than TiN (Table l), the second is related to the presence of carbon graphite at the surface of TiCN and DLC films which lubricates the contact and decreases the friction coefficient (Table 4). Consequently, the stresses developed during friction on and beneath the surface are reduced [ 171.

4. INVESTIGATION WITH SCANNING THERMOELASTIC MICROSCOPY

In the seventies, the photothermal investigation emerged as a new means to investigate the subsurface of opaque materials [ 181. Different methods have been

198 J. Takadourn and B. Cretin

n Fe

C

- 30 90 150 210 270

Sputtering time (s)

(a>

30 90 150 210 270 Sputtering time (s)

(b)

n

cb ? 3

1

36 108 180 252 324 Sputtering time (s)

(c) Figure 2. Depth profiles obtained by GDOS from: (a) TiN, (b) TiCN, (c) DLC.

Adhesion and tribological properties of some ceramic films

Y 20

10

100 200 300 Depth (nm)

Figure 3. Loading/unloading curve obtained by nanoindentation on TiN film.

Table 1. Hardness ( H ) and Young’s modulus (E) of the films deduced from nanoindentation measurements

TiN TiCN DLC

H (GPa) 21 * 2 31 + 2 45 & 2 E (GPa) 320 k 20 370 h 40 160 + 15

E(substrate) = 210 GPa.

Table 2. Internal stress values (MPa) measured for TiN, TiCN and DLC films using the curvature method

Present study Other studies

[-29001 [lo] [-I5001 [ l l ]

TiN -2300 [-500, -50001 [12] [-2100, -500001 [13] [-200, -lOOOO] [14]

TiCN -3600 -50000 [15] DLC -2200 -1500 [16]

199

developed in order to extract the thermomechanical properties of materials [ 19-22]. Our setup called SThEM (Scanning Thermo-Elastic Microscope) is based on a completely optical system for excitation and detection which gives a complete set of images by scanning the sample: optical, thermal and thermoelastic images with a micrometric resolution. The temperature of the surface is extracted from

200 J. Takadoum and B. Cretin

Table 3. Values of L , (N) deduced from scratch test experiments

Substrate Film

TiN TiCN DLC

3SCrMo4 1 2 2 ~ 2 1 5 ~ 3 6 & 2 X85WMoCrV6542 25 * 3 39 i 5 13 * 2

Figure 4. Detachment of a hard film deposited on a soft substrate produced by the scratch test.

Figure 5. A scratch test track on TiN coating deposited on 35CrMo4 steel substrate. Film detachment is clearly seen in the border of track.

the reflected intensity of the probe beam and the thermomechanical parameters are indirectly measured with a high sensitivity interferometer [23, 241,

The principle of the SThEM combining optical, photothermal and thermoelastic microscopies is shown in Fig. 7. The excitation laser beam (Ar+) is intensity

Adhesion and tribological properties of some cerainicfilms 20 1

Figure 6. The border of wear track on TiN coating deposited on 35CrMo4 steel after sliding against an alumina ball.

Table 4. Wear and friction coefficients for TiN, TiCN and DLC films deposited on two different substrates

~~ ~ ~

Wear coefficient (lop6 mm3 / N m) Friction coefficient

Substrate Substrate Substrate Substrate X85WMoCrV6542 35CrMo4 X85WMoCrV6542 35CrMo4

TIN No measurable 8.00 0.25 TiCN No measurable 5.25 0.15 DLC No measurable 5.00 0.10

0.15 0.15 0.10

modulated with an acousto-optic modulator (A.O.M.) driven by the low frequency generator (L.F. gene.). The two beams are focused by a specific microscope objective. The second laser beam (coming from HeNe laser of the laser probe) is used to detect both optical signal, photoreflectance (dynamic component of the reflected optical intensity) and thermally generated displacement normal to the surface (with a high sensitivity interferometer). Physically, the focused excitation beam locally heats the sample at a preset frequency. For most samples, the dynamic temperature resulting from the absorption of the modulated beam induces reflectance variations. The magnitude and phase of the photoreflectance signal are extracted with a double phase lock-in amplifier (lock-in amplifier 2). A part of the probe beam is used to detect the thermoelastic displacements related to thermal

202 J. Takadoum and B. Cretin

Figure 7. Setup of SThEM.

expansion with a heterodyne laser interferometer (laser probe). Demodulation is obtained with specific electronics. The magnitude and phase of the surface vibration are given with a second lock-in amplifier (lock-in amplifier 1) and stored in a microcomputer that also drives the scanning units. With this multi-acquisition microscope, the typical duration of an experiment in order to obtain a set of five low noise images is about 15 minutes. The resolution of the SThEM is given by the size at the photothermal source (radius of the optical beam: 5 p m here).

4.1. Application to the study of thinjlms

The first example concerns the observation of subsurface thin layers. In order to demonstrate the capacity for subsurface investigation we successively vapour deposited a 200 nm thick Si02 and 100 nm thick aluminium layers onto a polycrystalline nickel substrate (Fig. 8a). The bright strip on the right part of the image (Fig. 8b) reveals the presence of the subsurface Si02 layer which is optically invisible. This image has been obtained at 220 kHz modulation frequency of the excitation beam. The image contrast corresponds to about 25" phase shift.

As the SThEM makes it possible to observe the subsurface we decided to use it for the detection of thin films delamination. We used a 1 p m thick DLC film deposited on a steel substrate. Several lines of Vickers indentations were performed under an applied load of 4.5 N. A different spacing (25 to 140 pm) between indentations has been taken for each line. The SEM and thermoelastic images of the indentations spaced 25 p m are shown in Fig. 9. Due to the film delamination, an optically invisible bright area between the indentations (Fig. 9a) was observed by the SThEM at 100 kHz operating frequency (Fig. 9b). It is an indication of the excessive heating resulting from the film delamination. The latter is due to the tensile residual stresses which develop around each indentation. The bright area (film delamination) could not be detected both in the case of a single indentation or when the spacing between indentations was higher than 40 pm. In the latter case

Adhesion and tribological properties of some ceramic films 203

Figure 8. (a) Structure of the multilayered sample. (b) Thermoelastic phase image showing the presence of the subsurface layer. Image obtained at 220 kHz excitation frequency. Grey scale 0 to 255 equivalent to -180 to 180'. The white bar gives the spatial scale.

the elasto-plastically deformed zones around each indentation are not close enough to each other to interact leading to film detachment [ 251.

The minimum distance between indentations which may lead to film delamination observed by the SThEM may be used as a new adhesion parameter.

The thermoelastic microscope has also been successfully used for other metallur- gical applications [26]. In the case of thin films, delamination problems are clearly revealed with SThEM. In addition, with a frequency scanning and measurements of the magnitude and phase, the thickness or thermal properties of the observed layer can also been estimated [27-291. Thus, SThEM provides both qualitative and quan- titative data on thin films. This technique is promising for the investigation of the coatings used to improve the surface properties in many recent technologies.

204 J. Takadoum and B. Cretin

Figure 9. (a) SEM image of the indented DLC film deposited onto a steel substrate, (b) thermoelastic phase image showing the delamination between the Vickers indentations (bright area shown by arrows). Image obtained at 100 kHz excitation frequency.

5. CONCLUSION

Mechanical properties, internal stress, adhesion and tribological properties of TIN, TiCN and DLC coatings deposited on steel substrates were studied. An additional study using Scanning Thermoelastic Microscopy (SThEM) for subsurface charac- terization has also been presented. The following conclusions can be drawn from the data presented here.

Adhesion and tribological properties of some ceramic films 205

(1) Due to substrate deformation, both wear resistance and adhesion of the films (L& as measured by the scratch test, decreased when the substrate hardness decreased.

(2) TiCN and DLC films showed better mechanical and tribological properties than TiN .

(3) High compressive residual stresses were determined in all the films with the curvature method.

(4) The SThEM has been successfully used for thin film delamination imaging.

REFERENCES

1. S. J. Bull, Diamond. Relat. Mater. 4, 827 (1995). 2. K. Holmberg, H. Ronkainen and A. Matthews, in: New Directions in Tribology. I. M. Hutchings

3. D. Klaffke, R. Wasche and H. Czichos, Wear 153, 149 (1992). 4. S. Zhang and W. Zhu, J. Muter: Process. Technol. 39, 165 (1993). 5. S. J. Bull and D. S. Rickerby, Surf: Coat. Technol. 36. 743 (1988). 6. R. R. Manory, Surf: Eng. 4, 309 (1988). 7. J. Deng, M. Braun and I. Gudowska, J. Vac. Sci. Technol. A12, 733 (1994). 8. J. Takadoum. H. Houmid Bennani and M. Allouard. Surf: Coat. Technol. 88, 232 (1996). 9. H. Rudigier, E. Bergmann and J. Vogel, Surf: Coat. Technol. 36, 675 (1988).

(Ed.), pp. 251-267. MEP Publication, London (1997).

10. P. M. Ramsey, H. W. Chandler and T. F. Page, Surf: Coat. Technol. 43/44, 223 (1990). 11. A. Kinbara and S. Baba, Thin Solid Films 163, 67 (1988). 12. P. J. Burnett and D. S. Rickerby. Thin Solid Films 154, 403 (1987). 13. P. J. Burnett and D. S. Rickerby. Thin Solid Films 157, 233 (1988). 14. S. J. Bull. D. S . Rickerby, A. Mattews, A. Leyland, A. R. Pace and J. Perry, Surf: Coar. Techlzol.

15. M. Einzenberg, K. Littau, S. Ghanayen, M. Liao, R. Mosely and A. K. Sinha, J. Vac. Sci. Techrzol.

16. C. Weissmentel, K. Bewilogua and K. Breuer, Thin Solid Films, 96, 31 (1982). 17. H. Houmid Bennani and J. Takadoum, Surf: Coat. Technol. 111, 80 (1999). 18. A. Rosencwaig and A. Gersho. J. Appl. Phys. 47, 64 (1976). 19. T. Baumann, E Dacol and R. L. Melcher, Appl. Phys. Let]. 43, 7 (1983). 20. R. Santos and L. C. Miranda, J. Appl. Phys. 52, 4194 (1981). 21. E Lepoutre, D. Fournier and C. Boccara, J. Appl. Phys. 57. 1009 (1985). 22. W. B. Jackson, N. M. Amer, A. C. Boccara and D. Fournier, Appl. Optics 20, 1333 (1981). 23. B. Cretin, 0. Franquet, E. Farnault, D. Hauden and J. L. L,esne. J. Phys. IV 4, C7-7 (1994). 24. B. Cretin, AZP Conference Proceedings 463, 147 (1998). 25. K. Yamanaka, Y. Enomoto and Y. Tsuya. IEEE Trans. Soriics Ultrason. 32, 313 (1985). 26. B. Cretin. J. Takadoum, A. Mahmoud and D. Hauden, Thin Solid Films 209, 127 (1992). 27. L. Bincheng, J. P. Roger, L. Pottier and D. Fournier, J. Appl. Phys. 86, 5314 (1999). 28. K. L. Muratikov, A. L. Glazov and H. G. Walther, High Pmp., High Press. 31, 69 (1999). 29. L. Bertolottim. G. L. Liakhou, A. Ferrari, V. G. Ralchenko, A. A. Smolin, E. Obraztsova,

K. G. Korotoushenko, S. M. Pimenov and V. I. Konov. J. Appl. P l y . 75; 7795 (1994).

36. 503 (1998).

A13, 590 (1995).

Adhesion Aspects of Thiii Films, Vol. 1. pp. 207-216 Ed K. L. Mitral 0 VSP 2001

Properties of oxide coatings deposited on a plastic substrate by a successive pulsed plasma anodisation process

B. M. HENRY ',*, A. G. ERLAT ', C. R. M. GROVENOR G. A. D. BRIGGS ' and R. P. HOWSON2

Department of Materials, UniversiQ of Oxford, Parks Road, Oxford, OX1 3PH UK *Department of Physics, Loughbovough Universiv, Loughborough, Leics. LE11 3TU, UK

Abstract-The aim of this study was to investigate the properties of A10, layers deposited by mag- netron sputtering on poly(ethy1ene terephthalate) substrates for gas barrier packaging applications. Aluminium. a few monolayers thick, was initially sputtered and then was converted to an oxide by the application of an oxygen plasma. The cycle of depositing a very thin metal layer followed by anodis- ing was repeated several times until the desired thickness was achieved. This process utilised a co- axial dual magnetron to sputter the metal and then to anodise it through a self-biased oxygen plasma, a technique known as successive pulsed plasma anodisation. In films produced by this method, the adhesion of the oxide to the polymer was greater than the polymer strength such that it could not be quantitatively assessed. The gas barrier properties of the films were found to be good. Atomic force microscopy showed that the barrier layers had grown densely with an average grain size of 40 nm. Also, it was shown that the coating surfaces were smooth, a feature that is believed to facilitate good water vapour barrier properties. A combination of the compactness, low surface roughness of the bar- rier coating together with only a few macro-scale defects are thought to be responsible for the good barrier properties shown by these films.

Keywords: Successive pulsed plasma anodisation (SPPA); sputtering; gas barrier films: characterisa- tion: aluminium oxide.

1. INTRODUCTION

Thin metal oxide layers deposited by reactive evaporation onto poly(ethy1ene terephthalate) (PET) substrates are increasingly being used as flexible gas barrier layers in food packaging. These films offer many advantages over traditional alu- minium foil in that they offer transparency, microwave compatibility and reduced health concerns. Recent advances in reactive magnetron sputtering potentially of- fer alternative processes that can fabricate films with improved properties at eco-

*To whom all correspondence should be addressed. Phone: + 44 (0)1865 273758; Fax: + 44 (0) 1865 273764; E-mail: bernard.henry @materials.oxford.ac.uk

208 B. M. H e n T et al.

nomical rates favourable for specialized commercial applications. In conventional magnetron sputtering, the cathode target sputters material onto the substrate. The growing layer is supplied with energy arising from the inert gas plasma that is leaked from the cathode. The reactive gas incident on the surface converts the metal layer to a compound thin film. This process can be highly unstable and is susceptible to target poisoning and uncontrollable arcing when DC power is used. Two recent de- velopments, namely cathode-controlled reactive processing and the use of medium frequency (MF) power supplies has led to a vast improvement in the overall control of the sputtering process.

By observing the optical spectral-line emission of the cathode one can control target-poisoning [ 11. The emission spectra reflect the sputtering rate of the target and hence the degree to which it is poisoned. It, therefore, provides a means to control the partial pressure of oxygen that is responsible for causing the poisoning.

Reactive magnetron sputtering can produce a highly insulating film in regions of the cathode where the oxidation rate exceeds that due to sputtering, leading to uncontrollable arcing when using dc power. It has been realised that using medium- frequency power can overcome this problem. In using MF, the potential reverses in a time that is short compared to the electron traversal time of the dark space. The sputtering is done by the dc bias that appears across this dark space and ion flow is compensated by electron flow when the MF potential reverses and the target no longer requires to be conducting.

Apart from these advancements in the stability of the sputtering process there has also been reports of improved sputter rates through the adoption of a separate time-delayed delivery of the constituent parts of the compound material to the substrate which is subsequently followed by a reaction between them [ 1-31. One such process based on this approach is called successive pulsed plasma anodisation (SPPA). This process uses just one unbalanced planar magnetron to both sputter the metal and gaseously anodise it with a self-biased plasma [2, 31. In the SPPA an MF potential is used and the admission of the reactive gas is pulsed, hence the name, in response to a signal generated by optical spectra that indicate the state of the cathode surface. Oxide layers produced by this method would be expected to be more compact with improved adhesion compared to those fabricated by an evaporation process because of the higher energy of the arriving species. The main advantage of the evaporation process is its faster deposition rate, hundreds of P \ l s compared to tens of A / s for sputtering. But, the barrier coatings tend to feature a significantly high proportion of defects and, therefore, an inferior gas barrier compared to sputtered films: oxygen transmission rates of < 4 cm3/m2/day and < 2 cm3/m2/day respectively, which can be crucial for some specialized applications. With this in mind, an investigation of the gas barrier, microstructural and mechanical properties of SPPA layers would be of importance in the advancement of metal oxide barrier films. In this study, the gas barrier properties of oxide coatings on PET deposited by the SPPA process have been measured. The adhesion of the oxide to the polymer was assessed and

Properties of oxide coatings deposited on a plastic substrate 209

the microstructure of the coatings has been characterised using transmission and scanning electron microscopies, atomic force microscopy and ellipsometry.

2. EXPERIMENTAL PROCEDURE

A laboratory roll-to-roll coater was used to deposit A10, onto untreated Melinex@ PET substrates of 50 p m thickness and 10 cm width (see Fig. 1). The barrier layers were fabricated by SPPA using a co-axial dual magnetron source (see Fig. 2) and a range of web speeds between 0.06 m/s and 0.2 m/s. Initially, a few monolayers of aluminium are sputtered which are subsequently converted to an oxide by exposure to a reactive oxygen plasma. Pulsing controlled the inlet of oxygen gas into the coating chamber. A switching time of around 1s altered the oxygen partial pressure between two states, namely the level necessary to poison the target and subsequently the level allowing metal sputtering. This time allowed for the deposited metal film to be completely reacted with the oxygen. The self-bias appearing on the isolated surface exposed to the dense plasma provides the potential driving the ion bombardment. The potential is quite small, 20 to 30 V, and hence only a few monolayers can be converted. In order to build up a significant oxide coating it is necessary to repeat the metal deposition and plasma-anodisation processes several times. The unbalanced planar magnetron is used to both sputter the aluminium and to anodise it. For the reactive plasma, oxygen gas is admitted as a pulse controlled by discharge conditions determined by optical emission monitoring of the cathode. During the metal deposition, a medium frequency supply of 500 W powered the magnetron and the argon partial pressure was 3 mTorr and the chamber pressure during operation was between 3.2 and 3.4 mTorr. The deposition and oxygen flow

Figure 1. The pilot roll-to-roll coater used to fabricate the films studied.

210 B. M . H e n q et al.

Figure 2. The co-axial dual magnetron source.

rates were 0.85 nm/s and 4.5 sccm, respectively, such that the film produced was optically transparent.

Approximate thickness and refractive index measurements on the deposited barrier coatings were obtained by examining the films deposited on a reference silicon substrate using an Auto ELR-I1 Ellipsometer.

The measurement of the adhesion strength of the deposited oxide layers on PET was performed by the standard peel test method widely used in the literature [6-81. The polymer was attached to a metallic support using a double-sided tape. To grip the oxide layers, a polyethylene adhesive, which has a strong adhesion to the aluminium oxide, was laminated on the coated side. By peeling off the coated film at an angle of 180" and a speed of 5 cm/min in a tensile tester (Instron), the force (newtons per metre) required was monitored.

The films including the bare substrate were tested for their gas permeation properties. The oxygen and water vapour transmission rates were measured using an OxTran (Mocon Co., Minneapolis, MN, USA, model 10/50A) and a Permatran (Mocon Co., Minneapolis, MN, USA, model W 3/31) instruments using a sample size of approximately 100 cm2. In this study, all the oxygen and water vapour transmission rate values were obtained at atmospheric pressure over a range of temperatures from 10 up to 45°C and are reported in units of cm3/m2/day and g/m2 /day, respectively.

The surface morphology of the barrier coating was examined using SEM and AFM. A JSM 6300 scanning electron microscope operated at around 1 kV accel- erating voltage was used to investigate the coating with and without a deposited

Properties of oxide coatings deposited on 11 plastic substrate 21 1

conducting film. In order to scan a larger area of these samples and improve the sta- tistical interpretation of defects, a special stage was constructed to allow 40 x 40 mm areas to be examined.

For AFM studies a Model CP, Park Scientific Instruments (Sunnyvale, CA, USA), fitted with an ultralever@ of a triangular shape, nominal spring constant k, = 2.1 N/m and tip radius of curvature of R = 10 nm, was used. AFM images were obtained in both contact and non-contact modes using a scan speed of 1 line per second.

3. RESULTS AND DISCUSSION

3.1. Chemical analysis

Rutherford backscattering measurements on the deposited A10, layers led us to believe that their stoichiometry was close to that of A1203.

3.2. Adhesion study

It was found that for the films produced by the SPPA method (see Table 1) the adhesion of the oxide to the polymer was greater than the polymer strength. Standard peel tests on these films yielded fracture surfaces with exposed polymer material as established by X-ray photoelectron spectroscopy. For different film thicknesses the force at which separation occurred was similar, more than 100 N/m. Separation occurred within the PET and was estimated by SIMS depth profiling to be around 10 nm into the polymer substrate. It was observed that the separation position and load varied only a little with coating thickness. This meant that the adhesion of the oxide to the polymer could not be quantitatively assessed.

The failure mechanism of the films can be understood if we consider that the typical orientation process for PET involves stretching the length and the width of the film. The bi-axial process strongly aligns the polymer chains in the plane of the film resulting in a layered structure. Hence, the PET surface consists of highly oriented layers with only weak bonds into the bulk of the polymer. Consequently, the bi-axial films have high strength in the plane of the film but are much weaker in the thickness direction. It is thought that the cohesive failure observed was due to the delamination of the underlying polymer substrate.

3.3. Gas barrier properties

The measured transmission rates of oxygen (OTR) and water vapour (WVTR) through bare PET and the nanocomposite films at 30°C are listed in Table 1. The films have good gas barrier properties and show, for oxygen, more than an order of magnitude improvement in resisting gas transmission compared to un-coated PET. The oxygen and water vapour transmission values are competitive with those measured on electron beam evaporated A10, (30 nm) layers, which are typically

212

Table 1. The measured oxygen and water vapour transmission rates of films with a range of A10, layer thickness (nm)

B. M. H e n v et al.

Sample Barrier layer OTR WVTR structure thickness (nm) (cm3 / m2 /day) (g/ m2 /day)

(30°C, 0% RH) (30°C, 100% RH)

PET - A10, /PET 23 A10, /PET 26 A10, /PET 30 A10, /PET 34

38.70 3.89 3.48 1.69 1.94

3.86 1.18 0.75 0.42 0.57

around 4 cm3/m2/day and greater than 1 g/m2/day, respectively. Table 1 shows that the thicker A10, layer films exhibit better gas barrier properties.

The transmission rates of oxygen and water vapour through bare PET and the 34 nm thick A10, composite film were investigated over a temperature range from 10 to 45"C, see Figs 3 and 4. The data indicated that activated rate processes were responsible for gas permeation both for PET and the composite films. Using an Arrhenius-type equation, the activation energy for permeation was determined. It was found that the calculated energy did not vary with A10, coating thickness. Activation energies for the 34 nm oxide coated PET film, which is representative of the coated samples, together with the bare PET substrate are shown in Figs 3 and 4. It is seen that within experimental errors ( f 2 kJ/mol) the activation energies for oxygen permeation through the composite films, 30 kJ/mol, is similar to that obtained for the bare polymer substrate, 28 kJ/mol. This would imply that the oxide layers were defect dominated, and did not provide a complete obstacle to oxygen transport. For water vapour transmission the situation is not so clear as it is known that the permeation mechanism is significantly different from that of oxygen. However, it is noticeable that the activation energy for water vapour transmission is apparently slightly increased in the coated PET 50 kJ/mol compared to 43 kJ/mol for the uncoated PET.

3.4. Microstructural studies

Topographical examination of the barrier coatings on PET using scanning electron microscopy (SEM) established the presence of defects as shown in Fig. 5. These features are pinholes and are thought to extend into the PET substrate. These defects are typically 2-3 p m in size with a density of 200 defects/mm2, which is around an order of magnitude less than that found in evaporated films. The observed defect densities varied only slightly with barrier coating thickness and are thought to result from either contamination particles or electrostatic discharging. It is our belief that the influence of these defects, they account for an area fraction of around 0.1%, on the permeation properties of the nanocomposite films would be similar in all the samples. Their presence would account for the activation energy for oxygen permeation being similar for both coated and uncoated films. However,

Properties of oxide coatings deposited on ci plastic substrate

2 5

2 -

1 5 -

6 1 -

=- 0 5 -

2 c O -

U N-

0 7

I- >

E .

m

4 .5

-

Q

5

AE=43kJ /mol 0

213

3 1 5 3 2 3 2 5 3 3 3 3 5 3 4 3 4 5 3 5 3 5 5

1000iTemperature ( I i K )

Figure 3. The OTR as a function of measuring temperature and activation energy for permeation for bare and A10, (34 nm) coated PET. The observed linear relationships indicate that activated rate processes are responsible for gas permeation in these materials.

AE=SOkJ/mol

3 1 3 1 5 3 2 3 25 3 3 3 35

1000iTemperature ( I i K )

A

Figure 4. The WVTR as a function of measuring temperature and activation energy for permeation for bare and A10, (34 nm) coated PET. The observed linear relationships indicate that activated rate processes are responsible for gas permeation in these materials.

214 B. M. H e n q et al.

Figure 5. An SEM image of the 34 nm A1O,y barrier layer on PET featuring surface defects.

nm

0 u 4 0 8 1.2 pm

Figure 6. An AFM image of a 34 nm A10, sputtered barrier layer on PET.

the difference observed in the gas transmission rates between the films cannot be solely attributed to the presence of these features.

The intrinsic surface morphologies of the barrier coatings on PET were also investigated using AFM. Images of the different A10, layers were all similar in appearance. Figure 6 shows a high magnification AFM image of the 34 nm A10, coating on PET which is smooth and homogeneous with small granular features approximately 40 nm in size. The sputtered coating has an average RMS roughness value of 1.2 nm, which is similar to that of the bare PET substrate. The previous work by Garcia-Ayuso and co-workers [4, 51 hypothesised that smooth barrier

Properties of oxide coatings deposited 017 17 plastic substrate 215

layers were responsible for low water permeation by promoting a surface diffusion mechanism for water transport.

An AFM study of the early stages of the barrier layer growth showed that it was flat and uniform. The spacing between the nucleating columns of A10, was extremely narrow (less than 5 - 10 nm) on a scale that could not be resolved by the AFM tip [9]. The closeness of these nucleation sites is responsible, in part, for the good barrier properties and adhesion exhibited by these sputtered films. It is thought that the use of the plasma on the substrate surface was partially responsible for the observed layer compaction, as oxygen ion bombardment of the PET would tend to promote the formation of nucleation sites.

The refractive index ( n ) of the gas barrier coatings was measured using ellip- sometry. The measured refractive index increased with layer thickness and ranged from 1.5 (23 nm) to 1.62 (34 nm). The density of a film is related to n by the Lorenz-Lorenz expression [ lo- 121, d = k ( n 2 - l ) / (n2 +2), where d is the density and k is a constant. Thus, increasing the layer thickness resulted in the formation of a more densely packed coating with improved barrier properties.

4. CONCLUSIONS

The gas barrier, microstructural and mechanical properties of A10, layers deposited by the SPPA process on poly(ethy1ene terephthalate) substrates have been studied. It was determined that the adhesion of the oxide to the polymer was greater than the polymer strength and thus it could not be quantitatively assessed. Furthermore, the sputtered layers had good barriers to both oxygen and water vapour permeation. It was found that the oxide surfaces were intrinsically smooth and uniform. The layers contained a relatively low number of micrometer-size defects. The compactness of the layers increased with thickness and enhanced the gas barrier performance. It is thought that the combination of barrier layer compactness. low surface roughness and only a few macro-scale defects are responsible for the good gas barriers shown by the sputtered layers examined in this study.

Acknowledgements

The authors wish to thank Dr. F. Dinelli, Mrs K.-Y. Zhao and Mr. R. S. Kumar for their valuable contribution to this work.

REFERENCES

1. R. P. Howson, Pure Appl. Chem. 66: 13 1 1 (1994). 2. R. P. Howson, Nucl. Instrunz. Methods Phjs. Res. B 121. 65 (1997). 3. R. P. Howson, N. Danson and I. Safi, Thin Solid Filnzs 351, 32 (1999). 4. G. Garcia-Ayuso, R. Salvarezza. J. M. Martinez-Duart, 0. Sanchez and L. Vazquez, Ad$: Mater.

5 . G. Garcia-Ayuso, L. VBzquez and J. M. Martinez-Duart, Surface Coatings Techid . 80, 203 9, 654 (1997).

(1996).

216 B. M. Henry et al.

6. J. F. Silvain, J. J. Ehrhardt, A. Picco and S. P. Lutgen, in: Metallization of Polymers, E. Sacher, J.-J. Pireaux and S. Kowalczyk (Eds), ACS Symp. Ser. No. 440, p. 453. American Chemical Society, Washington DC (1990).

7. A. N. Gent, J. Adhesion 23, 115 (1987). 8. K. Bright, B. W. Malpass and D. E. Packhani, BK Polym. J . 3, 205 (1971). 9. B. M. Henry, F. Dinelli, K.-Y. Zhao, A. G. Erlat, C. R. M. Grovenor, G. A. D. Briggs, R. S. Kumar

and R. P. Howson, Proceedings of the Society of Vacuum Coaters, 42nd Annual Tech. Con$ Proc. 403 (1999).

10. J. H. Lee, D. S. Kim and Y. H. Lee, J. Electrochem. Soc. 143, 1443 (1996). 11. M. Born and E. Wolf, Principles of Optics, 3rd edn, p. 87. Pergamon Press, Oxford (1965). 12. B.-C. Wang, A. G. Erlat, R. J. Spontak, Y. G. Tropsha, E. A. Vogler, M. Dalvie and K. D. Mar,

ACS Polymer Preprints 38, 1018 (1997).

Adhesion Aspects of Thin Films, Vol. I , pp. 217-237 Ed. K. L. Mittal 0 VSP 2001

Practical adhesion of organic coatings to metals: The role of the interphase and its residual stresses

J. BOUCHET', A. A. ROCHE2,*, E. JACQUELIN' and G. W. SCHERER3

Universite' Claude Bernard, Lyon I, Laboratoire Me'canique Mate'riaux, IUT A Gknie Civil, 43 Boulevard du 11 novernbre 1918, F-69622 Villeurbanne Cedex, France Institut National des Sciences Applique'es de Lyon, Laboratoire des Mate'riaux Macrornole'culaires (CNRS, UMR 56271, 20 Avenue Albert Einstein, F-69621 Villeurbanne Cedex, France Princeton University, E-319 Engineering Quadrangle, Princeton, NJ 08.544, USA

Abstract-Internal stress and Young's modulus of different thickness organic layers made of DGEBA epoxy prepolymer and IPDA hardener were determined. Coatings were deposited on aluminum alloy (5754) after chemical etching or anodizing. Using the same formulation and the same curing conditions, coating, interphase and bulk properties (reaction extent and interphase thickness) were determined by using FTNIR spectroscopy. Young's modulus and curvature of the coated samples were determined by a three-point flexure test. For thick coatings (>200-250 pm), mechanical and chemical (amine and epoxy conversion) properties are found to be similar to those of bulk. For thin films, different gradients in both Young's modulus and chemical properties were observed depending on the surface treatment. Interphase thicknesses of 200 p m and 250 p m were obtained, respectively, for anodizing and chemical etching. Considering both the real interphase created between the organic coating having the bulk properties and the substrate and the gradient of mechanical properties observed experimentally, a three-layer model was developed to evaluate the residual stress profile generated in such three-layered material. This model was based on the identification of adhesional strains (Le. strains which can be of either chemical or thermal origin) using data for radius of curvature from layers thicker than interphase thickness of our systems. The maxima in residual stress intensities are obtained at the interphase/substrate interface for both surface treatments. To calculate the interphase residual stress intensities, only a model considering constant interphase properties (e.g. Young's modulus) can be used. Based on such model, when adhesional failure was observed, the practical adhesion increased when internal stresses at the interface between the metal and the formed interphase decreased.

Keywords: Residual stresses; Young's modulus; practical adhesion; thin and thick epoxy-diamine coatings; epoxy-metal interphase.

~

*To whom correspondence should be addressed. Phone: (33) 4 72 43 82 78; Fax: (33) 4 72 43 85 27; E-mail: Alain-Andre.Roche@insa-1yon.fr

218 J. Bouchet et al.

1. INTRODUCTION

It is well known that when epoxy resins are applied onto metallic substrates and cured, internal stresses are developed within the organic layers. These stresses reduce the practical adhesion (adherence) and may induce cracks in the coated materials or debonding of the coating [ 1, 21. Young’s modulus and residual stresses are the most important mechanical properties of coatings and their knowledge is of prime importance to understand and simulate their mechanical behavior. It is recognized that a third layer, called the interphase or interfacial region, which has different chemical, physical and mechanical properties from that of the bulk coating, is created between the coating and the substrate [3]. The polymer/ substrate interphase is a region where several reactions take place that induce gradients of residual stresses and rearrangement of the organic network [3]. In the following, we call “bulk coating” that part of the coating having the bulk properties. To determine residual stresses within the entire coating, many studies have been done [4, 51 in which authors have extensively used and improved the one-dimensional approach of Stoney [6] using beam theory. For strips undergoing spherical bending in two perpendicular directions (plates), Timoshenko [7] and Hoffman [8] introduced the bi-axial modulus. Similar study of residual stresses using the analysis of the strength of materials approach has been carried out by Roll [9]. For thick coatings (the coating thickness being approximately equal to the substrate thickness), Timoshenko [7] proposed the well-known solution of the bi-metallic strip undergoing a thermal expansion. Considering the limitation of the bi-metallic strip model developed by Timoshenko, Inoue and Kobatake [ 10, 1 11 analyzed, in terms of the plane stress, the residual stresses in composite beams (coatings applied to solid surfaces). In recent publications, Benabdi [12, 131 has studied mechanical properties of coating/ substrate systems as bi-layer materials having a perfect interface (with no thickness) to evaluate residual stresses at the coating/ metal interface by using one and two dimensional approaches based on plate and beam theories. The residual stresses within the coating layer (o**) were expressed as a function of geometrical and mechanical parameters of the two layers (coating and metal), leading to:

with a = E,/E, and ,!l = h,/h,, where: h, is the thickness of the substrate, h, is the thickness of the entire coating, E, is the longitudinal Young’s modulus of the entire coating, E , is the longitudinal Young’s modulus of the substrate, R1 is the radius of curvature of the coated system.

However, the thickness of the coating/ substrate interphase has to be considered. Therefore, Benabdi [ 131 considered a three-layer model for determining residual stresses within the interphase and the remaining part of the coating having the

Practical adhesion of organic coatings to metals 219

bi-layer three-layer A three-layer A

Figure 1. Schematic drawing of bi and three-layer models

bulk properties. This model requires a knowledge of differences in elongation at interfaces between the metal and the interphase and between the interphase and the bulk coating. The differences in elongation are functions of the radii of curvature. Two different tri-layer systems A and A’ of different thicknesses h and h’ need to be considered. For the A and A’ coated systems their respective thicknesses are h and h’ with h = h, + h, + hbc and h’ = h‘, + hi + hbc (h , and hi are the substrate thickness; h, and h: the interphase thickness and hbc and hb, the thickness of the remaining part of the coating having the bulk properties). Both bi and tri-layer models are represented in Fig. 1. When the nature and the surface treatment of the substrate, the polymer formulation and its curing cycle are kept constant, it can be assumed that the substrate and interphase properties remain constant as long as the entire coating thickness is thicker than the interphase. Then, hs = hk, Es = ES, h, = hi, El = E: but hbc # hbc. Equations obtained for residual stresses at the interphase/ metal interface (a lnt) and at the bulk coating/ interphase interface (a are:

a ln t - El - 6RlR; (dibcd,’, - d1kcdis)

X (2 ) A;,bCR1 (Eshsdis + Ebchbcdibc) - AsibcR; (Eshsd1’, + Ebchbcdibc)

Eshs + Eihi + Ebchbc

h: h,hj hi2 I S 3 2 3 +-I

K.’ = - + -

220 J. Bouchet et al.

, h12 hjhb, hb2, K . =-+ - + -3 3 2 3 ibc

h2 h,hbc hfe 3 2 3

K;, = 2 + - + - + hlh’:

d,’, = EbchL,[E,h:(h: f Abc) + Esh’,(h’ + hi )] ,

where: h, = hi is the thickness of the substrate, h, = hi is the thickness of the interphase, h b c is the thickness of the bulk coating layer, hb, is the thickness of the bulk coating layer, h is the total thickness of the three-layer material (h = hbc + h, + hs) , h’ is the total thickness of the three-layer material (h’ = h& + hi + hi), E, = E: is Young’s modulus of the interphase, E, = E: is Young’s modulus of the substrate, E b c =

However, Benabdi’s three-layer model assumes that the interphase mechanical properties (e.g. E,) remain constant, which is not observed experimentally [ 141.

To determine Young’s modulus of the coating without residual stresses within a bi-layer system, Roche and colleagues [15, 161 used a three-point flexure test with a beam theory approach. However, when an interphase layer is created between the coating and the substrate, the bi-layer model is no longer valid. A model that takes into account both the mechanical and geometrical properties of the three layers (coating, interphase and substrate) has to be used. For a three-layered system, Benabdi proposed a model satisfying the interphase formation [8]:

is Young’s modulus of the bulk coating layer.

E:(1,S1> + E,[~s(EsSS + E b c s b d -t S,(-x + ESSSHIS + EbcSbcHlhc ) ]

+ E , s s Ebc s h c Hibc - x (Es s s + E b c Sbc) = 0 (4)

with:

X = (EZ)eq - EsZs - EbcZbc and (EZ)eq = PLj/(48S)j,

His = (hs/2fhi/2)’, Hibc = (hi/2$-hbc/2)*5 Hsbc = (hs/2+hi+hbc/2I21

where: I , is the inertia moment of the substrate ( I , = b,h:/12), Zi is the inertia moment of the interphase (Zi = bih3/12),

b, is the width of the substrate, bi is the width of the interphase, bbc is the width of the bulk coating,

is the inertia moment of the coating (Ib, = bbchiC/12),

Practical adhesion of organic coatings to metals 22 1

S, is the cross section of the substrate (S, = b,h,), SI is the cross section of the interphase (S, = b,h,), S b c is the cross section of the coating (Sbc = bbchbc) ,

L, is the span length, P is the load, 6 is the deflection at mid-span.

The experimental determination of ( E & and a knowledge of Young's moduli of the substrate (E , ) and bulk coating ( E b c ) and geometrical parameters (Is, I,, Ibc , Ss, S,, Sbc, b,, b,, bbc , h,, h, and h b c ) of the three-layer system allow us to find roots of equation (4). Only the positive root has a physical meaning and, therefore, is considered and associated with the corresponding value of Young's modulus ( E , ) of the interphase layer. Benabdi [ 131 showed that E, was not significantly affected by residual stresses.

This work focuses on the determination of residual stresses within bulk coating/ interphase/ metal system by considering a gradient of mechanical properties and to point out the critical role of the interphase. Young's moduli of the interphase and the bulk coating as a function of substrate surface treatments (used to improve practical adhesion) are determined.

2. EXPERIMENTAL

2.1. Materials

The metallic substrate used in this study was a 0.5 mm thick commercial rolled aluminium alloy (5754 from Pechiney) prepared by die-cutting to provide identical sized strips (50 >: 10 mm2). Prior to deposit the DGEBA-IPDA mixture, two different surface treatments were used: (i) chromic-sulfuric etching: the metal samples were ultrasonically immersed in

acetone for 10 min, and wiped dry, then the samples were submerged in a solution of 250 g/l sulfuric acid, 50 g/ l chromium acid, 87 g/l aluminium sulphate octadecahydrate, at 60°C for 20 min, rinsed in running tap water for 1 min, kept in distilled water for 5 min, and wiped dry [17].

(ii) anodizing: the metal samples were ultrasonically degreased in acetone for 10 min, and wiped dry, then the samples were anodized at 10 V for 20 min in a 1 M phosphoric acid solution at 20"C, running tap water for 1 min, kept in deionized water for 5 min and wiped dry.

Immediately after surface treatments, the specimens were stored for 2 hours in an air-conditioned room (20 f 2°C and 50 f 5% RH). The epoxy prepolymer used was almost pure diglycidyl ether of bisphenol A (DGEBA, y1 = 0.03; DER 332 from Dow Chemical). The isophorone-diamine curing agent was (3-aminomethyl-3,5,5- trimethylcyclohexylamine or IPDA from Fluka). All chemicals were used without further purification. The stoichiometric ratio (Y) (aminohydroged epoxy) used was equal to 1. This ratio was calculated using a functionality equal to 4 for the diamine

222 J. Bouchet et al.

and 2 for the epoxy monomer. A homogeneous mixture of DGEBA and IPDA was achieved by stirring in vacuum (z 1 Pa) at room temperature (Rotavapor RE 211 from BUCHI, Switzerland) for 30 minutes to avoid any air bubble formation. The mixture was applied to treated metallic sheets to obtain the desired coating thickness using an automatic film applicator (from Sheen). The coating layer and the substrate had the same width. The adhesive curing cycle [18] was adapted to reach the maximum glass transition temperature ( Tg") (3 hours at room temperature; 20 -+ 60°C (2"C/min); 2 hours at 60°C; 60 -+ 140°C (2"C/min); 1 hour at 140°C; 140 -+ 190°C (2"C/min); 6 hours at 190°C; cooling (8 hours) in the oven down to 20" C) .

2.2. Characterization

2.2.1. Near-infrared spectroscopy. An infrared spectrometer (FTIR Magma-IR 550 from Nicolet) was used with Omnic FTIR software. The near-infrared spectra were recorded in the 4000-7000 cm-' range. Transmission mode was used for both bulk and organic coatings after detachment from the metallic substrate. For each spectrum, 64 scans were collected at 4 cm-' resolution. The conversion rates of epoxy and amine groups were calculated, respectively, by using the 4530 cm-' epoxy combination band and the 6500 cm-' amine band [19]. The 4623 cm-' aromatic C-H ring stretch combination band was considered as reference. Thus, the amine (X,) and the epoxy (X,) conversion rates were determined using the ratio of the respective band areas (A) by:

( 5 ) (A4530/A4623)r and X, = 1 -

(A6500/A4623)1=0 (A4530/ A46231 r=O '

(A6500/A4623) I x , = 1 -

To evaluate the X, and X, variation versus the coating thickness we calculated the (X,),, and (X,),, values at thickness ( i ) using:

(6) hi(Xa,e)i - hi-l(Xa,e)i-l

hi - hj-1 (Xa,e)eq =

2.2.2. Young's modulus. This work was performed with a three-point flexure machine (FLEX3, TechmCtal, Maizikres-les-Metz, France) [ 16-20]. The crosshead displacement speed was 0.1 mm/min. A 50 N full scale load cell with a sensitivity of 5.5 mN and a stiffness of 2.2 x lo5 N/m was fitted under the crosshead. The load ( P ) versus displacement (6) curves ( P / 6 curves) were recorded by a micro- computer and displayed on the PC screen in real time. The slope of the P / 6 curves, within the elastic region, was then computed using a linear regression program. Experimental curve slopes were corrected to take into account the load cell stiffness as described previously [20]. For various spans (Lj) , the apparent modulus (Eapp) can be calculated from the slope ( P I S ) of the load-displacement curve. According to Rippling et al. [21], the extrapolated Young's modulus of the substrate (E , ) and the coated system (E,) can be obtained from curves of (Eapp) as a function of (h/Lj) . It has been noticed that the extrapolated Young's modulus is independent of the ratio (b/ h ) and thus Poisson's effect can be neglected [20].

Practical adhesion of organic coatings to metals 223

To evaluate Young’s modulus variation versus the coating thickness, we calculated the (E)eq value at thickness (i) using:

2.2.3. Radius of curvature. Experiments were carried out with a flexure machine (FLEX3, Techmetal, Maizikres-les-Metz, France) equipped with a 50 N full-scale load cell. Samples were placed on a plane and rigid material. The curvature developed after the deposition onto the substrate and the curing of the coating layer were measured as reported elsewhere [12]. Assuming that the radius of curvature is large compared with the length and the thickness of the multi-layer beam ( R >> L >> h) , it can be considered that the length of the neutral axis is equal to its span. For such a curved multi-layer beam of neutral axis length ( L in mm) and maximal deflection (a,,, in mm) at the mid-span ( L / 2 ) , the radius of curvature (R1 in mm) is given by:

2.2.4. Practical adhesion measurement. This test was performed with a flexure machine (FLEX3, Techmetal, Maizihes-les-Metz, France) fitted with a 1000 N full-scale load cell, according to the AFNOR T 30 010 standard at a crosshead displacement speed of 0.5 mm/min. A solvent-free epoxy adhesive (3525 from 3M) was applied as a stiffener on the coated substrates. The curing condition of the stiffener was 24 h at room temperature. The ultimate load (Fmax) parameter was used to evaluate the practical adhesion of coatings to metallic substrates.

3. RESIDUAL STRESSES: THEORETICAL DEVELOPMENT

We consider here the interphase as an interlayer of the coating/substrate system in which an initial field of residual stress, denoted (a), is developed. Moreover, we consider a gradient of mechanical properties within the interphase as observed experimentally [ 141. The theoretical development is carried out by assuming that materials forming the coated system (with interphase layer) are elastic and isotropic.

3. I . Adhesional strain

Under the curing cycle, materials forming the tri-layer system determine the strain, which can be of either chemical or thermal origin. We called them adhesional strains ( ~ “ ( y ) ) . Chemical stresses are produced as a result of the mismatch between the active sites of the metallic substrate and the organic network and/or the formation of polymer network. Thermal stresses are developed during the cooling process and are the result of thermal expansion mismatch between the metallic substrate and the polymer or of cure-induced shrinkage within the organic layer. The objective

224 J. Bouchet et al.

Figure 2.

L pi Y

-L

Three-layer system.

BULK COATING

is to identify these adhesional strains by using the radius of curvature and to find the stress in each material. We assume that the total strain (Etot)at any point of the system represented in Fig. 2, is:

(9) &tot - - &rnech + &ad

with crnech is the mechanical strain. Considering the geometry and the size of the three-layer systems studied, the

beam theory can be used. For the one-dimensional approach, without lateral (width- wise) stresses, the following assumptions are made:

transverse sections of the beam are planar before, during, and after bending (Navier-Bernouilli’s hypothesis). Therefore, the effect of transverse shear (txy = 0) is neglected, the radius of curvature is large compared with transverse dimensions [width (b) and thickness (h ) of the three-layer system], leading to R1 >> bl, hl, longitudinal elements of the beam are subjected only to simple tension or compression inducing stresses in the x direction, Young’s modulii of the coating having bulk properties, the interphase and the substrate have the same value in both tension and compression (flexural modulus).

Based on these assumptions, final uni-axial residual stresses (a), in the x-direction of the three-layer system (bulk coating/ interphase/ substrate) are given by:

where: yo is the position where total strains are equal to zero = 0), y is the coordinate distance of any longitudinal fiber, R1 is the radius of curvature and E is Young’s modulus.

To determine the distribution of residual stresses in the tri-layer system from equation (10) requires a knowledge of the radius of curvature (R1) and the position of the zero deformation (yo) . Therefore, we consider two equilibrium conditions for the force (N) and the moment ( M ) for any cross section (area S ) of the coating/ interphase/ substrate system:

N = ] a d S = O (1 1) S

Practical adhesion of organic coatings to metals 225

M = I y a d S = O . (12) S

The geometrical and mechanical parameters of each layer (I) of the coating/ interphase/ substrate system are known: h, (metal thickness), hi (interphase thickness), hbc (bulk coating thickness), b, (metal width), bi (interphase width), bbc (bulk coating width), E , (Young's modulus of metal), Ei (Young's modulus of interphase), Ebc (Young's modulus of bulk coating).

To simplify writing we have adopted the following notation of a function f ( y ) is described as:

and

4 Y O f = f '

PO

For the force ( N ) , equations (10) and (1 l), yield:

the n order moment

(13)

For the bending moment ( M ) , the development and the rearrangement of equa- tion (12), yields:

Then, equations (15) and (16) yield:

226 J. Bouchet et al.

and, R1 and yo are obtained by using equation (18):

R1 =

Yo =

It is now possible to determine the adhesional strain evolution (cad) by using the radius of curvature as a function of the coating thickness curves.

4. RESULTS AND DISCUSSION

4.1. Interphase characterization

4.1.1. Bulk properties. The properties of aluminum substrates and DGEBA- IPDA bulk are listed in Table 1. Young’s modulus was determined using flexure tests and compared to the one obtained by tensile tests.

Practical adhesion of organic coatifigs to metals 227

1

0.98

P

25 P

25.

Q 0.96-

h CU 0.94

Table 1. Physical and mechanical properties of bulk materials

- x x ; 0 x 0 B - _=c Q---*-------- -i+-.+---8

0 , 0 ,';,a Fy ::

d :

, . , '_ ZY

- I

-G- Xa chemically etched - B - %a anodized

Xe chemically etched

I ,

d / 7' ; ; :

- ,; p $ 4

DGEBA/IPD Chemically etched Anodized A1-5754 A1-5754

Glass transition 163 temperature

Tensile Young's 3.2 70 modulus (GPa) Flexure Young's 3.3 71 modulus (GPa) Poisson's ratio 0.33 0.33

(TpOo) ("C) 70

71

0.33

".l

0 0.1 0.2 0.3 0.4 0.5 0.6 Coating thickness [mm]

Figure 3. Equivalent amine (X&, and epoxy (X&, extents as a function of the coating thickness for chemically etched and anodized aluminium substrates.

4.1.2. Chemical properties of coating. The equivalent amine (X,),, and epoxy (X,),, conversion rates are plotted versus the coating thickness in Fig. 3. (X,),, is independent of surface treatment and coating thickness, and is close to 1. This means that epoxy groups have completely reacted. For thick coatings (h , > 0.2- 0.25 mm), (X,),, reaches the value of the bulk. For (0.1 < h, < 0.2-0.25 mm), significant differences are observed for degreasing and anodizing processes. For the thinnest coating (h, < 0.05 mm), (X,),, is independent of the surface treatment and tends to have the same value (92%). In other words, some of amine groups (NH or NH2) have not been consumed during the amine-epoxy reaction.

Significant differences observed for chemical properties of thin coatings, com- pared to the bulk samples, mean that a polymer/substrate interphase is created. The interphase thickness in our systems is about 200 p m for anodizing and about

228 J. Bouchet et al.

250 p m for chemical etching. The obtained interphase thickness values corre- spond to the thickness needed to reach the bulk properties (i.e. (X,),, = 1). For 0.05 < h, < 0.200-0.250 mm, a gradient of properties within the interphase region is observed depending on surface treatment.

4.1.3. Mechanical properties of coatings. Young’s moduli of coatings calculated using equations (4) and (7) versus the coating thickness are shown in Fig. 4. For thin films (0 < h, < 0.2-0.25 mm), the modulus decreases as the coating thickness increases to reach the bulk value (h , =- 0.2-0.25 mm). Differences were observed between chemically etched and anodized samples. The highest value was obtained for the thinnest coating (0.04 mm) on anodized aluminium. It is six times greater than the bulk ( E ) e q . Similar differences in hardness of DGEBA-diamine (DICY) coatings on zinc were reported by Law et al. [22].

Bentadjine et al. [23] have explained the interphase formation for such systems. Because of the basic behavior of the curing agent (IPDA), the outer part of the oxide (and/or hydroxide) layer is dissolved when liquid monomer mixture is applied onto a metallic substrate leading to metallic ion diffusion within the liquid prepolymer coating. An organo-metallic complex is then formed between amine groups and metallic ions. When the concentration of the organo-metallic complex is higher than its solubility product, these complexes precipitate to form needle-shape crystals. However, the unprecipitated part of organo-metallic complex forms a new network with the epoxy prepolymer (DGEBA) during the curing cycle. Then after curing, a new network is associated to the initial one. This new bi-phase material contains crystals which act as short fibers randomly dispersed in a polymer matrix inducing an increase of the Young modulus ( 5 GPa) and a decrease of the elongation. It

20

I t , - 0 - anodized

0 0 0.1 0.2 0.3 0.4

Coating thickness [mm]

Figure 4. Equivalent Young’s modulus (E),q as a function of the coating thickness for chemically etched and anodized aluminium substrates.

Practical adhesion of organic coatirigs to metals 229

+ cryslals oriented close to the aluminium - Aluminium

Figure 5. Optical microscopy observation of the diamine-aluminium complex.

was also observed, close to the metallic surface, that organo-metallic crystals were aligned to surface, as represented in Fig. 5, leading to a large increase of the longitudinal Young’s modulus as determined by flexure test.

4.2. Residual stresses

4.2.1. Curvature measurement. To evaluate residual stresses, the radius of curvature for each coated system was determined. ‘These values are plotted versus the coating thickness in Fig. 6. R1 decreases with the coating thickness and remains constant for films thicker than 0.2 mm irrespective of the substrate surface treatment. To determine adhesional strain, a knowledge of Young’s modulus as a function of the coating thickness within the interphase is necessary. As represented in Figs 7a and 7b, the data obtained experimentally from Fig. 4 were approximated by dividing the interphase into n linear regions of thickness dhi ,

The adhesional strain within the interphase was considered to be linear within each region:

(19) ad ci ( y ) = a,y + b, with i := 1 . . . n

&pd(idh,) = ~ p : ~ ( i d h , ) with i = 1 . . . n - 1. (20)

Adhesional strains were considered as constant for both metal (E:) and coating having bulk properties (E;:). At the interphase/ metal interface, Young’s modulus of the aluminium substrate (Le. 70 GPa) was taken for both chemical etching and anodizing surface treatments.

Therefore, n + 1 parameters were obtained in the interphase and 2 parameters corresponding to the bulk coating and metal, lead to n + 3 parameters to be determined from Fig. 6. As the radius of curvature curve gives N equations with N 3 n + 3. unknowns were determined by a least squares method.

4.2.2. Adhesional strain determination by considering interphase and bulk coat- The adhesional strain determination by using 8-parameter approach is repre- ing.

230

80

70ct - cv 60-

v) 3

'0 0 E 40- v)

m 5 30 0

- 3 50-

2 0 -

10

0

J. Bouchet et ai.

METAL I INTERPHASE j BULKCOATIN1 I I

I I I I 1 I

I I I I I I , I , , I I I I I I I I I , I I I I I I I I I I I I , , ,

I I I I I , , I I I

Q

-

-

Y

0 0.5 0.75 1

- E E

rr' I ,.

- 0 - chemically etched - - anodized

0 0.1 0.2 0.3 0.4 0.5 0.6 Coating thickness [mm]

Figure 6. Curvature radius as a function of the coating thickness for chemically etched and anodized aluminium substrates.

Figure 7a. Linear approach of Young's modulus as a function of the coating thickness within the three-layer system for chemically etched aluminium substrate.

sented in Fig. 8. Only with eight or more parameters was the radius of curvature correctly described as a function of the coating thickness, especially in the inter- phase region. However, it was not possible to have stable adhesional strain by using the radius of curvature curve corresponding to coatings so thin that the interphase was not completely formed. Therefore, we have considered only the curvature curve for thicknesses exceeding the interphase thickness of our systems.

Practical adhesion of organic coatings to metals 23 1

>

0 0.5 0.7 1

Position within the three-layer model [mm]

Figure 7b. Linear approach of Young's modulus as a function of the coating thickness within the three-layer system for anodized aluminium substrate.

aooo

4-parameter model 6000

- E 4000 Y

2

2000

0 0.1 0.2 0.3 0.4 0.5 0.6 Coating thickness [mm]

Figure 8. Radius of curvature approximated by 4 and 8-parameter models for chemically etched substrate.

4.2.3. Adhesional strain determination by considering only bulk coating. Ad- hesional strain using 8-parameter approach and by considering only points in the bulk coating is represented in Figs 9 and 10, respectively, for chemical etching and anodizing surface treatments. In this case, adhesional strains obtained were stable at fl%. Therefore, it is now possible to calculate the stress within the three-layer system. To calculate it, we used the thicker sample (i.e. h , + hi + h, = 1 mm for both anodizing and chemical etching). The results are presented in Figs 11 and 12

232

7000

6000

5000 - E 4000-

3000-

Y

J. Bouchet et al.

-

-

0 -

0

0

1

8000-

7000

6000

- 8-parameter model -

-

1

" 0 0.1 0.2 0.3 0.4 0.5 0.6

Coating thickness [rnrn]

Figure 9. Radius of curvature used for a 8-parameter model for chemically etched substrate.

2000

1000

- 0 O 80

- --

E 4000

in term of adhesional strain and in Figs 13 and 14 in term of residual stresses for chemical etching and anodizing surface treatments, respectively. The adhesional strain results show that it can be approximated as varying linearly with the inter- phase thickness, so a 4-parameter model could be used without affecting the resid- ual stress profile results. The maximum stresses within the three-layer system were at the interphase/ metal interface for both anodizing and chemical etching surface treatments. In mechanical terms this means that the failure of such systems should take place at the interphase/ metal interface where residual stresses are maximum.

Practical adhesion of organic coatings to metals

1 - - E E 0.5 \

Y

S .- 2 0 - v)

m C 0 -0.5 v) a, c

- .-

2 -1

233

-

-

-

x 1 0 ' ~ METAL

1.5

4 .

3 - - E E 2 -

!!

\

Y

S .- Y v) 1 - m C 0 v) 0 - a, c

- .-

-1 2

-2

METAL ~ INTERPHAS~ BULK COATING I I I I , , , I I I , , , I I , , , , I I , I

I , , 1

I I I ,

-

- 0 0.5 0.7 1

INTERPHASE I BULK COATING 71

I I -L

0 0.5 0.75 1

Position within the three-layer system [mrn]

Figure 11. Adhesional strain as a function of the position within the three-layer system for chemical etchincr wrfwe trmtment

The stress intensities at the interphase/ substrate interface were higher for chemical etching than for anodizing.

234 J. Bouchet et al.

Figure 13. Profile of residual stress as a function of the position in the three-layer system for chemical etching surface treatment.

-80 1 I -1 00 1 , ,

0 0.5 0.7 1

Position within the three-layer system [mm]

Figure 14. Profile of residual stress as a function of the position in the three-layer system for anodizing surface treatment.

4.3. Practical adhesion

Since adhesional failure was always visually observed, so only residual stresses at the interphase/ substrate interface have to be considered. To correlate residual stresses to practical adhesion represented by the ultimate load (F,,,), the model developed by Benabdi [13] (i.e. a'"') was used. When the entire coating thickness is less than the interphase thickness, the thickness of the remaining part of the

Practical adhesion of organic coatings to metals

-

235

50 25 I

@ ' o 'b 15 2

LL -30

x

x * x

x x

m

x

- 2 0 10 . O 0

50

40

0

0 0 0 0 0

0 0

-

-

x

- -

0 0

x Y

- 5 0

40

- -30s

2 LL

x

T n 5 30- 'b I

20

0

-

0 0 0

10

0 0

om 0 - 10

8 m 0 0'"'

-

5 ' '10 0 0.05 0.1 0.15 0.2 0.25

Coating thickness [mm]

Figure 15. Ultimate load (Fmax) and interphase residual stress (aint) as a function of the coating thickness for chemically etched aluminium.

0

0

l o 0.05 0.1 0.15 0.2

O L 0

Coating thickness [mm]

Figure 16. Ultimate load (Fmax) and interphase residual stress (aint) as a function of the coating thickness for anodized aluminium.

coating having bulk properties is equal to zero. Then, a bi-layer system is assumed and Young's modulus of the entire coating is considered to be equal to 5 GPa. A good correlation was observed for both chemical etching (Fig. 15) and anodizing (Fig. 16) and indeed when aint increases, F,,, decreases and vice versa. However, Benabdi's model assumes that interphase mechanical properties remain constant. The proposed model can be used when the entire coating is thicker than the interphase. To consider residual stresses within the interphase region, without inducing an instability of the adhesional strain, a model admitting only constant

236 J. Bouchet et al.

mechanical properties (e.g. Young’s modulus) in the interphase has to be considered to approach the residual stress values that correspond to Benabdi’s work.

From Figs 15 and 16, it can be observed that practical adhesion (Fmax) is higher in the case of anodizing than of chemical etching. This can be explained by the difference between the interphase/ metal interface maximum residual stress intensities obtained for anodized and chemically etched substrates. Indeed the maximum residual stress intensities at the interphase/ metal interface are equal to 52 MPa and 83 MPa for anodizing and chemical etching, respectively. When the interphase/ metal interface residual stress value increases, practical adhesion decreases.

5. CONCLUSION

The residual stress and Young’s modulus of DGEBA-IPDA layers deposited onto 5754 aluminium following different surface treatments were determined. Using FTNIR spectroscopy, an interphase region between the substrate and the part of coating having bulk properties was observed. Interphase thicknesses of 200 p m and 250 p m were obtained for anodizing and chemical etching, respectively. The chemical and mechanical properties of the resulting interphase depend on its thickness and on the surface tieatment of the substrate. When the entire coating thickness exceeds the interphase thickness, mechanical properties of that part of the coating thicker than the interphase were similar to the bulk. To take into account the interphase in coated systems, a three-layer model (coating, interphase and substrate) was developed to evaluate the residual stress profiles generated in such materials. This model was based on the determination of the adhesional strains (i.e. strains of thermal or chemical origin) when the interphase was formed. Indeed, it is impossible to correctly numerically identify the interphase region until its complete formation. It was shown that the maximum residual stress intensities were present at the interphase/metal interface for both surface treatments. In the case of the chemically etched substrate, the interphase/ metal interface residual stress value is higher than for the anodized one. For coatings thinner than the interphase region, it was not possible to use the model developed by considering both the radius of curvature curve and the Young’s modulus of the coating curves. However, assuming a constant Young’s modulus within the interphase was possible leading to Benabdi’s model (dnt), and a good correlation was obtained between the practical adhesion and residual stresses. However, oint does not provide an evaluation of stress distribution within a sample. On the contrary, the model proposed here allows to calculate the stress profile within the sample and thus to foresee the system failure during the mechanical deformation to evaluate practical adhesion of such systems. Indeed, it seems possible to correlate the practical adhesion to the maximum residual stress values at the interphase/ metal interface. Since adhesional failure was always visually observed, only residual stresses at the interphase- metal interface have to be considered, and not the residual stresses within the entire coating.

Practical adhesion of organic coatings to metals 237

REFERENCES

1. H. Orsini and F. Schmit, J. Adhesion 43, 55 (1993). 2. M. D. Thouless and H. M. Jensen, J. Adhesion Sci. Techizol. 8, 579 (1994). 3. K. L. Mittal, in: Adhesion Measurement of Thin Films, Thick Films and Bulk Coatings,

K. L. Mittal (Ed.), STP 640, pp. 5- 17. ASTM, Philadelphia (1978). 4. R. J. Jaccodine and W. A. Schlegel, J. Appl. Phys. 37,2429 (1966). 5. Jwo-huei Jou and Li Hsu, J. Appl. Phys. 69, 1384 (1991). 6. G. G. Stoney, Proc. R. Soc. London, Ser: A 82, 172 (1909). 7. S.Timoshenko, J. Opt. Soc. Am. 11,233 (1925). 8. R. W. Hoffman, in: Physics ofThin Films, G. Hass (Ed.). Vol. 3, pp. 21 1-273. Academic Press,

9. K. Roll, J. Appl. Phys. 47, 3224 (1976). New York (1966).

10. Y. Inoui and Y. Kobatake, Kolloi’d-Z. 159, 18 (1958). 11. Y. Inoui and Y. Kobatake, Appl. Sci. Res. A7, 314 (1958). 12. M. Benabdi and A. A. Roche, J. Adhesion Sci. Technol. 11,281 (1997). 13. M. Benabdi, Ph.D. thesis, UniversitC de Lyonl (1998). 14. J. Bouchet, A. A. Roche and P. Hamelin, Thin So2id Films 355-356, 270 (1999). 15. A. A. Roche; J. Dumas and M. J. Romand, Proc. EURADH’92, held in Karlsruhe, Germany,

16. A. A. Roche, Ph.D. thesis, UniversitC de Lyonl (1983). 17. J. M. Cuntz, Engineer thesis, CNAM de Paris (1986). 18. F. F. De Nogarro, P. Guerrero, M. A. Corcuera and I. Mondragon, J. Appl. Polym. Sci. 56, 177

19. N. Poisson, G. Lachenal and H. Sautereau, Vibrational Spectroscopj 12, 237 (1996). 20. A. A. Roche, J. Dumas, J. E Quinson and M. Romand, in: Mechanics and Mechanisms

of Damage in Composites and Multi-materials, D. Baptiste (Ed.). Mechanical Engineering Publications, London (1991).

21. E. J. Ripling, J. S. Santner and P. B. Crosley, in: Adhesive Joints: Formation, Characteristics, and Testing, K. L. Mittal (Ed.), pp. 755-788. Plenum Press, New York (1984).

22. W. C. Law, D. I,. Chadwick and H. Taylor, Proc. EURADH’96, held in Cambridge, UK, p. 213 (1996).

23. S. Bentadjine, A. A. Roche and J. Bouchet, this volume, pp. 239-260.

p. 238 (1992).

(1995).

Adhesion Aspects of Thin Films, Vol. I , pp. 239-260 Ed. K. L. Mittal 0 VSP 2001

Epoxy- diamine adhesives on metals: The interphase formation and characterization

S. BENTADJINE ', A. A. ROCHE '.* and J. BOUCHET2 ' INSA de Lyon, Laboratoire des Mate'riaux Macromole'culaires (CNRS, UMR 56271, 20 Ave. Albert

Einstein, F-69621 Villeurbanne Cedex, France Universite' Claude Bernard Lyon I , IUT A Ge'nie Civil, Laboratoire Me'canique Mate'riaux, 43 Boulevard du 11 Novembre 1918, F-69622 Wlleurbanne Cedex, France

Abstract-Epoxy-diamine mixtures are extensively used as adhesives or paints in many industrial applications. When they are applied onto metallic substrates and cured, an interphase, having chemical, physical and mechanical properties quite different from those of polymer bulk, is created between the substrate and the polymer. Moreover, chemical reactions between diamine and metallic surfaces induce an increase in the practical adhesion. When epoxy-diamine mixtures are applied onto gold coated substrates, coating properties are the same as those of epoxy-diamine bulk. When epoxy- diamine mixtures are applied onto aluminum or titanium alloy surfaces, several material properties were found to be affected by the nature of amine (whether aromatic, aliphatic or cycloaliphatic), the stoichiometric ratio, the application conditions (duration and temperature), organic layer thickness and the metallic surface treatment. These properties include glass transition temperature, cure rates, interphase thickness, residual stresses within the interphase and its Young's modulus. Analyses (FTIR, FTNIR. DSC, DMTA, NMR, SEC, ICP and optical microscopy) suggest that diamine monomers chemically react and dissolve in the hydrated metallic oxide layer. Metallic ions then diffuse through the organic layer to form a coordination complex with diamine monomers (chelate or ligand). Since metal-diamine complexes are insoluble at room temperature in both diamine as well as DGEBA monomers, they induce phase separation during the cure cycle of the epoxy-diamine resin. Furthermore, the chemical bonding of diamine monomers with the metallic surfaces and the orientation of the diamine-metal complexes produce an oriented polymer within the interphase region leading to chemical, physical and mechanical properties different from those of epoxy- diamine bulk.

Keywords: Epoxy- diamine adhesives or paints; interphase formation; interphase characterization; interphase properties; diamine-metal complexes; diamine-metal chemical reactions.

'To whom correspondence should be addressed. Phone: (33) 4 72 43 82 78; Fax: (33) 4 72 43 85 27; E-mail: Alain-Andre.Roche@insa-lyon.fr

240 S. Bentadjine et al.

1. INTRODUCTION

Epoxy-diamine mixtures are extensively used as adhesives or paints in many industrial applications. When they are applied onto metallic substrates and cured, epoxy- amine liquid monomers react with the metallic oxide and/or hydroxide surface to form chemical bonds [ 1, 21 increasing practical adhesion (or adherence) between the epoxy polymer and the substrate [3, 41. Nevertheless, when epoxy resins are applied onto metallic substrates and cured, intrinsic and thermal residual stresses develop within the organic layers [5]. Intrinsic stresses are produced as a result of the mismatch between the active sites of the metallic substrate and the organic network and/or the formation of the polymer network. Thermal stresses are developed during the cooling process [6] and are the result of thermal expansion mismatch between the metallic substrate and the polymer or cure-induced shrinkage or buckling of the organic layer [7]. Whatever their source, these residual stresses reduce the practical adhesion and may induce cracks in the coating materials [8, 91 resulting in a drop of the overall performances of adhesives or paints. So far, to assume that such systems can be considered as bi-layer materials (i.e. that the epoxy/metal interface is perfect and that the chemical and physical properties of the epoxy layer are the same as that of the bulk) has been unsatisfactory and even unsuccessful in predicting the mechanical behavior and durability of such systems. To gain a greater understanding of epoxy/metal adhesion requires a full knowledge of the chemical and physical reactions which take place at the epoxy/metal interface or interphase [lo]. This interphase, which is created between the substrate and the polymer has been found to have chemical, physical and mechanical properties quite different from that of the bulk polymer network [ 111. Thus, the polymer/ substrate interphase is a complex region containing gradients of residual stresses, as well as allowing structural rearrangement, intermolecular and inter-atomic interactions and diffusion phenomena [lo]. When the adhesion of epoxy/metal systems failed, it was possible to not only correlate the residual stresses at the interphase/ metal interface to practical adhesion but also to correlate the adhesion and durability to the presence or not of some chemical species [12]. However, in such systems, chemical and physical mechanisms leading to the interphase formation are not well understood. The aim of this paper was to develop an understanding of this interphase formation and characterize it using model epoxy- amine systems.

2. EXPERIMENTAL

2.1. Materials

2.1.1. Substrates. The commercial metallic substrates used were 0.5 16 f 0.005 mm thick 5754 aluminum alloy from PCchiney and 0.600 f 0.005 mm thick Ti6A14V titanium alloy from ACrospatiale. Titanium and aluminum sheets were made into squares of dimensions 100 x 100 mm2. Before any polymer application, the aluminum and titanium substrate surfaces were treated as shown in Table 1. Af- ter surface treatment, all substrates were kept in an air-conditioned room at 22 f 2°C

Epoxy-diamine adhesives on metals 24 1

Table 1. Surface treatments of aluminum and titanium alloys

Alloy Treatment Description

Aluminum Degreasing

Chemical etching

Anodizing

Hydrolysis

Titanium Degreasing

Chemical etching

Anodizing

Hydrolysis

Ultrasonically degreased in acetone for 10 rnin and in ethyl acetate for 10 min, wiped dry Degreased and submerged in a solution of 250 g sulfuric acid, 50 g chromium acid, 44 g aluminum sulfate, distilled water to 1 1, 60°C for 20 min; rinsed in running tap water for 1 min, allowed to stand in deionized water for 5 min and wiped dry Degreased and submerged in a 1 M phosphoric acid solution, 20°C, potentiostatic anodization at 10 V for 20 min; rinsed in running tap water for 1 min, allowed to stand in deionized water for 5 min and wiped dry

Anodized and submerged in deionized water at 90°C for 10 min and wiped dry Ultrasonically degreased in acetone for 10 min and in ethyl acetate for 10 min, wiped dry Degreased and submerged in a solution of 10 g ammonium bifluoride, distilled water to 1 1, room temperature for 2 min; rinsed in running tap water for 1 min, allowed to stand in deionized water for 5 rnin and wiped dry Degreased and submerged in a 1 M phosphoric acid solution, 20°C, potentiostatically anodized at 10 V for 20 min; rinsed in running tap water for 1 min, allowed to stand in deionized water for 5 min and wiped dry Anodized and submerged in deionized water at 90°C for 10 rnin and wiped dry

and 55 f 5% RH for 2 h. Some aluminum sheets were coated with gold ( ~ 1 0 0 nm) using a SCDOO5 Sputter Coater from Bal-Tec.

2.1.2. Monomers and polymers. The bifunctional epoxy prepolymer used was a liquid diglycidyl ether of bisphenol A (DGEBA, M = 348 g/mol), DER 332 from Dow Chemical. The cyclo-aliphatic diamine curing agent used was isophorone- diamine (IPDA or 3-aminomethyl-3,5,5-trimethylcyclohexylamine) from Fluka. An aromatic diamine. 4,4'-methylenebis[3-chloro-2,6-diethylaniline] (MCDEA) from Lonza and an aliphatic diamine, polyoxypropylene diamine (D400) from Texaco were also used. The epoxy prepolymer and curing agents were used without further purification. A stoichiometric ratio ( r ) (aminohydroged epoxy) equal to 1 was used throughout the work and was calculated assuming a functionality of 4 for the diamine and 2 for the epoxy monomer. Homogeneous mixtures of DGEBA and diamine were achieved by stirring under vacuum ( ~ 1 Pa) at room temperature for 1 h (Rotavapor RE211 from Biichi, Switzerland) to avoid air bubble formation. The epoxy-amine adhesive cure cycle [13-151 was adapted to obtain both the maximum cure conversion, i.e. the highest glass transition temperature denoted

242 S. Bewtadjine et al.

Tgm, and the thickest interphase formation. For IPDA, the curing cycle was: 3 h at 20"C, 20 + 60°C (2"C/min), 2 h at 60"C, 60 + 140°C (2"C/min), 1 h at 140"C, 140 + 190°C (2"C/min), 6 h at 190"C, cooling (8 h) in the oven to 20°C; for MCDEA: 3 h at 20"C, 20 + 60°C (2"C/min), 2 h at 60"C, 60 + 160°C (2"C/min), 4 h at 160"C, 160 + 190°C (2"C/min), 9 h at 190"C, cooling (8 h) in the oven to 20°C; and for D400: 3 h at 20"C, 20 + 60°C (2"C/min), 2 h at 60"C, 60 + 100°C (2"C/min), 3 h at lOO"C, cooling (8 h) in the oven to 20°C.

2.1.3. Sample preparation. Several layers of an adhesive tape (5413 from 3M), approximately 50 p m thick, were applied all around the periphery of 100 x 100 mm2 treated metallic sheets to obtain the desired wet coating thickness. The epoxy- diamine mixtures were poured onto the metallic surfaces and spread with a cylindrical glass rod. For the bulk materials, 10 x 10 x 100 mm3 bars were prepared using a silicone (RTV 501 from RhBne-Poulenc) or PTFE mould. Only the central parts of the bars were used for analysis. After curing and cooling down, the coating thicknesses (from 40 to 1500 pm) were determined using a EG-100 Digital Linear Gauge (from Ono So& Co, Japan) having a f 2 p m sensitivity.

2.1.4. Monomers analysis. To characterize the change in the monomers, either the liquid monomer DGEBA or IPDA was applied between two chemically treated metallic substrates (100 x 50 mm2) to form a 110-150 p m thick wet film and kept at room temperature for 3 h in a desiccator under continuous nitrogen flow to prevent any monomer carbonization or oxidation. The modified monomers were then scraped from the metallic substrates with a PTFE spatula and stored in polyethylene vials under a nitrogen atmosphere until used.

2.2. Experimental techniques

2.2.1. Differential scanning calorimetry (DSC). DSC experiments were carried out in a Mettler DSC 30 apparatus to determine the glass transition temperature (T,) of epoxy resins. Sealed aluminum pans containing 15-20 mg epoxy materials were heated from -50°C to 250°C at a rate of 10"C/min under a continuous flow of U-grade argon. Samples were weighed using a Mettler balance having a f l p g sensitivity. The calorimeter was calibrated with both indium and zinc. The glass transition temperature was determined by the onset point with a 10.5"C sensitivity. To evaluate the variation in T, with the coating thickness, the (T,& value at thickness ( i ) was calculated using:

where Tgi corresponds to the Tg of a h; thick coating and Tgl-, value corresponds to the glass transition temperature of a h;-l thick coating. The glass transition amplitude (width) corresponding to the difference between the end temperature of the glass transition and the onset values of T, was denoted as AT,.

Epoxy-diamine adhesives on metals 243

2.2.2. Dynamic mechanical thermal analysis (DMTA). Dynamic viscoelastic measurements on coatings, after removal from the metallic substrate, were per- formed in a Rheoinetrics Solids Analyzer RSA I1 apparatus, using the tensile mode to determine the storage ( E ' ) and the loss ( E " ) moduli and the ratio E"/E ' = tan 6 (transition from glassy to rubbery behavior) as a function of temperature. Samples (4 x 40 mm2) were heated from -150°C to 250°C using a heating rate of 2"C/min by a forced convention oven using a nitrogen stream. The sample was deformed sinusoidally to a controlled strain amplitude of 0.05% at a fixed frequency of 1 Hz. Measurements were made at 30 s intervals.

2.2.3. Infrared and near-infrared spectroscopies (FTIR, FTNIRJ. An infrared spectrometer (FTIR Magna-IR 550 from Nicolet) was used with Omnic FTIR software. An Ever-GloTM source was used along with a KBr beam-splitter and DTGS-KBr or MCT/A detector. The mid-infrared spectra were recorded in the 400-4000 cm-' range and in the 4000-7000 cm-I range for the near-infrared. A transmission accessory was used for bulk or free-standing film characterization. For the bulk material, KBr was mixed in a 1 : 100 ratio with the various epoxy resins which had been ground cryogenically. This mixture was then pressed under vacuum to obtain disks. For analysis, pure KBr disks were used as background. For each analysis, 64 or 96 scans were collected at 4 cm-' resolution.

The cure conversions for epoxy and amine groups were calculated by using, respectively, the 4530 cm-' epoxy combination band and the 6500 cm-' amine band [16]. The 4623 cm-I aromatic C-H ring stretch combination band was considered as reference. Thus, the amine ( X , ) and the epoxy ( X , ) conversions were determined using the ratio of the respective peak areas (A) by:

(2) (A4530/A4623)t and X , = 1 - (A6500/A4623)t xa = 1 --

(A6500/A4623)t=O (A4530/A4623)t=0 Once again, to evaluate the X a and X e variation versus the coating thickness we calculated the ( X a ) e q and ( X , bq values at thickness (i) using:

(3) hi(Xa.,e)i - hi-'(Xaeli-1

h, - hl-1 ( X a , e ) e q

2.2.4. Nuclear magnetic resonance spectroscopy (NMR). For proton and carbon Nuclear Magnetic Resonance ('H and 13C NMR) a Bruker AC 400 spectrometer was used with the sodium salt of deuterated trimethyl silyl propionic acid (TSPd4) as the internal standard. Analyses were carried out at 293 K using deuterated water in which pure IPDA and modified IPDA were perfectly soluble. By comparing pure IPDA and modified IPDA spectra it was possible to determine the differences in chemical shifts for the equivalent carbon nuclei.

2.2.5. Inductively coupled plasma (ICP) spectroscopy. An ICP spectrometer (Modula by Spectro Analytical Instruments) was used with a 2.5 kW plasma generator at 27 MHz and with UV (0.75 m, 2400 groovesmm-'; 160-480 nm)

244 S. Bentadjine et al.

monochromator, UV (0.75 m, 3600 grooves mm-' ) polychromator and visible (0.75 m, 1200 grooves mm-') polychromator. A cross-flow nebulizer was used to introduce the liquid sample. Water was used as solvent.

2.2.6. Size exclusion chromatography (SEC). A Waters apparatus was used with double detection (UV Waters 484 Tunable Absorbance Detector at h = 254 nm and Waters 410 differential Refractometer) and a Waters 510 HPLC pump. The elution solvent used was tetrahydrofuran (THF) and the separation was carried out on two styrene- divinyl benzene columns (Nucleogel 100-5 and 500-5 from Macherey- Nagel) with a flow rate of 1 ml/min. The mass average molar mass calibration curve was constructed from monodispersed polystyrene standards.

2.2.7. Polarized optical microscopy (POM), scanning electron microscopy (SEM) and electron microprobe analysis (EMPA). Drops of diamine or epoxy- diamine mixture were confined between two glass plates and mounted on a hot plate under a POM apparatus (Laborlux 12POLS from Leica equipped with a hot plate FP82 from Mettler and a CCD-IRIS color video camera from Sony). Samples were heated from 30°C to 190°C at 10"C/min. A scanning electron microscope (Philips XL20) fitted with an electron microprobe analysis accessory (Edax-Econ4) was also used. Samples were neither coated with gold nor with carbon. The accelerating electron voltage was 5 kV, the electron spot diameter for micro-analyses was about 200 nm and the tilt angle used was 15".

3. RESULTS AND DISCUSSION

3.1. Bulk and coating characteristics

According to the curing cycle mentioned previously, the maximum glass transition temperatures (T,) of DGEBAAPDA, DGEBA/MCDEA and DGEBA/D400 bulk epoxies were 163"C, 188°C and 34°C respectively. For these bulk materials, the amine conversion (X,) and the epoxy conversion (X,) were equal to 1. These values are in good agreement with previous works [13-151. Variations in the glass transition temperatures ( Tg)eq and glass transition widths (AT,) versus DGEBA/IPD coating thicknesses applied onto both chemically etched titanium and gold coated substrates are represented in Fig. 1. Variations in the epoxy conversion (X,) and amine conversion (X,) versus the thickness of DGEBA/IPDA coatings applied onto chemically etched titanium alloy and gold coated substrates are shown in Fig. 2. Regardless of coating thickness, for the coatings applied onto a gold coated substrate, the (T& was equal to the (T,X) of the cured bulk DGEBA/IPDA material and the amine (X,) and epoxy ( X , ) conversions were equal to 1. Similarly for coatings applied onto chemically etched titanium that were thicker than 700 pm, (Tg)eq, (X,) and (X,) were all equal to the cured bulk DGEBA/IPDA system. For the thinnest coatings (< 400 pm), (Tg)eq values were quite constant (e 129°C) but very different from the bulk ones (163°C) and the amine conversion (X,) was 84%

Epoxy-diamine adhesives on metals

35

30

25 E

20

15

10

245

170

160

150 0 0 v

5 i h

M 140 e

130

1 20 0 0.2 0.4 0.6 0.8 1 1.2 1.4

Coating thickness (mm)

Figure 1. Variation of glass transition temperature ( Tg)eq and glass transition width ( A T g ) versus the thickness of DGEBA/IPDA coatings applied onto chemically etched titanium alloy and gold coated substrates.

100

95

h

5 x" 3 %

A

v

85

80

100

95

h

E

d w

90 A

85

80 0 0.3 0.6 0.9 1.2

Coating thickness (mm)

Figure 2. Variation of the epoxy conversion (X&, and amine conversion (Xa)eq versus the thickness of DGEBA/IPDA coatings applied onto chemically etched titanium alloy and gold coated substrates.

while the epoxy conversion ( X , ) = 100%. Thus it can be assumed that for these thin coatings, a different epoxy network was formed. For coating thicknesses between 400 < h < 700 pm, a gradient region was observed which is also well depicted

246 S. Bentadjine et al.

I

- - C - .Anodized

- - 0- - - Hjdroljized 128 - - + - NH I F,HF (1 OgL)

120 0 0,3 0 6 0,9 12 1.5

Coating thickness (mm)

Figure 3. Variation of glass transition temperatures (T&, versus DGEBA/IPDA coating thicknesses applied onto titanium alloy as a function of the substrate surface treatment.

by the (AT,) curve in Fig. 1. In all the following work, the interphase thickness (or width) will be defined as the region where the properties are different from those of the bulk. In the case of the DGEBADPDA coatings applied onto chemically etched titanium, the interphase thickness was about 700 pm. This means that for a 1 mm thick coating, the first 700 p m from the substrate surface corresponds to the interphase and that only the remaining 300 p m will have the same properties as the DGEBADPDA bulk epoxy. This surprising thick interphase formation will be explained in Section 3.2.

The variation of the glass transition temperatures ( Tg)eq versus DGEBADPDA coating thickness on titanium alloy as a function of the substrate surface treatment is shown in Fig. 3. Again, it can be observed that for thick DGEBA/IPDA coatings (T,)eq values are about the same as for the bulk material. For the thinnest coatings, (T,)eq values are similar to each other, regardless of surface treatment but much lower than for the bulk system. This suggests that the new epoxy network formed close to the metallic surface is the same irrespective of the surface treatment. On the other hand, the interphase width depends on the substrate surface treatment. According to previous work [ 171, after degreasing, chemical etching and anodizing, the surface oxide (Ti02) thickness was 5, 35 and 200 nm respectively. During hydrolysis, the outermost part of the oxide layer is transformed to hydroxide (TiO2, xH2O) without any significant thickness variation. Clearly then, the formation of the new epoxy network does not depend on the oxide layer thickness (i.e. the surface treatment), but its amplitude (or thickness) does in fact greatly depend upon the surface treatment. The thickness of this new epoxy network increases when the hydroxyl groups are present on the substrate oxide surface. The variation of the glass transition temperature ( T,)eq versus DGEBA/IPDA coating thickness applied onto aluminum alloy as a function of the substrate surface treatment is shown in Fig. 4. Similar behavior was observed as in the case of the titanium substrates.

Epoxy-diamine adhesives on metals

170 I 241

160

150

e 140 v Q)

I-" 130

120

110

- v

- - o- - chemical etching - - m- - degreasing - - 0- - anodization - - 0- - hydrolysis - . . - . bulk

100 0 0,2 0,4 0,6 0,8 1

Coating thickness (mm)

Figure 4. Variation of glass transition temperature (Tg)eq versus DGEBA/IPDA coating thickness applied onto aluminuin alloy as a function of the substrate surface treatment.

However, for the thinnest coatings, (T& values are lower ((Tg)eq % llO°C) than the ones obtained with titanium ((T,)eq % 129°C). Also, the interphase width is smaller when coatings are applied onto aluminum (150 < (AT,) < 250 pm) than in the case of titanium substrates (400 < (A Te) < 700 pm). Taking into account the basic nature of the diamine curing agent (pH * 13- 14), a dissolution of the metallic oxide or hydroxide by the amine is possible when the liquid epoxy-diamine mixture was applied onto the metallic substrate. The dissolution phenomenon should depend on both the nature of the oxide and the nature of the diamine agent. According to Ellingham [18], standard free energies at room temperature of formation of A1203 and Ti02 oxides are equal to 250 and 210 kcal/mol 02, respectively. In a first approximation, we can assume that the oxide layer formed on aluminum alloy was more stable than the one obtained on titanium alloy. Thus, for the same diamine monomer, the dissolution rate of Ti02 should be higher than that of A1203 leading to a wider interphase for Ti02. Experimentally, this is observed. After the partial oxide and/or hydroxide dissolution, metallic and hydroxyl ions are formed. Metallic ions may react with liquid monomers to form complexes leading to new networks, while hydroxyl ions increase the molecular mobility of the epoxy network leading to a decrease of the glass transition temperature [19]. Since the degree of hydration of A1203 is higher than Ti02 due to a minimization of surface free energy, the number of hydroxyl groups available to interact with the amine is also higher. This would then be expected to produce the largest decrease in (Tg)cq (compared to the bulk T,) for the aluminum alloy compared to the titanium alloy. Experimentally, this is also observed. The variation of glass transition temperature (T& and glass transition width (AT,) versus the thickness of DGEBA/MCDEA and DGEBA/D4OO coatings applied onto chemically etched titanium alloy are shown Figs 5 and 6 respectively. Both systems displayed a similar behavior to

248 S. Bentadjine et al.

25

9 2o W e

1s

lo

5

, I I \ - , I '!$ ,/' .-*/-= - 185

f 1 , I 1 , \ I , / e-4 / ' \ \ / , --+- AT8 - I8O

J i n

--e- V g j e q d' '1

I t 1 1

- 175 e - / + W

I I

I -<-*--..---- - 170 -* - Intekhase Width -*---- * - - e -

I

8 -e 165

chemically etched titanium alloy.

10

9

- 8 2

7

6

5

I S ,

#' 1 . / . I

' /

Interuhase width 4 II --=*=--

35

33

31 s v

29 5 jl

n i

21

25 0 0.3 0.6 0.9 1.2 1.5

Coating thickness (mm)

Figure 6. Variation of glass transition temperature (Tg)eq and glass transition width ( A T g ) versus the thickness of DGEBA/D400 coatings applied onto chem- ically etched titanium alloy.

that found for the DGEBA/IPDA system. That is, for thin coatings, (T& values were much lower than the bulk (T,X), but for thicker coatings the (T& values were

Epoxy-diamine adhesives on metals

o'8 r- 249

0.4 L I I I

D400 IPD MCDEA

Diamine curing agent

Figure 7. Variation of the interphase width for coatings applied onto chemically etched titanium alloy versus the nature of the diamine curing agent: IPDA (cyclo-aliphatic), MCDEA (aromatic) and D400 (aliphatic) diamines.

similar to (Tgx). The systems differed from the DGEBA/IPDA system, however, in that the DGEBA/D400 system displayed a very sharp transition zone while the DGEBA/MCDEA system displayed a wide transition zone. The variation of the interphase width (A Tg) for coatings applied onto chemically etched titanium alloy as a function of the nature of the diamine curing agent, i.e. IPDA (cyclo-aliphatic), MCDEA (aromatic) and D400 (aliphatic) diamines is presented in Fig. 7 . Indeed, it is observed that the interphase width depends strongly on the nature of diamine curing monomer.

3.2. Monomers characteristics

To characterize the chemical changes occurring, the liquid monomers DGEBA or IPDA were placed between two chemically treated metallic substrates and kept at room temperature for 3 h. Then, the modified monomers were scraped from their metallic substrates with a PTFE spatula and visually compared to the pure monomers. Both pure monomers (DGEBA and IPDA) were transparent. No color change of the DGEBA monomers after application onto titanium or aluminum or IPDA monomer after application onto a gold coated substrate was observed. The IPDA monomers from both aluminum and titanium became milky white (Fig. 8). After two hours within test tubes, separation was observed. The floating part remained transparent while the precipitate was milky white. One drop of the precipitate from the aluminum modified IPDA was placed between two glass plates and observed using a polarized optical microscope (Fig. 9). Crystals in the IPDA liquid were observed corresponding to the new chemical product formed when IPDA monomer reacted with the metallic oxide and/or hydroxide surface. The same phenomenon was observed with titanium. FTIR spectra of pure DGEBA and

250 S. Bentudjine et al.

Floating part

Precipitate

Figure 8. Visual aspect of pure IPDA monomer and modified IPDA monomer after application onto aluminum or titanium substrate.

Insoluble crystals

0.5 rnrn Figure 9. POM photograph of modified IPDA monomer after application onto aluminum substrate.

aluminum and titanium modified DGEBA are shown in Fig. 10. All spectra are identical suggesting that no reaction occurred between DGEBA and the metallic surfaces. FTIR spectra of pure IPDA and modified IPDA after application onto aluminum, titanium or gold coated substrate are shown in Fig. 11. Spectra of pure

Epoxy-diamine adhesives on metals 25 1

2.2

1.7

1.2

0.7

0.2

-0.3 3400 2900 2400 1900 1400 900 400

Wavenumber (cm-’)

Figure 10. FTIR spectra of pure DGEBA monomer and modified DGEBA monomer after application onto titanium and aluminum substrates.

IPDA and modified IPDA after application on a gold coated substrate or the floating part of modified IPDA on aluminum and titanium are identical. However, new bands (from 1100- 1350 cm-’ and 600-700 cm-’) are observed only when IPDA was applied onto either titanium or aluminum substrates. FTIR results confirmed that no chemical reaction occurred when coatings were applied onto the gold coated substrates and thus it is not surprising that (T&, X, and X, values were identical to the bulk values regardless of the coating thickness. Conversely, the diamine monomer (IPDA) clearly reacts with both titanium and aluminum substrates. Figure 12 shows the ICP spectra of pure IPDA and the precipitate part of the titanium modified IPDA. The presence of titanium stands out clearly in the modified IPDA monomer spectrum. The same findings were observed with aluminum. Thus the hypothesis of the dissolution of the titanium or aluminum oxide and/or hydroxide by the basic diamine monomer is strengthened. Figure 13 contains the UV and RI SEC spectra obtained from pure diamine monomer (MCDEA) and modified MCDEA with titanium and aluminum substrates. New small peaks are observed which correspond to a, = 1011 g/mol which is two times higher than the pure MCDEA monomer mass average molar mass (M, = 489 g/mol). To characterize chemical bindings that may appear between the IPDA monomer and metallic ions, nuclear magnetic resonance spectroscopy (NMR) was used. Pure IPDA monomer was actually a mixture of 25% “cis” and 75% “trans” isomers and the corresponding conformations and the carbon atom labeling are shown in Fig. 14. ‘H and I3C NMR spectra of pure IPDA monomer and modified IPDA applied onto titanium and aluminum are shown in Figs 14 and 15 respectively. After a complete peak assignment (the details will be published elsewhere), it was noted that some of the nucleus characteristic resonant frequencies varied (magnetic shielding) when IPDA was applied onto titanium or aluminum. From the differences between characteristic resonant frequencies, chemical shifts can be calculated [20].

252 S. Bentadjine et a1

0.41

0.36

2 0.31 4

4 1

0.26

0.21

0.16

750 700 650 600 550 500 450 400

Wavenumber ( cm")

0.64

0.44

0.24

0.04

1350 1250 1150 1050

Wavenumber (ern-') Figure 11. FTIR spectra of pure IPDA monomer and modified IPDA monomer after application onto titanium, aluminum and gold coated substrates and the floating part in Fig. 8.

Table 2 shows NMR chemical shifts for various carbon atoms when IPDA was applied onto titanium and aluminum. Only carbon atoms associated with nitrogen atoms or neighboring carbon atoms were affected when the monomer was applied onto metallic substrates and the shift was found to be the same regardless of the isomer. The chemical shift obtained for aluminum was about two to three times higher than the one observed on titanium. From these results, it can be assumed that coordination covalent bindings are formed between aluminum or titanium metallic

Epoxy-diamine adhesives on metals 253

2700

8 2250

*' 1800

v

c) h

3 1350

334.7 334.85 335 335.15

Wavelength (nm)

Figure 12. ICP spectra of pure IPIIA monomer and modified IPDA monomer after application onto titanium substrate.

1 9

1 7

1.1

0.9

14 15 16 17 18 19

Time (min.)

1 0.279

2 " 1 a .D": 0.275 v;

0 271

15 16 17 18 19

Time (min.)

Figure 13. UV and RI SEC spectra of pure MCDEA monomer and modified MCDEA monomer after application onto titanium and aluminum substrates.

254

Cis

S. Bentudjine et al.

Trans

a- Pure ZPDA

b- ZPDA/AI

c- ZPDA/Ti

4

b)

c) 3 1 3 0 2 9 2 8 2 7 2 6 ? 5 2 4 2 3 2 2 2 1 2 0 1 9 1 8 1 7 1 6 I S 1 4 I 3

Figure 14. “Cis” and “trans” pure IPDA isomer conformations and ‘H NMR spectra of pure IPDA and modified IPDA monomers after application onto titanium and aluminum.

ions and nitrogen atoms. The complexes formed are insoluble in the liquid monomer leading to the crystal formation. About thirty years ago, the same phenomenon (complex formation) was reported for organic pigments where metallic ions were introduced to stabilize the monomers [21] and more recently other authors have described the complex formation between amines and metals [22, 231. According

Epoxy-diamine adhesives on metals 255

X : trans 1

56 54 52 50 48 46 44 42 40 38 36 34 32 30 28 26 2 4 (ppm)

Figure 15. 13C NMR spectra of pure IPDA and modified IPDA monomers applied onto titanium and aluminum.

to Pourbaix et al. [24], the fluorine ions increase the amine dissolution rate of the metallic oxide layer by increasing the rate of complex formation. Since we know that the oxide layer formed after the chemical etching of the titanium alloy contains fluorine [17]; it is thus not surprising that we obtained the widest interphase with such surface treatment.

3.3. Modijied epoxy-diamine systems

To fully understand the properties of thin films when DGEBA-diamine mixtures are applied onto metallic surfaces and to verify our various hypotheses, we prepared some epoxy mixtures with pure DGEBA and both titanium and aluminum modified IPDA monomers. The stoichiometric ratio was equal to 1 and the curing conditions were the same as for the pure DGEBA/IPDA system. Figure 16 contains POM photographs obtained before and after the cure cycle for a pure DGEBA/aluminum modified IPDA mixture. The same photograph was obtained for a pure DGEBA/titanium modified IPDA mixture. Comparing Fig. 9 to Fig. 16, a dilution effect is observed before curing, but all crystals are insoluble in the pure DGEBA/aluminum modified IPDA mixture. After curing, crystals were still observed but around those crystals, phase separation (black parts in the photograph) occurred. It is quite evident that the presence of crystals and the phase separation

256 S. Bentadjine et al.

Table 2. NMR chemical shift [As (ppm)] for cis and trans modified IPDA isomers on titanium and aluminum

Trans isomer Cis isomer Applied onto Applied onto titanium aluminum

C1 0.22 0.23 0.18 c2 0.22 0.12 c3 0.10 0.16 c9 0.15 0.10 ClO

c;

c;

c;

C$

c; 0 0.10

0.44 0.47 0.6 0.68 0.30 0.24 0.44 0.39 0.23 0.29

Table 3. Physical (glass transition temperature ( Tg) ) , chemical (epoxy conversion (X,) and amine conversion (X , ) ) and mechanical (Young's modulus ( E ) ) properties of pure DGEBA/pure IPDA, pure DGEBA/Al modified IPDA, pure DGEBA/Ti modified IPDA mixtures compared to thin coatings applied onto titanium or aluminum

Tg ("C) x, X e E (GPa)

Bulk 163 1 1 3.2

on titanium Pure DGEBA/ 132 0.86 1 Ti modified IPDA

Thin coatings 129 0.84 1

Thin coatings 110 0.92 1 5*?18** on aluminum Pure DGEBA/ 108 0.90 1 5 A1 modified IPDA

*For a 100 k m thick coating For a 40 k m thick coating. **

lead to the formation of a new polymer. The properties of such modified polymers are reported in Table 3 and are compared to thin coatings on titanium and aluminum and to the pure bulk polymer. The properties are the same as the properties of the polymer network formed on aluminum and titanium prior to the transition zone of the interphase. This means that we have been able to reproduce within a modified bulk system the various phenomena observed for the thin coatings. It should be mentioned here that the increase of the mechanical properties (Young's modulus) can be attributed to the crystals acting as short fibers in a polymer matrix [25] as shown in Fig. 16. The highest values of Young's modulus are observed for the thinnest coatings where most of crystals are oriented parallel to the metallic

Epoxy-diamine adhesives on metals 251

Before curing After curing

500 ym

4- crystals not oriented 1 7 + crystals oriented closc to the aluminium - Aluminium

Figure 16. POM photographs of pure DGEBA/Al modified IPDA mixture (a/e = 1) before and after curing cycle and A1 modified IPDA crystals orientation close to the aluminum surface.

surface. Lastly, pure DGEBA/pure IPDA and pure DGEBA/Ti modified IPDA bulk materials and free coatings applied onto titanium were analyzed using dynamical mechanical thermal analysis. The tan 6 versus the temperature spectra are reported in Fig. 17. For pure DGEBA/pure IPDA bulk, the tan 6 curve displays a maximum at 172°C while for the pure DGEBA/Ti modified IPDA bulk, tan 6 curve presents a maximum at 151°C and a shoulder for high temperatures close to 170°C. For coatings upon removal from the metal substrate, a peak close to 170°C was observed and a shoulder at low temperatures close to 151°C. Thus, the findings from the DSC data were strengthened. It should be mentioned that the difference between tan6 peak values and DSC Tg values was due to the frequency effect and the corresponding definition (for DSC, Tg was determined by the onset point while for DMTA the maximum peak corresponds to the inflection point) [26].

25 8 S. Bentadjine et al.

0.98

0.78

0.58

(.o

8 0.38

0.18

-0.02

+Modified IPDA/Ti

+ 0.43 mm thick coating

80 100 120 140 160 180 200 220

Temperature ("C)

Figure 17. DMTA spectra of pure DGEBA/pure IPDA mixture, modified IPDA/pure DGEBA mixture and 0.43 mm and 1 mm thick coatings formed on titanium alloy.

4. SUMMARY AND CONCLUSION

When epoxy- diamine prepolymers are applied onto metallic substrates, an inter- phase between the coating part having the bulk properties and the metallic surface was created. Chemical, physical and mechanical properties of the interphase formed depend on the substrate nature and its surface treatment, the nature of the diamine hardener, and the curing cycle. Using an appropriate curing cycle of the prepolymer allowed us to reach the maximum conversion to the polymer and the full interphase formation. This enabled us to propose interphase formation mechanisms leading to a better understanding of its chemical, physical and mechanical properties. Mech- anisms describing the formation of the interphase were deduced from comparison of behaviors when epoxy- diamine mixtures were applied onto aluminum, titanium and gold coated surfaces. When either pure DGEBA or diamine monomers were applied onto gold coated surfaces, no chemical reaction was observed. In the same way, when the pure DGEBA monomer was applied onto the metallic surfaces, no chemical reaction was observed. On the contrary, when pure diamine monomers were applied onto metallic surfaces, chemical reactions occurred. According to the basic behavior of diamine monomers, a partial dissolution of the surface oxide and/ or hydroxide metallic substrate was observed. Then metallic ions diffuse within the liquid monomer mixture (epoxy-diamine) and react with the amine groups of the diamine monomers to form coordination complexes. When the complex concen- tration is higher than the solubility product, these complexes crystallized as sharp needles. They align themselves parallel to the metallic surfaces leading to an ori- ented crystalline layer in the vicinity of the substrate surface. These complexes and crystals are insoluble DGEBA/diamine mixtures. During the curing cycle, the

Epoxy-diamine adhesives on metals 259

crystals are not fully dissolved and a phase separation was observed to occur all around these crystals inducing the formation of a new epoxy network. Both the phase separation and the crystals orientation induce noteworthy mechanical prop- erties of the interphase. The same phenomena would be expected for all kinds of metallic substrates as soon as they are covered by an oxide or hydroxide layer. Also, since dissolution and diffusion phenomena are expected, the interphase formation is related to the contact duration between liquid diamine and metallic substrates. Lastly, the interphase properties (thickness, T,, E , . . .) depend on the nature of the metallic substrate and its surface treatment and on the diamine nature, the mixture viscosity, and the liquid contact duration.

Acknowledgements

We are very grateful to R. Pitiaud and V. Massardier for performing the NMR analyses and to R. Diemiaszonek for the ICP analyses.

REFERENCES

1. E. Peillex, Ph.D. thesis, UniversitC de Lyonl (1996). 2. C. Fauquet, Ph.D. thesis, UniversitC Paris 6 (1992). 3. P. Walker, J. Coat. Technol. 52,49 (1980). 4. R. D. Guminski and F. M. P. Meredith, J. Oil. Colour Chem. Assoc. 44, 93 (1961). 5. A. Entenberg, V. Lindberg, L.. Fendrock, Sang-Ki Hong, T. S. Chen and R. S. Horwath, in:

Metallized Plastics 1: Fundamental and Applied Aspects, K. L. Mittal and J. R. Susko (Eds), pp. 103- 113. Plenum Press, New York (1989).

6. F. J. Von Preissing, J. Appl. Phys. 66, 4262 (1989). 7. P. Scafidi and M. Ignat, J. Adhesion Sci. Technol. 12, 1219- 1242 (1998). 8. H. Orsini and F. Schmit, J. Adhesion 43, 55 (1993). 9. M. D. Thouless and H. M. Jensen, J. Adhesion Sci. Technol. 8, 579 (1994).

10. K. L. Mittal, in: Adhesion Measurement of Thin Films, Thick Films and Bulk Coatings, K. L. Mittal (Ed.), STP 640, pp. 5- 17. ASTM, Philadelphia (1978).

11. L. H. Sharpe, J. Adhesion 4, 51 (1972). 12. M. F. Finlayson and B. A. Shah, J. Adhesion Sci. Technol. 4, 431 (1990). 13. 0. Sindt, J. Perez and J. F. Gerard, Polymer 37, 2989 (1995). 14. L. Bonnaud, Ph.D. thesis, INSA de Lyon (1999). 15. H. Masood Siddiqi, Ph.D. thesis, INSA de Lyon (1997). 16. N. Poisson, G. Lachenal and H. Sautereau, Vibrational Spectros. 12, 237 (1996). 17. A. A. Roche, Thkse d’Etat, LniversitC de Lyonl (1983). 18. H. J. T. Ellingham, J. SOC. Chwz. Ind. 63, 125 (1944). 19. A. H. Sabra, Ph.D. thesis, INSA de Lyon (1985). 20. F. K. Bovey (Ed.), High Resolution NMR of Macromolecules. Academic Press, London (1972). 21. S. Wernick and R. Pinner (Eds), Les traitements de surface et lajinition de l’aluminium et de ses

22. Y. H. Kim, G. F. Walker, J. Kim and J. Park, J. Adhesion Sci. Technol. 1, 331 (1987). 23. M. Kinzler, M. Grunge, N. Blank, H. Shenkel and I. Scheffler, J. Vac. Sci. Technol. A10 (4),

24. M. Pourbaix, N. de Zoubov and J. Van Muylder (Eds), Atlas d’Equilibres Electrochimiques,

alliages. Eyrolles, Paris (1962).

2691 (1992).

pp. 216. Gauthier-Villas, Paris (1963).

260 S. Bentadjine et al.

25. J. Bouchet, A. A. Roche, E. Jacquelin and G. W. Scherer, this volume, pp. 217-237 26. A. Pierre, Ph.D. thesis, INSA de Lyon (1997).

Adhesion Aspects of Thin Films, Vol. 1, pp. 261 -270 Ed. K. L. Mittal 0 VSP 2001

Improving the adhesion of plasma polymerized thin fluorocarbon films on silicon using (CHF3 + SF6) radio-frequency discharge pretreatments

VASSIL IANEV and NORBERT SCHWESINGER* Ilmenau Technical University, Faculty of Mechanical Engineering, Department of Microsjstem Technology, PF 100 565, 0-98684 Ilmenau, Germany

Abstract-Thin fluorocarbon (FC) polymer films were deposited via plasma polymerization of trifluoromethane (CHF3) onto silicon substrates in a commercial reactive ion etching system (RIE, radio-frequency 13.56 MHz). In this way, plasma-polymerized fluorocarbon coatings have shown a low surface energy (sessile contact angle of water 106') and poor adhesion to the silicon wafers. To improve the adhesion of the FC films to the silicon substrates, pre-treatments of the silicon surfaces in CHF3 + SF6 plasma were applied before deposition in the same reaction chamber. The adhesion of plasma polymerized FC coatings to pre-treated and untreated silicon wafers was determined by using a simple test according to Yasuda, in which the coated silicon substrates were boiled in a 0.9% sodium chloride solution. The test results indicate that plasma pre-treatments are important and necessary for a good adhesion of FC coatings to silicon surfaces.

Selective FC coatings on silicon are useful in several chemical and biochemical applications. Therefore, the patterning of FC films is a very important task. Direct patterning of thin FC films is very difficult due to the low surface free energy of the FC polymers. Spin-coated resists do not adhere to the FC surface and so common lithography processes are often unsuitable. Therefore, various spin- coating process parameters were examined. A lift-off process was also investigated and used as a patterning technique for FC films.

Keywords: Fluorocarbon (FC) polymer films; plasma polymerization; pre-treatments of silicon surfaces; adhesion properties and measurements: patterning of FC films.

1. INTRODUCTION

1.1. Plasma polymerization and properties ofjuorocarbon jilms

Fluorocarbon films can be deposited by means of several dry or wet coating methods [ 1 - 101. The most used wet coating methods are spray deposition, spin-coating

*To whom correspondence should be addressed. Tel./Fax: ++49/3677-691296; E-mai:: ianev @ mb.stud.tu-ilmenau.de

262 K Ianev and N. Schwesinger

and dip-coating. Evaporation, sputtering and plasma polymerization belong to dry coating methods. Especially plasma polymerized FC films are very attractive because it is possible to achieve better adhesion properties to various substrates.

D’ Agostino and co-workers [2] have done intensive investigations on the plasma polymerization of fluorocarbon monomers. The feed gases must include sources of fluorine and carbon to prepare the plasma-polymerized fluorocarbon coatings. Many different fluorinated monomers, such as fluoroalkyls, fluorohydroalkyls, cyclo-fluoroalkyls, fluorobenzene, alkyl metal fluorides and others can be utilized as pure feed or in a mixture with specific additives. Many of these monomers have been used in the studies for basic understanding of the plasma polymerization mechanism by Yasuda [l], d’Agostino [2], Kay et al. [3].

The properties of plasma deposited FC films are dependent on the polymerization conditions. However, fluorocarbon plasma polymers are poly(tetrafluoroethy1ene)- like and have similar properties (low Young’s modulus (-4 x lo8 N/m2), low surface free energy (- 18 mJ/m2), low friction coefficient, high chemical stability and inertness, high thermal stability, very low dielectric constant and loss factor and physiological inertness).

In the present study, depositions of fluorocarbon films onto silicon substrates were carried out, using a trifluoromethane (CHF3) glow discharge in a conventional reactive ion etching (RF = 13.56 MHz) system. The plasma-polymerized fluoro- carbon films were characterized by the refractive index and the sessile contact angle of water.

1.2. Adhesion properties of plasma-polymerized fluorocarbon $films

The adhesion properties of plasma polymers are strongly dependent on the nature of the substrate, the pre-deposition treatment, and the working conditions during the plasma polymerization process (type of monomer, design of the reactor, the location of substrate in the reactor, discharge power, gas flow, pressure, substrate temperature, frequency, and so on). According to Yasuda [l], Chan et al. [4] and Jansen et al. [ 5 ] , the following conclusions regarding the adhesion properties of the plasma polymers can be derived: 0 The adhesion of plasma deposited FC films is much better than that of spin coated

and evaporated FC films. 0 The adhesion of plasma polymer films to common organic polymers is usually

good. 0 The adhesion of FC films to smooth substrates like silicon, metals, glasses, and

ceramics is difficult. It is necessary to implement a pre-deposition treatment to improve the adhesion of the plasma polymers to these substrates.

0 Surface characteristics (roughness, cleanliness, degree of oxidation, etc.) of the substrate are very important for the adhesion.

0 Good adhesion results can be achieved by using slow deposition rates, small film thickness and polar polymer substrates.

Plasma polymerized thin fluorocarbon jilms on Si 263

0 A good adhesion of relatively thick layers is more difficult to achieve than that of thin layers.

Plasma treatment processes involve interactions of electrical discharges with free surfaces of substrates. As result of these processes, a chemical modification of the surface, or a surface activation can be achieved by etch reactions. A deposition of a thin film is prevented due to the specific process conditions. This process can improve the adhesion of polymerized films to a given substrate in multiple ways. Plasma pre-treatments are important steps in microelectronic fabrication technologies. A large number of publications deal with plasma etching and modification of inorganic layers (e.g. silicon, silicon dioxide, [2, 14-20]) and organic coatings [2, 4, 7, 101. It is well known that perfluorocompounds (fluorine- containing gases) are used in the dry treatment, modification and etching of silicon. The most used process gases are CHF3, CF4, and SF6. The properties of these gases during the etching in plasmas have been investigated by many authors [2, 14-20]. Depending on the specific process conditions a given gas composition may lead to an activation, etching (isotropic or anisotropic) of the surface or to a deposition of thin films.

In the present study, several deposition processes with various plasma conditions were used for the generation of thin fluorocarbon coatings. In some cases plasma treatments were employed before the deposition to improve the adhesion. These treatments were carried out in the deposition process chamber in the presence of CHF3 + SF6.

1.3. Test methods for the measurement of adhesion properties of plasma polymerized thin juorocarbon jilms

The adhesion measurement method for a polymer to a substrate depends on the nature of the polymer-substrate system [6]. Typical methods are the “Adhesive tape tests” (ASTM D3359), “Butt joint tests” (ASTM D897), “Lap shear tests” (ASTM D1002), “90” peel tests” and “180” peel tests” (ASTM D903). The quantitative adhesion evaluation of plasma-deposited thin layers (thickness < 1 pm) is extremely difficult because of several artifacts involved in the measuring methods and procedures [ 11. Yasuda [ 11 explains that poor adhesion of plasma-deposited thin layers can be easily detected by the “Adhesive tape test” (ANWASTM D 3354-76).

Thin plasma-polymerized fluorocarbon coatings are used in several biomedical and chemical solutions as well as in water-like liquids [2, 5 , 71. Water sorption and diffusion processes in hydrophobic polymers are described by Polishchuk and Zaikov [ 111. The influence of water on the diffusion of electrolytes in the polymers was investigated and discussed by Moiseev and Zaikov [ 121. Unfortunately, very little information has been published regarding the water diffusion in thin plasma- polymerized FC coatings. Yasuda [ l ] explains that the failure of an adhering system is often caused by the action of water at the polymer-substrate interface. Under these circumstances, it is important to examine the adhesion of thin films on substrates in the presence of water. Yasuda and Sharma [ 131 have used simple test methods in which coated substrates were boiled in a 0.9% sodium chloride solution

264

Table 1. Grades and boiling test conditions in a 0.9% saline solution (from Yasuda and Sharma [13])

V Ianev and N. Schwesinger

5B 4B 3B 2B 1B OB

No lifting of the film after 3 cycles of 8 hr boiling and 16 hr room temperature soaking Lifting of the film after 3 cycles of 8 hr boiling and 16 hr room temperature soaking Lifting of the film after 1 cycle of 8 hr boiling and 16 hr room temperature soaking Lifting of the film after 60 min boiling Lifting of the film after 15 min boiling Lifting of the entire film almost immediately

Figure 1. Boiling test in 0.9% saline solution (after Yasuda and Sharma [13]).

to determine the adhesion properties of plasma-polymerized films. The test provides “semi-quantitative” results given in grades from 0 to 5 (0 is the worst, and 5 is the best adhesion, see Table 1 and Fig. 1).

In our studies, the adhesion of plasma-polymerized FC coatings to Si substrates was determined by using the test, according to Yasuda and Sharma [13]. The films were deposited under different conditions on plasma treated and untreated silicon wafers. After deposition the coated silicon substrates were boiled in a 0.9% sodium chloride solution.

1.4. Patterning of FCj lms

Thin fluorocarbon films can be etched at low pressures and relatively high powers in an RIE-system. Pure oxygen ( 0 2 ) , nitrogen (Nz), fluorocarbon ( e g SF6) or mixtures of these gases can be used as process gases [2]. The etch rates are different and strongly depend on the process conditions. However, direct patterning of FC films through masks, similar to microelectronic process techniques, is very difficult and often impossible due to the low surface free energy of the FC polymers. The spin-coated resist does not adhere to the FC surface. In order to overcome this problem surface modification techniques must be applied to increase the hydrophilicity and to improve the wettability. Thereby, adequate adhesion of the resist to the FC surface is ensured. Several methods have been developed to increase the surface free energy and improve the adhesion of FC polymers [ l , 2, 7, 10, 21, 221. Another method to pattern FC films is to use other masking materials such as metals. However, depending on the process conditions metals do

Plasma polymerized thin fluorocarbon j lms on Si 265

resist

Si substrate

mask resist

SI substrate

resist

Si substrate

FC film

resist SI substrate

FC film

Si substrate

Figure 2. Lift-off process for patterning of FC-films.

not adhere to the FC films in all cases or they diffuse into the polymer films and alter the surface properties.

Surface modifications can increase the adhesion of metallic films. These processes increase the surface free energy. Unfortunately, modifications of the polymer films or depositions of metallic films are not allowed in certain applications where the surface properties should remain as produced after patterning. Another pattern possibility for FC films is the lift-off technique (Fig. 2). A lithography step is carried out before the deposition is done. The FC film is deposited on the whole surface consisting of silicon and resist. After deposition the resist can be stripped in an organic solution. Thus, the FC film remains only on the free silicon areas.

In our studies conventional resist (AR-P35 1, ALLRESIST GmbH, Strausberg, Germany) was investigated as the masking material to pattern the thin FC films. Modification techniques were not applied after the spin-coating process. The FC films were etched in a pure oxygen plasma in an RIE-system. The lift-off process was investigated, too.

2. EXPERIMENTAL RESULTS

2.1. Deposition of FCJilrns on Si substrates

FC polymer films were deposited by plasma polymerization of trifluoromethane (CHF3) onto 4” Si wafers in a reactive ion etching system (RIE, radio-frequency, RF = 13.56 MHz, type STS 320 PC; Surface Technology Systems Ltd, Imperial Park, Newport NP1 9UJ, United Kingdom). Five different deposition conditions were used for the creation of thin FC films. The layer thickness and the refractive

266

Table 2. Parameters of the polymerization processes and properties of the deposited FC films

V Ianev and N. Schwesinger

No. Pretreat- Polymerization parameters d n 0 Rd ment P F P T t (nm) (deg) (nm/min)

(W) (sccm) (mTorr) ("C) (min)

1 No 80 25

2 No 50 25

3 Yes 80 25

4 Yes 50 25 35 25

5 Yes 35 25 80 25 50 25

130

300

130

300 130

130 130 130

20 4 116 1.41 105 29

20 4 72 1.41 106 18

20 4 115 1.41 105 29

20 8 20 4 200 1.40 106 17

20 4 20 4 20 4 250 1.42 106 21

index of the deposited layers were measured using an ellipsometer (SD 2302, Plasmos, Philips). Process parameters such as the discharge power, gas flow, pressure, temperature and process time as well as typical measurement results such as the thickness of the layers (d), refractive indexes ( a ) , sessile water contact angles (6) and calculated deposition rates (Rd) are shown in Table 2.

It is obvious from data in Table 2 that process No. 4 includes two processes with different discharge conditions, and process No. 5 consists of three partial processes.

2.1.1. Pre-treatment process. In processes 3, 4 and 5, plasma pre-treatments of the silicon samples were applied using CHF3 + SF6 as process gases before the deposition. The soft isotropic etching and the surface activation behavior during the pre-treatment process should improve the adhesion properties of the plasma- polymerized fluorocarbon films. The following process conditions of the plasma pre-treatment were used:

Discharge power: P = 5 5 W Pressure: p = 40 mTorr, Flow rate of CHF3: Flow rate of SF6: Temperature: T = 20°C, Treatment time:

FcHF3 = 5- 10 sccm, FsF6 = 15-20 sccm,

t = 2 min in process 3 t = 5 min in processes 4 and 5

All deposited fluorocarbon thin films have shown a low surface free energy (sessile contact angle of distilled water > 105"). The deposition rate is relatively high in all processes and is dependent on the discharge conditions. The refractive index of the various samples was always the same (1.40- 1.42).

Plasma polymerized thinJluorocarbon films on Si 261

Figure 3. Adhesion of plasma deposited FC thin films with and without plasma pretreatment.

2.2. Adhesion of the plasma-polymerized fluorocarbon j lms to silicon substrates

The adhesion properties of the plasma-polymerized FC coatings were determined by using a test, already employed by Yasuda and Sharma [ 131 (see Fig. 1 and Table 1) in which the silicon substrates coated with plasma FC-films were boiled in a 0.9% sodium chloride solution.

The FC thin films produced in the processes 1 and 2 were lifted after a very short time (15 minutes). Coatings generated in process 3 were lifted after the second cycle of boiling. The films produced in processes 4 and 5 withstood the complete test procedure. The results are shown in Fig. 3.

The poor adhesion of the polymerized films in the first two processes is due to the fact that these processes do not involve a plasma pre-treatment process. The difference between processes 1 and 3 is only in the plasma pre-treatment (process 1 does not contain the pre-treatment step of the silicon surface). The fluorocarbon films deposited by processes 4 and 5 have shown the best adhesion. These test results indicate that the plasma pre-treatment is very important and necessary for a good adhesion of the FC coatings to the silicon surfaces.

2.3. Patterning of FCjlms

The patterning of the FC films through a photoresist mask (conventional All resist AR-P35 1) was examined after deposition for process No. 5. Different coating parameters were investigated to improve the adhesion of the resist to the FC surface. The best adhesion results were obtained using the process parameters, shown in Table 3 . Differences in the thickness uniformity of so-deposited resists were in a range below 5%.

The samples were etched in a pure oxygen plasma in an RIE-system after the lithography steps (pre-bake, exposure, development, post-bake). A resolution of 2 p m was obtained. A significant increase in the surface energy was not observed after resist stripping. The sessile contact angle of water was 103".

2.3.1. Patterning through resist mask.

2.3.2. Lift-ofprocess for patterning thin plasma polymerized FCJilms. A lift-off process was also examined to pattern the thin FC films. The lithography steps were used before the plasma polymerization process was carried out (Fig. 2). A standard resist AR-P351 was coated directly onto the Si substrates. After all lithography

268 V Ianev and N. Schwesinger

Table 3. Optimal spin-coating process parameters

Spin-coating process Parameters

Speed (rpm) Time (s)

Spreading Forming Spinning

800 4 1000 3 3000 10

Figure 4. SEM photos of dry etched 250 nm thick FC films.

Figure 5. SEM photos of patterned FC films obtained using the lift-off process.

steps (pre-bake, exposure, development, post-bake) the FC films were grown on the samples with a thickness of about 1 pm. The photoresist was stripped in acetone after the deposition. Figure 5 shows lift-off patterned FC films. A resolution of 2 p m was obtained. An increase in the surface energy was not observed (sessile contact angle of water was nearly the same, 106").

Plasma polymerized thin juorocarbon films on Si 269

3. CONCLUSIONS

Fluorocarbon polymer deposition on pre-treated and untreated Si surfaces in CHF3 plasma with various process conditions has been investigated. All deposited thin FC films have shown a low surface free energy (sessile water contact angle >105") and refractive indexes between 1.40 and 1.42. The FC films deposited without pre- treatment of the substrate have shown poor adhesion. The adhesion test results indicate that the plasma pretreatment is very important and necessary for good adhesion of the fluorocarbon coatings to the silicon surfaces.

Direct patterning of fluorocarbon (FC) films was also investigated. The common lithography process is very difficult, and often impossible, to carry out due to the low surface free energy of the FC polymers. It was possible to coat the FC films with a resist using a modified spin-coating process. FC films patterns with a resolution of 2 p m were achieved via an etching through the resist mask. The lift-off process shows an acceptable reproducibility.

Acknowledgements

The authors would like to thank the staff of the Laboratory for Microstructuring Technologies at the Ilmenau Technical University, in particular Torsten Sandig for his help in deposition and etching processes, B. Hartmann for the sample preparation and K. Friedel for the SEM measurements.

REFERENCES

1. H. Yasuda, Plasma Polymeri.:ation. Academic Press, London (1985). 2. R. d' Agostino, Plasma Deposition, Treatment, and Etching of Polymers. Academic Press, San

3. E. Kay. J. Coburn and A. Dilks, Topics Current Chemist? 94,77 (1980). 4. C. M. Chan, T. M. KO and H Hiraoka, Suface Sci. Rep. 24, 1-54 (1 996). 5. J. Jansen, J. G. E. Gardeniers, J. Elders, H. A. C. Tilmans and M. Elwenspoek, Sensors and

6. F. Garbassi, M. Morra and E. Occhiello, Polymer Surfaces froin Physics to Technology. John

7. N. Inagaki, Plasma Surface Modijication and Plasma Polymerization. Technomic, Lancaster. PA

8. 3M Fluorad, Anti-migration Coating. 9. M. J. Keefe and J. M. Rigsbee, Mater. Res. Soc. Symp. Proc. 304. 179 (1993).

Diego (1990).

Actuators A41-42, 136- 140 (1994).

Wiley (1995).

(1996).

10. C. Chan, Polynzer Sugace Modification and Characterization. Hanser, Munich (1996). 11. A. Polishchuk and G. Zaikov. Multicomponent Transport in Polymer Systems for Controlled

Release, Polymer Science and Engineering Monographs: A State-of-the-Art Tutorial Series, Vol. 3. Gordon & Breach Science Publishers (1997).

12. Yu. Moiseev and G. Zaikov, Chemical Resistance of Polymers in Aggressive Media. Consultant Bureau, New York (1997).

13. H. Yasuda and A. K. Sharma, J. Adhesion 13, 201 (1982). 14. J. Bhardwaj. H. Ashraf and A. McQuarrie, Proceeding of the Symposium on Microstructures

and Microfabricated System.r at the Annual Meeting of the Electrochemical Socieh, Montreal, Canada (1997).

270 V lanev and N. Schwesinger

15. S. Vogel, U. Schaber, K. Kiihl, R. Schafflik, H. Pradel and F. Kozlowski, Proceeding OJ

16. R. d’Agostino, R. Lamendola, P. Favia and A. Gionel, J. Vac. Sci. Technol. A 12, 308-313

17. H. Sugawara and J. Minoru, Plasma Etching: Fundamentals and Applications. Oxford Univer-

18. B. N. Chapman, Glow Discharge Processes: Sputtering arid Plasma Etching. Wiley, New York

19. D. M. Manos, Plasma Etching: An Introduction. Academic Press, San Diego (1989). 20. S. M. Rossnagel, Handbook of Plasma Processing Technology: Fundamentals, Etching, Depo-

sition, and Su$ace. Noyes Publ., Westwood, NJ (1990). 21. I. Mathieson, Adhesion, 46,49-56 (1994). 22. J. Kelber, Mater. Res. SOC. Symp. Proc. 119, 255-263 (1988).

Microsystem Technology Symposium (Productronica 97) (1997).

(1994).

sity Press (1998).

(1980).