: rrv: “:th 4- tutti-4':- A . - ‘ '. A .‘ _ , o u‘” .I.\ v ”‘1‘“.1 - a.“ « ‘ A . V ‘:'~ f"“‘.n..'w.._;.._ '4 ‘. 9232": '- .p‘nfi-k ".... .k. ‘ . wwc'C-‘u 'v'< -.. 1 . ‘ ‘1‘ lllllllllllllulllllllllll\lllllllllllllllll 3 1293 01051 This is to certify that the thesis entitled Adhesion and Residual Stress in Sputter Deposited Nickel-Titanium Thin Films on Silicon presented by Bethany J. Walles has been accepted towards fulfillment of the requirements for Masters degree in Materials Science Major professor Date April 14, 1993 0.7639 MS U is an Affirmative Action/Equal Opportunity Institution LIBRARY Mlchlgan State Unlverslty PLACE ll RETURN BOX to romovo this chockoul from your rocord. 'ro AVOID FINES rotum on or Hot. duo duo. DATE DUE DATE DUE DATE DUE MSU lo An Affirmatlvo Action/Equal Opportunity Intuition . Wm: ——-—-——__. __ ADHESION AND RESIDUAL STRESS IN SPUTI'ER DEPOSIT ED NICKEL- TITANIUM THIN FILMS ON SILICON By Bethany J. Walles A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Materials Science and Mechanics Faculty Advisor Prof. David Grumman I993 ABSTRACT ADHESION AND RESIDUAL STRESS IN SPUTTER DEPOSITED NICKEL-TITANIUM THIN FILMS ON SILICON By Bethany J. Walles Nickel-titanium shape memory alloys are potential thin film candidates for microelectromechanical systems. NiTi thin films can be ion sputter deposited onto a silicon . substrate and can be activated by Joule heating. However. in order to develop NiTi microelectromechanical devices, fundamental characteristics of the film/substrate system, such as residual stress and adhesion must be understood. Residual film stress is an important parameter since thermoelastic transformation temperatures in NiTi are strongly affected by applied stress. Good adhesion between the film and substrate is also necessary for a robust microelectromechanical system. In the present work. Ni'I'i films were grown on (100) Si by ion sputtering and by ion beam assisted deposition (IBAD). The as- depositcd films were amorphous and were annealed after deposition at 600°C to produce a crystalline structure which was fully austenitic. Residual stress and adhesion were evaluated for both amorphous and crystalline films using profilometric and indentation methods. Films deposited at an I/A ratio (ion-to-atom arrival ratio) of zero were nearly stress-free, and as the [IA ratio increased the as-deposited film stress became more compressive due to an increasing ion-peening effect. After annealing the film stress became more tensile due to structural relaxation upon crystallization, but the films showed some memory of the pie-annealing stress state. Films deposited onto Si substrates cleaned with an HF/HzO solution prior to deposition, and ion cleaned prior to deposition, demonstrated good adhesion before and after annealing. However, adhesion of the as- deposited films was better than that of the annealed films. The deterioration of film adhesion upon annealing was attributed to a coefficient of thermal expansion mismatch between film and substrate. ACKNOWLEDGEMENTS The author is grateful to Professor David Grummon for his guidance through this research. Thanks also to Luiwen Chang and Sengho Nam for sharing their knowledge of NiTi alloys and sputter deposition. The author gratefully acknowledges the support of Ford Motor Company with special thanks to Bob Novak and Walter Winterbottom. Additional funding was provided by the National Science Foundation under grant #MSM8821755. TABLE OF CONTENTS Title Page LIST OF TABLES iv LIST OF FIGURES v CHAPTER 1. INTRODUCTION 1 CHAPTER 2. REVIEW OF LITERATURE 3 2.1 Residual Stress in Thin Films 3 2.1.1 Origins of Film Stress 4 2.1.2 Relations Between Processing Parameters and Film Structure and Stress 6 2.1.3 Ion Beam Assisted Deposition and Related Stress Effects 10 2.1.4 Methods of Measuring Film Stress 11 2.2 Adhesion of Thin Films 15 2.2.1 Film/Substrate Interface 16 2.2.2 Ion Bombardment Effects on Adhesion 17 2.2.3 Effects of Residual Film Stress on Adhesion 20 2.2.4 Methods of Measuring Adhesion 21 23 NiTi Thin Films 23 CHAPTER 3. EXPERIMENTAL PROCEDURES 28 3.1 Materials 28 3.2 Sputter Deposition Equipment 28 3.3 Sputter Deposition of NiTi 29 3.3.1 Sputtering Conditions 29 3.3.2 Substrate Preparation, Characterization and Holder Design 32 3.4 Adhesion Testing 33 3.5 Residual Stress Determination 37 3.5.1 Profilometry 37 3.6 Annealing 37 CHAPTER 4. RESULTS AND DISCUSSION 39 4.1 Deposition Temperature 39 4.2 Film Thickness 39 4.3 Ion-to-Atom Arrival Ratio Determination 40 l 4.4 X-Ray Diffraction and Atomic Structure 4.5 Residual Film Stress 4.6 Adhesion Measurements CHAPTER 5. CONCLUSIONS BIBLIOGRAPHY 11 42 52 62 LIST OF TABLES Table Page I Deposition conditions for Runs 1 through 4 36 2 Adhesion calculation parameters for all films tested 58 \OmleUtkb-DN '5 ll 12 13 14 15 16 17 18 LIST OF FIGURES Page Schematic displaying the effect of working gas pressure on the mean free path 7 of incoming sputtered atoms Substrate curvature due to residual film stress . . 12 Stoney equation parameters obtained from the profile of the film curvature l4 Schematic of unstressed film response to the indentation adhesion test 24 Schematic of the reaction of a residually stressed film to indentation 25 Schematic of sputter deposition chamber for Runs 1 and 2 3O Schematic of sputter deposition chamber for Runs 3 and 4 ., 31 Layout and dimensions of substrate holder used for Run 3 I 34 Layout and dimensions of substrate holder used for Run 4 35 Film thickness versus distance for Run 4 ion sputtered and Run 3 ion beam 41 assisted deposition films Plot of current density of assisting-ion beam versus distance from the central 43 axis of the assist flux for Run 3. X-ray intensity versus two-theta plots for as deposited and annealed films on Si 44 Residual film stress values versus I/A ratio for as—deposited films 47 Pressure versus film stress for as-deposited films 50 Residual fihn stress values versus I/A ratio for annealed films from Runs 3 51 md 4 Change in residual film stress between as-deposited and annealed films from 53 Runs 3 and 4 Adhesion testing indentation marks on Run 1 as-deposited film and Run 2 55 annealed film Adhesion testing indentation marks for as-deposited and annealed films of [BAD 56 Run 3 Iv I9 20 Adhesion indentation tesn'ng marks on as-deposited and annealed films of ion sputtered Run 4 Normalized adhesion values for as-deposited and annealed films from Runs 3 and 4 Page 57 60 CHAPTER 1. INTRODUCTION Nickel-titanium shape memory alloys are known to demonstrate rather unique shape recovery characteristics in the bulk form, and recently these characteristics have been found in NiTi thin films (Busch. et a1 1989). This shape memory alloy is a material which is capable of undergoing strains as high as 8% at a low temperature (in the martensitic state) and can completely recover to zero strain (original shape) when heated to the austenitic state. The mechanism of the shape memory effect is a diffusionless phase transformation from a low temperature martensitic phase with a monoclinic B 19' crystal structure to a high temperature austenitic phase with a CsCl 82 structure. The low temperature martensitic phase is a twinned structure which will deform to a favorable (most easily deformed) twin orientation with an applied strain. Upon heating above the austenitic transition temperature the material will regain the highly ordered parent structure, and thus its original shape. During the shape recovery process (heating), the material produces both displacement and potentially large forces. The transformation temperature at which the material recovers its shape is strongly dependent on composition, therefore this material can be alloyed to demonstrate the shape memory effect over a range of desired temperatures. Nickel-titanium shape memory alloys also have high fatigue resistance and relatively high strength. The most logical substrate to grow thin films of nickel-titanium onto is silicon, since the mechanical system (NiTi thin film) of the micromachine can be directly incorporated into the microcircuitry (integrated into the silicon). Another reason to employ silicon as a substrate is that silicon micrornachining techniques (etching) are highly developed and are easily found throughout technical handbooks and technical papers. One method by which nickel-titanium thin films can be produced is ion-sputter deposition. Sputter deposition of thin films involves bombarding a target (solid material from which the thin film will be made) with energetic particles, ions, such that atoms of the target material are ejected. A substrate is placed beneath the target surface so that the ejected atoms are condensed onto the substrate surface to form a thin film. The sputtering conditions, such as ambient pressure and temperature, greatly influence the structure and properties of the thin film. The characteristics of thin films can also be purposely affected by inadiating the growing film with ions supplied from a secondary source. This type of deposition is termed ion beam assisted deposition (IBAD). Two important characteristics of thin film systems which are greatly influenced by sputtering conditions are adhesion between the film and substrate, and residual stress in the film. Obviously, strong adhesion between a film and substrate is important since it will result in a robust micromechanical system. Residual stress is important because high film stresses can adversely affect adhesion. Residual film stress in nickel-titanium films can also affect the transformation behavior of the material. The intent of this work has been to explore the effects of thin film deposition conditions on the adhesion and residual stress in a NiTi/Si system. The research lays some groundwork for the development of nickel- titanium micromechanical devices. CHAPTER 2. REVIEW OF LITERATURE This chapter explores the behavior of thin film residual stress and adhesion demonstrated in various sputtered deposited materials. The effects of deposition conditions on thin film structures is also reviewed. Relatively few papers have been published on NiTi thin films, therefore thin film characteristics in materials other than NiTi are reviewed to aid in the understanding of residual stress and adhesion. The final section of this chapter reviews the behavior of NiTi thin films as reported in the literature. 2.1 Residual Stress in Thin Films Residual stress is an important parameter to be considered in thin films, as it can affect the integrity and properties of the film. Adhesion of films to substrates is strongly affected by residual stress (Drory, et al 1988). High residual stress can cause film blistering and cracking. Removal of residual stress (for example, by annealing) may produce holes, whiskers, hillocks or unfavorable microstructure (d'Heurle, 1989). Residual stress gradients in films will cause the films to curl after they are released from substrates. It is also believed that stress may play a role in the crystallization of some amorphous films (Pompe, et a1, 1986). Shape memory alloy (SMA) films are greatly affected by residual stress. Shape memory alloys exhibit phase transformations which can be excited by either applied stress or changes in temperature. Trann'ormation temperatures are generally determined by the chemical composition, however residual stress present in the films can also alter the transformation temperatures. Residual stress may also limit the range within which the film can be strained and fully recover. If a SMA film is well adhered to a substrate and has a relatively high amount of residual stress and is annealed for extended lengths of time, precipitates with a particular crystallographic orientation may form which will affect subsquent shape memory in the film. This effect is termed ‘all-around’ shape memory effect. 2.1.1 Origins of Film Stress Residual stress may be divided into two catagories: intrinsic and extrinsic. Extrinsic stress is due to a thermal mismatch between the film and substrate, whereas intrinsic stress incorporates all stresses which do not directly result from to thermal nrismatch. Virtually all films are in a state of stress, and it is not uncommon to observe stresses in thin films which exceed the yield point of the bulk materials. When a substrate is coated at an elevated temperature with a material having a thermal expansion coefficient different from the substrate, an extrinsic stress will form upon cooling. It should be noted that extrinsic stresses also form in films which have been annealed after deposition on substrates with mismatched thermal expansion coefficients to the films if structural relaxation occurs at the annealing temperature. In most applications, the substrate thickness is much greater than the film thickness. Under this condition, plastic flow in the substrate can be neglected since it occurs at a fraction of the total thickness, and if it is assumed that substrate/film adhesion is perfect, a one-dimensional approximation of the extrinsic (thermal) stress can be determined by: ac = Edaf - as)AT/(l - Vf) .......... ( l) where E and vr are the elastic modulus and Poission's ratio of the film, or and as are the thermal coefficient of expansion (CTE) of the film and substrate. Material constants for the film can be assumed to be the same as bulk values, however the effective value for AT is not that simple to determine for a given film. All materials display some degree of viscoelastic behavior, therefore a portion of the strain occurring from CTE mismatch is recovered when the mterial is returned to lower temperatures. The 4 extent of this behavior is a function of diffusion coefficients and to some extent the cooling rate. Some researchers have found the accurate AT is about half of the difference between the elavated temperature and room temperatue (Angilello, et a1, 1981). At elevated temperatures stress relaxation occurs due to high atomic mobility, therefore the intial stage of cooling from an elevated temperature does not contribute to the final extrinsic film stress. As the temperature decreases, atomic mobility decreases and the films are unable to relax, thus final stresses develop. Intrinsic stresses are difficult to measure and it can sometimes be difficult to determine their origin. As explained below, there are various factors which affect intrinsic stress: the formation of secondary phases such as silicides, and the microstructure of the film, which is greatly influenced by the deposition conditions. Generalizations about the sign of the intrinsic stress can be made, for example stresses developed during the silicide reactions are compressive. Metallic films deposited via evaporation develop tensile stresses; metallic films deposited via sputter deposition are compressive, and ion bombardment during deposition produces compressive stresses. Again, these are generalizations, not all films do coincide with these guidelines. Since intrinsic stress arises due to the accumulation of crystallographic flaws which are built into the film during deposition, it is somewhat analogous to internal stresses formed in a bulk material during cold working. However, the density of defects formed within a film during deposition can be two orders of magnitude larger than that formed in the most severe cold working treatment of a bulk material. For this reason, intrinsic stresses may exceed the yield stress of bulk materials. This is because: a) provided adhesion is good, thin films cannot easily yield independently of their substrates, and b) the small dimensions of thin films exhibit properties quite different from those of bulk materials due to a vastily different surface area to material volume ratio (Hardwick, 1987). Intrinsic stresses can usually be controlled during deposition by altering the substrate temperature, deposition rate, pressure, or by supplying supplemental ion bombardment as will be discussed below. 2.1.2 Relationships Between Processing Parameters and Film Structure and Stress At relatively low pressure the arriving atoms have high kinetic energy and the resulting film has a dense microstructure, thus experiencing compressive stress. Low temperatures and elevated pressures causes scattering and loss of kinetic energy of the f sputtered atoms by gas atoms (see figure 1). The sputtered atoms approach the substrate A in oblique directions which promotes a more open structure with tensile stresses (R.W. Hoffman, 1976). Gas scattering can also cause a reduction in the average energy per particle delivered to the substrate, which can in turn decrease the density of the growing film. Researchers have noted that during magnetron sputtering at low temperatures there exists a transition pressure above which films are tensile, and below films are in a state of compression (Hoffman and Thornton, 1977A and 19778; Thornton and Hoffman, 1977). It is believed that the films are more compressive at lower pressures due to an atomic peening effect (discussed below) by the energetic particles. At lower pressures these particles travel to the substrate without collision when the mean free path becomes sufficiently long. Some researchers had thought the compressive nature of the films was due in part to entrapped argon gas, however Thornton and coworkers (1979) proved the mere presence of Ar in the films does not cause compressive stresses. The atomic peening effect was first proposed by d'Heurle (1970). Bombardment of the film by neutral atoms (gas ions neutralized at the cathode and backseattered with energy close to that of accelerated ions), ions from an assist beam, and incoming sputtered atoms produce the peening effect. Peening causes the sputtered atoms to be incorporated into the growing film with a tighter structure than would be obtained otherwise. With sufficient 6 Large Mean Free Path of a) low pressure Mean Free Path of Sputtered Atom is Decreased by Gas Scattering Film , , $3 ,- 21,. b) high pressure Figure 1. Schematic displaying the effect of working gas pressure on the mean free path of incoming sputtered atoms (adapted from Thornton and Hoffman, 1 989). energy, atoms may be forced into spaces too small to accommodate them under thermal equilibrium conditions thus producing compressive intrinsic stresses. The relation between film microstructure and intrinsic tensile stress has been investigated. Muller (1987) and Itoh, et a1, (1991) proposed similar models formulated with two-dimensional molecular dynamics. The intrinsic stress was calculated as a function of incident kinetic energy of the adatoms (interatomic potential between nearest- neighbor film atoms was also considered in the calculation) . It was found that decreasing the incident kinetic energy by increasing working gas pressure, produced a decrease in E tensile stress which was caused by a microstructural change from a densely packed [1 network with atomic scale voids to a microcolumnar structure. The microcolumnar structure originated from a self-shadowing effect. According to this model the small voids in the densely packed network produced tension due to attractive forces between atoms across the void surface. ltol's model investigated intereolumnar interactions where all stress is assumed to be imposed at the column boundaries. It was found that as the microstructure becomes less dense and intercolumnar widths increase, the attractive forces between the columns decrease (same mechanism as interatornic attractive forces between nearest-neighbors) and tensile stress decreases. Vink, et al (1991) found that tensile stresses decreased with film thickness for high pressure (2.0 Pa) sputtered films, whereas films sputtered at lower pressures (1.0 Pa) did not display any variation in stress with film thickness. The stress gradients at higher pressures were attributed to microstructural changes throughout the thickness of the film. Near the substrate-film interface region the columns were separated by rather widely voided boundaries, therefore thinner films tend to have a larger percentage of widely separated boundaries (with nearly zero tensile stress) than thicker films. At low pressures no loss of directionality in the sputtered flux of atoms occurs by scattering, and the orientation of the substrate relative to the direction of the incoming 8 sputtered atoms has an effect on the film stress. As discussed above, sputtered metal films deposited at normal incidence often exhibit compressive stresses when the working pressure is sufficiently low (Hoffman and Thomton,1977A and 19778). The effects of oblique deposition have been investigated by Hoffman and Thornton (1979) and Castellano, et al (1978). Castellano and coworkers found inclination of the substrate up to 45° produced an increase in compressive stress in films which are in a state of compression when deposited at normal incidence, and films deposited at normal angles with tensile stresses displayed a transformation to compressive stresses with oblique deposition. They attributed this behavior to oxygen uptake in film structures with increased porosity. Hoffman and Thornton investigated the stress behavior of films which were compressive at normal incidence. They reported three regimes of stress behavior with incidence angle: a range of near-normal incidence angles where the compressive stress was essentially unaltered, followed by a second range of angles where the stress changed rapidly to less compressive or even tensile values, and finally at angles near grazing incidence the stresses vanish. The range of angles within each regime varied markedly with material. A correlation between entrapped gas (argon or oxygen) and compressive stress was not found in their investigation. Changes in stress were attributed to microstructure and degree of the peening effect. As the incidence angle increased the films became more porous due to self-shadowing during growth. At high incidence the films were so porous that they were incapable of supporting any stress, this accounted for the diminishing stresses in the third regime. The stress behavior in the first and second regime was explained by a decrease in the peening effect. According to the peening model, compressive stress is induced by a hail of energetic particles which strike against the growing film causing the atoms to become more tightly packed in the plane of the film. Tilting the substrate can thus nullify the peening effect (if the primary bombarding particles are traveling along the direction of the sputtering flux) and produce less compressive film stresses. 9 The sputter deposition rate also affectd the residual stress within the films. Thornton and Hoffman (1971A) reported that high deposition rates promoted larger compressive stresses. At high deposition rates the deposited atoms were immediately buried by other incoming atoms, and they were unable relax, thus compressive stress ensued. This effect was only seen in extreme differences between deposition rates. 2 .1 .3 Ion Beam Assisted Deposition and Related Stress Effects There exists a few studies on film stress effects in ion beam assisted deposition experiments (Hoffman and Gaettner, 1980; Cuomo et al , 1982; Ensinger and Wolf, 1989; Yee et a1, 1985; Wojciechowski, 1988). The general trend of film stress behavior with ion- to-atom arrival ratio (I/A ratio) in these investigations was that film stress became less tensile and more compressive with increasing l/A ratio. The VA ratio ranges in these studies were usually within 0 (no ion assistance during deposition) to 0.6. Hoffman and Gaettner found that Cr films deposited onto glass wafers and bombarded with Ar+ ions displayed tensile to compressive behavior with an l/A ratio from 0 to 0.01 . At an l/A ratio 0.01 the stress started to become less compressive up to an l/A ratio of 0.1, the highest I/A ratio investigated. None of the other investigations mentioned above showed a stress transition such as that found by Hoffman and Gaettner. Cuomo, et al observed a decrease in tensile stress to zero, or even compressive stress, with increasing ion bombardment to two factors: first, a removal of oxygen impurities which produced tensile stresses and second, an annealing effect or compaction of the structure by ions striking the growing film surface. Tang and Wehner ( 1972) studied the effects of ion bombardment on nucleation and agglomeration in sputter—deposited films. It was found that low energy ion bombardment aides in spreading atoms from islands over the substrate surface (islands form in film/substrate systems in which the interaction potentials between the substrate and adatoms is weaker than the potential between adatoms themselves). There exists a threshold energy 1 O of the bombarding ions above which the assisting ions resputter the adatoms rather than aid in surface migration. Lower ion energy (under normal incidence) favors the lateral movement of adatoms, whereas higher energy ions penetrate underneath the surface supplying the adatoms with enough energy to be sputtered. Wojciechowski found three regions of film growth which exhibited unique stress behavior with ion beam assisted deposition (IBAD). Region 1, an interface region approximately Snm thick which was dominated by interactions between the substrate incoming atoms. Region 2, an intermediate layer from 5 nm to approximately 40 nm thick where stress was compressive and very sensitive to HA ratio. Followed by region 3, theremaining film thickness where intrinsic stress changes were insignificant. 2.1.4 Methods of Measuring Film Stress There are many methods of measuring stress in thin films, the methods can be grouped into either in situ (during deposition) or ex situ (after deposition) measurements. Ex situ methods tend to be simpler to perform and are therefore more widely employed. In situ observations of stress are comparatively rare due to the difficulty of isolating the stress transducer from the energetic sputtering environment. Ex situ approaches do have the disadvantage of sometimes being cumbersome and less sensitive to finer details of stress behavior. Most methods of measuring stress in thin films utilize a classical curvature-stress relationship. Stoney (1909) formulated an equation which relates stress to the change in curvature of a substrate due to stress imposed by an adhered film (as in figure 2). The equation only requires knowledge of the substrate material properties and the film/substrate geometry as written below: .1- R .......... (2) where BS is the elastic modulus of the substrate, 1 l Substrate Resultant substrate/film curvature from a film with residual COMPRESSIVE stress. Sbstraeut Resultant substrate/film curvature from a film with residual TENSILE stress. Substrate without film Figure 2. Substrate curvature due to residual film stress. 12 vs is Poisson's ratio of the substrate, ts is the substrate thickness, tr is the film thickness. and R is the change in radius of curvature of the substrate. There are some modifications of this formula, such as using the plate modulus E3/(1- vsz), rather than the biaxial modulus Es/( l-vs) as done by Berry and Pritchet (1990). Assuming a Poissons’s ratio of 0.3, stress values calculated with the plate modulus formula are approximately 30% larger than those calculated from the original Stoney equation. Von Preissig (1989) has modified the Stoney equation to account for effects of gravity, substrates shape, film nonuniformity and substrate crystallinity on substrate curvature. Most of the in situ stress probes measure the distortion of substrate curvature, either mechanically with a displacement transducer (Hoffman and Kukla, 1985), or optically (Doljack and Hoffman, 1972). Some ex situ curvature measurements can be performed with shadow Moire interferometry (Park and Danyluk, 1990) or with profilometric techniques. The radius of curvature, R, can be determined from profilometric measurements with the following relation: 1 2w __ = _2_ R X .......... (3) where w is the displacement of the center of the substrate curvature and x is the length of the profilometric trace as seen in Figure 3. The stress in a film can also be measured ex situ by determining interplanar spacings of the film in a direction normal to the substrate and comparing this with the interplanar spacing of same material in a totally relaxed (stress-free) condition. The strain in the film can be calculated from the interplanar spacing, and by using a form of Hooke's law the stress can be determined. This method requires that one know the exact value for the relaxed interplanar spacing, which is not too difficult to determine for pure elements. 13 profile of flansubstrate curvature due to compressive film stress c—i—v w: change in substrate curvature (assuming substrate was initially flat) x: length of trace Figure 3. Stoney equation parameters obtained from the profile of the film curvature. 14 However, it is not always easy when working with films to know precisely what impurities might be present, and what effects they might have on the lattice parameter. This method cannot be used for amorphous films. Another disadvantage of this method is that the elastic modulus and Poisson's ratio of the film must be known. Schweitz (1991) has devise a micormechanical ex situ method which can be used to characterize a complete elasto-plastic relation for a thin film adhered to a substrate. The advantage of this approach is that no previous information on elastic constants of plasticity parameters for the film or substrate is required. The procedure also works equally well in the elastic or plastic film strain range. The method involves measuring the deflection of a stressed film/substrate cantilevered strip and measuring a load which must be applied to produce the same deflection in a cantilevered strip of the substrate material (no film). Using relatively simple beam mechanics, the stress-strain characteristics of the film can be determined from geometric parameters and the experimental load-deflection data. 2.2 Adhesion of Thin Films Strong, stable adhesion of thin films to substrates is important to the function of microelectronic and micromechanical systems. Poor adhesion can result in contact noise, reduction in thermal and electrical conductivity. Adhesion is a macroscopic property that depends on local stresses, bonding across the interfacial region, and the adhesive failure mode (the latter depending on type of stress which the interfacial region is subjected to, such as mechanical, chemical or electrical). Loss of adhesion between dissimilar materials involves deformation and fracture processes, therefore the fracture toughness of the materials involved is an important parameter to consider (Mattox, 1978). The term " good adhesion" is defined in the context of the required performance of the film, for example, the ability to withstand peeling, shear, normal pulling or scratching. Good adhesion is achieved with strong substrate-to-adatom bonding, uniform film coverage across the interfacial region, relatively low local stress, and absence of an easy 15 deformation or fracture path (Mattox, 1976). These factors depend strongly on the interactions at the interface between the film material and the substrate. Factors affecting adhesion are surface roughness, interfacial chemistry, contaminant layers, film stress and coefficient of thermal expansion mismatch between the film and substrate. 2 .2.1 F ilm/Substrate Interface The durability of the interface between the film and substrate has a strong effect on the adhesion. The integrity of the interface depends on freedom from contamination, chemical interactions, kinetic energy available during interface formation, and nucleation behavior of the depositing atoms. Contaminants or chemical interactions may assist or inhibit the strength of the interface. The strength of the interface depends on the degree of electronic bonding (ionic, covalent or metallic). A contaminant species can be regarded as a new contributor to the range of possible interface bonding configurations. It can be feasible for a reactive contaminant to participate in new electronically bonded configurations and assist adhesion. If an intermediate (contaminant) forms stronger bonds between itself and the film and substrate than would be obtained between the film and substrate alone, the interface will obviously be strengthened (Elliot, 1975; Sundahl, 1971). Conversely, contaminants may degrade interface bonding (Baglin, 1985). The quality of the interface depends on the kinetic energy available during its formation and also the type of nucleation which occurs. Atoms impinging on a surface lose energy to the surface and condense by forming stable nuclei. Adatoms (sputtered atoms which are adsorbed into the growing film) have a degree of mobility which is dependent upon their kinetic energy and the strength of their interaction with the substrate. A high degree of mobility will give rise to a high density of nuclei (Mattox, 1973). A high density of nuclei will lead to formation of a continuous film across the substrate, whereas a low 16 density of nuclei will produce a less effective interfacial contact area containing voids (Lloyd and Nakahara. 1977). It was found that higher nucleation density aids in the formation of relatively large islands over the substrate. As the islands coalesce, the "nose" of one island will contact another and consequently shield the substrate from depositing atoms, hence voids are formed. Therefore, a high island density results in a high void density. The researchers also observed that island density at coalescence varies with deposition rate; larger islands formed at lower deposition rates. An increase in the deposition rate increases the nucleation rate and the number of nuclei which form. 2.2.2 Ion Bombardment E fleets on Adhesion The assistance of an ion beam before or during deposition can greatly improve thin film adhesion. Prior to deposition, a substrate surface can be cleaned in situ by etching contaminants away with an ion beam. Chemical methods of cleaning substrates prior to deposition are usually not capable of preventing monolayers of water or carbon compounds from forming (Wie et al, 1988). Also, a chemically cleaned surface rapidly becomes covered with contaminants when exposed to atmosphere. Comfort, et a1 (1987) investigated silicon surface cleaning (removal of the oxide layer) by low-dose argon-ion bombardment. They found low-energy ion bombardment (hundreds of eV) was more efficient in removing surface oxides and also resulted in less substrate damage. Most ion lattice interactions induced by a high energy (1000 eV) ion bombardment of silicon occur relatively deep in the substrate and do not contribute directly to the sputtering process. It is therefore favorable to limit all collision events to the surface of the substrate where sputtering originates, this can be done with low energy ion beams. Carter and Armour (1981) have developed formulations to determine the limit of ion flux energy necessary for removal of surface contaminants. Excessive ion doses can be deleterious by causing argon accumulation, damage accumulation and severe roughening of the substrate 17 surface. It was also determined by Comfort that in order for the oxide-free silicon surface to be thermodynamically stable the partial pressures of oxygen and water should be held below approximately 10‘8 Torr. Exposing a substrate to low energy ion bombardment prior to film deposition can slightly texture the surface and may promote improved film adhesion by increasing the surface area of the interface and provide an interlocking mechanism. A rough interfacial region may also improve adhesion by "pinning” growing interfacial cracks. If the surface is too rough, detrimental effects may occur such as severe film stress, or formation of voids at the interface due to a shadowing effect. Baglin (1989) determined that the ion beam exposure should be limited to relatively short time periods (30 seconds at 500 eV on Cu substrate) for optimum adhesion benefits. Longer exposure times produced dendritic structures which weakened the substrate, thus providing a failure mechanism of substrate breakage rather than breaking the atomic-scale interfacial bonds. Energetic particle bombardment may also promote chemical reaction with the introduction of surface defects and by producing change in the surface chemistry. Nucleation sites may be introduced by lattice defect formation, adsorption of activated species, implantation of impinging energetic species, and generation of electrically charged sites. Energetic particle bombardment during film deposition can strongly modify the structural and chemical properties of the growing film, which can in turn effect adhesion. There is a collection of studies, published in Vacuum, pertaining to irradiation enhanced adhesion of films bombarded with secondary electrons (Colligon and Kheyrandish, 1989), photons ((Kellock et al, 1985) and, most commonly, ions (Ahmed and Colligon, 1988; Li et al, 1990; Rossnagel and Cuomo, 1988; Wie et a1, 1988). As mentioned in Section 1 of this chapter, energetic particle bombardment can effect the microstructure of the film: generally films become more dense and the stress and morphology change. 18 Ion beam mixing generally improves adhesion, however Mattox and Cuthrell (1988) found that if the interfacial material has a high density of voids or is more brittle than the film or substrate material, adhesion will be degraded. Ahmed and Colligon found that substrate and film materials which do not form solid solutions (Au on silca), can be mixed to produce mechanical diffusion and thus improve adhesion. It was found that adhesion was enhanced to levels indicative of chemical or metallic bonds at the interface. This was attributed to electronic collisions at the film-substrate interface where bond breaking and chemical reactions with a contaminant species (oxygen) proceeded, and also to atomic mixing at the interface where nuclear collisions produced a cascade of displaced adatoms inside the substrate. Wie et al also found that an oxide layer at the interface had a major effect on the adhesion enhancement and interface chemistry induced by ionizing radiation in a Au-GaAs system. Li et a1 utilized a Monte Carlo program simulating ion assisted deposition of Ti onto an Fe substrate and radiated with N+ ions, to shed light on processing parameters which strongly affect interface mixing. They found energy of the incident ions and the ion-to— atom arrival ratio (1/A ratio) are two of the most important parameters influencing interface mixing and adhesion. At a low energy range (0—5000 eV) thickness of the interface layer increases rapidly with ion energy, as ion energy increases beyond this range the interface layer thickness increases slowly and eventually the thickness reaches plateau with ion energies above 30 MeV. The effect of energetic particle bombardment on interface mixing during deposition takes place at the initial deposition stage for low energy ions because the displacement spike (collision cascade) of the ions is so small that the ions cannot reach the interface after a short deposition time. As the ion energy increases, the interface mixing process is prolonged, so that the interface mixing thickness grows with ion energy. As ion energy increases, two important factors play a role in the mixing process. First, most of the mixing energy is dissipated in collisions among like atoms, even though the cascade 19 volume induced by high energy particles is larger, so only a small part of the energy is contributed to interface mixing. Second, although an increase in ion energy can prolong the interface mixing process, collision mixing of unlike atoms at regions far from the interface is limited by the supply of unlike (substrate) atoms, which prevents effective mixing. In order to study the influence of ion-to—atom arrival ratio on the atomic redistribution layer thickness, the deposition rate was held constant and the ion energy dose rate was varied in the computations performed by Li. Interface mixing was found to be approximately proportional to the square root of the I/A ratio. The enhanced mixing behavior with an increase of VA ratio was attributed to raising the density of displacement spikes due to an increased number of incident ions, thereby enhancing the mixing efficiency during the earliest stages of film deposition. Overlap of displacement spikes will occur if the [IA ratio is large enough. As the [/A ratio increases so does the overlap of spikes. Beyond large value of [IA ratios (above 1.5) the atomic redistribution layer does not increase much due to the overlap of displacement spikes which reduce the effective number of atoms mixed by ion bombardment. Colligon and Kheyrandish examined the effect of secondary electrons on interface bonding. It was believed that interface mixing due to secondary electrons could not account for the entire effect on adhesion improvement since the Van der Waals forces between atoms is very small. Adhesion improvement was attributed to newly formed chemical bonds (much stronger than Van der Waals forces) resulting from electronic energy input and interfacial reactions involving impurities. 2.2.3 Eflects of Residual Film Stress on Adhesion When a crack near the film/substrate interface advances, energy is needed for the formation of new crack surfaces, and for defamation near the crack tip. Energy for crack propagation can be supplied by residual stress stored in the film/substrate system. The sign 20 of the residual film stress determines the decohesion mechanism. When the film stress is compressive, decohesion involves buckling above an initial interface separation, followed by delamination and eventual spallin g. On the other hand, decohesion under tensile film stress can initiate preferentially at specimen edges and propagate inward (Evans and Hutchinson, 1984). If film adhesion is high, fracture may occur in the substrate or film and not at the interface. Adhesion can be maintained in highly stressed films if the interface region is strong and the film and substrate have high fracture toughness. As discussed in Section 2.1, stress gradients at component edges increase with film thickness, therefore edge initiated decohesion is found to become more probable with thicker films. Drory, et al (1988) and Hunt and Gale (1972) have developed formulations which can be used to determine the maximum allowable film thickness to maintain adhesion at a certain tensile stress level, and Evans and Hutchinson (1984) have done the same with films in compression. The input parameters for both formulations consist of substrate and film material constants and the component geometry. 2.2.4 Methods of Measuring Adhesion Quantitatively measuring adhesion is a rather difficult task. Several methods of measuring adhesion have been developed, unfortunately the quantitative results of each test cannot be compared because different elements of adhesion are measured. Most of the methods work best when used on the basis of comparitive rankings. Several good reviews have been written on the measurement of thin film adhesion (Hull, et al, 1987; Chapman, 1974; and Sachs,et al, 1989). The commonly used methods to be discussed below are the peel test (or Scotch tape test), the pull test (or direct pull-off or direct shear), the scratch test and the indentation test. The peel test involves pressing adhesive tape onto the film and then measuring the force required to peel the film from the substrate at a 90° angle. The test can also be used qualitatively to allow adhesion to be classified according to whether the film is wholly 2 1 removed, partially removed or not removed at all. In order to obtain a quantitative measurement of adhesion, the film must be completely removed. Therefore this method is limited to systems with adhesion worse than that between the tape and film. The advantages are the test's simplicity, minimal cost, suitability for hard and soft films and independence of the adhesive quality of the tape (except that this will affect the usable test range). The pull test measures adhesion by applying a normal force to the interface via an attached rod or column to cause the interface to fail. Reproducibility is fairly poor due to requirement that loading must be perfectly normal to the interface. As with the peel test, the pull test is limited to systems with weaker adhesion than that between the film and adhesive used to attach the rod. The adhesive can pose a problem if it induces stress in the film by shrinking when it dries. The scratch test is the most widely utilized adhesion test. A smooth, finely pointed stylus is drawn across the film under increasing vertical loads. Adhesion is measured in terms of the critical load required to strip the film from the substrate leaving a clear channel revealing the substrate (usually this is done with a SEM). This method offers simplicity and a high degree of reproducibility. However, considerable controversy surrounds both the use and interpretation of the data. The scratching process is very complex and the failure mechanism varies from system to system. Butler, et a1 (1970) have shown that optical transparency is not a suitable measure of film deadherence since the film could become thinned and translucent under the stylus load without being detached, or could be detached from the substrate and not be transparent. The indentation test involves indenting the film with a diamond stylus (or hardened steel conical indentor) until it pierces the film. The edge of the indentation conforms to the angle of the stylus and lifts the film up off the substrate, see Figure 4. As the film is lifted from the substrate a interfacial lateral crack forms, and buckling of the film tends to deflect the interfacial crack into the film and toward the surface. The resistance to propagation of 22 the crack along the interface is used as a measure of adhesion. The analysis of the results is based on indentation fracture mechanics. The advantages of this method are simplicity of interpreting results. good reproducibility and ability to quantitatively measure adhesion. If the film is residually stressed in compression. the crack length should increase. Figure 5 depicts the indentation mechanism of lateral crack formation for residually stressed films. Marshall and Evans ( 1984), Rossington et al ( 1984) and Evans and Hutchinson (1984) have developed formulations which relate interfacial strength to crack length, indentation load, film thickness, film properites and initial film stress. To cancel out the effects of film stress the crack lengths, a, were normalized with respect to film stresses, 0, by the relation: a is proportional to 01/2. Chiang et al (1981) also evaluated the indentaion mechanics involved in this method of adhesion testing. Their formulations differ from the aforementioned authors' work in that film stress was not factored into the interface strength equation. The sensitivity of indentation testing for inferior adhesion was found to be rather poor. Poorly adhered films demonstrated large amounts of scatter in the lateral crack (the lateral crack was not circular), and in some instances indentation caused sections of the film to be removed. The variability in the extent of the lateral fracture around the indentation indicates that adhesion in poor quality films exhibits substantial variation, therefore very poor adhesion cannot be well quantified with the indentation method. 2.3 NiTi Thin Films Sputter deposited NiTi thin films are amorphous when grown at room temperature (Kim et al, 1986 and Busch et al, 1989). Busch and coworkers found that amorphous NiTi thin films do not display shape memory characteristics. Through differential scanning calorimetry (DSC) the crystallization temperature was found to be near 500°C. NiTi films have been crystallized by a post—deposition anneal or by heating during deposition. The transformation temperatures of the film were found to differ from those of the 23 Film : W b) —-— E§ .___ C) d) Figure 4. Schematic of the indentation adhesion test: a) film before indentation, b) film "elongates" due to strain from indentation stylus (section of film drawn outside of actual location for demonstration of strain due to indentation), c) actual length of indented film with arrows indicating stress due to indentation strain, and d) appearance of film after indentation showing lateral cracks between the film and substrate (adapted from Marshall and Evans, 1985), 24 I I —‘ *— —> W b) r i ------ I Y I r I I l | I l ' I r I W ——"i m—___ ——~ w __.- —>- 1:: d) WWW' Figure 5. Schematic of the reaction of a residually stressed film to indentation: a) residually stressed film before indentation, b) strain on film segment due to residual stress, c) additional strain on film segment due to indentation, d) actual length of residually stressed and indented film with arrows indicating stress due to indentation strain, and e) appearance of film after indentation. showing lateral cracks between the film and substrate (adapted from Marshall and Evans, 1985). 25 target material by up to 90°C (Busch et al, 1989). The discrepancy was attributed to oxygen pickup during deposition. Any oxygen deposited with the film reacts’ with titanium and effectively produces a nickel composition relative to the target material and in turn produce a lower transformation temperature. One percent of oxygen contamination in the film can lower the transition temperature by 100°C. The effects of ion bombardment during deposition of NiTi film was investigated by Grummon and Chang (1992). As l/A ratios increased, the titanium content of the film decreased. The drop in titanium content was attributed to resputtering which is dependent on l/A ratio and assist beam incidence angle. TiN precipitates were present in the films, their formation was due to nitrogen contamination during deposition. Busch and coworkers (1991) found that transformation temperatures are also affected by time and temperature of annealing. Films annealed at 540°C for 30 minutes showed a decrease in transformation temperature and an increase in grain boundary precipitates. The precipitates were determined to be TigNi and Ti 1 1Ni14, and their formation produced an effective composition change. Thus as annealing time and temperature are increased, more precipitates are formed and transformation temperatures differ. If the amount of precipitate formation is extensive, the shape memory effect can be lost. Moberlyol (1991) investigated the crystallization behavior of NiTi thin films in real time with a TEM. Crystallization progressed more rapidly in thicker portions of the film indicating that the rate is governed by the density of nucleation sites rather that by surface energy phenomenon. The grain size of the film was found to be approximately a micrometer in width, much smaller than that of bulk material solidified from melt. The stress-strain response of free standing crystalline films was studied by Busch et al (1990). The martensitic NiTi thin films yielded at 85 MPa and a 5% strain induced by 180 MPa stress was totally recovered (the upper recoverable strain limit was not determined). These properties are very similar to those found in bulk NiTi. Busch and Johnson (1990) 26 also experimentally determined that a 10 micron thick and l millimeter wide film is capable of exerting a 140 gram force upon heating. Nickel-titanium actuators have some advantages over electrostatic, piezoelectric and bimetallic actuators. NiTi actuators demonstrate an extremely high energy output per unit volume relative to the other actuators (Johnson, 1991), and the voltage required for actuation is much lower. Low voltage requirement of actuators is important in the design of hybrid rnicroelectrical systems. It is perceived that a limitation of NiTi actuators may be the frequency at which it can by cycled due to the rate of heat transfer. However, Johnson (1991) found that a NiTi thin film stretched over an orifice to form a valve could be cycled at twenty Hertz (the film was also cycled over two million times). NiTi thin films have also been produced by ion mixing alternating layers of crystalline nickel and titanium (Clemens, 1987; Rai and Bhattacharya, 1987; and Zheng and Dodd, 1990). The ion mixed alloyed films were amorphous and were crystallized at 460°C. In some cases it was desired that the films be amorphous to exploit their corrosion resistance, high fracture strength, and improved wear resistance. Films which were crystallized were not evaluated for shape memory characteristics. Although devices have been fabricated from NIT i thin films, there does not exsist any data pertaining to residual stress or adhesion. Such data is pertanant for the development of NiTi thin film applications. Those topics have been explored in this work. 27 28 CHAPTER 3. EXPERIMENTAL PROCEDURES 3.] Materials The following materials were used for the deposition of NiTi films: A) a sputtering target: 50Ni-501'i wrapped with Ti wire (net target composition: 47Ni- 53Ti), B) substrates: one-inch and two-inch diameter (100) Si wafers, C) operating gas for ion guns: 99.9999% Ar, and D) substrate cleaning solutions: methyl alcohol, acetone , ethyl alcohol and HF:H20 (3:40). 3.2 Sputter Deposition Equipment The deposition chamber was a 14—inch stainless-steel bell-jar. Inside the chamber were the following devices: A) an lon Tech MPS-3m0FC 3-cm ion gun with external tungsten filament used to sputter the NiTi target. B) an Anatech lS-3m filamentless 5-cm ion gun used to irradiate the substrate during ion beam assisted deposition (IBAD). C) Faraday cups connected to a shutter which could be rotated into the path of the 5-crn ion beam so that the ion current could be measured, D) a thermocouple located either directly atop the film surface or on the bottom surface of a substrate, E) an NiTi target shielded by titanium foil to prevent sputtering of the chamber wall material, and F) a substrate holder with a shutter. The vacuum chamber was pumped down with a roughing pump, a turbomolecular pump and a cryopump. The cryopump could be closed off from the chamber with an 8— inch gate valve so that it was engaged in the final pumping stages only. This valve also allowed the cryopump to remain at vacuum while the chamber was being opened to atmosphere. 3.3 Sputter Deposition of NiTi The nickel-titanium films on (100) silicon substrates evaluated in this thesis were produced in four separate sputter deposition runs. Run 1 involved growing an ion sputtered film on a two inch diameter Si substrate. Run 2 involved growing an ion beam assisted deposition (lBAD) film on a two inch diameter Si substrate. The chamber configuration for Runs 1 and 2 were identical although the sputtering conditions differed. Runs 3 and 4 utilized the same chamber configuration (this configuration was different from Runs 1 and 2), however Run 3 involved producing lBAD (ion beam assisted deposition) films on several one inch Si substrates, whereas Run 4 involved ion sputtering a film onto a couple of one inch Si substrates. lon sputter deposition for Runs 1 and 2, and Runs 3 and 4 was carried out in a vacuum chamber with a layouts depicted in Figures 6 and 7. Both chamber configurations employed the same equipment, however the geometrical set-up of the equipment differe 3.3.1 Sputtering Conditions The 3-cm ion gun operating conditions were kept constant throughout all sputter deposition runs. Argon gas was employed at an operating pressure (corresponds with the gas flow through the gun) of approximately 2x10‘5 torr. The Ari“ ion beam had a energy of 800 eV and a current of approximately 60 mA. The 5-cm gun was used to both sputter-clean the substrates and to assist the deposition of the flux of Ni and Ti atoms arriving at the substrate. When employed to sputter-clean the Si wafers, the ion energy was set at 50 V, and used for only 30 seconds, to prevent extensive radiation damage to the wafer. 29 Copper Heat Sink 25cm Rotating Substrate Holder Ti Apertures Figure 6. Schematic of sputter deposition chamber for Runs 1 and 2. 30 .v 93 m 95% .8“ 89 39:85 esgnonou Bean go as: 30m N 35E :3 gawk SE a. + a2 «853% v8 825 no a L . mau new <9. oi 8:28. 8:5er a 8o 5 m cooom CNN ‘70— x N v.8 >o Wm n.1\o~I m 0.39”. ocuflmm Logo—5% ..P m Ucm (Um Emwm :01 x_ ocoueua ecu .m. A o: mecooom om _ocmao.aow_ coEmoaoo . o m: Yo, x m .8 >o 0m .6552: o 83mm 3532 .3255 ..N N Jam 83m :0. mac 0:308 new A . $588 On Ecuaoaoe 5 GOP 0mm Top x N .8 >0 on .6552: m 2:9“. neat—am .8256 ..N F .(Um CO_ 050 38% 829.20 fieSEEV C8: ".0 c9330 :23. :0: .86 .6 . $9200 .6030 32593 SE mEE. oazmmmi 9:520 95520 a» . . 52:26 *0 e .F :o_u_moaco 203:2. :0. museums—em .¢ 3:95 F .83. ..8 2336.50 5:388 .— 298. 36 microsc0pe. lateral crack lengths were measured from the micrographs and these values were then used to quantify adhesion. 3.5 Residual Stress Determination Residual stress in the films was estimated using the Stoney method (Stoney, 1909) by comparing the curvature of the wafer before and after deposition. A full description of this method is reviewed in the Literature Review Section 2.1.5. 3.5 .1 Profilometry Surface curvature was measured with a Dektak 11A profilometer. Two 20 mm traces were taken across each side of the wafer before and after film deposition so an average curvature change could be calculated. The speed of the stylus was set at medium and the auto leveling tool was employed. Profilometry was also employed to determine film thickness, which was a variable needed for residual stress analysis. Profiles were taken across steps in the film which were created by masking the substrate with small pieces of glass microscope slide or silicon wafer. 3.6 Annealing The NiTi films on Si substrates were annealed in a MTS vacuum furnace. This vacuum furnace had tungsten heating elements fashioned in a cylindrical manner and the wafers were placed on a stage located in the middle of the heating elements. The wafers were thermally isolated from the stage with thin pieces of alumina. For films produced in Runs 3 and 4, clean glass slides were placed atop the film side of the wafers to aid in prohibiting the formation of oxides, especially TiOz. A thermocouple was placed adjacent to the wafer in order to monitor and control the temperature. 37 The annealing was carried out at 10'6 torr. The wafers were heated to 600°C in approximately 15 minutes and were held at this temperature for 8 minutes. The heating elements were then shut off and the wafers were cooled from 600°C to 400°C in approximately 15 minutes, and from 400°C to 100°C in approximately one hour, and finally from 100°C to room temperature in approximately 3 hours. The crystallinity of the films was checked via x-ray diffraction using a Scintag diffractometer The films were scanned between two-theta angles of 30° and 80° to look for [111] peaks. 38 39 CHAPTER 4. RESULTS AND DISCUSSION 4.] Deposition Temperature Temperature was estimated to be approximately 15°C higher than indicated by the thermocouple. This was determined from observing the temperature reading after the ion guns were fired (under vacuum). and comparing this value to the temperature approximately 30 seconds after the chamber was brought to near atmospheric pressure With nitrogen. The temperature reading generally increased by 15° C immediately after a large ambient pressure increase due to better thermal conduction between the thermocouple and the object to which it was connected. Substrate temperatures during deposition runs performed for this work were relatively low. Direct heating was not used for these runs. however indirect heating from the ion guns did occur. The substrate temperature in the longest run (Run 1. 360 minutes) only increased from 25°C to approximately 80°C. Whereas, the shortest run (Run 2. 115 minutes) had a maximum temperature of approximately 65°C. 4.2 Film Thickness The film thickness for Runs 1 and 2 were estimated from masked substrates adjacent to the two inch circular substrate. The thickness of the film from Run 1 was 2600 A and the thickness of the film produced in Run 2 was 800 A. The film thicknesses for Runs 3 and 4 were measured at various distances from the center of the NiTi sputtering target (this is the intersection between the substrate holder plate and the central axis of the sputtered flux, labeled B on Figure 8). Since the film on the one inch circular substrates had to be continuous. masking the film to produce a thickness step could not be done, hence no thickness measurements were taken directly from the one inch wafers. To determine the nominal thickness of the films on these substrates, the relationship between film thickness. determined from masked surveillance specimens adjacent to the one inch circular wafers, and distance of the substrate from a fixed point. was developed. Measured film thickness values were plotted against their location from the center of the sputtered flux and a best fit curve equation was carried out to allow interpolation of the film thickness on the object wafers (equation was developed by Cricketgraph software). Thickness curves for both ion sputtered films from Run 4, and lBAD films from Run 3, along with their corresponding curve fits. are shown in Figure 10. The nominal film thickness on each one-inch wafer was calculated by substituting their distance from the center of the sputter flux to the center of the wafer into the developed thickness curve equations. The film thickness across the wafers is not uniform, however due to symmetry the average film thickness was approximated at the center point of the wafer. 4.3 lon-to-Atom Arrival Ratio Determination lon-to-atom arrival ratios (l/A—ratios) from films prepared by ion beam assisted deposition were determined. The l/A-ratio for ion sputtered films was taken to be zero. The l/A-ratio for lBAD films were calculated by dividing the ion arrival rate with the atom arrival rate. The atom anival rate was calculated as follows: atom arrival rate (atom/sec-A) = deposition rate {A/sec) * film density (£043) ......... (3) average atomic mass (g/atom) where the deposition rate was the rate at which the film thickness increased without any influence from an assisting ion beam. Therefore, it was not the net deposition rate during an lBAD run, but rather. it was the net deposition rate obtained from an ion sputtered run having identical primary sputtering conditions (parameters set on the 3 cm ion gun which was creating the Ni + Ti sputtered flux) as the lBAD run. The deposition rate used to 40 y = 2757.5 - 32.755x - 155.73x"2 + 26.568x43 - 1.3418x"4 300“ 3 .9. '3 3 .2 :3 H . fl . T . , . O 2 4 6 8 Distance from center of NiTi sputtered llux (cm) a) Film thickness curve for ion sputtered Run 4 y = 3844.0 + 239.09): - 309.4492 + 49.4l6x"3 - 2.7570x44 3 g ‘6 3 3 ii [-1 r I j T ' I 0 2 4 6 8 Distance from center of NiTi sputtered flux (cm) b) Film thickness curve for lBAD Run 3 Figure 10. Film thickness versus distance for a) Run 4 ion sputtered and b) Run 3 ion beam assisted deposition films. The equation for the best fit curve is located above each plot. 41 calculate atom anival rates at a certain point on an lBAD films was the corresponding ion sputtered film thickness divided by the total deposition time. The atom anival rates of Runs 1 and 2 were identical. as were the rates for Runs 3 and 4. The film density for NiTi is 6.5x1024 g/A3 and the average atomic mass of the film is the average between the atomic mass of nickel and titanium which is 8.55x10'7-3 g/atom. The ion arrival rate was calculated as follows: ion arrival( ion/sec-AZ) 2 current density (A/A )*( 1 .6x101 9i on/coulomb )*( coulomb/sec~A ) ......... (4) where 1 .6*1019 ion/coulomb is a constant and coulomb/secA is included for unit conversion. The current density was obtained from the current measured by the Faraday cup divided by its surface area. Four Faraday cups located directly above the substrate holder were used to measure ion current. During Run 2 the assist-beam current was measured directly above the center of the two-inch Si substrate with a Faraday cup. However for Run 3 the current density was measured at four points above the substrates. but their locations did not correspond with the center location of the one-inch Si substrates. Therefore a plot of current density versus distance from the intersection of the central axis of the assisting ions and the substrate (labeled point A on Figure 8) was assembled and an equation of the best fit line was determined. This plot and equation are shown in Figure 11. From this equation the assist-beam current density at each of the six Si substrates during Run 3 could be determined. The HA ratio at each substrate could then be determined by substituting values from the above plots into the atom-arrival equation ( 3) and the ion-anival equation(4). 4.4 X-Ray Difli'action and Atomic Structure Intensity versus two-theta plots of as-deposited and annealed films are shown in Figure 12. Both plots represent the crystallinity of all as-deposited and annealed films. 42 f. : 6.2448022 - 9.1755e-23x + 1.4538e-24x‘2 + 2.76%e-25x‘3 6.00e-22 N A a J g S.OOe-22 _ s d on g 4.00e-22 g d V b .1 g 3.00e-22 _ E 4 g 2 009. 22 g 1.00e-22 * r ' l l r l a 1 F O i 2 3 4 Distance from central axis of ion-assist beam (cm) Figure 11. Plot of current density of assisting-ion beam versus distance from the central axis of the assist flux for Run 3. Equation above plot represents the best fit curve from which current density at any point on the substrate holder could be determined. 43 2.9. 2.25 1.02 1.54 1.34 1.20 72.0i ' ' L (100 64.83 i 290 . ‘ l 57.61 ‘ ‘80 .3‘ ’ x g 50.1 70 8 £5 43.: ‘0 ‘3 . £5 36.0l s, 50 E; 20.31 to <3 55 29 (degrees) a) as deposited films 2.90 2.25 1.02 1.54 1.34 1.20 103.0 . 1 . . »1oo 92.7. -so 2: am >00 g 12.1. in 23. z; 01... so 0 5 91.5. 50 E 41.21 ‘ E 4o 2 30.91 ‘ . 130 20.6 . V 320 10.3 , ‘ €10 0.0m. I l. , . . . . . , . . . . , . 0 so so so so 70 so 26 (degrees) b)annealedfilms Figure 12. X-ray intensity versus two-theta plots for a) as deposited and b) annealed films on Si. . 44 whether ion sputtered or lBAD. Because the NiTi films were less than 0.5 microns thick. intensity from the Si substrate was very large relative to the peak intensity corresponding to the film. For films less than 0.15 microns thick. no NiTi peaks could be distinguished due to very low peak intensity. It was assumed that these films were crystalline after a single annealing run, since all x-ray diffraction patterns of films which could be scanned indicated that the annealing procedure was successful in producing a crystalline structure. On the intensity versus two-theta plots for both the as-deposited and annealed films a large peak centered around 69° appears. This peak corresponds to the Si substrate, as do the smaller peaks on the as-deposi ted plot at two-theta values of 39°, 43 .5°, 47.5° and 485°. This was verified by scanning a Si substrate without a film. According to Busch and co—workers (1990), the austenitic B2 phase has peaks at two- theta values of 42° (highest intensity), 61° and 775°. The martensitic phase will produce two-theta peaks at 39.5°, 41° and 45°. The as-deposited film did not show any peaks corresponding to any NiTi crystalline structures, indicating an amorphous structure. The annealed film indicated a B2 crystalline phase with a peak centered at 42°. There was no evidence of any presence of the martensitic phase in any of the annealed films. it was not surprising to find that the as-deposited films were amorphous. This result was also documented by Busch, etal (1989) and Kim, etal (1985); they found that NiTi films sputtered deposited at room temperature were not crystalline. Since amorphous NiTi films obviously do not display shape memory characteristics. the films had to be annealed above the crystallization temperature of 500°C (Busch, etal, 1990 ). The films could have been crystallized either through an insitu anneal or a post- deposition anneal. lnsitu annealing was not performed because of difficulty in determining the true temperature inside the vacuum chamber using a fine wire thermocouple . The temperature range for annealing was limited between 550°C and 650°C. This was to insure that the temperature was high enough to obtain crystallization, but also low enough to limit the amount of silicide formation between the silicon substrate and the nickel-titanium film. 45 NiSig forms at 750800 °C (Tung, etal. 1984) and TiSig forms at 700 °C (Morgan, etal, 1988). Films sputter deposited by L. Chang in the same deposition chamber as used in this work, were grown under nearly identical conditions as films in this work. Compositional analysis of films produced by Chang indicated a nickel-rich composition, for both as— deposited and lBAD films (Grumman and Chang, 1992). According to their work, the lBAD films had nickel compositions ranging from 54% for VA ratio of 0.3 to 57% for [IA of 1.2, and the ion sputtered films had nickel compositions of approximately 52%. It was assumed that compositions of films used in this study had the same composition as films analyzed in the work of Grummon and Chang. Studies performed on bulk NiTi have shown that as the atomic percentage of nickel increases, the austenitic start temperature (temperature at which austenite starts to form upon heating) decreases (Mercier and Melton, 1979). Bulk alloys with 50% nickel have an austenitic start temperature above 100°C, whereas NiTi alloys with 56% nickel have an austenitic start temperature below 0°C. Since the compositions of the thin films in this work were nickel-rich. it was not surprising that they were completely austenitic at room temperature as was indicated by x-ray diffraction results. 4.5 Residual Film Stress Residual stress, a, was calculated with the Stoney equation: 0 = E h2_w 3 (I - v) x2 t ........... (5) where E/(I-v) for (100) Si is 181 GPa (Stuart, 1969), h is the wafer thickness, w is the change in curvature of the wafer, x is the length of the trace of the curvature, and t is the film thickness. Figure 13 displays the as-deposited film stresses. The origin of the as-deposited film stress is primarily intrinsic (non—thermal) since the ambient temperature during deposition 46 1000 y = 34.6507 - 1403.1 185x R 20.96 L Tj Stress (MPa) .2“ T l f r r r r I r 0.0 0.2 0.4 0.6 0.8 1.0 Ion-to-Atom-Arrival Ratino (I/A Ratio) I Run 1 A Run 2 0 Runs 3 and 4 Figure 13. Residual film stress values versus l/A ratio for as-deposited films. 47 was relatively low. it is apparent that film stress becomes more compressive with increasing ion—to—atom arrival rate. It has generally been found that sputtered deposited films are in a state of compressive stress. Sputtered deposited films are grown in an environment with many energetic particles, such as accelerated ions, neutral atoms and sputtered metal atoms, which bombard the surface with a shot-peening type effect. Peening causes the sputtered atoms to be incorporated into the growing film with a higher density than would be obtained otherwise. With sufficient energy, atoms may be forced into spaces too small to accomodate them under thermal equilibrium conditions thus producing compressive intrinsic stresses. The peening effect also explains the increasing compressive stress values with increasing l/A ratio values. It is not uncommon to find rather large intrinsic film stress values, such as the compressive stress value of -1350 MPa for the as sputtered film with l/A ratio of 0.96 (Thornton and Hoffman, 1989). During deposition an accumulation of crystallographic flaws build up within the film. The density of defects formed during deposition can be as large as two orders of magnitude greater than that formed in the most severe cold working treatment of the bulk material. Hardwick ( 1987) examined sputtered deposited films with intrinsic stress which exceeded the yield point of the bulk material. This behavior was in part explained by the fact that the very small dimensions of the thin films exhibit properties quite different from those of bulk materials (surface area per volume is much greater for thin films). Also, if adhesion between the film and substrate is good, the films cannot easily yield independently of their substrates, and thus experience very high stress. At l/A-ratio values between 0 and 0.2 the film stresses fluctuated between +150 (tensile) MPa and -200 (compressive) MPa. However these stress values are from films which were grown at three different ambient pressures: Run 1 at P=2x10'5 Torr, Run 2 at P=6x10'4 Torr and Runs 3 and 4 at P=2x104 Torr. There does appear to be a trend in the behavior of the film stress and ambient pressure; as pressure increased, the resultant film 48 stress became more tensile (or less compressive) as shown in Figure 14. Thorton and Hoffman ( 1977A and 19773) have noted that during magnetron sputtering at low temperatures there exists a transition pressure above which films are tensile, and below films are in a state of compression. Sputtered film stresses become tensile when incoming energetic particles have kinetic energy too low to produce the shot- peening effect. The kinetic energy of an incoming particle is dependent upon its mean free path (distance which particles can travel without collision with surrounding particles), and the mean free path is, in turn, dependent upon ambient pressure: as ambient pressure increases, a particle's mean free path is shortened and it experiences more collisions, thus losing kinetic energy. The effects of [BAD intensify the shot peening effect. thus a higher l/A ratio results in higher intrinsic compressive film stress. Based on this reasoning, it would seem that the lBAD film of Run 2 (VA ratio of 0.1) would have a less tensile film stress than the ion- sputtered (l/A ratio of 0) than the film of Run 1 and 3 shown in Figure 13 above. However, the above data does not support this logic, indicating that either the effect of pressure is a primary controlling factor (at least in combination with the very low assist beam energy used in Run 2), or it may indicate that the degree of the kinetic excitation provided by the ion assist beam at low l/A ratios is capable of promoting more efficient structural relaxation (effects of kinetic energy are similar to thermal energy of annealing) as opposed to densifying the film structure due to a high degree of bombardment. The stresses in the annealed films are shown in Figure 15. The stresses in the annealed films of Runs 1 and 2 are not included in the plot because these values could not be measured due to film delamination from the substrate. Although the stresses could not be quantified. it was apparent they were in a state of tensile stress due to the way in which the films curled up from the plane of the substrate. Annealed (crystalline) films from Runs 3 and 4 adhered to the substrates, therefore it was possible to determine film stresses. As the [IA ratio increased the film stresses became less tensile. The range of annealed film 49 Stress (MPa) -1m U T l I T I I I l I 1 I T U V‘r 10'5 104 10-3 Pressure (Torr) Figure 14. Pressure (Torr) versus stress (MPa) for as-deposited films with l/A ratios between 0 and 0. l. 50 Stress (MPa) y = 361.0216 - 884.8049x R = 0.96 ‘ l ' l ' I ' l 0.0 0.2 0.4 0.6 0.8 1 .0 Ion-to-Atom Arrival Ratio (I/A Ratio) Figure 15. Residual film stress values versus l/A ratio for annealed films from Runs 3 and 4. 51 stress values was +4CX) MPa to 600 MPa, smaller than the as—deposited film stress range. Since the films were all annealed at the same temperature, they all acquired the same extrinsic stress, according to the relation for extrinsic film stress, Ge = Eda; - as)AT/( l - w), as explained in the literature review chapter as equation 1. Extrinsic film stress occurs if there is a mismatch between the coefficients of thermal expansion (CTE or a) for the film and substrate, which holds for this work since the CTE for the film is approximately 10x10‘6 (Buehler and Wiley, 1961) and the in-plane CTE for the (100) Si substrate is approximately 3x10'6 (Mitra et al, 1989). E{, or, as, W are all constants of the film or substrate and AT is the only variable term for this material system, and it was constant for all film/substrate systems. Using the above relation, the films all acquired a positive-signed extrinsic film stress of approximately 600 megapascals upon annealing. As mentioned above, the films all acquired the same extrinsic stress since they experienced the same AT during annealing, and all the films crystallized upon annealing. However, the films did not recover to the same stress state after annealing; as the UA ratio decreases the annealed film structure is less compressed. The time and temperature of the annealing process may not have been long enough or high enough for the more severely ion bombarded films to recover to the same state as the lower l/A ratio films. Figure 16 shows a plot of the change in film stresses between the as-deposited and annealed states. After annealing the film stresses became less compressive, again this was most pronounced in films with higher l/A ratio values. 4.6 Adhesion Measurements Adhesion measurements were made by comparing the length of lateral cracks, between the delaminated film and substrate, produced in the films by indentation. The length of these cracks extends from the center of the indentation mark to the edge of the nearly circular "bubble". Transverse cracks (normal to the film surface) also formed in the film due to the pointed geometry of the Vickers indentor, but these were not used in the 5 2 2000 y = 201.4243 + 1485.4358x R = 0.96 Change in Film Stress (MPa) Ion-to-Atom Arrival Ratio (I/A Ratio) Figure 16. Change in residual film stress between as-deposited and annealed films from Runs 3 and 4. 53 adhesion analysis. Micrographs typical of the indentation morphologies from the as.deposited film from Run 1 and the annealed lBAD film from Run 2 are shown in Figure 17. The as-deposited lBAD film from Run 2 could not be de-adhered from the surface with a 03 kg load from the diamond Vickers stylus. and any load greater than 0.3 kg fractured the silicon substrate. Micrographs of indentation marks from as-deposited and annealed lBAD films from Run 3 are shown in Figure 18. and micrographs of indentation marks from as-deposited and annealed ion-sputtered films from Run 4 are shown in Figure 19. Very poor adhesion is demonstrated in the annealed lBAD film from Run 2 shown in Figure 17 in which gross delamination occurred around the indentation mark. Shown in the same figure is the indention mark made on the as-deposited 1S film from Run 1 in which adhesion was relatively satisfactory. indentation marks on as-deposited and annealed lBAD films from Run 3 indicated relatively good adhesion. The differing lengths of the lateral cracks between the films may be correlated with variation in film thickness. The level of adhesion is directly proportional to the crack length, however crack length is a function of film thickness. The film thickness for each film/substrate system tested varied, therefore the thickness had to be factored into a relationship with adhesion and crack length. The formulation for this adjustment was adapted from Rossington etal, 1984. These authors formulated the following relation between film adhesion, film thickness and crack length: a=P3’4 [52 y? E(1 -v2)/2H3tG]1/4 ........... (6) where a is crack length; P is indentation load; 8 and y are calibration constants; E is the elastic modulus of film; Vis Poisson's ratio of film; H is the film hardness; t is film thickness; and G is fracture resistance (level of adhesion). Since P, 8, y, H and E are constant for all the indentation tests performed, they can be represented by K and the following relation is established: a = 10"“ G'”4 or ....... (7) 54 Indentation Mark Transverse h i ‘ " Lateral Crack Crack 3) Run 1 Si Substrate b) Run 2 Figure 17. Adhesion testing indentation marks on Run 1 as-deposited film and Run 2 annealed film. 55 As-sputtered Films Annealed Films l/A = 0.1 I/A = 0.3 l/A = 0.6 I/A = 1.2 Figure 18. Adhesion testing indentation marks for as—deposited and annealed films of lBAD Run 3. 56 a) As-sputtered film b) Annealed film I/ / - Figure 19. Adhesion indentation testing marks on as-deposited and annealed films of ion sputtered Run 4. Thus G is proportional to a'4 t ". Lateral crack lengths of each film were averaged and the film thickness was detennincd from profilometry measurements (as discussed above). A value representing adhesion was calculated from the relationship in equation (7). Table 2 shows all values measured and calculated to determine the degree of adhesion. These values were normalized with respect to the largest value. The system with the largest adhesion value was the as-deposited lBAD film from Run 3 with an I/A ratio of 0.3 and no heat treatment. The adhesion of this system was also examined qualitatively with a cotton swab (same as above) and the pointed tips of forceps. The film did not de-adhere or scratch away from the substrate with either attempt, and it was therefore concluded that the adhesion of this system was good. 57 2:9, .33 ecu-.38.. 582:... .899 o. 3:93: e em a... .5292 9.8:... 8... 9-98..“ 8 8m. 8.. 8.88. 9.9 m ..8 998.8 3.. 88 8... 8.85.. 9.9 n «no 9.98." 8 88 8... 8.85... 9.9 m 8... 2888 8 88 9... 8.85.. 9.9 m 8... 9-9.3 8 8 o 8.88. 9 .. 8.8 ...-m8... 8 89 8.. .2289... 9.9 m 9.... 2-92. 8 88 8... 8.88.-.. 9.9 m 8.. ...-m8... 8 88 8... 8.8%-... 9.9 m 8... 2 9.} 8 .8. e. .o .8889.... 9.9 m cm... ...-mt... 8 88 o 8.3.8....... 9 e -- -- -- 8w 8.8 8.88...... 9.9 a 8... 9.9.9.. 8 88 e .8283. 9 _ {0:8 AmeouodEV $88893; .8... 8...... ...... ...... .. S8 628. 95: E. .co 98089.8 coma—:28 comma—=5. N 935. The same simple tests were performed on all films, these methods revealed that adhesion did vary. The as-deposited films (no heat treatment) demonstrated good adhesion, with the exception of the lBAD film with VA ratio of 1.2. All of the annealed films, on the other hand, could be scratched away from the substrate with the pointed tips of forceps. The normalized adhesion values for as-deposited and annealed systems are plotted in Figure 20. The films grown on substrates cleaned in a HgO/HF solution prior to deposition (Runs 3 and 4) demonstrated better adhesion than similarly deposited films grown on substrates cleaned with alcohol and acetone (Runs 1 and 2). This was especially evident in the annealed films, in which the annealed [BAD film of Run 2 peeled back from the substrate around the indentation mark. Evidently the HZO/l-IF solution was more aggressive in cleaning adhesion inhibiting contaminants from the substrate surface. Results of adhesion testing for films of Runs 3 and 4 (Figure 20) indicate that ion beam assisted deposition with VA ratios up to 0.8 provides better adhesion, for as- deposited (not annealed) films, than ion sputtering without an assisting ion beam. The optimum lBAD process for adhesion of as-deposited films occurs at an l/A ratio of 0.3. As the [IA ratio increases beyond 0.8 the adhesion deteriorates below the level achieved by ion sputtering alone. The poor adhesion is attributed to excessively high residual film stress which can provide enough energy necessary for formation and propagation of cracks. It was not surprising to find that films grown under concurrent ion bombardment showed improved adhesion over ion sputtered films. Concurrent ion bombardment aids in interfacial mixing between the film and substrate, and can also dislodge or remove adsorbed residual gas contaminants from the film surface resulting in cleaner films with improved adhesion. Annealing the films appears to decrease the adhesion between the film and substrate for all deposition conditions. The decrease in adhesion of the film from the substrate due to S9 .: 7‘\\l 0.60% 2122 1r“ 7* 11s 0 0 0.20 0.40 0.60 0.80 1.00 1.20 lon-to-Atom Arrival Ratio Normalized Adhesion X As-sputtered [BAD films 0 Annealed lBAD films Fr guze 20. Normalized adhesion values for as-deposited and annealed films from Runs 3 an . 60 annealing could be attributed to two factors: 1) coefficient of thermal expansion mismatch between the film and substrate, and 2) structure change of the film from amorphous to crystalline. Upon annealing the film experiences a larger strain than the substrate because of its higher coefficient of thermal expansion. As the film crystallizes the atomic spacing changes and in the case of highly ordered systems, such as NiTi, the atoms may also relocated from their deposited position. Under the process of film annealing (and crystallization), some of the initial chemical bonds formed during deposition between the film and substrate atoms break, weakening the interface and hence causing an overall decrease in film adhesion. The optimum deposition condition for adhesion of annealed films remains to be ion beam assisted deposition with an l/A ratio of 0.3, however the adhesion for the annealed film is only 30% of the as—deposited level. Although the adhesion levels in the annealed films were lower than the as-deposited levels, the adhesion between the annealed film and substrate remains satisfactory. 61 62 CHAPTER 5. CONCLUSIONS Both ion sputtered deposition and ion beam assisted deposition of NiTi onto silicon substrates at approximately 80'C produced amorphous films. Vacuum annealing at 6(D°C for 15 minutes caused the films to crystallize, and according to x-ray diffraction results all films were fully austenitic after annealing. The residual stress of the as-deposited films became more compressive with an increasing ion-to-atom arrival ratio within the range of [IA ratio values between 0 and 1.2. Some films deposited at very low l/A ratios (0 to 0.1) were grown under different ambient deposition pressures, within these parameters the film stress was less compressive (or more tensile, since the residual film stresses in this range were near zero) as pressure increased. The shot-peening efi'ect experienced by films deposited via ion-sputtering was intensified with higher I/A ratios and lower ambient pressures, which in turn caused the film stress to be more compressive. When the films were annealed they experienced a positive-signed change in residual stress (less compressive, or more tensile). Films with higher [IA ratios experienced a larger change than lower l/A ratio films, but still remained in a compressive stress state. NiTi films deposited onto silicon substrates cleaned with an l-IF/HzO solution prior to deposition demonstrated better adhesion than films deposited onto silicon subsuates cleaned with acetone and alcohol. Ion beam assisted deposition with VA ratios up to 0.8 provikd better adhesion than ion sputtering without an assisting ion beam. The assist ions promote interfacial mixing between the film and substrate thus enhancing adhesion, however as the VA ratio increases above 0.8 the resultant film stress becomes too high and adhesion deteriorates. Annealing the films caused the adhesion levels to decrease. The drop in adhesion is attributed to a coefficient of thermal expansion mismatch between the film and substrate, and to the change in atomic structure of film (from amorphous to crystalline), both of which produced weaker interfacial bonding relative to the initial as-deposited interfacial bonding. The optimum deposition conditions for both as-deposited and annealed film adhesion was achieved at an [M ratio of approximately 0.3. However, the adhesion of films deposited onto silicon substrates cleaned with HF/HgO was good for all films tested. 63 BIBLIOGRAPHY N .A. Ahmed and J.S. Colligon, "The Application of Dynamic Recoil Mixing to Enhance Adhesion of Gold Films on Silica Substrates", Vacuum, v.38 N.2 (1988) 83-88. J. Angilello, J.E. Baglin, F. Cardone, J. Dempsey, F.M. d'l-leurle, E.A. Irene, R. Maclnnes, C.S. Petersson, R. Savoy, A. Segmuller and E.Tiemey, "Tantalum Silicide Film Deposition by D C Sputtering", Journal of Electronic Materials, 10 ( 1981) 60-74. J.E.E. Baglin, "Thin films: The Relationship of Structure to Properties”. in Materials Research Society Symposium Proceedings (ed. C.R. Aita), 47 (1985) 3-10. J.E.E. Baglin, "Interface Structure and Thin Film Adhesion”, in Handbook of Ion Beam Processing Technology (ed. JJ. Cuomo, S.M. Rossnagel and HR. Kaufman). (1989) 279-299. 8.8. Berry and WC. Pritchet, "lntemal Stress and lntemal Friction inThin-Layer Microelectronic Materials”, Journal of Applied Physics, 67 (1990) 3661-3666. R.S. Bhattacharya, C.N. Raffoul and AK. Rai, ”A Potentiodynamic Evaluation of NiMo and NiTi Amorphous Thin Films", Corrosion v.42, N.4 (1986) 236-240. J.D. Busch, A.D. Johnson, DE. Hodgson, CH. Lee and D.A. Stevenson, ”Development of Shape Memory Porperties in Sputter Deposited Films of Nickel Titanium Alloys". presented at lCOMAT 1989. JD. Busch, A.D. Johnson, CH. Lee and D.A. Stevenson, ”Shape-Memory Properties in Ni-Ti Sputter-Deposited Film”, Journal of Applied Physics, v.68 N. 12 (19%) 6224- 6228. J.D. Busch and AD. Johnson, ”Prototype Micro-Valve Actuator", presented at MEMS in Napa Valley, Ca 1990. JD. Busch, M.H. Berkson and AD. Johnson, "Phase Transformations in Sputtered Ni-Ti Film: Effects of Heat Treatment and Precipitates", presented at Materials Research Society Symposium in Anaheim, CA 1991. D.W. Butler, C.T.H. Stoddart and PR. Struart, ”Evaluation of Film Adhesion Using Scratch Testing", Journal of Physics D, 3 (1970) 877-. G. Carter and D.G. Armour, "The Interaction of Low Energy [on Beams with Surfaces". Thin Solid Films, 80 (1981) 13-29. R.N. Castellano, M.R. Notis and G.W. Simmons, "Composition and Stress State of Thin Films Deposited by Ion Beam Sputtering”,Vacuum, 27 (1978) 109-1 17. B.N. Chapman,"Thin Film Adhesion", Journal of Vacuum Science and Technology, I l (1974) 106-1 13. ' S.S. Chiang, D.B. Marshall and A.G.Evans,"A Simple Method for Adhesion Measurements", Surfaces and Interfaces in Ceramic and C eramic/Metal Systems. v.14 (1981) 603-617. 64 B.M. Clemens, ”Effect of Sputtering Pressure on the Structure and Solid-State Reaction of Titanium-Nickel Compositionally Modualated Film", Journal of Applied Physics. V.6l N9 (1987) 4525-4529. J.S. Colligon and Kheyrandish, "Radiation Enhanced Adhesion of Thin Films", Vacuum,V.39 N.7 (1989) 705-710. J.H. Comfort, L.M. Garverick and R. Reif, "Silicon Surface Cleaning by Low Dose Argon-[on Bombardment for Low-Temperature (750°C) Epitaxial Silicon Deposition. 1. Process Considerations”, Journal of Applied Physics, 62 (1987) 3388-3397. J.J. Cuomo, J.M.E. Harper, C.R. Guamieri, D.S. Yee, L.J.Attianasio, C.T. Wu and RH Hammond, "Modification of Niobium Film Stress by Low-Energy [on Bombardment During Deposition”, Journal of Vacuum Science and Technology, 20 (1982) 349-354. F.M. d'Heurle, ”Aluminum Films Deposited by rf Sputtering”, Metallurgical Transactions, 1 (1970), 725. F.M. d'Heurle, " Metallurgical Topics in Silicon Device lnterconnections: Thin Film Stresses", International Materials Review, V34, N.2 (1989A) 53-68. F.M. d‘Heurle and J.M.E. Harper, ”Note on the Origin of Intrinsic Stresses in Films Deposited Via Evaporation and Sputtering“, Thin Solid F ilms, 171 (19898) 81-92. F.A. Doljack and R.W. Hoffman, "The Origins of Stress in Thin Nickel Film”, Thin Solid. Films, 12 (1972) 71-74. M.D. Drory, M.D. Thouless and A.G. Evans, "On the Decohesion of Residually Stresses Thin Films”, Acta Metallurgica, 36 (1988) 2019-2024. A.G. Elliot, "The Condenstion of Gold onto Tantalum (1m) Single Crystal Surfaces: Leed, AES Analysis”, Surface Science, 51 (1975) 489-503. W. Ensinger and GK. Wolf, ”Ion Assisted Coatings For Corrosion Protection Studies". Materials Science and Engineering, Al 16 (1%9) 1-10. A.G. Evans and J.W. Hutchinson, "On the Mechanics of Delamination and Spalling in Compressed Films", International Journal of Solids and Structures, 20 (1984) 455- 466. 0.8. Grummon and L. Chang, ”Effect of Concurrent Ion Bombardment on Structure and Composition of Ion Sputter-Deposited Ni-Ti Thin Films", presented at Materials Research Society Symposium in Boston, 1992. D.A. Hardwick, ”The Mechanical Properties of Thin Films: A Review”,Thin Solid F ilms. 154 (1987) 109-124. Heghkovitz, ”Stress Relaxation in Thin Aluminum Films”, Thin Solid Films, 130 ( 1985) -93. D.W. Hoffman and MR. Gaettner, ”Modification of Evaporated Chromium by Concurrent Ion Bombardment", Journal of Vacuum Science and Technology, 17 (1%0) 425-428. 65 D.W. Hoffman and CM. Kukla. ”Determination of Film Stresses During Sputter Deposition Using an In Situ Probe”. Journal of Vacuum Science and Technology A, V.3 N.6 (1985) 2600-2604. D.W. Hoffman and J .A. Thornton. "lntemal Stresses in Sputtered Chromium”, Thin Solid F ilms. 40 (1977A) 355-364. D.W. Hoffman and J .A. Thornton, "The Compressive Stress Transition in Al, V, Zr. Nb and W Metal Films Sputtered at Low Working Pressures", Thin Solid Films, 45 (19778) 387-96. R.W. Hoffman, ”Stresses in Thin Films: The Relavance of Grain Boundaries and Impurities", Thin Solid Films, 34 ( 1976) 185-190. T.R. Hull, J.S. Colligon and A.E. Hill, ”Measurement of Thin Film Adhesion”, Vacuum. V.37 N. 3/4 (1987) 327-330. R.A. Hunt and B. Gale, ”A model of Deadherence Due to Stresses in an Elastic Film”, Journal of Applied Physics D. 5 (1972) 359-372. M. Itoh, M. Hori and S. Nadahara, ”The Origin of Stress in Sputter Deposited Tungsten Films for X-Ray Masking", Journal of Vacuum Science and Technology 8, 9 ( 1991) 149-153. A.D. Johnson, "Vacuum-Deposited TiNi Shape Memory Film: Characterization and Applications in Micordevices", Journal of Micromechanics and Microengineering, l (1991) 34-41. . AJ. Kellock, G.L. Nyberg and J.S. Williams, "Thin Film Adhesion Improvements Under Photon Irradiation”, Vacuum, v.35 N. 12 (1985) 625-628. JJ. Kim, P. Moine and D.A. Stevenson, "Crystallization Behavior of Amorphous Ni-Ti Alloys Prepared by Sputter Deposition”, Scripta Metallurgica, 20 (1986) 243-248. W.-Z. Li, F.-Z. Cui, U Liao and H.D. Li, "A Simulation Study of Interface Mixing During Ion-Assisted Deposition", Vacuum, V.40 N3 (1990) 289-292. J.R. Lloyd and S. Nakahara. "Voids in Thin As-Deposited Gold Films Prepared by Vapor Deposition", Journal of Vacuum Science and Technology, 14 ( 1977) 655-659. 0.8. Marshall and A.G. Evans, ”Measurement of Adherence of Residually Stressed Thin Films by Indentation. I. Mechanics of Interface Delamination', Journal of Applied Physics, v.56 N. 10 (1984) 2632-2638. D.M. Mattox, "Thin Film Metallization of Oxides in Microelectronics”, Thin Solid Films, 18 (1973). 173-186. D.M. Mattox, “Thin-Film Adhesion and Adhesive Failure-A Perspective“, in Adhesion Measurement of Thin F ilnrs, Thick F ilms. and Bulk Coatings, AST M Special Technical Publication 640, KL Mittal, ed. (1978) 55-62. D.M. Mattox, "Particle Bombardment Effects on Thin-Film Depostion: A Review", Journal of Vacuum Science and Technology A, V.7 N3 (1989) 138- 146. 66 D.M. Mattox and RE. Cuthrell, ”Residual Stress, Fracture, and Adhesion in Sputter Deposited Molybdenm Films", in Materials Research Society Symposium Proceedings. V. 119 ( 1988) 132- 137. O. Mercier and K.N. Melton, "The Substitution of Cu for Ni in NiTi Shape Memory Alloys", Metallurgical Transactions A, 10A, (1979) 387-389. U.N. Mitra, P.W. Davies, R.K. Shukla and J.S. Multani, “Semiconductor Silicon/1986”. ed. H.R. Huff, T. Abe and D. Kolbesen, (1986). A.E. Morgan, E.K. Broadbent, K.N. Ritz, D.K. Sadana and BJ. Burrow, "Interactions of Thin Ti Films with Si, SiOZ. Si3N4, and SiOxNy Under Rapid Thermal Annealing". Journal of Applied Physics, 64 ( 1988) 344-353. K.H. Muller, ”Stress and Microstructure of Suptter-Deposited Thin Films: Molecular Dynamics Investigation", Journal of Applied Physics, 62 (1%7) 1796-1799. H.-W. Park-and S. Danyluk, ”Residual Stress Measurement in Filament-Evaporated Aluminum Films on Single Crystal Silicon Wafers”, ‘Journal of Materials Science, 20 (1990) 23-27. G.G. Paulson and A.C. Friedberg, "Coalescence and Agglomeration of Gold Films”. Thin Solid Films, 5 (1970) 47-52. W. Pompe, HJ. Scheibe, A. Richter, J .D. Bauer, K. Bewilogua and C. Weissmantel, "On the Investigation of Residual Stress and Density Flucuation on the Crystallization of Amorphous Carbon Films", Thin Solid Films, 144 (1986) 77-82. A.K. Rai and RS Bhattacharya, ”Sputter-Deposited and Ion-Mixed Ni-Ti and Ni-Cr Thin Films", Materials Science and Engineering, 85, (1987) 139-145. C. Rossington, A.G. Evans, D.B. Marshall and B.T. Khuri-Yakub, "Measurements of Adherence of Residually Stressed Thin Films by Indentation 11. Experiments with ZnO/Si“, Journal of Applied Physics, 56 (1984) 2639-2644. S.M. Rossnagel and JJ. Cuomo, "Ion Beam Bombardment Effects During Film Deposition", Vacuum, V38 N2 (1988) 73-81. PJ. Rudeck, J.M.E. Harper and RM. Fryer, ”The Angular Dependence of Preferential Sputtering and Composition in Aluminum-Copper Thin Films", Journal of Vacuum Science and Technology, V. A7, N3 (1989), 2289-2293. W. Sachse, K.Y. Kim and H.D. Conway, "Determination of the Bond Strength of a Metal Costing on a Substrate", in Materials Research Society Symposium Ptoceedings. V. 15.1 (1989) 241-259. J .-A. Schweitz, "A New and Simple Micromechanical Approach to the Stress-Strain Characterization of Thin Coatings“, Journal of Micromechanics and Microengineert‘nc. I ( 1990) 10-15. F.A. Smidt, “Use of Ion Beam Assisted Deposition to Modify the Microstructure and Properties of Thin Films", International Materials Review, V35 N2 (1990) 61-146. 67 P.R. Stuart, “Some Measurements of Stress in Thin Films Prepared by Low Pressure Triode Sputtering”, Vacuum, 19(1969) 507-512. R.C. Sundahl, "Relationship Between Substrate Surface Chemistry and Adhesion of Thin Films”, Journal of Vacuum Science and Technology, V.9 NI (1971) 181-185. M .L. Tarn g and GK. Wehner, "Auger Electron Spectroscopy Studies of Sputter Deposition and Sputter Removal of Mo from Various Metal Surfaces", Journal of Applied Physics, 43 ( 1972) 2268-2280. J.A. Thornton and D.W. Hoffman, ”Internal Stresses in Titanium, Nickel, Molybdenum and Tantalum Films Deposited by Cylindrical Magnetron Sputtering”, Journal of Vacuum Science and Technology, 14 (1977) 164-168. J.A. Thornton, J. Tabock and D.W. Hoffman, "Internal Stresses in Metallic Films Deposited by Cylindrical Magnetron Sputtering", Thin Solid Films, 64 (1979) 111-119. J.A. Thornton and D.W. Hoffman, "Stress-Related Effects in Thin Films", Thin Solid Films, 171 (1989)5-31. R.T. Tung, J.M. Poate, J.C. Bean, J.M. Gibson and D.C. Jacobson, "Epitaxial Silicides". Materials Research Society Symposium Proceedings, 81 (19%) 79-92. . TJ. Vink, M.A. Somers, J.L.C. Dams and A.G. Dirks, "Stress, Strain, and Microstructure of Sputter-Deposited Mo Thin Films", Journal of Applied Physics, v.70 N8 (1991) 4301-4308. FJ. von Preissi g, ”Applicability of the Classical Curvature-Stress Relation for Thin Films on Plate Substrates", Jounal of Applied Physics, 66 ( 1989) 4262-4268. J.A. Walker, KJ. Gabriel and M. Mehregany, ”Thin-film Processing of TiNi Shape Memory Alloy", Sensors and Actuators, A21-Zi (1990) 243-246. C.R. Wie, J.Y. Tang,T.A. Tombrello, R.W. Grant and R.M. Housley, "Ionizing Beam- lnduced Adhesion Enhacement and Interface Chemistry for Au-GaAs", Vacuum. v.38 N.3 (1988) 157-160. W.D. Westwood, ”Glow Discharge Sputtering", Progress in Surface Science, 7 (1976) 71-1 1 1. P.H.Wocjiechowski, “The Effects of lnsitu Ion Bombardment on Film Stress”, Journal of Vacuum Science and Technology A. 6 (1988) 1924-1930. D.S. Yee, J F'loro, DJ. Mikalsen, JJ. Cuomo, K.Y. Ahn and D.A. Smith, ”Stress Modification of WSi2 Films by Concurrent Low-Energy Ion Bombardment During Alloy Evaporation", Journal of Vacuum Science and Technology A, V3 (1985) 2121 - 2126. X.-C. Zheng and RA. Dodd, ”Epitaxial Recrystallization of Ni-Ti Films During Plasma Sourch Ion Mixing", presented at IBM 199), to be published in Nuclear Instruments and Methods. 68 "illlllllllllll'lllf