MICROSCOPY RESEARCH AND TECHNIQUE 40:136–151 (1998) High Resolution Electron Microscopy of Amorphous Interlayers Between Metal Thin Films and Silicon L.J. CHEN,* J.H. LIN, T.L. LEE, C.H. LUO, W.Y. HSIEH, J.M. LIANG, AND M.H. WANG Department of Materials Science and Engineering, National Tsing Hua University, Hsinchu, Taiwan, Republic of China KEY WORDS solid-state diffusion; growth kinetics; structure; stability; auto-correlation function analysis ABSTRACT High-resolution electron microscopy of amorphous interlayers (a-interlayer) formed by solid-state diffusion between metal thin films and silicon is reviewed. In this paper, an overview of the development is presented. Pertinent data obtained on the growth kinetics and structure of a-interlayers in polycrystalline metal thin films on single-crystal silicon are reported. For the Ti/Si, Zr/Si, Hf/Si, V/Si, Nb/Si and Ta/Si systems, the growth of a-interlayer was found to follow a linear law in the initial stage. Si atoms were found to be the dominant diffusing species in the solid phase amorphization in the Ti/Si, Zr/Si, and Hf/Si systems. For the Y/Si system, the stability of amorphous interlayer depends critically on the composition of the amorphous films. Auto-correlation function analysis was utilized to determine the structure of the amorphous interlayers. HRTEM in conjunction with the fast Fourier transform were applied to determine the first nucleated crystalline phase. Simultaneous presence of multiphases was observed to occur in a number of refractory metal/Si systems. Microsc. Res. Tech. 40:136–151, 1998. r 1998 Wiley-Liss, Inc. INTRODUCTION The study of amorphous metallic alloys dates back to investigations during the 1950s. The discovery of the formation of amorphous interlayer (a-interlayer) by solid-state diffusion in diffusion couples is relatively recent ( Johnson, 1988). For metal thin films on silicon, although the presence of an amorphous membrane at the metal-silicon interface was proposed to predict the first nucleated phase of the system in 1976, direct confirmation of solid phase amorphization was not realized until much later (Walser and Bene, 1976). Interfacial reactions of metal thin films on silicon have been a subject of continuing interest for many years (Nicolet and Lau, 1983; Tu and Mayer, 1978). The formation of a-interlayer is assumed to be driven by the large negative heat of mixing of the two components and can proceed in a broad concentration range as long as the temperature is high enough to allow sufficiently rapid interdiffusion of the atoms, but low enough to avoid any nucleation of the crystalline compound during the reaction time ( Johnson, 1988). The free energies of mixing between the undercooled liquid metals and silicon were estimated with the method proposed by Miedema et al. (1976). Negative free energy of mixing, which provides the driving force for the formation of a-interlayer, was obtained for all refractory metal/Si systems, with the exception of the W/Si system, and all rare-earth metal/Si systems investigated so far (Chen et al., 1994). As a rule, the crystalline compound phase appearing in the equilibrium phase diagram is more stable than the amorphous alloy of the same composition. An example is shown in Figure 1. For metal-silicon systems, the formation of amorphous metal-Si alloys between metal thin films and amorphous silicon during low-temperature annealing was first discovered in the Rh/a-Si system (Herd et al., 1983). Subsequently, it has been found to occur in all r 1998 WILEY-LISS, INC. refractory metal/Si and a number of rare-earth metal and platinum group metal and crystalline silicon systems (Abelson et al., 1988; Cheng and Chen, 1990, 1991; Hsieh et al., 1993; Lee and Chen, 1993; Liang and Chen, 1994; Lin and Chen, 1995; Lin et al., 1996; Lur and Chen, 1989; Wang and Chen, 1992). Both conventional and high-resolution transmission electron microscopy (HRTEM), reflection high-energy electron diffraction (RHEED), low-energy electron diffraction (LEED), Auger electron spectroscopy (AES), X-ray diffraction, differential scanning calorimetry, and electrical measurements have been applied to characterize the structures and properties of the systems. HRTEM has played a critical role in clarifying many key issues in the formation of amorphous layers in metal-silicon systems. In this paper, we present an overview of the development and report the results obtained on the growth kinetics and structure of a-interlayer in polycrystalline metal thin films on single-crystal silicon. Reactions in metal/amorphous silicon multilayers were reviewed recently (Sinclair and Konno, 1993). The scope of the present review is mostly confined to the reactions between metal thin films and single-crystal silicon unless it is deemed appropriate to use some of the results obtained for metal/a-Si multilayer reactions. MATERIALS AND METHODS For growth kinetics study, single crystal, 3–8 V-cm, 2 or 3 inches in diameter, phosphorus doped (001), (111), and (011) oriented silicon wafers were used. The wafers were cleaned chemically by a standard procedure. Thin metal films, 10–60 nm in thickness, were then deposContract grant sponsor: Republic of China National Science Council; Contract grant number: NSC84-2215-E007-011. *Correspondence to: L.J. Chen, Department of Materials Science and Engineering National Tsing Hua University, Hsinchu, Taiwan, Republic of China. E-mail: email@example.com Received 16 March 1996; accepted in revised form 21 March 1996. HREM OF AMORPHOUS INTERLAYERS Fig. 1. 137 Metastable free energy diagram at 500°C for the Ti/Si system. ited onto the wafers at room temperature in a UHV system (with a base pressure better than 1 3 10210 Torr) or in a non-UHV system. An a-Si layer, 10–20 nm in thickness, was deposited onto the metal layers to protect the metal thin films from oxidation during subsequent annealings. The vacua during depositions were maintained to be better than 1 3 1029 and 5 3 1027 Torr in UHV and non-UHV systems, respectively. To determine the effects of composition on the stability of amorphous interlayer, multilayer metal/Si samples were also prepared under UHV conditions. The deposited samples were annealed isothermally in an oil-free vacuum furnace or in a nitrogen gasflowing diffusion furnace at temperatures ranging from 200 to 625°C. HRTEM imaging was carried out in a JEOL (Peabody, MA) 4000EX electron microscope operating at 400 keV with a point-to-point resolution of 0.18 nm. Most of the cross-sectional TEM (XTEM) micrographs were taken under symmetric  diffraction condition. In situ RHEED, LEED, XRD, and AES analyses of the thin film systems were also carried out. For electrical characterization, sheet resistance, Schottky barrier height, contact resistance, and leakage current were measured (Liauh et al., 1993). GROWTH LAW AND ACTIVATION ENERGY For UHV deposited Ti/(111)Si samples, RHEED analysis revealed that after the first 0.2-nm-thick Ti layer Fig. 2. HRTEM image of an as-deposited sample, cross-sectional view (c-s). was deposited, the diffraction peaks corresponding to (7 3 7) Si surface reconstruction disappeared. An intermixed amorphous layer was formed during the deposition of the first 1.7-nm-thick Ti layer. The characteristic diffraction pattern of Ti showed up following 2-nmthick Ti deposition. XTEM examination of as-deposited samples revealed that a 2-nm-thick amorphous inter- 138 L.J. CHEN ET AL. Fig. 3. HRTEM images of V/(001)Si samples (a) as-deposited and annealed at 440°C for (b) 15, (c) 45, (d) 75, (e) 105 minutes, showing the growth of the amorphous interlayer. layer was indeed formed. The Ti thin film was also found to exhibit an (0, 1, 21, 0) texture. An example is shown in Figure 2 (Wang and Chen, 1992). For Ti/Si, Zr/Si, Hf/Si, V/Si, Nb/Si, and Ta/Si samples, amorphous interlayers were observed to form in as-deposited samples. The thickness of the ainterlayer was found to increase with annealing temperature and time (Chen et al., 1994; Cheng and Chen, 1990, 1991; Lur and Chen, 1989). Examples are shown in Figure 3. In the initial stage the growth follows a linear law. The growth then slows down and comes to a complete stop as a critical thickness is reached. The activation energies of the growth of a-interlayer are listed in Table 1. A 1-nm-thick a-interlayer was found to form in asdeposited Mo/Si samples. The maximum thickness was measured to be 4 nm, which is considerably thinner than that found in non-UHV samples. For as-deposited W/Si samples prepared under UHV environment, no continuous a-interlayer was detected to form. In samples annealed at 500°C for 1 hour, a 3-nm-thick disordered layer was detected at the W/c-Si interface. In samples annealed at a temperature as high as 700 or 750°C for 20 seconds, no a-interlayer was found to form (Chen et al., 1994). 139 HREM OF AMORPHOUS INTERLAYERS TABLE 1. Correlations Among Physical Parameters and Kinetic Data in the Formation and Growth of Amorphous Interlayers in Refractory Metal/Si and Y/Si Systems Metal Atomic radius (nm) Atomic size difference between metal and Si (%) Atomic weight (g) Electronegativity Chemical potentials for electrons (V) DG (kJ/g atom) Maximum thickness (nm) Activation energy (ev) Dominant diffusing species Ti Zr Hf Nb Ta V Y Si 0.147 11.3 47.9 1.5 3.65 239 10 1.6 — 0.16 21.2 91.2 1.4 3.4 253 17 1.4 Si 0.167 26.5 178.5 1.3 3.55 258 27 1.2 Si 0.146 10.6 92.9 1.6 4.25 223 8 0.8 — 0.149 12.8 181 1.5 4.00 221 7.5 0.9 — 0.134 1.5 50.9 1.6 4.05 215 4.5 1.1 — 0.178 34.9 88.9 1.3 3.20 239 .10 — — 0.132 — 28.1 1.8 4.70 — — — DOMINANT DIFFUSING SPECIES For the Zr/Si and Hf/Si systems, Kirkendall voids were found to be present beneath the a-interlayer/c-Si interfaces (Cheng and Chen, 1990). Similarly, the voids were seen to form at the a-interlayer/a-Si interface in Ti/a-Si multilayer system (Holloway and Sinclair, 1987). The observation indicated that silicon is the dominant diffusing species during the reactions in these systems. AVERAGE COMPOSITION OF THE AMORPHOUS INTERLAYERS A scheme using an ultrathin (,1 nm) oxide layer and capping Mo layer to define the reference planes for interdiffusion has been utilized to determine the average compositions of amorphous interlayer in ultrahigh vacuum (UHV) deposited V/c-Si samples. The buried oxide layer is present at the epitaxial Si/c-Si interface in UHV deposited samples. A thin Mo layer was deposited on the V layer to serve as an inert cap and define the upper surface of the metal layer. The ratios of participating Si and V atoms in the reactions were measured to be 5:6 and 5:9 in samples annealed at 465°C for 10 and 30 min, respectively. An example is shown in Figure 4 (Lin and Chen, 1996). A similar scheme using an ultrathin (,1 nm) oxide layer and capping W layer to define the reference planes for interdiffusion has been utilized to determine the average composition of amorphous interlayer in Zr/c-Si samples. The ratio of participating Si and Zr atoms in the reactions in samples annealed at 420°C for 50 minutes was measured to be 3:5. RARE-EARTH METAL/Si SYSTEMS For 10-nm-thick yttrium thin film on silicon, an 11-nm-thick amorphous interlayer located between the amorphous Si capping layer and Si substrate was observed. An example is shown in Figure 5. In samples annealed at 200°C, crystalline Y5Si3 and Si were always observed to nucleate within the amorphous interlayer. In addition, about 1.5-nm-thick epitaxial YSi2-x layer was found to form at the amorphous interlayer/c-Si interface in samples annealed for 30 minutes. An example is shown in Figure 6. In samples annealed at temperatures higher than 250°C, only epitaxial YSi2-x was found to form at the amorphous interlayer/c-Si interface. The amorphous interlayer was found to completely transform to epitaxial YSi2-x in a sample annealed at 300°C in 5 minutes. An example is shown in Figure 7 (Lee and Chen, 1993). Fig. 4. XTEM images of V/(111)Si samples (a) as-deposited and (b) annealed at 465°C for 10 minutes, showing the movement of the interfaces. c, d: Schematic diagrams of a and b, respectively. 140 L.J. CHEN ET AL. Fig. 5. HRTEM image of an as-deposited a 5 Si(20 nm)/Y(10 nm)/(111)Si sample viewed along the [112 ]Si direction. Fig. 6. HRTEM image of an a-Si(20 nm)/Y(10 nm)/Si sample annealed at 200°C for 30 minutes viewed along the Si direction. HREM OF AMORPHOUS INTERLAYERS Fig. 7. HRTEM image of an a-Si(20 nm)/Y(10 nm)/Si sample annealed at 300°C for 5 minutes, c-s view. Fig. 8. HRTEM image of an as-deposited a-Si(20 nm)/Y(30 nm)/Si sample viewed along the Si direction. 141 142 L.J. CHEN ET AL. Fig. 9. Schematic diagrams illustrating the phase formation in as-deposited samples. a: Initial configuration following deposition of a thin layer of Y. b: Growth of a-interlayer as the thickness of the deposited layer increases. c: Crystalline phases are formed when the critical thickness of the a-interlayer is exceeded. d: Same sample as that of c but annealed at 200°C. It is apparent that the deposited 10-nm-thick yttrium thin film was completely intermixed with atoms from the Si substrate to form an amorphous layer with a high Si concentration before the amorphous Si capping layer was deposited. As a result, the further intermixing of Si atoms from the amorphous capping layer becomes rather difficult. To check this conjecture, a-Si(20 nm)/ Y(20 nm)/Si, a-Si(20 nm)/Y(30 nm)/Si, and a-Si(20 nm)/Y(60 nm)/Si samples were also prepared. Crystalline phases were found to form in all these samples in as-deposited forms by both XRD and TEM. The XRD spectra revealed the phases formed in as-deposited samples free from any artifacts induced by ion milling. Considerable intermixing between the amorphous Si capping layer and yttrium thin films was found in these samples from AES concentration depth profiles. TEM observation and diffraction analysis revealed the simultaneous presence of polycrystalline Y, Y5Si3, YSi, Si, and amorphous interlayer. In as-deposited 20- to 60-nm-thick yttrium thin films on silicon samples, a 2.5-nm-thick amorphous interlayer was also found to form at the Y/c-Si interface. Y5Si3 was also found to be present. An example is shown in Figure 8. The thickness of the amorphous interlayer was found to increase with annealing time in samples annealed at 250°C up to 60 minutes. Compared to the results obtained for a-Si(20 nm)/Y(10 nm)/Si samples, it is apparent that as the critical thickness of the amorphous layer was exceeded, further intermixing of Y and Si was overtaken by the formation of crystalline phases at room temperature. Schematic diagrams of the interfacial reactions in the Y/Si system at low temperatures are shown in Figure 9. Amorphous interlayers were also found to form in Er/Si, Tb/Si, Dy/Si, and Gd/Si systems. For both Er/Si and Tb/Si samples, a 2.5-nm-thick amorphous interlayer was observed in the as-deposited samples. The growth followed a linear growth law in samples annealed at 190–250°C. The maximum thicknesses of the a-interlayers were measured to be 9 and 10 nm for the Er/Si and Tb/Si systems, respectively. Examples are shown in Figure 10 (Luo et al., unpublished data). PLATINUM GROUP METAL/Si SYSTEMS Amorphous interlayers were found to form in a Rh/a-Si diffusion couple as early as 1983 (Herd et al., 1983). More recently, solid state amorphization was also found to occur in the Rh/c-Si system (Tu et al., 1991). For metal/Si systems, Pt/Si was the first system HREM OF AMORPHOUS INTERLAYERS 143 Fig. 10. HRTEM images of Er/(001)Si samples annealed at 225°C for (a) 10, (b) 30, (c) 60, and (d) 90 minutes, showing the growth of the amorphous interlayer. in which amorphization by solid-state diffusion was discovered. However, the layer was rather thin and discontinuous (Abelson et al., 1988). A recent study showed that the amorphous interlayer is also present in as-deposited Ir thin film on silicon (Chen et al., 1994). EFFECTS OF ATOMIC COMPOSITION ON THE THERMAL STABILITY OF AMORPHOUS INTERLAYERS IN Y-Si MULTILAYER SYSTEMS Three sets of multilayer films with overall atomic composition ratios 1Y:2Si, 1Y:1Si, and 5Y:3Si were prepared. The films consisted of a total of 13 alternating Si and Y layers, with the starting and final layers being Si. The thickness of the Si layer was fixed at 5 nm, whereas the thicknesses of the Y layers varied from 5 to 17.5 nm to yield the desired average atomic concentrations. The top a-Si layer also served as a capping layer to protect the thin film assembly from oxidation during subsequent heat treatments. The configurations of the three sets of samples are depicted in Figure 11. A 65-nm-thick amorphous Y-Si intermixing layer was found to form in multilayer films with an overall composition of YSi2 at room temperature. An XTEM image and AES depth profile are shown in Figure 12. Homogenization of atomic compositions was found to proceed in samples annealed at 250–350°C. Examples are shown in Figure 13. The multilayer remained amorphous in this temperature range. In samples annealed at 400°C, the amorphous layer was found to convert to crystalline YSi2. For multilayer films with an overall composition of YSi, crystalline phases were found to be present between amorphous layers in as-deposited samples. Figure 14 shows an HRTEM image of an as-deposited sample. About 6-nm-thick amorphous interlayers were found to form between crystalline phases. From TEM diffraction analysis, crystalline Y5Si3, YSi, and Y are present. The average grain size was measured to be 20 nm. The results are consistent with those obtained from XRD spectrum. From the AES profile, significant interdiffusion was found to occur in as-deposited samples. In samples annealed at 400°C for 5 min, Y5Si3 and YSi were found to form at the expense of the amorphous layer. The average grain size was increased to 25 nm. In samples annealed at 500°C for 10 minutes, both Y5Si3 and Y2O3 were found to be present. In samples annealed at and higher than 600°C, Y2O3 was the only compound phase that remained. The results obtained for multilayer films with an overall composition of Y5Si3 are similar to those of the 1Y:1Si films. XRD spectrum of an as-deposited sample revealed that crystalline Y5Si3, YSi, and Y phases were 144 L.J. CHEN ET AL. Fig. 11. A schematic diagram showing the configurations of as-deposited multilayer samples. found to be present. About 5.5-nm-thick amorphous interlayers were found to form between crystalline phases. From TEM diffraction analysis, crystalline Y5Si3, YSi, and Y are present. The average grain size was measured to be 15 nm. Figure 15 shows an HRTEM image. Y5Si3 was found to be the only silicide phase formed in samples annealed at 400°C. The average grain sizes of Y5Si3 were found to increase from 20 nm to 1 µm as the annealing temperature was increased from 300 to 400°C. The formation of crystalline Y5Si3 and YSi was found in as-deposited samples prepared with excess Y so that the composition of the top intermixing layer exceeds a critical value. The Y overlayer was found to react with the amorphous layer to form polycrystalline Y5Si3, YSi, and Si in samples annealed at temperatures lower than 400°C. In 5Y:3Si films, Y5Si3 was found to be the only silicide phase formed in samples annealed at 400°C. The results indicated that the nucleation of the crystalline phase depends on the composition. The thickness of the amorphous layer was found to decrease with annealing temperature and/or annealing time. The height of nucleation barriers to form silicides in the Y-Si system was seen to depend critically on the atomic composition of the layer, which in turn significantly influences the phase formation sequence of the system. Metastable free energy diagrams provide a basis for predicting whether the solid-state amorphization can proceed. Based on a growth control model, it was concluded that the formation of the amorphous layer at room temperature was controlled by nucleation (Lee and Chen, 1994). CORRELATIONS WITH PHYSICAL PARAMETERS Miedema has expressed the energy in terms of differences in chemical potential and electron density at the Wigner-Seitz cell of the two elements (Miedema et al., 1976). A large difference in chemical potential is required for the significant intermixing of two elements. Previous studies revealed the existence of good correlations among differences in atomic size between metal and Si atoms, the calculated free energy difference in forming amorphous phase, and critical and maximum amorphous interlayer thickness for many refractory metals and silicon (Ti/Si, Zr/Si, Hf/Si, V/Si, Nb/Si, and Ta/Si) and Y/Si systems (Chen et al., 1994). Pertinent data are listed in Table 1. It is seen that the differences in both size and electronegativity between Y and Si atoms are considerably larger than those between refractory metal and Si atoms. In addition, the free energy of mixing for the Y/Si system is large and negative. As a result, a-interlayer is expected to be more prone to form in the Y/Si system. Indeed, the Y/Si system is the only system found up to date among all metal/Si systems in which the a-interlayer can be grown to a thickness of 10 nm during deposition at room temperature. STRUCTURE OF AMORPHOUS INTERLAYERS HRTEM in conjunction with optical diffractometry has been used to analyze amorphous interlayer and identify the first nucleated phase in metal thin films on silicon systems. However, crystallites less than 1 nm in HREM OF AMORPHOUS INTERLAYERS 145 Fig. 12. (a) HRTEM image and (b) AES depth profile of the alternating a-Si and Y multilayer films with an atomic composition ratio close to 1Y:2Si. size may be embedded in the amorphous interlayer and undetectable with direct imaging technique in HRTEM. On the other hand, a correlation technique is a statistical analysis of images in real space. The autocorrelation function (ACF) is equivalent to a twodimensional form of Patterson function, well known in X-ray crystallography in that it represents a map of interatomic vectors (Fan and Cowley, 1985). The correlation functions were introduced into electron microscopy when it became clear from optical diffraction experiments that different electron micrographs of the same specimen can be aligned with high accuracy despite their apparent differences due to contamination, radiation damage, electron and photographic noise, and change in defocus (Fan and Cowley, 1985; Frank, 1980). ACF analysis has been applied to the high-resolution transmission electron microscope images of amorphous interlayers formed in the interfacial reactions of ultrahigh vacuum deposited V, Zr, and Mo thin films on (111)Si. The results demonstrate the usefulness of the ACF analysis of atomic images in amorphous interlayers formed in the initial stage of reactions in metal thin films on silicon. 146 L.J. CHEN ET AL. Fig. 13. (a) HRTEM image and (b) AES depth profile of an 1Y:2Si multilayer film annealed at 300°C for 30 minutes. In V/Si samples, an amorphous interlayer was found to form. Figure 16 shows an example of a sample annealed at 430°C for 120 minutes. The presence of a crystalline phase and/or a short-range order in the amorphous region is not evident. ACF analysis was carried out on templates of 1.2 3 1.2 nm2 with each pixel corresponding to 0.019 3 0.019 nm2 cut along the interlayer in the HRTEM micrograph. The templates were patched on double in size square regions with a homogeneous background that is set to be equal to the average value of the original intensity so that the ACF-processed images are free from wrap-around arti- fact. Periodic structures were observed in some of the ACF images that revealed the short-range ordering in the amorphous interlayers. Typical examples of ACFprocessed images are shown in Figure 17. The crosshatch pattern was found to correspond to V3Si and V5Si3 from both interplanar spacing and angular measurements. It was found that the ACF-processed image and the simulated image following Patterson function treatment match rather well. Examples are shown in Figure 18. The calculated Patterson function images indicated that the thickness of the region with a short-range order is less than 2 nm. Similar analysis HREM OF AMORPHOUS INTERLAYERS Fig. 14. HRTEM image of an as-deposited 1Y:1Si multilayer film. Fig. 15. HRTEM image of an as-deposited 5Y:3Si multilayer film. . 147 148 L.J. CHEN ET AL. Fig. 16. minutes. HRTEM image of an a-Si (20 nm)/V(30 nm)/(001)Si sample annealed at 430°C for 120 Fig. 17. ACF-processed images of the outlined regions shown in Figure 16. was carried out for the amorphous interlayer formed in the Zr/Si system. Zr2Si, ZrSi, and ZrSi2 were found to form (Chen et al., 1994). For Mo/Si system, Mo3Si was identified to be the first nucleated crystalline phase that is correlated to the stable structure of the amorphous Mo-Si alloy. Both Mo3Si and Mo5Si3 were found to form simultaneously under certain annealing conditions (Liang and Chen, 1994). FORMATION OF CRYSTALLINE PHASES The first phase nucleation is one of the long-standing problems in the interfacial reactions of metal thin films HREM OF AMORPHOUS INTERLAYERS Fig. 18. V5Si3. 149 (a) HRTEM image, (b) ACF-processed image, (c) pattern function, and (d) simulated image of Fig. 19. XTEM micrograph, HRTEM image of a sample annealed at 450°C for 1 hour. Inset is the optical diffraction pattern corresponding to [ 110]Ti5Si3. on silicon. Theoretically, difficulties were encountered in predicting the reaction product of a basically nonequilibrium process from equilibrium thermodynamics. Experimentally, conventional techniques, such as X-ray diffraction and electron diffraction, are more appropriate for the detection of the growth phase rather than for the unambiguous identification of the first nucleated phase owing to their limited sensitivity and resolution in the determination of the phase. For the Ti/Si system, as many as four different crystalline phases (Ti5Si3, TiSi, C49-TiSi2, and C54-TiSi2 ) were reported to form in thin film reactions. The close proximity and/or overlap- 150 L.J. CHEN ET AL. Fig. 20. Schematic diagram of formation of Ti-silicides in samples annealed at 475°C for 30–60 minutes or at 500°C for 10–20 minutes. ping of diffraction rings of these silicides rendered the unambiguous identification of phases during the initial stages of reactions rather difficult. As a result, although the final stable phase is known to be C54-TiSi2, the phase formation sequence is still unclarified. Ti5Si3, TiSi, and C49-TiSi2 were variously reported to be the first nucleated phase. HRTEM in conjunction with optical diffractometry (or fast Fourier transform technique) were used to identify the first nucleated phase in silicide formation. The combined techniques are capable of unambiguous identification of a phase as small as 1 nm in size. The sensitivity compared favorably with conventional selected area electron diffraction and microdiffraction, which correlate diffraction patterns with areas about 500 and 100 nm in size in the specimens, respectively. Since Ti is a strong oxygen getter, to minimize the influence of oxygen on the interfacial reactions in the Ti/Si system, UHV deposition of the thin films is considered to be most appropriate. In the Ti/Si system, Ti5Si3, located at the Ti/a-interlayer interface, was identified to be the first nucleated phase. Ti5Si3, Ti5Si4, TiSi, and C49-TiSi2 along with the amorphous interlayer were observed to be present simultaneously in samples annealed at higher temperatures. Fundamental issues in silicide formation need to be addressed in light of the discovery of the formation of the amorphous interlayer and as many as four different silicide phases in the initial stages of interfacial reactions of UHV deposited Ti thin films on silicon. An example and a schematic diagram are shown in Figures 19 and 20, respectively (Wang and Chen, 1992). Simultaneous occurrence of multiphases was also observed in the interfacial reactions of the Zr/Si, Hf/Si, V/Si, Cr/Si, Mo/Si, and Y/Si systems (Chen et al., 1994; Hsieh et al., 1993; Lee and Chen, 1993; Liang and Chen, 1994; Lin et al., 1996). CONCLUSIONS HRTEM has been fruitfully applied to investigate the formation of the amorphous interlayers in metal/Si systems. Amorphous interlayers were found to form in all refractory metal and a number of rare-earth and platinum group metal/Si systems. For Ti/Si, Zr/Si, Hf/Si, V/Si, Nb/Si, and Ta/Si systems, the growth of a-interlayer follows a linear law in the initial stage. The maximum thickness of amorphous interlayers and activation energy of the growth are correlated with the free energy of mixing, in particular to the difference in electronegativity between elemental metals and silicon for the metals in the same group in the periodic table and silicon systems. Si atoms were found to be the dominant diffusing species in the solid phase amorphization in Ti/Si, Zr/Si, and Hf/Si systems. For Y/Si system, the stability of amorphous interlayer depends critically on the composition of the amorphous films. Auto-correlation function analysis was utilized to determine the structure of the amorphous interlayers. Taking advantage of the distinct contrast between amorphous interlayer and crystalline phase, HRTEM, in conjunction with the optical diffractometry (or fast Fourier transform technique), as applied to determine the first nucleated crystalline phase. The simultaneous presence of multiphases was observed to occur in a number of refractory metal/Si systems. ACKNOWLEDGMENTS The research was supported by the Republic of China National Science Council through grant NSC84-2215E007-011. REFERENCES Abelson, J.R., Kim, K.B., Mercer, D.E., Helms, C.R., Sinclair, R., and Sigmon, T.W. (1988) Disordered intermixing at the platinum-silicon interface demonstrated by high resolution cross-sectional transmission electron microscopy, Auger electron spectroscopy, and MeV ion channeling. J. Appl. Phys., 63:689–692. Chen, L.J., Hsieh, W.Y., Lin, J.H., Lee, T.L., Chen, J.F., Liang, J.M., and Wang, M.H. (1994) Solid state amorphization in silicide-forming system. In: Mater. Res. Soc. Symp. Proc., 320:343–354. Cheng, J.Y., and Chen, L.J. (1990) Growth kinetics of amorphous interlayers by solid-state diffusion in Zr/Si and Hf/Si systems. J. Appl. Phys., 68:4002–4007. Cheng, J.Y., and Chen, L.J. (1991) Growth kinetics of amorphous interlayers by solid-state diffusion in ultrahigh vacuum deposited Nb and Ta thin films on (111)Si. J. Appl. Phys., 69:2161–2168. Fan, G.Y., and Cowley, J.M. (1985) Auto-correlation function analysis of high resolution electron micrographs of near-amorphous thin films. Ultramicroscopy, 17:345–355. Frank, J. (1980) The role of correlation techniques in computer image processing. In: Computer Processing of Electron Microscope Images. P.W. Hawkes, ed. Springer, Berlin, pp. 187–222. Herd, S.R., Tu, K.N., and Ahn, K.Y. (1983) Formation of an amorphous Rh-Si alloy by interfacial reaction between amorphous Si and crystalline Rh thin films. Appl. Phys. Lett., 42:597–599. Holloway, K., and Sinclair, R. (1987) Amorphous Ti-Si alloy formed by interdiffusion of amorphous Si and crystalline multilayer film. J. Appl. Physiol., 61:1359–1364. Hsieh, W.Y., Lin, J.H., and Chen, L.J. (1993) Simultaneous occurrence of multiphases in the interfacial reactions of ultrahigh vacuum deposited Hf and Cr thin films on (111)Si. Appl. Phys. Lett., 62:1088–1090. Johnson, W.L. (1988) Thermodynamic and kinetic aspects of the crystal to glass transformation in metallic materials. Progr. Mat. Sci., 30:81–134. Lee, T.L., and Chen, L.J. (1993) Interfacial reactions in ultrahigh vacuum deposited Y thin films on atomically clean (111)Si. J. Appl. Physiol., 73:8258–8266. Lee, T.L., and Chen, L.J. (1994) Interfacial reactions in ultrahigh vacuum deposited Y-Si multilayer thin films. J. Appl. Phys., 75:2007– 2014. Liang, J.M., and Chen, L.J. (1994) Auto-correlation analysis for the determination of the structure of amorphous interlayers in ultra- HREM OF AMORPHOUS INTERLAYERS high vacuum deposited molybdenum thin films on silicon. Appl. Phys. Lett., 64:1224–1226. Liauh, H.R., Chen, M.C., Chen, J.F., and Chen, L.J. (1993) Electrical and microstructural characteristics of Ti contacts on (001)SI. J. Appl. Phys., 74:2590–2597. Lin, J.H., and Chen, L.J. (1995) Growth kinetics of amorphous interlayers by solid-state diffusion in ultrahigh vacuum deposited V thin films on (001)Si. J. Appl. Phys., 77:4425–4430. Lin, J.H., and Chen, L.J. (1996) The determination of average concentration of amorphous interlayer in the V/Si system with an ultrathin oxide layer and capping Mo layer as reference planes. Appl. Surf. Sci., 92:340–344. Lin, J.H., Hsieh, W.Y., and Chen, L.J. (1996) Simultaneous occurrence of multiphases in the interfacial reactions of ultrahigh vacuum deposited V and Zr thin films on (111)Si. J. Appl. Phys. (in press). Lur, W., and Chen, L.J. (1989) Growth kinetics of amorphous interlayer formed by interdiffusion of polycrystalline Ti thin film and single-crystal silicon. Appl. Phys. Lett., 54:1217–1219. Miedema, A.R., Buschow, K.H.J., and van Mal, H.H. (1976) Which 151 intermetallic compounds of transition metals form stable hydrides. J. Less-Common Metals, 49:463–472. Nicolet, M.A., and Lau, S.S. (1983) Formation and characterization of transition-metal silicides. In: Materials and Process Characterization. N.G. Einspruch and G.R. Larrabee, eds. Academic Press, New York, pp. 329–464. Sinclair, R., and Konno, T.J. (1993) Reactions in metal-metalloid multilayers. Mater. Res. Soc. Symp. Proc., 311:3–8. Tu, K.N., and Mayer, J.W. (1978) Silicide formation. In: Thin FilmsInterdiffusion and Reactions. J.M. Poate, K.N. Tu, and J.W. Mayer, eds. Wiley, New York, pp. 359–405. Tu, K.N., Herd, S.R., and Goesele, U. (1991) Metastability in slow thin-film reactions. Phys. Rev., B43:1198–1201. Walser, R.M., and Bene, R.W. (1976) First phase nucleation in silicon-transition-metal planar interface. Appl. Phys. Lett., 28:624– 625. Wang, M.H., and Chen, L.J. (1992) Phase formation in ultrahigh vacuum deposited titanium thin films on silicon. J. Appl. Phys., 71:5918–5925.