The process of suicide formation by contact reaction at metal/Si interfaces normally involves rather uniform motion of the growth fronts which separate metal, suicide, and Si regions, as has been observed for suicide growth in many transition-metal/Si systems. At lower temperatures, however, the reaction behavior can be complicated significantly by the presence of other material reactions which may compete with interfacial suicide formation. For refractory metals, strong interfacial mixing over considerable depth (∼ 100 Å or more) is observed at temperatures too low for the normal inlerlacial suicide formation process to contribute; the highly nonuniform character of this reaction, as shown by ion scattering and TEM studies, suggests that other material reactions (e.g., grain boundary diffusion) must dominate the interfacial chemistry at low temperature. In a similar way, anomalous and nonuniform reaction behavior during the low temperature deposition of initial transition metal layers on Si apparently involves surface diffusion processes which are faster than inlerlacial suicide formation.

Significant understanding of the processes occurring at the interface between transition metal ultrathin overlayers and silicon has been achieved in recent years. In general these interfaces are highly reactive and the formation of an interfacial suicide compound takes place. The reaction at the near-noble metal-silicon interface can occur even at liquid nitrogen temperature. Furthermore the reaction products can give rise to epitaxial interfaces. How can the covai ent bonds in silicon be broken at such a low temperature and eventually give rise to an epitaxial interface are intriguing questions, which cannot be answered without a full understanding of the electronic and atomic processes involved in the interface reaction.

Theoretical investigation‘through band structure calculations allows us to understand the modifications of the suicide electronic states in the interface region. Assuming various interface geometries (e.g. epitaxial growth or interstitial interdiffusion) the effect of different atomic processes on the electronic properties is analysed and available photoemission and Auger spectroscopie experimental informations are interpreted. The role of both electronic states and atomic processes in determining the interface physical properties is discussed.

We have studied the formation of palladium suicide on clean silicon <111> surfaces using in situ spectroscopie ellipsometry. Pd films of thicknesses from 5 to 50 A were studied both as deposited and after annealing to complete the reaction to Pd2Si. The measured dielectric function spectra were analyzed to determine the amounts of Pd2Si and unreacted Pd. Evidence of Pd-Si solid state reaction was detected after all depositions. In the as-deposited films the fraction of suicide increased with the thickness, up to about 40% for the thickest films (50 Å Pd), while the annealed films were completely reacted. These results are consistent with earlier Raman spectroscopy studies.

Thin-film Ni-Si lateral-diffusion couples were annealed at 450°C for 12 hours to cause formation of the suicide phases: Ni3Si, Ni5-Si2, Ni2Si, and NiSi. Scanning transmission electron microscopy (STEM) was used to investigate the suicide growth. The mechanism of Ni transport to the growth front was shown to be bulk phase diffusion, as evidenced by a bulk phase concentration gradient of 0.03 atomica Ni/nm in the NiSi phase. The transition from thin film kinetic behavior to bulk couple kinetic behavior was found to be greater than 200 nm. The absence of the Ni3Si2 phase in the lateral-diffusion couple was attributed to a nucleation barrier. An excess amount of Ni in the Ni2Si phase at the Ni2Si-NiSi interface indicates the interdiffusivity is lower in NiSi than in Ni2Si, thereby limiting Ni diffusion out of the Ni2Si phase.

The existence of a TiSi2 phase in the C49 (ZrSi2) structure prior to formation of the stable low resistivity C54 TiSi2 structure has been justified for Ti-Si thin film diffusion couples after low temperature (500°C to 650°C) anneal. The transition temperature (Ttr) to the C54 phase is shown to be dependent on layer thickness, thinner films yielding higher Ttr. We could not detect an influence of oxygen on Ttr. Moreover the growth rate of C49 TiSi2 is independent of oxygen contamination in the Ti layer up to a mean concentration of 23 at.%. It is postulated that the transition from the C49 to the C54 phase actually is a recrystallization process.

The reliable use of TiSi2 in integrated circuit metallizations will depend in part on an intimate knowledge of the effects of processing variables — such as annealing time, temperature, and ambient — on the phase formation sequence. For this reason, transmission electron microscopy has been used to investigate the formation of TiSi2 thin films on silicon substrates [1]. For films formed either by reacting titanium with a silicon substrate or by sintering a codeposited (Ti + 2 Si) mixture, we find that a high resistivity (∼60 μΩcm), metastable phase — TiSi2(C49 or ZrSi2 structure) — forms prior to the desired low resistivity (∼15 μΩcm), equilibrium phase — TiSi2(C54 structure). In titanium-silicon diffusion couples, a thin layer of TiSi is also present on top of the metastable TiSi2- For processing temperatures above 550°C, the available data suggest that the metastable TiSi2 forms first and acts as a template for subsequent nucleation of the TiSi phase. In codeposited (Ti + 2 Si) films, TiSi2(C49 structure) is the only intermediate phase. The temperature at which the C49 structure transforms to the C54 structure increases significantly as the film impurity content increases. Differences in earlier reports of TiSi2 formation [2–5] appear to be resolved if x-ray diffraction peaks attributed to Ti5Si3 and TiSi were actually from the metastable TiSi2

Interactions of evaporated Ni and Si thin films were investigated by a combination of backseat tering spectrometry and transmission electron microscopy. The presence of amorphous Si has no significant effects on Ni2Si and NiSi formation, but it drastically lowers the formation temperature of NiSi. Experiments with evaporated thin markers established that Ni is the dominant diffusing species in the growth of the three suicides. The stability of NiSi was examined by sequential evaporation of Ni34Si66 and Ni50Si50 thin films both on Si(100) and on evaporated Si substrates. The results showed that NiSi2 grows at the expence of NiSi when the stucture is in contact with evaporated Si, while it dissociates into NiSi and Si when in contact with single crystal Si.

Films of WSi, have been investigated by electron microscopy after annealing at different temperatures. It is shown that films annealed at 550°C contain a high density of stacking faults. The crystalline identity of these stacking faults has been established. Undoubtedly the high density of stacking faults is the cause of the resistivity maximum observed in films annealed at 550°C.

Epitaxial ruthenium, osmium, rhodium and iridium suicides have been successfully grown on silicon. Electroless chemical plating was used to deposit platinum-group metal thin films on silicon. Two step annealing was found to be effective in inducing the growth and improving the quality of the epitaxial suicide on silicon.

Transmission electron microscopy was applied to characterize the microstructures and determine the orientation relationships between epitaxial suicides and substrate Si. The compositions of deposited films were determined by scanning Auger electron spec-troscopy combined with depth profiling technique. The percentages of phosphorus were found to be in the range of 2–3 at. %.

After the review of dopant redistribution phenomena observed during formation of near noble metal suicides, we describe the results of our recent experiments to get a better understanding of a mechanism of the dopant redistribution phenomenon in Si substrates. The key factors to understand the dopant redistribution are dopant segregation at the suicide/ Si interface due to lower solubility limit of dopants in suicides, enhanced diffusion of dopants into the Si substrate at much lower temperatures than the ordinary thermal diffusion, and electrical activation of the redistributed dopants. The results of As and carrier concentration measurements before and after Pd2Si formation to make clear the third factor show that the electrical activity of the redistributed As atoms in Si is strongly dependent on the initial activity before Pd2Si formation which is controlled by the temperature for the pre-annealing of As implanted Si.

Shrinkage of extrinsic dislocation loops introduced by As implantation and subsequent annealing have been observed after Pd2Si formation, which is a good evidence of vacancy generation during Pd2Si formation. The role of the vacancies and interstitials on the second factor, the enhanced diffusion, has also been discussed. Finally we list a few issues to be answered in future by more detailed works in order to get a complete understanding of the redistribution phenomenon.

We have investigated the growth kinetics of platinum suicide compounds on undoped and heavily arsenic doped polycrystalline silicon. The results are compared with published data on the growth kinetics of these suicides on other silicon substrate materials. From this comparison we conclude that dopants and the microstructure of the substrate material do not affect the activation energy of the growth kinetics which means that the growth mechanism of these suicides is not altered. Only the preexponential factor is lowered in some cases, pointing to a slowing of the speed of growth.

Thermal stability of self supporting amorphous films is reviewed in terms of a simple kinetic model. A semi-empirical relationship between the crystallization temperature and the heat of hole formation is applicable only for systems with negative heat of mixing.

Interaction of amorphous diffusion barriers with Si substrates and Al overlayers is also reviewed for near noble-refractory alloys. If their crystallization temperature is higher than their reaction temperature, there is no correlation between the two parameters. The reaction takes place when the annealing temperature reaches a value at which compounds can be formed of both constituents of the alloy. Finally, comparison is made to reactions of crystalline alloys with Si.

High resolution core level photoemission results show the temperature evolution of the V/Ge(111) interface in the range from 300 to 600 K. Three well-defined chemical environments are present for Ge at 300K (the first is the substrate and the other two are reaction products with overall shifts of−0.5 and −0.95 eV). Increasing the temperature enhances Ge outdiffusion, and a homogeneous reacted layer forms when deposition and measurements are done isothermally at 475K. The activation energy for this diffusion process is very low (5 kcal/mole), indicating the importance of grain boundary diffusion at reacting, heterogeneous interfaces.

For modelling the reaction of chemically distinct materials in which intermediate alloy phases are formed a simple thermodynamic description is not adequate. Despite thermodynamic prediction of multiple product phases, a single phase generally forms first which is not necessarily that of greatest thermodynamic stability or of simplest structure. Such initial reaction processes may be modelled as metastable perltectoid (solid-solid) or perltectlc (solid-liquid) reactions, characterized by large thermodynamic driving forces with superimposed kinetic and morphological constraints. The interfacial reaction problem is reviewed in light of heterogeneous nucleation theory with emphasis on non-classical aspects.

Thin film interaction in the material systems of titanium and tantalum-layered Al- 1% Si films has been studied using advanced material characterization techniques. Compound phase formation was analyzed based on results from X-ray diffraction analysis. Al3Ti was identified in the titanium-layered film after 450° C sintering. TaSi2 and unreacted Ta. however, were identified after the sintering. Silicon participated in the phase formation process by incorporating itself as a solute in the Al3Ti while transforming Ta into its silicide. This behavior is to be interpreted based on available ternary phase diagrams. Cross-sectional transmission electron microscopy results indicate that well defined layered films and columnar grains of aluminum were obtained which are believed to be the primary reasons for the observed electromigration improvement over the base metal.

The crystallization of α-Si in α-Si(50nm) and Au(5nm) thin film bi-layers has been investigated during heat treatment in a transmission electron microscope (TEM). When crystallization of α-Si first begins at 130°C the Au-Si alloy reflections observed at lower temperatures vanish and several new reflections from metastable Au-Si compounds occur. Dendritically growing islands of poly-Si are observed after heating at 175°C. If the samples are held constant at 175°C for ten minutes, the poly-Si islands grow together. The formation of poly-Si depends on the diffusion of Au into α-Si and the formation of metastable Si-Au compounds. After crystallization Au segregates to the front and back surface of the poly-Si film.

The development of rapid thermal annealing (RTA) or sintering (RTS) techniques for processing of Si and compound semiconductor metallizations also opens a new avenue for investigations of the initial stages of high-temperature, solid state reactions in thin amorphous multilayers. Of particular interest in this regard are the nucleation and early growth of silicides. In this study we suggest a method to prepare large numbers of specimens for TEM, using deposition of Si/metal/Si tri layers on electron microscope grids and annealing the grids in an RTA system. The method enables quick determination of first silicide phases, nuclei shape and density, relative nucleation and growth rates and, for some systems, the diffusing species. Seven widely used metals are investigated: Co, Cr, Nb, Ni, Mo, Ti, and Ta. The first silicide phases determined from electron diffraction patterns are listed and compared to first phases reported in longer furnace annealings. TEM cross sections of reacted Si/Ta/Si and Si/Nb/Si show two different growth morphologies, which are discussed in terms of main diffusing species in a proposed model.

We have investigated the interactions of Al/Ni, NiAl3/Ni, and Ni2Al3/Ni thin films between 350 and 500°C. Sequential layer deposition and coevaporation were used to prepare the films which were then vacuum annealed and analyzed by Rutherford backseat tering and X-ray diffraction. Initially, NiAl3 is formed, followed by the more Ni-rich phases. All four intermetallic phases were formed.

Interactions between amorphous metal thin films and either a substrate or an overlayer can limit their effectiveness as diffusion barriers. We have found in previous studies that Au and Al polycrystalline thin films in contact with amorphous W-Si lowers the crystallization temperature of the a-(W-Si) by at least 100C. In contrast Cu and Mo have no apparent effect on the stability of the amorphous layer. The mechanisms leading to premature crystallization are not well understood. Amorphous W .72Si.28 was deposited by D.C. sputtering onto single crystal Si substrates. Overlayers of Al were then evaporated onto the W-Si. Using Auger electron spectroscopy depth profiling coupled with cross-section TEM, we have studied interfacial reactions between the amorphous layer and polycrystalline Al. Auger profiling results show that in the case of Al overlayers, W and Si diffuse out of the a-(W-Si) into the Al where WAl12 forms. These results can be explained in the context of three binary diffusion couples, W-Si, W-Al, Al-Si, and the individual interactions associated with these couples.

The interaction of Ti with A12O3 under UHV conditions has been studied by AES and UPS. Ti was deposited by iterated and successively thicker evaporations (up to a total thickness of 91 Å) under UHV conditions onto the alumina substrate at room temperature. The oxide substrate was grown on Al/Si(111) in an adjacent VHV preparation chamber. The last deposition was followed by in situ annealing. It was found that, contrary to indications of thermodynamic considerations using bulk equilibrium data, Ti interacts strongly with alumina at room temperature, dissociating it to metallic Al and forming TiO2 at the interface. This reaction is limited, resulting in an interface width of ∼10 Å and allowing subsequent evaporations to homogeneously cover the products. Annealing to 500°C does not produce any evidence of outdiffusion or extensive interface reactions, implying that the interface oxide is stable and an effective barrier to Al (and/or Ti) diffusion up to this temperature.