Nucleation is generally understood to refer to the initial process of creating a new phase, for example, liquid water, from another phase, water vapor, as a result of a temperature excursion which carries the system through some critical temperature (in the case at hand, at normal atmospheric pressure, 100°C). The processes involved in such transformations have been the object of careful observations for a long time, at least two centuries; a very good historical perspective on this subject is found at the beginning of Ref. 1. The special case of metal vapors condensing on substrates in order to form thin films is considered in all texts dealing with thin film preparation [2]. The precipitation of a new phase, such as CuAl2, from a supersaturated solution (of copper in aluminum) has immense technological importance which motivates the attention being paid to nucleation phenomena in all metallurgical treatises [3,4].(A particularly thorough discussion is found in Ref. 5.) In all cases, nucleation phenomena are controlled by the necessity of creating an interface between the nucleated phase and the original matrix. The energy of the interface imposes special thermodynamic constraints on the evolution of the system.

The initial phase formation at the interface of the near noble metals (Ni, Pd, and Pt) and Si is studied by Raman spectroscopy and Rutherford ion-backscattering. The results show that a crystalline Pd2Si forms immediately at the interface of Pd and Si while a disordered intermixed phase of composition ∼M2Si forms for Ni and Pt at 150°C. The results are discussed in terms of the disorder due to the kinetic process and the stability of the crystalline phase.

An effective heat of formation (δH') is formulated, enabling δH' to be calculated as a function of the concentration of metal and silicon available for interaction to form a silicide. A rule for first phase formation is proposed according to which, “The first silicide compound to form during metal-silicon interaction is the congruent phase with the most negative effective heat of formation at the concentration of the lowest eutectic temperature of the binary system.” The effective heat of formation concept is also extended to predict subsequent phase sequence.

Twenty years of research have now been devoted to investigating reaction products obtained by annealing metal-layer/silicon structures. A wide variety of cases have been analyzsed and a considerable amount of data has been produced. Despite the vast amount of information available, several aspects concerning phase formation and kinetic processes are not yet well established. The purpose of this paper is to investigate the mechanisms of phase formation and to show the importance of kinetic factors in the appearance of various compounds. Results will be shown for a single metal layer deposited on silicon, for bilayers. and for alloys. Depending upon the starting structure, metal-rich or silicon-rich silicides can be formed. Moreover, by modifying the boundary conditions, it is possible to change the growth kinetics of the silicide phase that forms.

We report on a study of thermal interactions between pairs of Ni, Cr, and Pt films and their silicides on a Si substrate. For each pair of metals M1, M2, four systems were investigated: Si−M1 −2, Si−M2 −M2, Si−M1 Si−M2 and Si−M2Si−M1 giving altogether twelve initial combinations. Samples were annealed in the temperature range of 300−700°C and the analysis was made by Backscattering Spectrometry (BS). It is found that double layers of uniform silicides can be grown when the Cr silicide is the top layer. The rate of CrSi2 formation in this case is less then that reported for CrSi2 growth on Si substrate. In all other cases we find some intermixing. Impurities are found to effect the growth kinetics and the layer morphology.

Gadolinium silicide with its attractive features of low formation temperature of about 350°C and low Schottky barrier height on n-type single-crystal silicon substrates (ϕnB1∼O.4ev,ϕpB ∼ 0.7ev) was chosen for studying the feasibility of forming shallow uniform contacts. Samples with various compositions prepared by both bilayer evaporation with a configuration of Si(α)/Gd/Si(xtl) and coevaporation with a Si−Gd /Si(xtl)structure were used for studying the contact formation as a function of composition and heat treatment. We found that shallow contact formation can be achieved provided that the following conditions are met: (a) for bilayer evaporation, the atomic ratio of Si(α)/Gd ≥ 2 should be maintained, (b) for coevaporation, the Si to Gd atomic ratio between 1.7 and 2.0 is desired. The bilayer deposition scheme appears to be a more convenient technique to use in practice.

Interdiffusion in the Si<100>/Pd2Si/Ni and Si<111>/Pd2Si/Ni thin film systems has been investigated using Rutherford backscattering spectrometry. Nickel is found to diffuse along the grain boundaries of polycrystalline Pd2Si upon which it accumulates at the Si<100>Pd2Si interface. The high mobility of Ni compared to that of si suggests that Pd diffuses faster than Si along the Pd2Si grain boundaries. An activation energy of 1.2 eV is determined for Ni grain boundary diffusion in Pd2Si.

Evaporated W, implanted Xe, and implanted 18O were used as markers to study the dominant moving species during (a) solid phase epitaxy (SPE) of evaporated Si, (b) silicide formation, and (c) oxidation of silicides on Si substrate.

MeV 4He+ backscattering spectrometry and 18O (p, α)15 N nuclear reaction were used to monitor the evolution of elemental profiles as well as the change in the marker position. In most cases, the dominant moving species in SPE is the same as that observed in the formation and oxidation of that silicide. However, in CrSi2 the dominant moving species is Si during silicide formation, but Cr during SPE or oxidation.

Rutherford backscattering has been used to study metal/disilicide thin film interactions for Ni and Co. Upon heating, the metal first reacted with the disilicide to form Ni2Si and Co2Si, respectively. After complete consumption of the free metal, the monosilicide phase was found to form at temperatures between 350°C and 550°C. In the case of Co it was found that after all the metal had been converted to CoSi in this way, the reaction stopped. In the Si<>/NiSi2/Ni system, however, all the disilicide converted to NiSi, even though the thickness of the deposited metal was insufficient to account for this. For this to occur, the disilicide had to dissociate into NiSi and Si, with the excess silicon regrown epitaxially on the silicon substrate.

The stability of NiSi2 under various boundary conditions was investigated to determine the factors affecting dissociation. Partial dissociation was found to occur when the NiSi2/Ni reaction proceeded on an inert SiO2 substrate. The disilicide was stable, however, in the Si<>/NiSi/NiSi2 structure. Argon sputtering at temperature was found to induce complete dissociation of NiSi2 on single-crystal Si. We believe that the NiSi2 instability is due to the very small heat of formation from the monosilicide. In such a case, the thermodynamic driving forces are small enough that the reaction can be significantly influenced by the presence of kinetic barriers.

Refractory metal silicides have been shown to form stable Schottky barriers on GaAs up to an annealing temperature of ∼850°C. In this study, the metallurgical and electrical stability of near noble metal (Pd and Pt) silicide contacts to GaAs have been investigated. It is observed that Pd and Pt silicides are metallurgically more stable than Pd and Pt alone on GaAs. A correlation between the stability of silicide contacts and the heat of formation of the silicides normalized to per metal atom is found, thus allowing the prediction of contact stability of other silicides on GaAs. We have also investigated the feasibility of using solid-phase epitaxy (SPE) to grow a highly doped Ge epitaxial layer on GaAs to form a non-alloyed ohmic contact. This approach of forming a Ge/GaAs heterojunction alleviates the high vacuum requirement of MBE techniques. Our preliminary results indicate that relatively low contact resistivities can be obtained by SPE using the Ge-Sb-Pd system.

The presence of a 500 Å Ge layer in Au/Ge/Au ohmic contacts to GaAs was found to hinder the agglomeration of Au upon heating. An ohmic contact as uniform as the asdeposited metallization was achieved after heating at 300°C. Heating at 350–500°C resulted in various morphological entities in a Au-rich matrix. The microstructure changed from one with a Au-rich matrix to one with a GaAs-rich matrix and dendritic Au-rich islands at 525±25°C. Heating at 400°C or above resulted in Au-Ga phases with the epitaxial relationships α' (110) ‖ GaAs (100) and Au2Ga (112) ‖ GaAs (100).

Silicide formation and growth kinetics have been investigated with lateral diffusion couples formed by deposition of Ni and Cr layers on patterned Si substrates and by deposition of Ni patterns on Si films. For annealing temperatures between 520 and 650°C the growth of CrSi2follows a (time)½ dependence with an activation energy of 1.4± 0.1 eV. In Ni-silicide formation at temperatures below 600°C, Ni was the predominant moving species. As the temperature increased, the motion of Si became significant. The apparent activation energy for silicide formation varied from Ea ≅ 1.4 eV for Ni motion at relatively low temperatures to Ea≅ 2.3 eV for Si motion that occurs at high temperatures. Lateral diffusion in device geometry structures resulted in degradation of contact planarity due to the penetration of metal silicides in Ni-Si structures or the erosion of silicon in Cr-Si structures.

The reaction of a metal film with polycrystalline silicon to form a metal silicide has been shown to occur very rapidly when using cw lamp annealing in contrast to conventional furnace annealing. The faster reaction kinetics implies a more efficient energy coupling (radiative heating) versus the conductive/convective heat transfer processes which dominate furnace annealing. From Rutherford backscattering determinations of the thickness of PtSi formed as a function of anneal time (t), we have found the relationship to be linear in t with growth rates in the rangg 10−7–10−6 cm/sec (10 – 100Å/sec) for the temperature range 375 – 450° C. From these data the activation energy for the formation of PtSi was calculated to be 1.8± 0.2eV for cw lamp annealing, in reasonable agreement with literature values from conventional furnace annealing experiments.

ErSi2 contacts formed by reacting Er with single crystal silicon using conventional furnace heat treatment are dominated by pits. Recently rapid electron and laser beam heating have been used to form pit-free ErSi 2 layers on Si by reacting Er with the Si substrate. For this investigation we studied the electrical properties of erbium silicide/Si<p,100> contact prepared by thermal, laser and electron beam annealing. We prepared and annealed three types of samples; (I) Er(600Å)/Si<p,100>, (II) Si(100Å)/ Er(600Å)/Si<p,100> and (III) Si(900Å)/Er(600Å)/Si<p,100>.The barrier height of the pit-free thermal annealed samples (type III) was ∼0.78 eV. All laser and e-beam annealed samples were observed to be pit-free. The barrier heights for the laser annealed samples varied from ∼ 0.63 eV for type I samples to ∼ 0.77 eV for type III samples. The e-beam annealed samples gave barrier heights ∼ 0.71 eV (type I and II) and ∼ 0.77 eV for type III. The barrier heights of beam processed diodes shifted towards the 0.78 eV range upon post-annealing. We model these results in terms of defects created at the ErSi2-Si interface.

Titanium disilicide was formed by multiply-scanned electron beam irradiation of titanium films of nominal thickness 1200Å on silicon substrates. Samples were annealed at power densities of 2 to 52.5Wcm−2 using times in the range of 1 to a few hundreds seconds. Rutherford backscattering analysis was used to study the metal redistribution and to estimate the approximate compositions and thicknesses of the films. Compounds were identified by X-ray and electron diffraction. Sheet resistance was measured by the four probe technique and surface topography inspected by scanning electron microscopy.

The silicide thickness achieved depends only on annealing time for power densities in the range of 20 to 50Wcm−2 and hence is independent of heating rate and peak temperature during the heating cycle.

A multiply-scanned electron beam was used to form nickel silicides from thin nickel films on unimplanted and arsenic implanted silicon (100) substrates. Films of thickness 300, 660 and 1000Å were annealed using a power density of 7.9Wcm−2 and irradiation times of 5 to 150 secs.

Rutherford backscattering and channelling techniques were used to determine the stoichiometry, thickness and epitaxial quality of the annealed films. The thinnest gave a Xminof ∼ 18% and the thickest a Xmin of ∼ 40%. Resistivities were measured by the four-probe technique. Chemical composition was determined by X-ray and glancing angle electron diffraction (RHEED). Transmission electron microscopy (TEM) was used to estimate grain size of the disilicide phase and scanning electron microscopy (SEM) to study surface textures. The effect of arsenic implantation on growth rate, epitaxial quality and surface texture is demonstrated.

Thin film interactions often show strong dependence on the annealing ambients used. The outdiffusion of a large number of metals and semiconductors through an overcoating gold film and subsequent formation of surface oxides have been observed to be greatly enhanced using a steam or air ambient compared with those using vacuum or inert ambients. The ambient effects become more puzzled when the overcoating metal consists of gold and silver layers, depending on the sequence of these layers. These ambient effects are shown to be well correlated by a surface potential model which considers both the relation between the overcoating metal and the underlying diffusing species, and the effects of ambient on the surface properties of the overcoating metal at the interface. Further ambient effects have been predicted and confirmed, including reduced interactions by oxygen for many systems with different overcoating metals. Different interaction mechanisms have also been observed, including competing ambient effects, competing interactions, and preferential reactions, all affected by the ambients used. Understandings of the various ambient effects not only provide important insights into the interface interactions in general, but also have great impact on metallurgical contacts in device technologies. Future studies for the ambient effects are also discussed.

Tantalum, being a refractory metal, is sensitive to ambient impurities when forming a silicide. By introducing nitrogen and oxygen impurities into a tantalum-silicon system, interesting chemical and physical effects are observed in their subsequent reactions. Nitrogen and oxygen behave similarly in such a system. If initially present in Ta, they segregate into the still unreacted Ta as the silicide layer grows at a somewhat retarded rate. The same impurities, initially present in Si, are immobile in the form of stable compouis and suppress TaSi2 growth. The rare isotopes 15N and 18O are introduced bY implantation and Profiled by 15N(P,α)12C and 18O(P,α)15N nuclear reaction analyses, respectively. In addition, unintentionally incorporated 18O is checked by the 16O(d,α) 14N nuclear reaction. The results are explained in terms of the moving species Si, and of the chemical affinity, solubility and diffusivity of the impurities in their host lattice.

18O was implanted in the Si substrate or in the Co overlayer to study the effect of oxygen and its redistribution during Co silicides formation. Implanted and unimplanted samples were annealed in vacuum at 450°C for different durations. Over the investigated dose range of 0.5 to 2 × 101618O/cm2 in the Co film, the phases formed upon annealing are Co2Si and CoSi, as for the unimplanted samples. The kinetics of the silicide growth is changed little by the presence of oxygen in the Co film. Oxygen accumulates at the Co2Si/Co interface. When the oxygen is implanted in the Si substrate, the higher the oxygen dose is, the slower (fasteil the growth rate of CoSi(Co2Si) becomes. Below 1 × 1016 O/cm, both phases (i.e. CoSi and Co2Si) still form; above 2 × 1016 O/cm2 , only Co2Si formsWhen two phases form, the oxygen moves deep into the Si substrate; when only Co2Si forms, the oxygen moves toward the surface.

The oxygen behaviour and its influence on the annealing properties of the TiO2/Si and Ti/TiO2/Si systems have been investigated. For the TiO2/Si system no reaction at all could be evidenced after vacuum annealing up to 900°C for 30'. In the Ti/TiO2/Si system metallic Ti reacts with the TiO2 film above 400°C and at 600°C a uniform oxygen solid solution at the solubility limit was obtained without any Si reaction. Silicide formation occurs for annealing temperatures higher than 650°C and causes oxygen expulsion from the reacted layer and consequently a rise in its concentration at the surface where Ti oxide precipitation takes place. This surface oxide layer prevents a further growth of the silicide up to 850°C. The reaction of the whole metal film is attained only by annealing at 900°C or above, when the oxide is completely reduced and an important oxygen loss takes place. A model explaining this behaviour is proposed.