No abstract

We have investigated the epitaxial growth and the electronic properties of a metallic metastable FeSi phase crystallizing with the CsCl structure on Si(l l l). Upon annealing below 500'C the stoichiometry of thin films ( (20 A) evolves towards FeSiz with no change of symmetry, i.e., the defect CsCl structure with a statistical occupation of metal sites remains epitaxially stable for all FeSi~q"(0~x~I). Films thicker than -20 A exhibit a transition to the cubic e-FeSi phase. The electronic band structure of FeSi (CsCI) has been calculated self-consistently using the full-potential linear augmented-plane-wave method.The epitaxial growth of an overlayer whose lattice constant ai deviates little from the substrate lattice constant an is well under control experimentally and well understood theoretically. For a misfit f =(a~an)/an of typically a few percent one expects an epitaxial film to grow coherently, i.e. , homogeneously strained with its lateral lattice parameter at perfectly matched to an. At some critical thickness h, the strain energy stored in the film becomes too large and coherency breaks down due to the formation of misfit dislocations. ' While the calculations of critical layer thickness for systems with small misfit make explicit use of linear elasticity theory, ' this is no longer justified for rnisfits above, say 10%. The overlayer may then adopt a crystal structure which is lattice matched well to the substrate but which differs from the usual bulk form. Examples of such systems are a-Sn deposited on lnSb or CdTe (Ref.3) and bcc Co on GaAs.The theory for such structural changes in epitaxial overlayers was developed in Ref. 5.In this paper we report on the successful synthesis of FeSi with the metastable CsC1 structure by molecularbeam epitaxy (MBE). The corresponding bulk phase, e FeSi, of this material is simple cubic and has a lattice constant of 4.4 A (Ref. 6) to be compared with the Si lattice parameter of 5.43 A. This latter phase also was previously found to grow epitaxially on Si (111) with (111) e FeSill(111) Si and (011)FeSill(211) Si. For these crystallographic directions the mismatch amounts to f = -6.4% at room temperature (RT). The geometric mismatch is not, however, the only factor relevant to epitaxy. The atomic arrangements in the (111) plane of e FeSi and of Si are vastly different. We hence have to expect the interfacial energy to be significantly greater for this epitaxial couple than for the simple overlayer/substrate couple we are going to discuss. FeSi was grown on clean (111) Si 7x7 substrates (n doped, 1-2000 0 cm) with parallel monolayer steps due to the unintentional misorientation of 0.1-0.3 . Details of the substrate preparation procedure can be found in Ref. 8. The silicide was grown at RT ( (100 C) by depositing 2 monolayers (ML) of pure Fe followed by the codeposition of Fe and Si at a stoichiometric ratio (1:1)until the desired thickness of 5-80 A was reached. The 1&&1 reflection high-energy electron-diffraction (RHEED) pattern exhibited well-defined Kikuchi bands of ...

First-principles and empirical pseudopotentials are used to study the effects of short-range and longrange atomic order on the electronic properties of III-V semiconductor alloys. The alloy structure with a given degree of long-or short-range order is modeled by two types of supercells: (a) Small (16 -32 atom) supercells are constructed in the fashion of the special quasirandom structures (SQS) used previously to simulate random alloys [A. Zunger et al. , Phys. Rev. Lett. 65, 353 (1990)]. Their electronic structure is treated via first-principles pseudopotential methods. (b) Large ( -1000 atom) supercells are found by a simulated-annealing technique which optimizes the atomic configuration until a given degree of short-range order is reproduced. The electronic structure is then determined using the empirical pseudopotential method. Statistical tests prove that the small cell SQS mimic the much larger supercells and thus provide an efficient means of studying the electronic band structure of disordered alloys in a non-mean-field approach. For the direct band gaps of ideally random Al& "Ga"As, Ga& In"P, and Al&, In As alloys, we find optical bowing parameters b=0.48, 0.46, and 0.52 eV, respectively. In the presence of short-range order in the form of cation clustering, we find the following: (i) Clustering elongates the Ga-P bond and shortens the In-P bond in Gao &Ino 5P and (ii) the optical bowing of the direct band gap is greatly enhanced. This leads to an indirect-gap to direct-gap crossover in Alo &Gao 5As with sufficient clustering. (iii) The band-gap reduction is accompanied by a localization of band-edge wave functions on certain types of clusters. The clusters act as "isoelectronic impurities" which localize states if their concentration (i.e., the degree of short-range order) is large enough. Electrons at the conduction-band minimum localize on the cations with lower s-orbital energies. The bandgap reduction and wave-function localization of alloys with short-range order is compared to the effects of long-range order, where the gap reduction is due to level repulsion between zone-folding conduction states. Numerical results are given for CuPt-type long-range order of A1GaAs2, GaInP"and A1InAs~. For complete ordering, the band-gap reduction relative to the random alloys are 0.36, 0.49, and 0.16 eV, respectively.

No abstract