THEORETICAL INVESTIGATION OF THE ATOMIC GEOMETRIES AND ELECTRONIC STRUCTURES OF RECONSTRUCTED SEMICONDUCTOR SURFACES

Abstract

Our aim in this work is to study the electronic structures and the optimum reconstructed surface structures, specially on the cleavage Si(001), Ge(001), GaAs(110), ZnSe(110) and the non-cleavage InP(001) surfaces using Chadi' s total energy algorithm and the self-consistent CNDO method. The Chadi's total energy algorithm consists of two terms: the band structure energy (Ebs) and the short-range-force-constant term (U). Two tight binding theoretical models, namely the sp3 and the sp3s* models are used to calculate the band structure energy through the introduction of the Chadi' s special-k-point scheme. With only the first-near-neighbour interactions considered, the Hamiltonian secular matrix of sp3 is of order eight, corresponding to ones and three p orbitals on each atom in the diamond crystal primitive cell; whereas that of sp3s* is of order ten with an excited s-state included on each atom. Atomic interactions are described in terms of tight binding parameters. To testify the accuracy of the sp3 tight binding parameters obtained by Chadi et al, we have calculated the band structures of some wellknown semiconductors and found that in general the valence bands of most crystals are well described by this model. The band structures of Si and Ge are also plotted using the sp3s* model, of which the conduction bands are more accurately reproduced. Chadi' s special-k-point total energy algorithm is validated through the calculation of some bulk properties of semiconductors. Results obtained show remarkable agreement with experimental findings. In applying the tight binding models in surface studies, a crystal system with surfaces is simulated with a two dimensional slab model of finite atomic layers. The tight binding surface bands calculated for Si and Ge (001)-2x1 asymmetric dimer structures show semiconducting features. Comparatively, the sp3s* surface band dispersions are more consistent with the experimental data. Si(001) surface bands of other studies all indicate semiconducting features of the surface. For Ge(001), a transition of metallic-insulating feature was observed in the ARP experiment when the surface temperature was reduced from room temperature to 77K. As our theoretical models are built on the zero-temperature basis, findings of semiconducting feature of Ge(001) surface suggested that at lower temperature, Ge(001) is actually non-metallic. Seven possible asymmetric dimer structures on Si (001) and Ge(001) are examined with the sp3 model, and the stable structures arc found to be c(2x2), c(4x2), p(2x2) and 2x1. The Si(001) surface is reinvestigated using CNDO method and c(4x2), 4x1, p(2x2) and 2x1 structures are found favored. Both Si(001) results agree with experimental data in advocating the existence of c(4xZ), p(2x2) and 2xl structures. However, there is a lack of experimental evidence on the appearance of c(2x2) and 4xl structures. The CNDO method is thus shown to be an alternative approach for surface study where cluster model of finite size and finite number of atoms is employed. In the case of compound semiconductors, the rippled structures of cleavage GaAs(110) and ZnSe(110) surfaces are confirmed through the use of the sp3 model. The noncleavage polar surface of InP(001) is more of our concerned. The tight binding parameters used to account for the interaction between In and In at the InP(001) surface are determined from the Harrison universal parameter scheme. Both p(2x2) and c(4x2) structures are found to be energetically stable and have surface states similar to experimental results obtained from UPS study.