Meetings & Events

Publishing Alliance

MRS publishes with Springer Nature

 

fall 1997 logo1997 MRS Fall Meeting & Exhibit

December 1 - 5, 1997 | Boston
Meeting Chairs:
 Harry A. Atwater, Peter F. Green, Dean W. Face, A. Lindsay Greer 
 

Symposium R—Tight-Binding Approach to Computational Materials Science

Chairs

Luciano Colombo, Univ di Milano
Antonios Gonis, Lawrence Livermore National Laboratory
Patrice Turchi, Lawrence Livermore National Laboratory 

Symposium Support 

  • CNR-GNSM 
  • ENEA 
  • Hewlett Packard Laboratories 
  • Lawrence Livermore National Laboratory 
  • MPI-Stuttgart 
  • Office of Naval Research 
  • Univ. of Milan, Dept. of Materials Science

* Invited paper

SESSION R1: FIRST-PRINCIPLES TIGHT-BINDING: PART I 
Chairs: Christian Mailhiot, Didier Mayou, Dimitrios A. Papaconstantopoulos and Alain Pasturel 
Monday Morning, December 1, 1997 
Fairfax A (S)

8:30 AM *R1.1 
THIRD-GENERATION TB-LMTO. O. K. Andersen, G. Krier, R. W. Tank, C. Arcangeli, T. Dasgupta, O. Jepsen, Max-Planck-Institut FKF, Stuttgart, GERMANY

Progress in our development of a third-generation Linear Muffin-Tin-Orbital method will be reported. The new basis set is correct to first order in energy, not only inside the MT-spheres, but also in the interstitial region. This has the following important consequences: (i) The Hamiltonian and overlap matrices are expresssed solely in terms of the screened KKR matrix and its first three energy derivatives at the energy chosen (Enu). (ii) The simple ASA formalism, e.g. the expansion of the Hamiltonian in the orthonormal representation in powers of a simple two-center Hamiltonian, now includes downfolding and the combined correction. (iii) Downfolding to few-band, orthonormal, low-energy Hamiltonians works exceedingly well, as will be demonstrated for HTSC materials. (iv) The MT-spheres may now overlap so much that there is no need for empty spheres. This opens the way for using the TB-LMTO-ASA method for ab initio molecular-dynamics calculations. Finally, we have used third-generation LMTOs also for efficient expansion of the charge density and solution of the Poissons equation.

9:00 AM *R1.2 
EFFICIENT ELECTRONIC ENERGY FUNCTIONALS FOR TIGHT-BINDING. Roger Haydock, Univ of Oregon, Dept of Physics and Materials Science Inst, Eugene, OR.

A generalized functional is constructed from the exchange-correlation energy by Legendre transformations which make the new functional stationary at the electronic charge density, potential, and wavefunctions for the ground-state. Using a generalized functional, the density, potential, and wavefunctions can be independently parameterized and varied to determine the ground-state energy-surface for a system of atoms. This eliminates the computationally awkward steps of constructing densities from wavefunctions or potentials from densities, and is particularly well suited to parameterizations using tight-binding orbitals together with atomic-like densities and potentials. For each choice of parameters, the only quantities which must be computed are the electron-electron energy for the density, the integral of the potential over the density, and the band structure energy for the wavefunctions. To second order in the density, potential, and wavefunctions, the energy for a configuration of atoms is given by the generalized functional evaluated at a superposition of atomic densities, a potential made by stitching together the atomic potentials where they are equal, and atomic wavefunctions. For more accurate stationary energies the densities, potentials, and wavefunctions can be improved by one or more conjugate gradient steps.

9:30 AM *R1.3 
A LCAO-LOCAL DENSITY APPROACH TO THE CALCULATION OF ELECTRONIC PROPERTIES OF MATERIALS. F. Flores, P. Pou, R. Pérez and J. Ortega, Departamento de Físca Teórica de la Materia Condensada, Universidad Autónoma, Madrid, SPAIN.

An ab-initio linear combination of atomic orbitals (LCAO) method is presented to calculate the electronic properties of materials. The method is based on the introduction of a many-body hamiltonian that is written in a localized atomic orbital basis, with the different parameters calculated using quantum chemistry techniques: this basic hamiltonian is suitably approximated by neglecting four-center integrals. Then, a Density Functional approach is introduced to analyze the many-body properties of the LCAO-hamiltonian. This is achieved by showing that the total electronic energy of the system can be written as a function of the different orbital occupancies; a exchange-correlation potential is then introduced for each orbital by taking the derivative of the total electronic energy w.r.t. the orbital occupancy. Using this local potential, the total energy of the system can be obtained by calculating self-consistently the orbital occupancies, avoiding the use of a local representation as is done in the conventional LDA-calculations. The self-consistent LCAO-LD solution of our initial hamiltonian allows us to calculate also the electronic bands of the materials and a set of parameters defining its tight-binding levels and hopping integrals. Examples ranging from simple molecules, FH, CH, NH, H2O, etc., to covalent and ionic semiconductors, Si and AlP, and ionic insulators like LiF will be presented.

10:30 AM *R1.4 
EFFICIENT AB INITIO TIGHT-BINDING. Andrew Horsfield, University of Oxford, Department of Materials, UNITED KINGDOM.

Tight-binding has been used very successfully as a qualitative theory for many years, providing insight into the behaviour of a wide variety of systems. Recently it has become popular as a quantitative theory. In spite of some spectacular successes there remain problems that prove to be very difficult to overcome, notably finding systematic ways of producing parameters that apply equally well to metallic and strongly covalent systems. The rise of tight-binding has been paralleled by a rapid development of ab initio methods. Attempts are now being made to bridge the gap between these two approaches so as to obtain the computational efficiency and intuitive nature of tight-binding with the accuracy and transferability of ab initio methods. A method based on the Harris-Foulkes functional, atomic basis sets and the local density approximation has been developed recently, and has been successfully applied to a number of molecular systems. The underlying theory, some of the key numerical methods employed and results obtained will be presented and discussed.

11:00 AM R1.5 
TIGHT-BINDING LINEAR MUFFIN-TIN ORBITAL IMPLEMENTATION OF THE DIFFERENCE EQUATION GREEN'S FUNCTION APPROACH FOR 2D-PERIODIC SYSTEMS. Walter R. L. Lambrecht, Dept. of Phys., Case Western Reserve University, Cleveland, OH; and Mark van Schilfgaarde, SRI International, Menlo Park, CA.

Many systems of interest are layered with 2D periodicity and contain segments of material in which the potential is close to bulk-like, separated by interfaces consisting of a few atomic layers of different potential and sometimes structure. The difference equation approach originally introduced by A. B. Chen et al. (1989) provides possibly the most efficient way of solving the electronic structure problem for such systems by noticing that the tight-binding equations are difference equations in the layer index. The Green's function for the bulk, semi-infinite asymptotic regions, and any finite piece of consecutive equal layers can be expressed in closed form in terms of the eigenvalues and eigenvectors of the quadratic characteristic equation. These eigenvalues are closely related to the complex band structure. The segments can then be coupled to obtain the Green's function of the entire system in order(N) matrix inversion steps where N is the number of couplings required. We have developed a linear muffin-tin orbital implementation of this method. Test results to some systems of interest will be presented. Supported by NSF-95-29376.

11:15 AM R1.6 
SIMPLICITY FOR COMPLEXITY - A DENSITY FUNCTIONAL BASED TIGHT-BINDING-LIKE THEORY WITH CHARGE TRANSFER. Alexander A. Demkovb, Otto F. Sankeya. (a) Department of Physics and Astronomy, Arizona State University, Tempe, AZ. (b) Semiconductor Products Sector, Motorola Inc., Mesa, AZ.

One can use density functional theory in a minimal basis set with orbitals that rapidly cut-off (Fireballs) to construct what mathematically acts like a non-orthogonal tight-binding model. This allows not only covalent effects to be included, but also ionic effects such as that occur in SiO2 and related materials. We describe this technique and apply it to zeolites and oxide systems, as well as to expanded semiconductor phases.

11:30 AM *R1.7 
EFFECTIVE INTERATOMIC INTERACTIONS VIA THE TB-LMTO METHOD. Václav Drchal, Josef Kudrnovský, Institute of Physics AS CR, Praha, CZECH REPUBLIC; Alain Pasturel, CNRS, Grenoble, FRANCE; Ilja Turek, Institute of Physics of Materials AS CR, Brno, CZECH REPUBLIC; Peter Weinberger, CMS, University of Technology, Vienna, AUSTRIA; Antonios Gonis, Patrice E.A. Turchi, LLNL, Livermore, CA.

The energetics of metallic alloys, their surfaces or interfaces, and magnetic multilayers is studied in terms of the effective interatomic (or interlayer) interactions that are determined from ab initio electronic structure calculations using the TB-LMTO method combined with the coherent potential approximation and the method of surface Green functions. First the theoretical background (force theorem, Lloyd formula, generalized perturbation method for bulk and surfaces, vertex cancellation theorem, method of infinitesimal rotations) will be discussed, and then the applications to phase stability of bulk alloys, surface segregation in disordered alloys, magnetism-induced ordering in two- and three-dimensional systems, phase diagram of two-dimensional alloys, interlayer exchange coupling in metallic multilayers, and the construction of Heisenberg-like Hamiltonians for magnetic systems will be presented.

SESSION R2: FIRST-PRINCIPLES TIGHT-BINDING: PART II 
Chairs: Ole K. Andersen, Alex M. Bratkovsky, Anthony T. Paxton and Patrice E.A. Turchi 
Monday Afternoon, December 1, 1997 
Fairfax A (S)

1:30 PM *R2.1 
ELECTRONIC STRUCTURE AND ATOMIC CONFIGURATION OF EXTENDED DEFECTS IN METALS BY FIRST-PRINCIPLES AND SEMIEMPIRICAL TB-LMTO METHODS. M. Sob (1,2), V. Vitek (2), I. Turek (1); (1) Institute of Physics of Materials, Academy of Sciences of the Czech Republic, Brno, CZECH REPUBLIC, (2) Department of Materials Science and Engineering, University of Pennsylvania, Philadelphia, PA.

We present two tight-binding linear muffin-tin orbitals (TB-LMTO) techniques for electronic structure calculations of extended defects (as grain boundaries, interphase interfaces etc.) in metals. The first one is based on the first-principles self-consistent surface Green function approach within the atomic-sphere approximation (ASA) utilizing two-dimensional periodicity in the layers parallel to the interface. For atomic relaxation a semiempirical method is used in which the Hamiltonian is constructed within the TB-LMTO-ASA as well, but semiempirical terms are employed to characterize the repulsive part of the interaction and to correct for the ASA. While the empirical parameters have only been fitted to the properties of ideal ground state structure, the semiempirical approach describes correctly the structural energy differences, phonon frequencies etc. Illustrative examples of extended defects in transition metals will be given.

2:00 PM *R2.2 
TIGHT-BINDING CALCULATIONS OF ELECTRONIC STRUCTURE AND RESISTIVITY OF LIQUID AND AMORPHOUS METALS. S.K. Bose, Physics Department, Brock University, St. Catharines, Ontario, CANADA.

We discuss various aspects of calculating the electronic structure of liquid and amorphous metals using the recursion method and the tight-binding linear muffin-tin orbitals (TB-LMTO) basis. Resistivity calculations for such systems based on the Kubo-Greenwood formula and the TB-LMTO-recursion method are presented and compared with similar calculations based on the linear combination of atomic and atomic-like orbitals (LCAO) and the chemical pseudopotential approach. Results for amorphous Fe and Co and liquid Hg, Pd, La and some 3d transition metals are presented. Sources of error in the calculation and ways to improve upon the present calculations are discussed. Finally, the advantage of the present approach over the Ziman-Faber Diffraction model approach is presented along with a comparis on of the results obtained via the two methods.

2:30 PM *R2.3 
COMPLEX MAGNETISM AND RELATED EFFECTS AT NANOSCALE. Alexander Bratkovsky, Hewlett-Packard Labs, Palo Alto, CA.

Recent developments in ab initio tight-binding methods will be discussed which allow for predictive studies of magnetism in imperfect and disordered structures, with the implications for transport properties. For ultrathin magnetic films we find that a submonolayer amount of Cu on a stepped Co/Cu(001) film changes dramatically the electronic and magnetic structure of the system. As a result, a noncollinear arrangement of magnetic moments and reentrant switching of the easy axis is promoted, as recently observed. Another example pertains to a new magnetic phenomenon where magnetic moment, absent in metallic crystal, sets in upon melting.This prediction has been made for Al-Mn liquids and then confirmed experimentally. Calculations show a clear tendency to forming a random magnetic order in these systems. Perhaps most interesting demonstration of importance of electronic structure and magnetism at nanoscale comes from recent studies of magnetoresistive systems. Spin-tunneling and tunnel magnetoresistance (TMR) will be analyzed with examples of half-metallic systems where the TMR effect may ideally be arbitrarily large.

3:30 PM R2.4 
DEFORMATION AND TRANSFORMATION PATHS IN TITANIUM: A COMPARATIVE STUDY OF FIRST-PRINCIPLES CALCULATIONS, BOND-ORDER AND FINNIS-SINCLAIR POTENTIALS. S. Znam, A. Girshick, V. Vitek, Department of Materials Science and Engineering, University of Pennsylvania, Philadelphia, PA; L.G. Wang, M. Sob, Institute of Physics of Materials, Academy of Sciences of the Czech Republic, Brno, CZECH REPUBLIC.

The energy of titanium deformed following the tetragonal (Bain's) and trigonal deformation paths as well as a transformation path between the hcp and bcc structures, is evaluated using three distinct approaches. The ab initio full-potential linearized augmented plane waves (FLAPW) method, recently developed bond-order potential (BOP) and a central-force many-body potential of the Finnis-Sinclair type. Comparison of these calculations reveals which features of these paths are sensitive to the details of the electronic structure, that is fully included in FLAPW, which depend on d-type directional bonding, taken into account in BOP, and which are generic and revealed by all three approaches. Configurations corresponding to energy minima are particularly important since they may represent metastable structures that can play a role in extended defects. These configurations will be discussed in detail and features of bonding that control them analyzed. This research was supported by the NSF Grant No. DMR-9626344 and US Czechoslovak Science and Technology Program Grant. No. 94048.

3:45 PM R2.5 
INTERACTIONS OF POINT AND EXTENDED DEFECTS IN STRUCTURAL INTERMETALLICS: REAL-SPACE LMTO-RECURSION CALCULATIONS. O.Yu. Kontsevoi, O.N.Mryasov and A.J. Freeman Northwestern Univ, Dept. of Phys and Astron, Evanston, IL;Yu.N. Gornostyrev,Inst of Metal Phy., Ekaterinburg, RUSSIA

A modified real-space TB-LMTO-recursion (TB-LMTO-REC) method for electronic structure calculations is applied to the study of interacting extended and point defects in NiAl and FeAl. Results of calculations with up to 104 atoms per cluster for the pure intermetallics and with ternary additions (within a supercell model) show good agreement with previous available band structure results. Further, electronic structure and total energy calculations of point (single impurity, M=Ti, V, Cr, Mn, Fe and Co) and planar defects such as antiphase boundaries (APB) were carried out and the interaction between them was studied. We found that for the 111>(110) APB in NiAl, ternary additions occupy exclusively the 3d-metal sublattice (except for Co) and decrease the APB energy. Finally, we employ the TB-LMTO-REC to study the electronic structure of the most complex extended defect, a dislocation. The coordinates of atoms in the dislocation core fo NiAl are determined within the Peierls-Nabarro model with dislocation structure parameters obtained from ab-initio calculations of the generalized stacking fault (GSF) energetics. % for NiAl. The effects of the electron density redistribution in the region with up to 100 atoms in the dislocation core are treated self-consistently with local charge neutrality. We demonstrate for the <001>(010) dislocation in NiAl that: (i) quasi-localized states may exist as a result of specific lattice distortions in the dislocation core with a type of broken bonds; (ii) the electronic structure changes appreciably in the process of dislocation motion; (iii) van-Hove singularities present in the ideal crystal may be shifted to the EF as a result of the dipolar character of the deformations in the dislocation core. Results are discussed in the context of electron localization and dislocation-impurity interactions. Supported by the AFOSR.

4:00 PM *R2.6 
QUANTUM MONTE CARLO SIMULATIONS OF MAGNETIC MATERIALS. R.T. Scalettar, C. Huscroft, A. McMahan, R. Pollock, M. Randeria, N. Trivedi, and M. Ulmke, University of California, Physics Dept, Davis, CA.

Over the last decade, Quantum Monte Carlo (QMC) calculations for tight binding Hamiltonians like the Hubbard and Anderson lattice models have made the transition from addressing abstract issues concerning the effects of electron-electron correlations on magnetic and metal-insulator transitions, to concrete contact with experiment. In this talk I will review some of the applications of determinant QMC methods to cuprate superconductors, and to the Kondo volume collapse in rare earth materials. I will also discuss recent work on disordered systems which has relevance to the conductivity of thin metallic films and the behavior of the magnetic susceptibility in doped semiconductors.

4:30 PM *R2.7 
MONTE CARLO STUDIES FOR STRONG CORRELATIONS IN HUBBARD-TYPE MODELS. E. Heeb, Institute for Theoretical Physics, ETH Zurich, SWITZERLAND.

The tight-binding model with repulsive Hubbard interactions represents an ideal prototype for the study of strong correlations. While exact numerical methods have been used with some success, they are typically limited by the size of the clusters that can be investigated or by the temperatures that can be reached. Variational methods, on the other hand, often require considerable advance knowledge of ground-state properties. The method presented here alleviates this problem by augmenting the variational approach with a scheme similar to Lanczos iterations thus bridging the gap between exact diagonalization and variational approaches. For the t-J model, the low-energy effective Hamiltonian of the Hubbard model, material properties like broken translational invariance or superconducting correlations are then investigated and a region of stability of a superconducting phase is found.

SESSION R3: SEMI-EMPIRICAL TIGHT-BINDING: PART I 
Chairs: Fernando Flores, Antonios Gonis, Roger Haydock and Masanori Kohyama 
Tuesday Morning, December 2, 1997 
Fairfax A (S)

8:30 AM R3.1 
ON-SITE CORRELATION IN NARROW BAND MATERIALS. F. Manghi, V. Bellini, Modena Univ, Dept of Physics, Modena, ITALY; C. Arcangeli, Max-Planck-Inst, Stuttgart, GERMANY.

It is well estasblished that the description of electronic states in narrow band materials - transition metals, transition metal oxides, cuprates etc. .. - requires improvements over the single particle approximation, with proper inclusion of many body effects associated to the on-site Coulomb interaction of localized electrons. In many of these systems the itinerant character of valence electrons, which is clearly shown by the k-dispersion seen in photoemission spectroscopy, coexists with strong local electron-electron correlation responsible of other observed features such as satellite structures, band narrowing and opening, in some cases, of a Mott-Hubbard gap. It seems therefore necessary to combine a realistic band structure description with an accurate treatment of many body effects. We present a theoretical method recently developed [1] which has been designed to treat highly correlated and highly hybridized systems including both the itinerant character of band electrons and the localized electron-electron repulsion: the single-particle band states are determined in a localized basis - either ab-initio LMTO or semi-empirical Tight-Binding - and used as input mean-field eigenstates for the calculation of self-energy corrections according to a 3-body scattering solution of a multi-orbital Hubbard Hamiltonian. Self-energy corrections, spectral functions and quasi-particle spectra for nickel, NiO and CuGeO3 are presented. The calculated quasi-particle spectra show a remarkable agreement with photoemission data in terms of band width, k-dispersion, exchange splitting and satellite structures. The correct insulating behaviour of NiO and CuGeO3 is also reproduced.

8:45 AM R3.2 
ELECTRONIC STRUCTURE OF TRANSITION METAL SYSTEMS BY QUANTUM MONTE CARLO. Lubos Mitas, National Center for Supercomputing Applications, University of Illinois, Urbana, IL.

Effects of electron-electron correlation on the electronic structure of transition metal (TM) systems such as TM clusters with small organic molecules and TM oxides are studied by variational and diffusion quantum Monte Carlo (QMC) methods. In particular, comparisons with Local Density and Hartree-Fock approaches show impact of correlation on calculated properties which are strongly affected by correlation, e.g., band gaps in TM oxides or energy ordering of low-lying states in small clusters. These calculations, which involve an explicit treatment of many-body effects, show an importance of a balanced approach to both exchange and correlation for an accurate and predictive description of electronic structure in these systems. Therefore QMC results and calculations, because of their accuracy and many-body information, provide excellent examples and test cases for development of simplified approaches such as tight binding Hamiltonians. From the fundamental point of view quantum Monte Carlo seems to be one of the most promising ab initio approaches because of accuracy, computational efficiency and favorable scaling in the number of particles.

9:00 AM *R3.3 
ELECTRONIC STRUCTURE AND TRANSPORT IN NON PERIODIC SYSTEMS: NEW O(N) METHODS. D. Mayou, LEPES-CNRS, Grenoble, FRANCE; P. Turchi, L.L.N.L., Livermore, CA; S. Roche, University of Tokyo, Dept of Applied Physics, Tokyo, JAPAN; J.P. Julien, LEPES-CNRS, Grenoble, FRANCE.

The mathematical theory of orthogonal polynomials and continued fractions provides very efficient tools, via the recursion method and related methods, for calculating diagonal elements of Green's function of tight-binding hamiltonians. We present here two new generalisations of this formalism. 
1) A new method for calculating linear response coefficients. We present several applications, and in particular the calculation of conductivity through the Kubo-Greenwood formula. 
2) A new approach to mean-field theories of alloys. We show that self consistent field equations, in particular the CPA equations, can be solved in a very efficient way through the use of continued fraction representation of Green's functions. This allows to calculate, via GPM or ECM theory, effective atomic interactions in any non periodic structure. Applications to the stability of complex phases are given.

9:30 AM *R3.4 
LET THERE BE LIGHT IN TIGHT BINDING. Peter Vogl, Martin Graf, Andreas Gorling, Walter Schottky Institute, Technical University of Munich, Garching, GERMANY.

We have recently developed a novel scheme that allows one to predict and understand the interaction of light with matter nonperturbatively within the empirical tight binding method without introducing extra free parameters [1]. Solids in extreme external electric, magnetic or intense laser fields exhibit fascinating new phenomena. A theory of these effects must be capable of treating external fields and the crystal or cluster potential on the same footing without relying on perturbation theory. We show that electromagnetic potentials can be incorporated in empirical tight binding theory in a general way that is gauge invariant, guarantees charge conservation and does not introduce any extra free parameters. The analogue of the optical f-sum rule for finite basis sets is shown to greatly facilitate the calculation of electronic properties of mesoscopic systems [1,2]. The applications that will be presented range from the Hofstaedter butterfly [3] in real solids, to optical response functions of solids, lattice dynamics, magnetotransport in superlattices, spin susceptibilities, and to the specific heat and entropy of ferroelectrics.

10:30 AM *R3.5 
AB INITIO CALCULATION OF TIGHT-BINDING PARAMETERS*. A. K. McMahan, J. E. Klepeis, Lawrence Livermore National Laboratory, Livermore, CA.

The authors have recently derived analytic expressions for the two-center Slater-Koster hopping parameters, effective site energies, and effective crystal field parameters in terms of the one-electron Hamiltonian matrix elements in any localized minimal basis, and analogous quantities for the overlap. These expressions were applied to four-fold coordinated phases of the Boron/Silicon system using spd, nonorthogonal FP-LMTO matrix elements obtained with a linked minimal basis. Here we extend these B/Si calculations to higher-coordination structures, and by Lorthogonalization, to orthogonal as well as nonorthogonal representations. We find the hopping parameters in more close packed structures to be slightly larger in magnitude than those in more open structures for the nonorthogonal representations. As expected, the transferability is significantly degraded for the orthogonal case where the more close packed structures have hopping parameters which are smaller in magnitude than the four-fold coordinated phases. This behavior may be understood in terms of low order expansions of the Lorthogonalization. That leading terms in such expansions are qualitatively meaningful can be demonstrated by calculating the orthogonal-basis hopping parameters, , over the range for scaled overlap matrices, . We also show analytically and numerically that L orthogonalization augments non-two-center contributions in the one-electron Hamiltonian, so that two-center approximations to orthogonal basis one-electron Hamiltonians are intrinsically less accurate than to the nonorthogonal basis Hamiltonians. Finally, we apply the present expressions to metallic Ce to show how the present method can be easily used to identify the relative sizes of f-f, f-valence, and f crystal field contributions in the band structure of f-electron metals.

11:00 AM *R3.6 
AN AB INITIO TWO-CENTER TIGHT-BINDING APPROACH TO SIMULATIONS OF COMPLEX MATERIALS PROPERTIES. Th. Frauenheim, D. Porczag, M. Elstner, G. Jungnickel, Technische Universität, Institut für Physik, Chemnitz, GERMANY; G. Seifert, Technische Universität Institut für Theoretische Physik, Dresden, GERMANY.

We describe the ab initio construction of two-center tight-binding (TB) hamiltonians, which upon non-selfconsistent solution of the related Kohn-Sham equations transform the energy within density-functional theory (DFT) into a tight-binding-like expression. At a properly chosen input density, all hamiltonian and overlap matrix elements are explicitly calculated within a non-orthogonal atomic orbital basis, thus avoiding the usual parametrization known as the major bottleneck of standard TB-variants. Assuming that the electron density of the interacting many-atom structure may be represented as a sum of atomic densities, the linear combination of excited or reasonably confirmned pseudoatomic orbitals (obtained within scf-LDA calculations) has been shown to yield a high transferability and accurate results. The calculated tight-binding energy differs from the true ground state energy only in second order of charge density fluctuations. Interatomic forces for molecular-dynamics simulations may easily be calculated. The method has been successful in predicting the structure and vibrational signatures of fullerene oligomers, amorphous carbons and carbon nitrides and in simulating elementary growth reactions on diamond surfaces. The uncertainties within the non-scf TB-variant increase if considerable fractions of charges are transferred between different atomic constituents, as e.g. in organic molecules, in polar semiconductors and in highly defective bulk and surface situations. Therefore, we extend the non-scf TB-approach to the operation in a selfconsistent-charge mode (sec DT-TB) in order to improve total energies, forces, and transferability in the presence of considerable long-range Coulomb interactions. We derive a transparent and readily calculate expression for the iterative modification of Hamiltonian matrix elements and show, that the final energy is second order approximation to the total energy in density-functional theory, see M. Elstner et al., this Symposium.

11:30 AM *R3.7 
TIGHT-BINDING INTERPOLATION OF FIRST-PRINCIPLES TOTAL ENERGIES. D.A. Papaconstantopoulos and M.J. Mehl, Complex Systems Theory Branch, Naval Research Laboratory, Washington, DC.

We have built non-orthogonal tight-binding Hamiltonians1 by fitting to both the energy bands and total energy of first-principles calculations for all transition metals and several s-p materials. This scheme gives as an output the elastic constants, vacancy formation energy, surface energies and phonon spectra in very good agreement with independent first-principles calculations and experiment. We have also extended this method to several binary systems where we are attempting to map out a complete phase diagram of the energetics of these systems.

SESSION R4: SEMI-EMPIRICAL TIGHT-BINDING: PART II 
Chairs: Shyamal K. Bose, Giulia Galli, Vittorio Rosato and Guy Treglia 
Tuesday Afternoon, December 2, 1997 
Fairfax A (S)

1:30 PM *R4.1 
ENVIRONMENT-DEPENDENT TIGHT-BINDING POTENTIAL MODEL. C.Z. Wang, Ames Lab and Dept of Physics, Iowa State Univ, Ames, IA.

In the past four decades, most tight-binding potential models are constructed using the two-center approximation for the hopping parameters, following the classic work of Slater and Koster. While the two-center approximation greatly simplifies the tight-binding parametrization, neglecting multi-center interactions is inadequate for systems where metallic bonding effects are significant. Recently, we have developed a new tight-binding model which goes beyond traditional two-center approximation and allows the tight-binding parameters and the repulsive potential to be dependent on the bonding environment. This new model preserves the two-center form of the traditional tight-binding model while taking into account the multi-center effects. Work on C, Si, Ge, Al, Mo, and Nb shows that the new approach improves remarkably the transferability of the tight-binding potentials. The properties of the higher-coordinated metallic structures are well described by the model in addition to those of the lower-coordinated covalent structures.

2:00 PM R4.2 
DEVELOPMENT OF INTERATOMIC POTENTIALS FOR MODELLING OF CVD DIAMOND GROWTH. Ivan I. Oleinik, David G. Pettifor, Adrian P. Sutton, University of Oxford, Dept of Materials, Oxford, UNITED KINGDOM.

A basic understanding of surface chemical reactions which take place during CVD diamond growth can only be achieved by reliable atomistic modelling of the reaction pathways and activation barriers. Ab-initio methods are restricted to simulating very small systems, therefore, the development of semi-empirical interatomic potentials which allow us to model much larger systems for longer times is the focus of current theoretical efforts. We develop analytical bond order potentials (BOPs) for hydrocarbon systems based on the two-center tight-binding (TB) approximation to the electronic structure. In the BOPs the electronic degrees of freedom are no longer treated explicitly but their influence is captured through the functional form which is devised by a well-defined set of approximations to the TB Hamiltonian. This helps to avoid the many ad hoc parameters and functional forms that enter empirical potentials. We demonstrate that the first term of the BOP expansion, which corresponds to a Tersoff type potentials, is unable to account correctly for the relative stability of different structures such as graphite, diamond simple cubic and FCC. Accurate structural predictions require inclusion of the second term of the BOP expansion which guarantees that the fourth moment of density of states is included exactly. This property of the analytical BOP is of paramount importance for the covalent systems with band gaps such as silicon and carbon. The remarkable feature of the novel BOP is that it predicts the environmentally dependent angular functions of the interatomic potentials which were recently introduced by inversion of ab-initio binding energy curves (M.Z. Bazant and E. Kaxiras, Phys Rev Let, 77, 4370 (1996)). As a result, analytical BOP gives the right ordering of different crystal structures and cohesive energies for open structures (diamond and graphite). We validate the analytical BOP against experiment and ab-initio predictions.

2:15 PM R4.3 
OPTICAL PROPERTIES OF MATERIALS USING THE EMPIRICAL TIGHT-BINDING METHOD. L.C. Lew Van Voon, Physics Dept, Worcester Polytechnic Institute, Worcester, MA.

The tight binding method has traditionally been used to investigate the electronic and structural properties of materials. More recently, application to optical properties has flourished, in part due to the discovery of the exact representation of the transition matrix element [Lew Yan Voon and Ram-Mohan, Phys. Rev. B47, 15500 (1993)]. In this talk, we compare our (exact) formalism with the (approximate) ones of others. Recent applications of the formalism to the structural stability of C60 and to the optical properties of semiconductor heterostructures will be reviewed. We will also present new results on the birefringence of multilayers which reveal the better treatment of symmetry of the tight binding over effective mass methods.

2:30 PM *R4.4 
ORDERING ENERGETICS IN THE TIGHT-BINDING APPROXIMATION. Alain Pasturel, LPMMC, CNRS, Grenoble, FRANCE.

Due to the presence of interatomic interactions, any alloy should eventually order or phase separate at low temperature. This is the case of many systems and for instance they lead to the order-disorder transitions observed in bulks or at surfaces or to the chemical short-range order found in liquid and amorphous alloys. Recent research indicates that it is possible to derive such interactions using electronic structure models based on various well-defined approximations. Our purpose is to discuss the advantages and the weakness of the different methods developed to obtain these interactions using the one-electron approximation as written within the tight-binding method. The coupling with statistical mechanics techniques to take into account finite temperatures effects will be also presented.

3:30 PM *R4.5 
STRUCTURE, BONDING AND STABILITY OF TRANSITION METAL SILICIDES: A REAL-SPACE PERSPECTIVE BY TIGHT BINDING POTENTIALS. Leo Miglio, University of Milano and INFM, Dept. of Materials Science, Milano, ITALY.

Recently, tight binding potentials have been successfully adopted for total-energy calculations and numerical simulations of semiconductors, transition metals and carbon-based materials. An increasing amount of applications are talking place whenever the complexity or the size of the system prevents the employment of first-principles approaches. Probably, the main reason of its prosperity is related to the low computational cost entailed by the semiempirical estimation of the electronic states entering the attractive (band structure) part of the potential and by the phenomenological description of the repulsive part through pair interactions. Still, the most interesting feature of this method rests in the possibility to exploit a real-space analysis of the system, a powerful interpretative tool which is really complementary to first principles calculations. This is accomplished by evaluating the orbital-and site- projected densities of states, both in frozen configurations and during a molecular dynamics simulation. Moreover, the partition of total energy into one attractive and one repulsive part provides a clear understanding of the stability trend between competing structures in terms of neighbors configuration (bonding) and atomic packing (Pauli repulsion), respectively. Transition metal silicides are particularly interesting from this point of view, since they display both a high sensitivity of the electronic features to the bond directions, as provided by the fairly covalent psi-dTM overlap, and a relevant polymorphic attitude, which is typical of metallic materials. Some examples relating structure, bonding and stability will be given.

4:00 PM *R4.6 
SELF-CONSISTENT TIGHT BINDING APPROXIMATION FOR METALS AND OXIDES. A.T. Paxton, M.W. Finnis, Atomistic Simulation Group, Department of Pure and Applied Physics, Queen's University, Belfast, UNITED KINGDOM.

We discuss parameters for tight binding models of transition metals and their oxides. We also consider various approaches to include charge self consistency.

4:30 PM *R4.7 
TIGHT-BINDING CALCULATION OF COMPLEX DEFECTS IN SEMICONDUCTORS: COMPARISON WITH AB INITIO RESULTS. M. Kohyama, Dept. of Material Phys., Osaka National Research Institute, Ikeda, Osaka, JAPAN; N. Arai and S. Takeda, Dept. of Phys., Osaka University, Toyonaka, Osaka, JAPAN.

By using the transferable tight-binding (TB) method [1,2], it is possible to deal with complex or extended defects in semiconductors more accurately than using previous TB methods or empirical potentials. However, one should be prudent about the quantitative accuracy of this semi-empirical method applied to complex systems, because the accuracy has been examined only for several crystal phases or simple defects. Thus, in this paper, we examine our TB results of various complex defects in Si and SiC by comparing those with our recent ab initio results. We deal with grain boundaries in Si and SiC, {113} planar defects in Si, and self-interstitial clusters in Si [3]. The supercells for the TB method contain hundreds of atoms, i.e. 724 atoms for the self-interstitial cluster. Ab initio calculations using conjugate-gradient techniques and optimized pseudopotentials must use supercells with reduced sizes, although we have tried to maintain essential features of defects. Results are compared in detail, and also with recent experimental results using HRTEM and EELS. We discuss the merits, limitations and roles of the TB method.

SESSION R5: POSTER SESSION: 
SEMI-EMPIRICAL TIGHT-BINDING: PART III 
Chairs: Andrew K. McMahan, J.P. Gaspard, G. Allan and V. Drchal 
Tuesday Evening, December 2, 1997 
8:00 P.M. Grand Ballroom (S)

R5.1 
TIGHT-BINDING CALCULATIONS OF BOND ENERGY FOR TRANSITION METALS: CONTRIBUTION OF THE SP-BAND. A. Aguayo, G.I. Canto and R. de Coss, Department of Applied Physics, CINVESTAV-IPN, Yucatán, MEXICO.

Recently it has been show the importance of the spd hybridization to obtain accurate site energies in Pd clusters [1]. In this work we study the sp-band contribution to the bond energy using a tight-binding model for bcc and fcc non-magnetic transition metals. We have calculated the bond energy from a semi-empirical Slater-Koster hamiltonian, which includes s-p-d orbitals within the two center aproximation. The used tight-binding parameters reproduce the ab initio band structures. We find that the sp contribution to the bond energy in the transition metals studies .In particular, the sp contribution in Ta reaches 41%. These results show the importance of the inclusion of the sp-band in tight-binding models to obtain accurate values of the total energy of transition metal systems.

R5.2 
NEGATIVE CAUCHY PRESSURES WITHIN THE TIGHT-BINDING APPROXIMATION. Duc Nguyen-Manh, David Pettifor, Oxford University, Department of Materials, Oxford, UNITED KINGDOM.

It is well-known that the Embedded Atom Method (EAM) predicts positive Cauchy pressures for cubic metals if physically-motivated embedding functions are used. Suprisingly, even if the angular character of the covalent bonding is included within an orthorgonal or non-orthogonal Tight-Binding (TB) description, the Cauchy pressure for most elemental and binary metallic systems remains positive. In this talk, we describe the results of a detailed breakdown of the different contributions to the Cauchy pressure within the Harris-Foulkes approximation to density functional theory. We show that negative values of the Cauchy pressure arise from the environment dependence of the local TB orbitals which leads to environment-dependent bonding intergrals and overlap repulsion. We illustrate the importance of this environment dependence for both elemental transition metals such as Ir and binary intermetallics such as TiAl and TiAl3. Finally, we discuss a general functional form for overlap repulsion and compare it with different fitting schemes proposed recently in TB theory.

R5.3 
TIGHT-BINDING ANALYSIS OF VALENCE BAND STRUCTURE FOR RELAXED GROUP IV SEMICONDUCTOR ALLOYS. C.Y.Lin, Dept. of Physics, National Chung Hsing Univ., Taichung, TAIWAN; and C.W. Liu, Dept. of Electrical Engineering, National Taiwan Univ., Taipei, TAIWAN.

Analytic valence band structures of relaxed group IV semiconductor alloys are obtained by using a combination of the tight-binding and k p methods. A 16 x 16 tight-binding Hamiltonian matrix is constructed for crystals of diamond lattice by employing direct product of eight s - p hybridization orbitals and the spin wave functions. The parameters A, B, and C appear in the k p valence band dispersion relation _k= -h^22m[Ak^2

8:30 AM *R6.1 
EMPIRICAL TIGHT-BINDING APPLIED TO SILICON NANOSTRUCTURES. G. Allan, C. Delerue and M. Lannoo, IEMN - Dept ISEN, Villeneuve d'Ascq, FRANCE.

The calculation of the electronic structure of silicon nanostructures is used to discuss the accuracy of results obtained by the tight-binding method. We first show that the level of refinement of the tight-binding approximation must be adapted to the calculated property. For example, an accurate description of both the valence and conduction bands which can be achieved with a 3rd-nearest neighbor approximation is necessary to calculate the variation of the gap energy with the silicon crystallite size. However the sp3s* model which gives a bad description of the conduction band underestimates the confinement energy but can give good results when it is used to determine the variation of the crystallite band gap with pressure. To study Si-III (BC-8) nanocrystallites, we show that a good description of the bulk band structure can be obtained with non-orthogonal tight-binding but due to the large number of nearest neighbors one must take into account forces to take analytical variations of the parameter with interatomic distances. The parameters involved in these expressions can be easily fitted to the bulk band structures using the k-point symmetry without requiring the use of group theory. Finally we discuss the effect of increasing the size of the minimal-basis set and we show that it would be possible to get the values of the tight-binding parameters from a first-principles localized states band structure calculation avoiding the fit to the energy dispersion curves.

9:00 AM *R6.2 
HOW FAR TO USE TIGHT-BINDING POTENTIALS FOR BIMETALLIC SURFACE MODELLING? Guy Treglia, CNRS, Marseille, FRANCE; B. Legrand, SRMP, CEA Saclay, Gif-sur-Yvette, FRANCE.

Modelling the behaviour of surfaces and interfaces in bimetallic systems (alloys, heteroepitaxial layers) from their electronic structure is a particularly exciting challenge in modern Material Science. By behaviour we mean, not only the equilibrium atomic (reconstruction, relaxation) and chemical (segregation, ordering) structures of these surfaces, but also their kinetic (out of equilibrium) properties, which are essential for understanding the growth of thin films deposited on a substrate and their (limited) incorporation under annealing. This requires to use numerical tools, such as Molecular Dynamics or Monte Carlo simulations, which are time-consuming, and therefore to derive interatomic potentials as analytical as possible ... while grounded on the electronic structure. This can be achieved in the framework of tight-binding formalism, at least under simplified forms which depend on the addressed problem: either effective hamiltonian of the Ising-type when only chemical exchanges are considered, or many-body potentials (defined within the second moment approximation of the density of states) when atomic displacements are also involved. However, deriving such tight-binding potentials requires more or less drastic assumptions. Among them is the possible charge redistribution at the surface which needs to be known from reference ab initio calculations. Note that, as a by-product, such calculations also give useful check of the analytical form of the semi-empirical potentials (square-root variation with coordination number, ...). Moreover, when the phenomena under study couple atomic and chemical rearrangements, both approximations have to be consistent which is far from being obvious, as shown from the influence of atomic relaxations around impurities on mixing interactions in the case of strong size-mismatch. We will illustrate these abilities (... and limitations) of tight-binding formalism in the context of both dynamical (surface diffusions close to steps, formation of surface alloys) and equilibrium (morphological and chemical structure of bimetallic systems such as clusters, alloy surfaces, ...) studies.

9:30 AM R6.3 
SUPERLATTICE CALCULATION IN AN EMPIRICAL spds* TIGHT-BINDING MODEL. Reinhard Scholz, Institut fuer Physik, Technische Universitaet Chemnitz, Chemnitz, GERMANY; Jean-Marc Jancu, Franco Bassani, Scuola Normale Superiore, Pisa, ITALY.

We propose an empirical tight-binding method for tetrahedrally coordinated cubic materials and apply it to group IV and III-V semiconductors, extending existing calculations by the inclusion of all five d-orbitals per atom in the basis set. The symmetry character of the conduction states at the surface of the Brillouin zone is considerably improved compared to calculations based on smaller basis sets, and the corresponding band positions can be obtained to meV precision. Because the distance dependence of the tight-binding parameters is derived from deformation potentials, the model is particularly suited for an investigation of strained superlattices where the states at direct or pseudo-direct conduction band minima are composed of wavefunctions of all the minima at , X, and L of the constituents. Calculations for GaAs/AlAs and GaSb/AlSb short-period superlattices indicate a strong mixing between the conduction band valleys in the miniband structure, and the results are in better agreement with experiments than state-of-the-art pseudopotential results.

9:45 AM R6.4 
SELF-CONSISTENT TIGHT-BINDING METHODS APPLIED TO SEMICONDUCTOR NANOSTRUCTURES. Aldo Di Carlo, Sara Pescetelli, Andrea Reale, Paolo Lugli, Istituto Nazionale Fisica della Materia (INFM) and Dipartimento di Ingegneria Elettronica, Universita di Roma ``Tor Vergata'', Roma, ITALY.

Electronic and optical properties of semiconductor nanostructures based on homo- and heterojunctions have been investigated theoretically by means of a variety of tools. The empirical tight binding method (TB) has been shown to be a valid alternative to common used envelope function approximation (EFA), since it improves the physical content in the description of nanostructures with respect to EFA, without requiring a much higher computational effort. So far, however, TB has been mainly used in the calculation of the electronic properties of nanostructures without taking into account self-consistent charge redistribution, which is an important requirement when we deal with real devices. In this communication we will show that the TB method can be implemented in a self-consistent fashion and we demonstrate its suitability for the calculation of optical and electronic properties of realistic nanostructured devices where the translational symmetry is broken in one direction. Self-consistent results for strained and unstrained quantum well systems, namely lattice matched quantum well and pseudomorphic high electron mobility transistor (P-HEMT), are given. Photoluminescence spectra are then calculated and the relation with selection rules of intersubband transitions and light polarization is analyzed. 
The tight binding method has be also applied to nanostructures where the electrons are confined in one-dimension, namely T- and V-shaped quantum wires. The large hamiltonian matrix which results from the realistic dimensions of the unit cell (more than 10000 atoms), has been diagonalized by using a generalization of the Lanczos algorithm without reorthogonalization. This analysis of QWR shows that band mixing between wire states and barrier X states can drastically change the electronic properties of the conduction band. Such mixing cannot be accounted for within the EFA since the band discontinuity is not simply given by the envelope of point. As a consequence, all EFA results related to the barrier height can be affected by the restriction of the envelope function model.

10:30 AM R6.5 
EFFECTS OF GRAIN BOUNDARIES IN SUPERCONDUCTING MATERIALS. J.F. Annett, J.J. Hogan-O'Neill, A.M. Martin, Bristol Univ, Dept of Physics, Bristol, UNITED KINGDOM.

Grain boundaries are important as weak links in high temperature superconductors. Critical currents of grain boundary junctions have been shown to depend strongly on grain orientation.1 Using a tight-binding model and the recursion method it is possible to calculate self-consistently the density of states, energy gap, and order parameter near grain boundaries. This enables us to calculate the critical current in grain boundaries as a function of orientation. This approach can be used for both conventional s-wave superconductivity and for d-wave pairing. For d-wave symmetry the junctions exhibit -Josephson effects at some angles, as observed by Tsuei et al..2 This effect is not seen in s-wave superconductors, and hence grain boundaries provide a useful tool for testing pairing symmetry.

10:45 AM R6.6 
CHARACTERIZATION OF INTERATOMIC POTENTIALS BY A CALCULATION OF DEFECT ENERGY. Yoshiaki Kogure, Masao Doyama, Teikyo University of Science & Technology, Uenohara, Yamanashi, JAPAN.

The embedded atom method potential has widely been used on the molecular dynamics simulation of metals and alloys. The potential parameters are determined from the experimental data of the cohesive energy, the lattice parameter, the elastic constants and the vacancy formation energy. A variety of potential function has been proposed. These potential functions are characterized through a molecular dynamics simulation of a crystal, which contains a interstitial, a dislocation and a stacking fault. The atomic configuration and the lattice vibration around the defects are carefully investigated. The fundamental values of defects, such as the interaction energy between vacancy and interstitial, and the Peierls stress of the dislocation, are also calculated. Several existing potentials and a newly developed potential by the present authors are used in the calculations and the results are compared. A theoretical calculation on the electronic state based on the tight-binding theory is also made to examine the reliability of the potential parameters and to find a new potential function.

11:00 AM *R6.7 
COMPARISON OF LINEAR SCALING TIGHT-BINDING METHODS. D.R. Bowler, M. Aoki, C.M. Goringe, A.P. Horsfield, D.G. Pettifor, Oxford University, Department of Materials, Oxford, UNITED KINGDOM.

Four linear scaling tight binding methods (the density matrix method, bond order potentials, the global density of states method, and the Fermi operator expansion) are described and compared to show relative computational efficiency for a given accuracy. Various example systems are explored: an insulator (carbon in the diamond structure), a semiconductor (silicon), a transition metal (titanium) and a molecular (benzene). The density matrix method proves to be most efficient for systems with energy gaps, while recursion based moments methods prove to be most efficient for metallic systems.

11:30 AM *R6.8 
QUANTUM SIMULATIONS USING LINEAR SCALING METHODS. Giulia Galli,IRRMA, Lausanne, SWITZERLAND.

After reviewing the basic features of orbital based linear scaling methods for quantum simulations [1], applications of these approaches within a tight-binding formulation will be presented. These include molecular dynamics simulations of the growth of a thin film composed of small fullerenes [2], and the investigation of ordered and disordered C solids and of collapsed fullerite.

SESSION R7: TIGHT-BINDING SIMULATIONS: PART II 
Chairs: Luciano Colombo, Leo Miglio, Richard T. Scalettar and R. Stanley Williams 
Wednesday Afternoon, December 3, 1997 
Fairfax A (S)

1:30 PM *R7.1 
TIGHT-BINDING SIMULATION OF DISORDERED SYSTEMS. V. Rosato, ENEA, High-Performance Computing and Networking Project, Centro Ricerche Casaccia, Roma, ITALY.

The reliability of the tight-binding approach to model complex disordered systems is discussed. Results of tight-binding molecular dynamics simulations for different systems (amorphous carbon, large structures based on fullerenes, surfaces and grain boundaries in silicon) will be reported, compared with experimental data and, wherever possible, with corresponding results generated by different model hamiltonians (e.g. Density Functional).lt is our aim to establish the current limits of transferability of the TB approach from ordered structures to complex systems, and to suggest possible alternative routes to increase the quality of the results.

2:00 PM R7.2 
TIGHT-BINDING ELECTRON-ION DYNAMICS. Roland E. Allen and John S. Graves, Texas A&M University, Physics Dept., College Station, TX.

A method is introduced for simulations of the coupled dynamics of electrons and ions in a material. It is applicable to general nonadiabatic processes, including interactions with an arbitrarily strong and time-dependent electromagnetic field. Since first-principles methods provide a poor description of the excited states, are computationally inefficent, and yield little chemical insight, the most suitable representation for the electronic states is a semiempirical tight-binding Hamiltonian. The interaction with the radiation field is treated through a time-dependent Peierls substitution, which introduces no additional parameters and is valid for the nonlinear effects associated with extremely intense fields. The time-dependent Schrodinger equation is solved with an adapted Cayley algorithm, which conserves probability and preserves orthogonality. The atomic forces are computed from a generalized Hellmann-Feynman theorem (which is also a generalized Ehrenfest theorem). Calculations for Si and GaAs demonstrate that the method is practical, reliable, and quantitative. It is also inherently O(N), once an intial configuration has been specified.

2:15 PM R7.3 
A NOVEL SCHEME FOR ACCURATE MD SIMULATIONS OF LARGE SYSTEMS. Alessandro De Vita and Roberto Car, Institut Romand de Recherche Numérique en Physique des Matériaux (IRRMA), Lausanne, SWITZERLAND.

We present a simple and informationally efficient approach to electronic-structure-based simulations of large material science systems. The algorithm is based on a flexible embedding scheme, in which the parameters of a model potential are fitted at run time to some precise information relevant to localised portions of the system. Such information is computed separately on small subsystems by electronic-structure ``black box'' subprograms, e.g. based on tight-binding and/or ab initio models. The scheme allows to enforce electronic structure precision only when and where needed, and to minimise the computed information within a desired accuracy, which can be systematically controlled. Moreover, it is inherently linear scaling, and highly suitable for modern parallel platforms, including those based on non-uniform processing. The method is demonstrated by performing computations of tight-binding accuracy on solid state systems in the ten thousand atoms size scale.

2:30 PM *R7.4 
COVALENT LIQUID ALLOYS. TIGHT-BINDING SIMULATION VERSUS EXPERIMENTAL RESULTS. Jean-Pierre Gaspard, Jean-Yves Raty, Liege Univ, Dept of Physics, Sart-Tilman, BELGIUM; Christophe Bichara, Vanessa Coulet, CNRS, CTM, Marseille, FRANCE; Gilles Prigent, CEA, LLB, Saclay, FRANCE.

Semiconducting elements (As, Sb,Se,Te) or compounds (III-V, II-VI and V-VI) undergo a semiconductor to metal transition at high pressures and at high temperatures. We first review the effects of external parameters (p, T) on their crystallographic structures. Neutron or X-ray diffraction and EXAFS are used to analyse the structure of the phases. The equation of state is determined as well as the bulk modulus and its pressure derivative. A simple tight-binding model and an empirical repulsive pair potential is used both to derive analytical results and to perform Monte-Carlo simulations in the liquid state. The partial pair correlation functions are obtained. Results are shown for different elements (columns V and VI) and compounds: (II-VI: CdTe, ZnTe, HgTe) and (V-VI: AsTe, SbTe). The crucial importance of the repulsive term is demonstrated. A Van der Waals interaction has to be added when planar or chain structures appear due to a strong Peierls distortion. The effect of the electrostatic Ewald term on the semiconductor to metal transition at the melting temperature is finally discussed.

3:30 PM *R7.5 
STRUCTURAL AND ELECTRONIC PROPERTIES OF a-GaAs: A TIGHT-BINDING-MOLECULAR-DYNAMICS-ART SIMULATION. Laurent J. Lewis and Normand Mousseau, Département de phyisque et Groupe de recherche en phyisque et technolgie des couche minces (GCM), Université de Montréal, Succ. Centre-Ville, Montréal, Québec, CANADA.

By combining tight-binding (TB) molecular dynamics (MD) with the recently-proposed activation-relaxation technique (ART), we have recently constructed structural models of -GaAs and -Si with an unprecedented level of quality: the models are almost perfectly four-fold coordinated and, in the case of -GaAs, exhibit a remarkably low density of homopolar bonds. In particular, the models are superior to structures obtained using melt-and-quench TB-MD or quantum MD. We find that -Si is best described by a Polk-type model, while -GaAs resembles closely the mechanical model proposed by Connell and Temkin, which is free of wrong bonds. In this talk, the structural, electronic, and dynamical properties of -GaAs based on this approach will be reviewed, and compared to experiment and other structural models. Our study provides much-needed information on the intermediate-range topology of amorphous tetrahedral semiconductors; in particular, we will see that the differences between the Polk and Connell Temkin models, while real, are difficult to extract from experiment, thus emphasizing the need for realistic computer models.

4:00 PM R7.6 
A STUDY OF DISLOCATIONS IN SEMICONDUCTORS USING TIGHT BINDING. Ricardo W. Nunesa, Jack Bennettob, and David Vanderbiltb, aNaval Research Lab and George Mason University, Washington, DC; bRutgers University, Piscataway, NJ.

In this talk we present the results of our investigation of the physics of dislocations in homopolar semiconductors [carbon (C), silicon (Si), and germanium (Ge)], using the density-matrix tight-binding technique of Li, Nunes, and Vanderbilt. This technique allows the study of large supercells ( 1000 atoms) while treating the electronic structure quantum-mechanically. In these systems, the predominant slip systems consist of 60 and screw dislocations oriented along and lying in a slip plane. Both are known experimentally to dissociate into pairs of partial dislocations bounding a ribbon of stacking fault. The resulting partials are the 30 and the 90 partial dislocations. In this talk, we consider the ground-state structure of the core of the 90-partial dislocation in these materials, in view of our recent proposal [J. Bennetto, R. W. Nunes, and D. Vanderbilt, Phys. Rev. Lett., in press] that the 90-partial in Si undergoes a period-doubling (DP) reconstruction, as opposed to the commonly accepted single-period (SP) one. Here this issue will be addressed also for C and Ge.

4:15 PM R7.7 
ELASTICITY, THERMAL PROPERTIES, AND MOLECULAR DYNAMICS USING NON-EMPIRICAL TIGHT BINDING. Ronald E. Cohen, Carnegie Institution of Washington and Center for High Pressure Research, Washington, D.C; Lars Stixrude, University of Michigan, Ann Arbor, MI; Evgeny Wasserman, Battelle, Pacific Northwest National Laboratory, Richland, WA.

We have applied the non-empirical tight-binding model of Cohen et al. (Phys. Rev. B 50, 14694, 1994) to the elastic and thermal properties of Fe, Si, and Xe at high pressures and have developed molecular dynamics programs and cellular models for studying the thermal properties of materials. The original model, which worked well for various transition and noble metals, as well as Xe, needed modification to work for Si, and we now have a Si model that includes s, p, and d states which gives the band structures and energies of Si over a wide pressure and structure range, including the diamond structure and close packed structures. Parameters are obtained by simultaneously fitting band eigenvalues and energies for a range of compressions and structures obtained using the LAPW method. With our molecular dynamics program we are studying the properties of Fe liquid at high pressures and temperatures and find liquid viscosities consistent with magnetohydromagnetic modeling of the Earth's outer core. We find that the model is generally applicable to metals, semiconductors, and rare gases. Recent developments will be discussed.

4:30 PM R7.8 
SIMULATIONS OF SOLID PHASE EPITAXY MECHANISMS IN AMORPHOUS-CRYSTALLINE INTERFACES IN SILICON. N. Bernstein, Div of Engineering and Applied Sciences, Harvard University, Cambridge, MA; E. Kaxiras, Physics Dept and Div of Engineering and Applied Sciences, Harvard University, Cambridge, MA.

We report simulations of the amorphous-crystalline interface of silicon, in an attempt to develop understanding of the microscopic processes that underlie solid phase epitaxial growth (SPEG), the process by which the crystal grows into the amorphous during annealing. The simulations are based on a recently developed tight-binding Hamiltonian for silicon which was optimized by fitting to a dataset of DFT/LDA total energies including the diamond and -Sn structures, unrelaxed point defects such as vacancies and interstitials and the concerted-exchange saddle point. This Hamiltonian accurately reproduces the relaxation energies of the various defects as well the the surface energies of some reconstructions of the (100) and (111) surfaces. The interface models consist of supercells containing 256 atoms, with 64 atoms in the crystalline and the remaining in the amorphous phase. Several samples were generated and the statistics of the interface structures were analyzed. The transition region from the bulk crustal to bulk amorphous is about 5 Åthick, and with a density lower than either of the adjacent phases. Focusing in on the atomistic structure of the interface, we were able to identify atoms in adjacent (110) chains in the crystal pairing up to form dimers, and short (110) chains beginning to form in the amorphous. We also report on calculations of saddle point configurations along paths connecting the initial configuration with configurations that have undergone a crystallization event. These calculations provide insight on the activation energy and state that can lead to SPEG.

4:45 PM R7.9 
THE EFFECTS OF THE ELECTRON-PHONON INTERACTION ON THE VIBRATIONAL ANOMALIES AND POLYMORPHISM IN TITANIUM. J.L. Gavartin and D. J. Bacon, Department of Materials Science and Engineering, The University of Liverpool, Liverpool, UNITED KINGDOM.

The group IV transition metals (titanium, zirconium and hafnium) possess a number of rather remarkable features in their properties attributed to strong anomalies in their phonon spectra. These include crystalline polymorphism, a high temperature saturation of the electrical conductivity, rapid increase of the constant pressure heat capacity at high temperature, and anisotropy in diffusion and thermal properties. The vibrational anomalies manifest themselves by a strong (and sometimes inverse) temperature dependence of the certain phonon frequencies even at the conditions far away from the phase transition. For example, the [0001]LO branch in the low temperature hexagonal -phase of Ti and Zr exhibits substantial softenning with temperature. The frequencies of the zone-boundary [001]T1 and [111]L () phonons in the high temperature bcc-structured -phase, also decrease drammatically while approaching the transition temperature from above. The established large sensitivity of the vibrational and elastic properties of d-metals to the d-band filling, as well as to the details of d-band structure, suggests strong electron-phonon interaction effects. We analyse both anharmonic and electron-phonon coupling effects in the different polytypes of Ti by means of tight-binding molecular dynamics and frozen phonon calculations. Based on these simulations, we discuss the static and dynamic variations of the electron chemical potential as applied to either the whole system or particular phonons, and the implications of these changes in the thermal frequency shifts and relative phase stability.