Structural and Electronic Properties of a -GaAs: A Tight-Binding-Molecular-Dynamics-Art Simulation
- PDF / 1,308,949 Bytes
- 10 Pages / 417.6 x 639 pts Page_size
- 5 Downloads / 175 Views
NORMAND MOUSSEAU[b] Computational Physics - Faculty of Applied Physics, Technische Universiteit Delft, Lorentzweg 1, 2628 CJ Delft, The Netherlands
ABSTRACT By combining tight-binding (TB) molecular dynamics (MD) with the recently-proposed activation-relaxation technique (ART), we have constructed structural models of a-GaAs and a-Si of an unprecedented level of quality: the models are almost perfectly four-fold coordinated and, in the case of a-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 a-Si is best described by a Polk-type model, while aGaAs resembles closely the mechanical model proposed by Connell and Temkin, which is free of wrong bonds. In this paper, the structural, electronic, and dynamical properties of a-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 emphasising the need for realistic computer models.
INTRODUCTION The energetics, and thus the physical properties, of multicomponent materials is notably difficult to describe accurately using simple empirical potentials because of the large number of interactions involved. Because of this, with perhaps the exception of water and SiO 2 (and related chalcogenide glasses), very little effort has been spent in developing such potentials. In light of this, model calculations based on a tight-binding (TB) or ab-initio description of the interactions are no luxury but, rather, a much-needed first step in the study of innumerable, and important, multicomponent systems. However, the computational cost of ab-initio calculations, and to a lesser extent TB, increases rapidly with system size. This becomes a serious problem for systems which exhibit little (or no) symmetry., such as nano-structures, surfaces or disordered (chemically or topologically) materials, where large models are required and for which, therefore, the number of relevant configurations increases rapidly. One possible solution to this problem is the use of different approaches at various stages of a calculation. Thus, for instance, a firstapproximation structural model for a particular material might be generated using a set of classical empirical potentials, which can then be improved or confirmed a posteriori through further relaxation using more appropriate (semi-empirical or first-principles) interactions. This is the philosophy we adopt here. Mixed approaches have been used with some success in the study of disordered materials, such as a-Si, a-C, etc.[1]. Although leaving open a priori the possibility that the empirical potentials bias the results in a significant manner, these approaches probably provide, at present, and until computers and m
Data Loading...
