First principles molecular dynamics study of liquid and amorphous silicon

main objective of this work was understanding the physics of the disordered silicon phases, i.e. liquid (1-Si) and amorphous (a-Si) silicon at the microscopic level using ab-initio methods. The reason for studying this subject is twofold. Silicon is the most studied prototype of elemental semiconductor. Its crystalline phases, including the high pressure metallic phases, have been extensively studied (Yin and Cohen 1982b, Chang and Cohen 1985). The metallic liquid phase, however, was relatively little explored, both experimentally ( Gabathuler and Steeb 1979, Waseda and Suzuki 1975, Hague et al 1980) and theoretically (Stillinger and Weber 1985, Car and Parrinello 1987, Hafner and Kahl 1984, Allan and Broughton 1987). The crystal to liquid transition occurs at the unusually high temperature of rv 1700K, which makes experiments difficult to make. Nevertheless, 1-Si is physically interesting. The atomic structure is dissimilar to that of most other liquid metals. The coordination number of rv 6.4 is intermediate between that characteristic of open tetrahedral systems and the highly coordinated closed packed systems, raising the question of the nature of chemical bonding. The reason for the metallic character of the melt and its electrical conductivity have been only poorly understood. The dynamical properties of 1-Si are essentially unknown. Much more work has been devoted to a-Si in the past twenty years. There is a vast number of physically interesting properties of a-Si. They include structural, dynamic, electronic, and defect-controlled properties (Elliot 1984, Zallen 1983). Until recently, the modeling of a-Si at the microscopic level was a highly empirical procedure. As a further step in this old but never fully understood field one would like to have a nonempirical scheme to study these properties. The other motivation for this work is that the disordered Si phases besides being physically interesting are also technologically important. Single crystals are grown from the liquid phase. Novel dopant profiles are generated by zone refining at advancing liquid interfaces in laser-melted Si surfaces. The metallic nature of the melt is used to determine the thickness of the laser-melted region and an alternating magnetic field is used to improve the quality of crystal growth. While crystalline Si ( c-Si) is of overwhelming importance in electronic industry and in special photovoltaic applications, it is too expensive for use in large-scale solar energy applications. On the other hand, large-area films of a-Si to be used in solar energy application can be prepared at a significantly lower cost than c-Si. The properties we intend to study range from structural to dynamical and electronic properties in a wide range of temperatures. A first principles approach that allows to treat all these properties on the same footing was pioneered by Car and Parrinello (1985). This method is a first-principles molecular dynamics scheme where the interatomic forces are calculated from the potentials that are derived from the instantaneous electronic ground-state calculated within density functional theory in the local density approximation. Thus no arbitrary assumptions are made on the form of the many-body potentials. We applied this technique to study the properties of 1- and a-Si. For both systems the resulting atomic structure compares very favorably with the experimental data. The knowledge of atomic coordinates enables to study other correlation functions, like, e.g., the triplet correlations, that are difficult to obtain experimentally. A vibrational spectrum for 1-Si was obtained which is quite dissimilar to that of most simple liquids. The calculated diffusion coefficients are in good agreement with indirect experimental estimates. The calculated electronic density of states and electrical conductivity show metallic behavior in agreement with experiments. A significant portion of covalent chemical bonds was found in the metallic 1-Si. The prevalence of broken bonds, however, leads to a high value of the diffusion coefficient and to the metallic nature of the melt. Both vibrational and electronic densities of states of a-Si agree well with available experimental data. Our MD simulation has revealed the existence of interesting mechanisms of defect dynamics that correlate well with proposed theoretical models. The thesis is organized as follows. The method is described in chapter 2. Chapter 3 contains a detailed study of 1-Si including structural, dynamical, bonding and electronic properties. The results on a-Si as well as the structural changes occurring upon cooling are presented in chapter 4. Finally, we present our conclusions in chapter 5.