Accuracy of existing atomic potentials for the CdTe semiconductor compound

CdTe and CdTe-based Cd(1-x)Zn(x)Te (CZT) alloys are important semiconductor compounds that are used in a variety of technologies including solar cells, radiation detectors, and medical imaging devices. Performance of such systems, however, is limited due to the propensity of nano- and micro-scale defects that form during crystal growth and manufacturing processes. Molecular dynamics simulations offer an effective approach to study the formation and interaction of atomic scale defects in these crystals, and provide insight on how to minimize their concentrations. The success of such a modeling effort relies on the accuracy and transferability of the underlying interatomic potential used in simulations. Such a potential must not only predict a correct trend of structures and energies of a variety of elemental and compound lattices, defects, and surfaces but also capture correct melting behavior and should be capable of simulating crystalline growth during vapor deposition as these processes sample a variety of local configurations. In this paper, we perform a detailed evaluation of the performance of two literature potentials for CdTe, one having the Stillinger-Weber form and the other possessing the Tersoff form. We examine simulations of structures and the corresponding energies of a variety of elemental and compound lattices, defects, and surfaces compared to those obtained from ab initio calculations and experiments. We also perform melting temperature calculations and vapor deposition simulations. Our calculations show that the Stillinger-Weber parameterization produces the correct lowest energy structure. This potential, however, is not sufficiently transferrable for defect studies. Origins of the problems of these potentials are discussed and insights leading to the development of a more transferrable potential suitable for molecular dynamics simulations of defects in CdTe crystals are provided.