(189q) Structural and Vibrational Properties of a Si- and Se- Induced 216-Atom Quasi-Random Ingaas

We are using a novel computational approach to study the structural and vibrational properties of Si and Se dopants and vacancies in randomly ordered 216-atom simulation boxes to represent In_0.5Ga_0.5As based on density function theory (DFT). Specifically, the random structure is modeled atomistically using a newly implemented efficient search and optimize algorithm. We want to compare and contrast the effect of Si and Se dopants in the vicinity of a vacancy to construct a model for a desirable n-type InGaAs. Different local environments for each case are determined using a modified cell list algorithm that I wrote. Vibrational frequencies are obtained from dynamic matrices using a Green’s function. And the dynamic matrices are calculated from weighted averages of local environments procured through two processes: a pure InGaAs system studied through the force field fitting method and a subsystem within the defect strain field investigated through the so-called “frozen phonon” method.

There are two unique aspects behind this project: the construction of a large quasi-random structure and obtaining vibrational frequencies through a “frozen phonon” approach. In the past, most researches have only studied ordered structures since it is surprisingly difficult to produce a truly random structure especially for a large supercell. A quasi-random structure considering multi-body clusters has never been constructed on such a large scale. Our approach combined an effective search of the multi-body (including two-body to four-body clusters) nearest neighbor (up to sixth nearest neighbor) with optimizing with a multi-objective simulated annealing to successfully generate a quasi-random system, that could contain thousands of atoms, at a very small computational cost. The second aspect that makes my research unique is the implementation of the frozen phonon approach. Due to the relatively large size of the system when compared to normal ab initio simulation systems, we cannot apply the phonon mode calculations in the existing modeling package. So I used an alternative way to solve this problem by applying the frozen phonon method for all the atoms within the dopant’s and/ or defects’ strain field to obtain the dynamic matrices and then further convert them to vibrational frequencies through a Green’s function.