Optimization of SiO2 with GHA and basin hopping

Abstract

In this paper we continue to develop our structural optimization algorithm built earlier on a numerical platform, the Genetic Hybrid Algorithm (GHA). Our goal now is to extend our algorithm to oxides, find an effective way to search for the known global minimum, alpha-quartz as a test case, and report our results and findings for this system. We studied unit cells of different sizes: 18, 36 and 72 atoms, but most of the presented results are for cases with 18 and 36 atoms. The algorithm makes heavy use of the basin hopping method for searching for the global minimum of the system. We show how we were able to apply basin hopping most effectively in this case and which variables were of importance.We identify three other low energy structures near the global minimum structure, that trap the search. We show that the energy guided basin hopping can be detrimental to the search and structure-based guiding works more reliably. Two different structure based guides were used, one that tries to maximize the shortest silicon–silicon bond in the cell, while the other tries to maximize the calculated order parameter. The guiding was implemented by generating multiple different options for the basin hopping jumps, and doing weighted choosing on those options based on their properties.