First Principles Global Optimization Method From Parallel Tempering Molecular Dynamics

Abstract

Global optimization techniques are often based on a stochastic method to explore the potential energy surface of the investigated system. The here-described global optimization is based on a temperature-driven potential energy surface exploration via Parallel Tempering Born-Oppenheimer Molecular Dynamics (PT-BOMD) simulations. Additionally, structure selection from the lowest temperature PT-BOMD trajectory for further optimization is performed with a scheme based on the Discrete Cosine Transformation (DCT). This step can be done automatically and therefore removes a human bias. The influence of parameters for the DCT and the length of the PT-BOMD simulations is investigated with regard to their impact on the result of the global optimization.