Accelerated Global Potential Energy Landscape Discovery: A Hybrid Genetic Algorithm and Stochastic Surface Walking Approach

Abstract

Efficiently exploring the global potential energy surface (PES) of different materials has been a challenging yet very important task for theoretical simulation. By accurately characterizing the global PES and strategically combining the genetic algorithm (GA) with the stochastic surface walking (SSW) method, here we propose a novel method for global PES exploration, achieving efficient localization of the global minimum. Notably, by leveraging the extensive configuration network sampled during the search, this framework further enables the automated identification of long-range, low-energy transition pathways between distant superbasins, providing deep insights into the structural evolution mechanisms. Through tests on different systems, such as atomic clusters, ligand-protected clusters, molecular clusters, surface-supported clusters, atomic crystals, and molecular crystals, it has been proven that the new algorithm is highly effective in describing the overall characteristics of the PES. It has shown significant improvements in exploring the global PES compared with standalone GA and SSW methods, especially for complex and large-scale chemical systems.