Stochastic Surface Walking Method for Reaction Sampling and the Combination with Machine Learning

Zhipan Liu
Fudan University

In this talk, I will introduce our latest progress for the Stochastic Surface Walking method (SSW), and apply the method for resolving the reaction pathways in catalysis and solid phase transition. The SSW method is designed for the global optimization of structure on potential energy surface (PES), while maintaining the pathway information during structure search. By adding bias potentials and performing local relaxation repeatedly, SSW method can perturb smoothly the structure from one minimum to another following a random direction. The SSW method in combination with double-ended transition state method can be utilized for finding unknown structures and predicting chemical reactivity from molecules to solids. Using these methods, we recently studied a number of important systems, e.g. ZrO2 tetragonal-to-monoclinic phase transition, heterophase junction structures in photocatalysts, and dynamic catalyst structure evolution in H2 evolution. At the end of the talk, I will discuss the possibility to combine SSW method with neural network technique to build neural network potential to speed up the search for global PES and reaction sampling.

Back to Workshop II: Stochastic Sampling and Accelerated Time Dynamics on Multidimensional Surfaces