| 초록 |
Computational chemistry is widely used to investigate the atomic-scale properties of materials. Given that these studies are conducted on materials in their most stable state, it is critical to determine the optimal structure of the materials prior to conducting the studies. The optimal structure is the global minimum of the potential energy surface (PES), which is a function of atomic positions. However, finding the global minimum of PES is extremely difficult because PES is a nonlinear and complex function with multiple local minima and the system’s configurational space exponentially scales with the number of atoms. In this study, we propose an accelerated search method for determining the optimal atomic structure based on Bayesian optimization. This method takes advantage of both Bayesian and local optimization. The method’s performance is validated by demonstrating the progress of the search for the optimal structure of alloy (Ag and Au) in a simulation environment and comparing it to the performance of other methods. |
