Automated Search Strategy for Novel Ordered Structures of Block Copolymers
Qingshu Dong, Zhanwen Xu, Qingliang Song, Yicheng Qiang, Yu Cao, Weihua Li
Abstract
Block copolymers with different architectures can possibly generate innumerable stable or metastable structures and thus provide an irreplaceable platform for theoretically exploring novel structures. Self-consistent field theory (SCFT) is a powerful tool to predict the ordered structures of block copolymers; however, it is sensitively dependent on its initial condition. Here we propose to use multiple symmetry-adapted basis functions to generate the initial conditions of SCFT and then apply Bayesian optimization to search for ordered structures by navigating the coefficient space of these basis functions. Without any prior knowledge, our scheme can automatically recover hundreds of ordered structures for two simple block copolymers, including most of the common structures and complex Frank-Kasper structures, together with many novel structures. By applying the automated scheme to various block copolymers, a huge number of novel structures can be obtained to expand the structural library, which may create new opportunities for the scientific community.