Anisotropic Collective Variables with Machine Learning Potential for Ab Initio Crystallization of Complex Ceramics

YP Deng and SB Fu and JR Guo and X Xu and H Li, ACS NANO, 17, 14099-14113 (2023).

DOI: 10.1021/acsnano.3c04602

Enhanced sampling molecular dynamics (MD) simulationshave beenextensively used in the phase transition study of simple crystallinematerials, such as aluminum, silica, and ice. However, MD simulationof the crystallization process for complex crystalline materials stillfaces a formidable challenge due to their multicomponent induced multiphaseproblem. Here, we realize the ab initio accuracyMD crystallization simulations of complex ceramics by using anisotropiccollective variables (CVs) and machine learning (ML) potential. Theanisotropic X-ray diffraction intensity CVs provide precise identificationof complex crystal structures with detailed crystallography information,while the ML potential makes it feasible to further perform enhancedsampling simulations with ab initio accuracy. Weverify the universality and accuracy of this method through complexceramics with three kinds of representative structures, i.e., Ti3SiC2 for the MAX structure, zircon for the mineralstructure, and lead zirconate titanate for the perovskite structure.It demonstrates exceptional efficiency and ab initio quality in achieving crystallization and generating free energysurfaces of all these ceramics, facilitating the analysis and designof complex crystalline materials.

Return to Publications page