Finding the stable structures of 2D hexagonal materials with Bayesian optimization: Beyond the structural relationship with 3D crystals in weakly-bonded binary systems

Abstract

The graphene-graphite relationship in structural geometry is a basic principle to predict novel two-dimensional (2D) materials. Here, we demonstrate that this is not the case in binary metallic systems. We use the Bayesian optimization framework combined with the density-functional theory approach to determine the stable configuration of atomic species on a hexagonal plane. We show that the optimized structure of 2D Cu-Au exhibits the hexagonal lattice of a hexagonal ring of Cu atoms containing one Au atom, where the number of the Cu atoms is larger than that of the Au atoms in the unit cell, which is difficult to speculate from the atomic distribution of CuAu in the L10 structure. We also show that 2D Cu-X with X = Be, Zn, and Pd have hexagonal or elongated rings containing different atoms in the unit cell. Based on the binary Lennard-Jones model, we propose that such structures can appear for weakly-bonded systems located in between the phase-separated and strongly-bonded systems with the interatomic interaction energy between different species.