Structure prediction of boron-doped graphene by machine learning.

Thaer M Dieb,Zhufeng Hou,Koji Tsuda

doi:10.1063/1.5018065

Abstract

Heteroatom doping has endowed graphene with manifold aspects of material properties and boosted its applications. The atomic structure determination of doped graphene is vital to understand its material properties. Motivated by the recently synthesized boron-doped graphene with relatively high concentration, here we employ machine learning methods to search the most stable structures of doped boron atoms in graphene, in conjunction with the atomistic simulations. From the determined stable structures, we find that in the free-standing pristine graphene, the doped boron atoms energetically prefer to substitute for the carbon atoms at different sublattice sites and that the para configuration of boron-boron pair is dominant in the cases of high boron concentrations. The boron doping can increase the work function of graphene by 0.7 eV for a boron content higher than 3.1%.

Highlights

Chemical composition modification of materials by incorporating additional elements is one of the commonly used approaches to tune the material properties
From the determined stable structures, we find that in the free-standing pristine graphene, the doped boron atoms energetically prefer to substitute for the carbon atoms at different sublattice sites and that the para configuration of boron-boron pair is dominant in the cases of high boron concentrations
We have employed the Monte Carlo tree search (MCTS) with Bayesian rollout to search the stable structures of B-graphene for the boron concentration up to 31.25%

Summary

INTRODUCTION

Chemical composition modification of materials by incorporating additional elements is one of the commonly used approaches to tune the material properties. To achieve the desired material properties, the optimal composition and the spatial distribution of incorporated atoms is demanded to be discovered. Because of the complexity introduced by the interplay between structural and chemical degrees of freedom, the atomic structure search of dopant atoms in the host material is very challenging, for the non-diluted doping level. To search the stable atomic structures of B-graphene more efficiently, in this work we propose to use a machine learning approach that utilizes Monte Carlo tree search, which was recently utilized for materials and chemical design with a rollout schema depending on Bayesian optimization (BO)..

MACHINE LEARNING BASED OPTIMIZATION METHOD

COMPUTATIONAL DETAILS

Obtaining optimal structures

Structure stability of boron-doped graphene

Electronic structures of boron-doped graphene

CONCLUSION