A Periodic Genetic Algorithm with Real-Space Representation for Crystal Structure and Polymorph Prediction

TL;DR

This paper introduces a novel genetic algorithm that employs real-space representation and a periodic crossover for accurate, ab initio crystal structure and polymorph prediction without prior assumptions on unit cell parameters.

Contribution

The paper presents a new genetic algorithm with a real-space approach and a periodic crossover, enabling true ab initio crystal and polymorph prediction without pre-set constraints.

Findings

Robust convergence to bulk structures from random initial states.

Successful discrimination between low enthalpy configurations.

Emergence of known polymorphs like Lonsdaleite and graphite in ab initio searches.

Abstract

A novel Genetic Algorithm is described that is suitable for determining the global minimum energy configurations of crystal structures and which can also be used as a polymorph search technique. This algorithm requires no prior assumptions about unit cell size, shape or symmetry, nor about the ionic configuration within the unit cell. This therefore enables true ab initio crystal structure and polymorph prediction. Our new algorithm uses a real-space representation of the population members, and makes use of a novel periodic cut for the crossover operation. Results on large Lennard-Jones systems with FCC- and HCP-commensurate cells show robust convergence to the bulk structure from a random initial assignment and an ability to successfully discriminate between competing low enthalpy configurations. Results from an ab initio carbon polymorph search show the spontaneous emergence of both Lonsdaleite and graphite like structures.