Applying Genetic Algorithm to Material Science Problems

Abstract

The design and optimization of materials is a common theme that occurs in many different energy-related problems. In particular, the prediction of crystal structures from their chemical composition has long been recognized as one of the outstanding challenges in theoretical solid state physics. In the past two decades a number of algorithms have been proposed to tackle the challenging problem of crystal structure prediction. These algorithms include simulated annealing, genetic algorithm (GA) and basin or minima hopping. Among these algorithms, the GA has proved to be a powerful approach to predict material structures using only first principles calculations and knowledge of the chemical composition of the system. GA mimics the biological evolution process to solve optimization problems. During the GA optimization process, inheritance, mutation, selection, and crossover operations are included to produce new and better structures from generation to generation. While all GA implementations follow the same general strategy, the details of the individual operations can vary a lot from problem to problem and can be critical to search efficiency. The speaker and his colleague David Deaven introduced a GA scheme for global structure optimization by adopting a real space representation of the problem and a geometrical cut-and-paste operation to generate off-springs from parent structures in the pool. Candidate structures are relaxed to the nearest local minimum before the selection process to simplify the structure configuration space. Originally this method was only applied to cluster systems. Recently this method has been generalized to explore systems with periodic boundary conditions (PBC) such as surfaces, interfaces, nanowires, and crystals. In this lecture, the speaker will review the development of this GA scheme and applications to a number of material science problems.