By: Amy Kaczmarowski, Shujiang Yang, Izabela Szlufarska, and Dane Morgan
Collaboration between groups within the Interdisciplinary Computational Group in the University of Wisconsin MRSEC program has led to the development of a real-space genetic algorithm tool that enables automated prediction of stable defect cluster structures embedded in a host matrix [1]. Very small clusters are important in many contexts, including being produced by irradiation, forming during doping, and representing the early stages of precipitation, but their structure is hard to predict given due to their many possible atomic configurations. The genetic algorithm technique exchanges chemical and positional data of the atoms in randomly generated interstitial and/or vacancy defect structures, ranks these different structures based on their potential energy relative to one another, and ultimately identifies a more globally stable structure based on the preferential survival of the lowest energy structures. This method efficiently reproduced small-size defects that had been previously identified as stable or metastable structures as well as predicted new mid-size defects in materials such as SiC, Fe, and Fe-Cr alloys. The structure optimization program (StructOpt) developed in this study is available under open source licensing as part of the MAterials Simulation Toolkit (MAST) [2], a multifaceted python-based platform for the computational analysis of materials also developed and maintained by ICG members.
[1] A. Kaczmarowski, S. Yang, I. Szlufarska, and D. Morgan, Genetic algorithm optimization of defect clusters in crystalline materials, Computational Materials Science 98, p. 234-244 (2015).
[2] T. Mayeshiba, H. Wu, and D. Morgan, MAterials Simulation Toolkit (MAST), 2014, https://pypi.python.org/pypi/MAST/