First Principles Prediction of Amorphous Phases Using Evolutionary Algorithms

TL;DR

This paper demonstrates that evolutionary algorithms combined with DFT can effectively predict amorphous structures, offering a stochastic alternative to traditional MD melt-quench methods with comparable accuracy.

Contribution

It introduces a Darwinian evolution-inspired stochastic approach for modeling amorphous materials, validated against established MD and experimental results.

Findings

Structural parameters match MD and experimental data within 2%

The method successfully models silicon and IGZO amorphous structures

Provides a viable alternative to deterministic melt-quench techniques

Abstract

We discuss the efficacy of evolutionary method for the purpose of structural analysis of amorphous solids. At present ab initio molecular dynamics (MD) based melt-quench technique is used and this deterministic approach has proven to be successful to study amorphous materials. We show that a stochastic approach motivated by Darwinian evolution can also be used to simulate amorphous structures. Applying this method, in conjunction with density functional theory (DFT) based electronic, ionic and cell relaxation, we re-investigate two well known amorphous semiconductors, namely silicon and indium gallium zinc oxide (IGZO). We find that characteristic structural parameters like average bond length and bond angle are within $\sim$ 2% to those reported by ab initio MD calculations and experimental studies.