Efficient Autoencoder Pipeline for Discovering High Entropy Alloys with Molecular Dynamics Data

Abstract

In this work, we utilize computationally efficient Molecular Dynamics (MD) simulations to create a machine learning pipeline for discovery of crystalline multi- component alloys. We employ high-quality interatomic potentials to create a dataset of NiFeCr structures and apply Crystal Diffusion Variational Autoencoder (CDVAE) to maximize their mechanical properties, i.e. bulk modulus. As part of the experiment, we utilize local search coupled with classical interatomic potentials to explore the local structure space and show that utilization of this procedure greatly improves optimization capability of the neural model. We also expand the model with an extra submodule, which attains 42% improvement on modeling the crystalline phase of the structures. Ultimately, we verify the global stability of the created structures with quantum mechanical calculation methods.