Generative deep learning for the inverse design of materials

TL;DR

This paper reviews how generative deep learning techniques enable inverse design of materials by modeling structure-property relationships, discussing methods, challenges, and future perspectives in the field.

Contribution

It provides a comprehensive overview of generative deep learning approaches for inverse materials design, focusing on structure-property mapping and key methodological elements.

Findings

Latent space construction is crucial for generative models.

Generative learning approaches vary based on material types.

Challenges include high computational costs and data limitations.

Abstract

In addition to the forward inference of materials properties using machine learning, generative deep learning techniques applied on materials science allow the inverse design of materials, i.e., assessing the composition-processing-(micro-)structure-property relationships in a reversed way. In this review, we focus on the (micro-)structure-property mapping, i.e., crystal structure-intrinsic property and microstructure-extrinsic property, and summarize comprehensively how generative deep learning can be performed. Three key elements, i.e., the construction of latent spaces for both the crystal structures and microstructures, generative learning approaches, and property constraints, are discussed in detail. A perspective is given outlining the challenges of the existing methods in terms of computational resource consumption, data compatibility, and yield of generation.