A Generative Approach to Materials Discovery, Design, and Optimization

Abstract

Despite its potential to transform society, materials research suffers from a major drawback: its long research timeline. Recently, machine-learning techniques have emerged as a viable solution to this drawback and have shown accuracies comparable to other computational techniques like density functional theory (DFT) at a fraction of the computational time. One particular class of machine-learning models, known as “generative models”, is of particular interest owing to its ability to approximate high-dimensional probability distribution functions, which in turn can be used to generate novel data such as molecular structures by sampling these approximated probability distribution functions. This review article aims to provide an in-depth understanding of the underlying mathematical principles of popular generative models such as recurrent neural networks, variational autoencoders, and generative adversarial networks and discuss their state-of-the-art applications in the domains of biomaterials and organic drug-like materials, energy materials, and structural materials. Here, we discuss a broad range of applications of these models spanning from the discovery of drugs that treat cancer to finding the first room-temperature superconductor and from the discovery and optimization of battery and photovoltaic materials to the optimization of high-entropy alloys. We conclude by presenting a brief outlook of the major challenges that lie ahead for the mainstream usage of these models for materials research.

Inverse Design of Next-Generation Superconductors Using Data-Driven Deep Generative Models

“…After superconductivity was discovered in 1911 by Onnes, the efforts to identify novel superconducting materials with high transition temperatures ( T c ) has been an intense area of research in materials science and condensed matter physics. , There have been systematic computational efforts to identify Bardeen–Cooper–Schrieffer (BCS) conventional superconductors , with high T c prior to costly experimental investigation, − where density functional theory-perturbation theory (DFT-PT) calculations have been performed to obtain the electron–phonon coupling (EPC) parameters. In addition, various machine learning approaches have been utilized to accelerate the search for high- T c superconductors. ,− However, these typical funnel-like screening-based approaches are not sufficient for inverse materials design, where, instead of engineering from structure to property, the goal is to engineer from a target property to the crystal structure.…”

The use of generative models to speed up the discovery of materials

“…The outcome of the models has also been validated through computational methods, demonstrating the utility of the AI-based tool. The presented work might be an important contribution to the Gen-erative-based methodologies (Menon & Ranganathan, 2022) for materials design and a first step towards the AI-driven design of high-performance materials based on industrial requirements or resource availability constraints. Once developed, the contrast between the time required for a compound discovery is profound.…”

Topological superconductors from a materials perspective

“…Despite the OOD problem, we can leverage various OOD detection models through ensemble learning [185] as a confidence predictor for absent, null phases. This predictor can be used to screen the existing material database or incorporated into a generative adversarial model, which has been highly successful in the field of image generation, to generate new candidates for materials not present in the training phases [186][187][188]. In the longer term, as TSC materials need to be confirmed, popular ML methods can further accelerate predictions for even more TSC candidates, forming a positive feedback loop toward accelerated discovery.…”