The design of innovative materials holds high potential for societal and technological advancement. In this context, the discovery of novel catalysts emerges as a prioritized need within many industrial applications to expedite the adoption of renewable energy sources, thereby contributing to mitigating global warming. However, exploring the whole chemical space for potential catalysts poses significant computational challenges. Furthermore, catalyst design involves a highly interdisciplinary technology that requires simultaneous control and comprehension of numerous interacting and conflicting design variables. The key limitation of traditional trial-and-error methods stems from their inefficiency and time-consuming nature, whereas computational quantum mechanical modeling methods incur substantial computational costs. A domain of growing importance pertains to the adoption of Artificial Intelligence (AI) and its related generative Machine Learning (ML) technologies, which represent a groundbreaking paradigm in materials science research. This paradigm aims to optimize the exploration of material relevance space, enhance knowledge about materials and their properties and accelerate the material discovery process through automated parallel/iterative procedures. This paper discusses the key challenges in AI-based material design/discovery and offers recommendations for future research directions.
Hosseini-Ardali et al. (Tue,) studied this question.