Materials Discovery Search Strategy
Novel functional materials are at the core of next-generation energy and environmental technologies. With rapidly improving materials synthesis and characterization capabilities comes increasing research interest in computational materials discovery.
Generally, computational materials discovery consists of four major components: structure generation, structure representation, properties calculation, and search strategy. While researchers have traditionally focused on the first three components, making exciting advancements through artificial intelligence technologies, the search strategy component hasn’t been extensively studied in the literature—despite being at the core of computational materials discovery.
Our research focuses on developing advanced search strategies for accelerated materials discovery. The main research theme is pushing the accuracy-efficiency Pareto front (see above) toward the high-accuracy and high-efficiency region. We have proposed various approaches, including mathematical optimization, model-based metaheuristic, and reinforcement learning methods.
It’s crucial to note that accuracy and efficiency are not the only important aspects of materials discovery. Other critical factors—such as scalability, generalizability, interpretability, and synthetic accessibility—must be considered in practical applications of computational materials discovery. From a methodological perspective, this means we need to go beyond applying advanced computational methods (e.g., “AI4Science” approaches) to incorporate physics into problem-solving methodologies (“Science4AI”). This integration enables us to tackle challenges that arise when not exploiting inherent physical problem structures and constraints.
Back to top