In recent times, the application of artificial intelligence (AI) in the discovery and development of new compounds and materials has demonstrated its effectiveness. Here, we present an AI-guided approach for expediting the virtual screening of two-dimensional (2D) materials, aimed at accelerating the identification of novel materials with specific physical and chemical attributes. As a demonstration, we create a publicly accessible database, the Virtual 2D Materials Database (V2DB), of potentially stable 2D materials and their AI-predicted essential physicochemical properties [1].

Next, we focus on identifying the most promising functional 2D materials for applications in energy conversion and storage. For instance, 2D materials are highly promising for advancing photocatalytic water splitting for hydrogen (H₂) production. However, efficiently exploring and identifying candidate 2D materials from the vast chemical space, including diverse structural and elemental combinations, is a daunting challenge in materials research. To address this challenge, we employ a data-driven approach, utilizing the V2DB as the primary source for 2D materials, to search for stable candidates possessing both appropriate band gaps and optimal photocatalytic properties for overall water splitting [2,3]. This approach includes a comprehensive virtual screening process that integrates machine learning, high-throughput density functional theory (DFT), hybrid-DFT, and GW calculations. As a key result of employing this approach, we have identified 27 novel 2D materials, with 16 of them exhibiting direct band gaps, as promising candidates for photocatalytic production of H₂. Further, an in-depth analysis of their properties related to solar water splitting, such as electronic and optical features, solar-to-hydrogen conversion efficiency, and carrier mobility, is conducted. These studies not only unveil new 2D photocatalysts but also showcase the effectiveness of a data-driven approach in systematically exploring materials. Therefore, this approach is useful in identifying materials with the desired properties for various energy applications within previously unexplored chemical spaces.

References

[1] M.C. Sorkun, S. Astruc, J.M.V.A. Koelman, S. Er npj Computational Materials, 2020, 6, 106.

[2] Y. Wang, M.C. Sorkun, G. Brocks, S. Er ChemRxiv, 2023, doi:10.26434/chemrxiv-2023-5pssn.

[3] Y. Wang, G. Brocks, S. Er ACS Catalysis, 2024, 14, 1336.