Active Learning of Microstructure-Property Relations in Hybrid Materials

Hybrid organic/inorganic materials play an important role in electronic technologies, where their functional properties are critically determined by the atomistic arrangement between compounds. The interface microstructure can be explored with first principle simulations at considerable computational cost. This procedure can be accelerated with active learning algorithms, by sampling configurations on-the-fly in the search for optimal structures. We encoded such a probabilistic algorithm into the Bayesian Optimization Structure Search (BOSS) Python tool for materials optimisation [1]. BOSS relies on a statistical surrogate model of materials properties to make smart decisions on sampling relevant microstructures. This makes it an effective tool for global exploration of materials energy and property spaces.<br/><br/>We combined BO with first-principles simulations to learn global energy landscapes and perform atomistic structure search [1]. This facilitated studies of hybrid functional materials such as ligand-protected clusters [2], surface adsorbates [3], thin film growth [4] and solid-solid interfaces [5] with modest dataset sizes. We also applied BOSS to map the energetics and dynamics of structural fluctuations inside halide perovskite lattices, and related them to changes in functional properties [6]. With recent multi-objective and multi-fidelity implementations for active learning, BOSS can make use of different information sources to learn materials properties at considerably reduced computational costs.<br/><br/>[1] M. Todorović, M. U. Gutmann, J. Corander, P. Rinke, npj Comput. Mater., 5, 35 (2019)<br/>[2] L. Fang, X. Guo, M. Todorović, P. Rinke, X. Chen, J. Chem. Inf. Model. 63, 745-752 (2023)<br/>[3] J. Järvi, B. Alldritt, O. Krejčí, M. Todorović, P. Liljeroth, P. Rinke, Adv. Func. Mater., 31, 2010853 (2021)<br/>[4] A. T. Egger, et al., Adv. Sci. 7, 2000992 (2020)<br/>[5] A. Fangnon, et al., ACS Appl. Mater. Interfaces 14 (10), 12758-12765 (2022)<br/>[6] J. Li, et al., accepted in Small Struct. (2024)