SPECIALS · Materials map for microstructures of inorganic materials

Machine learning (ML) is an emerging tool to accelerate the materials discovery, a crucial part of the global competitiveness of Europe. Conventional materials discovery has been based on voyages of chemists and materials scientists in the materials space. They have navigated themselves according to their experience-based maps in their brains to find novel materials from innumerable candidate materials. Then, how can novice researchers navigate themselves? The project SPECIALS (materialS maP for microstructurEs of inorganiC materIALS) will provide a map by leveraging expertise of the host and the experienced researcher (ER): inorganic chemistry, physical chemistry, and computer science. The project approach is distinct from the conventional ML approaches: the ER will focus on microstructures of target materials, because functionalities of the inorganic materials are microstructure-dependent. The target materials are battery materials considering the experiences of the ER. The map will be developed by ML to describe their structural changes upon charge and discharge in continuous manners so that novice researchers can easily compare similarities/dissimilarities of various materials. Required datasets will be provided by the host utilizing existing simulation results and new simulation to cover even amorphous structures. Because the simulation of microstructures is computation intensive, multi-fidelity methods will be applied to leverage various types of simulations at varying computational cost and accuracies. The research outcome has potential to be grown up to scientific and societal impact on the European battery research community. The project also focuses on the two-way transfer of knowledge by transferring data science knowledge of the ER to the host, and simulation and soft skills of the host to the ER. This two-way transfer will enhance the employability of the ER and the value of the host.