Synergies between Experimental and Computational Approaches in Materials Design: Importance and Challenges of Clean Data

Back

Annette Trunschke

Fritz-Haber-Institut der Max-Planck-Gesellschaft, Department of Inorganic Chemistry, Berlin, Germany, trunschke@fhi-berlin.mpg.de

Experimental data spaces are a key component of data-driven computational materials design.1 The data must meet high quality standards in order to be used as input or benchmark in data science applications. This means that the measurements must be reliable and reproducible and the data must be provided in a structured format that conforms to the FAIR (findable, accessible, interoperable and reusable) principles.

Data on functional materials in heterogeneous catalysis often do not meet these requirements.2 One intrinsic reason for this is to be found in the metastable and dynamic nature of catalysts in their working state. This will be illustrated using examples of complex catalytic reactions important for the sustainable synthesis of chemical intermediates and transformations relevant for energy conversion and storage in a future low-carbon economy.

In order to generate AI-ready data, there is a need to fundamentally change the way catalysis research is conducted. This includes the development of new methods for data acquisition, storage and transfer. We present a digitalization concept that involves working according to machine-readable Standard Operating Procedures (SOPs).3,4 The process of data collection, standardized analysis, uploading to a database, and establishing relationships between database entries is fully automated.4 Data exchange within a local data infrastructure and beyond to overarching repositories is enabled. This approach is instrumental in laying the experimental groundwork for the upcoming transition to autonomous materials development. It is shown that “clean data” generated in such a way in combination with interpretable machine learning methods lead to a deeper understanding of complex physico-chemical correlations (descriptors) that determine catalytic properties and thus can drive the discovery of new catalytically active materials.5

(1) Bauer, S.; Benner, P.; Bereau, T.; Blum, V.; Boley, M.; Carbogno, C.; Catlow, C. R. A.; Dehm, G.; Eibl, S.; Ernstorfer, R., et al. Roadmap on Data-Centric Materials Science. Modelling and Simulation in Materials Science and Engineering 2024, 32 (6), 063301, DOI: https://dx.doi.org/10.1088/1361-651X/ad4d0d

(2) Marshall, C. P.; Schumann, J.; Trunschke, A., Achieving Digital Catalysis: Strategies for Data Acquisition, Storage and Use. 2023, 62, (30), e202302971, DOI: https://doi.org/10.1002/anie.202302971

(3) Trunschke, A., et al., Towards Experimental Handbooks in Catalysis. 2020, 63, 1683-1699, DOI: https://dx.doi.org/10.1007/s11244-020-01380-2.

(4) Moshantaf, A.; Wesemann, M.; Beinlich, S.; Junkes, H.; Schumann, J.; Alkan, B.; Kube, P.; Marshall, C. P.; Pfister, N.; Trunschke, A., Advancing Catalysis Research through FAIR Data Principles Implemented in a Local Data Infrastructure – A Case Study of an Automated Test Reactor. 2024, DOI: https://dx.doi.org/10.1039/D4CY00693C.

(5) Foppa, L.; Rüther, F.; Geske, M.; Koch, G.; Girgsdies, F.; Kube, P.; Carey, S. J.; Hävecker, M.; Timpe, O.; Tarasov, A. V.; Scheffler, M.; Rosowski, F.; Schlögl, R.; Trunschke, A., Data-Centric Heterogeneous Catalysis: Identifying Rules and Materials Genes of Alkane Selective Oxidation. 2023, 145, (6), 3427-3442, DOI: https://dx.doi.org/10.1021/jacs.2c11117.