Applied machine learning in heterogeneous catalysis  

The discovery of novel heterogeneous catalysts is a formidable task as their design and synthesis have predominantly used heuristic guidelines, followed by DFT aided simulations for mechanistic insights. However, the former is laborious and time consuming, while the latter is computationally expensive. The recent progress in the field of data science and machine learning (ML) presents alternative to these traditional approaches. To this end, in alignment with the National Centre of Competence in Research, NCCR Catalysis’s drive towards the adoption of digital and AI technologies, my research endeavours are two-fold: as a first step, my work focuses on creating unified data infrastructures with standardized data formats for various catalytic reactions, by applying digital tools and concepts of text mining. Once such databases are developed, I intend to use the state-of-art ML and Bayesian optimization algorithms for the guided synthesis of catalysts via predictive and prescriptive inference. Thus, my research aims to present a closed-loop framework of ‘standardized data - machine intelligence - catalyst design’- thereby contributing to digitization in the field of heterogeneous catalysis.