A review article has been published in the prestigious journal Nature Reviews Chemistry, part of the renowned Nature Portfolio. The paper, entitled “Developing machine learning for heterogeneous catalysis with experimental and computational data”, highlights the transformative potential of machine learning (ML) in advancing heterogeneous catalysis — a key element of modern chemical and energy industries.

This work is the result of an international collaboration between IREC, the University of Toronto and The Matter Lab, and the article is led by Dr. Carlota Bozal-Ginesta along with Prof. ICREA Albert Tarancón from the Nanoionics and Fuel Cells Department at IREC.

The review comprehensively examines the emerging synergy between experimental and computational data in training ML models to optimize catalyst performance. Heterogeneous catalysis plays a critical role in sustainable energy and chemical processes, yet the rational design of new catalysts remains an immense scientific challenge. Traditional computational methods, particularly high-throughput quantum chemistry, have provided valuable insights, but their real-world impact has been limited by the complexity of catalyst structures and the scarcity of robust experimental datasets.

The review addresses this gap by systematically analyzing over 100 recent studies that integrate machine learning techniques with high-throughput experiments and simulations. The review explores trends in the types of descriptors used, materials and reactions studied, dataset sizes, and reported model performance, especially focusing on cases where ML led to verifiable experimental improvements.

The full article is available in Nature Reviews Chemistry and can be accessed at: https://www.nature.com/articles/s41570-025-00740-4

IREC acknowledges funding from a Marie Skłodowska Curie Actions Postdoctoral Fellowship grant (101064374) and support from the Generalitat de Catalunya (2021-SGR-00750, NANOEN).