Machine Learning Framework and Materials Database for the Discovery of Earth-Abundant Electrolyser Materials
Erik Griffioen a b, Rodrigo García-Muelas a, Vinzenz Koning c d, René van Roij b e, Nongnuch Artrith a
a Materials Chemistry and Catalysis, Debye Institute for Nanomaterials Science, Utrecht University. David de Wied building, University Road 99, 3584 CG Utrecht, the Netherlands.
b Institute for Theoretical Physics (ITF), Utrecht University. Princeton Square 5, 3584 CC Utrecht, the Netherlands.
c Copernicus Institute of Sustainable Development, Utrecht University. Princeton Av. 8a, 3584 CB Utrecht, the Netherlands.
d Centre for Complex Systems Studies, Utrecht University. Minnaert building, Leuvenlaan 4, 3584 CE Utrecht, the Netherlands.
e Soft Condensed Matter, Debye Institute for Nanomaterials Science, Utrecht University, Princeton Square1, 3584 CC Utrecht, the Netherland.
Proceedings of International Conference on Frontiers in Electrocatalytic Transformations (INTERECT22)
València, Spain, 2022 November 21st - 22nd
Organizers: Sara Barja, Nongnuch Artrith and Matthew Mayer
Poster, Rodrigo García-Muelas, 029
Publication date: 11th October 2022

Machine learning is a powerful tool to accelerate the computational characterization and prediction of new electrocatalysts and energy materials. Here we present progress toward a computational machine-learning framework for the discovery of low-cost, earth-abundant and stable electrolyser materials, aiming to identify economically optimal catalysts for the oxygen and hydrogen evolution reactions (OER, HER). Our framework makes use of first-principles thermodynamics data from the OC22 database [1], which includes transition and post-transition metals and their alloys, among other reactive compounds. Reaction kinetics data is approximated based on extended Bronsted-Evans-Polanyi relationships. The predicted OER and HER activity is evaluated in the context of the expected cost of the raw materials and their abundance in the Earth’s crust. Although the framework presently only evaluates transition metals, it is flexible and modular. This allows an iterative refinement to evaluate further reactive compounds through a general active learning process. Once completed, the database and framework will be shared with the scientific community in the spirit of the FAIR principles of Findability, Accessibility, Interoperability, and Reusability.

This work was funded through a Pathways to Sustainability - Energy in Transition Hub Seed project by the Utrecht University. We thank the Dutch supercomputing consortium SURFsara for providing generous computer resources in its flagship cluster, Snellius.  

We use our own and third party cookies for analysing and measuring usage of our website to improve our services. If you continue browsing, we consider accepting its use. You can check our Cookies Policy in which you will also find how to configure your web browser for the use of cookies. More info