The increasing demand for clean and renewable energy has intensified the exploration of advanced materials for efficient photocatalysis, particularly for water splitting applications. Among these materials, MXenes, a family of two-dimensional (2D) transition metal carbides and nitrides, have shown great promise. This study leverages machine learning (ML) to address the resource-intensive process of predicting the bandgap of MXenes, which is critical for their photocatalytic performance. Using an extensive data set of 4356 MXene structures, we trained multiple ML models and developed a robust classifier-regressor pipeline that achieves a classification accuracy of 92% and a mean absolute error (MAE) of 0.17 eV for bandgap prediction. This framework, implemented in an open-source Python package, MXgap, has been applied to screen 396 La-based MXenes, identifying six promising candidates with suitable band alignments for water splitting whose optical properties were further explored via optical absorption and solar-to-hydrogen (STH) efficiency. These findings demonstrate the potential of ML to accelerate MXene discovery and optimization for energy applications.

The authors acknowledge financial support from the Spanish Ministerio de Ciencia e Innovación and Agencia Estatal de Investigación (AEI) MCIN/AEI/10.13039/501100011033 and, as appropriate, by “European Union Next Generation EU/PRTR”, through grants PID2021-126076NB-I00 and TED2021-129506B-C22, the unit of excellence María de Maeztu CEX2021-001202-M granted to the IQTCUB, the COST Action IG18234, and the Generalitat de Catalunya 2021SGR00079 grant. The authors also acknowledge the computational resources provided by Consorci de Serveis Universitaris de Catalunya (CSUC), with financial support from Universitat de Barcelona. Also, F.V. thanks the ICREA Academia Award 2023 ref. Ac2216561, and D.O. thanks Universitat de Barcelona for a predoctoral contract (PREDOCS-UB).

Peer reviewed