Rapid Design of Efficient Mn3O4-Based Photocatalysts by Machine Learning and Density Functional Theory Calculations

Abstract

The development of efficient photocatalysts for visible-light-driven pollutant degradation contributes to sustainable and green solutions to environmental challenges. However, optimizing catalyst composition and structure remains a costly and time-consuming process. Here, a comprehensive design strategy is presented for the fast development of efficient Al-doped Mn₃O₄-based photocatalysts, combining density functional theory (DFT), machine learning (ML), and laboratory experiments. DFT-calculated effective mass and bandgaps, serving as indicators of charge mobility and light harvesting, respectively, are employed as descriptors to determine the optimal Al dopant amount. Al_0.5Mn_2.5O₄ is identified as a promising candidate due to its favorable bandgap and charge mobility. To further enhance performance, Al_xMn_3−xO₄/Ag₃PO₄ heterojunctions are synthesized, leveraging ML to optimize the ratios between Al_xMn_3−xO₄ and Ag₃PO₄. The best material is determined to be an Al_0.5Mn_2.5O₄/35 wt%-Ag₃PO₄ composite, which exhibits a 27-fold increase in photocatalytic efficiency for methylene blue degradation under visible light compared to pristine Mn₃O₄. This study not only provided promising photocatalysts for practical pollutant degradation but highlighted the potential of computational and ML-guided approaches to accelerate photocatalyst discovery. These computational methods provide a framework for the rational design of advanced materials for environmental remediation applications.