Density Functional Theory and Data-Driven Insights into Hydrogen Incorporation and Transport in Ternary ABO3 Perovskites

Abstract Hydrogen is a promising energy carrier due to its clean and versatile nature, offering opportunities for decarbonizing diverse sectors. Hydrogen utilization in many applications relies on identifying and understanding materials with exceptional hydrogen storage and transport properties. In this study, we conducted a comprehensive investigation of the thermodynamic and kinetic properties of hydrogenation reactions in ternary ABO3 perovskites using density functional theory (DFT) calculations and machine learning (ML) screening. We performed DFT simulations for a diverse range of perovskites featuring a variety of A-site cations (Ca, Sr, Ba) and B-site cations (Co, Fe, Mn, Ce, Ti, Zr) across various phases (cubic, tetragonal, orthorhombic, and hexagonal). By computing all data at the same theoretical level, we establish a consistent framework to compare how crystal phase, lattice-site chemistry, and host electronic structure govern hydrogen incorporation and diffusion. Therefore, we identify hexagonal SrMnO3, cubic SrFeO3, and hexagonal BaFeO3 as particularly promising candidates, combining favorable hydrogenation thermodynamics (ΔG < 1 eV) with low rotation and hopping barriers (<0.2 eV), while orthorhombic phases, although thermodynamically preferred, suffer from proton trapping, highlighting the importance of phase engineering toward cubic/hexagonal frameworks. Ab initio molecular dynamics further support this phase-dependent kinetic picture by showing relatively facile H diffusion in cubic SrFeO3 but much slower proton motion in orthorhombic BaCeO3 under the same conditions. By incorporating spatial and electronic descriptors with the DFT data into our ML models, we further develop a mechanistically motivated ML model based on structural and electronic descriptors to screen hydrogenation thermodynamics across 111 ternary ABO3 compositions. This model led to the identification of two additional promising candidates, hexagonal BaPdO3 and cubic BaSnO3, which were subsequently validated by DFT calculations. These findings provide a foundation for the rational design of ABO3 perovskites and demonstrate the potential of ML-driven frameworks with chemical insights to accelerate materials discovery in energetic and catalytic applications.