PCTpro: A Machine learning model for rapid prediction of Pressure-Composition-Temperature (PCT) isotherms

We present a machine-learning powered Pressure-Composition-Temperature isotherm Predictor (PCTpro) for metal compositions. To train the PCTpro, an experimental database of PCT isotherms (MH-PCT) is built from published literature. The database comprises over 14,000 data points extracted from 237 PCT isotherms, representing 138 distinct compositions. The dataset encom- passes more than 25 elements and spans a broad spectrum of absorption tem- peratures (263-653 K) and hydrogen pressures (0.001 to 40 MPa). The feature set includes weighted average of elemental properties, hydriding properties, and experimental parameters like absorption temperature and hydrogen pressure. The comprehensive feature set equips PCTpro to predict the PCT isotherms for a given composition. The model is validated on a wide range of alloy fam- ilies and its predictions are consistent with experimental results. The model also captures temperature-dependent variations in plateau pressure, enabling determination of enthalpy and entropy of hydride formation through Van’t Hoff plots. We also demonstrate that training PCTpro on a subset of relevant data provides improved PCT isotherm prediction. Hence, PCTpro can be used as an ML tool for guiding PCT experiments, offering PCT isotherm predictions and valuable thermodynamic insights into materials suitable for solid-state hydrogen storage.