Data-driven Design of High Pressure Hydride Superconductors using DFT and Deep Learning

Data and code associated with "Data-driven Design of High Pressure Hydride Superconductors using DFT and Deep Learning" by Daniel Wines and Kamal Choudhary, https://arxiv.org/abs/2312.12694raw-dft-hydride.tar.gz: Contains all of the raw DFT calculations reported in this study and used to train the ALIGNN modelalignn_hydride.tar.gz: Contains all relevant code and data for training both ALIGNN models presented in this work, one ALIGNN model trained only on hydride data and one trained on hydride+bulk JARVIS superconductor data (pretrained model from https://www.nature.com/articles/s41524-022-00933-1). The best pretrained models are provided for future use.ALIGNN_FF_hydride.ipynb: A notebook that uses ALIGNN-FF and ASE FIRE optimizer to optimize structures under pressure and predict Tc of the relaxed structures with the ALIGNN Tc models.phasediagram.tar.gz: Code and formation/total energy data (in eV/atom) to compute phase diagrams for selected systems.