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

Abstract

The observation of superconductivity in hydride-based materials under ultrahigh pressures (for example, H$_3$S and LaH$_{10}$) has fueled the interest in a more data-driven approach to discovering new high-pressure hydride superconductors. In this work, we performed density functional theory (DFT) calculations to predict the critical temperature ($T_c$) of over 900 hydride materials under a pressure range of (0 to 500) GPa, where we found 122 dynamically stable structures with a $T_c$ above MgB$_2$ (39 K). To accelerate screening, we trained a graph neural network (GNN) model to predict $T_c$ and demonstrated that a universal machine learned force-field can be used to relax hydride structures under arbitrary pressures, with significantly reduced cost. By combining DFT and GNNs, we can establish a more complete map of hydrides under pressure.

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.