Inverse Design of Next-Generation Superconductors using Data-Driven Deep Generative Models

Over the past few decades, finding new superconductors with a high critical temperature (T_c)has been a challenging task due to computational and experimental costs. In this work, we present a diffusion model inspired by the computer vision community to generate new superconductors with unique structures and chemical compositions. Specifically, we used a crystal diffusion variational autoencoder (CDVAE) along with atomistic line graph neural network (ALIGNN) pretrained models and the Joint Automated Repository for Various Integrated Simulations (JARVIS) superconducting database of density functional theory (DFT) calculations to generate new superconductors with a high success rate. We started with a DFT dataset of ~1000 superconducting materials to train the diffusion model. We used the model to generate 3000 new structures, which along with pre-trained ALIGNN screening results in 62 candidates. For the top candidate structures, we carried out further DFT calculations to validate our findings. We extended this approach to high-pressure hydride superconductors, utilizing our deep learning and DFT workflows to discover new high-T_c hydride-based structures outside of the initial training. Our approaches go beyond the typical funnel-like materials design approaches and allow for the inverse design of next-generation materials.<br/>[1] https://arxiv.org/abs/2304.08446<br/>[2] https://jarvis.nist.gov/<br/>[3] https://github.com/usnistgov/alignn<br/>[4] https://github.com/txie-93/cdvae