Diffusion Models for Conditional Generation of Hypothetical New Families of Superconductors

Samuel Yuan; S. V. Dordevic

doi:10.1038/s41598-024-61040-3

← Back to search

Diffusion Models for Conditional Generation of Hypothetical New Families of Superconductors

Samuel Yuan, S. V. Dordevic

arXiv:2402.00198 · DOI 10.1038/s41598-024-61040-3 · 2024-01-31 · Scientific Reports

T3experimental12 citations

Active bibliographic source — not scientific approval

Bibliographic access preserves source history; it does not approve extracted materials or validate reported claims. Review warnings on each occurrence separately.

Abstract

Effective computational search holds great potential for aiding the discovery of High-Temperature Superconductors (HTSs), especially given the lack of systematic methods for their discovery. Recent progress has been made in this area with machine learning, especially with deep generative models, which have been able to outperform traditional manual searches at predicting new superconductors within existing superconductor families but have yet to be able to generate completely new families of superconductors. We address this limitation by implementing conditioning -- a method to control the generation process -- for our generative model and develop SuperDiff, a Denoising Diffusion Probabilistic Model (DDPM) with Iterative Latent Variable Refinement (ILVR) conditioning for HTS discovery -- the first deep generative model for superconductor discovery with conditioning on reference compounds. With SuperDiff, by being able to control the generation process, we were able to computationally generate completely new families of hypothetical superconductors for the very first time. Given that SuperDiff also has relatively fast training and inference times, it has the potential to be a very powerful tool for accelerating the discovery of new superconductors and enhancing our understanding of them.