Guided Diffusion for the Discovery of New Superconductors

2025-09-29

AI Summary

Inverse design of high-temperature superconductors faces challenges due to the vast chemical and structural search space and low sampling efficiency. Method: We propose a classifier-free guided diffusion framework built upon a pre-trained DiffCSP model (pre-trained on the Alexandria database and fine-tuned on superconducting materials), augmented with first-principles–derived property labels for attribute-conditioned generation; integrated with multi-stage machine learning screening and density functional theory (DFT) validation to significantly improve feasibility prediction accuracy. Results: The method generated 200,000 candidate structures, identified 34,000 previously unreported compounds, and discovered 773 potential superconductors with *T*_c > 5 K—seven of which were successfully synthesized and experimentally validated. This work represents the first systematic application of classifier-free guided diffusion to superconducting inverse design, establishing a high-throughput, high-fidelity, and experimentally verifiable paradigm for novel material discovery.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Unsupervised & Self-Supervised Learning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

The inverse design of materials with specific desired properties, such as high-temperature superconductivity, represents a formidable challenge in materials science due to the vastness of chemical and structural space. We present a guided diffusion framework to accelerate the discovery of novel superconductors. A DiffCSP foundation model is pretrained on the Alexandria Database and fine-tuned on 7,183 superconductors with first principles derived labels. Employing classifier-free guidance, we sample 200,000 structures, which lead to 34,027 unique candidates. A multistage screening process that combines machine learning and density functional theory (DFT) calculations to assess stability and electronic properties, identifies 773 candidates with DFT-calculated $T_mathrm{c}>5$ K. Notably, our generative model demonstrates effective property-driven design. Our computational findings were validated against experimental synthesis and characterization performed as part of this work, which highlighted challenges in sparsely charted chemistries. This end-to-end workflow accelerates superconductor discovery while underscoring the challenge of predicting and synthesizing experimentally realizable materials.

Problem

• Accelerating discovery of novel superconductors via guided diffusion
• Inverse design of high-temperature superconducting materials
• Predicting synthesizable materials with DFT-validated Tc above 5K

Innovation

• Guided diffusion framework for superconductor discovery
• DiffCSP model pretrained and fine-tuned on superconductors
• Multistage screening combining machine learning and DFT