DMREF: NSF-DFG: AI-Driven Design of Superconducting Materials for Magnets (AISuper)

Non-technical abstract: This project will accelerate the discovery of new superconducting materials through a transformative approach that combines artificial intelligence (AI), quantum theory, and experimental synthesis. Superconductors are essential for technologies ranging from MRI systems and high-field magnets to quantum computing and sustainable energy. Yet analysis of known compounds suggests that only a small fraction of potential superconductors may have been discovered. This project aims to significantly expand the number of known superconductors and identify materials optimized for practical applications—specifically those with high critical temperatures and magnetic fields, ductility for wire fabrication, and three-dimensional electronic structures for enhanced performance. A core educational mission will train a group of students in AI-driven materials research and develop hands-on experiment kits for K–12 classrooms to promote STEM engagement. Partnerships with national laboratories, industry, and international collaborators will ensure timely and impactful transition of discoveries to real-world applications. Technical abstract: The research integrates two complementary AI methods to accelerate superconductor discovery. Property prediction models based on graph neural networks will estimate superconducting characteristics—such as electron-phonon spectral functions, critical temperatures, and critical fields—directly from crystal structures. In parallel, generative AI models using stochastic flow matching will design novel, synthesizable materials with targeted superconducting and mechanical properties. Predictions from both models will be evaluated using density functional theory (DFT) to assess thermodynamic stability, electronic structure, and superconducting potential. Selected candidates will undergo targeted synthesis and experimental characterization using high-throughput techniques. Both AI models will be iteratively refined using experimental feedback, forming a closed discovery loop that integrates theory, simulation, and validation. This approach will yield a comprehensive, open-access dataset of successful and unsuccessful candidates, providing a foundation for future AI-guided materials discovery. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.