Artificial intelligence and data science enabled predictive modeling of collective phenomena in strongly correlated quantum materials
Lead Principal Investigator: Steven Johnston (The University of Tennessee, Knoxville)
Principal Investigators: Adrian Del Maestro, Cristian D. Batista (The University of Tennessee, Knoxville); Adrian Feiguin (Northeastern University); Richard Scalettar (University of California, Davis); Ehsan Khatami (San Jose State University); Mark Dean (Brookhaven National Laboratory); Thomas A. Maier (Oak Ridge National Laboratory); Kipton Barros (Los Alamos National Laboratory)
Strongly correlated quantum materials represent a diverse family of materials whose novel properties are governed by the principles of quantum mechanics over macroscopic length and time scales. These materials host many novel states of matter, including high-temperature superconductivity, colossal magnetoresistance, and exotic magnetic orders, and have tremendous potential to revolutionize many science and technology sectors. Despite their importance, understanding and manipulating the properties of this diverse family of materials remains a forefront challenge for the scientific community.
This project develops and uses cutting-edge computational approaches, including methods in Machine Learning (ML), Artificial Intelligence (AI), and data science, to advance predictive modeling for quantum materials. The aim is the creation of new workflows to study these materials that tightly couple advanced computational theory to Resonant Inelastic X-Ray Scattering (RIXS) experiments conducted at national user facilities. The research is organized around three key thrusts. First, robust methods for identifying and validating low-energy effective models for different classes of correlated materials will be developed. Focusing on such models, enables the discovery of the emergent laws describing the behavior of these materials and the identification of new organizing principles. Second, powerful computational methods for solving these models will be developed, leveraging ML and AI algorithms and other state-of-the-art computational methods. Finally, new workflows will be developed, validated, and combined to create novel workflows for studying new materials.
This project is in its second funding cycle following a successful initial three-year campaign. Key outcomes include an ML-based approach for extracting effective models from experimental scattering data and theoretical simulations, acceleration of several quantum Monte Carlo algorithms, the discovery of new mechanisms for stabilizing novel magnetic states (skyrmions), and several joint theory/experiment studies on unconventional superconductors. Building on these achievements automated workflows for RIXS experiments will be developed and new frameworks for modeling materials and predicting measured RIXS spectra will be created. These efforts will considerably increase the range of systems targeted by this project and provide a foundation for the seamless integration of theory in experimental workflows, enabling near-real-time interpretation of data and feedback. In tandem with development and prototyping, the developed tools will be applied to tackle critical problems in quantum materials research. To execute this ambitious research project, the team consists of experimentalists, theorists, and experts working at the boundary of condensed matter physics and AI/ML.