CyberTraining: Implementation: Small: Modeling Quantum Dynamics of Excited States in Materials in the Era of Machine Learning

Abstract

Project Title: CyberTraining: Implementation: Small: Modeling Quantum Dynamics of Excited States in Materials in the Era of Machine LearningPI: Alexey AkimovAward number: NSF-OAC-2519325AbstractThis project addresses a critical national need for enhanced training in computational methods that support the rational design of advanced materials for applications in solar energy, quantum computing, sensing, and optoelectronics. The processes that determine materials’ performance—such as excitation energy and charge transfer, nonradiative relaxation of excited electronic states, photoinduced isomerization and reactions—are governed by an interplay of electrons and nuclei. The understanding of this interplay requires specialized simulations of coupled electronic and nuclear dynamics. However, researchers often lack specialized high-quality training in these complex methods as well as practical experience with the advanced software tools that implement them. This project meets that need by offering intensive training through four summer schools focused on the cyberinfrastructure (CI) for nonadiabatic quantum dynamics (NAQD) simulations and their integration with machine learning (ML) tools and excited-state electronic structure calculations. Online educational materials and a new university course will expand the project’s reach, helping to equip students and researchers across the country with the skills to use the cutting-edge computational tools for accurate modeling of broad range of quantum dynamics phenomena in complex systems. By building a stronger and better-prepared scientific workforce, by broadening participation, and by facilitating the adoption of the advanced CI for NAQD and ML, this project promotes the progress in material science, supports technological innovation and education, and enhances societal impacts and national prosperity.This implementation project will build capacity within the scientific community working on quantum-classical and quantum modeling of nonadiabatic dynamics in materials. It will do so by providing conceptual and hands-on training to approximately 100 graduate students, postdoctoral researchers, and early-career scientists through a series of four summer schools. The events will focus on advanced software packages that support NAQD, excited-state electronic structure calculations, and the application of machine learning to these problems. Sessions will be taught by leading experts, including original developers of more than 30 specialized tools and libraries. Online tutorials, input examples, and documentation will be created to promote broader adoption of these tools. The project will also contribute new educational content by developing and incorporating a "Machine Learning for Chemists" course into the undergraduate and graduate curriculum at the University at Buffalo. Additionally, the PI will develop workflows and tutorials for the comprehensive and modular Libra software package to facilitate integrated NAQD simulations. Overall, the project aims to bridge critical training gaps, advance computational capabilities, and catalyze the use of modern CI tools in chemical and materials research.