Structure Elucidation of Complex Liquid Solutions from Nuclear Magnetic Resonance Spectroscopy using Machine Learning

About This Grant

Electrochemical systems, such as batteries, fuel cells, and supercapacitors, are essential for creating cleaner energy, powering electric vehicles, and storing renewable energy efficiently. Improving these systems is important for making energy storage safer, longer-lasting, and more effective. At the core of these systems are electrolytes, special materials that allow ions to move and enable energy storage and transfer. By designing better electrolytes, it is possible to create improved batteries and energy systems that support sustainable energy solutions. This project focuses on understanding how electrolytes function at the molecular level. Using advanced tools like artificial intelligence, machine learning, and experimental techniques, the research aims to uncover new insights into how ions interact and move within these materials. These discoveries will help in designing better electrolytes, leading to more efficient and reliable energy storage technologies. Beyond advancing science, this project will benefit society by providing open-access tools and knowledge to researchers and educators. It will train the next generation of scientists and engineers in research, and contribute to the development of technologies that support cleaner energy. These efforts align with national priorities to drive innovation, foster sustainability, and expand the science and engineering workforce to meet future challenges in energy and materials science. The proposed research investigates the solvation structure of liquid organic electrolytes in lithium-ion batteries (LIBs) as a model system. The study aims to identify and understand multiple stable ionic species within multicomponent electrolyte solutions and their impact on transport properties. Using a combination of experimental spectroscopy, computational simulations, and machine learning (ML), the project will achieve the following goals: First, it will develop a large-scale database of experimental NMR spectra, capturing solvation environments in LIB electrolytes. Using advanced natural language processing (NLP) and vision-language models (VLM), data from the scientific literature will be extracted to create a high-fidelity resource for training ML models. Where gaps in the literature exist, new NMR data will be generated through collaborations with experimental researchers. Second, the project will elucidate solvation structures and their dynamics by combining molecular dynamics (MD) and density functional theory (DFT) simulations. This approach will overcome traditional limitations by automating predictions of NMR chemical shifts, enabling the identification of multiple stable species in complex solutions. The proposed work will result in the first systematic open-source augmented NMR spectroscopic database for both traditional and non-traditional nuclei, corresponding liquid composition and metadata, and detailed solvation structure obtained using simulations. Finally, the research will develop an ML-based framework to predict NMR chemical shifts for new electrolyte systems, bridging computational and experimental data. The framework will provide rapid and accurate mappings of solvation structures to observed NMR spectra, facilitating electrolyte design for various applications. 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.