CDS&E: Statistical Learning Tools for NMR Spectroscopy of Non-Crystalline Materials

With the support of the Chemical Measurement and Imaging Program in the Division of Chemistry, and partial co-funding from the Ceramics Program in the Division of Materials Research, Professor Philip Grandinetti and his group at the Ohio State University are developing machine learning tools to improve understanding of the physical and chemical properties of glass-containing materials using Nuclear Magnetic Resonance (NMR) Spectroscopy – the technique upon which Magnetic Resonance Imaging (MRI) is based. Specialty glasses continue to play critical roles in a large range of technological applications, such as glass substrates for handheld electronic device displays and lighting, optical fibers, nuclear waste storage, and bio-glass implants.

These applications have high societal impact across a wide range of environmental, energy, and health-related issues. A major challenge in tailoring the properties of new glass compositions is the inadequacy of available quantitative details about the structure of glasses, which determines their macroscopic (bulk) properties. Professor Grandinetti is developing more sensitive methods and open-source software tools that perform a deeper analysis of NMR data and give richer details about structure in glassy materials.

The work addresses a range of factors determining glass properties such as dimensional stability, strength, phase separation, hardness, and chemical durability. It is providing research opportunities for students from underrepresented groups, in part through a partnership with Berea College in Kentucky. Collaborations provide all students involved in the project with opportunities for interactions with scientists in industry as well as across national boundaries.

This project focuses on solving the ill-posed problem of inverting an NMR spectrum into its underlying distribution of nuclear interaction parameters, followed by a quantitative mapping of these parameters into structural distributions. In this effort, Professor Philip Grandinetti and his group are developing open-source Python programs, documentation, and tutorials, and associated progressive web apps to enable fast, easy-to-use, and versatile simulations and analyses of experimental one- and higher-dimensional solid-state NMR spectra.

They capitalize on their recent discovery that highly selective excitation of quadrupolar nuclei can extend NMR transition lifetimes and provide dramatic sensitivity enhancements. This advance in turn is expected to enable expanded applications of the statistical learning tools to natural abundance O-17 2D NMR spectra of inorganic oxide materials. Quantification of modifier cation clustering and tetrahedral framework network disorder in a series of alkali and alkaline earth silicate glasses is another aim.

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.