Collaborative Research: CDS&E: Computational Spectrometry and Hyperspectral Imaging Integrated with Machine Learning for Advanced Materials Characterization

This project will develop a new data-enabled hyperspectral imaging and artificial intelligence (AI) platform that would enable scientists and engineers to analyze highly concentrated particle suspensions that are currently beyond the reach of conventional optical measurement methods. Such systems are widely encountered in environmental monitoring, advanced manufacturing, energy technologies, biomedical diagnostics, and consumer products. The research will transform complex light interactions within dense materials from a long-standing measurement obstacle into a valuable source of information that could reveal particle size, composition, concentration, and optical properties in real time. The project would integrate advanced optical instrumentation, computational modeling, and AI to create a new generation of intelligent measurement technologies. Outcomes would be broadly disseminated through open datasets, software tools, educational modules, and research publications. Graduate, undergraduate, and postdoctoral researchers would receive interdisciplinary training in optics, computation, and data science, helping prepare a future workforce for data-intensive scientific and engineering fields. The project would also contribute to advances in AI by developing physics-informed machine-learning approaches capable of extracting meaningful scientific knowledge from large, complex, and highly nonlinear datasets. Technically, the project would establish a unified experimental–computational–AI framework for quantitative characterization of nano- and micro-particle dispersions with optical densities ranging from 3 to 100. A dual-track hyperspectral imaging system would be developed to acquire multidimensional, polarization-resolved hyperspectral image data cubes containing spatial, spectral, and polarization information. These measurements would be coupled with a computational hyperspectrometry engine based on electrodynamic theory and Monte Carlo photon transport simulations capable of generating physically realistic synthetic datasets that reproduce cascading photon–matter interactions, including multiple scattering, absorption, and emission. Experimental and simulated datasets would be integrated to train and validate machine-learning models that could predict particle size distributions, refractive indices, concentrations, and optical constants directly from complex hyperspectral signatures. Feature-selection and dimensionality-reduction methods would identify the most informative spectral and spatial variables governing photon transport in optically dense media. The resulting framework would establish a scalable, physics-informed, and AI-enabled methodology for quantitative optical characterization, while laying the foundation for future quantum-accelerated computational spectroscopy and automated intelligent measurement systems. 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.