Deep Learning Prediction and Optimization of Photon Upconversion in Lanthanide-Doped Nanoparticles

Deep learning prediction and optimization of photon upconversion in Lanthanide-doped nanoparticles<br/>Eric Sivonxay, Samuel M. Blau<br/><br/>Lanthanide-doped upconverting nanoparticles (UCNPs) have unique optical properties, capable of near-infrared excitation to yield visible and ultraviolet emissions. UCNPs have broad applications ranging from biosensing and drug delivery to super-resolution microscopy. Factors affecting the nonlinear photophysical properties of UCNPs include size, shape, doping, shell thickness, and surface ligands, defining a vast chemical design space. Due to the complexity of the energy transfer networks formed by the 4f electron shells of the lanthanide dopants, rationally designing UCNPs with high upconverting efficiency for tailored applications remains unfeasible. While Monte Carlo simulations allow for reasonably accurate in silico prediction of optical properties, calculations scale poorly with particle size and dopant loading, constraining the search for UCNPs with desirable properties to be fundamentally Edisonian. Here, we report efforts to overcome these challenges by combining high-throughput data generation with nanoparticle representation learning. We combine a high-performance implementation of lanthanide energy transfer Monte Carlo with automated workflows and HPC resources to generate the first large dataset of over 5,000 simulated spectra for UCNPs. We then develop a novel representation of nanoparticle features, enabling deep learning of spectra. Finally, we apply these models to optimize UCNP heterostructure, targeting high UV emission for 3D printing, and present preliminary results.