Predictability of Localized Plasmonic Responses in Nanoparticle Assemblies

Design of nanoscale structures with desired nanophotonic properties are key tasks for nanooptics and nanophotonics. Here, the correlative relationship between local nanoparticle geometries their plasmonic responses is established using encoder-decoder neural networks. In im2spec network, particle spectra via encoding observed to a small number latent variables subsequently decoding into spectra; in spec2im reversed. Surprisingly, these reduced descriptions allow high-veracity predictions based on fixed compositions chemical states surface. The analysis space distributions corresponding decoded closest (in space) encoded images yields insight generative mechanisms interactions arrays. Ultimately, this approach creates path toward determining configurations that can yield spectrum one, paving way stochastic design nanoplasmonic structures.