MetaField for Electromagnetic Evaluation and Design of Multilayer Nanophotonic Structures

Nanophotonic sensing and light manipulation require control over optical spectra and localized electromagnetic enhancement. Finite-difference time-domain (FDTD) simulations provide these responses, but repeated calculations limit exploration of multilayer geometries and material combinations. Artificial intelligence (AI) surrogates accelerate prediction; practical multilayer design also requires editable fabrication variables and spatially defined objectives. Here we present MetaField, a framework connecting fabrication-aware multilayer encoding, complex-field prediction, spectral readout, and constrained search. MetaField predicts spatially resolved complex electromagnetic fields and spectral responses, enabling the localization and quantification of local enhancement and joint geometry–material searches targeting specified spatial regions. The framework performs approximately one million candidate evaluations within hours, with microscopy and Raman measurements of coated nanopillars illustrating its application to practical device analysis. MetaField supports spatially targeted electromagnetic evaluation within a unified AI-assisted nanophotonic design workflow.