PRISM: Position-encoded Regressive Inverse Spectral Model for Multilayer Thin-Film Design

Researchers introduced PRISM, a decoder-only transformer for inverse design of multilayer thin-film optical coatings.

PRISM jointly predicts layer materials and thicknesses from target spectra using spectrum prefix conditioning and cumulative-depth positional encoding.

It outperforms baseline transformer methods on error and speed, and remains competitive with simulated annealing.

The approach could make optical coating development faster and more accurate.