Two-Stage Deformable-Convolutional Inverse Design of Nanophotonic Absorbers from Optical Spectra

2026-08-12

AI Summary

This work addresses the non-uniqueness of spectrum-to-geometry mapping and the challenge of reconstructing fine geometric details in inverse design of nanophotonic absorbers by proposing the first two-stage framework incorporating deformable convolutions. In the first stage, a supervised learning model maps an 80-dimensional absorption spectrum to a 64×64 resonator mask; the second stage refines this solution through least-squares adversarial fine-tuning initialized from the optimal supervised model. The approach enables adaptive sampling within the decoder and integrates latent-space projection with forward surrogate model evaluation. Experimental results demonstrate substantial improvements over baseline methods, achieving high performance across multiple metrics: PSNR of 20.79 dB, SSIM of 0.8501, Dice coefficient of 0.9623, IoU of 0.9342, spectral consistency RMSE of 0.0805, and R² of 0.7923.

Abstract

Data-driven inverse design enables efficient generation of nanophotonic structures with prescribed optical responses, but spectrum-to-geometry mapping remains challenging due to non-uniqueness and fine geometric features. This work presents a two-stage deformable-convolutional framework for reconstructing metal--insulator--metal resonator geometries from 80-dimensional absorption spectra. The spectrum is projected to a $150\times4\times4$ latent representation and decoded into a $64\times64$ resonator mask. Training combines supervised reconstruction with least-squares adversarial refinement initialized from the best supervised checkpoint. A three-run ablation compares deformable convolution with plain convolution, involution, Dynamic Conv, and ODConv under the same architecture. The proposed model achieves $20.79\pm0.31$~dB PSNR and $0.8501\pm0.0082$ SSIM, improving over plain convolution by 2.16~dB and 0.0831, respectively. It further achieves Dice $0.9623\pm0.0027$, IoU $0.9342\pm0.0038$, and boundary F-score $0.9550\pm0.0027$. Spectral consistency evaluated using a frozen forward surrogate yields RMSE $0.0805\pm0.0013$ and $R^2=0.7923\pm0.0065$. Learned offsets show stronger adaptive sampling at coarse and intermediate decoder stages. Overall, deformable sampling with supervised initialization and adversarial refinement improves spectrum-conditioned geometry reconstruction.

Problem

• inverse design
• nanophotonic absorbers
• spectrum-to-geometry mapping
• optical spectra
• geometric reconstruction

Innovation

• deformable convolution
• inverse design
• nanophotonic absorbers
• adversarial refinement
• spectrum-to-geometry mapping