Structure-embedding network for predicting the transmission spectrum of a multilayer deep etched grating

Area of Science:

• Photonics and optical engineering
• Computational physics
• Materials science

Background:

• Accurate spectral prediction is crucial for designing multilayer deep etched gratings (MDEGs).
• Current deep neural network (DNN) approaches face challenges due to dimensionality mismatches, limiting prediction accuracy for devices like MDEGs, nanoparticles, and metasurfaces.

Purpose of the Study:

• To introduce a novel structure-embedding network (SEmNet) for enhanced transmission spectrum prediction in MDEGs.
• To address and overcome the dimensionality mismatch issue in DNN-based spectral prediction.

Main Methods:

• Development of SEmNet, integrating a structure-embedding module with a DNN.
• The structure-embedding module employs a learnable matrix to augment the dimensionality of the structure parameter vector.
• The augmented vector serves as input to the DNN for spectral prediction.

Main Results:

• SEmNet effectively resolves the dimensionality mismatch problem inherent in existing DNN models.
• Experimental validation shows SEmNet achieves superior prediction accuracy for MDEG transmission spectra compared to state-of-the-art methods.

Conclusions:

• SEmNet offers a significant advancement in accurately predicting MDEG transmission spectra.
• The proposed network enhances the efficiency and reliability of MDEG design processes through improved spectral prediction.