Improving the performance of optical inverse design of multilayer thin films using CNN-LSTM tandem neural networks

2025-06-11

AI Summary

In multilayer thin-film optical inverse design, the thickness-to-spectrum mapping is inherently non-unique, and conventional optimization relies heavily on time-consuming numerical simulations. Method: This paper proposes a novel CNN-LSTM serial neural network architecture—the first to synergistically embed convolutional feature extraction and sequential modeling—thereby mitigating ambiguity in the one-to-many inverse mapping. The approach further integrates an MLP, a pretrained forward network, and a differentiable inverse network into a joint optimization framework. Contribution/Results: Among nine tested TNN configurations, the proposed method achieves the optimal trade-off between accuracy and training efficiency: average relative error <1.8%, and inference speed 3.2× faster than LSTM-LSTM. This significantly enhances the practicality and scalability of high-accuracy optical inverse design.

Technology Category

• Search and Optimization: Learning to Search
• Computer Vision: Learning & Optimization for CV
• Machine Learning: Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Optical properties of thin film are greatly influenced by the thickness of each layer. Accurately predicting these thicknesses and their corresponding optical properties is important in the optical inverse design of thin films. However, traditional inverse design methods usually demand extensive numerical simulations and optimization procedures, which are time-consuming. In this paper, we utilize deep learning for the inverse design of the transmission spectra of SiO2/TiO2 multilayer thin films. We implement a tandem neural network (TNN), which can solve the one-to-many mapping problem that greatly degrades the performance of deep-learning-based inverse designs. In general, the TNN has been implemented by a back-to-back connection of an inverse neural network and a pre-trained forward neural network, both of which have been implemented based on multilayer perceptron (MLP) algorithms. In this paper, we propose to use not only MLP, but also convolutional neural network (CNN) or long short-term memory (LSTM) algorithms in the configuration of the TNN. We show that an LSTM-LSTM-based TNN yields the highest accuracy but takes the longest training time among nine configurations of TNNs. We also find that a CNN-LSTM-based TNN will be an optimal solution in terms of accuracy and speed because it could integrate the strengths of the CNN and LSTM algorithms.

Problem

• Predicting thin film layer thicknesses accurately
• Reducing time-consuming traditional inverse design methods
• Solving one-to-many mapping in deep-learning designs

Innovation

• CNN-LSTM tandem neural networks for inverse design
• Solving one-to-many mapping in optical thin films
• Combining CNN and LSTM for accuracy and speed