Fabrication-conscious neural network based inverse design of single-material variable-index multilayer films

Abstract

Multilayer films with continuously varying indices for each layer have attracted great deal of attention due to their superior optical, mechanical, and thermal properties. However, difficulties in fabrication have limited their application and study in scientific literature compared to multilayer films with fixed index layers. In this work we propose a neural network based inverse design technique enabled by a differentiable analytical solver for realistic design and fabrication of single material variable-index multilayer films. This approach generates multilayer films with excellent performance under ideal conditions. We furthermore address the issue of how to translate these ideal designs into practical useful devices which will naturally suffer from growth imperfections. By integrating simulated systematic and random errors just as a deposition tool would into the optimization process, we demonstrated that the same neural network that produced the ideal device can be retrained to produce designs compensating for systematic deposition errors. Furthermore, the proposed approach corrects for systematic errors even in the presence of random fabrication imperfections. The results outlined in this paper provide a practical and experimentally viable approach for the design of single material multilayer film stacks for an extremely wide variety of practical applications with high performance.

Multifunctional Spaceplates for Optical Aberration Correction

“…During this iterative optimization process, the device is simulated using the transfer matrix method (TMM), and local gradients are calculated by autodifferentiating the TMM expressions. Analytic TMM autodifferentiation has been previously used for various thin-film stack design tasks and provides a straightforward way to calculate gradients in high-dimensional optimization landscapes. , Once the initial thin-film stack is optimized, it is duplicated, vertically stacked, and used as a starting point for further local gradient-based optimization. These duplication, stacking, and local optimization steps are repeated until the desired number of layers is achieved.…”

Designing Metasurfaces for Efficient Solar Energy Conversion

“…Deep learning based on multilayer artificial neural networks (Figure a) has recently attracted significant attention from the photonics’ community . Deep learning that brings substantial acceleration capability and forth a feasible avenue for global optimization, has been introduced in metasurface design problems, including multilayer perceptrons, , convolutional neural networks, generative adversarial networks, and variational autoencoders. − The method allows combination with other optimization techniques such as genetic algorithms, , topology optimization, , or adjoint optimization , to enable high-performance, large-scale metasurface designs.…”

Transfer Learning-Assisted Inverse Modeling in Nanophotonics Based on Mixture Density Networks

“…An inverse modeling approach avoids the need for coupling a forward model with an optimizer and directly performs the prediction of the optimal design parameters values. This paper focuses on the Mixture Density Network (MDN) architecture for inverse modeling [4], [9], [15], which is based on the composition of a Gaussian Mixture Model and a Neural Network. The output space of an MDN contains the mean and standard deviation values of multiple Gaussian probability distribution functions (pdfs) and the mixing coefficients that denote some weights (importance) associated with the Gaussian pdfs.…”