Meta-photonics: A bridge between physical association and digital models in photonics

Xiaohua Xing,Yuqi Ren,Die Zou,Qiankun Zhang,Bingxuan Mao,Jianquan Yao,Deyi Xiong,Liang Wu

DOI: https://doi.org/10.1063/5.0203698

IF: 4

2024-04-29

Applied Physics Letters

Abstract:Recently, the burgeoning field of metaverse has facilitated the convergence of digital scenarios and the real world to achieve unprecedented interaction patterns. Inspired by this digital approach, we propose the concept of meta-photonics to gain insight into the photonics underlying increasingly complex phenomena, while traditional physics theory is not keeping pace with current observations. Here, we are armed with different deep learning models to build intrinsic dimensions among different physical elements and bring together amplitude, phase, polarization conversion ratio, extinction ratio, unwrapping phase, surface electric field, and connected domain, most of which have never been reported, based on the results of the forward and inverse models. This approach reveals an opportunity to catalyze the discovery of more enigmatic associations among different concepts and paves the way for tailoring more flexible optical responses in the vibrant world of meta-photonics.

physics, applied

What problem does this paper attempt to address?

The paper attempts to address the issue in the field of photonics where traditional physical theories cannot keep up with the complexity of currently observed phenomena. Specifically, the paper proposes a new concept—meta-photonics, which aims to establish intrinsic dimensions between different physical elements by introducing various deep learning models and uncovering more unknown correlations. This helps accelerate the development of photonics and achieve more flexible optical responses.

The main contributions of the paper include:
1. **Establishing digital models**: By using deep learning models such as Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), Gated Recurrent Units (GRU), and Transformers, the paper constructs digital model relationships between various concepts in photonics.
2. **Improving prediction accuracy**: In both forward and inverse models, the prediction accuracy is significantly improved by optimizing the loss function module, especially in the application of phase unwrapping techniques, greatly reducing the Mean Squared Error (MSE) of predictions.
3. **Multi-objective optimization**: The design process considers multiple objectives, such as simultaneous optimization of phase and amplitude, making the designed optical devices both efficient and practical.
4. **Simplifying the design process**: The concept of meta-photonics eliminates the need to change datasets or redesign network architectures when switching target tasks, simplifying the design process of optical devices.

In summary, by introducing the concept of meta-photonics and utilizing deep learning techniques, the paper addresses some key bottleneck issues in traditional photonics, providing new ideas and methods for the future development of photonics.