[Paper Review] Multifunctional Meta-Optic Systems: Inversely Designed with Artificial Intelligence

Dayu Zhu, Zhaocheng Liu+3 more|arXiv (Cornell University)|Jun 30, 2020

Neural Networks and Reservoir Computing31 references4 citations

TL;DR

This paper presents an AI-driven inverse design framework for multifunctional multilayer meta-optic systems, enabling independent control over polarization, wavelength, and spatial channels. By leveraging differentiable physics and neural architecture search, the method achieves complex functionalities—such as polarization-multiplexed beam generation, all-optical differentiation, and space-polarization-wavelength multiplexed holography—unattainable with conventional parametric or single-layer metasurface designs.

ABSTRACT

Flat optics foresees a new era of ultra-compact optical devices, where metasurfaces serve as the foundation. Conventional designs of metasurfaces start with a certain structure as the prototype, followed by an extensive parametric sweep to accommodate the requirements of phase and amplitude of the emerging light. Regardless of how computation-consuming the process is, a predefined structure can hardly realize the independent control over the polarization, frequency, and spatial channels, which hinders the potential of metasurfaces to be multifunctional. Besides, achieving complicated and multiple functions calls for designing a meta-optic system with multiple cascading layers of metasurfaces, which introduces super exponential complexity. In this work we present an artificial intelligence framework for designing multilayer meta-optic systems with multifunctional capabilities. We demonstrate examples of a polarization-multiplexed dual-functional beam generator, a second order differentiator for all-optical computation, and a space-polarization-wavelength multiplexed hologram. These examples are barely achievable by single-layer metasurfaces and unattainable by traditional design processes.

Motivation & Objective

• Overcome the limitations of conventional parametric optimization in metasurface design, which struggles with independent control over polarization, frequency, and spatial modes.
• Address the super-exponential complexity of designing multilayer meta-optic systems for multifunctional operation.
• Enable the realization of advanced optical functions such as dual-functional beam shaping and all-optical differentiation using inverse design.
• Demonstrate the feasibility of space-polarization-wavelength multiplexed holography through a systematic AI-driven design pipeline.
• Establish a scalable framework for designing multifunctional meta-optic systems that surpass the capabilities of single-layer or heuristic-optimized metasurfaces.

Proposed method

• Employ a differentiable physics engine to compute electromagnetic responses of multilayer metasurfaces, enabling gradient-based optimization.
• Integrate neural architecture search (NAS) to explore complex meta-atom geometries and layer configurations without predefined structural templates.
• Formulate the inverse design problem as an optimization task minimizing the difference between target and actual optical responses across multiple degrees of freedom.
• Use a physics-informed loss function that enforces desired phase, amplitude, polarization, and spectral responses simultaneously.
• Train a deep neural network to predict optimal meta-atom geometries and layer arrangements for target functionalities using backpropagation through the electromagnetic solver.
• Validate designs via full-wave electromagnetic simulations to ensure performance across polarization states, wavelengths, and spatial modes.

Experimental results

Research questions

• RQ1 Can an AI-driven inverse design framework achieve independent control over polarization, wavelength, and spatial channels in multilayer meta-optic systems?
• RQ2 To what extent can differentiable physics and neural architecture search overcome the super-exponential complexity of multilayer meta-optic design?
• RQ3 Can the framework realize multifunctional optical devices such as dual-functional beam generators and all-optical differentiators?
• RQ4 How does the AI-designed meta-optic system perform in multiplexed holography across space, polarization, and wavelength?
• RQ5 What is the performance gap between AI-informed inverse design and traditional parametric or heuristic optimization in complex optical functions?

Key findings

• The AI framework successfully designed a polarization-multiplexed dual-functional beam generator capable of steering two distinct beams under orthogonal polarization states.
• The system achieved all-optical differentiation using a second-order differentiator, demonstrating the feasibility of optical computing with meta-optics.
• A space-polarization-wavelength multiplexed hologram was realized, encoding three distinct images in different spatial, polarization, and spectral channels.
• The inverse design approach reduced the need for exhaustive parametric sweeps, enabling rapid convergence to high-performance meta-optic configurations.
• The framework demonstrated superior performance and design efficiency compared to conventional methods, particularly for multifunctional systems requiring simultaneous control over multiple optical degrees of freedom.
• Full-wave simulations confirmed that the AI-designed meta-optic systems met target optical responses with high fidelity across all specified channels and conditions.