Stacked 3D Metasurface for Independent Multiwavelength Convolution

The metasurface is divided into multiple layers, with each layer responsible for processing a specific wavelength band. The first layer processes the first wavelength, the second layer processes the second wavelength, and so on. This segmentation allows each layer to be optimized for its specific wavelength while maintaining overall multiwavelength functionality, resolving the contradiction between fabrication simplicity and multiwavelength capability.

The solution transitions from a two-dimensional single-layer metasurface to a three-dimensional multi-layer structure. By adding the vertical dimension (stacking multiple layers), the system gains the ability to process multiple wavelengths simultaneously while each individual layer maintains relative fabrication simplicity. This dimensional expansion resolves the contradiction by distributing complexity across multiple simpler components.

The complex 3D metamaterial structure is segmented into multiple 2D metasurface layers. Each layer contains simplified patterns (kernels) that are easier to fabricate than full 3D structures. When stacked, these 2D layers collectively achieve the convolution operation functionality, significantly reducing fabrication time and cost while maintaining performance.

Instead of fabricating a single complex 3D metamaterial structure, the solution uses multiple copies of simpler 2D metasurface layers. Each layer is a simplified copy that processes a specific wavelength, and when combined, they replicate the functionality of a full 3D structure with much reduced fabrication complexity and time.

Each layer in the stacked metasurface is designed as a universal kernel that can process its assigned wavelength independently. The layers are structured similarly but tuned for different wavelengths, allowing the system to handle multiple wavelengths with a unified design approach. This universality reduces device complexity by using repeated modular units rather than entirely different structures for each wavelength.

The kernels in different layers are differentiated by parameter changes (such as geometric dimensions, material properties, or resonant frequencies) rather than fundamentally different structures. Each kernel maintains the same basic design paradigm but with adjusted parameters optimized for its specific wavelength, simplifying the overall device complexity while enabling multiwavelength functionality.