🔬 optics

Multi-wavelength polarisation imaging with inverse designed metasurfaces

This paper presents and experimentally demonstrates a compact, single-shot optical system using an inverse-designed metasurface that efficiently separates a scene into spectral and polarimetric measurements at 532 nm and 700 nm, offering a robust solution for lightweight multispectral polarisation imaging applications.

Original authors: Sarah E. Dean, Neuton Li, Josephine Munro, Benjamin Laudert, Thomas Siefke, Quyet Ngo, Robert Sharp, Dragomir N. Neshev, Falk Eilenberger, Andrey A. Sukhorukov

Published 2026-05-01

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Sarah E. Dean, Neuton Li, Josephine Munro, Benjamin Laudert, Thomas Siefke, Quyet Ngo, Robert Sharp, Dragomir N. Neshev, Falk Eilenberger, Andrey A. Sukhorukov

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to take a photo of a field to check the health of the crops. Normally, a camera just sees how bright things are (intensity). But light also has a hidden property called polarization, which is like the "vibration direction" of the light waves. By measuring this, you can learn extra secrets about the plants, like how healthy they are or if they are stressed, which a normal camera can't see.

The problem is that traditional cameras need bulky, heavy lenses and filters to see this polarization, and they often need to take multiple pictures or spin parts around to get the full data. This is terrible for drones, which need to be light and compact.

This paper introduces a clever solution: a metasurface. Think of this as a tiny, flat piece of glass (or plastic) covered in microscopic patterns, smaller than a human hair. It acts like a "magic prism" that does two things at once with a single snapshot:

It sorts light by color: It separates green light (532 nm) from red light (700 nm).
It sorts light by polarization: It splits the light into different beams based on how the light is vibrating.

The "Smart Mail Sorter" Analogy

Imagine a giant mail sorter at a post office.

Old Way: You have to run the mail through one machine to sort by color, then run it through a second machine to sort by size, then a third to sort by shape. It's slow, heavy, and the letters might get lost or mixed up between machines.
This New Way: The metasurface is like a single, super-smart sorter. You dump a mixed pile of letters (light from the scene) into it, and instantly, it sorts them into specific bins. Some bins catch the "green" letters, others catch the "red" letters. Inside those color bins, it further sorts them into "vertical vibration" piles and "horizontal vibration" piles.

Because it does all this in one step with one flat piece of material, the whole camera system can be tiny and lightweight—perfect for a drone.

How They Made It Work

The researchers didn't just guess the pattern; they used a computer "brain" (called topology optimization) to design the microscopic shapes.

They told the computer: "We need to see two specific colors (green and red) clearly, and we need to measure the polarization perfectly for both."
The computer tried millions of different microscopic shapes, like a sculptor chipping away at a block of stone, until it found the perfect pattern that splits the light exactly how they wanted.

What They Found

The team built a real version of this tiny pattern and tested it in a lab:

For Green Light (532 nm): It worked perfectly. It could measure the full "personality" of the light (all types of polarization), just like a high-end lab instrument, but in a tiny package.
For Red Light (637 nm): It worked well for measuring straight-line vibrations (linear polarization), though it couldn't do the full complex measurement as easily as with the green light.
Robustness: They checked if the device would break if the drone tilted or if the light wasn't a single pure color. They found it works over a wide range of angles and colors, meaning it's sturdy enough for real-world flying.

Why This Matters (According to the Paper)

The paper claims this technology allows for single-shot imaging. This means the drone doesn't need to stop, spin, or take multiple photos to get the data. It captures everything in one instant. This is crucial for applications like agricultural remote surveying (checking crop health from the air) and biological cell imaging, where you need to be fast, light, and efficient.

In short, they created a tiny, flat "smart filter" that lets a drone see the hidden polarization secrets of the world in a single snapshot, replacing heavy, complex camera gear with a simple, flat chip.

1. Problem Statement

Multispectral polarisation imaging is critical for applications ranging from biological cell imaging to agricultural remote sensing (e.g., crop health monitoring via drones). However, conventional systems face significant limitations:

Bulk and Complexity: Traditional full-Stokes polarimetry requires arrays of filters, prisms, or rotating waveplates, making systems bulky and heavy.
Multi-shot Limitations: Many existing methods require multiple exposures or mechanical rotation, which is unsuitable for moving platforms like drones where "single-shot" imaging is required to avoid motion artifacts and illumination changes.
Spectral Constraints: Existing metasurface-based polarimeters are typically limited to monochromatic light. Extending them to multispectral operation usually requires additional beam-splitting optics, increasing system size and reducing photon flux.

The authors aim to develop a compact, single-shot, lightweight optical system capable of performing multispectral polarimetry on a single optical component without additional filtering or beam-splitting.

2. Methodology

A. System Architecture

The proposed system utilizes a push-broom imaging technique, ideal for drone-based surveying. A narrow image swath is continuously imaged as the platform moves. The metasurface diffracts incident light in one direction to separate wavelengths into distinct spectral channels, while individual diffraction orders within those channels act as separate polarisation measurement channels. This allows simultaneous recording of all polarisation and spectral data.

B. Design Specifications

Target Wavelengths: 532 nm and 700 nm. These were selected based on chlorophyll reflection and absorption bands relevant to agricultural monitoring. Crucially, these wavelengths were chosen to ensure their diffraction orders do not overlap, allowing natural spectral separation.
Material & Structure: The metasurface consists of a 300 nm thick patterned Titanium Dioxide (TiO₂) layer on a 500 µm substrate. TiO₂ was chosen for its high refractive index and transparency in the visible spectrum.
Periodicity: A periodicity of 4500 nm × 250 nm was selected to produce diffraction angles that separate spectral components sufficiently (±8.9°/±18.1° for 700 nm; ±6.8°/±13.7° for 532 nm) while remaining within the capture range of standard optics (NA ≥ 0.32).

C. Inverse Design Framework

The authors employed topology optimisation to design the nanostructure:

Optimisation Process: Starting with a random distribution of refractive indices within a single period, the algorithm uses forward and adjoint simulations (via rigorous coupled-wave analysis in RETICOLO) to calculate gradients.
Figure-of-Merit (FoM): Instead of targeting specific polarisation states, the FoM maximizes the minimum singular value of the instrument matrix ( $M$ $M$ ) for both wavelengths simultaneously. This ensures robust polarisation reconstruction by minimizing error amplification (condition number).
- $FoM = \min(\mu_{532nm}) \cdot \min(\mu_{700nm})$
Binarisation: The continuous design space is binarised during iterations to ensure the final structure is fabricable.

3. Key Contributions

First Multispectral Single-Shot Metasurface: Demonstrated a single metasurface capable of separating and measuring polarisation at two distinct wavelengths (532 nm and 700 nm) simultaneously without external filters.
Inverse Design for Multispectral Robustness: Developed an optimisation strategy that balances polarisation reconstruction performance across multiple wavelengths, acknowledging that identical polarisation responses may not be achievable for both wavelengths simultaneously.
Comprehensive Performance Characterization: Provided detailed simulations and experimental validation of the system's Field-of-View (FoV) and spectral bandwidth, proving viability for real-world imaging.

4. Results

A. Simulation Results

Polarisation Span: The designed diffraction orders act as partial polarisers that span the Stokes space effectively.
- Condition Numbers: 3.29 at 532 nm and 3.69 at 700 nm (well below the threshold of 15, indicating low error amplification).
- Contrast: Average polarisation selectivity contrast of 82% (minimum 51%) across both wavelengths.
Field-of-View (FoV):
- 532 nm: Maintains acceptable performance over a 48° × 39° FoV.
- 700 nm: Maintains performance over a 93° (transverse) × >96° (diffraction direction) FoV.
Spectral Bandwidth:
- The design supports a viable bandwidth of ±7 nm at 532 nm and ±25 nm at 700 nm before polarisation analysis vectors vary too significantly to maintain accuracy.
- Image smearing due to diffraction limits the bandwidth to ±39.2 nm, but polarisation consistency is the primary limiting factor.

B. Experimental Results

The metasurface was fabricated using electron beam lithography and tested.

532 nm Performance: Successfully achieved full-Stokes polarimetry.
- Condition number: 14.
- Selectivity: Minimum 47% transmission contrast.
- Accuracy: Average Root Mean Square Error (RMSE) of 4.43% for test states (max error 8.51%).
637 nm Performance: While full-Stokes reconstruction was not achieved (likely due to material dispersion or fabrication tolerances shifting the response), the device successfully demonstrated linear polarimetry.
- Selectivity: >50% for key diffraction orders.
- Accuracy: Average RMSE of 2.31% for linear test states (max error 8.12%).

5. Significance

This work represents a significant step toward compact, lightweight polarimetric sensors suitable for unmanned aerial vehicles (UAVs) and other mobile platforms. By integrating spectral separation and polarisation analysis into a single, flat optical component, the system eliminates the need for bulky beam splitters and rotating components.

The ability to perform single-shot multispectral polarimetry enables real-time monitoring of vegetation health and crop identification with higher fidelity than intensity-only imaging. The inverse design framework presented offers a scalable path for developing multifunctional metasurfaces for complex optical tasks beyond the specific wavelengths tested here.