Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light

arXiv:2604.18879 · physics.optics, physics.app-ph · Submitted 2026-04-20 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light".

Mira: Here's the gist of the paper. Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light demonstrates an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum This work highlights the capacity of inverse design to generate multifunctional, non-intuitive flat optics that outperforms its traditional counterparts Pruning redundant reasoning tokens The paper presents an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum The authors demonstrate this by leveraging an adjoint optimization approach to independently control the geometry and rotation of each nanostructure, which breaks the aperture-sharing paradigm and utilizes the entire pixel area for all channels When coupled with a shallow neural network to automate peak identification and correct for hardware non-idealities, the device yields a mean polarization reconstruction error of only 0.046 across 100 test states on the Poincaré sphere while simultaneously maintaining the precise focal-spot tracking required for sensitive wavefront tilt detection Operating Concept and Inverse-Design Strategy The authors describe a single-zone metasurface array that transcends the aperture-sharing limitations of traditional zoned designs This design enables the metasurface to focus incident light into six polarization-resolved channels within a single device pixel by independently controlling the size and rotation angle of each nanostructure For their geometry (D = 100 µm, f = 1 mm), this corresponds to NA = 0.05, which is approximately 3× larger than the NA of a 1/3‑aperture zoned design at the same focal length The square zone design also enables tiling a plane without empty space in between the device pixels while sampling the wavefront uniformly in lateral directions Where the gains come from The optimization process involves an iterative adjoint method algorithm that maximizes the cost function defined as intensity at the weakest focal point The gradient of the cost function with respect to a geometry parameter is computed using propagation to the focal point via the Green’s transfer function This allows each element’s size and rotation to be independently tuned, maximizing the degrees of freedom available The optimization converges after a few hundred iterations, and the resulting simulated electric field distributions demonstrate a near-perfect polarization reconstruction capability, achieving an average error on the order of 10−7 across the Poincaré sphere Results and Discussion The functionality of the fabricated device in terms of polarization characterization was measured using a custom-built optical setup that allows for testing the metasurface’s response to various incident polarization states prepared by automatically rotating a half-wave plate and a quarter-wave plate in combination with a stationary polarizer The resulting focal spots were recorded with a camera, and the intensity of each focal point was estimated by summing the pixel values within a fixed radius around the peak centers By reframing the polarization evaluation as an inference task for a shallow neural network that maps six measured focal spot intensities on the input to the four Stokes parameters on the output, they effectively calibrate the system against experimental non-idealities including fabrication imperfections The resulting average inference error across the whole Poincaré sphere was 0.046, which is in-line or better than values reported for imaging polarimeters based on division-of-image approach Wavefront Sensing Performance To characterize the Shack–Hartmann wavefront-sensing performance of our metasurface, a custom optical setup was built that allowed them to achieve a wide range of incident illumination angles without the need for sample repositioning The measured data show a displacement of 126.1 px/deg on the sensor, which corresponds to a physical sensitivity of approximately 21.8 µm/deg within the metasurface focal plane Their design produces focal spots with a FWHM of around 4 µm, close to the diffraction-limited 2.7 µm, and spaced 25 µm apart to maintain clear separation This range is sufficient for measurement of typical planar distortions and can be increased further by packing the focal spots closer, which is permitted by the increased NA of the inverse-designed lens Simulations were performed using rigorous coupled-wave analysis (RCWA) implemented in the Python Inkstone library employing periodic boundary conditions with a period of p = 450 nm The high aspect ratio TiO2 nanostructures were fabricated using a method combining electron beam lithography with atomic layer deposition Stokes parameters were reconstructed using a shallow neural network (6-3-3) implemented in PyTorch, trained to minimize mean squared error over 10 000 iterations The work was supported by the Grant Agency of the Czech Republic (project No.

Kai: 25-18336M) The authors acknowledge CzechNanoLab Research Infrastructure (ID 90251), funded by MEYS CR, for the financial support of the sample fabrication We have demonstrated an inverse-designed, visible-light metasurface that integrates fullStokes polarimetry and Shack-Hartmann wavefront sensing into a single, compact aperture By moving beyond the division-of-aperture paradigm, our single-zone design utilizes the entire pixel area for every polarization channel, significantly increasing the device’s numerical aperture The resulting device achieves high-fidelity polarization reconstruction, yielding a mean error magnitude of just 0.046 over 100 test states while maintaining the sensitive wavefront tilt detection required for high-resolution optical metrology This work shows how optimization-driven meta-optics can outperform traditional analytical designs in both multi-functionality and precision References

1: N.

Mira: A.

Kai: Rubin, G.

Title and authors: Kai: So, we're looking at this paper today, "Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light." It sounds like they've managed to pack two really complex things—full polarization measurement and wavefront sensing—into one single device.

Mira: Exactly. It addresses the problem where you usually have to use two separate instruments, which means you’re increasing costs and taking longer for quality control. This paper is proposing a way to merge those functions directly into the hardware itself, using nanostructures as the core component <ref:2604.18879#pg2>.

Lev: And from an engineering standpoint, that merging is huge because it cuts down on external components and calibration steps, which is always a win for real hardware deployment. I wonder if they’ve actually built this thing at a fidelity that matters for industrial stress detection <ref:2604.18879#pg2>.

Kai: The paper dives into how they used an adjoint optimization approach to design this structure. They aren't just tiling the aperture in a standard way; they're breaking the aperture-sharing paradigm, which lets them control each nanostructure’s size and rotation independently <ref:2604.18879#pg3>.

Mira: That independent control is key because it allows them to focus light into six separate polarization channels using just one device pixel <ref:2604.18879#pg3>. It’s not just about fitting things in; it’s about how that design choice fundamentally changes the optics <ref:2604.18879#pg3>.

Lev: Breaking the aperture-sharing means they're using the entire pixel area for all channels, which is a significant increase in their numerical aperture, or NA, compared to traditional zoned designs <ref:2604.18879#pg2>. That higher NA is what gives them that better spatial sampling resolution for wavefront sensing <ref:2604.18879#pg3>.

Kai: And they claim this design lets them achieve a mean polarization reconstruction error of only zero point zero five over one hundred test states, which is pretty good, and it does this while simultaneously tracking the wavefront tilt precisely <ref:2604.18879#pg3>.

Mira: That error number is interesting because they used a shallow neural network to map the six focal spot intensities onto the four Stokes parameters <ref:2604.18879#pg3>. It’s essentially an automated calibration against hardware imperfections <ref:2604.18879#pg3>.

Lev: From a real hardware perspective, that neural network part is where things get tricky; you have to account for fabrication imperfections like the non-linear response of the sensor itself <ref:2604.18879#pg3>. If that calibration doesn't hold up when we move it off the lab bench, the whole system breaks down <ref:2604.18879#pg3>.

Title and authors: Kai: Now let's talk about how they actually measured the wavefront sensing part. They built a custom setup to test its performance across a wide range of incident angles without moving the sample <ref:2604.18879#pg3>.

Mira: The results show a displacement of one hundred twenty-six point per degree on the sensor, which translates to a physical sensitivity of approximately twenty-one point eight micrometers per degree within the metasurface focal plane <ref:2604.18879#pg3>. That sensitivity is what determines how much tilt they can actually detect <ref:2604.18879#pg3>.

Lev: Twenty-one point eight micrometers per degree is a solid number for spatial resolution in this context, and the focal spot size they achieved was around four micrometers, which is close to the diffraction-limited two point seven micrometers <ref:2604.18879#pg3>.

Kai: They spaced those focal spots about twenty-five micrometers apart to keep them distinct for measurement, and that spacing range is good enough to cover typical planar distortions <ref:2604.18879#pg3>.

Mira: What this means for us is that by packing those focal spots closer together, which they can do because of the increased numerical aperture from their inverse design, they can push that sensitivity even higher <ref:2604.18879#pg3>.

Lev: If they pack them tighter, you increase the spatial resolution for tilt detection, but you also increase the complexity of the required reconstruction algorithms <ref:2604.18879#pg3>.

Kai: So to summarize this paper, it shows how inverse design can create a single device that handles full polarization and wavefront sensing simultaneously by controlling every nanostructure independently <ref:2604.18879#pg3>.

Mira: And the key finding is that this unified approach gives them high-fidelity polarization reconstruction with an error of zero point zero five, while still providing the necessary precision for wavefront tilt detection <ref:2604.18879#pg3>.

Lev: It’s a lot of optimization work to get those parameters tuned correctly, but the result shows that this unified architecture is feasible for high-precision metrology <ref:2604.18879#pg3>.

Kai: So looking at how this works, it moves us away from needing separate polarimeters and wavefront sensors by merging them into one compact aperture <ref:2604.18879#pg2>.

Mira: And the design strategy itself is what’s novel—leveraging that adjoint-based inverse design to maximize the degrees of freedom for each element <ref:2604.18879#pg3>. It’s an optimization-driven approach rather than a purely analytical one <ref:2604.18879#pg3>.

Title and authors: Lev: I have to stress that that adjoint method is computationally intensive, but it allows them to independently tune the geometry and rotation of each nanostructure, which is what breaks the traditional aperture-sharing constraint <ref:2604.18879#pg3>. It’s a complex optimization landscape to navigate for real hardware implementation <ref:2604.18879#pg3>.

Kai: So when we look at the overall picture, this paper demonstrates that inverse design can generate multifunctional, non-intuitive flat optics that actually outperform traditional analytical designs <ref:2604.18879#pg1>.

Mira: That is the big implication—that optimization-driven meta-optics can achieve high fidelity in both polarization and wavefront sensing without relying on complex division-of-aperture paradigms <ref:2604.18879#pg2>.

Lev: For running this on a real system, we need to worry about the complexity of that optimization pipeline before it becomes practical for routine industrial use <ref:2604.18879#pg3>. The overhead lands in how robust the calibration is against real-world fabrication variations <ref:2604.18879#pg3>.

Kai: So to wrap up this discussion on "Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light," we see a compact device that combines polarization measurement and wavefront sensing in one shot <ref:2604.18879#pg2>.

Mira: The results are quite compelling because they show the system achieves a mean polarization reconstruction error of zero point zero five across one hundred test states while maintaining the required tilt detection sensitivity <ref:2604.18879#pg3>.

Lev: It’s a solid demonstration that this merged architecture is feasible for high-precision metrology, even if the design process itself is quite intricate with that adjoint method <ref:2604.18879#pg3>.

Kai: This work shows how optimization-driven meta-optics can outperform traditional analytical designs in both multi-functionality and precision, moving toward integrated sensing platforms <ref:2604.18879#pg1>.

Mira: It really points toward a future where we can build these complex sensors directly into the optical components we are testing or analyzing <ref:2604.18879#pg2>.

Lev: For practical application, it means the complexity of that adjoint optimization needs to be manageable so it doesn't require supercomputers for every single design iteration <ref:2604.18879#pg3>.

Kai: So that’s what we had on the table today regarding this paper, showing a single zone design that handles both polarization and wavefront sensing efficiently <ref:2604.18879#pg3>.

Mira: We’ll keep an eye on how they scale up the sensitivity with those focal spots, which is where the next challenge for this kind of design lies <ref:2604.18879#pg3>.

Lev: We'll be watching how they handle the noise and non-idealities when you try to move this from simulation to actual fabrication <ref:2604.18879#pg3>.

The paper's summary: Kai: So, we've seen how they designed this single device to do both full polarization measurement and wavefront sensing at the same time using an inverse design approach <ref:2604.18879#pg1>.

Mira: The core idea is that instead of designing a polarimeter and a sensor separately, they create one structure where every tiny piece controls both the light's polarization and how it bends <ref:2604.18879#pg3>. It’s about maximizing the function of each nanostructure for both tasks <ref:2604.18879#pg3>.

Lev: From an engineering standpoint, they use an adjoint optimization method to tune the geometry and rotation of every nanostructure independently <ref:2604.18879#pg3>. This breaks the rule where you have to share a pixel area between different functions <ref:2604.18879#pg3>.

Kai: That independent tuning lets them use the whole pixel area for every channel, which really boosts their numerical aperture compared to older designs <ref:2604.18879#pg2>.

Mira: And because of that higher NA, they can achieve a much better spatial sampling for the wavefront sensing part <ref:2604.18879#pg3>. It’s a direct trade-off where they get more light gathering ability for better resolution on the tilt detection <ref:2604.18879#pg3>.

Lev: They report that this unified architecture gives them a polarization reconstruction error of about zero point zero five across one hundred test states <ref:2604.18879#pg3>. That’s pretty tight for what they’re doing <ref:2604.18879#pg3>.

Kai: And that error comes with a shallow neural network to fix the hardware imperfections, like how the sensor itself responds non-linearly <ref:2604.18879#pg3>. That calibration step is crucial for making it usable in the real world <ref:2604.18879#pg3>.

Mira: The sensitivity number they got for wavefront sensing is about twenty one point eight micrometers per degree <ref:2604.18879#pg3>. That’s the physical distance they can resolve for tilt <ref:2604.18879#pg3>.

Lev: And they mentioned that if they pack those focal spots closer together, which their higher numerical aperture allows, they can increase that sensitivity further <ref:2604.18879#pg3>. That’s the potential gain for high-resolution tilt detection <ref:2604.18879#pg3>.

Kai: So what this means for us is that we're moving toward integrated sensing platforms where you don't need separate tools for polarization and wavefront information <ref:2604.18879#pg2>.

Mira: It suggests that complex, multifunctional optics aren't just theoretical puzzles anymore; they can be optimized into a single, compact chip <ref:2604.18879#pg3>.

Lev: The big question is how robust that whole optimization process is when you try to move it from simulation to actual fabrication on real hardware <ref:2604.18879#pg3>.

The paper's improvements: Kai: So we've talked about how they designed this single device to do both full polarization measurement and wavefront sensing at the same time <ref:2604.18879#pg1>. Now let's talk about what they suggest doing next with this design <ref:2604.18879#pg3>.

Mira: They are suggesting that because of that inverse design approach, the structure is actually very flexible, so you could potentially tune it for other things too <ref:2604.18879#pg3>. It’s not just fixed to polarization and tilt detection <ref:2604.18879#pg3>.

Lev: They are hinting that since they can control the geometry of each nanostructure independently, you could actually use this same framework to design things for other types of sensing <ref:2604.18879#pg3>. It suggests a more general method for building these integrated devices <ref:2604.18879#pg3>.

Kai: That means the next step isn't just to use this exact structure, but to leverage the optimization process itself as a design tool for other problems <ref:2604.18879#pg3>. It’s about using the learned design rules for more complex tasks <ref:2604.18879#pg3>.

Mira: They imply that we can use this same method to design metasurfaces that handle things beyond just polarization and wavefront sensing, maybe even integrating them into a larger system where they learn from the environment <ref:2604.18879#pg3>.

Lev: But realistically, building the next generation of these requires handling more complex constraints in that optimization pipeline <ref:2604.18879#pg3>. The computational cost for designing a third function might explode <ref:2604.18879#pg3>.

Kai: It means that the design isn't just about getting the best numbers for this one paper, but about creating a scalable process where you can generate these complex, multi-functional optical devices faster than ever before <ref:2604.18879#pg1>.

Mira: And if they manage to keep the error low while adding new functions, it opens up possibilities for real-time diagnostics in materials science or even high-precision metrology setups in industry <ref:2604.18879#pg3>.

Lev: I'm still focused on the practical side; scaling that adjoint method for a truly large array of elements without it taking an infinite amount of time is the real hurdle <ref:2604.18879#pg3>.

Conclusion: Kai: So to wrap up, this paper on "Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light" shows how you can merge polarization measurement and wavefront sensing into one compact aperture <ref:2604.18879#pg2>.

Mira: It's a really neat demonstration of how optimization-driven meta-optics can create multi-functional devices that actually outperform traditional analytical designs <ref:2604.18879#pg1>. It moves us toward integrated sensing platforms where you don't need separate tools for polarization and wavefront information <ref:2604.18879#pg2>.

Lev: From an error correction standpoint, the fact that they achieved a zero point zero five error while maintaining that tilt detection sensitivity is pretty strong evidence that this architecture is feasible <ref:2604.18879#pg3>. It shows the overhead of the calibration fits within what's expected for a complex system <ref:2604.18879#pg3>.

Kai: It really points toward a future where we can build these complex sensors directly into the optical components we are testing or analyzing <ref:2604.18879#pg2>. That kind of integrated capability could be huge in industrial settings <ref:2604.18879#pg3>.

Mira: I think it's important because it shows that the assumptions we make about how these structures will behave under different tasks—like polarization versus wavefront sensing—can be broken by a smart design approach <ref:2604.18879#pg3>.

Lev: I just wonder how robust that whole adjoint optimization process is when you try to move it from simulation to actual fabrication on real hardware <ref:2604.18879#pg3>. That’s the big unknown for scaling this up <ref:2604.18879#pg3>.

Ondžjƙ Červinka, Martin Hrtoůn, Štᮎpan Venos, Jakub Lelek, Libor ůlehla, •Tomśľ Šikola, Filip Ligmajer

Brno University of Technology · Central European Institute of Technology · Meopta s.r.o.

physics.optics, physics.app-ph

Submitted: 2026-04-20

Updated: 2026-04-20

Journal ref: ACS Photonics (2026) 13 (16) 4744-4752

DOI: 10.1021/acsphotonics.6c00985

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 90/100

The gist: Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light demonstrates an inverse-designed metasurface that integrates full-Stokes polarimetry and…

Key concepts

Full-Stokes Polarimetry
This technique measures all four components (Stokes parameters) of light's polarization—intensity and orientation—to fully characterize how light is polarized. The metasurface is designed to separate and analyze these polarization states simultaneously across a single device.
Shack-Hartmann Wavefront Sensing
This method detects small distortions in a surface by observing how they cause incoming light rays to focus at different points on a sensor array. The metasurface's focal spots are used as the input for this sensor to precisely measure wavefront tilt and surface shape.
Inverse Design with Adjoint Optimization
Instead of designing the structure forward, this method starts with a desired performance goal (like perfect polarization measurement) and iteratively optimizes the nanostructure's geometry and rotation. It uses an adjoint optimization algorithm to efficiently find the optimal settings for every element.
Aperture-Sharing Paradigm Break
Traditional designs limit how much of the aperture each element can use. This design breaks that rule by allowing each nanostructure to control its own size and rotation independently, utilizing the entire pixel area for all polarization channels, which significantly boosts performance.

Terminology

Summary

Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light demonstrates an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum This work highlights the capacity of inverse design to generate multifunctional, non-intuitive flat optics that outperforms its traditional counterparts

Pruning redundant reasoning tokens

The paper presents an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum The authors demonstrate this by leveraging an adjoint optimization approach to independently control the geometry and rotation of each nanostructure, which breaks the aperture-sharing paradigm and utilizes the entire pixel area for all channels When coupled with a shallow neural network to automate peak identification and correct for hardware non-idealities, the device yields a mean polarization reconstruction error of only 0.046 across 100 test states on the Poincaré sphere while simultaneously maintaining the precise focal-spot tracking required for sensitive wavefront tilt detection

Operating Concept and Inverse-Design Strategy

The authors describe a single-zone metasurface array that transcends the aperture-sharing limitations of traditional zoned designs This design enables the metasurface to focus incident light into six polarization-resolved channels within a single device pixel by independently controlling the size and rotation angle of each nanostructure For their geometry (D = 100 µm, f = 1 mm), this corresponds to NA = 0.05, which is approximately 3× larger than the NA of a 1/3‑aperture zoned design at the same focal length The square zone design also enables tiling a plane without empty space in between the device pixels while sampling the wavefront uniformly in lateral directions

Where the gains come from

The optimization process involves an iterative adjoint method algorithm that maximizes the cost function defined as intensity at the weakest focal point The gradient of the cost function with respect to a geometry parameter is computed using propagation to the focal point via the Green’s transfer function This allows each element’s size and rotation to be independently tuned, maximizing the degrees of freedom available The optimization converges after a few hundred iterations, and the resulting simulated electric field distributions demonstrate a near-perfect polarization reconstruction capability, achieving an average error on the order of 10−7 across the Poincaré sphere

Results and Discussion

The functionality of the fabricated device in terms of polarization characterization was measured using a custom-built optical setup that allows for testing the metasurface’s response to various incident polarization states prepared by automatically rotating a half-wave plate and a quarter-wave plate in combination with a stationary polarizer The resulting focal spots were recorded with a camera, and the intensity of each focal point was estimated by summing the pixel values within a fixed radius around the peak centers By reframing the polarization evaluation as an inference task for a shallow neural network that maps six measured focal spot intensities on the input to the four Stokes parameters on the output, they effectively calibrate the system against experimental non-idealities including fabrication imperfections The resulting average inference error across the whole Poincaré sphere was 0.046, which is in-line or better than values reported for imaging polarimeters based on division-of-image approach

Wavefront Sensing Performance

To characterize the Shack–Hartmann wavefront-sensing performance of our metasurface, a custom optical setup was built that allowed them to achieve a wide range of incident illumination angles without the need for sample repositioning The measured data show a displacement of 126.1 px/deg on the sensor, which corresponds to a physical sensitivity of approximately 21. p. 9

Where the paper says it

45 µm, this corresponds to a measured sensitivity of approximately 435 µm/deg at the camera plane. Accounting for the 20× magnification of the imaging system, this translates to a physical sensitivity of approximately 21.

8 µm/deg within the metasurface focal plane Their design produces focal spots with a FWHM of around 4 µm, close to the diffraction-limited 2. p. 9
Where the paper says it

In our metasurface-based implementation, the spot size and spacing are similarly critical: too large a spot or too small a separation could lead to overlaps, reducing measurement feasibility. Our design produces focal spots with a FWHM of around 4 µm, close to the diffraction-limited 2.

7 µm, and spaced 25 µm apart to maintain clear separation This range is sufficient for measurement of typical planar distortions and can be increased further by packing the focal spots closer, which is permitted by the increased NA of the inverse-designed lens

Methods

Simulations were performed using rigorous coupled-wave analysis (RCWA) implemented in the Python Inkstone library employing periodic boundary conditions with a period of p = 450 nm The high aspect ratio TiO2 nanostructures were fabricated using a method combining electron beam lithography with atomic layer deposition Stokes parameters were reconstructed using a shallow neural network (6-3-3) implemented in PyTorch, trained to minimize mean squared error over 10 000 iterations The work was supported by the Grant Agency of the Czech Republic (project No. 25-18336M) The authors acknowledge CzechNanoLab Research Infrastructure (ID 90251), funded by MEYS CR, for the financial support of the sample fabrication

Conclusion

We have demonstrated an inverse-designed, visible-light metasurface that integrates fullStokes polarimetry and Shack-Hartmann wavefront sensing into a single, compact aperture By moving beyond the division-of-aperture paradigm, our single-zone design utilizes the entire pixel area for every polarization channel, significantly increasing the device’s numerical aperture The resulting device achieves high-fidelity polarization reconstruction, yielding a mean error magnitude of just 0.046 over 100 test states while maintaining the sensitive wavefront tilt detection required for high-resolution optical metrology This work shows how optimization-driven meta-optics can outperform traditional analytical designs in both multi-functionality and precision

References

[1] N. A. Rubin, G. D’Aversa, P. Chevalier, Z. Shi, W. T. Chen, and F. Capasso, “Matrix Fourier optics enables a compact full-Stokes polarization camera,” Science (1979). p. 0

Where the paper says it

To achieve this in a compact footprint, existing metasurfaces typically partition their aperture into discrete zones, which inherently restricts the light-gathering efficiency and numerical aperture of the system. Here we demonstrate an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum.

[2] N. A. Rubin et al., “Imaging polarimetry through metasurface polarization gratings,” Optics Express, Vol. p. 1

Where the paper says it

Consequently, a unified architecture that maximizes light-gathering efficiency while delivering both complete polarization and wavefront metrology without compromises does not exist. In this work, we present an inverse-designed metasurface that enables simultaneous fullStokes polarimetry and Shack-Hartmann wavefront sensing in the visible spectrum.

30, Issue 6, pp. 9389-9412, vol. 30, no. 6, pp. 9389–9412

[3] E. Arbabi, S. M. Kamali, A. Arbabi, and A. Faraon, “Full-Stokes Imaging Polarimetry Using Dielectric Metasurfaces,” ACS Photonics, vol. p. 0

Where the paper says it

To achieve this in a compact footprint, existing metasurfaces typically partition their aperture into discrete zones, which inherently restricts the light-gathering efficiency and numerical aperture of the system. Here we demonstrate an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum.

5, no. 8, pp. 3132–3140

[4] X. Li, H. Wang, X. Xu, L. Ju, Z. Fan, and S. Chen, “Mid-infrared full-Stokes polarization detection based on dielectric metasurfaces,” Opt. p. 1

Where the paper says it

Contrary to the analytical designs used in previous works, the modern inverse design techniques do not require limiting the metasurface layout to zones or metamolecules, but instead they search a vast parameter space of nanostructure layouts to optimize a given cost function [15]. While inverse design approaches have recently shown promise for polarimetric metasurfaces[16], [17] even in the visible range [18], applying them to simultaneously resolve full-Stokes polarization and precise wavefront distortions without sacrificing efficiency or spatial resolution remains an outstanding challenge.

Commun., vol. 484, p. 126690

[5] Y. Ren et al., “Full-Stokes Polarimetry for Visible Light Enabled by an All-Dielectric Metasurface,” Adv. p. 0

Where the paper says it

To achieve this in a compact footprint, existing metasurfaces typically partition their aperture into discrete zones, which inherently restricts the light-gathering efficiency and numerical aperture of the system. Here we demonstrate an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum.

Photonics Res., vol. 3, no. 7, p. 2100373

[6] G. H. Go, D. G. Lee, J. Oh, G. Song, D. Lee, and M. Jang, “Meta Shack–Hartmann wavefront sensor with large sampling density and large angular field of view: phase imaging of complex objects,” Light: Science & Applications 2024 13:1

[7] C. Zhao et al., “High-resolution wavefront detection using binary phase grating modulation for large dynamic range,” Optics Express, Vol. p. 7

Where the paper says it

The overall power efficiency of our inversedesigned metasurface is the same as that of an equivalent metasurface designed with three separate zones using spatial multiplexing, which has been verified by integrating the power over the focal points in simulations of the two designs. However, the focal spots produced by the inverse-designed metasurface were significantly better confined, allowing the integration to be performed over a smaller area and improve the dynamic range of Shack-Hartmann wavefront sensing.

33, Issue 7, pp. 16082-16099

[8] Z. Wu, J. Yu, F. Zheng, J. Zhang, and H. Xu, “Diffractive encoding microlens arrays for real-time wavefront sensing with large dynamic range and high sensitivity,” Opt. p. 9

Where the paper says it

In a Shack-Hartmann sensor, the sensitivity of wavefront measurements is a fundamental tradeoff against the allowable range of incident angles, constrained by the numerical aperture of the individual lens, which must be chosen to avoid overlaps between neighboring focal spots. In our metasurface-based implementation, the spot size and spacing are similarly critical: too large a spot or too small a separation could lead to overlaps, reducing measurement feasibility.

Lasers Eng., vol. 194, p. 109130

[9] A. Martin Jimenez, M. p. 0

Where the paper says it

Inverse designed full-Stokes polarimetric metasurface with simultaneous wavefront sensing for visible light Ondřej Červinka1 , Martin Hrtoň1,2 , Štěpán Venos3 , Jakub Lelek3 , Libor Úlehla3 , Tomáš Šikola1,2, Filip Ligmajer1,2, * 1Brno University of Technology, Central European Institute of Technology, Purkyňova 123, 61200 Brno, Czech Republic 2Brno University of Technology, Faculty of Mechanical Engineering, Institute of Physical Engineering, Technická 2, 61669, Brno, Czech Republic 3Meopta s. , Kabelíkova 1, 75002 Přerov, Czech Republic *Contact author: filip.

Baltes, J. Cornelius, N. Aközbek, and Z. J. Coppens, “Single-shot phase diversity wavefront sensing in deep turbulence via metasurface optics,” Nature Photonics 2025 19:12

[10] X. Feng et al., “Dielectric metalens-based Hartmann-Shack array for a high-efficiency optical multiparameter detection system,” Photonics Research, Vol. p. 0

Where the paper says it

To achieve this in a compact footprint, existing metasurfaces typically partition their aperture into discrete zones, which inherently restricts the light-gathering efficiency and numerical aperture of the system. Here we demonstrate an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum.

8, Issue 4, pp. 482–489

[11] Z. Yang et al., “Generalized Hartmann-Shack array of dielectric metalens sub-arrays for polarimetric beam profiling,” Nature Communications 2018 9:1

[12] J. P. Balthasar Mueller, N. A. Rubin, R. C. Devlin, B. Groever, and F. Capasso, “Metasurface Polarization Optics: Independent Phase Control of Arbitrary Orthogonal States of Polarization,” Phys. p. 1

Where the paper says it

The second component of the quality control, wavefront measurements, has also been performed using metasurfaces, replacing microlens arrays with metalenses and providing improved spatial sampling resolution and field of view [6], [7], [8], [9]. In contrast to conventional optics, these two metasurface-based approaches can be merged together for simultaneous acquisition of both the polarization state and wavefront.

Rev. Lett., vol 118, no 11, p 113901

[13] J. Liu, J. Chu, R. Zhang, R. Liu, and J. Fu, “Wide field of view and full Stokes polarization imaging using metasurfaces inspired by the stomatopod eye,” Nanophotonics, vol 12, no 6, pp 1137–1146

[14] L. Li et al., “Dielectric Metalens Array for Simultaneous Polarization and Wavefront Mapping in the Visible Spectrum,” Nano Lett. p. 1

Where the paper says it

Consequently, a unified architecture that maximizes light-gathering efficiency while delivering both complete polarization and wavefront metrology without compromises does not exist. In this work, we present an inverse-designed metasurface that enables simultaneous fullStokes polarimetry and Shack-Hartmann wavefront sensing in the visible spectrum.

, vol. 25, no. 27, pp. 10879–10887

[15] Z. Li, R. Pestourie, Z. Lin, S. G. Johnson, and F. Capasso, “Empowering Metasurfaces with Inverse Design: Principles and Applications,” ACS Photonics, vol 9, no 7, pp 2178–2192

[16] W. Zhou et al., “Full Stokes polarimetry based on an inverse-designed multi-foci metalens,” Optics Letters, Vol. p. 0

Where the paper says it

To achieve this in a compact footprint, existing metasurfaces typically partition their aperture into discrete zones, which inherently restricts the light-gathering efficiency and numerical aperture of the system. Here we demonstrate an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum.

49, Issue 6, pp. 1595–1598

[17] Y. Wang et al., “Non-Interleaved Shared-Aperture Full-Stokes Metalens via PriorKnowledge-Driven Inverse Design,” Advanced Materials, vol. p. 1

Where the paper says it

While inverse design approaches have recently shown promise for polarimetric metasurfaces[16], [17] even in the visible range [18], applying them to simultaneously resolve full-Stokes polarization and precise wavefront distortions without sacrificing efficiency or spatial resolution remains an outstanding challenge. Analytical approaches still face performance constraints by dividing the aperture or relying on complex meta-molecules.

37, no 8, p 2408978

[18] Y. Hu, Y. Jiang, Y. Zhang, Q. Wang, M. Pan, and H. Duan, “Achromatic Full Stokes Polarimetry Metasurface for Full-Color Polarization Imaging in the Visible Range,” Nano Lett. p. 1

Where the paper says it

Contrary to the analytical designs used in previous works, the modern inverse design techniques do not require limiting the metasurface layout to zones or metamolecules, but instead they search a vast parameter space of nanostructure layouts to optimize a given cost function [15]. While inverse design approaches have recently shown promise for polarimetric metasurfaces[16], [17] even in the visible range [18], applying them to simultaneously resolve full-Stokes polarization and precise wavefront distortions without sacrificing efficiency or spatial resolution remains an outstanding challenge.

, 2024

[19] R. A. Chipman, W.-S. T. Lam, and G. Young, “Polarized Light and Optical Systems,” Polarized Light and Optical Systems, vol 7. p. 3

Where the paper says it

2a shows the library we obtained and used for our metasurface optimization, consisting of TiO2 nanostructures with lateral dimensions, width (𝑊𝑚) and length (𝐿𝑚), constrained to 130-330 nm and a height fixed to 𝐻 = 600 nm. Each entry of the library stores transmission amplitudes (𝐴) and phase delays (𝜙) for 𝑥- and 𝑦-polarized light at the target wavelength of 𝜆 = 532 nm.

4.1, Jul 2019

[20] W. T. Chen et al., “Integrated plasmonic metasurfaces for spectropolarimetry,” Nanotechnology, vol. p. 7

Where the paper says it

122 @ 550 nm, Ref. [18]), plasmonic metagratings (0.

27, no 22, p 224002

[21] A. Pors, M. G. Nielsen, and S. I. Bozhevolnyi, “Plasmonic metagratings for simultaneous determination of Stokes parameters,” Optica, Vol 2, Issue 8, pp 716-723

[22] P. A. Khorin, A. P. Dzyuba, and S. N. Khonina, “Optical wavefront aberration: detection, recognition, and compensation techniques – a comprehensive review,” Opt. p. 1

Where the paper says it

Contrary to the analytical designs used in previous works, the modern inverse design techniques do not require limiting the metasurface layout to zones or metamolecules, but instead they search a vast parameter space of nanostructure layouts to optimize a given cost function [15]. While inverse design approaches have recently shown promise for polarimetric metasurfaces[16], [17] even in the visible range [18], applying them to simultaneously resolve full-Stokes polarization and precise wavefront distortions without sacrificing efficiency or spatial resolution remains an outstanding challenge.

Laser Technol. p. 8
Where the paper says it

To characterize the Shack–Hartmann wavefront-sensing performance of our metasurface, we built a custom optical setup shown in Fig. The setup includes a rotatable arm carrying a laser source and a rotating frosted glass diffuser, which reduces temporal coherence and suppresses speckles in the image plane.

, vol 191, p 113342

[23] S. W. Moon, Y. Kim, G. Yoon, and J. Rho, “Recent Progress on Ultrathin Metalenses for Flat Optics,” iScience, vol 23, no 12, p 101877

[24] L. Novotny and B. Hecht, “Principles of Nano-Optics,” Principles of Nano-Optics, vol 9781107005464, pp 1–564

[25] Q. Jiang et al., “Multiwavelength Achromatic Metalens in Visible by Inverse Design,” Adv. p. 1

Where the paper says it

[3], [13], the aperture of each metalens remains limited to only one-third of the total pixel size, inherently constraining both the lightgathering efficiency and the numerical aperture of the system. Very recently, polarizationmultiplexing has been achieved in the visible domain over all six polarization states in a single zone [14] through the design of meta-molecules consisting of four nanostructures, providing sufficient degrees of freedom to simultaneously address all six polarization basis states.

Opt. Mater., vol 11, no 15, p 2300077

[26] Y. Yin et al., “Multi-Dimensional Multiplexed Metasurface Holography by Inverse Design,” Advanced Materials, vol. p. 1

Where the paper says it

Very recently, polarizationmultiplexing has been achieved in the visible domain over all six polarization states in a single zone [14] through the design of meta-molecules consisting of four nanostructures, providing sufficient degrees of freedom to simultaneously address all six polarization basis states. Contrary to the analytical designs used in previous works, the modern inverse design techniques do not require limiting the metasurface layout to zones or metamolecules, but instead they search a vast parameter space of nanostructure layouts to optimize a given cost function [15].

36, no 21, p 2312303

[27] R. C. Devlin, M. Khorasaninejad, W. T. Chen, J. Oh, and F. Capasso, “Broadband highefficiency dielectric metasurfaces for the visible spectrum,” Proc. p. 0

Where the paper says it

To achieve this in a compact footprint, existing metasurfaces typically partition their aperture into discrete zones, which inherently restricts the light-gathering efficiency and numerical aperture of the system. Here we demonstrate an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum.

Natl Acad Sci U S A., vol 113, no 38, pp 10473–10478

[28] B. Boulbry, J. C. Ramella-Roman, and T. A. Germer, “Improved method for calibrating a Stokes polarimeter,” Applied Optics, Vol 46, Issue 35, pp 8533-8541

[29] F. Wang and A. Theuwissen, “Linearity analysis of a CMOS image sensor,” IS and T International Symposium on Electronic Imaging Science and Technology, pp 84–90

[30] A. Erfanmanesh, “The Citation Advantage of AI-Related Publications,” Science Editor, Oct 2025

[31] A. Pors, M. G. Nielsen, and S. I. Bozhevolnyi, “Plasmonic metagratings for simultaneous determination of Stokes parameters,” Optica, Vol 2, Issue 8, pp 716–723

[32] Q. Fan et al., “Disordered metasurface enabled single-shot full-Stokes polarization imaging leveraging weak dichroism,” Nature Communications 2023 14:1

[33] “inkstone · PyPI.” Accessed: Nov. 19, 2025, [Online]. Available: https://pypi. p. 4

Where the paper says it

𝑇⃡𝑚 = [ 𝑇𝑚𝑥𝑥 𝑇𝑚𝑥𝑦 𝑇𝑚𝑦𝑥 𝑇𝑚𝑦𝑦] = 𝑅⃡(−𝜃𝑚) [ 𝐽𝑚𝑥 (𝑊𝑚, 𝐿𝑚) 0 0 𝐽𝑚𝑦 (𝑊𝑚, 𝐿𝑚) ]𝑅⃡(𝜃𝑚). (3) This parameterization allows the algorithm to treat both dimensions and orientation as continuous variables and allows each element’s size and rotation to be independently tuned, maximizing the degrees of freedom available.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Full-Stokes polarimetry and wavefront sensing integration allows for real-time observation of internal stresses and deformations of optical components as they are mounted or assembled. This enables real-time diagnostics of stress-induced birefringence and surface flatness deviations in industrial applications.

  2. The inverse design approach leverages an adjoint optimization method to independently control the geometry and rotation of each nanostructure, breaking the aperture-sharing paradigm to utilize the entire pixel area for all channels. This allows for a unified architecture that maximizes light-gathering efficiency while delivering both complete polarization and wavefront metrology without compromises.

  3. The system can perform high-fidelity polarization reconstruction with a mean error of 0.046 across 100 test states, superior to standard linear calibration methods which yield an average error of 0.049. This capability is achieved by coupling the metasurface with a shallow neural network that automates peak identification and corrects for hardware non-idealities like CMOS sensor non-linearity.

  4. The Shack-Hartmann wavefront sensing performance shows a physical sensitivity of approximately 21.8 micrometers per degree, which can be increased further by packing focal spots closer due to the increased numerical aperture achieved by the inverse-designed lens compared to spatial multiplexing designs. This provides high spatial resolution for precise wavefront tilt detection.

Abstract

Metasurfaces have emerged as a powerful platform for compact optical sensors by replacing bulky lenses with flat arrays of subwavelength nanostructures. In precision optical metrology, the simultaneous mapping of a beam's polarization state and wavefront is crucial for real-time diagnostics of stress-induced birefringence and surface flatness. To achieve this in a compact footprint, existing metasurfaces typically partition their aperture into discrete zones, which inherently restricts the light-gathering efficiency and numerical aperture of the system. Here we demonstrate an inverse-designed metasurface that integrates full-Stokes polarimetry and Shack-Hartmann wavefront sensing within a single, continuous aperture in the visible spectrum. By leveraging an adjoint optimization approach to independently control the geometry and rotation of each nanostructure, we break the aperture-sharing paradigm and utilize the entire pixel area for all channels. When coupled with a shallow neural network to automate peak identification and correct for hardware non-idealities, our device yields a mean polarization reconstruction error of only 0.046 across 100 test states on the Poincaré sphere, while simultaneously maintaining the precise focal-spot tracking required for sensitive wavefront tilt detection. This work highlights the capacity of inverse design to generate multifunctional, non-intuitive flat optics that outperforms its traditional counterparts.

Related papers