Multifunctional Imaging with an Inverse-Designed Nonlocal Metasurface
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: We're looking at the paper: "Multifunctional Imaging with an Inverse-Designed Nonlocal Metasurface".
Mira: Here's the gist of the paper. Nonlocal metasurfaces enable all-optical processing of spatial information in optical fields This work demonstrates a topology-optimised metasurface that switches between phase-contrast and brightfield imaging modalities via polarisation control Core Concept of the Device The paper presents a topology-optimised metasurface designed to perform switchable phase contrast/brightfield imaging at normal incidence by changing the illuminating polarisation The device switches between performing the identity operation, corresponding to conventional brightfield imaging, and a first-order derivative with respect to x when illuminated with y-polarised light for phase imaging This switching mechanism eliminates the need to reposition optical components or use computational techniques to image transparent samples Design and Optimization Methodology The metasurface design employs a topology optimisation algorithm that explores a very large design space to iteratively determine the distribution of material in the metasurface to achieve a desired optical response The optimization considers robustness to dilation and erosion fabrication defects by generating eroded and dilated variants and using a weighted average in the cost function The algorithm uses coupled wave analysis (RCWA) to evaluate the optical transfer function (OTF) against a target OTF, minimizing the root mean square error (RMSE) Optical Response and Characterization The spatial filtering is implemented in the object plane, where the transmission of a scalar system with a linear transfer function like H(kx, ky) = αkx+β can be converted into an intensity modulation that can be detected The metasurface is designed to produce a constant OTF for x-polarised light corresponding to brightfield imaging and a linear OTF in kx for y-polarised light, with the range of angles defining the numerical aperture (NA) Characterisation involves fabricating the structure using a top-down approach involving electron beam lithography and etching Imaging Performance and Applications The metasurface successfully visualises synthetic phase test targets, biological cells, and tissue samples by recovering the phase gradient along the x-direction using polarised light For biological samples like HeLa cells, switching to y-polarised light makes cell location and morphology much clearer, yielding images in good agreement with differential interference contrast (DIC) microscopy The ability to perform computational operations directly with light is demonstrated for applications in biological imaging, medical diagnostics, and materials characterisation Limitations and Future Directions Small differences exist between simulations and fabricated metasurface results due to non-uniform design defects and mismatches in optical properties The errors in phase gradient recovery are dominated by noise arising primarily from small defects in the metasurface Future work could focus on designing a metasurface to recover gradients along both x and y directions by changing the polarisation to enable quantitative phase imaging The approach can be extended to operations in the image plane for designing computational camera systems The paper successfully demonstrates a non-local freeform metasurface engineered to switch between a brightfield, identity operation and a phase imaging, first-order derivative operation simply by changing the polarisation Topology optimisation is shown to be a powerful approach for designing metasurfaces for image processing The use of non free-space propagating diffracted orders to generate asymmetry can be tailored to imaging different types of samples in both the object plane and image plane This work enables ultra-compact, low-cost imaging systems and energy-free image processing operations The fabrication involves depositing amorphous silicon, performing electron beam lithography, etching using a pseudo-Bosch process, and then lift-off with an Al2O3 layer Phase test targets are fabricated using greyscale lithography to control the phase by varying the thickness of the photoresist The phase is calculated using the expression ϕ(x, y) = 2π∆n/λT(x, y) Simulation Details Finite element method (FEM) simulations are used to model the optical properties of the metasurface and calculate the phase response of the 0th order which carries the processed image The 2D OTF is calculated by imaging the back-focal plane using a custom built microscope setup A finite size analysis shows that for a 300 µm x 300 µm metasurface, the diffraction minimised field of view for ∆I ≤ 0.1 is approximately 22 × 22 µm2 Experimental Validation Imaging experiments are performed using a custom built microscope setup where light is collimated and polarised before illuminating the metasurface with the phase object The recovered phase gradient is quantitatively extracted by solving a quadratic equation for the phase derivative, requiring near perfect alignment of the metasurface The results show that when switching to y-polarised light, the location and morphology of cells become much clearer, which is in good agreement with DIC microscopy The recovered phase gradient image and the metasurface y-polarised image are in good agreement with the DIC image Fabrication Robustness The robustness of the design is tested by generating eroded and dilated variants of the structure and using a cost function that is a weighted average of these variants The simulation shows that erosion and dilation have a minimal effect on the brightfield imaging mode, but they can cause small blue-shifting or red-shifting in the linear region of the OTF The fabrication process is designed to be robust to under or over etched designs through edge detection techniques Data Availability Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request The authors acknowledge support from the Australian Government through the Australian Research Council Centre of Excellence grant (CE200100010) and the Commonwealth through an Australian Government Research Training Program Scholarship The work was performed in part at the Melbourne Centre for Nanofabrication (MCN) in the Victorian Node of the Australian National Fabrication Facility (ANFF), and in part at the Materials Characterisation and Fabrication Platform (MCFP) at the University of Melbourne The authors would like to thank Dr Nitu Syed for performing the ellipsometry of the resist AZ5214E and Haiwei Wang for their useful discussions The paper is available on arXiv as 2606.21051v2
physics.optics: 24 Jun 2026 <ref:arXiv:2606.21051v2
physics.optics: >.
Paper summary: Kai: So to recap this paper, "Multifunctional Imaging with an Inverse-Designed Nonlocal Metasurface," the authors are proposing a topology-optimized metasurface that can perform all optical processing of spatial information. Their main thesis is that they’ve engineered a structure that switches between phase contrast and brightfield imaging modalities simply by changing the polarization of the light illuminating it > This eliminates the need to reposition optical components or use computational techniques to image transparent samples.
Mira: Essentially, for one polarization state, an asymmetric transfer function about normal incidence performs a first order derivative on the phase, which produces pseudo three dimensional phase-contrast images of transparent biological samples > And when you switch to the orthogonal polarization state, it returns the result of the identity operator > This is what they claim allows for this switching between imaging modes.
Lev: So, it’s about creating a single physical structure that acts as two different optical filters depending on which way you shine light on it > That means instead of needing two separate devices—one for brightfield and one for phase contrast—you have one device that does both > But I have to ask, how robust is the topology optimization against real-world fabrication issues like dilation and erosion defects?
Kai: They tackle that by generating eroded and dilated variants of the structure and using a weighted average in their cost function > The simulation shows that erosion and dilation have a minimal effect on the brightfield imaging mode, but they can cause small blue shifting or red shifting in the linear region of the transfer function > It seems like it handles fabrication imperfections reasonably well for basic modes.
Mira: That's where my theoretical perspective comes in, Lev. The results really depend on their assumptions about how that weighted average cost function truly reflects the physical robustness they need > If those defects cause significant phase noise, then you might lose the benefit of having a single device instead of two simpler ones > They are pushing inverse-design methods toward reconfigurable phase contrast microscopy and quantitative analogue optical computation.
Lev: I agree with Mira. To run this on real hardware, we need to know if that weighted average cost function actually minimizes the error in a way that’s useful for noise reduction, not just minimizing the structural deviation > If the noise from those small defects dominates the phase gradient recovery, then even this elegant switching mechanism won't give us much advantage over traditional DIC.
Kai: So they're pushing this idea because there remains an ongoing need to develop ultra-compact real-time imaging systems that can perform multiple image processing modalities directly with light > They want to show that you don't need additional optical components for these operations.
Mira: Right, the implication is they are targeting applications in biological imaging, medical diagnostics, and materials characterisation where compactness and direct computation are key.
Conclusion: Kai: Thinking about the title, "Multifunctional Imaging with an Inverse-Designed Nonlocal Metasurface," it really tells you the ambition here > It’s not just a new imaging device; it’s a design philosophy where you engineer the structure itself to handle different tasks.
Mira: And they are doing exactly that by using inverse design methods to create this metasurface, which is non-local, meaning its properties aren't just defined locally but across the whole structure > They are not just stacking materials; they are designing a complex optical system from the ground up to perform computation.
Lev: What this means practically for someone listening here is that instead of buying a new piece of hardware every time you want to switch between brightfield and phase contrast, you could potentially have one device that handles both modes electronically > But we need to remember the caveat about fabrication noise they mentioned, which suggests that as we move toward real-world deployment, the performance will depend heavily on how precisely those defects are controlled.
Kai: Exactly. The authors acknowledge that their simulation and fabricated results show small differences due to non uniform design defects and mismatches in optical properties > So it’s not a perfect solution yet; it's a path toward ultra-compact systems, but the engineering challenge is still perfecting that robustness.
Mira: It’s about showing that you can perform computations directly with light without needing extra hardware or heavy computation on a computer > This moves the capability closer to true in situ diagnostics and characterization of materials where space and energy are tight constraints.
Lev: So, the big picture is this: we're seeing a path toward integrated optical systems that handle complex image processing right at the light level, provided we can manage those small fabrication imperfections effectively.
Lincoln Clark, *, Yang Xu, , , , Cat-Uyen Phan, , David J. Byrne, , Shikun Ma, Lukas Wesemann, Elizabeth Hinde, Kylie L. Gorringe
ARC Centre of Excellence for Transformative Meta-Optical Systems, University of Melbourne ARC Centre of Excellence for Transformative Meta-optical Systems, Department of Electrical and Electronic Engineering, University of Melbourne ARC Centre of Excellence for Transformative Meta-optical Systems, School of Physics, The University of Melbourne · Sir Peter MacCallum Dept. Of Oncology, University Of Melbourne · Department Of Pathology, Peter MacCallum Cancer Centre
physics.optics
Submitted: 2026-06-19
Updated: 2026-06-24
Comments: Main text: 29 pages, 6 figures. Supplementary information included: 19 pages, 16 figures
Journal ref: Optica 13, (2026), 2342-2353
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 80/100
The gist: Nonlocal metasurfaces enable all-optical processing of spatial information in optical fields This work demonstrates a topology-optimised metasurface that switches between phase-contrast and…
Key concepts
- Nonlocal Metasurfaces
- These are specially designed surfaces that process spatial information within optical fields. They allow light to interact with itself in complex ways, enabling advanced optical processing like switching imaging modes and performing computations directly with light.
- Topology Optimization
- This is a computational method used to design the physical structure of the metasurface. The algorithm explores many possible designs iteratively, using simulations to find the material distribution that achieves a specific desired optical response, such as switching between imaging types.
- Phase-Contrast/Brightfield Imaging
- These are two different ways to capture an image from a sample. Brightfield imaging captures intensity information directly, while phase-contrast imaging recovers the phase variations of the light to reveal fine details and structures that are invisible in standard brightfield images.
Terminology
Summary
Nonlocal metasurfaces enable all-optical processing of spatial information in optical fields This work demonstrates a topology-optimised metasurface that switches between phase-contrast and brightfield imaging modalities via polarisation control
Core Concept of the Device
The paper presents a topology-optimised metasurface designed to perform switchable phase contrast/brightfield imaging at normal incidence by changing the illuminating polarisation The device switches between performing the identity operation, corresponding to conventional brightfield imaging, and a first-order derivative with respect to x when illuminated with y-polarised light for phase imaging This switching mechanism eliminates the need to reposition optical components or use computational techniques to image transparent samples
Design and Optimization Methodology
The metasurface design employs a topology optimisation algorithm that explores a very large design space to iteratively determine the distribution of material in the metasurface to achieve a desired optical response The optimization considers robustness to dilation and erosion fabrication defects by generating eroded and dilated variants and using a weighted average in the cost function The algorithm uses coupled wave analysis (RCWA) to evaluate the optical transfer function (OTF) against a target OTF, minimizing the root mean square error (RMSE)
Optical Response and Characterization
The spatial filtering is implemented in the object plane, where the transmission of a scalar system with a linear transfer function like H(kx, ky) = αkx+β can be converted into an intensity modulation that can be detected The metasurface is designed to produce a constant OTF for x-polarised light corresponding to brightfield imaging and a linear OTF in kx for y-polarised light, with the range of angles defining the numerical aperture (NA) Characterisation involves fabricating the structure using a top-down approach involving electron beam lithography and etching
Imaging Performance and Applications
The metasurface successfully visualises synthetic phase test targets, biological cells, and tissue samples by recovering the phase gradient along the x-direction using polarised light For biological samples like HeLa cells, switching to y-polarised light makes cell location and morphology much clearer, yielding images in good agreement with differential interference contrast (DIC) microscopy The ability to perform computational operations directly with light is demonstrated for applications in biological imaging, medical diagnostics, and materials characterisation
Limitations and Future Directions
Small differences exist between simulations and fabricated metasurface results due to non-uniform design defects and mismatches in optical properties The errors in phase gradient recovery are dominated by noise arising primarily from small defects in the metasurface Future work could focus on designing a metasurface to recover gradients along both x and y directions by changing the polarisation to enable quantitative phase imaging The approach can be extended to operations in the image plane for designing computational camera systems
Conclusion
The paper successfully demonstrates a non-local freeform metasurface engineered to switch between a brightfield, identity operation and a phase imaging, first-order derivative operation simply by changing the polarisation Topology optimisation is shown to be a powerful approach for designing metasurfaces for image processing The use of non free-space propagating diffracted orders to generate asymmetry can be tailored to imaging different types of samples in both the object plane and image plane This work enables ultra-compact, low-cost imaging systems and energy-free image processing operations
Methods Summary
The fabrication involves depositing amorphous silicon, performing electron beam lithography, etching using a pseudo-Bosch process, and then lift-off with an Al2O3 layer Phase test targets are fabricated using greyscale lithography to control the phase by varying the thickness of the photoresist The phase is calculated using the expression ϕ(x, y) = 2π∆n/λT(x, y)
Simulation Details
Finite element method (FEM) simulations are used to model the optical properties of the metasurface and calculate the phase response of the 0th order which carries the processed image The 2D OTF is calculated by imaging the back-focal plane using a custom built microscope setup A finite size analysis shows that for a 300 µm x 300 µm metasurface, the diffraction minimised field of view for ∆I ≤ 0.1 is approximately 22 × 22 µm2
Experimental Validation
Imaging experiments are performed using a custom built microscope setup where light is collimated and polarised before illuminating the metasurface with the phase object The recovered phase gradient is quantitatively extracted by solving a quadratic equation for the phase derivative, requiring near perfect alignment of the metasurface The results show that when switching to y-polarised light, the location and morphology of cells become much clearer, which is in good agreement with DIC microscopy The recovered phase gradient image and the metasurface y-polarised image are in good agreement with the DIC image
Fabrication Robustness
The robustness of the design is tested by generating eroded and dilated variants of the structure and using a cost function that is a weighted average of these variants The simulation shows that erosion and dilation have a minimal effect on the brightfield imaging mode, but they can cause small blue-shifting or red-shifting in the linear region of the OTF The fabrication process is designed to be robust to under or over etched designs through edge detection techniques
Data Availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request The authors acknowledge support from the Australian Government through the Australian Research Council Centre of Excellence grant (CE200100010) and the Commonwealth through an Australian Government Research Training Program Scholarship The work was performed in part at the Melbourne Centre for Nanofabrication (MCN) in the Victorian Node of the Australian National Fabrication Facility (ANFF), and in part at the Materials Characterisation and Fabrication Platform (MCFP) at the University of Melbourne The authors would like to thank Dr Nitu Syed for performing the ellipsometry of the resist AZ5214E and Haiwei Wang for their useful discussions The paper is available on arXiv as 2606.21051v2 [physics. p. 0 au
1
arXiv:2606. 21051v2 [physics. au
1
arXiv:2606. 21051v2 [physics.Where the paper says it
Where the paper says it
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
-
Develop an ultra-compact real-time imaging system capable of performing multiple image processing modalities directly with light by switching between brightfield imaging and phase contrast via polarisation control. This eliminates the need to reposition optical components or use computational techniques to image transparent samples.
-
Implement quantitative analogue optical computation by using a metasurface where one polarisation state performs a first order derivative on the phase, producing pseudo-3D phase-contrast images of transparent biological samples while the orthogonal state returns the result of the identity operator.
-
Design and fabricate non-local, free-form dielectric metasurfaces using topology optimisation to achieve switchable phase contrast/brightfield imaging at normal incidence by exploiting diffraction effects for asymmetry generation.
-
Create a robust design methodology that incorporates robustness to fabrication defects by iteratively generating eroded and dilated variants of the metasurface during the topology optimisation process, minimizing the difference between design and fabrication outcomes.
-
Enable object recognition tasks by tailoring an asymmetric transfer function using topology optimisation to detect specific features in a sample, extending this approach to computational camera systems where the metasurface is adjacent to a camera sensor.
Abstract
Nonlocal metasurfaces enable all-optical processing of spatial information in optical fields. Here, we demonstrate a topology-optimised metasurface that switches between phase-contrast and brightfield imaging modalities via polarisation control, eliminating the need to reposition optical components or use computational techniques to image transparent samples. Specifically, for one polarisation state, an asymmetric transfer function about normal incidence performs a first order derivative on the phase, producing pseudo-3D phase-contrast images of transparent biological samples while the orthogonal state returns the result of the identity operator. This work extends inverse-design methods to reconfigurable phase-contrast microscopy and quantitative analogue optical computation with applications in biological imaging, medical diagnostics, and materials characterisation.
Related papers
- Depth-enhanced molecular imaging with two-photon oblique plane microscopy
- Attosecond soft X-ray pulses generated by chirp-dispersed manipulation in an XFEL reveal nonlinear core-electron dynamics in neon
- Raman optical activity and helicity-dependent phonon splitting in chiral CoSi
- Stability and optical quality of "windmill"-formed 8CB liquid crystal films for replenishable plasma mirrors
- Experimental Scaling of Diffraction Efficiency in Laser-Induced Plasma Gratings
- Photonic Exceptional Topological Bound States in the Continuum