Polarization-Multiplexed Spatial Filtering in a Symmetric Metasurface with an Anisotropic van der Waals Gap Fill

arXiv:2606.01983 · physics.optics · Submitted 2026-06-01 · 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: We're looking at the paper: "Polarization-Multiplexed Spatial Filtering in a Symmetric Metasurface with an Anisotropic van der Waals Gap Fill".

Mira: Here's the gist of the paper. Finite-difference time-domain simulations of a symmetric TiO2 nanobar-pair metasurface with a 60 nm α-MoO3 gap fill give two Fano resonances that are selected by the input polarization angle, at 885.5 nm (Q = 50.5) and 893.8 nm (Q = 44.8) Both channels act as spatial high-pass filters and differ in their cutoff spatial frequency, 0.44 and 1.17 µm−1.

Kai: Resonance Characteristics The simulation reveals two Fano resonances at different wavelengths with distinct linewidths The TE resonance is at 885.5 nm with Q = 50.5, and the TM resonance is at 893.8 nm with Q = 44.8, showing a separation of 8.3 nm.

Mira: These features are described as Fano resonances rather than quasi-BICs because the quality factors are well below the values that motivate the quasi-BIC description and a diagonal permittivity tensor preserves both mirror planes.

Kai: The linewidth of such a resonance is set by the coupling between a nominally dark mode and the radiation continuum.

Paper summary: Kai: So we're looking at this paper now, "Polarization-Multiplexed Spatial Filtering in a Symmetric Metasurface with an Anisotropic van der Waals Gap Fill". It basically claims they can use polarization to select different spatial filtering channels in this tiny structure.

Mira: Right. The main thing is that they have a symmetric titanium dioxide nanobar pair with a sixty nanometer gap filled with alpha molybdenum trioxide, and they found two Fano resonances selected by the input polarization angle at eight hundred eighty-five point five nanometers and eight hundred ninety-three point eight nanometers.

Lev: So, what does this mean for us in terms of what's actually been built? These are these Fano resonances, right? They aren't just some theoretical prediction; they are real features they simulated.

Kai: Yeah, the simulations show that these two channels act as spatial high-pass filters and they differ in their cutoff spatial frequency. One channel has a cutoff of zero point four four mu meter inverse and the other has a cutoff of one point one seven mu meter inverse.

Mira: And it’s interesting because the broader cutoff belongs to the lower quality factor resonance. So, so the TM resonance at eight hundred ninety-three point eight nanometers with a quality factor of forty-four point eight is broader than the TE resonance at eight hundred eighty-five point five nanometers with a quality factor of fifty point five.

Lev: The paper also breaks down how this separation between channels happens. It says the shape anisotropy of the bar pair sets the scale for that channel separation, and an out-of-plane index contrast from the crystal accounts for most of it.

Kai: That’s cool because we’re talking about geometry setting a fundamental limit on how well you can separate those modes before material properties even get involved too.

Mira: And the in-plane birefringence adds just a little bit to that separation, about zero point nine nanometer, but the shape anisotropy and the out-of-plane index contrast do most of the work.

Lev: From an engineering standpoint, if you were trying to build this on real hardware, you'd have to worry about those tolerances. They mention that a bar width of one hundred twenty nanometers is tight because a ten nanometer increase moves that TE resonance by nine point four nanometers.

Kai: That’s a pretty big sensitivity for the resonance position, so keeping the fabrication precise is definitely going to be hard if you want to hit those exact wavelengths.

Mira: And they also mentioned some experimental challenges. For instance, when it comes to power transmission in the TM channel versus the TE channel, it’s quite different; the TM channel transmits zero point one five of the incident power in its passband against zero point seven for the TE channel.

Lev: That difference in transmission is a real constraint on how much optical budget you have available from that specific polarization path when you try to implement this design.

Kai: So, if we're talking about practical realization, these transmission differences are something the authors flagged as a limitation in their simulation study.

Mira: Exactly. It’s not just about getting the resonance in the right place; it’s about how much light you can actually get through once you've got it there.

Lev: So, to wrap up this first part, this paper, "Polarization-Multiplexed Spatial Filtering in a Symmetric Metasurface with an Anisotropic van der Waals Gap Fill", shows two Fano resonances at eight hundred eighty-five point five nanometers and eight hundred ninety-three point eight nanometers with specific quality factors and spatial cutoff frequencies.

Kai: And the main implication is that these two channels offer distinct spatial filtering capabilities, which is relevant for things like image processing where you want to separate different features spatially based on polarization.

Mira: It’s about achieving this polarization-multiplexed spatial filtering using a symmetric titanium dioxide nanobar pair with an anisotropic van der Waals gap fill.

Lev: So, the final part of this discussion is how these results change things for someone who just listens to the show. This work shows a specific way to use polarization multiplexing in metasurfaces for analog computing operations, which is important for understanding how we might build hardware that processes information spatially.

Conclusion: Kai: So, this paper is about taking polarization to control spatial filtering in a symmetric structure made of titanium dioxide nanobars with an anisotropic molybdenum trioxide gap fill and they found two Fano resonances at eight hundred eighty-five point five nanometers and eight hundred ninety-three point eight nanometers.

Mira: Right. The title itself points to the core idea, which is using polarization to multiplex spatial filtering. It sounds like something that could be useful for separating different things in an image or a signal spatially.

Lev: From what I'm seeing, it’s about using the shape of the structure and how light enters it—the polarization—to choose which spatial frequency gets through. That’s a very specific way to do filtering.

Kai: Exactly. It’s not just one filter anymore; you have two distinct channels, each with its own cutoff frequency, meaning you can filter for two different spatial scales at once by choosing the input polarization.

Mira: And I think it's interesting that they didn't call these quasi-BIC modes because the quality factors are low enough that they fit better as standard Fano resonances. That means the coupling to light is strong enough to make them visible features, not just theoretical traps.

Lev: For me, the real question is how much overhead this kind of multiplexing would add if we tried to build this on actual hardware for error correction; the geometry separation they found—the shape anisotropy—that sets a real constraint on how close those two channels can actually be before they merge.

Kai: That's a good point about the geometry setting that scale. And remember, they pointed out that while both channels act as high-pass filters, the TM channel has a much lower transmission in its passband, only zero point one five against zero point seven for the TE channel.

Mira: That transmission difference is crucial because it tells us which polarization path you have more available light from when you try to implement this design; it’s a real practical limitation on the optical budget.

Lev: So, if we look at the authors and what they did, S. Debnath handled the whole concept and writing while S. Saha focused on validation and resources, which is standard for getting a solid experimental result from simulations.

Kai: It sounds like they've mapped out a clear path from the theoretical setup to seeing how those specific spatial filtering effects manifest in the simulation data for this nanobar pair.

Mira: And looking at it broadly, this work shows a pathway toward using polarization multiplexing in these metasurfaces for analog computing operations, which is relevant for how we might build hardware that processes information spatially.

Lev: So, the next thing we need to figure out is whether those specific spatial filtering capabilities translate into something usable when you start talking about real-world data streams and processing.

Shoumik Debnath, Sudipta Saha

Bangladesh University of Engineering and Technology

physics.optics

Submitted: 2026-06-01

Updated: 2026-09-14

Journal ref: RSC Advances (2026)

DOI: 10.1039/d6ra07836b

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

Importance score: 82/100

The gist: Finite-difference time-domain simulations of a symmetric TiO2 nanobar-pair metasurface with a 60 nm α-MoO3 gap fill give two Fano resonances that are selected by the input polarization angle…

Key concepts

Fano Resonance
A resonance phenomenon where a discrete energy level interacts with a continuum, leading to an asymmetric spectral line shape. In this metasurface, it occurs at specific wavelengths (885.5 nm and 893.8 nm) and is used to create sharp filtering effects in the structure.
Spatial High-Pass Filter
A type of filter that allows higher spatial frequencies (finer details) to pass while blocking lower frequencies (broader features). Both polarization channels in the simulation function this way, with different cutoff spatial frequencies determining which features are transmitted.
Channel Separation Mechanism
The distinct behavior between the two polarization channels is caused by geometric and material differences. The primary separation comes from the shape anisotropy of the bar pair, supplemented by out-of-plane index contrast and in-plane birefringence, which collectively determine how far apart the resonance modes are spatially.
Cutoff Spatial Frequency
This value defines the boundary between what spatial features can pass through a specific polarization channel. The TE channel has a lower cutoff (0.436 $\mu$m$^{-1}$), meaning it filters out coarser spatial variations compared to the TM channel's broader cutoff (1.165 $\mu$m$^{-1}$).

Terminology

Summary

Finite-difference time-domain simulations of a symmetric TiO2 nanobar-pair metasurface with a 60 nm α-MoO3 gap fill give two Fano resonances that are selected by the input polarization angle, at 885. p. 0

Where the paper says it

JournalName Polarization-multiplexed spatial filtering in a symmetric TiO2 metasurface with an anisotropic van der Waals gap fill† Shoumik Debnath a and Sudipta Saha ∗a,b Finite-difference time-domain simulations of a symmetric TiO2 nanobar-pair metasurface with a 60 nm α-MoO3 gap fill give two Fano resonances that are selected by the input polarization angle, at 885. 5 nm (Q = 50.

5 nm (Q = 50.5) and 893. p. 0
Where the paper says it

5 nm (Q = 50. 5) and 893.

8 nm (Q = 44.8) Both channels act as spatial high-pass filters and differ in their cutoff spatial frequency, 0. p. 0
Where the paper says it

8 nm (Q = 44. Both channels act as spatial high-pass filters and differ in their cutoff spatial frequency, 0.

p. 0
Where the paper says it

8 nm (Q = 44. Both channels act as spatial high-pass filters and differ in their cutoff spatial frequency, 0.

44 and 1.17 µm−1. p. 0
Where the paper says it

Both channels act as spatial high-pass filters and differ in their cutoff spatial frequency, 0. 17 µm−1 , with the broader cutoff belonging to the lower-Q resonance.

Resonance Characteristics

The simulation reveals two Fano resonances at different wavelengths with distinct linewidths The TE resonance is at 885.5 nm with Q = 50.5, and the TM resonance is at 893. p. 1

Where the paper says it

5 nm with Q = 50. 5, and the TM resonance at 893.

8 nm with Q = 44.8, showing a separation of 8. p. 0
Where the paper says it

The simulations use measured anisotropic optical constants for α-MoO3 with the crystal axes assigned as they lie in an exfoliated flake. A control series with isotropic and uniaxial gap fillings indicates that the shape anisotropy of the bar pair sets the scale of the channel separation, that the anisotropy of the crystal acts mainly through its out-of-plane index contrast, and that the in-plane birefringence contributes a small part of it.

3 nm. These features are described as Fano resonances rather than quasi-BICs because the quality factors are well below the values that motivate the quasi-BIC description and a diagonal permittivity tensor preserves both mirror planes. p. 2
Where the paper says it

A diagonal permittivity tensor aligned with the structural axes preserves both mirror planes of the nanobar pair, so filling the gap does not break a protecting symmetry, and the mechanism that produces a symmetry-protected quasi-BIC is therefore absent. The quality factors, 50.

The linewidth of such a resonance is set by the coupling between a nominally dark mode and the radiation continuum. p. 0
Where the paper says it

8–15 Many dielectric analog-computing platforms use Fano resonances, and in the high-Q limit quasi-bound states in the continuum (quasi-BICs), to produce sharp spectral features. 16–20 The linewidth of such a resonance is set by the coupling between a nominally dark mode and the radiation continuum.

Channel Separation Mechanism

The separation between the two channels is determined by several factors related to geometry and material anisotropy

  1. The shape anisotropy of the bar pair sets the scale of the channel separation, as an empty cell already separates the two channels by 14. p. 8

    Where the paper says it

    Channel selection is by input polarization angle, and the results support the use of a single symmetric cell for two spatial filtering operations of different bandwidth. The shape anisotropy of the bar pair sets the scale of the channel separation: the empty cell already separates the two channels by 14.

    4 nm.

  2. The out-of-plane index contrast of the crystal accounts for most of the anisotropic contribution, contributing 7. p. 5

    Where the paper says it

    2 separates the contributions to the channel separation. The shape anisotropy of the bar pair sets its scale, the out-of-plane index contrast of the crystal accounts for most of the anisotropic contribution, and the in-plane birefringence for about 0.

    8 nm to the separation. p. 0
    Where the paper says it

    The simulations use measured anisotropic optical constants for α-MoO3 with the crystal axes assigned as they lie in an exfoliated flake. A control series with isotropic and uniaxial gap fillings indicates that the shape anisotropy of the bar pair sets the scale of the channel separation, that the anisotropy of the crystal acts mainly through its out-of-plane index contrast, and that the in-plane birefringence contributes a small part of it.

  3. The in-plane birefringence contributes a small part, accounting for about 0. p. 0

    Where the paper says it

    The simulations use measured anisotropic optical constants for α-MoO3 with the crystal axes assigned as they lie in an exfoliated flake. A control series with isotropic and uniaxial gap fillings indicates that the shape anisotropy of the bar pair sets the scale of the channel separation, that the anisotropy of the crystal acts mainly through its out-of-plane index contrast, and that the in-plane birefringence contributes a small part of it.

    9 nm.

Spatial Filtering Operation

Both polarization channels function as spatial high-pass filters, with the broader cutoff belonging to the lower-Q resonance. The TE channel has a cutoff of 0. p. 0

Where the paper says it

17 µm−1 , with the broader cutoff belonging to the lower-Q resonance. The unit cell retains a mirror plane normal to the swept direction, so the momentum-space transfer function is even in kx; over the accessible momentum range the broader channel follows |H| ∝ |kx|, which is the kernel used for edge enhancement.

436 µm−1, while the TM channel has a broader cutoff of 1. p. 0
Where the paper says it

17 µm−1 , with the broader cutoff belonging to the lower-Q resonance. The unit cell retains a mirror plane normal to the swept direction, so the momentum-space transfer function is even in kx; over the accessible momentum range the broader channel follows |H| ∝ |kx|, which is the kernel used for edge enhancement.

165 µm−1. p. 1
Where the paper says it

8 Cutoff κ (µm−1 ) 0. 165 Passband transmittance 0.

p. 1
Where the paper says it

8 Cutoff κ (µm−1 ) 0. 165 Passband transmittance 0.

Over the accessible momentum range, both channels follow H ∝ kx, which is the kernel used for edge enhancement. p. 0
Where the paper says it

17 µm−1 , with the broader cutoff belonging to the lower-Q resonance. The unit cell retains a mirror plane normal to the swept direction, so the momentum-space transfer function is even in kx; over the accessible momentum range the broader channel follows |H| ∝ |kx|, which is the kernel used for edge enhancement.

Experimental Challenges and Benchmarks

Practical realization faces difficulties regarding gap filling and crystal orientation The TM channel transmits 0. p. 6

Where the paper says it

208 at the TE operating wavelength. In the passband the TE channel transmits a power fraction of 0.

15 of the incident power in its passband against 0. p. 8
Where the paper says it

The TM channel transmits 0. 15 of the incident power in its passband against 0.

p. 8
Where the paper says it

The TM channel transmits 0. 15 of the incident power in its passband against 0.

70 for the TE channel, which limits the optical budget available from that channel. p. 8
Where the paper says it

15 of the incident power in its passband against 0. 70 for the TE channel, which limits the optical budget available from that channel.

The bar width is identified as the tighter of two lithographic tolerances because a 10 nm increase moves the TE resonance by 9. p. 6
Where the paper says it

A 10 nm increase in bar width moves the TE resonance by 9. 4 nm, so bar width is the tighter of the two lithographic tolerances.

4 nm.

Conclusion

A symmetric TiO2 nanobar-pair metasurface with a 60 nm αMoO3 gap fill supports two Fano resonances, at 885.5 nm with Q = 50.5 and at 893. p. 0

Where the paper says it

5 nm (Q = 50. 5) and 893.

8 nm with Q = 44.8. Both channels act as spatial high-pass filters and differ in their cutoff spatial frequency, 0.436 and 1.165 µm−1. The shape anisotropy of the bar pair sets the scale of the channel separation. p. 0
Where the paper says it

The simulations use measured anisotropic optical constants for α-MoO3 with the crystal axes assigned as they lie in an exfoliated flake. A control series with isotropic and uniaxial gap fillings indicates that the shape anisotropy of the bar pair sets the scale of the channel separation, that the anisotropy of the crystal acts mainly through its out-of-plane index contrast, and that the in-plane birefringence contributes a small part of it.

The TM channel transmits 0.15 of the incident power in its passband against 0.70 for the TE channel. p. 0
Where the paper says it

17 µm−1 , with the broader cutoff belonging to the lower-Q resonance. The unit cell retains a mirror plane normal to the swept direction, so the momentum-space transfer function is even in kx; over the accessible momentum range the broader channel follows |H| ∝ |kx|, which is the kernel used for edge enhancement.

The paper is a simulation study examining how polarization-multiplexed spatial filtering can be achieved in a symmetric TiO2 metasurface with an anisotropic van der Waals gap fill, which is relevant for analog optical computing operations like image processing. p. 0

Where the paper says it

JournalName Polarization-multiplexed spatial filtering in a symmetric TiO2 metasurface with an anisotropic van der Waals gap fill† Shoumik Debnath a and Sudipta Saha ∗a,b Finite-difference time-domain simulations of a symmetric TiO2 nanobar-pair metasurface with a 60 nm α-MoO3 gap fill give two Fano resonances that are selected by the input polarization angle, at 885. 5 nm (Q = 50.

Author Contributions

S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing. p. 8

Where the paper says it

Author contributions S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing.

p. 8
Where the paper says it

Author contributions S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing.

S. Saha: conceptualisation, methodology, validation, supervision, resources, writing – review and editing. p. 8
Where the paper says it

Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing. Saha: conceptualisation, methodology, validation, supervision, resources, writing – review and editing.

Conflicts of Interest

There are no conflicts to declare.

Data Availability

The data supporting this article are included in the manuscript and in the Supplementary Information. The unit-cell geometry, the material tensor, the mesh specification and the convergence, control, crosstalk, tolerance and fit data are given in Section 2 and in Tables S1 to S5. p. 8

Where the paper says it

Data availability The data supporting this article are included in the manuscript and in the Supplementary Information. The unit-cell geometry, the material tensor, the mesh specification and the convergence, control, crosstalk, tolerance and fit data are given in Section 2 and in Tables S1 to S5.

The measured optical constants of α-MoO3 are those published by Toksumakov et al. p. 8
Where the paper says it

The unit-cell geometry, the material tensor, the mesh specification and the convergence, control, crosstalk, tolerance and fit data are given in Section 2 and in Tables S1 to S5. The measured optical constants of α-MoO3 are those published by Toksumakov et al.

31.

Acknowledgements

The authors thank the Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology for computational resources. The authors thank the anonymous referees for their constructive comments. p. 8

Where the paper says it

31 Acknowledgements The authors thank the Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology for computational resources. The authors thank the anonymous referees for their constructive comments.

Notes and References

The references are listed in the bottom section of the paper, such as Reference 1 A. Silva, F. Monticone, G. p. 8

Where the paper says it

Notes and references 1 A. Monticone, G.

p. 8
Where the paper says it

Notes and references 1 A. Monticone, G.

p. 8
Where the paper says it

Notes and references 1 A. Monticone, G.

p. 8
Where the paper says it

Notes and references 1 A. Monticone, G.

Castaldi, V. p. 8
Where the paper says it

Monticone, G. Castaldi, V.

p. 8
Where the paper says it

Monticone, G. Castaldi, V.

p. 8
Where the paper says it

Monticone, G. Castaldi, V.

p. 8
Where the paper says it

Monticone, G. Castaldi, V.

Galdi, A. Alù and N. p. 8
Where the paper says it

Castaldi, V. Alù and N.

p. 8
Where the paper says it

Castaldi, V. Alù and N.

p. 8
Where the paper says it

Castaldi, V. Alù and N.

p. 8
Where the paper says it

Castaldi, V. Alù and N.

Engheta, Science, 2014. p. 8
Where the paper says it

Alù and N. Engheta, Science, 2014, 343, 160–163.

The references cover topics from plasmonic structures to dielectric metasurfaces. p. 0
Where the paper says it

3–7 Early work used plasmonic structures and photonic crystal slabs. 6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty.

p. 0
Where the paper says it

3–7 Early work used plasmonic structures and photonic crystal slabs. 6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty.

p. 0
Where the paper says it

3–7 Early work used plasmonic structures and photonic crystal slabs. 6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty.

p. 0
Where the paper says it

3–7 Early work used plasmonic structures and photonic crystal slabs. 6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty.

The work builds upon prior studies on Fano resonances and quasi-BICs in analog computing platforms. p. 0
Where the paper says it

6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty. 8–15 Many dielectric analog-computing platforms use Fano resonances, and in the high-Q limit quasi-bound states in the continuum (quasi-BICs), to produce sharp spectral features.

p. 0
Where the paper says it

6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty. 8–15 Many dielectric analog-computing platforms use Fano resonances, and in the high-Q limit quasi-bound states in the continuum (quasi-BICs), to produce sharp spectral features.

p. 0
Where the paper says it

6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty. 8–15 Many dielectric analog-computing platforms use Fano resonances, and in the high-Q limit quasi-bound states in the continuum (quasi-BICs), to produce sharp spectral features.

p. 0
Where the paper says it

6,7 Dielectric metasurfaces have since become the more common platform, since their low absorption and high transmission allow strong modulation of the transfer function without a large throughput penalty. 8–15 Many dielectric analog-computing platforms use Fano resonances, and in the high-Q limit quasi-bound states in the continuum (quasi-BICs), to produce sharp spectral features.

References

The references are listed at the end of the paper, such as Reference 1 A. Silva, F. Monticone, G. Castaldi, V. Galdi, A. Alù and N. Engheta, Science. p. 8

Where the paper says it

Alù and N. Engheta, Science, 2014, 343, 160–163.

The references cover topics from plasmonic structures to dielectric metasurfaces. The work builds upon prior studies on Fano resonances and quasi-BICs in analog computing platforms. The work is compared with related work in Table 2. p. 6
Where the paper says it

3, so they should be read as shifts relative to their own nominal row rather than as absolute values. 5 Comparison with related work Table 2 places this work among selected multi-channel analog computing demonstrations.

S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing. S. Saha: conceptualisation, methodology, validation, supervision, resources. p. 8

Where the paper says it

Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing. Saha: conceptualisation, methodology, validation, supervision, resources, writing – review and editing.

The references are listed in the bottom section of the paper, such as Reference 1 A. Silva, F. Monticone, G. Castaldi, V. Galdi, A. Alù and N. Engheta. The references cover topics from plasmonic structures to dielectric metasurfaces. The work builds upon prior studies on Fano resonances and quasi-BICs in analog computing platforms.

The references are listed at the end of the paper, such as Reference 1 A. Silva, F. Monticone, G. Castaldi, V. Galdi, A. Alù and N. Engheta. The references cover topics from plasmonic structures to dielectric metasurfaces. The work builds upon prior studies on Fano resonances and quasi-BICs in analog computing platforms.

S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft. p. 8

Where the paper says it

Author contributions S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing.

S. Saha: conceptualisation, methodology, validation. p. 8
Where the paper says it

Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing. Saha: conceptualisation, methodology, validation, supervision, resources, writing – review and editing.

p. 8
Where the paper says it

Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing. Saha: conceptualisation, methodology, validation, supervision, resources, writing – review and editing.

p. 8
Where the paper says it

Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing. Saha: conceptualisation, methodology, validation, supervision, resources, writing – review and editing.

S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation. p. 8

Where the paper says it

Author contributions S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing.

S. Saha: conceptualisation, methodology, validation.

S. Debnath: conceptualisation, methodology, software, formal analysis, investigation. p. 8

Where the paper says it

Author contributions S. Debnath: conceptualisation, methodology, software, formal analysis, investigation, visualisation, writing – original draft, writing – review and editing.

S. Saha: conceptualisation, methodology, validation.

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. Spatial High-Pass Filtering Operation: The AI system can perform spatial differentiation by applying a transfer function proportional to ikx, but it must be noted that an even transfer function cannot take that form, meaning the system can only implement an edge-enhancement kernel with |H| proportional to |kx| rather than a true first-order spatial differentiator.

  2. Polarization Multiplexing Control: The AI system can select between two distinct spatial filtering operations by controlling the input polarization angle, as Channel selection is by input polarization angle. This allows for two different high-pass filters with different cutoffs, specifically 0.436 µm-1 for the TE channel and 1.165 µm-1 for the TM channel.

  3. Channel Separation Scaling: The system can be optimized to understand how shape anisotropy sets the scale of separation, as The shape anisotropy of the bar pair sets its scale of the channel separation. This knowledge allows for predicting how changes in geometry affect channel separation based on the empty cell already separates the two channels by 14.4 nm.

  4. Material Anisotropy Utilization: The AI system can leverage anisotropic material properties to modify optical response without changing geometry, specifically by using a diagonal anisotropic medium with εxx = n2a, εyy = n2b and εzz = n2c to tune the channel separation based on the crystal's in-plane birefringence.

Abstract

Finite-difference time-domain simulations of a symmetric nanobar-pair metasurface with a 60 nm α - gap fill give two Fano resonances that are selected by the input polarization angle, at 885.5 nm (Q=50.5) and 893.8 nm (Q=44.8). Both channels act as spatial high-pass filters and differ in their cutoff spatial frequency, 0.44 and 1.17 μ m-1, with the broader cutoff belonging to the lower- Q resonance. The unit cell retains a mirror plane normal to the swept direction, so the momentum-space transfer function is even in k x; over the accessible momentum range the broader channel follows H proportional tok x, which is the kernel used for edge enhancement. Both operations are applied to a USAF 1951 resolution chart in a simulated 4 f arrangement. The simulations use measured anisotropic optical constants for α - with the crystal axes assigned as they lie in an exfoliated flake. A control series with isotropic and uniaxial gap fillings indicates that the shape anisotropy of the bar pair sets the scale of the channel separation, that the anisotropy of the crystal acts mainly through its out-of-plane index contrast, and that the in-plane birefringence contributes a small part of it.

Related papers