Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 45 of the 47 references it cites were found in a registry or an index; 12 of the 23 sources its claims cite were checked for support.

Shape optimization for high efficiency metasurfaces: theory and implementation

Open access
Dainese, Paulo, Marra, Louis, Cassara, Davide, Portes, Ary, Oh, Jaewon, Yang, Jun, Palmieri, Alfonso, Rodrigues, Janderson Rocha, Dorrah, Ahmed H., Capasso, Federico · 2024 · Light: Science & Applications

Abstract

Complex non-local behavior makes designing high efficiency and multifunctional metasurfaces a significant challenge. While using libraries of meta-atoms provide a simple and fast implementation methodology, pillar to pillar interaction often imposes performance limitations. On the other extreme, inverse design based on topology optimization leverages non-local coupling to achieve high efficiency, but leads to complex and difficult to fabricate structures. In this paper, we demonstrate numerically and experimentally a shape optimization method that enables high efficiency metasurfaces while providing direct control of the structure complexity through a Fourier decomposition of the surface gradient. The proposed method provides a path towards manufacturability of inverse-designed high efficiency metasurfaces.

Overview

CoChat Summary

## TL;DR

This paper presents a shape-optimization framework for metasurfaces that keeps the benefits of inverse design while giving direct control over fabrication complexity via Fourier decomposition of boundary gradients. The authors numerically and experimentally show that it can raise efficiencies of library-based metasurfaces while preserving smooth, manufacturable shapes.

## Key findings

  • For a resonant 51° metagrating, an initial library design had only about 30% TE efficiency and 85% TM efficiency; shape optimization improved this to 82% TE and 84% TM with higher Fourier orders, while the simpler m = 0 case improved TE but could not recover TM as well.
  • For a non-resonant 51° metagrating, the initial library design had 76% TE and 88% TM efficiency; shape optimization improved this to 85% TE and 90% TM.
  • For metalenses at 1.55 μm, shape optimization improved initial library designs to 88% and 90% efficiency for the two starting geometries mentioned, with smooth boundaries and fabrication constraints respected.
  • A fabricated 70° metagrating showed measured peak efficiency close to 70%, and the shape-optimized m = 6 designs had substantially higher efficiencies than the library for both TE and TM.

## Methods

  • The method is an adjoint-based shape optimization approach: each iteration uses a forward and adjoint electromagnetic simulation to compute shape gradients for the meta-atoms, then updates boundaries; the gradients are decomposed in a Fourier basis before updating.
  • Numerical examples used 1 μm tall amorphous silicon pillars on glass, typically with 1.55 μm operation, 90 nm minimum gap, and 80 nm minimum feature size; metagratings were optimized for deflection angles from 10° to 70°, and metalenses were also demonstrated.
  • Experimental validation used conventional top-down fabrication on a 500 μm fused silica wafer with PECVD a-Si, e-beam lithography, and RIE; characterization measured absolute diffraction efficiencies for TE and TM with a tunable laser.

## Limitations

  • The paper states that adjoint shape optimization is computationally expensive because it requires solving Maxwell’s equations for both forward and adjoint fields.
  • It also states that the method converges to a local optimum, so different initial conditions can lead to different final designs.
  • Evident from the text: the method is shown on specific metasurface examples (metagratings and metalenses) and does not demonstrate a broader class of devices beyond what is described here.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41377-024-01629-5

Cites

45 of 46 references verified0 citation errors · 1 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Metasurface design methodologies. In the library method (a), the phase and amplitude response of pre-determined meta-atoms is calculated as a function of its parameters (e.g. diameter for a circular pillar). The response is then used to create devices by dividing the surface in unit cells, each containing a specific element of the library. In topology optimization (b), the design domain is typically initialized with a randomized refractive index distribution, and the adjoint algorithm is used to converge towards a final binarized geometry. In shape optimization (c), only smooth variations in existing geometrical features are allowed
Fig. 1 Metasurface design methodologies. In the library method (a), the phase and amplitude response of pre-determined meta-atoms is calculated as a function of its parameters (e.g. diameter for a circular pillar). The response is then used to create devices by dividing the surface in unit cells, each containing a specific element of the library. In topology optimization (b), the design domain is typically initialized with a randomized refractive index distribution, and the adjoint algorithm is used to converge towards a final binarized geometry. In shape optimization (c), only smooth variations in existing geometrical features are allowedCC BY · PubMed Central
Fig. 2 Flow chart of the optimization process. The design domain is initialized with a given set of meta-atoms, for example with a library-based metasurface. Forward and adjoint simulations are performed and the figure of merit computed. If the figure of merit satisfies the target or if it has converged the optimization stops. If not, the shifting boundary gradients are calculated from the simulated fields, followed by Fourier decomposition and possibly additional fabrication constraints. The shape is then deformed according to the gradients and a new simulation is performed. The process repeats until convergence
Fig. 2 Flow chart of the optimization process. The design domain is initialized with a given set of meta-atoms, for example with a library-based metasurface. Forward and adjoint simulations are performed and the figure of merit computed. If the figure of merit satisfies the target or if it has converged the optimization stops. If not, the shifting boundary gradients are calculated from the simulated fields, followed by Fourier decomposition and possibly additional fabrication constraints. The shape is then deformed according to the gradients and a new simulation is performed. The process repeats until convergenceCC BY · PubMed Central
Fig. 3 Details of the shape optimization algorithm. The forward and adjoint fields are computed in the design domain, and are then used to calculate a gradient function g on the surface of all existing meta-atoms. a, b Show the fields and the derived boundary gradients for TM and TE polarizations. The gradient function determines the magnitude and direction that each point on the surface must be displaced to increase the device efficiency. In (c), the resulting shape gradient is plotted (blue) along with its Fourier decomposition (black) for the pillar highlighted in (b). One can observe that the gradient containts significant contributions up to order m = 6
Fig. 3 Details of the shape optimization algorithm. The forward and adjoint fields are computed in the design domain, and are then used to calculate a gradient function g on the surface of all existing meta-atoms. a, b Show the fields and the derived boundary gradients for TM and TE polarizations. The gradient function determines the magnitude and direction that each point on the surface must be displaced to increase the device efficiency. In (c), the resulting shape gradient is plotted (blue) along with its Fourier decomposition (black) for the pillar highlighted in (b). One can observe that the gradient containts significant contributions up to order m = 6CC BY · PubMed Central
Fig. 4 Optimization of a dual-polarization 51-degree beam deflector. Shape optimization was applied to two library designs differing only by the choice of the first pillar diameter: in (a), the initial design exhibits low efficiency due to the appearance of a resonance for TE polarization while in (b) the initial structure exhibited the highest efficiency possible for such library. The curves show the evolution of the absolute efficiency at each iteration, where dashed curves represent restricting the gradient function to m = 0 Fourier order (i.e. maintaining circular shape), while solid line allows up to m = 6. The final structures for both Fourier orders are shown in the insets. In (c), shape optimization was applied to design meta-gratings from 0 to 70 degrees (final geometries shown as insets). The color legend on top of the figure applies to all plots
Fig. 4 Optimization of a dual-polarization 51-degree beam deflector. Shape optimization was applied to two library designs differing only by the choice of the first pillar diameter: in (a), the initial design exhibits low efficiency due to the appearance of a resonance for TE polarization while in (b) the initial structure exhibited the highest efficiency possible for such library. The curves show the evolution of the absolute efficiency at each iteration, where dashed curves represent restricting the gradient function to m = 0 Fourier order (i.e. maintaining circular shape), while solid line allows up to m = 6. The final structures for both Fourier orders are shown in the insets. In (c), shape optimization was applied to design meta-gratings from 0 to 70 degrees (final geometries shown as insets). The color legend on top of the figure applies to all plotsCC BY · PubMed Central
Fig. 5 Shape optimization of a high NA metalens. Evolution of the absolute efficiency for two different initial structures is shown in (a); cuts of the z-component of the Poynting vector in the xz plane in (b) and in the xy plane in (c). The color scale shown in (c) also applies for (b). The full metalens structure is shown in (d), and a zoomed region at the metalens center before and after shape optimization is shown in (e)
Fig. 5 Shape optimization of a high NA metalens. Evolution of the absolute efficiency for two different initial structures is shown in (a); cuts of the z-component of the Poynting vector in the xz plane in (b) and in the xy plane in (c). The color scale shown in (c) also applies for (b). The full metalens structure is shown in (d), and a zoomed region at the metalens center before and after shape optimization is shown in (e)CC BY · PubMed Central
Fig. 6 Comparison between simulation and measurement for the resonant 51-degree beam deflector. Simulated (a) and measured (b) TE spectral responses. Different colors show the first order diffraction efficiency for the initial and shape optimized designs with Fourier order m = 0 and m = 6 (the color legend applies to both (a) and (b) plots). A scanning electron microscope image of the fabricated structures for m = 6 is shown in the inset (scale bar is 200 nm). In (c), the efficiency at the design wavelength is plotted for TE and TM polarizations
Fig. 6 Comparison between simulation and measurement for the resonant 51-degree beam deflector. Simulated (a) and measured (b) TE spectral responses. Different colors show the first order diffraction efficiency for the initial and shape optimized designs with Fourier order m = 0 and m = 6 (the color legend applies to both (a) and (b) plots). A scanning electron microscope image of the fabricated structures for m = 6 is shown in the inset (scale bar is 200 nm). In (c), the efficiency at the design wavelength is plotted for TE and TM polarizationsCC BY · PubMed Central
Fig. 7 Comparison between simulation and measurement for the non-resonant 51-degree beam deflector. Simulated (a) and measured (b) TE spectral responses. Different colors show the first order diffraction efficiency for the initial and shape optimized designs with Fourier order m = 0 and m = 6 (the color legend applies to both (a) and (b) plots). A scanning electron microscope image of the fabricated structures for m = 6 is shown in the inset (scale bar is 300 nm). In (c), the efficiency at the design wavelength is plotted for TE and TM polarizations
Fig. 7 Comparison between simulation and measurement for the non-resonant 51-degree beam deflector. Simulated (a) and measured (b) TE spectral responses. Different colors show the first order diffraction efficiency for the initial and shape optimized designs with Fourier order m = 0 and m = 6 (the color legend applies to both (a) and (b) plots). A scanning electron microscope image of the fabricated structures for m = 6 is shown in the inset (scale bar is 300 nm). In (c), the efficiency at the design wavelength is plotted for TE and TM polarizationsCC BY · PubMed Central
Fig. 8 Comparison between simulation and measurement for the broadband 70-degree beam deflector. Simulated (a) and measured (b) spectral responses for both TE and TM polarizations. Different colors show the first order diffraction efficiency for the initial and shape optimized designs with Fourier order m = 6 (the color legend applies to both (a) and (b) plots). A scanning electron microscope image of the fabricated structure is shown in the inset (scale bar is 200 nm). In (c), the efficiency at the design wavelength is plotted for TE and TM polarizations
Fig. 8 Comparison between simulation and measurement for the broadband 70-degree beam deflector. Simulated (a) and measured (b) spectral responses for both TE and TM polarizations. Different colors show the first order diffraction efficiency for the initial and shape optimized designs with Fourier order m = 6 (the color legend applies to both (a) and (b) plots). A scanning electron microscope image of the fabricated structure is shown in the inset (scale bar is 200 nm). In (c), the efficiency at the design wavelength is plotted for TE and TM polarizationsCC BY · PubMed Central

Verification

Claims the paper makes

6 Valid · 0 Not valid · 4 Not checked · 32 Own results and interpretations

12 more claims

Create a free account to see all 42 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 21
  • Probably not a substantial claim. 8
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

1 of 1

Citations

Cites

Show all 47 references

Research works printed but not linked: 1

  1. Landau, L. D. et al. Electrodynamics of continuous media, 8 (Elsevier, 2013).

Cited by

See all 75

Metadata

Authors
Dainese, P., Marra, L., Cassara, D., Portes, A., Oh, J., Yang, J., Palmieri, A., Rodrigues, J. R., Dorrah, A. H., & Capasso, F.
Affiliations
  • Corning (United States) (US)
  • Harvard University (US)
Venue
Light: Science & Applications
Published
2024-10-29
Volume
13
Issue
1
Pages
300-300
ISSN
2047-7538 (ISSN-L)
OpenAlex
W4403882296
Cited by
75
References
47
Citations by year
2026: 38 · 2025: 32 · 2024: 5
FWCI
5.84
Type
journal article
Language
en
Open access
Licence
Locations
  • Light Science & Applications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • Figshare (submittedVersion, cc-by)
  • arXiv (Cornell University) (cc-by)
Topics
  • Physical Sciences › Materials Science › Electronic, Optical and Magnetic Materials › Metamaterials and Metasurfaces Applications
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Antenna and Metasurface Technologies
  • Physical Sciences › Physics and Astronomy › Acoustics and Ultrasonics › Acoustic Wave Phenomena Research
Keywords
metasurfaces, shape optimization, inverse design, topology optimization, meta-atoms, nonlocal coupling, Fourier decomposition, surface gradient, manufacturability

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Shape optimization for high efficiency metasurfaces: theory and implementation. https://app.cochat.ai/papers/10.1038/s41377-024-01629-5
The paper
Dainese, P., Marra, L., Cassara, D., Portes, A., Oh, J., Yang, J., Palmieri, A., Rodrigues, J. R., Dorrah, A. H., & Capasso, F. (2024). Shape optimization for high efficiency metasurfaces: theory and implementation. Light: Science & Applications. https://doi.org/10.1038/s41377-024-01629-5

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.