nature photonics

Review Article https://doi.org/10.1038/s41566-022-01126-4

Enabling smart vision with metasurfaces

Received: 8 August 2022 Dragomir N. Neshev 1

& Andrey E. Miroshnichenko 2

Accepted: 7 November 2022

Published online: 22 December 2022 Optical metasurfaces—subwavelength-patterned surfaces that interact

strongly with light—have been an active area of research for more than a
Check for updates
decade. The field has been driven by the key advantages of this concept,
including the ultimate miniaturization of optical elements, empowering
novel functionalities that process hidden modalities of light, and the
opportunity to tune their properties on demand. A large number of
applications with a focus on smart vision have emerged, foreseeing a
meta-optical device under the hood of any robotic system. Nowadays, the
field is experiencing a solid industry pull that defines the challenges and
research directions. This review overviews the application focus of the field
defined by the growing number of optical functionalities. We describe the
current challenges and outline the research frontiers in the field.

Light carries an enormous amount of information that enables by engineering—with unprecedented spatial resolution—the light
humans and autonomous machines to see and understand the world. wavefront, emission and absorption.
However, this information is often hidden from our eyes, which only The field of meta-optics has emerged from pioneering works on
react to light intensity and colours in the visible spectral range. The wavefront and polarization control1,2. The name ‘meta-optics’ was
hidden attributes of light, including phase, polarization or angular coined by Ross McPhedran and colleagues in 20113 and has found wide
momentum, can be extracted by smart optical devices that enable acceptance worldwide. Over the past decade, the field has grown expo-
machine vision to surpass conventional vision. Novel smart optical nentially, with over 160 groups working in the field, including major
systems can also allow us to see such hidden information through research centres worldwide. We are witnessing thousands of research
human–machine interfaces, such as virtual and augmented reality publications and industrial patents every year. Although this growth
(VR/AR) displays, wearable sensors and artificial intelligence in the reflects the importance of meta-optical development, it also presents
photonic domain. But what is under the hood of a robotic system challenges where novel and disruptive ideas cannot easily be noticed4.
that can enable superior vision across multiple spectral bands? This Review aims to map the active research areas in the field over the
Conventional optical technologies are too bulky to allow advanced past five years, identify key concepts, discuss the existing challenges
functionalities in a small device. For example, how many more lenses and provide a vision for the impact of the technology.
can one fit on a mobile phone? Such miniaturized advanced-vision By mapping the activities in the field, we have identified three main
technologies require radically new approaches for smart optical stages of importance. These are presented as an artist-impression map
systems. The concept of meta-optics might provide the awaited in Fig. 1. At the focal spot of this map is the fundamental research on
solution. Meta-optics is based on subwavelength-patterned sur- understanding the physics of metasurface elements. Their response
faces, called metasurfaces, which acquire optical functionalities is driven by optical resonance (localized or lattice-type5) and can be
through the scattering of nanoparticles rather than by refraction. tailored by their geometry for efficient light manipulation, generation
Metasurfaces offer extreme miniaturization of optical components, and detection. A composition of metasurface elements forms a metade-
together with smart functionalities beyond what is possible with vice that can deliver enhanced optical functionalities. The research on
conventional optics. The name stems from the Greek word μετά metadevices represents an increasingly large set of activities. It focuses
(meta), meaning ‘after’ or ‘beyond’, referring to optics beyond the on metasurface functionalities that cannot be achieved conventionally
three-millennia-old conventional optics. The meta-optics concept or exhibit improved performance with a miniaturized footprint. Exam-
provides a viable pathway for developing advanced machine vision ples include (Fig. 1, middle) flat metalenses6–9, polarization routers10,

ARC Centre of Excellence for Transformative Meta-Optical Systems (TMOS), Research School of Physics, The Australian National University, Canberra,
1

Australian Capital Territory, Australia. 2School of Engineering and Information Technology, University of New South Wales Canberra, Canberra, Australian
Capital Territory, Australia. e-mail: [email protected]

Nature Photonics | Volume 17 | January 2023 | 26–35 26

Review Article https://doi.org/10.1038/s41566-022-01126-4

ns
atio
plic
Ap

3D
es imaging
vic
De

Flat
AR and VR
lenses
nts
me
Ele
Holographic
Polarization
displays

vision
routers

LIDAR
Microscopy
coverslips

Image
Biosensors Amp. (a.U.)
1 cm classification

Light sail

Fig. 1 | Metasurfaces are dramatically enhancing the vision of robotic Images adapted with permission from: flat lenses, ref. 7 under a Creative
and autonomous systems. Schematic representation of the meta-optics Commons licence CC BY 4.0; polarization routers, ref. 10, AAAS; microscopy
research field. Left: metasurface elements. Middle: metadevices. Right: optical coverslips, ref. 11, Springer Nature Ltd; biosensors, ref. 111, Wiley; 3D imaging
system applications. Metadevices (from top to bottom) include metalenses7, (right), ref. 112, American Chemical Society; LIDAR, ref. 18, Springer Nature Ltd;
polarization routers10, microscopy coverslips11 and biosensors111. Applications image classification, ref. 19, AAAS. Holographic display image reproduced from
include 3D imaging for face identification112, AR and VR, holographic displays17, ref. 17 under a Creative Commons licence CC BY 4.0.
LIDAR beam steering18, image processing and classification19 and light sails.

specialty microscopy coverslips11,12 and biosensors13,14. The product seeking new applications. One can conclude that a strong industry drive
development of such metadevices has been the focus of several start-up currently defines the progress in the field. Namely, the new industrial
companies, such as Metalenz (www.metalenz.com), NILT Technologies applications define the research problems the field needs to solve. Sev-
(www.nilt.com) and Meta Materials Inc. (www.metamaterial.com), to eral other inspiring applications beyond vision can be identified. These
mention a few. In the search for markets, this industrial uptake also include using metasurfaces in light sail technologies20,21, light-fidelity
sets new challenges and seeks solutions from the research community. (LiFi) communications22 and thermal management23.
However, the biggest driver for the meta-optics field comes from To understand how such applications can benefit from the integra-
integrating metadevices into optical systems to offer new consumer tion of optical metasurfaces, we next review their basic functionalities.
optoelectronics applications. Importantly, meta-optical systems ena- We then describe how these functionalities can be achieved and outline
ble novel applications not conceivable before, adding to so-called the challenges in recent research.
Industry 4.015. Such applications include the Internet of Things (IoT),
vision for autonomous cars, wearable devices, AR and remote sensing. Metasurface functionalities
For example, an autonomously driving car requires three-dimensional Metasurfaces have gained a great deal of attention for their ability
(3D) vision enabled by light detection and ranging (LIDAR) devices in to control the spatial wavefront of light1,2, linking the fields of dif-
the infrared. Wearable sensors must see under the skin tissue, and AR fractive and nano-optics. Subsequently, metasurface functionalities
systems must display objects in conjunction with eye movements. have greatly increased, as summarized in Fig. 2 and described in the
Several examples of current industrial applications of metasurfaces following.
are listed in Fig. 1 (right), and many more system-level applications
are likely to emerge. Without a doubt, vision applications dominate, Spectral and dispersion control
including (from top to bottom) 3D imaging by point-cloud illumina- The frequency selectivity of metasurfaces stems directly from their res-
tion, AR vision16, holographic displays17, 3D LIDAR vision18, and image onant behaviour and, to a large extent, borrows ideas from the field of
processing and classification19. Over recent years, big industry players frequency-selective surfaces in radio waves. It is one of the oldest appli-
such as Apple, Google, Samsung, Meta (Facebook), Huawei, Sony and cations of optical metasurfaces, aiming at realizing notch or bandpass
STMicroelectronics have explored such meta-optical applications. filters in a single nanostructured layer. Application examples include
These and other companies have been actively investing in integrating eye protection from harmful lasers, such as in existing commercial
meta-optical systems in their products, hiring graduates in the field and products by Meta Materials Inc. (www.metamaterial.com). Over recent

Nature Photonics | Volume 17 | January 2023 | 26–35 27

Review Article https://doi.org/10.1038/s41566-022-01126-4

a c e g and microwave regions. Importantly, it allows for implementing spati-

Spectral Wavefront Emission Detection otemporal light modulations to manipulate the electromagnetic waves
control control control control in both spectral and spatial domains37. Spatiotemporal programmable
metasurfaces bridge the physical and digital worlds, leading to an
entirely new platform of information-operating metamaterials, which
can be coined ‘informational metasurfaces’. The reprogrammable
capability offers numerous applications ranging from LiFi communi-
cations to self-adaptive and cognitive radars empowered by artificial
b d f h intelligence algorithms.
Temporal Polarization Quantum Control of
control control state control chemistry Notably, the time variation of the metasurface properties can be
explored for breaking optical reciprocity, leading to applications in
free-space optical isolators. The required ultrafast time modulation
can be achieved using Kerr-type optical nonlinearity38. However, the
small modulation depth limits the overall applicability of the approach
Fig. 2 | Metasurface functionalities. a, Spectral control, including spectral in the optical spectral range.
filtering, colour routing and dispersion (spectral and angular) engineering. b,
Temporal control of wavepackets, including time-variant metasurfaces. The Wavefront control
combination of a and b represents the emerging field of space–time-variant Wavefront control of light beams is the driving-force functionality of
metasurfaces. c, Wavefront control, including lensing, beam shaping and optical metasurfaces. It relies on spatially tailoring the metasurface
holograms. d, Polarization control, including polarimetry and polarization properties and altering the transmitted or reflected light wavefront.
imaging. e, Emission control, including shaping of spontaneous and thermal As such, spatial engineering can be achieved at a subwavelength
emission, lasing and nonlinear frequency conversion. f, Control of quantum scale, diffraction is suppressed, and the wavefront manipulation
states, including generation of single photons, bi-photons and quantum
can have close-to-unity efficiency. We note that wavefront control
entanglement. g, Detection control, including perfect absorbers and
can be accomplished by either amplitude or phase modulation (or
photocurrent enhancement. h, Controlling chemical reactions by light-driven
a combination of both). For increased efficiency, phase modulation
charge exchange or strong coupling with metasurfaces.
is preferred and can enable beam focusing and imaging32,39, as well
as holographic projections40. Achieving full 2π phase modulation in
reflection is more accessible, as it only requires control of a single
years, researchers have extended the concepts to filters in the infrared resonant mode41. In transmission, more than one mode is necessary
and terahertz spectral regions, where conventional materials do not for a 2π phase modulation. This multimode response is achieved in
perform well24. Furthermore, narrow-bandpass filters with flat angular dielectric metasurfaces using Huygens-condition33 or waveguiding39
dispersion are of great importance in machine vision applications25. approaches. The latter offers higher efficiency and is commonly used
Another important application of the spectral selectivity of in industrial designs.
metasurfaces is the development of structural colours26,27, where the
frequency selectivity of metasurfaces has been explored to create Polarization control
bright colours spanning the entire chromaticity diagram. Examples Metasurfaces can enable the polarization control of light waves,
include plasmonic26 and Mie resonant metasurfaces28. Several design because anisotropic meta-atoms can have vastly different responses to
techniques have been used to enhance the visual performance of such orthogonal polarizations. Ideally, the meta-atoms should have unitary
colours, including forward, inverse and machine learning designs. transmission/reflection for both polarizations, but different phases.
Some recent works have made a further nontrivial extension, aiming The meta-atoms can therefore act as nanoscopic waveplates for the
at routing different spectral components to selected pixels29 instead incoming light. Placing such nano-waveplates at different spatial posi-
of filtering them. Such colour-routing schemes direct all incom- tions and orientations offers unprecedented polarization control39.
ing photons towards the detector and could benefit applications in Critical applications include polarimetry and polarization imaging10.
low-light conditions. In addition, metasurfaces can be engineered to enhance chirality
The metasurface frequency response is highly dispersive (having and chiral near-fields42, as well as to be achiral but enhance the sensi-
a π phase shift across the resonance), allowing for effective dispersion tivity to chiral molecules in their proximity43. The latter opens new
control. This functionality enables achromatic30 or hyper-dispersive31 applications in biosensing chiral molecules that are important for the
metasurfaces, with achromatic metalenses being one of the most pharmaceutical industry.
prominent examples8,32. Dispersion control is achieved by careful
engineering of the meta-atom geometry. However, the obtained Emission control
group delays are relatively small, limiting the transverse size of the Metasurfaces can alter the rate of spontaneous emission processes
meta-optical elements8. through resonant Purcell enhancement. This enhancement is propor-
tional to the quality factor of the resonance and inversely proportional
Temporal control of light wavepackets to the mode volume. The latter, however, requires special considera-
Temporal light control in metasurfaces can be realized by two tions, as metasurfaces are open systems and radiate into free space44.
approaches. The first comprises engineering of the group-delay The Purcell enhancements can be considerably higher in plasmonic
dispersion, thereby imposing temporal reshaping of the transmit- metasurfaces due to their localized mode volume. For optically excited
ted wavepackets. This opportunity was proposed in ref. 33 and was emitters, the emission rate can be further enhanced by resonances
recently demonstrated using a Fourier-transform set-up34 and by direct at the pump wavelength. In addition, the directionality of emission
dispersion engineering35. Although the overall group-velocity dis- could be engineered through multipolar interference in dielectric
persion remains low, metasurfaces can still outperform conventional metasurfaces45. For delocalized emitters, such as 2D transition-metal
pulse-shaping techniques. dichalcogenides46,47, or thermal emission23,48, careful engineering of
The second approach for temporal control is achieved by realizing the metasurface modes is required to take advantage of their near-field
metasurfaces with time-varying characteristics36. This approach is enhancement. Critical applications include brighter light-emitting
difficult at optical frequencies. However, it is feasible in the terahertz diodes and thermal sources.

Nature Photonics | Volume 17 | January 2023 | 26–35 28

Review Article https://doi.org/10.1038/s41566-022-01126-4

a b c g

500 nm 1 µm
500 nm

d e f

UV
500 nm SHG SHG
θ = 30°

5 µm 500 nm 300 nm

Fig. 3 | Novel metasurface geometries and materials. a–c, Novel metasurface Top: scanning electron microscopy (SEM) image of the TiO2 nanostructures.
geometries: topology-optimized metasurfaces for large-angle beam deflection71 Bottom: photograph of the fabricated 1-cm-diameter metalens79. Panels
(a); aberration-corrected metasurface spectrometer65 (b); three-dimensionally reproduced with permission from: a, ref. 71, Optica Publishing Group; e, ref. 74,
printed metasurfaces for independent amplitude and phase manipulation70 (c). American Chemical Society; f, ref. 77 under a Creative Commons licence CC BY
d,e, Novel materials for metasurfaces: lithium niobate metasurface for SHG73 4.0. Panels adapted with permission from: b, ref. 65, Wiley; c, ref. 70, Springer
(d); gallium phosphide metasurface for continuous-wave SHG74 (e). f, High- Nature Ltd; d, ref. 73, American Chemical Society; g(bottom), ref. 79 under a
performance hafnium oxide metasurface operating at ultraviolet wavelengths77. Creative Commons licence CC BY 4.0.
g, Large-area TiO2 metalens fabrication suitable for virtual-reality applications.

In the case of high-quality-factor metasurfaces, lasing action can Detection control

also be obtained49 where the directionality50 and mode profile51 can The engineering of the modes of metasurfaces allows for enhancing
be engineered. The process of nonlinear frequency conversions, such light absorption in ultrathin layers. The absorbed energy can be con-
as second harmonic generation (SHG) to a large extent, can also be verted into heat, photocurrent or hot carriers. Plasmonic enhanced
considered an emission process due to the excitation of nonlinear photo-thermoelectric conversion61 and plasmon-induced hot carri-
currents inside the metasurfaces. SHG has been explored in plasmonic ers62 are showing great promise for applications in infrared detection,
and dielectric metasurfaces52. However, the dielectric metasurfaces despite their lower efficiency.
enable substantially higher frequency conversion rates due to their The resonance phenomena in dielectric metasurfaces containing
larger mode volume and higher pumping powers. Although the achiev- p–n junctions have been used to enhance the absorption of specific col-
able conversion efficiencies remain below 1%, possible applications ours at the pixel level63. Dielectric metasurfaces have further enabled
include frequency upconversion of infrared imaging53 for night- the detection of different light modalities, including angle selectivity64.
vision applications. The exploration of such enhanced detection has allowed the realization
of important devices, such as spectrometers65 and cameras63.
Manipulating the quantum properties of light
Optical metasurfaces have been explored for quantum light gen- Control of chemical reactions
eration, including single- or bi-photon states and for manipulat- Metasurfaces have recently attracted attention to control chemical
ing quantum states54. The brightness of single-photon emitters reactions based on light-driven charge exchange or strong coupling.
can also be enhanced following the emission-control functionality Such control is mainly suitable for plasmonic metasurfaces66, although
described above. However, such emitters are highly localized, requir- semiconductor systems, such as TiO2 and other metal oxides67, could
ing high-precision placement on the metasurface. Such placement also be employed. Plasmonic tunnel junctions for redox chemis-
remains a substantial challenge, and the benefits of integrating try have been tested68, opening up critical applications in CO2 and
single-photon emitters with metasurfaces remain elusive. The genera- water splitting.
tion of bi-photon quantum states by spontaneous parametric down- Furthermore, light–matter hybridization can alter the chemical
conversion from nonlinear quadratic metasurfaces shows greater properties of the coupled system. Strong coupling between resonant
promise. Mie resonant55 and high-quality-factor metasurfaces56 have metasurfaces and the material in their proximity can modify molecular
been successfully used to generate bi-photon states and complex transitions, chemical reactivity and superconductivity69.
quantum states57. A metalens array, combined with a bulk nonlinear
crystal, has also been used to generate multiphoton quantum states58. Challenges for metasurface development
The generation of quantum entanglement from metasurfaces has also Although the meta-optics field has matured and is testing market
been demonstrated recently56. opportunities, there are several challenges ahead. These can be divided
Finally, metasurfaces have shown a great ability to manipu- into two broad categories: (1) materials and fabrication challenges,
late the state of quantum light after its generation. Manipulations including large-scale and scalable fabrication, incorporating differ-
of single-photon59 and multiphoton states60 have been successfully ent materials platforms, and (2) physics and engineering challenges,
demonstrated. These pioneering experiments are paving the way for focusing on novel concepts, developing new designs with enhanced
novel quantum applications of metasurfaces, such as quantum imag- functionalities, and hybridizing metasurfaces with other materials—
ing, sensing and communications. multi-physics metasurfaces.

Nature Photonics | Volume 17 | January 2023 | 26–35 29

Review Article https://doi.org/10.1038/s41566-022-01126-4

a Tuning of resonator material b Tuning of environment c Tuning of geometry

d Engineering of dispersion e Multi-layer and 3D designs f Multi-physics metasurfaces

Fig. 4 | Metasurface challenges. a–c, Tunability by altering the resonator f, Multi-physics metasurfaces enabled by coupling to the materials’ degrees of
materials (a), the surrounding environment (b) and mechanically modifying the freedom, such as excitons98, polaritons and phonons. Panel d reproduced with
metasurface geometry (c). d, Engineering the spectral and spatial dispersion permission from ref. 90, APS. Panels adapted with permission from: e, ref. 95,
properties90. e, Large-area inverse design of the metasurface properties95. American Chemical Society; f, ref. 98, APS.

Materials and fabrication challenges off-diagonal nonlinear susceptibility tensor, and metasurfaces from
Metasurfaces are fabricated by planar fabrication technology, effec- other transparent crystals, have been sought. An enhanced SHG has
tively merging optics and chip-making. This planar fabrication presents been demonstrated in lithium niobate73 (Fig. 3d) and in GaP metas-
a tremendous advantage for translation to industrial applications. In urfaces74 (Fig. 3e). Beyond nonlinear applications, other dielectric
comparison, nanostructuring of bulk 3D materials remains a considera- materials, such as TiO232, GaN40, SiN75 and SiC76, have also been tested.
ble challenge. Although techniques such as 3D laser writing70 are getting Although many of these semiconductor metasurfaces can operate from
closer to sub-100-nm resolution, volumetric metamaterials operating the visible to the infrared spectral regions, the extension of their opera-
in the visible and near-infrared remain out of reach. As a result, planar tion to the ultraviolet (UV) range has been a challenge. This challenge
metasurfaces dominate research and applications. However, many lies in the transparency of the materials in the UV range and the required
materials used in the research laboratories for metasurface fabrica- small feature sizes. Nevertheless, the first demonstrations of functional
tion are incompatible with the industry standards of semiconductor UV metasurfaces using hafnium oxide have been presented77. Vacuum
fabrication foundries. It is therefore often challenging to translate the ultraviolet nonlinear metalenses based on ZnO meta-atoms have also
technology to scalable manufacturing and industry-ready standards. been demonstrated recently78.
For example, common plasmonic metals like gold and silver are incom- Finally, the fabrication of large-area metasurfaces that can
patible with complementary metal–oxide–semiconductor (CMOS) fab- replace existing optical elements is not without challenges. Most
rication. Dielectric metasurfaces, often made of silicon, have bridged electron-beam lithography tools used in the laboratory cannot pat-
that gap. However, although conventional CMOS chip-making uses tern large areas, and advanced photomasks are required for deep-UV
silicon-on-insulator wafers, silicon on transparent substrates is required lithography. Nevertheless, large-area meta-optical elements have
for optical metasurfaces. The technology of amorphous silicon (a-Si) on recently been demonstrated79 (Fig. 3g), and this is expected to increase
glass wafers has recently been demonstrated by Metalenz together with industry interest further.
STMicroelectronics, using extreme ultraviolet (XUV) lithography (www.
metalenz.com). However, this fabrication is not without challenges, Physics and engineering challenges
including the reproducibility of the a-Si-layer deposition, its robustness It is impossible to identify all the challenges in a field with so many
to environmental changes, and process challenges when using glass diverse applications, so we focus here on some of the global ones. As
wafers. Implementing other materials for operation in the visible and the major challenge, we identify the tunability of the optical response
UV spectral ranges will further challenge the fabrication processes. of metasurfaces, including the tunability of transmission amplitude,
In terms of the design aspect, metasurfaces have made enormous phase and polarization. Such tunability can enable the realization of
progress over recent years, figuratively named by Federico Capasso as programmable meta-optical elements, which can perform arbitrary
a transition from realism to surrealism. Indeed, the initial ‘work-horse’ operations depending on an external stimulus. For example, spatially
nano-cylinder meta-atoms have been replaced by high-performing variant tuning of the transmitted phase can allow a single metasurface
inverse design structures. Figure 3a shows one example used for to become a length lens, a beam shaper or a hologram. Because the
large-angle beam deflection71. Complex dispersion-engineered meta-atoms have nanoscale volumes, it should be easier to tune their
meta-atoms65 have also been explored (Fig. 3b), and full 3D designs properties. However, so far, no unified approach exists, and most works
achieved by laser printing (Fig. 3c) have been employed for moving consider only amplitude modulation. To our knowledge, no continu-
holographic images70. ous 2π phase modulation (with unity efficiency) nor full polarization
With the expansion of meta-optics applications, the community control has been reported. The difficulty is that for both phase and
is actively exploring the fabrication of different high-index dielectric polarization tuning, simultaneous control of the spectral positions of
materials. For example, crystalline materials with second-order non- two optical modes is needed, and this requires two control parameters.
linearity are required in nonlinear applications52, such as SHG. One of The tunability of metasurfaces can be enabled by three generic
the most used materials in this regard is GaAs and its Al compounds72. principles: (1) altering the resonator material (Fig. 4a) through a mate-
However, challenges due to the limited transparency of GaAs and its rial phase change80,81, thermo-optics82, electro-optics83 or carrier

Nature Photonics | Volume 17 | January 2023 | 26–35 30

Review Article https://doi.org/10.1038/s41566-022-01126-4

a b
Ex
Iout
Iin
Hy
kz
e
fac
ur
tas
Me p
(–) p
ge
lta
vo
Gate tGST dAu
tAu

P-GST

+) Hard-
e( Mild-
ltag annealing annealing
vo
te

yz
Ga
te A-GST
tra C-GST
x su
bs
Super-cell geometry Output spectrum Si
Melt-quenching Au Al2O3 GST W HfO2

c d Photon-pair Polarization
θ1
θ2 θ1 source
Probe Reference
θ3 θ2 θ1 θ2 θ3 θ4
θ
θ4 θ3 AFR Object
1.0 MP MR
Reference Ω

Light intensity (a.u.)

1.0
Varia θ4
ble
Reflectance

angle
θ
0.5 ∆/(k) +1st
–1st 0th
Metasurfaces
0th
Analyte Coincidence
0 0 counter
–1
Wavenumber (cm ) Light incidence angle, θ

e f g
σ+ σ–
0.4

Exceptional
±|ω|

0
point
–0.4
–0.1 0 0.1
ε

L3 MoS2-meta.
1,500
PL intensity (counts)

MoS2
1,200 Meta.
L2
L1 Subs.
900 10 µm
Bound exciton
600
A exciton T = 87 K
300
640 680 720 760 800
Wavelength (nm)

Fig. 5 | Research frontiers in the metasurface field. a, Inverse photonics a, ref. 101, Optica Publishing Group; b, ref. 81 under a Creative Commons licence
design101. b, Tunable metasurfaces81. c, Nonlocal metasurfaces103. d, Quantum CC BY 4.0; d, ref. 104, APS; e, ref. 105, Electromagnetics Academy; f, ref. 113, Wiley;
metasurfaces104. e, Non-Hermitian metasurfaces105. f, 2D materials metasurfaces g, ref. 106 under a Creative Commons licence CC BY 4.0. Panel c adapted with
(coupling to materials’ degree of freedom)113. PL, photoluminescence. g, Space– permission from ref. 103, AAAS.
time-variant metasurfaces106. Panels reproduced with permission from:

injections37,84; (2) modifying the surrounding environment (Fig. 4b) Finally, the coupling of optical metasurfaces to other material
by liquid crystals85,86 or optofluidics87; (3) changing the geometry degrees of freedom, including atomic96, excitonic97, phononic88 and
(Fig. 4c) by mechanical actuations88 or stretching89. Which of these polaritonic excitations98, is facilitating completely unknown opportu-
tuning mechanisms will meet the industry standards remains to be nities for multi-physics metasurfaces (Fig. 4f). In the strong coupling
seen, but is likely to be defined by the specific application constraints. regime, the optical and materials properties are hybridized, leading
Another major challenge in the field is the complete control of to novel polaritonic regimes and phenomena. The modification of
angular and spectral dispersion90 (Fig. 4d). Such control is expected to chemical reactions, discussed in Fig. 2h, is an important example of
have a powerful impact on applications of dispersive31 and achromatic such a novel phenomenon.
meta-optics91,92. The angular dispersion also plays a vital role in optical
image processing, including edge detection by image differentiation11,93 Frontiers in metasurface research
(Fig. 4d). To meet the above challenges, research frontiers have focused on sev-
As meta-optical elements are becoming integral to various optical eral fundamental and applied directions. On the fundamental side,
systems, it is becoming increasingly important to have tools for a full these include the implementation of deep learning and inverse design
system design. This need has triggered strong activity regarding inverse approaches99,100 for obtaining the desired spectral, phase, polarization
design, including end-to-end designs94, large-scale meta-optics79 and and transverse momentum responses. For example, Fig. 5a shows a
3D inverse designs95 (Fig. 4e). neural network technique for tailoring the metasurface transmission

Nature Photonics | Volume 17 | January 2023 | 26–35 31

Review Article https://doi.org/10.1038/s41566-022-01126-4

a b

Al

DAl Ag
IAl Au
DAg
Test chart
t SiO IAg
DAu Objective
2
IAu Lens Camera

15 µm

30 µm 4 µm

λ = 1,120 nm
700
600
Scattering (a.u.)

500
400
300
200
100
0

0
0
0

0
0

0
40

60

80
50
45

65

70
55

75

Wavelength (nm)

NIR-LED
VIS
Eye Cam. NIR

NIR SiO2 Diffracted orders

pSi

Si3N4
w (900 nm)
VIS Metasurface
pSi (3 nm)

Quartz
Si3N4

p (1,000 nm)
5 µm 500 nm
θ = 0° 40° 80°
Outside world

Fig. 6 | Metasurfaces enabling smart vision. a, Multicolour imaging with multi- and eye-tracking by guided-wave metasurfaces110. Figure reproduced with
layer metasurfaces108. Scale bars, 35 μm (middle row); 10 μm (bottom-left image); permission from: a, ref. 108 under a Creative Commons licence CC BY 4.0; b,
and 20 μm (bottom-right inset). b, Image differentiation and edge detection as ref. 11, Springer Nature Ltd; c, ref. 110, Springer Nature Ltd.
examples of image processing11. c, Spectrally decoupled wavefront manipulation

spectrum101. The next frontier is focused on achieving metasurface quantum light direction promises important applications. Metas-
tunability on a pixel level. Here, challenges associated with pixel cross- urfaces with gain and loss can also be considered as an upcoming
talk, tuning speed and electrical control need to be solved to enable frontier. Such metasurfaces can be designed to exhibit unconventional
electrically programmable mega-pixel metasurfaces. Figure 5b shows non-Hermitian properties, supporting exceptional points for unidirec-
one promising approach based on phase-change materials81. Another tional propagation and adiabatic mode conversion. Figure 5e presents
important frontier is spatial dispersion engineering through the design an example of non-Hermitian metasurfaces, which allow for ultrasensi-
of nonlocal metasurfaces102. Figure 5c shows the use of a spatially dis- tive measurements105. The integration of 2D materials47 also opens new
persive metasurface in biosensing103. opportunities for frontier optoelectronic applications, enabled, for
Arguably, one of the most exciting research frontiers is the appli- example, by precision control of excitonic excitations97 (Fig. 5f). Finally,
cation of optical metasurfaces for quantum light generation56,57, the concept of space–time metasurfaces is important for complete
detection104 (Fig. 5d) and classification60. Although still in its infancy, spatiotemporal wavefront control. This has been demonstrated in

Nature Photonics | Volume 17 | January 2023 | 26–35 32

Review Article https://doi.org/10.1038/s41566-022-01126-4

the microwave spectral range106 (Fig. 5g), and it is highly desirable to References
translate it to the optical domain. 1. Bomzon, Z. E., Kleiner, V. & Hasman, E. Pancharatnam-Berry
On the application side, research frontiers are focusing on devel- phase in space-variant polarization-state manipulations with
oping smart vision, with three key aspects in mind. The first is the reali- subwavelength gratings. Opt. Lett. 26, 1424–1426 (2001).
zation of better imaging devices through multi-layered metasurfaces. 2. Yu, N. et al. Light propagation with phase discontinuities:
Although single-layer metasurfaces have been successfully employed generalized laws of reflection and refraction. Science 334,
in many meta-optical applications, multi-layered metasurfaces are 333–337 (2011).
required for full aberration correction and novel functionalities. 3. McPhedran, R. C., Shadrivov, I. V., Kuhlmey, B. T. & Kivshar, Y. S.
Stacking of multiple metasurfaces is still challenging, but it offers an Metamaterials and metaoptics. NPG Asia Mater. 3, 100–108 (2011).
increased degree of design freedom and allows for multi-wavelength107 4. Chu, J. S. G. & Evans, J. A. Slowed canonical progress in
and white-light focusing108 (Fig. 6a) or multicolour holograms109. large fields of science. Proc. Natl Acad. Sci. USA 118,
The second aspect of smart vision is focused on image-processing e2021636118 (2021).
metasurfaces. An important example of image processing is the edge 5. Ko, Y. H. & Magnusson, R. Wideband dielectric metamaterial
detection of images, as shown in Fig. 6b. However, more complex reflectors: Mie scattering or leaky Bloch mode resonance? Optica
imaging functions, including image classification19, are also being 5, 289–294 (2018).
investigated. The third aspect focuses on machine–human interfaces, 6. Khorasaninejad, M. & Capasso, F. Metalenses: versatile
such as AR vision sets, where the display, processing and tracking func- multifunctional photonic components. Science 358,
tions must be combined. A recent example of simultaneous display eaam8100 (2017).
and eye-tracking functions by guided-wave metasurfaces110 is shown 7. Wang, S. et al. Broadband achromatic optical metasurface
in Fig. 6c. We expect this set of frontier technologies to expand fur- devices. Nat. Commun. 8, 187 (2017).
ther, as metasurfaces prove their potential and reliability in end-user 8. Engelberg, J. & Levy, U. Achromatic flat lens performance limits.
optical systems. Optica 8, 834–845 (2021).
9. Zou, X. et al. Imaging based on metalenses. PhotoniX 1,
Perspective 2 (2020).
Overall, optical metasurfaces present several essential benefits to mod- 10. Rubin, N. A. et al. Matrix Fourier optics enables a compact
ern optical technologies. First, they allow for the ultimate miniaturiza- full-Stokes polarization camera. Science 365, eaax1839 (2019).
tion of optical components, reducing the size, weight and power (SWaP) 11. Zhou, Y., Zheng, H., Kravchenko, I. I. & Valentine, J. Flat optics for
requirements for optical systems. Such miniaturization is essential in image differentiation. Nat. Photon. 14, 316–323 (2020).
modern consumer optoelectronic devices, drones and space applica- 12. Wesemann, L. et al. Nanophotonics enhanced coverslip for phase
tions. Second, they enable smart functions that are not readily achiev- imaging in biology. Light Sci. Appl. 10, 98 (2021).
able with conventional optics. These include visualizing the hidden 13. Balaur, E. et al. Colorimetric histology using plasmonically active
information carried by light (for example, polarization, phase and microscope slides. Nature 598, 65–71 (2021).
other modalities) and integrating multiple functions in a single optical 14. Altug, H., Oh, S.-H., Maier, S. A. & Homola, J. Advances and
element, which is at the essence of Industry 4.0 applications. Third, the applications of nanophotonic biosensors. Nat. Nanotechnol. 17,
nanoscale volumes of the resonators offer opportunities for fast and 5–16 (2022).
real-time reconfigurability of optical functions, something that remains 15. Coffey, V. C. Machine vision: The Eyes of Industry 4.0. Opt. Photon.
to be accomplished. We should also flag the advantages metasurfaces News 29, 42–49 (2018).
offer in exploring other spectral ranges, such as long and mid-wave 16. Lee, G.-Y. et al. Metasurface eyepiece for augmented reality. Nat.
infrared, which are not easily accessible with conventional glass optics Commun. 9, 4562 (2018).
and CMOS cameras. The development of novel optical components 17. An, J. et al. Slim-panel holographic video display. Nat. Commun.
is particularly important for detecting infrared light, where no good 11, 5568 (2020).
detectors exist. Metasurface-enhanced infrared detectors can find 18. Kim, I. et al. Nanophotonics for light detection and ranging
important applications in night vision and the spectral identification technology. Nat. Nanotechnol. 16, 508–524 (2021).
of chemical compounds. 19. Lin, X. et al. All-optical machine learning using diffractive deep
The research field of meta-optics has experienced an exponential neural networks. Science 361, 1004–1008 (2018).
growth of ideas over the past five years. These ideas have attracted 20. Atwater, H. A. et al. Materials challenges for the Starshot lightsail.
industry interest that, in return, has defined new scientific questions Nat. Mater. 17, 861–867 (2018).
and is driving the modern development of the field. Among the numer- 21. Tung, H.-T. & Davoyan, A. R. Low-power laser sailing for fast-transit
ous applications, machine vision is the most important market oppor- space flight. Nano Lett. 22, 1108–1114 (2022).
tunity, offering capabilities surpassing human vision and enabling 22. Abumarshoud, H. et al. LiFi through reconfigurable intelligent
machine autonomy, for example, in autonomous vehicles. We are surfaces: a new frontier for 6G? IEEE Veh. Technol. Mag. 17,
also seeing opportunities for meta-optics in enhancing human vision 37–46 (2022).
through AR technologies, and the formation of new machine–human 23. Overvig, A. C., Mann, S. A. & Alù, A. Thermal metasurfaces:
interfaces. Of course, the path from research to market is much more complete emission control by combining local and nonlocal
complex than the published laboratory demonstrations. Market analy- light-matter interactions. Phys. Rev. 11, 021050 (2021).
sis must consider the technology’s cost, reproducibility, scalability 24. Chang, C.-C., Huang, L., Nogan, J. & Chen, H.-T. Invited Article:
and long-term stability. For example, stringent space testing must be narrowband terahertz bandpass filters employing stacked
performed for space-flight stability and compatibility to deploy light- bilayer metasurface antireflection structures. APL Photonics 3,
weight meta-optics in space. We envision that the first metadevices 051602 (2018).
to be translated to the market will be metalenses and holographic 25. Zheng, Z. et al. Planar narrow bandpass filter based on Si resonant
elements for monochromatic imaging, such as in face ID or driver metasurface. J. Appl. Phys. 130, 053105 (2021).
fatigue monitoring applications. These will probably be followed by AR 26. Kristensen, A. et al. Plasmonic colour generation. Nat. Rev. Mater.
applications about five years from now. However, one thing is sure—we 2, 16088 (2016).
will increasingly see optical metasurfaces under the hood of many 27. Daqiqeh Rezaei, S. et al. Nanophotonic structural colors. ACS
consumer devices and Industry 4.0 applications. Photonics 8, 18–33 (2021).

Nature Photonics | Volume 17 | January 2023 | 26–35 33

Review Article https://doi.org/10.1038/s41566-022-01126-4

28. Flauraud, V., Reyes, M., Paniagua-Domínguez, R., Kuznetsov, A. I. & 53. Camacho-Morales, R. et al. Infrared upconversion imaging in
Brugger, J. Silicon nanostructures for bright field full color prints. nonlinear metasurfaces. Adv. Photonics 3, 036002 (2021).
ACS Photonics 4, 1913–1919 (2017). 54. Solntsev, A. S., Agarwal, G. S. & Kivshar, Y. S. Metasurfaces for
29. Zou, X. et al. Pixel-level Bayer-type colour router based on quantum photonics. Nat. Photon. 15, 327–336 (2021).
metasurfaces. Nat. Commun. 13, 3288 (2022). 55. Santiago-Cruz, T. et al. Photon pairs from resonant metasurfaces.
30. Aieta, F., Kats, M. A., Genevet, P. & Capasso, F. Multiwavelength Nano Lett. 21, 4423–4429 (2021).
achromatic metasurfaces by dispersive phase compensation. 56. Zhang, J. et al. Spatially entangled photon pairs from lithium
Science 347, 1342–1345 (2015). niobate nonlocal metasurfaces. Sci. Adv. 8, eabq4240 (2022).
31. Arbabi, E., Arbabi, A., Kamali, S. M., Horie, Y. & Faraon, A. 57. Santiago-Cruz, T. et al. Resonant metasurfaces for generating
Controlling the sign of chromatic dispersion in diffractive optics complex quantum states. Science 377, 991–995 (2022).
with dielectric metasurfaces. Optica 4, 625–632 (2017). 58. Li, L. et al. Metalens-array-based high-dimensional and
32. Chen, W. T. et al. A broadband achromatic metalens for focusing multiphoton quantum source. Science 368, 1487–1490 (2020).
and imaging in the visible. Nat. Nanotechnol. 13, 220–226 (2018). 59. Stav, T. et al. Quantum entanglement of the spin and orbital
33. Decker, M. et al. High-efficiency dielectric Huygens’ surfaces. Adv. angular momentum of photons using metamaterials. Science
Opt. Mater. 3, 813–820 (2015). 361, 1101–1104 (2018).
34. Divitt, S., Zhu, W., Zhang, C., Lezec, H. J. & Agrawal, A. Ultrafast 60. Wang, K. et al. Quantum metasurface for multiphoton
optical pulse shaping using dielectric metasurfaces. Science 364, interference and state reconstruction. Science 361,
890–894 (2019). 1104–1108 (2018).
35. Ossiander, M. et al. Slow light nanocoatings for ultrashort pulse 61. Horikawa, M., Fang, X. & Kubo, W. Metamaterial perfect
compression. Nat. Commun. 12, 6518 (2021). absorber-enhanced plasmonic photo-thermoelectric conversion.
36. Galiffi, E. et al. Photonics of time-varying media. Adv. Photonics 4, Appl. Phys. Express 13, 082006 (2020).
014002 (2022). 62. Tanzid, M. et al. Combining plasmonic hot carrier generation
37. Shirmanesh, G. K., Sokhoyan, R., Wu, P. C. & Atwater, H. A. with free carrier absorption for high-performance near-
Electro-optically tunable multifunctional metasurfaces. ACS Nano infrared silicon-based photodetection. ACS Photonics 5,
14, 6912–6920 (2020). 3472–3477 (2018).
38. Guo, X., Ding, Y., Duan, Y. & Ni, X. Nonreciprocal metasurface with 63. Park, H. et al. Filter-free image sensor pixels comprising silicon
space-time phase modulation. Light Sci. Appl. 8, 123 (2019). nanowires with selective color absorption. Nano Lett. 14,
39. Arbabi, A., Horie, Y., Bagheri, M. & Faraon, A. Dielectric 1804–1809 (2014).
metasurfaces for complete control of phase and polarization with 64. Yi, S. et al. Subwavelength angle-sensing photodetectors inspired
subwavelength spatial resolution and high transmission. Nat. by directional hearing in small animals. Nat. Nanotechnol. 13,
Nanotechnol. 10, 937–943 (2015). 1143–1147 (2018).
40. Ren, H. et al. Metasurface orbital angular momentum holography. 65. Zhu, A. Y. et al. Compact aberration-corrected spectrometers
Nat. Commun. 10, 2986 (2019). in the visible using dispersion-tailored metasurfaces. Adv. Opt.
41. Pors, A., Albrektsen, O., Radko, I. P. & Bozhevolnyi, S. I. Gap Mater. 7, 1801144 (2019).
plasmon-based metasurfaces for total control of reflected light. 66. Zhou, L. et al. Hot carrier multiplication in plasmonic
Sci. Rep. https://doi.org/10.1038/srep02155 (2013). photocatalysis. Proc. Natl Acad. Sci. USA 118, e2022109118 (2021).
42. Basiri, A. et al. Nature-inspired chiral metasurfaces for 67. Jia, Q. et al. Metal and metal oxide interactions and their catalytic
circular polarization detection and full-Stokes polarimetric consequences for oxygen reduction reaction. J. Am. Chem. Soc.
measurements. Light Sci. Appl. 8, 78 (2019). 139, 7893–7903 (2017).
43. Solomon, M. L., Hu, J., Lawrence, M., García-Etxarri, A. & Dionne, J. 68. de Nijs, B. et al. Plasmonic tunnel junctions for single-molecule
A. Enantiospecific optical enhancement of chiral sensing redox chemistry. Nat. Commun. 8, 994 (2017).
and separation with dielectric metasurfaces. ACS Photonics 6, 69. Garcia-Vidal, F. J., Ciuti, C. & Ebbesen, T. W. Manipulating
43–49 (2019). matter by strong coupling to vacuum fields. Science 373,
44. Sauvan, C., Hugonin, J. P., Maksymov, I. S. & Lalanne, P. Theory eabd0336 (2021).
of the spontaneous optical emission of nanosize photonic and 70. Ren, H. R. et al. Complex-amplitude metasurface-based orbital
plasmon resonators. Phys. Rev. Lett. 110, 237401 (2013). angular momentum holography in momentum space. Nat.
45. Vaskin, A., Kolkowski, R., Koenderink, A. F. & Staude, I. Nanotechnol. 15, 948–955 (2020).
Light-emitting metasurfaces. Nanophotonics 8, 1151–1198 (2019). 71. Wang, E. W., Sell, D., Phan, T. & Fan, J. A. Robust design of
46. Tran, T. T. et al. Deterministic coupling of quantum emitters topology-optimized metasurfaces. Opt. Mater. Express 9,
in 2D materials to plasmonic nanocavity arrays. Nano Lett. 17, 469–482 (2019).
2634–2639 (2017). 72. Rahmani, M. et al. Nonlinear frequency conversion in optical
47. Huang, L. et al. Enhanced light-matter interaction in nanoantennas and metasurfaces: materials evolution and
two-dimensional transition metal dichalcogenides. Rep. Prog. fabrication. Opto Electron. Adv. 1, 180021 (2018).
Phys. 85, 046401 (2022). 73. Fedotova, A. et al. Second-harmonic generation in resonant
48. Fan, S. & Li, W. Photonics and thermodynamics concepts in nonlinear metasurfaces based on lithium niobate. Nano Lett. 20,
radiative cooling. Nat. Photon. 16, 182–190 (2022). 8608–8614 (2020).
49. Kodigala, A. et al. Lasing action from photonic bound states in 74. Anthur, A. P. et al. Continuous wave second harmonic
continuum. Nature 541, 196–199 (2017). generation enabled by quasi-bound-states in the continuum
50. Ha, S. T. et al. Directional lasing in resonant semiconductor on gallium phosphide metasurfaces. Nano Lett. 20,
nanoantenna arrays. Nat. Nanotechnol. 13, 1042–1047 (2018). 8745–8751 (2020).
51. Huang, C. et al. Ultrafast control of vortex microlasers. Science 75. Colburn, S. et al. Broadband transparent and CMOS-compatible
367, 1018–1021 (2020). flat optics with silicon nitride metasurfaces [Invited]. Opt. Mater.
52. Grinblat, G. Nonlinear dielectric nanoantennas and metasurfaces: Express 8, 2330–2344 (2018).
frequency conversion and wavefront control. ACS Photonics 8, 76. Schaeper, O. et al. Monolithic silicon carbide metalenses.
3406–3432 (2021). ACS Photonics 9, 1409–1414 (2022).

Nature Photonics | Volume 17 | January 2023 | 26–35 34

Review Article https://doi.org/10.1038/s41566-022-01126-4

77. Zhang, C. et al. Low-loss metasurface optics down to the deep 102. Overvig, A. & Alù, A. Diffractive nonlocal metasurfaces. Laser
ultraviolet region. Light Sci. Appl. 9, 55 (2020). Photon. Rev 16, 2100633 (2022).
78. Tseng, M. L. et al. Vacuum ultraviolet nonlinear metalens. Sci. Adv. 103. Leitis, A. et al. Angle-multiplexed all-dielectric metasurfaces
8, eabn5644 (2022). for broadband molecular fingerprint retrieval. Sci. Adv. 5,
79. Li, Z. et al. Inverse design enables large-scale high-performance eaaw2871 (2019).
meta-optics reshaping virtual reality. Nat. Commun. 13, 104. Vega, A., Pertsch, T., Setzpfandt, F. & Sukhorukov, A. A.
2409 (2022). Metasurface-assisted quantum ghost discrimination of
80. Leitis, A. et al. All-dielectric programmable Huygens’ polarization objects. Phys. Rev. Appl. 16, 064032 (2021).
metasurfaces. Adv. Funct. Mater. 30, 1910259 (2020). 105. Li, Z. et al. Non-Hermitian electromagnetic metasurfaces at
81. Abdollahramezani, S. et al. Electrically driven reprogrammable exceptional points (Invited Review). Progress Electromagn. Res.
phase-change metasurface reaching 80% efficiency. Nat. 171, 1–20 (2021).
Commun. 13, 1696 (2022). 106. Zhao, J. et al. Programmable time-domain digital-coding
82. Zangeneh Kamali, K. et al. Reversible image contrast metasurface for non-linear harmonic manipulation and new
manipulation with thermally tunable dielectric metasurfaces. wireless communication systems. Natl Sci. Rev. 6, 231–238 (2018).
Small 15, 1805142 (2019). 107. Zhou, Y. et al. Multilayer noninteracting dielectric metasurfaces
83. Benea-Chelmus, I.-C. et al. Gigahertz free-space electro- for multiwavelength metaoptics. Nano Lett. 18, 7529–7537 (2018).
optic modulators based on Mie resonances. Nat. Commun. 13, 108. Avayu, O., Almeida, E., Prior, Y. & Ellenbogen, T. Composite
3170 (2022). functional metasurfaces for multispectral achromatic optics. Nat.
84. Karst, J. et al. Electrically switchable metallic polymer Commun. 8, 14992 (2017).
nanoantennas. Science 374, 612–616 (2021). 109. Hu, Y. et al. 3D-Integrated metasurfaces for full-colour
85. Komar, A. et al. Electrically tunable all-dielectric optical holography. Light Sci. Appl. 8, 86 (2019).
metasurfaces based on liquid crystals. Appl. Phys. Lett. 110, 110. Song, J.-H., van de Groep, J., Kim, S. J. & Brongersma, M. L.
071109 (2017). Non-local metasurfaces for spectrally decoupled wavefront
86. Li, S.-Q. et al. Phase-only transmissive spatial light modulator based manipulation and eye tracking. Nat. Nanotechnol. 16,
on tunable dielectric metasurface. Science 364, 1087–1090 (2019). 1224–1230 (2021).
87. Li, Q. et al. Metasurface optofluidics for dynamic control of light 111. Aoni, R. A. et al. Resonant dielectric metagratings for response
fields. Nat. Nanotechnol. 17, 1097–1103 (2022). intensified optical sensing. Adv. Funct. Mater. 32, 2103143 (2022).
88. Ajia, I. A. et al. Gigahertz nano-optomechanical resonances in 112. Ni, Y. et al. Metasurface for structured light projection over 120°
a dielectric SiC-membrane metasurface array. Nano Lett. 21, field of view. Nano Lett. 20, 6719–6724 (2020).
4563–4569 (2021). 113. Li, Z. et al. Tailoring MoS2 valley-polarized photoluminescence
89. Zhang, X. et al. Reconfigurable metasurface for image with super chiral near-field. Adv. Mater. 30, 1801908 (2018).
processing. Nano Lett. 21, 8715–8722 (2021).
90. Long, O. Y., Guo, C., Jin, W. & Fan, S. Polarization-independent Acknowledgements
isotropic nonlocal metasurfaces with wavelength-controlled We thank F. Capasso, S. Maier, K. Crozier and A. Davoyan for
functionality. Phys. Rev. Appl. 17, 024029 (2022). valuable discussions and M. Alaloul for feedback regarding
91. Li, Z. et al. Meta-optics achieves RGB-achromatic focusing for metasurface detectors. This work was supported by the Australian
virtual reality. Sci. Adv. 7, eabe4458 (2021). Research Council Centre of Excellence (CE200100010), Discovery
92. Ren, H. et al. An achromatic metafiber for focusing and imaging (DP200101353) and Linkage (LP180100904) programmes, as well as
across the entire telecommunication range. Nat. Commun. 13, through a DARPA ENVision grant (HR00112220006).
4183 (2022).
93. Guo, C., Xiao, M., Minkov, M., Shi, Y. & Fan, S. Photonic crystal Competing interests
slab Laplace operator for image differentiation. Optica 5, The authors declare no competing interests.
251–256 (2018).
94. Lin, Z. et al. End-to-end nanophotonic inverse design for imaging Additional information
and polarimetry. Nanophotonics 10, 1177–1187 (2021). Correspondence should be addressed to Dragomir N. Neshev.
95. Wang, H. et al. Design of compact meta-crystal slab for general
optical convolution. ACS Photonics 9, 1358–1365 (2022). Peer review information Nature Photonics thanks Sui Yang and the
96. Bekenstein, R. et al. Quantum metasurfaces with atom arrays. other, anonymous, reviewer(s) for their contribution to the peer
Nat. Phys. 16, 676–681 (2020). review of this work.
97. Chen, Y. et al. Metasurface integrated monolayer exciton
polariton. Nano Lett. 20, 5292–5300 (2020). Reprints and permissions information is available at
98. Jha, P. K., Shitrit, N., Ren, X., Wang, Y. & Zhang, X. Spontaneous www.nature.com/reprints.
exciton valley coherence in transition metal dichalcogenide
monolayers interfaced with an anisotropic metasurface. Phys. Publisher’s note Springer Nature remains neutral with regard to
Rev. Lett. 121, 116102 (2018). jurisdictional claims in published maps and institutional affiliations.
99. Xu, L. et al. Enhanced light-matter interactions in dielectric
nanostructures via machine-learning approach. Adv. Photonics 2, Springer Nature or its licensor (e.g. a society or other partner) holds
026003 (2020). exclusive rights to this article under a publishing agreement with
100. Chen, M. K., Liu, X., Sun, Y. & Tsai, D. P. Artificial intelligence in the author(s) or other rightsholder(s); author self-archiving of the
meta-optics. Chem. Rev. 122, 15356–15413 (2022). accepted manuscript version of this article is solely governed by the
101. Deng, Y., Ren, S., Fan, K., Malof, J. M. & Padilla, W. J. Neural-adjoint terms of such publishing agreement and applicable law.
method for the inverse design of all-dielectric metasurfaces. Opt.
Express 29, 7526–7534 (2021). © Springer Nature Limited 2022

Nature Photonics | Volume 17 | January 2023 | 26–35 35