A Review of Metasurfaces - Physics and Applications

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/303521094
A review of metasurfaces: Physics and applications
Article in Reports on Progress in Physics · June 2016

DOI: 10.1088/0034-4885/79/7/076401
CITATIONS READS
1,838 7,798
3 authors, including:
Hou-Tong Chen Nanfang yu

Los Alamos National Laboratory Columbia University
211 PUBLICATIONS 18,268 CITATIONS 202 PUBLICATIONS 33,199 CITATIONS
SEE PROFILE SEE PROFILE
All content following this page was uploaded by Hou-Tong Chen on 28 June 2016.
The user has requested enhancement of the downloaded file.

Home Search Collections Journals About Contact us My IOPscience
A review of metasurfaces: physics and applications
This content has been downloaded from IOPscience. Please scroll down to see the full text.
2016 Rep. Prog. Phys. 79 076401
(http://iopscience.iop.org/0034-4885/79/7/076401)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 192.12.184.7
This content was downloaded on 24/06/2016 at 18:17
Please note that terms and conditions apply.

Reports on Progress in Physics
Rep. Prog. Phys. 79 (2016) 076401 (40pp) doi:10.1088/0034-4885/79/7/076401
Review
A review of metasurfaces: physics and

applications
Hou-Tong Chen1, Antoinette J Taylor2 and Nanfang Yu3
1
Center for Integrated Nanotechnologies, Los Alamos National Laboratory, Los Alamos, NM 87545, USA
2
Associate Directorate for Chemistry, Life, and Earth Sciences, Los Alamos National Laboratory,
Los Alamos, NM 87545, USA
3
Department of Applied Physics and Applied Mathematics, Columbia University, New York,
NY 10027, USA
E-mail: [email protected], [email protected] and [email protected]
Received 15 December 2015, revised 10 March 2016

Accepted for publication 14 March 2016
Published 16 June 2016
Invited by Masud Mansuripur
Abstract
Metamaterials are composed of periodic subwavelength metal/dielectric structures
that resonantly couple to the electric and/or magnetic components of the incident
electromagnetic fields, exhibiting properties that are not found in nature. This class of
micro- and nano-structured artificial media have attracted great interest during the past 15
years and yielded ground-breaking electromagnetic and photonic phenomena. However,
the high losses and strong dispersion associated with the resonant responses and the use
of metallic structures, as well as the difficulty in fabricating the micro- and nanoscale 3D
structures, have hindered practical applications of metamaterials. Planar metamaterials
with subwavelength thickness, or metasurfaces, consisting of single-layer or few-layer
stacks of planar structures, can be readily fabricated using lithography and nanoprinting
methods, and the ultrathin thickness in the wave propagation direction can greatly suppress
the undesirable losses. Metasurfaces enable a spatially varying optical response (e.g.
scattering amplitude, phase, and polarization), mold optical wavefronts into shapes that
can be designed at will, and facilitate the integration of functional materials to accomplish
active control and greatly enhanced nonlinear response. This paper reviews recent progress
in the physics of metasurfaces operating at wavelengths ranging from microwave to
visible. We provide an overview of key metasurface concepts such as anomalous reflection
and refraction, and introduce metasurfaces based on the Pancharatnam–Berry phase and
Huygens’ metasurfaces, as well as their use in wavefront shaping and beam forming
applications, followed by a discussion of polarization conversion in few-layer metasurfaces
and their related properties. An overview of dielectric metasurfaces reveals their ability to
realize unique functionalities coupled with Mie resonances and their low ohmic losses. We
also describe metasurfaces for wave guidance and radiation control, as well as active and
nonlinear metasurfaces. Finally, we conclude by providing our opinions of opportunities and
challenges in this rapidly developing research field.
Keywords: metamaterials and metasurfaces, wavefront shaping and beam forming,

polarization, active metasurfaces, nonlinearity, dielectric metamaterials,
surface and guided waves
0034-4885/16/076401+40$33.00 1 © 2016 IOP Publishing Ltd Printed in the UK

Rep. Prog. Phys. 79 (2016) 076401 Review
1. Introduction metasurfaces and can be considered as the two-dimensional

(2D) equivalent of bulk metamaterials. Because the subwave-
Optical devices control and manipulate light by altering its length thickness introduces minimal propagation phase, the
amplitude, phase, and polarization states in a desired manner, effective permittivity, permeability and refractive index are of
which result in steering the beam propagation direction, shap- less interest in metasurfaces. In contrast, of significant impor-
ing the wavefront (e.g. focusing), and imparting information tance are the surface or interface reflection and transmission
for applications such as sensing, imaging and communication. resulting from the tailored surface impedance, including their
Conventional optical components are based on refraction, amplitude, phase, and polarization states. The ultrathin thick-
reflection, absorption, and/or diffraction of light, and light ness in the wave propagation direction can greatly suppress
manipulation is achieved via propagation through media of the undesirable losses by using appropriately chosen mat
given refractive indices, which can be engineered to control erials and metasurface structures. Overall, metasurfaces can
the optical path of light beams. In this way, phase, amplitude, overcome the challenges encountered in bulk metamaterials
and polarization changes are accumulated through propaga- while their interactions with the incident waves can be still
tion in optical components based on refraction and reflection, sufficiently strong to obtain very useful functionalities. For
such as lenses, waveplates, and optical modulators. Ancient this reason, we envision that metasurfaces will dominate the
people already used ice lenses to focus sunlight and start fires general field of metamaterials research given their high poten-
[1], one example of controlling light propagation. They still tial in applications.
prevail in today’s optical laboratories and many consumer- Metasurfaces diminish our dependence on the propaga-
based optical products, but are bulky and heavy, unsuitable tion effect by introducing abrupt changes in optical properties
for the increasingly demanding requirements of integration [12–14]. At microwave and terahertz (THz) frequencies, one
and miniaturization in modern electromagnetic and photonic can take advantage of subwavelength metallic resonators such
systems. The propagation effect is also used in transformation as split-ring resonators (SRRs) [15, 16] and a variety of elements
optics [2, 3], which utilizes optical materials structured on a typically used in frequency selective surfaces [17]. Abrupt and
subwavelength scale to produce spatially varying refractive controllable changes of optical properties are achieved by engi-
indices that can range from positive to negative. neering the interaction between light and an array of optical
Metamaterials are composed of periodic subwavelength scatterers called ‘optical antennas’ [18, 19], which can take a
metal/dielectric structures (i.e. meta-atoms) that resonantly variety of forms, including metallic or dielectric micro/nano-
couple to the electric or magnetic or both components of particles [20, 21], apertures formed in metallic films [22, 23],
the incident electromagnetic fields, exhibiting effective elec- and their multi-layer structures [24]. The most critical feature
tric (represented by electric permittivity ε) and/or magnetic of metasurfaces is that they provide degrees of freedom in
(represented by magnetic permeability μ) response not found designing spatial inhomogeneity over an optically thin interface.
in nature. This class of micro- and nano-structured artificial Arrays of antennas with subwavelength separation between
media have attracted great interest during the past 15 years adjacent elements can have spatially varying structural features
and yielded ground-breaking electromagnetic and photonic or material compositions. Thus, metasurfaces are able to intro-
phenomena [4, 5]. Their electromagnetic properties are duce a spatially varying electromagnetic or optical response (i.e.
mainly determined by the subwavelength structures together scattering amplitude, phase, and polarization), and mold wave-
with the integrated functional materials, therefore, producing fronts into shapes that can be designed at will.
the desirable electromagnetic response and device function- As metasurfaces comprise a rapidly growing field of
alities by structural engineering. The initial overwhelming research, there have been a few good review articles focus-
interest in metamaterials lies in the realization of simultane- ing on different areas [25–31]. Here we provide our perspec-
ously negative electric and magnetic responses and, thereby, tive on this research field by reviewing the progress during
negative refractive index [6, 7], which can be used to accom- the past few years, where metasurfaces are broadly defined
plish super-resolution in optical imaging [8, 9]. The capability as planar metamaterial structures with subwavelength thick-
of tailoring inhomogeneous and anisotropic refractive index ness, operating at wavelengths ranging from microwave to
resulted in electromagnetic invisibility [10], another hallmark visible. The paper is organized as described below. In sec-
accomplishment using metamaterials. These promising poten- tion 2 we overview the concept and provide demonstrations
tial applications are, however, hindered in practice due to the of anomalous reflection and refraction, which have largely
high losses and strong dispersion associated with the resonant stimulated and reformed worldwide research interest in meta-
responses and the use of metallic structures. surfaces. In section 3 we introduce metasurfaces based on
Another challenge in metamaterials is the difficult micro- the Pancharatnam–Berry phase and Huygens’ metasurfaces,
and nano-fabrication of the required three-dimensional (3D) as well as their use in wavefront shaping and beam forming
structures [11], as permittivity, permeability and refractive applications. In section 4 we review polarization conversion
index are essentially properties of bulk materials. Planar met- in few-layer metasurfaces and their related properties. This
amaterials, however, can be readily fabricated using existing section is followed by an overview in section 5 of dielectric
technologies such as lithography and nanoprinting methods, metasurfaces that not only reduce the ohmic losses in metal-
driving many metamaterial researchers to focus on single- lic metasurfaces but also realize some other unique proper-
layer or few-layer stacks of planar structures that are more ties and functionalities. In section 6 we describe metasurfaces
accessible particularly in the optical regime. They are called for wave guidance and radiation control. We also summarize
2
active and nonlinear metasurfaces in section 7, and in the last

section we provide concluding remarks and an outlook on
future research directions.
2. Anomalous reflection and refraction
2.1. Generalized laws of reflection and refraction
When a plane electromagnetic wave encounters a boundary

between two homogeneous media with different refractive
indices, it is split into a reflected beam that propagates back
to the first medium and a transmitted beam that proceeds into
the second medium. The reflection and transmission coeffi-
cients and their directions are determined by the continuity of
field components at the boundary, and are given by the Fresnel
equations and Snell’s law, respectively. If we add to the inter-
face an array of subwavelength resonators of negligible thick-
ness forming a metasurface, the reflection and transmission
coefficients will be then dramatically changed because the
boundary conditions are modified by the resonant excitation
of an effective current within the metasurface. The reflection
and transmission waves carry a phase change that can vary
from −π to π, depending on the wavelength of the incident
wave relative to the metasurface resonance. When the resona-
Figure 1. A gradient of interfacial phase jump dΦ/dr provides an
tors are anisotropic, the polarization state may be also altered.
effective wavevector along the interface that can bend transmitted
When the phase change is uniform along the interface, the and reflected light into arbitrary directions.
directions of reflection and refraction are unaltered; in con-
trast, one of the merits provided by metasurfaces is that we where the definition of angles is shown in figure 1, and dΦ/dx
can create spatial phase variation with subwavelength resolu- and dΦ/dy are, respectively, the components of the phase gra-
tion to effectively control the direction of wave propagation dient parallel and perpendicular to the plane of incidence.
and the shape of wavefront. Looking at the problem from an alternative point of view,
We can understand quantitatively the control of wave prop- the interfacial phase gradient functions as an effective wave-
agation direction using Fermat’s principle, which states that vector along the interface, and is imparted to the transmitted
the route for the propagation of a light beam should be sta- and reflected waves. The above generalized laws can thus be
tionary in the total accumulated phase with respect to small derived by considering the conservation of wavevector along
variations of the route. Now we consider a specific case where the interface. These generalized laws indicate that the trans-
a metasurface introduces a spatial distribution of phase jumps mitted and reflected light beams can be bent in arbitrary direc-
due to electromagnetic scattering at its constitutive antennas. tions in their respective half space, depending on the direction
The actual optical path in the presence of these phase jumps and magnitude of the interfacial phase gradient, as well as the
should be stationary in the total accumulated optical phase. refractive indices of the surrounding media.
This law of stationary phase ensures that wavelets starting
from a source point with the same initial phase will arrive at the
2.2. Demonstration of generalized optical laws
point of destination in phase after reflecting from or transmit-
ting through the metasurface, and thus constructively interfere, To experimentally demonstrate the generalized laws, one
which makes the route a physical path of optical power. A set has to devise miniature scatterers that are able to controlla-
of generalized laws of refraction and reflection can be derived bly change the phase of the scattered waves and to place such
from Fermat’s principle of stationary phase [12, 13, 32]: scatterers into an array, forming an artificial interface. The
⎧ 1 dΦ
scattering amplitudes should be the same for all scatterers and
⎪ n sin(θt ) − n i sin(θi ) =
⎪ t
the spacing between neighboring scatterers in the array should
⎨
k 0 dx be much less than the wavelength. These conditions ensure
(1)
⎪ cos(θ ) sin(ϕ ) = 1 d Φ
⎪
that the superposition of spherical waves emanating from the
⎩
t t
n tk 0 dy scatterers gives rise to refracted and reflected waves with pla-
nar wavefronts, following Huygens’ principle.
⎧ 1 dΦ One approach to design the phase response of scatterers is
⎪
⎪
sin(θr ) − sin(θi ) = to use antenna dispersion. That is, for a fixed electromagnetic
⎨
n ik 0 dx
(2) wavelength and a variation of antenna geometries, or for a fixed
⎪ cos(θ ) sin(ϕ ) = 1 dΦ
⎪ antenna geometry and a variation of excitation wavelengths,
⎩
r r
n rk 0 dy there is an associated phase variation of the waves scattered
3
Figure 2. (a) SEM image of a mid-infrared metasurface consisting of an array of V-shaped gold optical antennas patterned on a silicon
wafer, with the unit cell highlighted and Γ = 11 μm. It creates a constant gradient of phase jump along the metasurface for the control of the
propagation direction of light transmitted through or reflected from the metasurface. (b) Under normal incidence, measured far-field intensity
profiles show the ordinary (co-polarized) and anomalous (cross-polarized) refraction generated by metasurfaces like the one shown in (a) and
with different interfacial phase gradients (from 2π /13 μm to 2π /17 μm). The far-field profiles are normalized with respect to the intensity
of the ordinary beams located at θt = 0°. The arrows indicate the calculated angular positions of the anomalous refraction according to
θt = −arcsin(λ /Γ). (c) A metasurface used to demonstrate generalized laws of reflection and refraction in the near-infrared. Upper panel depicts
one unit cell of the fabricated structure and lower panel reveals a schematic of the metasurface. (d) Measured far-field intensity profiles of the
metasurface in (c) showing reflection angle θr versus wavelength for cross-polarized light with 65° incidence angle. (a) used with permission
from [12], reprinted with permission from AAAS, (b) used with permission from [33], copyright 2012 by American Chemical Society, (c) and
(d) Reprinted with permission from [13] reprinted with permission from AAAS.
from the antenna (there are also associated amplitude and polar- antenna resonances, or geometric effects (see discussion of the
ization changes that can be utilized or otherwise managed for Pancharatnam–Berry phase in section 3.1) are able to extend the
designing metasurfaces). For example, when a beam of light phase response to cover the entire 2π range, which is necessary
impinges on a metallic optical antenna, the optical energy is for complete control of the wavefront. In addition to metallic
coupled into surface electromagnetic waves propagating back antennas, dielectric ones are also able to introduce phase varia-
and forth along the antenna surface. These are accompanied by tions to the scattered light associated with Mie resonances (i.e.
charge oscillations inside the antenna. These coupled surface establishment of standing wave patterns in the dielectric anten-
electromagnetic waves and oscillating charges are known as sur- nas, see the discussion of dielectric metasurfaces in section 5).
face plasmons. For a fixed excitation wavelength, the antenna Generalized optical laws were first demonstrated using
resonance occurs when the antenna length L res ≈ λ /2, where λ V-shaped optical antennas in the mid-infrared spectral range
is the surface plasmon wavelength; under this condition the inci- [12] and later confirmed in the near-infrared [13] (see figure 2).
dent electromagnetic wave is in phase with the excited antenna Such optically anisotropic antennas support two plasmonic
current. When the antenna length is smaller or larger than L res, eigenmodes with different resonant properties. The geometry
the current leads or lags the driving field. Therefore, the phase and orientation of antennas in the array are properly chosen so
of the antenna current and the phase of electromagnetic waves that for an incident wave at around 8 μm wavelength, over a wide
created by the oscillating current (i.e. scattered waves from the range of incident angles, and polarized along the x-axis (see fig-
antenna) can be controlled by choosing the appropriate antenna ure 2(a)), the component of the scattered wave polarized along
length. The tuning range of phase is up to π if a single antenna the y-axis has an incremental phase of π /4 between adjacent
resonance is involved. Multiple independent resonances, coupled V-antennas in the unit cell of the metasurface. The amplitude of
4
the component polarized along the y-axis is also tuned to be uni- A completely different approach to introducing phase jumps
form across the antenna array. The antenna spacing is between is to use the so-called Pancharatnam–Berry phase [40, 41].
one tenth and one fifth of the free space wavelength. The metas- The latter is associated with polarization change and can be
urface creates anomalously refracted and reflected beams satis- created by using anisotropic, subwavelength scatterers with
fying the generalized laws over a wide wavelength range, with identical geometric parameters but spatially varying orien-
negligible spurious beams and optical background, as shown in tations. The recent development of metasurfaces based on
figures 2(b) and (d). The broadband performance is due to the Pancharatnam–Berry phase has been largely following the
fact that the two eigenmodes supported by the V-antennas form a innovative early works by Hasman and co-workers [42], who
broad effective resonance over which the scattering efficiency is used continuous or discrete subwavelength gratings to con-
nearly constant and the phase response is approximately linear trol the polarization states for the generation of vector beams
[33, 34]. The scalability of metasurfaces allows the extension of and manipulation of wavefronts. The easiest way to reveal the
this concept to other frequency ranges, e.g. broadband anoma- relation between polarization and phase is to use Jones cal-
lous refraction was also demonstrated at THz frequencies using culus [43, 44]. In general, the Jones matrix of an anisotropic
C-shaped metallic resonators [35]. scatterer can be written as
The generalized law of reflection has also been demon-
⎛t 0 ⎞
strated using reflect-arrays [14, 36, 37], which consist of Mˆ = Rˆ (−α)⎜ o ⎟ Rˆ (α),
⎝ 0 te ⎠
(3)
metallic antennas separated from a back metallic plane by a
thin layer of dielectric material (see figure 3). Such reflect-
array metasurfaces are inspired by initial work on micro- where to and te are, respectively, the coefficients of forward
wave and millimeter wave reflect-array antennas [38, 39]. scattering for incident light linearly polarized along the two
Figure 3(a) shows a near-infrared reflect-array metasurface principal axes of the anisotropic scatterer,
based on patch antennas and figure 3(c) shows one that is ⎛ cos(α) sin(α) ⎞
Rˆ (α) = ⎜ ⎟
⎝−sin(α) cos(α) ⎠
based on H-shaped antennas for microwaves. The essence (4)
of reflect-arrays is to use antennas coupled with their dipo-
lar images in the back mirror to achieve a phase coverage of is the rotation matrix and α is the rotation angle. Given an
2π. Ideally, all incident power will be coupled into anomalous incident wave of right/left circular polarization EIR / L, the scat-
reflection, which will have the same polarization as that of the tered light from the anisotropic scatterer in the forward direc-
incident light; the transmission and specular reflection will be
tion ERT / L can then be written as [45]:
absent. Experimentally demonstrated efficiency in generating
anomalous reflection in reflect-array metasurface is as high ER/L ˆ R/L
T = M ⋅ EI
as 80%, significantly higher than the initial proof-of-principle
t +t t −t
demonstrations in figure 2, which are based on a single = o e EIR/L + o e exp(im 2α)EIL/R. (5)
antenna layer, rely on polarization rotation to achieve the 2π 2 2
phase coverage, and have an efficiency of 10–20%. The first term represents circularly polarized scattered
Figure 3(b) shows three regimes of operation for a reflect- waves with the same handedness as the incident light, and the
array metasurface: negative angle of reflection (θr and θi of dif- second term represents circularly polarized scattered waves
ferent signs), angle of incidence and angle of reflection of the with opposite handedness and an additional Pancharatnam–
same sign but not equal to each other, and coupling of incident Berry phase of m2α, where m is ‘−’ for right-handed and ‘+’
light into evanescent waves propagating on the metasurface for left-handed circularly polarized incident light. The second
(θr beyond 90°). In the last case, the interaction between the component can be selected in experiments by using a quar-
incident light and the metasurface leads to a lateral wavevec- ter-wave plate and a polarizer. Its phase can cover the entire
tor that is larger than the free space wavevector; as a result, 2π range if the anisotropic scatterer is rotated from 0 to 180°.
no reflection exists and the incident optical power can only be Based on this principle, a phase-gradient metasurface has been
coupled into surface waves. The work shown in figures 3(c) demonstrated to steer light into different directions depending
and (d) confirms the existence of such surface waves by exper on the handedness of the incident circular polarization (see
imentally measuring their near-field characteristics. A num- figure 4(a)) [45]. The unit cell of the metasurface consists
ber of variations of the reflect-array metasurface have been of U-shaped aperture antennas with an incremental angle of
also demonstrated. For example, birefringent reflect-array rotation between adjacent elements, with the total rotation
metasurfaces that steer incident light into different directions angle being 180° within the unit cell. Similar U-shaped aper-
according to its polarization state have been demonstrated in ture antennas have been used to create a planar lens, which
simulations [37] (see figures 3(e) and (f)). either functions as a focusing or diverging lens depending
on the handedness of the incident circular polarization [45],
as shown in figure 4(b). A broadband phase plate generating
3. Arbitrary phase gradient and beam forming
optical vortex beams has been demonstrated by using an array
3.1. Metasurfaces based on Pancharatnam–Berry phase of rod antennas with different orientations (figures 4(c) and
(d)) [46]. A bi-layer metallic aperture metasurface was also
In the previous examples, variations in phase or amplitude demonstrated to accomplish the simultaneous manipulation of
response are introduced by the dispersion of antenna resonance. polarization and phase of the transmitted light [47].
5
Figure 3. (a) Schematic of a near-infrared reflect-array metasurface consisting of gold patch antennas separated from a gold back plane
by a MgF2 spacer with subwavelength thickness. Left inset shows a basic building block, and right inset is an SEM image of part of the
metasurface. (b) Anomalous reflections at different incident angles for the metasurface shown in (a). The shaded quadrant indicates ‘negative’
reflection. (c) Photograph of a fabricated microwave reflect-array consisting of H-antennas separated from a metallic back plane by a dielectric
spacer. The reflect-array introduces an interfacial phase gradient ξ = 1.14k 0, where k0 is the wavevector of the incident beam corresponding to
a wavelength of 2 cm. (d) Scattering phase profile from the metasurface in (c) showing the phase gradient along the x-direction. (e) Schematic
of part of a birefringent reflect-array metasurface working at λ = 8.06 μm. (f) Depending on the polarization of the incident light, the phase
gradient is either positive or negative along the x-direction for the metasurface in (e), leading to polarization-dependent anomalous reflection.
(a) and (b) reprinted by permission from [36] copyright 2012 by American Chemical Society, (c) and (d) used with permission from Macmillan
Publishers Ltd: Nature Materials [14], copyright 2012, (e) and (f) used with permission from [37] copyright 2013 the Optical Society.
The metasurfaces based on the Pancharatnam–Berry phase is no wavefront distortion resulting from antenna dispersion.
work for circularly polarized incident light and control the The operating bandwidth is limited on the long-wavelength
component of the circularly polarized transmission with the side by reduced scattering efficiency and on the short-wave-
opposite handedness. A major advantage of the approach based length side by the requirement that the wavelength has to be at
on the Pancharatnam–Berry phase is ultra-broadband perfor- least several times larger than the spacing between scatterers
mance: given a certain antenna geometry, the magnitude of (i.e. metasurface regime). In early demonstrations of broad-
the phase jump is only a function of the orientation angle of band metasurfaces based on Pancharatnam–Berry phase, the
the antenna and the sign of the phase jump is determined by presence of scattered waves that do not carry Pancharatnam–
the handedness of the incident circularly polarized light; there Berry phase inevitably decreases their efficiency. In a new
6
Figure 4. (a) Upper panel: schematic of the super-unit-cell of a metasurface consisting of an array of identical U-shaped apertures with
gradually increasing rotation angles. Lower panels: simulation rusults showing that the metasurface in (a) bends a circularly polarized
incident beam under normal incidence into left or right direction according to the handedness of the incident beam. (b) Upper panel: super-
unit-cell of a planar cylindrical lens consisting of an array of identical U-apertures with different orientations. Lower panels: schematics
and simulation results showing that the lens focuses right-handed circularly polarized transmission component when the incident light
is left-handed circularly polarized, and that the same lens defocuses left-handed circularly polarized transmission component when the
incident light is right-handed circularly polarized. (c) SEM image of a metasurface consisting of an array of gold rod antennas with
identical geometry but spatially varying orientations, which is designed for generating an optical vortex beam with L = 1 (incidence: right-
handed circularly polarized; detection: left-handed circularly polarized). (d) Measured intensity distribution of vortex beams generated
by the metasurface in (c) at different wavelengths from 670 to 1100 nm. (e) Pioncaré sphere used to derive the phase difference between
scattered waves of left-handed circular polarization from rod antennas located at points A and B in (c), with right-handed circularly
polarized incident light. (a) and (b) reproduced with permission from [45] Copyright 2012 by The Optical Society, (c) and (d) reproduced
with permission from [46], copyright 2012 American Chemical Society.
generation of metasurfaces, Luo et al were able to suppress because the antenna at point A on the metasurface preferen-
these scattered components and created metasurfaces based on tially scatters vertically polarized waves; the trace ends at the
Pancharatnam–Berry phase with efficiency approaching unity south pole because left-handed circularly polarized transmis-
[48]. They demonstrated two different metasurfaces that sepa- sion is selectively monitored. Similarly, the dashed red trace
rate a linearly polarized incident microwave into a left-handed in figure 4(e) corresponds to light passing through point B on
circularly polarized beam and a right-handed circularly polar- the metasurface shown in figure 4(c). The solid angle enclosed
ized beam over a frequency range of 11–14 GHz, within which by the two traces is π; therefore the phase difference between
the linearly polarized background is very weak. The metas- left-handed circularly polarized light scattering from points A
urface design is based on rigorous Jones matrix analyses that and B on the metasurface is π /2. Similar analyses show that
provide a set of criteria for achieving 100% efficiency [48]. the phase difference between points A and C is π and between
A completely different perspective to understand the A and D is 3π /2. Therefore, the metasurface in figure 4(c)
operation of metasurfaces based on the Pancharatnam–Berry introduces a constant phase gradient in the azimuthal direction
phase results from tracing the evolution of polarization in the and the phase variation is 2π during one circle around the cen-
Poincaré sphere [40, 49–51]. The phase difference between tral point of the metasurface. The metasurface thus imprints a
the scattered waves from any two points on the metasurface is spiral phase distribution to the transmitted wavefront, creating
equal to the solid angle enclosed by their corresponding traces a vortex beam with orbital angular momentum of L = 1.
in the Poincaré sphere divided by two [49]. For example,
the solid red trace in figure 4(e) corresponds to light passing
3.2. Huygens’ surfaces
through point A in figure 4(c): The trace starts at north pole
of the Poincaré sphere representing right-handed circularly To boost the efficiency of a metasurface in controlling the
polarized incident light; the trace passes a point on the equa- transmitted light, one has to match its impedance with that of
tor that represents linear polarization in the vertical direction free space. Complete elimination of reflection can be realized
7
Figure 5. (a) Upper panel: photograph of a fabricated microwave metasurface that can redirect an incident beam with nearly 100%
efficiency into a refracted beam. It is made of a stack of identical circuit board stripes, the top and bottom sides of which are printed with
copper traces. Bottom panel: one unit cell of the metasurface consists of capacitively and inductively loaded traces to realize desired electric
sheet reactance (on the top side of each stripe) and capacitively loaded loops to realize desired magnetic sheet reactance (on the bottom
side of each stripe). (b) Measured magnetic field distribution of the beam-deflecting metasurface in (a). (c) Schematic of an optically thin,
isotropic Huygens’ metasurface that efficiently refracts a normally incident beam at telecommunication wavelengths. Inset: schematic of a
unit cell. (d) Simulated electric field distribution of a beam deflector based on the metasurface in (c).(e) Left panel: basic building block of
a metasurface made of plasmonic (AZO: aluminum-doped zinc oxide) and dielectric (silicon) materials, with l = 250 nm and h = 250 nm.
Right panel: metatransmit-array made of three stacked metasurfaces with center-center distance of d = λ 0 /8 = 375 nm. (f) and (g)
Simulated electric field distributions of a beam deflector and a flat lens based on the metatransmit-array shown in (e). (a) and (b) reproduced
with permission from [52] copyright 2013 by the American Physical Society, (c) and (d) reproduced with permission from [53] copyright
2014 by the American Chemical Society, (e)–(g) used with permission from [54] copyright 2013 by the American Physical Society.
by controlling the surface electric and magnetic polarizabili- to free space and able to fully control the phase of the trans-
ties, αe and αm, of the metasurfaces so that [52] mitted light was proposed in a recent paper [54]. It is designed
based on optical nano-circuit concepts and is comprised of
αm /αe = η0,
(6) three planarized arrays stacked together, as shown in fig-
where η0 is the impedance of the surrounding media. The ure 5(e), where the building blocks of the array are subwave-
effective electric and magnetic surface currents, which are length components made of metallic and dielectric materials
proportional to αe and αm, respectively, change the bound- with different filling ratios and function as LC nano-circuit
ary conditions at the metasurface and lead to the new scat- elements. A beam deflector and a flat lens with high transmis-
tered wavefronts. The complex transmission coefficient of the sion efficiency were demonstrated in simulations, as shown
metasurface is [52] in figures 5(f) and (g), by engineering the effective surface
impedance of the metasurface via tuning of the filling ratios.
2 − jωαeη0
T=
(7) .
2 + jωαeη0
3.3. Wavefront shaping and beam forming
If αe is predominantly real, one can vary αe and αm simul-
taneously at each point on the metasurface to ensure that the Metasurfaces provide us with an unprecedented opportunity
waves transmitted through the metasurface acquire a phase to design the wavefronts of light at will, as we have seen from
jump anywhere from −π to +π according to (7) and that the the descriptions of anomalous reflection/refraction and beam
transmission efficiency is close to unity by satisfying (6) eve- focusing in the previous sections. Figure 6 further shows a few
rywhere on the metasurface. The above design concept has planar devices based on metasurfaces. To realize flat lenses, a
been implemented in the microwave spectral region by using metasurface should impose a phase profile
spatially varying copper traces supporting both electric and
2π
magnetic polarization currents (figure 5(a)) [52]. A transmis- ϕ L (x , y ) =
(8) ( x 2 + y2 + f 2 − f )
sion efficiency of 86% was experimentally demonstrated in a λ
beam deflector shown in figure 5(b). Although the demonstra- to convert incident planar wavefronts into spherical ones,
tions are in the microwave regime, the concepts can be adapted which converge at a distance f from the lenses. The optical
to the optical regime and one example is shown in figures 5(c) wavefronts in transmission or reflection remain spherical as
and (d) [53]. Another metasurface that is impedance matched long as the incident plane wave impinges normal to the flat
8
Figure 6. (a) Left panel: SEM image of a fabricated metasurface lens with 3 cm focal length, consisting of an array of V-antennas. Right
panel: phase profile of the lens discretized according to the phase responses of eight constituent antennas. Insets: zoom-in view of fabricated
antennas. (b) 3D plots of the simulated (top panel) and measured (middle panel) and 1D plots (cross-sectional planes along the lines) of
intensity distribution of the lens in (a) on the focal plane. (c) SEM image of a planar plasmonic metalens consisting of V-shaped apertures and
with a focal length f = 2.5 μm at an operational wavelength of 676 nm. (d) Intensity distributions for two cross-sectional planes (top panel)
cutting through the center of the metalens in (c), and on the focal plane of the metalens (bottom panel). (e) Schematic of a metasurface lens
consisting of an array of 21 × 21 scatterers each made of two silver concentric loops (gray). The scatterers are placed on both sides of the
substrate (yellow). (f) Simulated intensity distribution on the focal plane of the metasurface lens in (e). (a) and (b) used with permission from
[55] copyright 2012 by the American Chemical Society, (c) and (d) used with permission from Macmillan Publishers Ltd, [58] Light: Science
and Applications, copyright 2013, and (e) and (f) used with permission from [60] copyright 2011 by The Optical Society.
9
lenses. It is therefore straightforward to achieve high numer The most complex and general wavefront shaping is to cre-
ical-aperture (NA) focusing without spherical aberration. ate a holographic image in the far-field. Metasurfaces provide
Flat lenses at telecom wavelengths have been experimentally the degrees of freedom to engineer the local amplitude, phase,
demonstrated using V-shaped antennas (see figures 6(a) and and polarization response on an interface, and thus are a good
(b)) [55]. The efficiency of these flat lenses is, however, platform to realize all types of computer-generated holograms
rather small (i.e. 1% of the incident optical power is focused) (CGHs) (e.g. binary holograms, phase-only holograms, ampl
because of the use of only a single scatterer layer, the small itude and phase modulation holograms). Figure 7(a) shows a
surface filling factor, and focusing only the component of the metasurface consisting of arrays of nanoaperture antennas that
scattered light that is cross-polarized with respect to the inci- produce a spatially varying transmission coefficient [22]. By
dent polarization. At THz and microwave frequencies, high- utilizing the dispersion of aperture antennas, the metasurface
performance planar components can benefit from few-layer was designed to operate as two distinctive binary transmis-
metasurfaces, which have enabled highly efficient and ultra- sion holograms at two different wavelengths, λ1 = 905 nm and
broadband polarization conversion and anomalous refraction λ2 = 1385 nm. It creates a word ‘META’ shown in figure 7(b)
[56], and highly efficient reflect-array metasurface lenses at λ1 = 905 nm and a word ‘CGH’ shown in figure 7(c) at
[57]. V-shaped apertures allow similar control of scattering λ2 = 1385 nm in the far-field. In another metasurface holo-
polarization, amplitude and phase as in their complementary gram, V-shaped aperture antennas shown in figure 7(d) were
V-antennas according to Babinet’s principle; they have been used to introduce an eight-level phase distribution and a two-
used to demonstrate flat lenses to focus visible light (figures level amplitude distribution [66]. The amplitude and phase
6(c) and (d)) [58] and THz waves [59], with one of the advan- distributions approximated the required near-field amplitude
tages being significant suppression of the background light. and phase distributions on the metasurface plane, so that a
Based on the Pancharatnam–Berry phase, U-shaped and certain holographic image was obtained in the far-field, as
other nano aperture antennas have been used to create a planar shown in figure 7(f). Additionally, a reflect-array metasurface
lens, which either functions as a focusing or diverging lens that introduced a 16-level Pancharatnam–Berry phase has
depending on the handedness of the incident circular polar- been demonstrated to create complex holographic images in
ization (see figures 4(b)) [45, 61]. A flat lens design at tel- the far-field (figures 7(g)–(i)) [67]. The antenna-orientation-
ecom wavelengths with potentially high efficiency has been controlled Pancharatnam–Berry phase combined with the
demonstrated in simulations (figures 6(e) and (f)) [60]. The reflect-array design led to broadband performance and high
design uses concentric loop antennas placed on both sides of efficiency of the hologram. Experimentally demonstrated effi-
a substrate to enhance scattering efficiency and to increase ciency reaches 80% at λ = 825 nm and the hologram operates
the range of phase coverage. Flat lenses working in the near- between 630 nm and 1050 nm.
infrared with high efficiency have been demonstrated exper
imentally by using reflect-arrays of patch antennas [62]. Note 4. Polarization conversion
that except for spherical aberration, monochromatic aberra-
tions are still present in the above demonstrated flat lenses. Polarization state is an intrinsic property of electromagnetic
For example, when incident light is not perpendicular to the waves, and the conversion between polarization states is very
lenses, the transmitted or reflected wavefront is no longer often highly desirable (or even necessary) for many mod-
spherical because its phase distribution is that of (8) plus a lin- ern electromagnetic and photonic applications. For instance,
ear phase distribution introduced by the non-normal incidence in advanced communication and sensing, converting linear
angle. Flat lenses also have chromatic aberration, although polarization to circular polarization makes a beam resistant
one can design antennas with multiple resonances to elimi- to environmental variation, scattering and diffraction. During
nate it by engineering their dispersion [63] (see discussions recent years, conversion among polarization states using
in section 5). metasurfaces has attracted increasing interest due to their
It is of particular interest to focus Gaussian beams to non- design flexibility and compactness. The accompanied capabil-
diffracting and long focal depth Bessel beams that are tradi- ity of tuning a phase delay spanning the entire 2π range over
tionally generated using axicons and have been widely used a broad bandwidth and with a deep subwavelength resolution
in microscopy imaging. A metasurface approach to generate could potentially address some critical issues in the develop-
a Bessel beam (axicons) [55, 64] has been demonstrated by ment of flat optics.
creating a linear phase gradient along the radial direction of Highly symmetric simple meta-atoms can be advanta-
the metasurface. An arbitrary spatially varying phase profile geous in maintaining polarization states. Breaking the sym-
can be created in the azimuthal direction. A V-shaped antenna metry can, however, provide additional degrees of freedom
based metasurface has been used to create a vortex beam (i.e. to achieve customized functionality that enables the manipu-
Laguerre–Gaussian modes) from a Gaussian beam [12, 65], lation of polarization states. Through tailoring the two eigen-
resulting in optical singularity at the beam center and a heli- modes corresponding to orthogonal linear polarizations, it is
coidal equal-phase wavefront carrying orbital angular momen- possible to have equal transmission magnitude but a relative
tum. Using a phase profile based on the Pancharatnam–Berry phase delay ∆φ at a specific frequency. Narrowband polari-
phase, a broadband phase plate generating optical vortex zation conversions between linear and circular polarization
beams has been demonstrated using an array of rod antennas states (∆φ = π /2, quarter wave plates), or linear polarization
with different orientations (see figures 4(c) and (d)) [46]. rotation (∆φ = π, half wave plates) have been realized using
10
Figure 7. Metasurface holograms. (a) SEM image of part of a metasurface hologram consisting of nanoaperture antennas. Different colors
represent pixels with distinctive transmission coefficients. (b) and (c) Transmitted light intensity of the metasurface in (a) recorded in the
far-field at λ1 = 905 nm and λ2 = 1385 nm, respectively. (d) SEM image of a fabricated metasurface for generating a holographic image
of the letter ‘P’. Inset: zoomed-in view of the hologram. (e) Simulated and (f) measured holographic image created by the metasurface
holograms similar to (d) with an eight-level phase modulation and a two-level amplitude modulation. (g) One-pixel cell structure of a
nanorod-based hologram. The nanorod can rotate in the x–y plane with an orientation angle φ to create different Pancharatnam–Berry phase
delays. (h) 16-level phase distribution of the nanorod-based hologram (100 × 100 pixels shown). (i) Experimentally obtained image in the
far field created by the nanorod-based hologram at 632.8 nm wavelength. (a)–(c) used with permission from [22] copyright 2012 by John
Wiley and Sons, (d)–(f) used with permission from Macmillan Publishers Ltd, [66] Nature Communications copyright 2013, and (g)–(i)
used with permission from Macmillan Publishers Ltd, Nature Nanotechnology [67] copyright 2015.
single-layer metasurfaces [68–73] or multi-layer cascading For circularly polarized incident fields, it becomes
metasurfaces [74–76] operating from microwave to optical
⎛ E t ⎞ ⎛T++ T+−⎞⎛ E i ⎞
⎜ + ⎟ = ⎜⎜ ⎟⎜⎜ + ⎟⎟
frequencies. However, the efficiency is limited, in general, up
⎟ i
⎝ E− ⎠ ⎝ ⎠⎝ E − ⎠
to 50% with a bandwidth comparable to a meanderline quarter t T −+ T −−
wave plate [68, 77]. The low level of polarization conversion
⎛ Ei ⎞

= Tˆcirc⎜⎜ + ⎟⎟,
efficiency can be addressed by the implementation of few-
(10)
⎝ E−i ⎠
layer metasurfaces.
Following the Jones matrix description [43, 44] the trans-
mission of linearly polarized incident fields (Ex, E y ) through a
metasurface can be described as where T±± = 12 (Txx + Tyy ) ± 2i (Txy − Tyx ) and T±∓ = 12 (Txx − Tyy )
i
∓ 2 (Txy + Tyx ). Under normal incidence and in general, x and
⎛ E t ⎞ ⎛Txx Txy ⎞⎛ E i ⎞
⎜⎜ tx ⎟⎟ = ⎜ ⎟⎜⎜ i ⎟⎟
x y directions do not necessarily coincide with the structure’s
⎝ ⎠
E y ⎝ T yx T yy ⎠ ⎝ y⎠
E principal axes. There are a few properties of Jones matrices
that are related to metasurface structural symmetries:
⎛ Ei ⎞

= Tˆlin⎜⎜ i ⎟⎟,
x
(9)
⎝ y⎠
E (i) All components in the Jones matrices could be different if
the metasurface lacks reflection or rotational symmetries;
11
When designing metasurfaces for polarization conversion

between the same kinds (x and y linear polarizations or left-
and right-handed circular polarizations), we need to maxi-
mize the off-diagonal components of the Jones matrices. For
the conversion between linear and circular polarizations, the
metasurfaces need to enable π /2 phase difference between the
orthogonal components.
4.1. Linear-to-circular polarization conversion
An antenna array backed with a ground plane has been widely

exploited at microwave frequencies to enhance the radiation
efficiency and beam directionality. This configuration also
enhances the polarization conversion in reflection for aniso-
tropic subwavelength metallic resonator arrays. Early work at
microwave frequencies demonstrated that narrowband conver-
sion to various polarization states, including linear-to-circular
polarization and linear polarization perpendicular to the inci-
dent one, is possible depending on the structural parameters,
incident angle, and frequency [78]. It has also been shown that
a pair of perpendicularly oriented and detuned electric dipoles
(e.g. rectangular, elliptical, squeezed cross resonators, etc) can
be used to manipulate polarization states including the con-
struction of quarter-wave plates operating in reflection at optical
wavelengths [79, 80]. This type of structure is similar to those
widely used in metamaterial perfect absorbers [81], where the
Fabry–Pérot-like interference plays an important role [82].
New device functionalities could be realized by control-
ling spatial distribution of polarization response using meta-
surfaces. Figure 8 show a metasurface-based quarter-wave
plate [33] that generates high-quality circularly polarized
light (degree of circular polarization or ellipticity >0.97) over
a broad wavelength range (λ = 5–12 μm) (figure 8(c)). The
unit cell of the metasurface comprises two subunits (colored
pink and green in figures 8(a) and (b)). Upon excitation by
linearly polarized incident light, the subunits generate two
Figure 8. (a) Schematic of a metasurface quarter-wave plate,
co-propagating waves with equal amplitudes, orthogonal
with the unit cell of the metasurface consisting of two subunits
(pink and green). Each subunit contains eight V-antennas. Upon linear polarizations, and a π /2 phase difference (when offset
excitation by linearly polarized incident light, the subunits generate d = Γ/4), which produce a circularly polarized anomalously
two copropagating waves with equal amplitudes, orthogonal linear refracted beam that bends away from the surface normal.
polarizations, and a π /2 phase difference (when offset d = Γ/4), By increasing the number of layers to two or three, the near
which produce a circularly polarized anomalous refraction that is
field or Fabry–Pérot-like coupling can significantly enhance
separated from the normal beam. (b) SEM image of a portion of
the fabricated metasurface quarter-wave plate with a footprint of the efficiency of linear-to-circular polarization conversion as
230 × 230 μm2 to accommodate the plane-wave like incident light. well as the operation bandwidth. This property is realized in
Antenna orientation angles are indicated by β1 and β 2, and dashed the few-layer metasurface structures illustrated in figures 9 and
lines represent the antenna symmetry axes. (c) Calculated degree 10. An ABA-type, anisotropic tri-layer metasurface, shown in
of circular polarization and intensity of the anomalously refracted
figure 9(a), has enabled narrowband, highly efficient linear-to-
beam as a function of wavelength, showing the broadband and high
efficiency properties of the quarter-wave plate. Used with permission circular polarization conversion in transmission at microwave
from [33] copyright 2012 by the American Chemical Society. frequencies [83]. Here layer A is an electric metasurface with
periodically arranged resonant microstructures, while layer B
is a metallic mesh. There are two mechanisms that are respon-
(ii) If the metasurface structure has a mirror symmetry, sible for transparency. The first one is the electromagnetic
Txy = Tyx and T++ = T−−, and if the incident linear polari- wave tunneling [84] (a mechanism that is essentially equiva-
zation is further parallel or perpendicular to the symmetry lent to Fabry–Pérot-like resonance [85]), and the second one
plane, Txy = Tyx = 0; is the extraordinary optical transmission (EOT) of layer B that
(iii) For metasurface structures with a C4 or C3 rotational is mediated by the periodic structure of layer A [83]. Through
symmetry with respect to the z-axis, we have Txx = Tyy, structural tailoring, these two transparency bands, corresp
Tyx = −Txy, and T+− = T−+. onding to the two orthogonal linear polarization directions
12
ellipticity is about 0.99, as shown in figure 10(b). Circular-to-

circular polarization conversion was demonstrated employing
a tri-layer metasurface designed through the approach devel-
oped by Pfeiffer and Grbic [88], with the unit cell illustrated
in figure 10(c). The measured and simulated Jones matrix of
the metasurface [89], shown in figure 10(d), reveals a high
transmittance of 50% for right-handed to left-handed circu-
lar polarization conversion, while all other components in the
Jones matrix are below 2.5%, suggesting an extinction ration
of ∼ 20 : 1 at the designed wavelength of 1.5 μm. It was also
observed that the circular-to-circular polarization conversion
extends over a quite broad wavelength range.
4.2. Linear polarization rotation
Planar chiral response can yield optical activity, rotating the

direction of linear polarization. While the polarization rota-
tion power may significantly exceed naturally occurring mat
erials per unit thickness, typically it is insufficient to obtain
the desirable 90° polarization rotation. Increasing the number
of layers can yield half wave rotation; however, in general,
this approach cannot sustain the polarization rotation power
through increasing the number of layers by simple stacking,
due to the near-field coupling or interference of the multire-
flections. In the past, efficient linear polarization conversion
Figure 9. (a) Unit cell of a tri-layer ABA-type narrowband microwave
linear-to-circular polarization converter, with the transmission still employed anisotropic properties of metamaterials.
amplitude and phase shown in (b) at two orthogonal directions. The A simple structure is shown in figure 11(a) where an array
operation frequency is indicated by the dashed vertical line. Used of cut-wires was separated from the ground plane by a poly
with permission from [83] copyright 2011 by The Optical Society. imide spacer [56]. Under normal incidence, the incident x
polarized THz waves were converted to y polarized waves
(x and y), can overlap and, at the same time, have a phase in reflection with a conversion efficiency higher than 80%
difference of π /2, as shown in figure 9(b) at the frequency over an ultrabroad bandwidth, as shown in figure 11(b). The
indicated by the dashed vertical line. This means that an inci- co-polarized reflection approaches zero at several individual
dent electromagnetic wave linearly polarized at 45° has been frequencies where the destructive interference conditions
transformed to a circularly polarized one, with a conversion [56, 82] are largely satisfied, as illustrated in figures 11(c) and
efficiency greatly exceeding any single-layer metasurface. (d), in which the superposition seems to be responsible for the
Bi-layer metasurfaces have enabled high-efficiency and observed broadband performance. Following this concept, a
broadband conversions from linear to circular polarizations variety of metasurface structures, mostly at microwave fre-
[86, 87]. A bi-layer metasurface comprised of stacked and quencies, have been demonstrated to accomplish multi-band
twisted metallic wire grids shown in figure 10(a) was devel- and ultra broadband linear polarization conversion in reflec-
oped to operate at THz frequencies [86]. For normal incidence tion [90]; even at visible wavelengths the high efficiency can
and linearly polarized light in the x direction, the first wire be still largely maintained according to the simulation results
grid is aligned at 45° with respect to x direction. The wire grid in [91]. The observed linear polarization rotation is consistent
is designed such that the transmission amplitude of orthogo- with an earlier contribution using a similar structure to con-
nal components |txx| and |txy| are approximately constant and trol optical polarization in a reflection geometry [92], while
equal, while the linear phase retardance is frequency depend- the bandwidth was much improved and different theoretical
ent. This frequency dependent phase retardance is compen- models were used. In order to avoid the increasing metallic
sated through tailoring the geometric parameters of the second loss in the optical frequency range, dielectric metasurfaces for
wire grid, which also has simultaneously high transmission linear polarization conversion in reflection were also demon-
coefficients |txx| and |tyy|. The metallic grids were embedded strated, based on the same principle [93] (see discussions in
within a polyimide film so there are 4 interfaces: front air/ section 5).
polyimide, front wire grid, back wire grid, and back polyimide/ It is more desirable to have linear polarization converters
air. Through combining the multiple reflections due to these operating in the transmission mode. There have been a few bi-
interfaces and the dispersion of specially designed wire grids, layer or tri-layer metasurfaces demonstrated to realize cross
the overall output of the two orthogonal x and y components polarization conversion operating at a narrow single band or
have approximately equal amplitude and a phase delay of at multiple bands [94–96], where the polarization rotation
about 90°, resulting in circularly polarized transmission over is insensitive to the azimuthal angle of the incident polari-
a relatively broad bandwidth from 0.98 to 1.36 THz where the zation due to the use of structures with four-fold rotational
13
Figure 10. (a) Optical images with different levels of zooming for a fabricated bi-layer THz metasurface embedded within a polyimide
film. (b) Experimentally measured transmission amplitude, phase retardation, and ellipticity under horizontally polarized incidence, for the
sample shown in (a). (c) Unit cell of a tri-layer metasurface for circular-to-circular polarization conversion operating at the near infrared,
with simulated and measured transmittance shown in (d). (a) and (b) used with permission from [86] copyright 2014 by John Wiley and
Sons, (c) and (d) used with permission from [89] copyright 2014 by the American Physical Society.
symmetry. Figure 12(a) shows the anisotropic unit cell of a predict a conversion efficiency of 50% (cross-polarized trans-
bi-layer metasurface for 90° rotation of the THz linear polar- mission magnitude 0.71) and a polarization-conversion ratio
ization, consisting of a front array of asymmetric split-ring (PCR) up to 99.9% using lossless PET spacer, the exper
resonators (ASRR) for polarization conversion and a rear imental values realized are 23% (magnitude 0.48) and 97.7%,
array of S-shaped resonators (SR) for polarization selection respectively, at 1.04 THz as shown in figure 12(b), due to the
[97]. For the ASRR metasurface, the incident x-polarized THz significant loss within the PET spacer [97].
waves induces currents and forms a net electric dipole in the Increasing the conversion efficiency and/or bandwidth
y-direction, providing both x- and y-polarized components in becomes particularly interesting when metasurfaces are used
reflection and transmission. The SR metasurface, however, to realize a new class of flat optical components where the
exhibits negligible polarization conversion; it was tailored to transmission phase can be simultaneously controlled. An
have a resonance frequency coinciding with the ASRRs for intriguing example for linear polarization rotation is a tri-
x-polarized waves, allowing only y-polarized waves to pass layer THz metasurface demonstrated by Chen and co-work-
through and blocking the x-polarized waves. Due to the dis- ers [56]. It consists of a pair of identical gratings that are
persion of the metasurfaces and through carefully optimizing aligned in orthogonal directions, and an array of cut-wires
the PET spacer thickness, a Fabry–Pérot resonance occurs tilted at an azimuthal angle of 45°, as shown in figure 12(c).
within the ultrathin polarization rotator, which can enhance The front grating is transparent when the incident THz field
the polarization conversion efficiency exceeding that of the is linearly polarized along the x direction. As it continues
ASRR metasurface alone. Although numerical simulations to propagate and excite the cut-wires, the scattering results
14
Figure 11. Metasurface broadband polarization conversion in reflection. (a) Schematic metasurface structure. The incidence angle θi = 25°,
and the incident electric field E0 is linearly polarized in the x direction with an angle α = 45° with respect to the cut-wire orientation. (b) Exper
imentally measured co- and cross-polarized reflectance. (c) Cross- and (d) co-polarized multiple reflections theoretically calculated at 0.76 THz,
revealing the constructive and destructive interferences, respectively. Used with permission from [56]. Reprinted by permission from AAAS.
in both x and y polarized components. For forward scatter- broadband, high-efficiency linear polarization rotators.
ing, the back grating allows the newly generated y polarized Furthermore, in the structure shown in figure 12(c), the
component to pass through while blocking the x polarized transmission phase can be finely tuned to span an entire 2π
component; for back scattering, the front grating reflects the range and with subwavelength resolution through replacing
y polarized component and allows the x polarized component the cut-wires with a variety of anisotropic resonators with
to pass. This process continues due to a multireflection pro- varying geometric dimensions [56]. Combining this prop-
cess within this multi-layer structure. When the thicknesses erty and the high polarization conversion efficiency promises
of the polyimide spacer layers are carefully tuned, a con- great potential in wavefront control, resulting in a new class
structive interference enhances the polarization conversion of practical flat optical devices.
and a destructive interference of the co-polarized reflections A similar broadband THz polarization rotator was demon-
largely reduces the reflection loss (insertion loss) at multiple strated by Cong et al [98], where the middle cut-wire array
frequencies, as shown in figure 12(d), a mechanism simi- was replaced by a wire grating, as schematically shown in
lar to metamaterial antireflection coatings [85] and perfect figure 12(e). The formation of a Fabry–Pérot cavity makes
absorbers [82]. The back grating also guarantees a purely y this metasurface structure perform in remarkable contrast to
polarized output—there is practically no co-polarized trans- cascading wire polarizers with consecutive 45° rotation. The
mission. The overall result is that the x polarized incident latter does rotate the incident linear polarization by 90° but
THz waves can be completely converted to its orthogonal allows only up to 25% power transmission. This metasurface
y polarization, over a bandwidth exceeding 2 octaves and showed a conversion efficiency up to 85%, and the output
with a conversion efficiency up to 80%. Simply by scaling, waves exhibit extremely clean cross linear polarization over a
a variety of similar structures [99, 100] were employed in broad bandwidth, as shown in figure 12(f), although the trans-
the microwave and infrared frequency ranges to demonstrate mission phase cannot be controlled.
15
Figure 12. (a) Unit cell (left panel) and optical images (right panel) of a bi-layer polarization rotator, and (b) measured co- and cross-
polarized transmission coinciding with the simulated results, together with the polarization conversion ratio. (c) Schematic of the unit cell
of a tri-layer metasurface linear polarization converter and (d) cross-polarized transmittance obtained through experimental measurements,
numerical simulations, and theoretical calculations, together with the numerically simulated co-polarized reflectance. (e) Schematic of a
tri-layer metasurface polarization rotator consisting of three metallic gratings, and (f) experimental transmittance spectra. (a) and (b) used
with permission from [97] copyright 2013, AIP Publishing LLC, (c) and (d) used with permission from [56] reprinted with permission from
AAAS, (e) and (f) used with permission from [98] copyright 2013, AIP Publishing LLC.
4.3. Asymmetric transmission full-wave numerical simulations are used for their physical
realization. One such metasurface exhibiting strongly asym-
By reducing the structural symmetry and converting between
metric transmission of circularly polarized millimeter waves is
polarization states, metasurfaces have yielded a polarization
shown in figures 13(a) and (b) [88]. As shown in figure 13(c),
sensitive and asymmetric transmission with respect to the
the S21 parameter (i.e. transmission) is below −10 dB for ++,
direction of wave propagation [101]. Asymmetric polariza-
+−, and −−, and it is above −0.8 dB for −+, resulting in
tion conversion and transmission were observed in planar chi-
an asymmetric response of 0.99 over a bandwidth of 20% at
ral metasurfaces for circularly polarized incident fields with
f f f b the designed millimeter wavelengths. Similar behaviors were
T± ∓ ≠ T ∓ ± and T ± ∓ ≠ T ± ∓, where the superscripts ‘f’ and observed in tri-layer metasurfaces operating at near infrared
‘b’ denote the forward and backward propagation directions, wavelengths [89], as shown in figures 10(c) and (d).
f b b
respectively, though T ±± = T ±± and T ±f ∓ = T ∓ ± as required A variety of bi-layer metasurfaces have been also reported
by Lorentz Reciprocity Lemma. The planar chiral metasur- to exhibit asymmetric transmission for linearly polarized
faces are more transparent to a circularly polarized wave from incident light [105, 106]. Further developments showed that
one side than from the other side, with an experimentally mea- bi-layer metasurface structures can be used to demonstrate
sured transmission difference up to 40% at microwave [101] increased bandwidth of the asymmetric transmission in the
and 15% at visible [102, 103] frequencies. This effect is caused near infrared [107, 108]. It was shown that the interlayer
by the different efficiencies of polarization conversion in the alignment could have very little effect on the asymmetric
opposite propagation directions for lossy metasurfaces, in transmission [107], which indicates that the near-field cou-
remarkable contrast to the optical activity and Faraday effect. pling is negligible. This advantageous property is particularly
It implies that when circularly polarized light passes through useful in the optical regime where the interlayer alignment is
the metasurface and then retraces its path after reflection from challenging. In order to take full advantage of the asymmetric
a mirror, the final polarization state will be different from that transmission, it is necessary to suppress other components and
of the initial state [104]. only obtain a high contrast asymmetric component (e.g. tyx)
Bi-layer and multi-layer metasurfaces can increase the within the Jones transmission matrix [109–111]. Tri-layer
polarization conversion and consequently enhance the trans- metasurfaces have demonstrated the best performance in both
mission asymmetry. Pfeiffer and Grbic recently presented the efficiency and bandwidth. Excellent examples include
systematic methods to analyze and synthesize bianisotropic the ultra-broadband THz linear polarization rotator shown in
metasurfaces realized by cascading anisotropic, patterned figures 12(c)–(f), which exhibits a bandwidth over two octaves
metallic sheets. This design approach starts with the desir- [56, 98]. Another tri-layer metasurface is shown in figures 13(d)
able S-parameters and solves for the necessary admittances and (e), which demonstrates highly efficient, broadband asym-
of the metallic sheets. Once the required sheet admittances metric transmission of linearly polarized millimeter waves [88].
are known, the theory of frequency-selective surface and The simulated results show that a 1 dB transmission bandwidth
16
Figure 13. (a) Schematic of a tri-layer metasurface unit cell and (b) optical image of its top metallic sheet, which exhibits asymmetric
transmission of circularly polarized millimeter waves, with transmission coefficients shown in (c). Solid curves: measured data; dashed
curves: simulated data. (d) The unit cell of a tri-layer metasurface which enables (e) broadband and highly asymmetric transmission of
linearly polarized millimeter waves. Used with permission from [88] copyright 2014 by the American Physical Society.
of 2.43 : 1 for the desired polarization is achieved, and that the required size d of the resonators, which is related to the free
rejection of the unwanted polarization exceeds 30 dB in this space resonant wavelength λ 0 by d ∼ λ 0 / ε . However, increas-
band. ing the dielectric constant also reduces the radiation efficiency
and narrows the operational bandwidth, which is inversely
5. Dielectric metasurfaces related to the dielectric constant. Typical values of the dielectric
constant used range from 8 to 100 in order to balance the com-
The majority of metasurface research has focused on using pactness, radiation efficiency and bandwidth requirements. Very
subwavelength metallic structures, where ohmic losses pose a often dielectric resonators are mounted on top of a metal ground
severe issue, particularly in the optical frequency range, limit- plane, which improves the radiation efficiency and acts as an
ing the performance of arguably any desirable functions. Low- electrical symmetry plane to improve the compactness. Early
loss, high-refractive-index dielectric materials have received work in resonant dielectric antennas at microwave frequencies
much attention during recent years partially due to their abil- has been summarized in review articles [114, 115].
ity in addressing the efficiency issue in metallic metasurfaces. In the optical regime, low loss dielectric particles support
Furthermore, the capability of tuning the magnetic and elec- strong electric and magnetic scattering known as Mie reso-
tric resonances through tailoring the geometry and spacing nances, which can be decomposed into a multipole series.
of dielectric resonators enables device functionalities beyond The modes are determined by the particle size and structural
metallic metasurfaces. properties [116–119], in contrast to metallic particles where
the resonance scattering is dominated by the electric reso-
nances. In most dielectric resonators of regular shapes such as
5.1. Dielectric resonators
spheres, cubes, cylindrical disks and rods, the lowest resonant
Dielectric resonators can be traced back to the discussions by mode is the magnetic dipole resonance and the second lowest
Richtmyer [112]. Due to the excitation of the resonant modes as mode is the electric dipole resonance [119, 120]. Figure 14
well as their leaky nature, dielectric resonators can serve as radi- shows the fundamental magnetic and electrical dipole modes
ative antennas, as developed theoretically and experimentally for a cubic dielectric resonator [121]. The magnetic resonance
in the 1980’s by Long et al at microwave frequencies [113]. mode originates from the excitation of circulating displace-
Increasing the dielectric constant ε can significantly reduce the ment currents, resulting in the strongest magnetic polarization
17
Figure 14. Electric and magnetic modes in a cubic dielectric resonator. (a) and (b) Magnetic dipole resonance mode, showing electric field (a)
and magnetic field (b) distributions. (c) and (d) Electric dipole resonance mode, showing electric field (c) and magnetic field (d) distributions.
The incident fields are indicated in the insets. Reproduced with permission from [121] Copyright 2008 by the American Physical Society.
at the center, similar to the case of magnetic resonant response

in metallic SRRs. The contribution from other higher order
modes can be ignored as the coefficients of these modes are
often orders of magnitude lower [122].
Subwavelength dielectric resonators can be used as the
basic building blocks of metamaterials and metasurfaces, as
first suggested by O’Brien and Pendry to obtain magnetic
activity in dielectric composites [123]. A class of Mie reso-
nance-based dielectric metamaterials have been consequently
demonstrated, with some early work reviewed in [124], where
high dielectric constant materials are used to create subwave-
length resonators for the realization of negative electric and
magnetic responses. Ferroelectric barium strontium titanate
(BST or Ba0.5Sr0.5TiO3) was used to demonstrate dielectric
metamaterials because of its high dielectric constant (∼600)
at microwave frequencies. Left-handed behavior was observed
in prisms formed by an array of periodic or random subwave-
length BST rods [125], and negative magnetic response was
also observed in a bulk metamaterial consisting of an array
of subwavelength BST cubes [121]. In the optical frequency
range, materials used to form dielectric metamaterials include
tellurium (Te) cubes on barium fluoride (BaF2) [126], cubic
(β) phase silicon carbide (SiC) whiskers on zinc selenide
(ZnSe) [127, 128], in the mid-infrared; silicon cylindrical Figure 15. Dielectric metasurface for broadband polarization
nano disks embedded within silicon dioxide [129] in the near conversion in reflection. (a) Schematic and (b) SEM image of the
infrared; silicon nano spheres on glass [130] and titanium dielectric metasurface structure. (c) Experimentally measured
dioxide cylindrical disks on silver [21] at visible frequencies. (solid lines) and numerically simulated (dotted lines) co- and cross-
polarized reflectance. Used with permission from [93]. Reprinted with
The loss reduction enabled by dielectric metasurfaces
permission from [93], copyright 2014 the American Chemical Society.
becomes clear when functioning as a linear polarization rotator
as shown in figure 15, where an array of anisotropic (rectangu-
lar) silicon resonators is separated from a metal ground plane shown in figure 15(c). This result exemplifies the significant loss
by a thin layer of PMMA. In experiments, linear polarization reduction enabled by the use of dielectric metasurfaces instead
conversion with more than 98% conversion efficiency was dem- of metallic resonators shown in figure 11, particularly in the
onstrated over a 200 nm bandwidth in the near infrared [93], as infrared and visible frequency ranges.
18
Figure 16. (a) Schematic of silicon nanodisks embedded into a low-index (SiO2) medium. (b) SEM image of fabricated silicon nanodisks
before embedding them into SiO2. The insets show the close-up top and oblique views. (c) Optical transmittance and (d) reflectance spectra
of the fabricated sample, where the white dashed ellipses indicate the regions where the back scattering is significantly reduced. Used with
permission from [129] copyright 2013 the American Chemical Society.
In general, dielectric resonators offer only up to π phase direction. The magnetic Mie resonance overcomes the absence
variation in transmission when the electric and magnetic of magnetic materials at optical frequencies and enables the
resonances are at different frequencies. By overlapping the investigation of directional optical scattering using dielectric
electric and magnetic dipole resonances through varying the metasurfaces. The complete cancellation of back scattering
geometry of dielectric resonators, however, it is possible to was also theoretically predicted in [118] at an off-resonance
achieve a phase variation covering the entire 2π range [131]. frequency in an array of silicon nano spheres where the elec-
This was experimentally verified even without satisfying the tric and magnetic polarizabilities have equal values. Such a
condition of equal electric and magnetic resonance width phenomenon corresponds to a ‘Huygens’ secondary source,
[132]. In cylindrical dielectric disks, the tuning parameter and was experimentally demonstrated using nonmagnetic
could be the disk height, diameter, and period (spacing). The dielectric spherical and cylindrical scatters with moderate
spacing between resonators further facilitates the tuning of dielectric constants at microwave [135] and visible [136, 137]
resonance coupling [118], which affects the dispersion of the frequencies.
scattering phase resulting from the different transverse elec- The resonant directional scattering is more interesting
tric and transverse magnetic modes, and also enables electro because of the large field enhancement and concentration.
magnetically induced transparency in dielectric metasurfaces Resonant response usually accompanies large back scatter-
with an ultra high quality factor [133]. ing, which makes it more feasible for dielectric metasurfaces
to operate in a reflection configuration [21, 93]. This ena-
5.2. Directional scattering
bled the demonstration of broadband dielectric metasurface
mirrors [138–140] and optical magnetic mirrors [141, 142],
In 1983 Kerker et al discussed electromagnetic scattering by without reflection phase reversal in the latter. Using geometric
magnetic spheres. It was shown that back scattering can be shapes other than spherical or cubic dielectric resonators, one
reduced to zero by spheres with equal permeability μ and could have more degrees of freedom to tune independently
permittivity ε [134]. In such a situation the particle exhibits the frequencies of electric and magnetic resonances to real-
equal electric and magnetic multipole coefficients, resulting ize resonant directional scattering. This is exemplified by the
in destructive interference in the backward propagating direc- closer electric and magnetic dipole resonances when squeez-
tion and constructive interference in the forward propagating ing the silicon spheres in the z-direction, which results in a
19
larger forward-to-backward scattering ratio [136]. An array of the expected 20° (figure 17(f)) [21]. It was shown that con-
silicon cylindrical nano disks, as shown in figures 16(a) and siderable dissipation of optical power occurs within the TiO2
(b), was used to demonstrate resonant directional scattering resonators, partially because this configuration can also
in the visible wavelength range [129]. By varying the diam- function as a metamaterial absorber [145]. Even more opti-
eter of the silicon disks, it was observed that the electric and cal power is coupled to surface waves, which was described
magnetic resonances overlap, resulting in enhanced forward in a recent theoretical proposal of directional launching of
scattering and cancellation of backward scattering, as shown surface waves [146].
in figures 16(c) and (d). A metasurface that converts a Gaussian beam into a vortex
An ideal dielectric Huygens’ metasurface requires over- beam was demonstrated; it consists of four quadrants with a
lapping electric and magnetic dipole resonances of equal phase increment of π /2 and each quadrant consists of an array
resonance strength and width in order to completely can- of cylindrical silicon nano disks of the same geometry but dif-
cel the reflection and obtain near unity transmission [132]. ferent separations between adjacent disks [147]. By varying
High transmittance of 55% at resonance was experimentally the diameter of silicon nanoposts to control the phase profile,
demonstrated in the near-infrared using a silicon metasur- a high-efficiency lens was demonstrated with measured focus-
face consisting of an array of cylindrical resonators embed- ing efficiency in transmission up to 82% [148]. Through vary-
ded within an SiO2 environment, where the condition of ing the geometric dimensions and coupling strength between
equal width of the electric and magnetic resonances was dielectric resonators, it is possible to create the required phase
not yet satisfied [132]. By tuning the dielectric constant of profiles to simultaneously control the wavefront at multiple
the environment and the geometric dimensions of the reso- wavelengths. This approach was exploited in the demonstra-
nators, it is possible to achieve spectral overlap and equal tion of a multi-wavelength dielectric metasurface lens operat-
width of the resonances. Post-fabrication active tuning of ing near telecommunication wavelengths [63, 149]. To achieve
the resonances is attractive for this purpose as well. For equal focal lengths at different wavelengths, the metasurface
instance, a layer of liquid crystals was added on top of the lens imparts a wavelength dependent phase contribution to
silicon nano disks, providing temperature-dependent refrac- compensate for the dispersive accumulated propagation phase.
tive indices when the liquid crystals were switched between This is achieved by designing the dispersive phase response
the nematic and isotropic phases [143]. It was shown that of coupled dielectric ridges patterned on a fused silica (SiO2)
the electric resonance has a larger tuning range because of substrate, as shown in figure 18(a). It creates a phase profile
extended fringing fields outside the resonators, while the that realizes the same focal length for wavelengths at 1300,
magnetic resonance has smaller tuning capability because 1550, and 1800 nm as shown in figures 18(b)–(d). The focus-
of the better-confined field distribution within the dielectric ing efficiency, defined as the ratio of power at the beam focal
resonators. Reconfigurable directional scattering can be also waist and the input power, is still rather low, in part due to the
accomplished using metasurfaces consisting of semicon- reflection loss. Few-layer metasurfaces introduced in previ-
ducting resonator arrays through injection of free charge ous sections could potentially address this issue of impedance
carriers by optical excitation [144]. mismatch and improve the focusing efficiency. For wave-
lengths other than these specific values, the operation of the
lens follows the normal dispersion curves, which indicates
5.3. Beam forming and wavefront control enabled
by dielectric metasurfaces that a dielectric metasurface lens that eliminates chromatic
aberration over a broad range of wavelengths is still challeng-
Similar to metasurfaces consisting of plasmonic resonators, ing to accomplish.
wavefront control and beam forming can be accomplished An alternative approach to create a spatially-varying phase
using dielectric metasurfaces. By varying the dimensions profile is through the use of Pancharatnam–Berry phase [50].
of the rectangular silicon resonators shown in figure 15, a The key is the conversion between left- and right-handed cir-
phase variation can span the entire 2π range. This enables cular polarization states via different routes on the Poincaré
the generation of a near infrared optical vortex beam in sphere. The required polarization control can be achieved
reflection with high efficiency when a phase gradient profile by the excitation of electric and magnetic resonances in
was created in the azimuthal direction using 8 elements of dielectric resonators. Using silicon nanobeams with appro-
different sizes [93], as shown in figures 17(a)–(c). The use priate geometric dimensions, it was shown that the incident
of a PMMA spacer layer between the silicon resonators and circularly polarized light is partially converted into circularly
a metallic back plane not only provides the desirable inter- polarized light with opposite handedness with an imparted
ference resulting from the Fabry–Pérot-like multiple reflec- Pancharatnam–Berry phase depending on the orientation of
tions, but also effectively prevents the incident light from the silicon nanobeams [150]. The nanobeam metasurface
coupling to surface waves. This is in remarkable contrast exhibits anomalous refraction when forming a constant phase
to the situation where the dielectric resonators are directly gradient. Linearly polarized incident light is split into right-
mounted onto the metallic surface [21]. In the latter work, and left-handed circularly polarized beams that propagate in
a linear phase gradient at wavelength of 633 nm was created different directions. Transmission spatial phase profiles have
by using six cylindrical TiO2 cylindrical resonators of vari- been also experimentally demonstrated, functioning as lenses
ous diameters sitting on top of a silver plane (figure 17(d)), for focusing and axicons for creating a Bessel beam (see fig-
demonstrating a deflection from the specular reflection by ure 19) [150].
20
Figure 17. (a) A phase profile in the azimuthal direction with an increment of π /4, created using (b) silicon rectangular resonators on
top of a metal mirror with a PMMA spacer, and enabling the formation of a near infrared optical vortex beam. The pattern in (c) is the
interference between the vortex beam and a reference Gaussian beam. (d) Schematic of part of a reflect-array metasurface consisting of
dielectric resonators patterned on a metallic substrate and operating at λ = 633 nm. (e) Simulated electric and magnetic field distributions
in a dielectric resonator antenna. (f) Simulation showing that at zero-degree angle of incidence the metasurface in (d) generates a reflected
wave propagating along 20° direction from the surface normal. (a)–(c) used with permission from [93] copyright 2014 American Chemical
Society, (d)–(f) used with permission from [21] copyright 2013 by The Optical Society.
It is essential to realize simultaneous and complete control beamsplitters, holograms and arbitrary vector beam genera-
of polarization and phase with subwavelength resolution and tors. Two examples are illustrated in figure 20 for incident
high transmission. In the optical regime plasmonic metasur- polarization-dependent focusing.
faces partially accomplish this goal with limited efficiency
[47]. In a recent paper from Faraon’s group, a dielectric metas-
uface platform was demonstrated based on elliptical high- 6. Metasurfaces for wave guidance and radiation
contrast dielectric nanoposts that provide complete control of
transmissive polarization and phase with measured efficiency In the previous sections we mainly focus on the physics and
ranging from 72% to 97%, achieved through varying the ellip- applications of metasurfaces in controlling waves that prop
ticity, size, as well as orientation of the nanoposts [151]. It was agate in free space. The present section reviews the emerging
shown that most free space high-performance transmissive research on using metasurfaces to control guided waves and to
optical elements can be realized, such as lenses, wave plates, couple between guided waves and waves propagating in free
21
Figure 18. Multi-wavelength dielectric metasurface cylindrical lens. (a) False colored side-view SEM image of the metasurface lens.
Each unit cell is identified by a different color. (b)–(d) Measured intensity distributions in the plane perpendicular to the silicon ridges at
wavelengths (b) 1300 nm, (c) 1550 nm, and (d) 1800 nm. Used with permission from [63], copyright 2015 American Chemical Society.
Figure 19. (a) SEM image of a fabricated dielectric metasurface axicon consisting of silicon nanobeams. (b) Measured intensity profile
of the nondiffractive Bessel beam generated behind the axicon in (a) in the xz plane. Used with permission from [150]. Reprinted with
permission from AAAS.
22
Figure 20. (a) Dielectric metasurface that separates the x- and y-polarized incident light, deflecting and focusing them to two different
spots. (b) Dielectric metasurface that focuses the incident circularly polarized light to a diffraction-limited spot or a doughnut-shaped spot
depending on its handedness. Left column: schematic illustration of the devices; Mid-column: simulated and experimental results; Right
column: SEM images of the dielectric metasurfaces. Used with permission from Macmillan Publishers Ltd: Nature Nanotechnology [151]
copyright 2015.
space. Because of the spatial inhomogeneity of metasurfaces, over a broad spectral range and won’t be greatly affected
they do not support any eigen guided modes. That is, waves by small structural changes to the metasurfaces. On the
propagating along metasurfaces are at a transient state and are contrary, conventional grating couplers provide positive
constantly evolving. Thus, metasurfaces are most suitable for and negative reciprocal lattice vectors, ±2π /Λ, where Λ
realizing mode conversions. By designing the in-plane effec- is the grating period. The coupling between two modes
tive wavevector using metasurface structures, one is able to is symmetric, and thus the phase matching condition,
realize conversion between two different guided modes or β1 − β 2 = 2π /Λ, has to be strictly satisfied to ensure that
between a guided mode and a mode propagating in the free electromagnetic energy is transferred from one mode to
space. the other.
There are a couple of major differences between mode con- (2) The spacing between adjacent constituent elements in a
version using metasurfaces and using conventional grating- metasurface is subwavelength. Therefore, metasurfaces
based mode convertors: are able to modify the wavevector of a guided wave adi-
abatically. The absence of abrupt variation of wavevectors
(1) Metasurfaces can be designed to provide a unidirectional prevents scattering of electromagnetic energy into free
phase gradient, or a unidirectional effective wavevector. space or into the substrate. Grating couplers, however,
The latter leads to an asymmetric coupling between modes: have a periodicity comparable to the wavelength. Guided
electromagnetic energy is transferred preferentially from waves are likely to be scattered, which makes in-plane
one mode to the other, while the inverse process can be confinement of electromagnetic energy a challenge.
highly inefficient. Such asymmetric electromagnetic
energy transfer between modes is maintained even when 6.1. Coupling between free space and surface waves
the conventional phase matching condition is not strictly
satisfied (i.e. phase gradient dΦ/dr not equal to the differ- The pioneering work on using metasurfaces to control guided
ence in wave number between two modes, β1 − β 2). This waves was conducted by Sievenpiper and colleagues in the
property ensures that mode conversion can be realized microwave spectral range [152]. They used the concept of
23
Figure 21. (a) Schematic showing the concept of holographic leaky wave antenna. Surface waves (undulating arrows) are excited on a
metasurface impedance surface, and are scattered by variations in the surface impedance to produce the desired radiation (straight arrows).
(b) Unit cell of the impedance surface consisting of a patch antenna patterned on a metal grounded layer of insulator. (c) A section of the
designed scalar impedance surface that scatters a cylindrical surface wave produced by a point source into a plane wave propagating along
60° from the surface normal. (d) A section of the designed tensor impedance surface that scatters a cylindrical surface wave produced by a
point source into a plane wave propagating along 45° from the surface normal. (e) Black and gray curves are, respectively, radiation patterns
of a monopolar antenna placed on the scalar holographic impedance surface and on a smooth metal surface. (f) Black and gray curves show,
respectively, measured radiation patterns with left-handed and right-handed circular polarization produced by a monopolar antenna placed
on the tensor holographic impedance surface. Used with permission from [152], copyright 2010 IEEE.
holography to design impedance surfaces that convert a given of surface current. The function of the impedance surface is to
surface wave into a freely propagating wave with desired translate this surface current to a distribution of electromagnetic
far-field radiation pattern and polarization. The impedance waves on the surface, which matches the desired radiative wave.
surface is essentially a hologram, which is the interference In Sievenpiper and colleagues’ work, square patch anten-
pattern between a reference beam and an object beam, and nas (figure 21(b)) were used to construct scalar impedance
carries information of the phase, amplitude and polariza- surfaces and square patches with an additional slice were used
tion of the desired object beam. The object beam is recon- for tensor impedance surfaces. The three independent terms
structed when the reference beam impinges on the hologram. in the impedance tensor, Zxx, Z xy = Z yx and Zyy, are controlled
In Sievenpiper’s implementation, a source antenna produces by the three degrees of freedom in antenna design: the slice
the reference beam in the form of a surface wave, Esurf, and width, its orientation angle, and the gap between neighboring
the object beam is the desired wave, Erad, propagating in square patches. In the case of scalar impedance surfaces, the
the half space above the surface (figure 21(a)); microwave value of surface impedance of patch antennas was determined
holograms are created according to the interference pattern by the following procedure:
produced by the two waves and consist of a square lattice
of dissimilar sub-wavelength conductive patches on a metal- (i) Calculating dispersion relation of surface waves
grounded dielectric substrate. Both scalar and tensor forms of propagating on a 2D periodic array of patch antennas.
the impedance surfaces were experimentally demonstrated. Specifically, Bloch boundary conditions are applied to
Surface impedance provides an appropriate language to char- a unit cell of the impedance surface, and eigen surface
acterize the properties of the metasurface. It is defined as the wave modes and their eigen wavevectors are determined
ratio between the electric and magnetic fields near the surface. for a range of frequencies.
For transverse magnetic (TM) waves (i.e. magnetic field trans- (ii) Calculating the surface impedance for a given operation
verse to the propagation direction) that propagate in the x-direc- frequency ω 0, Z (ω 0 ) = ∫ (Ex /Hy )dx dy.
unit cell
tion, the surface impedance is Z (x, y ) = Ex (x, y )/Hy(x, y ). The
surface magnetic field is proportional to the surface current, A library that relates the surface impedances and patch
which is provided by the electromagnetic source. For exam- antenna geometries can be created by repeating the above pro-
ple, a monopole antenna produces a cylindrical distribution cedure for patch antennas of different sizes.
24
Figure 22. (a) Section of an impedance surface near the central monopolar antenna. (b) Right-handed circularly polarized radiation profiles
produced by the impedance surface antenna near 17 GHz. Inset is the entire antenna with a radius of 9.7 cm. Used with permission from [153],
copyright 2011 IEEE.
The distribution of surface impedance Z(x, y) is determined of patch antennas, Sievenpiper and coworkers demonstrated
by the following holographic technique. In the case of a scalar a scalar impedance surface that scatters the current generated
impedance surface, with a surface current Jsurf (x, y ) produced by a monopolar antenna into a linearly polarized plane wave
by the electromagnetic source and the object far-field radia- propagating along 60° from the surface normal (figures 21(c)
tion E rad(x, y, z ), the required surface impedance is and (e)). The surface current has a cylindrical distribution and
⎧ ⎡ ⎛ Jsurf, x ⎞⎤ ⎫
1
can be described by Jsurf = exp(−jk 0 ns r )(x, y, 0), where
Z (x, y ) = j ⎨X + M Re ⎢(E rad, x, E rad, y ) ⎜ ⎟⎥ ⎬.
⎪ ⎪ r2
2 1/2
r = (x + y ) , k0 is the free space wavevector, and ns is the
2
⎝ Jsurf, y ⎠⎦ ⎭
(11)
⎩
⎪
⎣ ⎪
effective index of the surface current, which is assumed to be

In the case of a tensor impedance surface, we have a constant and is a function of the thickness and materials of
the dielectric spacing layer between the metal patches and the
⎛ ⎞ M ⎡⎛ E rad, x ⎞ metallic ground. They also experimentally demonstrated a
Z (x, y ) = j ⎜ X 0 ⎟ + j Im ⎢⎜ ⎟ (Jsurf, x, Jsurf, y )
⎝ 0 X⎠ 2 ⎣⎝ E rad, y ⎠ tensor impedance surface that converts the current generated
by a monopolar antenna to a circularly polarized far-field radi-
⎛ Jsurf, x ⎞ ⎤
−⎜ ⎟(E rad, x, E rad, y )⎥ . (12) ation propagating along 45° direction (figures 21(d) and (f)).
⎝ Jsurf, y ⎠ ⎦ Maci and colleagues used the same holographic principle
to demonstrate metasurfaces with modulated surface imped-
In the above two equations, X represents the average ance [153, 155]. They used square patch antennas of different
impedance value, and M spans the entire available imped- sizes to create a spiral distribution of surface impedance that
ance range. Using the holographic technique and the library converts a surface current produced by a monopolar antenna
25
Figure 23. (a) Planar 2D leaky-wave antenna consisting of radially directed and width-modulated microstrip lines and is able to transform
a cylindrical surface wave into leaky waves. (b) Broadside beam pattern of the leaky-wave antenna. Solid and dashed curves are,
respectively, measured and simulated beam patterns at ∼22 GHz. Used with permission from [154], copyright 2013 IEEE.
to a collimated right-handed circularly polarized far-field radi- wave with a wavevector larger than the free space wavevector
ation (figure 22). Podilchak and collaborators demonstrated is excited; the surface wave becomes even more evanescent
experimentally [154] that width-modulated microstrip lines as it further interacts with the gradient metasurface, which
patterned on a grounded dielectric slab introduce a sinusoi- prevents decoupling of the wave back to the free space; (3)
dally modulated surface impedance and provide appropriate the impedance mismatch between the supercells of the meta-
conditions for leaky wave radiation (figure 23). Figure 24 surface coupler is reduced to prevent scattering of the surface
show a holographic metasurface that detects optical vortex wave. Through these approaches, the authors were able to
beams with specific orbital angular momentum (OAM) [156]. demonstrate the coupling of an incident wave from free space
The nano-structured binary holograms shown in the left panel into a surface wave with efficiencies of ∼94% in simulations
of figure 24(a) were created by calculating the interference and ∼73% in experiments using microwaves [158].
pattern between a converging surface plasmon wave and an
incident optical vortex beam. The simulated results in fig- 6.2. Control of surface waves
ure 24(a) show that a converging surface plasmon wave is
generated only when an optical vortex beam with the correct The examples that have been discussed in this section so far
OAM is scattered by the hologram. Experimental results in are all about coupling surface waves and waves propagating in
figure 24(b) show that a hologram can distinguish an optical free space using metasurfaces. The subject of controlling the
vortex beam with OAM of −1 from optical vortex beams with propagation of surface waves confined to a 2D plane is a new
other values of OAM. frontier of metasurface research.
The major challenges in coupling an incident wave from Vakil and Engheta proposed using graphene as an ultra-
free space into a surface wave with high efficiency are to sup- thin platform for controlling in-plane propagation of infrared
press the reflection of the incident wave on the device surface electromagnetic waves [159]. They demonstrated theoretically
and to prevent decoupling of the surface wave back into free that by designing and manipulating spatially inhomogene-
space. In a series of work from the Zhou group [14, 157, 158], ous conductivity patterns on a sheet of graphene using the
a few strategies were devised to address these challenges: (1) electric field effect, one can realize a number of transforma-
the entire surface of the metasurface coupler is designed to be tion optical devices. The example of a graphene metasurface
impedance matched with free space to minimize direct reflec- Luneberg lens is shown in figure 25(a). The research group
tion; (2) the lateral effective wavevector provided by the meta- of S. Maci used patch antennas of different sizes or metallic
surface is designed to be sufficiently large so that a surface pins of different heights to demonstrate two types of in-plane
26
Figure 24. (a) Left panel: metasurface holograms for detecting optical vortex beams. Right panel: simulation results of the intensity
distribution of surface plasmon waves generated by illuminating the holograms at normal incidence with different optical vortex beams.
(b) Photocurrent as a function of incident polarization measured for a metasurface hologram designed for detecting optical vortex beams
with orbital angular momentum of L i = −1. Used with permission from Macmillan Publishers Ltd: Nature Communications [156]
copyright 2012.
planar lenses (Luneberg lenses and Maxwell’s fish-eye [153]; optical coupling from the second mode back to the first one
the latter is shown in figure 25(b)). In their pioneering work, is highly inefficient. As a result, the phase matching condi-
Gok and Grbic used the concept of transformation electro tions are greatly relaxed, which enables the demonstration of
magnetics to demonstrate independent control of the power extremely broadband and robust waveguide mode conversion
flow and phase progression of electromagnetic fields in a 2D (figures 25(e)–(g)) [161, 162].
space (figures 25(c) and (d)) [160]. The resulting metasurface
is a highly inhomogeneous, anisotropic media where each 7. Active metasurfaces
unit cell is characterized by a 2 × 2 permeability tensor in the
plane and a scalar permittivity in the surface normal direction. Active devices and components play a critical role in mod-
These parameters were judicially chosen to create stipulated ern electromagnetic and photonic systems. Active control of
2D distributions of wavevector and Poynting vector, as well as metamaterials and metasurfaces extends their exotic passive
to ensure impedance matching between the adjacent unit cells properties by allowing fine resonance tuning to adapt to the
so that there is no reflection and scattering of the surface wave operational conditions, and enabling a switchable resonant
as it propagates on the metasurface. response, for instance, for signal modulation in communi-
The concept of metasurfaces has been introduced into the cation and imaging. Furthermore, the concentration of opti-
field of integrated photonics where 1D phased array antennas cal power in metasurface resonators integrated with optical
patterned on optical waveguides enable the control of opti- nonlinear materials can dramatically enhance the nonlinear
cal power flow and mode coupling in the waveguides. The response, as predicted in Pendry’s original work on SRRs [15].
1D antenna array introduces a unidirectional phase gradient As compared to bulk metamaterials, the planar configuration
dΦ/dx, where dΦ is the difference in phase response between of metasurfaces facilitates the integration of active functional
adjacent antennas that are separated from each other by a materials. A variety of functional materials providing tunable
subwavelength distance of dx. The phase gradient is equiv- refractive indices through thermal excitation, voltage bias,
alent to a unidirectional effective wavevector ∆k along the magnetic field, optical pump, or mechanical deformation have
waveguide, which leads to directional coupling of waveguide been successfully incorporated into metasurfaces. In part
modes. That is, optical power couples preferentially from icular, semiconductors and graphene become the materials of
one waveguide mode to a second waveguide mode, whereas choice for electrically tunable active metasurfaces.
27
Figure 25. (a) Luneburg lens based on graphene metasurface. Shown is the simulated phase of Ey of the surface plasmon at 30 THz on the
graphene. D = 1.5 μm, w = 75 nm, and L = 1.6 μm. (b) Snapshot of the field in a metasurface Maxwell’s fish-eye lens consisting of pins
of different heights on a grounded slab and curvilinear trajectory of the real part of the Poynting vector. (c) Snapshot of simulated, vertical
electric field (Ez) of a metasurface that transforms a cylindrical surface wave into a surface wave with trapezoidal power density and linear
phase progression. (d) Upper panel: simulated and ideal power densities along boundary 1 and boundary 2. Lower panel: phase profiles
along boundary 1 and boundary 2. (e) Schematic of a telecom TE00-to-TM10 mode converter consisting of silicon phased array antennas
patterned on a Si3N4 waveguide. The phase response is due to the optical Mie resonance in the silicon nanorod. (f) Simulated field evolution
in the mode converter. (g) Purity of the converted TM10 mode as a function of wavelength, showing that the mode converter works over a
broad wavelength range. (a) used with permission from [159] reprinted with permission from AAAS, (b) used with permission from [153]
copyright 2011 IEEE, (c) and (d) used with permission from [160] copyright 2013 by the American Physical Society, (e)–(f) used with
permission from [161, 162] copyright 2014, 2015 by The Optical Society.
7.1. Actively switchable and frequency tunable bars form a part of the capacitive gap in an electric SRR unit
metal/semiconductor hybrid metasurfaces cell. Under photoexcitation with near-infrared light, the sili-
con bars become metallic , which increases the SRR capaci-
The conductivity of semiconductors can be increased by
orders of magnitude through doping, and thus semiconductors tance. Therefore, the frequency of the SRR LC resonance is
can be converted into plasmonic materials in the infrared and tuned to a lower frequency with the tuning range of about 20%
spectral ranges with longer wavelengths. Active tuning of the [167], as shown in figure 26(b). A variety of similar structures
conductivity can be realized by carrier injection and deple- were demonstrated resulting in a blue shift of the resonance
tion through photoexcitation and voltage bias. Such a unique frequency [168].
capability makes semiconductors ideal materials for integra- Optically modifying the metasurface geometric structure
tion into metamaterial structures to accomplish active and enables the transition between different types of resonances.
dynamic functionalities, particularly in the microwave and In figure 26(c) silicon is integrated at the gaps of a metal
THz frequency range. Varactor diodes have been widely used patch array that exhibits a dipolar resonance without photo-
to realize frequency tunable and nonlinear response [163, excitation and allows high transmission below the resonance
164] in microwave metasurfaces. At THz frequencies, SRR frequency. Under photoexcitation, the metallic silicon con-
arrays can be directly fabricated on top of semiconducting nects the metal patches, effectively forming a metal wire
substrates such as intrinsic silicon and gallium arsenide, and grating that blocks the low frequency THz waves, as shown
the resonant response can be tuned through photoexcitation in figure 26(d), and resulting in ultra broadband THz modu-
of free charge carriers at the substrate surface [165], result- lation [169]. Recently, optically tunable THz metamaterial
ing in an ultrafast switching speed [166]. Furthermore, semi- perfect absorbers [170] were demonstrated, as shown in fig-
conductors can be used as part of the resonant structure. In ures 26(e) and (f), where silicon islands are located at the
this case, photoexcitation dynamically modifies the structural gaps of electric SRRs. Using such an approach, a variety
geometry of the resonator, enabling switchable or frequency of optical responses can be switched/tuned via photoexcita-
tunable response. As shown in figure 26(a), a pair of silicon tion, such as the handedness of chiral metasurfaces [171,
28
Figure 26. Optically tunable THz metal/semiconductor hybrid metasurfaces. (a) SEM image of an electric SRR unit cell of a frequency
tunable THz metasurface. (b) Upon photoexcitation of the silicon bars, the gap capacitance increases, which results in a lower resonance
frequency. (c) Optical microscopy image of an ultra broadband THz modulator. (d) Without photoexcitation, the transmission is high at
the low frequency side of the Lorentzian resonance of the gold grid; upon photoexcitation of the silicon region, it becomes effectively
a wire grating showing low transmission. (e) Schematic (left panel) and optical microscopy image (right panel) of a THz metamaterial
absorber consisting of silicon pads integrated at the gaps of SRRs. (f) Photoexcitation dramatically tunes the property from dual-band to a
broadband absorption. (a) and (b) used with permission from Macmillan Publishers Ltd: Nature Photonics [167] copyright 2008, (c) and
(d) used with permission from [169] copyright 2014 AIP Publishing LLC, (e) and (f) used with permission from [170] copyright 2014 by
John Wiley and Sons.
172] and plasmonic electromagnetically induced transpar- development of high performance THz wireless communica-
ency (EIT) [173]. tion and imaging systems.
Semiconducting hybrid metasurfaces feature electri- In recent work, an electrically driven THz metasurface
cal tuning of resonances via the application of a voltage active diffraction grating was demonstrated to realize back-
bias, which is more convenient and practical for applica- ground-free THz modulation with an unprecedented 22 dB
tions. The most prominent examples are the integration of of dynamic range [180]. Each ‘grating finger’ consists of an
varactor diodes for microwaves and Schottky junctions for array of electrically connected and switchable SRRs forming
THz frequencies. The first demonstration of an electrically a column that is controlled by an independent voltage bias, as
switchable THz metasurface featured an unprecedented 50% shown in figure 27(c). The diffractive metasurface grating is
modulation depth [174], which was further improved to 80% created by applying a voltage bias to alternate columns within
through structural optimization [175]. Together with the the 32-column metasurface structure, resulting in a frequency
causally connected phase modulation (up to 0.55 rad), this dependent diffraction angle for the incident broadband THz
device allows broadband THz modulation [174] that can be radiation. At the metasurface resonance frequency of 0.4 THz,
used to replace a mechanical optical chopper in a lock-in THz the diffraction is strongest because of the largest transmission
detection scheme with modulation speed in the MHz range contrast between two neighboring columns. However, when
[176–178], limited either by the large device area accom- the same voltage bias is applied to each of the columns, the
panied by high stray capacitance or parasitic capacitance structure behaves as a uniform metasurface with no observ-
from the bonding electrodes and wires. Very recently, GHz able diffraction. Therefore, application of an AC voltage to
electronic modulation speed has been demonstrated by alternate columns results in background-free diffractive mod-
using double-channel heterostructures supporting nanoscale ulation of the incident THz radiation, as shown in figure 27(d).
2DEGs with high carrier concentration and mobility [179], Spatial light modulators have been realized by pixelating
shown in figure 27(a). Through designing a composite hybrid the metasurface for independent control of reflection, trans-
metasurface structure to reduce the stray capacitance, 1 GHz mission, or their phase. A prototype THz metasurface spatial
modulation speed, 85% modulation depth (figure 27(b)), and light modulator with 4 × 4 pixels was realized to demonstrate
a phase shift of 1.19 rad were experimentally realized dur- reconfigurable interference patterns of double slits [181]. THz
ing real-time dynamic tests. Furthermore, a wireless free metasurface spatial light modulators with a larger number of
space modulation THz communication system based on pixels are possible, although the increasing number of elec-
this external THz modulator was tested using 0.2 Gbps eye trical connecting wires makes them more complicated. One
patterns. This accomplishment opens an avenue toward the solution to this problem is a reflection-mode metasurface
29
Figure 27. Electrical modulation of metal/semiconductor hybrid metasurfaces. (a) Schematic of a unit cell of a high-speed THz
metasurface modulator based on double-channel heterostructures. (b) Modulation performance at different frequencies. (c) Schematic of an
active THz metasurface diffraction grating formed by 32 columns controlled by independent voltage biases, where different colors indicate
different voltage biases. (d) Dynamic diffraction is enabled by applying reverse voltage biases to alternate columns. An unprecedented
modulation depth of 22 dB is accomplished at the designed operation frequency of 0.4 THz. (e) THz spatial modulator with 8 × 8 pixels
based on electrically switchable THz metasurface absorbers, and used for THz compressive imaging, and (f) the corresponding device
schematic consisting of a linked array of metallic resonators making Schottky contacts with an underlying n-doped semiconductor spacer,
a metal ground plane serving as the ohmic contact, as well as the accessory structure enabling independent voltage biases to the pixels. (a)
and (b) used with permission from [179] copyright 2015 America Chemical Society, (c) and (d) used with permission from [180] copyright
2014 AIP Publishing LLC, (e) and (f) used with permission from [178] copyright 2013 by John Wiley and Sons.
30
spatial modulator based on an electrically tunable metamat graphene-antenna array (figure 28(c)). This approach provides
erial absorber structure [178]. As shown in figure 27(f), a a modulation speed in the tens of MHz range and an opti-
linked array of resonators and an underlying semiconductor cal modulation depth close to 100% with the latter defined as
layer create Schottky junctions, and a metal ground plane 1 − Rmin(λ )/Rmax(λ ) where Rmin(λ ) and Rmax(λ ) are the mini-
serves as the ohmic contact. Application of a reverse voltage mum and maximum achievable reflectivity at a certain wave-
bias enables tuning the frequency of the resonant absorption, length λ [193].
with modulation speeds up to 10 MHz. This type of THz spa- A narrow spectral width is essential for realizing a high
tial modulators based on metamaterial absorbers, shown in modulation depth based on resonance frequency tuning. For
figure 27(e), have been recently successfully employed in this purpose one may integrate graphene into metasurfaces
THz compressive imaging [182]. that exhibit high Q-factor Fano resonances, such as the one
shown in figure 28(d), which consists of an array of connected
7.2. Graphene hybrid metasurfaces dipole and monopole resonators fabricated on top of a thin sil-
icon dioxide layer on a silicon substrate [194]. This metasur-
Except for fabrication of metallic metasurface structures face structure exhibits double plasmonic electromagnetically
directly on a substrate such as those shown in figures 27(a) and induced transparency (EIT) as illustrated by the two near-field
(c), integration of crystalline semiconductor films or islands intensity enhancement peaks in figure 28(e) and two reflection
into the critical regions of more complex metasurfaces (e.g. the minima in figure 28(f). The graphene-SiO2-silicon structure
structure shown in figures 26 and 27(f) as well as other few- also enables back-gating to tune the graphene carrier density,
layer metasurfaces) poses significant fabrication challenges which consequently tunes the Fano resonances (figure 28(f)).
[183] mainly due to the requirement of nano-lithography or At a specific wavelength, it results in high reflection ‘ON’ and
transferring fragile semiconductor thin films. In this sense, the low reflection ‘OFF’ states with an experimentally measured
excellent mechanical properties and the tunable carrier den- modulation depth up to 90%, though the insertion loss of 81%
sity of graphene make it an excellent material to enable active is still rather high and the bandwidth is also rather small (a few
metasurfaces [184]. Graphene has largely tunable optical con- percent of the operational wavelength) [194]. Phase modula-
ductivity in the mid-infrared and THz frequency ranges. The tion has been also shown recently in a similar graphene hybrid
doping of graphene can be adjusted through changing the bias metasurface in the mid-infrared, which can be potentially
voltage by a factor of 10 at room temperature, which leads used for motion sensing and tunable waveplates [196].
to a large change in its sheet conductivity σ and therefore Electrically tunable metasurfaces can be made of struc-
the in-plane electric permittivity ε∥ = 1 + iσ /(ε 0ωt ), where tured graphene sheets without utilizing metallic plasmonic
t = 0.33 nm is the thickness of single-layer graphene. antennas. Figure 28(g) shows a reflect-array mid-infrared
The resonant response of metasurfaces is of particular modulator consisting of graphene nanoapertures (i.e. voids cut
importance to enhance interactions between atomically thin into a graphene sheet) separated from a metallic back mirror
graphene sheets and mid-infrared and THz radiation. Metallic by a thin film of Si3N4 [197]. The width of the nanoapertures
plasmonic antennas are able to capture light from free space is chosen to be in the range of 20–60 nm, so that incident mid-
and concentrate optical energy into subwavelength spots. The infrared light can excite plasmonic resonances in the nanoap-
electric field at these spots can be two to three orders of mag- ertures, leading to strongly enhanced optical absorption of up
nitude larger than the incident field. By placing graphene in to 25%. As the bias voltage changes the carrier doping of the
the hot spots created by metallic plasmonic antennas and by perforated graphene sheet, the plasmonic resonances shift,
tuning the optical conductivity of graphene, one can switch giving rise to tunable amplitude and spectral position of the
the resonance or tune the resonance frequency of the compos- absorption peaks (figure 28(h)).
ite over a wide range. In the THz frequency range, intraband
transitions in graphene sheets have been used to demonstrate
7.3. Other resonance switchable and frequency
broadband electrical modulation [185], and patterned gra-
tunable metasurfaces
phene structures have been shown to exhibit resonant plas-
monic response [186, 187]. Integrating graphene into metallic While electrically tunable metasurfaces integrated with
resonators has enabled the demonstration of THz metasurface semiconductors and graphene are of utmost importance in
electrical modulators [188–190]. applications, there are a variety of other functional mat
Mid-infrared metasurfaces with electrically tunable spec- erials and structures that have been successfully used to real-
tral properties have been experimentally demonstrated by ize active metasurfaces. When a THz metallic SRR array
controlling the carrier density of graphene [191]. Optimizing was fabricated directly on top of a strontium titanate (STO)
optical antenna designs has improved both the frequency tun- substrate (a phase transition material), the resonance fre-
ing range and the modulation depth [192]. In figure 28(a), the quency experiences a red-shift with decreasing temperature,
upper metasurface layer is separated from a back aluminum due to the increasing refractive index of the STO substrate,
mirror by a thin aluminum oxide film. Such a reflect-array although it suffers from significant insertion loss due to the
structure exhibits nearly perfect absorption [193]. That is, at high refractive index of the STO substrate [199]. Vanadium
the OFF state the reflection is nearly zero, and the frequency dioxide (VO2) exhibits thermally driven insulator-to-metal
at which near-zero reflection occurs can be tuned by apply- phase transition and has attracted great interest in realizing
ing a voltage bias that modifies the dispersion of the top thermally active metasurfaces at THz and infrared frequencies
31
Figure 28. Electrically tunable graphene-based metasurfaces. (a) Schematic of an ultrathin mid-infrared modulator based on a tunable
metasurface absorber. (b) SEM image of the metasurface absorber. Inset: a zoomed-in view of a portion of the device. (c) Measured
reflection spectra (normalized to the reflection spectrum of an aluminum mirror) of the metasurface absorber in (b) at different gate
voltages |VG − VCNP|, where VCNP is the gate voltage when the concentrations of electrons and holes in the graphene sheet are equal, i.e.
charge neutral point (CNP). (d) SEM image of a metasurface structure exhibiting dual Fano resonances. (e) Average near-field intensity
enhancement η on graphene surface. Insets: spatial distribution of η inside the gap for the two Fano resonances. (f) Measured reflection
spectra of the device in (d) as a function of the Fermi energy. (g) Schematic of a metasurface modulator based on a graphene Salisbury
screen. The inset illustrates the device with the optical waves at the resonance condition. (h) Change in absorption with respect to the
absorption at CNP in 40 nm-wide graphene nanoapertures at various doping levels. The solid black curve corresponds to bare (unpatterned)
graphene. The inset: AFM image of graphene nanoapertures with 40 nm width. (a)–(c) used with permission from [193] copyright 2014
American Chemical Society, (d)–(f) used with permission from [194] copyright 2015 American Chemical Society, (g) and (h) used with
permission from [195] copyright 2014 by the American Physical Society.
[200–205]. The hysteresis of its phase transition has been uti- frequency tuning behaviors via thermal control [208–211] or
lized to demonstrate metasurface memory devices [201]. The photoexcitation [212], although this only applies to micro-
resonance of metasurfaces based on VO2 can be also switched wave and THz frequencies, limiting the applications of active
through the application of a voltage bias [198]. In the latter superconducting metasurfaces. Last but not least, integration
case, a thin layer of ionic gel was applied on the surface of of micro-electro-mechanical systems (MEMS) into metasur-
the metasurface (figure 29(a)). Application of positive (nega- faces has enabled reconfigurable resonances by changing the
tive) voltage selectively tunes the metasurface resonance into geometry of the resonant elements through thermal or electro-
the ‘OFF’ (‘ON’) state by inducing the VO2 film into a more static actuation [213–217].
conductive (insulating) state. In particular, a positive volt
age drives the following electrochemical reaction: VO2 + 2x 7.4. Nonlinear metasurfaces
e− → VO2−x + x O2−, so that VO2 goes across the boundary
between the insulating and metallic states following the pink The abilities of metasurfaces to promote light-matter inter-
solid arrow in figure 29(b). As a result, application of posi- action and manipulate local optical polarizations are ideally
tive voltages damps the SRR resonance, with 3 volts yielding suited to enhance nonlinear optical effects. This is particularly
complete suppression of the resonance (figure 29(c)). significant in the THz frequency range due to the difficulties
Liquid crystals can be also conveniently integrated with in generating high-power THz radiation. The concentration of
metasurface structures, and help realize electrically tun- incident THz waves relaxes the requirement of a strong THz
able spectral properties when the refractive index of liquid source, and further reveals the ultrafast dynamics of electronic
crystals is adjusted [206, 207]. The frequency tuning responses initiated by the intense THz pulses. An excellent
range is, however, quite limited and the operation speed is example is the THz-field-induced insulator-to-metal transition
slow. THz superconducting metasurfaces consisting of reso- in metal/VO2 hybrid metasurfaces, where an array of electric
nant elements made of superconducting films instead of the SRRs were fabricated on top of a VO2 film (figure 30(a))
typically used metals have shown outstanding switching and [218]. It was shown that the transmission spectra depend on
32
has been further applied to investigate nonlinear metasurfaces

integrated with semiconductors such as gallium arsenide and
indium arsenide [219, 220], where the electric field induces
intervalley scattering or impact ionization, resulting in a
reduced carrier mobility or increased carrier density, thereby
either damping or strengthening the metasurface resonant
response. Strong THz nonlinear response was also observed
in superconducting metasurfaces under intense THz radiation
[221, 222]. Although the energy of a THz photon is well below
that required to directly break a Cooper pair upon absorp-
tion and the applied THz pulses do not significantly raise the
sample temperature, the transmission measurements reveal
significant field-strength-dependent transmission spectra at
various temperatures. It would be expected that the intense
THz field can accelerate electrons that gain sufficiently high
kinetic energy to induce Cooper pair breaking, which damps
the resonance similar to the cases of resonance switching and
frequency tuning under thermal and optical excitation.
Conventionally, the phase matching condition in nonlinear
processes, such as second harmonic generation (SHG), has to
be satisfied in bulk nonlinear crystals to achieve efficient non-
linear optical generation. Under the condition of perfect phase
matching, nonlinearly generated optical signals constructively
build up, and optical power is continuously transferred from
the pump(s) to the nonlinear optical signal. Metasurfaces
greatly relax the requirement for phase matching as nonlinear
processes occur within metasurfaces that have significantly
reduced thicknesses. Giant second-harmonic (SH) response
(figures 31(a)–(c)) has been experimentally demonstrated
in plasmonic metasurfaces integrated with nonlinear media
[223]. Specifically, InGaAs/AlInAs multiple quantum wells
were used as the nonlinear media, which exhibit giant and
electrically tunable nonlinear coefficients in the mid-infrared
[225–227]. The plasmonic metasurfaces were designed to not
only enhance the local fields of both the pump and SH signal,
but also to manipulate the near-field polarization, as the rel-
evant field components involved in the harmonic generation
are the ones normal to the quantum wells (due to the selec-
tion rules for intersubband transitions within quantum wells
[228]). The nonlinear metasurfaces achieved a nonlinear con-
version efficiency of ∼2 × 10−6 using a pump intensity of
only 15 kW cm−2 [223], corresponding to an effective second-
order nonlinear coefficient of χ(2) ∼ 30 nm V−1, about three
orders of magnitude larger than that of LiNbO3. Even larger
χ(2) ∼ 250 nm V−1 has been experimentally demonstrated in
Figure 29. (a) Schematic showing electrolyte gating of VO2-based
metasurfaces. (b) Phase diagram of VO2. (c) Voltage dependent THz a nonlinear metasurface consisting of an array of SRRs and
transmission spectra of the device in (a) at 315 K. Insets: photos of InGaAs/AlInAs multiple quantum wells (figures 31(d) and (e))
gold SRRs sitting on VO2 before the ionic gel is applied. Used with [224]. The two plasmonic resonances of the SRRs enhance,
permission from [198] copyright 2014, AIP Publishing LLC. respectively, the pump and SH signal (figure 31(d)).
These demonstrations have followed the original proposal
of using SRRs by Pendry in 1999, where the resonance would
the incident field strength, as shown in figure 30(b), due to the localize electromagnetic energy within the small split-gaps
phase transition of VO2 that is initiated by Poole–Frenkel elec- and dramatically enhance nonlinear response in materials
tron liberation, followed by lattice equilibration on a picosec- being introduced [15]. Most experimental demonstrations
ond timescale. The incident few hundreds kV cm−1 THz field of nonlinear metamaterials have been mainly focused in the
is resonantly enhanced to the MV cm−1 level within the split microwave frequency range using packaged nonlinear elec-
gap, which is then sufficient to induce irreversible damage of tronic elements, such as varactor diodes, to introduce non-
the VO2 film, as shown in figure 30(a). Such a methodology linearity into the gaps of the metal resonators, resulting in
33
Figure 30. (a) Top panel: optical image of an array of gold electric SRRs fabricated on top of a VO2 film showing THz-field-induced
damage illustrated by the black spots at the split gaps. The dashed blue circle approximates the THz beam waist, and the red curve
approximates the THz intensity profile. Bottom panels: SEM images of a single SRR show that VO2 is damaged by the vertically polarized
THz field, with an expanded view of damage at the edge of the THz beam (right top) and near the beam centre (right bottom). (b)
Experimental data showing incident field-dependent nonlinear transmission spectra of SRRs on VO2 at 324 K, for in-gap fields ranging
from 0.3 to 3.3 MV cm−1. Used with permission from Macmillan Publishers Ltd: Nature [218] copyright 2012.
Figure 31. (a) Unit cell of the SHG metasurface. Dimensions of the gold nanocross are given in nm, and the unit cell has a dimension
of 1000 nm × 1300 nm. (b) Conduction band diagram of one period of an In0.53Ga0.47As/Al0.48In0.52As coupled quantum well structure
designed as the nonlinear medium for highly efficient SHG. The moduli squared of the electron wavefunctions for subbands 1, 2 and 3 are
shown and labelled accordingly. Transitions between pairs of electron subbands are marked with double-headed red arrows, and the values
of the transition energies (E21 and E32) and dipole moments (Z21, Z32 and Z31) are shown next to each arrow. (c) SHG from metasurfaces
based on (a) and (b). Shown are SH peak power (left axis) and intensity (right axis) as a function of pump peak power squared (bottom
axis) or peak intensity squared (top axis) at a pump wavenumber of 1240 cm−1 for different input/output polarization combinations.
(d) Left panel: schematic of a metasurface consisting of SRRs on top of a stack of MQWs. Upper right panel: top view of one SRR. Lower
right panel: schematic showing the two main resonant modes of the SRR at the pump and SH frequency, respectively. (e) Intensity of
the SH signal propagating in the forward and backward directions with respect to the nonlinear metasurface in (d) as a function of pump
intensity. (a)–(c) used with permission from Macmillan Publishers Ltd: Nature [223] copyright 2014, (d) and (e) used with permission from
[224] copyright 2014 AIP Publishing LLC.
nonlinear functionalities such as bistability [229, 230], reso- fabricate bulk metamaterials consisting of complex three-
nance tunability [163, 231], and harmonic generation [232, dimensional nanostructures, and utilize their thickness to
233]. In the optical frequency regime, in addition to the dif- enhance nonlinear response. Experimental work has been
ficulty in packaging nonlinear materials, it is challenging to scarce as compared to its microwave counterpart. One of
34
Figure 32. (a) SEM image of a portion of a nonlinear metasurface that radiates SH signal of different polarizations into different
directions (i.e. polarizing beam splitter for SH signal). A unit cell of the metasurface is denoted by the red rectangle. (b) Measured far-
field profiles for the metasurface in (a) for two orthogonal polarizations of the SH radiation when the pump beam is polarized along the
vertical direction. (c) SEM image of a portion of a nonlinear metasurface Fresnel zone plate (FZP) showing mirror inversion of SRRs in
adjacent zones that radiate SH waves with opposite phases. Arrows mark the effective χ(2) direction. (d) Recorded images of SH signal at
a distance of Z = 0 and Z = 1 mm from the metasurface in (c). (e) Upper and lower panels are, respectively, simulation and measurement
results showing the focusing of SH signal by the FZP (m denotes focusing order). (a) and (b) used with permission from Macmillan
Publishers Ltd: Nature Communications [239] copyright 2015, (c)–(e) used with permission from Macmillan Publishers Ltd: Nature
Photonics [240] copyright 2015.
the focuses has been harmonic generation using single layer they generate two v-polarized beams at ±40° as shown in the
metal SRR arrays (tens of nanometer thickness) excited at left panel of figure 32(b), where the radiation angles are deter-
their magnetic resonance [234, 235], where the nonlinearity mined by the period of the metasurface structure. The same
is associated with the dynamics of free and bound charges, effect was used to demonstrate complex wavefront engineer-
particularly at the metal surface [236]. This seems to be also ing of the SH signal generated from metasurfaces consisting
responsible for the recently observed broadband THz genera- of gold SRRs (figures 32(c)–(e)) [240]. Here the surface sec-
tion under femtosecond near-infrared laser excitation in an ond-order nonlinearity of gold leads to SHG. In experiments a
array of gold SRRs [237]. It was also shown that the geometry nonlinear Fresnel zone plate (FZP) was demonstrated, which
of resonators, particularly the asymmetry ratio, plays a critical focuses the SH signal to the focal spots of the plate, leading to
role as it governs the spatial overlap of the resonant modes a large enhancement of the SH intensity.
at the pump and harmonic frequencies [238]. Although the
nonlinear coefficients were often enhanced by orders of mag- 8. Summary and outlook
nitude compared to common nonlinear crystals, the absolute
conversion efficiency is still rather low. Further improvement Metamaterials and metasurfaces have led to the realization of
is non-trivial and cannot be realized by simply stacking multi- novel electromagnetic properties and functionalities through
layers for larger interaction thickness due to the impedance tailoring subwavelength structures and integrating functional
and propagation phase mismatches. materials. In this paper we have reviewed the recent develop-
A step further is the demonstration of nonlinear phased ment of two-dimensional metamaterials—metasurfaces—by
arrays that radiate generated SH signals into different direc- introducing the fundamental concepts, physical realization,
tions depending on their polarization states [239]. In the and their promising applications in the control and manipu-
metasurface structure shown in figure 32(a), the top row of lation of electromagnetic waves at frequencies ranging from
six identical resonators within the unit cell generate a sin- microwave to visible light. One of our focuses is on the cre-
gle u-polarized broadside beam, shown in the right panel of ation of an arbitrary phase profile for wavefront control and
figure 32(b), at the SH frequency. In the bottom row of four beam forming, using both metallic and dielectric metasur-
resonators within the unit cell, the left two resonators have a π faces. Another focus is on the few-layer metasurfaces that
phase difference as compared to the right two resonators, as the address the efficiency issue encountered during the earlier
local effective second-order nonlinear coefficient χ(2) changes development of metasurfaces. Active and nonlinear meta-
sign when the orientation of a SRR rotates by 180°. Together surfaces represent an important research direction that will
35
greatly expand metasurface functionalities and applications. that a monolayer of densely packed hairs with triangular
As a rapidly developing research field that has attracted world cross-sections, in some sense a biological ‘metasurface’,
wide interest, it would be impossible (and not necessary) to enhances not only the ant body’s reflectivity in the visible
include every aspect of its past success. For instance, we have and near-infrared, where solar radiation culminates, but
not included metasurfaces for antireflection coatings [85, also its emissivity in the mid-infrared [248]. The com-
241], photonic spin Hall effects in metasurfaces [242] and bined effect enables the ants to minimize absorption from
ultrathin invisibility cloaks [243]. solar radiation, and to efficiently dissipate heat back to the
We see a number of promising areas in fundamental surroundings via blackbody radiation. Animals and plants
research and practical applications where metasurfaces could living in extreme environments could provide us valuable
have an important impact: scientific and engineering lessons on optical design and
thermal management. In general, by designing the struc-
(1) Dispersionless flat lenses. Flat lenses that are able to
tural hierarchy, compositional heterogeneity, and local
correct chromatic aberration over a broad wavelength
anisotropy of metasurface structures, one could create
range, and reduce spherical aberration, coma, and other
coatings that are optically thin and have desired spectral
monochromatic aberrations, could revolutionize optical
properties (reflectivity, absorptivity, transmissivity, and
instrumentation. Substantially shrinking the complexity
emissivity) over an extremely broad electromagnetic
and size of optical instruments (e.g. replacing the entire
spectral range. Such ultra-thin and ultra-broadband
set of compound lenses in a camera lens with a few dis-
metasurfaces will open doors to a variety of new applica-
persionless and aberration-corrected flat lenses) seems
tions, including control of radiative heat transfer, infrared
feasible in view of recent developments of metasurface
camouflage and structural coloration.
lenses.
(4) New material platforms for metasurfaces. Investigations
(2) Optical modulators and spatial light modulators (SLMs)
of materials with low losses, tunability, high melting
in the mid-infrared and THz spectral range. The lack
point, and CMOS compatibility for metamaterials and
of compact and fast modulators and SLMs has been
metasurfaces have been very active in recent years.
a big challenge that prevents the wide applications of
Transition-metal nitrides such as TiN show comparable
mid-infrared and THz technology in free space commu-
optical properties as gold in the visible and infrared but
nications, imaging, LIDAR (light detection and ranging),
have much higher melting points [249–251], a prop-
and homeland security (e.g. remote sensing, surveillance,
erty that can be explored for metasurface applications
and navigation in severe environments, such as foggy and
involving high optical intensity. Transparent conducting
dusty weather). Metasurfaces provide an ideal platform
oxides (TCOs) such as indium-tin-oxide enable one to
to create flat modulators in the mid-infrared and THz
control the spectral location of the epsilon-near-zero
regimes as they enable a strong interaction between light
point, which is associated with enhanced optical near-
and materials with tunable optical properties, and allow
fields; the resulting strong interaction between light
for introducing spatially-varying optical response. Strong
and TCOs can be exploited for optical modulation
light-material interactions enabled by metasurfaces allow
[252, 253] and nonlinear optics. Phase-change materials
for reducing the amount of tunable materials used so that
such as chalcogenide alloys that have been used in rewrit-
one can increase the modulation speed.
able CDs, DVDs, and Blu-ray discs, can be switched
(3) Radiative cooling metasurfaces. Metasurfaces that pos-
between the amorphous and crystalline states by laser
sess exceptional thermoregulatory properties have been
or electrical current pulses with controlled duration and
an emerging field of research and have the potential to
intensity [254, 255]. This material system has recently
make an important technological impact. Fan and col-
been used to demonstrate all-optical, non-volatile,
leagues are pioneering the research on radiative cooling
metasurface switch [256], and high-resolution solid-state
metasurfaces [244–246], which have strong reflectivity
displays [257]. SmNiO3, a prototypical phase-transition
in the solar radiation spectrum and enhanced emissivity
perovskite nickelate, exhibits non-volatile and revers-
in the thermal radiation spectrum. Metasurfaces based on
ible large refractive index changes over an ultra-broad
multilayered thin films have demonstrated in experiments
spectral range, from the visible to the long-wavelength
passive cooling of objects to a few degrees below the
mid-infrared. The super broadband performance is due
ambient air temperature under direct sunlight [246]. Chen
to strong electron correlation effects [258], and this new
and colleagues recently proposed a fabric that blocks sun-
mechanism can be exploited to create a variety of active
light and provides passive cooling via the transmission
photonic devices.
of thermal radiation emitted by the human body [247].
It is interesting to note that radiative cooling has always
been essential for the survival of animals living in harsh Acknowledgments
environmental conditions. Yu and co-workers recently
reported the thermoregulatory strategies that enable HTC acknowledges support in part from the Los Alamos
Saharan silver ants to forage in the midday sun on the National Laboratory LDRD Program. NY acknowledges
desert surface where temperatures can reach 70 °C (which support from NSF (grant ECCS-1307948), the AFOSR
is not survivable by their primary predators). It was found Multidisciplinary University Research Initiative program
36
(grant FA9550-14-1-0389), and DARPA Young Faculty Award [28] Yu N F and Capasso F 2014 Nat. Mater. 13 139–50
(grant D15AP00111). This work was performed, in part, at the [29] Walia S, Shah C M, Gutruf P, Nili H, Chowdhury D R,
Withayachumnankul W, Bhaskaran M and Sriram S 2015
Center for Integrated Nanotechnologies, a U.S. Department of
Appl. Phys. Rev. 2 011303
Energy, Office of Basic Energy Sciences Nanoscale Science [30] Minovich A E, Miroshnichenko A E, Bykov A Y,
Research Center operated jointly by Los Alamos and Sandia Murzina T V, Neshev D N and Kivshar Y S 2015 Laser
National Laboratories. Los Alamos National Laboratory, an Photonics Rev. 9 195–213
affirmative action/equal opportunity employer, is operated by [31] Genevet P and Capasso F 2015 Rep. Prog. Phys.
78 024401
Los Alamos National Security, LLC, for the National Nuclear
[32] Aieta F, Genevet P, Yu N, Kats M A, Gaburro Z and Capasso F
Security Administration of the U.S. Department of Energy 2012 Nano Lett. 12 1702–6
under Contract No. DE-AC52-06NA25396. [33] Yu N, Aieta F, Genevet P, Kats M A, Gaburro Z and Capasso F
2012 Nano Lett. 12 6328–33
[34] Kats M A, Genevet P, Aoust G, Yu N, Blanchard R, Aieta F,
References Gaburro Z and Capasso F 2012 Proc. Natl Acad. Sci.
109 12364–8
[35] Zhang X Q, Tian Z, Yue W S, Gu J Q, Zhang S, Han J G and
[1] Wu L A, Long G L, Gong Q and Guo G C 2015 AAPPS Bull. Zhang W L 2013 Adv. Mater. 25 4567–72
25 6–10 [36] Sun S et al 2012 Nano Lett. 12 6223–9
[2] Pendry J B, Schurig D and Smith D R 2006 Science [37] Farmahini-Farahani M and Mosallaei H 2013 Opt. Lett.
312 1780–2 38 462–4
[3] Leonhardt U 2006 Science 312 1777–80 [38] Pozar D M, Targonski S D and Syrigos H 1997 IEEE Trans.
[4] Cai W and Shalaev V 2010 Optical Metamaterials: Antennas Propag. 45 287–96
Fundamentals and Applications (Heidelberg: Springer) [39] Gagnon N, Petosa A and McNamara D 2013 IEEE Antennas
[5] Engheta N and Ziolkowski R W 2006 Metamaterials: Physics Propag. Mag. 55 29–48
and Engineering Explorations (New York: Wiley) [40] Pancharatnam S 1956 Proc. Indian Acad. Sci. A 44 247–62
[6] Smith D R, Padilla W J, Vier D C, Nemat-Nasser S C and [41] Berry M V 1984 Proc. R. Soc. Lond. A 392 45–57
Schultz S 2000 Phys. Rev. Lett. 84 4184–7 [42] Hasman E, Biener G, Niv A and Kleiner V 2005 Space-
[7] Shelby R A, Smith D R and Schultz S 2001 Science Variant Polarization Manipulation (Progress in Optics
292 77–9 vol 47) ed E Wolf (Amsterdam: Elsevier)
[8] Pendry J 2000 Phys. Rev. Lett. 85 3966–9 [43] Menzel C, Rockstuhl C and Lederer F 2010 Phys. Rev. A
[9] Fang N, Lee H, Sun C and Zhang X 2005 Science 308 534–7 82 053811
[10] Schurig D, Mock J J, Justice B J, Cummer S A, Pendry J B, [44] Armitage N P 2014 Phys. Rev. B 90 035135
Starr A F and Smith D R 2006 Science 314 977–80 [45] Kang M, Feng T, Wang H T and Li J 2012 Opt. Express
[11] Soukoulis C M and Wegener M 2011 Nat. Photon. 5 523–30 20 15882–90
[12] Yu N F, Genevet P, Kats M A, Aieta F, Tetienne J P, Capasso F [46] Huang L, Chen X, Mühlenbernd H, Li G, Bai B, Tan Q, Jin G,
and Gaburro Z 2011 Science 334 333–7 Zentgraf T and Zhang S 2012 Nano Lett. 12 5750–5
[13] Ni X J, Emani N K, Kildishev A V, Boltasseva A and [47] Li J, Chen S, Yang H, Li J, Yu P, Cheng H, Gu C, Chen H T
Shalaev V M 2012 Science 335 427 and Tian J 2015 Adv. Funct. Mater. 25 704–10
[14] Sun S, He Q, Xiao S, Xu Q, Li X and Zhou L 2012 Nat. [48] Luo W, Xiao S, He Q, Sun S and Zhou L 2015 Adv. Opt.
Mater. 11 426–31 Mater. 3 1102–8
[15] Pendry J B, Holden A J, Robbins D J and Stewart W J 1999 [49] Bomzon Z, Kleiner V and Hasman E 2001 Opt. Lett.
IEEE Trans. Microw. Theory Tech. 47 2075–84 26 1424–6
[16] Chen H T, O’Hara J F, Taylor A J, Averitt R D, Highstrete C, [50] Bomzon Z, Biener G, Kleiner V and Hasman E 2002 Opt.
Lee M and Padilla W J 2007 Opt. Express 15 1084–95 Lett. 27 1141–3
[17] Munk B A 2000 Frequency Selective Surfaces: Theory and [51] Li G, Kang M, Chen S, Zhang S, Pun E Y B, Cheah K W and
Design (New York: Wiley) Li J 2013 Nano Lett. 13 4148–51
[18] Bharadwaj P, Deutsch B and Novotny L 2009 Adv. Opt. [52] Pfeiffer C and Grbic A 2013 Phys. Rev. Lett. 110 197401
Photonics 1 438–83 [53] Pfeiffer C, Emani N K, Shaltout A M, Boltasseva A,
[19] Novotny L and van Hulst N 2011 Nat. Photon. 5 83–90 Shalaev V M and Grbic A 2014 Nano Lett. 14 2491–7
[20] Svirko Y, Zheludev N and Osipov M 2001 Appl. Phys. Lett. [54] Monticone F, Estakhri N M and Alù A 2013 Phys. Rev. Lett.
78 498–500 110 203903
[21] Zou L, Withayachumnankul W, Shah C M, Mitchell A, [55] Aieta F, Genevet P, Kats M A, Yu N, Blanchard R, Gaburro Z
Bhaskaran M, Sriram S and Fumeaux C 2013 Opt. Express and Capasso F 2012 Nano Lett. 12 4932–6
21 1344–52 [56] Grady N K, Heyes J E, Chowdhury D R, Zeng Y, Reiten M T,
[22] Walther B, Helgert C, Rockstuhl C, Setzpfandt F, Azad A K, Taylor A J, Dalvit D A R and Chen H T 2013
Eilenberger F, Kley E B, Lederer F, Tuennermann A and Science 340 1304–7
Pertsch T 2012 Adv. Mater. 24 6300–4 [57] Li X, Xiao S, Cai B, He Q, Cui T J and Zhou L 2012 Opt. Lett.
[23] Lin J, Mueller J P B, Wang Q, Yuan G, Antoniou N, Yuan X C 37 4940–2
and Capasso F 2013 Science 340 331–4 [58] Ni X, Ishii S, Kildishev A V and Shalaev V M 2013 Light Sci.
[24] Cheng H, Liu Z, Chen S and Tian J 2015 Adv. Mater. Appl. 2 e72
27 5410–21 [59] Jiang X Y, Ye J S, He J W, Wang X K, Hu D, Feng S F, Kan Q
[25] Holloway C, Kuester E F, Gordon J, O’Hara J, Booth J and and Zhang Y 2013 Opt. Express 21 30030–8
Smith D 2012 IEEE Antennas Propag. Mag. 54 10–35 [60] Memarzadeh B and Mosallaei H 2011 Opt. Lett. 36 2569–71
[26] Yu N, Genevet P, Aieta F, Kats M, Blanchard R, Aoust G, [61] Chen X, Huang L, Mühlenbernd H, Li G, Bai B, Tan Q, Jin G,
Tetienne J P, Gaburro Z and Capasso F 2013 IEEE J. Sel. Qiu C W, Zhang S and Zentgraf T 2012 Nat. Commun.
Top. Quantum Electron. 19 4700423 3 1198
[27] Kildishev A V, Boltasseva A and Shalaev V M 2013 Science [62] Pors A, Nielsen M G, Eriksen R and Bozhevolnyi S I 2013
339 1232009 Nano Lett. 13 829–34
37
[63] Khorasaninejad M, Aieta F, Kanhaiya P, Kats M A, Genevet P, [101] Fedotov V A, Mladyonov P L, Prosvirnin S L,
Rousso D and Capasso F 2015 Nano Lett. 15 5358–62 Rogacheva A V, Chen Y and Zheludev N I 2006
[64] Pfeiffer C and Grbic A 2014 Phys. Rev. Appl. 2 044012 Phys. Rev. Lett. 97 167401
[65] Genevet P, Yu N, Aieta F, Lin J, Kats M A, Blanchard R, [102] Schwanecke A S, Fedotov V A, Khardikov V V,
Scully M O, Gaburro Z and Capasso F 2012 Appl. Phys. Prosvirnin S L, Chen Y and Zheludev N I 2008 Nano Lett.
Lett. 100 013101 8 2940–3
[66] Ni X, Kildishev A V and Shalaev V M 2013 Nat. Commun. [103] Fedotov V A, Schwanecke A S, Zheludev N I,
4 2807 Khardikov V V and Prosvirnin S L 2007 Nano Lett.
[67] Zheng G, Mühlenbernd H, Kenney M, Li G, Zentgraf T and 7 1996–9
Zhang S 2015 Nat. Nanotechnol. 10 308–12 [104] Drezet A, Genet C, Laluet J Y and Ebbesen T W 2008
[68] Strikwerda A C, Fan K, Tao H, Pilon D V, Zhang X and Opt. Express 16 12559–70
Averitt R D 2009 Opt. Express 17 136–49 [105] Menzel C, Helgert C, Rockstuhl C, Kley E B,
[69] Khoo E H, Li E P and Crozier K B 2011 Opt. Lett. Tünnermann A, Pertsch T and Lederer F 2010 Phys. Rev.
36 2498–500 Lett. 104 253902
[70] Zhao Y and Alù A 2011 Phys. Rev. B 84 205428 [106] Huang C, Feng Y, Zhao J, Wang Z and Jiang T 2012 Phys.
[71] Cong L Q, Cao W, Tian Z, Gu J Q, Han J G and Zhang W L Rev. B 85 195131
2012 New J. Phys. 14 115013 [107] Li Z et al 2014 Appl. Phys. Lett. 105 201103
[72] Zhu L, Meng F Y, Dong L, Fu J H, Zhang F and Wu Q 2013 [108] Li Z, Chen S, Liu W, Cheng H, Liu Z, Li J, Yu P, Xie B and
Opt. Express 21 32099–110 Tian J 2015 Plasmonics 10 1703–11
[73] Wang D, Gu Y, Gong Y, Qiu C W and Hong M 2015 Opt. [109] Han J, Li H, Fan Y, Wei Z, Wu C, Cao Y, Yu X, Li F and
Express 23 11114–22 Wang Z 2011 Appl. Phys. Lett. 98 151908
[74] Chin J Y, Lu M Z and Cui T J 2008 Appl. Phys. Lett. 93 251903 [110] Shi J H, Ma H F, Guan C Y, Wang Z P and Cui T J 2014
[75] Li T, Wang S M, Cao J X, Liu H and Zhu S N 2010 Appl. Phys. Rev. B 89 165128
Phys. Lett. 97 261113 [111] Mutlu M, Akosman A E, Serebryannikov A E and Ozbay E
[76] Peralta X G, Smirnova E I, Azad A K, Chen H T, Taylor A J, 2012 Phys. Rev. Lett. 108 213905
Brener I and O’Hara J F 2009 Opt. Express 17 773–83 [112] Richtmyer R D 1939 J. Appl. Phys. 10 391–8
[77] Young L, Robinson L A and Hacking C A 1973 IEEE Trans. [113] Long S A, McAllister M and Liang S 1983 IEEE Trans.
Antennas Propag. 21 376–8 Antennas Propag. 31 406–12
[78] Hao J M, Yuan Y, Ran L X, Jiang T, Kong J A, Chan C T and [114] Mongia R K and Bhartia P 1994 Int. J. Microw. Millim. Wave
Zhou L 2007 Phys. Rev. Lett. 99 063908 Comput. Aided Eng. 4 230–47
[79] Pors A, Nielsen M G, Valle G D, Willatzen M, Albrektsen O [115] Petosa A and Ittipiboon A 2010 IEEE Antennas Propag.
and Bozhevolnyi S I 2011 Opt. Lett. 36 1626–8 Mag. 52 91–116
[80] Wang F, Chakrabarty A, Minkowski F, Sun K and Wei Q H [116] Wheeler M S, Aitchison J S and Mojahedi M 2005 Phys. Rev.
2012 Appl. Phys. Lett. 101 023101 B 72 193103
[81] Watts C M, Liu X L and Padilla W J 2012 Adv. Mater. [117] Ahmadi A and Mosallaei H 2008 Phys. Rev. B 77 045104
24 OP98–120 [118] Evlyukhin A B, Reinhardt C, Seidel A, Luk’yanchuk B S and
[82] Chen H T 2012 Opt. Express 20 7165–72 Chichkov B N 2010 Phys. Rev. B 82 045404
[83] Sun W J, He Q O, Hao J M and Zhou L 2011 Opt. Lett. [119] Evlyukhin A B, Reinhardt C and Chichkov B N 2011 Phys.
36 927–9 Rev. B 84 235429
[84] Zhou L, Wen W J, Chan C T and Sheng P 2005 Phys. Rev. [120] Kuznetsov A I, Miroshnichenko A E, Fu Y H, Zhang J and
Lett. 94 243905 Luk’yanchuk B 2012 Sci. Rep. 2 492
[85] Chen H T, Zhou J F, O’Hara J F, Chen F, Azad A K and [121] Zhao Q, Kang L, Du B, Zhao H, Xie Q, Huang X, Li B,
Taylor A J 2010 Phys. Rev. Lett. 105 073901 Zhou J and Li L 2008 Phys. Rev. Lett. 101 027402
[86] Cong L Q, Xu N N, Gu J Q, Singh R, Han J G and Zhang W L [122] Vynck K, Felbacq D, Centeno E, Căbuz A I, Cassagne D and
2014 Laser Photonics Rev. 8 626–32 Guizal B 2009 Phys. Rev. Lett. 102 133901
[87] Li Y F, Zhang J Q, Qu S B, Wang J F, Zheng L, Pang Y Q, [123] O’Brien S and Pendry J B 2002 J. Phys.: Condens. Matter
Xu Z and Zhang A X 2015 J. Appl. Phys. 117 044501 14 4035
[88] Pfeiffer C and Grbic A 2014 Phys. Rev. Appl. 2 044011 [124] Zhao Q, Zhou J, Zhang F and Lippens D 2009 Mater. Today
[89] Pfeiffer C, Zhang C, Ray V, Guo L J and Grbic A 2014 Phys. 12 60–9
Rev. Lett. 113 023902 [125] Peng L, Ran L, Chen H, Zhang H, Kong J A and
[90] Chen H Y, Wang J F, Ma H, Qu S B, Xu Z, Zhang A X, Grzegorczyk T M 2007 Phys. Rev. Lett. 98 157403
Yan M B and Li Y F 2014 J. Appl. Phys. 115 154504 [126] Ginn J C et al 2012 Phys. Rev. Lett. 108 097402
[91] Dai Y M, Ren W Z, Cai H B, Ding H Y, Pan N and Wang X P [127] Schuller J A, Zia R, Taubner T and Brongersma M L 2007
2014 Opt. Express 22 7465–72 Phys. Rev. Lett. 99 107401
[92] Hao J, Ren Q, An Z, Huang X, Chen Z, Qiu M and Zhou L [128] Schuller J A, Taubner T and Brongersma M L 2009 Nat.
2009 Phys. Rev. A 80 023807 Photon. 3 658–61
[93] Yang Y M, Wang W Y, Moitra P, Kravchenko I I, Briggs D P [129] Staude I et al 2013 ACS Nano 7 7824–32
and Valentine J 2014 Nano Lett. 14 1394–9 [130] Evlyukhin A B, Novikov S M, Zywietz U, Eriksen R L,
[94] Ye Y Q and He S 2010 Appl. Phys. Lett. 96 203501 Reinhardt C, Bozhevolnyi S I and Chichkov B N 2012
[95] Mutlu M and Ozbay E 2012 Appl. Phys. Lett. 100 051909 Nano Lett. 12 3749–55
[96] Shi H Y, Zhang A X, Zheng S, Li J X and Jiang Y S 2014 [131] Cheng J, Ansari-Oghol-Beig D and Mosallaei H 2014 Opt.
Appl. Phys. Lett. 104 034102 Lett. 39 6285–8
[97] Chiang Y J and Yen T J 2013 Appl. Phys. Lett. 102 011129 [132] Decker M, Staude I, Falkner M, Dominguez J, Neshev D N,
[98] Cong L Q, Cao W, Zhang X Q, Tian Z, Gu J Q, Singh R, Brener I, Pertsch T and Kivshar Y S 2015 Adv. Opt. Mater.
Han J G and Zhang W L 2013 Appl. Phys. Lett. 103 171107 3 813–20
[99] Li Y F, Zhang J Q, Qu S B, Wang J F, Zheng L, Zhang A X [133] Yang Y, Kravchenko I I, Briggs D P and Valentine J 2014
and Xu Z 2014 Electron. Lett. 50 1658–9 Nat. Commun. 5 5753
[100] Liu D J, Xiao Z Y, Ma X L and Wang Z H 2015 [134] Kerker M, Wang D S and Giles C L 1983 J. Opt. Soc. Am.
Opt. Commun. 354 272–6 73 765–7
38
[135] Geffrin J M et al 2012 Nat. Commun. 3 1171 [170] Seren H R, Keiser G R, Cao L, Zhang J,
[136] Fu Y H, Kuznetsov A I, Miroshnichenko A E, Yu Y F and Strikwerda A C, Fan K, Metcalfe G D, Wraback M,
Luk’yanchuk B 2013 Nat. Commun. 4 1527 Zhang X and Averitt R D 2014 Adv. Opt. Mater.
[137] Person S, Jain M, Lapin Z, Sáenz J J, Wicks G and 2 1221–6
Novotny L 2013 Nano Lett. 13 1806–9 [171] Zhang S, Zhou J, Park Y S, Rho J, Singh R, Nam S,
[138] Slovick B, Yu Z G, Berding M and Krishnamurthy S 2013 Azad A K, Chen H T, Yin X, Taylor A J and Zhang X
Phys. Rev. B 88 165116 2012 Nat. Commun. 3 942
[139] Moitra P, Slovick B A, Gang Yu Z, Krishnamurthy S and [172] Zhou J, Chowdhury D R, Zhao R, Azad A K, Chen H T,
Valentine J 2014 Appl. Phys. Lett. 104 171102 Soukoulis C M, Taylor A J and O’Hara J F 2012 Phys.
[140] Moitra P, Slovick B A, Li W, Kravchencko I I, Briggs D P, Rev. B 86 035448
Krishnamurthy S and Valentine J 2015 ACS Photonics [173] Gu J et al 2012 Nat. Commun. 3 1151
2 692–8 [174] Chen H T, Padilla W J, Zide J M O, Gossard A C, Taylor A J
[141] Liu S et al 2014 Optica 1 250–6 and Averitt R D 2006 Nature 444 597–600
[142] Headland D, Nirantar S, Withayachumnankul W, Gutruf P, [175] Chen H T, Padilla W J, Cich M J, Azad A K, Averitt R D and
Abbott D, Bhaskaran M, Fumeaux C and Sriram S 2015 Taylor A J 2009 Nat. Photon. 3 148–51
Adv. Mater. 27 7137–44 [176] Chen H T et al 2008 Appl. Phys. Lett. 93 091117
[143] Sautter J, Staude I, Decker M, Rusak E, Neshev D N, [177] Shrekenhamer D, Rout S, Strikwerda A C, Bingham C,
Brener I and Kivshar Y S 2015 ACS Nano 9 4308–15 Averitt R D, Sonkusale S and Padilla W J 2011 Opt.
[144] Iyer P P, Butakov N A and Schuller J A 2015 ACS Photonics Express 19 9968–75
2 1077–84 [178] Shrekenhamer D, Montoya J, Krishna S and Padilla W J
[145] Liu X, Zhao Q, Lan C and Zhou J 2013 Appl. Phys. Lett. 2013 Adv. Opt. Mater. 1 905–9
103 031910 [179] Zhang Y et al 2015 Nano Lett. 15 3501–6
[146] Zou C, Withayachumnankul W, Shadrivov I V, Kivshar Y S [180] Karl N, Reichel K, Chen H T, Taylor A J, Brener I, Benz A,
and Fumeaux C 2015 Phys. Rev. B 91 085433 Reno J L, Mendis R and Mittleman D M 2014 Appl. Phys.
[147] Chong K E et al 2015 Nano Lett. 15 5369–74 Lett. 104 091115
[148] Arbabi A, Horie Y, Ball A J, Bagheri M and Faraon A 2015 [181] Chan W L, Chen H T, Taylor A J, Brener I, Cich M J
Nat. Commun. 6 7069 and Mittleman D M 2009 Appl. Phys. Lett.
[149] Aieta F, Kats M A, Genevet P and Capasso F 2015 Science 94 213511
347 1342–5 [182] Watts C M, Shrekenhamer D, Montoya J, Lipworth G,
[150] Lin D M, Fan P Y, Hasman E and Brongersma M L 2014 Hunt J, Sleasman T, Krishna S, Smith D R and Padilla W J
Science 345 298–302 2014 Nat. Photon. 8 605–9
[151] Arbabi A, Horie Y, Bagheri M and Faraon A 2015 [183] Fan K, Zhao X, Zhang J, Geng K, Keiser G, Seren H,
Nat. Nanotechnol. 10 937–43 Metcalfe G, Wraback M, Zhang X and Averitt R 2013
[152] Fong B, Colburn J, Ottusch J, Visher J and Sievenpiper D IEEE Trans. Terahertz Sci. Technol. 3 702–8
2010 IEEE Trans. Antennas Propag. 58 3212–21 [184] Emani N K, Kildishev A V, Shalaev V M and Boltasseva A
[153] Maci S, Minatti G, Casaletti M and Bosiljevac M 2011 IEEE 2015 Nanophotonics 4 214–23
Antennas Wirel. Propag. Lett. 10 1499–502 [185] Sensale-Rodriguez B, Yan R, Kelly M M, Fang T, Tahy K,
[154] Podilchak S, Matekovits L, Freundorfer A, Antar Y and Hwang W S, Jena D, Liu L and Xing H G 2012 Nat.
Orefice M 2013 IEEE Trans. Antennas Propag. 61 4957–72 Commun. 3 780
[155] Minatti G, Caminita F, Casaletti M and Maci S 2011 IEEE [186] Ju L, Geng B, Horng J, Girit C, Martin M, Hao Z,
Trans. Antennas Propag. 59 4436–44 Bechtel H A, Liang X, Zettl A, Shen Y R and Wang F
[156] Genevet P, Lin J, Kats M A and Capasso F 2012 Nat. 2011 Nat. Nanotechnol. 6 630–4
Commun. 3 1278 [187] García de Abajo F J 2014 ACS Photonics 1 135–52
[157] Qu C, Xiao S, Sun S, He Q and Zhou L 2013 Europhys. Lett. [188] Lee S H et al 2012 Nat. Mater. 11 936–41
101 54002 [189] Gao W et al 2014 Nano Lett. 14 1242–8
[158] Sun W, He Q, Sun S and Zhou L 2016 Light Sci. Appl. [190] Miao Z, Wu Q, Li X, He Q, Ding K, An Z, Zhang Y and
5 e16003 Zhou L 2015 Phys. Rev. X 5 041027
[159] Vakil A and Engheta N 2011 Science 332 1291–4 [191] Yao Y, Kats M A, Genevet P, Yu N, Song Y, Kong J and
[160] Gok G and Grbic A 2013 Phys. Rev. Lett. 111 233904 Capasso F 2013 Nano Lett. 13 1257–64
[161] Yu N 2014 Conf. on Lasers and Electro-Optics [192] Yao Y, Kats M A, Shankar R, Song Y, Kong J, Loncar M and
(San Jose, CA) Capasso F 2014 Nano Lett. 14 214–9
[162] Li Z, Kim M H and Yu N 2015 Conf. on Lasers and [193] Yao Y, Shankar R, Kats M A, Song Y, Kong J, Loncar M and
Electro-Optics (San Jose, CA) Capasso F 2014 Nano Lett. 14 6526–32
[163] Shadrivov I V, Morrison S K and Kivshar Y S 2006 [194] Dabidian N, Kholmanov I, Khanikaev A B, Tatar K,
Opt. Express 14 9344–9 Trendafilov S, Mousavi S H, Magnuson C, Ruoff R S and
[164] Zhu B O, Zhao J and Feng Y 2013 Sci. Rep. 3 3059 Shvets G 2015 ACS Photonics 2 216–27
[165] Padilla W J, Taylor A J, Highstrete C, Lee M and Averitt R D [195] Jang M S, Brar V W, Sherrott M C, Lopez J J, Kim L, Kim S,
2006 Phys. Rev. Lett. 96 107401 Choi M and Atwater H A 2014 Phys. Rev. B 90 165409
[166] Chen H T, Padilla W J, Zide J M O, Bank S R, Gossard A C, [196] Dabidian N, Dutta-Gupta S, Kholmanov I, Lu F, Lee J,
Taylor A J and Averitt R D 2007 Opt. Lett. 32 1620–2 Lai K, Jin M, Fallahazad B, Tutuc E, Belkin M A and
[167] Chen H T, O’Hara J F, Azad A K, Taylor A J, Averitt R D, Shvets G 2015 arXiv:1510.04260
Shrekenhamer D B and Padilla W J 2008 Nat. Photon. [197] Mousavi S H et al 2013 Nano Lett. 13 1111–7
2 295–8 [198] Goldflam M D et al 2014 Appl. Phys. Lett. 105 041117
[168] Shen N H, Massaouti M, Gokkavas M, Manceau J M, [199] Singh R, Azad A K, Jia Q X, Taylor A J and Chen H T 2011
Ozbay E, Kafesaki M, Koschny T, Tzortzakis S and Opt. Lett. 36 1230–2
Soukoulis C M 2011 Phys. Rev. Lett. 106 037403 [200] Driscoll T et al 2008 Appl. Phys. Lett. 93 024101
[169] Heyes J E, Withayachumnankul W, Grady N K, [201] Driscoll T, Kim H T, Chae B G, Kim B J, Lee Y W,
Chowdhury D R, Azad A K and Chen H T 2014 Appl. Jokerst N M, Palit S, Smith D R, Di Ventra M and
Phys. Lett. 105 181108 Basov D N 2009 Science 325 1518–21
39
[202] Dicken M J, Aydin K, Pryce I M, Sweatlock L A, Boyd E M, [228] Helm M 2000 Chapter: The basic physics of intersubband
Walavalkar S, Ma J and Atwater H A 2009 Opt. Express transitions Intersubband Transitions in Quantum Wells:
17 18330–9 Physics and Device Applications (Semiconductors and
[203] Wen Q Y, Zhang H W, Yang Q H, Xie Y S, Chen K and Semimetals vol 62) (New York: Academic)
Liu Y L 2010 Appl. Phys. Lett. 97 021111 [229] Powell D A, Shadrivov I V, Kivshar Y S and Gorkunov M V
[204] Goldflam M D et al 2011 Appl. Phys. Lett. 99 044103 2007 Appl. Phys. Lett. 91 144107
[205] Kats M A, Blanchard R, Genevet P, Yang Z, Qazilbash M M, [230] Wang B, Zhou J, Koschny T and Soukoulis C M 2008 Opt.
Basov D N, Ramanathan S and Capasso F 2013 Opt. Lett. Express 16 16058–63
38 368–70 [231] Huang D, Poutrina E and Smith D R 2010 Appl. Phys. Lett.
[206] Zhao Q, Kang L, Du B, Li B, Zhou J, Tang H, Liang X and 96 104104
Zhang B 2007 Appl. Phys. Lett. 90 011112 [232] Shadrivov I V, Kozyrev A B, van der Weide D W and
[207] Shrekenhamer D, Chen W C and Padilla W J 2013 Phys. Rev. Kivshar Y S 2008 Appl. Phys. Lett. 93 161903
Lett. 110 177403 [233] Rose A, Huang D and Smith D R 2011 Phys. Rev. Lett.
[208] Chen H T, Yang H, Singh R, O’Hara J F, Azad A K, 107 063902
Trugman S A, Jia Q X and Taylor A J 2010 Phys. Rev. [234] Klein M W, Enkrich C, Wegener M and Linden S 2006
Lett. 105 247402 Science 313 502–4
[209] Gu J, Singh R, Tian Z, Cao W, Xing Q, He M, Zhang J W, [235] Klein M W, Wegener M, Feth N and Linden S 2007 Opt.
Han J, Chen H T and Zhang W 2010 Appl. Phys. Lett. Express 15 5238–47
97 071102 [236] Ciracì C, Poutrina E, Scalora M and Smith D R 2012 Phys.
[210] Wu J, Jin B, Xue Y, Zhang C, Dai H, Zhang L, Cao C, Rev. B 85 201403
Kang L, Xu W, Chen J and Wu P 2011 Opt. Express [237] Luo L, Chatzakis I, Wang J, Niesler F B P, Wegener M,
19 12021–6 Koschny T and Soukoulis C M 2014 Nat. Commun.
[211] Kurter C, Tassin P, Zhang L, Koschny T, Zhuravel A P, 5 3055
Ustinov A V, Anlage S M and Soukoulis C M 2011 Phys. [238] O’Brien K, Suchowski H, Rho J, Salandrino A, Kante B,
Rev. Lett. 107 043901 Yin X and Zhang X 2015 Nat. Mater. 14 379–83
[212] Singh R, Xiong J, Azad A K, Yang H, Trugman S A, [239] Wolf O et al 2015 Nat. Commun. 6 7667
Jia Q X, Taylor A J and Chen H T 2012 Nanophotonics [240] Segal N, Keren-Zur S, Hendler N and Ellenbogen T 2015
1 117–23 Nat. Photon. 9 180–4
[213] Tao H, Strikwerda A C, Fan K, Padilla W J, Zhang X and [241] Zhang B, Hendrickson J, Nader N, Chen H T and Guo J 2014
Averitt R D 2009 Phys. Rev. Lett. 103 147401 Appl. Phys. Lett. 105 241113
[214] Zhu W M, Liu A Q, Zhang X M, Tsai D P, Bourouina T, [242] Yin X, Ye Z, Rho J, Wang Y and Zhang X 2013 Science
Teng J H, Zhang X H, Guo H C, Tanoto H, Mei T, Lo G Q 339 1405–7
and Kwong D L 2011 Adv. Mater. 23 1792–6 [243] Ni X, Wong Z J, Mrejen M, Wang Y and Zhang X 2015
[215] Ma F, Lin Y S, Zhang X and Lee C 2014 Light Sci. Appl. Science 349 1310–4
3 e171 [244] Rephaeli E, Raman A and Fan S 2013 Nano Lett.
[216] Han Z, Kohno K, Fujita H, Hirakawa K and Toshiyoshi H 13 1457–61
2014 Opt. Express 22 21326–39 [245] Zhu L, Raman A, Wang K X, Anoma M A and Fan S 2014
[217] Kan T, Isozaki A, Kanda N, Nemoto N, Konishi K, Optica 1 32–8
Takahashi H, Kuwata-Gonokami M, Matsumoto K and [246] Raman A P, Anoma M A, Zhu L, Rephaeli E and Fan S 2014
Shimoyama I 2015 Nat. Commun. 6 8422 Nature 515 540–4
[218] Liu M et al 2012 Nature 487 345–8 [247] Tong J K, Huang X, Boriskina S V, Loomis J, Xu Y and
[219] Fan K et al 2013 Phys. Rev. Lett. 110 217404 Chen G 2015 ACS Photonics 2 769–78
[220] Seren H R, Zhang J, Keiser G R, Maddox S J, Zhao X, abd [248] Shi N N, Tsai C C, Camino F, Bernard G D, Yu N and
Seth R. Bank K F, Zhang X and Averitt R D 2016 Light Wehner R 2015 Science 349 298–301
Sci. Appl. (doi: 10.1038/lsa.2016.78) [249] Naik G V, Kim J and Boltasseva A 2011 Opt. Mater. Express
[221] Grady N K et al 2013 New J. Phys. 15 105016 1 1090–9
[222] Zhang C, Jin B, Han J, Kawayama I, Murakami H, Wu J, [250] Boltasseva A and Atwater H A 2011 Science 331 290–1
Kang L, Chen J, Wu P and Tonouchi M 2013 Appl. Phys. [251] Guler U, Boltasseva A and Shalaev V M 2014 Science
Lett. 102 081121 344 263–4
[223] Lee J, Tymchenko M, Argyropoulos C, Chen P Y, Lu F, [252] Yi F, Shim E, Zhu A Y, Zhu H, Reed J C and Cubukcu E
Demmerle F, Boehm G, Amann M C, Alù A and 2013 Appl. Phys. Lett. 102 221102
Belkin M A 2014 Nature 511 65–9 [253] Park J, Kang J H, Liu X and Brongersma M L 2015 Sci. Rep.
[224] Campione S, Benz A, Sinclair M B, Capolino F and Brener I 5 15754
2014 Appl. Phys. Lett. 104 131104 [254] Hudgens S and Johnson B 2004 MRS Bull. 29 829–32
[225] Ahn D and Chuang S L 1987 IEEE J. Quantum Electron. [255] Wuttig M and Yamada N 2007 Nat. Mater. 6 824–32
23 2196–204 [256] Gholipour B, Zhang J, MacDonald K F, Hewak D W and
[226] Capasso F, Sirtori C and Cho A Y 1994 IEEE J. Quantum Zheludev N I 2013 Adv. Mater. 25 3050–4
Electron. 30 1313–26 [257] Hosseini P, Wright C D and Bhaskaran H 2014 Nature
[227] Gmachl C, Belyanin A, Sivco D L, Peabody M L, 511 206–11
Owschimikow N, Sergent A, Capasso F and Cho A 2003 [258] Shi J, Zhou Y and Ramanathan S 2014 Nat. Commun.
IEEE J. Quantum Electron. 39 1345–55 5 4860
40
View publication stats

A Review of Metasurfaces - Physics and Applications

Uploaded by

Copyright:

Available Formats

A Review of Metasurfaces - Physics and Applications

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Review of Metasurfaces - Physics and Applications

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

A review of metasurfaces: Physics and applications

Article in Reports on Progress in Physics · June 2016

Hou-Tong Chen Nanfang yu

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

A review of metasurfaces: physics and applications

2016 Rep. Prog. Phys. 79 076401

Please note that terms and conditions apply.

Rep. Prog. Phys. 79 (2016) 076401 (40pp) doi:10.1088/0034-4885/79/7/076401

A review of metasurfaces: physics and

E-mail: [email protected], [email protected] and [email protected]

Received 15 December 2015, revised 10 March 2016

Invited by Masud Mansuripur

Keywords: metamaterials and metasurfaces, wavefront shaping and beam forming,

0034-4885/16/076401+40$33.00 1 © 2016 IOP Publishing Ltd Printed in the UK

1. Introduction metasurfaces and can be considered as the two-dimensional

active and nonlinear metasurfaces in section 7, and in the last

2. Anomalous reflection and refraction

2.1. Generalized laws of reflection and refraction

When a plane electromagnetic wave encounters a boundary

When designing metasurfaces for polarization conversion

4.1. Linear-to-circular polarization conversion

An antenna array backed with a ground plane has been widely

ellipticity is about 0.99, as shown in figure 10(b). Circular-to-

4.2. Linear polarization rotation

Planar chiral response can yield optical activity, rotating the

at the center, similar to the case of magnetic resonant response

effective index of the surface current, which is assumed to be

has been further applied to investigate nonlinear metasurfaces

View publication stats

You might also like