Journal of

Materials Chemistry C
View Article Online
REVIEW View Journal | View Issue
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Information metamaterials and metasurfaces

Cite this: J. Mater. Chem. C, 2017,
Tie Jun Cui,*ab Shuo Liuab and Lei Zhangab
5, 3644
Traditionally, ‘‘metamaterials’’ have been described by effective medium parameters due to the sub-
wavelength nature of unit particles. The continuous nature of medium parameters makes traditional
metamaterials behave as analog metamaterials. Recently, the concept of coding metamaterials or
‘‘metasurfaces’’ has been proposed, in which metamaterials are characterized by digital coding particles
of ‘‘0’’ and ‘‘1’’ with opposite phase responses. It has been demonstrated that electromagnetic waves can
be manipulated by changing the coding sequences of ‘‘0’’ and ‘‘1’’. The coding particles provide a link
between the physical world and digital world, leading to digital metamaterials and even field programmable
metamaterials, which can be used to control electromagnetic waves in real time. The digital coding
representation of metamaterials or metasurfaces can also allow the concepts and signal processing
methods in information science to be introduced to physical metamaterials, thereby realizing extreme
Received 3rd February 2017, control of electromagnetic waves. Such studies have set the foundation of information metamaterials
Accepted 14th March 2017 and metasurfaces. In this review article, the coding, digital, and field programmable metamaterials and
DOI: 10.1039/c7tc00548b metasurfaces are systematically summarized and analyzed with particular emphases on the information
and digital convolution aspects. The future trend of information metamaterial/metasurface is predicted,
rsc.li/materials-c including software-defined metamaterials/metasurfaces and cognitive metamaterials/metasurfaces.

1. Introduction
‘‘Metamaterials’’ are artificial structures engineered to achieve
a
State Key Laboratory of Millimeter Waves, Department of Radio Engineering,
unique electromagnetic (EM) properties that are unavailable
Southeast University, Nanjing 210096, People’s Republic of China.
E-mail: [email protected]
in nature. In the past decade, metamaterials described by the
b
Synergetic Innovation Center of Wireless Communication Technology, effective medium theory have been widely used to manipulate
Southeast University, Nanjing 210096, People’s Repulic of China EM waves.1 They have elicited many exciting physical phenomena,

Tie Jun Cui received his doctorate Shuo Liu received a BE degree in
from Xidian University, China, in information engineering from the
1993. He was an Alexander von School of Information Science and
Humboldt Foundation Research Engineering, Southeast University,
Fellow at the University of Nanjing, China, in 2010. He is
Karlsruhe, Germany (2005–2007), currently working towards his
a Postdoctoral Research Associate doctorate at the State Key
(1997–2000) and Research Scien- Laboratory of Millimeter Waves,
tist (2001–2002) at the University Department of Radio Engineering,
of Illinois at Urbana-Champaign, Southeast University, Nanjing,
USA. In 2002, he joined the China. In 2016, he received the
Department of Radio Engineering, grand prize of the Baosteel
Tie Jun Cui Southeast University, China, as Shuo Liu Student Scholarship. A series of
Cheung-Kong Professor. Dr Cui works related to coding meta-
is the author of two books (Springer, 2009; CRC, 2016), and has materials was selected as the top 30 best works in 2016 (Optics in
published over 400 peer-reviewed journal articles with more than 2016) by Optics & Photonics. His research includes metamaterials,
14 800 citations. His research interests include metamaterials, metasurfaces, spoof surface plasmon polaritons, and wireless
plasmonics in microwaves, and computational electromagnetics. power transfer.

3644 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

such as negative refraction,2,3 ‘‘perfect lenses’’,4–6 and ‘‘invisi- into SW with high efficiency.34 Surface plasmon polaritons
bility cloaking’’.7–13 (SPPs) are a type of surface wave tightly bound on a surface.
‘‘Metasurfaces’’ are the two-dimensional (2D) version of They can also be readily obtained by imparting an appropriate
metamaterials. They have attracted increasing attention due gradient phase profile to metasurfaces.35–37 Among the various
to their superior abilities to manipulate wavefronts of trans- types of metasurfaces, the geometric metasurfaces that encode
mitted and reflected EM waves.14–21 In general, metasurfaces a geometric phase into the orientation angle of the constituent
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

are composed of an array of artificial ‘‘meta-atoms’’, which are meta-atoms have shown great potential in manipulating the
commonly made of metal structures with specially designed polarization of light.37–42 They were developed based on the
size, geometry, and orientation to generate specific responses. concept of the Pancharatnam–Berry phase,43,44 which consists
Compared with three-dimensional (3D) bulk metamaterials, of an array of identical anisotropic subwavelength scatterers.
metasurfaces have a negligible layer thickness with respect to When the orientation angles of their optical axes vary, a specific
the working wavelength, thus occupying much less physical space phase is generated for circularly polarized waves, resulting in a
and with lower insertion loss. There is a significant difference space-variant phase-front modification. Based on convenient
between the characterization of bulk metamaterials and thin local-phase controls, geometric metasurfaces have been employed to
metasurfaces. The effective medium theory is usually adopted for demonstrate photonic spin-Hall effects at microwave frequencies.45
metamaterials,22,23 whereas the generalized sheet transition Holography has been widely used to record and reconstruct
condition14 is more appropriate for the modeling of metasurfaces wavefronts by storing and releasing phase information. Recently,
due to their ultrathin nature. One of the benefits of the planar highly efficient, broadband and 3D holographic imaging has
configuration of metasurfaces is that it can be easily imple- been realized using metasurfaces.46–49 In addition, several tech-
mented at terahertz waves and visible light spectra using standard niques for multiplexing information into a single metasurface
photolithography.21,24–28 hologram have been demonstrated,50–53 which allow the storage
Generally, the reflection and refraction of light or EM waves of multiple images at the same frequency. A metasurface con-
are governed by Snell’s law on the interface between two homo- sisting of an array of C-shaped split-ring resonators has also been
geneous and isotropic media. In 2011, the generalized Snell’s presented to control the phase and amplitude profiles of the
laws of reflection and refraction were proposed by introducing transmitted waves,54,55 which paves the way towards high-quality
an abrupt phase shift on a metasurface,24 resulting in a phase holographic imaging.56
discontinuity that could be used to manipulate the wavefronts of However, most metasurfaces can manipulate only the electric
lights. Since then, metasurfaces have experienced rapid develop- field. Hence, there is a tradeoff between an attainable transmis-
ment, producing many interesting devices to manipulate micro- sion phase and amplitude, especially if a single metallic layer is
waves, terahertz waves, and visible light.25–28 For instance, an utilized.24,54 In 2013, it was pointed out from a microwave net-
optical vortex was generated to demonstrate the concept of work perspective that it is theoretically impossible to obtain 2p
interfacial phase discontinuities.24 Instead of using a spiral phase coverage with unity co-polarized transmission for a single
phase plate or computer-generated hologram, vortex beams non-magnetic metasurface.57 Nevertheless, the beam bending
with helical equal-phase wavefronts can be easily achieved by and focusing devices implemented by such transmission-type
metasurfaces.29–33 The vortex beam carrying an orbital angular metasurfaces suffered from low efficiency. It was also mentioned
momentum can be used to improve the transmission speed in that, by cascading three layers of metasurfaces, it is possible to
optical communication systems.31 realize 2p phase coverage without sacrificing the amplitude for
Another interesting application of metasurfaces is the con- co-polarized transmission. Also in 2013, the Huygens principle
version of spatially propagating waves (PW) to surface waves was applied for the first time to design a dual-layered reflection-
(SW), in which a gradient phase profile is used to couple PW less metasurface that could easily provide a phase span of 2p and
which could realize beam shaping, steering, and focusing cap-
abilities with almost 100% efficiency.58 However, the non-planar
configuration of the Huygens surface structure poses serious
Lei Zhang was born in Anhui, fabrication challenges at micro- and nano-scales, greatly restrict-
China, in 1992. He received a ing its application in terahertz and optical spectra. Later,
BE degree in electromagnetic wave several attempts on the planarization of Huygens metasurfaces
propagation and antennae from were made to reduce the thickness and simplify the fabrication
Xidian University, Xi’an, China, process.59–64
in 2015. He is currently working Metasurfaces can also be used to construct planar devices,
towards his doctorate in electro- such as flat lenses, which are imparted with a specially designed
magnetics and microwave tech- phase profile to transform planar wavefronts into spherical
nology at Southeast University, wavefronts and to converge incident waves at a distance away
Najjing, Jiangsu, China. His from the lenses.65–69 The optical wavefronts in transmission or
research interests include meta- reflection types remain spherical as long as the incident plane
surfaces, metamaterials and wave is normally incident on the flat lenses. They can, therefore,
Lei Zhang antenna arrays. achieve high numerical-aperture focusing.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3645
 View Article Online

Journal of Materials Chemistry C Review

Polarization is one of the basic properties of EM waves.

Conventional methods for polarization control are extremely
demanding of material properties, and their performance is
limited. In contrast, metasurfaces have been demonstrated to
convert the polarization of EM waves with the advantages of
ultrathin thickness, broad bandwidth, and high efficiency.
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Ref. 70–72 show that circular polarizers have been developed

based on metasurfaces in microwave, terahertz, and optical frequen-
cies. In addition, several metasurface-based polarization converters
have been reported,27,73,74 which can rotate a linear polarized wave
into an orthogonal one.
Another fascinating research area is ‘‘tunable’’ metasurfaces,
which are mainly controlled by electrical,75–81 optical,82 mecha-
nical,83 and fluid84 means. By elaborately designing tunable
meta-atoms, the reconfigurable metasurfaces can manipulate
EM waves dynamically. In addition to the various applications of
Fig. 1 Relationship between physical world metamaterials, digital world, and
the metasurfaces mentioned above, they have been developed information metamaterials. They are linked through coding meta-particles,
for many other practical applications, including reductions of physical principles and signal processing algorithms.
radar cross sections (RCSs),85,86 ultrathin absorbers,87,88 high-
efficient quarter-wave plates,89 terahertz modulators,90–92 Bessel
beam generation,93,94 asymmetric transmissions,95,96 and non- coding metasurfaces proposed for generating Bessel beams and
linear properties.97,98 anomalous refractions94 at a terahertz frequency. More impor-
In 2014, Giovampaola and Engheta proposed the concept of tantly, physical principles and digital signal processing methods
‘‘digital metamaterials’’,99 which make use of appropriate can be applied on coding metasurfaces (Fig. 1). As a simple
spatial mixtures of ‘‘metamaterial bits’’ to construct elemental example, the digital convolution operation was performed on a
‘‘metamaterial bytes’’ with the desired effective medium para- coding metasurface to realize the arbitrary rotation of radiation
meters. Nevertheless, this digital representation still belongs to beam and continuous beam scanning in the upper half space.105
the effective medium theory. Also in 2014, Cui et al. introduced the Information entropy was also introduced to coding metasurfaces106
concepts of coding, digital, and programmable metamaterials.100 to measure the information of the coding patterns and scattering
Instead of using effective medium parameters, digital coding fields. Such findings set the foundation for information meta-
particles with specified phase responses were used to charac- materials and metasurfaces.119
terize metamaterials. For example, 1-bit coding metamaterials In this review article, we first introduce the concept of coding
are constructed by a sequence of binary coding particles ‘‘0’’ metasurfaces in the microwave frequency and show their powerful
and ‘‘1’’, which correspond to ‘‘0’’ and ‘‘p’’ phase responses, manipulations to EM waves with several examples. Then, we present
respectively. Also, 2-bit coding metamaterials are constructed several types of coding metasurfaces in the terahertz frequency,
by a sequence of four coding particles ‘‘00’’, ‘‘01’’, ‘‘10’’, and including a broadband diffusion metasurface, frequency-dependent
‘‘11’’, which have ‘‘0’’, ‘‘p/2’’, ‘‘p’’, and ‘‘3p/2’’ phase responses, dual-functional coding metasurface, highly efficient transmission-
respectively, and so on. type coding metasurface, and polarization-controlled anisotropic
The coding representation of metasurfaces simplifies the coding metasurface. By switching the spatial coding particles in a
design and optimization procedures of metamaterials owing to digital way, we further introduce digital and programmable meta-
the digitalization of meta-atom geometry. However, it also pro- materials and their new applications, including programmable
vides a link between the physical world and digital world (Fig. 1), antenna, and single-sensor and single-frequency imaging. Taking
leading to digital metamaterials and field programmable meta- advantage of digital coding representation, two information features
materials.100 By arranging coding particles on a 2D plane with of the coding metasurfaces, Shannon entropy and digital convolu-
optimized coding sequences, the coding metamaterials can be tion, are presented. Finally, we summarize future trends in informa-
used for manipulating microwaves in a more simple and efficient tion metamaterials and metasurfaces.
way, such as anomalous reflections and random diffusions.100
Then, the coding metasurfaces are extended to a terahertz
frequency to realize broadband and wide-angle diffusions based 2. Coding metamaterials and
on Minkowski closed-loop particles,101 and anomalous reflec- metasurfaces
tions with ring-shaped particles.102 The concept of coding has
also been extended from isotropic to anisotropic, and results in 2.1 Concept and method
polarization-controlled dual EM features.103 Based on the flexible The concept of coding metamaterials and metasurfaces was
designs of coding particles, frequency-dependent coding meta- initially proposed and demonstrated experimentally in the micro-
surfaces have been presented for realizing multiple function- wave frequency due to relatively easy fabrication and measure-
alities with a single coding pattern,104 and transmission-type ment.100 Conventional metamaterials are described by the

3646 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

effective medium theory,22,23 in which the constitutive parameters to construct 1-bit and 2-bit coding metasurfaces. We note from
are continuous, either in homogeneous (for periodic structures) or Fig. 2(c) that the phase difference deviates from the desired value
inhomogeneous (for non-periodic structures) ways. In contrast, 1801 outside the working frequency range, which could lead to
the coding metamaterial and metasurface are characterized in a a degenerated performance. Non-Foster element with negative
much simpler manner, in which the basic unit cell constituting dispersion characteristics could help obtain a constant phase
the entire metamaterial array is designated with digital states. In the difference over a broad bandwidth.
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

simplest 1-bit coding, binary digits ‘‘0’’ and ‘‘1’’ with EM phases of One of the obvious advantages of a coding metasurface is
01 and 1801 are used to control EM behaviors by different sequences that the manipulation of EM waves can be simply achieved by
of ‘‘0’’ and ‘‘1’’ (Fig. 2(a) and (b)). Only the phase difference between changing the coding sequence, which determines how the
‘‘0’’ and ‘‘1’’ digital states should be kept as 1801, whereas their coding particles with different digital states are arranged on a
absolute phases are unimportant. In 2-bit coding, the four digits 2D plane. Fig. 2(d)–(f) demonstrate the code-controlled behaviors
‘‘00’’, ‘‘01’’, ‘‘10’’, and ‘‘11’’ are required, representing the EM phases of the coding metasurface with three coding patterns (i.e., coding
of 01, 901, 1801, and 2701, respectively. The generation of multi-bit sequences expressed in the 2D form), which are ‘‘000000. . .’’,
coding obeys a similar rule.100 ‘‘010101. . .’’, and chessboard distribution, respectively. When
There are many subwavelength structures that can be used to they are normally illuminated by plane waves, the scattering
implement coding particles with the required phase difference. To patterns vary from a single beam to dual beams, and finally
achieve high-level and uniform amplitudes for digital states, in the quad beams. The scattering pattern (i.e., directivity function)
initial attempt the reflection-type coding particles are considered can be calculated by:
using the simplest square patch structures. These are constructed
4pj f ðy; jÞj2
by printing a metallic square patch with variable side length on the Dirðy; jÞ ¼ Ð 2p Ð p=2 (1)
topside of a dielectric substrate (F4B, e = 2.65, d = 0.001) (Fig. 2(c)). 0 0 j f ðy; jÞj2 sin ydydj
Due to the existence of a metallic cover on the backside of the
substrate, the structure typically has a reflection amplitude 40.85. X
N
f ðy; jÞ ¼ fe ðy; jÞ expfifjðm; nÞ þ kD sin y½ðm  1=2Þ cos j
With an appropriate thickness of the dielectric substrate, it can n¼1
cover a reflection-phase range over 2701 at the frequency of interest
by changing the side length of the square patch. This is sufficient þ ðn  1=2Þ sin jgg (2)

Fig. 2 (a and b) The various functionalities controlled by coding sequences (schematic). (c) Reflection phases of the ‘‘0’’ and ‘‘1’’ coding particle, and their
phases difference. (d–f) Simulated scattering patterns when the coding sequences are ‘‘000000. . .’’, ‘‘010101. . .’’ and a chessboard distribution,
respectively. Adapted from ref. 100 with permission of the Nature Publishing Group.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3647
 View Article Online

Journal of Materials Chemistry C Review

The functions shown above allow rigorous calculation of the the beam directions of periodic coding patterns and the level
scattering pattern from the coding pattern. However, the main- of random diffusions, which will be discussed in detail in
beam direction of a periodic coding pattern can be calculated Sections 2.3 and 4.3.
by a much simpler function, which will be given in detail in
Section 2.3. 2.2 Terahertz coding metasurfaces
For the other important application of a coding metasurface, Recently, coding metasurfaces were extended to the terahertz
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Cui et al. proposed to reduce the RCSs of metallic objects with frequency to manipulate both the reflections and transmis-
random coding patterns, as presented in Fig. 3(a) and (d) for sions of terahertz waves with high efficiency. The first terahertz
1-bit and 2-bit cases. The normally incident EM wave is coding metasurface was designed to reduce the scattering of
randomly redistributed to all directions in the upper-half space, terahertz waves using multi-bit coding particles,101 as presented
forming scattering patterns with random signatures, as shown in Fig. 4(a). Compared with the square patch design shown in
in Fig. 3(b) and (c). This technique is completely different Fig. 2(c),100 the Minkowski closed-loop coding particle could
from that of metamaterial absorbers,87,88,107–109 which are con- deliver wider phase coverage in a broader bandwidth. By changing
structed by the same unit cell and designed to minimize the the loop width L of the Minkowski structure, the reflection phases
reflections at the specific frequency and direction. Because the can be tuned from 01 to 3151 (see Fig. 4(b)), constituting the
coding metasurfaces do not require an exact 1801 phase differ- eight coding particles in Fig. 4(c) that could be used to build the
ence between the ‘‘0’’ and ‘‘1’’ digital states for RCS reductions, 1-bit, 2-bit, and 3-bit terahertz coding metasurfaces. Numerical
they can typically work in a wide frequency band. To quantita- simulations were carried out to characterize the scattering
tively evaluate the performance of wideband random diffu- diffusion effect of the 1-bit and 3-bit coding metasurfaces, as
sions, a random pattern encoded with 2-bit coding particles shown in Fig. 4(d)–(g), where we can see that the normal inci-
is considered in Fig. 3(d), and the RCS reduction in the back- dences are randomly scattered in numerous directions. The back
ward scattering direction is illustrated in Fig. 3(e). We clearly scatterings are effectively reduced by Z20 dB compared with that
observe that RCSs are effectively reduced by Z10 dB from 7.5 reflected by a perfectly electric conductor (PEC) board with the
to 15 GHz. same dimension.
One may notice from Fig. 3(a) and (d) that the coding The sample was fabricated by a standard photolithography
patterns are formed by clusters of identical coding particles. procedure, as depicted in Fig. 5(a), which mainly involves the
This is called as a ‘‘super unit cell’’, which is introduced to preparation of a polyimide layer and a lift-off procedure to create
design coding metasurfaces based on two main considerations. the Minkowski metallic pattern. The 30 mm-thick polyimide layer
First, because the reflection phase of a coding particle strongly was fabricated by repeatedly spin-coating liquid polyimide onto a
relies on the gap between its adjacent particles, a sub-array of silicon wafer and solidifying it on a hot plate at 80, 120, 180, and
identical coding particles could help generate more accurate 250 1C for 5 min each. The lift-off resist (LOR) photoresist was
phase for each individual and reduce mutual coupling. Second, used to form a T-shaped cross-section of the developed photo-
the size of the super unit cell can be adjusted to control resist pattern, which helps improve the quality of the metallic

Fig. 3 (a and d) Coding patterns for 1-bit and 2-bit microwave coding metasurfaces. (b and c) Scattering patterns for a 1-bit random coding pattern at
8 and 11.5 GHz, respectively. (e) RCS reduction of a 2-bit random coding pattern from 7 to 16 GHz. Adapted from ref. 100 with permission of the Nature
Publishing Group.

3648 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 4 (a) The Minkowski structure for the THz coding metasurface (schematic). (b) The dependence of reflection phases of the THz coding particles on
a loop of width L. (c) Structures for 1-bit, 2-bit and 3-bit THz coding particles. (d and f) The 3-D and 2-D (E-plane) scattering patterns for a 1-bit random
coding pattern at 1.1 THz. (e and g) The 3-D and 2-D (E-plane) scattering patterns for a 3-bit random coding pattern at 1.0 THz. Adapted from ref. 101 with
permission of the Nature Publishing Group.

pattern generated in the final acetone bath. Fig. 4(b) shows the digital states at two distinct frequencies. A dual-layered structure
microscopy image of the fabricated sample, in which the gray incorporating two electric-field-coupled (ELC) resonators is
pattern is the metallic Minkowski structure. designed as the coding particle (see Fig. 6(b)) to implement a
The diffusion effect of the sample was characterized by 1-bit frequency-dependent coding metasurface, which can exhibit
custom-built time-domain spectroscopy (TDS). An ultrafast laser ‘‘0’’ and ‘‘1’’ states at 0.78 and 1.19 THz.
pulse (width, 84 fs), after passing through a series of optical Three coding patterns together with their scattering patterns
modules, is separated into two beams, a pump beam and a probe are presented in Fig. 6(c)–(h) to demonstrate the dual-functionality
beam, and reaches the terahertz photoconductive emitter and performance of the proposed design. In the first example, the
detector, respectively. Two off-axis parabolic mirrors mounted on sub-coding pattern is ‘‘010101. . .’’ along the x direction at the
a rotary stage help collimate the terahertz beam to the sample at lower frequency, whereas it is along the y direction at the higher
an arbitrary incident angle and receive the scattered wave in the frequency, so the dual-beam scattering patterns are located in
specular direction. The normalized reflection under a 201 oblique the x–z (see Fig. 6(c)) and y–z (see Fig. 6(d)) planes accordingly.
incident angle is shown in Fig. 5(c), where we can observe a The sub-coding pattern at the higher frequency is replaced by a
significant reduction of the scattering in the specular direction. chessboard distribution, so the normal incidence is scattered
Fig. 5(d) further confirms the excellent diffusion effect by provid- into four equal beams, as shown in Fig. 6(f). For the last example,
ing the measured reflections in a broad range of receiving angles the sub-coding patterns at the lower and higher frequencies are
from 201 to 801. set with different sizes of super-unit-cell, which are 5  5 and
Based on the orthogonality of different frequencies, Liu et al. 2  2, respectively. As a result, the normal incidence is reflected
developed a frequency-dependent dual-functional coding meta- to the anomalous angles of 321 (see Fig. 6(g)) and 641 (see
surface in the terahertz range that could accomplish two Fig. 6(h)) at the two frequencies, respectively. This phenom-
independent functionalities at two designed frequencies,104 enon is somewhat counter-intuitive to the generalized Snell’s
as shown in Fig. 6(a), where the scattering patterns are different law because the scattering angle decreases, in principle, with
at lower (red beam) and higher frequencies (blue beam). Such an increase in the frequency for a conventional metasurface
a dual-functional performance is attributed to the dual-state with a gradient phase profile. Such a dual-band coding meta-
property of the coding particle, which possesses independent surface may have several possible applications, for example,

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3649
 View Article Online

Journal of Materials Chemistry C Review

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 5 (a) Lift-off process for fabricating THz samples. (b) Microscopy image of the THz sample. (c) Normalized reflection of a 2-bit sample measured
under 201 oblique incidence. (d) Measured scatterings in a wide angle range from 201 to 801 when the 2-bit sample is illuminated by a 201 oblique
incidence. Adapted from ref. 100 with permission of the Nature Publishing Group.

spatial separation of signals modulated at different frequen- free-standing sample, as shown in Fig. 7(a).94 Because the
cies, and multi-color holography.48 rotational twist of SRRs provides a 901 polarization rotation for
It has always been a challenge to design a metasurface to the outgoing terahertz wave (see Fig. 7(b)), the proposed coding
manipulate the transmission of terahertz waves due to two metasurface can manipulate the cross-polarized component of
reasons. First, the coding particle for a transmission-type coding the transmitted terahertz wave, as illustrated in Fig. 7(a), where
metasurface requires 3601 phase coverage with high trans- the y-polarized normal incidence is refracted to the anomalous
mittance, which is almost impossible to achieve if the number direction with cross-polarization. By designing a coding pattern
of metallic layers is o3. Second, the multi-layered metallic with a gradient phase profile along the radial direction, it could
patterns stacked between deep subwavelength dielectric spacers be used to focus the terahertz wave in the near-field region to
pose serious challenges to fabrication using standard photolitho- generate a non-diffraction Bessel beam, as seen from the electric-
graphy. An initial attempt was made by Zhang et al. in 2013 to field distributions in Fig. 7(c) with two examples having different
manipulate the phase profile of transmitted terahertz wavefronts non-diffraction lengths.
by fabricating differently oriented C-shaped split ring resonators The free-standing design of the fabricated sample (see Fig. 7(d))
(SRRs) on a silicon substrate.54 A further improvement of this can naturally deliver much higher efficiency for controlling the
approach was accomplished by the same research team in 2014 transmitted terahertz waves compared with previously reported
by taking control of transmission amplitude through variation of works.24,54,55,92 The sample has a total thickness of E90 mm,
the opening angle of C-shaped SRRs.55 Unfortunately, these making it easy to attach to curved surfaces. In addition, both
designs suffer from low transmittance due to the serious mis- sides of the ultrathin sample are covered by a 5 mm-thick
match in impedance between the silicon substrate (e = 11.49) and polyimide layer to protect the entire sample against mechanical
free space, as well as the Fabry–Perot resonance caused by the scratches and chemical corrosion.
thickness (E500 mm) of a silicon wafer, thus limiting their
applications in real terahertz systems. 2.3 Anisotropic coding metasurface
There is no good way to solve the issues mentioned above Polarization is an important parameter of EM waves that deter-
without discarding the thick silicon wafer. Liu et al. attempted mines the geometrical orientation of the oscillation. Due to the
to improve the transmittance by designing a triple-layered isotropic geometry of the coding particle design in coding

3650 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 6 (a and b) Schematic illustration and coding particle design of a frequency-dependent dual-functionality coding metasurface. (c and d) The sub-
coding patterns are ‘‘010101. . .’’ varying along the x and y directions, respectively. (e and f) The sub coding patterns are ‘‘010101. . .’’ (varying along the
x direction) and a chessboard distribution, respectively. (g and h) The sub coding patterns are ‘‘010101. . .’’ with a super-unit-cell size of 5  5 and 2  2,
respectively. (c, e and g) At a lower frequency. (d, f and h) At the higher frequency. Adapted from ref. 104 with the permission of Wiley VCH.

metasurfaces mentioned above, a fixed coding pattern can have required to build up the n-bit anisotropic coding metasurface.
only one function that does not change with the polarization Fig. 8(d) shows the 16 coding particles needed to construct a
of incidence. Based on the orthogonality of two mutually 2-bit anisotropic coding metasurface. The four coding particles
perpendicularly polarized EM waves, Liu et al. presented an on the diagonal line have the same digital states under both
anisotropic coding metasurface in the terahertz frequency.103 polarizations, and so they have the same square patch design as
Fig. 8(a) and (b) schematically demonstrate the polarization- those in ref. 100. Due to the reciprocity of the digital states
controlled behavior of an anisotropic coding metasurface. For under x and y polarizations, the six anisotropic coding particles
example, under the y-polarized normal incidence, the coding in the upper triangular matrix can be obtained by making a
pattern is ‘‘010101. . .’’ and generates two equal beams (see 901 rotation of those in the lower triangular matrix.
Fig. 8(a)). However, as the polarization changes to the x direc- Three differently anisotropic coding patterns are given in
tion, the coding pattern changes to a chessboard distribution, Fig. 9 to demonstrate the polarization-controlled performance.
showing four equal beams, as illustrated in Fig. 8(b). The dual Fig. 9(a) and (b) are the scattering patterns of the 1-bit coding
functionality of the anisotropic coding metasurface originates pattern M1 under the x and y polarizations, respectively. For the
from the geometrical anisotropy of the coding particle shown in x polarization, the coding pattern is ‘‘010101. . .’’, resulting in
the inset of Fig. 8(c), where the square patch in the isotropic two equal beams whereas, for the y polarization, both the coding
design is replaced by a dumb-bell shaped structure. The reflec- and scattering patterns become random. Fig. 9(c) and (d) further
tion curves of the ‘‘1/0’’ coding particle plotted in Fig. 8(c) show demonstrate the 2-bit case with coding pattern M2, which
that the phase difference between the x and y polarizations is composed of two gradient coding patterns varying along
reaches 1801 at the designed frequency of 1 THz. Because each perpendicular directions. As a result, the normal incidence is
coding particle should have independent digital states under anomalously reflected in the y–z plane under the x polarization,
x and y polarizations, a total number of 22n coding states is and in the x–z plane under the y polarization. Here, a function

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3651
 View Article Online

Journal of Materials Chemistry C Review

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 7 (a and b) Schematic illustration and coding particle design of a transmission-type coding metasurface. (c) Electric-field distributions of two Bessel
beams having different non-diffraction lengths. (d) Fabricated sample and microcopy image of a transmission-type coding metasurface. Reprinted
(adapted) with permission from ref. 94. Copyright (2017) American Chemical Society.

is given to calculate the scattering direction from a gradient enables the sample to be ‘‘peeled off’’ from the silicon wafer
coding pattern to become a free-standing and flexible sample, as shown in
  Fig. 10(a). A fiber-based rotary THz-TDS (see Fig. 9(b)) was
l
y ¼ sin1 (3) employed to characterize the polarization-dependent perfor-
G
mance of the fabricated sample experimentally. Under the
in which l and G are the free-space wavelength and periodi- x polarization, the normal incidence is reflected to the anomalous
city of the gradient coding pattern, respectively. All simulated direction, as shown in Fig. 10(c). Under the y polarization,
anomalous reflection angles have excellent agreements with the it is randomly scattered to multiple directions, significantly
values predicted by eqn (3). The last example in Fig. 9(e) suppressing the RCS in the specular direction under different
functions as a quarter-wave plate, which reflects the 451 linearly incident angles, as illustrated in Fig. 10(d). The proposed aniso-
polarized incidence (with respect to the x axis) to the anomalous tropic coding metasurface may have broad applications from
direction with circular polarization. The axial ratio keeps o1.26 microwave to terahertz and visible light spectra. For example,
from 0.8 to 1.2 THz, and is only 1.03 at 1 THz, as can be observed it can be used as a beam splitter at microwave and terahertz
in Fig. 9(f). frequencies to improve the transmission speed of wireless com-
The sample was fabricated using the same procedure des- munication systems. This dual-state feature can also be utilized
cribed in Section 2.2. Titanium was not deposited beneath the to double the storage density of optical data storage media,
gold layer, so the low adhesion between the gold and silicon such as light disks.

3652 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 8 (a and b) Anisotropic coding metasurface under x and y polarizations, respectively (schematic). (c) Reflection phases of the dumb-bell shaped
anisotropic coding particle (inset) under x and y polarizations. (d) Structures of the 16 coding particles for the 2-bit anisotropic coding metasurface.
Adapted from ref. 103 with permission of the Nature Publishing Group.

3. Digital and programmable One of the benefits of the digital characterization of coding
metasurfaces metasurfaces is that the ‘‘0’’ and ‘‘1’’ digital states are electri-
cally controlled by the pin-diode through DC voltage. Then, one
3.1 Digital and field programmable features of coding can construct a single metasurface using the designed unit
metasurfaces structure, in which the digital states of all particles are switched
In previous discussions, different unit cells have been used to between ‘‘0’’ and ‘‘1’’ electrically. This metasurface has numerous
characterize the digital states of coding particles. To link the functions with various coding-sequence possibilities controlled by
physical world of metamaterials to the digital world, as shown the biased pin-diodes, thereby realizing the digital metasurface.100
in Fig. 1, the coding states ‘‘0’’ and ‘‘1’’ should be realized by a When all possible coding sequences are stored in a field pro-
single unit particle. To this end, an elaborately designed struc- grammable gate array (FPGA), a variety of functionality can be
ture was proposed to reach a real digital coding particle,100 as dynamically generated in a programmable way,100 as illustrated in
shown in Fig. 11(a). Here, a pin-diode (Skyworks, SMP1320- Fig. 11(c). This will result in a field programmable metasurface.
079LF) is welded between two separated metallic structures on As a proof of principle, a programmable metasurface proto-
the topside of a dielectric substrate (F4B), which is controlled type with 30  30 coding particles was fabricated and measured,
by DC voltage. Different from the square-patch design shown in which was divided into 6 control lines, each consisting of 5  30
Fig. 2(c) with a fully covered metal background on the bottom, coding particles. In this condition, only 6 coding sequences are
two separated metallic strips are printed on the backside of the required to control the one-dimensional (1D) programmable
substrate to serve as the bias line, which are connected to metasurface. Fig. 11(d)–(f) are the scattering patterns when the
the top structures through the two metallic strips via holes. coding sequences are ‘‘000000’’, ‘‘010101’’, and ‘‘001011’’, which
Note that there is a small transmission (less than 15 dB) for generate a single beam, two beams, and multiple beams, respec-
this structure due to the gap between the bias lines. The struc- tively, controlled by FPGA in real time. Many other functions can
ture is optimized by the numerical simulation software CST be accomplished by the same metasurface corresponding to
Microwave Studio to ensure that the reflection phase difference other coding sequences. It is expected that a multi-bit program-
reaches 1801 when the diode is biased ‘‘on’’ and ‘‘off’’ by the mable metasurface could be realized in the near future by
DC voltage, as can be observed from the reflection phases in loading varactor diodes to the coding particle and adjusting
Fig. 11(b). the value of the applied voltage.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3653
 View Article Online

Journal of Materials Chemistry C Review

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 9 Three examples to demonstrate the polarization-controlled behavior of an anisotropic coding metasurface. (a and b) The scattering pattern for
coding pattern M1 under x and y polarizations, respectively. (c and d) The scattering pattern for coding pattern M2 under x and y polarizations,
respectively. (e) The scattering pattern for coding pattern M3 under 451 polarized normal illumination (with respect to the x and y axes). (f) The axial ratio at
a main beam direction from 0.8 to 1.2 THz. Adapted from ref. 103 with permission of the Nature Publishing Group.

We remark that there is a big difference between the pro- near-field region. Wan et al. sought to solve this problem by
grammable metasurface and phased-array antenna,110,111 which letting the programmable metasurface work under the illumina-
has been widely used in radar applications. First, the electrical tion of a point source112 whose size is sufficiently small to avoid
size of the unit cell of programmable metasurface is, in princi- distortion to the scattering pattern of the programmable meta-
ple, between l/8 and l/4, whereas it is typically larger than l/2 for surface. As the waves reaching the programmable metasurface
a phased-array antenna. The smaller size of the unit cell comes experience different light paths, as sketched in Fig. 12(a), a
with two benefits: (1) more identical unit cells can be arranged in ‘‘compensation’’ approach is proposed to offset the phase delay
a super-unit-cell, leading to a more accurate phase response; for each coding particle so that the programmable metasurface
(2) a gradient coding sequence with periodicity smaller than the can work with the same coding sequence as it is under the plane-
free-space wavelength could convert a spatial wave to a surface wave excitation. Similar to eqn (1), the pattern function under the
wave. Second, the digital description enables the coding meta- point source illumination is written as
surface to be studied from the information perspective, as will be
X
N
discussed in Section 4. f ðy; jÞ ¼ fe ðy; jÞ expfifj0 ðm; nÞ þ kD sin y½ðm  1=2Þ cos j
n¼1
3.2 Programmable antennae
þ ðn  1=2Þ sin jgg (4)
The programmable metasurface presented in the section above is
designed to work for plane wave excitation. In real applications, in which j 0 (m,n) = j(m,n)  krm,n 0 . It implies that the reflection
however, feeding the reflection-type programmable metasurface phases j 0 (m,n) of the ‘‘0’’ and ‘‘1’’ coding particles no longer
by a plane wave is impractical because the source is located in the maintain the ideal 01 and 1801 phases, but have a gradient item

3654 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 10 (a) Fabricated sample of an anisotropic coding metasurface. (b) Fiber-based rotary THz-TDS for characterization of the anomalous reflection.
(c) The measured anomalous reflection of the sample with coding pattern M1 under x polarization. (d) The measured scattering reduction of the sample
with coding pattern M1 under y polarization. Adapted from ref. 103 with permission of the Nature Publishing Group.

of krm,n 0 . Fortunately, such a non-planar wavefront can be 2-bit programmable metasurface in the microwave frequency for
compensated by varying the geometry of the coding particles, as single-antenna and single-frequency imaging.81 Fig. 13(a) schema-
shown in Fig. 12(b), which maintains a similar structure from tically illustrates the working mechanism of the single-sensor
the previous design.100 It should be noted that, because the imaging system, which is mainly composed of a transmission-
position of the point source is fixed with respect to the program- type 2-bit programmable metasurface electrically controlled
mable metasurface, the compensation in structure geometry is by a FPGA, a feeding antenna connected to a vector network
valid for any coding pattern. analyzer (VNA), and a computer. When the FPGA inputs a series
The compensation process was accomplished by extracting of random coding sequences into the programmable meta-
the reflection coefficients first for all combinations of geometrical surface, several independently random radiation patterns are
parameters S and W, and then replacing each coding particle with generated, illuminating the object to be imaged. Each time the
the one having the required phase to offset the gradient term FPGA changes the coding sequence, the same horn antenna
krm,n 0 . The distributions of parameters S and W used to measures the signal reflected back from the object. All the
compensate the non-planar phase caused by the point source recorded data are sent to the computer for reconstructing the
excitation are presented in Fig. 12(c) and (d). The validity of the image of the object.
compensation approach was confirmed by simulating the same Here, the single-sensor image reconstruction algorithm is
coding pattern under the excitation of a plane wave and a small briefly described. Consider an object that is divided into N
rectangular-horn antenna placed 250 mm above the meta- equal-sized partitions s = (si), i = 1, 2,. . ., N, and a total number
surface, as shown in Fig. 12(e) and (f), respectively. We observe of P times of observations. Then, the measured data V = (V(p) i ),
excellent consistency between the two results. This work was p = 1, 2,. . ., P, are analytically expressed as
experimentally verified by a fabricated programmable meta- 2 ð1Þ 3 2 32 3
V G11 G12    G1N s1
surface including 4  4 control units (each control unit con- 6 7 6 76 7
6 ð2Þ 7 6 G 76 s 7
tains 5  5 coding particles), advancing a step towards a 2D 6 V 7 6 21 G22 76 2 7
6 7 6 76 7
programmable antenna.112 The radiation pattern measured in 6 . 7¼6 .. 76 . 7 (5)
6 . 7 6 76 . 7
an anechoic chamber showed that the main lobes appear close 6 . 7 6 . 76 . 7
4 5 4 54 5
to the predicted directions in the programmable way.
V ðPÞ GP1 GPN sN
3.3 New imaging approach based on a programmable in which G = (Gpj), p = 1, 2,. . ., P, j = 1, 2,. . ., N, is a generalized
metasurface system response matrix. The matrix G will become a square
Taking advantage of dynamically controlled radiation patterns of a matrix if P = N, making it convenient to retrieve the image of an
programmable metasurface, Li et al. developed a transmission-type object with r = G1V.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3655
 View Article Online

Journal of Materials Chemistry C Review

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 11 (a and b) Structure design for the coding particle of a programmable metasurface and its reflection phases when the pin-diode is switched ‘‘on’’
and ‘‘off’’, respectively. (c) The programmable metasurface prototype to demonstrate the dynamic control of scattering patterns through different coding
sequences. (d–f) The simulated scattering patterns when the input coding sequences are ‘‘000000’’, ‘‘010101’’ and ‘‘001011’’, respectively. Adapted from
ref. 100 with permission of the Nature Publishing Group.

To dynamically switch the radiation patterns, a dual-layered in experiments. A 2-bit programmable metasurface with 10 
structure displayed in Fig. 13(b) is designed for the coding 10 = 100 elements occupying an area of 140  140 mm2 was
particle of the transmission-type programmable metasurface. placed in front of a horn antenna. Only the 5  5 coding
It is composed of two dielectric layers (F4B) separated by an air particles in the center of the programmable metasurface were
gap of 3 mm. The outer side of each dielectric substrate is printed electrically controlled, and were shown to generate sufficiently
with a metallic square patch loaded by a pin-diode (Skyworks, random scattering patterns. To save time spent during the
SMP1320-079LF).81 Different from the design shown in ref. 100, calibration for the generalized system response matrix G,
a ‘‘metallic fishnet’’ is added at the edge of each coding particle, the object was divided into 5  5 = 25 sub-areas. Based on the
and serves as the ground line. Another two metallic lines are theory mentioned above, a total of 25 measurements, each time
printed on the inner side of two dielectric layers with orthogonal with a different random coding sequence, will be required for
orientations to control the voltage applied to the pin-diodes. successful image reconstruction. Fig. 13(c) presents the photo
In this configuration, the entire programmable metasurface and reconstructed images of the objects, from which we can see
can be controlled by rows and columns, resulting in 2-bit binary that the objects were correctly reconstructed. The image resolu-
states of ‘‘11’’, ‘‘01’’, ‘‘10’’, and ‘‘00’’. Two important issues tion can be improved by using a large array of programmable
should be noted here. First, the layout of the feeding lines metasurfaces and performing a more careful calibration for
does not support independent control of all coding particles. the response matrix. The cost and complexity of this new
Second, the phase span for the four digital states is E1001 at a approach for single-sensor imaging is greatly reduced due to
working frequency of 9 GHz. However, these constraints have the simpler feeding configuration. The problem of object dis-
little effect on generation of the random radiation patterns persion can also be avoided because the object is imaged at a
required for the single-sensor imaging, as will be verified in the single frequency.
experiments. Recently, the single-sensor and single-frequency imaging
As a proof of the new imaging principle, I-shaped and methods have been modified using a compressed-sensing tech-
T-shaped metallic patches were used as the objects to be imaged nique to handle large-scale imaging problems, and have resulted

3656 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 12 (a) A programmable metasurface under a point source excitation (schematic). (b) The structure design for the coding particle of a programmable
metasurface. (c and d) Distributions of the parameters S and W used to compensate the non-planar phase caused by the point source excitation. (e and f)
Simulated scattering patterns of the same coding pattern under the excitations of a plane wave and a rectangular-horn antenna, respectively. Adapted
from ref. 112 with permission of the Nature Publishing Group.

in a 40  40 programmable metasurface imaging system.113 This coding metasurface. The proposed reprogrammable hologram
single-sensor imager has a reduced data-acquisition time with represents a key enabler for future intelligent devices with
improved obtainable temporal and spatial resolutions. Besides, reconfigurable functionalities, which may lead to advances in
it was demonstrated that the proposed imager has a ‘‘theoretical various applications, such as microscopy, display, security, data
guarantee’’ on successful recovery of a sparse or compressible object storage, and information processing.
from its reduced measurements by solving a sparsity-regularized
convex optimization problem. The excellent performance of the
proposed imager was validated by numerical simulations and 4. Information operations on
experiments for high-resolution microwave imaging. metasurfaces
In addition, the programmable metasurface has also been used
to produce programmable hologram imaging.114 The proof- 4.1 Information entropy on metasurfaces
of-concept experiments showed that multiple holographic As mentioned in Fig. 1, coding metasurfaces have become a ‘‘bridge’’
images can be realized in real time by applying only a single linking the physical metamaterial and information science.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3657
 View Article Online

Journal of Materials Chemistry C Review

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 13 (a) System diagram for single-antenna and single-frequency imaging. (b) Structure for the coding particle of a transmission-type
2-bit programmable metasurface. (c) Photos and reconstructed images of the objects. Adapted from ref. 81 with permission of the Nature
Publishing Group.

More recently, Cui et al. proposed to measure the amount of Before introducing the information entropy, the relationship
information carried by a coding metasurface using Shannon between the far electric field of the coding metasurface and
entropy, which was originally used to estimate the uncertainty the corresponding coding pattern can be written in the closed-
(i.e., average value of information) of an information source.106 form as
To demonstrate this idea, a reflection-type information system
iom0
is considered, as sketched in Fig. 14(a), which includes a coding E s ðr; y; jÞ ¼ Pðk sin y cos j; k sin y sin jÞ (6)
4pr
metasurface and several receiving antennae. In this configu-
ration, the information modulated in scattering patterns of where r, y and j indicate the distance, elevation angle, and
coding metasurfaces can be received by the receiving antennae. azimuth angle in the spherical coordinates, respectively, k is the

3658 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 14 (a) A reflection-type information system using a programmable metasurface (schematic). (b) Group G(i,j) in a coding pattern. GR(i,j) and GU(i,j) are
considered from the row and column, respectively. (c–e) Coding patterns with all-zero, ‘‘010101. . .’’ and a chessboard distribution, respectively. The entire
coding pattern includes 64  64 coding particles and each coding particle has a periodicity of 7 mm. (f–h) The calculated scattering patterns at 10 GHz.
Adapted from ref. 106 with permission of the Nature Publishing Group.

wavenumber in free space, and P(k sin y cos j,k sin y sin j) is column (H2U). With respect to the entropy of the far-field
the Fourier transform of the electric current distribution on the scattering pattern, it is obtained by calculating the 2D entropy
coding metasurface. This implies that the scattering pattern of the scattering pattern image using the following function
of the coding metasurface is determined by the fast Fourier
256 X
X 256
transform (FFT) of the coding pattern, which can be used to H2 ¼  Pij log2 Pij (8)
estimate the amount of information carried in the scattering i¼1 j¼1
pattern.
First, the entropy of the coding pattern is studied by adopting where the G(i,j) has a maximum number of 256  256 = 65 536
the 2D Shannon entropy: combinations, due to the 8-bit gray level of the scattering pattern
image. Several examples are given to show that the physical
2 X
X 2
entropy of a coding metasurface can be directly estimated by the
H2 ¼  Pij log2 Pij (7)
i¼1 j¼1 geometrical entropy of the coding pattern.
Three periodic coding patterns (Fig. 14(c)–(e)) are first selected
in which Pij is the joint probability of a group G(i, j) containing to show the calculation process of entropy, each of which includes
two adjacent coding elements, as illustrated in Fig. 14(b). For 64  64 coding particles. A fast method based on FFT was
a 1-bit coding metasurface, there are only four combinations developed to calculate the scattering patterns from coding pat-
G(0,0), G(0,1), G(1,0) and G(1,1). According to eqn (7), the entropy terns with sufficient accuracy in a time Z1000 times faster than it
is related to the probability of the above four cases that appear takes by numerical simulations. Fig. 14(f)–(h) are the theoretically
in a coding pattern, which could reflect the spatial distribution calculated scattering patterns of the coding patterns at 10 GHz.
characteristics of the ‘‘0’’ and ‘‘1’’ digital states within it. For the uniform coding pattern (Fig. 14(c)), the geometrical
The neighboring code in group G(i, j) can be on either the right entropy is zero and the scattering pattern (Fig. 14(f)) also has
side (GR(i, j)), or the upper side (GU(i, j)), of the current code the smallest physical entropy of 0.9273. For the ‘‘010101. . .’’
(indicated in red), the final 2D entropy should be the average coding pattern, the entropy directly calculated from eqn (7)
value of the entropies calculated along the row (H2R) and is 0.7028. If the coding sequence can be considered as a

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3659
 View Article Online

Journal of Materials Chemistry C Review

2  2 matrix (0,1; 0,1) with four symbols as an entity, then the The corresponding scattering patterns expressed in the polar
entropy will become zero. Hence, there are four independent coordinate are shown in Fig. 15(d)–(f). It can be seen that
codes in this case. The entire coding pattern contains 8  8 = 64 there will be more scatterings as the coding pattern becomes
of such matrices, the final geometrical entropy should be calcu- more random, leading to a larger physical entropy. The pro-
lated as 0.7028  4/64 = 0.0439. The physical entropy of the portional relationship between the geometrical entropy and
scattering patterns is 1.3923, which is larger than the all-zero physical entropy is clearly observed in Fig. 15(g), in which both
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

pattern. Using the same method, the geometrical entropy and the geometrical and physical entropies of the 99 iterations
physical entropy of the third coding pattern are calculated as (each iteration includes 500 steps of element switching) are
0.0566 and 2.0467, respectively. provided.
It can be noticed that as the geometrical entropy of coding Experiments were performed with a ‘‘010101. . .’’ coding
pattern increases, the physical entropy of the scattering pattern sample fabricated at microwave frequencies, and the measured
also increases. To further verify this conjecture, more calcula- results were highly consistent with numerical simulations.106
tions were conducted with non-periodic coding patterns and The proposed concept may help the programmable metasurface-
random coding patterns. A model of cellular automata machine, based information system to estimate the average information
which mimics the diffusion process of gas molecules, was sent by a certain coding pattern.
utilized to generate a series of random coding patterns with
increasing randomness. Fig. 15(a)–(c) show three random coding 4.2 Digital convolution operations on metasurfaces
patterns with different randomness, which are at the initial, For the coding schemes adopted previously,94,100,103,104 the
middle, and final steps of the generation process, respectively. scattering direction is limited to several discrete angles because

Fig. 15 (a–c) Three coding patterns with increasing randomness. (d–f) The calculated scattering patterns at 10 GHz. (g) The geometrical entropy and
physical entropy for a series of coding patterns generated by the model of cellular automata machine. Adapted from ref. 106 with permission of the
Nature Publishing Group.

3660 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

the gradient coding sequence with the minimum periodicity the computational complexity of the calculation for the mixed
is multiplied by an integer. To circumvent this limitation, Liu coding pattern.
et al. proposed a new coding strategy, called as the ‘‘scattering To better understand the principle of scattering pattern
pattern shift’’, which can rotate the scattering pattern to a shift, an example is provided in Fig. 16(a)–(f) with three coding
desired direction with little distortion and realize beam scanning patterns and their scattering patterns. The cross-shaped coding
with a continuous scanning angle.105 The idea was inspired by pattern (Fig. 16(a)) has a five-beam scattering pattern (Fig. 16(d)).
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

the convolution theorem in the Fourier transform, which origin- When it is added by a gradient coding pattern (Fig. 16(b)) having
ally describes the Fourier relationship between the ordinary single-beam pointing in a certain direction (Fig. 16(e)), the mixed
multiplication of two signals in the time domain and the coding pattern (Fig. 16(c)) will have the same five-beam scattering
convolution of their frequency spectra in the frequency domain. pattern but is rotated to the direction of the single beam, as
This theorem can be degenerated as the frequency-shift function shown in Fig. 16(f).
when one of the product terms is a Dirac-delta function, which is Four 2-bit coding patterns are presented in Fig. 17 to
analytically expressed as demonstrate the performance of the new coding scheme. All
coding patterns are composed of 64  64 coding particles, and
FFT
f ðtÞejo0 t ! f ðoÞ  dðo  o0 Þ ¼ f ðo  o0 Þ (9) each coding particle has a periodicity of 70 mm. The scattering
patterns are calculated by FFT at 1 THz. The first coding pattern
Fig. 16(g)–(i) illustrate the convolution of a signal f (o) (Fig. 16(g)) shown in Fig. 17(a) is encoded by ‘‘0202. . .’’ (super-unit-cell size
and a Dirac-delta function d(o  o0) (Fig. 16(h)) in the frequency is 8  8) varying along the vertical direction, whose scattering
spectrum domain, resulting in a shifted signal f (o  o0) at a patterns are simply two beams located in the x–z plane, as
higher frequency (Fig. 16(h)). If the arguments o and t in eqn (9) shown in Fig. 17(e). When a ‘‘01230123. . .’’ coding sequence
are replaced by sin y and xl = x/l, respectively, we obtain the that varies along the horizontal direction (super-unit-cell size
function of scattering pattern shift is 3  3) is added, resulting in the mixed coding pattern in
Fig. 17(b), the two-beam scattering pattern is tilted toward
FFT
E ðxl Þ  ejxl sin y0 ! E ðsin y0 Þ  dðsin y  sin y0 Þ ¼ E ðsin y  sin y0 Þ the +y direction, as can be observed in Fig. 17(f). The tilted
(10) angle is just equal to the scattering angle of the gradient coding
pattern.
in which the item ejxlsiny0 is a coding pattern with gradient Another two examples are given in Fig. 17(c) and (d) to
phase distribution. Based on the understanding of eqn (9), it is further demonstrate the ability of continuous scanning. Fig. 17(c)
clear from eqn (10) that the multiplication of the coding pattern is obtained by two gradient coding patterns ‘‘01230123. . .’’ with
E(xl) and a gradient coding pattern ejxlsiny0 can rotate the different super-unit-cell sizes of 2  2 and 3  3, which have
scattering pattern E(sin y0) to a desired direction by a quantity of single-beam scatterings pointing at 32.41 and 20.91, respectively.
sin y0. Most importantly, the multiplication of two coding patterns The scattering pattern of the mixed coding pattern is shown
is simply the modulus of their digital codes, significantly reducing in Fig. 17(g), which has a single beam pointing at a larger

Fig. 16 Principle of scattering pattern shift by making an analogy to the convolution theorem in the Fourier transform (schematic). (a–c) Cross-shape
coding pattern, gradient coding pattern, and mixed coding pattern, respectively. (d–f) Corresponding scattering patterns calculated by the FFT method.
(g–i) The convolution theorem in signal processing. Adapted from ref. 105 with the permission of Wiley VCH.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3661
 View Article Online

Journal of Materials Chemistry C Review

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 17 Four examples to demonstrate the rotation of scattering pattern and ability of continuous scanning of the new coding scheme. (a) ‘‘010101. . .’’
with a super-unit-cell size of 8  8. (b) The addition of the coding pattern in (a) with a gradient coding pattern ‘‘01230123. . .’’ (super-unit-cell size of 3  3)
varying along the horizontal direction. (c) The addition of two gradient coding patterns ‘‘01230123. . .’’ with different super-unit-cell sizes of 2  2 and
3  3. (d) Subtraction of two gradient coding patterns ‘‘01230123. . .’’ with different super-unit-cell sizes of 2  2 and 3  3. (e–h) Corresponding scattering
patterns for (a–d), respectively. Adapted from ref. 105 with the permission of Wiley VCH.

angle of 63.21. According to eqn (10), the new angle can be pattern is observed. If this coding pattern is added by a gradient
analytically calculated by the following function coding sequence ‘‘0123456701234567. . .’’ (with super-unit-cell
size of 3  3) that varies along the x direction, we obtain a
y = sin1(sin y1  sin y2) (11) mixed coding pattern, as shown in Fig. 18(b). In this situation,
the ring-shaped pattern is shifted to the right of the center
where y1 and y2 are the scattering angles of the two gradient
by a certain value, as clearly illustrated in Fig. 18(e). Based
coding patterns that are added. If these two gradient coding
on the discussions shown above, it is expected that the ring-
patterns subtract with each other, the mixed coding pattern
shaped pattern could scan the entire upper-half space with
(see Fig. 17(d)) will have a single beam scattering pointing at a
continuous angles. More interestingly, a four-ring pattern
small angle of 10.31. Obviously, an arbitrary scattering angle can
(Fig. 18(f)) is readily achieved by adding a coding pattern with
be realized by changing the periodicity of these two gradient
a chessboard distribution (super-unit-cell size of 6  6) to the
coding sequences. In addition, the beam can scan in the azi-
coding pattern in Fig. 18(a), resulting in the mixed coding
muthal direction by adding two gradient coding sequences that
pattern shown in Fig. 18(c). The versatile ring-shaped scattering
vary along perpendicular directions. In this regard, single-beam
patterns may have possible applications in medium-range
scattering can scan continuously in the whole upper-half space
fast radar detection. A circularly polarized ring-shaped scatter-
using a 2-bit programmable metasurface.
ing pattern can also be achieved by combining the anisotropic
design.103
4.3 Two applications Such a function was implemented with a real structure
Two applications were presented by Liu et al. based on the displayed in Fig. 18(g), which maintains the similar design
principle of scattering pattern shift. The first application is to from the square patch structure.100 However, it has two new
generate a ring-shaped scattering pattern with controllable features. First, four metallic frames are added at the edge of
shapes and directions.115 To realize a scattering pattern with the coding particle, serving as the common ground for two
uniform scattering in the azimuthal direction, which looks like adjacent coding particles, which could effectively reduce the
a ring when plotted in the polar coordinate system, a gradient EM coupling between neighboring coding particles with differ-
coding sequence should vary along the radial direction, instead of ent geometries. The size of the super-unit-cell could, therefore,
along the x or y directions as shown previously.101,103–108 An be reduced accordingly, enabling coding metasurfaces with
example is given in Fig. 18(a), which is generated by a 3-bit coding real structures to have more accurate scattering patterns.
metasurface with the coding sequence ‘‘0123456701234567. . .’’, Second, the substrate material is chosen to be crystal quartz,
in which the super-unit-cell size is 2  2. To keep a uniform which has a much smaller loss tangent of permittivity (0.0004)
scattering in the azimuthal direction, the entire coding pattern than polyimide. As a result, all the eight coding particles have
is cut into a round shape with 128 coding particles along the a reflection amplitude of 40.98 at the working frequency,
diameter. Each coding particle has a periodicity of 250 mm. which delivers (so far) the highest efficiency for the reflection-
Using the FFT method, the scattering pattern calculated at type coding metasurfaces and reflectarrays at the terahertz
0.24 THz is provided in Fig. 18(d), where a small ring-shaped frequency.

3662 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 18 (a) A coding pattern with ‘‘0123456701234567. . .’’ (super-unit-cell size of 2  2) varying along the radial direction. (b) A mixed coding pattern
obtained by adding a ‘‘0123456701234567. . .’’ (super-unit-cell size of 3  3) varying along the x direction to the coding pattern in (a). (c) A mixed coding
pattern obtained by adding a chessboard distribution (super-unit-cell size of 6  6) to the coding pattern in (a). (d–f) The calculated scattering patterns
for the coding pattern in (a–c), respectively. (g) Coding particle design. (h) The numerically simulated scattering pattern for the coding pattern shown in
(c). Adapted from ref. 115 with the permission of IEEE.

The second application is a controllable random metasurface coding pattern maintains both features of random scattering
which could control the probability of random scatterings and beam splitting from the constituent coding patterns, the
appearing in certain directions.116 For conventional random calculated scattering pattern shows two regions of random
surfaces reported in the literature,100,101,103 or the rough sur- scatterings, as clearly illustrated in Fig. 19(e). In contrast to the
faces at the visible light frequency, the normal incidence is periodic coding patterns that have definite scatterings in certain
randomly scattered around the normal direction of the meta- directions, the controllable random surface aims to control the
surfaces, as illustrated schematically in Fig. 19(a). This is probability of random scatterings appearing in certain angle
not the case for the controllable random surface shown in ranges. Therefore, it can be compared with the electron cloud in
Fig. 19(b). Unlike any other rough surfaces from natural materials, an atom, where only the probability of the electron’s position can
the random scatterings are diverted to a certain direction. be predicted. In this regard, the cluster of random scatterings is
Fig. 19(c) shows the procedure for obtaining a mixed coding called the ‘‘scattering cloud’’.
pattern of a controllable random surface (the right coding pattern), The relationship between the size of super-unit-cell and size
which is simply the addition of a random coding pattern (the of scattering cloud (i.e., the level of diffusion) is also revealed. It is
left coding pattern) and gradient coding pattern (the middle found that the smaller the super-unit-cell size, the larger is the size
coding pattern). of scattering cloud.116 A structure featuring lower interference with
An example is given in Fig. 19(d) and (e) to demonstrate the the neighboring coding particles was designed for the implemen-
performance of a controllable random surface in redirecting tation of controllable random metasurfaces. Numerical simula-
the random scatterings to desired directions. The mixed coding tions show excellent agreement with theoretical calculations. The
pattern in Fig. 19(d) is composed of two sub-coding patterns, a proposed design could be utilized as a new diffusion material in
random pattern and a ‘‘0101. . .’’ pattern (super-unit-cell size of both EM and acoustic fields, and may also be utilized to improve
3  3), both having 64  64 coding particles. Because the mixed the imaging quality of single-sensor imaging systems.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3663
 View Article Online

Journal of Materials Chemistry C Review

Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

Fig. 19 (a and b) A conventional rough surface and a proposed controllable random surface (schematic). (c) The generation of the mixed coding pattern
for the controllable random surface. (d) A mixed coding pattern obtained by the addition of a ‘‘010101. . .’’ (super-unit-cell size of 3  3) and a random
coding pattern. (e) The calculated scattering pattern. Adapted from ref. 116 with permission of the Nature Publishing Group.

5. Conclusion and outlook ‘‘0’’ and ‘‘1’’ with opposite phase responses are employed as
the meta-atoms to create 1-bit coding metasurfaces. This new
The past 20 years have witnessed the rapid progress of meta- representation sets up a bridge between physical meta-atoms
materials characterized by effective medium parameters, which and digital codes, thereby connecting the metamaterial to infor-
are named here as physical metamaterials, or analog meta- mation science. The coding metamaterial can be easily extended
materials. In a recent perspective article, N. I. Zheludev reviewed from 1-bit to n-bit coding with the 1801/(2n  1) phase difference
some significant achievements,117 and gave the ‘‘roadmap’’ of for 2n coding particles. It has been demonstrated that the
physical metamaterials, as shown in Fig. 20 (the left branch). EM waves could be controlled by simply changing the coding
Information science was originally evolved from physical sequences on metasurfaces, including the wideband diffusions
science, but has developed rapidly to an independent discipline of terahertz waves. The concept of anisotropy in effective medium
since the 1960s using digital codes for efficient message trans- theory was also applied to coding metasurface, yielding an
mission.118 Can we introduce information metamaterials so anisotropic coding metasurface. Under different polarizations,
that they are used to control information directly? To this end, the same coding particle has different coding states, which
the coding representation of the meta-atoms was presented. increases the information capacities. In this way, the same coding
Instead of using effective medium parameters, coding particles metasurface can fulfill completely distinct functionalities, such as

3664 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

principle. The Shannon entropy of the coding pattern on a coding

metasurface (geometrical entropy) is approximately proportional
to the Shannon entropy of its far-field pattern (physical entropy).
This relationship enables the rapid design of a high-information-
entropy metamaterial, which helps realize new imaging systems
with a single sensor at a single frequency through the program-
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

mable metasurface. The future direction in this area is to reveal

the quantitative relationship between the coding entropy and
imaging quality, and to generate large-aperture imaging systems
with more pixels. The Shannon entropy of coding metamaterials
can also be used to increase the information channels of wireless
communications. The other important feature of the coding
description is the convolution operation on the coding meta-
surface, which can achieve extreme controls of anomalous reflec-
tions to arbitrarily designed directions in the upper space. In the
future, more digital signal processing operations, such as digital
filters and correlations, could be performed on the coding meta-
Fig. 20 Two branches of the ‘‘metamaterial tree’’. The past and future of
surfaces to reach other functionalities.
metamaterials from the perspectives of physical science and information
science. A bigger step towards the future will be software-defined
metamaterials, which is the natural development of the coding,
digital, and programmable metamaterials (see Fig. 20, the right
the directive radiationships for the horizontal polarization and branch). The software-defined metamaterial should be a com-
diffusions for the vertical polarization. The other method to bination of a digital coding metamaterial and various modern
double the information process is a dual-band coding meta- digital modular components conformed to certain software.
surface, which can make different EM operations by designing Instead of being simply controlled by software, the metamaterial
the frequency band. functions will be directly defined by the software through the
A coding metasurface can evolve to digital metasurface if ‘‘0’’ distributed digital modules. The actual capabilities of the digital
and ‘‘1’’ coding particles share the same unit cell controlled by modules are only realized in software running on computing
the digital pulse of voltage through a biased diode. When the systems, which could support novel organizations such as the
biased voltage is in the ‘‘0’’ level, the coding state is ‘‘0’’; sharing of computational resources between different digital
whereas when the biased voltage is in the ‘‘1’’ level, the coding modules and dynamic reconfiguration in response to changing
state is ‘‘1’’. Hence the coding state of each meta-atom on situations. The software-defined metamaterial will be an opti-
the metasurface is fully manipulated by the digital state of the mization of many advanced functions and lower hardware
controlling network, yielding completely digital controls to the requirements.
EM properties of metasurfaces. This is quite different from Along this direction of evolution, the final target will be
the digital representation of effective medium parameters. The cognitive metamaterials (see Fig. 20, the right branch). On the
digital metasurface is easily connected with digital hardware in basis of software-defined features, a cognitive metamaterial
information technology, such as FPGA, to produce a program- should have online learning ability to be cognitive to surrounding
mable metasurface. The significance of a field programmable environments and be adaptive to various conditions. It should
metasurface is real-time manipulations to EM waves in a pro- contain a smart digital signal processing system to interact with
grammed way through FPGA. Multiple distinctly different func- the surrounding environments, and an information-feedback
tionalities can be instantly switched by a single programmable system to automatically improve the ability. The cognitive meta-
metasurface, such as directive radiations, multi-beam reflections material should also be able to store information and data, which
with designed numbers, and even random scattering to avoid will be used for the machine learning. Different from traditional
radar detections. According to the programmable metasurface, metamaterials, future software-defined metamaterials and cogni-
new-concept radar can be realized. The next stage in developing tive metamaterials will be complicated systems, which may cover
the programmable metasurface towards practical applications all functions of radar, wireless communications, and high-
should solve a series of technological and methodological diffi- resolution imaging simultaneously, thereby breaking the barrier
culties, including wideband operations, new designs of digital of modern information systems.
meta-atoms, new principles of digital meta-atoms based on We remark that the programmable metasurfaces have been
liquid crystals and graphene, large-pixel controlling network, presented only in the microwave frequency using pin-diodes.
and multi-bit digital controls. However, with appropriate active materials, we could (in principle)
The digital coding representation makes it convenient to implement programmable metasurfaces at higher frequencies
introduce the Shannon entropy on metasurfaces to measure such as terahertz, infrared, and visible light spectra. For example,
their information capacities quantitatively, which can provide by doping a thin layer of p-type gallium arsenide (GaAs) on an
an information-based explanation to the single-sensor imaging intrinsic GaAs substrate, one could obtain an amplitude

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3665
 View Article Online

Journal of Materials Chemistry C Review

modulation of terahertz waves by tuning the applied voltage.120 21 N. F. Yu and F. Capasso, Nat. Mater., 2014, 13, 139–150.
Although there is currently an absence of research on the active 22 D. R. Smith, S. Schultz, P. Markoš and C. M. Soukoulis,
control of light in a pixel-by-pixel manner at infrared and visible Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 65, 195104.
light spectra, the theory of the coding metasurface presented in 23 R. Liu, T. J. Cui, D. Huang, B. Zhao and D. R. Smith, Phys.
this review serves as a general guide to design programmable Rev. E: Stat., Nonlinear, Soft Matter Phys., 2007, 76, 026606.
metasurfaces at higher frequencies in the future. 24 N. F. Yu, P. Genevet, M. A. Kats, F. Aieta, J. P. Tetienne,
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

F. Capasso and Z. Gaburro, Science, 2011, 334, 333–337.

25 F. Aieta, P. Genevet, N. Yu, M. A. Kats, Z. Gaburro and
Acknowledgements F. Capasso, Nano Lett., 2012, 12, 1702–1706.
26 X. Ni, N. K. Emani, A. V. Kildishev, A. Boltasseva and
This work was supported by the National Natural Science Foun-
V. M. Shalaev, Science, 2012, 335, 427.
dation of China (61138001, 61302018, 61401089, 61401091,
27 N. K. Grady, J. E. Heyes, D. R. Chowdhury, Y. Zeng,
61571117, 61501112, 61501117, and 61631007), National Instru-
M. T. Reiten, A. K. Azad, A. J. Taylor, D. A. Dalvit and
mentation Program (2013YQ200647), and 111 Project (111-2-05).
H. T. Chen, Science, 2013, 340, 1304–1307.
28 Z. Y. Li, E. Palacios, S. Butun and K. Aydin, Nano Lett.,
References 2015, 15, 1615–1621.
29 G. Li, M. Kang, S. Chen, S. Zhang, E. Y. Pun, K. W. Cheah
1 D. Smith and R. Liu, Metamaterials: theory, design, and and J. Li, Nano Lett., 2013, 13, 4148–4151.
applications, Springer, New York, 2010. 30 E. Karimi, S. A. Schulz, I. De Leon, H. Qassim, J. Upham
2 V. G. Veselago, Sov. Phys. Uspekhi, 1968, 10, 509–514. and R. W. Boyd, Light: Sci. Appl., 2014, 3, e167.
3 R. A. Shelby, D. R. Smith and S. Schultz, Science, 2001, 292, 31 Y. Yan, G. Xie, M. P. Lavery, H. Huang, N. Ahmed, C. Bao,
77–79. Y. Ren, Y. Cao, L. Li, Z. Zhao, A. F. Molisch, M. Tur,
4 J. B. Pendry, Phys. Rev. Lett., 2000, 85, 3966–3969. M. J. Padgett and A. E. Willner, Nat. Commun., 2014,
5 N. Fang, H. Lee, C. Sun and X. Zhang, Science, 2005, 308, 5, 4876.
534–537. 32 J. Jin, M. Pu, Y. Wang, X. Li, X. Ma, J. Luo, Z. Zhao, P. Gao
6 Z. Liu, H. Lee, Y. Xiong, C. Sun and X. Zhang, Science, 2007, and X. Luo, Adv. Mater. Technol., 2016, 1600201.
315, 1686. 33 M. Q. Mehmood, S. Mei, S. Hussain, K. Huang, S. Y. Siew,
7 J. B. Pendry and D. Schurig, Science, 2006, 312, 1780–1782. L. Zhang, T. Zhang, X. Ling, H. Liu, J. Teng, A. Danner,
8 U. Leonhardt, Optical conformal mapping, Science, 2006, S. Zhang and C. W. Qiu, Adv. Mater., 2016, 28, 2533–2539.
312, 1777–1780. 34 S. Sun, Q. He, S. Xiao, Q. Xu, X. Li and L. Zhou, Nat. Mater.,
9 D. Schurig, J. J. Mock, B. J. Justice, S. A. Cummer, 2012, 11, 426–431.
J. B. Pendry, A. F. Starr and D. R. Smith, Science, 2006, 35 J. Wang, S. Qu, H. Ma, Z. Xu, A. Zhang, H. Zhou, H. Chen
314, 977–980. and Y. Li, Appl. Phys. Lett., 2012, 101, 201104.
10 J. Li and J. B. Pendry, Phys. Rev. Lett., 2008, 101, 203901. 36 J. Lin, J. P. B. Mueller, Q. Wang, G. H. Yuan, N. Antoniou,
11 R. Liu, C. Ji, J. J. Mock, J. Y. Chin, T. J. Cui and D. R. Smith, X. C. Yuan and F. Capasso, Science, 2013, 340, 331–334.
Science, 2009, 323, 366–369. 37 L. Huang, X. Chen, B. Bai, Q. Tan, G. Jin, T. Zentgraf and
12 T. Ergin, N. Stenger, P. Brenner, J. B. Pendry and S. Zhang, Light: Sci. Appl., 2013, 2, e70.
M. Wegener, Science, 2010, 328, 337–339. 38 L. Huang, X. Chen, H. Muhlenbernd, G. Li, B. Bai, Q. Tan,
13 H. F. Ma and T. J. Cui, Nat. Commun., 2010, 1, 21. G. Jin, T. Zentgraf and S. Zhang, Nano Lett., 2012, 12,
14 E. F. Kuester, M. A. Mohamed, M. Piket-May and C. L. 5750–5755.
Holloway, IEEE Trans. Antennas Propag., 2003, 51, 2641–2651. 39 X. Ding, F. Monticone, K. Zhang, L. Zhang, D. Gao,
15 F. Falcone, T. Lopetegi, M. A. G. Laso, J. D. Baena, S. N. Burokur, A. de Lustrac, Q. Wu, C. W. Qiu and
J. Bonache, M. Beruete, R. Marques, F. Martin and A. Alu, Adv. Mater., 2015, 27, 1195–1200.
M. Sorolla, Phys. Rev. Lett., 2004, 93, 197401. 40 M. Tymchenko, J. S. Gomez-Diaz, J. Lee, N. Nookala,
16 C. L. Holloway, M. A. Mohamed, E. F. Kuester and M. A. Belkin and A. Alu, Phys. Rev. Lett., 2015, 115, 207403.
A. Dienstfrey, IEEE Trans. Electromagn. Compat., 2005, 47, 41 E. Maguid, I. Yulevich, D. Veksler, V. Kleiner, M. L. Brongersma
853–865. and E. Hasman, Science, 2016, 352, 1202–1206.
17 C. L. Holloway, A. Dienstfrey, E. F. Kuester, J. F. O’Hara, 42 K. Chen, Y. Feng, Z. Yang, L. Cui, J. Zhao, B. Zhu and
A. K. Azad and A. J. Taylor, Metamaterials, 2009, 3, 100–112. T. Jiang, Sci. Rep., 2016, 6, 35968.
18 C. L. Holloway, E. F. Kuester and A. Dienstfrey, IEEE 43 M. V. Berry, J. Mod. Opt., 1987, 34, 1401–1407.
Antennas Wireless Propag. Lett., 2011, 10, 1507–1511. 44 Z. Bomzon, G. Biener, V. Kleiner and E. Hasman, Opt. Lett.,
19 C. L. Holloway, E. F. Kuester, J. A. Gordon, J. O’Hara, 2002, 27, 1141–1143.
J. Booth and D. R. Smith, IEEE Antennas Propag. Mag., 45 W. Luo, S. Xiao, Q. He, S. Sun and L. Zhou, Adv. Opt. Mater.,
2012, 54, 10–35. 2015, 3, 1102–1108.
20 A. V. Kildishev, A. Boltasseva and V. M. Shalaev, Science, 46 X. J. Ni, A. V. Kildishev and V. M. Shalaev, Nat. Commun.,
2013, 339, 6125. 2013, 4, 2807.

3666 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017
 View Article Online

Review Journal of Materials Chemistry C

47 L. Huang, X. Chen, H. Mühlenbernd, H. Zhang, S. Chen, 71 Y. Zhao, M. A. Belkin and A. Alu, Nat. Commun., 2012,
B. Bai, Q. Tan, G. Jin, K.-W. Cheah, C.-W. Qiu, J. Li, 3, 870.
T. Zentgraf and S. Zhang, Nat. Commun., 2013, 4, 2808. 72 Z. Li, W. Liu, H. Cheng, S. Chen and J. Tian, Sci. Rep., 2015,
48 Y. W. Huang, W. T. Chen, W. Y. Tsai, P. C. Wu, C. M. Wang, 5, 18106.
G. Sun and D. P. Tsai, Nano Lett., 2015, 15, 3122–3127. 73 X. Gao, X. Han, W. P. Cao, H. O. Li, H. F. Ma and T. J. Cui,
49 G. Zheng, H. Muhlenbernd, M. Kenney, G. Li, T. Zentgraf IEEE Trans. Antennas Propag., 2015, 63, 3522–3530.
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

and S. Zhang, Nat. Nanotechnol., 2015, 10, 308–312. 74 J. Y. Yin, X. Wan, Q. Zhang and T. J. Cui, Sci. Rep., 2015,
50 W. T. Chen, K. Y. Yang, C. M. Wang, Y. W. Huang, G. Sun, 5, 12476.
I. D. Chiang, C. Y. Liao, W. L. Hsu, H. T. Lin, S. Sun, 75 S. N. Burokur, J. P. Daniel, P. Ratajczak and A. de Lustrac,
L. Zhou, A. Q. Liu and D. P. Tsai, Nano Lett., 2014, 14, Appl. Phys. Lett., 2010, 97, 064101.
225–230. 76 B. O. Zhu, J. Zhao and Y. Feng, Sci. Rep., 2013, 3, 3059.
51 D. Wen, F. Yue, G. Li, G. Zheng, K. Chan, S. Chen, M. Chen, 77 C. Zeng, X. Liu and G. Wang, Sci. Rep., 2014, 4, 5763.
K. F. Li, P. W. Wong, K. W. Cheah, E. Y. Pun, S. Zhang and 78 B. O. Zhu, K. Chen, N. Jia, L. Sun, J. Zhao, T. Jiang and
X. Chen, Nat. Commun., 2015, 6, 8241. Y. Feng, Sci. Rep., 2014, 4, 4971.
52 L. Huang, H. Muhlenbernd, X. Li, X. Song, B. Bai, Y. Wang 79 Z. Miao, Q. Wu, X. Li, Q. He, K. Ding, Z. An, Y. Zhang and
and T. Zentgraf, Adv. Mater., 2015, 27, 6444–6449. L. Zhou, Phys. Rev. X, 2015, 5, 041027.
53 W. Ye, F. Zeuner, X. Li, B. Reineke, S. He, C. W. Qiu, J. Liu, 80 H. X. Xu, S. Sun, S. Tang, S. Ma, Q. He, G. M. Wang, T. Cai,
Y. Wang, S. Zhang and T. Zentgraf, Nat. Commun., 2016, H. P. Li and L. Zhou, Sci. Rep., 2016, 6, 27503.
7, 11930. 81 Y. B. Li, L. L. Li, B. B. Xu, W. Wu, R. Y. Wu, X. Wan,
54 X. Zhang, Z. Tian, W. Yue, J. Gu, S. Zhang, J. Han and Q. Cheng and T. J. Cui, Sci. Rep., 2016, 6, 23731.
W. Zhang, Adv. Mater., 2013, 25, 4567–4572. 82 J. Yang, C. Gong, L. Sun, P. Chen, L. Lin and W. Liu,
55 L. Liu, X. Zhang, M. Kenney, X. Su, N. Xu, C. Ouyang, Sci. Rep., 2016, 6, 38732.
Y. Shi, J. Han, W. Zhang and S. Zhang, Adv. Mater., 2014, 83 P. Gutruf, C. Zou, W. Withayachumnankul, M. Bhaskaran,
26, 5031–5036. S. Sriram and C. Fumeaux, ACS Nano, 2016, 10, 133–141.
56 Q. Wang, X. Zhang, Y. Xu, J. Gu, Y. Li, Z. Tian, R. Singh, 84 J. A. Gordon, C. L. Holloway, J. Booth, S. Kim, Y. Wang,
S. Zhang, J. Han and W. Zhang, Sci. Rep., 2016, 6, 32867. J. Baker-Jarvis and D. R. Novotny, Phys. Rev. B: Condens.
57 F. Monticone, N. M. Estakhri and A. Alu, Phys. Rev. Lett., Matter Mater. Phys., 2011, 83, 205130.
2013, 110, 203903. 85 J. Zhao, Q. Cheng, X. K. Wang, M. J. Yuan, X. Zhou, X. J. Fu,
58 C. Pfeiffer and A. Grbic, Phys. Rev. Lett., 2013, 110, 197401. M. Q. Qi, S. Liu, H. B. Chen, Y. Zhang and T. J. Cui, Adv.
59 C. Pfeiffer and A. Grbic, IEEE Trans. Microwave Theory Opt. Mater., 2016, 4, 1773–1779.
Tech., 2013, 61, 4407–4417. 86 D. S. Dong, J. Yang, Q. Cheng, J. Zhao, L. H. Gao, S. J. Ma,
60 C. Pfeiffer, N. K. Emani, A. M. Shaltout, A. Boltasseva, S. Liu, H. B. Chen, Q. He, W. W. Liu, Z. Fang, L. Zhou and
V. M. Shalaev and A. Grbic, Nano Lett., 2014, 14, 2491–2497. T. J. Cui, Adv. Opt. Mater., 2015, 3, 1405–1410.
61 S. L. Jia, X. Wan, D. Bao, Y. J. Zhao and T. J. Cui, Laser 87 S. Liu, H. Chen and T. J. Cui, Appl. Phys. Lett., 2015,
Photonics Rev., 2015, 9, 545–553. 106, 151601.
62 J. P. S. Wong, M. Selvanayagam and G. V. Eleftheriades, 88 J. Hao, J. Wang, X. Liu, W. J. Padilla, L. Zhou and M. Qiu,
Photonic. Nanostruct., 2014, 12, 360–375. Appl. Phys. Lett., 2010, 96, 251104.
63 A. Epstein, J. P. Wong and G. V. Eleftheriades, Nat. Commun., 89 N. Yu, F. Aieta, P. Genevet, M. A. Kats, Z. Gaburro and
2016, 7, 10360. F. Capasso, Nano Lett., 2012, 12, 6328–6333.
64 X. Wan, S. L. Jia, T. J. Cui and Y. J. Zhao, Sci. Rep., 2016, 90 W. L. Chan, H.-T. Chen, A. J. Taylor, I. Brener, M. J. Cich
6, 25639. and D. M. Mittleman, Appl. Phys. Lett., 2009, 94, 213511.
65 F. Aieta, P. Genevet, M. A. Kats, N. Yu, R. Blanchard, 91 Z. Xie, X. Wang, J. Ye, S. Feng, W. Sun, T. Akalin and
Z. Gaburro and F. Capasso, Nano Lett., 2012, 12, 4932–4936. Y. Zhang, Sci. Rep., 2013, 3, 3347.
66 X. Chen, L. Huang, H. Mühlenbernd, G. Li, B. Bai, Q. Tan, 92 C. M. Watts, D. Shrekenhamer, J. Montoya, G. Lipworth,
G. Jin, C.-W. Qiu, T. Zentgraf and S. Zhang, Adv. Opt. J. Hunt, T. Sleasman, S. Krishna, D. R. Smith and
Mater., 2013, 1, 517–521. W. J. Padilla, Nat. Photonics, 2014, 8, 605–609.
67 X. J. Ni, S. Ishii, A. V. Kildishev and V. M. Shalaev, Light: Sci. 93 C. Pfeiffer and A. Grbic, Phys. Rev. Appl., 2014, 2, 044012.
Appl., 2013, 2, e72. 94 S. Liu, A. Noor, L. L. Du, L. Zhang, Q. Xu, K. Luan,
68 Q. Wang, X. Zhang, Y. Xu, Z. Tian, J. Gu, W. Yue, T. Q. Wang, Z. Tian, W. X. Tang, J. G. Han, W. L. Zhang,
S. Zhang, J. Han and W. Zhang, Adv. Opt. Mater., 2015, 3, X. Y. Zhou, Q. Cheng and T. J. Cui, ACS Photonics, 2016, 3,
779–785. 1968–1977.
69 M. Khorasaninejad, W. T. Chen, R. C. Devlin, J. Oh, 95 C. Pfeiffer, C. Zhang, V. Ray, L. J. Guo and A. Grbic, Phys.
A. Y. Zhu and F. Capasso, Science, 2016, 352, 1190–1194. Rev. Lett., 2014, 113, 023902.
70 J. K. Gansel, M. Thiel, M. S. Rill, M. Decker, K. Bade, 96 J. H. Shi, H. F. Ma, C. Y. Guan, Z. P. Wang and
V. Saile, G. von Freymann, S. Linden and M. Wegener, T. J. Cui, Phys. Rev. B: Condens. Matter Mater. Phys., 2014,
Science, 2009, 325, 1513–1515. 89, 165127.

This journal is © The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 3644--3668 | 3667
 View Article Online

Journal of Materials Chemistry C Review

97 J. Y. Kim, B. J. Kang, J. Park, Y. M. Bahk, W. T. Kim, J. Rhie, 108 X. Shen, Y. Yang, Y. Zang, J. Gu, J. Han, W. Zhang and
H. Jeon, F. Rotermund and D. S. Kim, Nano Lett., 2015, 15, T. J. Cui, Triple-band terahertz metamaterial absorber:
6683–6688. Design, experiment, and physical interpretation, Appl.
98 G. Li, S. Chen, N. Pholchai, B. Reineke, P. W. Wong, Phys. Lett., 2012, 101, 154102.
E. Y. Pun, K. W. Cheah, T. Zentgraf and S. Zhang, Nat. 109 S. Liu, J. Zhuge, S. Ma, H. Chen, D. Bao, Q. He, L. Zhou and
Mater., 2015, 14, 607–612. T. J. Cui, J. Appl. Phys., 2015, 118, 245304.
Published on 14 March 2017. Downloaded by Southeast University - Jiulonghu Campus on 03/05/2017 11:58:30.

99 C. Della Giovampaola and N. Engheta, Nat. Mater., 2014, 110 P. Nayeri, F. Yang and A. Z. Elsherbeni, IEEE Antennas
13, 1115–1121. Propag. Mag., 2015, 57, 32–47.
100 T. J. Cui, M. Q. Qi, X. Wan, J. Zhao and Q. Cheng, Light: Sci. 111 P. Nayeri, F. Yang and A. Z. Elsherbeni, IEEE Trans.
Appl., 2014, 3, e218. Antennas Propag., 2013, 61, 4588–4597.
101 L. H. Gao, Q. Cheng, J. Yang, S. J. Ma, J. Zhao, S. Liu, 112 X. Wan, M. Q. Qi, T. Y. Chen and T. J. Cui, Sci. Rep., 2016,
H.-B. Chen, Q. He, W. X. Jiang, H. F. Ma, Q. Y. Wen, 6, 20663.
L. J. Liang, B. B. Jin, W. W. Liu, L. Zhou, J. Q. Yao, P. H. Wu 113 D. Lao, L. Li, J. Ding, Y. B. Li and T. J. Cui, Large-aperture
and T. J. Cui, Light: Sci. Appl., 2015, 4, e324. computational single-sensor microwave imager using 1-bit
102 L. Liang, M. Qi, J. Yang, X. Shen, J. Zhai, W. Xu, B. Jin, programmable coding metasurface at single frequency,
W. Liu, Y. Feng, C. Zhang, H. Lu, H. T. Chen, L. Kang, IEEE Trans. Antennas Propag., 2016, submitted.
W. Xu, J. Chen, T. J. Cui, P. Wu and S. Liu, Adv. Opt. Mater., 114 L. Li, T. J. Cui, W. Ji, S. Liu, J. Ding, X. Wan, Y. B. Li,
2015, 3, 1374–1380. M. Jiang, C. W. Qiu and S. Zhang, Electromagnetic repro-
103 S. Liu, T. J. Cui, Q. Xu, D. Bao, L. Du, X. Wan, W. X. Tang, grammable coding-metasurface holograms, Nat. Commun.,
C. Ouyang, X. Y. Zhou, H. Yuan, H. F. Ma, W. X. Jiang, J. Han, 2016, submitted.
W. Zhang and Q. Cheng, Light: Sci. Appl., 2016, 5, e16076. 115 S. Liu and T. J. Cui, IEEE J. Sel. Top. Quantum Electron.,
104 S. Liu, L. Zhang, Q. L. Yang, Q. Xu, Y. Yang, A. Noor, 2017, 23, 1–12.
Q. Zhang, S. Iqbal, X. Wan, Z. Tian, W. X. Tang, Q. Cheng, 116 S. Liu and T. J. Cui, Sci. Rep., 2016, 6, 37545.
J. G. Han, W. L. Zhang and T. J. Cui, Adv. Opt. Mater., 2016, 117 N. I. Zheludev, Science, 2016, 348, 973–974.
4, 1965–1973. 118 R. Day, The modern invention of information: Discourse,
105 S. Liu, T. J. Cui, L. Zhang, Q. Xu, Q. Wang, X. Wan, J. Q. Gu, history, and power, 2001.
W. X. Tang, M. Qing Qi, J. G. Han, W. L. Zhang, X. Y. Zhou 119 T. J. Cui and S. Liu, Opt. Photonics News, 2016, 59,
and Q. Cheng, Adv. Sci., 2016, 3, 1600156. December issue.
106 T. J. Cui, S. Liu and L. L. Li, Light: Sci. Appl., 2016, 5, e16172. 120 H.-T. Chen, W. J. Padilla, M. J. Cich, A. K. Azad, R. D. Averitt
107 H. Tao, N. I. Landy, C. M. Bingham, X. Zhang, R. D. Averitt and A. J. Taylor, A metamaterial solid-state terahertz phase
and W. J. Padilla, Opt. Express, 2008, 16, 7181–7188. modulator, Nat. Photonics, 2009, 3, 148–151.

3668 | J. Mater. Chem. C, 2017, 5, 3644--3668 This journal is © The Royal Society of Chemistry 2017