6 IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 10, NO.

1, MARCH 2020

Programmable Metamaterials for Software-Defined

Electromagnetic Control: Circuits,
Systems, and Architectures
Sergi Abadal , Member, IEEE, Tie-Jun Cui , Fellow, IEEE, Tony Low,
and Julius Georgiou , Senior Member, IEEE
Abstract— Metamaterials and their two-dimensional ana- However, due to their resonant nature, most existing meta-
logues, metasurfaces, have recently attracted enormous attention materials have been typically tailored for a single application
because of their powerful control over electromagnetic (EM) working under preset conditions (e.g. electromagnetic cloaking
waves from microwave to visible range. Moreover, by introducing for a fixed angle of incidence and at a particular frequency)
explicit control of its sub-wavelength unit cells, a metamaterial and cannot be reused. This clearly limits their applicability
can become programmable. Programmable metamaterials may
not only host multiple EM functionalities that can be cho-
and, considering the complexity of the current metamaterial
sen or combined through simple software directives, but also be design processes, increases their cost.
provided with means to adapt to the environment or communicate In response to these limitations, recent years have witnessed
with other metamaterials, thereby enabling a myriad of appli- the advent of different types of tunable and reconfigurable
cations in sensing, imaging, or communications. The realization metamaterials [11], [12]. Reconfigurability is achieved by
of such a software-driven cyber-physical vision comes, however, using components that tune the metamaterial response or that
at the cost of significant hardware requirements. In this paper, completely change the current flows within the metama-
recent progress in the field of programmable metasurfaces is terial, including varactors or variable resistors [13], PIN
reviewed, cutting across layers from the application down to diodes or switches [14], microelectromechanical systems
the device and technology levels. The main aim is to present (MEMS) [15], [16], mechanical parts [17], or advanced mate-
the current status, main benefits, and key challenges of this
thriving research area with a tutorial spirit and from a hardware
rials such as graphene [18], [19] or liquid crystals [20].
perspective. A natural step further in this domain has been the proposal
of the programmable metamaterial paradigm. To achieve pro-
Index Terms— Programmable metamaterials, programmable grammability, metamaterials need to be provided with ways to
metasurfaces, software-defined electromagnetic control, cyber- control the components that realize the reconfigurability. Cur-
physical systems, embedded digital systems. rent developments generally account for a Field Programmable
I. I NTRODUCTION Gate Array (FPGA) that drives an appropriately digitized
version of a reconfigurable metamaterial [14], [21], [22].
M ETAMATERIALS are artificial structures that enable
the realization of novel electromagnetic (EM) com-
ponents with engineered and even unnatural functionali-
An evolution of this approach has been envisaged recently,
where the external FPGA is replaced by a network of tiny
ties. Through the unprecedented control over electromagnetic sensors and controllers that are integrated within the structure
waves, achieved thanks to the manipulation of impedance at of the metamaterial and interfaced with the real world [23],
subwavelength scale, metamaterials have led to significant [24]. The ultimate goal of this cyber-physical system vision
breakthroughs in fields such as imaging, radar, or wireless is to turn metamaterials into software-driven, intelligent,
communications [1]–[4] across the whole spectrum [5]–[10]. autonomous yet interconnectable devices capable of host-
ing multiple functionalities and using them while adapting
Manuscript received January 27, 2020; revised February 16, 2020; accepted to the needs of the system.
February 18, 2020. Date of publication February 24, 2020; date of cur- Fundamentally, the behavior of programmable metamateri-
rent version March 12, 2020. This work was supported by the European
Commission under Grant H2020-736876-VISORSURF, and in part by the als hinges on the collective response of the unit cells. This,
National Key Research and Development Program of China under Grant effectively, represents a discretization of both the space (with
2017YFA0700201, Grant 2017YFA0700202, and Grant 2017YFA0700203. subwavelength granularity) and the response of each unit cell
This article was recommended by Guest Editor A. Wu. (Corresponding (with a discrete set of amplitudes and phases). Due to their
author: Sergi Abadal.)
Sergi Abadal is with the NaNoNetworking Center in Catalonia (N3Cat), inherent “digital” nature, programmable metamaterials have
Universitat Politècnica de Catalunya, 08034 Barcelona, Spain (e-mail: been associated to the coding metamaterial paradigm where
[email protected]). the device is encoded using a discrete set of unit cell states.
Tie-Jun Cui is with the School of Information Science and Engineering, The state of the metamaterial can be thus elegantly described
Southeast University, Nanjing 210096, China (e-mail: [email protected]). as a state matrix, thus providing a powerful and intuitive design
Tony Low is with the Department of Electrical and Computer Engi-
neering, University of Minnesota, Minneapolis, MN 55455 USA (e-mail: perspective while drawing a clear parallelism with information
[email protected]). theory. This opens new ways to model, compose, and design
Julius Georgiou is with the Department of Electrical and Com- metamaterials [25]–[28], and has proven helpful in tackling the
puter Engineering, University of Cyprus, 1678 Nicosia, Cyprus (e-mail: challenge of packing multiple functionalities within a single
[email protected]).
Color versions of one or more of the figures in this article are available device and serving multiple purposes concurrently.
online at https://ieeexplore.ieee.org. The cyber-physical view of programmable metamaterials,
Digital Object Identifier 10.1109/JETCAS.2020.2976165 with the added properties of intelligence and interconnectivity,
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
ABADAL et al.: PROGRAMMABLE METAMATERIALS FOR SOFTWARE-DEFINED ELECTROMAGNETIC CONTROL 7

of circuits and systems, thus encouraging the proposal of

smart, metamaterial-aware and opportunistic, highly stream-
lined solutions at the circuit, system, and full architecture
levels. In a longer term, the use of emerging technologies
such as graphene, nanoscale integrated sensors, opportunistic
energy harvesting, or neural networks for metamaterial control
may be a critical enablers of fully autonomous programmable
metamaterials.
This paper, as well as the special issue where it belongs to,
provides an overview of the research avenue of programmable
metamaterials and describes recent advances from a hardware
perspective. We cover the areas ranging from tunable meta-
material engineering, low-power circuits and systems design,
nanoelectronics, and system architecture, encompassing both
contributions close to production and more prospective pro-
posals with longer-term potential.
The paper is organized in a top-down approach. First,
in Section II, the fundamentals of programmable metamaterials
are reviewed from the system perspective, delving into the
Fig. 1. Graphical abstract of this tutorial paper. Applications demand prospects and applications of such systems. In Section III,
the deployment and use of programmable metasurfaces, whose architecture recent trends are surveyed from the architectural stand-
includes a number of tunable unit cells. Unit cells integrate different com- point, analyzing the existing FPGA-based and fully integrated
ponents, materials, and technologies. Finally, several cross-cutting paradigms
can impact the design of the whole system. approaches. Next, in Section IV, we focus on challenges and
advances at the component and circuit levels for practical
computation, communication, sensing, and tuning at the unit
cells. In Section V, we dive deeper down to the technol-
ogy, emphasizing novel materials, fabrication methods, and
integration approaches allowing to extend the programmable
metamaterial paradigm towards THz and optical frequencies.
Before concluding, cross-cutting issues and advanced topics
such as energy harvesting, approximate computing, or the
application of neural networks to coding metamaterials are
discussed in Section VI. Finally, the paper is concluded in
Section VII.
Fig. 2. From the application to the metasurface coding: applications require
the use of multiple functionalities (e.g. beam steering) with programmable
inputs (e.g. direction of reflected waves) that, depending on the metasurface II. A P RIMER ON P ROGRAMMABLE M ETASURFACES :
characteristics, lead to a particular metasurface coding. Coding refers to the O UTLOOK AND A PPLICATIONS
assignment of states to unit cells to achieve the target functionality and inputs
as accurately as possible. Any change to the functionality or inputs leads to Programmable metasurfaces, as the thin-film counterpart of
a different coding, hence the need for programmability. programmable metamaterials, serve to exemplify the process
from the application to the design and coding of the device.
is expected to open the door to a plethora of applications Such a process is illustrated in Fig. 2 and developed herein.
in areas such as sensing, communications, or imaging [12], We note that, although similar concepts can be applied to
[29]–[31]. A good example in communications would be the 3D metamaterials, issues related to fabrication complexity,
concept of programmable wireless environment, which aims to losses, cost, or bulkiness encourage the use of metasurfaces
revolutionize wireless communications with significant impact (i.e. metamaterials with subwavelength thickness) instead [36].
in 5G networking and AR/VR applications [32], [33]. In imag- Metasurfaces enable a new set of applications (Sec. II-A)
ing, the analogy between the unit cells of a programmable emerging from the unique and programmable control of the
metamaterial and the pixels of a screen allow to envisage EM response. In essence, an application requires the meta-
practical ways to create dynamic holograms [22]. Further, surface to implement a certain EM functionality (Sec. II-B)
implementing programmable metamaterials in the terahertz with a given set of inputs. The metasurface is then encoded
(THz) band could aid in closing the so-called terahertz gap, to that end, this is, the unit cells are set to the states that lead
with important implications in imaging or wireless communi- to the targeted functionality and inputs with the least possible
cations applications [9], [34], [35]. deviation (Sec. II-C). The ultimate aim of the programmable
While the potential of the programmable metamaterials metamaterials field is to be able of doing so with the minimum
is high, harnessing software-defined electromagnetic control effort, being able to automatically adapt to varying conditions
poses significant challenges at all levels, from the system over time, with little to no power consumption, among other
down to the circuit or the technology. For these devices desired properties listed in Section II-D.
to be practical, the associated control circuits must comply
with stringent limitations in terms of area, power consump- A. Emerging Applications
tion, or interference. At the same time, the control system By virtue of the unique real-time reconfigurability of pro-
needs to be powerful enough to cover a representative pool of grammable metasurfaces, many novel ideas and solutions
potential electromagnetic functionalities and applications. This have been introduced into traditional industries with numerous
imposes severe restrictions on the practicable design space emerging applications appearing. Additionally, extending the
8 IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 10, NO. 1, MARCH 2020

Fig. 3. Emerging applications enabled by the programmable metamaterial approach, including (a) programmable and self-adaptive wireless environments
[37], (b) simplified architectures for multi-channel wireless transceivers [38], (c) holographic images [22], and (d) self-adaptive smart metasurfaces [30].

use metasurfaces in the millimeter-wave (mmWave) and THz communication systems. This is of special importance in
bands is expected to open the door to multiple opportunities the mmWave and THz bands, where high spreading losses
across multiple fields. Here we list some of these applications, and the inability of EM waves to penetrate walls and most
emphasizing the key role of real-time reconfigurability and the objects creates a distance problem [42]. In this application
potential of mmWave-THz solutions when appropriate. context, especially at high frequencies, the main challenge is
1) Wireless Communications in 5G and Beyond: The field to develop metasurfaces that offer fine-grained programma-
of wireless communications is seeing a revolution in the bility but, at the same time, are cost-efficient and scalable
use of programmable metasurfaces as intelligent reflecting to relatively large walls. Otherwise, lower cost alternatives to
surfaces. This enables the concept of programmable wireless the same problem (e.g. using active relays) will probably be
environments as seen in Fig. 3(a). A wireless environment is adopted instead [43].
defined as a group of physical objects that can significantly Another application related to wireless communications is
alter the propagation of EM waves among communicating the development of simplified architectures for wireless
devices [32], [37], [39]–[41]. In general, the environment communication transmitters. In the traditional wireless com-
shows a negative effect on the communication efficiency due to munication transmitters, both in the heterodyne or homodyne
the energy loss from material absorption and free-space propa- architectures, the baseband signal processing module and the
gation, and the multipath effects caused by EM wave scattering radio-frequency (RF) chain module are separated from each
from the diffraction, reflection, and refraction phenomena. other. The information is modulated in the baseband and
However, the emergence of programmable metasurfaces fea- then up-converted to the RF frequency through the RF chain.
turing EM-wave re-engineering enables the implementation of As more and more protocols and standards are proposed to
programmable wireless environments. By introducing different satisfy the increasing demands of stability, high speed and low
functions like focusing, scattering and steering at selected latency in the information era, the RF components are facing
spots and at different rates, the programmable metasurfaces higher complexity requirements in design and manufacturing,
make the wireless environment capable of handling the EM resulting in extremely high hardware cost and energy con-
propagation in a manner reminiscent of routers and fire- sumption in practical implementation. Recently, several sim-
walls in classical networking, which mitigate both propagation plified wireless-communication transmitter architectures based
loss and multipath effects existing in the current wireless on programmable metasurfaces have been proposed with low
ABADAL et al.: PROGRAMMABLE METAMATERIALS FOR SOFTWARE-DEFINED ELECTROMAGNETIC CONTROL 9

hardware complexity, high cost effectivity, and good perfor-

mance [38], [44]. As shown in Fig. 3(b), the metasurfaces are
utilized to directly modulate the carrier wave in multiple chan-
nels using the baseband signal without the need of core com-
ponents of the RF chain such as mixers, filters, amplifiers, etc.
At high frequencies, this is crucial as these components can
be lossy and significantly complex. In any case, the required
refresh rate of the metasurface is thus the modulation rate.
Although the principle is similar to that of direct antenna
modulation (DAM) technology [45], the programmable meta-
surface approach possesses better performance and higher effi-
Fig. 4. Summary of electromagnetic functionalities achievable with pro-
ciency due to more flexible tunability of EM waves. The novel grammable metasurfaces. Note that, depending on the unit cell structure,
architecture has shown great potential to bring a new paradigm such functionalities can be achieved in reflection, transmission, or both.
that naturally integrates the signal processing algorithms in Programmable metasurfaces can also host multiple functionalities, implement
information science and the hardware of the programmable distributed intelligence, and react to the environment.
metasurfaces for future-generation wireless communications.
Eventually, a great challenge will be to prove whether sim- coding and programmable metamaterials in the literature,
plified transceiver architectures, together with metasurface- classified in Figure 4.
based steering mechanisms, lead to mmWave-THz chipsets • Wavefront shaping. Metasurfaces have shown their pow-
that, in contrast to classical architectures, are practical enough erful capabilities in engineering the EM wavefronts. Cod-
for their take up in future wireless communication systems. ing and programmable metasurfaces have the advantages
2) Holographic Imaging: Holography is one of the most of simplifying the design procedures and of switching
promising imaging techniques that attract tremendous attention among different functions in real time, which can be
covering many applications in display, security, data storage, used to shape both reflection and transmission wavefronts
etc. However, most of the conventional holograms suffer from on a shared aperture [47]. Furthermore, the full-space
low resolution, limited image quality, and fixed holographic wavefront shaping can be dynamically switched between
image. The programmable metasurface-based holograms, illus- reflection or transmission modes by changing the PIN
trated in Fig. 3(c), feature two major advantages compared diodes states [48].
with the conventional holograms [22]. Firstly, the outstanding • Focusing. Programmable metasurfaces have also been
recording ability for both amplitude and phase information used to focus the EM fields. By applying the specially-
of wavefront in the subwavelength scale provide excellent designed coding matrices, a programmable metasurface
spatial resolution. Secondly, the pixel (i.e. the metasurface can be used to generate arbitrary focusing points dynam-
unit cell) level programmability greatly contributes to real- ically, and field intensity at each point can also be
time and arbitrary holographic images. Metasurfaces should controlled independently [49]. Besides, a programmable
thus be programmable at typical refresh rates of displays, this Huygens’ metasurface was realized to focus the transmit-
is, around 60 frames per second. If achieved, these remarkable ted fields dynamically with high efficiency [50].
properties make the programmable metasurface one of the • Anomalous reflection. Metasurfaces can introduce an
competitive candidates in the future holographic imaging and abrupt phase shift on the interface by designing the
even holographic movies. corresponding unit cell, which will result in phase discon-
3) Self-Adaptive Systems: Recently, researchers have pur- tinuities. Based on the idea of generalized Snell’s laws,
sued the implementation of intelligent metasurfaces, which the programmable metasurface can realize anomalous
are self-adaptive to control the presented functions by the reflection by using the phase-gradient space-coding
metasurface itself [30], [46]. As shown in Fig. 3(d), the pro- sequence [51].
grammable metasurface constructs a closed-loop system that, • Asymmetric transmission. Metamaterials can be used
in collaboration with some specific sensors and feedback to realize the asymmetric transmission by using chiral
circuits, can operate on its own without human intervention. structures. Programmable chiral metamaterials integrated
The speed at which the metasurface needs to be updated will with PIN diodes have been proposed to achieve dynamic
depend on the end-user application and the main challenge control of the asymmetric transmission by changing the
resides in the building of the lightweight, robust, real-time control voltages [52].
control algorithms that will close the loop. In any case, such • Polarization control. Polarization is a very important
systems are expected to open up a new avenue for future dimension of EM waves. How to control the polarization
unmanned devices and systems that are consistent with the state has always been the concern of researchers. Pro-
ambient environment, which further pave the way towards grammable metasurfaces also have an outstanding ability
intelligent and cognitive metasurfaces. in controlling the polarization, which can be used for
reconfigurable linear polarization conversion [14]. Fur-
B. Electromagnetic Functionalities thermore, the varactors-based programmable metasurface
Digital coding and programmable metamaterials can be can be dynamically switched among the functionalities
applied to realize diverse EM functionalities due to their of linear-to-linear, linear-to-elliptical, and liner-to-circular
advantages. By changing the coding matrices pre-stored in polarization conversion [53].
an FPGA, many functionalities can be switched in real time, • Vorticity. Vortex waves carrying orbital angular momen-
leading to many promising applications. In fact, the possibility tum (OAM) can be used for orthogonal multiplexing
of manipulating the response in the time domain opens the communication via different OAM modes, which has
door to multiple new opportunities. Next, we briefly describe the feature of helical wavefront and the property of
a comprehensive set of EM functionalities implemented by mode orthogonality. Recently, coding metasurfaces were
10 IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 10, NO. 1, MARCH 2020

developed to generate vortex beams with different modes,

including spin-controlled vortex beams [54].
• Absorption. Metamaterials have been widely studied
for absorption in the past decade. Most metamaterial
absorbers are designed in a fixed frequency band.
Programmable metamaterials can be utilized to absorb
the incident waves at different frequencies by electrically
controlling the biasing voltages on the varactors [55],
which can extend the working frequency range of
absorbers.
• Diffuse scattering. By optimizing the coding pattern,
coding metasurfaces can be used for scattering-
signature reduction by the strategy of diffuse scattering
in both microwave and terahertz regime [56]. The
recently proposed space-time-coding modulation in Fig. 5. Illustration of the coding of a metasurface with a set of four states
and existing methods to encode a metasurface for different functionalities.
programmable metasurfaces can also redistribute the
scattered power in both space and frequency domains
uniformly, thereby leading to unprecedented levels of
Radar Cross-Section (RCS) reduction [28]. responses satisfy the conditions for the 1-bit and multi-bit
• Frequency shifting. By modulating the metasurface digital states. These concepts can also be promoted to
properties in the time domain, a programmable polarization- or frequency-dependent coding metamateri-
metasurface can control the reflected waves in the als to form multi-channel coding strategies [61], [62].
frequency domain. For instance, frequency shifting • Amplitude coding states. The coding principle also
can be realized by imparting suitable time modulation enables digital descriptions of amplitudes, which can con-
signals on programmable metasurface, such as frequency trol the radiating power by using indium tin oxide (ITO)
conversion [38], and harmonic frequency generation [28]. films or amplifiers [63]. A combination of amplitude and
• Nonreciprocity. By introducing the time-varying phase coding further enables full-space radiation control,
approach, the space-time-coding programmable metasur- in which the incident polarizations or PIN diodes are
face was successfully used to break reciprocity which utilized for the real-time modulations for the amplitude
relied on inducing suitable spatiotemporal phase gradients coding states [48].
in a programmable way via digital coding modulation 2) The Fundamental Methods to Map Metasurface States to
[28], [57]. Besides, the amplifier transistors-based Specific Functionalities: Due to the digital expressions of EM
programmable metasurface was also proposed to realize properties, the EM functionalities of coding and programmable
forward and backward nonreciprocal transmissions, but metamaterials are decided by the distributions of the coding
without using the time-varying modulation [58]. states, namely the coding patterns. Some of the coding patterns
It is worth noting that, depending on the unit cell struc- rely on known, often analytical solutions, derived from funda-
ture, the functionalities are achieved either in reflection [18], mental rules such as the generalized Snell’s laws of reflection.
[56] or transmission [59], [60]. Specific designs may support Such common coding patterns are the following:
both modes simultaneously [47] or switch among them [48]. • Linear phase gradient for anomalous reflection and
transmission. The directions of the scattering waves can
be controlled by forming linear phase gradient coding
C. Metamaterial Coding
patterns on the interfaces, according to the generalized
Coding and programmable metamaterials describe the EM Snell’s laws of reflection [64]. The higher the number of
performance from the information aspect, providing us novel unit cell states, the more anomalous manipulations (and
and brief approaches to design functional metamaterials [21], even negative reflection) can be achieved [60], [65]. The
[36]. The process of coding consists in assigning a par- detailed deflecting angle also relies on the period of the
ticular state to each unit cell of the metasurface among phase gradient and the working wavelength [61].
the set of possible states, so that final coding pattern • Concentric phase gradient for focusing and beam
attains the targeted functionality with the least error. For forming. When the coding states are arranged in concen-
instance, let us consider a unit cell designed to take four tric phase gradient, the incident planar EM wave can be
discrete values. Thus, the coding set is  = {s0 , s1 , s2 , s3 }. focused to a certain spotlight. The focal length depends
Depending on the desired functionality, different strategies can on the period of the phase gradient, whereas the lateral
be used for the mapping. Here, we briefly summarize the main position of the focal point is dependent on the position
aspects involved in the coding of programmable metamaterials, of the center of the concentric gradient [19]. Inversely,
as illustrated in Fig. 5. this coding pattern is also capable to form pencil beams
1) The Coding States of the Unit Cells: The coding states in far field by converting the spherical waves to planar
of the unit cells are usually employed to represent the EM waves [14].
coefficients, which are the fundamental factors when designing • Angular gradient for vortex beams. The phase cod-
the coding and programmable metamaterials. There are two ing states in angular gradient enable the radiations of
major coding schemes: vortex beams. The digital states with single or several
• Phase coding states. The phase coding scheme is the periods of 0–2π are placed spirally on the aperture,
most common one in coding and programmable metama- clockwise or anticlockwise, to construct different OAM
terials. By adopting two unit cells with different geome- modes. [54].
tries or introducing PIN diodes in the single unit cell with • Random pattern for RCS reduction. By estab-
dynamical modes [21], the reflected or transmitted phase lishing the coding states in random patterns other
ABADAL et al.: PROGRAMMABLE METAMATERIALS FOR SOFTWARE-DEFINED ELECTROMAGNETIC CONTROL 11

than the abovementioned gradient cases, the input TABLE I

power is reflected to arbitrary directions. Hence, R EPRESENTATIVE P ROGRAMMABLE M ETASURFACES IN THE L ITERATURE
the RCS is reduced notably [66]. Several optimiza-
tion methods including particle optimization, genetic
algorithms, or Golay-Rudin-Shapiro (GRS) polynomials
model have also been introduced into the coding pattern
design and accomplished almost homogeneously anom-
alous reflection [67].
The methods above are powerful, but do not generalize to
complex functionalities or arbitrary wavefronts. To that end,
several alternatives have arisen to facilitate the programming
of metasurfaces. Besides the use of analytical operations on
the coding patterns, explained in the subsection below, we find
that programmable metasurfaces are amenable to widely used
algorithms and optimization techniques, namely:
• Gerchberg-Saxton algorithm for holograms. Numer-
ical algorithms such as Gerchberg-Saxton (GS) allow
to build arbitrary holograms for near-field imaging and
information transmission [22], [49]. Especially in the pro-
grammable designs, the holograms are also reprogram-
mable, indicating that the multiple desired holographic
images can be realized in real time.
• Optimization algorithms and neural networks. To
achieve arbitrary complex wavefronts that cannot be
described as the combination of the patterns above, classi-
cal optimization algorithms can be employed. Evolution-
ary methods such as genetic algorithms or particle swarm
optimization have been applied in several works [14], in fact, an explosion of works attempting to demonstrate
[66]. Further, the use of neural networks as a powerful the benefits of programmable metasurfaces and the multiple
tool to program and control metasurfaces is an emerging possibilities they enable across the spectrum. Table I shows
possibility (see Section VI for more details). a selection of representative designs of the state of the art in
3) Analytical Operations on Multiple Coding Patterns: The coding metasurfaces.
methods described above can achieve a comprehensive set of However, the vision of programmable metasurfaces for
coding patterns for multiple functionalities. If more advanced software-driven EM control goes beyond the achievement of
functionalities are desired, more complicated analytical opera- particular EM functionalities and involves groundbreaking as
tions should be implemented on multiple coding patterns. The well as, in occasions, much-anticipated non-trivial features.
following two methods allow to build complex coding patterns We next list some of such features and relate them with the
as the combination of two or more simpler patterns, using emerging applications outlined in Section II-A:
intuitive mathematical formalisms: • Space-time reconfigurability: as we have seen in
• Convolution operations. The purpose of convolution
Section II-C, metasurfaces have traditionally encoded
operations is to realize more precise EM control, breaking spatial patterns within their structure in the form of
through the limitation of deflecting angle from the period spatial phase or amplitude profiles giving rise to the
of the phase gradient and the working wavelength. Con- required functionalities [64]. The advent of program-
volution operations perform Fourier transformations on mable metasurfaces, however, introduced the dimension
coding and programmable metamaterials, allows steering of time [28], [77]: metasurfaces can now be temporally
of the scattering pattern to an arbitrarily pre-designed modulated to implement non-reciprocal functionalities
direction only through the modulus of several coding [78], compensate for relativistic motion [79], integrate
patterns. Theoretically, the 1D and 2D convolution oper- signal and beam modulation within the antenna [38],
ations permit the radiating beams to radiate in the whole enable new modulation schemes [80], and many others.
• Multi-functionality: the idea of a metasurface plat-
half space [26].
• Addition theorem. In fact, the essence of convolution
form that can implement arbitrary spatiotemporal patterns
operations is the phase multiplication of the coding directly leads to think on the implementation of multiple
states. The other operation, addition starts from the basic functions or channels within the same metasurface, be it
EM theorem and originally involves all the complex in a discrete/multiplexed or a combined/concurrent fash-
phase parts e j ϕ for their addition, implying the final ion [23], [70], [81]. Designs are thus reusable for mul-
result would contain all the added components. Hence, tiple tasks or a single complex task where the metasur-
it means a superposition of the coding patterns, i.e., the face needs to exhibit different behaviors simultaneously,
corresponding scattering patterns and functionalities can as in programmable wireless environments [32], [39].
appear simultaneously on the shared aperture [27], [68]. Combined with space-time reconfigurability, this leads to
very powerful control of EM waves [77].
• Learnability: the addition of software mechanisms
D. Desired System Properties abstracting the physics enables the use of not only
The methods described in the previous section have allowed optimization techniques, but also machine learning algo-
to achieve a wide range of functionalities. There has been, rithms to model, predict, and control the operation of the
12 IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 10, NO. 1, MARCH 2020

metasurfaces for arbitrary functionalities and environment

conditions. Moreover, with the embedding of intelligence
within the metasurface, the metamaterial could automat-
ically adapt to changes in the environment without the
intervention of external components.
• Interconnectivity: by providing metamaterials with
means of internal and external communication, we are
opening the door to the coordination of multiple meta-
materials to cover large areas or implement heteroge-
neous, complex, emerging behaviors. Thanks to multi-
functionality, this does not imply the deployment of a set
of specialized metasurfaces, but rather the use a single
type of metamaterials that changes its behavior according
to the role that it is assigned within the swarm. This
has obvious implications in the programmable wireless
environments paradigm [32], [39].
• Resilience: enabling self-adaptivity of metamaterials to
changes in the environment also opens the door, naturally,
Fig. 6. Schematic representation of the two general types of programmable
to changes in the internal structure of the metamaterial metasurface architectures.
due to transient or persistent faults. Although it has been
shown that metasurfaces are inherently tolerant to faults
of a fraction of unit cells [82], one can envisage designs physical core and algorithm core), which can be essentially
where the impact can be minimized by incorporating, classified into two main architectures. They are illustrated
within the learning/optimization framework, their possi- in Figure 6 and described next.
ble occurrence.
• Perpetual Operation: as metasurfaces become active A. FPGA-Based Architectures
elements that consume power, a concern is whether
The general FPGA-based architecture of a programmable
the power consumption will impact on the auton-
metasurface system is represented in Figure 6(a) as seen in
omy or applicability of metamaterials. Through the
multiple works [14], [21], [28], [57]. The metasurface itself is
embedding of intelligence and self-adaptive optimization
responsible for providing programmable electromagnetic char-
frameworks, however, one can envisage platforms that
acteristics such as reflection/transmission amplitude, phase,
minimize power consumption by, for instance, keeping
polarization, etc. The control sub-system, on the other hand,
only the indispensable features alive. Power-gated unit
is mainly composed of an FPGA-based control circuit, which
cells can be considered as faults and resort to the
is assembled to implement the functional algorithm and drive
resilience of the platform to still adapt performance to the
the programmable metasurface. The specific configuration
power restrictions [82]. Furthermore, perpetual operation
of the interface or connector between the two sub-systems
is a desirable trait arising from the combination of ultra-
depends on the tuning technology used in the metasurface.
low power consumption and the likely incorporation of
Several examples include the voltage control circuit for the
energy harvesting means within the platform, as discussed
varactor/PIN diode based metasurface [14], [21], [28], [57],
in Sec. VI.
the light intensity control circuit for the photosensitive materi-
• Terahertz Operation: metamaterial and metasurface
als based metasurface [83], and the temperature control circuit
designs in the terahertz band already exist [9], although
for thermo-sensitive materials-based metasurfaces [84], [85].
most of them are either theoretical or with limited
The separated architecture allows the two sub-systems to
reconfigurability properties. Given the multiple opportu-
be designed independently, which is beneficial to the function
nities arising in the fields of imaging [35] or wireless
expansion and performance upgrade of the entire system
communications [34], among others, there is an interest to
through the use of different types of metasurfaces and FPGA
take the programmable metasurface paradigm to the THz
models. Therefore, considering cost, difficulty, convenience,
band and beyond. However, this poses important chal-
etc., numerous proof-of-concept systems have been devel-
lenges in the technological front, as discussed in Sec. V.
oped using the FPGA-based architecture [14], [21], [28],
III. A RCHITECTURES [57]. Desired attributes such as space-time multi-functionality,
learnability, and interconnectivity are also possible. However,
The programmability in metasurfaces is enabled by the like the other side of a coin, such rough and non-customized
addition of both tuning elements at the unit cells and appro- architecture is usually bulky and suffers from significant
priate control circuits across the metasurface. The control power consumption – hence turning perpetual operation into a
sub-system, in essence, may implement a control loop that complex possibility and, in overall, limiting the applicability
is able to modify the response of the metasurface with a and integrability of such a platform in real-world scenarios.
given maximum rate, e.g. a few kHz. Such a speed will be Also, the FPGA is a potential single point of failure, rendering
determined by the characteristics of the control sub-system the metasurface useless if the FPGA breaks down.
and by how it communicates with the metasurface sub-system,
always trying to meet the reconfiguration requirements of the
final application. B. Integrated Architectures
There are different ways to combine and interconnect the Opposed to the FPGA-based approach, researchers and
metasurface and control sub-systems (also referred to as engineers have started to propose integrated architecture-based
ABADAL et al.: PROGRAMMABLE METAMATERIALS FOR SOFTWARE-DEFINED ELECTROMAGNETIC CONTROL 13

systems [24], [29], [32], [37]. This paradigm, represented both costly and impractical in terms of the space required
in Figure 6(b) and sometimes referred to the HyperSurface on the auxiliary Printed Circuit Board (PCB). This leads
(HSF), indeed hinges on the integration of a network of to the necessity to take advantage of integrated circuit (IC)
communicating chips within the metasurface containing tuning technology, whereby a single integrated circuit can contain
elements, control circuits, and even sensors. Attached to the digitally programmable adaptive impedance loading elements
metasurface, a gateway device with wireless connectivity can and networking communication circuitry. This is especially
interface the EM device with the real-world in a cyber- important if the system is to be scalable so as to cover an
physical/IoT manner. Essentially, the metasurface receives arbitrarily large surface.
reconfiguration requests that are disseminated to the internal The requirements for a practical programmable metasur-
control logic via chip-to-chip interconnects and routing logic face are quite challenging. In the most basic of forms a
[86], [87]. Controllers may consist of simple circuitry that programmable metasurface requires a 3-layer PCB where the
reads the unit cell state and Digital/Analog (D/A) converters top layer is used for a patterned conductive array, the middle
that set the appropriate voltage to the tuning elements, or more layer is a ground plane, used to decouple the EM effects of
complex circuits implementing the desired intelligence. Addi- where the metasurface will be installed, and finally the bottom
tionally, embedded sensors can pick up data from the environ- layer is where the active control components will be placed.
ment and send it to the control logic or external devices again Space on the bottom layer is extremely limited, especially
via this communications plane. if independent control of each unit cell is required. Solutions
Integrated architectures are custom-made and therefore are with conventional IC packages are not viable as they are not
much more optimized than FPGA-based architectures. This cost effective, but also the in-package wirebonds tend to intro-
means that the control sub-system is clearly less intrusive in duce significant parasitic inductances and resistances into the
terms of EM interference, less bulky, and with lower potential system that have large random variations. Thus, wafer-level-
power consumption. For instance, movement of control data chip-scale-packaging (WLCSP) is the only practical option.
among controllers can be measured and predicted, opening the The size of these components tend to limit the maximum
door to streamlined interconnect designs [88]. Furthermore, frequency of a system [90], given that, as the frequency goes
the addition of an internal network of controllers instead of a up, the unit cell pitch becomes smaller.
centralized FPGA provides metasurfaces with means to imple- The custom ASIC alleviates the space problem on the
ment the so-called distributed intelligence allowing controllers bottom of the PCB, however this comes with other challenges.
to sense, cooperate, and act locally without the intervention of For instance, on-chip inductors are relatively lossy, as are
an external entity [89]. This avoids central-point bottlenecks other key components such as on-chip varactors and varistors.
and improves resilience by allowing information to be re- A technology must be selected that balances economic cost
routed to reach the intended destinations even if a set of with maximum frequency performance. From the digital con-
connections fail [86]. trol side, there are also challenges with regards to the digital
All these features, however, come at the cost of increased inter-chip communication schemes, given pin constraints and
engineering efforts and cost to design and embed the necessary the difficulty of producing a global clock on a arbitrarily
elements within the metasurface. This is especially true as scalable metasurface that will not be skewed enough to contain
the unit cell density increases with the metasurface frequency, problems. Furthermore, clocking a large surface is very energy
which requires a proportional miniaturization of the control intensive and, in some applications such as EM absorbers, this
circuitry [29]. Further, the integrated approach lacks the flex- is totally unacceptable given that each clock cycle will produce
ibility and convenience of the FPGA-based alternative. More extensive self-EM emissions. Given these challenges, asyn-
details are given in the next section. chronous digital architectures appear to be the best solution,
especially if global unit cell reconfiguration does not occur
frequently. A more detailed discussion on these issues can be
IV. D EVICES AND C IRCUITS found in [90].
From the perspective of devices and circuits, the fully The future is bringing exciting new passive two terminal
integrated architecture [Fig. 6(b)] presents multiple challenges devices, which will enable programmable metasurfaces to
and considerations related to the tuning devices and methods reach a new level, both in terms of spatial resolution as well
[11], [12], as well as the efficient integration of such elements as zero power consumption once configured. The Chua Mem-
and possible interference of control circuits with the operation components, namely Memristor, Memcapacitor and Memin-
of the metasurface itself. We discuss a few of these challenges ductor will enable the reconfiguration of the unit cells through
next. a crossbar array using pulses generated by a device on the
Much has been achieved so far just through the use of the perimeter, and these passive devices will hold their resistance,
“humble” PIN diode in adaptive metasurfaces [14], [21]. The capacitance and inductance without consuming any power,
diode essentially behaves as a capacitor when it is under zero until another charge packet is sent to the two-terminal device.
forward bias voltage and it can behave as a resistive element The concept was first described in 2018 [91], and new devel-
for small signals when forward biased. By carefully combining opments using conductive filament based-memristors show
the diode with a binary DC-voltage driven radial stub and a promise for low power configurations, that can work at very
square patch creates unit cells whose reflection phase can be high frequencies, giving tunable resistance, inductance and
switched between 0 and π. Though a simple but very powerful capacitance in nm-sized passive devices.
technique, this has room for improvement when it comes
to power consumption and fine adaptation; because forward-
V. T ECHNOLOGY
biased diodes consume much current. Intermediate reflection
phases can be achieved by changing the bias voltage, however The concept of programmable metamaterials is relatively
having a discrete D/A converter for each unit cell would be recent and, as such, most of existing realizations are rough
14 IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 10, NO. 1, MARCH 2020

prototypes, proof-of-concept FPGA-based architectures that do The fundamental and limiting plasmon damping pathways
not push the limits in terms of the involved technologies. This in graphene nanostructures are via the graphene intrinsic
is because being mostly in the microwave band, unit cells phonons and scattering from the edges [104]. Much like
are large enough to fit the required tuning elements and the the case of metallic nanoparticles [112], the inhomogeneity
back-side connectors. Moreover, the metasurface is tested in of the ribbon width can lead to plasmon damping due to
several steady-state conditions or for functionalities that do not the atomic edge defects. Although plasmon excitations in
require a high refresh rate, e.g., several tens of Hz is more than most graphene plasmonic devices lie outside of the Landau
enough for single-beam steering [88]. Under these conditions, damping regime, atomic edge defects can provide the required
programmable metasurfaces could be built using affordable momentum transfer leading to plasmon decay into electron-
components and relatively old technologies. hole pairs. The damping rate goes as a/W , where a is of
However, the quest for fully-integrated intelligent multifunc- order of the Fermi velocity in graphene, and W is the ribbon
tional metasurfaces with space-time programmability or sim- width. The intrinsic phonons in graphene and those from the
ply aiming at the mmWave-THz bands for much-anticipated substrate can also strongly influence the plasmon lifetimes
applications is likely to bring technological challenges at when the plasmons are tuned to these phonon frequencies.
different levels. One important challenge is to scale the fabrica- In fact, a plasmon lifetime of 20 femto-seconds and smaller
tion process of programmable metasurfaces to relatively large is observed when damping through the emission of an optical
areas and arbitrary shapes. A possible solution comes from phonon is allowed [104], [113]. The use of hexagonal boron
the architecture perspective: one could resort to modular tile nitride (hBN) as a substrate for graphene has also been
designs that leverage interconnectivity to be clipped together demonstrated to significantly improve the plasmon quality
in the required application size and shape [90]. This way, factor [114]. However, as with any materials development,
widespread fabrication and integration techniques could be quality will continue to improve. Large scale growth of both
employed, reducing costs and enabling scalability. graphene and hBN has progressed in recent years [115], [116],
This and other issues such as the choice of substrate leading to centimeter scale single crystals.
materials [92], the practicality of the tuning elements [11], Another practical technological consideration in 2D materi-
[12], or the space left to integrate controllers and I/O pins at als metasurface design is appropriate photonic engineering to
the unit cell level are greatly dependent on the metasurface boost light-matter interactions. A homogeneous graphene sheet
frequency, whereas the computational requirements of the has a free space optical absorption of less than 1% in the mid-
control logic hinge on the refresh rate imposed by the final infrared regime [117]. Graphene plasmonic resonators, on the
applications. Within this context, we next review the prospects other hand, have resonant absorption of order 10% in the
of promising technologies towards high-speed high-frequency mid-infrared regime [104]. To achieve total light absorption,
programmable metasurfaces. a common approach is to include a back metal reflector
Conventional materials used for tunable/programmable positioned at quarter wavelength underneath graphene [118],
metasurface are conducting oxides (e.g. ITO) [94], [95], phase [119]. However, it is not efficient to use the same metal
change materials [96], liquid crystals [97], and approaches reflector as the back gate for electrostatic doping control
that integrate dielectrics with electro-mechanical systems [15]. of graphene, since the gate capacitance is small. There are
Recent emergence of a class of atomically thin materi- various approaches in the literature in circumventing this issue.
als [98], most notably graphene, transition metal dichalco- One approach involves the use of thin metallic layers placed
genides, among many others are most natural materials plat- at nanometer scale underneath the graphene layer [120] as
form. Their atomic thickness implies that their electronic the gate electrodes. It was experimentally demonstrated that
properties are very amenable to the effect of electric field. total light absorption can be achieved in this 2D materials
For example, the electron surface densities in these materials metasurface design [120].
are highly tunable in a typical back-gated device geome- Fig. 7(a) illustrates a particular design of graphene based
try [99], [100]. These materials also exhibit strong light-matter metasurface which is theoretically shown to dynamically tune
interactions via polaritonic resonances [101], [102], due to the scattered wavefront over a full 2π phase space [93].
their inherent dipole excitations such as plasmons, phonons or The unit cell of the proposed structure is denoted as a
excitons. metamolecule, which consists of a pair of noble metal antennas
In recent years, there has been significant advancement each independently contacting a strip of graphene nanoribbon
in graphene plasmonics [103] for various terahertz and i.e. plasmonic resonator. Electrical biasing between the noble
mid-infrared applications, such as infrared detectors, filters, metal antenna and the global back ground plane would then
modulators, biosensing, to general metasurface applications. enable the independent tuning of the Fermi level (i.e. E F 1
Plasmons in graphene are excited by patterning graphene and E F 2 ) in each of the graphene nanoribbons within the
into nanostructures [8], [104], and their plasmon resonances metamolecule. The pair of Fermi levels thus constitute the
reside in the terahertz to mid-infrared and are also tunable two parameters allowing for the modulation of both ampli-
electrically. General reflection and amplitude phase modu- tude and phase of each metamolecule. Using experimentally
lation of these graphene plasmonic resonators enable very realistic design parameters like conservative graphene carrier
diverse beam manipulation applications [18], [19], [65]. This mobility of 1000 cm2 V-1 s-1 , simulation results show that the
includes general beam forming [105], cloaking [106], tunable two parameters E F 1 and E F 2 allow one to independently
absorption [107], or polarization control [108], among many control the phase and amplitude of the reflected beam of an
others. Experimental demonstration of these various concepts incoming free space 7 µm light, as illustrated in Fig. 7(b)-(e).
are also emerging [109]–[111]. The performance of these Further, experimental efforts to-date have demonstrated key
devices are mainly limited by the graphene or substrate losses, advancements in integrating high-quality graphene into back-
hence, restricting the reflection phases to less than the ideal gated devices [114], [121], integration of ultra-smooth metallic
2π. components utilizing atomic layer deposition achieving strong
ABADAL et al.: PROGRAMMABLE METAMATERIALS FOR SOFTWARE-DEFINED ELECTROMAGNETIC CONTROL 15

Fig. 7. (a) Schematic of the unit cell structure, denoted as metamolecule, composed of a pair of independently controlled graphene plasmonic metaatoms.
(b) Amplitude map and (c) phase map of the reflection as independently adjusting the graphene Fermi levels (E F1 and E F2 ) of the meta-atoms. Reflected
transverse magnetic field components (H y /H0 ) displaying (d) phase modulation with a constant amplitude (|r| = 0.5) along the solid blue line and (e) amplitude
modulation including perfect absorption with a constant phase (ϕ = 0) along the solid red line. See [93] for more details.

light-matter interaction [120], independent meta-element gate already appeared proposing to use Convolutional Neural Net-
tuning [122], among many others. However, a full experimen- works (CNNs) to design broadband unit cells for anisotropic
tal demonstration of complete phase and amplitude modulation coding metamaterials [129], or Generative Adversarial Net-
for efficient beam forming is still in the works. works (GANs) for the inverse design of unit cells with
In summary, graphene, 2D materials, or semiconductor arbitrary spectra [130]–[132]. More recently, interest has been
quantum wells [123], present tremendous new device opportu- shown in using CNNs to automatically find the most suitable
nities in the field of terahertz and mid-infrared metasurfaces. coding pattern for a particular functionality –an approach
However, in order to fully exploit their unique properties or to that has been proven for beam steering [74], but that is
harness other opportunities such as the use of flexible elec- also theoretically applicable to arbitrary functionalities. At the
tronics, it may become necessary to go beyond conventional application level, machine learning has been also proposed
lithography, etching, or deposition processes. The main hurdle for the control and coordination of multiple metasurfaces in
for the manufacturing of THz metasurfaces is the scaling to programmable wireless environments [133] or the optimization
surface areas that make sense for the applications at hand. of metasurface-based imaging in real time [134].
In this direction, the modular approach discussed above is
also applicable here. Moreover, besides conventional microma-
chining methods [9], researchers are pushing towards methods B. Energy Harvesting
such as transfer printing [124] or deterministic nano-assembly When relating energy harvesting with metamaterials,
[125]. In the particular case of graphene, wafer-level man- the main direction has been the use of metamaterials to enable
ufacturing and integration at the graphene qualities required high-efficiency energy harvesting [135]. However, the advent
for metasurface applications remains as an outstanding chal- of programmable metasurfaces with low-power integrated
lenge. Many efforts have been devoted recently in this respect architectures has opened a new perspective. Metasurfaces are
[126]–[128] which could allow, in the near future, to electri- constantly impinged by EM waves and, theoretically, part
cally gate small graphene patches in relatively large wafers, of this energy could be absorbed and reused to drive the
in the pathway to the much-sought local biasing of graphene integrated electronics, as proposed in [29], [136]. This is also
–key to achieve the necessary phase modulation in THz encouraged by some wireless communication applications,
metasurfaces. where users may transmit intermittently in predictable time-
slots as defined by the underlying communication protocols;
thus opening the door to harvesting the rest of the time.
VI. C ROSS -C UTTING I SSUES
It thus becomes apparent that achieving the objective of
The field of programmable metamaterials is thriving and perpetual operation may hinge on efforts at the device level
new research works continue to appear which propose novel (i.e., the energy harvesting device), architecture level (i.e.
cross-cutting approaches with disruptive potential in this which energy source to harvest and where/how to place the
emerging field. Next, we briefly describe a few of such aspects harvesting devices to that end), and application level (i.e. when
and attempt to anticipate its significance. to opportunistically harvest).

A. Neural Networks C. Fault Tolerance and Approximate Computation

Machine learning methods have become all-pervasive in Recent studies suggest that metasurfaces are naturally
engineering and have been recently proposed for RF/photonics resilient to errors in a fraction of the unit cells [82]. In fact,
engineering in general and for the modelling, prediction, it has been observed that non-trivial phase deviations in the
and control of metasurfaces in particular. The key benefit unit cells may be tolerated. This can be exploited in ultra-
of machine learning is that it can simplify tedious design low power designs by operating at low voltages and allowing
processes like those found in programmable metamaterials, the system to make transient faults that may be acceptable
from the unit cell up to the application. First proposals have in certain applications. Leveraging fault tolerance will thus
16 IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 10, NO. 1, MARCH 2020

impact the design of the controller circuits, as well as the [10] W. Cai and V. M. Shalaev, Optical Metamaterials, vol. 10, no. 6011.
complete architecture in ways that maximize the efficiency New York, NY, USA: Springer-Verlag, 2010.
benefit while limiting the performance degradation within [11] G. Oliveri, D. H. Werner, and A. Massa, “Reconfigurable electromag-
netics through Metamaterials—A review,” Proc. IEEE, vol. 103, no. 7,
acceptable bounds. pp. 1034–1056, Jul. 2015.
[12] F. Liu et al., “Programmable metasurfaces: State of the art and
prospects,” in Proc. IEEE Int. Symp. Circuits Syst. (ISCAS), 2018,
D. Internet of Things (IoT) pp. 1–5.
Providing programmable metasurfaces with wireless com- [13] Z. Luo, J. Long, X. Chen, and D. Sievenpiper, “Electrically tunable
munication capabilities enables the cyber-physical/IoT vision metasurface absorber based on dissipating behavior of embedded var-
actors,” Appl. Phys. Lett., vol. 109, no. 7, Aug. 2016, Art. no. 071107.
mentioned in the introduction. As such, programmable meta- [14] H. Yang et al., “A programmable metasurface with dynamic polariza-
materials can be introduced in the control or management tion, scattering and focusing control,” Sci. Rep., vol. 6, no. 1, Oct. 2016,
loop of IoT contexts, enabling unique approaches in a variety Art. no. 35692.
of applications, e.g. operators remotely managing wide-area [15] E. Arbabi, A. Arbabi, S. M. Kamali, Y. Horie, M. Faraji-Dana, and
programmable wireless environments. A. Faraon, “MEMS-tunable dielectric metasurface lens,” Nature Com-
mun., vol. 9, no. 1, 2018, Art. no. 812.
[16] M. Manjappa et al., “Reconfigurable MEMS fano metasurfaces with
VII. C ONCLUSION multiple-input–output states for logic operations at terahertz frequen-
cies,” Nature Commun., vol. 9, no. 1, p. 4056, Oct. 2018.
The field of programmable metamaterials is thriving. Mul- [17] S. Liu, L. Zhang, G. D. Bai, and T. J. Cui, “Flexible controls of broad-
tiple working prototypes have appeared which illustrate the band electromagnetic wavefronts with a mechanically programmable
potential of providing metasurfaces with tuning elements and metamaterial,” Sci. Rep., vol. 9, no. 1, Feb. 2019, Art. no. 1809.
the appropriate control circuitry. Further, a myriad of concep- [18] S. R. Biswas et al., “Tunable graphene metasurface reflectarray for
tual works have shown that, with the introduction of modeling cloaking, illusion, and focusing,” Phys. Rev. Appl., vol. 9, no. 3,
Mar. 2018, Art. no. 034021.
strategies such as the coding metamaterial approach or the use [19] S. E. Hosseininejad et al., “Reprogrammable graphene-based meta-
of neural networks, the programmable metamaterial paradigm surface mirror with adaptive focal point for THz imaging,” Sci. Rep.,
may not only be generalized to virtually any electromagnetic vol. 9, no. 1, Feb. 2019, Art. no. 2868.
functionality, but also become intelligent and self-adaptive. For [20] S. Savo, D. Shrekenhamer, and W. J. Padilla, “Liquid crystal metama-
this to happen beyond the proof-of-concept stage and in a terial absorber spatial light modulator for THz applications,” Adv. Opt.
Mater., vol. 2, no. 3, pp. 275–279, Jan. 2014.
fully-integrated way, however, advances are required at the
[21] T. J. Cui, M. Q. Qi, X. Wan, J. Zhao, and Q. Cheng, “Coding metama-
technology, circuit, and system levels. Significant challenges terials, digital metamaterials and programmable metamaterials,” Light,
will be faced as we push the electromagnetic frequency Sci. Appl., vol. 3, no. 10, p. e218, Oct. 2014.
towards mmWave-THz bands and as we scale the refresh [22] L. Li et al., “Electromagnetic reprogrammable coding-metasurface
rate towards kHz-MHz magnitudes, both necessary for the holograms,” Nature Commun., vol. 8, no. 1, pp. 1–7, Aug. 2017.
realization of compelling applications in the wireless commu- [23] F. Liu et al., “Intelligent metasurfaces with continuously tunable local
surface impedance for multiple reconfigurable functions,” Phys. Rev.
nications or holography fields, among many others. Appl., vol. 11, no. 4, Apr. 2019, Art. no. 044024.
[24] A. Pitilakis et al., “Software-defined metasurface paradigm: Concept,
ACKNOWLEDGMENT challenges, prospects,” in Proc. 12th Int. Congr. Artif. Mater. Novel
Wave Phenomena (Metamaterials), Aug. 2018, pp. 483–485.
The authors would like to thank the members of the [25] T.-J. Cui, S. Liu, and L.-L. Li, “Information entropy of coding meta-
consortia of their respective projects for the fruitful technical surface,” Light, Sci. Appl., vol. 5, no. 11, p. e16172, Jun. 2016.
discussions on the matter. [26] S. Liu et al., “Convolution operations on coding metasurface to reach
flexible and continuous controls of terahertz beams,” Adv. Sci., vol. 3,
no. 10, Jul. 2016, Art. no. 1600156.
R EFERENCES [27] R. Y. Wu, C. B. Shi, S. Liu, W. Wu, and T. J. Cui, “Addition theorem
for digital coding metamaterials,” Adv. Opt. Mater., vol. 6, no. 5,
[1] J. B. Pendry, “Controlling electromagnetic fields,” Science, vol. 312, Jan. 2018, Art. no. 1701236.
no. 5781, pp. 1780–1782, Jun. 2006. [28] L. Zhang et al., “Space-time-coding digital metasurfaces,” Nature
[2] S. B. Glybovski, S. A. Tretyakov, P. A. Belov, Y. S. Kivshar, and Commun., vol. 9, no. 1, p. 4334, 2018.
C. R. Simovski, “Metasurfaces: From microwaves to visible,” Phys. [29] S. Abadal et al., “Computing and communications for the software-
Rep., vol. 634, pp. 1–72, May 2016. defined metamaterial paradigm: A context analysis,” IEEE Access,
[3] A. Li, S. Singh, and D. Sievenpiper, “Metasurfaces and their applica- vol. 5, pp. 6225–6235, 2017.
tions,” Nanophotonics, vol. 7, no. 6, pp. 989–1011, Jun. 2018.
[30] Q. Ma, G. D. Bai, H. B. Jing, C. Yang, L. Li, and T. J. Cui, “Smart
[4] S. Vellucci, A. Monti, M. Barbuto, A. Toscano, and F. Bilotti, “Satel-
metasurface with self-adaptively reprogrammable functions,” Light, Sci.
lite applications of electromagnetic cloaking,” IEEE Trans. Antennas
Appl., vol. 8, no. 1, pp. 1–12, Oct. 2019.
Propag., vol. 65, no. 9, pp. 4931–4934, Sep. 2017.
[5] N. I. Landy, S. Sajuyigbe, J. J. Mock, D. R. Smith, and W. J. Padilla, [31] S. Hu, F. Rusek, and O. Edfors, “Beyond massive MIMO: The poten-
“Perfect metamaterial absorber,” Phys. Rev. Lett., vol. 100, no. 20, tial of data transmission with large intelligent surfaces,” IEEE Trans.
May 2008, Art. no. 207402. Signal Process., vol. 66, no. 10, pp. 2746–2758, May 2018.
[6] M. Gokkavas et al., “Experimental demonstration of a left-handed [32] C. Liaskos, S. Nie, A. Tsioliaridou, A. Pitsillides, S. Ioannidis, and
metamaterial operating at 100 GHz,” Phys. Rev. B, Condens. Matter, I. Akyildiz, “A new wireless communication paradigm through
vol. 73, no. 19, May 2006, Art. no. 193103. software-controlled metasurfaces,” IEEE Commun. Mag., vol. 56, no. 9,
[7] Z. Zhao et al., “Experimental demonstration of 16-Gbit/s millimeter- pp. 162–169, Sep. 2018.
wave communications link using thin metamaterial plates to generate [33] O. Abari, D. Bharadia, A. Duffield, and D. Katabi, “Cutting the cord
data-carrying orbital-angular-momentum beams,” in Proc. IEEE Int. in virtual reality,” in Proc. 15th ACM Workshop Hot Topics Netw.
Conf. Commun. (ICC), Jun. 2015, pp. 1392–1397. (HotNets), 2016, pp. 162–168.
[8] L. Ju et al., “Graphene plasmonics for tunable terahertz metamaterials,” [34] I. F. Akyildiz, J. M. Jornet, and C. Han, “Terahertz band: Next frontier
Nature Nanotechnol., vol. 6, no. 10, pp. 630–634, 2011. for wireless communications,” Phys. Commun., vol. 12, pp. 16–32,
[9] S.-W. Qu, H. Yi, B. J. Chen, K. B. Ng, and C. H. Chan, “Terahertz Sep. 2014.
reflecting and transmitting metasurfaces,” Proc. IEEE, vol. 105, no. 6, [35] D. M. Mittleman, “Twenty years of terahertz imaging [Invited],” Opt.
pp. 1166–1184, Jun. 2017. Express, vol. 26, no. 8, pp. 9417–9431, Apr. 2018.
ABADAL et al.: PROGRAMMABLE METAMATERIALS FOR SOFTWARE-DEFINED ELECTROMAGNETIC CONTROL 17

[36] S. Liu and T. J. Cui, “Concepts, working principles, and applications [60] T. J. Cui, R. Y. Wu, W. Wu, C. B. Shi, and Y. B. Li, “Large-scale
of coding and programmable metamaterials,” Adv. Opt. Mater., vol. 5, transmission-type multifunctional anisotropic coding metasurfaces in
no. 22, 2017, Art. no. 1700624. millimeter-wave frequencies,” J. Phys. D, Appl. Phys., vol. 50, no. 40,
[37] C. Liaskos, S. Nie, A. Tsioliaridou, A. Pitsillides, S. Ioannidis, and Sep. 2017, Art. no. 404002.
I. Akyildiz, “Realizing wireless communication through software- [61] S. Liu et al., “Anisotropic coding metamaterials and their powerful
defined HyperSurface environments,” in Proc. IEEE 19th Int. Symp. manipulation of differently polarized terahertz waves,” Light, Sci. Appl.,
World Wireless, Mobile Multimedia Networks (WoWMoM), Jun. 2018, vol. 5, no. 5, p. e16076, Jan. 2016.
pp. 14–15. [62] G. D. Bai et al., “Multitasking shared aperture enabled with multiband
[38] J. Zhao et al., “Programmable time-domain digital-coding metasurface digital coding metasurface,” Adv. Opt. Mater., vol. 6, no. 21, Aug. 2018,
for non-linear harmonic manipulation and new wireless communication Art. no. 1800657.
systems,” Nat. Sci. Rev., vol. 6, no. 2, pp. 231–238, Nov. 2018. [63] J. Luo et al., “2-bit amplitude-modulated coding metasurfaces based
[39] C. Liaskos, A. Tsioliaridou, A. Pitsillides, S. Ioannidis, and on indium tin oxide films,” J. Appl. Phys., vol. 126, no. 11, Sep. 2019,
I. Akyildiz, “Using any surface to realize a new paradigm for Art. no. 113102.
wireless communications,” Commun. ACM, vol. 61, no. 11, pp. 30–33, [64] N. Yu et al., “Light propagation with phase discontinuities: General-
Oct. 2018. ized laws of reflection and refraction,” Science, vol. 334, no. 6054,
[40] Y. Han, W. Tang, S. Jin, C.-K. Wen, and X. Ma, “Large intelligent pp. 333–337, Sep. 2011.
surface-assisted wireless communication exploiting statistical CSI,” [65] S. E. Hosseininejad, K. Rouhi, M. Neshat, A. Cabellos-Aparicio,
IEEE Trans. Veh. Technol., vol. 68, no. 8, pp. 8238–8242, Aug. 2019. S. Abadal, and E. Alarcon, “Digital metasurface based on graphene:
[41] M. D. Renzo et al., “Smart radio environments empowered by recon- An application to beam steering in terahertz plasmonic antennas,” IEEE
figurable AI meta-surfaces: An idea whose time has come,” EURASIP Trans. Nanotechnol., vol. 18, pp. 734–746, 2019.
J. Wireless Commun. Netw., vol. 2019, no. 1, p. 129, May 2019. [66] K. Wang, J. Zhao, Q. Cheng, D. S. Dong, and T. J. Cui, “Broadband
[42] I. F. Akyildiz, C. Han, and S. Nie, “Combating the distance problem and broad-angle low-scattering metasurface based on hybrid optimiza-
in the millimeter wave and terahertz frequency bands,” IEEE Commun. tion algorithm,” Sci. Rep., vol. 4, no. 1, Aug. 2014, Art. no. 5935.
Mag., vol. 56, no. 6, pp. 102–108, Jun. 2018. [67] L. R. Ji-Di, X. Y. Cao, Y. Tang, S. M. Wang, Y. Zhao, and X. W. Zhu,
[43] K. Ntontin et al., “Reconfigurable intelligent surfaces vs. Relay- “A new coding metasurface for wideband RCS reduction,” Radioengi-
ing: Differences, similarities, and performance comparison,” 2019, neering, vol. 27, no. 2, pp. 394–401, Jun. 2018.
arXiv:1908.08747. [Online]. Available: http://arxiv.org/abs/1908.08747 [68] H. Rajabalipanah, A. Abdolali, J. Shabanpour, A. Momeni, and
[44] J. Y. Dai et al., “Wireless communications through a simplified A. Cheldavi, “Asymmetric spatial power dividers using phase–
architecture based on time-domain digital coding metasurface,” Adv. amplitude metasurfaces driven by Huygens principle,” ACS Omega,
Mater. Technol., vol. 4, no. 7, Feb. 2019, Art. no. 1900044. vol. 4, no. 10, pp. 14340–14352, Aug. 2019.
[45] A. Babakhani, D. B. Rutledge, and A. Hajimiri, “Transmitter architec- [69] H. Yi, S.-W. Qu, B.-J. Chen, X. Bai, K. B. Ng, and C. H. Chan, “Flat
tures based on near-field direct antenna modulation,” IEEE J. Solid- terahertz reflective focusing metasurface with scanning ability,” Sci.
State Circuits, vol. 43, no. 12, pp. 2674–2692, Dec. 2008. Rep., vol. 7, no. 1, pp. 2–9, Jun. 2017.
[46] N. Ashraf, T. Saeed, R. I. Ansari, M. Lestas, C. Liaskos, and [70] S. Liu et al., “Full-state controls of terahertz waves using tensor
A. Pitsillides, “Feedback based beam steering for intelligent metasur- coding metasurfaces,” ACS Appl. Mater. Interfaces, vol. 9, no. 25,
faces,” in Proc. 2nd IEEE Middle East North Afr. Commun. Conf. pp. 21503–21514, Jun. 2017.
(MENACOMM), Nov. 2019, pp. 1–6. [71] S. Sui et al., “Absorptive coding metasurface for further radar cross
section reduction,” J. Phys. D, Appl. Phys., vol. 51, no. 6, Jan. 2018,
[47] L. Zhang et al., “Transmission-Reflection-Integrated multifunctional
Art. no. 065603.
coding metasurface for full-space controls of electromagnetic waves,”
Adv. Funct. Mater., vol. 28, no. 33, Jun. 2018, Art. no. 1802205. [72] Y. Zhou et al., “Design of phase gradient coding metasurfaces for
broadband wave modulating,” Sci. Rep., vol. 8, no. 1, Jun. 2018,
[48] R. Y. Wu et al., “Digital metasurface with phase code and reflection–
Art. no. 8672.
transmission amplitude code for flexible full-space electromagnetic
[73] K. Rouhi, H. Rajabalipanah, and A. Abdolali, “Multi-bit graphene-
manipulations,” Adv. Opt. Mater., vol. 7, no. 8, 2019, Art. no. 1801429.
based bias-encoded metasurfaces for real-time terahertz wavefront
[49] X. Wan et al., “Multichannel direct transmissions of near-field infor- shaping: From controllable orbital angular momentum generation
mation,” Light, Sci. Appl., vol. 8, no. 1, pp. 1–8, Jul. 2019. toward arbitrary beam tailoring,” Carbon, vol. 149, pp. 125–138,
[50] K. Chen et al., “A reconfigurable active Huygens’ metalens,” Adv. Aug. 2019.
Mater., vol. 29, no. 17, Feb. 2017, Art. no. 1606422. [74] T. Shan and M. Li, “Coding programmable metasurfaces based on deep
[51] L. Liang et al., “Anomalous terahertz reflection and scattering by learning techniques,” in Proc. AP-S/URSI, 2019.
flexible and conformal coding metamaterials,” Adv. Opt. Mater., vol. 3, [75] Q. Wang et al., “Millimeter-wave digital coding metasurfaces based on
no. 10, p. 1311, Jul. 2015. nematic liquid crystals,” Adv. Theory Simulations, vol. 2, no. 12, 2019,
[52] K. Chen, Y. Feng, L. Cui, J. Zhao, T. Jiang, and B. Zhu, “Dynamic Art. no. 1900141.
control of asymmetric electromagnetic wave transmission by active [76] J. Li et al., “Frequency-switchable VO2 -based coding metasurfaces at
chiral metamaterial,” Sci. Rep., vol. 7, no. 1, Feb. 2017, Art. no. 42802. the terahertz band,” Opt. Commun., vol. 458, 2020, Art. no. 124744.
[53] X. Gao, W. L. Yang, H. F. Ma, Q. Cheng, X. H. Yu, and T. J. Cui, [77] X. Wang, A. Diaz-Rubio, H. Li, S. A. Tretyakov, and A. Alu,
“A reconfigurable broadband polarization converter based on an “Multifunctional space-time metasurfaces,” 2019, arXiv:1910.11812.
active metasurface,” IEEE Trans. Antennas Propag., vol. 66, no. 11, [Online]. Available: http://arxiv.org/abs/1910.11812
pp. 6086–6095, Nov. 2018. [78] A. Shaltout, A. Kildishev, and V. Shalaev, “Time-varying metasurfaces
[54] L. Zhang, S. Liu, L. Li, and T. J. Cui, “Spin-controlled multiple pencil and Lorentz non-reciprocity,” Opt. Mater. Express, vol. 5, no. 11,
beams and vortex beams with different polarizations generated by pp. 2459–2467, Oct. 2015.
pancharatnam-berry coding metasurfaces,” ACS Appl. Mater. Interfaces, [79] D. Ramaccia, D. L. Sounas, A. Alu, A. Toscano, and F. Bilotti,
vol. 9, no. 41, pp. 36447–36455, Oct. 2017. “Phase-induced frequency conversion and Doppler effect with time-
[55] J. Zhao, Q. Cheng, J. Chen, M. Q. Qi, W. X. Jiang, and T. J. Cui, modulated metasurfaces,” IEEE Trans. Antennas Propag., to be pub-
“A tunable metamaterial absorber using varactor diodes,” New J. Phys., lished.
vol. 15, no. 4, Apr. 2013, Art. no. 043049. [80] E. Basar, “Reconfigurable intelligent surface-based index modulation:
[56] D. S. Dong et al., “Terahertz broadband low-reflection metasurface A new beyond MIMO paradigm for 6G,” IEEE Trans. Commun., to be
by controlling phase distributions,” Adv. Opt. Mater., vol. 3, no. 10, published.
pp. 1405–1410, Jun. 2015. [81] V. S. Asadchy et al., “Flat engineered multichannel reflectors,” Phys.
[57] L. Zhang et al., “Breaking reciprocity with space-time-coding digital Rev. X, vol. 7, no. 3, Sep. 2017, Art. no. 031046.
metasurfaces,” Adv. Mater., vol. 31, no. 41, 2019, Art. no. 1904069. [82] H. Taghvaee, S. Abadal, J. Georgiou, A. Cabellos-Aparicio, and
[58] Q. Ma et al., “Controllable and programmable nonreciprocity based on E. Alarcon, “Fault tolerance in programmable metasurfaces: The beam
detachable digital coding metasurface,” Adv. Opt. Mater., vol. 7, no. 24, steering case,” in Proc. IEEE Int. Symp. Circuits Syst. (ISCAS),
Oct. 2019, Art. no. 1901285. May 2019, pp. 1–5.
[59] Y. B. Li et al., “Transmission-type 2-Bit programmable metasurface [83] K. M. Kossifos, M. A. Antoniades, J. Georgiou, A. H. Jaafar, and
for single-sensor and single-frequency microwave imaging,” Sci. Rep., N. T. Kemp, “An optically-programmable absorbing metasurface,” in
vol. 6, no. 1, pp. 1–8, Mar. 2016. Proc. IEEE Int. Symp. Circuits Syst. (ISCAS), May 2018, pp. 1–5.
18 IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 10, NO. 1, MARCH 2020

[84] D. Wang et al., “Multiband switchable terahertz quarter-wave plates via [110] N. Dabidian et al., “Experimental demonstration of phase modulation
phase-change metasurfaces,” IEEE Photon. J., vol. 8, no. 1, Feb. 2016, and motion sensing using graphene-integrated metasurfaces,” Nano
Art. no. 5500308. Lett., vol. 16, no. 6, pp. 3607–3615, May 2016.
[85] K. Sun et al., “VO2 thermochromic metamaterial-based smart optical [111] Z. Miao et al., “Widely tunable terahertz phase modulation with
solar reflector,” ACS Photon., vol. 5, no. 6, pp. 2280–2286, 2018. gate-controlled graphene metasurfaces,” Phys. Rev. X, vol. 5, no. 4,
[86] D. Kouzapas et al., “Towards fault adaptive routing in metasurface con- Nov. 2015, Art. no. 041027.
troller networks,” J. Syst. Archit., vol. 106, Jun. 2020, Art. no. 101703. [112] U. Kreibig and M. Vollmer, Optical Properties of Metal Clusters,
[87] A. C. Tasolamprou et al., “Exploration of intercell wireless millimeter- vol. 25. Berlin, Germany: Springer-Verlag, 2013.
wave communication in the landscape of intelligent metasurfaces,” [113] G. X. Ni et al., “Fundamental limits to graphene plasmonics,” Nature,
IEEE Access, vol. 7, pp. 122931–122948, 2019. vol. 557, no. 7706, pp. 530–533, May 2018.
[88] T. Saeed et al., “Workload characterization of programmable meta- [114] A. Woessner et al., “Highly confined low-loss plasmons in graphene–
surfaces,” in Proc. the 6th Annu. ACM Int. Conf. Nanosc. Comput. boron nitride heterostructures,” Nature Mater., vol. 14, no. 4,
Commun. NANOCOM, 2019, pp. 1–6. pp. 421–425, Dec. 2014.
[89] C. Liaskos et al., “ABSense: Sensing electromagnetic waves on meta- [115] Y. Hao et al., “The role of surface oxygen in the growth of large single-
surfaces via ambient compilation of full absorption,” in Proc. 6th Annu. crystal graphene on copper,” Science, vol. 342, no. 6159, pp. 720–723,
ACM Int. Conf. Nanosc. Comput. Commun. NANOCOM, 2019, pp. 1–6. Oct. 2013.
[90] L. Petrou, P. Karousios, and J. Georgiou, “Asynchronous circuits as an [116] L. Wang et al., “Epitaxial growth of a 100-square-centimetre single-
enabler of scalable and programmable metasurfaces,” in Proc. IEEE crystal hexagonal boron nitride monolayer on copper,” Nature, vol. 570,
Int. Symp. Circuits Syst. (ISCAS), May 2018. no. 7759, pp. 91–95, May 2019.
[91] J. Georgiou, K. M. Kossifos, M. A. Antoniades, A. H. Jaafar, and [117] Z. Q. Li et al., “Dirac charge dynamics in graphene by infrared
N. T. Kemp, “Chua mem-components for adaptive RF metamaterials,” spectroscopy,” Nature Phys., vol. 4, no. 7, pp. 532–535, Jun. 2008.
in Proc. IEEE Int. Symp. Circuits Syst. (ISCAS), May 2018, pp. 1–5. [118] E. Carrasco, M. Tamagnone, J. R. Mosig, T. Low, and
[92] D. Manessis et al., “High frequency substrate technologies for the J. Perruisseau-Carrier, “Gate-controlled mid-infrared light bending
realisation of software programmable metasurfaces on PCB hardware with aperiodic graphene nanoribbons array,” Nanotechnology, vol. 26,
platforms with integrated controller nodes,” in Proc. 22nd Eur. Micro- no. 13, Mar. 2015, Art. no. 134002.
electron. Packag. Conf. Exhibit. (EMPC), Sep. 2019, pp. 1–7. [119] M. S. Jang et al., “Tunable large resonant absorption in a midinfrared
[93] S. Han, S. Kim, S. Kim, T. Low, V. W. Brar, and M. S. Jang, “Com- graphene Salisbury screen,” Phys. Rev. B, vol. 90, no. 16, Oct. 2014,
plete complex amplitude modulation with electronically tunable Art. no. 165409.
graphene plasmonic metamolecules,” ACS Nano, vol. 14, no. 1, [120] I.-H. Lee, D. Yoo, P. Avouris, T. Low, and S.-H. Oh, “Graphene
pp. 1166–1175, Jan. 2020. acoustic plasmon resonator for ultrasensitive infrared spectroscopy,”
[94] Y.-W. Huang et al., “Gate-tunable conducting oxide metasurfaces,” Nature Nanotechnol., vol. 14, no. 4, pp. 313–319, Feb. 2019.
Nano Lett., vol. 16, no. 9, pp. 5319–5325, Sep. 2016. [121] S. S. Sunku et al., “Photonic crystals for nano-light in moiré graphene
[95] G. Kafaie Shirmanesh, R. Sokhoyan, P. Chieh Wu, and H. A. Atwater, superlattices,” Science, vol. 362, no. 6419, pp. 1153–1156, Dec. 2018.
“Electro-optically tunable universal metasurfaces,” 2019, [122] M. Tamagnone et al., “Graphene reflectarray metasurface for tera-
arXiv:1910.02069. [Online]. Available: http://arxiv.org/abs/1910. hertz beam steering and phase modulation,” 2018, arXiv:1806.02202.
02069 [Online]. Available: http://arxiv.org/abs/1806.02202
[96] K. Chaudhary et al., “Polariton nanophotonics using phase-change [123] P. C. Wu et al., “Dynamic beam steering with all-dielectric electro-optic
materials,” Nature Commun., vol. 10, no. 1, pp. 1–6, Oct. 2019. III–V multiple-quantum-well metasurfaces,” Nature Commun., vol. 10,
[97] J. Lee et al., “Ultrafast electrically tunable polaritonic metasurfaces,” no. 1, p. 3654, Aug. 2019.
Adv. Opt. Mater., vol. 2, no. 11, pp. 1057–1063, Jul. 2014. [124] S. Lee et al., “Heterogeneously assembled metamaterials and metade-
[98] P. Avouris, T. F. Heinz, and T. Low, 2D Materials. Cambridge, U.K.: vices via 3D modular transfer printing,” Sci. Rep., vol. 6, no. 1,
Cambridge Univ. Press, 2017. Jun. 2016, Art. no. 27621.
[99] K. S. Novoselov, “Electric field effect in atomically thin carbon films,” [125] B. Kim et al., “Deterministic nanoassembly of quasi-three-dimensional
Science, vol. 306, no. 5696, pp. 666–669, Oct. 2004. plasmonic nanoarrays with arbitrary substrate materials and structures,”
[100] Q. H. Wang, K. Kalantar-Zadeh, A. Kis, J. N. Coleman, and Nano Lett., vol. 19, no. 8, pp. 5796–5805, Jul. 2019.
M. S. Strano, “Electronics and optoelectronics of two-dimensional [126] Y. Lee et al., “Wafer-scale synthesis and transfer of graphene films,”
transition metal dichalcogenides,” Nature Nanotechnol., vol. 7, no. 11, Nano Lett., vol. 10, no. 2, pp. 490–493, Feb. 2010.
pp. 699–712, Nov. 2012. [127] J. S. Lee et al., “Wafer-scale single-crystal hexagonal boron nitride
[101] T. Low et al., “Polaritons in layered two-dimensional materials,” Nature film via self-collimated grain formation,” Science, vol. 362, no. 6416,
Mater., vol. 16, no. 2, pp. 182–194, Nov. 2016. pp. 817–821, Nov. 2018.
[102] D. N. Basov, M. M. Fogler, and F. J. Garcia de Abajo, “Polaritons in [128] D. Neumaier, S. Pindl, and M. C. Lemme, “Integrating graphene
van der Waals materials,” Science, vol. 354, no. 6309, p. aag1992, into semiconductor fabrication lines,” Nature Mater., vol. 18, no. 6,
Oct. 2016. pp. 525–529, May 2019.
[103] T. Low and P. Avouris, “Graphene plasmonics for terahertz to mid- [129] Q. Zhang et al., “Machine-learning designs of anisotropic digital
infrared applications,” ACS Nano, vol. 8, no. 2, pp. 1086–1101, coding metasurfaces,” Adv. Theory Simul., vol. 2, no. 2, Nov. 2018,
Jan. 2014. Art. no. 1800132.
[104] H. Yan et al., “Damping pathways of mid-infrared plasmons in [130] Z. Liu, D. Zhu, S. P. Rodrigues, K.-T. Lee, and W. Cai, “Generative
graphene nanostructures,” Nature Photon., vol. 7, no. 5, pp. 394–399, model for the inverse design of metasurfaces,” Nano Lett., vol. 18,
Apr. 2013. no. 10, pp. 6570–6576, Sep. 2018.
[105] A. Fallahi and J. Perruisseau-Carrier, “Design of tunable biperiodic [131] J. Jiang and J. A. Fan, “Simulator-based training of generative neural
graphene metasurfaces,” Phys. Rev. B, Condens. Matter, vol. 86, no. 19, networks for the inverse design of metasurfaces,” Nanophotonics,
Nov. 2012, Art. no. 195408. Nov. 2019. [Online]. Available: https://www.degruyter.com/view/j/
[106] P.-Y. Chen, J. Soric, Y. R. Padooru, H. M. Bernety, A. B. Yakovlev, nanoph.ahead-of-print/nanoph-2019-0330/nanoph-2019-0330.xml?
and A. Alù, “Nanostructured graphene metasurface for tunable ter- lang=en, doi: 10.1515/nanoph-2019-0330.
ahertz cloaking,” New J. Phys., vol. 15, no. 12, Dec. 2013, [132] W. Ma, F. Cheng, Y. Xu, Q. Wen, and Y. Liu, “Probabilistic represen-
Art. no. 123029. tation and inverse design of metamaterials based on a deep generative
[107] X. Wang and S. A. Tretyakov, “Toward ultimate control of terahertz model with semi-supervised learning strategy,” Adv. Mater., vol. 31,
wave absorption in graphene,” IEEE Trans. Antennas Propag., vol. 67, no. 35, Jul. 2019, Art. no. 1901111.
no. 4, pp. 2452–2461, Apr. 2019. [133] C. Liaskos, A. Tsioliaridou, S. Nie, A. Pitsillides, S. Ioannidis, and
[108] T. Guo and C. Argyropoulos, “Broadband polarizers based on graphene I. Akyildiz, “An interpretable neural network for configuring pro-
metasurfaces,” Opt. Lett., vol. 41, no. 23, pp. 5592–5595, Nov. 2016. grammable wireless environments,” in Proc. IEEE 20th Int. Work-
[109] M. C. Sherrott et al., “Experimental demonstration of >230 shop Signal Process. Adv. Wireless Commun. (SPAWC), Jul. 2019,
phase modulation in gate-tunable graphene–gold reconfigurable mid- pp. 1–5.
infrared metasurfaces,” Nano Lett., vol. 17, no. 5, pp. 3027–3034, [134] L. Li et al., “Machine-learning reprogrammable metasurface imager,”
Apr. 2017. Nature Commun., vol. 10, no. 1, pp. 1–8, Mar. 2019.
ABADAL et al.: PROGRAMMABLE METAMATERIALS FOR SOFTWARE-DEFINED ELECTROMAGNETIC CONTROL 19

[135] Z. Chen, B. Guo, Y. Yang, and C. Cheng, “Metamaterials-based He is an Active Reviewer for Science, Nature Materials, Nature Photonics,
enhanced energy harvesting: A review,” Phys. B, Condens. Matter, Nature Physics, Nature Communications, Physical Review Letters, Advanced
vol. 438, pp. 1–8, Apr. 2014. Materials, and a series of the IEEE T RANSACTIONS. He was an Associate
[136] F. Lemic, R. U. Akbar, J. Marquez-Barja, and J. Famaey, “Assessing Editor of the IEEE T RANSACTIONS ON G EOSCIENCE AND R EMOTE S ENS -
the reliability of energy harvesting terahertz nanonetworks for con- ING and a Guest Editor of Science China – Information Sciences. He served as
trolling software-defined metamaterials,” in Proc. 6th Annu. ACM Int. on the editorial staff for the IEEE Antennas and Propagation Magazine, and
Conf. Nanosc. Comput. Commun. (NANOCOM), 2019, pp. 1–6. is on the Editorial Boards of Progress in Electromagnetics Research (PIER)
and Journal of Electromagnetic Waves and Applications.

Sergi Abadal (Member, IEEE) received the B.Sc.

degree in telecommunication engineering from the
Telecommunication Engineering School (Telecom Tony Low received the B.Sc. and Ph.D. degrees
BCN), Universitat Politècnica de Catalunya (UPC), in electrical and computer engineering from the
Barcelona, Spain, in 2010, the M.Sc. degree in infor- National University of Singapore in 2002 and 2008,
mation and communication technologies (MINT) respectively. After that, he was a Post-Doctoral
from UPC in 2011, and the Ph.D. degree in com- Associate with Purdue University until 2011, and
puter architecture from the Department of Com- then, he worked as a Theorist with various experi-
puter Architecture, UPC, in July 2016. He has held mental groups at Columbia University, Yale Univer-
several visiting positions at Georgia Tech in 2009, sity, and IBM Thomas J. Watson Research. While at
the University of Illinois at Urbana–Champaign from IBM, from 2011 to 2014, he served as an industry
2015 to 2016, and the Foundation of Research and Technology – Hellas liaison to various universities under the Nanoelec-
in 2018. He is currently a Coordinator of the WIPLASH FET-OPEN Project. tronics Research Initiative with the goal of finding
His current research interests are in the areas of chip-scale wireless commu- the next electronics switch. He is currently an Associate Professor at the
nications, including channel modeling and protocol design, the application Department of Electrical and Computer Engineering, University of Minnesota.
of these techniques for the creation of next-generation wireless networks for He was also a member of the technical program committees of over ten con-
on-chip or within metamaterials, and their implications at the system design ferences, given more than 100 invited talks, and published over 100 articles.
level. He has served as a TPC member for more than ten conferences and He has been included in the 2019 Global List of Highly Cited Researchers,
has organized three special sessions, and published over 60 articles in top-tier compiled by the Web of Science Group. He received the Presidential McK-
journals and conferences. He was a recipient of the Nano Communication night Fellowship Award in 2019, the IBM Pat Goldberg Memorial Best Paper
Networks Young Researcher Award in 2019, the INTEL Doctoral Fellowship Award in 2014, the IBM Invention Award in 2013, the KITP Rice Family
in 2013, the Accenture Award for M.Sc. students in 2012, and the Vodafone Fund Fellowship in 2012, the Singapore Millennium Fellowship in 2007, and
Fellowship for B.Sc. students in 2010. He is also an Area Editor of the the IEEE Electron Device Society Fellowship in 2005, among other honors.
Nano Communication Networks (Elsevier), where he was named “Editor of He leads a research group in ECE that works on expanding the understanding
the Year” in 2019. and design of nanomaterials and nanodevices. In recent years, they have
focused on the class of two-dimensional crystals and their heterostructures,
and topological and magnetic materials. His team has revealed their basic
Tie-Jun Cui (Fellow, IEEE) received the B.Sc.,
electronic and optical properties, and their opportunities for novel electronics,
M.Sc., and Ph.D. degrees in electrical engineering
spintronics, optoelectronics, nanophotonics, and plasmonics. He has served
from Xidian University, Xi’an, China, in 1987, 1990,
as a Guest Editor for Nanotechnology (IOP) and an Associate Editor for
and 1993, respectively.
the Nanophotonics journal (De Gruyter) and Journal of Physics: Condensed
In March 1993, he joined the Department of Elec-
Matter (IOP), as well as the Co-Chair of the “2D Materials Nanophotonics”
tromagnetic Engineering, Xidian University, where
in 2017, 2019, and 2020.
he was promoted to an Associate Professor in
November 1993. From 1995 to 1997, he was a
Research Fellow with the Institut für Hochfrequen-
ztechnik und Elektronik (IHE), University of Karl-
sruhe, Germany. In July 1997, he joined the Center Julius Georgiou (Senior Member, IEEE) received
for Computational Electromagnetics, Department of Electrical and Com- the M.Eng. degree in electrical and electronic engi-
puter Engineering, University of Illinois at Urbana–Champaign, first as a neering and the Ph.D. degree from Imperial College
Post-Doctoral Research Associate and then a Research Scientist. In Sep- London in 1998 and 2003, respectively. He worked
tember 2001, he was a Cheung-Kong Professor with the Department of as the Head of Micropower Design at Toumaz Tech-
Radio Engineering, Southeast University, Nanjing, China, where he became nology, a technology-based start-up company, for
a Chief Professor in January 2018. He is the first author of the books two years. In 2004, he joined Johns Hopkins Uni-
Metamaterials – Theory, Design, and Applications (Springer, November 2009) versity as a Post-Doctoral Fellow, before becoming
and Metamaterials: Beyond Crystals, Noncrystals, and Quasicrystals (CRC a Faculty Member with the University of Cyprus
Press, March 2016). He has published over 400 peer-review journal articles in 2005. He is currently an Associate Professor at the
in Science, Proceedings of the National Academy of Sciences (PNAS), Nature University of Cyprus. His research interests include
Communications, Physical Review Letters, Advanced Materials, and the IEEE low-power analog and digital ASICs, implantable biomedical devices, bioin-
T RANSACTIONS, which have been cited more than 20 000 times (H-Factor spired electronic systems, electronics for space, brain–computer interfaces
73). According to Elsevier, he is one of the Most Cited Chinese Researchers. (BCIs), memristive devices, electronics for adaptive metamaterials, inertial
Dr. Cui received the Research Fellowship from the Alexander von Humboldt and optical sensors, and related systems. He is a member of the IEEE
Foundation, Bonn, Germany, in 1995, the Young Scientist Award from the Circuits and Systems Society and the IEEE Circuits and Systems Society
International Union of Radio Science (URSI) in 1999, the Cheung Kong Analog Signal Processing Technical Committee. He was a recipient of the
Professorship under the Cheung Kong Scholar Program by the Ministry of Best Paper Award at the IEEE ISCAS 2011 International Symposium and at
Education, China, in 2001, the National Science Foundation of China for Dis- the IEEE BioDevices 2008 Conference. In 2016, he received the 2015 ONE
tinguished Young Scholars in 2002, the Special Government Allowance from Award from the President of the Republic of Cyprus for his research accom-
the Department of State, China, in 2008, the Award of Science and Technology plishments related to the development of an “Infrared Fluorescence-Based
Progress from the Shaanxi Province Government in 2009, and the May 1st Cancer Screening Capsule for the Small Intestine.” He served as the General
Labour Medal from Jiangsu Province Government in 2010. He also received Chair for the 2010 IEEE Biomedical Circuits and Systems Conference, and is
the First Prize of Natural Science Awards from Ministry of Education, China, currently the Action Chair of the EU COST Action ICT-1401 on “Memristors-
in 2011, the Second Prize of National Natural Science Awards, China, in 2014, Devices, Models, Circuits, Systems and Applications - MemoCIS.” He is
and the First Prize of Military Science and Technology Progress Awards also the Chair of the IEEE Biomedical and Life Science Circuits and
in 2016. His researches have been selected as one of the “Optics in 2016” Systems (BioCAS) Technical Committee. He has been selected as an IEEE
by Optics and Photonics News Magazine (OSA), “10 Breakthroughs of China Circuits and Systems Society Distinguished Lecturer from 2016 to 2017.
Optics in 2016,” “10 Breakthroughs of China Science in 2010,” “Best of 2010” He is also an Associate Editor of the IEEE T RANSACTIONS ON B IOMEDICAL
in New Journal of Physics, Research Highlights in Europhysics News, Journal C IRCUITS AND S YSTEMS and the Frontiers in Neuromorphic Engineering
of Physics D: Applied Physics, Applied Physics Letters, and Nature China. journal.