1

Reconfigurable Intelligent Surfaces and

Metamaterials: The Potential of Wave Propagation
Control for 6G Wireless Communications
George C. Alexandropoulos, Senior Member, IEEE, Geoffroy Lerosey,
Mérouane Debbah, Fellow, IEEE, and Mathias Fink
arXiv:2006.11136v1 [eess.SP] 19 Jun 2020

Abstract—The future 6th Generation (6G) of wireless com- mainly include massive amounts of antennas or other elec-
munication networks will have to meet multiple requirements tromagnetically excited elements, whose implementation cost
(e.g., throughput, latency, positioning accuracy, energy efficiency, and power consumption is extensively improved compared
massive connectivity, reliability, and networking intelligence) in
increasingly demanding levels, either individually or in combina- to conventional massive Multiple Input and Multiple Output
tions in small groups. This trend has spurred recent research ac- (MIMO) systems [2]. Hence, their role in 6G networks has the
tivities on transceiver hardware architectures and novel wireless potential to be prominent via providing improved throughput,
connectivity concepts. Among the emerging wireless hardware ar- coverage extension, security, and positioning with lower power
chitectures belong the Reconfigurable Intelligent Surfaces (RISs), consumption footprint. In achieving these overarching goals,
which are artificial planar structures with integrated electronic
circuits that can be programmed to manipulate an incoming novel approaches for manipulating the wireless propagation
ElectroMagnetic (EM) field in a wide variety of functionalities. signals and revolutionary networking schemes are required. To
Incorporating RISs in wireless networks has been recently advo- this end, there have been recently increased research interests
cated as a revolutionary means to transform any naturally passive (in most of the flagship magazines and conferences of the
wireless communication environment to an active one. This can be ComSoc as well as the Signal Processing and Antennas and
accomplished by deploying cost-effective and easy to coat RISs
to the environments objects (e.g., building facades and indoor Propagation Societies) in wireless connectivity concepts and
walls/ceilings), thus, offering increased environmental intelligence signal processing algorithms incorporating nonconventional
for the scope of diverse wireless networking objectives. In this transceiver and ElectroMagnetic (EM) wave control architec-
paper, we first provide a brief history on wave propagation tures, like:
control for optics and acoustics, and overview two representative
indoor wireless trials at 2.47GHz for spatial EM modulation • Load modulated arrays and electronically steerable an-
with a passive discrete RIS. The first trial dating back to 2014 tenna radiators (e.g., [3]–[5]);
showcases the feasibility of highly accurate spatiotemporal focus- • Hybrid Analog and Digital (A/D) beamforming structures
ing and nulling, while the second very recent one demonstrates
that passive RISs can enrich multipath scattering, thus, enabling
(e.g., [6]–[8]); and
throughput boosted communication links. Motivated by the late • Reconfigurable Intelligent Surfaces (RISs) (e.g., reflec-
research excitement on the RIS potential for intelligent EM tarrays [9], metasurfaces [10]–[12], and antennas made
wave propagation modulation, we describe the status on RIS from metamaterials [13]).
hardware architectures and present key open challenges and
future research directions for RIS design and RIS-empowered The common philosophy of the works falling into this research
6G wireless communications. direction is that: the more antennas deployed per transceiver
Index Terms—Coverage extension, localization, spatiotemporal or the more elements per electromagnetically excited struc-
focusing, time reversal, transceiver beamforming. ture, the more dominating hardware-imposed limitations will
be present to the overall system design. These stricter and
stricter limitations have to be tracked by sophisticated methods
I. I NTRODUCTION of channel modeling and compensated by intelligent sig-
nal/information processing algorithms and wireless connectiv-
The increasingly demanding objectives for 6th Generation
ity techniques, including advanced machine learning tools and
(6G) wireless communications have spurred recent research
dedicated Artificial Intelligence (AI) methods with affordable
activities on novel wireless hardware architectures and con-
computational complexity.
nectivity concepts [1]. The transceiver hardware architectures
Over the last few years, metamaterials have emerged as
G. C. Alexandropoulos is with the Department of Informatics and Telecom- a powerful technology with a broad range of applications,
munications, National and Kapodistrian University of Athens, Panepistimiopo- including wireless communications. Metamaterials comprise a
lis Ilissia, 15784 Athens, Greece. (e-mail: alexandg@di.uoa.gr) class of artificial materials whose physical properties, and par-
G. Lerosey and M. Fink are with Greenerwave, Incubator ESPCI, 75005
Paris, France. M. Fink is also with the Institut Langevin, ESPCI, 75005 Paris, ticularly their permittivity and permeability, can be engineered
France. (e-mails: {geoffroy.lerosey, mathias.fink}@espci.fr) to exhibit various desired characteristics [14]. When deployed
M. Debbah is with both the CentraleSupélec, University Paris-Saclay, 91192 in planar structures (a.k.a. metasurfaces), their effective pa-
Gif-sur-Yvette, France and the Mathematical and Algorithmic Sciences Lab,
Paris Research Center, Huawei Technologies France SASU, 92100 Boulogne- rameters can be tailored to realize a desired transformation on
Billancourt, France. (email: merouane.debbah@huawei.com) the transmitted, received, or impinging EM waves [15]. Such
 2

structures have been lately envisioned as a revolutionary means in [24]–[26] permitting wave focusing from the far field, way
to transform any naturally passive wireless communication below the diffraction limit.
environment (the set of objects between a transmitter and In optics, experiments with Spatial Light Modulators
a receiver constitute the wireless environment) to an active (SLMs) showed the possibility to focus light in media exhibit-
one [16]–[18]. Their extremely low hardware footprint enables ing multiple media (e.g. in [27]); this concept has been used
their cost-effective embedding in various components of the in [28] for imaging applications. SLMs consist of matrices of
wireless propagation environment (e.g., building facades and micro-mirrors or liquid crystal cells, which impose a physical
room walls/ceilings), thus, enabling manmade EM wave prop- phase shift to the portion of light they reflect. A simple
agation control and environmental AI. Due to latter reasons, incoherent energy-based feedback technique together with an
RIS-empowered wireless communications are lately gaining optimization algorithm were proposed in [29] to focus light
booming attention for the upcoming 6G broadband networks. on a single speckle grain (i.e., a random wave field) through
In this paper, we commence in Section II with a brief a thick layer of commercial paint. The idea in that work was
history on wave propagation control, which has been a well- to control the phase and/or amplitude of independent speckle
known concept in optics and acoustics. In Section III, we grains at the input of a multiple scattering medium in order
describe two indoor wireless trials at the WiFi frequency to add them in phase at its output, thus obtaining a focal spot
band for EM wave control with a fabricated passive RIS. whose intensity varies linearly with the number of controlled
The state-of-the-art in RIS hardware architectures for wireless grains.
communications is overviewed in in Section IV. Section V The aforementioned ideas of wave phase/amplitude control
includes key open challenges and future research directions with reflectarrays, metamaterials, and SLMs have recently
for RIS design and RIS-empowered wireless communication. inspired the concept of RISs, as a revolutionary means for
The paper is concluded in Section VI. real-time reconfiguration of EM wave propagation in wireless
communications. We next overview the first passive RIS
(reflectarray) hardware architecture acting as a spatial EM
II. B RIEF H ISTORY OF W WAVE P ROPAGATION C ONTROL
modulator at 2.47GHz [11], and its experimentation results
The control of propagation waves has been of significant for spatiotemporal focusing/nulling and multipath scattering
interest in many domains ranging from medical imaging enrichment in indoor room settings.
and therapy to wireless communications and nanolithography.
Controlling waves in homogeneous media such as air is III. I NDOOR RIS-E MPOWERED W IRELESS T RIALS
relatively easy and has long been realized using lenses in
In this section, we describe the design of an RIS structure
optics [19]. These apply a path difference to every ray going
(reflectarray) with nearly passive discrete elements, as pre-
from one point to another, in order to allow for constructive
sented in the seminal work of [11] dating back to 2014, and
wave interference at a specific 3-dimensional location, thus
discuss its two representative indoor wireless trials at the WiFi
achieving highly accurate spatial signal focusing. Similarly,
frequency band.
in microwave frequencies, reflectarrays have been proposed
in order to steer EM waves to specific spatial directions [9].
In both optics and microwave, the same principle has been A. RIS Design
applied. The waves are reflected from a planar matrix of Each unit cell element in the RIS design of [10] was
resonators of different sizes, which applies a phase shift to designed as a planar resonator intended to reflect the impinging
the incoming wave that depends on the physical dimensions EM waves with a controllable phase shift. In particular, each
of its resonators. Reflectarrays are widely used for satellite resonator was fabricated as a rectangular patch sitting on a
communications and are the ancestors of the concept of ground plane and having two distinct states, resulting in binary
metasurfaces (i.e., planar panels of metamaterials for wave phase modulation. According to these two states, the resonator
control). They are extensively studied nowadays for free space reflects the waves either positively or negatively. The two
applications in microwave, acoustics, and optics [20]. resonator states were realized as follows. Suppose a resonance
In heterogeneous and complex media, the control of wave frequency fref which can be shifted using an electronic circuit.
propagation becomes more complex due to scattering and If fref is set such that it corresponds to the working frequency
diffraction that can turn a plane wave into a completely random f0 , the resonant unit cell reflects the waves at this frequency
wave field. Those media have been considered, up to some with a π phase shift compared to the bare ground plane. When
years ago, as extremely difficult to tackle by wave physicists. It fref is shifted away from f0 , the unit cell is non-resonant at
was shown, however, nearly two decades ago that those seem- this frequency, and the ground plane reflects the EM waves
ingly useless media can be tamed and used for the profits of with a 0 relative phase shift. It is noted that the phase shift of
highly accurate spatial focusing or imaging purposes through the reflected waves was defined relatively to that of the non-
the concept of time reversal. Time reversal constituted the resonant unit cell, since it is general and can be applied to any
broadband equivalent of phase conjugation, enabling scattering kind of unit cell.
and reverberation harnessing in order tofocus waves far below The schematic diagram of the designed RIS unit cell
the Rayleigh limit, which is given by the transmitting source comprising of two strongly couple (or hybridized) resonators,
aperture in free space [21]–[23]. The time reversal technique and consuming only 50mW, is illustrated in Fig. 1. The first
has been also associated with locally resonant metamaterials resonator is called the reflector patch resonator. It is a patch
 3

-STATE parasitic strip

std of the tansmission [a.u.]

0-STATE 1.8 - 0-state

STRIP FREQUENCIES

STRIP
resonator (par)

Occurences
25 - -state
1.6
1.4 15
1.2

FREQUENCIES
RESONANT

RESONANT
f+ 1 5
f+ fpar
d fpar 0.8
0.6
f ~ 2.34 2.38 2.42 2.46 2.50 2.54 2.58
fref
reflector patch resonator 100 MHz Frequency (GHz)
fref f- (ref) f- 0.4
L
0.2
0
1 W 1 2 2.2 2.4 2.6 2.8
0.8 0.8
SPECTRUM

SPECTRUM
Frequency (GHz)
0.6 0.6
0.4 b) 0.4
0.2 0.2 Fig. 3. Standard deviation of the communication between the transmit (S)
0 0 and receive (R) antennas for 11000 random configurations and 10 S positions,
2 2.2 2.4 2.6 2.8 3 2 2.2 2.4 2.6 2.8 3
Frequency (GHz) Frequency (GHz) and one RIS panel (the inset left in Fig. 2), as well as the distribution of the
resonance frequencies of the 102 resonators for both states measured with
Fig. 1. An RIS unit cell with binary phase tuning. A patch is sitting on near field probes (inset right) [11]. A portion of the fabricated 102-element
a ground plane and is strongly coupled to a parasitic resonator that can be RIS is also illustrated.
controlled with a PIN diode and a power feed [10]. When the two resonators
hybridize (diode forward mode), the resonator reflects the waves with a 0
phase shift, while the main resonator only reflects microwaves with a π phase
shift (diode reverse mode). a wavelength at the working frequency of 2.47GHz (i.e., the
distance between adjacent unit cells is 6cm). The experimental
setup for RIS-empowered indoor wireless communication is
BIN CUPBOARD sketched in Fig. 1. Figure 3 illustrates a portion of the designed
RIS together with the standard deviation of the transmission
SCANNED X
between the transmit Source (S) and Receive (R) antennas for
AREA
(Z=0) 11000 random configurations, 10 positions of the S antenna,
CUPBOARD

DOOR

and one RIS panel (the inset left in Fig. 1). In this figure, the
DESK distribution of the resonance frequencies of the 102 resonators
reception source
antenna (R) antenna (S) for both states measured with near field probes (inset right)
Y Z is also depicted. All RIS elements were controlled through
PARTIALLY two 54-channels Arduino digital controllers, and an Agilent
CUPBOARD
SHIELDED TABLE network analyzer was used to monitor the communication
WINDOWS
between the S and R antennas. Commercial monopole WiFi
antennas polarized along the same axis were used as the RIS
Fig. 2. The RIS-empowered indoor wireless experimental setup of [11] for
highly accurate spatial focusing and nulling. resonators. The S antenna was placed far away and out of
the Line-Of-Sight (LOS) of R and the RIS in a furnished, and
hence, rich scattering office room of dimensions 3m×3m×4m.
The R antenna was placed 1m away from RIS. The conducted
sitting on a ground plane, polarized along its short axis, and experimental results have showcased that the fabricated RIS
whose resonance frequency fref is set to f0 . The second can improve spatial focusing of the radiated EM wave onto
resonator is the parasitic one being a strip line sitting on the well designed λ/2 isotropic focal spots (see Fig. 4), or can be
ground plane and coupled to the reflector in the near field. Its alternatively used for minimizing the EM field on the receiving
resonance frequency fpar can be electronically tuned from f0 R antenna placed at any location inside the room (see Fig. 5).
to a higher frequency f1 using a diode. When fpar is set to Interestingly and contrary to free space, it was shown that RIS
f1 , the reflector resonance frequency fref is unchanged and results in an isotropic shaped EM field around the receiving
it reflects the waves with a π phase shift compared to the antenna, which is attributed to the reverberant nature of the
bare ground plane (this is the π-state). In contrast, when the wireless propagation medium.
fpar is shifted to f0 , the two resonators hybridize and a dimer
presenting two resonant frequencies f − and f +, respectively
below and above f0 , is created. In this state (i.e., at f0 ), C. Multipath Scattering Enrichment
the dimer is again transparent and the waves are reflected by The fabricated RIS design of [11], and specifically a 65-
the ground plane with a 0 relative phase shift (this is the 0 element portion of it, has been very recently deployed in [12]
state). This design presents notable advantages. The reflection in a 1.45m×1m×0.75m chaotic aluminium cavity of volume
properties of each RIS unit cell are insensitive to both the 1.1m3 , as shown in Fig. 6. The RIS covered the 4% of the
losses and impedance variations of the electronic components cavitys surfaces and was deployed in order to boost the rate
and to the soldering, which are only placed on the parasitic performance of a LOS link between two MIMO nodes each
resonator. equipped with 8 antenna elements. Identical commercial WiFi
monopole antennas separated by 10cm (a little more than λ/2
at the working frequency of 2.47GHz) were used at both
B. Spatiotemporal Focusing and Nulling antenna arrays. All antennas were in the same orientation (i.e.,
The fabricated 0.4m2 and 1.5mm of thickness RIS in [11] no polarization diversity applied), and each antenna array was
consisted of 102 controllable EM reflectors spaced by half connected to a Radio Frequency (RF) switch, which in turn
 4

1
−150 −150
0.8
Y (mm)

Y (mm)
0.6
0 0
0.4
0.2
150 150
0
−200 −100 0 100 200 −200 −100 0 100 200
X (mm) X (mm)

Fig. 4. Without RIS (left) and with RIS (right) measured EM field intensity
maps in the XY-plane averaged over 30 realizations of the experiment for the Fig. 6. The aluminium cavity experimental setup of [12] incorporating [11]’s
spatiotemporal focusing use case [11]. The R antenna is placed at the (0, 0) fabricated RIS coated on the wall in order to enrich multipath scattering in
point. a LOS N × N MIMO wireless communication link at the WiFi frequency
2.47GHz.

1
−150 −150 6
0.8
Y (mm)

Y (mm)

0.6
0 0
0.4 5
0.2
150 150
0
−200 −100 0 100 200 −200 −100 0 100 200 4
X (mm) X (mm)

Reff
Fig. 5. Without RIS (left) and with RIS (right) measured EM field intensity 3
maps in the XY-plane averaged over 30 realizations of the experiment for the
spatiotemporal nulling use case [11]. The R antenna is placed at the (0,0) Rayleigh matrix
point. Orthogonal matrix
2
Average over realizations
Single realization
was connected to a vector network analyzer. The 65 RIS 1
100 200 300
unit cells in the experimental setup of Fig. 6 were designed Iteration number
according to the iterative sequential optimization algorithm in
[38] in order to improve the rank of the N ×N MIMO channel Fig. 7. The evolution of the effective rank (Reff ) of the LOS N × N channel
matrix (red for N = 2, blue for N = 4, and orange for N = 6) over
propagation matrix with 2 ≤ N ≤ 8,compared to the case the number of algorithmic steps for a single channel realization and for the
where the RIS is not used. For each iteration of the deployed average over 30 channel realizations with different MIMO channel geometries.
algorithm, the binary configuration (0 and π phase states) of a Benchmarks for Rayleigh fading and perfect channel orthogonality are also
included.
unit cell was changed, the new channel matrix was measured,
and the new effective channel rank was calculated. If that
phase change resulted in higher channel rank, the unit cell’s
configuration was updated accordingly. With this iterative way, propagation channels reached the maximum number 3, which
the configurations of all unit cells were updated. It is noted was only equal to 2 when the RIS was not used. This fact was
that the configuration of each unit cell was iterated multiple translated to improvement in the achievable rate performance.
times to deal with the long-range correlations between the
RIS optimal configurations, due to the reverberation inside the IV. S TATE - OF - THE -A RT I N
cavity. The obtained experimental results are demonstrated in RIS H ARDWARE A RCHITECTURES
Fig. 7 for N = 2, 4, and 6. As shown for all considered In this section, we overview the late advances in RIS
values for N , the optimized RIS configuration leads to a full hardware architecture designs for wireless communications
rank channel as the orthogonal one, which models the richest based on three different categorizations [30]. The first cate-
multipath scattering conditions. It can be also seen that the gory focuses on whether an RIS includes active or passive
convergence to the highest possible channel rank depends on components, which consequently determines its overall power
the value of N . consumption. The second category is based on whether an RIS
It is obvious from Fig. 7 that the optimal channel diver- is a contiguous surface or is comprised of discrete elements,
sity can be achieved by physically shaping the propagation and the third category discusses the available RIS modes of
medium itself via [11]’s fabricated RIS design having nearly operation when deployed in wireless communication systems.
passive unit cells. The disorder of the random EM propa-
gation environment inside the cavity was actually tamed to
impose perfect orthogonality of the wireless channels. As A. RIS Power Consumption
further demonstrated in [12] through an indoor wireless image 1) Active RISs: To realize reconfigurable wireless envi-
transmission trial using a LOS 3 × 3 MIMO link empowered ronments, an RIS can serve as a transmitter, receiver, or an
by the fabricated RIS, the number of effective independent EM wave modulator. When the transceiver role is considered,
 5

and thus energy-intensive RF circuits and signal processing the interference principle of the EM propagation wave. The
units are embedded in the surface, the term active RIS is recorded EM field can be then utilized for reconstructing the
adopted [30]–[36]. Many RF chains realizing a special form initial field based on the diffraction principle. It should be
of hybrid A/D beamforming have been considered in [34] noted that wireless communications over a continuous aperture
(currently, only for transmission), whereas [36] proposes that is inspired by optical holography. Since a continuous aperture
all RIS unit elements are attached to a single RF chain to benefits from the theoretical integration of an infinite number
enable efficient wireless channel estimation at the RIS side. On of antennas which can be viewed as the asymptotic limit of
another note, active RIS systems comprise a natural evolution Massive MIMO, its potential advantages include the highly
of conventional massive MIMO systems, by packing more accurate spatial resolution, and the creation/ detection of EM
and more software-controlled antenna elements onto a 2- waves with arbitrary spatial frequency components, without
dimensional surface of finite size. In [33], where the spacing undesired side lobes.
between the RIS unit elements reduces when their number 2) Discrete RISs: A discrete RIS is usually composed
increases, an active RIS is also termed as large intelligent of many discrete unit cells made of low power software-
surface. A practical implementation of active RISs can be a tunable metamaterials [11], [17], [18], [31]–[34]. The means
compact integration of a large number of tiny antenna elements to electronically modify the EM properties of the unit cells
with reconfigurable processing networks realizing a continuous range from off the shelves electronic components to using
antenna aperture. This structure can be used to transmit and liquid crystals, MicroElectroMechanical Systems (MEMS) or
receive communication signals across the entire surface by even electromechanical switches, and other reconfigurable
leveraging the hologram principle [31], [32]. Another active metamaterials. This structure is substantially different from the
RIS implementation can be based on discrete photonic antenna conventional multi-antenna antenna array. One embodiment
arrays that integrate active optical-electrical detectors, convert- of a discrete surface is based on discrete meta-atoms with
ers, and modulators for performing transmission, reception, electronically steerable reflection properties [16]. As men-
and conversion of optical or RF signals [31]. tioned earlier, another type of discrete surface is the active one
2) Passive RISs: As previously discussed, a passive RIS based on photonic antenna arrays. Compared with contiguous
acts like a passive metal mirror or wave modulator, and RISs, discrete RISs have some essential differences from the
can be programmed to change an impinging EM field in a perspectives of implementation and hardware [30]. Current
customizable way [11], [17], [18]. Compared with its ac- RIS fabricated designs comprise of discrete unit elements.
tive counterpart, a passive RIS is usually composed of low
cost passive elements that do not require dedicated power
C. RIS Operation Modes
sources. Their circuitry and embedded sensors can be powered
with energy harvesting modules, an approach that has the The following four RIS operating modes are lately consid-
potential of making them truly energy neutral. Regardless of ered in wireless communications: 1) continuous RISs acting
their specific implementations, what makes the passive RIS as active transceivers; 2) discrete RISs used also as active
technology attractive from an energy efficiency standpoint, is transceivers; 3) discrete RISs with a single RF chain intended
its inherent capability in controlling waves impinging upon for wireless channel estimation; and 4) discrete passive RISs
it, and forwarding the incoming signal without employing operating as EM wave modulators.
any power amplifier nor RF chain, and also without applying 1) Continuous RISs as Active Transceivers: According to
sophisticated signal processing techniqes. Moreover, passive this mode of operation, a continuous RIS operates as an active
RISs can work in full duplex mode without significant self- transceiver. The RF signal is generated at its backside and
interference or increased noise level, and require only low propagates through a steerable distribution network to the
rate control link or backhaul connections. Finally, passive contiguous surface constituted by a large number of software-
RIS structures can be easily integrated into the wireless defined and electronically steerable elements that generate
communication environment, since their extremely low power multiple beams to the intended users. A distinct difference
consumption and hardware costs allow them to be deployed between and active continuous RIS and a passive one is
into building facades, room and factory ceilings, laptop cases, that the beamforming process of the former is accomplished
or even human clothing [17], [18]. based on the holographic concept, which is a new dynamic
beamforming technique based on software-defined EM wave
modulators with low cost/weight, compact size, and a low-
B. RIS Hardware Structures power circuitry.
1) Contiguous RISs: A contiguous RIS integrates a virtu- 2) Discrete RISs as Active Transceivers: Discrete active
ally infinite number of elements into a limited surface area RISs, also known as Dynamic Metasurface Antennas (DMAs),
in order to form a spatially continuous transceiver aperture have been recently proposed as an efficient realization of
[31], [32]. For a better understanding of the operation of massive antenna arrays for wireless communications [13],
contiguous surfaces and their communication models, we [34]. They provide beam tailoring capabilities using simplified
commence with a brief description of the physical operation transceiver hardware, which requires less power and cost
of the optical holography concept. Holography is a technique compared to conventional hybrid A/D/ antenna arrays (i.e.,
that enables an EM field, which is generally the result of a those based on patch arrays and phase shifters), eliminating
signal source scattered off objects, to be recorded based on the need for complicated corporate feed as well as active phase
 6

shifters. DMAs may comprise of a large number of tunable antenna design, and wireless communications are required.
metamaterial antenna elements that can be packed in small Luckily, and in contrast to the unfortunate parallel research
physical areas [38] for a wide range of operating frequencies. paths up to date, the RIS research topic has pronounced the
This feature makes them an appealing technology for the next common ground of the latter fields witnessing the increased
generation extreme massive MIMO transceivers. needs for higher convergence between electromagnetics, com-
3) Discrete RISs for Channel Estimators: An RIS architec- munication, and wireless systems theories.
ture comprising of any number of passive reflecting elements, In the following, we emphasize on some of the key open
a simple controller for their adjustable configuration, and a challenges and future research directions (see also [30], [39])
single RF chain for baseband measurements was presented with the fabrication of individual RIS hardware structures as
in [36]. Capitalizing on this architecture and assuming sparse well as the design of wireless networking schemes empowered
wireless channels in the beamspace domain, an alternating by RISs as EM wave controllers.
optimization approach for explicit estimation at the RIS side 1) Electromagnetics Information Theory: It is natural to
of the channel gains at its elements (which are all attached to expect that RIS-empowered wireless communication systems
the single RF chain) was proposed. will exhibit different features compared with traditional com-
4) Discrete RISs as EM Wave Modulators:: Another opera- munications based on conventional multi-antenna transceivers.
tion mode of RISs is the mirror or wave modulator, where the Recall that current communication systems operate over un-
surface is considered to be discrete and passive. In this case, controllable wireless environments, whereas wireless systems
an RIS includes unit cells that can reconfigured in real time (as incorporating EM wave modulating RISs will be capable of
[11]’s reflectarray that was previously described), which makes reconfiguring signal propagation. This fact witnesses the need
their synergetic operation resembling that of spatiotemporal for new information theoretical methodologies to characterize
focusing. [10], unlike the late considerations for continuous the physical channels in RIS-empowered systems and analyze
transceiver RIS systems. It is worth noting that most of the their ultimate capacity gains (e.g., study optimal signaling and
existing works (e.g., [5], [7], [8]) focus on this RIS operation multi-user communcation), as well as for new signal process-
mode which is simpler to implement and analyze. ing algorithms and networking schemes for realizing wireless
communications via RISs. For example, continuous RISs can
be used for the reception and transmission of the impinging
V. O PEN C HALLENGES AND F UTURE D IRECTIONS
EM field over its continuous aperture using the hologram
As highlighted in the previous sections and as witnessed concept. Differently from the massive MIMO systems, RIS
from the recent RIS-focused demonstrations and increasingly operation can be described by the Fresnel-Kirchhoff integral
booming research interests, RISs have high potential in en- that is based on the Huygens-Fresnel principle [11], [31]. It
abling smart wireless propagation environments via their deli- also interesting to devise and analyze schemes based on time
cate EM wave control capabilities. In addition, their hardware reversal (e.g., [22], [24], [40]) in the context of future large-
design principles can pave the way for the fabrication of the bandwidth communications empowered by RISs (especially
next generation ultra massive MIMO antenna systems with in the millimeter and THz bands), in order to enable highly
acceptable power consumption levels and affordable hardware accurate multi-spot spatiotemporal focusing for diverse com-
complexity. Although, RIS-enabled wave control dates back munication objectives. To this end, the implications of the
to acoustics, recent experimental results and hardware designs RIS role over different frequency bands need to be further
showcased the dynamics of the various RIS architectures investigated. For example, at very high frequencies, the EM
(especially those stemming from metamaterials) for higher field is not as diffuse in nature as it is in low frequencies.
frequency communications (from sub6GHz to THz), which This entails RISs to be deployed as highly localized access
will be a core component of 6G wireless communication point extenders, rather than structures that are capable to enrich
networks. multipath scattering (as has been done for throughput boosting
Owing to the nearly zero power consumption of passive in WiFi frequencies).
RISs and their envisioned extremely low fabrication cost and 2) Modeling of RIS-Empowered EM Wave Propagation:
compact size, the ubiquitous deployment (i.e., ultra densifi- Realistic models for the EM wave propagation of signals
cation) of intelligent EM wave control structures becomes bouncing on RISs are needed. Additionally, the adoption of
feasible, which will eventually lead to the realization of the large RISs challenges the common far-field EM propagation
holographic concept [30], [32] for 6G wireless communi- assumption. The sources of information signals can be close
cation. This concept will enable highly accurate multi-spot to RISs, and particularly, in distances smaller than the RIS
spatiotemporal focusing for diverse communication objectives structure size, giving rise to near-field EM propagation. It
(e.g., increase highly localized throughput, accurate position- is thus of paramount importance to devise physics-inspired
ing, improved security, and reduced EM field exposure) with models for EM wave propagation in the RIS proximity, which
minimal realization overhead. It is also noted that holographic account for interactions in the RIS circuitry. It also necessary
wireless systems can be further reinforced with active RISs to study realistic pathloss models for RIS-empowered wireless
made from power efficient metamaterials. However, to unveil networks (especially for the passive RIS case), in order to
the true potential of holographic communications and devise unveil the RISs true link budget potential as well as their
efficient relevant algorithmic approaches, synergies among the optimal placement in space. Moreover, the interactions of
theorists and practitioners in the fields of EM propagation, the RIS unit elements, which are placed in subwavelength
 7

distances in RISs made from metamaterials, need to accurately beams for various communication objectives (e.g, massive
modeled and incorporated in the channel matrix model in order data streams for multiple spatial spots when throughput is
to be accounted for in the communication theory analyses and the objective). In addition, the future designs need to address
wireless signal processing designs. the provisioned requirements for the millimeter and the THz
3) Channel Tracking in RIS-Empowered Networks: To bands. In such cases, efficient hardware designs are necessary,
date, studies on RIS-empowered wireless communications as- which currently quite challenging.
sume that the transmitters/receivers have full channel informa- Since both active and passive RISs are lately gaining
tion knowledge. In practice, however, the channel coefficients increased interest for both EM wave propagation control
need to be efficiently estimated, which is a challenging task and transmission/reception, it is reasonable to envision hy-
with either passive or active RISs. Channel estimation cannot brid passive and active RISs. Such structures will notably
be implemented at the side of a passive RIS, but rather at strengthen the design flexibility for RISs, either for enabling
one end (transmitter or receiver) of the communication link. programmable wireless environments, or realizing ultra mas-
This makes the channel estimation task challenging and has sive MIMO antenna arrays, or both. For instance, having such
motivated the introduction of channel estimation protocols a hybrid RIS acting as a receiving device [36] can significantly
for the case of passive-RIS-empowered wireless communica- facilitate channel estimation via machine learning tools [7],
tions. However, current approaches require lengthy channel which is still a major challenge and a source of substantial
estimation protocols, and low overhead channel estimation communication control overhead in purely passive RISs. In
frameworks are needed. When channel estimation is carried addition, hybrid RISs will enable more advanced relaying
out in a time-division duplexing manner, active RISs offer strategies, overcoming the dominating impact of pathloss in
the possibility of tuning their elements to facilitate channel the applications of their passive versions.
estimation via pilot signals, and to adapt in a manner which 6) Use Cases for RIS-Empowered Connectivity: The use
optimizes data reception in light of the estimated channel. The cases and applications where passive and active RISs can pro-
design and analysis of efficient algorithms for active RISs, vide substantial improvement compared to current transceiver
which have to estimate features of the wireless channel and and network architectures have not yet been thoroughly iden-
reliably communicate, have not yet been properly treated. tified. For example, the RIS planar shape and compact size for
4) Algorithms for RIS-Enabled EM Wave Control: Chan- the passive versions as well as the active versions with small
nel dependent beamforming has been extensively considered numbers of RF chains, facilitate their deployment in indoor
in massive MIMO systems. However, realizing environment- environments, like buildings, factories, malls, train stations,
aware designs in RIS-empowered wireless communication hospitals, and airports. In such setups, RISs are expected to
systems is extremely challenging, since the RIS unit cells, communicate with multiple users in close to LOS conditions,
which can be fabricated from metamaterials, impose demand- possibly operating in the near-field regime. As previously
ing online tuning constraints. The latest RIS design formu- discussed, such near-field scenarios bring forth the possibility
lations include large numbers of reconfigurable parameters of spatiotemporal focusing, namely, the ability to focus the
with nonconvex constraints, rendering their optimal solution signal towards a specific location in space, instead of a
highly nontrivial. For the case of continuous RISs, intelligent specific direction as in the far-field conditions via conventional
holographic beamforming is an approach to smartly target and beamforming. Moreover, the potential of passive and active
follow individual or small clusters of devices, and provide RISs in outdoor network setups needs to be further investigated
them with high fidelity beams and smart radio management. and demonstrated, and for different operating frequencies.
However, self-optimizing holographic beamforming technolo- Finally, a large body of works that combines RISs with
gies that depend on complex aperture synthesis and low level various other communication technologies (e.g., physical layer
modulation are not available yet. security, unamanned aerial vehicles, energy harvesting, and
Current algorithmic designs for active RISs focus either cognitive networking) have lately appeared, identifying various
on narrowband communications or ignore their capability to relevant design challenges.
dynamically configure the frequency-selective profile of each 7) Joint Communication and Computing RIS Platforms:
unit metamaterial element. This unique property, which does Late advances in chipset design and computational effective-
not exist in any conventional hybrid A/D architecture, provides ness of AI approaches have enabled the incorporation of basic
increased flexibility for the design of wideband operation by AI functionalities in 5th Generation (5G) base stations and
matching the spectral behavior of each element to optimize the mobile handsets. Following this trend and the basic computing
equivalent wideband channel. Consequently, the true potential and storage capabilities of the current RIS designs, we envision
of extreme massive MIMO systems implemented with active future RISs to being capable of training local Artificial Neural
RISs in achieving ultra-reliable and ultra-high rate communi- Networks (ANNs) to obtain models for their local wireless
cations is not yet fully explored [39]. connectivity environment. Each ANN can operate on the unit
5) Design of Passive and Active RIS Architectures: A cell level, where each cell trains a model and all derived
large body of fabricated designs and experimental works is still models from the unit cells are used to design the global
required in order to transit the RIS concept into an established model for the RIS. This model can be used for efficient
technology for 6G wireless communications. As previously online configurations as per the desired EM wave control.
discussed, densely deployed EM wave modulating RISs have Alternatively, each RIS model can be trained from all the unit
the potential to enable massive numbers of highly focused cells simultaneously.
 8

In achieving the latter overarching goal, further advances in R EFERENCES

low complexity AI approaches are needed. To this direction
[1] W. Saad, M. Bennis, and M. Chen, “A vision of 6G wireless systems:
belong the binary neural networks, which are lately receiving Applications, trends, technologies, and open research problems,” IEEE
significant attention for smart mobile handsets. These ANNs Network, 2019.
have binary weights and are activated at run time. At the [2] F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L.Marzetta, O. Edfors,
and F. Tufvesson, “Scaling up MIMO: Opportunities and challenges with
training time, the weights and activations are used for com- very large arrays,” IEEE Signal Process. Mag., vol. 30, no. 1, pp. 40–60,
puting gradients, however, the gradients and true weights are Jan. 2013.
stored in full precision. This procedure permits effective ANN [3] A. K. Khandani, “Media-based modulation: A new approach to wireless
transmission,” in Proc. IEEE ISIT, Istanbul, Turkey, Jul. 2013, pp. 3050–
training on systems with limited resources. The availability of 3054.
an individual ANN model per RIS structure can be used for [4] G. C. Alexandropoulos, V. I. Barousis, and C. B. Papadias, “Precoding
future configurations of the values of all the deployed RISs in for multiuser MIMO systems with single-fed parasitic antenna arrays,”
the RIS-empowered wireless network. The individual ANNs in Proc. IEEE GLOBECOM, Austin, USA, Dec. 2014, pp. 3656–3661.
[5] M. A. Sedaghat, V. I. Barousis, R. R. Müller, and C. B. Papadias, “Load
can be also shared to a central network entity that gathers modulated arrays: A low-complexity antenna,” IEEE Commun. Mag.,
in a compressed manner the sensing information from the vol. 54, no. 3, pp. 46–52, Mar. 2016.
available dense network of AI-enabled RISs. This sensing [6] A. F. Molisch, V. V. Ratnam, S. Han, Z. Li, S. L. H. Nguyen, L. Li,
and K. Haneda, “Hybrid beamforming for massive MIMO: A survey,”
information can be used for network monitoring, management, IEEE Commun. Mag., vol. 55, no. 9, pp. 134–141, Sep. 2017.
and optimization purposes. [7] E. Vlachos, G. C. Alexandropoulos, and J. Thompson, “Wideband
MIMO channel estimation for hybrid beamforming millimeter wave
systems via random spatial sampling,” IEEE J. Sel. Topics Signal
VI. C ONCLUSIONS Process., vol. 13, no. 5, pp. 1136–1150, Sep. 2019.
[8] S. S. Ioushua and Y. C. Eldar, “A family of hybrid analogdigital
The concepts of artificial EM wave propagation control and beamforming methods for massive MIMO systems,” IEEE Trans. Signal
Process. vol. 67, no. 12, pp. 3243–3257, Jun. 2019.
tunable reflecting metamaterials, which naturally constitute the [9] S. V. Hum and J. Perruisseau-Carrier, “Reconfigurable reflectarrays and
ancestors of smart programmable wireless environments and array lenses for dynamic antenna beam control: A review,” IEEE Trans.
RISs, have been initially conceived in the acoustics and optics Antennas Prop., vol. 62, no. 1, pp. 183–198, Jan. 2014.
[10] H. Yang, X. Cao, F. Yang, J. Gao, S. Xu, M. Li, Z. Chen, Y. Zhao,
fields dating back to more than ten years ago. Due to the Y. Zheng, and S. Li, “A programmable metasurface with dynamic
increased potential of RISs for 6G wireless communication polarization, scattering and focusing control,” Scientific Reports, vol.
networks, as witnessed by the recent proof of concepts with 6, pp. 1–11, Oct. 2016.
both passive reflectarrays and active metasurface antennas, [11] N. Kaina, M. Dupre, G. Lerosey, and M. Fink, “Shaping complex mi-
crowave fields in reverberating media with binary tunable metasurfaces,”
there has been lately a surprisingly increasing attention on the Scientific Reports, vol. 4, pp. 1–7, Aug. 2014.
RIS topic from both academia and industry working in antenna [12] P. del Hougne, M. Fink, and G. Lerosey, “Optimally diverse commu-
design and wireless communications. In fact, a large body nication channels in disordered environments with tuned randomness,”
Nature Electronics, vol. 2, pp. 36–41, Jan. 2019.
of research papers and special issues in prestigious ComSoc [13] I. Yoo, M. F. Imani, T. Sleasman, H. D. Pfister, and D. R. Smith, “En-
and Antenna Propagation Society periodicals has appeared hancing capacity of spatial multiplexing systems using reconfigurable
in the last 2 years, as well as novel RIS demonstrators and cavity-backed metasurface antennas in clustered MIMO channels,” IEEE
Trans. Commun., vol. 67, no. 2, pp. 1070–1084, Feb. 2018.
collaborative R&D projects. [14] D. R. Smith, J. B. Pendry, and M. C. Wiltshire, “Metamaterials and
In this paper, we provided a brief historical descrip- negative refractive index,” Science, vol. 305, no. 5685, pp. 788–792,
tion on wave propagation control for optics and acoustics, 2004.
[15] D. R. Smith, O. Yurduseven, L. Pulido-Mancera, P. Bowen, and N. B.
and overviewed two representative indoor wireless trials at Kundtz, “Analysis of a waveguide-fed metasurface antenna,” Physical
2.47GHz for EM wave modulation with a fabricated passive Review Applied, vol. 8, no. 5, Nov. 2017.
discrete RIS. The first trial dating back to 2014 showcased [16] C. Liaskos, S. Nie, A. Tsioliaridou, A. Pitsillides, S. Ioannidis, and I. F.
Akyildiz, “A new wireless communication paradigm through software-
the feasibility of highly accurate spatiotemporal focusing and controlled metasurfaces,” IEEE Commun. Mag., vol. 56, no. 9, pp. 162–
nulling, while the second very recent one in 2019 demonstrated 169, Sep. 2018.
that passive RISs can enrich multipath scattering, thus, en- [17] Huang, A. Zappone, G. C. Alexandropoulos, M. Debbah, and C. Yuen,
“Reconfigurable intelligent surfaces for energy efficiency in wireless
abling throughput boosted wireless connectivity. We also dis- communication,” IEEE Trans. Wireless Commun., vol. 18, no. 8, pp.
cussed the current status in RIS hardware designs emphasizing 4157–4170, Aug. 2019.
the key features of the different approaches. We concluded the [18] M. Di Renzo, M. Debbah, D.-T. Phan-Huy, A. Zappone, M.-S. Alouini,
C. Yuen, V. Sciancalepore, G. C. Alexandropoulos, J. Hoydis, H.
paper with a detailed list of key open challenges and future Gacanin, J. de Rosny, A. Bounceu, G. Lerosey, and M. Fink, “Smart
research directions for the design of individual RIS structures radio environments empowered by reconfigurable AI meta-surfaces: an
as well as for connectivity approaches in RIS-empowered idea whose time has come,” EURASIP J. Wireless Commun. Netw., vol.
2019, no. 1, pp. 1–20, May 2019.
wireless networks. As advocated in numerous parts of this [19] J. W. Goodman, Introduction to Fourier optics, Englewood, Roberts &
paper, the RIS topic triggers fascinating synergies among the Company Publishers, 2005.
fields of EM propagation, antenna design, communication [20] N. Yu, P. Genevet, M. A. Kats, F. Aieta, J.-P. Tetienne, F. Capasso, and
theory, and signal processing for wireless communications. Z. Gaburro, “Light propagation with phase discontinuities: Generalized
laws of reflection and refraction,” Science, vol. 334, no. 6054, pp. 333–
More importantly, we believe that potential of RISs in EM 337, 2011.
wave propagation control will enable highly accurate multi- [21] M. Fink, “Time reversed acoustics,” Physics Today, vol. 50, no. 3, Mar.
spot spatiotemporal focusing towards the ultimate goal for 1997.
[22] C. Draeger, and M. Fink, “One-channel time reversal of elastic waves
revolutionary 6G wireless communication networks with em- in a chaotic 2D-silicon cavity,” Physical Review Lett., vol. 79, Article
bedded environmental intelligence. ID 407, Jul. 1997.
 9

[23] A. Derode, A. Tourin, J. de Rosny, M. Tanter, S. Yon, and M. Fink,

“Taking advantage of multiple scattering to communicate with time-
reversal antennas,” Physical Review Lett., vol. 90, Article ID 014301,
Jan. 2003.
[24] G. Lerosey, J. de,Rosny, A. Tourin, and M. Fink, “Focusing beyond the
diffraction limit with far-field time reversal,” Science, vol. 315, no. 5815,
pp. 11–20, Feb. 2007.
[25] F. Lemoult, M. Fink, and G. Lerosey, “Acoustic resonators for far-field
control of sound on a subwavelength scale,” Physical Review Lett., vol.
107, Article ID 064301, Aug. 2011.
[26] F. Lemoult, G. Lerosey, J. de Rosny, and M. Fink, “Resonant metalenses
for breaking the diffraction barrier,” Physical Review Lett., vol. 104,
Article ID 203901, May 2010.
[27] Y. M. Wang, B. Judkewitz, C. A. Dimarzio, and C. Yang, “Deep-tissue
focal fluorescence imaging with digitally time-reversed ultrasound-
encoded light,” Nature Commun., vol. 3, no. 928, Jun. 2012.
[28] O. Katz, E. Small, Y. Bromberg, and Y. Silberberg, “Focusing and
compression of ultrashort pulses through scattering media,” Nature
Photonics, vol. 5, pp. 372–377, May 2011.
[29] I. M. Vellekoop and A. P. Mosk, “Focusing coherent light through
opaque strongly scattering media” Optics Lett., vol. 32, no. 16, pp.
2309–2311, 2007.
[30] C. Huang, S. Hu, G. C. Alexandropoulos, A. Zappone, C. Yuen, R.
Zhang, M. Di Renzo, and M. Debbah, “Holographic MIMO surfaces
for 6G wireless networks: Opportunities, challenges, and trends,” IEEE
Wireless Commun., accepted for publication, 2020.
[31] O. Yurduseven, D. L. Marks, T. Fromenteze, and D. R. Smith, “Dynam-
ically reconfigurable holographic metasurface aperture for a mills-cross
monochromatic microwave camera,” Optics Express, vol. 26, no. 5, pp.
5281–5291, 2018.
[32] A. Pizzo, T. L. Marzetta, and L. Sanguinetti, “Spatially-stationary
model for holographic MIMO small-scale fading,” [online]
https://arxiv.org/abs/1911.04853, 2019.
[33] S. Hu, F. Rusek, and O. Edfors, “Beyond massive MIMO: The potential
of positioning with large intelligent surfaces,” IEEE Trans. Signal
Process., vol. 66, no. 7, pp. 1761–1774, Apr. 2018.
[34] N. Shlezinger, O. Dicker, Y. C. Eldar, I. Yoo, M. F. Imani, and D.
R. Smith, “Dynamic metasurface antennas for uplink massive MIMO
systems,” IEEE Trans. Commun., vol. 67, no. 10, pp. 6829–6843, Oct.
2019.
[35] A. Taha, M. Alrabeiah, and A. Alkhateeb, “Enabling large intelli-
gent surfaces with compressive sensing and deep learning,” [online]
https://arxiv.org/abs/1904.10136v2, 2019.
[36] G. C. Alexandropoulos and E. Vlachos, “A hardware architecture for
reconfigurable intelligent surfaces with minimal active elements for
explicit channel estimation,” in Proc. IEEE ICASSP, Barcelona, Spain,
May 2020, pp. 9175–9179.
[37] I. F. Akyildiz and J. M. Jornet, “Realizing ultra-massive MIMO (1024×
1024) communication in the (0.06-10) terahertz band, Nano Commun.
Netw., vol. 8, pp. 46–54, 2016.
[38] I. M. Vellekoop and A. P. Mosk, “Phase control algorithms for focusing
light through turbid media,” Optics Commun., vol. 281, no. 11, pp. 3071–
3080, 2008.
[39] N. Shlezinger, G. C. Alexandropoulos, M. F. Imani, Y. C. Eldar,
and D. R. Smith, “Dynamic metasurfaces antennas for 6G massive
MIMO communications,” IEEE Commun. Mag., under review, [online]
https://arxiv.org/pdf/2006.07838.pdf, 2020.
[40] G. C. Alexandropoulos, R. Khayatzadeh, M. Kamoun, Y. Ganghua, and
M. Debbah, “Indoor time reversal wireless communication: Experimen-
tal results for localization and signal coverage,” in Proc. IEEE ICASSP,
Brighton, UK, May 2019, pp. 1–5.