Modeling the Mutual Coupling

of Reconfigurable Metasurfaces
Marco Di Renzo∗ , Vincenzo Galdi† , and Giuseppe Castaldi†
∗ Université Paris-Saclay, CNRS, CentraleSupélec, Laboratoire des Signaux et Systèmes, 91192 Gif-sur-Yvette, France
† University of Sannio, Department of Engineering, 82100, Benevento, Italy

[email protected]

Abstract—Recently, a circuits-based approach for modeling

the mutual coupling of reconfigurable surfaces, which comprise
sub-wavelength spaced passive scattering elements coupled with
electronic circuits for enabling the reconfiguration of the surface,
has been introduced. The approach is based on a finite-length
discrete dipole representation of a reconfigurable surface, and
on the assumption that the current distribution on each thin
wire dipole is a sinusoidal function. Under these assumptions,
the voltages at the ports of a multi-antenna receiver can be
formulated in terms of the voltage generators at a multi-antenna
transmitter through a transfer function matrix that explicitly
depends on the mutual coupling and the tuning circuits through
the mutual impedances between every pair of thin wire dipoles. In
currently available works, the mutual impedances are formulated
in an integral form. In this paper, we show that they can be
formulated in a closed-form expression in terms of exponential
integral functions. Fig. 1. Schematic illustration of the region of influence of evanescent waves.
Index Terms—Metasurfaces, reconfigurable intelligent sur-
faces, sub-wavelength design, mutual coupling.
representation of the electromagnetic field under consideration.
I. I NTRODUCTION For the exact reconstruction of an electromagnetic field, it is
Dynamic (i.e., reconfigurable) metasurfaces have recently known that the sampling period depends on the observation
attracted major interest from the wireless research community plane with respect to a reference plane, e.g., the plane that
[1], [2], [3]. These structures can be utilized for different contains a given natural or digitally controllable scatterer. As
applications in wireless networks, which include the imple- sketched in Fig. 1, at distances from the reference plane where
mentation of smart reflectors, holographic transceivers, and the evanescent waves (i.e., waves that attenuate exponentially
single-RF multiple-input multiple-output modulators [4]. In along the z-axis) that constitute the electromagnetic field are
spite of the many potential applications in wireless commu- negligible and can be ignored, a sampling period equal to half-
nications, the integration of dynamic surfaces into wireless wavelength is enough to perfectly recover the electromagnetic
systems needs to tackle many open challenges, which include field. At distances from the reference plane where the evanes-
the control overhead to enable the configuration of the surfaces cent waves that constitute the electromagnetic field are not
every channel coherence interval [5]. From a communication- negligible and cannot be ignored, however, a sampling period
theoretic point of view, major open problems include the equal to half-wavelength is not enough to perfectly recover
development of communication models for dynamic surfaces the electromagnetic field because the evanescent waves create
that are accurate enough but sufficiently tractable for per- rapid oscillations that cannot be reconstructed with a half-
formance analysis and optimization [6], the development of wavelength sampling period [13, Section 1.4.2]. An illustration
efficient algorithms for optimization [7], and the development of the relation between the bandwidth of an electromagnetic
of analytical frameworks to unveil fundamental performance field in the wavenumber domain and the observation plane (the
trends and scaling laws [8], [9], [10]. distance with respect to the xy-plane in Fig. 1) is illustrated
In general terms, a metasurface is defined as a reflectarray in [14, Fig. 2]. Even though the evanescent waves decay
with sub-wavelength inter-distances between the array ele- exponentially along the z-axis and are typically ignored in
ments [11], [12]. Sub-wavelength dynamic surfaces play an wireless communications, their control is, however, instrumen-
important role for controlling the wavefront of electromag- tal for realizing high-performance metasurfaces, e.g., perfect
netic waves with a high power efficiency. In this context, a anomalous reflectors with high power efficiency and a large
fundamental question is: how finely an electric field needs to field of view (i.e., a large angle of reflection with respect to
be sampled in order to recover it exactly? [13]. The answer lies the angle of incidence) [11].
in the sampling theorem applied to the wavenumber domain Reconfigurable surfaces with sub-wavelength inter-distances

This paper's
Authorizedcopyright islimited
licensed use held to:
byQueen
the author(s). It isofpublished
Mary University in theseonproceedings
London. Downloaded March 12,2024 and included
at 22:01:25 UTC in any
from archive
IEEE Xplore. such as IEEE
Restrictions Xplore
apply.
under the license granted by the "Agreement Granting EurAAP Rights Related to Publication of Scholarly Work."
 between the scattering elements are, however, not easy to
model and analyze. Due to the closely spaced scattering
elements, the mutual coupling among them cannot be ignored
and need to be taken in account at the design stage. It
is known, in fact, that an appropriate optimization of the
mutual coupling of antenna arrays may result in super-directive
designs in which the gain scales with the square of the
number of antenna elements, provided that the Ohmic losses
are sufficiently low [15]. In the context of modeling the
mutual coupling of sub-wavelength reconfigurable surfaces,
the authors of [16] have recently introduced a circuits-based
approach for application to relay-type reconfigurable surfaces,
which are known as reconfigurable intelligent surfaces (RIS), Fig. 2. Schematic illustration of the impact of sub-wavelength designs on
that explicitly accounts for the sub-wavelength spacing among utilizing a given surface area.
the passive scattering elements and the circuital model for
enabling their configuration. The approach is based on a finite-
length discrete dipole representation [17] for a reconfigurable
surface and on the assumption that the current distribution on
each thin wire dipole is a sinusoidal function. The authors of
[18] and [19] have subsequently shown that the performance of
RIS-aided systems can be substantially improved if the mutual
coupling is taken into account by design. In particular, mutual
coupling aware designs enable either a better utilization of
the aperture of the surface, since a larger number of densely
spaced scattering elements can be deployed on the same
surface area, or a reduction of the form factor of the surface,
since the same number of densely spaced scattering elements
can be deployed on a smaller surface area. This tradeoff is
sketched in Fig. 2. Fig. 3. Schematic illustration of the impedance-based model for relay-type
surfaces introduced in [16], [6].
In [16], the end-to-end transfer function matrix between
a pair of transmitter and receiver is formulated in terms of where zRT is the mutual impedance between the transmitter
mutual impedances that depend on the scattering elements and the receiver, zRS is the N ×1 vector of mutual impedances
and the tuning circuits for enabling the configuration of between the N reconfigurable elements of the surface and
the surface. In [16], however, the mutual impedances are the receiver, zST is the N × 1 vector of mutual impedances
formulated in an integral form. In this paper, we show that between the transmitter and the N reconfigurable elements of
they can be formulated in a closed-form expression by uti- the surface, ZRIS is the N × N diagonal matrix containing
lizing a simple approach borrowed from [20]. Thanks to the the N equivalent impedances of the tuning circuits enabling
obtained analytical expressions, the analysis and optimization the reconfigurability of the surface, and ZSS is the N × N
of reconfigurable surfaces in the presence of mutual coupling (full) matrix containing the mutual impedances between every
is greatly simplified. pair of scattering elements of the surface. Under the modeling
II. C LOSED -F ORM E XPRESSION OF THE M UTUAL assumptions in [16], ZSS is independent of ZRIS .
I MPEDANCES We are interested in computing the generic element of
For simplicity, we depart from [16, Corollary 1] that is ZSS . For ease of writing, let us denote the (qp)th element
applicable to single-antenna transmitters and single-antenna of ZSS by zqp . Specifically, zqp denotes the contribution of
receivers under the assumption that they are in the far-field of the electric field generated by the current flowing through the
each scattering element that constitutes an electrically large pth reconfigurable element of the surface when it is observed
reconfigurable surface made of N reconfigurable elements. on the qth reconfigurable element of the surface. From [16],
Therefore, the model is applicable to the near-field and the we have the following:
far-field regions of the whole surface. For self-consistency, jη
Z +hq
a schematic illustration of the impedance-based model intro- zqp = Eqp (z) fq (z) dz (2)
4πk −hq
duced in [16] is reported in Fig. 3 [6].
The end-to-end channel can be written as follows1 : where
Z +hp  
he2e = zRT − zTRS (ZSS + ZRIS )
−1
zST exp (−jkRqp (ξ, z))
(1) Eqp (z) = ∂z2 + k 2


fp (ξ) dξ
−hp Rqp (ξ, z)
1 (·)T denotes the transpose operator. (3)

This paper's
Authorizedcopyright islimited
licensed use held to:
byQueen
the author(s). It isofpublished
Mary University in theseonproceedings
London. Downloaded March 12,2024 and included
at 22:01:25 UTC in any
from archive
IEEE Xplore. such as IEEE
Restrictions Xplore
apply.
under the license granted by the "Agreement Granting EurAAP Rights Related to Publication of Scholarly Work."
 with where cq = 1/sin (khq ), cqp = 2 cos (khp ) cq , and
sin (k (hp − |ξ|)) sin (k (hq − |z|)) Iqp (ξp ; s0 )
fp (ξ) = , fq (z) =
sin (khp ) sin (khq )  q 
2
(4) 2
Z 0 exp −jk ρqp + (z + zqp − ξp )
and = q exp (js0 kz) dz
−hq 2
ρ2qp + (z + zqp − ξp )
q
2
Rqp (ξ, z) = d2qp + (zq − zp + z − ξ) if p 6= q
q (5)  q
2

2 2
Rqp (ξ, z) = a2q + (z − ξ) if p=q Z +hq exp −jk ρqp + (z + zqp − ξp )
2 2
+ q exp (−js0 kz) dz
where d2qp = (xq − xp ) +(yq − yp ) . In (2)-(5), the following 0 ρ2qp + (z + zqp − ξp )
2

notation is utilized: hp and hq denote the half-length of the (9)

pth and qth reconfigurable element, respectively; aq denotes
the radius of the qth reconfigurable element; (xp , yp , zp ) and with ρ2qp = d2qp if p 6= q and ρ2qp = a2q if p = q, respectively,
(xq , yq , zq ) are the center locations of the pth and qth recon- and zqp = zq − zp .
figurable element, respectively, which are all aligned along the The integral in (9) can be formulated in terms of exponential
z-axis; k = 2π/λ where λ is the wavelength; η is the intrinsic integral functions by applying the change of variables in
impedance of free space; j is the imaginary unit; and ∂z2 is [20, Appendix G]. Specifically, let us consider the following
the second-order partial derivative with respect to z. notable integral:
The computation of zqp requires the calculation of the J (s0 , d0 , z0 ; L, U )
electric field in (3) and then the integral in (2). q  
2 2
Z U exp −jk d0 + (ζ − z0 )
= exp (−jks0 ζ) q dζ
2
A. Computation of (3)
L d20 + (ζ − z0 )
= s0 exp (−jks0 z0 ) E1 (jkL0 )
The electric field in (3) can be formulated in a closed-form − s0 exp (−jks0 z0 ) E1 (jkU0 )
expression by applying the approach proposed in [20, Section (10)
25.1]. The final result is the following:
where
k exp (−jkRqp (ξ = +hp , z)) q
Eqp (z) = 2
sin (khp ) Rqp (ξ = +hp , z) L0 = d2 + (L − z0 ) + s0 (L − z0 )
q 0 (11)
k exp (−jkRqp (ξ = −hp , z)) 2
U0 = d20 + (U − z0 ) + s0 (U − z0 )
+ (6)
sin (khp ) Rqp (ξ = −hp , z)
2k cos (khp ) exp (−jkRqp (ξ = 0, z)) and Z ∞
− exp (−u)
sin (khp ) Rqp (ξ = 0, z) E1 (c) = du (12)
c u
B. Computation of (2) is the exponential integral function with c being a complex
number whose phase is such that |arg (c)| < π.
Given the closed-form expression of the electric field in Then, Iqp (ξp ; s0 ) can be formulated as follows:
(6), zqp can be formulated in terms of exponential integral
Iqp (ξp ; s0 ) = J (−s0 , ρqp , ξp − zqp ; −hq , 0)
functions. Specifically, we first rewrite the function fq (z) as (13)
follows: + J (+s0 , ρqp , ξp − zqp ; 0, +hq )
1 The integral in (13) can be applied to any setup provided
fq (z) =
2j sin (khq ) that ρqp 6= 0, because E1 (c) is not defined for c = 0. If ρqp =
X (7) 0, e.g., the scattering elements are in a collinear formation [21,
× s0 exp (js0 khq ) exp (−js0 k |z|)
Fig. 8.20(b)], a specific closed-from expression is available
s0 ={−1,+1}
[21, Eq. (8-72)]. Alternatively, (9) can be directly computed
Then, zqp can be rewritten as follows: numerically.
ηcq X
zqp = s0 exp (js0 khq ) Iqp (ξp = +hp ; s0 ) III. C ONCLUSION
8π
s0 ={−1,+1}
ηcq X In this paper, we have discussed the engineering relevance of
+ s0 exp (js0 khq ) Iqp (ξp = −hp ; s0 ) considering reconfigurable surfaces whose scattering elements
8π
s0 ={−1,+1} are spaced more densely than half-wavelength, in order to
ηcqp X
model the presence of evanescent waves in the close vicinity of
− s0 exp (js0 khq ) Iqp (ξp = 0; s0 )
8π the surface, and to be able to control them for realizing extreme
s0 ={−1,+1}
(8) manipulations of the electromagnetic waves with high power

This paper's
Authorizedcopyright islimited
licensed use held to:
byQueen
the author(s). It isofpublished
Mary University in theseonproceedings
London. Downloaded March 12,2024 and included
at 22:01:25 UTC in any
from archive
IEEE Xplore. such as IEEE
Restrictions Xplore
apply.
under the license granted by the "Agreement Granting EurAAP Rights Related to Publication of Scholarly Work."
 efficiency. In these implementations, the mutual coupling [7] N. S. Perovic, L. -N. Tran, M. Di Renzo, and M. F. Flanagan, “Achiev-
among the scattering elements cannot be ignored. Departing able rate optimization for MIMO systems with reconfigurable intelligent
surfaces”, IEEE Trans. Wireless Commun., vol. 20, no. 6, pp. 3865-3882,
from the approach recently introduced in [16] for modeling June 2021.
the mutual coupling among closely-spaced scattering elements, [8] X. Qian, M. Di Renzo, J. Liu, A. Kammoun, and M.-S. Alouini,
we have shown that the end-to-end channel transfer function “Beamforming through reconfigurable intelligent surfaces in single-user
MIMO systems: SNR distribution and scaling laws in the presence of
matrix can be formulated in a closed-form expression in terms channel fading and phase noise”, IEEE Wireless Commun. Letters, vol.
of exponential integral functions, which are built-in functions 10, no. 1, pp. 77-81, Jan. 2021.
in state-of-the-art programming and numeric computing plat- [9] M. Di Renzo, F. Habibi Danufane, X. Xi, J. de Rosny, and S. Tretyakov,
“Analytical modeling of the path-loss for reconfigurable intelligent
forms. This greatly simplifies the analytical calculation of surfaces – Anomalous mirror or scatterer”, Int. Workshop on Signal
the end-to-end channel transfer function matrix in [16], and Processing Advances in Wireless Communications, 2020, pp. 1-5.
the corresponding optimization of the tuning elements of the [10] F. H. Danufane, M. Di Renzo, J. de Rosny, and S. Tretyakov, “On the
path-loss of reconfigurable intelligent surfaces: An approach based on
surface for maximizing the system performance [18], [19]. Green’s theorem applied to vector fields”, IEEE Trans. Commun., vol.
69, no. 8, pp. 5573-5592, Aug. 2021.
ACKNOWLEDGMENT [11] A. Dıaz-Rubio, V. S. Asadchy, A. Elsakka, and S. A. Tretyakov, “From
This work was supported by the European Commission the generalized reflection law to the realization of perfect anomalous
reflectors”, Science Advances, vol. 3, no. 8, p. e1602714, 2017.
through the H2020 ARIADNE (grant 871464) and RISE-6G [12] M. Moccia, S. Liu, R. Y. Wu, G. Castaldi, A. Andreone, T. J. Cui,
(grant 101017011) projects, and the Fulbright Foundation. and V. Galdi, “Coding metasurfaces for diffuse scattering: Scaling laws,
bounds, and suboptimal design”, Advanced Optical Materials, vol. 5,
R EFERENCES no. 19, p. e1700455, 2017.
[13] W. Yu, W. Li, A. Elsherbeni, Y. Rahmat-Samii Advanced Computational
[1] M. Di Renzo, A. Zappone, M. Debbah, M.-S. Alouini, C. Yuen, J. de Electromagnetic Methods and Applications, Artech House, 2015.
Rosny, and S. A. Tretyakov, “Smart radio environments empowered by [14] A. Pizzo, A. d. J. Torres, L. Sanguinetti, and T. L. Marzetta, “Nyquist
reconfigurable intelligent surfaces: How it works, state of research, and sampling and degrees of freedom of electromagnetic fields”, IEEE Trans.
the road ahead”, IEEE J. Sel. Areas Commun., vol. 38, no. 11, pp. 2450- Sig. Process., vol. 70, pp. 3935-3947, June 2022.
2525, Nov. 2020. [15] T. L. Marzetta, “Super-directive antenna arrays: Fundamentals and new
[2] M. Di Renzo and L. Song, “Reflection probability in wireless networks perspectives”, IEEE Asilomar Conf. Signals, Systems, Computers, pp.
with metasurface-coated environmental objects: An approach based on 1-4, Nov. 2019.
random spatial processes”, J. Wireless Commun. Networks, 2019, 99 [16] G. Gradoni and M. Di Renzo, “End-to-end mutual coupling aware
(2019). communication model for reconfigurable intelligent surfaces: An
[3] Z. Wan, Z. Gao, F. Gao, M. Di Renzo, and M. -S. Alouini, “Terahertz electromagnetic-compliant approach based on mutual impedances”,
massive MIMO with holographic reconfigurable intelligent surfaces”, IEEE Wireless Commun. Lett., vol. 10, no. 5, pp. 938-942, 2021.
IEEE Trans. Commun., vol. 69, no. 7, pp. 4732-4750, July 2021. [17] M. Johnson, P. Bowen, N. Kundtz, and A. Bily, “Discrete-dipole
[4] M. Di Renzo, A. Ahmed, A. Zappone, V. Galdi, G. Gradoni, M. Moccia, approximation model for control and optimization of a holographic
and G. Castaldi, “Digital reconfigurable intelligent surfaces: On the metamaterial antenna”, Appl. Opt., vol. 53, pp. 5791-5799, 2014.
impact of realistic reradiation models”, arXiv, 2022. [Online]. Available: [18] X. Qian and M. Di Renzo, “Mutual coupling and unit cell aware
https://arxiv.org/abs/2205.09799. optimization for reconfigurable intelligent surfaces”, IEEE Wireless
[5] C. Pan, G. Zhou, K. Zhi, S. Hong, T. Wu, Y. Pan, H. Ren, M. Di Renzo, Commun. Lett., vol. 10, no. 6, pp. 1183-1187, 2021.
A. L. Swindlehurst, R. Zhang, and A. Y. Zhang, “An overview of signal [19] A. Abrardo, D. Dardari, M. Di Renzo, and X. Qian, “MIMO interference
processing techniques for RIS/IRS-aided wireless systems”, IEEE J. Sel. channels assisted by reconfigurable intelligent surfaces: Mutual coupling
Topics Signal Process., vol. 16, no. 5, pp. 883-917, Aug. 2022. aware sum-rate optimization based on a mutual impedance channel
[6] M. Di Renzo, F. H. Danufane, and S. Tretyakov, “Communication mod- model”, IEEE Wireless Commun. Lett., vol. 10, no. 12, pp. 2624-2628,
els for reconfigurable intelligent surfaces: From surface electromagnetics 2021.
to wireless networks optimization”, Proc. of the IEEE, vol. 110, no. 9, [20] S. J. Orfanidis, Electromagnetic Waves and Antennas, ECE Department,
pp. 1164-1209, Sep. 2022. Rutgers University, 2016.
[21] C. A. Balanis, Antenna Theory: Analysis and Design, Wiley, 2nd edition,
1996.

This paper's
Authorizedcopyright islimited
licensed use held to:
byQueen
the author(s). It isofpublished
Mary University in theseonproceedings
London. Downloaded March 12,2024 and included
at 22:01:25 UTC in any
from archive
IEEE Xplore. such as IEEE
Restrictions Xplore
apply.
under the license granted by the "Agreement Granting EurAAP Rights Related to Publication of Scholarly Work."