PHYSICAL REVIEW B 91, 195444 (2015)
Terahertz carpet cloak based on a ring resonator metasurface
B. Orazbayev,1,* N. Mohammadi Estakhri,2 M. Beruete,1 and A. Alù2
1
2
Antennas Group-TERALAB, Universidad Pública de Navarra, Campus Arrosadı́a, 31006 Pamplona, Spain
Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, Texas 78712, USA
(Received 18 March 2015; revised manuscript received 7 May 2015; published 29 May 2015)
In this work we present the concept and design of an ultrathin (λ/22) terahertz (THz) unidirectional carpet
cloak based on the local phase compensation approach enabled by gradient metasurfaces. A triangular surface
bump with center height of 4.1 mm (1.1λ) and tilt angle of 20° is covered with a metasurface composed of
an array of suitably designed closed ring resonators with a transverse gradient of surface impedance. The ring
resonators provide a wide range of control for the reflection phase with small absorption losses, enabling efficient
phase manipulation along the edge of the bump. Our numerical results demonstrate a good performance of
the designed cloak in both near field and far field, and the cloaked object mimics a flat ground plane within a
broad range of incidence angles, over 35° angular spectrum centered at 45°. The presented cloak design can be
applied in radar and antenna systems as a thin, lightweight, and easy to fabricate solution for radio and THz
frequencies.
DOI: 10.1103/PhysRevB.91.195444
PACS number(s): 42.25.Fx, 41.20.Jb, 78.67.Pt
I. INTRODUCTION
Metamaterials have opened new directions to tailor at
will the intrinsic electromagnetic parameters of a composite,
such as its permittivity and permeability, providing interesting
solutions for one of the oldest quests of electromagnetism—
controlling electromagnetic waves at will [1,2]. Many devices
have been proposed based on metamaterials, ranging from
lenses [3,4] and beam steering structures [5] to cloaking
devices [6–9] able to hide objects from an external observer.
After years of intensive studies, several cloaking mechanisms
and designs have been proposed, mostly among the two leading
categories of transformation optics [10,11] and scattering
cancellation [12–14]. The techniques based on transformation
optics impose strong demands on bulk metamaterial designs,
such as a specific profile of inhomogeneity and anisotropy of
the material parameters. These constraints make such cloaking
devices difficult to realize in practice, due to their high
complexity. The scattering cancellation cloaks are simpler to
realize and more robust [15], yet, like any passive cloaking
technique, they suffer from fundamental limitations on the
size of the object to be cloaked [16].
A cloaking approach that relaxes the causality limitations
on size and bandwidth is known as carpet cloaking or ground
cloaking, with the idea of hiding in reflection a bump on a mirror. This problem has gained the interest of many researchers,
due to its inherently relaxed constraints, its simplified design,
and wide range of applications. Transformation-based carpet
cloaks exploit quasiconformal mapping [17,18], which allows
to minimize the anisotropy of the required material and can
be implemented using isotropic dielectrics [19], simplifying
the design and minimizing absorption losses. However, the
proposed cloak is still volumetric and non-trivial to realize.
Another important hurdle for practical applications of carpet
cloaks is the lateral shift they introduce under the isotropic
approximation. Unfortunately, the introduced lateral shift is
comparable to the case when a ground plane is placed above
*
Corresponding author: b.orazbayev@unavarra.es
1098-0121/2015/91(19)/195444(5)
the cloaked object to suppress its scattering. Since the cloaking
medium is typically denser than free space, the beam inside the
cloak is refracted into a smaller angle. However, if we require
that a finite-size beam emerges from the cloak at the same
location as it would when reflected by a flat ground plane, it
should be refracted into a larger angle, which is only possible
in anisotropic materials. This lateral shift presents a serious
problem, since an external observer still can notice that the
beam emerges from a different point [20] and it also begs the
question of whether another flat reflector on top of the bump
would not provide a simpler solution to the problem.
To overcome the issues associated with conventional
ground cloaking approaches, recently it has been proposed that
covering a bump with a specially designed surface can reduce
unwanted scattering from an arbitrary bump, creating an
effective ultrathin cloaking surface [21,22]. The metasurface
is used to build a phase distribution on the bump edge equal to
the phase response created upon reflection from a conducting
ground plane, i.e., when no bump is present. In this case, for
the external observer the reflected wave will have the same
phase distribution as if it were reflected from the ground
plane, creating an ultrathin and relatively simple cloaking
configuration for practical implementation. Moreover, due to
surface phase compensation, the metasurface cloak does not
create a lateral shift [21]. Obviously this approach is dependent
on the object to be hidden and the illumination, but, given
its robustness, it may provide a viable solution for several
practical applications, as we discuss in the following.
In this work, we propose an ultrathin metasurface-based
carpet cloak suitable for THz regime, we report successful
concealment of an electrically large object in the near and far
field, and we provide a thorough analysis of the angle and
frequency dependence of the device performance. Our cloak
is based on the concept originally introduced in Refs. [21,22]
applied to a triangular bump with center height of 4.1 mm
(1.1λ0 ) at 80 GHz. However, in Ref. [21] the cloak is designed
for a normal incidence with small range of the incident angles
(∼5°). In contrast to the results in Ref. [22] we obtain the carpet
cloak for an electrically large object—the height of the bump
is 4.1 mm (1.1λ0 ). We demonstrate a good performance within
195444-1
©2015 American Physical Society
PHYSICAL REVIEW B 91, 195444 (2015)
ORAZBAYEV, MOHAMMADI ESTAKHRI, BERUETE, AND ALÙ
h
(a)
y
θ
g
β
z
x
w
β = θ+ψ
β = θ-ψ
metasurface
θ
Ey
h
l
ψ
FIG. 1. (Color online) Scheme of the carpet cloak with metasurface.
(b)
(c)
35° range of the incidence angle and a frequency bandwidth
of 8 GHz, with 10% fractional bandwidth (FBW).
II. CARPET CLOAK DESIGN
Figure 1 shows the general scheme of the cloak and of the
object. The incoming oblique wave, with angle of incidence
θ (with respect to the horizontal ground plane), illuminates a
PEC bump with a tilt angle ψ. Any arbitrary object we aim
to conceal can be placed inside the bump, as long as it fits
in its volume. The goal is to create a field distribution on
the external boundary of the object (dashed line), identical
to the case when no bump is presented, in view of the field
equivalence principle. This technique is different from carpet
cloaks based on the quasiconformal mapping, where the object
is concealed by controlling the propagation of the incident
waves and effectively isolating the hidden region from the
incident wave. Here, reconstruction of the field can be done by
introducing an abrupt phase variation on the boundary, which
can be calculated at each point of the bump’s edge as Ref. [22]:
δ = π − 2k0 h cos θ
(1)
where k0 is the free space wave vector at the operation
frequency, h is the height of the unit cell center from the
ground plane, and θ is the angle of incidence of the incoming
wave with respect to the back-plane normal.
III. NUMERICAL RESULTS
A. Cloaking metasurface
In order to obtain the required phase distribution, a
metasurface based on pairs of closed ring resonators (CRR)
is used, whose unit cell is shown in the inset of Fig. 2(a).
A clear advantage of such topology is its insensitivity to
the polarization of the incident electromagnetic field. To
realize the effect, we need to control the phase response of the
reflected wave from each block over the entire 2π phase range.
To this purpose, we simulated the reflection from infinite
planar arrays of such closed rings using the commercial
software CST Microwave StudioTM [23], using unit cell
boundaries and frequency-domain solver. A fine tetrahedral
mesh was chosen with maximum edge length 0.285 mm
(0.076λ0 ) and minimum edge length 0.0007 mm (0.0002λ0 ).
To create a high-resolution surface with better control over the
phase distribution of the reflected beam, the lateral dimension
FIG. 2. (Color online) (a) Phase response of the unit cell for
different angles of incidence. (Inset indicates the unit cell geometry
and its corresponding parameters.) Color-map for the amplitude (b)
and phase (c) of the reflection coefficient as a function of the incidence
angle and radius of the inner ring.
of the unit cell was fixed at 400 μm ≈ λ0 /10, for the working
frequency f = 80 GHz (λ0 = 3.75 mm). Each unit cell
consists of two concentric metallic rings with a fixed width
w = 10 μm, separated by a gap g = 10 μm. The radius of
the outer ring is then found as r2 = r1 + w + g, where r1 is
the radius of the inner ring. The rings are separated from the
ground plane by a thin silicon layer of thickness h = 165 μm
(≈λ0 /22) with dielectric permittivity εr = 11.2 and loss
tangent tanδ = 4.7 × 10−6 . The metal used for the rings is
aluminum with a conductivity σAl = 3.56 × 107 S/m.
Due to the geometry of the carpet cloak, the incoming wave
has two possible incidence angles (β) on each block of the
metasurface, depending on which side of the bump the block
is located. In our particular case for an obliquely incident wave
with angle θ = 45° and tilt angle of the bump ψ = 20°, the
incidence angles at the bump edges are β1 = θ –ψ = 25◦ and
β2 = θ + ψ = 65◦ for the left and right side, respectively.
The amplitude and phase of the reflection coefficient was
obtained as a function of radius of the inner ring and the
angle of incidence β, shown in Figs. 2(b) and 2(c). In order to
increase the design precision, instead of the normal reflection
coefficients, here we design the cloaking layers on each side
of the bump based on the data calculated for the corresponding
incidence angles. As observed in Figs. 2(a) and 2(b), the
amplitude of the reflected beam slightly drops around r1 =
135 μm, due to the CRR resonance. Apart from the resonance,
however, the metasurface is operating almost as an ideal mirror
with close to unitary efficiency. Fig. 2(a) also shows the phase
response of the unit cell for two incidence angles β1 (solid
blue line) and β2 (dash dotted blue line). The range of phase
195444-2
PHYSICAL REVIEW B 91, 195444 (2015)
TERAHERTZ CARPET CLOAK BASED ON A RING . . .
variation attainable by changing the inclusion radius spans
almost 2π , confirming that this unit cell can adequately control
the local phase response of the cloaking metasurface. It is also
clearly seen that for a larger incidence angle the phase response
has a steeper slope and, therefore, it is more sensitive to the
radius variation of the rings, as it may be expected. This means
that a bump with a larger tilt angle requires a metasurface with
a smaller variation of radii, in the order of a few μm.
B. Final design
Once the phase response of the unit cell has been characterized, it is possible to put together the carpet cloak using the
design equation (1), in order to hide a perfectly conducting
bump with triangular shape lying in the xz plane and infinite
in the transverse y direction. For practical realizations, it
is preferred to cloak electrically large objects with bigger
tilt angles, which allows us to more efficiently utilize the
space under the cloak. Extreme shapes and large corner
angles may require nonlocal and active surfaces [24], but for
slowly varying configurations, including the current proposed
design, surface phase engineering is adequate [21]. In this
example, the ground cloak was designed for a bump with
a tilt angle ψ = 20°, height of 4.1 mm (1.1λ0 ), edge
length of 12 mm (3.2λ0 ) and base length 22.5 mm (6λ0 ).
Full-wave simulations of the structure were performed using
the transient solver CST Microwave StudioTM . The structure
was illuminated by an obliquely incident (θ = 45°) Gaussian
beam with TE polarization. To this end, an array of electric
dipoles was used with Gaussian distribution of amplitudes,
providing a quasi-Gaussian beam excitation. The ground plane
was emulated by using an electric boundary (perfect electric
conductor) in the xy plane (z = 0). Given the symmetry
of the structure, an electric symmetry was applied in the
xz plane (y = l/2) in order to reduce computational time.
A fine hexahedral mesh was applied with minimum cell
length of 0.1 mm (0.026λ0 ) and maximum of 0.44 mm
(0.112λ0 )
As explained above, the phase response of the each block
was obtained using unit cell boundaries, assuming that all cells
in each simulation have the same parameters. On the contrary,
to successfully mimic the ground plane, the radii of the rings
must change according to the phase distribution determined
by (1). The transverse inhomogeneities modify the mutual
coupling between adjacent blocks and, therefore, its response
to the incident wave. Hence, an optimization procedure is
required to fine tune the design based on (1), imparting a
local variation to the radii of surface blocks (60 blocks in the
current design). Due to the reciprocity principle [25], we need
to optimize only half of the bump, so only 30 unit cells need
to be considered in the optimization process.
Figures 3(a)–3(c) show the spatial distribution of the
electric field (Ey component) at the operation frequency f0 =
80 GHz for three cases: ground plane [(Fig. 3(a)], bare bump
[Fig. 3(b)], and cloaked bump [Fig. 3(c)]. As it can be seen,
when an irregularity is introduced on the ground plane the
Gaussian beam is scattered over a wide range of angles.
After the cloak is applied, the near-field distribution of the
reflected wave is restored to the original Gaussian beam. The
small disturbance of the reflected beam is caused by the finite
FIG. 3. (Color online) Electric field distribution on xz plane for
(a) ground plane, (b) bare bump, and (c) cloaked bump.
discretization of the cloaking metasurface and high (β = 65°)
incidence angle at the second edge of the bump, which may be
mitigated with active cloaking surfaces [24]. As it was shown
in Fig. 2(a) absorption losses are higher for high angles of
incidence, provoking higher scattering level. Despite all these
factors, Fig. 4 demonstrates that we are able to obtain a similar
far-field radiation pattern from the cloaked beam as if a bare
ground plane were interacting with the incident wave.
FIG. 4. (Color online) Radiation pattern for the reflected Gaussian beam from the ground plane (dash-dotted blue line), the bump
without cloak (dotted green line), and from the cloaked bump (solid
red line).
195444-3
PHYSICAL REVIEW B 91, 195444 (2015)
ORAZBAYEV, MOHAMMADI ESTAKHRI, BERUETE, AND ALÙ
(a)
(d)
root mean square error (RMSE) function, which determines
the goodness of the fit, and can be defined as [26]:
N
1
[Es (f,θ,ϕi ) − Eb (f,θ,ϕi )]2 (2)
RMSE(f,θ ) =
N i=1
(c)
(b)
(e)
(f)
FIG. 5. (Color online) Amplitude of the far-field component of
the electric field as a function of azimuth angle ϕ and frequency
(a)–(c) and as a function of the incidence angle θ and azimuth angle
ϕ (d)–(f). Left column refers to the beam reflected from the ground
plane, center from the bare bump, and right column from the bump
covered with the cloak.
C. Carpet cloak bandwidth
Next, the performance of the optimized cloak was analyzed
in terms of the angle and frequency bandwidth. For this study,
the numerical simulations were run within a frequency span
from 75–85 GHz with a step of 0.2 GHz and steering the
incidence angle from 25°–60° with a step of 1°. The resulting
color maps for the far-field scattering electric field magnitude
as a function of azimuth angle and frequency are shown in
Figs. 5(a)–5(c). The analogous color maps for the far-field
magnitude as a function of azimuth angle and incident angle θ
of the incoming wave are shown in Figs. 5(d)–5(f). The cloak
works close to ideally at the designed frequency and angle
of incidence, yet, and in spite of the original unidirectional
design, it is able to significantly reduce the scattering level
in the whole simulated range of frequencies and incidence
angles. The multilobe pattern created in the presence of the
bump is converted into a directive beam, as desired for an
ideal ground plane. In order to quantitatively define the region
over which the beam reconstruction is acceptable we use a
where N is the number of sample azimuth angles, Es (f,θ,ϕi )
and Eb (f,θ,ϕi ) are the far-field magnitude of the reflected
wave from the bump and the ground plane at given azimuth angle ϕi , frequency f , and incident angle θ . The
calculated RMSE(f,θ ) for the scattered beam without and
with the cloak are shown in Figs. 6(a) and 6(b). A
sufficiently accurate fit is considered for the RMSE less
than 10% [27]. Hence, our parametric study reveals that
the ground cloak maintains a reasonable performance in
a frequency span of about 8 GHz (fractional bandwidth
FBW = 10%) and angular span of 35°. In addition, it is
noticeable that the cloak has a wider bandwidth for the
lower incident angles, which is in good agreement with our
previous study for the phase response as a function of the
incident angle β. For applications in which the accuracy of the
recovered beam is important, for example for high-precision
measurements, an RMSE below 5% [dashed contour line
in Fig. 6(b)] would still provide a bandwidth of 2.2 GHz
(FBW = 3%) and angular span of 10°.
These results have been obtained for relatively small
substrate losses (tanδ = 4.7 × 10−6 or σ = 2.5 × 10−4 S/m).
However, in practical realizations the losses can be several
orders higher and may disrupt the cloak operation. Therefore,
an analysis of the performance of the cloak was done for
higher values of dielectric losses. The recovery of the Gaussian
beam at the operational frequency 80 GHz was estimated for
different values of conductivity, using the same RMSE metric,
and corresponding results are plotted in Fig. 6(c). Interestingly,
even for very low RMSE (<5%) the cloaking metasurface is
able to recover the far-field pattern of the original reflected
Gaussian beam for values of conductivity up to 1.2 S/m
(tanδ = 0.023), as it may be expected due to the inherently
nonresonant nature of the proposed cloaking technique. Silicon
samples with an equivalent value of resistivity ρ = 83 cm
are relatively cheap and commercially available, making
the proposed cloaking device appealing for low-cost THz
(b)
(a)
Uncloaked
(c)
Cloaked
FIG. 6. (Color online) RMSE distribution vs incident angle and frequency for (a) uncloaked and (b) cloaked bump. (c) RMSE distribution
as a function of the conductivity σ of dielectric substrate and frequency.
195444-4
PHYSICAL REVIEW B 91, 195444 (2015)
TERAHERTZ CARPET CLOAK BASED ON A RING . . .
devices such as automotive radar systems, to reduce unwanted
scattering from electrically large objects.
in radar and antenna systems, where other techniques may be
unpractical due to their excessively large volumes.
IV. CONCLUSIONS
ACKNOWLEDGMENTS
To conclude, in this paper we have presented a thorough
numerical study of an ultrathin (λ/22) carpet cloak based
on a smartly engineered graded metasurface. The proposed
cloaking metasurface is polarization insensitive and has a
simple geometry, which facilitates the fabrication and reduces
cost. Moreover, the proposed cloak demonstrates an acceptable
beam reconstruction over a good angular span of 35° and a
moderately broad bandwidth FBW = 10%. Such a cloak can
conceal electrically large objects and may find applications
This work was supported in part by the Spanish Government under Contract Consolider Engineering Metamaterials
CSD2008-00066 and Contract TEC2011-28664-C02-01. B.O.
is sponsored by the Spanish Ministerio de Economı́a y
Competitividad under Grant No. FPI BES-2012-054909. M.B.
is sponsored by the Spanish Government via RYC-201108221. N.M.E. and A.A. have been supported by the NSF
CAREER Award No. ECCS-0953311 and the AFOSR Grant
No. FA9550-13-1-0204.
[1] N. Engheta and R. W. Ziolkowski, Metamaterials: Physics and
Engineering Explorations (Wiley, Hoboken, 2006).
[2] L. Solymar and E. Shamonina, Waves in Metamaterials (Oxford
University Press, Oxford, 2009).
[3] J. B. Pendry, Phys. Rev. Lett. 85, 3966 (2000).
[4] Z. Liu, H. Lee, Y. Xiong, C. Sun, and X. Zhang, Science 315,
1686 (2007).
[5] V. Pacheco-Peña, V. Torres, B. Orazbayev, M. Beruete, M.
Navarro-Cı́a, M. Sorolla Ayza, and N. Engheta, Appl. Phys.
Lett. 105, 243503 (2014).
[6] D. Schurig, J. J. Mock, B. J. Justice, S. A. Cummer, J. B. Pendry,
A. F. Starr, and D. R. Smith, Science 314, 977 (2006).
[7] M. Silveirinha, A. Alù, and N. Engheta, Phys. Rev. E 75, 036603
(2007).
[8] D. Shin, Y. Urzhumov, Y. Jung, G. Kang, S. Baek, M. Choi,
H. Park, K. Kim, and D. R. Smith, Nature Commun. 3, 1213
(2012).
[9] N. Landy and D. R. Smith, Nature Mater. 12, 25 (2013).
[10] U. Leonhardt, Science 312, 1777 (2006).
[11] J. B. Pendry, D. Schurig, and D. R. Smith, Science 312, 1780
(2006).
[12] A. Alù and N. Engheta, Phys. Rev. E 72, 016623 (2005).
[13] A. Alù, M. G. Silveirinha, and N. Engheta, Phys. Rev. E 78,
016604 (2008).
[14] D. Rainwater, A. Kerkhoff, K. Melin, J. C. Soric, G. Moreno,
and A. Alù, New J. Phys. 14, 013054 (2012).
[15] A. Alù and N. Engheta, Opt. Express 15, 3318 (2007).
[16] F. Monticone and A. Alù, Phys. Rev. X 3, 041005 (2013).
[17] J. Li and J. B. Pendry, Phys. Rev. Lett. 101, 203901
(2008).
[18] R. Liu, C. Ji, J. J. Mock, J. Y. Chin, T. J. Cui, and D. R. Smith,
Science 323, 366 (2009).
[19] P. Zhang, M. Lobet, and S. He, Opt. Express 18, 18158 (2010).
[20] B. Zhang, T. Chan, and B.-I. Wu, Phys. Rev. Lett. 104, 233903
(2010).
[21] J. Zhang, Z. Lei Mei, W. Ru Zhang, F. Yang, and T. Jun Cui,
Appl. Phys. Lett. 103, 151115 (2013).
[22] N. M. Estakhri and A. Alu, IEEE Antennas Wirel. Propag. Lett.
13, 1775 (2014).
[23] CST Computer Simulation Technology AG, www.cst.com,
(2015).
[24] N. Mohammadi Estakhri, C. Argyropoulos, and A. Alù, Philos.
Trans. R. Soc. London (to be published).
[25] D. L. Sounas and A. Alù, Opt. Lett. 39, 4053 (2014).
[26] I. Moreno and C.-C. Sun, Opt. Express 16, 1808 (2008).
[27] C. Daniel and F. S. Wood, Fitting Equations to Data: Computer
Analysis of Multifactor Data for Scientists and Engineers
(Wiley, New York, 1971).
195444-5
View publication stats