4998 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO.

11, NOVEMBER 2012

A Circularly-Polarized Isoflux Antenna Based on

Anisotropic Metasurface
Gabriele Minatti, Stefano Maci, Fellow, IEEE, Paolo De Vita, Angelo Freni, Senior Member, IEEE, and
Marco Sabbadini

Abstract—Theory, design, realization and measurements of an This simple, cost-effective and extremely thin antenna is
X-band isoflux circularly polarized antenna for LEO satellite plat- founded on an inhomogeneous anisotropic metasurface. A
forms are presented. The antenna is based on a metasurface com- metasurface is a thin metamaterial layer backed by a ground
posed by a dense texture of sub-wavelength metal patches on a
grounded dielectric slab, excited by a surface wave generated by plane and characterized by unusual reflection and dispersion
a coplanar feeder. The antenna is extremely flat (1.57 mm) and properties. Metasurfaces are obtained by a dense periodic
light (less than 1 Kg) and represents a competitive solution for texture of small elements printed on a grounded slab with or
space-to-ground data link applications. without shorting vias [2], [3]. These have been used in the
Index Terms—Leaky wave antennas, metasurface antennas, sur- past to realize electromagnetic bandgaps (EBGs) or equivalent
face impedance. magnetic-walls [3].
As described in [2]–[5] when a metasurface is inhomoge-
neous, namely when its characteristic parameters vary in space,
I. INTRODUCTION
it is able to control the surface wave propagation path, as well
as polarization/radiation properties. This phenomenon, recently

E FFICIENT low-mass and low-cost antennas for payload

data handling and transmissions (PDHT) with optimal
shaped beams are a major asset for space missions. These
denoted “metasurfing” [5], [6], can produce high-gain pencil
beam with circularly polarized radiation [6], [5].
The antenna presented here makes use of a planar metasur-
antennas play an important role in Earth Observation, where face to provide an isoflux shaped beam with circular polariza-
high transmission rate is necessary to acquire Earth images. tion. Specifically, a cylindrical surface-wave (SW) is excited at
Examples of these antennas are recently manufactured by the center of the antenna by a small circularly polarized radiator,
Thales Alenia Space for RADARSAT and COSMO/SkyMed and it is gradually transformed into a leaky-wave (LW) by the
missions, but their design dates back about ten years [1]. interaction with the quasi-periodically modulated metasurface.
Future satellite missions will require new antennas with more The metasurface is anisotropic to control the polarization of the
demanding performances for the radiation pattern, especially radiated field. The implementation of this anisotropy is similar
in terms of cross-polarization discrimination. The antenna on to that presented in [4]; however it makes use of circular patches
the satellite platform shall distribute a uniform power density in place of square patches and the design criterion is different,
over a well-defined portion of the visible Earth surface. The essentially because here the beam generated by the antenna is
relevant shaped beam is referred to as isoflux pattern. LEO shaped.
satellite antennas shall provide EM isoflux coverage to the The paper is structured as follows. Section II presents the
Earth surface over a visibility cone with an apex angle of about isoflux mask used for LEO satellite platform. Section III
120 . This requirement is extremely difficult to satisfy for presents the criteria of conical radiation for isotropic uniform
many typologies of antennas, as a significant portion of the and non-uniform (modulated) continuous metasurfaces excited
radiation needs to be spread over a very large angular region. by a point source. Section IV extends the criteria to uniform
The prototype antenna we present here, despite the use of a anisotropic metasurfaces, while Section V is indeed devoted
new technology, closely satisfies the gain pattern requirements to modulated anisotropic metasurface and to the condition for
in the requested frequency band (8.5–8.7 GHz), while showing which the dominant LW contribution is circularly polarized.
excellent circular polarization over the wide angular range. Section VI specifies how to carry out the circularly polarized
anisotropic metasurface by small (sub-wavelength) patches.
Manuscript received January 16, 2012; revised April 08, 2012; accepted July Section VII describes the design procedure for the analysis and
02, 2012. Date of publication July 13, 2012; date of current version October 26, optimization of the prototype, and Section VIII presents its
2012.
G. Minnatti and S. Maci are with Department of Information Engineering,
manufacturing and measurements. Conclusions are drawn in
University of Siena, 53100 Siena, Italy (e-mail: macis@ing.unisi.it). Section IX.
P. De Vita is with the Ingegneria dei Sistemi, Pisa, Italy (e-mail: p.devita@ids.
it). II. ISOFLUX GAIN MASK
A. Fremi is with the Department of Electronics and Telecommunications,
University of Florence, Florence IT-50139, Italy (e-mail: freni@unifi.it). In our reference scenario, an antenna located on a LEO satel-
M. Sabbadini is with the European Space Agency, Noordwijk, The Nether- lite platform must radiate a beam with homogeneous power den-
lands (e-mail: sabbadini@esa.int).
sity at the earth surface (Fig. 1). According to our design speci-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. fications, a satellite platform S is at distance over the sea level
Digital Object Identifier 10.1109/TAP.2012.2208614 in an altitude range comprised between 500 and 1000 Km. The

0018-926X/$31.00 © 2012 IEEE

MINATTI et al.: A CIRCULARLY-POLARIZED ISOFLUX ANTENNA BASED ON ANISOTROPIC METASURFACE 4999

Fig. 2. Envelope of the isoflux pattern for several orbital altitudes. All the pat-
terns are comprised between upper and lower limit masks. The gain is consid-
ered as normalized with respect to the total input power.

Fig. 1. Geometrical description of the reference scenario for a LEO satellite

isoflux antenna. The power density radiated inside the visible cone must be ho-
mogenous at the ground level.

satellite is considered visible from a ground station when it is at

a minimum elevation angle .
To provide a homogeneous power density on ground, the an-
tenna must compensate the power loss due to the propagation
path by increasing the gain toward directions where the path is
longer. That is, the antenna gain pattern must satisfy the condi-
tion where is the maximum
distance between antenna and ground station, and is the obser-
vation angle versus the direction of the Earth centre. The relation Fig. 3. Increasing slope effect when augmenting the radius of an isoflux an-
tenna having field distribution given by (5).
between the distance and the other geometrical variables
is

(1) A. Conventional Definition of Aperture Efficiency

Implementing a planar solution for wide-angle isoflux an-
Solving (1) determines the relation between the satellite to tennas with a steep roll-off sacrifices the antenna dimension.
Earth surface distance, and the gain profile to provide a homo- The aperture efficiency defined in a conventional sense as
geneous power density on ground:
(4)
(2)
where is the aperture field distribution, loses its meaning for
where aperture antennas that should respect the mask in Fig. 2. The
wide coverage angle and the rapid roll-off at the coverage edge,
(3) imply extremely small efficiency when using (4). Actually, for
this application, it is by far more important to have a good
where and . For satellite drop-off than a high efficiency defined by (4). As an example,
altitudes between and Km, the relevant let us consider the case of isoflux beam realized by an aperture
isoflux patterns are comprised between a lower and an upper distribution of the type
limit gain masks as shown in Fig. 2.
The requirements are satisfied whenever the radiation pattern (5)
will be comprised inside the upper and lower limits in the op-
erational bandwidth 8.5–8.7 GHz. The cross polarization level where is the drop-off angle (in our case ) and
is requested to be lower than dB all over the angular range. is a parameter that adjusts the level at broadside (in our case
Roughly speaking, the radiation pattern is in fact conical with ). It is seen (Fig. 3) that to change the slope in the range
peak around 55 –65 , a steep roll-off outside 65 and a broad- 65 –75 from 5 dB/10 to 13 dB/10 the aperture radius must be
side level around 0 dB. We will obtain this condition by playing increased from 3 to 7 wavelengths. As a direct consequence the
simultaneously with the metasurface and with the direct feed ra- aperture efficiency for the aperture distribution (5) for a radius
diation. To this end, we analyze first the possibility to obtain a of is , showing how misleading the use of the
conical radiation by a metasurface. standard definition of efficiency is.
5000 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 11, NOVEMBER 2012

In fact, the efficiency in (4) does not give any information

about how close its pattern is to the desired isoflux one or, equiv-
alently, how close to the optimum is the distribution of (equiv-
alent) currents on the radiating aperture.

B. Weighted Tapering Efficiency

A figure of merit that better describes the antenna perfor-
mance for this kind of application is obtained by

(6) Fig. 4. Radial periodic modulation of impedance and identification of an ele-

mental sector with a one dimensional sinusoidal reactance problem.
which is denoted by weighted aperture efficiency. In (6)
is a reference (ideal) tapered field illumination which gives the
desired isoflux pattern as provided by the mask. It can be easily where and are the free-space wavenumber and impedance,
seen, from the Schwarz inequality with a scalar integral product respectively. Equation (8) implies a relation between reactance
over the aperture, that this efficiency is always less than or equal and propagation constant: the larger the average reactance, the
to unity. It is unity when and only when . Whenever shorter the SW wavelength.
the reference aperture distribution is chosen as uniform over the
aperture, recovers the standard efficiency definition in (4). B. Cylindrically Modulated Periodic Reactance
By using as a reference aperture distribution the inverse Fourier
transform of (3) (with km), the Assume now a periodic radial modulation of the reactance
weighted efficiency relevant to the distribution (5) with with period (this will be implemented later by changing the
radius is , which is a better figure of merit for the dimensions of the printed patches still maintaining them very
distribution under the intended purpose. small in terms of a wavelength). We assume that the reactance
variation is described by the sinusoidal function
III. ISOTROPIC UNIFORM AND MODULATED CONTINUOUS
(9)
METASURFACE
The metasurface that forms the radiating part of the antenna where is the average value, is the modulation index, and
is made up with a thin grounded dielectric slab periodically . It is expected that the interaction of the SW with
loaded by sub-wavelength printed patches. We assume here the radial sinusoidal modulated reactance transforms the SW in
that the printed elements are so small that the set of patches a LW, thus producing a conical radiation. As in [5], the structure
plus grounded slab can be equivalently modeled by continuous may be thought as obtained by rotating a one-dimensional peri-
boundary conditions of reactive type, which we suppose located odic surface around the axis (Fig. 4). The interaction between
in and, for the moment, to be isotropic. We consider the cylindrical SW and the cylindrical modulated reactance is
first a uniform inductive reactance, and next a cylindrically therefore locally modeled by the interaction between a plane
modulated inductive reactance. SW and a surface reactance sinusoidally modulated in one di-
mension. Oliner and Hessel solved the latter problem in ‘59 [7].
A. Uniform Impedance of Inductive Reactance Type Their results are briefly summarized below for the reader’s con-
Let us first assume uniform impedance boundary conditions venience.
of inductive reactance type; i.e., assume a relationship between Modulating the surface reactance with a sinusoidal function
the tangential components of the electric and magnetic fields of as in (9) produces a small perturbation on both the real and imag-
the type where the reactance inary parts of obtained for a uniform surface reactance (i.e.,
is a positive scalar quantity. In this case, the structure sup- ); that is, . In [7], a simple
ports the propagation of a cylindrical bounded SW, whose algorithm to calculate is presented; this algorithm
electric and magnetic tangent fields at the metasurface are given shows that depends on the normalized impedance
by , the modulation index , the period .
For the problem described by the radial periodic impedance in
(9), the aperture SW field in the homogeneous reactance trans-
forms in
(7)

where denotes a Hankel function of the second kind of (10)

order one, and are the unit vectors of the radial coordinates.
The wavenumber can be found by imposing a transverse This field can be represented by a series of n-indexed Floquet
resonance with the free space TM impedance [5]; leading to waves whose wavenumbers are given by
. We assume to be the one found by the
(8) Oliner-Hessel procedure. Among all the Floquet modes, those
MINATTI et al.: A CIRCULARLY-POLARIZED ISOFLUX ANTENNA BASED ON ANISOTROPIC METASURFACE 5001

Fig. 5. Normalized average surface reactance as a function of the average pe-

riodicity to obtain a single beam in a direction for both forward and
backward modes. Curves are referred to beams ranging from 40 to 70 off axis
angle. Fig. 6. Behavior of as a function of for different pointing angles
in the backward configuration and for different values of
modulation index (0.3, 0.4, 0.5).
having where k is the free-space wavenumber,
namely

(11) Being the structure cylindrical, a conical radiation with aper-

ture angle is obtained for both cases in which the elemental
are LWs and contribute to the radiating beam. It is worth noting sector radiation is backward or forward, provided that each el-
that (11) can be satisfied only for negative-index Floquet modes. emental sector has a single beam. On the one hand, the for-
The dominant LW mode has index and leads to a beam ward mode operation implies a larger period for a fixed level of
pointing at defined by impedance, which is a desirable property for the reconstruction
of gradual impedance variation with small patches. On the other
(12) hand, being the period equal, the forward mode requires much
higher impedance. For instance, a single beam at re-
Depending on whether the beam is directed in the same or in
quires . Namely, it requires a surface reactance
the opposite sense with respect to the exciting SW propagation
around a thousand Ohms, which is difficult to implement in
direction, the LW is referred to as forward LW,
printed technology. For this reason, we will use here a backward
or backward LW,
mode radiation. It is also clear from Fig. 5 that low reactance
, respectively [7]. By analogy, we will refer to the relevant
levels imply a more sensitive variation of the beam angle with
radiated beam as forward or backward beam.
frequency (i.e., a smaller antenna bandwidth). The choice of the
It is important to find when the indexed LW is the only
impedance surface is actually a compromise between having a
radiating mode (we refer to this condition as unimodal regime).
surface with large reactance values and a period as large as pos-
We start for simplicity by considering for which
sible to have a smooth variation. A good compromise is found
can be neglected in (12). From (11) and (12) it is straightforward
for .
to demonstrate the following:
Fig. 6 presents as a function of for different pointing
• a single backward beam in direction is produced by a
angles in the backward configuration and
surface impedance given by
for different values of the modulation index. The curves are ob-
tained using the Oliner-Hessel method. The values of in-
(13)
creases for increasing modulation index and for decreasing ,
when where is the free space wave- as expected; however, it remains within acceptable small values.
length; The deviation obtained by the Oliner-Hessel method, pre-
• a single forward beam in direction is produced by a sented in Fig. 7, has a very minor impact in comparison with
surface impedance . For instance, in the case of modulation index
and unit normalized impedance, one obtains
(14) .
Using an isotropic metasurface, the conical beam cannot be
when or circularly polarized. Control of the polarization implies the use
Fig. 5 shows the normalized average reactance versus the nor- of an anisotropic metasurface. In the following, we first intro-
malized periodicity for several ranging from 40 to 70 re- duce an anisotropic uniform metasurface and continue next with
sulting in a backward or a forward single beam. an anisotropic modulated metasurface.
5002 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 11, NOVEMBER 2012

which has been cast in reactive form since we are looking for a
bounded SW. Since

(20)

transverse resonance leads to ,

which admits non-trivial solutions only for .
The latter equation can be re-written as follows

(21)

Its solution admits two eigenvalues. The dominant

(quasi-TM) eigenvalue is given by
Fig. 7. Behavior of and as a function of for different values
of average normalized reactance and backward pointing angle (22)
.
where

IV. ANISOTROPIC CONTINUOUS UNIFORM METASURFACE

Let us consider a surface that is uniform and anisotropic.
Namely, the tangential fields on the surface are sub- (23)
jected to anisotropic boundary condition
, where the impedance dyad is written in Equation (22) is equal to (8), except for the use of an equiv-
matrix form, in terms of its cylindrical components along the alent reactance in place of . If (18) holds, (23) can be
unit vectors and as follows approximated as follows

(15) (24)

The dominant SW is expected to have at tangential elec- The equivalent reactance slightly deviates from those
tric and magnetic components of the form of the isotropic cases for the presence of the correction term
. This means that under (18) the
anisotropic impedance supports a dominant mode which is the
(16) same as that of an isotropic impedance of effective value .
This allows the use the plots in Figs. 5 and 7 also for the design
where the modal coefficients are related of the anisotropic surface just considering the vertical axis as
each other by .
(17) B. Aperture Field Expression for SW
in which apex denotes transpose. In absence of losses the In the assumption purely imaginary (this hap-
impedance matrix is anti-Hermitian, (i.e., ) [8], pens in the practical case that we will see in Section VI), using
which implies and real and . In prac- (17), (18), (20) and (22), the SW dominant quasi-TM mode can
tical situation, being the element that realizes the anisotropic be approximated as
impedance extremely small, the anisotropic effect is quite weak,
thus obtaining (25)

(18) where , and

.
This condition implies that the cylindrical hybrid TE-TM SW
contains a large TM component and a small TE com- V. ANISOTROPIC MODULATED METASURFACE
ponent . The optimum way to obtain a circularly polarized antenna
A. Resonance Equation is to require that each point of the surface (namely each small
patch of the metasurface) radiates a circularly polarized field.
The propagation constant along can be obtained by con- However, this is extremely difficult to get since in (25) is
sidering the resonance along between the surface impedance a real number in most cases. It is simpler to devise a surface
and the dyadic free space impedance composed by circularly polarized sub-surfaces whose area is
smaller than . This area can be identified with a single
(19)
period of sinusoidal modulation along and a small angle
MINATTI et al.: A CIRCULARLY-POLARIZED ISOFLUX ANTENNA BASED ON ANISOTROPIC METASURFACE 5003

along . The radiation from the sub-surfaces plays a similar

role to that played by the element-factor in the product (element
factor) (array factor) of array theory.
Achieving such a behavior is greatly simplified if one con-
siders an azimuth-rotating first-order excitation .
In fact, one can exploit the following identity

(26)

which expresses the possibility to circularly polarize the aper-

ture in Cartesian coordinates by equalizing the and quadra-
ture components.
Fig. 8. Behavior of the field in one radial period of an anisotropic non-uniform
periodic metasurface.
A. Metasurface-Tensor Modulated Components
Let us assume that the normalized components of the reac-
tance tensor are described by the functions The field is linearly polarized at each point of the surface
and when one moves along it rotates and makes a
(27) complete turn in a period . The situation is sketched
(28) in Fig. 8. This property, together with the fact that each incre-
(29) mental strip is excited with initial phase , allows one to iden-
tify through (26) a -independent circular polarization of type
where is chosen as shown in for each subsurface.
Section III to produce a conical beam at in backward mode,
and and are modulation indexes whose ratio B. Condition for Circularly Polarized Sub-Surface
has to be chosen in order to obtain circularly polarized aper- The radiation from the continuous surface can be seen as the
ture sub-surfaces. We anticipate that the relationship between superposition of elementary sub-surfaces of angular dimension
and in (28) and (29) can be automatically satisfied by and radial length . The effective
exploiting a symmetry of the single element that constitutes the contribution from each sub-surface is obtained by averaging the
homogenized surface. Equations (27)–(29) exhibit an average field along , i.e.,
equivalence surface impedance

(30)

The above expression is used in Fig. 5 to find the value of

and thus that of appearing in (27)–(29). However, for
and values of reactance , the correction (33)
term proportional to is negligible.
Using (27)–(29) in (25), and neglecting the term proportional where
to in the product , leads to the following approxima-
tion for the electric field tangential to the surface (34)

(35)

(31)
with . Each sub-surface
where the term required for the evaluation of is therefore elliptically polarized, with axial ratio
can be obtained by using the Oliner-Hessel . Since
method within the approximation (24). Equation (31) is the sum- (13) and (22) imply
mation of a TM modal term and an oscillating and hybrid , the circular polarization
TE-TM contribution . The term does not provide a condition can be rewritten as
significant radiation contribution since its phase velocity is less
than the speed of light in free-space. The field forms an (36)
angle with the radial direction given by
where we have approximated by assuming to hold on
(32)
the conditions under which the -depending correction term
5004 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 11, NOVEMBER 2012

Fig. 10. Sequential-rotation excited patch feed embedded in the metasurface.

(Fig. 10). The patch is printed at the same level of the meta-
surface and it is excited in sequential rotation by four pins dis-
Fig. 9. Quantities involved in the design as a function of the normalized placed symmetrically with respect to the center. A square area
impedance. all around the circular patch is left without metasurface and its
dimension serves as a design parameter. The role of the patch
is not only that to excite a surface wave along the metasurface
in (30) is negligible. From (26), the sub-surface effective but also to radiate directly in the broadside direction for ad-
field becomes justing the radiation pattern inside the mask of Fig. 2.

(37)
VI. “PIXEL-TYPE” DESIGN THROUGH PRINTED PATCHES
where . In the previous section we have treated the basic design
Fig. 9 shows as a function of under (36) from various issues of non-uniform anisotropic metasurface without consid-
values of . For the sake of convenience, the same figure also ering its practical implementation. In this section we show the
shows the parameters design process of the normalized reactance tensor components
(27)–(29) through printed patches. To emphasize that these
(38) patches are small in terms of wavelength, we will sometimes
denote them as well as the cell that they occupy as “pixel”. The
and . pixel design process consists of three steps. We first choose
a suitable patch geometry for the anisotropic impedance pat-
C. Aperture Radiated Field terning. Then, we create reactance maps of the values of the
impedance tensor components versus two geometrical parame-
Under condition (36), for a circular aperture field of radius
ters of the pixel. Finally we synthesize a path on the reactance
, the far field is obtained from the radiation
maps that reconstruct the needed value (27)–(29).
integral
A. Pixel Shape
To produce anisotropic impedance, the patch geometry must
not be symmetric with respect to the SW direction of propaga-
(39) tion. Also, the cross-diagonal terms of the reactance tensor
must be real according to the considerations made in the pre-
vious section about circular polarization.
which can be manipulated to yield
Several geometrical shapes have been investigated (Fig. 11).
The solution of Fig. 11(b) is the one suggested in [4]. All the
investigated solutions exhibit two non-dimensional parameters
and which primarily affect the co-diagonal and cross-di-
(40) agonal reactance levels, respectively. The first parameter
is the ratio between one characteristic length of the geom-
etry normalized to the side of the periodic cell. Roughly
This expression is used to get the aperture radius that allows us speaking, the larger is the ratio the larger is the area of the
to reach the sufficient level of gain at . We observe that these patch. Increasing this parameter implies increasing the value
dimensions should be large enough for allowing a very high of co-polar reactances and . The second parameter
slope in the angular range 65 –75 . Also, it is worth emphasize is an angular parameter , which is the orientation of the slot
that (40) accounts for the leaky attenuation constant in . direction (Fig. 11(a), (b) and (d)), or the ellipse axis direction
(Fig. 11(c)). This parameter influences dominantly the value
D. Sequential-Rotation Patch-Feed
of . Due to the limitation of space available, we do not
To excite a surface wave with phase , a resonant circular present here the performance comparison for all the elements
patch of diameter is placed at the center of the metasurface investigated. We focus only to the solution that exhibits the best
MINATTI et al.: A CIRCULARLY-POLARIZED ISOFLUX ANTENNA BASED ON ANISOTROPIC METASURFACE 5005

Fig. 11. Patch geometries for anisotropic impedance surface. Each geometry
possesses two specific non-dimensional parameters and that are taken
for constructing the reactance maps. The arrow denotes the direction of SW
propagation.

performance for our purposes: the circular patch with a narrow

cut of Fig. 11(d) that we denote as screw-head patch (SHP).

B. Parameter-Plane Map for SHP

To evaluate the impedance tensor of the screw-head patch
this is assumed embedded in a periodic Cartesian lattice. A peri-
odic Method of Moment (MoM) analysis is applied for a limited
but sufficiently dense number of pairs. The pole-zero
matching (PZM) method [9] is then used to interpolate the re-
sults obtained on the design-plane .

C. Surface Texture
To texture the antenna surface with the appropriate elements,
one has to find the right parameter-plane path (PPP) which sat-
isfies both conditions and
(namely (27)–(28)), once one has fixed
, so that respects the circular polarization
condition (36) and satisfies (38). A least-square optimiza-
tion performed on the entire design-plane can be conveniently
used to find the best path. It is worth noting that, because of
the symmetry properties of , the same path also ensures
that the relation is satisfied.
These paths are shown by dashed lines in Fig. 12(a)–(c) for the
case and (i.e., ). The
anisotropic metasurface is textured on a Cartesian rectangular
lattice starting from a square region of side around the source
(see Fig. 10). The period of the Cartesian lattice is chosen as
an entire fraction of the requested (e.g., ranging from 1/4
to 1/6, which corresponds to ranging approximately from Fig. 12. Impedance map and relevant path to obtain an anisotropic surface
impedance in (27)–(29).
1/8 to 1/12). The choice of the period is actually dictated by
technological considerations, since very small elements collide
with manufacturing tolerance.
The position in the rectangular lattice of each pixel is associ- VII. PROTOTYPE DESIGN, ANALYSIS AND OPTIMIZATION
ated to its distance from the origin, and, therefore, to a specific
A. Surface Design
position along the PPP. The latter defines the slot orientation
and the patch radius. The final picture is a circular patch whose The overall design process to synthesize an antenna operating
slot continuously rotates and gradually changes its dimension in the frequency band from 8.5 to 8.7 GHz, and having a gain
within a modulation period , when we move along the radial pattern satisfying the mask of Fig. 2, is defined herein after. The
coordinate. fundamental parameters are the normalized impedance surface
Fig. 13 shows the construction geometry and the behavior of , the angle , and the couple of modulation indexes and
the field along a radial path. This picture emphasizes that the . As a starting point, we have used and
LW propagating along the radial direction leaks energy back- . From (38) (plotted in Fig. 9) we have derived the
ward, due to the modulation in patch sizes, and its electric field period of the modulation normalized to wavelength obtaining
component tangent to the surface rotates by interacting with the and mm. From (36), also plotted in Fig. 9,
slot of the patches. we obtain the ratio that ensures a circular polarization of
5006 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 11, NOVEMBER 2012

Fig. 14. Numerical analysis of the complete antenna. Sketch of the meshed
structure. In the inset, the gain pattern of the central patch feed, when the meta-
surface is not present.

Fig. 13. Metasurface texture and field behavior on an elementary subsurface

of one modulation period.

the aperture (i.e., ). By using the approximated field

in (40) one can define the radius of the aperture that allows
for a co-polar gain of 9 dB at (which is the maximum
level of the mask in Fig. 2). Note that (40) takes into account
the leaky attenuation constant inside . With this method we
obtained and an aperture of mm
(about at 8.6 GHz). Fig. 15. Simulated gain pattern (RHCP co-polar, and LHCP cross-polar com-
After identifying the substrate to print the patches (we have ponents) of the optimized antenna at three frequencies. The SHP is characterized
by mm, mm, ranging from 3.78 mm to 4.18 mm.
chosen Arlon AR 1000, and thickness 1.575 mm), we
textured the surface by using SHPs through the design-plane
path process illustrated in Section VI.C. The dimension of
the pixels (i.e., the period of the constant lattice) should be an
entire fraction of the period . This value should range from
to depending on the technology used. In our case, we
decided for mm.
The circularly polarized feeding patch has been designed with
the structure of Fig. 10. The patch diameter of mm has
been chosen to resonate at 8.6 GHz. The gain pattern of the feed
when the metasurface is not present (i.e., no SHPs are printed on
the upper face of the substrate) is shown in the inset of Fig. 14.
The maximum gain at broadside is 4 dBi. We emphasize that
the feed gain level at broadside is also approximately the gain Fig. 16. Printed layout of the antenna feed circuit (before 50 loads are
level of the overall antenna. With the above parameters we need mounted). (b) Return loss at the input port of the antenna.
around 12000 SHPs to cover the entire surface.
PPP to the new value of (see Fig. 12). The optimization has
B. Analysis and Optimization been performed at 8.6 GHz by using a standard optimization
The overall antenna structure is analyzed by using a method algorithm, starting from mm and mm. The op-
of moment code based on the multilayer adaptive integral timization process led to mm, mm and an
method (MLayAIM) [10], [11], which is highly suitable for the antenna diameter of 540 mm. The gain pattern obtained numer-
analysis of multilayered patch antenna arrays with a limited ically for a few frequencies is shown in Fig. 15. The result is
number of vertical conductors. Fig. 14 shows a sketch of the fully satisfactory if we consider that the prototype has been op-
mesh of the structure. Each SHP has been meshed with about 60 timized by using only two parameters.
RWG basis functions for a total of around 720 000 unknowns.
The structure has been optimized by using as parameters the VIII. PROTOTYPE REALIZATION AND MEASUREMENTS
period and the side of the square area outside which the A prototype of the designed antenna has been manufactured
SHP starts (see Fig. 10). When is changed, the SHP dimen- using a GIL GML 1034 substrate ( , thickness 0.508
sioning has been maintained perfectly scaled by adapting the mm) for the feeding circuit (Fig. 16(a)). For the metasurface
MINATTI et al.: A CIRCULARLY-POLARIZED ISOFLUX ANTENNA BASED ON ANISOTROPIC METASURFACE 5007

Fig. 17. Photograph of the realized prototype (above) and a detail of the feeding
zone without exciter pins (below).

(Fig. 17), a substrate Arlon AR 1000 ( and thickness

1.575 mm) has been used, which has the same permittivity and
thickness of the TMM 10i accredited for space.
A rotational SW is launched on the surface by feeding the
central patch, with diameter 6.2 mm, with four small vertical
dipoles excited with 0 , 90 , 180 , 270 phasing. The correct
phase excitation is provided by a 180 hybrid coupler connected
with other two 90 hybrid couplers (Fig. 16(a)). The input port
of the 180 hybrid is fed by a coaxial line. A matched 50
load is connected to the decoupled port of each hybrid coupler
to absorb reflected power. Fig. 16(b) shows the measured reflec-
tion coefficient at the input port of the antenna. Fig. 17 gives an
overview of the antenna, with a detail of the feed.
Fig. 18 shows the measured and simulated gain patterns,
respectively, for a few frequencies, including one outside the
band of interest (8.8 GHz, Fig. 18(d)). A good agreement
between numerical and experimental results is shown for the
co-polar components, while a higher value of the cross-polar
component is found around broadside in the measured pat-
terns. This is probably due to the imperfection of the feeder.
Cross-polar components are anyway satisfactory in the overall
angular range. Fig. 19 shows the measured gain at 8.6 GHz in
the plane . This gives an idea
of the symmetry of the pattern in the azimuth plane.
By using as a reference aperture distribution the inverse
Fourier transform of (3) with
km, the weighted aperture efficiency of the realized antennas
is (against a conventional aperture efficiency of Fig. 18. Comparisons between measurements and experimental results in the
working band at (a) 8.5 GHz, (b) 8.6 GHz, (c) 8.7 GHz, and outside the working
, which compares very favorably with that of a band at (d) 8.8 GHz. The Gain is normalized with respect to the total power. The
continuous aperture with the illumination law given in (5) and cuts shown are on a symmetry plane of the feed.
5008 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 11, NOVEMBER 2012

pins with sequential phase rotation, is used for exciting the sur-
face. The feeder also contributes to the overall radiation, mostly
in the broadside direction. The final prototype of the antenna
has a radius of 7.5 wavelengths, which is a result comparable
to the one obtained in [1] (6.5 wavelengths), with a weighted
tapering efficiency of 88%. It is realized according to the ana-
lytical design, provides satisfactory performances and the nu-
merical results obtained in latest design stages are proven to be
very accurate. The losses have been seen to be very low due to
the sub-wavelength dimension of the elements which implies
weak currents on the antenna surface.
Most of all, the antenna is extremely flat (1.57 mm), light (less
than 1 Kg) and it is built with the same PCB process used for
standard printed circuit. Thus, the final device has a significant
potential as a low-cost, simple solution for space application.

ACKNOWLEDGMENT

We thank Prof. M. Orefice and his team at Polytechnic of

Turin for having carried out the measurement of the prototype.
We also thank M. Bandinelli of Ingegneria dei Sistemi (IDS),
Pisa, Italy for leading the IDS activity in the frame of the project
supported by European Space agency contract (ESA Contract
No. 22808/09/NL/JD/al).

Fig. 19. Gain measurements at 8.6 GHz in the horizontal plane, (a) RHCP com-
ponent and (b) LHCP component.
REFERENCES

[1] R. Ravanelli, C. Iannicelli, N. Baldecchi, and F. Franchini, “Multi-ob-

shows that our solution is quite close to the optimum for the jective optimization of an isoflux antenna for LEO satellite down-han-
chosen aperture diameter. dling link,” presented at the 18th Int. Conf. on Microwave Radar and
Wireless Communications (MIKON), Jun. 14–16, 2010.
[2] D. Sievenpiper, L. Zhang, F. J. Broas, N. G. Alexopulos, and E.
Yablanovitch, “High-impedance electromagnetic surfaces with a
forbidden frequency band,” IEEE Trans. Microw. Theory Tech., vol.
IX. CONCLUSION 47, no. 11, pp. 2059–2074, Nov. 1999.
[3] P.-S. Kildal, A. Kishk, and S. Maci, “Special issue in artificial mag-
netic conductors, soft/hard surfaces, and other complex surfaces,”
A planar metasurface isoflux-type antenna has been success- IEEE Trans. Antennas Propag., vol. 53, no. 1, pt. 1, Jan. 2005.
fully designed and manufactured for LEO satellite application. [4] B. H. Fong, J. S. Colburn, J. J. Ottusch, J. L. Visher, and D. F. Sieven-
piper, “Scalar and tensor holographic artificial impedance surfaces,”
The antenna on the satellite platform shall radiate a uniform IEEE Trans. Antennas Propag., vol. 58, no. 10, pp. 3212–3221, Oct.
power density over a well-defined portion of the visible Earth 2010.
[5] S. Maci, G. Minatti, M. Casaletti, and M. Bosiljevac, “Metasurfing: Ad-
surface. The radiation mechanism exploited in the design is dressing waves on impenetrable metasurfaces,” IEEE Antennas Wire-
based on the transformation of a bounded SW mode into less Propag. Lett., vol. 10, pp. 1499–1502, 2011.
an unbounded LW due to the interaction with the reactance of a [6] G. Minatti, F. Caminita, M. Casaletti, and S. Maci, “Spiral leaky-wave
antennas based on modulated surface impedance,” IEEE Trans. An-
sinusoidally modulated anisotropic metasurface. The reactance tennas Propag., vol. 59, no. 12, pp. 4436–4444, Dec. 2011.
is realized by a dense texture of sub-wavelength metal patches [7] A. Oliner and A. Hessel, “Guided waves on sinusoidally-modulated
printed on a grounded dielectric slab and excited by an in-plane reactance surfaces,” IRE Trans. Antennas Propag., vol. 7, no. 5, pp.
201–208, Dec. 1959.
feeder. The patches have a circular shape with a thin slit cut [8] J. Hoppe and Y. R. Samii, Impedance Boundary Conditions in Electro-
along their diameter: the reactance tensor depends on both the magnetics. Washington, DC: Taylor and Francis, 1995, pp. 135–137.
area covered by the patch and the slit tilt angle with respect to [9] S. Maci, M. Caiazzo, A. Cucini, and M. Casaletti, “A pole-zero
matching method for EBG surfaces composed of a dipole FSS printed
the incidence direction of the surface wave. Roughly speaking, on a grounded dielectric slab,” IEEE Trans. Antennas Propag., vol.
changing the area of the patch is responsible for the intensity of 53, no. 1, pp. 70–81, Jan. 2005.
[10] P. De Vita, A. Freni, F. Vipiana, P. Pirinoli, and G. Vecchi, “Fast anal-
the radiation, whereas the rotation of the slit controls the polar- ysis of large finite arrays with a combined multiresolution—SM/AIM
ization of the radiated field. approach,” IEEE Trans. Antennas Propag., vol. 54, no. 12, pp.
By an appropriate metasurface modulation, the indexed 3827–3832, Dec. 2006.
[11] P. De Vita, F. De Vita, A. De Maria, and A. Freni, “An efficient tech-
Floquet-wave mode creates a circularly polarized aperture field nique for the analysis of large multilayered printed arrays,” IEEE An-
that radiates a conical beam. A small circular patch, fed by four tennas Wireless Propag. Lett., vol. 8, no. 1, pp. 104–107, Jan. 2009.
MINATTI et al.: A CIRCULARLY-POLARIZED ISOFLUX ANTENNA BASED ON ANISOTROPIC METASURFACE 5009

Gabriele Minatti was born in Florence, Italy, on ESoA counts about 150 among the best teachers of Europe (which include 11
1982. He received the M.S. degree (cum laude) IEEE Fellows) and it is frequented by an average of 220 students per year. The
in electronic engineering from the University of ESoA consortium presently comprises 33 European research centres and offers
Florence, Italy, in 2008 and the Ph.D. degree in twelve one-week courses per year. In 2005–2007, he was Italian National
information engineering from the University of representative of the NATO SET-TG 084 “Emerging Technology for Sensor
Siena, Siena, Italy, in 2012. and Front-ends,” and he is presently involved as co-representative in the NATO
He is presently working under a scholarship SET-181 RTG on “Metamaterials for Defense and Security Applications.”
from the University of Siena. His research interests He was co-founder of two spinoff-companies and since 2008 he is honorary
include periodic structures, artificial surfaces and President of LEAntenne e Progetti SPA (Valeggio sul Mincio, VE). His h index
holographic leaky wave antennas. is 25, with a record of more than 2000 citations (source Google Scholar).
Dr. Minatti was awarded with other coauthors with
the Best Paper on Antenna Theory at the 5th European Conference on Antennas
and propagation (EuCAP-2011, Rome, Italy, 2011) and with the Antenna In-
novation Award at the 33rd ESA antenna Workshop (ESTEC, Noordwijk, The Paolo De Vita received the M.S. degree in electronic engineering and the Ph.D.
Netherlands, 2011). degree in information and telecommunication engineering from the University
of Florence, Florence, Italy, in 1999 and 2004, respectively.
From 1999 to 2007, he has served as a Research Assistant at the Department
of Electronics and Telecommunication, University of Florence. Since 2008, he
Stefano Maci (M’92–SM’99–F’04) was born in is with IDS Ingegneria dei Sistemi, Pisa, Italy. His main research interests are
Rome, Italy, in 1961. He received the Ph.D. degree on numerical techniques for electromagnetic radiation and scattering problems.
in electronic engineering from the University of
Florence, Florence, Italy, in 1987.
He is a Full Professor the University of Siena,
with scientific responsibility of a group of 15 Angelo Freni (S’90–M’91–SM’03) received the
researchers (http://www.dii.unisi.it/~lea/). He is Laurea (Doctors) degree in electronics engineering
the Director of the Ph.D. School of Information from the University of Florence, Florence, Italy, in
Engineering and Science (which presently includes 1987.
about 60 Ph.D. students). His present research Since 1990, he is with the Department of Elec-
interests are focused on high-frequency and beam tronic Engineering, University of Florence, Italy, first
representation methods, computational electromagnetics, large phased arrays, as Assistant Professor and from 2002 as Associate
planar antennas, reflector antennas and feeds, metamaterials and metasurfaces. Professor of electromagnetism. From 1995 to 1999,
Since 2000, he was responsible of five projects funded by the European Union he has also been an Adjunct Professor at the Univer-
(EU). In 2004–2007, he was WP leader of the Antenna Center of Excellence sity of Pisa, and on 2010/2012 a Visiting Professor
(ACE, FP6-EU) and in 2007–2010 he was International Coordinator of a at the TU Delft University of Technology, Delft, The
24-institution consortium of a Marie Curie Action (FP6). He also carried out Netherlands. During 1994, he was involved in research at the Engineering De-
several projects supported by the European Space Agency (ESA-ESTEC), partment of the University of Cambridge, U.K., concerning the extension and
the European Defense Agency (EDA), the US-Army Research Laboratory the application of the finite element method to the electromagnetic scattering
(ARL), and by various industries and research institutions: EADS-MATRA, from periodic structures. Between 2009 and 2010, he also spent one year as a
(Tolosa, Francia), IDS (Pisa, Italia), TICRA (Copenhagen), SELEX SISTEMI researcher at the TNO Defence, Security and Safety, The Hague, The Nether-
INTEGRATI (Rome, Italy), THALES (Paris, France), TNO (Den Hague, lands, working on the electromagnetic modeling of kinetic inductance devices
The Netherlands), OTO MELARA (La Spezia, Italia), OFFICINE GALILEO and their coupling with array of slots in THz range. His research interests include
(Florence, Italy), SELEX Communication (Florence), Thales Alenia Space meteorological radar systems, radiowave propagation, numerical and asymp-
(Rome). His research activity is documented in 10 book chapters, 110 papers totic methods in electromagnetic scattering and antenna problems, electromag-
published in international journals, (among which 76 on IEEE journals), and netic interaction with moving media and remote sensing. In particular, part of his
about 300 papers in proceedings of international conferences. research concerned numerical techniques based on the integral-equation, with
Prof. Maci was elected Fellow of IEEE in 2003. Since 2000, he was focus on domain-decomposition and fast solution methods.
a member the Technical Advisory Board of 11 international conferences,
member of the Review Board of 6 International Journals. He organized 23
special sessions in international conferences and he held 10 short courses in
the IEEE Antennas and Propagation Society (AP-S) Symposia about metama-
terials, antennas and computational electromagnetics. He was an Associate Marco Sabbadini received the M.Sc. degree in
Editor of the IEEE TRANSACTIONS ON ELECTROMAGNETIC CAPABILITY and of 1983.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION and two times Guest Since 1988, he has been with the European Space
Editor of special issues of the latter journal. He is presently Director of ESoA, Agency, currently in the post of Senior Antenna
a member of the Board of Directors of the European Association on Antennas Engineer, at ESTEC in Noordwijk (NL). Previously,
and Propagation (EuRAAP), a member of the Technical Advisory Board of the he was with the Antenna Group, Thales Alenia Space
URSI Commission B, a member elected of the AdCom of IEEE Antennas and Italy, Roma (IT). His main areas of professional
Propagation Society, a member of the Governing Board of the European Sci- activity are: electromagnetic modelling algorithms,
ence Foundation (ESF) Project “NewFocus,” a member of the Governing board computational electromagnetics and computer-aided
of the FP7 coordination action “CARE” (Coordinating the Antenna Research antenna engineering tools as well as creative solu-
in Europe), a Distinguished Lecturer of the IEEE Antennas and Propagation tions in space antenna technology. Among others
Society (AP-S), a member of the Antennas and Propagation Executive Board his main contribution in the space antenna field include: the conception of the
of the Institution of Engineering and Technology (IET, UK), and a member antenna design framework and the management of its development over the
of the Focus Group on METAMATERIALS in the Finmeccanica project past 20 years, a leading role in the development of the electromagnetic data
“Mind-Share.” In 2004, he founded the European School of Antennas (ESoA), exchange language, the initial work on leaky-lens antennas. Currently he is
a post graduate school that presently comprises 30 courses on Antennas, involved in space antenna “miniaturisation” finding ways to make antennas use
Propagation, Electromagnetic Theory, and Computational Electromagnetics. in the best way the limited space and mass allocation available on spacecraft.