Signals and Communication Technology

Artem Boriskin
Ronan Sauleau Editors

Aperture Antennas
for Millimeter
and Sub-Millimeter
Wave Applications
Signals and Communication Technology
More information about this series at http://www.springer.com/series/4748
Artem Boriskin Ronan Sauleau
•

Editors

Aperture Antennas
for Millimeter
and Sub-Millimeter Wave
Applications

123
Editors
Artem Boriskin Ronan Sauleau
Technicolor R&D France IETR
Cesson-Sévigné Université de Rennes 1
France Rennes
France

ISSN 1860-4862 ISSN 1860-4870 (electronic)

Signals and Communication Technology
ISBN 978-3-319-62772-4 ISBN 978-3-319-62773-1 (eBook)
DOI 10.1007/978-3-319-62773-1
Library of Congress Control Number: 2017946656

© Springer International Publishing AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Contents

Part I Antenna Systems

1 Integrated Lens Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Artem V. Boriskin, Ronan Sauleau, Jorge R. Costa
and Carlos Fernandes
2 Multi-shell Radially Symmetrical Lens Antennas . . . . . . . . . . . . . . . 37
Zvonimir Sipus and Tin Komljenovic
3 Advanced Feeds for mm-Wave Antenna Systems . . . . . . . . . . . . . . . 75
Jorge Teniente, Juan Carlos Iriarte, Iñigo Ederra and Ramon Gonzalo
4 Flat Corrugated and Bull’s-Eye Antennas . . . . . . . . . . . . . . . . . . . . . 111
Miguel Beruete, Unai Beaskoetxea and Tahsin Akalin
5 Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Sean Victor Hum
6 Transmitarray Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Laurent Dussopt
7 Fabry-Perot Cavity Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Alexandros Feresidis, Konstantinos Konstantinidis and Peter Gardner
8 Near-Field Focusing by Non-diffracting Bessel Beams . . . . . . . . . . . 243
Mauro Ettorre, Santi Concetto Pavone, Massimiliano Casaletti,
Matteo Albani, Agnese Mazzinghi and Angelo Freni
9 Metasurface Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Gabriele Minatti, Marco Faenzi, Mario Mencagli,
Francesco Caminita, David González Ovejero,
Cristian Della Giovampaola, Alice Benini, Enrica Martini,
Marco Sabbadini and Stefano Maci

v
vi Contents

10 Terahertz Antennas and Feeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335

Goutam Chattopadhyay, Maria Alonso-delPino, Nacer Chahat,
David González-Ovejero, Choonsup Lee and Theodore Reck
11 Transformation Optics Applied to Antennas and Focusing
Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
Rhiannon C. Mitchell-Thomas and Oscar Quevedo-Teruel

Part II Antenna Measurements

12 Antenna Measurements at Millimeter and Submillimeter
Wavelengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
Antti V. Räisänen, Juha Ala-Laurinaho, Thomas Crowe,
Sergiy Pivnenko, Manuel Sierra Castañer and Ville Viikari
13 Holographic Imaging Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
Christian Schildbach and Lorenz-Peter Schmidt
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
Introduction

This book is about challenges and solutions: the fundamental challenges associated
with the exploration of the millimeter (mm) wave band and the antenna solutions
introduced to overcome these challenges.
The rapid development of the wireless communication and radar technologies
that took place in the second half of the twentieth century, thanks to the growth
of the semiconductor industry, resulted in a fast overload of the available frequency
bandwidth at microwaves.
A number of technical solutions were explored to answer this challenge,
including multiple-input and multiple-output (MIMO) methods, frequency and
spatial multiplexing, and finally increase of the operational frequency towards the
mm- and submm-wave ranges that are capable of providing higher data rates
compatible with the needs of today and future applications.
Although the increase of the operational frequencies promised a natural solution
for the limited bandwidth problem, the exploration of the mm-wave band raised a
number of fundamental issues that handicapped development of mm-wave antenna
technologies. This includes the low radiation efficiency of planar antennas, sig-
nificant material loss intrinsic to both metals and dielectric materials, and high
atmospheric absorption in some portions of the mm-wave band. These fundamental
problems together with the lack of powerful mm-wave solid state generators
required development of electrically large radiating structures, whose total effi-
ciency and physical dimensions were not always compatible with specifications
of the target applications, such as high-data-rate communications, intelligent
transport systems, and security imaging and sensing applications.
These challenges have determined the main trends in the research and devel-
opment of mm-wave antennas. Getting a high radiation efficient became of para-
mount importance to compensate for the high losses and low output power of
mm-wave generators. For the same purpose, a point-to-point communication con-
cept based on high gain antennas was proposed instead of an omnidirectional
broadcasting used at microwaves. This research direction naturally leads to
exploration of various types of focusing and collimating systems, including metallic
reflectors and dielectric lenses that can be considered as two milestone technologies

vii
viii Introduction

that created a base for a majority of the antenna solutions existing at the mm-wave
band until now.
However, being comparatively simple to design and fabricate, reflector and lens
antennas were often too heavy and bulky. In the original designs, the primary feed
of such antennas should have been placed at a certain distance from the phase
correcting element (i.e., shaped (or not) reflector acting in the reflection mode or a
lens acting in the transmission mode) that required additional volume and structure
complexity.
A solution for this problem was found in the introduction of the so-called
integrated lens antennas (ILAs) that assume a direct coupling between the feed and
dielectric lens. This solution was first introduced in the late 1980s and became one
of the most reliable antenna technologies used from lower mm-wave to infrared
ranges. In particular, it has been widely studied for point-to-point communications,
automotive radars, wireless backhauling, and imaging systems.
Although being quite effective in terms of aperture efficiency and directivity,
ILAs often suffer from a significant material loss and large dimensions. These weak
points are often considered as a bottleneck for the use of ILAs for mobile and
on-board applications requiring lower profile and reduced size/weight solutions.
Implementation of multi-beam and scanning ILAs was also associated with certain
difficulties due to the limited field of view (FoV), which is intrinsically restricted by
the focusing capabilities of extended hemielliptical lenses typically used in ILAs.
A solution for the latter problem (i.e., limited FoV) was found by using radially
symmetrical lenses that were already well-know at that time in the optical domain.
The aperture of such lenses, typically having the shape of a multi-shell sphere, can
be easily shared between several primary feeds, each generating an independent
beam in the far zone. In such a way, an angular coverage of up to ±90° and even
more can be achieved, being limited only by an aperture blockage by other feeds
and elements of the lens supports. However, the size of such lenses, limited number
of standard low-loss dielectric materials with the desired dielectric permittivity
values, and difficulties associated with the fabrication and assembly of multi-shell
lenses affected the attractivity of radially symmetrical lens antennas (RSLA).
Because of this, their usage has been largely limited to multi-beam automotive
radars and base stations for wireless communication systems requiring a
point-to-multi-point broadcasting.
A solution for the oversize and overweight problems of lens antennas was also
found in optics in the form of Fresnel-type diffractive lens antennas (DLAs). The
idea of using diffractive-type zone plates, capable of producing a required phase
shift for a portion of the wave front of an incident electromagnetic wave, dates back
to the nineteenth century. This concept widely used in optics was successfully
scaled down to the mm-wave range. Thanks to their planar topology and simple
fabrication, DLAs became an attractive solution for applications sensitive to the
weight and size of the antenna system, but still requiring a high directivity, such as
security imaging and space communication. More advanced configurations of
DLAs were successfully integrated in conformal dielectric radomes. However, the
Introduction ix

operational principle of DLAs naturally restricts their total efficiency, which may
not be sufficient for some applications.
As a next step in the evolution, two complementary antenna technologies were
proposed, namely the reflect array (RA) and transmit array (TA) antennas, whose
operational principles rely on the use of discrete arrays of phase correcting elements
operating in reflection or transmission, respectively. These antennas assumed illu-
mination of the entire array by a single feed, same as it was in the original designs
of reflector and lens antennas. However, the discrete structure of the
reflecting/transmitting phase correcting elements enabled a new functionality that
was not possible in any of earlier configurations, namely the reconfigurability (e.g.,
variation of the antenna radiation pattern) that may be enabled via electronic tuning
of the phase delay provided by each element (unit cell) of the array. Unlike active
phased arrays, which are widely used at lower frequencies but less effective at
mm-wave range due to the low radiation efficiency, RA and TA antennas are
capable of providing a twofold advantage, namely a higher radiation efficiency and
reconfigurability. Thanks to their advanced performance characteristics and reduced
weight, RA and TA antennas are widely used nowadays for on-board and space
applications.
Another antenna solution that was introduced aiming at producing low-profile
antennas with a simultaneous increase of its directivity was the so-called Fabry–
Perot antenna (FPA). Its operational principles rely on the use of a cavity covered
with a semi-transparent mirror that causes multiple reflections inside the cavity, thus
allowing to increase the electrical size of the radiating aperture over a limited
bandwidth, whose bound mainly depend on the antenna directivity. Although the
operational concepts of FPAs were introduced in the microwave community in the
1950s, research on FPA really started in the mid-1990s. Many high-performance
designs have been proposed, most of them below 100 GHz, with the main goal to
enlarge the bandwidth and provide multiband capabilities.
Of course, the progress in the development of mm-wave antennas was not solely
associated with the evolution of the focusing/collimating systems. It was also
supported by the advances in the primary feed design. Indeed, the mismatch
between the operational bandwidth of primary feeds, whose function was originally
limited to matching a waveguiding system with a free space in order to minimize
the reflection loss, and quasi-optical focusing systems aimed at correcting the phase
front of the radiated wave remained (and still remains) a bottleneck for design of all
types of mm-wave antennas. To handle this fundamental problem, significant efforts
were devoted to development of advanced feeds, having different topologies and
operational principles and, thus, capable of providing a required radiation effi-
ciency, bandwidth, and illumination conditions for the phase correcting elements
(e.g., lenses and reflectors) used to increase the directivity and/or shape of radiation
pattern of a primary feed antenna.
In some cases, the research into the primary feeds resulted in the development of
antennas that do not require an additional focusing/collimation element. Among
those are flat horn antennas (FHA), whose operational principle relies on the
excitation of leaky waves (LW) that propagate along the corrugated flanges of the
x Introduction

horn, thus contributing to the formation of the total radiation pattern. Optimization
of corrugated surface parameters enables one to support desired LWs that, in its
turn, defines the phase and amplitude distribution along the antenna aperture. In the
limit, one can open the mouth of a corrugated horn antenna to get a totally flat
profile. Such flat horn antennas can be fed by standard waveguide feeds as well as
narrow slots. The latter family is often referred as the bulls-eye antennas, thanks to
the shape of the annular grooves surrounding the central slot. Its operational
principle is linked to the extraordinary transmission through sub-wavelength
apertures. The low-profile corrugated horn antennas were recognized as being of
potential interest for indoor communication systems.
In recent years, the idea of using functional surfaces, capable of supporting
different types of waves (e.g., surface and leaky waves) or, instead, preventing their
propagation, has been attracting a growing attention that was boosted by the
advances in the metamaterial domain. Indeed, the possibility of changing locally the
surface impedance of the functional surface and thus manipulating with the phase
and amplitude of the fields in the entire aperture of a metasurface antenna (MSA),
opened the door for development of a new class of antennas with very attractive
performance characteristics, including shaped-beam patterns and near-field focus-
ing capabilities. Moreover, an active control over the surface impedance can
potentially enable development of reconfigurable and scanning MSAs capable of
competing with alternative approaches, namely the RA and TA antennas. Such a
functionality may be achieved, for instance, thanks to the use of advanced elec-
trically and/or optically tunable materials, like graphene.
In this book, the aforementioned antenna technologies are reviewed from the
point of view of their maturity and potential for future development. Practical
examples are provided in order to illustrate the evolution, current trends, and
specific features intrinsic to each antenna technology. The cross-links are provided
between the chapters to facilitate benchmarking of different antenna solutions. The
last two chapters are devoted to the challenges associated with antenna measure-
ment at mm-wave bands. The pitfalls in the design and characterization of different
types of mm-wave aperture antennas are identified in order to help readers in the
selection of an appropriate antenna solution and its effective implementation.
In such a way, the book represents a survey of the state-of-the-art and recent
developments in the field of the mm- and submm-wave antennas technologies being
in the focus of interest of the 5-year ESF RNP Newfocus program (2010–2015).
The authors gratefully acknowledge the financial support provided by the European
Science Foundation (ESF) through the RNP Newfocus program, which enabled
76 research projects in between 43 research teams from 19 countries worldwide,
whose outcomes constitute the core of the current book.

Artem Boriskin
Ronan Sauleau
 Part I
Antenna Systems
Chapter 1
Integrated Lens Antennas

Artem V. Boriskin, Ronan Sauleau, Jorge R. Costa

and Carlos Fernandes

Abstract This chapter presents the Integrated Lens Antenna (ILA) technology as it
evolved since its introduction aiming to respond to the needs of emerging appli-
cations such as high-data-rate communication, intelligent transport, and mm-wave
imaging. The topics covered include the ILA design concepts as well as the elec-
tromagnetic phenomena intrinsic to dielectric lenses that may affect ILA perfor-
mance. The aspects of the ILA technology related to selection of the primary feeds,
lens materials, and fabrication methods are also revised. A few practical examples
are provided to illustrate the current and future trends of this technology.

Keywords Integrated lens antennas Geometrical optics Physical optics Lens

feeds
Dielectric materials

Lens manufacturing

Lens profile design

Optimization

A.V. Boriskin (&)

Research & Innovation Division, Technicolor R&D France,
975 avenue des Champs Blancs, 35576 Cesson-Sévigné, France
e-mail: [email protected]
R. Sauleau
Université de Rennes 1, Institut d’Electronique et de Télécommunications
de Rennes (IETR), UMR CNRS 6164, Campus de Beaulieu, 263 Avenue
du Général Leclerc - CS 74205, 35042 Rennes Cedex, France
e-mail: [email protected]
J.R. Costa
Instituto de Telecomunicações, Instituto Universitário de Lisboa
(ISCTE-IUL), Lisbon, Portugal
e-mail: [email protected]
C. Fernandes
Instituto de Telecomunicações, Instituto Superior Técnico,
Universidade de Lisboa, Lisbon, Portugal
e-mail: [email protected]

© Springer International Publishing AG 2018 3

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_1
4 A.V. Boriskin et al.

1.1 Introduction

Integrated lens antennas (ILAs) were originally introduced in the late 1980s for
high-resolution imaging and sensing applications to compensate the low directivity
and low radiation efficiency of printed microstrip antennas at the millimeter
(mm) wave bands.
A partial remedy for the low-radiation efficiency problem was found in a direct
integration between printed antennas and dielectric lenses. This approach helped
one to minimize losses associated with modes supported by a dielectric substrate,
which is an essential counterpart of a printed antenna. Such modes (i.e., both first
TE and TM modes of a dielectric slab) have no cutoff frequencies and, thus, can be
excited even for very thin substrates. In case of a groundless microstrip antenna, the
presence of the lens on top of the planar feed also improved distribution of the
radiated power between the upper and lower half spaces. As demonstrated in [1],
the ratio of power radiated by a microstrip antenna printed at the interface of two
dielectric media towards the upper and lower half space is defined by the index ratio
at this interface:

P1 =P2 ¼ ðe1 =e2 Þ3=2 ; ð1:1Þ

where P1 and P2 is the total power radiated towards the upper and lower half space
filled in with the dielectric material having relative permittivity e1 and e2 ,
respectively.
This physical phenomenon suggests the use of electrically large dielectric lenses
made of high permittivity materials capable of imitating unbounded dielectric
medium. In some embodiments typically used at upper mm and THz ranges, the
primary feeds were printed directly on the flat base of a plano-convex lens. Such
lenses were often referred as “substrate lens” [2].
Optimization of the lens shape and structure provides additional degrees of
freedom in optimizing ILA performance characteristics such as directivity and
shape of the radiation pattern, phase center position, as well as radiation and
aperture efficiencies. In the 90s, the ILA concept has resurfaced as an effective
configuration to produce highly shaped radiation patterns for mm-wave mobile
applications [3,4, 60–62]: for these applications the objective was no longer
enhancing the directivity by producing a collimated beam but instead reshaping the
output beam to match demanding amplitude templates. For instance, shaped ILAs
were used to obtain isoflux illumination of areas having specific shapes, like a
rectangle, which is a typical scenario for mm-wave wireless LANs and mm-wave
mobile communications [4, 5].
Collimated beams lenses may by of primary importance for mm waves
point-to-point links due to the need to compensate the high material and propa-
gation losses to achieve a desired power link budget. To reach the highest possible
directivity, dielectric lenses with elliptical profiles were proposed [6, 7]. This shape
was suggested by the ray-tracing theory, which predicts that a spherical wave
1 Integrated Lens Antennas 5

radiated by a point source can be transformed into a locally plane wave (or vice
versa) by an elliptical lens, provided that the source is located in one of the ellipse
foci and the ellipse eccentricity is properly related to the lens material permittivity

f
e¼ ¼ e1=2 ; ð1:2Þ
a

where f ¼ bðe  1Þ1=2 is the focal length of the ellipse and b is its minor semi-axis
(Fig. 1.1a).
As the correction occurs only for the portion of the wave front impinging on the
front part of the lens, it was suggested to cut the ellipse through its rear focus and
attach the resulting flat-bottom lens directly on top of a primary feed, as a way to
solve simultaneously the radiation efficiency, beamforming, and assembly prob-
lems. In another embodiment, a hemielliptical lens was extended using a cylindrical

(a) (b)
1,5 1,0

Normalized focal length, f/b

Rexolite (ε = 2.54)
Ellipse axes ratio, a/b

1,4 0,8
Quartz ( ε = 3.8)

1,3 Macor ( ε = 5.6) 0,6

Silicon ( ε = 11.7)
1,2 0,4

1,1 0,2

1,0 0,0
2 4 6 8 10 12
Relative permittivity, ε

(c) 40 80
Critical angle, θ cr (degree)

Rexolite Quartz Macor Silicon

ε = 2.54 ε = 3.8 ε = 5.6 ε = 11.7
30 60
Reflectivity, R (%)

20 40

10 20

0 0
2 4 6 8 10 12

Relative permittivity, ε

Fig. 1.1 a Topology and notation of an extended hemielliptical lens, b parameters of the elliptical
lens versus relative permittivity of the lens material, c reflectance at normal incidence and critical
angle for plane wave incident on an air-dielectric interface from the denser medium side
6 A.V. Boriskin et al.

extension, whose height was selected so that the primary feed occurs close to the
rear focus of the ellipse. Dimensions of the ellipse semi-axes and its extension
length suggested by Eq. (1.2) are presented in Fig. 1.1b for some standard dielectric
materials. In other embodiments, the elliptical shape of the lens was approximated
by a hemispherical one, whereas the difference in the height of the hemiellipse and
hemisphere was again compensated by the height of the cylindrical extension.
As follows from Eq. (1.2), the lens material may be selected rather arbitrarily,
whereas Eq. (1.1) suggests the use of dielectrics with the highest possible permit-
tivity. A larger electrical size of the lens aperture is also favorable because it may
provide a higher directivity value for the same physical size of the lens. On the
other hand, high-index materials usually have larger density, which may cause
increase of the lens weight. In some cases, it is partly compensated by a smaller
height of the lens (see Fig. 1.1b) but may still be critical for some applications.
What is more important, an abrupt change of the refractive index value at the
air-dielectric boundary may cause a strong backreflection and excitation of multiple
internal reflections inside the lens that may deteriorate impedance matching of the
primary feed and the ILA radiation pattern. As shown in Fig. 1.1c, reflectivity value
for the plane wave normally incident on the surface of standard dielectric materials
may vary from about 6% observed for low-permittivity materials, such as Rexolite,
up to about 30% observed for dense materials, such as crystalline silicon. Because
of this, and also due to a reduced value of the critical angle (Fig. 1.1c, right axis)
that causes a larger amount of energy being trapped inside the lens, the modal
features in the electromagnetic behavior of lenses made of dense dielectric materials
may dominate over the ray-tracing focusing thus strongly affecting their perfor-
mance characteristics [8, 9].
A solution for the latter problem related to internal reflections was found in using
one or more matching layers in the form of quarter-wavelength layer made of a
material with a lower permittivity [10] or in a form of corrugations, acting as a layer
with an effective permittivity lower than that of the lens material [11]. These
solutions have been widely used in optics for decades and are well established
nowadays at upper mm wave and THz ranges. At the same time, they are less
popular at lower frequencies mostly because of the fabrication and assembly dif-
ficulties associated with multi-shell structures. This makes the use of
low-permittivity-shaped lenses to be a preferable solution for low mm-wave bands.
Regarding the optimal lens dimensions and choice of materials, there is no
universal guideline derived so far. This selection is usually made based on a
trade-off approach with respect to a technical specification derived for a certain
application. The factors typically taken into account are the loss tangent of the
dielectric material, its density, its homogeneity, permittivity, ease and cost of
manufacturing, including the cost of the supports and matching layer(s), if
applicable.
As follows from above, the hemielliptical (or hemispherical) shape of the lens
suggested by ray optics takes into account neither the actual size of the feed
aperture, nor its radiation pattern. Thus, it should be considered only as an initial
guess for the optimization process that may include optimization of the lens shape,
1 Integrated Lens Antennas 7

size, material, as well as the primary feed radiation pattern and position. Some hints
on how to reach the optimal match between the feed and lens parameters have been
provided in [12], based on the analysis of the lens edge illumination conditions.
All of the previous issues related to lens size, energy coupling from the feed to
the lens, material permittivity choice and internal reflections also apply for arbi-
trarily shaped ILAs. These kinds of lenses can be used to compensate for the effects
associated with finite size of the lens and nonnegligible size of the primary feed
aperture, as well as a possibly nonuniform and nonsymmetrical pattern of the
primary feed [13, 14]. ILAs can also be used to provide highly “shaped-beam”
radiation patterns according to given amplitude templates.
While enabling an effective control over the ILA radiation pattern at a certain
frequency, in certain designs the use of shaped lenses may restrict the operation
bandwidth of ILAs, especially if it is defined with respect to the shape of the
radiation pattern or position of the ILA phase center. To overcome this problem, the
use of multi-material shaped lenses has been recommended. In [15] it was
demonstrated that a double-layer shaped lens can provide a higher radiation effi-
ciency as well as a more stable radiation pattern and phase center position achieved
thanks to the lower contrast at the air-dielectric boundaries and additional degree of
freedom in optimizing the antenna characteristics provided by the additional
interfaces between different shells of the lens. The same conclusions have been
reached in [16], where a complementary focusing structure in a form of a
multi-shell lens-like dome has been reported.
Another trend in the design of ILAs is related to the reduction of the lens weight
and dimensions. Solutions proposed in this context include the use of lenses with air
cavities enabling reduction of the lens focal distance [17]. For some applications,
requiring fan-type beams with different angular width in the azimuthal and eleva-
tion planes, the use of planar slab lenses was suggested [18].
There is also a number of mm-wave applications, which require multi-beam or
scanning performance. Among those are point-to-point and point-to-multi-point
communications, imaging, and intelligent transport systems. The sought
multi-beam functionality was enabled using beam-switching arrays with each feed
producing an independent beam in a given direction [19, 20]. This approach pro-
vided a good reliability and fast scanning rates with a field of view (FoV) of
about ±25°. For larger scan angles, a rapid degradation of the antenna directivity
occurs that is explained by a shift of the feed position from the focal arc of the lens
observed for the off-axis feeds as well as a mismatch in the orientation of the
primary feed (radiating at broadside) and the desired scan angle [21].
At present, the enlargement of the ILA scan angle is probably one of the most
critical challenges associated with ILAs, which handicap further advances of this
technology. Solutions proposed in this context include multi-objective optimization
over the lens and feed parameters [22, 23], mechanical scanning [24], and the use of
shaped bifocal lenses [25] and double lens focusing systems [26, 27].
Another available solution is based on the use of multi-shell spherical [28] and
hemispherical [29] lenses, however, this solution is not always effective at upper
8 A.V. Boriskin et al.

mm waves because of the fabrication and assembly difficulties associated with the
multi-shell structure of such lenses [30].
There exists another solution for resolving the limited FoV problem that has
been developed by nature as a result of the million-year evolution, i.e., the insect
compound eye, which comprises plurality of low-resolution sensors arranged along
a curvilinear surface [31]. Technical implementation of such a device at mm waves
is difficult but, potentially, feasible using soft polymer materials.
Of course, the progress of ILA technology was not only associated with the
design of focusing elements but was also tightly linked with the advances in pri-
mary feed designs. The role of a primary feed of any aperture antenna is twofold:
impedance matching between a primary feed and its feeding system (e.g., hollow
metallic waveguide or microstrip line) and the optimal illumination conditions of
the focusing and/or beamforming element (e.g., a lens). The types of feeds used as
counterparts of ILAs include various types of microstrip patch antennas printed on a
PCB board [26] or directly at the bottom of the lens [1, 2, 23], as well as waveguide
feeds [4, 32, 16]. The latter typically provides a wider bandwidth but require
additional space for the waveguide feeding structure or a microstip-to-waveguide
transition that may spoil the total ILA bandwidth. The matching problem between a
waveguide feed and a lens is typically solved using dielectric tapers integrated
inside the waveguide. However, the bandwidth provided by these solutions is not
sufficient for some applications, especially at upper mm and submm-wave ranges.
Solutions for ultra-wideband ILAs have been provided in [33] using a crossed
exponential tapered slot antenna configuration, or in [34] using non-dispersive
leaky-wave planar feeds, both cases intended for imaging systems.
Finally, the use of primary feeds in a form of active arrays enabled another
interesting functionality not possible with a single-feed ILA, namely a reconfig-
urable radiation pattern, whose shape (e.g., half power beam width) can be changed
by switching between different combinations of array elements [35]. Such ILAs
with a reconfigurable FoV were recently proposed for automotive radars requiring
switching between long-range and short-range scanning regimes.
Looking back at the evolution of the ILA technology, we can see that after about
30 years of development the ILA technology has reached a certain level of maturity
and became an essential part of many established and emerging mm wave and THz
applications. At the same time, there are still some functionalities that have not been
enabled due to certain fundamental issues intrinsic to the technology, such as a wide
FoV and reconfigurability. Being of high practical importance, these constrains may
define future trends in the development of the ILA technology.
In a summary for this overview of the ILA technology evolution, it is worth
mentioning various numerical techniques that have been developed for the analysis
and synthesis of ILAs, among those are: high frequency techniques based on the
geometrical optics (GO) and physical optics (PO) [13, 4, 36, 37, 5], frequency
domain techniques based on two-dimensional and three-dimensional boundary in-
tegral equation (BIE) methods [8, 38], and time domain techniques, based on the
two-dimensional, two-dimensional-BoR, and three-dimensional FDTD methods
[39]. The GO/PO techniques were widely used for fast analysis and synthesis of
1 Integrated Lens Antennas 9

elliptical and shaped ILAs. The integral equation methods enabled accurate analysis
of the interplay between the geometrical optics focusing and modal (resonant)
features of compact-size dielectric lenses. In particular, use of Muller’s BIEs
enabled one, for the first time, to investigate the drastic influence of the so-called
half-bowtie resonances on the performance of elliptical ILAs [40, 9] as well as to
assess efficiency of GO/PO [8] and FDTD [41] methods in characterization of such
resonances. Finally, the FDTD techniques were effectively used for wideband
analysis of body-of-revolution (BoR) and three-dimensional single- and
multi-material structures metal-dielectric structures, e.g., [42, 16]. Nowadays, these
complementary techniques are implemented in a number of in-house [43] and
commercial software.
In the following sections, we present general guidelines for the design and
implementation of ILAs intended for various mm-wave applications. The topics
covered include: ILA design steps, lens materials, primary feeds and ILA fabri-
cating methods. The discussion is followed by a few practical examples of ILAs
illustrating recent advances in the ILA technology.

1.2 Lens Antenna Design

Figure 1.2 presents the generic lens geometry that of an ILA. It consists of a single-
or double-shell dielectric lens, shaped in the elevation coordinate and eventually
also in azimuth coordinate. It may be fed by a single feed or by an array of feed
elements attached or not to the lens base. In the general case, feed elements are
either on-axis (x = 0; y = 0; z = 0) or off-axis. Different options exist for the feed
element, as will be discussed in the following section, and of course its configu-
ration and radiation characteristics influence the final lens shape.
The adopted design procedure for shaped lenses is summarized in Fig. 1.3. Two
alternative approaches are presented: one provides a fast first guess of the lens
shape, based on closed-form analytical expressions—labeled (a); the second more
elaborate approach—labeled (b)—combines lens performance analysis tools with
an optimization algorithm that successively generates and evaluates meaningful
populations of test lenses until a specified figure of merit is reached.

Fig. 1.2 Generic geometry z

of a substrate integrated lens
antennas (ILA)

x
10 A.V. Boriskin et al.

INPUT Feed characteristics; materials,

design specifications; (b)
(a)
LENS SYNTHESIS
CLOSED-FORM EXPRESSIONS ITERATIVE PROCESS
Spline lens Parameterized
Analytical lens profiles profiles analytical lenses

LENS ANALYSIS

Lens performance evaluation and diagnostic tools

OPTIMIZATION
Optimization Algorithm; cost function;

OUTPUT

Fig. 1.3 Block diagram of lens design steps

Closed-form lens synthesis expressions can be obtained as a first guess for some
classes of target specifications, based upon Geometrical Optics (GO) formulation.
The lens’ overall dimensions and the principal radii of curvature at every point on
the surface are required to be much larger than the wavelength to ensure the GO
validity. The feed is characterized only by its phase center and by its far-field
pattern in an unbounded medium with the same permittivity of the lens material.
The GO lens designs are inherently broadband as long as the lens dimensions are
large compared to the wavelength, the material dispersion is negligible, and the feed
radiation pattern remains constant versus frequency.
The lens shape is directly determined from the solution of a set of partial dif-
ferential equations corresponding to Snell laws at the lens interfaces and additional
conditions related to the output requirements. The calculation time is generally in
the order of seconds, which makes it attractive as a first guess and as a seed for a
subsequent refinement optimization loop if required. Alternatively, the GO for-
mulation can include shaping parameters that enable running the closed-form
expressions in the optimization loop (b) of Fig. 1.3 to match an appropriate cost
function [43]. This narrows down the search space to physically meaningful lenses.
The GO-based formulation is inherently independent from frequency. An
appropriate analysis method is required to evaluate the lens actual performance
taking into account diffraction effects, frequency dependence of the feed radiation
pattern, and material dispersion.
The two-step Geometrical Optics/Physical Optics (GO/PO) method [6] is widely
used in the literature to determine the radiation pattern of lens antennas (also of
reflector and other aperture antennas). It offers a very good compromise between
quality of results and computation time, making it attractive for integration into lens
design optimization loops. It takes as input the lens shape, the material permittivity,
1 Integrated Lens Antennas 11

the feed position, and the far-field pattern from the feed in the lens material.
The GO is used for evaluating the field distribution over the lens/reflector surface,
and PO is used to calculate its contribution to the near- or far-field radiation. GO/PO
is especially useful for very large structures in terms of wavelength and there is no
upper size limit for its applicability. On the contrary, GO fails for small lenses [8,
42], where the feed can no longer be replaced by its far-field radiation pattern and
the ray propagation approach is no longer valid.
An alternative formulation was developed to circumvent some of the indicated
limitations of GO/PO in the characterization of small lenses [44]. It is a PO-based
method that uses point-like sources as basis functions to decompose the
pre-calculated feed aperture fields at the lens base and calculate the resulting
incident near-field at the lens surface. The proposed method was implemented only
for axial-symmetric lenses with on-axis feed.
Another limitation of the GO approach is that it fails to characterize completely
the total internal reflection phenomenon [37]. However, analysis showed that this
effect is more important in collimating lenses with canonical shapes, like the el-
liptical lens; it is imperceptible in most of the shaped lenses.
Full-wave numerical solvers like the CST Microwave StudioTM transient solver
[45] can be used to analyze the final lens design prior to fabrication. Such tool takes
into account all the effects that are handled with some approximations in the pre-
viously discussed methods but, due to the involved computation time, it is not
viable option for integration in lens design optimization loops.
Genetic Algorithms (GA) optimization method has long been used because of its
flexibility, robustness, global search character, inherent capability to avoid trapping
in local minima and complete absence of restrictions on the behavior of the function
to be optimized, e.g., [22, 14]. The drawback is its relatively slow convergence, and
hence two different approaches were used to reduce the number of shape parameters
that model the unknown lens surfaces (parallel boxes in the (b) path of Fig. 1.3):
• Splines-based representation of the lens surface;
• Parametric optimization of selected parameters from the GO closed-form design
expressions.
The splines-based representation is the most flexible of the two solutions,
allowing for a limitless search space. As few as 4–5 nodal points npn of the lens
surface are used as the target for optimization (see Fig. 1.4) while a cubic spline
approximation is used to represent the lens surface between consecutive pairs of
these wide-spaced nodal points. Especially with double-shell lenses, this strategy
often leads to the presence of caustics near the outer lens surface. These lenses are
automatically eliminated from the test population but this may significantly reduce
the number of “individuals” in the test population, with a negative impact on the
convergence of the optimization method.
The alternative strategy was to construct the test population based on lens shapes
obtained directly from the previously described GO closed-form expressions, where
the shaping parameters are incorporated. These shaping parameters were used as the
12 A.V. Boriskin et al.

Fig. 1.4 Geometry for spline parameterization of a generic lens interface

Fig. 1.5 Screenshot of ILASH software tool [43]

target for optimization, thus ensuring that all the generated lenses are necessarily
viable and none needs to be eliminated a priori. This favors convergence.
In order to quantify the merit of a lens in the optimization procedure, versatile
cost functions were defined combining different target conditions like: far-field
amplitude template, maximum power transfer, main beam power transfer,
Gaussicity, phase center position or beam scanning. For broadband and for scan-
ning lens design, multi-frequency and/or multi-feed position optimization runs were
required.
A software tool was developed ILASH [43] that implements the previously
described methods for the design, analysis, and optimization of axial-symmetric
multi-shell lens antennas with homogenous refractive index. Its interface, presented
in Fig. 1.5, provides the user with a simple way to interact with the kernel, to
generate and manipulate lens design data, to fully characterize the lens performance
and to export and import results. It is also possible to monitor in real time several
aspects of the optimization process like the cost function value, lens parameter
evolution, and convergence.
1 Integrated Lens Antennas 13

1.3 Lens Antenna Implementation

1.3.1 Feeds

For any lens output specification, at least two items need to be carefully addressed
prior to the lens synthesis process: the lens feeding structure and the lens dielectric
materials.
Like the feeds for reflectors, lens feeds have specific requirements in terms of
radiation pattern, polarization, and phase center depending on the application. The
shape of an integrated lens is tailored to obtain a prescribed output radiation pattern,
and thus it obviously depends on the feed radiation pattern inside the lens body.
Three main technologies are the most common for the lens integrated feeds:
waveguide, coaxial, and planar printed feeds.
Figure 1.6a shows one example of a dedicated V-band (50–75 GHz) waveguide
feed [32] that is used in Sect. 1.4. Near its open end, the waveguide is loaded with a
high permittivity dielectric wedge and tapered down to a narrow 1.4 mm square
aperture designed to present a good power transfer to the lens. The square aperture
produces a circularly symmetric radiation pattern inside the lens body, equivalent to
a printed double-slot radiation pattern in the same conditions [6], with linear
polarization and about 8 dBi directivity (depending on lens material permittivity).
This directivity is adequate to illuminate most part of the useable aperture from our
shaped integrated lenses. The reflection coefficient at the waveguide port is lower
than −10 dB.
However, some lenses have much wider bandwidth than this V-band waveguide
feed. A similar feed was fabricated for the Q-band (30–50 GHz) with the same
external dimensions, Fig. 1.6b, producing exactly the same symmetric radiation
pattern and the same directivity [32] to allow testing wideband lenses by separate
sub-bands.
The outer wall of the developed waveguide feeds is threaded and screwed into a
matching hole opened at an extended part of the lens base. This strategy enables

Fig. 1.6 Photograph of the

fabricated waveguide feeds
[32]: a for 62.5 GHz
operation; b for 43 GHz
operation
14 A.V. Boriskin et al.

simple and reliable exchange of different test feeds in the same lens, or the reuse of
the same test feed in different lenses without compromising alignment and mea-
surement repeatability in lab evaluation tests.
However, an actual ultra-wideband low profile feed is also required for the
end-application, compatible with simple integration of ultra-wideband bolometer
detectors used, for example, in radio astronomy applications [46, 47]. Planar
antenna structures printed directly at the lens base [48, 49] are the preferred solution
for submillimeter wave applications where waveguides are difficult to manufacture.
However, once the feed is printed or glued to the lens, it is not possible to reuse it in
a different lens, making this type of feeds less attractive for laboratory systematic
testing. A very common integrated lens feed is the double-slot [6], however its
bandwidth is limited. Wideband self-complementary log-periodic printed antennas
have been used as integrated lens feeds in the literature, but its polarization is not
sufficiently stable versus frequency [48].
A dedicated broadband printed antenna was developed, which is especially
adequate for integration with bolometers or mixer diodes. It is based on a new
crossed exponentially tapered slot configuration combined with an intersecting
square slot, which is onwards referred to as the XETS antenna. A fabricated pro-
totype for 30–70 GHz operation attached to a MACORTM lens [33] is shown in
Fig. 1.7. The feed is developed in two layers of metallization [33]. The slot layer is
in direct contact with the lens base; the other layer (Fig. 1.7b) contains a replica of
two opposing central petals from the front face, printed at the same position in order
to provide a capacitive coupling with the front layer. These petals are used to mount
a Schottky diode (used as a mixer). The Intermediate Frequency (IF) retrieving
circuit with soldering pads for the coaxial cable is also shown in Fig. 1.7b.
Depending on the lens permittivity, more than 60% impedance bandwidth can be
achieved with reasonably stable radiation pattern and almost pure linear polariza-
tion across the whole bandwidth; cross-polarization level is in the order of −15 dB.

(a) (b)

1.7 mm

Fig. 1.7 XETS prototype for 30–70 GHz operation when integrated with a MACORTM lens [33]:
a photograph of the front face; b back face
1 Integrated Lens Antennas 15

Fig. 1.8 Front side of the

multiple XETS prototype [50]

1 2 3 4 5 6

This is a clear advantage over self-complementary log-periodic printed feeds [49].

The phase center of this radiating element is very stable across the whole frequency
band and lies in the antenna plane.
These broadband XETS elements can be closely packed for scanning lens
applications, see Fig. 1.8. The multiple XETS elements are aligned along the feed
H-Plane [50]. Simulations have shown that a separation between XETS centers of
only 2 mm in this plane ensures better than 18 dB isolation between adjacent feeds
across the operating bandwidth, despite the presence of the very close IF circuits on
the backface.

1.3.2 Lens Materials

The second issue to be addressed prior to the lens design is the selection and
characterization of the dielectric material. Common materials are PTFE, poly-
styrene, fused quartz, ceramics, silicon, or alumina. In most cases, the electro-
magnetic characteristics of candidate materials are specified by the manufacturer for
microwave frequencies only, with tolerances that are incompatible with an accurate
lens design in the most stringent applications. Thus, the dielectric complex per-
mittivity must be evaluated in-house to ensure a good agreement between predic-
tions and measurements. Besides the electrical properties, the homogeneity and the
isotropy of the raw material must also be checked.
Two most used methods for permittivity measurement are: the waveguide
method [51] and the open resonator Fabry–Perot method [52, 53]. The waveguide
method is very convenient for low to moderate permittivity (up to 5) and loss
tangent (in the 10−4 to 10−2 range). Parallelepiped samples of the dielectric material
are precisely cut to fill completely the cross section of the rectangular waveguide
short-circuited sample holder, see Fig. 1.9. Under single-mode TE10 conditions, the
16 A.V. Boriskin et al.

Fig. 1.9 Example of waveguide short-circuited sample holder for V-band and a few dielectric
samples [51]

Fig. 1.10 Fabry-Perot open

resonator for V-band complex
permittivity measurements
[53]

complex permittivity is calculated from the measured s11 frequency responses for
the (i) dielectric filled and (ii) empty sample holder. The method is quite accurate,
provided that the air gaps between the sample and waveguide walls are less than
50 lm in the V-band. Because the required samples are very small (typically
3.9  1.9  5.0 mm3 in V-band), material non-homogeneity or anisotropy can be
evaluated by cutting samples from different parts and different orientations of the
raw material block.
Excitation of higher order modes in the dielectric filled sample holder for relative
permittivity values typically larger than 5 tends to introduce significant error in the
complex permittivity determination, especially in the loss tangent. The Fabry–Pérot
open resonator method [52] does not have this limitation. In Fig. 1.10, a
1 Integrated Lens Antennas 17

Table 1.1 Measured permittivity values of different materials at 60 GHz

Material Dielectric constant Loss tangent
ABS-M30 (three-dimensional printed) 2.48 0.008
Acrylic glass 2.5 0.0118
Alumina 9.3 0.0013
Fused quartz 3.8 0.0015
MACOR 5.5 0.0118
PLA (three-dimensional printed) 2.85 0.014
Polyethylene 2.3 0.0003
Polypropylene 2.2 0.0005
Polystyrene (Rexolite) 2.5 0.0004
Teflon 2.2 0.0002

plane-concave configuration was selected, since it involves the fabrication of only

one spherical mirror while allowing precise positioning of the dielectric samples
directly on the planar mirror, without any perturbation from sample holders. The
resonator dimensions were chosen to produce at the planar mirror a 6 mm beam
waist fundamental Gaussian mode with well-defined linear polarization. The quality
factor of the empty Fabry–Perot resonator is 106. The complex permittivity of the
sample dielectric material disk is calculated from the S21 frequency response
measured between the feeding and pick-up waveguide probes [53] for the empty
and for the dielectric loaded resonator. The measurement is sensitive to polariza-
tion, and thus, by rotating the sample disk about the cavity z-axis, it is possible to
clearly identify possible anisotropies of the material sample and calculate the
complex permittivity along the anisotropy axes [53].
The measured permittivity for typical dielectrics is shown in Table 1.1. High
permittivity values of the lens material allow more flexibility to comply with beam
shaping requirements and better power coupling from bidirectional planar feeds to
the lens. But this comes at the cost of higher lens internal reflection loss detuning of
the feed input impedance [54]. Matching layers [10] or multiple shell lenses can
cope with these opposite dependencies on material permittivity.

1.3.3 Fabrication

Besides the electromagnetic properties, the choice of the dielectric material may be
limited by the selected process for lens fabrication. Several techniques exist to
manufacture a dielectric lens antenna. The most common approaches use either
computer numerical control milling machine (CNC), moulding or
three-dimensional additive manufacturing. Milling is a common approach, but it
may not be appropriate for hard materials, typically with high permittivity values
like alumina.
18 A.V. Boriskin et al.

Fig. 1.11 Lens fabrication by CNC milling machine technique

Fig. 1.12 Lens fabrication by molding technique

CNC enables to produce physical objects from its digitized three-dimensional

description, by automatically excavating the object shape out from a block of raw
material, Fig. 1.11. This type of fabrication process is especially appropriate for
antennas with complex shapes, tight dimensional tolerances and good surface finish
as happens for instance with shaped lens antennas at millimeter waves. It is more
indicated for small series production or for lab prototyping as opposed for instance
to the injection molding process, which is cost effective for mass production. CNC
milling may, however, be used for mold production for the latter technology.
There are no special limitations on the materials that can be used for CNC
milling, except that they must be machineable. Some examples of commercially
available machineable dielectrics are listed in Table 1.1. The appropriate cutting
tool, its rotation speed, the cutting speed, feed rate, and cutting depth must be
appropriate for the selected material. The specific values for these cutting param-
eters can be found in manufacturers’ material data sheets. Depending on the
dielectric material and CNC milling machine characteristics, 50 lm fabrication
accuracies can be achieved. For some type of materials it is possible to fabricate the
lens using a moulding technique, Fig. 1.12. In this case, a mold with the shape of
the lens has to be manufactured where the dielectric, in liquid phase, is poured or
injected into. Usually, the assembly must be left for a few hours before the mold can
be removed. Depending on the material, the lens dimensions may change
1 Integrated Lens Antennas 19

marginally after material curing; this has to be considered when designing the mold.
Moulding is generally considered a cost-effective technique for mass-production,
but it is of limited advantage for lens lab prototyping due to the mold manufacturing
cost. Furthermore, it is not easy to avoid trapped gas bubbles or to control the
homogeneity of the mixture as it becomes solid, without an appropriate facility.
In the case of multilayer lenses particular attention has to be given to the
matching and alignment of the layers produced by milling or molding technologies.
Air gaps between the layers should be avoided when assembling the lens. The
presence of air gaps as small as 0.03 wavelengths in multilayer lenses can reduce
the antenna gain and increase side lobe level of the radiation pattern. This effect is
more intense with the increase of the number of layers [55] or with higher per-
mittivity materials [56]. A glue can be used to mitigate the effect of the air gaps as
long as it has approximately the same dielectric constant of one of the adjacent
layers.
Three-dimensional printing is becoming an attractive technique to fabricate
elaborate dielectric prototypes at relatively low cost. As with CNC, it starts with a
three-dimensional digitized model of the object, but the rendering is based on an
additive process where successive layers of a material are laid down dot by dot to
construct the desired shape. Various materials can be considered, but thermoplastic
based three-dimensional printers are the most popular. Although the commonly
used materials tend to have moderately high losses and the fabrication tolerance of
low-end printers is presently of the order of 200 lm, the viability of a
three-dimensional printed lens antenna made of ABS (tan d = 0.008) for
short-range indoor wireless link at 60 GHz has been recently demonstrated [57, 58].
A similar technique can be used to manufacture lenses using higher permittivity
lower loss materials like Alumina [11].

1.4 Lens Design Examples

This section presents selected lens examples for different applications. These and
other examples are discussed in full detail in the indicated references.

1.4.1 Constant Flux and Flat-Top ILAs for mm-Wave

WLAN

The cellular Mobile Broadband Systems (MBS) concept was introduced in 1992
and developed throughout the 90s [59]. It was intended to transport, in a transparent
way, Asynchronous Transfer Mode cells (ATM) over the air interface at data rates
up to 155 Mbit/s, withstanding mobile velocities up to 100 km/h, while maintaining
20 A.V. Boriskin et al.

Fig. 1.13 a Geometry of a street sectored cell; b photograph of a shaped lens antenna for a
constant flux illumination of the street at 60 GHz [60]

the quality of service of the fixed network. Provisional sub-bands were allocated for
these systems at 39.5–40.5, 42.5–43.5 GHz, and 62–63, 65–66 GHz.
At the base station, a shaped beam was required to approximate a constant flux
illumination of the cell to favor the link budget, with reasonably sharp drop at the
cell edge to control cell spillover (see Fig. 1.13a). This characteristic corresponds to
cosec2 type gain contribution from the antennas to the link budget within the cell,
which ideally compensates for the free space loss at each observation direction. An
acrylic three-dimensional lens was designed using a modification of GO
closed-form expressions to produce constant flux illumination across a long street
section when hanging from a lamp post located 10 m above the street level at the
center of the cell [60]. The lens was fed by the embedded open-end aperture of a
dielectric loaded rectangular waveguide operating in the TE10 mode. The lens
prototype is shown in Fig. 1.13b [60]. The dimensions are 10 cm  4 cm  2 cm
and the lens produces a bidirectional radiation pattern with 13.2 dBi gain at
62.5 GHz at the maximum distance direction (hmax, u = ±90°), with linear
polarization. The lens was successfully used in an outdoor system demonstrator
operating at 32 Mbit/s gross bit rate. In a follow-up project, these lenses were
redesigned for the 40 GHz band with the same specification but using a lower loss
dielectric material, polystyrene. A similar radiation pattern was obtained in this
case, but with slightly higher gain of 14 dBi. The received power over the street cell
showed that by using a single-shaped lens antenna at the center of the cell it is
possible to produce a rectangular-type of coverage above the threshold level, almost
confined to the cell limits [61]. No similar type of performance has been presented
before for any antenna. In a different configuration, a three-dimensional lens was
developed to produce a constant flux illumination with sharp square boundaries,
appropriate for indoor room coverage with controlled wall illumination [62],
Fig. 1.14a. The lens was made of acrylic with the size 84 mm  32 mm (diago-
nal  depth) for operation at 62.5 GHz. The lens was fed by the circularly
polarized TE11 mode of a circular waveguide immersed in the lens body and
mounted near the ceiling. Figure 1.14b shows the received power distribution
beneath the lens, measured while the receiving lens antenna, with a flat-topped
radiation pattern, traveled on a kart to scan a 4 m  8 m floor area. The difference
in height between base station and the mobile terminal antenna was 0.5 m. The
figure shows a constant received power in the illuminated region and a sharp drop at
1 Integrated Lens Antennas 21

Fig. 1.14 a Dielectric lens antenna for a square cell fed by a circular waveguide; b received
power distribution measured in the laboratory at 62.5 GHz [62]

the cell edge as required. The sec2 radiation pattern has the interesting property that
the cell dimensions can be scaled just by changing the height difference between the
fixed and the mobile antennas. This topic is further developed in reference [62],
where the calculated and measured radiation patterns, and also the resulting channel
time dispersion characteristic, are given.

1.4.2 Broadband ILAs as a Feed for Reflectors

The most common reflector feeds are based on horn antennas. Despite many
variants that horns may have, they all share the same feature of aperture antennas,
which is a radiation pattern that increases its directivity with the frequency. When
large bandwidths are involved, this may represent an important reduction of the
reflector aperture illumination efficiency. This problem is often circumvented by
splitting the wideband into multiple sub-bands, each with its own feed horn; but
then a complex set-up is required to minimize aberrations due to off-axis posi-
tioning of the multiple horns.
A challenging specification for a reflector feed is the ability to maintain a
constant radiation pattern beam width with stable phase center over a 1:3 band-
width. A double-shell shaped dielectric lens can be designed to meet such speci-
fication [15]. The lens is intended to feed a 90° offset reflector, Fig. 1.15a. The lens
output beam produces a virtual focus located far below the lens and close to the
corresponding reflector focal point. A 20-k diameter lens prototype was designed
and fabricated using MACORTM and acrylic (Fig. 1.15b). The optical transparency
of the acrylic reveals the inner shell, which appears distorted due to refraction
effects. The lens is fed by the waveguide device described in Sect. 1.3.1 (Fig. 1.6).
The measured lens radiation patterns are shown in Fig. 1.16a for two distinct
frequencies in the Q- and V-band (40 and 62.5 GHz). The measured lens radiation
22 A.V. Boriskin et al.

Fig. 1.15 a Problem geometry; b fabricated MACOR™/acrylic lens prototype [15]

Fig. 1.16 a Measured radiation patterns of the lens at 40 and 62.5 GHz; b calculated directivity
and aperture efficiency of the reflector with frequency plotted on log scale [15]

pattern agrees very well with the Gaussian template with a0 = 23° Gaussian width
(dashed curve). It is possible to see that the shape and beam width of the lens
radiation pattern are identical in both bands as desired. The same behavior was
demonstrated by full-wave simulation between 30 and 90 GHz [15]. The lens
output beam Gaussicity is better than 94% over the band. The measured phase
center position of the lens agrees with the value imposed during the design
(z0 = 68.7 mm) and it is identical in both frequencies. Full-wave simulations
confirm phase center stability over the 1:3 bandwidth. The performance of the
lens-reflector assembly was simulated using the ILASH software [43]. Figure 1.16b
shows that the aperture efficiency is practically constant versus frequency as
1 Integrated Lens Antennas 23

Fig. 1.17 Fabricated shaped

double-shell lens prototype
[63]

required and, in agreement with that property, the reflector directivity in dB

increases almost linearly with logarithm of the frequency.
In a second example, a shaped double-shell dielectric lens (Fig. 1.17) has been
designed as a primary feed for a multi-beam single reflector system operating in the
Ka-band satellite up-link (30 GHz) and down-link (20 GHz) bands, complying
with gain and edge-of-coverage directivity (EoC) requirements [63]. An assembly
of dual-band printed feeds is integrated at the base of a single lens, each feed
producing a virtual focus far behind the lens base and coincident with the reflector
focal arch.
The lens is axial-symmetric, formed by two embedded shells of Polystyrene
(inner shell) and MACORTM (outer shell). The used double-shell lens approach,
instead of a single-material lens, allows an extra degree of freedom to accommodate
an aberration mitigation condition. In fact, two refracting surfaces are required to
impose two design conditions, one associated with the phase center and the other
with low aberration. Further, it was verified that selecting a higher permittivity for
the outer shell material favors lower beam tilt for the off-axis feeds (lower spillover
losses) while allowing more space between adjacent feeds at the lens base.
Therefore, this lens based primary feed system is proposed as a low-complexity
solution to enable fitting more beams per solid angle than conventional
single-feed-per-beam systems based on a cluster of focal-plane horns. The proof of
concept lens prototype with 87 mm diameter and 62 mm height, fed by a linear
arrangement of five dual-band printed feeds was fabricated and tested at the
Ka-band [63]. The lens measured radiation patterns were post-processed to evaluate
the combined performance of the lens with an offset F/D  1 reflector system
designed for 45 dBi EoC directivity, see Fig. 1.18. It is shown that it duplicates the
reflector aperture efficiency compared to horn fed systems with same feed
separation.
24 A.V. Boriskin et al.

Fig. 1.18 Radiation patterns of an F/D = 0.95 reflector obtained by illuminating a simulated CST
reflector with the post-processed measured lens radiation patterns and lens simulated radiation
patterns: a at 20 GHz; b at 30 GHz [63]

(a) (b)

Fig. 1.19 Multi-shell lens-like dome: a topology in the E-plane cross-section, b prototype

1.4.3 Broadband Lens-Like Dome Antenna

Another example of a broadband ILA is presented in Fig. 1.19. Unlike earlier

examples, this antenna comprised a multilayer lens-like dome, whose profile was
optimized using a genetic algorithm aiming at producing a stable Gaussian-type
radiation pattern across the entire V-band (50–75 GHz). The desired functionality
was achieved thanks to the simultaneous optimization of the shape for both internal
and external surfaces of the dome [16]. The backreflection from both air-dielectric
interfaces was minimized using quarter-wavelength matching layers. The dome
optimization method was similar to the one represented in Fig. 1.3 (loop b) with a
number of cubic spline nodes uniformly distributed along both surfaces
(Fig. 1.19a).
The antenna was fed with E-plane sectoral horn having a rectangular aperture
and a nearly symmetrical radiation pattern that justified the use of a rotationally
symmetrical dome.
1 Integrated Lens Antennas 25

Fig. 1.20 Measured radiation patterns of the optimized LLDA with two matching layers:
a E-plane, b H-plane. The families of 6 curves corresponds to different frequencies:
50 GHz; 55 GHz; 60 GHz; 65 GHz; 70 GHz; 75 GHz
(borrowed from [16] with permission from IEEE)

The dome core layer was made of Rexolite (e = 2.53), with two matching layers
made of a low-permittivity foam (e = 1.59). All counterparts of the shaped dome
have been fabrication via micro-machining (Fig. 1.19b). The feasibility of a
single-step fabrication of each shell and their assembly was guaranteed by the
constrains to the dome feasible shape that have been implemented in the opti-
mization routine.
The measured radiation patterns in both planes presented in Fig. 1.20 for five
frequency points reveal successful achievement of the optimization goal. The main
advantages of the lens-like dome antenna include wideband performance and light
weight, which may be important for on-board applications. Note that a certain
constrain on the shape of external surface can be easily added in the optimization
routine. For instance, it may be optimized to provide optimal aerodynamic char-
acteristics. Possible weak points include complexity of mechanical assembly of the
multilayer structure and rigidity of the foam matching layers, which may prevent
scaling this design to upper mm-wave bands.

1.4.4 Multi-beam ILAs

The first example refers to an integrated multi-beam lens antenna intended to

produce 3 dB beamwidths of the order of 6° at the 60 GHz band, with high
Gaussicity, gain scan loss lower than 1 dB over ±20° scan angle and complying
with a maximum diameter of 70 mm [64]. The challenge was to minimize internal
reflections and to ensure a reasonably stable Gaussian beam shape across the
scanned angular interval. A double-shell configuration was adopted to allow
imposing both a beam collimation condition and the Abbe sine condition to min-
imize aberration of the output beam. The selected lens materials were the
MACORTM/acrylic combination (permittivity 5.5/2.53), and the feed was the
especially designed waveguide aperture described in Sect. 1.3.1 for 62.5 GHz
26 A.V. Boriskin et al.

Fig. 1.21 a Fabricated MACOR™/acrylic lens prototype, showing the attached 60 GHz band
waveguide feed; b simulated and measured gain radiation patterns at 62.5 GHz, in the H-plane,
with feed at 0, 1.1, 2.2, 3.3, 4.4, and 5.5 mm from the lens axis [64]

operation. Figure 1.21a shows the lens antenna prototype which was designed
using ILASH [43]. The lens dimensions are 60 mm diameter at the base and 37 mm
height; the outer shell material is transparent, and thus it reveals the inner shell,
which appears distorted due to refraction effects.
Figure 1.21b shows the measured and simulated radiation pattern results for
different feed positions at 1.1 mm steps along the lens base in the H-plane. This
produces overlapping of consecutive beams near −1.5 dB level with respect to
maximum. Beam overlap at −3 dB occurs for wider separation between consecu-
tive feeds, compatible with the used feed aperture size (not shown). The figure
shows excellent agreement between ILASH predictions and measurements and
confirms the effectiveness of the proposed double-shell lens concept regarding
beam scan linearity, internal reflections, beam shape and Gaussicity preservation
with extremely low scan loss up to 20° elevation. The achieved gain was of the
order of 18–19 dBi and the scan loss better than 1 dB. As a consequence of the
GO-based lens design, the scan angle dependence on the feed off-axis position was
practically independent of frequency and the beam Gaussicity was better than 95%.
The second example of a multi-beam ILA concerns an antenna with a recon-
figurable FoV (Fig. 1.22). Such a functionality may be of particular interest of
automotive radar systems requiring switching between short-range (SR) and
long-range (LR) regimes. The presented antenna relies on the concept proposed in
[35] but, instead of using a standard extended cylindrical lens, it is based on a
shaped lens, whose profile was optimized in three cut-planes to provide a desired
beam width in both horizontal and elevation planes and both operational regimes
(Fig. 1.22). The lens was fabrication via micro-machining (Fig. 1.11c) in Rexolite
and excited by a five-element non-equally-spaced linear array (in H-plane) of
aperture-coupled microstrip patch antennas. The array is symmetric (the central
patch is located at the origin of the coordinate system) and all patches are excited in
phase. Modification of the lens surface illumination laws is provided by a selective
1 Integrated Lens Antennas 27

Fig. 1.22 Shape ILA with a reconfigurable field of view: a CAD model, b shaped lens
micromachined in Rexolite material, c, d back view of the BFM networks used to excite the
primary feed array for SR and LR regimes, respectively

excitation of the array elements. To simplify the architecture of the beamforming

network (BFN), the excitation amplitude of the array elements was limited to two
states ON or OFF. The lens shape was optimized to transform a sub-array pattern
(corresponding to a specific state of the BFN) into a prescribed far-field beam. The
design goal is to enable switching between two beams with half power beam widths
of 10°10° and 10°40°, respectively.
To facilitate fabrication, the antenna was designed to operate at 24 GHz with
two operational regimes implemented using two different BFN configurations
comprising three central ones for the SR regime (Fig. 1.22c) and four offset ones
for the LR regime (Fig. 1.22d), respectively. The simulated and measured patterns
for both regimes are represented in Fig. 1.23 and Fig. 1.24, respectively. The slight
asymmetry of the measured curves was attributed to the manual alignment toler-
ances in the assembly of the multilayer BFN. Nevertheless, good agreement
between the simulated (FDTD) and measured curves was obtained with the desired
beam width values achieved in both regimes. The other radiation characteristics of
the prototype operating in the wide and narrow-beam regimes were determined to
be the following: measured gain is 12.6 and 16.6 dB with the operational band-
width S11 < −10 dB is 14.6% (22.3–25.8 GHz) and 5% (24.7–25.5 GHz),
28 A.V. Boriskin et al.

Fig. 1.23 Radiation patterns of the ILA operating at 24 GHz in the wide-beam regime:
a elevation E-plane, b azimuth H-plane. Co-polar measured: , co-polar simulated: ,
X-polar measured: , template:

Fig. 1.24 Same as in Fig. 1.23 for the ILA operating in the narrow-beam regime

respectively. The matching difficulties in the narrow-beam regime were associated

with fabrication intolerances, and thus, in principle, can be avoided. The radiation
patterns fit well the target template within the entire operational bandwidth. The
realized antenna radiation efficiency was estimated as 58 and 53% for the wide- and
narrow-beam regimes, respectively. A quite high insertion loss was attributed to the
material losses associated with the lens and BFN.
This example is reported as a proof of concept for ILAs with a highly recon-
figurable FoV, which can be achieved thanks to a combination of a reconfigurable
primary feed array and shaped dielectric lens. This concept can be generalized for
other frequency bands and types of primary feeds. It can also be implemented to
support a double-band functionality with the same shaped lens enabling two dif-
ferent beamforming functions for different excitation conditions. Such antennas
may find application in a number of application domains including intelligent
transport.
1 Integrated Lens Antennas 29

1.4.5 Beam-Steering ILAs

Beam-steering ILAs are attractive solutions for automotive radars and

point-to-point (or point-to-multipoint) communication systems. An established
solution for such antennas comprises an extended hemielliptical (or hemispherical
lens) fed by a focal array (Fig. 1.2). The weak point of this design is a rapid drop of
the directivity value for scan angles beyond about 20° [6, 20]. The following
examples illustrate three possible alternatives to overcome this limitation.
The first example is a mechanical scanning ILA proposed in Ref. [65] for Low
Earth Orbit satellite data link communications at 26 GHz. The lens was required to
produce a sector beam with shaping in elevation, and to enable a simple mechanical
azimuth scanning. The target gain versus elevation angle h is given approximately
by sec(kh). With an appropriate choice of the k shaping parameter, this elevation
radiation pattern compensates path loss attenuation taking into account the Earth
curvature.
The designed shaped lens is axial-symmetric but a spherical air cavity is exca-
vated through its base at a calculated offset distance with respect to the lens
symmetry axis, Fig. 1.25a. The feed aperture is at the center of this spherical air
cavity. The asymmetry introduced by the feed position produces an adequate
directive shaped beam. This is an extremely simple solution where the feed is fixed
and the lens rotates about the feed axis to provide the required beam scanning. The
feed is not in physical contact with the lens, thus avoiding the need for fault prone
rotary joints in long continuous usage. The assembly is very compact and light-
weight (lens diameter is less than 5 wavelengths and height is less than 3 wave-
lengths) complying with mass and volume restrictions for satellites.
The same concept of a moving lens with fixed feed was extended for azimuth
plus elevation mechanical scanning in a different application context. Wireless
indoor transmission of uncompressed HD video signal between a video device like

Fig. 1.25 a MACORTM axial-symmetric 50 mm diameter shaped lens; b radiation performance

(in polar coordinates) of the lens computed with the GO/PO method [65]
30 A.V. Boriskin et al.

Fig. 1.26 a Photograph of the manufactured polyethylene lens plus horn feed; b measured
radiation patterns of the lens antenna for several lens tilt angles [24]

a camcorder and a HD TV Display may soon become a requirement for mass

market home applications. It proposes to use the unlicensed spectrum from 57 to
66 GHz for such applications. For mass consumer applications, the radio link must
use inexpensive low power sources and hence high gain antennas (>20 dBi) are
required to favor the link budget. The narrow beam needs to be steered to allow for
the user mobility. A new configuration of a mechanical steerable beam antenna has
been developed where a dielectric lens pivots in front of a single stationary mod-
erate gain feed (see Fig. 1.26a) [24]. Again the feed is not in physical contact with
the lens. As discussed previously, this eliminates the need for rotary joints and the
associated drawbacks. The designed lens is made of polyethylene and is such that
both the input and output lens surfaces are shaped to accommodate a beam colli-
mation condition and a scan angle maximization condition. The lens output beam is
aligned with the lens axis for all lens orientations within the operation cone. Thus,
pivoting the lens around two of its main axis allows steering the beam in elevation
and azimuth. The important difference with respect to conventional scanning lenses
is that for all inclinations of the lens the focus is always exactly coincident with the
aperture phase center. In this way, aberrations are strongly reduced allowing wider
scan angles than conventional solutions. The fabricated lens assembly demonstrated
a −45° to +45° elevation scan capability over full azimuth, with 21 dB gain and
better than −1.1 dB gain scan loss (see Fig. 1.26b). The radiation efficiency was
always above 95%. A simple system demonstrator has been recently assembled to
prove the feasibility of the antenna concept, of the tracking mechanism and its
algorithms [66].
The third example of a beam-steering ILA proposed for automotive radar
applications at 77 GHz band [26] is presented in Fig. 1.27. This antenna comprises
two plano-convex dielectric lenses having the same diameter but different radius of
curvature: one attached on top of a planar feed (labeled as substrate lens) and
another one placed at a certain distance above the first one (labeled as objective
lens). As illustrated in Fig. 1.27a via a ray-tracing diagrams, such a topology
enables one to obtain a flatter shape of the focal arc than that observed in case of a
1 Integrated Lens Antennas 31

(a) (b)
33 30

Scan angle (Degree)

DL system 25
32

Directivity (dBi)
31 20

30 15
H-plane
29 E-plane 10
Reference ILA
28 5

27 0
0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7
Feed position, x/R

Fig. 1.27 Double-lens ILA: a topology and ray-tracing diagrams for the proposed DL-ILA and a
standard synthesized elliptical lens considered as a reference solution, b simulated directivity in
both principal planes and scan angle of the DL-ILA and the reference antenna versus feed position
at 77 GHz

(a) (b)
0
x/R = 0.0
Radiation pattern, dB

x/R = 0.2
-5 x/R = 0.4
x/R = 0.6
x/R = 0.8
-10

-15

-20

-25
-45 -30 -15 0 15 30 45
Scan angle (Degree)

Fig. 1.28 a Prototype of the DL-ILA, b measured radiation pattern in H-plane at 77 GHz

standard synthesized elliptical lens reported in [6]. Reciprocally, a lower scan loss is
achieved for off-axis feed positions (Fig. 1.27b). In this study, both ILAs have
diameter 2R = 16  k0, which corresponds to a maximum directivity of 34 dBi (for
a uniform field distribution). In case of the double lens (DL) system, the distance
between the lenses and thickness of the objective lens have been optimized aiming
at achieving a minimum scan loss for the angular range of ±30°. The optimal
design reported in [26] provided 1.1 and 0.9 dB scan loss for the most offset
position in the E and H planes, respectively. Comparison with the reference syn-
thesized elliptical ILA [6] antenna shows a 3.9 dB advantage in terms of the
directivity scan loss at 30° for the DL system that has been achieved by the expense
of 1 dB loss in terms of the broadside directivity.
The DL-ILA design concept was experimentally validated via prototyping
(Fig. 1.28). The two 62.4 mm diameter lenses were micromachined in a thermo-
plastic polyetherimide resin (Ultem™, e = 3.01). The off-axis characteristics of the
antenna were measured by changing the feed position with respect to optical axis of
the DL focusing system. For this purpose, a special support and positioning system
was used, enabling synchronized displacement of the lenses with respect to the
32 A.V. Boriskin et al.

primary feed position. The antenna was fed by an aperture-coupled patch antenna
excited through a waveguide-to-microstrip transition, coupled (S11 < −10 dB) from
74.5 to 80 GHz.
The radiation pattern of the antenna was measured at three frequency points, i.e.,
75, 77, and 79 GHz. The realized scan loss in both principal planes and all fre-
quency points does not exceed 1.2 and 1.0 dB for the E and H-planes, respectively.
For brevity, only the 77-GHz pattern in presented in Fig. 1.28b.
The obtained results confirmed a strong potential of the DL concept for ILAs
with a wide FoV and/or low scan loss performance. Additional advantages of the
DL-ILA revealed by this example include a smaller total height of the focusing
system and, consequently, reduced weight and smaller material loss. Analysis
showed that the latter advantage may fully compensate the additional scattering loss
associated with the objective lens. A potential weak point of DL-ILA is complexity
with assembly and alignment of the DL focusing system. A possible extension of
this concept may include integration of a substrate ILA with a shaped dome acting
as an objective lens. This approach may help further improve FoV as well as
enlarge the operational bandwidth.

Acknowledgements The authors acknowledge contributions of the former students and postdocs
at IETR (France), namely: Barbara Barès, Gaël Godi, Anthony Rolland, Tinh Ngyen, and oth-
ers and at IT (Portugal): Eduardo Lima, Carla Medeiros, Catarina Cruz, and others. The authors
also thank Laurent Le Coq for antenna measurements. The authors acknowledge the collaboration
from Vasco Fred, Jorge Farinha, and Carlos Brito for prototype construction, and António Almeida
and Jorge Silva for prototype measurements. The authors also thank Maarten van der Vorst for
discussions on some aspects of the lens design. The authors are also indebted to numerous students
that actively contributed to this work in the framework of their theses.

References

1. M.R. Rebeiz, Millimeter-wave and terahertz integrated circuit antennas. Proc. IEEE 80(11),
1748–1770 (1992)
2. J.V. Rudd, D.M. Mittleman, Influence of substrate-lens design in terahertz time-domain
spectroscopy. J. Opt. Soc. Am. B 19(2), 319–329 (2002)
3. C.A. Fernandes, P.O. Francês, A. Barbosa, Shaped coverage of elongated cells at millimetre
waves using a dielectric lens antenna, in European Microwave Conference (EMC), 1995,
Bolonha, Italy, September 1995, vol. 1 (1995), pp. 66–70
4. C.A. Fernandes, Shaped-beam antennas, in Handbook of Antennas in Wireless
Communications, Chapter 15, ed. by L.C. Godara (CRC, New York, 2002)
5. C. Salema, C.A. Fernandes, R.K. Jha, Solid Dielectric Horn Antennas (Artech House, Boston,
1998)
6. D.F. Filipovic, S.S. Gearhart, G.M. Rebeiz, Double-slot antennas on extended hemispherical
and elliptical silicon dielectric lenses. IEEE Trans. Microw. Theory Tech. 41(10), 1738–1749
(1993)
7. D.B. Rutledge, M.S. Muha, Imaging antenna arrays. IEEE Trans. Antennas Propag. AP-30
(4), 535–540 (1982)
1 Integrated Lens Antennas 33

8. A.V. Boriskin, G. Godi, R. Sauleau, A.I. Nosich, Small hemielliptic dielectric lens antenna
analysis in 2-D: boundary integral equations versus geometrical and physical optics. IEEE
Trans. Antennas Propag. 56(2), 485–492 (2008)
9. A.V. Boriskin, R. Sauleau, Drastic influence of the half-bowtie resonances on the focusing
and collimating capabilities of 2-D extended hemielliptical and hemispherical dielectric
lenses. J. Opt. Soc. Am. A 27(11), 2442–2449 (2010)
10. N.T. Nguyen, R. Sauleau, C.J. Martínez Pérez, Very broadband extended hemispherical
lenses: role of matching layers for bandwidth enlargement. IEEE Trans. Antennas Propag. 57
(7), 1907–1913 (2009)
11. N.T. Nguyen, N. Delhote, M. Ettorre, D. Baillargeat, L. Le Coq, R. Sauleau, Design and
characterization of 60-GHz integrated lens antennas fabricated through ceramic stereolithog-
raphy. IEEE Trans. Antennas Propag. 58(8), 2757–2762 (2010)
12. A.V. Boriskin, R. Sauleau, A.I. Nosich, Performance of hemielliptic dielectric lens antennas
with optimal edge illumination. IEEE Trans. Antennas Propag. 57(7), 2193–2198 (2009)
13. B. Barès, R. Sauleau, L. Le Coq, K. Mahdjoubi, A new accurate design method for
millimeter-wave homogeneous dielectric substrate lens antennas of arbitrary shape. IEEE
Trans. Antennas Propag. 53(3), 1069–1082 (2005)
14. G. Godi, R. Sauleau, L. Le Coq, D. Thouroude, Design and optimization of three-dimensional
integrated lens antennas with genetic algorithm. IEEE Trans. Antennas Propag. 55(3), 770–
775 (2007)
15. C.A. Fernandes, E.B. Lima, J.R. Costa, Broadband integrated lens for illuminating reflector
antenna with constant aperture efficiency. IEEE Trans. Antennas Propag. 58(12), 3805–3813
(2010). doi:10.1109/TAP.2010.2078463
16. N.T. Nguyen, A.V. Boriskin, A. Rolland, L. Le Coq, R. Sauleau, Shaped lens-like dome for
UWB antennas with a Gaussian-like radiation pattern. IEEE Trans. Antennas Propag. 61(4),
1658–1664 (2013)
17. N.T. Nguyen, A.V. Boriskin, A. Rolland, L. Le Coq, R. Sauleau, Size and weight reduction of
integrated lens antennas using a cylindrical air cavity. IEEE Trans. Antennas Propag. 60(12),
5993–5998 (2012)
18. L. Xue, V. Fusco, Patch fed planar dielectric slab extended hemi-elliptical lens antenna. IEEE
Trans. Antennas Propag. 56(3), 661–666 (2008)
19. D.F. Filipovic, G.P. Gauthier, S. Raman, G.M. Rebeiz, Off-axis properties of silicon and
quartz dielectric lens antennas. IEEE Trans. Antennas Propag. 45(5), 760–766 (1997)
20. X. Wu, G.V. Eleftheriades, T.E. van Deventer-Perkins, Design and characterization of single-
and multiple-beam mm-wave circularly polarized substrate lens antennas for wireless
communications. IEEE Trans. Microw. Theory Tech. 49(3), 431–441 (2001)
21. A. Karttunen, J. Ala-Laurinaho, R. Sauleau, A.V. Räisänen, Extended hemispherical
integrated lens antenna with feeds on a spherical surface, in Proceedings European
Conference Antennas Propagation (EuCAP), Gothenburg, Sweden (2013), pp. 2539–2543
22. A.V. Boriskin, R. Sauleau, Numerical investigation into the design of shaped dielectric lens
antennas with improved angular characteristics. Prog. Electromagn. Res. B 30, 279–292
(2011)
23. G.C. Trichopoulos, G. Mumcu, K. Sertel, L.H. Mosbacker, P. Smith, A novel approach for
improving off-axis pixel performance of Terahertz focal plane arrays. IEEE Trans. Microw.
Theory Tech. 58(7), 2014–2021 (2010)
24. J.R. Costa, E.B. Lima, C.A. Fernandes, Compact beam-steerable lens antenna for 60-GHz
wireless communications. IEEE Trans. Antennas Propag. 57(10), 2926–2933 (2009). doi:10.
1109/TAP.2009.2029288
25. A.L. Peebles, A dielectric bifocal lens for multibeam antenna applications. IEEE Trans.
Antennas Propag. 31(5), 599–606 (1988)
26. N.T. Nguyen, A.V. Boriskin, L. Le Coq, R. Sauleau, Improvement of scanning performance
of integrated lens antenna using a double lens focusing system. IEEE Trans. Antennas Propag.
64(8), 3698–3702 (2016)
34 A.V. Boriskin et al.

27. X. Wu, G.V. Eleftheriades, Two-lens and lens-fed reflector antenna systems for MM-wave
wireless communications, in Proceedings of the IEEE Antennas Propagation Symposium
(APS), Salt Lake City, UT (2000), pp. 660–663
28. B. Schoenlinner, X. Wu, J.P. Ebling, G.V. Eleftheriades, G.M. Rebeiz, Wide-scan
spherical-lens antennas for automotive radars. IEEE Trans. Microw. Theory Tech. 50(9),
2166–2175 (2002)
29. J. Thornton, S. Gregson, D. Gray, Aperture blockage and truncation in scanning lens-reflector
antennas. IET Microw. Antennas Propag. 4(7), 828–836 (2010). doi:10.1049/iet-map.2009.
0283
30. Z. Sipus, et al., This book/Chapter 2
31. K.H. Jeong, J. Kim, L.P. Lee, Biologically inspired artificial compound eyes. Science 312,
557–561 (2006)
32. C.A. Fernandes, E.B. Lima, J.R. Costa, Tapered waveguide feed for integrated dielectric lens
antenna performance tests, in EUROCON—International Conference on Computer as a Tool
(EUROCON), 27–29 April 2011 (IEEE, 2011), pp. 1–4
33. J.R. Costa, C.A. Fernandes, Broadband slot feed for integrated lens antennas. IEEE Antennas
Wirel. Propag. Lett. 6, 396–400 (2007). doi:10.1109/lawp.2007.900954
34. A. Neto, UWB, Non dispersive radiation from the planarly fed leaky lens antenna—Part 1:
Theory and design. IEEE Trans. Antennas Propag. 58(7), 2238–2247 (2010)
35. N.T. Nguyen, R. Sauleau, M. Ettorre, L. Le Coq, Focal array fed dielectric lenses: an
attractive solution for beam reconfiguration at millimeter waves. IEEE Trans. Antennas
Propag. 59(6), 2152–2159 (2011)
36. D. Lemaire, C.A. Fernandes, P. Sobieski, C.A. Barbosa, A method to overcome the
limitations of G.O. in the design of axis-symmetrical lenses. Int. J. Infrared Millimetrew. 17
(8), 1377–1390 (1996)
37. D. Pasqualini, S. Maci, High-frequency analysis of integrated dielectric lens antennas. IEEE
Trans. Antennas Propag. 52(3), 840–847 (2004)
38. P. Otero, G.V. Eleftheriades, J.R. Mosig, Integrated modified rectangular loop slot antenna on
substrate lenses for millimeter- and submillimeter-wave frequencies mixer applications. IEEE
Trans. Antennas Propag. 46(10), 1489–1497 (1998)
39. H. Hoteit, R. Sauleau, B. Philippe, Ph Coquet, J.-P. Daniel, Vector and parallel
implementations for the FDTD analysis of millimeter wave planar antennas. Int. J. High
Speed Comput. 10(2), 209–234 (1999)
40. A.V. Boriskin, A.I. Nosich, S.V. Boriskina, T.B. Benson, P. Sewell, A. Altintas, Lens or
resonator? Electromagnetic behavior of an extended hemielliptical lens for a sub-mm wave
receiver. Microw. Opt. Technol. Lett. 43(6), 515–518 (2004)
41. A.V. Boriskin, A. Rolland, R. Sauleau, A.I. Nosich, Assessment of FDTD accuracy in the
compact hemielliptic dielectric lens antenna analysis. IEEE Trans. Antennas Propag. 56(3),
758–764 (2008)
42. G. Godi, R. Sauleau, D. Thouroude, Performance of reduced size substrate lens antennas for
millimeter-wave communications. IEEE Trans. Antennas Propag. 53(4), 1278–1286 (2005)
43. E.B. Lima, J.R. Costa, M.G. Silveirinha, C.A. Fernandes, ILASH—software tool for the
design of integrated lens antennas, in 2008 IEEE Antennas and Propagation Society
International Symposium, vols. 1–9 (IEEE, New York, 2008), pp. 863–866
44. J.R. Costa, C.A. Fernandes, Analysis of small integrated lens antennas using multiple point
source physical optics discretization of the feed aperture field. Journees Internationales de
Nice sur les Antennes—JINA 1, 466–467 (2004)
45. Computer Simulation Technology, CST, http://www.cst.com/
46. D.S. Alexei, N.G.T. Gregory, S. Roman, Hot-electron effect in superconductors and its
applications for radiation sensors. Supercond. Sci. Technol. 15(4), R1 (2002)
47. P.H. Siegel, THz instruments for space. IEEE Trans. Antennas Propag. 55(11), 2957–2965
(2007). doi:10.1109/TAP.2007.908557
1 Integrated Lens Antennas 35

48. L. Liu, H. Xu, R.R. Percy, D.L. Herald, A.W. Lichtenberger, J.L. Hesler, R.M. Weikle,
Development of integrated terahertz broadband detectors utilizing superconducting
hot-electron bolometers. IEEE Trans. Appl. Supercond. 19(3), 282–286 (2009). doi:10.
1109/TASC.2009.2018268
49. A.D. Semenov, H. Richter, H.W. Hubers, B. Gunther, A. Smirnov, K.S. Il’in, M. Siegel, J.
P. Karamarkovic, Terahertz performance of integrated lens antennas with a hot-electron
bolometer. IEEE Trans. Microw. Theory Tech. 55(2), 239–247 (2007). doi:10.1109/TMTT.
2006.889153
50. J.R. Costa, C.A. Fernandes, Integrated imaging lens antenna with broadband feeds, in The
Second European Conference on Antennas and Propagation, 2007, EuCAP 2007 (2007),
pp. 1–6
51. M.G. Silveirinha, C.A. Fernandes, J.R. Costa, A graphical aid for the complex permittivity
measurement at microwave and millimeter wavelengths. IEEE Microw. Wirel. Compon. Lett.
24(6), 421–423 (2014). doi:10.1109/LMWC.2014.2310470
52. M.N. Afsar, X. Li, H. Chi, An automated 60 GHz open resonator system for precision
dielectric measurements. IEEE Trans. Microw. Theory Tech. 38(12), 1845–1853 (1990).
doi:10.1109/22.64565
53. C.A. Fernandes, J.R. Costa, Permittivity measurement and anisotropy evaluation of dielectric
materials at millimeter-waves, in XIX IMEKO World Congress: Fundamental and Applied
Metrology, Proceedings, Budapest 2009 (IMEKO, 2009), pp. 673–677
54. M.J.M. van der Vorst, P.J.I. de Maagt, A. Neto, A.L. Reynolds, R.M. Heeres, W. Luinge, M.
H.A.J. Herben, Effect of internal reflections on the radiation properties and input impedance
of integrated lens antennas-comparison between theory and measurements. IEEE Trans.
Microw. Theory Tech. 49(6), 1118–1125 (2001)
55. K.W. Kim, Y. Rahmat-Samii, Spherical Luneburg lens antennas: engineering characteriza-
tions including air gap effects, in Antennas and Propagation Society International
Symposium, 21–26 June 1998, vol. 2064 (IEEE, 1998), pp. 2062–2065
56. N.T. Nguyen, R. Sauleau, C.J. Martinez Perez, M. Ettorre, Finite-difference time-domain
simulations of the effects of air gaps in double-shell extended hemispherical lenses. IET
Microw. Antennas Propag. 4(1), 35–42 (2010). doi:10.1049/iet-map.2008.0255
57. A. Bisognin, D. Titz, F. Ferrero, G. Jacquemod, R. Pilard, F. Gianesello, D. Gloria, D.
Lugara, E.B. Lima, J.R. Costa, C.A. Fernandes, C. Luxey, Noncollimating MMW
polyethylene lens mitigating dual-source offset from a Tx/Rx WiGig module. IEEE Trans.
Antennas Propag. 63(12), 5908–5913 (2015). doi:10.1109/TAP.2015.2484420
58. A. Bisognin, A. Cihangir, C. Luxey, G. Jacquemod, R. Pilard, F. Gianesello, J.R. Costa, C.A.
Fernandes, E.B. Lima, C.J. Panagamuwa, W.G. Whittow, Ball grid array-module with
integrated shaped lens for wigig applications in eyewear devices. IEEE Trans. Antennas
Propag. 64(3), 872–882 (2016). doi:10.1109/TAP.2016.2517667
59. L. Fernandes, Developing a system concept and technologies for mobile broadband
communications. IEEE Pers. Commun. 2(1), 54–59 (1995). doi:10.1109/98.350865
60. C.A. Fernandes, Design of shaped lenses for non-symmetric cells in MBS, in Antennas and
Propagation Society International Symposium, 1999, 11–16 July 1999, vol. 2444 (IEEE,
1999), pp. 2440–2443
61. C.A. Fernandes, Shaped dielectric lenses for wireless millimeter-wave communications. IEEE
Antennas Propag. Mag. 41(5), 141–150 (1999). doi:10.1109/74.801527
62. C.A. Fernandes, L.M. Anunciada, Constant flux illumination of square cells for
millimeter-wave wireless communications. IEEE Trans. Microw. Theory Tech. 49(11),
2137–2141 (2001). doi:10.1109/22.963149
63. C.C. Cruz, J.R. Costa, C.A. Fernandes, S.A. Matos, Focal-plane multi-beam dual-band
dielectric lens for Ka-band. IEEE Antennas Wirel. Propag. Lett. 16, 432–436 (2017). doi:10.
1109/LAWP.2016.2582263
64. J.R. Costa, M.G. Silveirinha, C.A. Fernandes, Evaluation of a double-shell integrated
scanning lens antenna. IEEE Antennas Wirel. Propag. Lett. 7, 781–784 (2008). doi:10.1109/
lawp.2008.2008403
36 A.V. Boriskin et al.

65. J.R. Costa, C.A. Fernandes, G. Godi, R. Sauleau, L. Le Coq, H. Legay, Compact Ka-band
lens antennas for LEO satellites. IEEE Trans. Antennas Propag. 56(5), 1251–1258 (2008).
doi:10.1109/tap.2008.922690
66. J.R. Costa, E.B. Lima, C.R. Medeiros, T. Radil, R.C. Martins, P.M. Ramos, C.A. Fernandes,
Development of an indoor wireless personal area network based on mechanically steered
millimeter-wave lens antenna, in Instrumentation and Measurement Technology Conference
(I2MTC), 3–6 May 2010 (IEEE, 2010), pp. 1202–1206
Chapter 2
Multi-shell Radially Symmetrical Lens
Antennas

Zvonimir Sipus and Tin Komljenovic

Abstract In this chapter, we review several aspects of the analysis and the design
of multi-shell spherical and circular cylindrical lens antennas. Such lens antennas
are attractive for implementation in communication and radar systems, in particular
in the millimeter-wave frequency band, due to their broadband behavior, excellent
focusing properties, possibility of beam scanning, and the ability to form multiple
beams. In order to develop an efficient analysis tool, needed for successful design of
multi-shell lens antennas, we first demonstrate the principles of the analysis algo-
rithm for calculating the EM field distribution in general multilayer structures (i.e.,
inside a structure with an arbitrary number of layers). This algorithm is designed for
spherical and circular-cylindrical geometries with elementary excitation. To model
realistic lens antennas, we introduce additional flexibility that allows the analysis of
actual feed antennas that usually do not follow the symmetry properties of the lens.
Finally, by connecting the algorithm with an optimization subroutine, a powerful
analysis and design tool is created. All the aspects of the proposed analysis
approach are explained and illustrated with examples. Furthermore, some practical
problems which are encountered in the design of these types of lens antennas are
highlighted and common solutions are presented and compared to the ideal
situations.

Z. Sipus (&)
Faculty of Electrical Engineering and Computing, University of Zagreb,
Unska 3, HR-10000 Zagreb, Croatia
e-mail: [email protected]
T. Komljenovic
Department of Electrical and Computer Engineering, University of California,
Santa Barbara, CA 93106, USA

© Springer International Publishing AG 2018 37

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_2
38 Z. Sipus and T. Komljenovic

2.1 Introduction

In recent years, there is a growing interest for antennas in communication and radar
systems at higher microwave and millimeter-wave frequencies. Different kinds of
spherical and cylindrical lens antennas are attractive solutions for such applications
due to their focusing properties, capability of beam scanning, ability to form
multiple beams, and broadband behavior. Furthermore, the typical problems with
microwave lenses in the past, such as their bulkiness and weight are no longer an
obstacle for practical implementation since by moving to higher frequencies and by
using novel materials the size and weight have been considerably reduced. This
added degree of freedom in terms of weight and size has in many cases led to the
increase in the electrical size of the lenses in order for them to meet requirements
for more demanding applications. In addition, the feed can be any type of antenna
(horn, dipole, microstrip patch), and typically it is a part of the mechanical system
that points the main beam in the desired direction. Altogether, the lens antenna
system is a complex electromagnetic design problem that requires efficient and
accurate analysis tools that can take into account various practical problems and
demands.
The inhomogeneous lenses have first been analysed and designed using geo-
metrical optics and ray-tracing techniques [1–4]. However, these are approximate
methods valid for structures with all the dimensions large compared to the wave-
length. Furthermore, they cannot take into account internal reflections. The general
numerical methods, such as the finite difference time domain [5] or the finite
element method in [6], can be applied to the lens analysis. However, such methods
are suitable for structures having limited electric dimensions, usually not more than
20 free space wavelengths, and are time- and memory-consuming. Therefore, there
is a need for specialized software in which the geometrical properties of the con-
sidered structures are analytically taken into account, and which can be efficiently
connected with an optimization routine.
The analytical solution for spherical and circular-cylindrical multilayer structures
is derived from the solution of the homogeneous Helmholtz equation, known as
Mie series. It has been extensively used and reported for plane wave scattering by a
conducting and dielectric sphere [7, 8]. This approach has been generalized for
multilayered dielectric lens using different formulations: Mie series augmented
matrix solution [9, 10], scalar potentials [11], mode matching technique [12–15] or
dyadic Green functions [16–19]. The similar analysis approach can be applied for
analyzing other types of structures (e.g., conformal microstrip antennas [20–23]).
2 Multi-shell Radially Symmetrical Lens Antennas 39

2.2 Analysis of Spherical and Circular Cylindrical

Multilayer Structures

For many structures and technical problems, one requires the knowledge of the field
distribution scattered and/or inside multilayer spherical or cylindrical structures.
Examples would be lens antennas, radomes, optical fibers, geophysical probing,
scattering reduction structures (cloaking), etc. In all the considered cases we would
like to determine the EM field distribution inside and around the multilayer
structure while the position and/or existence of the source can be arbitrary. For
example, in scattering problems the sources are faraway from the structure, in
waveguide problems we search for eigenmode solutions, while in the antenna
problems the sources are a part of the structure. In general, all of these problems can
be described using the Green’s function approach, i.e., we can consider the physical
or equivalent sources as a part of the structure (the waveguide problems can be
treated as poles of the Green’s functions). The knowledge of the most appropriate
Green’s function is often essential for implementing the numerical method we
would like to apply. For example, planar, cylindrical, and spherical patch antennas
and periodic surfaces are frequently analysed by means of the electric field integral
equation and the moment method. There the kernel of the integral operator is a
Green’s function, which takes into account the considered curved multilayer
structure.

2.2.1 Green’s Functions for Layered Media

In principle, there are two basic approaches for calculating the Green’s function of
a general multilayer structure: either to analytically derive an expression for it and
then code this expression, or to develop a numerical routine for the complete
calculation. The analytic approach requires less computer resources than the
numerical approach; however, it is very laborious to analytically determine the
Green’s functions for structures with more than two layers. In such cases, it is more
convenient to use a numerical algorithm which determines the Green’s function
directly. Furthermore, the analytic approach often requires a new derivation of the
Green’s functions for practically every slightly different problem, such as for dif-
ferent number of layers or for different source locations inside the layers.
In most cases, lenses follow spherical or cylindrical geometry, or a slight vari-
ation of these. Therefore, it is reasonable to focus our attention on these geometries
and use the inherent properties of these structures to our advantage. Ideal planar,
circular cylindrical, and spherical multilayer structures have one property in com-
mon: the structure is homogeneous in two dimensions, and varies in the third
dimension. For example, the spherical structure in Fig. 2.1 varies in r direction and
is homogeneous in h and / directions. Thus, planar, cylindrical and spherical
structures can be referred to as one-dimensional structures since they vary only in
40 Z. Sipus and T. Komljenovic

Fig. 2.1 Spherical lens

antenna illuminated by a horn
antenna

one dimension [24]. We can simplify the problem of determining the field distri-
bution Green’s functions for one-dimensional structures if we perform the
two-dimensional (2D) Fourier transformation in the coordinates for which the
structure is homogeneous. As a result, our original three-dimensional problem is
transformed into a series of one-dimensional problems, which are much easier to
solve. Instead of using Fourier transformation one can search for modal represen-
tation of the EM field, which at the end gives the same EM field expression.
When determining the Green’s functions of multilayer one-dimensional struc-
tures one can select two paths. The first is to select one component of electric and
magnetic potential and then to express the field components at the boundaries to
determine the considered potentials. In practice, the algorithm determines the
reflection and transmission coefficients of TE and TM waves which are connected
(simply by multiplication) in a final matrix that describes the structure (see, e.g.,
[16]). For example, in the spherical case we can use the Debye potentials we and wm
to characterize the TM and TE waves (note that r ¼ r^r )

ETE ¼ r  rwm ð2:1aÞ

HTM ¼ r  rwe ð2:1bÞ

The Debye potentials we and wm are determined by solving the scalar Helmholtz
differential equation (in spherical coordinate system in this case)

wmn ¼ zn ðkrÞPm
n ðcos hÞe
jm/
ð2:2Þ

Here wn is the elementary solution of the Helmholtz differential equation, i.e.,

Pm
n ðcos hÞ are the associated Legendre functions of the first kind and zn denotes
ð2Þ
spherical Bessel jn or Hankel hn functions. Using the following notation:
2 Multi-shell Radially Symmetrical Lens Antennas 41

Mmn ¼ r  rwmn ð2:3aÞ

1
Nmn ¼ r  Mmn ð2:3bÞ
k

we can express the electromagnetic field as

XX
E¼ amn Mmn þ bmn Nmn : ð2:4aÞ
n m

j XX
H¼ bmn Mmn þ amn Nmn : ð2:4bÞ
g n m

The coefficients are determined by fulfilling the boundary conditions for the
tangential EM field components (e.g., for electric field)
1 X h
X i jm jmj

Etan ¼  jn ðkrÞ þ ahnm hð2Þ ^ ^ d Pjmj ðcos hÞ ejm/
n ðkrÞ Pn ðcos hÞ  /
j
anm h
n¼0 jmj  n
sin h dh n
 
1 dh j ð2Þ
i
^ d jmj ^ jm jmj
þ b rjn ðkrÞ þ bnm rhn ðkrÞ h Pn ðcos hÞ þ /
h
P ðcos hÞ ejm/
kr dr nm dh sin h n
ð2:5Þ

By fulfilling the boundary conditions at each boundary (continuity of the tan-

gential components) the reflection coefficients can be determined, see [16] for
details.
The second approach is based on determining the tangential components of the
EM field (so-called propagator matrix approach [16]). This approach actually
corresponds to the implementation of the Huygens’s principle or more precisely, the
Love’s equivalence theorem [8]. Here the original problem is divided into equiv-
alent subproblems, one for each layer. The field inside each layer is then determined
from equivalent currents at the subproblem boundaries and from the sources inside
that subproblem. As a theoretical background, in parallel to the Love’s equivalence
theorem, we can modify the curl Maxwell’s equation to “eliminate” the normal
component of the EM field, i.e., to get the equations that depend only on the
tangential field and source components. This modification is given in [25] and the
resulting equations are (e.g., for H-field; by duality principle the equivalent equa-
tion can be derived for the E-field)
 
@ jg rt rt
 Ht ¼ 1þ 2  ð^n  Et Þ þ ^
n  Jte
@n k k
ð2:6Þ
j
Jte ¼ Jt  r  Mn
gk
42 Z. Sipus and T. Komljenovic

Here subscripts t and n denote tangential and normal directions (related to the
considered boundary), ^n is the unit vector in the normal direction, and rt is the
transverse part of the r operator. In other words, in order to determine the EM
fields inside the considered multilayer structure we need to know the tangential EM
field components at each boundary as well as tangential component of the excitation
currents inside the structure. Note that the excitation current in Eq. (2.6) contains
two parts: the tangential electric current Jt and the replacement tangential electric
current with the origin in the normal component of the excitation magnetic current
Mn .

2.2.2 Description of the G1DMULT Algorithm

As indicated, a canonical three-dimensional problem can be efficiently decomposed

into one-dimensional subproblems by applying Fourier transformation for the
directions where the structure is homogeneous. This so-called spectral domain
approach is the core of an algorithm called G1DMULT for calculating the Green’s
function in the spectral domain (representing a one-dimensional (1D) spatial
domain) for planar, circular cylindrical, and spherical multilayer structures. The
G1DMULT algorithm calculates the Green’s functions in the same way for all three
types of geometries and it will be explained by considering the spherical geometry.
The solution procedure makes use of the Fourier transformation technique. Since
the problem is described in spherical coordinate system, we use the vector Legendre
transformation in h and / directions, defined by [26, 24].

Zp Z p 
1
~Jðr; n; mÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi Lðn; m; hÞJðr; h; /Þ sin h ejm/ dh d/ ð2:7aÞ
2pSðn; mÞ
p 0

X
1 X
1
1  
Jðr; h; /Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Lðn; m; hÞ J ðr; n; mÞejm/ ð2:7bÞ
m¼1 n¼jmj 2pSðn; mÞ
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3
P jnmj ðcos hÞ nðn þ 1Þ 0 0
 6 jmj jmj 7
Lðn; m; hÞ ¼ 6
4 0 @P n ðcos hÞ
@h
jmPn ðcos hÞ 7
sin h 5 ð2:7cÞ
jmj jmj
jmPn ðcos hÞ @P n ðcos hÞ
0 sin h @h

2nðn þ 1Þðn þ jmjÞ!

Sðn; mÞ ¼ : ð2:7dÞ
ð2n þ 1Þðn  jmjÞ!

By applying the vector Legendre transformation, the three-dimensional excita-

tions are transformed into harmonic current shells. If the source is infinitely thin in
r direction, we get one discrete current shell per source, otherwise we get a
2 Multi-shell Radially Symmetrical Lens Antennas 43

continuous distribution of current shells in r direction. The E- and H-fields induced

by the harmonic current sources have the same harmonic variations in h and / as
the source. Therefore, only the field variation in the direction perpendicular to the
boundaries is unknown, and we have a harmonic one-dimensional (1D) field
problem. In this way, the spectral domain problem is interpreted as a
one-dimensional spatial domain problem consisting of one-dimensional multilayer
structure and harmonic one-dimensional sources in the form of current shells. This
is visualized in Fig. 2.2, where also all other steps of the algorithm are shown. We
describe the steps of the algorithm below.
The harmonic one-dimensional problem is solved by making use of the equiv-
alent problems, one for each layer. The unknowns are the tangential E- and H-fields
at layer boundaries. Since the variation of the E- and H-fields in the direction
tangential to the boundaries is harmonic with known periodicity, we only need to
determine the complex field amplitudes at the interfaces, i.e., we have four
unknowns per boundary. For example, the E-field in the layer j is expressed as

~ homo ~Jj1 þ G
~j ¼ G
E ~ homo ~Jj þ G
~ homo M ~ homo M
~ j1 þ G ~ homo ~
~ jþG exci
~ homo M
þG ~ exci ;
EJ EJ EM EM EJ Jj EM j
ð2:8Þ

where ~Jj and M ~ j are equivalent electric and magnetic current sheets at boundary j,
~Jj and M
exci ~ j are excitation electric and magnetic currents in layer j (if any), and
exci

~
Ghomo is the Green’s function of the homogeneous problem (given below). By using
J~j ¼ ^n  H~ j and M~ j ¼ ^n  E
~ j Eq. (2.8) can be expressed in terms of the unknown
~ ~
EM field amplitudes Ej and Hj at the boundary j between layers j and j + 1 and the
known excitation currents. Since the tangential E- and H-fields are continuous at the
layer boundaries we obtain 4 linear equations per boundary.
The algorithm connects all equivalent subproblems into a system of 4Nboundary
linear equations with the same number of unknowns (Nboundary denotes the number
of boundaries). Once the amplitudes of the tangential fields have been determined,
it is easy to determine the field amplitudes anywhere in the multilayer structure by
applying the homogeneous region equivalent of the layer inside which we want to
determine the field value.
The core subproblem in the formulation is to calculate the E- and H-fields due to
a harmonic current shell excitation of radius rs in a homogeneous region. This is the
only part (i.e., the only subroutine) that is different for planar, cylindrical and
spherical case. In the spherical case the formulation is as follows:
(a) from ~Jh ðrs ; m; nÞ

(
^ nð2Þ0 ðkrs Þ^
Jn0 ðkrÞ r  rs
~ h ðr; n; mÞ ¼  k rs ~Jh ðn; mÞ 
E
H
ð2:9aÞ
^Jn ðkrs ÞH
0 ð2Þ0
^ n ðkrÞ
xe r r  rs
44 Z. Sipus and T. Komljenovic

(a) 3-D problem

z θ
exci
J (x,y) z y

x
φ
φ
ρ
exci
z J (φ,z) x
Jexci (θ,φ)

(b) Harmonic 1-D problem

_ ~ jmφ
~exci
J (kx ,ky ) e-jkxx e-jkyy L(n,m,θ) Jexci(n,m) e

~exci
J (m,kz ) e-jmφ e-jkz z

(c) Definition of material boundaries and regions

ρ r
4 ρ 3 r 3
z3 ρ 2 r1 2
3 1
2 z2 1 1
z1 2 2
1 3 3
4 4

(d) Equivalent problem of region j

~ ~
Mj Jj ~ ~ ~ ~
Mj Jj Mj Jj

~ ~ ~ ~
~ ~ Mj-1 Jj-1 Mj-1 Jj-1
Mj-1 Jj-1

(e) Basic homogeneous harmonic 1-D field problem

_ ~ jmφ
~ -jkxx -jkyy L(n,m,θ) J(n,m) e
Je e

~ -jmφ -jkz z
Je e

Fig. 2.2 Structure of the G1DMULT algorithm

2 Multi-shell Radially Symmetrical Lens Antennas 45

(
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ^ nð2Þ ðkrs Þ^Jn ðkrÞ r  rs
~ r ðr; n; mÞ ¼  1 rs nðn þ 1Þ ~Jh ðn; mÞ 
E
H
ð2:9bÞ
xe r 2 ^
Jn ðkrs ÞH^ nð2Þ ðkrÞ r  rs
(
^ nð2Þ ðkrs Þ^
Jn ðkrÞ r  rs
~ / ðr; n; mÞ ¼ jrs ~Jh ðn; mÞ 
H
H
ð2:9cÞ
r ^Jn ðkrs ÞH^ nð2Þ ðkrÞ r  rs

(b) from ~J/ ðrs ; m; nÞ

(
^ nð2Þ ðkrs Þ^
Jn ðkrÞ r  rs
~ / ðr; n; mÞ ¼  k rs ~J/ ðn; mÞ 
E
H
ð2:10aÞ
xe r ^Jn ðkrs ÞH^ nð2Þ ðkrÞ r  rs
(
^ nð2Þ0 ðkrs Þ^
Jn0 ðkrÞ r  rs
~ h ðr; n; mÞ ¼ jrs ~J/ ðn; mÞ 
H
H
ð2:10bÞ
^Jn ðkrs ÞH
0 ð2Þ0
^ n ðkrÞ r  rs
r
(
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ^ nð2Þ ðkrs Þ^
Jn ðkrÞ r  rs
~ r ðr; n; mÞ ¼ jrs nðn þ 1Þ ~J/ ðn; mÞ 
H
H
ð2:10cÞ
kr 2 ^Jn ðkrs ÞH^ nð2Þ ðkrÞ r  rs

ð2Þ
Here ^Jn ðkrÞ and H^ n ðkrÞ are the Schelkunoff type of spherical Bessel and Hankel
^ nð2Þ ðxÞ ¼ x hð2Þ
functions ð^Jn ðxÞ ¼ x jn ðxÞ and H n ðxÞÞ, and k is the wave number of the
considered layer. The fields excited by the magnetic shell can easily be determined
by applying the duality concept. The fields caused by the r-directed sources are
evaluated using the transverse replacement currents (see [24] for details).
Once we have the fields inside the structure the procedure to obtain the radiation
pattern is quite straightforward. If we consider, e.g., the /-component of the electric
field in the outermost region with the r-coordinate larger than the r-coordinate of
the patch, we have only outward-traveling waves described by ainm H ^ nð2Þ ðk0 rÞ.
Therefore, in the outermost region we can connect the /-component of the electric
field with different r-coordinates r1 and r2 as

^ ð2Þ nþ1
~ / ðr2 ; n; mÞ r2 Hn ðk0 r1 Þ  E
~ / ðr1 ; n; mÞ ¼ E
E ~ / ðr2 ; n; m Þ j r2 expðjk0 r1 Þ
r1 H ð2Þ
^ n ðk0 r2 Þ ð2Þ
^ n ðk0 r2 Þ r1
H
ð2:11aÞ
0
^ ð2Þ ejk0 r1
E ~ h ðr2 ; n; mÞ r2 Hn 0 ðk0 r1 Þ  E
~ h ðr1 ; n; mÞ ¼ E ~ h ðr2 ; n; mÞ jn r2
0
^ nð2Þ ðk0 r2 Þ
r1 H ^ nð2Þ ðk0 r2 Þ r1
H
ð2:11bÞ
46 Z. Sipus and T. Komljenovic

Here r1 represents the r-coordinate of the far-field pattern. This avoids the
problem of evaluating Hankel functions with very large argument. Notice that in the
lens analysis a slightly larger value than the feed position is chosen for the radius r2.
The final solution is obtained by superposing the spectral solutions, see Eqs. (2.7a–
2.7d).
In order to calculate the directivity radiation pattern it is necessary to calculate
the radiated power. This is done in spectral domain by using Parseval theorem [8],
i.e., the radiated power is equal
2 3
X X
Prad ¼ r22 Re4 ~h H
Sðn; mÞðE ~ / E ~ h Þ5:
~/ H ð2:12Þ
8m n  jmj

To summarize, the presented modal analysis approach (i.e., the spectral domain
approach or the Mie series approach) is completely rigorous (there are no
approximations in the formulation) and suitable for analyzing lenses with arbitrary
number of layers. However, its application is limited only to canonical geometries
(spherical, circular-cylindrical or planar) and one should be aware that numerical
problems can appear for electrically very large structures. Two major sources of
possible numerical difficulties are connected with the problem of the associated
Legendre functions Pm n ðcos hÞ becoming very large (especially if m  n,
Pjnmj ðcos hÞ  njm1=2j , making the solution numerically unstable), and with the
problem of calculating Bessel/Hankel functions of large order. The first numerical
difficulty linked to associated Legendre functions can be solved using suitable
normalization process [22]. The second numerical problem (Bessel/Hankel func-
tions of large order) can be mitigated by implementing the Debye’s asymptotic
formulas for Bessel and Hankel functions of large order, by which either the pro-
duct of Bessel and Hankel function, or the ratio of two Hankel functions, are
calculated with extracted exponential parts [22].
Multilayer canonical one-dimensional structures can also be analyzed by
expressing the field quantities inside each layer as the superposition of forward and
backward propagating waves (as mentioned at the beginning of this chapter). The
reflection and transmission coefficients inside each layer can be determined in a
recursive way, resulting in the generalized reflection and transmission coefficient of
the whole structure (i.e., of the generalized scattering matrix). A detailed descrip-
tion of this method is given in [16]. Although in the presented approach the forward
and backward propagating waves are also considered, the way of determining the
field quantities is different—the presented G1DMULT algorithm simultaneously
determines the tangential field quantities at the interfaces between the layers. Unlike
in [16], in the presented method the field solutions to all the subproblems in the
procedure are given in terms of Green’s functions, i.e., the excitation within each
subproblem are equivalent or physical current sources. This makes it easy to modify
the code. For example, it is easy to implement the metasurface impedance boundary
2 Multi-shell Radially Symmetrical Lens Antennas 47

conditions into the routine, which makes the algorithm suitable for analyzing
multilayer metasurface structures (e.g., metasurface lenses).

2.3 Spherical Multi-shell Lenses

Spherical lens antennas in general have many attractive features like broadband
behavior, excellent focusing properties, wide scan coverage without scanning los-
ses, possibilities of multiple beams on (in general) different working frequencies,
etc. Such properties make lens antennas attractive solution in many applications, in
particular in millimeter frequency range where the weight of such antennas is no
longer an obstacle for practical implementation due to the increased frequency of
operation. Two most common designs follow the spherical Luneburg and half
Maxwell fisheye variation of index of refraction (see Fig. 2.3 and [27]), i.e., the
respective radial dependency of index of refraction is equal
.
n2 ðrÞ ¼ 2  r 2 and n2 ðrÞ ¼ 4 ð1 þ r 2 Þ2 : ð2:13Þ

For example, spherical lens antennas can be used for automotive radars [28],
radar cross-section (RCS) enhancers [29], satellite communications [30, 31], or for
astronomy applications [32].

2.3.1 Modeling of Feed Antennas

As explained in the previous section, spherical lens antennas are usually rigorously
analysed using spherical wave decomposition. However, the feed was usually taken
only in an approximate way. For example, the Huygens source (approximation of
low-gain horns) was considered in [10], dipole sources were considered in [17], and
in [12, 13] the feeding horn antenna was modeled by an array of dipoles repre-
senting the aperture field distribution of the horn. In more details, in the case of horn
feeding, the horn is replaced with equivalent aperture currents by using free space

Fig. 2.3 Ray tracing illustration of the properties of a Luneburg lens, b half Maxwell fisheye lens,
c hemispherical lens with a ground plane
48 Z. Sipus and T. Komljenovic

equivalent principle [8]. For example, if the rectangular horn is used, the field at the
horn aperture in the local coordinate system is
 
py
ejk½x =qx þ y =qy
=2
2 2
Eðx; yÞ ¼ ^x E0 cos ð2:13aÞ
Wy
 
E0 py
ejk½x =qx þ y =qy
=2
2 2
Hðx; yÞ ¼ ^y cos ð2:13bÞ
g Wy

where Wx and Wy are the width of the horn in x- and y-direction, and qx and qy are
distances of the centers of radial waveguides from the horn opening. Therefore, the
equivalent currents are
 
E0 py
ejk½x =qx þ y =qy
=2
2 2
Jeq ðx; yÞ ¼ ^x cos ð2:14aÞ
g Wy
 
py
ejk½x =qx þ y =qy
=2
2 2
Meq ðx; yÞ ¼ ^y E0 cos ð2:14bÞ
Wy

Small horns can be approximated with an elementary area (so-called Huygens

source [33]) with E- and H-fields equal to

dðhÞ dð/Þ dðr  rfeed Þ

Eeq ðrfeed ; h; /Þ ¼ ^h 2 sin h
ð2:15aÞ
rfeed

^ dðhÞ dð/Þ dðr  rfeed Þ

Heq ðfeed ; h; /Þ ¼ / 2 sin h
: ð2:15bÞ
g rfeed

We can assume that the small horn is located at the h = 0° axis with r-coordinate
r = r feed. The vector-Legendre transformation of all described models can be found
almost completely analytically (for the rectangular horn numerical integration in h-
direction is needed, or alternatively one can consider the array of electric and
magnetic dipoles as explained in [13]). For example, the vector-Legendre trans-
formation of the Huygens source is equal
2 3
0
~J ðrfeed ; n; m ¼ 1Þ ¼ 
eq 1 2n þ 1 4 1 5
2 nðn þ 1Þ
ð2:16aÞ
8pg rfeed
jm
2 3
0
~ ðrfeed ; n; m ¼ 1Þ ¼ 
eq 1 2n þ 1 4 jm 5
M 2 nðn þ 1Þ
ð2:16bÞ
8p rfeed
1

Note that only the terms with m = ±1 are different from zero.
2 Multi-shell Radially Symmetrical Lens Antennas 49

Rigorous model, based on closing the feed antenna in a small volume and
representing the near-field distribution via spherical harmonics, was presented in
[15]. The idea is to project the equivalent current distribution (at the interface of a
volume that encloses the source) on the spherical wave basis
Z  
k2 g 1 j eq
amn ¼  J  Mmn  M  Nmn dV;
eq
ð2:17aÞ
pð1 þ dm Þ Sðn; mÞ g
V

Z  
k2 g 1 j
bnm ¼   Jeq  Nnm  Meq  Mnm dV; ð2:17bÞ
pð1 þ dm Þ Sðn; mÞ g
V

where dm is the Krönecker symbol. The equivalent electric and magnetic currents
are determined mostly from computed data. However, in order to implement the
proposed method, details of the feeding antenna structure should be known in order
to determine the equivalent currents (e.g., by a general EM solver), which is not
always the case—one often receives only the far-field measurements from the
antenna manufacturer.
In many cases it is beneficial to use a feed model which is based on spherical
mode representation of a far-field radiation pattern of the feed antenna and then
using an efficient way of connecting fields in local and global coordinate systems.
With this a fast and accurate method for analyzing spherical stratified lens antenna
systems can be obtained. Note that the arbitrary feed antenna for which we want to
find the model can be characterized either by measurements (often performed by the
manufacturer) or by analysis using general EM solver. The proposed feed model
contains three steps.
First step
The far-field of the feed antenna can be simply represented using vectorial spherical
harmonics [7]:
XX
E¼ amn Mmn þ bmn Nmn ; ð2:18Þ
n m

where the functions Mmn and Nmn are defined in Eqs. (2.3a)–(2.3b).
Second step
The spherical harmonics representation of the feed antenna is given in the local
coordinate system, so we need to rewrite it in the global coordinate system. If we
make a translation and/or rotation of the coordinate system, a new basis can be
defined with respect to the new coordinate system, and any of the considered
functions has to be expressed in the new basis. From addition theorems we get
expansion coefficients for Mmn and Nmn in terms of M′mn and N′mn which are
referenced to the second coordinate system
50 Z. Sipus and T. Komljenovic

m 
1 X
X 
0 mn 0
Mmn ¼ Amn
lm M lm þ Blm lm ;
N ð2:19aÞ
m¼0 l¼m

m 
1 X
X 
0 mn 0
Nmn ¼ Amn N
lm lm þ Blm M lm ; ð2:19bÞ
m¼0 l¼m

where Amn mn
lm and Blm are coefficients that connect the global and local coordinate
systems [34, 35]. A way to efficiently calculate them will be discussed at the end of
this section.
Third step
The excitation field is rewritten in terms of equivalent currents (suitable for
implementation into the code for analyzing spherical multilayer structures [8]):

Jeq ¼ ð^rÞ  H
ð2:20Þ
Meq ¼ E  ð^rÞ

The equivalent currents have the following form in the spectral domain (for
example M~ eq ):
" !
XX X m pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 X 
~ eq ¼  mn ^ mn ^
M amn 2pSðn; mÞ Alm / þ Blm h
n m m¼0 l¼m
!# ð2:21Þ
X m pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 X 
mn ^ mn ^
þ bmn 2pSðn; mÞ Blm / þ Alm h
m¼0 l¼m

Field representation in local and global coordinate systems.

In order to simplify mathematical description of the electromagnetic field distri-
bution it is desirable to expand field solutions in different coordinate sets which
describe the same space. The connection between different representations is made
by using addition theorems, i.e., by using formulas which describe representation of
a basis set of one coordinate system in terms of a basis set of one other [34]. There
is a practical problem with calculating coupling coefficients, in particular with the
Gaunt coefficients. This was first achieved by Stein [34] in terms of the Wigner 3jm
symbols, and it turns out to be very complicated and lengthy task [10] (for this
reason no results were shown in [10] that are obtained using transformation from
local to global coordinate systems).
Coupling coefficients can be efficiently calculated for special case when the
center of the local coordinate system is located at the z-axis (which is our case;
multiple sources can be easily treated by rotating the coordinate system [36, 37]). In
that case the coupling coefficients in Eqs. (2.19a), and (2.19b) are calculated from
expressions
2 Multi-shell Radially Symmetrical Lens Antennas 51

m mn 2m þ 1 X p
mm ¼ ð1Þ j
Amn i ½nðn þ 1Þ þ mðm þ 1Þ  pðp þ 1Þ

2mðm þ 1Þ p ð2:22Þ
 aðm; nj  m; mjpÞ zp ðkrfeed Þ

m mn 2m þ 1 X p

mm ¼ ð1Þ j
Bmn i 2jmkrfeed  aðm; nj  m; mjpÞ zp ðkrfeed Þ;
2mðm þ 1Þ p
ð2:23Þ

where a(m,n|l,m|p) are Gaunt coefficients for which efficient recursion formulas
exist for translation in the z-direction (i.e. for l = −m, see [38] for details).

2.3.2 Examples of Spherical Multi-shell Lens Antennas

12-layer Luneburg lens antenna

To demonstrate the results of the presented methodology, let us consider a 12-shell
Luneburg lens antenna developed at EPFL [12]. The lens parameters are given in
Table 2.1. The operating frequency is 10 GHz. The outer radius of the lens antenna
is 15.9 cm, therefore the antenna is approximately 10k0 in diameter. As a feed
antenna let us consider a rectangular horn for which the radiation pattern is
expanded in spherical harmonics. The opening of the rectangular horn is
2.0  3.0 cm and the horn is placed at h = 2.3 cm distance from the lens surface,
see [12] for details.

Table 2.1 Lens geometry and electrical characteristics

Shell EPFL Univ. of Rennes
number Radius Permittivity Shell Radius Permittivity
(cm) number (cm)
1 3.755 1.93 1 8.2 1.92
2 5.52 1.74 2 11.6 1.76
3 7.41 1.71 3 14.2 1.61
4 8.38 1.65 4 16.4 1.46
5 9.415 1.63 5 18.4 1.31
6 9.94 1.56 6 20.0 1.15
7 11.84 1.54
8 12.5 1.50
9 13.4 1.40
10 14.27 1.28
11 15.15 1.20
12 15.9 1.10
52 Z. Sipus and T. Komljenovic

Fig. 2.4 Comparison of the calculated and measured radiation patterns of the 12-layer Luneburg
lens antenna developed at EPFL. a E-plane, b H-plane

The comparison of the calculated radiation pattern of the whole lens antenna
with measurements is given in Fig. 2.4. Two analysis methods are compared—the
considered analysis method (the presented feed model + G1DMULT algorithm)
and results obtained using CST Microwave Studio [39]. There is an excellent
agreement between calculated results obtained by two analysis methods.
Furthermore, the comparison between calculated results and measurements is also
very good. In [12] only the H-plane measurements were provided; this is the reason
2 Multi-shell Radially Symmetrical Lens Antennas 53

why there are no measured results in Fig. 2.4a. For this specific antenna, the
spherical wave expansion approach is more than four orders of magnitude faster
comparing to the time needed for CST calculations. Therefore, the developed
program is suitable for coupling with a global optimization algorithm (e.g. with the
Particle Swarm Optimization (PSO) algorithm) to allow synthesis of spherically
stratified lens antennas [18].
6-layer Luneburg lens antenna
As a second test example, a six-shell Luneburg lens antenna developed at
University of Rennes is considered [14, 15]. The lens parameters are also given in
Table 2.1. The operating frequency is 6 GHz, i.e., the lens diameter is 8 k0. The
feeding antenna is simply a waveguide opening with directivity 7.4 dBi placed at
distance h = 1.0 cm distance from the lens surface. Figure 2.5 shows comparison
of the measured and calculated far-field pattern obtained by our method and by the
general electromagnetic solver (CST Microwave Studio [39]). The agreement
between CST results and spherical harmonic representation is excellent, and with
very good agreement with measurements. Here we have also tested two feed
models (the one based on far-field radiation pattern and the rigorous one based on
projecting the equivalent currents on the spherical wave basis [15]), and the dif-
ference between results obtain by using two feed models is negligible (therefore, the
results obtained by both feed models are represented with the same line in Fig. 2.5
—spectral domain method).
The analyzed examples of step-index realization of Luneburg lens antenna are
two rare examples for which all antenna parameters are given in the open scientific
literature (to the best knowledge of the authors, only two additional examples can
be found in open scientific literature in [40, 41]). In most of the cases, only a short
description of the realized lens is given (e.g., in [42] it is simply stated “8-inch
diameter, ten-step Styrofoam lens at 16.65 GHz”).

2.3.3 Multi-shell Lens Optimization

When designing the lens antenna usually there are a lot of parameters to be
determined. Since the proposed algorithms for analyzing the lens antenna are very
fast it is practical to connect the developed algorithm with some global optimization
routine, for example with the Particle Swarm Optimization (PSO) algorithm. This is
an evolutionary algorithm similar to the genetic algorithm and to the simulated
annealing, but it operates on a model of social interaction between independent
agents and utilizes swarm intelligence to achieve the goal of the optimization
problem. It is rather easy and straightforward to implement into the program [43],
and its performance is at least comparable to the widely used genetic algorithms
[44]. All the optimizations done in this chapter were made using the PSO algorithm.
54 Z. Sipus and T. Komljenovic

Fig. 2.5 Comparison of the calculated and measured radiation patterns of the 6-layer Luneburg
lens antenna developed at University of Rennes. a E-plane, b H-plane

Characterization of lens and feed antenna

The original Luneburg lens antenna with the permittivity variation er ðrÞ ¼
2  ðr=rlens Þ2 has the feed antenna optimally positioned at the lens surface [1]. For
practical reasons, it is desirable to move the feed position from the lens surface, i.e.,
to introduce a gap between the feed and the lens. In order to optimize the lens
antenna for new feed position, a modified lens design should be applied [45, 46]
2 Multi-shell Radially Symmetrical Lens Antennas 55

0 1
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi  Zrfeed
 2 B 2 arcsinðt  rlens rfeed Þ C
er ðrÞ ¼ 1 þ 1  er r=rfeed  exp@ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dtA;
p
1 t2  er ðr=rlens Þ2
ð2:24Þ

where r and rfeed represents the radial coordinate inside the lens and the position of
the feed, respectively. Figure 2.6 shows the optimal permittivities of Luneburg lens
as a function of normalized radial coordinate in case if the feed is placed at the lens
surface (rfeed = 1.0  rlens) and if it is placed at 1.1  rlens. Manufactured antenna is
always realized from several layers of material with constant permittivity. In other
words, the practical realization is just a stepwise approximation of the ideal lens.
The theoretical curves are compared to permittivities of two manufactured lenses,
with the permittivity value of each layer shown for a radius corresponding to the
middle of the considered layer (ðri1 þ ri Þ=2). It can be seen that in both cases there
is some uncertainty about the obtained electromagnetic parameters of the realized
lens antennas. We address this issue at the end of this section (see Fig. 2.8).
In order to study the effect of feed distance from lens surface and the directivity
of feed antenna on total directivity of lens antenna and sidelobe levels, we model an
“ideal” type of feed antenna, where the far-field is approximated with a cosine type
of pattern

ejkr h^ ^
i
Eðr; h; /Þ ¼ E0 hðcos ðh=2ÞÞq1 cos /  /ðcos ðh=2ÞÞq2 sin / : ð2:25Þ
r

With varying factors q1 and q2 we can model different types of feed antennas.
We have selected ðcos ðh=2ÞÞq type of function (instead of the ðcos hÞq dependence
[47]), since it better approximates the low-directivity antennas which are usually

Fig. 2.6 Comparison of

permittivity of practical lens
realizations and ideal designs
for rfeed = rlens (Luneburg
lens) and for rfeed = 1.1  rlens;
diamonds (◊)—lens antenna
from EPFL, stars (*)—lens
antenna from University of
Rennes
56 Z. Sipus and T. Komljenovic

used as feeding elements of lens antennas (some authors have applied ðcosðh=2ÞÞq
dependency for reflector antenna analysis as well [33]). Similar results can also be
obtained using a complex Huygens source [48, 49]. First step in antenna system
characterization is to determine the optimum directivity of feeding antenna as a
function of feed position rfeed. We have calculated directivity of modified Luneburg
lens antenna (permittivity variation is obtained using Eq. (2.24); the continuous
permittivity variation is approximated with 20-layer step-wise distribution) for
different q values, i.e., for different directivities of feeding antenna (Fig. 2.7a). For
simplicity, we have taken q1 = q2 = q. The diameter of the considered lens antenna
is 10 k0. The optimum directivity of the feed antenna is between 8 and 10 dBi
(depending on the designed antenna position rfeed), and the directivity of the whole
lens antenna does not vary much by moving the feed antenna away from the lens
surface (for larger rfeed/rlens a more directive feed should be used). Therefore, the
low-directive horn antennas or waveguide openings are very good elements for
feeding the Luneburg lens antennas. In Fig. 2.7b the dependence of sidelobe level
and illumination of the edge of the lens antenna is shown for the optimized case.
Both, the sidelobe level and edge illumination, do not vary much with the change of
feed position; they are around 22 dB and −13 dB, respectively. It is interesting to

Fig. 2.7 a Optimized feed

directivity and complete lens
antenna directivity as a
function of the feed antenna
position. b Sidelobe level and
edge illumination of the
optimized case. The diameter
of the considered Luneburg
lens antenna is 10 k
(reproduced with permission
from Ref. [19])
2 Multi-shell Radially Symmetrical Lens Antennas 57

note that the edge illumination is approximately 3 dB lower comparing to the

rule-of-thumb value for reflector antennas (–10 dB edge illumination [47]).
To show the effect of feed selection and placement on a practical realization of
Luneburg lens we have calculated the total lens antenna directivities as a function of
the feed directivity and position for the two considered realizations (see Table 2.1).
Due to uncertainties about the obtained electromagnetic parameters of the realized
lens antennas (see Fig. 2.6), it is advisable to make such characterization, i.e., to
optimize the feed type as well as its position. Results are shown in Fig. 2.8. It can
be seen that there is a significant freedom in selecting the feed antenna type and
position for optimal performance. This is in particular important since the phase
center of the feed antenna is usually located inside the antenna. In the EPFL case
the used horn antenna has 9 dBi directivity and the normalized feed position is
rfeed = 1.145  rlens. The obtained sidelobe level and lens edge illumination is
19.5 dB and around −10 dB, respectively, which is comparable with the optimized
feed properties. Similar result is also obtained for the antenna developed at
University of Rennes. The used waveguide opening has directivity of 7.4 dBi and
the normalized feed position is rfeed = 1.05  rlens. The directivity determined by the
spherical wave expansion technique equals 27.6 dBi whereas 27.5 dBi are predicted
by CST. The obtained sidelobe level and lens edge illumination is 16.5 dB and
between −6.0 dB (E-plane) and −11.3 dB (H-plane), respectively, which is also
comparable to the optimized feed properties.

2.3.4 Fabrication and Experimental Results

The challenge of producing gradient index (GRIN) lenses is already mentioned in

the articles of James Clerk Maxwell. Spherical Luneburg lenses have been fabri-
cated using various techniques. Crushed Styrofoam was used to build 186 identical
wedges—the lens was sliced through the North–South axis and the gradient vari-
ation of index of refraction is obtained by compressing Styrofoam pieces [50].
Polystyrene powder, solidified in heated hemispherical molds, was used to build
10-layer step-index Luneburg lens [40]. Ten-layer step-index Luneburg lenses were
also built from Styrofoam and Foamed glass [41, 42]. The needed variation of the
refractive index can in addition be obtained by drilling holes in the dielectric
structures, which can be easily fabricated by using the traditional printed circuit
board technique [51, 52]. Recently, a rapid prototyping machine was used to build a
Luneburg lens in X-band by changing the size of plastic blocks centered on the
junctions of the plastic rod frame [53].
Lens antennas are attractive solutions for many applications due to their focusing
properties, capability of beam scanning, ability to form multiple beams and
broadband behavior. For example, spherical lens antennas can be used for auto-
motive radars [28], satellite communications [30, 31], for systems that require
communication or/and multiple object tracking through a shared aperture [54],
radar cross-section (RCS) enhancers [29], or for astronomy applications [32]. One
58 Z. Sipus and T. Komljenovic

Fig. 2.8 Lens antenna

directivity as a function of
feed position and directivity.
The optimum selection that
results with the lens antenna
with maximum directivity is
marked with X. The
parameters of the lens
antennas are given in
Table 2.1; a lens antenna
from EPFL, b lens antenna
from University of Rennes
(reproduced with permission
from Ref. [19])

of the most attractive applications is satellite communication systems for receiving

signals from multiple satellites [30]. With this application in mind, a very useful
variation of the Luneburg lens antenna is the hemispherical lens in conjunction with
the ground plane. Furthermore, this is also an attractive solution for satellite
communications with moving vehicles (e.g., trains, see Fig. 2.3c). If applied on the
train, the antenna is mounted on the roof of the train and the mechanical system for
positioning the feed antenna enables Internet connection via satellite [31]. Also, an
important application of the Luneburg lens is a radar cross-section (RCS) enhancer,
realized by placing a reflecting cap on the lens surface [29]. An incoming plane
wave is focused at a point on the cap and reflected; it is transmitted in the opposite
direction after propagating through the lens the second time. By this, a broadband
calibrated scatterer with wide angular coverage is obtained, which is needed for
calibrating scattered field measurement systems.
Since the construction of multilayer lenses is not easy, for many applications a
constant-n lens or a two–layer lens is proposed that is much easier to construct
2 Multi-shell Radially Symmetrical Lens Antennas 59

comparing to Luneburg lenses. For fabrication of optimum constant-index lenses, a

material with permittivity of 3.5 is desired, which can be obtained using slip-cast
fused silica with small additions of rutile (er = 90) [55]. An investigation on the
construction of two layer-lenses with off-the-shell materials was performed in [31],
where the considered materials were Rexolite (er = 2,53), polyethylene (er = 2.28)
and Fused silica (er = 3.8).
Finally, there is a class of two-dimensional Luneburg lenses where the focusing
of electromagnetic waves is obtained inside a parallel-plate waveguide. The lens
effect can be obtained by changing the thickness of the dielectric plate [56, 57], by
drilling holes in a dielectric slab to control the effective permittivity [58], or by
printing metallic patterns on dielectric substrate (i.e., variation of index of refraction
is obtained using the metasurface concept [59–61]). Recently, the new imple-
mentation consisting of two mirrored metallic holey metasurfaces, having
ultra-wideband lens property, was introduced [62].

2.4 Cylindrical Multi-shell Lenses

Circular-cylindrical dielectric lenses are attractive antennas because they allow

launching multiple fan-beams, each of them originating from one primary feed.
Like in the spherical case, the beams can be scanned by moving mechanically the
feed around the lens surface or by switching between the feeds. They are also
inherently wideband antennas. Such radiation characteristics are of particular
interest for many applications at millimeter waves, like Doppler-weather radar,
aircraft landing system or imaging systems. These lens antennas have been mostly
neglected in the open literature, in contrast to the spherical ones.
In the past, research has been limited to homogeneous cylindrical lenses [63] and
cylindrical Luneberg lenses [56, 57] placed between parallel-plate waveguides, i.e.,
the effect of the finite lens height is not present in this case. There has also been
some research regarding antennas with different directivities in two orthogonal
planes based on cylindrical Luneberg lens [64], and similar concept has been
applied for automotive radars [65]. Cylindrical lenses with special requirements on
radiation pattern have also been analyzed as an EM missile launcher, i.e., case when
the decay of the energy per unit length is slower than 1/r [66]. Finally, the influence
of the whispering-gallery modes on obtained directivity of cylindrical Luneberg
lens antennas has been studied in [67]. In this section design of multilayered
cylindrical dielectric lens antennas of finite height is discussed (a detailed study
made by University of Rennes and University of Zagreb is given in [68]).
60 Z. Sipus and T. Komljenovic

2.4.1 Lens Geometry

The generic lens antenna configuration is represented in Fig. 2.9a (the lens antenna
geometry will be described using the classical cylindrical coordinate system (r, /, z)).
It consists of a multilayer cylindrical dielectric lens illuminated by a feed antenna.
The total lens height and radius are labeled h are rlens, respectively. The multilayer
lens structure is schematized in Fig. 2.9b. The relative permittivity and thickness of
layer #i are labeled er,i and ai, respectively (1  i  N). Without loss of generality
we have selected a pyramidal horn as a feed antenna. The horn (of length L) is fed by a
standard metallic rectangular waveguide linearly polarized along z-axis and d de-
notes the distance separating the horn aperture (Wy  Wz) from the lens boundary.
The multilayer cylindrical lens can be approximated with an infinite circular
cylinder. In that case we can apply the method of analysis described in the previous
section, this time for cylindrical structures. The major advantage of this method of
analysis is that the determination of the lens radiation performance is extremely fast.
As a consequence, the iterative full-wave optimization or synthesis of cylindrical
lenses can be made in reasonable amount of time. For lenses of moderate size
(around 10  k0 in diameter), the G1DMULT-based program needs less than one
second to compute the three-dimensional far-field radiation patterns, whereas
general-purpose commercial software like CST Microwave Studio® [39] requires
approximately one hour, depending on the lens height and permittivity of the
dielectric layers.

Fig. 2.9 Geometry of the circular-cylindrical dielectric lens antenna. a 3-D view. h, rlens, and
d denote the lens height, the total lens radius, and the distance between the horn aperture
(Wy  Wz) and the lens surface, respectively. L is the length of the horn. b Cross-section view of
the lens. Definition of the thickness ai and permittivity er,i of each layer
2 Multi-shell Radially Symmetrical Lens Antennas 61

2.4.2 Range of Validity of Modal Analysis Approach

By essence the modal analysis approach assumes that the cylindrical structures are
infinitely long, whereas, in practice, the objective is to minimize the height of the
antenna prototype while keeping the lens radiation characteristics close to those of
an infinite lens. To determine the range of validity of modal analysis approach, we
have carried out an exhaustive parametric study consisting in comparing the
far-field radiation patterns of a set of lenses analyzed using both the
G1DMULT-based program and CST Microwave Studio®.
An empirical formula has been derived to estimate the minimum cylindrical lens
height hmin that would provide, in far-field zone, nearly the same results as for
lenses of infinite height

hmin ¼ 2ð2rlens þ dÞ þ Wz ð2:26Þ

Here rlens is the total lens radius, d is the distance separating the horn aperture
from the lens, and Wz is the height of the feed horn. The formula has been verified
for low permittivity materials (er,i < 3) as, for our applications, there was no need to
use materials with even higher permittivities. The validity of this relation is clearly
demonstrated in Fig. 2.10 where the fan-beam pattern of an infinite lens at 30 GHz
is compared to the radiation patterns of three finite lenses (h = hmin, h = 0.7 
hmin, h = 0.5  hmin). In cases where the lens height is too small, both bases of the
cylinder start to contribute to the antenna pattern. This results in a slight variation of
the antenna gain, and in strongly pronounced ripples in E-plane.

2.4.3 Numerical and Experimental Results

The objective of the optimization problems studied here is to maximize the antenna
gain while keeping low sidelobe levels. The corresponding fitness function is
defined as

Fitness ¼ a  Gain  b  SLðHÞ  c  SLðEÞ: ð2:27Þ

Here Gain is the antenna gain at broadside (i.e., along x-axis), and SL(H) and SL
(E) are the highest sidelobe levels in the H- and E-planes, respectively. All quan-
tities are expressed in dB. a, b, and c are weighting coefficients that can be tuned to
maximize the gain, or produce fan-beam with low sidelobes. Their typical values
are 1, 0.3, and 0.2, respectively. In all cases, the feed and lens parameters are
optimized simultaneously since this approach has been shown to be powerful [69].
Influence of the number of layers on the radiation characteristics of small lenses
To the authors’ knowledge, in contrast to multilayer spherical lenses (e.g., [14]),
there are very few papers dealing with the influence of number of layers on the
62 Z. Sipus and T. Komljenovic

Fig. 2.10 Range of validity

of modal analysis approach.
Comparison of the radiation
patterns computed at 30 GHz
with G1DMULT and CST
Microwave Studio®.
a H-plane. b E-plane. The
lens is made in Teflon
(er,1 = 2.1,
rlens = a1 = 2.86 cm) and is
illuminated by a pyramidal
horn fed by a standard WR-28
waveguide (L = 2 cm,
Wy = 2.4 cm, Wz = 1 cm,
d = 0.65 cm)

radiation characteristics of a cylindrical lens (see, e.g., [67]). The study described
here is restricted to lenses with a moderate gain (*18 dBi). Their total diameter is
equal to 5 cm, i.e., 5  k0, at 30 GHz. The lens and feed horn parameters have
been optimized simultaneously, and the number of layers was varied from 1 to 3.
The range of variation of the optimization parameters are the following:
0.05 cm  d  2 cm, 0.4 cm  Wz  2 cm, 0.8 cm  Wy  3 cm,
0.5 cm  L  3 cm, 1.4  er,i  5, 8 i = 1, …, 3. Both the gain and the
sidelobe level were optimized (a = 1, b = 0.3, c = 0.2).
Up to seven optimization runs per lens configuration were launched, and the best
result was selected based on the value of the fitness function and the visual
inspection of the far-field radiation patterns. The geometry of the best designs and
their corresponding patterns in H- and E-planes are given in Table 2.2 and
2 Multi-shell Radially Symmetrical Lens Antennas 63

Table 2.2 Influence of the number of layers

1 layer 2 layers 3 layers
Layer no. 1: thickness (cm) 2.50 0.10 0.73
Layer no. 1: permittivity 1.43 1.56 1.93
Layer no. 2: thickness (cm) – 2.40 0.72
Layer no. 2: permittivity – 1.52 1.77
Layer no. 3: thickness (cm) – – 1.05
Layer no. 3: permittivity – – 1.63
d (cm) 0.82 0.77 0.63
Wz (cm) 1.05 1.03 1.07
Wy (cm) 2.82 2.66 1.57
L (cm) 3.00 2.84 3.00
Gain (dBi) 17.76 17.63 17.62
Sidelobes (dB) −22.30 −22.95 −23.71

Fig. 2.11, respectively. The radiation characteristics of the best three lens config-
urations are nearly identical. This demonstrates that there is no need to use more
than one layer to optimize the fan-beam characteristics of moderate-sized cylin-
drical lenses.
Experimental results
To validate experimentally the previous results, a single-layer homogeneous
cylindrical lens has been optimized in Ka-band. In this case, the results given in
Table 2.2 suggest using a bulk material whose dielectric constant equals 1.43.
Although customized materials are available commercially or could be synthesized
using effective medium theory [51, 58] their fabrication cost remains very expensive.
To keep the antenna manufacturing as simple as possible, a cylindrical lens made
from Teflon (er,1 = 2.1) has been synthesized. The final dimensions of the antenna
prototype are the following: rlens = a1 = 2.82 cm, L = 2 cm, Wz = 1 cm,
Wy = 2.4 cm, and d = 0.65 cm. The minimum height hmin of the lens, guarantying
that the infinite and finite lenses have nearly the same radiation characteristics, is
equal to 13.6 cm (according Eq. (2.26)).
To keep some margin with respect to this threshold value, a 16 cm long lens was
manufactured. As the optimized dimensions of the feed horn are not standard, a
specific horn was fabricated in aluminum using electrical discharge machining
techniques. Both the lens antenna and the feed horn were fabricated and measured
at IETR, University of Rennes [68].
The radiation patterns measured at 30 GHz in H- and E-planes are represented in
Figs. 2.12a, b, respectively. They are in excellent agreement with those predicted by
G1DMULT-based program and CST Microwave Studio®, confirming thereby the
relevance of the empirical relation (2.26). Additional measurements have been
performed from 26 to 40 GHz (Fig. 2.13). They show that the lens patterns are very
stable over the whole frequency band proving that the design is wideband.
64 Z. Sipus and T. Komljenovic

Fig. 2.11 Radiation patterns

of the best designs at 30 GHz.
The patterns have been
computed using
G1DMULT-based program.
a H-plane. b E-plane. The
corresponding lens
geometries are detailed in
Table 2.2

2.4.4 Size Reduction

Cylindrical lenses whose height is defined according to Eq. (2.26) produce nearly
the same fan-beam patterns as lenses of infinite height. Nevertheless, applying this
empirical criterion leads to large and bulky lens configurations. To overcome this
limitation, the lens geometry has been optimized to minimize the influence of both
lens bases, in particular to reduce the substantial ripples that are generated in
E-plane for lenses of small height (Fig. 2.10b). These ripples originate from lon-
gitudinal resonant modes that contribute significantly to radiation in E-plane. This
phenomenon is clearly highlighted on the amplitude near-field map represented in
Fig. 2.14.
2 Multi-shell Radially Symmetrical Lens Antennas 65

Fig. 2.12 Synthesized lens

made from Teflon: radiation
patterns measured and
computed at 30 GHz.
a H-plane. b E-plane

In order to reduce as much as possible the effects of lens truncation, two conical
shapes are added to the cylinder bases (Fig. 2.15). As the resulting lens shape is
rotationally symmetric, the ability to launch multiple beams and scan them around
the lens axis is maintained. The height and permittivity of both cones are labeled hc
and er,c, respectively. Their permittivity must be higher than the one of the cylin-
drical part of the lens to prevent possible total reflection that can happen when
electromagnetic waves travels from a denser to a less dense medium.
Two cones made in Rexolite (er,c = 2.53) have been designed to improve the
radiation characteristics of the small lens represented in Fig. 2.15a. The height of
the circular-cylindrical part is 4.8 cm. The lens is illuminated by the same horn as in
Sect. 2.4.3. The cone height has been optimized using CST Microwave Studio® to
minimize the ripple levels in E-plane (hc = 2 cm). As a result the total antenna
height equals 8.8 cm.
66 Z. Sipus and T. Komljenovic

Fig. 2.13 Synthesized lens

made in Teflon: radiation
patterns measured in
Ka-band. a H-plane.
b E-plane. These patterns are
quite frequency independent

The radiation patterns computed at 30 GHz are represented in Fig. 2.16, and
they are compared to those of two pure cylindrical lenses: the first one has the same
height (8.8 cm) as the lens with conical caps, and the height of the second one
(16 cm) fulfills relation (2.26). The three lenses have nearly the same patterns in
H-plane (Fig. 2.16a). However, by introducing two conical caps at the bases of
small circular-cylindrical lenses, the ripple level in E-plane can greatly be reduced
(Fig. 2.16b). This technique also provides a faster roll-off, compared to pure
cylindrical lenses. The influence of cones on the amplitude near-field distribution is
highlighted in Fig. 2.14b. It can be seen that the waves can propagate into the
denser cone (without total reflection) and finally in the surrounding air as they
approach the Rexolite-Air boundary at smaller incident angles.
The lens with conical caps has been fabricated and measured at University of
Rennes [68]. Both cones have been assembled on the cylinder bases using a thin
film of dielectric glue. The distance d separating the feed horn and the lens surface
2 Multi-shell Radially Symmetrical Lens Antennas 67

Fig. 2.14 Amplitude near-field maps (arbitrary units). a Original design. The lens parameters are
the same as for the experimental model (Sect. 2.4.3), except the lens height that is roughly three
times smaller than hmin (h = 4.8 cm). b Proposed design. The lens is the same as in (a) with
addition of two cones. The cone parameters are the following: hc = 2 cm, er,c = 2.53

Fig. 2.15 Reduced-height circular-cylindrical dielectric lens with conical caps. The lens is
illuminated by a pyramidal feed horn linearly polarized along z-axis. a 3-D view. b Cross-section
view

is equal to 0.4 cm. The radiation patterns measured in both principal planes at
30 GHz are represented in Fig. 2.17. The agreement between numerical and
experimental results is excellent, especially in E-plane where the ripple level in the
main beam is smaller than 1.5 dB. The cross-polarization level is smaller than
–21 dB and the measured antenna gain equals 17.66 dBi.
The height reduced design is also wideband as can be seen in Fig. 2.18.
Measurements performed from 26 to 40 GHz show that the radiation patterns are
really frequency independent over Ka-band with an expected increase of the gain as
operating frequency increases. It should be noted that height reduced design shows
even higher insensitivity of the radiation pattern to change in frequency compared
to the original circular-cylindrical design.
68 Z. Sipus and T. Komljenovic

Fig. 2.16 Radiation

patterns (computed at
30 GHz) of the reduced-size
Teflon lens (h = 8.8 cm,
er,1 = 2.1) with conical caps
(hc = 2 cm, er,c = 2.53).
Comparison with two pure
cylindrical Teflon lenses of
variable height (h = 8.8 cm or
16 cm). a H-plane. b E-plane.
All lenses have the same
radius (rlens = a1 = 2.82 cm)
and are illuminated by the
same horn as in Sect. 2.4.3

2.5 Concluding Remarks

A fast and reliable synthesis method is presented for designing multilayered

spherical and circular-cylindrical dielectric lens antennas. In order to make the
analysis program fast, thus enabling efficient merging with the global optimization
routine, the multi-shell lens structure is analytically taken into account using
Green’s functions approach. The proposed analysis approach can successfully
analyze lens structures with arbitrary number of layers and arbitrary type of exci-
tation antenna. The latter is obtained by representing the measured or calculated
far-field radiation pattern in terms of spherical harmonics in local coordinate
system.
2 Multi-shell Radially Symmetrical Lens Antennas 69

Fig. 2.17 Reduced-height

Teflon lens with conical caps
from Rexolite: radiation
patterns measured and
computed at 30 GHz.
a H-plane. b E-plane

In the cylindrical case, the analysis approach assumes that the cylindrical
structure is infinitely long, so we assess the range of validity of the analysis method.
An empirical formula to estimate the minimum cylindrical lens height that provides
similar far-field to the lens of infinite height is derived. For the cases where the lens
size should further be reduced, a new height-reduction technique is proposed. It
consists in adding conical caps to both bases of the cylinder. This technique allows
a significant reduction of the ripple level in the elevation pattern of small cylindrical
lenses, leading to high-quality fan-beams. Theoretical predictions are successfully
verified by fabricating and characterizing various lens prototypes.
70 Z. Sipus and T. Komljenovic

Fig. 2.18 Reduced-height

Teflon lens with conical caps
made in Rexolite: radiation
patterns measured in
Ka-band. a H-plane.
b E-plane. The radiation
pattern is quite similar at all
frequencies, with expected
gain increase as frequency
increases (reproduced with
permission from Ref. [68])

References

1. R.K. Luneburg, Mathematical Theory of Optics (Brown University Press, Providence, RI,
1944)
2. E. Braun, Radiation characteristics of the spherical Luneberg lens. IRE Trans. Antennas
Propag. 4(2), 132–138 (1956)
3. S.P. Morgan, Generalizations of spherically symmetric lenses. IRE Trans. Antennas Propag.
7(4), 342–345 (1959)
4. T.L. Ap Rhys, The design of radially symmetric lenses. IEEE Trans. Antennas Propag. 18(5),
497–506 (1970)
5. G. Godi, R. Sauleau, D. Thouroude, Performance of reduced size substrate lens antennas for
millimeter-wave communications. IEEE Trans. Antennas Propag. 53(4), 1278–1286 (2005)
6. A.D. Greenwood, J.-M. Jin, Finite-element analysis of complex axisymmetric radiating
structures. IEEE Trans. Antennas Propag. 47(8), 1260–1266 (1999)
7. J.A. Stratton, Electromagnetic Theory (McGraw Hill, New York, 1941)
8. R.F. Harrington, Time Harmonic Electromagnetic Fields (McGraw Hill, New York, 1961)
2 Multi-shell Radially Symmetrical Lens Antennas 71

9. J.J. Mikulski, E.L. Murphy, The computation of electromagnetic scattering from concentric
spherical structures. IEEE Trans. Antennas Propag. 11, 169–177 (1963)
10. H. Mieras, Radiation pattern computation of a spherical lens using Mie series. IEEE Trans.
Antennas Propag. 30(6), 1221–1224 (1982)
11. S.S. Vinogradov, E.D. Vinogradova, P.D. Smith, Accurate modelling of a scanning Luneberg
lens antenna Mie series approach, in Proceedings of the International Conference on
Electromagnetics in Advanced Applications, Turin, Italy, Sep. 1999 (1999), pp. 277–280
12. J.R. Sanford, Spherically Stratified Microwave Lenses, Ph.D. Thesis No. 1065, École
polytechnique fédérale de Lausanne (EPFL), 1992
13. J.R. Sanford, Scattering by spherically stratified microwave lens antennas. IEEE Trans.
Antennas Propag. 42, 690–698 (1994)
14. B. Fuchs, L. Le Coq, O. Lafond, S. Rondineau, M. Himdi, Design optimization of multishell
Luneburg lenses. IEEE Trans. Antennas Propag. 55, 283–289 (2007)
15. B. Fuchs, S. Palud, L. Le Coq, O. Lafond, M. Himdi, S. Rondineau, Scattering of spherically
and hemispherically stratified lenses fed by any real source. IEEE Trans. Antennas Propag.
56, 450–460 (2008)
16. W.C. Chew, Waves and Fields in Inhomogeneous Media (Van Nostrand Reinhold, New
York, 1990)
17. H. Mosallaei, Y. Rahmat-Samii, Non-uniform Luneburg and 2-shell lens antennas: radiation
characteristics and design optimization. IEEE Trans. Antennas Propag. 49, 60–69 (2001)
18. T. Komljenovic, Z. Sipus, J.-P. Daniel, Scanning vehicular lens antennas for satellite
communications, in Proceedings of EuCAP’2010, Barcelona, Spain (2010), pp. 1–4
19. Z. Sipus, D. Bojanjac, T. Komljenovic, Electromagnetic modeling of spherically stratified
lenses illuminated by arbitrary sources. IEEE Trans. Antennas Propag. 63(4), 1837–1843
(2015)
20. S. Rondineau, Modélisation de lentilles sphériques à gradient d’indice et sources conformes
associées, Ph.D. dissertation, University of Rennes 1, Rennes, France, Dec. 2002
21. S. Rondineau, A.I. Nosich, J.-P. Daniel, M. Himdi, S.S. Vinogradov, MAR analysis of a
spherical-circular printed antenna with finite ground excited by an axially symmetric probe.
IEEE Trans. Antennas Propag. 52(5), 1270–1280 (2004)
22. Z. Sipus, N. Burum, S. Skokic, P.-S. Kildal, Analysis of spherical arrays of microstrip
antennas using moment method in spectral domain. IEE Proc. Microw. Antennas Propag. 153,
533–543 (2006)
23. Z. Sipus, S. Skokic, M. Bosiljevac, N. Burum, Study of mutual coupling between circular
stacked-patch antennas on a sphere. IEEE Trans. Antennas Propag. 56, 1834–1844 (2008)
24. Z. Sipus, P.-S. Kildal, R. Leijon, M. Johansson, An algorithm for calculating Green’s
functions for planar, circular cylindrical and spherical multilayer substrates. Appl. Comput.
Electromagn. Soc. J. 13, 243–254 (1998)
25. L.B. Felsen, N. Marcuvitz, Radiation and Scattering of Waves (Prentice-Hall Inc., New
Jersey, 1973)
26. W.Y. Tam, K.M. Luk, Resonances in spherical-circular microstrip structures of
cylindrical-rectangular and wraparound microstrip antennas. IEEE Trans. Microw. Theory
Tech. 39, 700–704 (1991)
27. B. Fuchs, O. Lafond, S. Palud, L. Le Coq, M. Himdi, M.C. Buck, S. Rondineau, Comparative
design and analysis of Luneburg and Half Maxwell fish-eye lens antennas. IEEE Trans.
Antennas Propag. 56, 3058–3062 (2008)
28. B. Schoenlinner, X. Wu, J.P. Ebling, G.V. Eleftheriades, G.M. Rebeiz, Wide-scan
spherical-lens antennas for automotive radars. IEEE Trans. Microw. Theory Tech. 50(9),
2166–2175 (2002)
29. J.R. Sanford, Analysis of spherical radar cross-section enhancers. IEEE Trans. Microw.
Theory Tech. 43(6), 1400–1403 (1995)
30. C.S. Lee, M.E. Rayner, Luneberg lens antenna with multiple gimbaled RF feeds, U.S.A.
patent No. US 6266029 B1
72 Z. Sipus and T. Komljenovic

31. J. Thornton, Wide-scanning multi-layer hemisphere lens antenna for Ka band. IEE Proc.
Microw. Antennas Propag. 153(6), 573–578 (2006)
32. A.J. Parfitt, J.S. Kot, G.L. James, The Luneburg lens as a radio telescope element, in
Proceedings of the IEEE Antennas and Propagation Society International Symposium, Salt
Lake City (2000), pp. 170–173
33. P.-S. Kildal, Foundations of Antennas—A Unified Approach (Studentlitteratur AB, 2000)
34. S. Stein, Addition theorems for spherical wave functions. Q. Appl. Math. 19, 15–24 (1961)
35. O.R. Cruzan, Translational addition theorems for spherical vector wave functions. Quart.
Appl. Math. 20, 33–40 (1962)
36. D.L. Sengupta, T.M. Smith, R.W. Larson, Radiation characteristics of spherical array of
circularly polarized elements. IEEE Trans. Antennas Propag. 16, 2–7 (1968)
37. J.R. Sanford, Z. Sipus, Sidelobe reduction with array fed spherical lenses, in Proceedings of
the IEEE Antennas and Propagation Society International Symposium, Newport Beach
(1995), pp. 670–673
38. J.H. Bruning, Y.T. Lo, Multiple scattering of EM waves by spheres. Part I—Multiple
expansion and ray optical solutions. IEEE Trans. Antennas Propag. 19, 378–390 (1971)
39. CST Microwave Studio™, Computer Simulation Technology AG, www.cst.com, Darmstadt,
Germany (2013)
40. G.D.M. Peeler, H.P. Coleman, Microwave stepped-index Luneburg lenses. IRE Trans.
Antennas Propag. 6, 202–207 (1958)
41. L.C. Gunderson, J.F. Kauffman, A high temperature Luneburg lens. IEEE Proc. 56, 883–884
(1968)
42. R.E. Webster, Radiation patterns of a spherical luneburg lens with simple feeds. IRE Trans
Antennas Propag. 6, 301–302 (1968)
43. J. Robinson, Y. Rahmat-Samii, Particle swarm optimization in electromagnetics. IEEE Trans.
Antennas Propag. 52(2), 397–407 (2004)
44. D.W. Boeringer, D.H. Werner, Particle swarm optimization versus genetic algorithms for
phased array synthesis. IEEE Trans. Antennas Propag. 52(3), 771–778 (2004)
45. A.S. Gutman, Modified Luneberg lens. J. Appl. Phys. 25, 855–859 (1954)
46. D.K. Cheng, Modified Luneberg lens for defocused source. IRE Trans. Antennas Propag. 8,
110–111 (1960)
47. Y. Rahmat-Sammii, Reflector antennas, in Antenna Engineering Handbook, ed. by J.L.
Volakis, Chap. 15 (McGraw-Hill, 2007)
48. V.S. Bulygin, T.M. Benson, Y.V. Gandel, A.I. Nosich, Full-wave analysis and optimization
of a TARA-like shield-assisted paraboloidal reflector antenna using a Nystrom-type method.
IEEE Trans. Antennas Propag. 61(10), 4981–4989 (2013)
49. D. Bojanjac, A.I. Nosich, Z. Sipus, Design of spherical lens antennas using complex Huygens
element, in Proceedings of EuCAP’2014, The Hague, The Netherlands (2014), pp. 1–3
50. G.P. Robinson, Three-dimensional microwave lens, in Tele-Tech & Electronic Industry, Nov.
1954 (1954), p. 73
51. S. Rondineau, M. Himdi, J. Sorieux, A sliced spherical Lüneburg lens. IEEE Antennas Wirel.
Propag. Lett. 2, 163–166 (2003)
52. H.F. Ma, B.G. Cai, T.X. Zhang, Y. Yang, W.X. Jiang, T.J. Cui, Three-dimensional
gradient-index materials and their applications in microwave lens antennas. IEEE Trans.
Antennas Propag. 61, 2561–2569 (2013)
53. M. Liang, W.-R. Ng, K. Chang, K. Gbele, M.E. Gehm, H. Xin, A 3-D Luneburg lens antenna
fabricated by polymer jetting rapid prototyping. IEEE Trans. Antennas Propag. 62, 1799–
1807 (2014)
54. M.A. Mitchell, J.R. Sandford, Luneburg lens revival, in Electronics&Wireless World, May
1989 (1989), pp. 456–458
55. W.R. Free, F.L. Cain, C.E. Ryan Jr., C.P. Burns, E.M. Turner, High-power constant-index
lens antennas. IEEE Trans Antennas Propag. 22, 582–584 (1974)
56. G.D.M. Peeler, D.H. Archer, A two-dimensional microwave Luneburg lens. IRE Trans.
Antennas Propag. 1, 12–23 (1953)
2 Multi-shell Radially Symmetrical Lens Antennas 73

57. X. Wu, J. Laurin, Fan-beam millimeter-wave antenna design based on the cylindrical luneberg
lens. IEEE Trans. Antennas Propag. 55, 2147–2156 (2007)
58. K. Sato, H. Ujiie, A plate Luneberg lens with the permittivity distribution controlled by hole
density. Electr. Commun. Jpn. 85(9), 1–12 (2002)
59. C. Pfeiffer, A. Grbic, A printed, broadband Luneburg lens antenna. IEEE Trans. Antennas
Propag. 58, 3055–3059 (2010)
60. S. Maci, G. Minatti, M. Casaletti, M. Bosiljevac, Metasurfing: addressing waves on
impenetrable metasurfaces. IEEE Antennas Wirel. Propag. Lett. 10, 1499–1502 (2011)
61. M. Bosiljevac, M. Casaletti, F. Caminita, Z. Sipus, S. Maci, Non-uniform metasurface
luneburg lens antenna design. IEEE Trans. Antennas Propag. 60, 4065–4073 (2012)
62. O. Quevedo-Teruel, M. Ebrahimpouri, M.N.M. Kehn, Ultrawideband metasurface lenses
based on off-shifted opposite layers. IEEE Antennas Wirel. Propag. Lett. 15, 484–487 (2016)
63. L.C. Gunderson, An electromagnetic analysis of a cylindrical homogeneous lens. IEEE Trans.
Antennas Propagat. 20, 476–479 (1972)
64. G. Dubost, Flat linear radiating array applied on a cylindrical lens, in MELECON’85, Madrid,
Spain, vol. 3 (1985), pp. 215–218
65. P. Wenig, R. Weigel, Analysis of a microstrip patch array fed cylindric lens antenna for
77 GHz automotive radar, in IEEE AP-S International Symposium Digest, San Diego (CA),
5–11 July 2008
66. T.T. Wu, R.W.P. King, H.-M. Shen, Circular cylindrical lens as a line-source
electromagnetic-missile launcher. IEEE Trans. Antennas Propag. 37(1), 39–44 (1989)
67. A.V. Boriskin, A.I. Nosich, Whispering-gallery and Luneburg-lens effects in a beam-fed
circularly layered dielectric cylinder. IEEE Trans. Antennas Propag. 50(9), 1245–1249 (2002)
68. T. Komljenovic, R. Sauleau, Z. Sipus, L. Le Coq, Layered circular-cylindrical dielectric lens
antennas—synthesis and height reduction technique. IEEE Trans. Antennas Propag. 58(5),
1783–1788 (2010)
69. R. Sauleau, B. Barès, A complete procedure for the design and optimization of
arbitrarily-shaped integrated lens antennas. IEEE Trans. Antennas Propag. 54(4), 1122–
1133 (2006)
Chapter 3
Advanced Feeds for mm-Wave Antenna
Systems

Jorge Teniente, Juan Carlos Iriarte, Iñigo Ederra

and Ramon Gonzalo

Abstract Millimeter-wave antenna systems have traditionally required high per-

formance feeds in order to fulfill its stringent requirements. Therefore, this goal has
been achieved by corrugated horns. However, in the last years new applications
mainly in the communication systems have driven the use of other types of antenna
feed with slightly reduced performance but simpler manufacturing at mm-wave and
submm-wave frequencies with improvements in cost reduction. These advanced
profiles are usually based in smooth waveguide advanced profiles. Besides, the
chapter describes different feed configurations based on metamaterial structures and
several examples of metamaterial based or inspired antennas are considered. Then,
this chapter covers the different alternatives currently used for mm-wave antenna
feed: corrugated horns, spline horns, and metamaterial horns. At the end of each
section, it includes some research successful results.

3.1 Introduction

Millimeter-wave antenna systems have traditionally required high performance

feeds in order to fulfill its stringent requirements. Therefore, this goal has been
achieved by corrugated horns. However, in the last years new applications mainly
in the communication area have driven the use of other types of antenna feed with

J. Teniente (&)
J.C. Iriarte
I. Ederra
R. Gonzalo

Electric and Electronic Engineering Department, Public University of Navarra,
Campus de Arrosadía, 31006 Pamplona, Navarra, Spain
e-mail: [email protected]
URL: http://antenas.unavarra.es
J.C. Iriarte
e-mail: [email protected]
I. Ederra
e-mail: [email protected]
R. Gonzalo
e-mail: [email protected]

© Springer International Publishing AG 2018 75

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_3
76 J. Teniente et al.

slightly reduced performance but with simpler manufacturing and reduced cost. The
simplicity of manufacturing is especially critical as frequency increases and in fact
corrugated horns are extremely difficult and expensive to manufacture at
submm-wave frequencies.
This chapter will cover the different alternatives currently used for mm-wave and
submm-wave antenna feeds: corrugated horns, smooth-walled horns where
multi-flare angle horns and splined horns will be introduced as alternatives to cor-
rugated horns at mm-wave and submm-wave frequencies. Finally, the chapter will
finish with the perspectives and actual research results regarding metamaterial-based
feeds toward future applications.

3.2 Advanced Corrugated Horn Antennas

This section describes the actual corrugated feed horn technologies employed at
mm-wave frequencies. It begins explaining the principles of the electrical behavior
of the electrical fields inside a corrugated waveguide. The hybrid mode basis as a
tool to help in the analysis of corrugated horns and the relation between waveguide
modes and free space gaussian modes are explained to understand the behavior of
the different types of corrugated horn antennas.
In fact, corrugated feed horns are difficult to manufacture as frequency increases
and usually are not used above W-band because their manufacture results quite
difficult. Corrugated horn fabrication above W-band usually needs using expensive
electroforming techniques or even stacked rings if the weight is not an important
parameter.

3.2.1 Introduction to Corrugated Horn Antennas

Nowadays, corrugated horns have become the preferred choice of feed antennas for
use in high restrictive applications. This is because of their superior radiation
performance, their high co-polar radiation pattern symmetry, and their low
cross-polarization. Usually they are the preferred choice when a superior radiation
pattern is needed for a specific application and the main market is spaceborne
communications and science missions in fact.
The operation principle of corrugated horns can be physically explained by
considering the way in which the corrugated wall affects the field distribution inside
a corrugated waveguide (see Fig. 3.1). It can be demonstrated that the corrugations
change the fields traveling through the waveguide to produce the desirable radiating
properties of axial beam symmetry, low sidelobes and low cross-polarization [1].
A linear electric field for low cross-polar level will be desirable but it cannot be
obtained with smooth waveguides that only support pure transverse electric (TE) or
3 Advanced Feeds for mm-Wave Antenna Systems 77

Fig. 3.1 Corrugated

waveguide

Fig. 3.2 TE11 and TM11 aperture electric fields

a pure transverse-magnetic (TM) modes. These modes have the aperture electric
field lines curved (see Fig. 3.2). Therefore, a multimode horn should be designed.
In [2], a special horn design to obtain an appropriate mode mixture by the addition
of TE11 and TM11 modes in a particular proportion and phase was presented, but its
bandwidth was very narrow.
It is well known that to define properly the field inside a circular waveguide the
most known basis used is the TE and TM mode family which are a direct solution of
the wave equation inside a smooth circular waveguide [3]. But if the waveguide is
corrugated it could be also useful to define the field inside the waveguide by the
family of hybrid modes HE and EH (those modes which does not present pure
transverse components along the waveguide). So, in fact we can choose to define
the field inside the corrugated horn antennas in terms of TE and TM modes or in
terms of HE and EH modes.
Theoretically, the hybrid modes HE1n present at the aperture of a circular
waveguide perfectly linear electric field lines, (see Fig. 3.3). It is interesting to aid
in the knowledge of corrugated horn antennas to deepen in the understanding of
hybrid modes of a corrugated waveguide.
As a corrugated horn is fed via a circular waveguide that propagates the fun-
damental mode of a circular waveguide (TE11) and considering that a corrugated
horn is composed of symmetrical radius variations of such circular waveguide at its
throat, the only hybrid modes to consider in a corrugated waveguide or horn
antenna are those with m = 1, HE1n and EH1n.
78 J. Teniente et al.

Fig. 3.3 HEmn and EHmn aperture electric fields

Table 3.1 HE11 mode decomposition in terms of TE1n and TM1n modes
   
TE modes 84:496% 0:082% 3:58
103 % 4:94
104 %
TE11  TE12 TE TE
0 180 13
180 14
180
   
TM 14:606% 0:613% 0:121% 0:039%
TM11 TM12 TM13 TM14
modes 0 0 0 0

Then, the HE11 mode is the fundamental mode of a corrugated waveguide and
presents linear electric field at the aperture, (see Fig. 3.3). The mode can be
expressed as a combination of TE and TM modes, as a generating basis that they
are. Commonly, in the bibliography [1], the HE11 mode is supposed to be a
combination of 85% of TE11 and 15% of TM11 with the adequate phase shift
between them. But in fact, this mode mixture is not perfect, (99.19% efficient with
HE11 mode), the perfect mode mixture in terms of smooth waveguide modes can be
seen in Table 3.1.
It is also important to know the radiation properties of hybrid modes to
understand the behavior of corrugated horn antennas. In Fig. 3.4, the radiation
diagrams of the first HE1n and EH1n modes are represented. In that figure we can
see, for example, the effect on the cross-polar component if a EH1n mode is excited
inside a corrugated horn. Or also from the same figure, the excellent radiation
properties of the HE11 with nearly null cross-polar component and with a radiation
diagram with quite low sidelobes can be observed.
The HE11 fundamental corrugated waveguide hybrid mode is then an excellent
mode for radiation purposes. For a common corrugated horn with oversized
3 Advanced Feeds for mm-Wave Antenna Systems 79

Fig. 3.4 Radiation properties of several HE1n and EH1n modes

80 J. Teniente et al.

Fig. 3.5 Radiation properties

of the HE11 hybrid mode
against aperture diameter (D)

aperture, the radiated cross-polarized level will be very low, the aperture illumi-
nation efficiency will be high and the sidelobe level will be also quite low, (see
Fig. 3.5).
As a conclusion of the radiation properties of the HE11 hybrid mode, it can be
said that this mode is an excellent mode to be excited at the aperture of a corrugated
waveguide, and it offers low cross-polar levels and quite low sidelobe levels.
However, if very low sidelobes are required for a high restrictive antenna, and it is
necessary to maintain low cross-polar levels, we would need several higher order
HE1n modes to be present at the aperture of the horn with appropriate amplitude and
phase shift between them. This is in fact the purpose of advanced corrugated feed
horn designs via optimization of the profiles and the corrugation parameters.
However, it is well known that one of the best ways to define a free space
radiation from a horn antenna is by means of the paraxial free space modes, the
gaussian modes, which are a solution of the paraxial free space equation. It is
important also to remark the radiation similarity between the fundamental gaussian
mode of a certain radiation pattern and the HE11 mode at a certain diameter
aperture. In fact, the fundamental gaussian beam mode can be decomposed in terms
of smooth waveguide modes (TEmn and TMmn) and also in terms of corrugated
waveguide hybrid modes (HEmn and EHmn) at a certain horn aperture radius
(R) with respect to the beamwaist (w0) of the gaussian beam at such horn aperture.
From Fig. 3.6, we can see that the fundamental gaussian mode can be expressed
completely as a combination of TE1n and TM1n smooth waveguide modes and also
as a combination of HE1n hybrid modes. At this point, it should be noted that HE11
mode has been always known as a gaussian-like mode because its radiated field is
nearly a pure gaussian, in fact it is up to 98.1% efficient with a fundamental
gaussian beam of R/w0 = 1.554 (w0/R = 0.6435) (see Fig. 3.6b). To obtain at the
aperture of a corrugated horn antenna a high efficient fundamental gaussian beam
mode (this implies lower sidelobes), more hybrid modes must be present at the horn
3 Advanced Feeds for mm-Wave Antenna Systems 81

Fig. 3.6 a Fundamental gaussian mode decomposition in terms of TE and TM b Fundamental

gaussian mode decomposition in terms of HE and EH

aperture as it has been introduced. This aspect will be covered in the following lines
regarding advanced corrugated feed horn design.

3.2.2 Radially Corrugated Feed Horn Antennas

The design of circular horn antennas was based, for a long time, on the control of
the waveguide mode mixture to excite an HE11 hybrid mode. It is well known from
the previous part that this hybrid mode can be made up of approximately a com-
bination of 85% TE11 and 15% TM11 smooth circular waveguide modes with an
appropriate relative phasing between them. The starting field distribution is usually
the TE11 mode of the circular waveguide under monomode operation, and by means
of a proper step or taper in the horn radius, the right amount of TM11 (amplitude and
phase) was excited (Potter-type horns) [1, 4, 5]. To get this mixture with nice
radiating features, two main parameters had to be considered: the output diameter
and the horn length. Since the coupling coefficient between waveguide modes is
directly related to the waveguide slope change, for a given output radius that fixes
the desired beamwidth, the change in horn length allows the designer to select the
appropriate phasing in the 85% of TE11 and 15% of TM11 mode mixture obtaining
the appropriate sidelobe and cross-polarization minimum levels. This type of horn
antennas has been extensively used in the past and is known as Potter-type horns,
[4]. Its drawback is the reduced bandwidth a design of this type could cover.
Another technique is based on corrugated circular waveguides and takes profit of
the fact that this mode mixture corresponds to the fundamental mode of a circular
corrugated waveguide, the HE11 mode. This technique reported in [1, 5–7] involves
a gradual matching of the smooth circular guide to another corrugated one wherein
the corrugation depth is smoothly tapered from k/2 to k/4. These two outlined
82 J. Teniente et al.

techniques are combined in the so-called radially corrugated horn antennas with a
matching device at their input port. In principle, corrugated horn antennas present a
wider frequency response than Potter-type horns. Their design parameters are
basically: corrugation parameters (period, duty cycle, depth, shape, etc.); length and
profile of the k/2-to-k/4 impedance matching transformer; and the horn geometry in
order to optimize the global performance of the horn. In the past, many of the
applications involving high performance horn antennas have been equipped with
radially corrugated horn antennas. Radially corrugated horn antennas are one of the
best possibilities to accomplish very high radiation pattern requirements and they
have been extensively used at mm-wave radiation applications with the only
drawback of their difficulty of manufacture as the frequency increases.
The most common radially corrugated horns are the ones where the tapering of
the horn is a constant slope. In this type of horns, the intention is to generate the
HE11 fundamental hybrid mode inside the corrugated waveguide and to guide it
smoothly to a certain aperture diameter. This type of horns avoided the coupling to
any other hybrid mode because the designer thought that the rest of hybrid modes
would ruin the radiation pattern. But every horn designer knows nowadays that this
is not completely true since the addition of HE1n modes does not affect cross-polar
level but can improve the main beam radiation pattern lowering the sidelobes and
helping to improve the Gaussian efficiency in the final radiation pattern, EH1n
hybrid modes must be avoided in any case since they add a lot of cross-polar
radiation (see Fig. 3.3). Radially profiled corrugated feed horn antennas that
improve the radiation pattern by means of optimization of their profile will be
presented in the next section. In this section, as a simpler way to understand
corrugated horns, the design of the ones that have a constant slope tapering is being
covered.
To design a radially corrugated horn antenna for a specific application, the first
thing we must define are the corrugation parameters. Such parameters that are
frequency related can be chosen according to the corrugated waveguide (see
Fig. 3.1) as follows:
• Input radius, R, must be of enough size to allow TE11 circular waveguide mode
to be above cutoff at the lowest usable frequency and as small as possible to be
in monomode operation at the highest cutoff frequency. Sometimes, this second
condition cannot be met and the TM01 mode could be present at the input radius.
This aspect is usually defined by the reflection coefficient requirement we must
achieve since bigger input radius facilitates this parameter.
• The corrugation depth, d, should be around k/4 with k de free space wavelength
at the central frequency (if there are several bands use the central frequency
between although such frequency is not considered). This is a parameter to be
optimized for every corrugation, so begin with a number around this k/4
mentioned.
• The corrugation period, p, should be around k/3 with k the lowest frequency.
Use a value rounded to the smallest integer in mm or use at least only one
3 Advanced Feeds for mm-Wave Antenna Systems 83

decimal value. The corrugation period is constant along all the corrugated horn,
if its dimension is optimized it complicates a lot the fabrication.
• The corrugation tooth width, w, should be around in between p/2 to p/5 with
p the corrugation period. Use a value rounded to the nearest integer in mm or
use at least only one decimal value. The thinner the corrugation tooth width, the
lower the weight of the corrugated horn, but too thin values complicate a lot the
fabrication, do not use values lower than 0.4 mm for this parameter if you plan
to manufacture the mm-wave corrugated horn with a lathe. The corrugation
tooth width is also constant along all the corrugated horn, if its dimension is
optimized it complicates a lot the fabrication.
Once the corrugation parameters have been decided, the corrugated horn profile
must be prepared. This preparation can be divided in to two parts, throat region and
flare region:
• The throat region controls mainly the reflection coefficient result. To obtain a
nice reflection coefficient for a corrugated horn antenna, the most common
technique employed is to adapt the transition between smooth circular waveg-
uide to corrugated waveguide. This is usually made via an impedance trans-
former with consists usually in a taper between k/2 and k/4 corrugation depths in
the first 4 to 10 corrugations, (see Fig. 3.7). As a beginning point, use a linear
taper in the first five corrugations and optimize this tapering for reflection
coefficient improvement. The designer must consider that the first corrugation
depth cannot be bigger than the input radius because the horn could not be
manufactured in one single piece via a lathe. This aspect is in fact the main
drawback of this type of corrugated horns.
• The flare region, (see Fig. 3.7), controls the radiation parameters. The designer
should use as a starting point, the diameter given in Fig. 3.8 for a certain
directivity. The flare angle selected must be the minimum possible considering
the sidelobe level or the spillover power the application can assume. The

Fig. 3.7 Detail of a k/2 to

k/4 impedance transformer at
the throat of a corrugated horn
antenna
84 J. Teniente et al.

designer must consider that a low flare angle means longer corrugated horn and
more difficult to manufacture, especially as frequency increases.
As a practical example of how to design a linear taper corrugated feed horn, let’s
say that we want to design a 22 dB directivity corrugated horn antenna with
sidelobes lower than –25 dB and as low return loss and cross-polar level as possible
in the most compact profile. Analyzing carefully Fig. 3.8 and by means of a simple
optimization via a mode matching software code [8, 9], the result leads to an
antenna of 6.6° taper profile with the first corrugation depth of 0.52
k decreasing
linearly with an impedance transformer of 1.9
k long to a corrugation depth for the
rest of the antenna of 0.25
k, see Fig. 3.9. The corrugation parameters have been
selected as p = k/5 and w = p/3. The resultant length of the profile is 17.8
k and
presents a diameter of 4.86
k. It is a quite long antenna for 22 dB directivity. This
length cannot be shortened if we must have the sidelobe level at –25 dB or less by

Fig. 3.8 a Directivity design curves for linear taper corrugated feed horns b Max. sidelobe level
design curves for linear taper corrugated feed horns

Fig. 3.9 Linear taper

corrugated horn antenna
design for 22 dB directivity
and −25 dB maximum
sidelobe level
3 Advanced Feeds for mm-Wave Antenna Systems 85

any method if we maintain the linear taper corrugated profile, but in the next section
we will learn that this can be made by means of an optimization of the profile.

3.2.3 Profiled Radially Corrugated Horn Antennas

In the last section, the linear taper corrugated feed horns have been covered, but
there are not much parameters to improve their performance. To overcome the
optimization limitations of the linear profiled corrugated horn antennas, during the
90s, several profiles by means of different formulas appeared in the scientific lit-
erature. The reason of such formulas was the availability of mode matching tech-
niques (MM) for the design of corrugated horns. However, since the computational
speed was not too much, the designers used several profile formulas to reduce the
number of unknowns in the profiled corrugated feed horn design and then increase
optimization speed.
A certain number of formulas appeared to solve this problem. In fact, every
research group in corrugated horn antennas had their own preferences [10], such
formulas were based in square roots, exponentials, sine-squared, polynomials,
gaussian, etc., see Fig. 3.10.
One of the formulas used to profile radially corrugated horn antennas was the
gaussian beam expansion formula that is in fact a form of a square root, [11]. This
type of formula aroused to implement a perfect match between corrugated
waveguide modes (mostly HE11 mode or similar mode mixtures) and the funda-
mental free space modes (fundamental gaussian mode, W00). By using this formula,
the matching between the waveguide and the free space was almost perfect, being
the most “natural” way to match the two media.
As a practical example of how to design a profiled corrugated feed horn with the
profile defined via a gaussian beam expansion formula, let’s say that we want to

Fig. 3.10 Examples of profiled corrugated horn antenna designs

86 J. Teniente et al.

design again a 22 dB directivity corrugated horn antenna but now we are asking for
sidelobes lower than –35 dB and as low return loss and cross-polar level as possible
in the most compact profile. We use for the profiled radially corrugated feed horn
definition two parts, the first part presents three sections: The first section at the
beginning presents a quick and smooth change in the slope at the throat, (less than
10% of the total first part length). The second section has a linear taper profile to get
quickly to the needed diameter (more than 65% of the total first part length). The
third section (less than 25% of the total length) is a smoothed end to allow con-
nectivity to the second part corrugated feed horn profile with a gaussian beam
expansion formula, see Fig. 3.11.
Analyzing carefully Fig. 3.11, the result leads to an antenna of 11.07
k length
and 5.64
k aperture diameter. The reduction in length, compared to the linear taper
profile, is 38%. It must be also considered the improvement in radiation pattern with
10 dB lower sidelobe level. In fact the bigger aperture diameter is caused by such
requirement of lower sidelobe level, the same radiation pattern would lead to even
shorter solution with the same aperture diameter.
Nowadays, the availability of optimized mode matching software packages [8, 9]
and the increased speed of computers make possible the optimization of a corru-
gated feed horn without the necessity to use a formula that defines the profile. In
fact, usually the corrugated feed horn designers use a formula to define just the
initial profile and after that, all corrugation radiuses (inner and outer) are optimized
individually, the final appearance of the optimized radially corrugated feed horn is
radically different from any formula and in fact the designer can adapt such opti-
mization to generate the smallest size possible profiled radially corrugated horn and
at the same time comply with all of the stringent requirements usually demanded for
this type of feed horns.

Fig. 3.11 Profiled radially

corrugated feed horn designed
and optimized with a
combination of two sections
defined by formulas for 22 dB
directivity and −35 dB
sidelobe level
3 Advanced Feeds for mm-Wave Antenna Systems 87

3.2.4 Corrugated Horn Antennas that Combine Axial

and Radial Corrugations

One of the main drawbacks for manufacture of radially corrugated feed horns is the
impedance transformer at their throat. This part is difficult to manufacture in a piece
by means of a lathe since the farther corrugations are also the deepest and indeed
the ones where the fabrication tolerance is worse. This can be solved making such
horn in several parts or by making the throat by means of stacked irises but in this
section an elegant solution to this problem is given. This solution not only solves
the common manufacture techniques; besides it even results in shortening the total
profile by a significant amount.
The solution we are referring to, is a horn antenna that combines horizontal
corrugations (known also as axial corrugations or chokes) for the throat region and
vertical corrugations (known also as radial corrugations) for the flare region [12].
The design guidelines for this type of corrugated horn antennas were fully
explained in [13].
This type of horn antenna achieves significant improvements on the following
four parameters, when compared to a normal radially corrugated horn, maintaining
the same radiation performance:
1. Shorter horn axial length
2. Improved return loss over a wide bandwidth
3. Smaller design computation complexity
4. Reduction of the manufacturing complexity (avoids the deep first radial
corrugations)
Again, as a practical example of how to design a corrugated horn antenna that
combines axial and radial corrugations, let’s say that we want also to design a
22 dB directivity corrugated horn antenna but now we are asking for sidelobes
lower than –40 dB and as low return loss and cross-polar level as possible in the
most compact profile. We use for the throat region an axially corrugated part that
presents a linear taper and six axial corrugations. For the flare region, we select a
profile defined by a gaussian beam expansion formula, see Fig. 3.12. The solution
leads to an antenna of only 7
k length and 6.3
k aperture diameter. The reduction in
length compared to the optimized radially corrugated profile is a 37% and compared
to the linear taper radially corrugated profile the reduction is a 61% and it should
also be considered the improved radiation pattern with −40 dB sidelobe level, in
fact again the bigger aperture diameter is caused by such requirement of lower
sidelobe level, the same radiation pattern would lead to even shorter solution with
the same aperture diameter as the previous solutions.
88 J. Teniente et al.

Fig. 3.12 Complete

corrugated GPHA with a
choked corrugated waveguide
input for 22 dB directivity
and −40 dB sidelobe level

3.2.5 Examples

In this section, two examples of advanced corrugated feed horn antennas are being
presented. Both have been manufactured at mm-wave frequencies with two dif-
ferent manufacturing techniques and tested obtaining the expected performance
simulations predicted.

3.2.5.1 Axial and Radial Corrugated Feed Horn Antenna Design

for a mm-wave 80–100 GHz Body Scanner

An 80−100 GHz corrugated horn design was optimized to act as feed of a rotating
reflector for a body scanner. The design was implemented by means of axial and
radial corrugations and the main requirement was a very low spillover to achieve as
pure as possible image from the body emission at mm-wave frequencies.
The design parameters were a FWHM (Full-Width Half-Maximum) of 14°,
gaussian beam decay radiation pattern till 26 ° in azimuth (view angle from the
subreflector) and sidelobes from 26° in azimuth lower than −30 dB. The maximum
measured cross-polar level should be below −25 dB and the measured reflection
coefficient should be lower than −20 dB.
The feed horn was designed with these specifications resulting in a total length
of 46.3 mm and an output diameter of 36 mm, (see Fig. 3.13). The manufacture
was made by means of iris rings of 0.7 mm and 0.2 mm sequentially stacked to
form the radially corrugated part. The axially corrugated part was manufactured
with a high-precision milling machine.
3 Advanced Feeds for mm-Wave Antenna Systems 89

Fig. 3.13 Manufactured axial and radial corrugated feed horn antenna design for a mm-wave 80–
100 GHz body scanner

Fig. 3.14 Simulated and measured reflection coefficient a and gain b for the axial and radial
corrugated feed horn antenna design for a mm-wave 80–100 GHz body scanner

Simulated and measured reflection coefficient can be seen in Fig. 3.14a. The
measured reflection coefficient is lower than −22 dB in the whole band. Regarding
the gain, see Fig. 3.14b, losses lower than 0.2 dB were measured as well. Measured
radiation patterns can be checked in Fig. 3.15, the measured cross-polar level is
always below −28 dB for the whole band. The measured main beam radiation
pattern decay is as expected from simulations, approximately of 14° FWHM.

3.2.5.2 Radially Profiled Corrugated Feed Horns for ESA

Funded MARSCHALS Airborne System

MARSCHALS was a three-channel limb sounder for the European Space Agency
(ESA) that flew on an aircraft. The designer of the mission was especially con-
cerned about the sidelobes generated by the feed horns or the rest of the optics. The
aim of the MARSCHALS project is the limb observation, and therefore any power
introduced via a sidelobe pointing the earth should be consider as noise decreasing
90 J. Teniente et al.

Fig. 3.15 Measured radiation patterns at 80 GHz (a), 90 GHz (b) and 100 GHz (c) for the axial
and radial corrugated feed horn antenna design for a mm-wave 80–100 GHz body scanner

the performance of the whole system. The sidelobe requirement was –35 dB in both
feed horn designs.
The requirements for these corrugated horn antennas (called antenna in band C
and antenna in band D) are as follows:
Other parameters common to both antennas are as follows:
• Sidelobes: It was required sidelobes to be less than −35 dB (with no “shoulders”
on the main lobe). Obviously, the lower the better.
• Peak cross-polar: below −35 dB (again, the lower the better)
• Feed waveguide: Full height rectangular waveguide (0.762  0.381 mm).
• Horn Length (flange to aperture): Maximum 40 mm.
The manufacture of the complete corrugated feed horn was made by means of
high-precision electroforming including the rectangular to circular transition to the
rectangular feed waveguide of 0.762  0.381 mm dimension. To reduce the
manufacture complexity, the corrugation tooth width should be exactly half of
the corrugation period and constant along the whole antenna.
The bandwidth requirements were not too tight, 2.8% for the C-band antenna
and 1.9% for the D-band antenna. The resultant directivity, by means of translation
3 Advanced Feeds for mm-Wave Antenna Systems 91

Table 3.2 MARSCHALS feed horns requirements

Band Frequency range (GHz) Center frequency (GHz) Beamwaist radius (mm)
C 316.5–325.5 321.00 2.078
D 342.2–348.8 345.50 2.003

Fig. 3.16 a MARSCHALS feed horn proposed profile for C-band b MARSCHALS feed horn
proposed profile for D-band

of the beamwaist radius values of Table 3.2, was quite high; 2.078 mm beamwaist
radius at 321 GHz in C-band antenna means an illumination at 16° of –35 dB and
therefore a directivity of 26 dB. As well, the 2.003 mm beamwaist radius at
345.5 GHz in D-band antenna means an illumination at 15.5° of –35 dB and
therefore a directivity of 26.3 dB.
The transition from rectangular to circular waveguide was designed ending in a
circular diameter of 0.762 mm. This diameter was used as the input waveguide
diameter for both antennas. After the optimization of both profiles, these were the
results:
• The total length of the C-band antenna is 39.1 mm (41.9
k) and the output
diameter 8.29
k (7.74 mm). A picture of the antenna is shown in Fig. 3.16a.
• The total length of the D-band antenna is 39.1 mm (45.1
k) and the output
diameter 8.53
k (7.398 mm). A picture of the antenna is shown in Fig. 3.16b.
The rectangular to circular transition design was very simple and it was a direct
cut of 15° angle of the circular input waveguide of each antenna in a 0.711 mm
length to result in the 0.762  0.381 mm input rectangular waveguide (see
Fig. 3.17).
92 J. Teniente et al.

Fig. 3.17 MARSCHALS feed horns throat region with rectangular to circular transition

Fig. 3.18 Mandrel of one of the MARSCHALS feed horns

The antenna was manufactured using an electroforming technique.

Electroforming is a metal forming process that forms parts through electrodeposi-
tion or electroplating on a model, known in the industry as a mandrel. The mandrel
was made by a milling machine and a picture of it can be seen in Fig. 3.18. Metallic
mandrels are pretreated chemically to allow subsequent separation of the finished
electroform. The outer surface of the mandrel forms the inner surface of the form.
A thick layer of electroplating copper is applied until the plate itself is strong
enough to be self-supporting. The mandrel is dissolved away after forming. The
surface of the finished part that was in intimate contact with the mandrel is rendered
in fine detail with respect to the original, and is not subject to the shrinkage that
3 Advanced Feeds for mm-Wave Antenna Systems 93

Fig. 3.19 Simulated far-field radiation pattern of both MARSCHALS feed horns

Fig. 3.20 Simulated far-field behavior of both MARSCHALS feed horns

would normally be experienced in a foundry cast metal object, or the tool marks of
a milled part. It can be observed in such Fig. 3.18, the extremely high-precision
manufacturing of the mandrel remembering that the corrugation tooth width is only
0.1 mm.
Simulated results for both antennas at central frequency can be found in
Figs. 3.19 and 3.20. Both antennas meet all the specifications widely.

3.3 Smooth-Walled Feed Horns

This section describes the technology concerning the smooth-walled feed horn
technology for mm-wave and submm-wave frequencies. It begins with a short
introduction to smooth-walled feed horn profiles and continues with the main
techniques to design this type of feed horns, multi-flare angle and spline profiled.
94 J. Teniente et al.

This type of horns presents worse radiation characteristics compared to corru-

gated profiled ones and are much longer. However, these are much simpler and
cheaper to manufacture at high frequencies, so many times are the preferred choice.
This problem regarding manufacture of corrugated horns at frequencies above
100 GHz becomes particularly acute when many tens, hundreds, or even thousands
of horns are required for large format focal plane array receivers. Therefore, there
has been much recent interest in the use of easy-to-fabricate smooth-walled horns
[14] at these shorter wavelengths.

3.3.1 Introduction to Smooth-Walled Feed Horn Profiles

Typically, smooth-walled horns either have step or flare angle discontinuities, or

smoothly varying interior profiles. The shapes of the horn interiors are optimized
numerically by mode matching techniques [8, 9] to generate a balance of higher
order waveguide modes which leads to a uniform aperture illumination and hence
far-field patterns with high beam circularity, low sidelobe levels and low
cross-polarization.
The simplest form of a smooth-walled horn is a Potter-type horn [15], which has
only one step discontinuity near the throat followed by a conical flaring section,
(see Fig. 3.21). The step discontinuity excites the TM11 mode at about 15% of the
total incident power of the incoming TE11 mode. Both fields propagate along the
conical flaring section until they arrive in phase at the horn aperture.
Figure 3.22 shows the electric field of the TE11 and TM11 modes, and it can be
seen clearly that the addition of both in the correct proportion and in phase produces
a linearly polarized electric field at the aperture of the horn. The combination of
85% of TE11 and 15% of TM11 leads in fact to an HE11 hybrid waveguide mode,
(see Table 3.1), being the mode mixture that generates automatically a linear taper
corrugated horn. As it was indicated in the section regarding corrugated horns, this
mode combination generates a highly uniform field that produces a radiation pattern
with low sidelobes level and low cross-polarization approximating to the perfor-
mance of a corrugated feed horn. But the main disadvantage of the Potter horn is
that it has a narrow operating bandwidth (5–8%).
However, the performance of a Potter-type horn can be substantially improved
by increasing the number of discontinuities of the horn. These types of horns are
known nowadays as multi-flare angle feed horns [16] and are being covered in the

Fig. 3.21 Principle of

operation of a Potter-type
horn
3 Advanced Feeds for mm-Wave Antenna Systems 95

Fig. 3.22 Combination of modes pursued in a Potter-type horn

Fig. 3.23 Principle of

operation of a Turrin-type
horn

following section. On the other hand, there is another method of optimizing a

smooth-walled feed horn profile with even better results, this method consists in the
optimization of the coefficients of a spline curve to meet the specific radiation
parameters needed for a certain application [14], these types of horns will be
covered in the next to the following section.

3.3.2 Smooth-Walled Multi-flare Angle Feed Horns

The Potter-type horn antenna exhibits nice radiation properties, but in a reduced
bandwidth. To improve the performance, the designer increases the number of
discontinuities of the horn. With some more parameters to optimize, the designer
can generate a carefully chosen combination of the higher order modes which could
widen the operating bandwidth of the horn and maintain the nice radiation
properties.
The simplest multi-flare angle horn is the Turrin-type horn [17], see Fig. 3.23.
This type of horn starts with a short conical section with a cone angle and then is
followed by another conical section of lower angle.
After the Turrin-type horn that only has two different angles, (see Fig. 3.23), the
designer can add as many discontinuities in the throat region of the horn as he
needs, these types of horns are then known as multi-flare angle horns, (see
Fig. 3.24). The optimized depth of these discontinuities (R0, R1, R2, R3…), the horn
length and flare angles (L1, L2, L3…), can be predicted using modal matching in
conjunction with optimization algorithms to derive the desired radiation properties
in a certain bandwidth.
Step or flare angle discontinuities near the throat of the horn will, unavoidably,
excite other higher modes (TE1n and TM1n) in addition to the desired TM11 mode.
96 J. Teniente et al.

Fig. 3.24 Multi-flare angle smooth waveguide feed horn with three steps

Fig. 3.25 Direct metallic laser sintering manufactured 2  2 antenna array of multi-flared angle
smooth-walled feed horns for a communications system working at 330 GHz (courtesy of
ANTERAL S.L.)

These higher order modes will affect the aperture field in a complicated, frequency
dependent way, making the design of multi-flare angle horns that give good per-
formance over a finite bandwidth more difficult. Fortunately, the effect of these
modes on the far-field pattern of a particular horn can be predicted very accurately
using the numerical modal matching techniques. Since such modal matching
techniques can be used to calculate the far-field patterns, it is possible also to use
this technique in conjunction with suitable optimization algorithms to determine the
optimum horn profiles for good performance over a particular bandwidth [7, 8].
These types of horns are very suitable for manufacture at mm-wave and
submm-wave frequencies since they can be made very quickly and cheaply by
repeated drilling with a properly shaped electrode as machine tool, [18] or even by
additive manufacturing techniques with a certain postprocessing of the manufac-
tured result. In Fig. 3.25, a set of four feed multi-flare angle smooth-walled feed
horns for a communication system at 330 GHz is presented. Such submm-wave
horn array was manufactured by means of an additive manufacturing technique
called direct metallic laser sintering, the array needed a postprocessing with
properly shaped electrode to be fully functional.
3 Advanced Feeds for mm-Wave Antenna Systems 97

3.3.3 Smooth-Walled Spline Profiled Feed Horns

The smooth-walled profiles defined in the previous section are formed by steps, but
it could be interesting to have the possibility to design a smooth-walled profile
whose taper is smoothly opening without any discontinuity. To fill this gap, in
2004, the researchers from CSIRO in Australia presented their contribution, and
they called it the smooth-walled spline profile [14]. This profile was developed as a
substitute of corrugated horns for mm-wave and submm-wave frequencies since the
manufacturing complexity is much reduced. Their performance is not as good as a
corrugated horn but it is quite acceptable considering a bandwidth around 30% can
be easily achieved for cross-polar levels below −30 dB. In fact, in many cases (as
feed for multibeam reflector antennas) smooth-walled spline profiled feed horns
have replaced corrugated horns for onboard satellite payload [19].
The spline profile is a numeric function that is piecewise defined by polynomial
functions and which possesses a high degree of smoothness at the places where the
polynomial pieces connect, see Fig. 3.26. In fact, the term spline was adopted from
the name of a flexible strip of metal commonly used by drafters to assist in drawing
curved lines.
To design a smooth-walled spline profile feed horn antenna for a specific
application, we must define first the initial dimensions of the horn we want to
develop. We need as an initial approach three parameters: input radius, output
aperture radius and total length. Once they are defined, we should decide how many
optimization points we need. Many optimization points will allow a better result,
but the optimization time will be longer; a few optimization points (in the Fig. 3.26
only five variables are used) allow a quicker optimization time, but perhaps they are
not enough to reach the requirements needed.
If the designer needs to achieve a better result, he must increase the number of
optimization radiuses, in fact in [20], the authors increase the number of optimiz-
able radiuses till 20 unknowns improved the result significantly.
As an example, a K/Ka-band smooth-walled spline profile feed horn for a
communications satellite is presented, see Fig. 3.27.

Fig. 3.26 Spline profile feed horn geometry

98 J. Teniente et al.

Fig. 3.27 Optimized smooth-walled spline profile feed horn for onboard a communications
satellite (courtesy of ANTERAL S.L.)

Such a feed horn has more than a 42.3% bandwidth and the partial bandwidths
were 3.5% and 2.4%. Therefore, it is a wideband design and its requirements are
extremely severe for a smooth-walled horn, since a return loss value better than
40 dB and a cross-polar level below −26 dB were required for both frequency
bands, and they are quite apart in frequency. In addition, it is very short for a spline
profile.
The design employs 40 radiuses to define the profile (see Fig. 3.27). The out-
come is that of a rather curved inside for smooth-walled horn, but the results (see
Figs. 3.28 and 3.29) show that the requirements are met.
The resulting radiation patterns can be seen in Fig. 3.29. This horn antenna was
manufactured and is successfully operating onboard a geostationary satellite, where
it is part of a feed horn array for spot beam communications.

3.4 Metamaterial Based Feeds

This section describes different feed configurations based on metamaterial struc-

tures. The field of metamaterials is very broad and several examples of metamaterial
based or inspired antennas are considered along this book. In this case, metama-
terial antennas which resemble or create horn-like antennas are treated. First the
metallic walls of the horns will be substituted by a tailored surface so that different
properties are achieved. These antennas are based on the so-called hard and soft
surfaces, which can be considered as special cases of metasurfaces. Finally, horn
antennas will be constructed based on 3D electromagnetic bandgap
(EBG) structures. Their fundamentals and several examples are given in the fol-
lowing sections.
3 Advanced Feeds for mm-Wave Antenna Systems 99

Fig. 3.28 Results of the optimized smooth-walled spline profile: a Return loss b Cross-polar level
c Aperture efficiency d Spillover loss above 13°

3.4.1 Soft and Hard Horns

Corrugated horn antennas can be considered as a particular case of the more general
concept of hybrid mode horns [21]. These antennas support linear polarized modes
provided that an anisotropic boundary condition exists in the horn walls. The
condition to be fulfilled by the boundary impedances is called balanced hybrid
condition:

ZTE ZTM ¼ g20 ;

where ZTE and ZTM are the TE and TM boundary impedances and g0 is the free
space wave impedance.
Two families of antennas satisfy this condition, leading to the so-called hard and
soft horns [21]. In the soft horn case, the boundary conditions correspond to:
100 J. Teniente et al.

Fig. 3.29 Far-field radiation patterns of the optimized smooth-walled spline profile: a At fmin1
b At 1.04
fmin1 c At 1.5
fmin1 d At 1.54
fmin1

Ex
Z TE ¼ Zx ¼ ¼0
Hz
Ez
Z TM ¼ Zz ¼ ¼1
Hx

whereas in the hard horn they are:

Ex
Z TE ¼ Zx ¼ ¼1
Hz
Ez
Z TM ¼ Zz ¼ ¼0
Hx

Soft horns support tapered aperture distributions which lead to very low side
lobe radiation patterns. Conversely, hard horns provide uniform aperture distribu-
tions which correspond to very high aperture efficiencies.
These hybrid mode horns have been implemented with different techniques,
which are reviewed in [22]. These include longitudinal and transversal corrugated
horns, dielcore horns, strip loaded and metamaterial or metasurface wall horns.
3 Advanced Feeds for mm-Wave Antenna Systems 101

This last case constitutes the most novel implementation and, thanks to the flexi-
bility of metamaterials, a very promising alternative to the conventional imple-
mentations. The following sections describe different implementations of these
concepts.

3.4.2 Metamaterial Horns

Metamaterial-based horn antennas, as schematically shown in Fig. 3.30, can be

employed for alternative implementations of the dielcore horn designs. Since
metamaterials allow obtaining values of the dielectric constant lower than 1, both
hard and soft surface designs can be constructed based on the dual dielcore horn. In
both cases the core dielectric can be taken as air, which requires the outer dielectric
to have dielectric constant lower than 1, which can be achieved by a metamaterial.
This has additional advantages, since removing the central core allows reducing
mass and losses and improves reflections in the interface between the air-filled
feeding waveguide and the dielectric loaded horn antenna.
The concept behind these antennas was first demonstrated by [22]. Simulation
results showed that the performance of these antennas complied with the expec-
tations. However, the results were based on theoretical materials with ideal
parameters.
The challenge of finding a metamaterial realization which complied with the
required parameters was sorted out by [22]. The developed metamaterial is based on
a wire grid with wires in the three coordinate axis directions, so that the structural
elements of the metamaterial affect all the field components and full control of the
surface impedance can be achieved. The first experimental demonstration of a horn
antenna based on this metamaterial wall was carried out by the same group [23].
A photograph of the manufactured antenna is shown in Fig. 3.31a. The achieved
pattern was symmetric, with low sidelobes along the full super-extended C-band.
A comparison between the E-plane pattern of this antenna and a trifurcated and
standard pyramidal horn is presented in Fig. 3.31b. Lower sidelobes than in the
conventional counterparts have been achieved by means of the metamaterial
surface.

Fig. 3.30 Schematic of hard

and soft antenna
implementations using
metamaterials as alternatives
to the dual dielcore horns
102 J. Teniente et al.

Fig. 3.31 a Photograph of metamaterial horn prototype, with inset showing the wire grid used to
create the metamaterial surface. b Comparison of the radiation pattern on the metamaterial horn, a
standard pyramidal horn and trifurcated horn [23]

In order to simplify the antenna manufacturing a planar metasurface was pre-

sented in [24], where a metasurface implemented on a printed circuit board (PCB)
was used to create a soft horn antenna. Photographs of the prototype are shown in
Fig. 3.32. The achieved performance shows sidelobes at the −30 dB level, in
agreement with the soft nature of the horn. These sidelobes are nearly 20 dB lower
than those obtained with a similar smooth wall pyramidal horn and are maintained
in the whole Ku-band.
An example of the measured patterns is presented in Fig. 3.32b. Very sym-
metrical patterns are achieved leading to low cross-polarization levels (around
30 dB).
One problem associated with these antennas is the need of a matching section to
reduce reflections at the interface between the smooth wall and the metamaterial
sections of the horn. In this case, a dielectric section with an optimized profile is
3 Advanced Feeds for mm-Wave Antenna Systems 103

Fig. 3.32 a Photographs of the fabricated metahorn antenna and the metasurface liners on the
PCB. b Measured and simulated E-plane, H-plane, and / ¼ 45 plane cut co- and cross-polarized
radiation patterns of the metahorn at 14 GHz [24]

used for this purpose. This requires a brute force optimization which can be avoided
if a tapered transition is implemented in the metamaterial wall liner. This approach
has been followed in [25].
Finally another implementation of this concept was used in [25], applied to a
conical horn antenna. The metasurface was based on a mushroom structure, see
Fig. 3.33.
104 J. Teniente et al.

Fig. 3.33 Schematic of the

conical metamaterial horn
antenna proposed in [25]

3.4.3 EBG Horn Antennas

Electromagnetic band gap materials can be considered as a type of metamaterials.

They are 3D periodic arrangements of elements, usually dielectric, which, by
properly choosing their shape and spatial position present a frequency range where
no propagation of electromagnetic radiation is allowed [26]. This frequency range is
called the bandgap and its use has been proposed to reduce the effect of surface
waves in planar antennas [27]. This effect can be obtained with planar configura-
tions, usually based on printed board technologies or with 3D structures, where the
material must be tailored in three dimensions.
In this last case, if defects are created in the otherwise perfect periodic structure,
cavities and waveguides can be created. These can be considered as linear defects
and their performance is similar to that of metallic waveguides. Therefore, the same
techniques that are used in order to create horn antennas can also be applied in this
case.
Horn-like structures have been proposed based on 2D and 3D EBG structures
[28]. However, we will focus on those horn configurations based on 3D EBG
structures. Among them, the most used EBG for these experiments is the woodpile
or layer-by-layer structure [29, 30], due to its simplicity and relatively easy man-
ufacturing up to the submillimeter wave range [31, 32]. As a matter of fact, it is in
these high frequency ranges where this type of configurations can be of interest
thanks to their low losses due to the absence of metals (provided the EBG structure
is fabricated with sufficient precision).
3 Advanced Feeds for mm-Wave Antenna Systems 105

3.4.4 Sectoral Horn Antennas

The simplest way to create a horn antenna based on a woodpile structure is by

altering the bars of one of its layers. Removing one bar will create a waveguide and
by tilting those bars that form the waveguide walls a sectoral antenna is created.
Given the polarization of the mode in the woodpile feeding waveguide, this antenna
corresponds to a H-plane sectoral horn. The radiation properties of the antenna
depend on the flare angle and the antenna length, as in standard metallic sectoral
horns.
These configurations were first studied by Weily et al. [33]. An example of the
radiation pattern achieved can be seen in Fig. 3.34b. It is worth noting the non-
symmetric E-Plane cut, due to the asymmetry of the woodpile in the stacking
direction. Moreover, given the adiabatic nature of the transition, good impedance
matching was achieved between the EBG waveguide section and the horn. In [34],
the same group proposed a transition between rectangular and EBG waveguides
with good performance. This configuration was employed to create an array of such
antennas with which scanning performance was demonstrated. The radiation pattern
of the 16-element array for different scanning angles is shown in Fig. 3.34c.

3.4.5 Evanescently Fed EBG Horn Antenna Arrays

Another way of controlling the radiation E-plane is by arraying two sectoral EBG
antennas. They can be simultaneously fed by means of the so-called evanescent
coupling method [35]. In this configuration, an EBG waveguide is used to couple
the two sectoral horns by properly adjusting their relative position, as shown in
Fig. 3.35. This way, narrower E-plane cut can be obtained while maintaining the
same H-plane of the individual sectoral antennas, as shown in Fig. 3.36, where the
simulated and measured antenna radiation patterns obtained with this antenna are
compared.

3.4.6 Pyramidal Horn Antennas

As mentioned above, the most natural sectoral horn antenna which can be created
taken as base the woodpile structure is a sectoral H-horn antenna, where the electric
field is parallel to the woodpile stacking direction. Thus, the E-plane dimensions of
such horn antennas are determined by the thickness of the woodpile bars, since
106 J. Teniente et al.

Fig. 3.34 a Perspective view of the woodpile EBG sectoral horn antenna [34]. b Radiation pattern
c Array theory and FDTD computed E-plane radiation patterns for the 16-element linear array for
scan angles of: a 0º, b 60º, c 30º

these horns are created within a single woodpile layer. Such a narrow and prede-
termined aperture leads to a broad beam in the E-plane, which is nonsymmetrical
due to the absence of mirror symmetry in the stacking direction of the woodpile
structure.
Reducing the E-plane beamwidth would require creating a pyramidal horn
antenna. The difficulty of creating a pyramidal horn based on a defect-containing
woodpile structure lies in the complexity of the embedding medium itself.
3 Advanced Feeds for mm-Wave Antenna Systems 107

Fig. 3.35 Front and side view of the evanescently fed double-EBG horn antenna. Dark areas
represent the feeding EBG waveguide region and the EBG horn antenna layers [35]

Fig. 3.36 Comparison between simulation and measurements [35]

108 J. Teniente et al.

Fig. 3.37 Woodpile based symmetrical EH-horn antenna (a—front view, b—3D view) [36]

Fig. 3.38 Radiation pattern of the EBG pyramidal horn in Fig. 3.37 [36]

As shown in [36], simple solutions based on cutting several woodpile layers lead to
resonant structures. A solution to his problem can be obtained if the woodpile layers
are tilted as shown in Fig. 3.37.
This EBG pyramidal horn allows simultaneously controlling the E- and H-planes
of the radiation patterns. An example of the pattern that can be obtained is presented
in Fig. 3.38. Its main drawback comes from the manufacturing difficulties, since the
layer-by-layer manufacturing approach followed for the woodpile is not valid
anymore.
3 Advanced Feeds for mm-Wave Antenna Systems 109

References

1. A.D. Olver, P.J.B. Clarricoats, A.A. Kishk, L. Shafai, Microwave horns and feeds, in IEE
Electromagnetic waves series 39 (The Institution of Electrical Engineers, 1994)
2. P.D. Potter, A new horn antenna with suppressed sidelobes and equal beamwidths. Microw.
J. 6, 71–78 (1963)
3. S. Ramo, J.R. Whinnery, T. van Duzer, in Fields and Waves in Communication Electronics
(Wiley, 1994)
4. P. D. Potter, A new horn antenna with suppressed sidelobes and equal beamwidths. Microw.
J. VI, 71–78 (1963)
5. P.J.B. Clarricoats, A.D. Olver, Corrugated horns for microwave antennas, in IEE
Electromagnetics Waves Series 18, Chap. 5, 6 and 7 (Peter Peregrinus, 1984)
6. G.L. James, Analysis and design of TE11 to HE11 corrugated cylindrical waveguide mode
converters, in IEEE Transactions on Microwave Theory and Techniques, vol. MTT-29,
pp. 1059–1066, 1981
7. B.M. Thomas, G.L. James, K.J. Greene, Design of high-performance wideband corrugated
horns for Cassegrain antennas, IEEE Transactions on Antennas and Propagation, vol. AP-34,
pp. 750–757, 1986
8. Mician lWave Wizard Software. http://www.mician.com
9. Ticra’s CHAMP Software. http://www.ticra.com/products/software/champ
10. C. Granet, G.L. James, Design of corrugated horns: a primer. IEEE Antennas Propag. Mag.
47(2), 76–84 (2005)
11. R. Gonzalo, J. Teniente, C. del Río, Improved radiation pattern performance of horn antennas
with gaussian profiled shapes. IEEE Trans. Antennas Propag. 50(11), 1505–1513 (2002)
12. J. Teniente, R. Gonzalo, C. del-Río, Choked gaussian antenna: extremely low sidelobe
compact antenna design, in IEEE Antennas and Wireless Propagation Letters, vol. 1,
pp. 200–202, 2002
13. J. Teniente, A. Martínez, B. Larumbe, A. Ibañez y R. Gonzalo, Design guidelines of horn
antennas that combine horizontal and vertical corrugations for satellite communications, in
IEEE Transactions on Antennas and Propagation, vol. 63, No. 4, pp. 1314–1323, 2015
14. C. Granet, G.L. James, R. Bolton, G. Moorey, A smooth-walled spline-profile horn as an
alternative to the corrugated horn for wide band millimeter-wave applications. IEEE Trans.
Antennas Propag. 52(3), 848–854 (2004)
15. P.D. Potter, A new horn antenna with suppressed sidelobes and equal beamwidths. Microw.
J. 6, 71–78 (1963)
16. J. Leech, B.K. Tan, G. Yassin, Smooth walled feed horns for mm and submm radio
astronomy, in 6th UK, Europe, China Millimeter Waves and THz Technology Workshop
(UCMMT), 2013
17. R.H. Turrin, Dual mode small-aperture antennas. IEEE Trans. Antennas Propag. 15(2),
307–308 (1967)
18. J. Leech, B.K. Tan, G. Yassin, P. Kittara, S. Wangsuya, J. Treuttel, M. Henry, M.L. Oldfield,
P.G. Huggard, Multiple flare-angle horn feeds for sub-mm astronomy and cosmic microwave
background experiments. Astron. Astrophys. 532 (2011)
19. J. Teniente, D. Valcázar, B. Larumbe, A. Martínez, A. Ibañez, R. Gonzalo, Feed horn
antennas for data downlink and uplink spaceborne communications, in 10th European
Conference on Antennas and Propagation, EuCAP, 2016
20. L. Zeng, C.L. Bennett, D.T. Chuss, E.J. Wollack, A low cross-polarization smooth-walled
horn with improved bandwith, IEEE Trans. Antennas Propag. 58(4) (2010)
21. E. Lier, P.S. Kildal, Soft and hard horn antennas. IEEE Trans. Antennas Propag. 36(8),
1152–1157 (1988)
22. E. Lier, Review of soft and hard horn antennas, including metamaterial-based hybrid-mode
horns. IEEE Antennas Propag. Mag. 52(2), 31–39 (2010)
110 J. Teniente et al.

23. C.P. Scarborough, Q. Wu, D.H. Werner, E. Lier, R.K. Shaw, B.G. Martin, Demonstration of
an octave-bandwidth negligible-loss metamaterial horn antenna for satellite applications.
IEEE Trans. Antennas Propag. 61(3), 1081–1088 (2013)
24. Q. Wu, C.P. Scarborough, B.G. Martin, R.K. Shaw, D.H. Werner, E. Lier, X. Wang, A
Ku-Band dual polarization hybrid-mode horn antenna enabled by printed-circuit-board
metasurfaces. IEEE Trans. Antennas Propag. 61(3), 1089–1098 (2013)
25. Q. Wu, C.P. Scarborough, D.H. Werner, E. Lier, R.K. Shaw, Inhomogeneous metasurfaces
with engineered dispersion for broadband hybrid-mode horn antennas. IEEE Trans. Antennas
Propag. 61(10), 4947–4956 (2013)
26. J.D. Joannopoulos, R.D. Meade, J.N. Winn, Photonic Crystals (Princeton University Press,
1995)
27. P. de Maagt, R. Gonzalo, J.C. Vardaxoglou, J.M. Baracco, Photonic bandgap antennas and
components for microwave and (sub)millimetre wave applications. IEEE Trans. Antennas
Propag. 51(10), 2667–2677 (2003)
28. A.R. Weily, K.P. Esselle, B.C. Sanders, Photonic crystal horn and array antennas. Phys. Rev.
E. 68, 016609-1-016609-6 (2003)
29. K.M. Ho, C.T. Chan, C.M. Soukoulis, R. Biswas, M. Sigalas, Photonic band gaps in three
dimensions: New layer-by-layer periodic structures. Solid State Commun. 89(6), 413–416
(1994)
30. H.S. Sözüer, J. Dowling, Photonic band calculations for woodpile structures. J. Mod. Opt. 41
(2), 231–239 (1994)
31. R. Gonzalo, B. Martínez, C.M. Mann, H. Pellemans, P. Haring-Bolivar, P. de Maagt, A
low-cost fabrication technique for symmetrical and asymmetrical layer-by-layer photonic
crystals at submillimeter-wave frequencies. IEEE Trans. Microw. Theory Tech. 50(10),
2384–2392 (2002)
32. B. Martínez, I. Ederra, R. Gonzalo, B. Alderman, L. Azcona, P.G. Huggard, B. de Hon, A.
Hussain, S.R. Andrews, L. Marchand, P. de Maagt, Manufacturing tolerance analysis,
fabrication, and characterization of 3-D submillimeter-wave electromagnetic-bandgap crys-
tals. IEEE Trans. Microw. Theory Tech. 55(4), 672–681 (2007)
33. A.R. Weily, K.P. Esselle, B.C. Sanders, Layer-by-layer photonic crystal horn antenna. Phys.
Rev. E 70, 4–037602 (2004)
34. A.R. Weily, K.P. Esselle, T.S. Bird, B.C. Sanders, Linear array of woodpile EBG sectoral
horn antennas. IEEE Trans. Antennas Propag. 54(8), 2263–2274 (2006)
35. I. Khromova, I. Ederra, J. Teniente, R. Gonzalo, K.P. Esselle, Evanescently-fed electromag-
netic band gap horn antennas and arrays. IEEE Trans. Antennas Propag. 60(6), 2635–2644
(2012)
36. I. Khromova, I. Ederra, R. Gonzalo, B.P. de Hon, Symmetrical pyramidal horn antennas
based on EBG structures. PIER B 29, 1–22 (2011)
Chapter 4
Flat Corrugated and Bull’s-Eye Antennas

Miguel Beruete, Unai Beaskoetxea and Tahsin Akalin

Abstract This chapter is focused on a family of antennas recently proposed, whose

common denominator is that they consist in a flat metallic plane with a central
narrow aperture surrounded by corrugations. This new family of antennas has
occupied a very important place in the development of communication technologies
and solutions for several communication applications since, compared to higher
volume structures, as horn or parabolic antennas, they present equal or even higher
radiation characteristics. An introduction, which also serves as an historical over-
view, is presented in the first place, emphasizing the relation with both extraordi-
nary transmission structures demonstrated initially at optical wavelengths as well as
with leaky-wave antennas, developed mainly in microwaves. Afterwards, the
physical mechanism for radiation and the main guidelines for the design of these
antennas are discussed. Next, an overview of some of the most appealing designs
and results related with this technology is presented, putting special emphasis in
terahertz-band applications. Finally, tips on foreseen future trends are summarized
to conclude the chapter.

M. Beruete (&)
U. Beaskoetxea

Antennas Group-TERALAB, Universidad Pública de
Navarra, 31006 Pamplona, Spain
e-mail: [email protected]
U. Beaskoetxea
e-mail: [email protected]
T. Akalin
Institute of Electronics, Microelectronics and Nanotechnology (IEMN),
Lille University, Lille, France
e-mail: [email protected]

© Springer International Publishing AG 2018 111

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_4
112 M. Beruete et al.

4.1 Introduction: Extraordinary Transmission Resonance

and Evolution Towards Metallic Corrugated Antennas

Daily experience teaches us that the amount of light able to penetrate an aperture
perforated on a metallic plane is proportional to the size of the aperture. More
technically, the power coupled through the hole is limited by diffraction and, for
sufficiently small apertures, diffraction rings will appear at the output, a fact that is
customarily exploited to demonstrate the wave nature of light in typical lab
experiments. Quantifying the fraction of power that goes through a hole perforated
in an infinite metal film is a classical problem that was solved by Bethe and
Bouwkamp some time ago [1, 2]. Using standard diffraction theory, they estab-
lished that this power is proportional to (r/k)4, where r is the radius of the aperture
and k is the operation wavelength. The previous expression confirms our intuition
putting in evidence that for wavelengths much larger than the hole diameter only a
small fraction of the incoming power can traverse the aperture.
The theory of Bethe and Bouwkamp provided a satisfactory solution and
remained unquestioned for more than 30 years. However, in 1998 the scenario
changed radically when Ebbesen et al. [3] reported an experiment where high
transmittance peaks appeared in an aluminium plate perforated with a periodic
matrix of sublambda holes at wavelengths as large as ten times the diameter of the
circular apertures. This result came as a surprise, since the holes were operating in
cutoff and still were able to produce a high transmittance, in apparent contradiction
with Bethe-Bouwkamp’s theory.1 For this reason, this phenomenon was called
Extraordinary Optical Transmission (EOT). Ebbesen’s experiment was done in the
near infrared band where metals admit a Drude model and have a negative real part
of permittivity, as corresponds to an overdense plasma. This negative permittivity
property allows the coupling of surface plasmon polaritons (SPPs), which are
surface waves sustained at the metal-dielectric (also known as metal-insulator, MI)
interface [4]. SPPs are bound modes and hence are not directly accessible by an
external incident wave. The extra momentum of the tangent wave vector necessary
to excite SPPs can be achieved by an external prism or by using a periodic structure,
as was the case in the experiment by Ebbesen et al.. Thus, the explanation proposed
to elucidate the high transmittance was the coupling of light with SPPs excited
through the periodic matrix of holes.
Ebbesen’s result ignited the spark of modern SPP research, leading to the so-called
“surface-plasmon resurrection” [5] (excellent reviews can be found in [6, 7]).
A model able to embody the physics involved in the extraordinary transmission of
light through an array of subwavelength holes in a metal film was presented by
Martín-Moreno et al. [8]. With this analysis, it was established that the phenomenon

1
Strictly speaking, Ebbesen’s results do not contradict Bethe-Bouwkamp’s theory. Recall that the
latter is valid for an isolated hole in an infinite metal film, rather different from the hole matrix
studied by Ebbesen.
4 Flat Corrugated and Bull’s-Eye Antennas 113

depended crucially on the periodic structure. Even more, a minimal model able to
catch the main underlying physics of the phenomenon was presented.
Following the path initiated by Ebbesen, other structures capable of enhancing
the transmission through small apertures were investigated. In [9], one-dimensional
metallic gratings with very narrow slits periodically distributed over a dielectric
substrate were studied for impinging light beams with polarization perpendicular to
the aperture. The transmission resonances obtained were related to the excitation of
two types of electromagnetic modes: (i) the coupling between the SPPs in the two
surfaces of the grating, when the impinging wavelength was approximately equal to
the grating period; and (ii) waveguide modes in the slits when the wavelength was
much larger than the period. It was pointed out that those effects appeared for any
range of frequencies, provided the width of the slits was small in comparison to the
period and the frequency of the incident wave was lower than the plasma frequency
of the metal.
Further studies were carried on in [10], analyzing structures formed only by
holes and by holes and periodically inserted dimples. It was observed that the
zero-order transmission could be improved at certain wavelengths for the holes and
dimples structure, although the rest of the spectrum remained almost equal. This
proved that the light coupling depended mainly on the period of the corrugations
rather than on their shape. Given that the dimples do not emit but do contribute to
the peaks, the authors suggested that the surface modulation caused by the dimples
helped to couple the incident photons to SPPs, whereas the holes also contributed to
the opposite operation coupling again the surface plasmons to photons. Taking this
idea to the limit, they demonstrated that a single sub-wavelength aperture sur-
rounded by a dimple array had a transmittance enhancement proportional to k4/r2
compared to a hole in a flat plane and that the transmission peak could be tuned in
frequency by just changing the period of the dimples. In [11], it was showed that the
use of metamaterial shaped apertures rather than typical geometries (ellipsoidal or
quadrilateral holes), i.e., Babinet metamaterials [12, 13], led to a reinforcement of
the EOT effect. More specifically, a higher transmittance was observed at lower
frequencies, while the transmission efficiency at higher frequencies remained
almost still.
The natural step forward was to test other periodic structures flanking a central
aperture. In [14], two different structures were analyzed: a small hole surrounded by
concentric annular grooves, a structure called Bull’s-Eye and a narrow slit sur-
rounded by straight grooves (respective schemes are shown in Fig. 4.1a and b). In
either case, an enhancement of transmission was demonstrated both when the
grooves where carved on the input and output faces. It was found that grooves on
the input face contribute to a more efficient light harvesting from the incident wave,
which was then directed towards the central aperture and radiated almost isotrop-
ically, with a clear diffractive pattern. On the other hand, with grooves on the output
face, light harvesting at the input did not change compared to a single aperture on a
flat metallic plane, but radiated a directive beam at the output with its maximum
pointing perpendicularly to the surface. It was also confirmed that when the output
observation angle was shifted away from the normal direction, the maximum
114 M. Beruete et al.

Fig. 4.1 Schematic of a Bull’s-Eye with circular aperture and b slit aperture with parallel
grooves. Insets observed optical response of the corresponding sample illuminated from the back
at its wavelength of maximum transmission. Artwork inspired in the results of Lezec et al.

transmission intensity dropped and the peak splitted into two peaks, moving each
one towards lower and higher wavelengths respectively.
As a theoretical complement, in [15] the beaming mechanism for this structure
was described in detail, analyzing the angular distribution of light at the exit side. It
was shown that the beaming intensity saturated for a certain number of grooves. It
was also concluded that the emission process basically consists of two steps: first,
the output side of the slit emits a primary beam directly into free space and into the
grooves; then, these grooves reradiate light to vacuum and into other corrugations.
It was demonstrated that the grooves’ depth plays an important role in the beaming,
as it shifts the wavelength at which the maximum occurs, as well as the achievable
maximum beam intensity. These cavity modes—controlled by the grooves’
dimensions—and the coupling between indentations—maximum at wavelengths
commensurable with the period—lead to the surface resonances responsible for the
observed beaming properties. These results were experimentally confirmed after-
wards in [16]. It was proved that the maximum boost of transmission takes place
when three radiation mechanisms coincide at the same frequency: the previously
mentioned groove cavity mode excitation, the slit waveguide mode and a third one
corresponding to the in-phase groove reemission, dependent on the period of the
groove array.
An important fact disregarded up to now in this brief historical introduction, is
that the theoretical analyses derived to explain extraordinary transmission through
hole arrays or through small apertures on corrugated planes showed high trans-
mittance peaks even assuming a perfect electric conductor instead of a Drude model
for the metallic plane. This suggested that the phenomenon was far more general
and that it could be reproduced in other frequency regimes where genuine SPPs do
not exist, like microwaves, millimeter-waves or even the terahertz range in which
metals admit a high conductivity model. The first demonstration of EOT through
hole arrays at millimeter-waves was published in 2004 [17] showing definitely that
the phenomenon was universal and it could appear in any band of the
4 Flat Corrugated and Bull’s-Eye Antennas 115

electromagnetic spectrum, regardless of the exact metal model employed. The key
to get a high transmittance peak is to design a coarse hole matrix, ensuring that the
cutoff frequency of the apertures falls above the onset of the first diffraction order. If
this condition is fulfilled and the metal plate is thin enough so as to avoid a strong
decay of the power transmitted through the holes, the EOT peak will appear at a
frequency slightly below the onset of the first diffraction order. An experimental
study showing the importance of the number of holes in the EOT resonance was
reported soon after [18] putting into evidence that the strength of the transmission is
basically controlled by the number of periods in the array. A theoretical analysis of
the finite structure effects was presented in [19]. Further numerical and experi-
mental studies in [19, 20] proved that the enhancement and bandwidth also
depended on the hole diameter.
Regarding corrugated structures, enhanced transmission was demonstrated for a
single slit surrounded by two narrow and deep corrugations at both sides of the slab
[21]. Using a scaled version of the single slit surrounded by multiple corrugations
presented in [14], enhanced transmission and strong beaming was demonstrated at
microwaves [22]. Later on, a thorough experimental study showing the enhance-
ment with input and output corrugations was reported at millimeter-waves [23]. In
[24, 25], the microwave enhanced transmission and its angle-dependence was
experimentally studied for a subwavelength circular aperture surrounded by annular
concentric grooves. All these experiments are the precursors of the flat metallic
antennas that we will consider in the next section.
Finally, it has to be remarked that the corrugated structures have been presented
here as an evolution of extraordinary transmission, which is in the roots of the
antennas discussed in the chapter. However, an alternative point of view is to
employ the leaky-wave (LW) formalism to explain the behavior of these structures.
Recently, in [26], a thorough study of the beam-forming physical mechanism of flat
horns was presented. There, the regimes of the surface waves and the leaky waves
are clearly distinguished and the behavior of corrugated flat horn antennas is
meticulously analyzed and depicted by means of representations of near- and
far-fields at each regime. A LW analysis will be presented later in this chapter, for
the sake of completeness.

4.2 A New Antenna Family: Small Apertures

on Corrugated Metallic Planes

4.2.1 First Generation: Antennas at Microwaves

Up to now, all the structures discussed are similar to frequency selective surfaces or
spatial filters, in the sense that they act on an incident wave and produce a response
at the output. However, the beaming property of corrugated structures makes them
interesting candidates for high-gain antennas with the advantage of having a much
116 M. Beruete et al.

reduced volume and weight than conventional solutions. The capability of pre-
senting beaming characteristics comparable to horn or parabolic antennas, among
others, has made them a field of intensive research for more than a decade. Indeed,
the results obtained in [22, 23] served as a basis for a new family of antennas based
on a small aperture on a corrugated metallic plane. These antennas differ from the
previous structures in that power is injected by a waveguide attached to the back of
a metallic plane, and is coupled to the output through a resonant slot instead of a
narrow slit.
In [27] the first member of this family was presented: a metallic Bull’s-Eye
antenna characterized by a central radiating slot surrounded by annular corrugations
and optimized to a wavelength approximately equal to the period of the grooves,
see Fig. 4.2a. A relatively high gain beam of approximately 21 dB was measured at
the operation frequency of 16.5 GHz, as shown in the gain curve of Fig. 4.2b. For
the comparison, an EMCO-3115 horn antenna was used (black curve). This
antenna, from now on called Test Antenna 1, is a 24.4 cm  15.9 cm
double-ridged waveguide horn which operates in the band from 750 MHz to
18 GHz. For all the band of interest (10–18 GHz) it presents an almost constant
gain of more than 12 dB. This corresponds to an aperture efficiency ea = 6.3%.
Comparing Fig. 4.2c and d it is evident that the antenna has different radiation
patterns in the principal planes. In the E-plane a narrow main lobe pointing at
broadside with well-defined nulls can be identified, whereas in the H-plane the
radiation pattern is smoother. This is a general characteristic of this kind of antennas
and can be explained by attending to the current distribution shown in Fig. 4.2e,
calculated with the commercial simulator CST Microwave Studio. The current
density is higher in vertical direction, which corresponds to E-plane, than in the
horizontal direction, which corresponds to the H-plane. In the E-plane (Fig. 4.2c)
the antenna presents an angular width of 6° and less than −15 dB side lobe level. In
the H-plane, by contrast, it presents a wider beamwidth at −3 dB of 12.3°,
(Fig. 4.2d). It has relatively low cross-polarization levels for E- and H-plane (30 dB
and 25 dB, respectively), as shown in Fig. 4.2c and d respectively. In Fig. 4.2f, a
three-dimensional plot of the radiation pattern has been included for the sake of
completeness.
This design opened the path to other schemes such as the antenna with 6 straight
grooves on each side depicted in Fig. 4.3a and published in [28]. The experimental
results showed a gain of 18 dB (ea = 6.5%), 6 dB higher than the Test Antenna 1,
see Fig. 4.3b. In this case, due to the absence of revolution symmetry, the antenna
presented a wider beam at −3 dB of 12° in the E-plane and 30° in the H-plane.
Besides, the prototype showed lower side lobes for a range of about ±30° at the
E-plane, see Fig. 4.3c. As it was pointed out before, the current density is dis-
tributed preferentially along the vertical (E-field) direction, which is the reason for
the larger beamwidth observed in the H-plane, Fig. 4.3d.
With the aim of reducing the size of the antenna and hence improve the aperture
efficiency, a miniaturized 1-D planar antenna operating at 17 GHz and consisting of
4 Flat Corrugated and Bull’s-Eye Antennas 117

Fig. 4.2 a Picture of the metallic Bull’s-Eye antenna. b Measured gain versus frequency for
prototype (blue curve) and test horn antenna (black curve). Measured co- (solid line) and
cross-polar (dashed line) E- and H-planes, (c) and (d), respectively. e Surface current distribution
on the Bull’s-Eye antenna surface. f Three-dimensional plot of the radiation pattern. Reprinted
with permission from IEEE

a resonant slot and two straight corrugations, was also experimentally and
numerically studied in [29], see picture in Fig. 4.4a. A relatively high gain of near
10 dB (ea = 17.2%) was measured, a value slightly lower than that of the Test
Antenna 1, see Fig. 4.4b. The radiation pattern showed a –3 dB angular width at of
34° in the E-plane and 60° in the H-plane, Fig. 4.4c and d respectively. Attending
to the characteristics of the prototype it can be described as a weakly directive
antenna with broadside gain enhancement of 4.3 dB compared to the
non-corrugated metallic plane. Nevertheless, it must be taken into account that the
lower gain and wide beaming is balanced out by the reduced physical dimensions
compared to those of the horn antenna. It must be emphasized that the behavior of
this antenna is more similar to a parasite-element radiating structure than to a LW
antenna, where a periodic grating is needed. This explains the observed relatively
low gain, compared to the previous antenna with 6 grooves on each side.
A compact antenna design was discussed in [30]. By means of the use of the
slot’s transversal resonance, instead of the longitudinal resonance, the antenna
thickness can be reduced. This was applied to the antenna shown in the bottom
antenna of Fig. 4.5a. The thickness can be further reduced by filling the corruga-
tions with a dielectric material so that the grooves’ depth can be reduced. In the
antenna shown on the top of Fig. 4.5a the corrugations were filled with glass-fiber
118 M. Beruete et al.

Fig. 4.3 a Pictures of the metallic low profile 12 corrugations antenna. b Measured gain versus
frequency for the antenna prototype (blue curve) and test horn antenna (black curve). E-plane
(c) and H-plane (d) radiation pattern for the simulated prototype (dashed blue), measured
prototype (solid blue) and measured test horn antenna (solid black)

Fig. 4.4 a Pictures of the metallic miniaturized antenna. b Measured gain versus frequency for the
antenna prototype (blue curve) and test horn antenna (black curve). c E-plane and d H-plane
radiation patterns for simulated prototype (dashed blue curve) and measured prototype (solid blue
curve)
4 Flat Corrugated and Bull’s-Eye Antennas 119

Fig. 4.5 a Picture of the metallic miniaturized flat antennas: dielectric filled grooves (top) and
air-filled grooves (bottom). b Measured gain versus frequency for the dielectric filled grooves
prototype (blue curve) and test horn antenna (black curve). c E-plane and d H-plane radiation
patterns for the simulated dielectric filled grooves prototype (dashed blue curve) and measured
prototype (solid blue curve)

(with permittivity er = 4) to reduce the grooves’ depth by half. By applying these

modifications, a reduction by a factor of over 3 of the thickness compared with the
previous prototype was achieved. According to the experimental results of
Fig. 4.5b, there is a clear improvement in the radiation pattern with a gain of
11.3 dB (ea = 39%), more than 5 dB larger than a flat metallic plane and barely
1 dB less than the Test Antenna 1. Figure 4.5c and d show the E- and H-plane
radiation diagrams, respectively. The measured –3 dB angular width for E-plane is
30° and for H-plane is 52°.
A dual-band low-profile antenna was reported in [28], taking advantage of the
double k/2 resonance of the slot located at f1 = 13 GHz and f2 = 16.5 GHz. The
transversal resonance, which corresponds to f1 = 13 GHz is fixed by the slot width,
whereas the longitudinal resonance, given at f2 = 16.5 GHz, is fixed by the slot’s
depth. In this case, two sets of corrugations were disposed at the output surface,
each one separated from the slot a distance equal to nearly the wavelengths fixed by
the resonances, see Fig. 4.6a. The patterning of the surface led to a gain near 10 dB
at both frequencies of design (aperture efficiency over 23%), Fig. 4.6b, nearly the
same values obtained for the Test Antenna 1 at f1 and f2. The measured beamwidth
at –3 dB was 20° and 30° in the E-plane (Fig. 4.6c and d), and 50° and 40° in the
H-plane (Fig. 4.6e and f), at f1 and f2, respectively.
120 M. Beruete et al.

Fig. 4.6 a Picture of the metallic dual band flat antenna. b Measured gain versus frequency for the
dual band prototype (blue curve) and test horn antenna (black curve). Simulated (dashed curves)
and measured (solid curve) E-plane at c f1 = 13 GHz and d f2 = 16.5 GHz. Idem for H-plane at
e f1 = 13 GHz and f f2 = 16.5 GHz

These antennas are included in the first generation of waveguide fed corrugated
antennas developed in the Public University of Navarre (UPNA), being a landmark
and design line for further coming structures. Figure 4.7 gathers all these antennas,
characterized by their low profile and operating at low frequencies in the microwave
range, and makes a comparison of the observed gain for each of them with a
standard Horn antenna.

4.2.2 Towards Terahertz Frequencies

One-dimensional (1-D, i.e., slit + grooves) and two-dimensional (2-D, i.e.,

Bull’s-Eye) antennas based on a radiating slot surrounded by corrugations were
designed and experimentally measured in the terahertz range in [31]. First, the
feasibility to develop antennas and the enhancement in the transmission at this
range of frequencies was proved, with the prototype shown in Fig. 4.8a. Two
structures with 10 corrugations working at f = 0.56 THz were designed and ana-
lyzed, differing only in the corrugation shape: triangular and rectangular.
Simulations showed a gain enhancement compared to a flat plane of 10 dB for both
antennas with a narrow beamwidth of h–3dB = 5.8° and 6.8° and side lobe levels of
8.6 and 9.7 dB in the E-plane for the triangular and square grooves, respectively.
Due to fabrication restraints, only the antenna with triangular grooves was fabri-
cated and experimentally characterized. As expected, the structure showed large
transmission enhancement at the working frequency, fixed by the period of the
4 Flat Corrugated and Bull’s-Eye Antennas 121

Fig. 4.7 Compendium of the first generation of low profile leaky wave antennas operating at
microwave frequencies

corrugations and the transversal resonance of the slot. A secondary peak at 0.8 THz
and with much lower amplitude was noticed in the spectrum, corresponding to the
longitudinal slot resonance. However, it was clear that the corrugations, optimized
for the design frequency of 0.56 THz, had little effect at 0.8 THz, Fig. 4.8b. In the
same article a Bull’s-Eye antenna was also experimentally analyzed, see Fig. 4.9a.
In this case, the time-domain waveforms for different sets of concentric annular
corrugations with variable depths were recorded, see Fig. 4.9b. It was first con-
sidered the electric field at the input face of the aperture, arising from the directly
incident THz beam, which evanescently tunneled through the subwavelength
aperture. Then, from these measurements, it was clear that the beaming mechanism
could be explained as the sum of the directly transmitted and
time-delayed-groove-contributions at the output face, Fig. 4.9c.
Even though there is a vast research on LW antennas, the THz range is still a
relatively unexplored field concerning to leaky corrugated structures, with few
designs as the one presented above or the one in [32] where a 560 lm  280 lm
terahertz low-profile pyramidal horn antenna was loaded with parallel straight
V-grooves of different depth, see a picture in Fig. 4.10a and b. The corrugated
structure presented at 0.49 THz a narrower beam, (Fig. 4.10c), enhancing its gain
more than 3 dB compared to the non-loaded horn (15.6 dB).
122 M. Beruete et al.

Fig. 4.8 a Picture of the metallic low profile antenna operating at 0.56 THz. b THz Bull’s-Eye
antenna. c Schematic of output beaming forming

Fig. 4.9 a Picture of the metallic low profile antenna operating at 0.56 THz. b THz Bull’s-Eye
antenna. c Schematic of output beaming forming

Fig. 4.10 a Micromachined V-grooves antenna. b Close view of manufactured antenna.

c Normalized E- and H-plane for patterned antenna (black and blue dotted curves) and horn
antenna (red and pink dashed curves). Reprinted with permission from John Wiley & Sons, Ltd

A typical drawback of quantum cascade lasers is their large beam divergence due
to the small emission aperture, Fig. 4.11c. In [33–36] this problem was overcome
by integrating periodic subwavelength corrugations onto the output facet of semi-
conductor lasers, Fig. 4.11a. With this solution, the beam divergence could be
highly reduced, Fig. 4.11d. This structure, named laser collimator or laser beam
4 Flat Corrugated and Bull’s-Eye Antennas 123

Fig. 4.11 a Detail of a 1D beam collimator. b 2D (Bull’s-Eye) beam collimator (Inset detail of the
structure). c Measured 2D far-field intensity distributions of the original unpatterned device.
d Measured 2D far-field intensity distributions of the device patterned with a 1D slit-grating
structure. Reprinted with permission from John Wiley & Sons, Ltd

shaper, achieves the reduction of divergence from *180° to *10°, an over 10 dB

directivity improvement and an increase factor of about six for the power collection
efficiency. It was also proved that using a concentric ring grating rather than a 1D
grating, Fig. 4.11b, the beam divergence was reduced not only for the laser
polarization direction, but also for the parallel one. As it happens with the previ-
ously mentioned antennas, the collimator design can be scaled to work from the
visible to the far-IR regimes.
124 M. Beruete et al.

Fig. 4.12 a Photograph of

THz Bessel beamformer.
b Detail of the central slot in
the beamformer. c Schematic
of the Bessel beam generation
mechanism. Reprinted with
permission from AIP
Publishing LLC

In [37] it was shown the usefulness of Bull’s-Eye corrugated structures as Bessel

beam generators embeddable in solid-state terahertz sources, in contrast to con-
ventional axicon lenses approach typically used for Bessel beam generation which
are usually bulky systems. In that work, a Bull’s-Eye operating at 0.29 THz was
fabricated, Fig. 4.12a and b, and characterized through terahertz time-domain
spectrometry, observing a non-diffractive behavior within a distance greater than
20k. In Fig. 4.12c it is shown how the beam generation mechanism is identical to
that observed for the antennas discussed above. A much more exhaustive review of
the Bessel beam sources as well as the physical mechanism involved in the beam
generation is carried out in Chap. 8.

4.2.3 Latest Developments

Nearly a decade after the first Bull’s-Eye paper, an antenna of this type was pro-
posed as a good solution to embed in CubeSat devices (miniaturized satellites for
space research composed of multiple small cubic units) taking advantage of the
antenna’s low profile [38]. In this publication, the effect of shifting the center of
each of the 7 annular corrugations was studied, observing a variation of the radiated
beam direction. For the concentric corrugations case, the antennas displayed a gain
of 19.6 dB, whereas it was reduced as the center was shifted and the radiated beam
gradually pointed away from broadside, although the beamwidth barely varied. We
can see how, once again, this kind of antennas surpasses in the field of space
4 Flat Corrugated and Bull’s-Eye Antennas 125

Fig. 4.13 Schematic of patch over a corrugation loaded substrate

communications (among others) classical parabolic antennas, providing high gain,

unconventional beaming properties and lightweight, which results of high interest
in this area.
Another example of the application of a Bull’s-Eye geometry in a metallic plane
is that reported in [39], in which a patch over a substrate was placed on top of a
metallic annular corrugated structure, increasing strongly the directivity, as shown
schematically in Fig. 4.13. An extra annular soft-surface configuration placed after
the Bull’s-Eye served to reduce the backward radiation, preventing the diffraction
of the surface wave on the edge of the structure. This diffraction produces a
degradation of the antenna pattern and a decrease of the antenna efficiency. This is
due to the nearly infinite surface impedance observed in the soft surface and the null
value of all components of the electric field in any polarization direction. This
antenna presented a gain of 19.45 dB, more than 12 dB larger gain than the
non-patterned ground + patch, and a narrow beamwidth of 7.9°.
In [40] a Bull’s-Eye antenna with sinusoidal corrugations working at millimeter
waves was presented, see picture in Fig. 4.14a. In this case, the aim of the study
was the comparison between a sinusoidal pattern and the previously analyzed
rectangular groove antennas. Similar results were obtained for a high number of
periods, with the added value of a soft surface with no corners in the case of the
sinusoidal profile, which could reduce multipactor effects, interesting for applica-
tions where medium/high power handling is needed. The manufactured 20 period
antenna presented a gain of 28.9 dB (Fig. 4.14b) with a –24 dB side lobe level and
a very narrow 1.2° beam in the E-plane as shown in Fig. 4.14c. The H-plane is
shown in Fig. 4.14d for completeness.
In [41–44] two Bull’s-Eye antennas in planar technology were thoroughly
analyzed. These antennas consisted of a series of annular concentric metallic strip
gratings placed over a grounded dielectric slab and fed by a surface wave launcher
(SWL) (scheme shown in Fig. 4.15) or an aperture coupled slot (ACS). There, the
126 M. Beruete et al.

Fig. 4.14 a Picture of the metallic sinusoidal grooves Bull’s-Eye antenna. b Simulated (red
curve) and measured (blue curve) gain versus frequency. Simulated (red) and measured (blue)
E- and H-planes, (c) and (d) respectively

Fig. 4.15 Schematic of a

metallic strip grating over a
grounded dielectric slab fed
by a SWL antenna

performance of different geometries as well as suitable optimizations were pro-

posed. For the Bull’s-Eye fed by SWL, a beam of near 15 dB and approximately
10° beamwidth was measured at 19.8 GHz.
Closer in time, an off-axis beaming 3D-printed antenna operating at 96 GHz was
proposed [45]. This structure was manufactured by means of SLA (stereolithog-
raphy) and then metal coated by a several step chemical and plating process,
Fig. 4.16a. The antenna presented 7 annular off-center corrugations which per-
mitted a tilted beaming pointing towards 16.5° and a 17 dB peak gain, with −10 dB
side lobe level. Figure 4.16b displays the measured gain vs. frequency for E-plane,
where the tilted beaming at 96 GHz is observed. This tilted beaming is possible due
to the in-phase radiation of upper and lower halves of the structure at the same
angle. This structure is a clear example of where the interest is veering to: corru-
gated leaky wave antennas were initially proposed as an alternative solution to
classical high-volume antennas as horns, since they offered better radiation
4 Flat Corrugated and Bull’s-Eye Antennas 127

Fig. 4.16 a 3D-printed off-axis beaming Bull’s-Eye antenna. b Measured realized gain versus
frequency in dB for E-plane

characteristics and, at the same time, presented lower profiles and lighter weights.
The following natural step is to reduce the weight of this new family of antennas.
Evidently, the metallization of plastic structures (fabrication method that also
facilitates manufacturing of complex structures) is a very interesting solution which,
sooner or later, will surely become a hot research topic. This result demonstrates
that this type of antennas can be produced employing additive manufacturing which
is nowadays envisioned as the fabrication technology of the future and a major actor
of the fourth industrial revolution.
The following table sums up some of the radiation characteristics at E-plane for
all the above mentioned antennas (Table 4.1).
A quick glance at the table enables us to verify that the best performance in
terms of gain is obtained, among the presented structures, for the sinusoidal cor-
rugations antenna. Nevertheless, note that in this case, it is also the structure with
the highest number of periods (N = 20 periods) which leads us to a low aperture
efficiency antenna, typical of corrugated leaky wave antennas. Meanwhile, the
Patch + Bull’s-Eye antenna, while showing high gain and relatively low side lobe
level, displays a wider beam compared to that presented by the sinusoidally
modulated antenna. Thus, it is easy to understand that there is no ideal design which
matches any requirement, and it is highly important to carry out a preliminary study
so as to choose the design that correctly fits our needs.

4.3 Classical Leaky-Wave Antennas

(LWA) and the Problem of Broadside Radiation

As mentioned in the Introduction, corrugated structures can be analyzed using the

LW formalism [46]. Indeed Jackson et al. argued in [47], that the theories based on
extraordinary transmission could be complemented with the LW theory to explain
the mechanism of narrow beaming in corrugated structures [47, 48]. In short, they
argued that when a metallic surface is loaded with a periodic array of grooves with
periodicity larger than the wavelength and the operating frequency is raised enough,
128 M. Beruete et al.

Table 4.1 Flat leaky wave antennas overview table

Antenna Frequency Gain Side lobe Beamwidth Physical aperture
(GHz) (dB) level (dB) (°) (normalized to k2)
Low profile 16.5 21 –15 6 158
Bull’s-Eye [27]
12 straight grooves 16.5 18 –10 12 48
[28]
Miniaturized [29] 17 10 –8 34 4.62
Dielectric loaded 13 11.3 –9 30 4.62
grooves [30]
Dual-band [28] 13/16.5 10.4/9 –9/–9 20/30 4.62
Wedge corrugations 560 15.88 –8.6 5.8 94
[31]
Micromachined 490 15.6 *–12 18.5 223
V-grooves [32]
CubeSat Bull’s-Eye 60 19.6 –14 *10 400
[38]
Patch + Bull’s-Eye 12.64 19.45 –15 7.9 17.75
[39]
Sinusoidal corrugated 77 28.9 –24 1.2 1370
Bull’s-Eye [40]
Surface wave 19.8 *15 *15 *10 126
launcher Bull’s-Eye
[42]
Tilted beaming 96 17 –10 3.5 5
3D-printed
antenna [45]

a leaky plasmon mode is achieved, which radiates as it propagates along the

periodic structure [46, 49].
A leaky wave antenna (LWA) is a member of the traveling-wave antennas
family [50]. It is a waveguiding structure with a mechanism able to leak power
along its length [51]. Due to this leakage, the waveguide presents a complex
propagation wavenumber consisting in a phase constant b, which controls the
beaming angle, and a leakage constant a. A large or small a corresponds to a large
or low leakage affecting the size of the effective aperture of the antenna, which
results in a wide or narrow beamwidth, respectively. This kind of antennas present a
LW that attenuates in the propagating direction (a [ 0) and an amplitude that either
increases transversely away from the guiding surface (at \0), forward leaky wave,
or decreases (at [ 0), backward leaky wave. For this reason, forward LWs have
been called improper waves and backward LWs proper waves. Anyway, in the
forward case, as the LW only exists in a portion of space in the direction normal to
the structure, the amplitude increases up to a point and then decreases [52].
Furthermore, due to the fact that the phase constant b varies with frequency, the
beam direction can be changed sweeping the frequency, as discussed in [53, 54].
4 Flat Corrugated and Bull’s-Eye Antennas 129

Fig. 4.17 a Dispersion diagram for non-corrugated metallic surface. b Dispersion diagram for
periodically corrugated metallic surface. c Fast modes within the radiation region of the dispersion
diagram for a corrugated metallic surface

The width of the emerging beam is also controllable by the configuration of the
antenna. Thus, for a given beam direction h and knowing L, the angular width is:
Dh ¼ k0 =ðL cos hÞ. The easiest and earliest example of this kind of structures is a
simple rectangular waveguide with a continuous slit cut along its side, presented by
W.W. Hansen in 1940 [55].
Depending on the geometry of the guiding structure two kinds of LWAs are
distinguished: uniform and periodic LWAs. Although the literature surrounding
LWAs is extensive, only the structures that serve to the purpose of this chapter will
be reviewed here. Thus, more emphasis will be made on periodic antennas:
• Uniform LWAs: in these type of antennas, no periodic modulation is present
along the length of the guiding structure. In uniform structures the dominant
mode, the space harmonic n ¼ 0, is a fast wave that radiates whenever the
structure is open. This kind of antennas can also be catalogued in two groups:
1-D and 2-D LWAs.
• Periodic LWAs: are loaded with a periodic structure along the guiding struc-
ture’s length. In comparison with uniform antennas, periodic LWAs possess a
dominant slow-wave mode. The phase velocity of this mode is greater than the
velocity of light. This implies that even though the structure is open, it does not
radiate. This is why it needs a periodic array to radiate. Although these antennas
can scan from backward endfire into part of the forward quadrant, there is a
narrow region around broadside where radiation is not allowed due to an open
stop band. This issue will be explained thoroughly further in this section.
Earliest structures of this kind of antennas consisted in the periodically modu-
lation of a dielectric waveguide like the one presented by King [56]. A typical
example is that analyzed in [57–60], which basically consists in a dielectric
rectangular rod loaded with periodically distributed metal strips. In another
geometry, the metallic strips were replaced by grooves [61].
It is easy to see the resemblance between the periodic LWAs and the antennas
previously discussed, in particular [22, 23, 28–31, 40, 62, 63]. All of these antennas
consisted of a central radiating slot perforated on a metallic slab surrounded by
periodically distributed corrugations with periodicity d. For a non-corrugated flat
slab, the dispersion curve is like that displayed in Fig. 4.17a. As all these antennas
are made of good electric conductors, we can approximate b0  k, where b0 is the
130 M. Beruete et al.

phase constant of the n = 0 space harmonic (or fundamental mode), where n is the
index of the space harmonic, and k = 2p/k is the free-space wavenumber. When a
metallic slab is properly modulated, for example, by periodically disposing corru-
gations, a propagating surface wave is generated.
Thus, the dispersion diagram shows two straight lines at ±45°. The corrugation
patterning permits the coupling of the power emanating from the slot to a TMz
mode surface wave that propagates away from the aperture. The presence of the
grating produces an infinite number of space harmonics (also called Floquet
modes), Fig. 4.17b, that depending on the frequency, will be fast or slow. All these
space harmonics are related to each other by:

bn d ¼ b0 d þ 2pn ð4:1Þ

where bn is the phase constant of the n-th space harmonic and b0 is the phase
constant of the n = 0 space harmonic, corresponding to the travelling wave sup-
ported by a non-corrugated flat metallic plane (valid only for weak modulations).
As stated in [54, 64], the beam direction of LWAs can be swept varying the
frequency, due to the fact that the phase constant changes as frequency is swept. In
general, and when the surface modulation is weak, it is possible to guess the beam
direction hn of the n-th radiated space harmonic following the next equation:

bn b0 2pn
sinðhn Þ ¼ ¼ þ ð4:2Þ
k k kd

where k = 2p/k the free-space wavenumber. If the wave propagates along a

metal-air interface we can approximate k0  k. Furthermore, for a practical
antenna, it is desired only the n = –1 space harmonic to radiate inside the radiation
region, Fig. 4.17c. The radiation region is the area of the dispersion diagram for
which kd > bd, i.e., the modes possess a phase velocity larger than the light. So,
recalling that b0  k = 2p/k and the relation between b0 and bn, for the mode
n = −1, Eq. (4.2) becomes:

k
sinðh1 Þ  1  ð4:3Þ
d

Thus, for a beam pointing at broadside, h−1 = 0°, d must be chosen equal to k.
Nevertheless, as discussed in [49], pure broadside radiation is not possible.
Radiation at h−1 = 0° would imply b−1 = 0, which would also imply that the leaky
mode becomes a standing wave, as attenuation constant drops to zero, resulting in
an open stop band in which radiated power sharply decreases [49, 64].
Already in [65], it was attempted to overcome the stop-band inconvenient pre-
sent in most periodic LWAs. It was foreseen that for a structure like those in [66,
67], basically a dielectric rod with a single metallic strip grating, the stop-band
could be highly reduced if a double-strip was used per unit cell. Later on, methods
4 Flat Corrugated and Bull’s-Eye Antennas 131

Fig. 4.18 Radiated beams

for a frequencies f < f0,
b frequencies f * f0 and
c frequencies f > f0

to optimize broadside radiation in uniform [68] and periodic [69] LWAs were
discussed.
As explained in [49], the open stop band effect in periodic LWAs is inevitable.
But in the antennas analyzed in the previous section, this limitation was circum-
vented by choosing a proper b−1 so that the beams corresponding to each half of the
structure point towards a scan angle slightly diverted from h−1 = 0°. The maximum
of radiation appears just before or immediately after the null point, depending on
the wavelength-period relation. With regard to the radiation diagram, at the opti-
mized frequency, opposed beams will be so close that they will appear to merge as a
single broadside beam. For a certain range around that frequency, the emitted beam
will appear as a less directive wider beam and will ultimately be detached into two
different beams, corresponding to each half.
This characteristic was experimentally studied in [40] and in more detail in [70],
where the behavior of the radiated beams in the vicinity of the operation frequency
f0 was analyzed. For frequencies f < f0, the leaky-mode corresponded to a
backward-wave. Given the symmetry of the antenna, the upper and lower halves
radiated in the opposite half: the upper-half in the lower-half plane direction and
vice versa, Fig. 4.18a. At f * f0, an apparently single beam was obtained,
Fig. 4.18b. Then, for f > f0, a forward leaky wave is supported by each half and the
beams split again in two, radiating each half of the structure in its own half,
Fig. 4.18c. Both beams diverged as frequency was shifted towards higher
frequencies.
132 M. Beruete et al.

4.4 Design Tips and Guidelines

In this section it is intended to offer a few tips and basic guidelines to design flat
leaky wave corrugated antennas. Two types of geometries will be considered:
straight parallel and annular corrugations. For both cases, the possibility of
designing a tilted beam radiating structure will be shown. An intuitive procedure to
design a leaky-wave flat corrugated antenna was presented in [27–30, 40]. Here, we
summarize the main tips and guidelines that must be followed for the design of this
type of antennas.
The first goal that must be accomplished is ensuring a good matching between
the input waveguide and the antenna. This is usually done by designing a central
resonant slot, with a relatively high-quality factor. Depending on the needs, the
resonant aperture can be designed to resonate at one or two frequencies. The
transversal resonance is governed by the slot width, sx * k0/2, whereas the lon-
gitudinal resonance is given by the slot depth (or metal plate thickness) sz * k0/2,
although in this case the approximation is rougher since the resonance is due to an
open cavity, Fig. 4.19a. In either case, the height of the slot must be relatively small
sy  k0. This ensures a high-quality factor and hence a good matching which is of
interest in this type of antennas. Although a high-quality factor implies a reduced
bandwidth, in this case this is not a strong limitation, since the periodic structure
automatically imposes a narrow bandwidth for broadside radiation. For dual band
operation, the chosen frequencies must be relatively far apart. In fact, as stated in
[71], when both sx and sz tend to k0/2, longitudinal and transversal resonances might
be affected and differ from their isolated behaviour. This is due to the necessary
fulfillment of Foster’s Theorem, which forces a zero to exist in between two poles
(resonances) in the frequency response of passive structures, such as slots (iden-
tifiable as an open Fabry-Perot cavity) [72]. This means that there cannot be two

Fig. 4.19 a Detail of slot. b Front view detail of straight parallel corrugations. c Front view detail
of asymmetric straight parallel corrugations
4 Flat Corrugated and Bull’s-Eye Antennas 133

identical resonances at the same frequency in the same cavity. Thus, in order to
design an antenna with dual frequency operation, sz must be slightly larger than k0.
Designing the grating is probably the most complex issue and some consider-
ations must be taken into account. For a given periodicity d, there is a certain
frequency range for which only the n = −1 mode radiates and a narrower range for
which the beam points close to broadside. Thus, from (4.3) it is easily obtained that
for broadside radiation d  k0, recalling that beaming at exactly h−1 = 0° is not
possible. Regarding the optimal number of periods, it was pointed out in [28] that
for approximately n = 12 grooves (6 at each side of the slot) a “gain saturation”
point is reached. So, n = 12 can be chosen as a first guess to obtain a large
directivity with a reasonable side length.
Thus, on the basis that the maximum directivity is obtained at broadside radi-
ation for periods d  k0, the metallic plate should have a side length of at least
Ly = (n + 2)k0 in the direction of the currents, i.e., the E-plane. As a first guess, a
width of the metallic plate of Lx = 7sx (with slot width sx * k0/2) has been used in
previous works, although the final width is fixed further on, as a part of the opti-
mization of the structure. The slab thickness Lz usually coincides with the slot
depth, sz.
For a broadside beaming structure, the first groove must be placed at an offset
distance d0 from the central slot. The rest of the grooves are periodically distributed
following the first one with period d, Fig. 4.19b. The offset distance d0 for the first
pair of opposite corrugations must be such that the intense reactive field near the
slot discontinuity does not affect the excited LW. In addition, for broadside radi-
ation this distance must be such that all the corrugations radiate in-phase.
A convenient way to optimize the offset distance is to place the first pair of grooves
almost touching the slot and then sweep the separation with a numerical simulator,
monitoring the return losses response, until matching is obtained.
For off-axis beaming, i.e., beam pointing in a direction other than broadside, the
groove distribution is a little bit trickier, as corrugations must be asymmetrically
distributed at both sides of the slot, Fig. 4.19c. This implies that different offset
distances must be separately optimized for both halves. In this case, period dlong for
one half of the structure for a beam pointing to h−1 6¼ 0° is obtained using
Eq. (4.3). The period dshort for the other half must be calculated for a beam pointing
towards the opposite direction, i.e.,h1 . Thus, from Eq. (4.3) we extract two
equations:

k
dlong 
ð1  sinðh1 Þ
ð4:4Þ
k
dshort 
ð1 þ sinðh1 Þ

where dlong corresponds to the half structure which supports a forward leaky-wave
and dshort to the part which supports a backward leaky-wave. For these asymmetric
structures, the number of grooves is not equal and an optimization can help to
134 M. Beruete et al.

Fig. 4.20 a Front view detail of annular concentric corrugation. b Front view detail of off-center
annular corrugations

design the antenna with little effort. Normally, a larger number of grooves is needed
for the half-side which presents a lower period.
However, a better performance can be obtained for both broadside and off-axis
beaming antennas if annular corrugations are employed, rather than parallel straight
corrugations, Fig. 4.20a. This is due to the in-phase interaction between the
cylindrical wavefront launched by the slot and the annular grooves. For the
h−1  0° case, concentric annular corrugations surrounding the central slot must be
placed with a separation of d, obtained with (4.3). Period and offset values obtained
for the straight corrugations case can be taken in this configuration as a seed for a
new optimization.
If h1 6¼ 0 is desired, the equation governing elliptical corrugations, taking the
slot as the origin of coordinates, is as follows:

d0 þ nk
d ðnÞ ðuÞ ¼ ð4:5Þ
1  sin h1 sin u

where d ðnÞ ðuÞ is the distance from the slot to the n-th groove at angle u, contained
on the plane of the surface, d0 is the distance from the slot to the first groove at
u ¼ 0 (x axis) and h1 is the elevation angle at which the beam points, Fig. 4.20b.
It has been previously shown in Sect. 4.2 that the grooves’ design can vary
depending on the needs and the available resources. Here, for simplicity, reference
will only be made to the rectangular geometry. Normally, the depth of the grooves
is less than k=4 and the width is very small compared to the wavelength at the
operation frequency. An adequate way to obtain these values, is to carry out an
optimization where k=4 and k=8 are taken as seeds for the depth and the width,
respectively. It must be taken into account that, for the off-axis beaming case,
grooves’ dimensions differ at both halves. Thus, the optimization of the whole
4 Flat Corrugated and Bull’s-Eye Antennas 135

structure for this case becomes more complex than for the broadside (symmetric)
one, as more parameters must be included in the calculation. Anyway it was
recently shown, in the optical regime, that it was possible to harvest even more light
by means of a corrugated structure if wider grooves were used [73]. We have
recently applied this concept for a miniaturized antenna working at
millimeter-waves with a high aperture efficiency of 32% [74].
If any possible pitfall should be pointed out, it would be the need of an opti-
mization routine for each modification introduced in the structure. For example, the
sole introduction of an extra corrugation in an optimized N-period 1D structure,
would require a review of the corrugations’ parameters, so as to recover the opti-
mum throughput, as well as the resizing of the slot due to a possible
frequency-shifting of the cavity resonance.
Regardless of the potential need of several optimizations to obtain an efficient
structure, following these few tips it is possible to design a basic corrugated leaky
wave antenna whose performance can be fine-tuned by successive numerical
optimizations. It is not necessary to remark the scalability of this kind of structures,
as its presence in different frequency regimes has already been shown in Sect. 4.2.
However, it is noteworthy that the design of the corrugated surfaces at any fre-
quency, from microwaves to the optical regime, follows the same tips above
mentioned and only the need of different sources and the proper metallic modelling
at each frequency must be separately addressed.

4.5 Discussion and Future Trends

Leaky wave antennas came to light several decades ago. Despite being a classical
subject, novel antennas appear from time to time, based on different topologies and
coming up with new designs which merge design characteristics specific to other
antennas, achieving unparalleled throughputs taking into account their low profile.
Among them, corrugated antennas stand out in a privileged position due to their
interesting radiation performance, showing higher gains and narrower beams
compared to those emitted by standard horn antennas. This has led to the devel-
opment of several geometries and different capabilities, which is evidenced with the
wide range of antennas shown in Sect. 4.2. The large diversity of these antennas
and the possibility of scaling them in frequency, allow their use in a large variety of
applications, overcoming the challenge of having antennas which presented good
radiation characteristics and reduced volume and weight (essential for certain
applications, as space communications).
Although corrugated LWAs are also known for presenting a relatively low
aperture efficiency, they compensate this issue by a large gain comparable or even
higher than that presented in larger volume antennas, such as horn antennas.
Anyway, latest designs and research on LWAs are focused in attaining lower side
lobe levels and exploiting in a better way the surface of the corrugated structure. It
136 M. Beruete et al.

is also of high interest the development of antennas which achieve pure radiation at
broadside, and several solutions have been proposed using different technologies.
Modern 3D printing technologies can also be applied to corrugated antennas,
giving rise to lighter structures. In that respect, in the recent literature are found
purely plastic structures with a thin metallic coating, whose configuration allows the
radiation of beams at desired angles. This off-axis beaming had been previously
achieved [75–78], but the possibility of fast manufacturing, low price, light weight
and easiness of its design are added values that stress the interest of 3D-printed
corrugated antennas.
Finally, the interest of this type of structures goes beyond simple corrugated
antennas oriented towards communication applications. In recent times, similar
structures been proposed in the field of renewable energies, as heat harvesting
devices for the photosensitive cells arranged in photovoltaic panels [79, 80].
Micro lenses play a major role in the nowadays thriving sector of the minia-
turized optics, with applications in fiber coupling, optical switching, laser colli-
mation, imaging… The joint use of lenses and corrugated structures may result in
sensors with increased responsivity and noise decrease [81–83].
Corrugated structures have also found application in imaging in the visible and
infrared spectrum. Gratings have been incorporated to multicolor sensing devices or
colour filters, replacing typically used filters and glass prisms, which were used to
disperse light, and taking advantage of their capability of selective conversion
between free-space waves and spatially confined modes [84].

4.6 Conclusions

In this chapter the high interest of corrugated antennas has been shown through the
analysis of a vast number of prototypes and applications over a large range of the
electromagnetic spectrum. It has been shown the evolution of these antennas since
the extraordinary optical transmission was replicated in the microwave and mil-
limeter ranges until the latest designs at THz frequencies.
A brief analysis of the leaky mechanism has been presented, the problem that
these structures encounter when broadside beaming is needed and the existing
solutions to bypass it, as the pointing of opposed radiating beams to a direction
close to broadside. Furthermore, a basic guideline to design broadside and off-axis
beaming corrugated antennas has been proposed for both 1D and 2D geometries.
Applications which go far beyond the field of the communications have been
introduced, as efficient frequency selective devices found in the range of optics or
heat harvesting devices in the field of renewable energies, becoming manifest the
large number of possibilities which are on the fringes of the antennas field.
As the drive for miniaturization increases, more efficient and power saving
systems will be demanded in fields like optical computing, imaging, sensing and,
without a doubt, telecommunication. Thus, as long as the manufacturing and
handling processes allow it, patterned surfaces will surely find application in the
4 Flat Corrugated and Bull’s-Eye Antennas 137

upcoming technological developments. We have already demonstrated that these

antennas are compatible with modern additive manufacturing techniques even at
relatively high frequencies. These facts allows us to foresee a brilliant future for
planar corrugated antennas all along the electromagnetic spectrum.

References

1. H.A. Bethe, Theory of diffraction by small holes. Phys. Rev. 66, 163–182 (1944). doi:10.
1103/PhysRev.66.163
2. C.J. Bouwkamp, Diffraction theory. Rep. Prog. Phys. 17, 35–100 (1954). doi:10.1088/0034-
4885/17/1/302
3. T.W. Ebbesen, H.J. Lezec, H.F. Ghaemi, T. Thio, P.A. Wolff, Extraordinary optical
transmission through sub-wavelength hole arrays. Nature 391, 667–669 (1998)
4. W.L. Barnes, A. Dereux, T.W. Ebbesen, Surface plasmon subwavelength optics. Nature 424,
824–830 (2003). doi:10.1038/nature01937
5. Surface plasmon resurrection, Nat. Photon. 6, 707 (2012)
6. F.J. García de Abajo, Colloquium: light scattering by particle and hole arrays. Rev. Mod.
Phys. 79, 1267–1290 (2007). doi:10.1103/RevModPhys.79.1267
7. F.J. Garcia-Vidal, L. Martin-Moreno, T.W. Ebbesen, L. Kuipers, Light passing through
subwavelength apertures. Rev. Mod. Phys. 82, 729–787 (2010). doi:10.1103/RevModPhys.
82.729
8. L. Martín-Moreno, F.J. García-Vidal, H.J. Lezec, K.M. Pellerin, T. Thio, J.B. Pendry, T.W.
Ebbesen, Theory of extraordinary optical transmission through subwavelength hole arrays.
Phys. Rev. Lett. 86, 1114–1117 (2001). doi:10.1103/PhysRevLett.86.1114
9. J.A. Porto, F.J. Garcia-Vidal, J.B. Pendry, Transmission resonances on metallic gratings with
very narrow slits. Phys. Rev. Lett. 83, 2845–2848 (1999). doi:10.1103/PhysRevLett.83.2845
10. D.E. Grupp, H.J. Lezec, T. Thio, T.W. Ebbesen, Beyond the Bethe limit: tunable enhanced
light transmission through a single sub-wavelength aperture. Adv. Mater. 11, 860–862 (1999).
doi:10.1002/(SICI)1521-4095(199907)11:10<860::AID-ADMA860>3.0.CO;2-V
11. W.C. Chen, N.I. Landy, K. Kempa, W.J. Padilla, A subwavelength extraordinary-
optical-transmission channel in babinet metamaterials. Adv. Opt. Mater. 1, 195–195
(2013). doi:10.1002/adom.201370017
12. F. Falcone, T. Lopetegi, M.A.G. Laso, J.D. Baena, J. Bonache, M. Beruete, R. Marqués, F.
Martín, M. Sorolla, Babinet principle applied to the design of metasurfaces and metamaterials.
Phys. Rev. Lett. 93, 197401 (2004). doi:10.1103/PhysRevLett.93.197401
13. T. Zentgraf, T.P. Meyrath, A. Seidel, S. Kaiser, H. Giessen, C. Rockstuhl, F. Lederer,
Babinet’s principle for optical frequency metamaterials and nanoantennas. Phys. Rev.
B Condens. Matter Mater. Phys. 76, 4–7 (2007). doi:10.1103/PhysRevB.76.033407
14. H.J. Lezec, A. Degiron, E. Devaux, R.A. Linke, L. Martin-Moreno, F.J. Garcia-Vidal, T.W.
Ebbesen, Beaming light from a subwavelength aperture. Science 297, 820–822 (2002).
doi:10.1126/science.1071895
15. L. Martín-Moreno, F.J. García-Vidal, H.J. Lezec, A. Degiron, T.W. Ebbesen, Theory of
highly directional emission from a single subwavelength aperture surrounded by surface
corrugations. Phys. Rev. Lett. 90, 167401 (2003). doi:10.1103/PhysRevLett.90.167401
16. F.J. García-Vidal, H.J. Lezec, T.W. Ebbesen, L. Martín-Moreno, Multiple paths to enhance
optical transmission through a single subwavelength slit. Phys. Rev. Lett. 90, 213901 (2003).
doi:10.1103/PhysRevLett.90.213901
17. M. Beruete, M. Sorolla, I. Campillo, J.S. Dolado, L. Martín-Moreno, J. Bravo-Abad, F.
J. García-Vidal, Enhanced millimeter-wave transmission through subwavelength hole arrays.
Opt. Lett. 29, 2500–2502 (2004). doi:10.1364/OL.29.002500
138 M. Beruete et al.

18. M. Beruete, M. Sorolla, I. Campillo, J.S. Dolado, Increase of the transmission in cut-off
metallic hole arrays. IEEE Microw. Wirel. Compon. Lett. 15, 116–118 (2005). doi:10.1109/
LMWC.2004.842852
19. M. Beruete, M. Sorolla, I. Campillo, J.S. Dolado, L. Martín-Moreno, J. Bravo-Abad, F.
J. García-Vidal, Enhanced millimeter wave transmission through quasioptical subwavelength
perforated plates. IEEE Trans. Antennas Propag. 53, 1897–1903 (2005). doi:10.1109/TAP.
2005.848689
20. J. Bravo-Abad, A. Degiron, F. Przybilla, C. Genet, F.J. García-Vidal, L. Martín-Moreno, T.
W. Ebbesen, How light emerges from an illuminated array of subwavelength holes. Nat. Phys.
2, 120–123 (2006). doi:10.1038/nphys213
21. A.P. Hibbins, J.R. Sambles, C.R. Lawrence, Gratingless enhanced microwave transmission
through a subwavelength aperture in a thick metal plate. Appl. Phys. Lett. 81, 4661 (2002).
doi:10.1063/1.1527704
22. M. Beruete, I. Campillo, J.S. Dolado, J.E. Rodríguez-Seco, E. Perea, M. Sorolla, Enhanced
microwave transmission and beaming using a subwavelength slot in corrugated plate. IEEE
Antennas Wirel. Propag. Lett. 3, 328–331 (2004). doi:10.1109/LAWP.2004.839461
23. M. Beruete, M. Sorolla, I. Campillo, J.S. Dolado, Subwavelength slotted corrugated plate
with enhanced quasioptical millimeter wave transmission. IEEE Microw. Wirel. Compon.
Lett. 15, 286–288 (2005). doi:10.1109/LMWC.2005.845753
24. M.J. Lockyear, A.P. Hibbins, J.R. Sambles, C.R. Lawrence, Surface-topography-induced
enhanced transmission and directivity of microwave radiation through a subwavelength
circular metal aperture. Appl. Phys. Lett. 84, 2040–2042 (2004). doi:10.1063/1.1688001
25. M.J. Lockyear, A.P. Hibbins, J.R. Sambles, C.R. Lawrence, Enhanced microwave
transmission through a single subwavelength aperture surrounded by concentric grooves.
J. Opt. A Pure Appl. Opt. 7, S152–S158 (2005). doi:10.1088/1464-4258/7/2/020
26. S. Steshenko, M. Zhadobov, R. Sauleau, A.A. Kirilenko, A.V. Boriskin, Beam-forming
capabilities of waveguide feeds assisted by corrugated flanges. IEEE Trans. Antennas Propag.
63, 5548–5560 (2015). doi:10.1109/TAP.2015.2487990
27. M. Beruete, I. Campillo, J.S. Dolado, J.E. Rodríguez-Seco, E. Perea, F. Falcone, M. Sorolla,
Very low-profile “bull’s eye” feeder antenna. IEEE Antennas Wirel. Propag. Lett. 4, 365–368
(2005). doi:10.1109/LAWP.2005.851104
28. M. Beruete, I. Campillo, J.S. Dolado, E. Perea, F. Falcone, M. Sorolla, Dual-band low-profile
corrugated feeder antenna. IEEE Trans. Antennas Propag. 54, 340–350 (2006). doi:10.1109/
TAP.2005.863380
29. M. Beruete, I. Campillo, J.S. Dolado, J.E. Rodríguez-Seco, E. Perea, F. Falcone, M. Sorolla,
Low-profile corrugated feeder antenna. IEEE Antennas Wirel. Propag. Lett. 4, 378–380
(2005). doi:10.1109/LAWP.2005.857297
30. M. Beruete, I. Campillo, J.S. Dolado, J.E. Rodríguez-Seco, E. Perea, F. Falcone, M. Sorolla,
Very low profile and dielectric loaded feeder antenna. IEEE Antennas Wirel. Propag. Lett. 6,
544–548 (2007). doi:10.1109/LAWP.2007.909969
31. M. Beruete, U. Beaskoetxea, M. Zehar, A. Agrawal, S. Liu, K. Blary, A. Chahadih, X.-L.
Han, M. Navarro-Cía, D. Etayo, A. Nahata, T. Akalin, M. Sorolla, Terahertz corrugated and
bull’s-eye antennas. IEEE Trans. Terahertz Sci. Technol. 3, 740–747 (2013). doi:10.1109/
TTHZ.2013.2287096
32. H. Lu, X. Lv, Z. Gao, Y. Liu, Experimental radiation characteristics of micromachined
terahertz low-profile corrugated horn antenna. Microw. Opt. Technol. Lett. 57, 364–367
(2015). doi:10.1002/mop.28849
33. N. Yu, J. Fan, Q.J.Q.J. Wang, C. Pflügl, L. Diehl, T. Edamura, M. Yamanishi, H. Kan, F.
Capasso, Small-divergence semiconductor lasers by plasmonic collimation. Nat. Photonics 2,
564–570 (2008). doi:10.1038/nphoton.2008.152
34. N. Yu, Q.J. Wang, M.A. Kats, J.A. Fan, S.P. Khanna, L. Li, A.G. Davies, E.H. Linfield, F.
Capasso, Designer spoof surface plasmon structures collimate terahertz laser beams. Nat.
Mater. 9, 730–735 (2010). doi:10.1038/nmat2822
4 Flat Corrugated and Bull’s-Eye Antennas 139

35. N. Yu, R. Blanchard, J. Fan, Q.J. Wang, C. Pflügl, L. Diehl, T. Edamura, S. Furuta, M.
Yamanishi, H. Kan, F. Capasso, Plasmonics for laser beam shaping. IEEE Trans.
Nanotechnol. 9, 11–29 (2010). doi:10.1109/TNANO.2009.2029099
36. N. Yu, Q. Wang, F. Capasso, Beam engineering of quantum cascade lasers. Laser Photon.
Rev. 6, 24–46 (2012). doi:10.1002/lpor.201100019
37. Y. Monnai, D. Jahn, W. Withayachumnankul, M. Koch, H. Shinoda, Terahertz plasmonic
Bessel beamformer. Appl. Phys. Lett. 21101 (2015). doi:10.1063/1.4905445
38. C.J. Vourch, T.D. Drysdale, V-band “bull”s eye’ antenna for cubeSat applications. IEEE
Antennas Wirel. Propag. Lett. 13, 1092–1095 (2014). doi:10.1109/LAWP.2014.2327852
39. C. Huang, Z. Zhao, X. Luo, Application of “bull’s eye” corrugated grooves integrated with
artificially soft surfaces structure in the patch antenna to improve radiation performance.
Microw. Opt. Technol. Lett. 51, 1676–1679 (2009). doi:10.1002/mop.24443
40. U. Beaskoetxea, V. Pacheco-Peña, B. Orazbayev, T. Akalin, S. Maci, M. Navarro-Cía, M.
Beruete, 77 GHz high gain bull’s-eye antenna with sinusoidal profile. IEEE Antennas Wirel.
Propag. Lett. 14, 205–208 (2015). doi:10.1109/LAWP.2014.2360215
41. P. Baccarelli, P. Burghignoli, G. Lovat, S. Paulotto, A novel printed leaky-wave “bull-eye”
antenna with suppressed surface-wave excitation, in IEEE Antennas and Propagation Society
Symposium, 2004. IEEE (2004), pp. 1078–1081. doi:10.1109/APS.2004.1329861
42. S.K. Podilchak, S. Paulotto, P. Burghignoli, Y.M.M. Antar, P. Baccarelli, A.P. Freundorfer,
G. Lovat, A printed “bull-eye” leaky-wave antenna fed by a non-directive surface-wave
launcher, in 2nd European Wireless Technology Conference, EuWIT (2009), pp. 81–83
43. S.K. Podilchak, P. Baccarelli, P. Burghignoli, A.P. Freundorfer, Y.M.M. Antar, Analysis and
design of annular microstrip-based planar periodic leaky-wave antennas. IEEE Trans.
Antennas Propag. 62, 2978–2991 (2014)
44. S.K. Podilchak, Y.M.M. Antar, A. Freundorfer, P. Baccarelli, P. Burghignoli, S. Paulotto, G.
Lovat, Planar antenna for continuous beam scanning and broadside radiation by selective
surface wave suppression. Electron. Lett. 46, 613 (2010). doi:10.1049/el.2010.0074
45. U. Beaskoetxea, S. Maci, M. Navarro-Cía, M. Beruete, 3D-printed 96 GHz bull’s-eye antenna
with off-axis beaming. IEEE Trans. Antennas Propag. 65, 17–25 (2017). doi:10.1109/TAP.
2016.2628322
46. A.A. Oliner, A. Hessel, Guided waves on sinusoidally-modulated reactance surfaces. IRE
Trans. Antennas Propag. 7, 201–208 (1959). doi:10.1109/TAP.1959.1144771
47. D.R. Jackson, T. Zhao, J.T. Williams, A.A. Oliner, Leaky surface-plasmon theory for
dramatically enhanced transmission through a subwavelength aperture, Part II: Leaky-wave
antenna model, in IEEE Antennas Propagation Society International Symposium, Columbus,
OH (2003), pp. 1095–1098. doi:10.1109/APS.2003.1219426
48. A.A. Oliner, D.R. Jackson, Leaky surface-plasmon theory for dramatically enhanced
transmission through a subwavelength aperture, Part I : Basic features, in Proceedings of
IEEE AP-S Symp. Radio Science Meeting, Columbus, OH (2003), pp. 1095–1098. doi:10.
1109/APS.2003.1219425
49. D.R. Jackson, A.A. Oliner, T. Zhao, J.T.T. Williams, Beaming of light at broadside through a
subwavelength hole: leaky wave model and open stopband effect. Radio Sci. 40, 1–12 (2005).
doi:10.1029/2004RS003226
50. C.H. Walter, Traveling Wave Antennas (McGraw Hill, 1965)
51. R.C. Johnson, H. Jasik, Antenna Engineering Handbook (McGraw Hill, 2007)
52. A. Ishimaru, Electromagnetic Wave Propagation, Radiation and Scattering (Prentice Hall,
Englewood Cliffs, 1991)
53. L.O. Goldstone, A.A. Oliner, Leaky-wave antennas I: rectangular waveguides. IRE Trans.
Antennas Propag. 7 (1959). doi:10.1109/TAP.1959.1144702
54. F. Monticone, A. Alù, Leaky-wave theory, techniques, and applications: from microwaves to
visible frequencies. Proc. IEEE 103, 793–821 (2015). doi:10.1109/JPROC.2015.2399419
55. W.W. Hansen, Radiating Electromagnetic Waveguide (1946)
56. D.D. King, Dielectric image line. J. Appl. Phys. 23, 699–700 (1952). doi:10.1063/1.1702285
140 M. Beruete et al.

57. M. Chen, B. Houshmand, T. Itoh, FDTD analysis of a metal-strip-loaded dielectric

leaky-wave antenna. IEEE Trans. Antennas Propag. 45, 1294–1301 (1997). doi:10.1109/8.
611250
58. S. Kobayashi, R. Lampe, R. Mittra, S. Ray, Dielectric rod leaky-wave antennas for
millimeter-wave applications. IEEE Trans. Antennas Propag. 29, 822–824 (1981). doi:10.
1109/TAP.1981.1142669
59. K.L. Klohn, R.E. Horn, H. Jacobs, E. Freibergs, Silicon waveguide frequency scanning linear
array antenna. IEEE Trans. Microw. Theory Tech. 26, 764–773 (1978). doi:10.1109/TMTT.
1978.1129484
60. R.E. Horn, H. Jacobs, K.L. Klohn, E. Freigbergs, Single-frequency electronic-modulated
analog line scanning using a dielectric antenna. IEEE Trans. Microw. Theory Tech. 30, 816–
820 (1982). doi:10.1109/TMTT.1982.1131144
61. F.K. Schwering, S.-T. Peng, Design of dielectric grating antennas for millimeter-wave
applications. IEEE Trans. Microw. Theory Tech. 31, 199–209 (1983). doi:10.1109/TMTT.
1983.1131458
62. M. Beruete, I. Campillo, J.S. Dolado, E. Perea, F. Falcone, M. Sorolla, Very low-profile
“bull’s-eye” feeder antenna. IEEE Antennas Wirel. Propag. Lett. 4, 365–368 (2005). doi:10.
1109/LAWP.2005.851104
63. M. Beruete, I. Campillo, J.S. Dolado, J.E. Rodríguez-Seco, E. Perea, M. Sorolla, Enhanced
microwave transmission using a subwavelength slot in corrugated plate, in IEEE International
Symposium on Antennas and Propagation, vol. 3A (2005), pp. 14–17. doi:10.1109/APS.
2005.1552161
64. A.A. Oliner, D.R. Jackson, Leaky-wave antennas, in Antenna Engineering Handbook, ed. by
J.L. Volakis (Mc Graw-Hill, New York, 2007), pp. 11-1-56
65. M. Guglielmi, D.R. Jackson, Broadside radiation from periodic leaky-wave antennas. IEEE
Trans. Antennas Propag. 41, 31–37 (1993). doi:10.1109/8.210112
66. T. Rozzi, R. De Leo, A. Morini, Analysis of the “microstrip-loaded inset dielectric
waveguide”, in IEEE MTT-S International Microwave Symposium Digest (1989), pp. 923–
926. doi:10.1109/MWSYM.1989.38873
67. T. Rozzi, L. Ma, Scattering by dipoles in inset dielectric guide and application to millimetric
leaky wave antennas, in 17th European Microwave Conference 1987, pp. 543–548. doi:10.
1109/EUMA.1987.333662
68. G. Lovat, P. Burghignoli, D.R. Jackson, Fundamental properties and optimization of
broadside radiation from uniform leaky-wave antennas. IEEE Trans. Antennas Propag. 54,
1442–1452 (2006). doi:10.1109/TAP.2006.874350
69. P. Burghignoli, G. Lovat, D.R. Jackson, Analysis and optimization of leaky-wave radiation at
broadside from a class of 1-D periodic structures. IEEE Trans. Antennas Propag. 54, 2593–
2604 (2006). doi:10.1109/TAP.2006.880725
70. U. Beaskoetxea, M. Navarro-Cía, M. Beruete, Broadband frequency and angular response of a
sinusoidal bull’s eye antenna. J. Phys. D Appl. Phys. 49, 265103 (2016). doi:10.1088/0022-
3727/49/26/265103
71. M. Beruete, P. Rodríguez-Ulibarri, V. Pacheco-Peña, M. Navarro-Cía, A.E. Serebryannikov,
Frozen mode from hybridized extraordinary transmission and Fabry-Perot resonances. Phys.
Rev. B. 87, 205128 (2013). doi:10.1103/PhysRevB.87.205128
72. N. Engheta, R.W. Ziolkowski, Metamaterials: Physics and Engineering Explorations (Wiley,
2006)
73. D.Y. Na, K.-Y. Jung, Y.B. Park, Transmission through an annular aperture surrounded with
corrugations in a PEC plane. IEEE Antennas Wirel. Propag. Lett. 14, 179–182 (2015). doi:10.
1109/LAWP.2014.2360130
74. U. Beaskoetxea, M. Beruete, High aperture efficiency wide corrugations bull’s-eye antenna
working at 60 GHz. IEEE Trans. Antennas Propag. 65(6), 3226–3230 (2017). doi:10.1109/
TAP.2017.2696423
4 Flat Corrugated and Bull’s-Eye Antennas 141

75. M. Nannetti, F. Caminita, S. Maci, Leaky-wave based interpretation of the radiation from
holographic surfaces, in IEEE Antennas and Propagation Society, AP-S International
Symposium (Digest) (2007), pp. 5813–5816. doi:10.1109/APS.2007.4396873
76. S. Cakmakyapan, A.E. Serebryannikov, H. Caglayan, E. Ozbay, Spoof-plasmon relevant
one-way collimation and multiplexing at beaming from a slit in metallic grating. Opt. Express
20, 26636–26648 (2012). doi:10.1364/OE.20.026636
77. B.H. Fong, J.S. Colburn, J.J. Ottusch, J.L. Visher, D.F. Sievenpiper, Scalar and tensor
holographic artificial impedance surfaces. IEEE Trans. Antennas Propag. 58, 3212–3221
(2010). doi:10.1109/TAP.2010.2055812
78. G. Minatti, M. Faenzi, E. Martini, F. Caminita, P. De Vita, D. Gonzalez-Ovejero, M.
Sabbadini, S. Maci, Modulated metasurface antennas for space: synthesis, analysis and
realizations. IEEE Trans. Antennas Propag. 63, 1288–1300 (2015). doi:10.1109/TAP.2014.
2377718
79. J.J. Greffet, R. Carminati, K. Joulain, J.P. Mulet, M. Stéphane, Y. Chen, Coherent emission of
light by thermal sources. Lett. Nat. 416, 61–64 (2002). doi:10.1103/PhysRevB.69.155412
80. S. Han, Theory of thermal emission from periodic structures. Phys. Rev. B. 80, 1–10 (2009).
doi:10.1103/PhysRevB.80.155108
81. S. Krishna, S.D. Gunapala, S.V. Bandara, C. Hill, D.Z. Ting, Quantum dot based infrared
focal plane arrays. Proc. IEEE 95, 1838–1852 (2007). doi:10.1109/JPROC.2007.900969
82. J.A. Shackleford, R. Grote, M. Currie, J.E. Spanier, B. Nabet, Integrated plasmonic lens
photodetector. Appl. Phys. Lett. 94, 8–11 (2009). doi:10.1063/1.3086898
83. A. Harrer, B. Schwarz, R. Gansch, P. Reininger, H. Detz, T. Zederbauer, A.M. Andrews, W.
Schrenk, G. Strasser, Plasmonic lens enhanced mid-infrared quantum cascade detector. Appl.
Phys. Lett. 105 (2014). doi:10.1063/1.4901043
84. T. Xu, Y.-K. Wu, X. Luo, L.J. Guo, Plasmonic nanoresonators for high-resolution colour
filtering and spectral imaging. Nat. Commun. 1, 59 (2010). doi:10.1038/ncomms1058
Chapter 5
Reflectarrays

Sean Victor Hum

Abstract High-gain antennas are a crucial component of microwave, millimetre,

and sub-millimetre wave communication and sensing systems. Many systems, such
as satellite communication links, point-to-point terrestrial links, deep-space com-
munication links, RADAR systems, and remote-sensing systems depend on
high-gain antennas that can be realized in low-profile form at low cost. Combining
the best features of reflector antennas and antenna arrays, reflectarrays are
promising planar antenna solution for these applications, providing good efficiency
and a high degree of flexibility in synthesizing arbitrary antenna patterns.
Furthermore, they can be engineered to provide reconfigurable beam-forming
capabilities using a variety of approaches. This chapter describes the fundamental
operating and design principles of reflectarrays, both in fixed and reconfigurable
form. Several practical design examples are included to help readers build confi-
dence with designing this type of space-fed array. Important developments in the
field, particularly in the active research area of reconfigurable reflectarrays, are
summarized and important future challenges enumerated. Finally, the state of the art
in reflectarray applications is reviewed, illustrating the promising future of this
emergent antenna architecture.

5.1 Reflectarray Fundamentals

5.1.1 Motivation

High-gain antennas are a crucial component of microwave, millimetre, and

sub-millimetre wave communication and sensing systems. These antennas are
paramount in many systems including satellite communication links, point-to-point
terrestrial links, deep-space communication links, RADAR (including synthetic

S.V. Hum (&)

Edward S. Rogers Sr. Department of Electrical & Computer Engineering,
University of Toronto, 10 King’s College Road, Toronto, ON M5S 3G4, Canada
e-mail: [email protected]

© Springer International Publishing AG 2018 143

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_5
144 S.V. Hum

aperture RADAR) systems, and remote-sensing systems. Generally, these antennas

either take the form of aperture antennas such as reflectors and lenses, or antenna
arrays, depending on system requirements. On the one hand, aperture antennas offer
a straightforward solution for producing high-gain patterns. However, they tend to
be bulky and heavy, which can be a major drawback in spaceborne platforms where
launch weight and volume are crucial driving factors. In situations where
beam-steering is needed, traditional aperture antennas require mechanical scanning
to implement this capability. On the other hand, antenna arrays offer a flexible
antenna platform from which high-gain antenna patterns, including potentially
adaptive ones, can be realized. Phased arrays, in particular, provide adaptive
beam-scanning and beam-shaping capabilities. However, the main drawbacks of
arrays are the high complexity and potential losses in the beam-forming network.
The latter are particularly pronounced at millimetre and sub-millimetre wave fre-
quencies. In the case of phased arrays, there is also a substantial hardware footprint
as each array element (or sub-array in some cases) requires a dedicated RF signal
chain to handle transceiver functions. This leads to very high implementation costs
in phased arrays since this hardware needs to be duplicated at the element or
sub-array level.
The issues facing both apertures and arrays have spurred research into alternative
antenna platforms. Most significantly, reflectarray antennas have emerged as
promising candidates for realizing high-gain, high-performance antenna platforms.
As the name suggests, the reflectarray combines features of both reflector antennas
and antenna arrays. A typical microstrip reflectarray is shown in Fig. 5.1a. In its
simplest form, a reflectarray can be thought of as an array of electromagnetic
scatterers illuminated by a feed antenna, much like a traditional reflector antenna.
The array and scatterers are usually planar. In order to emulate the behaviour of a
curved reflector, each scatterer must be configured to produce a phase shift to

x x

P Q


rmn z r  θ z
O O
y Feed y Feed

Reflector
Reflectarray Aperture plane Aperture plane
(a) Reflectarray antenna (b) Paraboloidal reflector

Fig. 5.1 Comparison of a reflectarray and a reflector antenna

5 Reflectarrays 145

compensate for the path lengths and subsequent phase delays introduced between
the feed antenna and reflector surface. As such, a reflectarray can be thought of as
an array of spatial phase shifters that electrically deform the surface to resemble the
electromagnetic response of a physical reflector, over a given frequency range.
A myriad of techniques for realizing these phase shifters have emerged over the
past few decades, and have inspired research along many fronts. While most
research focuses on realizing fixed-beam reflectarrays, increasingly research has
focused on reconfigurable reflectarrays capable of tuning the response of the
scatterers to provide adaptive beam-steering and beam-forming capabilities.

5.1.2 Brief History of Development

Flat reflectors have been long sought after as replacements for bulkier curved
reflectors. The first reflectarray implementation was reported by Berry et al. in 1963
[1]. The unit cells consisted of short-circuited rectangular waveguide sections,
whose lengths were chosen appropriately to realize the correct delays to collimate
radiation from the primary feed. However, the concept did not capture widespread
interest, likely because the waveguide array itself was bulky, similar to the
shaped-reflector it aimed to replace. The advent of microstrip antenna technology
was a game-changer for reflectarrays, since it allowed them to be realized in true
low-profile form. Advances in microstrip antenna technology revived interest in
reflectarrays in the late 1980s/early 1990s [2, 3], sparking considerable advances in
the design of various planar units.
Early reflectarray elements were based on resonators, based on the fact that
pronounced shifts in the phase of the scattered field can be produced in the vicinity
of the resonant frequency of the scatterer. Such designs employed printed
dipoles/patches slots, rings, and other interesting shapes to realize the resonator.
Equally popular in early designs was using delay lines in the form of stubs con-
nected to microstrip patches, to achieve the required phase shifts. Physical (se-
quential) rotation of the elements could be used to realize phase shifting of
circularly polarized waves. These phase-shifting techniques are described more
detail in Sect. 5.2.2. Basically, by manipulating the physical characteristics of the
elements, the desired phase profiles for beam collimation, shaped beams, and other
radiation patterns could be realized by treating the reflectarray as a type of antenna
array with only control of the phase of the radiated fields from each element. Early
reflectarrays tended to follow standard rules for arrays to avoid grating lobes,
usually employing half-wavelength spacing between elements. Much of the early
effort in reflectarrays throughout the 1990s was motivated by their potential in
satellite communications.
In the 2000s, important advances were made to expand the functionality and
utility of reflectarrays. One important advance was that researchers recognized that
the bandwidth of reflectarrays needed to be significantly improved if they were to
realistically compete with reflectors. This was initially done by coupling multiple
146 S.V. Hum

resonators together, either to support dual/multi-band operation or to increase the

bandwidth about a single operating frequency. However, researchers soon saw that
in order to achieve good performance in terms of phase control and polarization
purity over large bandwidths, the reflectarray had to be seen as less as a discrete
array of scatterers and more as an inhomogeneous surface impedance, which ini-
tiated the development of reflectarrays based on sub-wavelength (versus
half-wavelength) unit cells. Many modern wideband implementations of reflectar-
rays employ sub-wavelength unit cells and are analyzed in terms of this framework.
A second major advance was the realization of reconfigurable functionality from
reflectarrays. By adding electronic tuning capability to the unit cell, the phase
profile of the reflectarray can be dynamically manipulated, allowing for function-
ality such as adaptive beam-scanning, on-the-fly pattern synthesis, and even
advanced features such as frequency reconfigurability. Advances in both micro-
wave components as well as tunable microwave materials have spurred research
and development in a great variety of beam-forming platforms that includes not
only reflectarrays, but related space-fed arrays such as transmitarrays, as discussed
in Chap. 6. We will consider developments in this area in more detail in Sect. 5.4.

5.1.3 Basic Operating Principle

The operating principle of a reflectarray can be rationalized from both the per-
spective of a reflector antenna and an antenna array. In their most basic form,
reflectarrays emulate curved reflectors to synthesize pencil beams. As such, con-
sider a basic axially symmetric reflector shown in Fig. 5.1b that produces a pencil
beam along the reflector ðzÞ axis. The reflector achieves collimation of radiation
from a feed placed at the origin O by ensuring that for each ray emanating from the
feed and subsequently reflected from the reflector, that the total path length tra-
versed by the ray is constant. Referring to Fig. 5.1b, for each point P considered on
the reflector surface,

OP þ PQ ¼ r 0 þ r 0 cos h0 ¼ constant; ð5:1Þ

where the point Q is located in an imaginary aperture perpendicular to the reflector

axis, in front of the reflector. A standard paraboloidal reflector results if the constant
in Eq. (5.1) is set to be 2f , where f is the focal length of the reflector. Then

0
h
0
r ¼ f sec 2
; h0  h0 ; ð5:2Þ
2

where h0 is the subtended half-angle of the reflector. Incidentally, from a ray optics
perspective, this reflector shape also produces a local angle of reflection such that
all rays emanating from the feed point O are parallel to the reflector axis.
5 Reflectarrays 147

The path length relation given by Eq. (5.1) can also be stated in terms of
electrical path length. If the free space wavenumber is denoted as k0 , then
k0 ðOP þ PQÞ ¼ constant, where a negative sign has been shown explicitly to
indicate phase delay. The reflectarray in Fig. 5.1a can be configured to achieve the
same behaviour as a reflector antenna by making all the electrical path lengths at
one frequency or a range of frequencies constant for each element on the reflec-
tarray surface. Each element of the reflectarray produces a phase shift wmn between
the incident and reflected waves. Furthermore, the second term of the equation is
not needed since the aperture plane can be defined to be coincident with the
reflectarray surface. If we consider a reflectarray in which the reflectarray elements
have been configured in a rectangular lattice and use a standard row–column
indexing scheme as shown in Fig. 5.1a, then the design equation for reflectarrays
can be written as
0
k0 rmn þ wmn ¼ constant; ð5:3Þ

where the indices m and n are used to index each element. wmn is typically con-
strained to fall within an interval ½0; 2p, and thus a factor of 2pN; N 2 Z is typi-
cally added to the constant so that wmn 2 ½0; 2p. It is worth noting that the
remaining constant in the equation is a phase offset that can be used as a degree of
freedom in determining where the phase wraps in the resulting phase mask of the
reflectarray are produced.
At this point, the choice of wmn produces a constant phase across the aperture
plane after the incident fields have been reflected from the reflectarray surface.
From an array perspective, it is as if the element excitations are uniform in phase.
Continuing with this interpretation, it is straightforward to see that the beam can be
steered off the reflector axis to a direction given by a unit vector ^r0 by superim-
posing a phase gradient on top of the existing phases,

0
k0 ðrmn ~
Rmn  ^r0 Þ þ wmn ¼ constant þ 2pN; ð5:4Þ

where ~ Rmn is a position vector of the mnth element within the plane of the
reflectarray. This equation is the standard pencil beam-forming equation for
reflectarrays. This antenna array perspective offers a simple way to see how the
phase shifts are derived. However, a corresponding reflector perspective is also
possible: just as a curved reflector employs Snell’s law to achieve beam collimation,
the same is also true of reflectarrays. However, a more generalized version of
Snell’s Law is required [4].
Consider a cross section of a 1D reflectarray surface shown in Fig. 5.2. The
element spacing is Dx. Two rays are shown that are associated with an incoming
plane wave incident upon the reflectarray. A phase shift wðxA Þ is applied by a
reflectarray element at point A, and a phase shift wðxB Þ by the corresponding
element at point B. If the angle of incidence is hi , the phase gradient between
points A and B determines the angle of reflection produced by the reflectarray, hr .
148 S.V. Hum

z
ray 1 ray 2

R
n̂ ñ
O
−θi θr

x
xA A xB B
Δx

Fig. 5.2 Generalized Snell’s law

The generalized Snell’s law can be found by observing that the phase shift incurred
along path AR plus that from wðxA Þ must be equal to that incurred along path OB
plus wðxB Þ,

wðxA Þ  k0 Dx sin hr ¼ wðxB Þ þ k0 Dx sin hi : ð5:5Þ

Rearranging this equation and taking the limit as Dx ! 0,

1 @wðxÞ
lim sin hr þ sin hi ¼  : ð5:6Þ
Dx!0 k0 @x

Clearly, we can see that Snell’s law for normal media ðhr ¼ hi Þ does not apply
here because of the phase gradient imposed by the reflectarray surface. In fact, the
effect of this spatial phase shift is to produce an effective normal ~ n that is not
coincident with the normal to the physical reflectarray surface, about which the
incident and reflected rays appear to obey Snell’s law. Hence, reflectarrays can be
seen as synthesizing a set of effective normals such that the emerging rays from the
surface are parallel to the desired direction of propagation. For a two-dimensional
reflectarray,

1 @wðx; yÞ
sin hr cos hr þ sin hi cos hi ¼  ð5:7aÞ
k0 @x

1 @wðx; yÞ
sin hr sin hr þ sin hi sin hi ¼  ð5:7bÞ
k0 @y

With the basic calculation of the required reflectarray element phases concluded,
we turn our attention to the realization of the required phase shifts by the reflec-
tarray elements. This forms the core part of the reflectarray design process: devising
and implementing a suitable reflectarray element or unit cell for implementing the
phase-shifting operation needed on the reflector surface. Literally, hundreds of
different designs have been proposed over the years, including both fixed and
reconfigurable variations. A survey of different element types will be presented in
5 Reflectarrays 149

more detail in Sect. 5.2.2. For illustration purposes, here, we explain the operation
of the reflectarray using a simple reflectarray element, a square microstrip patch.
The element geometry is shown in Fig. 5.3a. The element consists of a con-
ductive patch of dimensions l  w printed on a dielectric substrate. The substrate
has a thickness h and a dielectric constant r , and is backed by a conducting ground
plane. Patches in the reflectarray are situated in a periodic arrangement corre-
sponding to a rectangular unit cell of dimensions a  b. For simplicity, we consider
a simple square patch ðw ¼ lÞ situated within a square unit cell ða ¼ bÞ.
The scattering characteristics of the patch are determined by analyzing the patch
in a periodic environment. This is accomplished using Floquet analysis and is
presented in more detail in Sect. 5.2.1. The patch is illuminated with a plane wave,
for example, at normal incidence, and the scattered field is analyzed in terms of the
resulting reflection coefficient C. Typically, with low loss conductors and dielec-
trics, the magnitude of the reflection coefficient is close to unity, and the phase is of
most interest. An example of the phase shifts produced by cells of different lengths l
is plotted in Fig. 5.3b. It can be seen that for a given patch length l, an (inverted)
S-shaped phase curve is produced, resulting from the resonant behaviour of the
patch. This results in a phase curve that nearly spans a 2p radian phase range, and it
is this pronounced phase shift that allows patch-like scatterers to act as spatial phase
shifters. As the patch length increases, the resonant frequency decreases, resulting
in the multitude of curves shown in the plot. At the design frequency f0 , it is
possible to evaluate the phases as a function of patch length, resulting in the phase
curve plotted in Fig. 5.4.

π
arg(Γ) [rad]

w
f
h l
r
a −π
b Increasing l
f0
(a) Element geometry (b) Plot of scattered phase versus frequency

Fig. 5.3 Microstrip patch reflectarray element and its scattering characteristics

Fig. 5.4 Plot of scattered

phase versus patch length
π
arg(Γ) [rad]

f1 < f0
f0
l
Lmin f > f Δ Lmax
2 0
−π
150 S.V. Hum

This plot is a design curve for a reflectarray unit cell. It informs the designer
which patch length l must be chosen to yield a given phase shift. In general, the
realizable patch lengths are limited ðl 2 ½Lmin ; Lmax Þ, given that the patch must fit
within the confines of the periodic cell ðLmax \aÞ. This, coupled with the fact that a
single-pole resonator like a patch cannot fully span a 2p scattered phase range even
if its length was unconstrained, means that there will a range of phase shifts that are
not achievable with a cell of this type. Generally, the region is small and leads to
only small phase errors across the aperture and subsequent changes in the radiation
pattern from the reflectarray. A more significant issue is the fact that the required
phases (in the achievable range) can only be realized at the design frequency f0 . As
seen in the plot, at frequencies in the vicinity of f0 , a different S-curve is produced,
leading to an error D between the design phase at f0 and the actual phase at the
frequency of interest (e.g. f1 in the plot). A common rule of thumb is that the
bandwidth of the reflectarray is constrained to frequencies where the phase error is
D   p=8. Beyond this range, phase errors become pronounced enough to sig-
nificantly impact the ability of the reflectarray to form a pencil beam in the required
direction with the required gain. As a result, single-pole reflectarrays such as those
based on the unit cell here can only achieve very narrow bandwidths of operation,
typically on the order of a few percent.
Fundamentally, the bandwidth problem stems from the fact that reflectarray
element must not only compensate for the phase shift in reflectors but also correct
for the difference in delay between the flat reflector and curved-reflector surrogate.
Hence, fundamentally, true time delay (TTD) reflectarray elements are required to
achieve wideband operation. In the case of a broadside pencil beam, for example,
0
the ideal reflection phase from the element is wmn ¼ k0 rmn , which is a linear
function of frequency. Such a phase response cannot be realized by a
tuned-resonator design with a finite number of resonators, and necessitates other
design types which will be discussed in Sect. 5.2.2.
Reflectarrays can realize a diverse range of radiation patterns in addition to
traditional pencil beams, owing to the flexibility in manipulating the phase of the
aperture field. A wide variety of patterns have been demonstrated, including con-
toured beams, cosecant beams, multiple beams, and many others. While the detailed
presentation of synthesis methods is beyond the scope of this chapter, we briefly
overview a few techniques here. One simple technique is to use a shaped-reflector
as a surrogate for a reflectarray and determine the phase shifts using ray-tracing [5].
A number of phase-only synthesis techniques have also been proposed that are
applicable to reflectarrays, such as the intersection approach [6], which can be
subsequently implemented using alternating projections [7], and the matrix pro-
jection method [8]. There is also a class of methods that have been successfully
applied to reflectarray synthesis problems that use evolutionary algorithms to
determine the element phase distributions. These include particle swarm opti-
mization (PSO) [9], genetic algorithms [10], hybrid techniques, and others.
5 Reflectarrays 151

5.1.4 Reflectarray Variations

The majority of reflectarray designs are single-reflector designs as shown in

Fig. 5.1a whereby a moderate-directivity antenna such as a horn antenna is used as
the feed. Such reflectarrays consist of a single reflecting surface which can be fed
from a point centred in front of the reflector (a centre- or prime focus-fed reflec-
tarray), or at a point offset from the centre point (offset-fed reflectarray). Offset
reflectors are useful in situations where one wishes to minimize the impact of feed
blockage, particularly when the feed size is a significant fraction of the reflectarray
size. Offset reflectarray geometries can be related to their curved reflector coun-
terparts [11].
Another reflectarray architecture uses multiple reflectors, usually two, to
accomplish beam-forming as shown in Fig. 5.5a. Most such reflectarrays are
designed to emulate classical two-reflector designs such as Cassegrain or Gregorian
reflectors. This is done to reduce the distance between the feed antenna and
reflecting surface. In such cases, the sub-reflectarray phase distribution is chosen so
that it produces an effective hyperboloidal shape (for Cassegrain reflectors) or an
ellipsoidal shape (for Gregorian reflectors), while the main reflectarray maintains a
paraboloidal phase distribution. The corresponding design equations for the
sub-reflectarray are easily found [12].
Finally, a special type of dual-reflector antenna is known as the folded reflec-
tarray, which is illustrated in Fig. 5.5b. Similar to a folded reflector, a polarizing
grid is placed in front of the reflector, and is used to reflect radiation from the feed
back on to the reflectarray, and in some cases, to manipulate the amplitude distri-
bution on the reflectarray. As such, the main purpose of such designs is to reduce
the physical size of the reflector system while maintaining the equivalent focal
length of a single reflector. In reflectarray form, the main reflector must accomplish
a polarization-twisting operation so that the scattered fields, which have been
collimated by the reflectarray, can pass through the polarizing grid. The details of
how this can be accomplished will be discussed in Sect. 5.2.2.
Main reflectarray

Reflectarray

Polarizing grid +
optional reflectarray
Sub-reflectarray

Virtual
Feed
feed

Virtual
Real feed
feed
(a) Dual reflector (b) Folded reflector

Fig. 5.5 Reflectarray architectures

152 S.V. Hum

5.2 Reflectarray Design at the Unit Cell Level

5.2.1 Floquet Analysis

Reflectarray unit cell analysis revolves around treating the unit cells as if they were
in an infinite array, or periodic, environment. Reflectarrays are not strictly periodic
structures, because the cell geometry changes from cell to cell in order to achieve
feed collimation and beam-forming. Nevertheless, the changes from cell to cell are
generally small enough that locally, each cell can be treated as if it were surrounded
by identical unit cells. The use of the so-called local periodic assumption greatly
simplifies the analysis of reflectarray unit cells while fairly accurately capturing the
mutual coupling effects with adjacent unit cells. Approaches for accounting for
perturbations in nearby cells in order to yield more accurate phase curves have been
proposed, but employing the periodic assumption generally yields acceptable
results.
Before discussing reflectarray unit cell analysis further, a brief review of
two-dimensional periodic structures is in order. Consider a planar surface current
that flows in a plane parallel to the xy-plane. This current is defined spatially using a
function f ðx; yÞ which is definable over one period of the unit cell, a=2  x  a=2
and b=2  y  b=2. The current density is repeated periodically in both the x- and
y-directions. For example, the x-component of the current density can be written as

X
1 X
1
Jx ðx; yÞ ¼ f ðx  ma; y  nbÞ: ð5:8Þ
m¼1 m¼1

Without loss of generality, we can impose a linear phase gradient on the current
distribution, since linear phase gradients are also periodic, albeit generally with a
different spatial period than that of the unit cell itself. Let the scalar current dis-
tribution be

X
1 X
1
Jx ðx; yÞ ¼ f ðx  ma; y  nbÞ expðjkx0 ma  jky0 nbÞ; ð5:9Þ
m¼1 m¼1

where kx0 and ky0 denote the phase gradients in the x- and y-directions between two
successive unit cells. Selecting the phase shifts to be

kx0 ¼ k0 sin h0 cos /0 ð5:10aÞ

ky0 ¼ k0 sin h0 sin /0 ð5:10bÞ

establishes a beam from the infinite array pointing in the direction ðh0 ; /0 Þ. The
two-dimensional Fourier transform of this current distribution is
5 Reflectarrays 153

2 X1 X1
~Jx ðkx ; ky Þ ¼ 4p ~f ðkx ; ky Þ dðkx  kxmn Þdðky  kymn Þ; ð5:11Þ
ab m¼1 m¼1

where ~f denotes the two-dimensional Fourier transform of f ðx; yÞ, dðÞ is the Dirac
delta function, and

2mp
kxmn ¼ kx0 þ ð5:12aÞ
a
2np
kymn ¼ ky0 þ : ð5:12bÞ
b

It can be seen from Eq. (5.11) that the current distribution in the spectral domain
is defined at discrete spectral points. If we transform back to the spatial domain,
then

4p2 X1 X1
~f ðkxmn ; kymn Þ expðjkxmn x  jkymn yÞ
Jx ðx; yÞ ¼ ð5:13Þ
ab m¼1 m¼1

which is known as the Floquet series of Jx ðx; yÞ. It can be seen that each discrete
spectral point corresponds to a corresponding plane wave weighted by a factor of
~f ðkxmn ; kymn Þ, which is known as a Floquet mode. If

2
kxmn þ kymn
2
 k02 ð5:14Þ

then the Floquet mode is a propagating plane wave whose direction in the spherical
coordinate system is described according to

kxmn ¼ k0 sin hmn cos /mn ð5:15aÞ

kymn ¼ k0 sin hmn sin /mn ð5:15bÞ

kzmn ¼ k0 cos hmn : ð5:15cÞ

If Eq. (5.14) is not satisfied, then the Floquet mode is evanescent. Knowledge of
propagating Floquet modes is important in identifying the ways in which a periodic
structure can radiate, since large unit cell sizes can produce grating lobes which are
associated with higher order propagating Floquet modes. For reflectarray designs,
we are principally interested in the behaviour of the fundamental propagating
Floquet modes.
Our analysis of a reflectarray unit cell begins by defining the scatterer to be
within a periodic environment as shown in Fig. 5.6b. The scatterer is shown as a
free-standing planar structure in the z ¼ 0 plane that supports symmetric scattering
on both sides of z ¼ 0. The scatterer can be thought of as being placed with a
periodic waveguide whose walls correspond to periodic boundary conditions
154 S.V. Hum

Floquet port 1

z
T E|T M T E|T M
bmn amn

b
a
Floquet port 1
J

y
T E|T M T E|T M
x bmn amn
T E|T M T E|T M
dmn cmn
b
a
J

Floquet port 2 h
r y
x
(a) Two-port periodic structure (b) One-port periodic structure

Fig. 5.6 Periodic unit cell and excitation

(PBCs) and whose axis is along z-axis. A Floquet port is placed at both ends of the
waveguide to introduce incident waves towards the structure. Each mode represents
an incident wave propagating at an angle ðh0 ; /0 Þ towards the scatterer, establishing
incident wavenumbers according to Eq. (5.10a). Structuring the analysis of the unit
cell in this way, the goal is to derive the generalized scattering matrix
(GSM) relating the incident and scattered waves associated for all the Floquet
modes, which is written as
2 3 2 TE;TE 32 3
bTE S11 STE;TM
11 STE;TE
12 STE;TM
12 aTE
6 bTM 7 66
TM;TE
STM;TM STM;TE STM;TM
76 TM 7
76 a 7
6 TE 7 ¼ 6 S11
TE;TM 74 TE 5; ð5:16Þ
11 12 12
4 d 5 4 STE;TE STE;TM STE;TE S22 5 c
21 21 22
dTM STM;TE STM;TM STM;TE TM;TM
S22 cTM
21 21 22

TEjTM
where amn and bTEjTM
mn are vectors containing the amplitudes of the incident and
TEjTM
reflected waves for each Floquet mode from port 1, respectively, and cmn and
dTEjTM
mn are the corresponding amplitudes for port 2. Generally, unit cells are chosen
5 Reflectarrays 155

to be k0 =2 or smaller to avoid grating lobes (where k0 =2 is the free space

wavelength), so only two fundamental Floquet modes can be excited and propagate:
the TE00 mode and the TM00 mode.
Next, we consider the reflectarray case shown in Fig. 5.6b, whereby a ground
plane has been introduced at z ¼ h and a substrate with dielectric constant er
occupies the space between z ¼ h and z ¼ 0. In general, the scatterer could also
be a multi-layer structure with multiple dielectrics forming stratified layers below
z ¼ 0. In either case, the waves cmn and dmn are constrained by the reflection
coefficient seen looking in the z\0 direction at z ¼ 0. The reflectarray unit cell can
be thought of as a one-port device with
      
bTE STE;TE STE;TM aTE aTE
¼ 11 11 ¼ S11 : ð5:17Þ
bTM STM;TE
11 STM;TM
11
aTM aTM

The GSM of the reflectarray unit cell is usually first evaluated for normal
incidence (h0 ¼ 0). In the consideration of unit cell designs, it is also important to
evaluate the GSM at oblique angles, which is readily facilitated using the GSM by
changing h0 and /0 . This leads to a GSM that is angle-dependent, and subsequently
S-curves that are a function of the angle of incidence. Generally, there is a region of
angles of incidence over which the GSM is moderately stable, allowing the S-curve
at normal incidence to be used for first-cut design purposes. However, rigorous
designs consider all possible angles of incidence for a particular reflectarray and
feed geometry, and employing individual S-curves to account for the various dif-
ferent angles. This is more time-consuming, but leads to more accurate results.
In the literature, most numerical formulations for determining the GSM employ
the method of moments (MoM) in the spectral domain [13, 14]. These formulations
solve the EFIE on the metallic surfaces of the unit cell to yield the surface current
densities, which can, in turn, be used to calculate the GSM. Modern commercial
computer-aided design (CAD) tools generally support periodic structure analysis
within a Floquet waveguide, and support the computation of the periodic structure
GSMs. Numerical techniques are powerful and accurate, but at the early design
stages, analytical models can also be useful. A presentation of a potential analytical
analysis technique is deferred until reconfigurable reflectarrays are introduced in
Sect. 5.4.

5.2.2 Unit Cell Design Types and Methods

Over the past few decades, many different reflectarray unit cell designs have been
proposed for meeting a multitude of design requirements. To add to this mix, there
has recently been significant activity in the area of reconfigurable reflectarray unit
cells, which allow the scattered field phases of the reflectarray to be dynamically
manipulated. To explain the operation of these unit cells, we begin with fixed unit
156 S.V. Hum

cell designs and divide them according to the basic principle by which the phase
shifting action is realized. In Sect. 5.4, we will see how the design methods
translate to reconfigurable reflectarrays elements.

5.2.2.1 Tuned-Resonator Unit Cells

Most of the first planar reflectarray designs exploited the fact that a planar res-
onator, such as a microstrip patch or dipole, resonates when the electrical length of
the resonator corresponds to one-half of a wavelength (higher order resonances are
possible, but cell sizes are usually limited to be around a half-wavelength to avoid
the onset of grating lobes). About the resonance point, there are dramatic changes in
the phase of the scattered field of the unit cell as either frequency or the resonant
dimensions of the cell are changed. This is the basic principle of the variable-size
patch element discussed in Sect. 5.1.3. Exploiting the resonant behaviour of the
scatterer, phase ranges of close to 360 can be realized using basic shapes such as
dipoles, patches, loops, dog-bone shapes, and many others. A few of these shapes
are shown in Fig. 5.7a. The phase of the scattered field is manipulated by modifying
one or more dimensions of the metallic scatterers. Generally, the slope of the phase
curve versus scatterer dimension is influenced by the overall shape of the scatterer.
Sharp phase slopes are usually avoided, since they increase the sensitivity of the
phase to the manufacturing tolerances of the scatterer. More importantly, sharp
S-curves also translate to sharp S-curves as a function of frequency, whose non-
linear shape has a significant impact on the phase bandwidth of the unit cell.
The bandwidth of the unit cell can be significantly improved by linearizing the
phase response of the cell. This is accomplished by using a design with multiple
scatterers with closely spaced, coupled resonances. This has the effect of making the
phase curve more linear in a region, while simultaneously extending the phase
range to make the cell emulate its true time delay counterpart. Conversely, res-
onators with substantially different resonant frequencies can be integrated together
to realize multi-band unit cell designs.
Unit cell designs based on the tuned-resonator approach can be further classified
into how they manipulate the incoming polarization of the wave. First, unit cells
designed using this approach is generally linearly polarized. Most unit cells are
ψ2

Γ1 Γ2 ψ
ψ1

(a) Various tuned-resonator LP cells (b) Resonator twist (c) Guided-wave (d) Guided-wave
cell LP cell twist cell

Fig. 5.7 Reflectarray unit cell design methods

5 Reflectarrays 157

co-polarized, meaning that the reflected field is the same polarization as the incident
field. It is possible to devise unit cell designs that phase orthogonally polarized
waves independently, realizing a dual-polarized unit cell design. For example, the
rectangular patch element shown in Fig. 5.7a can control the reflection phase of two
polarizations: the vertical polarization is phase-shifted by modifying the length of
the patch, and the horizontal polarization by the width of the patch. Finally, there
are some unit cells that twist the incoming polarization. Such cells are referred to as
cross-polarized. Usually, the need to twist the polarization is needed for specific
purposes, such as implementing a so-called “trans-reflector” for folded reflectarrays
discussed in Sect. 5.1.4. This requires that a dual-polarized reflectarray element that
is designed to achieve specific phase shifts to the two orthogonal polarizations. In
the case of a linearly polarized reflectarray element, if the incident wave is polarized
at a 45 angle with respect to the reflectarray element axis, then the incoming
electric field vector can be decomposed into two orthogonal components, as shown
in Fig. 5.7b. These two components are then acted upon by two different reflection
coefficients C1 and C2 controlled by the length and width of the patch. The
dimensions of the element can be chosen to produce a relative phase shift of 180
between the two components in order to twist the polarization of the incoming
signal by 90 . Meanwhile, the choice of dimensions is also chosen to manipulate
overall phase shift applied commonly to both components, which can be tailored for
collimating the feed radiation similar to a single-polarization reflectarray.

5.2.2.2 Guided-Wave Unit Cells

A guided-wave unit cell design transitions the space-wave impinging upon the
reflectarray element to a guided-wave. The phase shifting operation is implemented
in the guided-wave portion of the element, using traditional guided-wave phase
shifters. The cell is shown in Fig. 5.7c, where a transmission line is coupled to each
polarization of the patch and connected to corresponding reflective phase shifters
producing phase shifts w1 and w2 for each polarization (only one phase shifter and
line are needed for a single-polarization cell). In the case of fixed reflectarray
elements, the phase shifter is usually an open-circuited stub connected to a patch
element. By varying the length of the stub, the phase shift is equal to 2b‘, where ‘
is the length of the stub and b is the phase constant along the line.
This technique can be seen as part of a class of true time delay (TTD) reflectarray
elements that attempt to reproduce the required delay profile required of the
reflectarray, rather than simply the phase. This is important in the design of
broadband reflectarrays. However, it is worth pointing out that there are limits to the
delays that can be realized (since there are limits to how long the stub can be made
given the confines of the unit cell) and the fact that coupling between the patch and
stub tends to still be narrowband. Furthermore, the introduction of stubs can
interfere with the scattering response of the cell, particularly the cross-polarization
level, necessitating the need for symmetric reflectarray layouts to reduce
cross-polarization.
158 S.V. Hum

Polarization-twisting elements can also be realized using the approach of

manipulating the two reflection coefficients associated with each polarization of the
cell in a dual-polarization version. However, an easier technique is to simply
cross-connect the two polarization ports of the antenna as shown in Fig. 5.7d.
A suitable phase shifter or delay line can be introduced in the path to change the
phase of the outgoing wave.

5.2.2.3 Circularly Polarized, Sequential Rotation Unit Cells

Circular polarization (CP) reflectarray elements fall into a special category because
they are easily implemented using the so-called sequential rotation technique [15].
Instead of changing the size of the element to change the associated reflection
coefficients, in the sequential rotation technique the physical rotation angle a of the
element is used to control the phase shift, as illustrated in Fig. 5.8. Here a split ring
resonator is used to illustrate a generic cell which can realize different responses in
the principal directions, while maintaining symmetry about a local coordinate
system.
Mathematically, the operation of the cell can be explained using the diagram and
coordinate system illustrated in Fig. 5.8. An incident wave, nominally travelling in
the z direction, impinges upon the cell at normal incidence. This wave may be
composed of both left-hand and right-hand circularly polarized waves in general,
and is described according to

~
Ei ¼ arh ð^x þ j^yÞejk0 z þ alh ð^x  j^yÞejk0 z ; ð5:18Þ

where arhjlh denote the amplitudes of each circularly polarized component. The cell
is rotated physically by an angle a as shown. To determine the reflected wave, the
wave must be decomposed into its x- and y-components. Locally, the unit cell is
described using a GSM S0 aligned with the x0  y0 coordinate system shown in
Fig. 5.8, which the incident wave can be translated into using a rotation matrix. It
can then be shown that the reflected wave is given by [16]

z z

x α
y y

x x
y
(a) Unrotated cell (b) Rotated cell
Fig. 5.8 Example geometry and coordinate system for CP reflectarray unit cell
5 Reflectarrays 159

~
Er ¼ ðCrh rh rh lh
x  j^yÞejk0 z þ ðClh
c a þ Cx a Þð^ x þ j^yÞejk0 z ;
c a þ Cx a Þð^
lh lh rh
ð5:19Þ

where Clhjrh
c and Clhjrh
x refer to the co- and cross-polarized components of the
corresponding polarization components, respectively, and
 
1 0
Crh
c ¼ ðS11  S022 Þ þ jS012 ej2a ð5:20aÞ
2
 
1 0
Cc ¼ ðS11  S22 Þ  jS12 ej2a
lh 0 0
ð5:20bÞ
2

1 0 0
x ¼ Cx ¼ ðS11 þ S22 Þ:
Clh rh
ð5:20cÞ
2

An ideal CP cell is  based on a unit cell that has perfect isolation between the
TE and TM modes jS12 j ¼ jS021 j ¼ 0 , and maximum co-polarization reflection
0

for each mode with a 180 phase shift between them jS011 j ¼ jS022 j ¼ 1;
j\S011  \S022 j ¼ pÞ. With this phase shift, polarization sense of the reflected
component wave is the same as the incident wave. For example, for a purely
right-hand circularly polarized (RHCP) wave, these conditions lead to Crh c ¼ e
j2a

and Cx ¼ 0. We see that the phase shift introduced to the RHCP wave is pro-
rh

portional to 2a, twice the physical rotation angle of the cell. In fact, according to
Eq. (5.20b), an incident LHCP wave will undergo a similar phase shifting process,
so it makes no difference whether the cell is used for RHCP or LHCP systems. We
also see in both cases that the cross-polarization level is established purely by S011
and S022 , and that for minimum cross-polarization level their magnitudes should be
equal while the phase difference between them should be 180 .
The sequential rotation technique is an easy and straightforward way to imple-
ment phase shifting in CP reflectarrays. Once a suitable polarization-twisting cell
has been defined, it is a simple matter to rotate the elements to achieve the required
phase shift. A number of cells have been proposed for this purpose, including cells
based on stubs, rings, ring-slot resonators, and others. Generally, only
single-polarization cells have been proposed, since achieving dual-polarization CP
designs is quite challenging. Reconfigurable variants will be considered in
Sect. 5.4.

5.3 Reflectarray Analysis at the Array Level

5.3.1 Pattern Calculation

Once a suitable reflectarray element has been designed and characterized, the next
step is to assemble the unit cells and feed system into the desired reflectarray
160 S.V. Hum

geometry, and analyze the performance of the resulting system. The simplest way to
predict reflectarray patterns is using antenna array analysis, which employs the
principle of superposition. Knowing the individual excitation amplitudes and
phases of the scatterers composing the reflectarray, the radiated fields from the
reflectarray can be easily found via an array factor calculation. As such, the fol-
lowing assumptions are made:
• Mutual coupling between elements, beyond that predicted using the periodic
analysis (infinite array assumption), is not considered;
• The element factor is usually considered to be a uniformly illuminated rectan-
gular aperture whose dimensions correspond to that of the unit cell a  b; and
• The array itself does not interact with the feeding system, i.e. there is no
near-field coupling between the array and the feed system.
Consider the reflectarray shown in Fig. 5.1a, which has the feed located at the
origin. To calculate the radiated fields from the reflectarray, the electric field
incident upon each element must be determined. For real (non-analytical) feeds, this
can be determined from full-wave simulations or measured experimentally. At the
early design stages of a reflectarray, an analytical feed model is usually easier to
work with. In this case, the feed is assumed to be a source of spherical waves
weighted spatially by the radiation pattern of the feed. Mathematically, the incident
field upon the mnth reflectarray element can then be expressed as
 1=2 jkr0
~ Pf 0 0 e mn
Emn ¼ ^emn g0 Gf ðhmn ; /mn Þ
i i
0
; ð5:21Þ
2p rmn

where g0 is the intrinsic impedance of free space, Pf is the power applied to the
feed, Gf ðh0 ; /0 Þ is the gain pattern of the feed and ðh0mn ; /0mn Þ and rmn
0
are the angles
and distance from the phase centre of the feed to the mnth reflectarray element. The
unit vector ^eimn depends on the polarization of the feed ^ p, and is equal to

ð^r 0  ^pÞ  ^rmn

0
^eimn ¼  mn 0 
: ð5:22Þ
ð^rmn  ^pÞ  ^rmn
0

Locally, at each reflectarray unit cell along the surface of the reflectarray, the
incident electric field is treated as a plane wave arriving at an angle of incidence
ðh0mn ; /0mn Þ. As such, the incident field consists of only two fundamental propa-
gating Floquet modes: the TE00 mode and the TM00 mode, which are designated
aTE TM
mn and amn . These can be determined solely from the tangential components of the
incident field at each cell. The GSMin Eq. (5.17) is then used to compute the
reflected electric field amplitudes bTE TM
mn and bmn . Since the tangential components are
oriented in the xy-plane, we can write the GSM relation as
5 Reflectarrays 161

   0 0 0 0  
bxmn xx ðhmn ; /mn Þ
Cmn Cmn
yx ðhmn ; /mn Þ axmn
¼ 0 0 0 0 : ð5:23Þ
bymn xy ðhmn ; /mn Þ
Cmn Cmn
yy ðhmn ; /mn Þ aymn

This process can also be used to treat feeds not located at the origin (e.g. in the
case of offset reflectors), by defining a local feed coordinate system, expressing the
radiated fields from the feed in that coordinate system, and transforming these fields
into the incident fields on the reflectarray using a transformation matrix.
The far-field radiation pattern can now be easily computed using array theory.
Each reflectarray element can be treated as an aperture of dimensions a  b, which
is considered to be uniformly illuminated by the reflected electric fields for each
polarization. The element factor is found through asymptotic evaluation of the
radiated fields from a rectangular aperture, which is assumed to be embedded in an
infinite ground plane. Essentially, the equivalent currents are evaluated over the
aperture, and the radiated (reflected) electric field from the mnth element is given by

~ k0 ejk0 r h^ x  
^ cos h bx sin / þ bt cos /
i
r
Emn jF h bmn cos / þ bymn sin / þ / mn mn
2pr
exp½jk0 ðxmn sin h cos / þ ymn sin h sin /;
ð5:24Þ

where



kx a ky b
F ¼ ab sinc sinc expðjk0 z0 cos hÞ: ð5:25Þ
2 2

Here, sincðxÞ ¼ sinðxÞ=x and z0 is the location of the reflectarray. The reflected
electric amplitudes bxmn and bymn act as the weights for the array factor, so that the
total radiated field can be written as
jk0 r
k0 e X Xh    i
~
E r ¼ jF ^h bx cos / þ by sin / þ /
^ cos h bx sin / þ bt cos /
mn mn mn mn
2pr m n
exp½jk0 ðxmn sin h cos / þ ymn sin h sin /;
ð5:26Þ

A rectangular lattice has been assumed for the reflectarray geometry, and hence
row–column indexing has been employed in the summations. It is worth noting that
each component of ~ E r can be computed quite quickly using the inverse 2D discrete
Fourier transform [17]. This can be a major advantage when evaluating the radiated
fields from large reflectarrays, since the Fast Fourier Transform can be employed to
greatly accelerate the calculations.
The array factor method for computing the radiated fields from the reflectarray is
quite accurate. Since the GSM captures the co- and cross-polarization behaviour of
the constituent elements of the reflectarray, most of the dominant contributors to
pattern features are included. Diffraction from the edges of the reflectarray is not
162 S.V. Hum

accounted for. Interestingly, despite the assumption of local periodicity among the
reflectarray unit cells, this simplified analysis does a very good job in predicting the
pattern when comparing to measurements or full-wave simulations of the entire
reflectarray. More accurate formulations have been proposed, whereby the spectral
domain Green’s function of the array elements is considered and used to form a
continuous spectrum of equivalent currents over the entire reflectarray aperture
[18]. This mitigates the problem of having discontinuities in the equivalent currents
between unit cells, which leads to more accurate results, especially in the case of
highly aperiodic reflectarrays.

5.3.2 Gain and Aperture Efficiency

The peak gain G of a reflectarray is related to the aperture efficiency gap via

4p
G ¼ gap Ap ð5:27Þ
k20

where Ap is the physical area of the reflectarray. The aperture efficiency is deter-
mined from

gap ¼ gspillover gtaper gohmic gdiel gphase gblockage gspecular ; ð5:28Þ

where only the dominant contributors to efficiency have been listed. Each term is
described as follows.
The spillover efficiency gspillover is determined from the ratio of the power col-
lected from the reflectarray to that emitted from the feed. This is easily approxi-
mated by treating reflectarray unit cells as ideal rectangular collectors, so that
PP
Pinc
mn ab
m n
gspillover ¼ R 2p R p ; ð5:29Þ
0 0 Uf ðh0 ; /0 Þ sin h0 dh0 d/0

mn is the incident power density at the mnth cell and Uf ðh; /Þ is the
where Pinc
radiation intensity from the feed. A straightforward way for calculating the taper
efficiency gtaper is to employ the array factor approximation of the radiated field
5.26) for computing the radiated fields with and without the feed-induced taper in
place. The taper efficiency can then be found from the ratio

jEr ðh0 ; /0 Þj2actual

gtaper ¼ ; ð5:30Þ
jE r ðh0 ; /0 Þj2incident field uniform

where ðh0 ; /0 Þ refers to the beam pointing direction and E r refers to the desired
co-polarized field component. The reflected fields from the unit cells in this
5 Reflectarrays 163

calculation should not include ohmic or dielectric losses from the element which
would further modify the taper levels. Instead, the losses are removed and accounted
separately in the efficiency terms gohmic and gdielectric separately, considering the
analysis of the reflectarray with the appropriate loss mechanisms in place. Ohmic loss
here considers both conductor losses and thermal losses within potential reconfigu-
ration mechanisms (e.g. the parasitic resistance of varactor diodes).
Phase errors over the aperture contribute to a phase error efficiency term gphase
in (5.35). These phase errors arise from imperfect phasing of the elements, resulting
from limited phase range of the elements, the discrete number of phase states
available from the library of elements (or finite number of bits available in the case
of digitally controlled reconfigurable elements), and phase errors arising in general
from imperfections, frequency dispersion, etc. The phase error efficiency can also
be determined using array factor techniques using

jE r ðh0 ; /0 Þj2actual
gphase ¼ : ð5:31Þ
jEr ðh0 ; /0 Þj2ideal phases

where the numerator is evaluated with the actual phases employed in the reflec-
tarray, and the denominator is evaluated assuming perfect (ideal) phases without
errors.
The final two terms in (5.35) are the feed blockage and specular reflec-
tion losses, respectively. Feed blockage results from a shadow cast by the feed
when it is placed in front of the reflector surface. It is usually mitigated by using an
offset feed configuration. It can only be predicted using full-wave simulations of the
entire reflectarray and feed system. Specular reflection losses result from imbal-
ances between the reflectarray element and ground plane currents, particularly away
from the operating frequency of the reflectarray. This imbalance gives rise to stray
reflections in the specular direction creating a beam known as a feed image lobe
[19]. Similar to feed blockage, full-wave analysis of the entire reflectarray is needed
to predict the exact loss to specular reflections. Generally specular reflection losses
do not significantly reduce the gain of the reflectarray, but they can increase the
stray radiation in other directions if not controlled carefully.

5.4 Reconfigurable Reflectarrays

The addition of reconfigurable capabilities to reflectarrays opens the possibility to

dynamic beam-steering and shaping, allowing for on-the-fly reconfiguration of the
antenna pattern. There are many scenarios that can benefit from adaptive, high-gain
patterns including satellite systems (especially satcom-on-the-move systems),
RADAR, point-to-point microwave links, and remote-sensing (e.g. synthetic
aperture RADAR). Beam-forming capabilities will also be foremost in
next-generation (5G) communications systems, which employ higher frequencies
164 S.V. Hum

and hence have stronger needs for high-gain antennas. With the associated narrower
antenna beams, adaptive beam-forming becomes a natural requirement.
Reflectarrays and other space-fed arrays offer compelling advantages over other
antennas in this arena. Unlike reflector antennas, which must be mechanically
scanned and offer very limited possibilities for adaptive re-shaping of the beam, the
aperture fields can be modified in electronically tunable reflectarrays by modifying
the reflectarray phase distribution. While phased arrays offer the same capability,
this is achieved at a much higher cost due to the replication of transmit/receive
modules for each antenna (or sub-array, at least) in the array. Additionally, the
transmission line feed networks associated with phased arrays can become signif-
icantly lossy at high frequencies, especially in the mm-wave range. Beam-forming
networks also become very complex in the structure for very large phased arrays.
Spatially fed arrays remove the need to lay out complex feed networks and mitigate
much of the loss associated with transmission line feeds, replacing it partly with
taper and spillover losses typical of reflector antennas, which are easier to control.
Reconfigurable reflectarrays can be placed into two classes in terms of methods
used to realize beam reconfiguration. The first method uses a scanned feed and fixed
(non-reconfigurable) reflectarray surface. The feed is moved to different locations in
front of the reflectarray. This scans the beam off the reflector axis, and is commonly
used in parabolic reflectors and their reflectarray counterparts [20]. Parabolic,
spherical, and multi-focal phase distributions can be used on reflectarray aperture to
manage the scanning range of the beam. Hence, the reconfiguration is still
mechanical, albeit in a different sense than a reflector which is mechanically rotated
to point the beam. Some degree of “electronic” beam-steering can be achieved by
electronically switching between multiple feeds located at required focal points to
produce the corresponding beams, allowing for multi-beam capabilities [9]. The
second method is to use a stationary feed, but a reflectarray surface that can be
electronically manipulated to change the phase distribution and corresponding
aperture fields [21]. This method offers the most flexibility in terms of the patterns
that can be realized from the antenna, as well as the potential speed at which the
aperture can be reconfigured, since large macroscopic movement of parts is not
required. This method is the focus of this section.

5.4.1 Enabling Technologies

At the heart of a reconfigurable reflectarrays is a unit cell design with electronically

tunable characteristics. These characteristics in turn depend on microwave and
mm-wave technologies that allow for electronic reconfiguration of the cell in some
way [21]. A summary of a wide range of enabling technologies is tabulated in
Table 5.1. An ideal reflectarray obviously should be low cost, consume little or no
power, exhibit low loss, be fast to reconfigure, and offer low implementation
complexity. Additionally, generally digital control is preferred to analogue control
 5 Reflectarrays

Table 5.1 Enabling technologies for reconfigurable reflectarrays

Type Technology Maturity Digital/analogue Complexity Cost Loss (l-wave / Bias power Linearity Configuration
control mm-wave) consumption time
Lumped-elements PIN diodes High D Low Low H/H High Low Fast
Varactor diodes High A Low Low H/H Low Low Fast
RF-MEMS Mod. D Low Mod. L/M Low High Mod.
Hybrid Ferroelectric Mod. A Mod. Mod. M/L Low Mod. Fast
films
Tunable materials Liquid crystal Low A High High H/L Low Low Mod.
Graphene Low A High High H/L Low Low Fast
Photoconductive Low A High Mod. H/H High Low Fast
Mechanical Fluidic Low A High Mod. M/H Low High Slow
Micro-motors Low A High Mod. L/– Low High Slow
165
166 S.V. Hum

schemes, and in certain applications, mature technologies are preferable (e.g.

space-qualified devices for satellite communications). However, as we will see, as
one migrates to the millimetre wave frequencies, we will see that more of these
ideals must be sacrificed.
Each technology is broadly classified according to its type in the first column.
The first technology type, lumped-element, refers to electronic components that can
be introduced into the reflectarray unit cell at discrete, spatially localized points.
Radio frequency (RF) switches are the most obvious example, which can be real-
ized both in semiconductor form (such as from PIN diodes and FET switches), and
in the form of RF microelectromechanical systems (MEMS) switches. Tunable
reactances are also commonly used, usually as capacitances which can be realized
from semiconductors (varactor diodes), and also in MEMS form, where both tun-
able capacitors and tunable inductors have been realized. Semiconductor-based
technologies exhibit significant levels of loss which reduce the reflection efficiency
of reflectarray unit cells. They also exhibit nonlinear behaviour at large RF signal
levels, which can lead to signal distortion problems when used in transmitters. For
these reasons, MEMS devices have grown in popularity, especially MEMS
switches, since they generally mitigate these issues.
At the other end of the spectrum are tunable materials, which instead provide
tunable constitutive parameters (permittivity, permeability, or conductivity) in the
underlying material platform order to realize distributed tuning of the reconfigurable
unit cell. An excellent example of these is liquid crystal (LC) materials. These
materials have a tunable permittivity tensor which is dependent on a DC or
low-frequency AC signal biasing field applied to the material. Any structure, such as
patch, using LC as its substrate will have a resonant frequency (and corresponding
scattered field phase) dependent on the properties of the substrate permittivity.
Graphene is an emerging technology, which in the terahertz frequency range exhibits
a surface impedance which can be tuned by manipulating the doping parameters of
the graphene as well an applied electric field (the so-called graphene “field effect”).
Photoconductive materials, such as semiconductors, can exhibit sensitivity to an
applied light source to change the carrier concentration (and hence conductivity) of a
predefined region. This region can range from being localized to a very small area
(realizing a photoconductive switch, which is more of a lumped-element type of
reconfiguration) to the entire element itself (or absence thereof, when the light is off).
Somewhere between these two extremes lie hybrid technologies, which are
neither distributed nor spatially localized as in the case of lumped-elements.
Devices based on ferroelectric films fall into this category, since the ferroelectric
films are a type of tunable material, but the corresponding devices are often realized
in component form.
Finally, mechanical solutions tend to fall into their own unique category, though
they largely achieve distributed type of tuning. For example, we have seen in
Sect. 5.2.2 that physical rotation of CP elements can be used to directly manipulate
the phase shift of the unit cell. Mounting such cells on micro-motors, for example,
can be used to change the phase shifter. Micro-fluidic devices generally have been
used to move highly conductive liquid “slugs” of galistan and other materials in the
5 Reflectarrays 167

vicinity of resonant reflectarray unit cells to change the resonance characteristics

and induce a phase shift.

5.4.2 Realizing Phase-Tunable Reflectarray Unit Cells

As in the case of fixed reflectarray cells, reconfigurable unit cells can be classified
according to the same schemes used in Sect. 5.2.2. Examples of the application of
these techniques are shown via examples in Fig. 5.9 and discussed in detail in the
sections below. With each technique, varying degrees of phase control can be
achieved depending on the underlying technology. For example, devices with a
discrete number of states (e.g. those relying on switches) may only have limited
phase resolution. While devices with as low as 1-bit (2-state) phase control have
been successfully realized, designers must be mindful in the tradeoff that exists
between the effective number of bits in the phase shifter, and the corresponding
impact on the antenna pattern, with particular attention to the gain and sidelobe
levels [22]. For example, phase quantization at the 1-bit level generally leads to a
3-4 dB reduction in gain. The loss is reduced to less than 1 dB if 2-bit control is
used, and less than 0.2 dB for 3-bit control.

5.4.2.1 Tuned-Resonator Cells

Tuned-resonator cells can be implemented using the widest range of enabling

technologies discussed in the previous section, since perturbing the resonant fre-
quency of a cell can be realized by virtually any technique. Beginning with lumped
components, tunable reactances can easily be integrated with resonators, and the

(a) Varactor-tuned res- (b) Tuned stub [24] (c) Sequential rotation [25]
onator [23]

Fig. 5.9 Reconfigurable reflectarray element examples

168 S.V. Hum

reactance electronically manipulated to produce the necessary changes in resonant

frequency and corresponding phase shift in the scattered field. An example of this
type of cell is illustrated conceptually in Fig. 5.9a, where a pair of varactor diodes is
used to change the resonant frequency of a patch antenna. Through the proper
coupling of the tunable reactances to the patch, very large phase ranges close to
360 can be achieved. Furthermore, MEMS-tunable reactances such as MEMS
capacitors can also replace varactor diodes to decrease losses and improve linearity,
while maintaining a high effective number of bits.
Tunable reactances integrated with reflectarray elements change the resonant
frequency of the resonator by changing the effective length of the resonator.
Dual-LP variations are also possible [26], and even photo-induced plasmas have
been used in place of switches [27]. To mitigate the loss, linearity, and power
consumption issues with these approaches, numerous examples of RF-MEMS
switch-controlled reflectarray elements have emerged in previous years which
coupled switches and resonators together to achieve phase control [28, 29]. In
general, switch-based designs are often preferred, since they are readily compatible
with control circuits and do not need continuously variable voltages to tune the
phase.
Distributed tuning techniques also have the most impact on tuned-resonator
cells, since small changes in constitutive parameters can be translated to large
changes in the scattered phase with proper designs of the unit cell. Tunable
dielectrics have the most potential in this regard. For example, liquid crystal ma-
terials employing LC molecules in the nematic state possess an anisotropic per-
mittivity tensor which can be tuned between k  ? : Placing a resonator such as a
patch [30–32] on such a material can lead to pronounced changes in the phase as the
magnitude of the DC or AC biasing field is varied. Similarly, ferroelectric films also
produce changes in the permittivity tensor in the presence of an applied biasing field
and can also be used as substrate layers to produce phase shifters and phase shifting
reflectarray unit cells [33].

5.4.2.2 Guided-Wave Cells

Guided-wave reconfigurable unit cells work very similarly to their fixed counter-
parts. Instead of coupling to a delay line, the scatterer is coupled to a tunable
guided-wave phase shifter. The phase shifter must be a one-port reflection-type
phase shifter, so very often tunable delay lines are used. One such as in the example
shown in Fig. 5.9b, where a rectangular patch is aperture-coupled to a continuously
tunable delay line. Switched-delay lines are also used to produce a more wideband
response from the reflectarray. PIN diode-based versions have been realized [34,
35]. Minimizing the loss of the delay lines often means that MEMS switches are
preferable for this purpose, and several experimental demonstrations have been
produced [36, 37]. Since reconfigurable capabilities add cost to each unit cell, a
5 Reflectarrays 169

useful strategy that can be employed is to share the same switched-delay line
between a group (or sub-array) of elements.
Most guided-wave reconfigurable reflectarray elements are co-polarized.
However, cross-polarized (polarization-twisting) elements open up the possibility
of using the two-port microwave and mm-wave circuits to interconnect the two
polarization ports, as shown in Fig. 5.7d. An important function that can be added
to reflectarrays in this configuration is amplification. This function is traditionally
difficult to add to co-polarized arrays, because of two major challenges: (1) the need
for a one-port reflection-mode amplifier, which is difficult to design and make
stable; and (2) a fundamental requirement of isolation between the incident and
reflected wave ports. Cross-polarized designs are more amenable to the introduction
of amplification since two-port amplifiers are readily available, and the
cross-polarization discrimination of the element can be made very high. As a result,
a number of interesting and truly “active” designs (i.e. those with a reflection
coefficient magnitude larger than 1) have emerged: those based on amplifiers alone
(fixed-beam reflectarrays) [38, 39] and those combining amplification and phase
shifting [40, 41]. Such designs offer advantages in transmitters where the spatial
combining feature of reflectarrays mitigates losses normally present in transmission
line manifolds. This advantage is more compelling at mm- and sub-mm wave
frequencies, where the need for power combining is stronger and the losses in these
networks are most pronounced.

5.4.2.3 Sequential Rotation Cells

Finally, reconfigurable variations on the sequential rotation unit cell introduced in

Sect. 5.2.2 are a straightforward extension of the idea. The cell can be physically
rotated using an actuator such as a micro-motor. Novel MEMS micro-motors have
been realized for this purpose [42]. However, a slightly easier approach is to
electromagnetically rotate the element through the use of RF switches, which is the
approach illustrated in Fig. 5.9c. Here, a series of switches are placed along the
circumference of a slot, and a pair of switches diametrically opposite one another
are actuated to produce the rotation effect shown earlier in Fig. 5.8. This concept
has been successfully deployed in a number diode-based [25] and MEMS-based
demonstrators [16, 43].
A recent innovation in this area is to make dual-polarization CP reflectarray
cells. That is, the reflectarray cell produces a unique phase shift for the left-handed
component and another unique phase shift for the right-handed component. Such
reflectarrays can be useful in applications utilizing circular polarization, such as
satellite systems. This can be achieved by stacking two circular polarization
selective surfaces (CPSSs) on top of each other: one layer which is transparent to
one polarization type (e.g. RHCP) while being reflective to the other, and a second
is having the reverse behaviour. The first layer is structured to reflect and manip-
ulate the phase of the LHCP signal through the principle of sequential rotation,
while an RHCP signal incident upon this layer passes through unaffected to the
170 S.V. Hum

second layer. The second layer also modifies the reflection phase of the RHCP
signal through sequential rotation; it then passes unaffected through the first layer.
This concept has been successfully demonstrated using a unique combination of
CPSSs and can potentially be realized in a switched, reconfigurable form [44].

5.4.3 Modelling Considerations

There are several frameworks that assist in the modelling of reconfigurable

reflectarray elements. Both apply to the case where lumped-elements such as
switches, tunable reactances, and other components are introduced in the unit cell.
The first is the concept of a mixed-mode generalized scattering matrix, which
assists in predicting the overall scattering behaviour of the unit cell without
resorting to exhaustive full simulations with the lumped components in different
states. The second is an equivalent circuit model that can be used to predict the
scattering behaviour of simple unit cells even before the numerical simulation stage.
Both are described next.

5.4.3.1 Mixed-Mode Generalized Scattering Matrix

The GSM representation of reflectarray unit cells can be extended to the case where
lumped-element components are introduced in the reflectarray. For example, the
capacitively loaded cell shown in Fig. 5.9a employs two varactor diodes as tunable
capacitances. A straightforward way to determine the S-curve for such an element is
to introduce a reactance in the electromagnetic description of the unit cell (e.g. in
the impedance matrix for MoM formulations), and sweep the capacitance over a
series of simulations run to trace out the S-curve. While this process is correct, it
unnecessarily lengthens the simulation time of structure, since the underlying
scatterer is linear and hence obeys network theory. Furthermore, when the scatterer
involves a large number of reconfigurable elements, the number of configurations to
simulate quickly can become unmanageable. For example, a cell equipped with N
switches has 2N unique states.
A much more computationally efficient approach is to extract a mixed-mode
GSM from the unit cell. This approach is illustrated in Fig. 5.10a, which considers a
unit cell which has L lumped-elements introduced in its structure. Here, we assume
that only the fundamental Floquet modes are of interest, and hence we only have
two Floquet ports at the input to the periodic waveguide. A series of L discrete ports
3; 4; . . .k; . . .N; N ¼ L þ 2 are introduced at the locations of the lumped-element
component loading points.
Using this description, we then describe the overall scattering response of the
unit cell in terms of the two spatial ports and L lumped ports using
5 Reflectarrays 171

z
z
b
a

Floquet port
Floquet ports 1,2

T E|T M T E|T M
b00 a00
w
b
a
3 J N l
k r y
r y
x x
(a) Mixed-mode GSM of a reflectarray unit cell (b) Dipole in driven mode

Fig. 5.10 Scattering scenarios for modelling

    
b12 SA SB a12
¼ ; ð5:32Þ
b3N SC SD a3N

where we see we now have a mix of Floquet ports ð1; 2Þ and lumped-element ports
ð3; 4; . . .k; . . .NÞ associated with the incident and reflected waves. Each port has its
own associated port impedance Z0k . For the Floquet ports, these impedances are the
associated TE and TM wave impedances associated with ports 1 and 2, while for
the lumped ports, an arbitrary reference impedance can be employed (usually an
impedance of 50 X is used.) Note that the incident and reflected waves are nor-
malized according to the conventional power wave definitions, i.e. ak and bk are
normalized by the square root of the port impedance associated with the kth port.
When actual components are introduced to load each port, the associated
reflection coefficients Ck are determined for each state the lumped components can
be in. A reflection coefficient vector is then defined,

C ¼ diagðC3 ; C4 ; . . .; Ck ; . . .; CL Þ ð5:33Þ

which relates a3N and b3N through

a3N ¼ Cb3N : ð5:34Þ

Imposing this constraint on Eq. (5.32) and solving yields

h i
b12 ¼ SA þ SB CðI  SD CÞ1 SC a12 ; ð5:35Þ
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
SM
172 S.V. Hum

where I is the identity matrix and SM is the resulting 2  2 GSM resulting from the
loading of the reflectarray unit cell in a given state of lumped-element configura-
tions C. This approach to modelling is very powerful, because it only requires the
determination of the initial mixed-mode scattering matrix shown in Eq. (5.32) by
electromagnetic simulations or other means. Once this matrix is known, SM can be
evaluated algebraically using Eq. (5.35). Usually, C can be easily computed by
analytical means or using circuit simulators in the case of complicated loading
networks (e.g. amplifiers). This technique opens the possibility up of optimizing the
unit cell when many load combinations are possible, for example, using genetic
algorithms or particle swarm optimization.

5.4.3.2 Equivalent Circuit Modelling of Reconfigurable

Reflectarray Elements

For tuned-resonator types of reconfigurable reflectarrays, a further tool one can use
in determining a “first-cut” reflectarray design is an equivalent circuit model
(ECM) for the element. Such tools have an advantage of being faster to evaluate
during the initial design stages, when a design is trying to determine the optimal
size and loading of a potential reflectarray element. Here, we employ a
Floquet-based framework for deriving an equivalent circuit model that is versatile
and easily to evaluate [45]. In particular, unlike many other models proposed in the
literature, this model does not require full-wave simulations to determine compo-
nent values, and can treat a variety of reconfigurable element cases.
The ECM is derived by first considering the behaviour of a dipole scatterer in the
driven mode, in an infinite array environment. An example of a dipole oriented
along the x-direction of dimensions l  w is shown in Fig. 5.10b. The dipole is
initially treated in the driven mode where an excitation is placed across the ter-
minals shown. Infinite dipole arrays have been considered at length in the literature.
The dominant current component on the dipole is expanded in terms of a Floquet
series as
XX
~
Iðx; yÞ ¼ ^x ~f ðkxmn ; kymn Þejkxmn xjkymn y ; ð5:36Þ
m n

where ~f is the Fourier transform of the x-directed current distribution f ðx; yÞ on the
dipole. In general, the current distribution on the patch is assumed to be separable
such that f ðx; yÞ ¼ gðxÞhðyÞ. Using a common formulation, the active input
impedance seen looking into the dipole for a particular scan angle ðh; /Þ can
subsequently be derived as [46]
" #
l2 4 X X kymn
2 2
kxmn G2 H 2
Z ðkx0 ; ky0 Þ ¼
FL
þ TM 2 mn mn þ2
; ð5:37Þ
2 TE
ab p m n ymn ymn k0  kzmn
5 Reflectarrays 173

where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ
kzmn ¼ k02  kxmn 2  kymn 2 ð5:38Þ

is the z-component of the wavenumber in the z [ 0 region, and Gmn and Hmn are the
Fourier transforms of the current distribution in the x- and y-directions evaluated at
the corresponding wavenumbers kxmn and kymn , respectively. Meanwhile, the
quantities yTE TM
mn and ymn represent the equivalent admittance experienced by a TE zmn
and TM zmn current source at the air-dielectric interface, respectively. This admit-
TE þ jTM þ
tance is the sum of the modal admittance in the z [ 0 region, Ymn , and the
input admittance of the grounded dielectric slab,


jBTEjTM
mn ¼ jYmn
TEjTM
cotðkzmn hÞ: ð5:39Þ

TEjTM
Ymn are the TE zmn j TM zmn model admittances for the z\0 region. Hence,

TE þ jTM þ
yTEjTM
mn ¼ Ymn þ jBTEjTM
mn : ð5:40Þ

The modal admittances are given by

þ
TE þ kzmn k
Ymn ¼ ; Ymn
TE
¼ zmn ð5:41aÞ
xl0 xl0

TM þ x 0 TM x 0 r
Ymn ¼ þ
; Ymn ¼  : ð5:41bÞ
kzmn kzmn

Equation (5.44) can be reorganized to explicitly show the contributions from the
TE and TM modes,
XX 1 XX
TM 1
Z FL ðkx0 ; ky0 Þ ¼ TE
Tmn þ Tmn ; ð5:42Þ
m n
yTE
mn m n
yTM
mn

where
2
l2 4 kymn
TE
Tmn ¼ G2 H 2
þ 2 mn mn
ð5:43Þ
ab p2 k02  kzmn

and

l2 4 kxmn2
TM
Tmn ¼ G2 H 2 :
þ 2 mn mn
ð5:44Þ
ab p k0  kzmn
2 2
174 S.V. Hum

(ZL )

A B
ZF L
   
TE
T00 TM
T00 TE
T01 TM
T01

TE TM TE TM
jB00 jB00 jB01 jB01

T E+ T M+ T E+ T M+
Y00 Y00 Y01 Y01

TE
(Zin,00 )

Fig. 5.11 Equivalent circuit for a dipole an infinite array environment. In the driven mode
TE þ
(Fig. 5.10b), ZL is absent, Y00 is present, and ZFL is measured. In the scattering mode, ZL is
TE þ TE
present, Y00 is absent, and Zin;00 is measured

A close examination of (5.37) reveals that the input impedance Z FL to the dipole
is an infinite series summation of impedances 1=yTEjTM scaled by transformers with
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi mn
TE=TM
turns ratios Tmn . This suggests the equivalent circuit shown in Fig. 5.11,
TE þ
where ZL is absent, Y00 is present, and ZFL is measured.
This circuit can easily be turned around to treat the scattering scenario whereby
the dipole in Fig. 5.10 is excited externally by a Floquet mode and a load places
across the dipole terminals. For example, consider excitation of the TE00 mode. In
TE þ TE
Fig. 5.11, ZL is now introduced, Y00 is removed, and Zin;00 is measured as shown.
The input impedance seen looking upward across the interface AB is
XX 1 XX
0
TM 1
TE
Z00 ¼ ZL þ TE
Tmn TE
þ Tmn : ð5:45Þ
m6¼0 n6¼0
ymn m n
yTM
mn

Then the input impedance of the desired TE00 Floquet mode can be calculated
as
"
0 1 #1
TE
Z00  

TE
Zin;00 ¼ TE jY00
TE
cot kz00 h : ð5:46Þ
T00

Knowing the input impedance of the TE00 mode, the associated reflection
coefficient and corresponding S-curve can be found.
5 Reflectarrays 175

The ECM is straightforward to evaluate. To populate the ECM, the transformer

TEjTM
turns ratios Tmn must be found, which in turn relies on a reasonably accurate
description of the current distribution on the scatterer when illuminated by the
co-polarized field. For the case of a rectangular patch scatterer in which width w in
Fig. 5.10b becomes comparable to the cell width b, the patch current distribution is
specified as follows. First, in the x-direction, the current distribution is described by
a half-wave sinusoid along a lossless TEM transmission line with effective phase
constant beff , characteristic impedance Zc and fringing capacitance Cf associated
with common patch antennas. These are calculated using the standard quasi-static
formulae for microstrip lines. The capacitance Cf loads each end of the transmission
line. This allows the current distribution in the x-direction to be found as
Vi
gðxÞ ¼ Ii coshðjbeff xÞ  sinhðjbeff xÞ; 0  jxj  L=2; ð5:47Þ
Zc

where Vi is an arbitrary applied voltage to the circuit and Ii is the resulting input
current at the input to the transmission line. In the y-direction, the patch currents are
non-uniform with a singularity near the edges of the patch. In fact, the current
distribution, which is determined by the tangential magnetic fields along the patch,
is the dual of the electric field distribution that occurs in the gaps of the comple-
mentary structure, a metal strip grating. Therefore, for a TE-polarized incident
wave, the surface current density on the patch is equal to
Cejby
hðyÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5:48Þ
ðW=2Þ2  y2

where C is a constant. Gmn and Hmn is readily found by numerically taking the
Fourier transforms of Eqs. (5.47) and (5.48), after which the turns ratios can be
found from Eqs. (5.43) and (5.44). The remainder of the components in the ECM of
Fig. 5.11 are found analytically from Eqs. (5.41a) and (5.39).

5.5 Design Examples

In this section, we consider two design examples for reflectarrays designed at 10

GHz. The first is a fixed reflectarray based on varying the shape of a square patch,
while the second is reconfigurable reflectarray, implemented using the
tuned-capacitor method outlined in the previous section. In both cases the scatterer
is single patch-like scatterer printed on a grounded dielectric substrate. The key
parameters of the reflectarray design are summarized as follows. The reflectarray is
realized on a 1-mm thick substrate with
r ¼ 3:02: It consists of 625 elements
arranged in a 25  25 array with a cell period of 15 mm. The f =D ratio is 1, the
reflectarray is linearly polarized, and the desired beam direction is h ¼ 30 ; / ¼ 0 .
For the examples considered here, conductor and dielectric losses are ignored.
176 S.V. Hum

5.5.1 Fixed Reflectarray

The first reflectarray example is a fixed-beam reflectarray realized using

patch-based scatterers of varying size, corresponding to the tuned-resonator
approach described in Sect. 5.2.2. For simplicity, the patch is kept square in size,
though dipole-shaped and rectangular-shaped patches can also be used. The size of
the patch is used to manipulate the phase of the fields co-polarized with a linearly
polarized incident field.
To create the S-curve for the reflectarray, both analytical and numerical
approaches can be employed. The equivalent circuit model described in Sect. 5.4.3.
can be evaluated for a variety of patch dimensions to produce the desired S-curve.
Since there is no lumped-element loading present, the load ZL is simply set to 0 X
(a short circuit), and the model evaluated for a variety of patch dimensions
according to a parametric sweep. The ECM is evaluated using 60 TE and 60 TM
spatial harmonics and results in the S-curve shown in Fig. 5.12. For comparison, a
spectral domain method of moments code employing Rao-Wilton-Glisson basis
functions is used to compute the GSM of the same cell, and the results are overlaid
in the same plot. It can be seen that the results of both methods align well, and that
the ECM can be used to accurately predict the response of the unit cell.
The analysis shows that the cell possesses approximately 330 of phase range,
which is adequate for forming a beam. The missing 30 of phase range has only a
very mild impact on the sidelobe levels. The moderate slope in middle of the phase
curve is a good feature, since high slopes increase the sensitivity of the design to
fabrication errors. The slope can be reduced at the expense of phase range by, for
example, increasing the substrate height, and hence the designer is afforded some
freedom in the design process.
Next, the full array is designed based on the reflectarray unit cell. The required
phases to collimate the fields from the feed and produce the beam in the required
direction are computed using Eq. (5.4). The required phases are then reconciled
against the phase curve and mapped to the corresponding patch lengths required at

Fig. 5.12 Phase curve for the 200

fixed reflectarray unit cell as ECM
150 MoM
predicted by the ECM and
MoM techniques 100
arg(Γ) [deg]

50

-50

-100

-150

-200
2 4 6 8 10 12 14
Patch length [mm]
5 Reflectarrays 177

y
40
AF
30 MoM

20

Directivity [dB]
10
0

x -10
-20
-30
-40
-100 -80 -60 -40 -20 0 20 40 60 80 100
θ [deg]
(a) Phase mask (b) Directivity

Fig. 5.13 Fixed reflectarray example

10 GHz via a lookup table, and a corresponding mask for the reflectarray is shown
in Fig. 5.13a. Note that a variable in the mapping is the arbitrary phase offset on the
right-hand side of Eq. (5.4), which can be used to determine where the phase-wraps
occur in the reflectarray element distribution. Further, this can be used as a degree
of freedom in the case where the reflectarray element has loss, in order to best
distribute the lossy (lower amplitude) cells across the reflectarray to yield the best
performance according to the required specifications. This tends to be more
meaningful in when using reconfigurable reflectarray elements, which can have
more loss compared to fixed elements in general.
Once the phase distribution across the reflectarray has been established, the feed
is usually optimized to yield a desired taper level across the array. With the given
f =D ratio and a 12:5-wavelength reflectarray diameter, a cosq ðhÞ feed with q ¼ 6 is
chosen to produce an approximate 12 dB edge taper in the principal planes of the
reflectarray. The feed is chosen to be linearly polarized in the x-direction. Using
Eq. (5.21), the incident field levels at the centre of each reflectarray element can be
computed and used as inputs to the local reflected field calculations via the GSM in
Eq. (5.17).
The radiated fields from the reflectarray are determined from Eq. (5.26), and the
corresponding directivity from the array is computed. Loss terms such as spillover
loss, surface wave losses, reflection efficiency, feed blockage, etc., have not been
considered here, since a simplified model of the reflectarray and feed is being
analyzed, but these calculations can be included in order to determine the overall
gain of the reflectarray. The resulting directivity of the array is plotted in Fig. 5.13b,
which clearly shows a pencil beam pointing in the h ¼ 30 ; / ¼ 0 direction.
To complete the analysis, the results are compared with a method of moments
simulation of the same reflectarray, which more accurately accounts for mutual
coupling effects between the elements. The feed is assumed to be completely
decoupled from the reflectarray. The simulation employs an integral formulation
involving the Green’s function for a radiating element on a grounded dielectric slab,
178 S.V. Hum

200 40
ECM, 10 GHz AF
150 MoM, 10 GHz 30 MoM
ECM, 9.75 GHz
100 20

Directivity [dB]
MoM, 9.75 GHz
arg(Γ) [deg]

ECM, 10.25 GHz

50 MoM, 10.25 GHz 10
0 0
-50 -10
-100 -20
-150 -30
-200 -40
10 100 1000 10000 -100 -80 -60 -40 -20 0 20 40 60 80 100
Capacitance [pF/m] θ [deg]
(a) Phase curve of reconfigurable unit cell as predicted by (b) Directivity pattern of the reconfigurable reflectarray as
the ECM and MoM techniques predicted by the AF and MoM techniques

Fig. 5.14 Reconfigurable reflectarray example

and as such, the finite nature of the substrate and ground plane in a real reflectarray
is not accounted for. The simulation results are overlaid with the array factor results
in Fig. 5.14b. We see that the AF technique is generally quite good at predicting the
overall shape of the beam, though the more rigorous MoM analysis reveals stray
scattering that increases the overall sidelobe levels (and correspondingly reduces
the peak directivity). Overall, very good correlation is obtained between the two
methods.

5.5.2 Reconfigurable Reflectarray

The second reflectarray example is based on a unit cell whose phase can be tuned
dynamically using a tunable capacitance such as a varactor diode. Modelled after a
successful reconfigurable reflectarray implementation and demonstration [47], the
unit cell is based on the design shown in Fig. 5.9a whereby two halves of a patch
are serially loaded by a variable capacitance. For simplicity in the analysis, the
capacitance is assumed to uniformly load the gap between the two patch-halves, but
in reality this can be replaced with a number of varactor diodes placed across the
gap (nominally two). In this situation, the equivalent circuit model should be
modified to account for the gap capacitance and current crowding inductance in the
vicinity of the varactor diodes, by incorporating these parasitics into the diode
model [47]. The parasitics intrinsic to the diode itself can also be included.
To determine the patch dimensions, the equivalent circuit model described in
Sect. 5.4.3 can be tested for a variety of patch dimensions to produce the desired
S-curve. For the example, the patch dimensions here are chosen to be 8:3 mm 8:3
mm. The ECM is evaluated using the same number of spatial harmonics as before
and results in the S-curve shown in Fig. 5.14a. For comparison, results from the
spectral domain MoM codes are overlaid in the same plot. It can be seen that the
5 Reflectarrays 179

results of both methods align well, and that the ECM can be used to accurately
predict the response of the unit cell. Also shown are the corresponding S-curves at
adjacent frequencies, to illustrate the frequency-dependent behaviour of the unit
cell. We see that as the frequency is changed, the phase curves deviate from the
nominal 10 GHz case, which introduces phase errors and eventual pattern and
directivity degradation as the difference between the operating and actual fre-
quencies becomes larger. However, for a narrowband reflectarray, the cell design is
sufficient.
Using an analogous process as for the fixed reflectarray case, the beam is syn-
thesized by mapping the required phase shifts as required by Eq. (5.4) to the
capacitances at 10 GHz. In practice, the required capacitances can be further
mapped to control voltages in the case that a varactor diode is used to implement
the variable capacitance, using the corresponding CðVÞ relation for the diode.
Results of both the array factor calculation and the MoM analysis are shown in
Fig. 5.14b, again illustrating good correlation between the two methods.
The practical design of reconfigurable reflectarrays involves an additional step
whereby a suitable control network needs to be introduced for managing the
reconfigurable components of the antenna. The realization of the bias network is
highly implementation-specific, and while an exhaustive discussion is not possible
here, some best practices can be deduced from studying the example here. For the
diode-based cell employed in the example, a local biasing network needs to be
introduced similar to the scheme shown in Fig. 5.9a. In this case, the bias network
consists of a pair of RF chokes, in the form of self-resonant inductors, and an RF
short composed of a self-resonant capacitor, through which a DC bias voltage is
applied. The bias network serves to form both a high-impedance load to prevent RF
currents on the patch from travelling to the bias wires, and a low-impedance load to
short out such currents should they make it to the interface with the wires. This is
not the only way to realize bias networks; distributed components such as stubs can
also be used in place of lumped components. Furthermore, in the case of compo-
nents with zero DC current draw (such as varactor diodes), resistors serve equally
well in place of inductive RF chokes, and are less frequency-dependent.
The presence of the bias network can also introduce undesired scattering,
especially if the feeding traces are long. Efforts can be undertaken to minimize this
scattering. The use of a symmetric bias network, such as that shown in Fig. 5.9a,
maintains low cross-polarization. In the case of reflectarrays, the bias network itself
can often be routed to be behind the ground plane of the reflectarray, effectively
shielding it from incoming radiation. However, if the bias network must co-exist
near or on the same layer as the scatterers, resistively loading the lines has been
shown to be effective, especially in transmitarrays where bias lines cannot be
conveniently tucked behind ground planes. Resistive lines can be realized by
depositing resistive material on the substrate and patterning it into the required bias
lines, or by periodically loading metallic lines with resistors.
180 S.V. Hum

5.6 Practical Examples

To round out the design examples presented in Sect. 5.5, this section presents some
samples of practical implementations taken from the literature. The samples illus-
trate each of the element design approaches discussed in Sect. 5.2.2. Emphasis is
placed on full-scale experimental demonstrations of reconfigurable reflectarrays,
since they exemplify the operating principles of reflectarrays while simultaneously
illustrating useful beam-scanning capabilities.
The performance of reflectarray designs based on lumped-element tuning is
strongly affected by packaging and parasitic issues at millimetre wave and sub-mm
wave frequencies. As such, distributed tuning based on the tunable resonator
approach mitigates many of these issues. As discussed in Sect. 5.4.2, liquid crystal
materials possess a tunable bulk permittivity tensor which can be exploited to tune
the resonant frequency of microstrip elements. An example of a recent F-band
design is shown in Fig. 5.15 [48]. As illustrated in Fig. 5.15a, the unit cell employs
a multi-resonant design based three unequal length parallel dipoles placed within
the unit cell to extend the phase bandwidth of the unit cell around 100 GHz. The
unit cells include a bias line to allow a low-frequency biasing field to be applied to
the LC substrate beneath the dipoles, allowing the permittivity tensor to be
manipulated. LC materials with D
¼ 4:6: were employed in the design. A 52  54
element offset-fed reflectarray was fabricated with one-dimensional scanning
capabilities. Some examples characteristics of the radiation patterns for different
scan angles are shown in Fig. 5.15b. Overall, the work achieved a very wide
scanning range and an aperture efficiency of 18:5%. Most of the inefficiency is
attributable to absorption losses in the LC material, but considering the frequency of
operation, the results are quite impressive.
The guided-wave approach has also proved useful in the millimetre wave
frequency range. A 60 GHz design employing PIN diodes is shown in Fig. 5.16
[34]. As shown in Fig. 5.16a, the unit cell employs a stub-based phase shifter
design, whereby the length of the stub is controlled by the state of the switch.

(a) Unit cell and array layout (b) Measured radiation pattern for elevation scanning

Fig. 5.15 Liquid-crystal based reflectarray operating at 100 GHz [48]

5 Reflectarrays 181

Biasing of the diode is accomplished by placing control circuits behind the

reflectarray ground plane, as shown in Fig. 5.16b. This implements one-bit phase
control, providing the reflectarray elements with 0 or 180 of phase shifting
capability. A 160  160 array with independent control of each element was fab-
ricated and characterized at 60:25 GHz. Typical beam patterns illustrating the
beam-scanning capabilities of the array are shown in Fig. 5.16c, d. The achieved
aperture efficiency of the array is about 9% (10:5 dB). The bulk of the losses
originates from phase quantization losses due 1-bit control (gphase 4 dB) and
diode losses (gohmic 4 dB). These underscore the balance designers try to
achieve between complex, high-performance elements (with high phase resolution
and low loss) and low-complexity elements which can be implemented at a much
lower cost. Keeping costs down is especially important in large reconfigurable
apertures such as the one in this work.
The final example considered employs a reconfigurable element utilizing the
sequential rotation concept. CP designs are important for satellite applications,
where an additional and imposing requirement is operation at two frequencies
(transmit and receive). A dual-band K/Ka-band reflectarray design has been
developed working at 24:4 and 35:5 GHz [16]. The reflectarray utilizes a unit cell
whereby ring-slots of two different radii are interleaved to form the reflectarray
surface, as shown in Fig. 5.17a. Each ring-slot contains 6 MEMS switches that can
be actuated in pairs to achieve electronic rotation of the cell, allowing for 3 phase

(a) Unit cell top view (b) Unit cell side view

(c) Patterns showing scanning in azimuth (d) Patterns showing scanning in elevation
Fig. 5.16 Switched-stub reflectarray operating at 60 GHz [34]
182 S.V. Hum

(b) Normalized pattern at 24.4 GHz for MEMS switches in

broadside state (left) and scanned state (right). Measure-
ments are in orange and simulations in blue.

(a) Unit cell design

(c) Normalized pattern at 35.5 GHz for MEMS switches

in broadside state (left) and scanned state (right). Mea-
surements are in purple and simulations in red.

Fig. 5.17 Dual-band reflectarray based on sequential rotation [16]

states per ring-slot. A prototype was realized with actual MEMS devices (albeit in a
frozen state), and the achieved co-polarized radiation patterns are shown for each
frequency in Fig. 5.17b, c. A grating lobe appears in the 35:5 GHz results since an
interleaved design results in a fixed 6:86-mm cell period that is used for both
frequency bands. This results in an electrical periodicity of 0:56k0 at 24:4 GHz and
0:86k0 at 35:5 GHz, and highlights the challenges of designing dual-band reflec-
tarrays. Nevertheless, this is an acceptable compromise and this example illustrates
the potential of MEMS devices in future generations of reflectarrays.

5.7 Applications

Reflectarrays are promising candidates as replacements for reflector antennas in a

number of current and next-generation applications. As low-profile, high-gain
antennas with potential adaptive beam-forming capabilities, they offer a number
compelling advantages, especially at millimetre and sub-mm wave frequencies
where array antennas become increasingly lossy and complex. Reflectarrays offer
the best features of both reflector antennas and arrays, providing the gain and
efficiency of reflectors while offering the flexibility and potentially adaptability of
an antenna array. The economical nature of reflectarrays can also make them
attractive in applications where there are high setup costs (e.g. as replacements for
shaped reflectors, which require costly moulds) or a large amount of hardware at the
5 Reflectarrays 183

(a) (b)

Fig. 5.18 Reflectarray application examples

unit cell level (e.g. in phased arrays, where a large amount of RF hardware is
duplicated at an element or sub-array level).
As high-gain antennas, reflectarrays are potential antenna candidates in RADAR
applications. Perhaps one of the most successful applications of reflectarrays has
been in automotive RADAR systems at 77 GHz, where the technology has been
successfully commercialized into compact, mechanically steered RADAR units
[49]. Folded reflectarrays offer further advantages here by allowing the radar unit to
be reduced in size while maintaining a narrow-beam. An example of a commercial
reflectarray RADAR is shown in Fig. 5.18a.
Reflectarrays have always held the promise of being replacements for reflector
antennas in space applications. Indeed, microstrip reflectarrays had their humble
beginnings as a research topic at the Jet Propulsion Laboratory in the 1990s where it
was hoped that so-called “inflatable” reflectarrays could be deployed on satellites
and other spacecraft to reduce launch volume and mass. Fast-forwarding to the
present day, this dream has been fulfilled with the launch of the NASA ISARA
(Integrated Solar Array and Reflectarray Antenna) cubesat in 2016, which was the
first demonstration in space of a multi-faceted reflectarray on the back side of a
solar array that deploys in space after launch [50]. This concept, shown in Fig. 5.18
(b), has become a central feature on NASA’s upcoming MarCO (Mars Cube One)
mission to Mars, which uses an X-band reflectarray as a bent-pipe telecom link to
transmit entry, descent, and landing (EDL) data to Earth [51]. The MarCO
reflectarray is based on the square patch element discussed in the example, and is a
deployable design.
Many other features of reflectarrays have been highlighted as advantageous for
space applications. Foremost of these is the ease at which the beam can be shaped,
opening up the possibility of shaped coverage patterns for satellites [52, 53]. As
mentioned earlier, the ability to use reflectarrays in such applications could be a
potential cost-saving measure as the need for fabricating shaped-reflector moulds is
eliminated. Reflectarrays have also been identified as promising candidates in a
number of remote-sensing applications. For example, reflectarrays have been pro-
posed as an interferometric instrument for measuring ocean height [54], as a
replacement for cylindrical reflectors in the Titan cloud precipitation radar and
altimeter [55], and for use as antennas in synthetic aperture RADAR (SAR) [56].
184 S.V. Hum

It is possible to realize reflectarrays that are optically transparent, either from

materials that are optically transparent, such as transparent conducting oxides [57],
or from elements that are only partially opaque over the extent of the unit cell [58].
It is possible to overlay such elements over solar panels on spacecraft to allow the
large surface area to be used for both solar power collection and as a large aperture
for communications. This enables a number of additional operational modes. For
spacecraft missions at the L2 Lagrangian point between the Earth and the Sun,
where the spacecraft maintains the same relative position between these two bodies,
pointing the solar panel and communications array in the same direction is a
necessary requirement that such an antenna fulfills. In other satellite missions not
requiring full-time communication with the ground station, the satellite’s attitude
control system could be used to point the solar array/reflectarray at the Earth only
when it is in the range of the ground station, and pointed at the sun the rest of the
time.
Terrestrial communication systems can also greatly benefit from reflectarray
antennas. Reflectarrays have been proposed for local multi-point distribution
(LMDS) communication systems at mm-wave frequencies, since they offer an easy
way to shape the beam according to the requirements of the base station [59]. An
interesting capability of reflectarrays is their ability to manipulate the propagation
characteristics of wireless channels. For example, the frequency selective nature has
been shown to improve the characteristics of non-line-of-sight channels when
reflectarrays are introduced as passive reflectors in the environment [60]. A similar
concept can be employed to control radio wave propagation across different bands
within buildings [61].
Finally, as reflectarrays move into the upper-mm-wave and sub-mm wave ran-
ges, possibilities emerge for imaging, particularly among reconfigurable reflectar-
rays. Large 1-bit electronically controlled reflectarrays have been proposed for
imaging at 60 GHz [34], and recent advancements in liquid crystal-based devices
could enable imaging and communications within the F-band [62].

5.8 Future Trends

The wide scope and variety of applications discussed in Sect. 5.7 illustrates the
progress reflectarrays is making in industrial and scientific fields. Still, the field is
germane for future research, especially in the mm-wave and sub-mm wave fre-
quency range.
Bandwidth is a well-known limitation of reflectarrays, especially when they are
compared to the curved reflectors they aim to replace, which are very wideband.
Many reflectarray designs capitalize on resonances with the scatterer to achieve the
pronounced changes in phase about an operating point. Even guided-wave designs,
which can employ wideband phase-shifting devices, are ultimately constrained by
the bandwidth of the antennas they couple to. As a result, elementary reflectarray
designs are usually constrained to have bandwidths on the order of 10% or less.
5 Reflectarrays 185

Dual- and multi-band designs have been explored to increase the attractiveness of
reflectarrays, particularly in satellite applications, which separate uplink and
downlink frequencies. This can be achieved by integrating multiple resonators with
significantly differing resonant frequencies together [63, 64]. Alternatively, the
bandwidth of a single band can be increased by coupling resonators closely spaced
in frequency, such as by stacking patches together [65] or printing dipoles in close
proximity to each other in the same layer [32], which can be advantageous in
simplifying the layout of reconfigurable reflectarrays based on tunable materials
[62].
Guided-wave designs can have an advantage here provided the delay devices
(e.g. delay lines) can be integrated within the confines of the unit cell, and the
antenna portion engineered to be suitably wideband. However, planar wideband
antennas are difficult to realize, and the introduction of large delay lines can cause
secondary scattering effects. Recently, a class of TTD reflectarrays has emerged
that address these issues. The first is reflectarray design where the elements are
designed to emulate the response of a Bessel filter, which achieves as flat a group
delay as possible in the passband. Using planar scatterers to realize the required
inductive and capacitive elements of a Bessel filter, reflectarray designs with
bandwidths in excess of 67% and fidelity factors close to 1 can be achieved [66].
Another approach employs miniaturized-element FSSs to approximate a TTD
response by using multiple layers of sub-wavelength square patches, leading to a
40% fractional bandwidth. A key idea in both concepts is the use of sub-wavelength
elements, which in turn realizes something analogous to an impedance surface (or
stack of impedance surfaces), which has long been known to be linked to improving
reflectarray bandwidth [67]. Translating the bandwidths achieved from these
approaches to reconfigurable reflectarrays remains an open problem.
As frequencies increase to the mm-wave and especially sub-mm wave ranges,
the behaviour of metals moves into the plasmonic regime, necessitating the change
in the analysis of reflectarray elements at these frequencies [68–71]. In some cases,
it is preferable to move to dielectrics for realizing reflectarray elements [72]. While
the much shorter wavelengths entail significant novel practical considerations, the
operating principle is essentially the same as discussed in this chapter, at microwave
frequencies.
An equally important future research direction is the development of reconfig-
urable reflectarray technologies in the mm- and sub-mm wave ranges. For example,
lumped-element loading as a reconfiguration mechanism, has only been demon-
strated up to around 60 GHz for the case of PIN diodes, and 120 GHz in the case of
guided-wave-based MEMS devices [73]. Beyond this frequency, tunable material
platforms are required, since the electrical size and parasitics associated with
component-based tuning become too difficult to deal with. Liquid crystal technol-
ogy has been the most promising thus far, achieving practical operation in the
F-band [62]. Still, loss remains an important issue here and efforts are underway to
engineer LC materials with lower losses in the mm-wave range.
In the sub-mm wave range, graphene has shown significant potential as a tunable
material platform for realizing reconfigurable reflectarrays [74, 75]. Employing the
186 S.V. Hum

so-called graphene “field effect”, the phase of a unit cell at 1:3 THz can in principle
be controlled by manipulating the complex conductivity of a graphene sheet from a
localized bias voltage. Such novel devices may open the doors to reconfigurable
THz reflectarrays in the future.

References

1. D.G. Berry, R.G. Malech, W.A. Kennedy, The reflectarray antenna. IEEE Trans. Antennas
Propag. 11(6), 645–651 (1963)
2. R.E. Munson, H.A. Haddad, J.W. Hanlen, Microstrip reflectarray for satellite communication
and radar cross-section enhancement or reduction, US Patent 4,684,952 (1987)
3. J. Huang, Microstrip reflectarray, in 1991 Antennas Propagation Society International
Symposium Digests, vol. 2 (1991), pp. 612–615
4. R. Leberer, W. Menzel, A dual planar reflectarray with synthesized phase and amplitude
distribution. IEEE Trans. Antennas Propag. 53(11), 3534–3539 (2005)
5. D.M. Pozar, S.D. Targonski, R. Pokuls, A shaped-beam microstrip patch reflectarray. IEEE
Trans. Antennas Propag. 47(7), 1167–1173 (1999)
6. O.M. Bucci, G. Franceschetti, G. Mazzarella, G. Panariello, Intersection approach to array
pattern synthesis. IEE Proc. H Microw. Antennas Propag. 137(6), 349–357 (1990)
7. J.A. Zornoza, J.A. Encinar, Efficient phase-only synthesis of contoured-beam patterns for very
large reflectarrays. Int. J. RF Microw. Comput. Aided Eng. 14(5), 415–423 (2004)
8. A.K. Bhattacharyya, Projection matrix method for shaped beam synthesis in phased arrays
and reflectors. IEEE Trans. Antennas Propag. 55(3), 675–683 (2007)
9. P. Nayeri, F. Yang, A.Z. Elsherbeni, Design of single-feed reflectarray antennas with
asymmetric multiple beams using the particle swarm optimization method. IEEE Trans.
Antennas Propag. 61(9), 4598–4605 (2013)
10. M. Mussetta, P. Pirinoli, G. Dassano, R.E. Zich, M. Orefice, Experimental validation of a
genetically optimized microstrip reflectarray, in Antennas and Propagation Society
International Symposium, 2004, vol. 1. (IEEE, 2014), pp. 9–12
11. P. Nayeri, A.Z. Elsherbeni, and F. Yang, The analogy between offset configurations of
parabolic reflectors and reflectarrays, in 2015 IEEE International Symposium on Antennas
and Propagation USNC/URSI National Radio Science Meeting (2015), pp. 2127–2128
12. E. Almajali, D. McNamara, J. Shaker, M.R. Chaharmir, Derivation and validation of the basic
design equations for symmetric sub-reflectarrays. IEEE Trans. Antennas Propag. 60(5),
2336–2346 (2012)
13. C. Wan, J.A. Encinar, Efficient computation of generalized scattering matrix for analyzing
multilayered periodic structures. IEEE Trans. Antennas Propag. 43(11), 1233–1242 (1995)
14. M. Zhou, E. Jorgensen, O.S. Kim, S.B. Sorensen, P. Meincke, O. Breinbjerg, Accurate and
efficient analysis of printed reflectarrays with arbitrary elements using higher-order
hierarchical Legendre basis functions. IEEE Antennas Wirel. Propag. Lett. 11, 814–817
(2012)
15. H.R. Phelan, Spiraphase reflectarray for multitarget radar. Microw. J. 20, 67 (1977)
16. C. Guclu, J. Perruisseau-Carrier, O. Civi, Proof of concept of a dual-band circularly-polarized
RF MEMS beam-switching reflectarray. IEEE Trans. Antennas Propag. 60(11), 5451–5455
(2012)
17. C. Tienda, M. Arrebola, J.A. Encinar, G. Toso, Analysis of a dual-reflect array antenna. IEE
Proc. Microw. Antennas Propag. 5(13), 1636–1645 (2011)
18. M. Zhou, S.B. Sorensen, E. Jorgensen, P. Meincke, O.S. Kim, O. Breinbjerg, An accurate
technique for calculation of radiation from printed reflectarrays. IEEE Antennas Wirel.
Propag. Lett. 10, 1081–1084 (2011)
5 Reflectarrays 187

19. E. Almajali, D.A. McNamara, J. Shaker, M.R. Chaharmir, Feed image lobes in offset-fed
reflectarrays: diagnosis and solution. IEEE Trans. Antennas Propag. 62(1), 216–227 (2014)
20. S.R. Rengarajan, Scanning and defocusing characteristics of microstrip reflectarrays. IEEE
Antennas Wirel. Propag. Lett. 9, 163–166 (2010)
21. S.V. Hum, J. Perruisseau-Carrier, Reconfigurable reflectarrays and array lenses for dynamic
antenna beam control: a review. IEEE Trans. Antennas Propag. 62(1), 183–198 (2014)
22. B. Wu, A. Sutinjo, M.E. Potter, M. Okoniewski, On the selection of the number of bits to
control a dynamic digital MEMS reflectarray. IEEE Antennas Wirel. Propag. Lett. 7, 183–186
(2008)
23. S.V. Hum, M. Okoniewski, R.J. Davies, Realizing an electronically tunable reflectarray using
varactor diode-tuned elements. IEEE Microw. Wirel. Compon. Lett. 15(6), 422–424 (2005)
24. M. Riel, J. Laurin, Design of an electronically beam scanning reflectarray using
aperture-coupled elements. IEEE Trans. Antennas Propag. 55(5), 1260–1266 (2007)
25. J. Rodriguez-Zamudio, J.I. Martinez-Lopez, J. Rodriguez-Cuevas, A.E. Martynyuk,
Reconfigurable reflectarrays based on optimized spiraphase-type elements. IEEE Trans.
Antennas Propag. 60(4), 1821–1830 (2012)
26. R. Pereira, R. Gillard, R. Sauleau, P. Potier, T. Dousset, X. Delestre, Dual linearly-polarized
unit-cells with nearly 2-bit resolution for reflectarray applications in X-band. IEEE Trans.
Antennas Propag. 60(12), 6042–6048 (2012)
27. M.R. Chaharmir, J. Shaker, M. Cuhaci, A. Sebak, Novel photonically-controlled reflectarray
antenna. IEEE Trans. Antennas Propag. 54(4), 1134–1141 (2006)
28. J. Perruisseau-Carrier, A.K. Skriverviky, Monolithic MEMS-based reflectarray cell digitally
reconfigurable over a 360° phase range. IEEE Antennas Wirel. Propag. Lett. 7, 138–141
(2008)
29. H. Legay, Y. Cailloce, O. Vendier, G. Caille, J. Perruisseau-Carrier, M. Lathi, J.P. Polizzi, U.
Oestermann, P. Pons, N. Raveu, Satellite antennas based on MEMS tunable reflectarrays, in
Proceedings of the 2007 European Conference on Antennas Propagation (EuCAP2007)
(2007)
30. A. Moessinger, R. Marin, S. Mueller, J. Freese, R. Jakoby, Electronically reconfigurable
reflectarrays with nematic liquid crystals. Electron. Lett. 42(16), 899–900 (2006)
31. W. Hu, R. Cahill, J. Encinar, R. Dickie, H. Gamble, V. Fusco, N. Grant, Design and
measurement of reconfigurable millimeter wave reflectarray cells with nematic liquid crystal.
IEEE Trans. Antennas Propag. 56(10), 3112–3117 (2008)
32. G. Perez-Palomino, J.A. Encinar, M. Barba, E. Carrasco, Design and evaluation of
multi-resonant unit cells based on liquid crystals for reconfigurable reflectarrays. IEE Proc.
Microw. Antennas Propag. 6(3), 348–354 (2012)
33. R.R. Romanofsky, Advances in scanning reflectarray antennas based on ferroelectric thin-film
phase shifters for deep-space communications. Proc. IEEE 95(10), 1968–1975 (2007)
34. H. Kamoda, T. Iwasaki, J. Tsumochi, T. Kuki, O. Hashimoto, 60-GHz electronically
reconfigurable large reflectarray using single-bit phase shifters. IEEE Trans. Antennas Propag.
59(7), 2524–2531 (2011)
35. E. Carrasco, M. Barba, J.A. Encinar, X-band reflectarray antenna with switching-beam using
PIN diodes and gathered elements. IEEE Trans. Antennas Propag. 60(12), 5700–5708 (2012)
36. O. Bayraktar, O.A. Civi, T. Akin, Beam switching reflectarray monolithically integrated with
RF MEMS switches. IEEE Trans. Antennas Propag. 60(2), 854–862 (2012)
37. E. Carrasco, M. Barba, B. Reig, C. Dieppedale, J.A. Encinar, Characterization of a
reflectarray gathered element with electronic control using ohmic RF MEMS and patches
aperture-coupled to a delay line. IEEE Trans. Antennas Propag. 60(9), 4190–4201 (2012)
38. M.E. Bialkowski, A.W. Robinson, H.J. Song, Design, development, and testing of X-band
amplifying reflectarrays. IEEE Trans. Antennas Propag. 50(8), 1065–1076 (2002)
39. R.W. Clark, G.H. Huff, J.T. Bernhard, An integrated active microstrip reflectarray element
with an internal amplifier. IEEE Trans. Antennas Propag. 51(5), 993–999 (2003)
188 S.V. Hum

40. L. Cabria, J.Á. García, J. Gutiérrez-Ríos, A. Tazón, J. Vassal’lo, Active reflectors possible
solutions based on reflectarrays and Fresnel reflectors. Int. J. Antennas Propag. Article ID
653952, 13 (2009)
41. K.K. Kishor, S.V. Hum, An amplifying reconfigurable reflectarray antenna. IEEE Trans.
Antennas Propag. 60(1), 197–205 (2012)
42. R.H. Phillion, M. Okoniewski, Improving the phase resolution of a micromotor-actuated
phased reflectarray, in 2008 Microsystems and Nanoelectronics Research Conference (MNRC
2008) (2008), pp. 169–172
43. H. Legay, B. Pinte, M. Charrier, A. Ziaei, E. Girard, R. Gillard, A steerable reflectarray
antenna with MEMS controls, in Proceedings of the 2003 IEEE International Symposium
Phased Array System and Technology (2003), pp. 494–499
44. S. Mener, R. Gillard, R. Sauleau, C. Cheymol, P. Potier, Design and characterization of a
CPSS-based unit-cell for circularly polarized reflectarray applications. IEEE Trans. Antennas
Propag. 61(4), 2313–2318 (2013)
45. S.V. Hum, B. Du, Equivalent circuit modeling for reflectarrays using Floquet modal
expansion, IEEE Trans. Antennas Propag., 65(3), 1131–1140 (2017)
46. A.K. Bhattacharyya, Phased Array Antennas: Floquet Analysis, Synthesis, BFNs and Active
Array Systems (Wiley-Interscience, 2006)
47. S.V. Hum, M. Okoniewski, R.J. Davies, Modeling and design of electronically tunable
reflectarrays. IEEE Trans. Antennas Propag. 55(8), 2200–2210 (2007)
48. G. Perez-Palomino, M. Barba, J.A. Encinar, R. Cahill, R. Dickie, P. Baine, M. Bain, Design
and demonstration of an electronically scanned reflectarray antenna at 100 GHz using
multiresonant cells based on liquid crystals. IEEE Trans. Antennas Propag. 63(8), 3722–3727
(2015)
49. W. Menzel, A. Moebius, Antenna concepts for millimeter-wave automotive radar sensors.
Proc. IEEE 100(7), 2372–2379 (2012)
50. R.E. Hodges, M.J. Radway, A. Toorian, D.J. Hoppe, B. Shah, A.E. Kalman, ISARA—
integrated solar array and reflectarray CubeSat deployable Ka-band antenna, in 2015 IEEE
International Symposium on Antennas and Propagation USNC/URSI National Radio Science
Meeting (2015), pp. 2141–2142
51. R.E. Hodges, D.J. Hoppe, M.J. Radway, N.E. Chahat, Novel deployable reflectarray antennas
for CubeSat communications, in 2015 IEEE MTT-S International Microwave Symposium
(2015)
52. J.A. Encinar, J.A. Zornoza, Three-layer printed reflectarrays for contoured beam space
applications. IEEE Trans. Antennas Propag. 52(5), 1138–1148 (2004)
53. M. Zhou, O. Borries, E. Jørgensen, Design and optimization of a single-layer planar
transmit-receive contoured beam reflectarray with enhanced performance. IEEE Trans.
Antennas Propag. 63(4), 1247–1254 (2015)
54. R.E. Hodges, M. Zawadzki, A reflectarray antenna for use in interferometeric ocean height
measurement, in 2005 IEEE Aerospace Conference (2005), pp. 1131–1139
55. C. Han, S. Hsu, K. Chang, J. Huang, A Ku/Ka-dual band reflectarray to emulate a cylindrical
reflector for titan cloud precipitation radar and altimeter, in 2007 IEEE Antennas and
Propagation Society International Symposium (2007), pp. 1445–1448
56. C. Tienda, M. Younis, P. Lopez-Dekker, P. Laskowski, Ka-band reflectarray antenna system
for SAR applications, in The 8th European Conference on Antennas and Propagation
(EuCAP 2014) (2014), pp. 1603–1606
57. C. Kocia, S.V. Hum, Design of an optically transparent reflectarray for solar applications
using indium tin oxide. IEEE Trans. Antennas Propag. 64(7), 2884-2893 (2016)
58. P. Dreyer, M. Morales-Masis, S. Nicolay, C. Ballif, J. Perruisseau-Carrier, Copper and
transparent-conductor reflectarray elements on thin-film solar cell panels. IEEE Trans.
Antennas Propag. 62(7), 3813–3818 (2014)
59. J.A. Zornoza, R. Leberer, J.A. Encinar, W. Menzel, Folded multilayer microstrip reflectarray
with shaped pattern. IEEE Trans. Antennas Propag. 54(2), 510–518 (2006)
5 Reflectarrays 189

60. Q. Chen, Reflectarray development for improving NLOS radio channel, in 2013 Asia-Pacific
Microwave Conference Proceedings (APMC) (2013), pp. 654–656
61. L. Li, Q. Chen, Q. Yuan, K. Sawaya, T. Maruyama, T. Furuno, S. Uebayashi, Frequency
selective reflectarray using crossed-dipole elements with square loops for wireless commu-
nication applications. IEEE Trans. Antennas Propag. 59(1), 89–99 (2011)
62. G. Perez-Palomino, P. Baine, R. Dickie, M. Bain, J.A. Encinar, R. Cahill, M. Barba, G. Toso,
Design and experimental validation of liquid crystal-based reconfigurable reflectarray
elements with improved bandwidth in F-band. IEEE Trans. Antennas Propag. 61(4),
1704–1713 (2013)
63. C. Han, C. Rodenbeck, J. Huang, K. Chang, A C/Ka dual frequency dual layer circularly
polarized reflectarray antenna with microstrip ring elements. IEEE Trans. Antennas Propag.
52(11), 2871–2876 (2004)
64. M.R. Chaharmir, J. Shaker, H. Legay, Dual-band Ka/X reflectarray with broadband loop
elements. IET Microw. Antennas Propag. 4(2), 225–231 (2010)
65. J.A. Encinar, J.A. Zornoza, Broadband design of three-layer printed reflectarrays. IEEE Trans.
Antennas Propag. 51(7), 1662–1664 (2003)
66. L. Liang, S.V. Hum, Design of an UWB reflectarray as an impedance surface using Bessel
filters, IEEE Trans. Antennas Propag. 64(10) 4242-4255 (2016)
67. D.M. Pozar, Wideband reflectarrays using artificial impedance surfaces. Electron. Lett. 43(3),
148–149 (2007)
68. M. Farmahini-Farahani, H. Mosallaei, Birefringent reflectarray metasurface for beam
engineering in infrared. Opt. Lett. 38(4), 462–464 (2013)
69. T. Niu, W. Withayachumnankul, B.S.-Y. Ung, H. Menekse, M. Bhaskaran, S. Sriram, C.
Fumeaux, Experimental demonstration of reflectarray antennas at terahertz frequencies. Opt.
Express 21(3), 2875–2889 (2013)
70. X. Chen, L. Huang, H. Mühlenbernd, G. Li, B. Bai, Q. Tan, G. Jin, C.-W. Qiu, S. Zhang, T.
Zentgraf, Dual-polarity plasmonic metalens for visible light. Nat. Commun. 3, 1198 (2012)
71. N. Yu, N. Yu, P. Genevet, M.A. Kats, F. Aieta, Light propagation with phase discontinuities:
generalized laws of reflection and refraction. Science 334(6054), 333–337 (2011)
72. L. Zou, W. Withayachumnankul, C.M. Shah, A. Mitchell, M. Bhaskaran, S. Sriram,
C. Fumeaux, Dielectric resonator nanoantennas at visible frequencies. Opt. Express 21(1),
1344–1352 (2013)
73. Z. Du, A. Tamminen, J. Ala-Laurinaho, J. Säily, P. Rantakari, A. Luukanen, A.V. Räisänen,
Design and optimization of reconfigurable reflectarray element with MEMS phase shifter, in
2013 7th European Conference on Antennas and Propagation (EuCAP) (2013),
pp. 2422–2426
74. E. Carrasco, J. Perruisseau-Carrier, Reflectarray antenna at terahertz using graphene. IEEE
Antennas Wirel. Propag. Lett. 12, 253–256 (2013)
75. E. Carrasco, M. Tamagnone, J. Perruisseau-Carrier, Tunable graphene reflective cells for THz
reflectarrays and generalized law of reflection. Appl. Phys. Lett. 102(10), 104103 (2013)
Chapter 6
Transmitarray Antennas

Laurent Dussopt

Abstract Transmitarray antennas have attracted a lot of interest for the last 10
years to address the challenges of high-directivity reconfigurable antenna designs
for emerging applications at millimeter-wave frequencies. This chapter presents the
main principles of operation of transmitarray antennas, their main advantages and
challenges, and typical envisioned applications. The main achievements and
research orientations are summarized. Next, a theoretical model is presented
enabling an accurate and efficient simulation and optimization; generic theoretical
results are included to illustrate the main optimization trade-offs. In the final sec-
tion, practical examples taken from the most recent state of the art are presented
covering passive, beam-steering, and reconfigurable antennas in different frequency
ranges as well as innovative illumination schemes to reduce the total height of these
antennas. Finally, the conclusion highlights some important perspectives and
research topics toward future developments.

6.1 Introduction

6.1.1 Principles and Development

For the last 10–15 years, transmitarray antennas have triggered a lot of interest and
became popular to enable high-gain, high-efficiency, wideband, low-cost, and
reconfigurable antennas for emerging millimeter-wave applications. They achieve a
similar function as dielectric lenses, but are fabricated on planar technologies (e.g.,
Printed Circuit Boards) and materials, which makes them lightweight, conformable,
and suitable for integrating active devices for reconfiguration. Transmitarray
antennas are typically composed of one or several focal sources illuminating a
planar arrangement of sub-wavelength phase-shifting unit cells designed to define a
predetermined phase distribution across the aperture and therefore a predetermined

L. Dussopt (&)
CEA, LETI, MINATEC Campus, 17 rue des martyrs, 38054 Grenoble, France
e-mail: [email protected]

© Springer International Publishing AG 2018 191

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_6
192 L. Dussopt

Fig. 6.1 Different structures of artificial lens antennas based on phase-shifters (a), multiple
coupled layers (b), planar lens (c), Fresnel lens (d), metal waveguide lens (e), delay lines based on
constrained lenses (f)

radiation characteristic in far-field or near-field. A wide variety of structures based

on this principle have been investigated and named using different terminologies
such as discrete lens, phase-shifting surface, planar lens, flat lens, array lens, or
transmitarray (Fig. 6.1). In this document, the generic term transmitarray will be
preferred for its analogy with reflectarray antennas.
In contrast, we exclude from this family frequency-selective surfaces which are
designed primarily to achieve a selected frequency response for filtering applica-
tions without specific requirement on the transmission phase distribution across the
aperture. We exclude also metasurfaces or metamaterial structures which are
composed of periodic cells having a very small period as compared to the wave-
length, making them equivalent to a homogeneous material with effective permit-
tivity and permeability. Fresnel lenses are also well-known structures not
considered in this chapter as they belong to a different class of fabrication tech-
nology although a lot of analogy can be found in their design, especially because
transmitarrays implement phase-correcting Fresnel zones to focus a radiation beam
[1, 2].
Transmitarray antennas currently developed in the literature actually inherit from
a long history of large-size transmitting antenna structures such as constrained
lenses and bootlace antennas developed in the 1940s–1980s for space and military
applications (Fig. 6.1) [3–7]. Grid arrays belong also to this family of transmitting
apertures and were investigated for beam-steering linearly polarized waves [8].
Transmitting structure gained a renewed interest in the 1990s for power combining
applications with amplifiers integrated inside the transmitting panel in order to
overcome the power limitations of integrated solid-state amplifiers [9].
Transmitarray antennas are able to achieve high directivity and gain, wideband
performance and good power efficiency. These are mainly outcomes of the spatial
feeding configuration which represents a major difference as compared to tradi-
tional phased arrays with corporate feed networks; in contrast, phased arrays
6 Transmitarray Antennas 193

usually enable a much higher flexibility in phase/magnitude control of each radi-

ating element and therefore more accurate radiation control but the design of
high-gain (e.g., above 30 dBi) phased arrays in the millimeter-wave domain
remains highly challenging. In contrast to dielectric lenses, transmitarray antennas
are spatially fed in free space instead of a dielectric material, and the beam colli-
mating is achieved through sub-wavelength circuits instead of propagation; as a
result, they generally offer a wider bandwidth, a higher efficiency, a lower weight,
and more opportunities for electronic reconfiguration of the radiation properties
(polarization, beam shape, and direction) at microwave and low millimeter-wave
frequencies. Beyond 100–200 GHz, dielectric lenses certainly offer better efficiency
and manufacturability while their weight and size become a minor issue. Unlike
reflectarrays, transmitarray antennas do not suffer from masking effect from the
focal source. This is an important point as there is a strong interest in integrating
focal sources and active circuits (e.g., amplifiers, frequency conversion) at
millimeter-wave frequencies, resulting in large systems of several wavelengths in
size. Like other antenna array structures, transmitarrays offer also excellent polar-
ization properties with a capability to generate linear or circular polarization with
very low cross-polarization. They can be implemented, usually, on standard
printed-circuit-board technologies which can be planar, conformed, or flexible, in
order to guarantee low manufacturing costs in large volumes with good repro-
ducibility. Some of the shortcomings of transmitarray antennas are their size (area
and height), implementation in the upper millimeter-wave range or submillimeter
wave range, and the implementation of large reconfigurable aperture with fine phase
control due to the transmission topology which leaves little room for biasing cir-
cuits and biasing lines routing.
The electronic reconfiguration of transmitarray antennas to achieve
beam-forming and beam-steering functions is certainly one of today’s main chal-
lenges. It can be implemented at the focal source level using a reconfigurable focal
array, which enables limited beam-steering capabilities but is still suitable for some
applications, or it can be implemented at the transmitarray panel level, i.e.,
reconfiguration of each unit cell of the transmitarray, which enables a high radiation
flexibility. In this case, the integration of active devices (switches, varactors,
phase-shifters) significantly impacts the losses, fabrication complexity, and cost of
these structures, especially at millimeter-wave frequencies.
Transmitarray antennas are attractive solutions for many current or emerging
applications such as follows:
• Point-to-point links in telecommunication networks for long-haul and short-haul
wireless links, where very high transmission capacity is needed. These markets
cover a wide range of frequency bands from 20 GHz up to 90 GHz currently
and additional bands beyond 90 GHz under investigation. Typical gain levels of
30–50 dBi are required and first industrial prototypes of transmitarray antennas
developed at 60 GHz indicate a good level of maturity of this technology.
194 L. Dussopt

• Base stations in future mobile networks providing millimeter-wave mobile

access; these systems will require compact and cost-effective base station
antennas with beam-steering capabilities over a wide angular sector, multi-beam
or beam-forming capability and gain levels in the range of 20–30 dBi.
• Satellite communications in Ku–Ka-bands, in particular at the level of mobile
(vehicles) or portable terminals where lightweight compact and low-cost
structures are needed.
• Automotive radars: today a variety of short-range or long-range radars equip
modern vehicles for security (crash avoidance), comfort (auto-cruise), or
autonomous driving. Gain ranges from 10 to 40 dBi and wide or narrow
beam-scanning capabilities are needed depending on the application.
• Imaging and security applications where the use of millimeter-wave frequency
bands becomes more and more popular for imaging (security screening in public
transportations), surveillance (radars), and industrial control.

6.1.2 General Overview of the State of the Art

Many passive (fixed-beam) transmitarrays have been demonstrated to date in a wide

range of frequency bands from C- to W-bands in linear, circular, or dual polar-
ization [10–15]. Most of these designs are based on standard PCB fabrication which
is well suited to low-cost, large-area, multilayer printed circuits in these frequency
bands where standard line/slot resolution of about 100 µm is appropriate. First
commercial products are expected soon for V- and E-band point-to-point
communications.
Most devices in the past works were designed using geometric optics between
the focal source and the transmitarray panel to calculate the appropriate phase
correction of each unit cell to synthesize the linear phase distribution required for
simple pencil-beam patterns. In fact, transmitarray antennas can be quite easily
modeled analytically and simulated at low computation cost as shown in Sect. 6.2
[16–18]. Full-wave electromagnetic simulations are possible using current com-
mercial softwares and standard computer stations. However, the simulation time
can reach several hours, which is acceptable in a validation phase but hardly in an
iterative design or optimization procedure. Analytical models can be easily coupled
with optimization algorithms to customize radiation properties and soon,
researchers demonstrated transmitarrays with synthesized beams, e.g., shaped
beams [19], multiple beams [20], and near-field focusing.
Transmitarrays are based on elementary unit cells implementing a selected time
delay or phase shift with low transmission/reflection losses. However, the imple-
mentation constraints usually limit the feasibility of any transmission phase value so
that unit cell designs often exhibit a limited phase range (i.e., smaller than 360°)
and/or a limited phase resolution (i.e., phase quantization). Several authors dis-
cussed these design issues and highlighted in particular the trade-offs between the
6 Transmitarray Antennas 195

phase variation range, the transmission loss, and the complexity of multilayer
transmitarrays [21, 22]. Phase quantization, i.e., number of phase states available in
a 360° range, is also an important parameter, especially in reconfigurable trans-
mitarrays. Hence, the losses, as well as the complexity, cost, and power con-
sumption, are increasing with the number of switching active devices and therefore
the number of phase states.
Achieving a wide transmission bandwidth is often a major requirement to meet
the needs of practical communication and radar applications. Most transmitarrays
based on resonating radiators—e.g., half-wavelength patches, slots, dipoles, etc.—
and phase-shifters, typically achieve a 3-dB bandwidth in the range of 10–20%
suitable for many applications. However, some applications like imaging systems
or multifunction radars require a much wider bandwidth, e.g., in order to transmit
broadband pulsed signals. In this domain, multilayer sub-wavelength
metallo-dielectric structures with true-time delay responses across a wideband
have shown to operate over 30–50% bandwidth [23–25]. Dual-band transmitarrays
have been demonstrated as well, a typical application being satellite communica-
tions (SATCOM) where transmit and receive frequency bands are separated by a
large gap (e.g., 20/30 GHz for K/Ka-band systems) [26].
The focal distance between the focal source(s) and the transmitarray panel is
often seen as a drawback in many applications where thin and low-profile
requirements are important. Multiple options to mitigate this issue have been
investigated by several groups, such as the use of multiple focal sources, guided
feeds, and metasurfaces; several examples are discussed in Sect. 6.3.5.
The electronic reconfiguration of antennas radiation properties is clearly a major
technical challenge and enabler for future millimeter-wave applications in order to
cope with mobility and system reconfigurability needs with lower cost, mass, and
size than mechanically steered systems. In the case of transmitarray antennas,
reconfiguration can be implemented either at focal source or transmitarray panel
levels with different capabilities in terms of scanning range or beam-forming
flexibility. The reconfiguration of the focal source enables beam-scanning over
narrow angular sectors around the nominal beam direction and can be implemented
with simple switching circuits at low cost and with high efficiency; applications
include automotive long-range radar or backhaul/fronthaul point-to-point systems
where auto-alignment of the antenna beam is valuable to save time in installation
phase and to mitigate small displacements due to wind or other causes. In contrast,
the reconfiguration at the transmitarray panel enables much higher flexibility of the
radiation properties including beam-scanning over a large angular sector, beam-
forming adapted to arbitrary spatial constraints in terms of gain, sidelobes,
null-steering, etc. Such flexibility comes at the cost of higher complexity, cost, and
losses but is still valuable in many applications including SATCOMs,
millimeter-wave access in future mobile networks, point-to-multi-point communi-
cations, radars, imaging, etc. Reconfigurable transmitarrays were considered and
investigated since the early days of this antenna technology. Most of the investi-
gations reported in the literature are performed in frequency bands from 4 to
15 GHz where few applications exist but technical implementation is easier. A few
196 L. Dussopt

other examples are demonstrated around 30 GHz with the growing interest for
SATCOM and 5G mobile communication networks. Several examples are dis-
cussed in Sect. 6.3.4. These works are based on a wide variety of reconfiguration
technologies:
• PIN diodes [27–29] are currently the best switching technology available for
millimeter-wave frequency bands in terms of transmission losses, power han-
dling, linearity, and switching time. They lead to the best results demonstrated
so far in terms of antenna efficiency despite the lower phase resolution (phase
resolution is often limited to 90°–180°).
• MEMS switches have a potential for very low losses at high millimeter-wave
frequencies; as compared to PIN diodes, higher power handling and linearity
can be expected but switching times are orders of magnitude higher (in the range
of microseconds). To date, this technology is still limited in terms of maturity
and reliability [30] so that few demonstration works exist and practical appli-
cations seem far away.
• Varactor diodes are a widely available semiconductor technology enabling
continuously tunable circuits. They enable a continuous phase control and, in
that respect, can provide more accurate radiation capabilities. Their tuning range
is often limited but they can be combined with a switching technology, such as
PIN diodes, to extend the phase-shifting range. Several demonstrations were
performed based on this technology at 2–15 GHz [31–37]. However, varactor
diodes are characterized with a limited quality factor, i.e., significant losses at
millimeter-wave frequencies. MEMS varactors are an attractive alternative in
principle but have not been applied to transmitarray antennas to date, probably
due to lack of a reliable and mature technology. Ferroelectric materials are
promising for the implementation of tunable capacitors with better quality factor
than semiconductor devices at millimeter-wave frequencies. First demonstra-
tions were presented with attractive performance at 12 GHz [37].
• MMIC phase-shifters based on silicon or III-V semiconductor technologies are
an obvious option to integrate phase-shifting capabilities inside transmitarray
unit cells. The design of low-loss integrated phase-shifters is a very active field
for many decades, driven mostly by phased-array developments in defense and
spatial applications and more recently for civilian wireless applications. One of
the challenges is the integration of these integrated circuits within the limited
area of a transmitarray unit cell (typically k0/2
k0/2 in size). This approach is
investigated in defense applications but few examples are available in the open
literature [38]. In contrast to stand-alone devices like switches or varactors,
integrated phase-shifters suffer from higher costs but enable more accurate phase
control and can be combined with attenuation/amplification circuits.
• Microfluidics is a technology relying on the displacement of liquid metal in
microchannels in order to reconfigure a radiating element or phase-shifting
circuit. It has been investigated quite recently, e.g., at 8.8 GHz [39], and remains
an exploratory area so far with little application potential. Hence, microfluidics
is characterized by a certain fabrication and actuation complexity, as well as
6 Transmitarray Antennas 197

very low switching times. On the upside, they can operate in static mode with
zero power consumption and therefore may address applications with very low
power consumption requirements and low reconfiguration rate.
• Liquid crystal materials exhibit a tunable dielectric permittivity which may be
interesting for the design of reconfigurable transmitarrays. They are lossy at
microwave frequencies and exhibit a small tuning range, but they offer inter-
esting properties at millimeter-wave frequencies. To date, they were success-
fully used for reflectarrays but not for transmitarrays to our knowledge.

6.2 Modeling and Design

Transmitarray antennas are electrically large structures so their full-wave electro-

magnetic simulation is usually challenging in time and computer resources; most
developments to date are based on simplified models at least in the first design steps
and full-wave simulations are limited to validation and verification of main per-
formances but with limited accuracy.
Transmitarray antennas can be modeled analytically to derive first-order
performance metrics and guide the main design choices; such basic modeling is
presented briefly in the following subsections. More accurate simulations can be
performed using a mix of analytical modeling and full-wave electromagnetic
simulations; such a methodology is presented in the following where focal source(s)
and each individual unit cell are simulated with a 3D electromagnetic simulation
software and these results are combined analytically to extract the global trans-
mitarray antenna performances. Finally, some design guidelines are given on the
two main families of transmitting structures, namely designs based on coupled
layers or separate Tx–Rx antennas.

6.2.1 Theoretical Model

A generic configuration of a transmitarray antenna is shown in Fig. 6.2a. It is

illuminated by a focal source with a complex radiation pattern H ~FS ðh; /Þ. Each unit
cell n is modeled by two antennas, of radiation patterns H ~1n ðh; /Þ and H~2n ðh; /Þ,
connected through a phase-shifter or time delay device represented by a 2-port
network [Sn]. In this work, the radiation patterns H ~ are normalized to the square
root of the isotropic gain so that the antenna efficiency η is given by
ZZ
1  
g¼ H~ ðh; /Þ2 sin hdhd/: ð6:1Þ
4p
198 L. Dussopt

(a) (b)
P6

Transmitarray Transmitarray P5
unit-cell
[Sn ]
P4
Unit cell
Pr P3
a 2n b 2n
Rn [S n]
a 1n b 1n

Focal source P2
Focal source P1

Fig. 6.2 Principle of a transmitarray antenna and generic model of a unit cell (a), power budget (b)

Assuming the focal source is fed with the input power P1, the incident wave a1n
received by unit cell n is given by

pffiffiffiffiffi kej2pRn =k
a1n ¼ P1 ~FS ðhn ; /n Þ  H
H ~1n ðhn ; /n Þ; ð6:2Þ
4pRn

where Rn is the distance between the focal source and ðhn ; /n Þ is the direction of the
unit cell n with respect to the focal source. The reflected and transmitted waves b1n
and b2n are computed as

b1n ¼ S11n a1n and b2n ¼ S21n a1n : ð6:3Þ

~TA ðh; /Þ is computed by

Finally, the radiation pattern of the transmitarray H
summing the contribution of each unit cell:
X
~TA ðh; /Þ ¼
H ~2n ðh; /Þ:
b2n H ð6:4Þ
n

The optimization and analysis of a transmitarray is best understood through a

detailed power budget of the structure which is represented schematically in
Fig. 6.2b. We can identify seven power figures:
• P1: input power of the focal source.
• P2: power radiated by the focal source.
• P3: incident power from the focal source on the transmitarray.
• P4: power collected by the transmitarray.
6 Transmitarray Antennas 199

• P5: power transmitted through the 2-port networks of the unit cells.
• P6: power radiated by the transmitarray.
• Pr: power reflected by the transmitarray.
P2 and P6 can be computed by integration of the radiation patterns of the focal
source and the transmitarray, respectively. P3 can be computed by the integration of
the focal source radiation pattern over the area of the transmitarray. P4 and P5 are
the sum of the incident (a1n) and transmitted (b2n) waves, respectively. There is no
rigorous technique to compute the reflected power Pr analytically since part of the
incident radiation from the focal source is directly reflected by the transmitarray
structure without going through the unit cell receiving antenna, i.e., the receiving
antenna effective aperture is lower than the unit cell area (P4 < P3). However, if the
losses in the receiving antennas are low, it may be approximated by summing the
power not captured by the unit cells (P3 − P4) and the sum of the reflected waves
b1n:
X
Pr  P3  P4 þ jb1n j2 : ð6:5Þ
n

Based on these power figures, the total efficiency of the transmitarray can be
defined as

P6
g¼ ¼ gFS  gSO  gCA  gIL  gAL ; ð6:6Þ
P1

where
• gFS ¼ P2 =P1 is the efficiency of the focal source taking into account the resistive
losses and mismatch losses.
• gSO ¼ P3 =P2 is the illumination efficiency taking into account the spill-over
losses.
• gCA ¼ P4 =P3 is the efficiency of the receiving array. This term shall take into
account resistive, impedance mismatch, and polarization mismatch losses. It can
be considered as the effective aperture of the receiving array.
• gIL ¼ P5 =P4 is the efficiency of the 2-port networks taking into account their
insertion losses.
• gAL ¼ P6 =P5 is the efficiency of the emitting array. This term shall take into
account the resistive losses and impedance mismatch losses.
The total efficiency of the transmitarray can be computed classically by inte-
~TA ðh; /Þ given by Eq. 6.4.
gration of the radiation patterns H
The phase-shift values phðS21n Þ of each unit cell are selected to generate a phase
distribution of the emerging waves phðb2n Þ corresponding to a desired radiation
pattern:
200 L. Dussopt

phðb2n Þ ¼ phða1n Þ þ phðS21n Þ ¼ uFS ðhn ; /n Þ  k0 Rn þ u1n ðhFS ; /FS Þ þ phðS21n Þ;

ð6:7Þ

where uFS ðhn ; /n Þ is the phase of the focal source radiation pattern in the direction
of unit cell n and u1n ðhFS ; /FS Þ is the phase of the unit cell n radiation pattern in the
direction of the focal source. In many cases, a linear phase distribution is chosen to
generate a pencil beam in a direction ðh0 ; /0 Þ and reach the maximum directivity:

r!
phðb2n Þ ¼ k0~ rn ¼ k0 sin h0 cos /0 xn  k0 sin h0 sin /0 yn ; ð6:8Þ

where ~ r is the unit vector in the direction ðh0 ; /0 Þ of the beam and !
rn ¼ ðxn ; yn Þ is
the coordinate vector of unit cell n.
The model described above is generic and can be used to investigate most, if not
all, transmitarray configurations. A critical point is the implementation of the
transmitting unit cells; as discussed in the following (Sect. 6.2.4), two main families
can be identified. In the first one, the unit cells are composed of separate receive and
transmit antennas with an intermediate interconnection network and correspond to
the equivalent model presented in Fig. 6.2a. In these cases, each subpart of the unit
cell can be designed and simulated independently in order to extract its charac-
teristic parameters. Such structures usually include a ground plane isolating each
side of the transmitarray panel. Some examples are the works presented in [28, 34,
36]. In the second family, unit cells are composed of coupled resonators or radiating
elements (slots, patches, strips) and such unit cells cannot be divided into inde-
pendent subblocks; typical examples are structures presented in [10, 11, 30]. The
equivalent model of Fig. 6.2a can still be used by considering fictitious receive and
transmit antennas having the same radiation patterns as the actual unit cell but ideal
efficiency and impedance matching, while the intermediate 2-port network includes
the S-parameters of the unit cell. These radiation patterns and S-parameters are
computed from a Floquet analysis of the unit cell with periodic conditions and
incident plane waves. We can note in this case that P5 = P6.
It is also worth to mention the case of multisource or focal-array-fed transmi-
tarrays. As shown further (Sect. 6.3), the illumination of the transmitarray by
multiple sources can be of interest to reduce the focal distance and make it more
compact for instance, or to implement multi-beam functionalities. The same model
can be applied readily in these cases using the superposition principle by summing
the incident radiation from each focal source at the level of each unit cell.

6.2.2 Theoretical Results

A theoretical analysis of transmitarray antennas can be easily performed to derive

generic performance results for a better understanding of the basic trade-offs and
provide design guidelines to reach particular requirements.
6 Transmitarray Antennas 201

Table 6.1 Maximum Index n 1 4 10 15 49

directivity and half-power
Directivity 6 10 13.4 15 20
beamwidth of an ideal focal
(dBi)
source in cosnh
Beamwidth 120° 65.5° 42° 34.6° 19.2°
(3 dB)

The far-field radiation of a focal source antenna, very often, can be approached
by a cosn h function where the directivity Dðh; /Þ and half-power beamwidth Dh3dB
are determined by the index n:
rffiffiffi!
1 n 1
Dðh; /Þ ¼ 2ðn þ 1Þcos h and Dh3dB ¼ 2 cos
n
: ð6:9Þ
2

Typical values of the maximum directivity and half-power beamwidth are given
in Table 6.1. For such focal source, the spill-over losses can be easily calculated in
the case of a circular transmitarray of diameter Dt and focal distance F by inte-
gration of the directivity over the array aperture:
 
nþ1 1 Dt
gso ¼ 1  cos h0 with h0 ¼ tan : ð6:10Þ
2F

This expression is important as one can note from Eq. 6.6 that the antenna total
efficiency η is equal to the spill-over efficiency gso if we assume a lossless focal
source and transmitarray.
In [40], the theoretical directivity of an aperture illuminated by a cosnh focal
source is derived:
 h 2p sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
Z 0 Z 
2F 2  cos n þ 2h sinh 

D ¼ 2 ð n þ 1Þ  dhd/ : ð6:11Þ
k  1 þ tan hsin / cos h
2 2 3

0 0

This expression can be integrated numerically for a continuous aperture or

transformed into a discrete sum in the (usual) case of a transmitarray composed of
individual unit cells with a size in the order of 0.2–1k0. Very close numerical results
are obtained using both approaches, showing that the discretization of the aperture
into unit cells has no significant influence on the achievable directivity as long as
the unit cell size remains below about k0.
From the directivity, one can easily compute the aperture efficiency of the
transmitarray:
 
Dk2 pDt 2
gap ¼ ¼D ; ð6:12Þ
4pA k
202 L. Dussopt

where A is the aperture area of the transmitarray and Dt is its diameter. In the case of
a lossless transmitarray, the aperture efficiency gap is also the product of the
spill-over efficiency gso , accounting for the power radiated by the focal source and
not captured by the transmitarray, and the taper efficiency gtaper , accounting for the
reduction of directivity due to a nonuniform amplitude distribution across the
aperture; it follows that gtaper can be computed from Eqs. 6.10 and 6.12. In the case
of a lossy transmitarray, the average loss of the unit cells can be taken into account
to extract the taper efficiency.
Figure 6.3 reports theoretical plots of the spill-over efficiency, aperture effi-
ciency, and taper efficiency for an ideal (lossless) circular planar transmitarray
antenna illuminated by a cosnh focal source. It clearly shows the trade-off between
taper and spill-over losses as a function of the focal distance and the maximum
aperture efficiency as a function of the focal source index.
The frequency bandwidth of an antenna is usually an important requirement in
most applications. Figure 6.4 presents the frequency response of an ideal planar
transmitarray illuminated by a cosnh focal source with phase-shifting implemented
as true-time delay (TTD) lines or constant phase-shifters. In the former case, TTD
lines perfectly match the electrical delay between the focal source and each unit cell
so that the transmitarray achieves its highest gain value at any frequency. In the
latter case, the constant phase response of phase-shifters can be tuned to the optimal

(a) 1 (b) 1
n = 49 n = 49
Spillover Efficiency (%)

n = 15 n = 15
Taper Efficiency (%)

0.8 n = 10 0.8 n = 10
n=4 n=4
n=1 n=1
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Focal Ratio (F /D) Focal Ratio (F /D)
(c) 1 (d) 40
n = 49 n = 49
Aperture Efficiency (%)

n = 15 n = 15
0.8 n = 10 n = 10
n=4 30 n=4
n=1 n=1
Taper (dB)

0.6
20
0.4

0.2 10

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Focal Ratio (F /D) Focal Ratio (F /D)

Fig. 6.3 Spill-over efficiency (a), taper efficiency (b), aperture efficiency (c) and edge amplitude
taper (d) of an ideal transmitarray antenna illuminated by a cosnh focal source
6 Transmitarray Antennas 203

Fig. 6.4 Normalized 5

frequency response of
transmitarrays with true-time
0
delays (TTD) or

Norm. Gain (dB)

TTD
frequency-constant
phase-shifters with cosn h -5
focal sources with different n = 49
indexes n (see Table 6.1 for -10 n = 15
gain versus index
n = 10
correspondence). Focal ratio
-15 n=4
F/D equals 0.25, 0.45, 0.7,
0.85, 1.55 for n = 1, 4, 10, 15, n=1
49, respectively -20
0.5 0.7 0.9 1.1 1.3 1.5
Norm. Frequency ( f/f 0)

electrical delays at the center frequency only and phase errors occurring at other
frequencies limit the antenna bandwidth. This bandwidth reduction is more
important for short focal distances (focal sources with a low index n in Fig. 6.4)
since the path length difference between the center and edge cells is more important.
True-time delay lines are a very attractive option theoretically in terms of band-
width but have not been implemented in transmitarrays so far due to the very long
delay values that are challenging to implement in practice.

6.2.3 Transmitarray Simulation

As with reflectarrays and other large-aperture antennas, full-wave electromagnetic

simulation of transmitarray antennas is feasible but is usually not the method of
choice in the design and optimization phase considering the considerable amount of
computation resources necessary to reach a suitable accuracy. With computation
times on the order of several hours at least using current mainstream computing
stations and softwares, full-wave simulations can be very useful in the verification
stage and for the investigation of the impact of the transmitarray environment
(mechanical fixture, surrounding objects in the operational environment, etc.).
In contrast, the theoretical model presented in Sect. 6.2.1 enables a fast and
accurate simulation of most transmitarray structures as it requires only a few
electromagnetic full-wave simulations of small complexity (focal source(s) and unit
cells) and derives accurate results on radiation patterns and main antenna charac-
teristics through simple analytical calculations. A flow chart of the simulation
procedure is presented in Fig. 6.5 and this procedure has been validated in
numerous test cases presented in the articles of the author. A short simulation time
is especially useful when optimization techniques with a large number of iterations
are used, for instance, for radiation pattern synthesis. Several effects are not taken
into account in this theoretical model such as the impact of spill-over radiation,
204 L. Dussopt

EM simulations:
Focal source: radiation patterns
Unit-cells: radiation patterns, [S] param.

Optimization of the phase-shift

Calculation of incident waves a1n
distribution for targeted radiation pattern
(eq. 6.2).
(e.g. eq. 6.8).

Calculation of radiation patterns Calculation of emerging waves b2n

(eq. 6.4). (eq. 6.3).

Calculation of power levels ( P 1 … P 6 , Pr )

and efficiency terms (η FS . η SO . η CA . η IL . η AL ).

Fig. 6.5 Simulation procedure

diffraction, and scattering on the fixture or edges of the transmitarray; although

these effects can be modeled, they would add a significant complexity to the
simulation software and they usually can be mitigated through suitable design of the
transmitarray structure (e.g., using shielding and absorbing materials).
Unit cells are simulated using periodic conditions and a Floquet analysis to
derive the S-parameters and radiation patterns assuming an infinite array of iden-
tical cells. Such analysis is available in many electromagnetic simulation softwares
and allows the investigation of large periodic structures through the simulation of a
single unit cell. It is based on the theory established by G. Floquet to solve periodic
linear differential equations. In the context of periodic electromagnetic structures
excited by a plane wave, the fields’ distribution is periodic as well and the
reflected/transmitted waves are expressed as a sum of plane waves (Floquet modes)
[41]. While the environment of each unit cell may generally be composed of
different unit cells in the actual transmitarray, the assumption of infinite periodicity
is generally good enough to allow for an accurate design. The size of unit cells is
typically of k0/2
k0/2 so that only the first order of Floquet modes is relevant.
As mentioned in Sect. 6.2.1, in cases where unit cells are based on coupled
layers and cannot be divided into independent subblocks (antennas and intermediate
circuit), the Floquet simulation is used to derive the S-parameters of the unit cell
which is then a 4
4 matrix taking into account the two orthogonal Floquet modes
associated with the two polarization directions X and Y [42, 43]:
6 Transmitarray Antennas 205

0 1
S11xx S11xy S12xx S12xy
B S11yx S11yy S12yx S12yy C
½ S ¼ B
@ S21xx
C: ð6:13Þ
S21xy S22xx S22xy A
S21yx S21yy S22yx S22yy

In some cases, unit cells are circularly polarized on either or both sides and
circularly polarized S-parameters can be deduced from linearly polarized ones
(Eq. 6.13) using classical expressions not reproduced here for brevity. Circular
polarization can be also generated using sequential rotation schemes where the
assumption of infinite periodicity of the individual unit cells is not met. In this case,
it can be necessary to simulate sub-arrays to derive accurate radiation patterns of
each unit cell and of the transmitarray [28].
Finally, the model, and in particular Eq. 6.2, is based on an important
assumption that the transmitarray is located at far-field distance of the focal source
since it assumes a 1/R field attenuation. The experience shows that a suitable
accuracy was obtained with focal distances chosen for maximum aperture efficiency
corresponding to typical applications. The replacement of Eq. 6.2 with a near-field
model of the focal source radiation could be envisioned in specific cases with very
short focal distances.

6.2.4 Transmitarray Unit Cells Design

A major challenge in the development of a transmitarray antenna is the design of

unit cells with low transmission losses and a wide range of transmission phase,
possibly up to 360° for short focal length arrays, beam-steering reconfigurable
transmitarrays, or accurate beam synthesis. To reach this goal, one can identify two
main strategies, which we will call separate Tx–Rx antennas and coupled-layers
approaches, respectively. In the former, the unit cell can be clearly separated into
two separate antennas on each side of the transmitarray panel and an intermediate
circuit or interconnection as represented in Fig. 6.2a. In the latter, the multiple
layers composing the unit cell operate collectively and the whole unit cell shall be
simulated as a single structure.
Separate Tx–Rx antennas: the phase shift is generated in lumped or distributed
circuits that have to be integrated within the transmitarray unit cell. As the area of
the unit cell is usually limited to about k0/2
k0/2 and a small thickness is also
desired, the main challenge lies in the miniaturization and minimization of losses of
this circuit. This challenge is even more important if one would implement
true-time delay unit cells which would be required for ultra-wideband operation.
Hence, in large arrays, the time delay difference between unit cells in the center and
at the edges amounts to several wavelengths (unless in very long focal transmi-
tarrays); such an array has not been demonstrated so far to our knowledge.
206 L. Dussopt

Coupled-layers approach: this approach is inherited from frequency-selective

surfaces (FSSs) which have been investigated for a long time for filtering purposes.
It relies on the collective response of multiple layers of resonators or scatterers
which exhibits a dispersive response much like a multistage filter. Small dimension
changes or lumped capacitance/inductance changes shift the frequency response
and result in a phase change. As it is often desired to minimize the number of layers
for the sake of low manufacturing cost, it can be shown that a quarter-wavelength
spacing between layers will maximize the phase-shift range and minimize
reflection/insertion losses. The design principles and limitations of such structures
have been investigated by several authors; in particular, the trade-off between the
phase-shift range and transmission losses was investigated in [21] and it was shown
that at least 3 and 4 layers were required for a 360° phase range within a 3- and
1-dB transmission loss limit, respectively. However, in [35], the authors showed
that coupled-layers transmitarrays designed for beam-steering result in significant
sidelobes in fixed directions dependent on the main beam direction but independent
of the transmitarray size. This effect is attributed to higher order modes triggered in
the structure at phase discontinuities and is emphasized by the large thickness
(typically k0/4 between each layer) of the structure.
The approach described above based on resonant structures (half-wavelength
scatterers, quarter-wavelength layer spacing) is usually narrowband due to a
strongly dispersive frequency response; they may also exhibit a significant angular
dispersion making the structure sensitive to incidence angles and prone to scanning
losses. This lead several groups to investigate the use of non-resonant miniaturized
(smaller than k0/4
k0/4) designs combining inductive and capacitive responses
with reduced spacing between each layer to reach wideband responses. In this case,
the phase shift per layer is reduced and a full 360° phase range would require a
large number of layers resulting in a larger fabrication complexity. In [23], a 5-layer
transmitarray with a total thickness of 0.08k0 only was demonstrated with a 20%
3-dB gain bandwidth; in this example, the transmission phase range was about 180°
only but this was sufficient for a transmitarray with a focal ratio of about 1. It can be
shown that these structures usually result in a quasi-linear phase response or
true-time delays which can be used to design ultra-wideband lenses. In [24], the
authors demonstrated a 4-layer transmitarray with 0.24k0 total thickness and a
maximum true-time delay of about 40 ps at 8.5 GHz (equivalent to 123° phase
shift). Although this delay is quite limited, this was suitable for a long focal lens
(F/D = 1.5) with a very broad frequency response compatible with the transmission
of pulsed signals with fractional bandwidth as large as 50%.
One can observe that such transmitarrays with a large number of layers come
close to designs based on the effective-index approach, which indeed can provide
very broadband performances but at a significant fabrication cost. The large number
of layers makes very difficult the design of a reconfigurable transmitarray that
would require either low-loss tunable materials or a very large number of tunable
lumped devices distributed in several layers. So far, examples available in the
literature have been limited to passive structures.
6 Transmitarray Antennas 207

6.3 Practical Examples and State of the Art

A selection of transmitarrays taken from the literature is presented hereafter to

illustrate the main achievements and most representative developments performed
to date in the domains of passive transmitarrays, beam-steering or reconfiguration
schemes, and feeding schemes.

6.3.1 Passive Transmitarrays

Numerous passive transmitarrays have been demonstrated to date in frequency

bands ranging from 5 to 90 GHz. They are usually based on standard
printed-circuit-board technologies and differ mainly by their phase-shifting schemes
using either separate Tx–Rx antennas or coupled layers. Overall, most structures
exhibit similar performances with 3-dB fractional bandwidths of 10–20%, aperture
efficiencies of 25–45% and total efficiencies of 50–70%.
In [44], a 30-GHz circularly polarized transmitarray using the separate Tx–Rx
antennas scheme was demonstrated on organic substrates with a thickness of
1.3 mm (0.13 k0) and three metal layers (Fig. 6.6). The transmitarray is a 10k0
10k0 square array of unit cells having an insertion loss of 0.2 dB only. Each unit
cell is linearly polarized on the focal source side and circularly polarized on the free
space side. The total gain is 22.8 dBi which corresponds to an aperture efficiency of
15% only due to the 180° quantization used in the phase compensation scheme
(about 3.1 dB of quantization loss) and to a significant taper loss due to the short

(a) Sequential rotation (b)

Circularly angle
polarized patch

Ground plane

Linearly
polarized patch
Phase correction
angle

Fig. 6.6 30-GHz circularly polarized transmitarray unit cell (a) and prototype with sequential
rotations (b) [44]
208 L. Dussopt

focal distance (F/D = 0.52). However, the antenna efficiency is very good (63.1%)
for this frequency band and the bandwidth properties are excellent (3-dB bandwidth
of 20%, 3-dB axial ratio bandwidth of 24%).
In [17], linearly and circularly polarized transmitarrays were demonstrated for
the first time at 60 GHz. The advantages of the structure are its simplicity, made of
only three metal layers, leading to very low loss (about 0.5 dB insertion loss) and
wide bandwidth (7% 1-dB bandwidth). One of the critical fabrication difficulties in
high millimeter-wave bands is the realization of vertical via interconnection
between metal layers, which impacts the fabrication cost as well. This issue was
overcome in [45] where via-less unit cells were demonstrated at 60 GHz with still
only three metal layers and low thickness (735 lm). A transmitarray composed of
seven different unit cells was demonstrated. These seven unit cells effectively
realize 2.8 bits of quantization, i.e., seven phases state from 45° to 315° with 45°
increments. Hence, it was found that it was not possible to implement a 0° phase
state with the selected design options (only three metal layers, no via).
A fixed-beam antenna was demonstrated with a broadside gain of 32 dBi at
60 GHz, a 3-dB bandwidth of 20%, and a very low cross-polarization level.
Recently, the same design was demonstrated in E-band (71–86 GHz) with similar
performance [46].
In [16], a 35-GHz linearly polarized filter-lens array is presented based on the
coupled-layer approach. It is fabricated on two glass wafers with a total thickness of
1 mm (0.12 k0), and three metal layers with a diameter of 7.8 cm (*9k0). It is
composed of bandpass unit cells inspired from frequency-selective surfaces with a
stack of patch-slot-patch resonators exhibiting insertion losses in the range of 2–3
dBi; their dimensions are tuned to provide the required phase shift at the central
frequency to focus the beam in the broadside direction. The 3-dB bandwidth is
8.2%, the maximum gain is 25.6 dBi, the aperture efficiency is 45%, and the total
efficiency is estimated to 52%.
In [47], another transmitarray based on the coupled-layer approach was
demonstrated at 30-GHz The array is composed of stacked parallel continuous
metallic strips diffracting the incident wave. The structure is made of three metal
and two dielectric layers with a total thickness of 1 mm (0.1k0). It was shown that
the number of metal layers and their relative distance (thickness of the structure)
determine the achievable transmission phase range and the transmission magnitude
variation [10]. In the presented example, a transmission phase range of *300° and
a transmission magnitude variation of *2 dB were obtained. A difficulty in this
structure is that the transmission magnitude and phase are not independent and a
trade-off has to be made between the minimization of phase errors and the opti-
mization of the transmission coefficient. Among the advantages of this design are
the simplicity of the fabrication process and the possibility to tune the transmission
phase quasi-continuously along the strip directions, thereby minimizing aperture
discretization losses. In [48], the authors reduced the structure to two metal layers
only with a transmission phase range of about 180° and used the amplitude cor-
rection scheme of Fresnel Zone Plate lenses to complete the 360° phase correction
range. They demonstrated a gain of 28.6 dBi and an aperture efficiency of 28.9%
6 Transmitarray Antennas 209

Fig. 6.7 Gradient-index dielectric flat lens: cross section and effective dielectric profile (a),
prototype at 60 GHz (b) [11]

with a 3-dB bandwidth of 16.5%. In contrast, the same authors increased the
number of metal layers to four in [19] to reach a more accurate phase and amplitude
control and enable a beam-shaping capability with the demonstration of a flat-top
beam.
In [11], transmitarrays implemented as gradient-index dielectric flat lenses were
demonstrated at 60 and 77 GHz (Fig. 6.7). The lenses are made of an organic
substrate with a permittivity er = 6.0, a diameter of 25 mm (5k0 at 60 GHz), and a
thickness of 7 mm (1.4k0). Holes are drilled in the dielectric substrate with a
variable diameter and spacing in order to synthesize an equivalent permittivity
tunable in the range of 2.25–6. The antenna gain is 18.3 dBi and 18.9 dBi at
60 GHz and 77 GHz, respectively, which corresponds to an aperture efficiency in
the range of 19–27%. The antenna efficiency is estimated by the authors to 60–70%.

6.3.2 Mechanical Beam-Steering

Beam-steering is an important functionality in many applications and can be

implemented mechanically. Instead of moving the whole antenna structure, trans-
mitarray antennas offer the opportunity to keep the focal source(s) fixed and move
only the transmitarray panel, which is interesting since the displacements of the RF
and electronic interconnections of the focal source(s) can trigger significant com-
plexity (rotary joint, flexible cables), cost, weight, and reliability issues, while the
transmitarray panel does not have any electrical connection if it is passive.
As for dielectric lenses or reflectors, a common beam-steering technique is to
rotate the transmitarray panel around its center so that the focal source remains on
the focal arc (Fig. 6.8a) [49]. Two axes of rotations can be combined to steer in
azimuth and elevation. This approach has been used in many works to demonstrate
the capabilities of passive transmitarrays. It is usually limited to about ±30°
because of increased spill-over losses and degraded radiation patterns (beam
210 L. Dussopt

distortion, sidelobes) due to a non-optimal amplitude and phase illumination of the

transmitarray by the focal source.
In the case of a large-aperture transmitarray, the rotation can induce a significant
increase of the total height of the antenna system and a large volume requirement,
which is inconvenient for many applications where low-profile antennas are desired
such as vehicles or aircrafts. A similar beam-steering effect can be obtained by
translation of the transmitarray panel above the focal source (Fig. 6.8b). In [50], a
Ka-band circularly polarized transmitarray is demonstrated for satellite communi-
cations with in-plane translation of the transmitarray panel above the focal source to
steer the beam in elevation over 0–50° with less than 3 dB scanning loss (Fig. 6.9).
This translation is combined with a 360° in-plane rotation of the transmitarray panel
to steer the beam in azimuth; hence, the focal source can be kept in fixed position
because it is circularly polarized. The transmitarray is optimized for a beam tilted at
32.5° elevation angle when in center position and the 0–50° scanning range is
covered for a translation of [−0.3F, 0.6F], where F is the focal distance; this
optimization takes into account the scan loss, the beam distortion, and the height
requirements. The antenna aperture is 195
145 mm2 and includes an extension to
translate the panel without increasing spill-over losses, the focal ratio is
F/D * 0.55, and the gain is 27.3 dBi.

Fig. 6.8 Mechanical beam-steering schemes: rotation in the focal arc (a), translation (b), in-plane
rotation (c)
6 Transmitarray Antennas 211

(a) (b) 30
RHCP
LHCP
20

Gain (d Bi)
10

-10
-90 -60 -30 0 30 60 90
Angle (Deg.)

Fig. 6.9 Mechanical beam-steering through translation of the transmitarray; Ka-band prototype
(a) and experimental radiation patterns with beam steered at 0–50° [50]

In-plane rotations of the transmitarray are easy to realize mechanically and do

not degrade the overall size of the antenna. In [51], a 30-GHz linearly polarized
transmitarray antenna composed of two rotating transmitting panels was presented
(Fig. 6.8c). The panels are separated by about k0/5; each one exhibits a linear phase
distribution; and the system is analogue to rotatable dielectric wedges, or Risley
prisms, which were already known but seldom used in the microwave domain due
to excessive thickness, weight, and reflection losses. This transmitarray mitigates
these issues and beam-scanning capabilities up to 57° and 72° from the zenith were
successfully demonstrated with scanning losses of 2.8 dB and 5.8 dB, respectively.
As compared to a fixed-beam transmitarray of similar technology, the two-panel
configuration results in about 1.8 dB additional losses at 30 GHz.

6.3.3 Switched-Beam Transmitarrays

In contrast to the mechanical beam-steering presented above, electronic switching

or scanning offers much faster operation, which can be useful in many applications,
especially radars and future mobile communication systems. In addition, it usually
improves also the system in terms of cost, weight, and reliability. As previously
known from dielectric and Fresnel lenses or reflectors, an array of sources in the
focal plane can be used for a simple implementation of multi-beam and
beam-switching capabilities (Fig. 6.10a). This approach is less complex and usually
provides better efficiency than fully reconfigurable antennas (next section) at the
cost of less radiation flexibility; these particularities make this configuration very
attractive for millimeter-wave systems such as automotive radars or point-to-point
communications.
This approach was first implemented at 60 GHz with circularly polarized
small-aperture transmitarrays (11–14 dBi gain) targeting indoor short-range com-
munications [52]. A focal array of five antennas integrated on high-resistivity
212 L. Dussopt

Fig. 6.10 Beam-switching transmitarray with switched focal sources (a), 60-GHz switched-beam
transmitarray [45]

silicon was developed and a beam-switching capability of ±22° was demonstrated.

In [45], a 60-GHz linearly polarized switched-beam transmitarray was developed
approaching the specifications of point-to-point communications in wireless net-
works (Fig. 6.10b). A linear focal array of five patch antennas was designed
associated to a SP5T switching network. The five sources are separated by 2.6 mm,
the focal ratio F/D is 0.55, and this leads to five beams with a beamwidth of 3.2°,
oriented at 0°, ±2.3°, ±4.6° and an inter-beam crossing level at about −1.5 dB; the
overall −3 dB angular coverage is about ±6.1° in the steering plane. The maximum
gain is 29.3 dBi and the 57–66 GHz band is covered with a gain above 25 dBi.
In the above examples, the transmitarrays were optimized for a source centered
in the focal plane; off-center sources result in a steered beam but also in gain,
beamwidth, and sidelobe degradation due to the oblique illumination. A specific
optimization of the transmitarray can be considered to mitigate this issue and find a
performance trade-off between broadside and steered beam [53].

6.3.4 Reconfigurable Transmitarrays

Reconfigurable transmitarray antennas correspond to structures where switching or

tuning capabilities are integrated inside each unit cell in order to change the
transmitting phase/amplitude/polarization and thereby reconfigure the antenna
radiation patterns to steer the beam, change its spatial distribution (null-steering,
sidelobe level control, beamwidth, multiple beams) or change its polarization. Most
works to date have been focused on beam-steering performance which is the main
applicative need and reconfigurable transmitarrays offer the perspective for more
accurate beam-steering and larger scanning range than using switched focal sources
6 Transmitarray Antennas 213

(i.e., previous section), and faster steering than mechanical schemes. The design of
reconfigurable unit cells with a high efficiency, small size (0.5k0
0.5k0 in area,
low thickness), low power consumption, and simple control has been the focus of
many works for the last 15 years in frequency bands between 5 and 40 GHz.
Targeting the highest frequency bands is also a challenge in terms of integration and
efficiency. Few full-size reconfigurable transmitarrays have been demonstrated to
date with attractive performances and two examples are described hereafter.
In [35], a 6
6-element transmitarray at 5 GHz using varactor diodes as
tuning devices was demonstrated (Fig. 6.11). The array unit cell is composed of
patch antennas on both sides of the panel with a phase-shifter circuit in an
intermediate level [34]; this phase-shifter uses six varactor diodes but a
single-biasing voltage. The unit cell exhibits a 400° phase range and a 3.6 dB
average insertion loss (1.7–5.2 dB). One interest of this structure from the fabri-
cation point of view is the lack of interlayer via interconnections; the signal
transmission between layers is performed using slot coupling or proximity cou-
pling. In contrast, coupling techniques can result in parasitic radiation degrading
the radiation performance of the array in particular in terms of cross-polarization
as observed by the authors. The authors mention that each unit cell was indi-
vidually characterized in transmission phase to generate per-element lookup tables
to control accurately the bias voltages of each varactor diode. The array generates
a peak gain of 15 dBi, a scanning window of 100°
100° and a fractional
bandwidth of about 10% for 2 dB gain variation.
In [28], Di Palma et al. demonstrated a 20
20-element electronically recon-
figurable transmitarray operating in Ka-band (27.4–31.7 GHz) with a beam-steering
range of ±60°, full polarization reconfiguration (linear H/V, circular left/right), a
broad 3-dB bandwidth (14.6%) and a good radiation efficiency of 58% (Fig. 6.12).
The array is based on linearly polarized unit cells with 180° phase switching (1 bit
phase quantization using PIN diodes) and the unit cells are arranged with a random
0°/90°/180°/270° sequential rotation. The random distribution mitigates the issue of

(a) (b) 20
15
10
Gain (dBi)

5
0
-5
-10
-15
-20
-60 -40 -20 0 20 40 60
Elevation (deg.)

Fig. 6.11 6
6-element 5-GHz reconfigurable transmitarray (a) based on varactor diodes
phase-shifters and radiation patterns in the E-plane with beam steered at ±50° (b) [35]
214 L. Dussopt

(a) (b)
Bias lines Radiating aperture 25

15

Gain (dBi)
5

-5

-15

-25
-90 -60 -30 0 30 60 90
Angle (degrees)

Fig. 6.12 Ka-band 20
20-element reconfigurable transmitarray (a), measured radiation patterns
(realized gain) at 29 GHz for right-hand circularly-polarized beams in the horizontal plane (b) [28]

spurious grating lobes occurring in periodic arrays when steering the beam off the
broadside direction. This sequential rotation scheme associated with the 180° phase
switching enables a full polarization reconfigurability. In the case of circular
polarization, the axial ratio remains lower than 2 dB over the full operating
bandwidth of the antenna. The beam can be steered in all azimuth planes up to
±60° with 5 dB scanning loss and sidelobe levels below –12 dB. The array was
demonstrated first with a 10 dBi focal horn source with a focal ratio F/D = 0.60 and
next with a 2
2 planar focal array of SIW slot sub-arrays enabling a focal ratio
reduction down to F/D = 0.36.
While the two selected examples above are based on varactor diodes or PIN
diodes, it is important to note also the works performed using less mature tech-
nologies like RF-MEMS [30], ferroelectric [37], or microfluidics [39].

6.3.5 Focal Distance and Thickness Reduction

The space-feeding architecture of transmitarrays makes them quite bulky and is a

drawback as compared to low-profile antenna solutions like phased arrays in terms
of integration in applicative environments (e.g., vehicles, aircraft, building, portable
systems, etc.). Depending on the focal source beamwidth, the maximum gain and
aperture efficiency are reached for a typical focal ratio F/D of 0.3–1 as illustrated in
Fig. 6.3 and therefore the focal distance is proportional to the transmitarray area. In
this respect, the thickness of high-gain, large-aperture antennas can become sig-
nificantly large and the idea emerged to divide the array into multiple sub-arrays of
smaller area, so shorter focal distance, to make the structure thinner (Fig. 6.13a).
The concept was introduced in the domain of Fresnel lenses, which suffer from the
same issue [54] and was implemented in transmitarrays at X band with four focal
sources and a focal distance reduction of about 1.8 [55]. As a rule of thumb, since
the transmitarray aperture is divided into N sub-arrays (N focal sources), the focal
6 Transmitarray Antennas 215

(a) (b)

X Polar.
Y Polar.

Polar. twisting reflector

(c) (d)

Fig. 6.13 Different schemes for reducing the focal distance and thickness of transmitarray
antennas: multisource architecture (a), folded transmitarray (b), waveguide feed (c), metasurface
feed (d)

pffiffiffiffi
distance can be reduced by about N . In principle, the number N of focal sources
can be increased as necessary to reduce dramatically the structure height but the
benefit of a quasioptically fed structure will be reduced as well by the requirement
of a feed network of the focal sources which need appropriate equal amplitude and
phasing. It can be shown that the multiple focal source configuration can result in an
increase of the transmitarray bandwidth since the distance between each unit cell
and its closest focal source is reduced as compared to a single-source structure of
equivalent area and therefore the differential time delays are reduced as well. It is
worth to note that a multisource configuration can serve other purposes such as
power combining (reducing the output power requirement of each focal source to
reach a targeted radiated power level) or implementing multi-beam schemes for
monopulse radars [56], imaging systems or spatial multiplexing schemes in com-
munication systems.
Another approach is to illuminate the transmitarray indirectly by using a
reflection in a so-called folded transmitarray architecture, which is similar to the
folded reflectarray presented in Chap. 5. The focal source radiates in a polarization
orthogonal to the transmitarray polarization so that the signal is reflected toward a
polarization-twisting reflector. In contrast to the folded reflectarray, the phase
correction to focus the beam is implemented in the transmitting surface and not in
the reflector. To our knowledge, this approach has not been demonstrated so far
with transmitarrays but only with a dielectric lens at 60 GHz in [57] where the focal
distance has been reduced by a factor of about 3.
In [58], the transmitarray panel is no longer fed through a quasioptic radiation
from the source but through guided waves (Fig. 6.13c). A feed network is com-
posed of metallic hollow rectangular waveguides operating in the fundamental
216 L. Dussopt

TE10 mode. The unit cells are on the narrow side of the rectangular waveguides and
each waveguide feeds in series a line of unit cells by coupling (leaking) a small
amount of power to each of them. This approach is demonstrated with a recon-
figurable transmitarray of 6
6 elements at 4.8 GHz and the performances are
shown to be very similar in gain and efficiency to the same array previously
demonstrated with a quasioptic feed [35]. This feeding scheme allows to reduce the
thickness of the array to 0.75k0 only. This comes at the cost of a significant increase
in design complexity to ensure a proper phasing and power distribution across the
aperture. The series feeding scheme leads also to a reduced bandwidth and a more
significant beam-squint effect as in series-fed phased array. This architecture,
composed of a guided feed network, phase-shifters, and radiating elements, comes
close to classical phased arrays with the same difficulties in terms of scaling in size,
weight, design complexity, and scaling to the millimeter-wave domain.
In [59, 60], the feed source is a metasurface antenna, of similar area as the
transmitarray, generating a leaky-wave which illuminates the transmitarray panel
located in its near-field (Fig. 6.13d). The metasurface source is fed in its center, the
guided substrate mode propagates radially, and the surface is designed to generate a
Bessel beam, i.e., with a nearly uniform amplitude, with a radial polarization. The
transmitarray is also a metasurface and is designed to convert the radially polarized
incident wave into the desired (linear or circular) polarization for the radiated beam.
This scheme enables a very thin structure of 0.25k0 only (0.14k0 separation
between the transmitarray and the metasurface feed). The antenna efficiency and
aperture efficiency for a passive (fixed-beam) transmitarray at 9.9 GHz were about
40% and 25%, respectively. As in the previous example (waveguide feed), the main
challenge in this structure is the accurate design and control of the phase and
amplitude distribution across the aperture.

6.4 Conclusion and Perspectives

Transmitarray antennas offer very promising opportunities for future systems,

especially in the millimeter-wave range with a wealth of emerging applications in
the telecommunication sector as well as radar and imaging systems. They have seen
a lot of developments over the last 15 years related to the design of low-loss
structures, especially in the millimeter-wave domain, the investigation of bandwidth
limitations, and development of beam-steering schemes and fully reconfigurable
antennas.
First commercial exploitation of passive transmitarray antennas is expected soon
in the domain of point-to-point millimeter-wave communications but there are still
many innovations and challenges expected to expand their capabilities, such as dual
polarization, dual-band, and multiband operation, as well as filtering properties. The
integration of transmitarrays with their environment is also a topic that has hardly
been addressed so far: codesign with radomes, conformation on nonplanar surfaces,
packaging and codesign of custom focal sources to reduce/cancel spill-over
6 Transmitarray Antennas 217

radiation and optimize the aperture efficiency. New and more complex transmi-
tarray architectures are also a promising path for innovation: some of them have
been described in Sect. 6.3.5 for size reduction like folded transmitarrays or dis-
tributed sources. The emergence of applications in the high millimeter-wave and
submillimeter-wave ranges is also a good opportunity for transmitarrays to replace
bulky and expensive dielectric lenses and mirrors; such development will challenge
current manufacturing technologies to realize multilayer, thin, large-area,
metal-dielectric structures with feature resolutions in the 10-µm range or below.
Most existing works are based on simple modeling and simulation principles
which are good enough for proof of concept but lack the accuracy required for
large-aperture antennas with synthesized beams in some challenging applications,
like satellite communications. In this domain, new simulation and optimization
tools based on multiscale electromagnetic modeling techniques will certainly be
required. Such tools will certainly play a major role in the development of the
complex architectures mentioned above.
Fully reconfigurable transmitarrays are certainly a major challenge for the future
as current structures remain limited in performances and scalability in large aper-
tures. This requires the design of transmitting unit cells with low loss, accurate
phase control over a full 360° range, and simple and scalable control schemes. In
the microwave and low millimeter-wave range, semiconductor switches and tunable
devices will certainly remain the technology of choice to implement such systems.
In the high millimeter-wave range, there are certainly important opportunities for
innovating reconfiguration technologies likes MEMS, NEMS, ferroelectric devices,
or tunable materials.

References

1. Y.J. Guo, S.K. Barton, Fresnel Zone Antennas, (Kluwer academic publishers, 2002)
2. I.V. Minin, O.V. Minin, Basic Principles of Fresnel Antenna Arrays, (Springer, 2010)
3. D.T. McGrath, Planar three-dimensional constrained lenses. IEEE Trans. Antennas Propag.
34(1), 46–50 (1986)
4. W.E. Kock, Metal lens antennas. Proc. I.R.E. 34(11), 826–836 (1946)
5. D.M. Pozar, Flat lens antenna concept using aperture-coupled microstrip patches. Electron.
Lett. 32(23), 2109–2111 (1996)
6. J. Ruze, Wide angle metal-plate optics. Proc. I.R.E. 38(1), 53–59 (1950)
7. L. Schwartzman, L. Topper, Analysis of phased array lenses. IEEE Trans. Antennas Propag.
16(6), 628–632 (1968)
8. C. Chekroun, D. Herrick, Y.M. Michel, R. Pauchard, P. Vidal, Radant—new method of
electronic scanning. Microw. J. 24, 45–53 (1981)
9. Z. Popovic, A. Mortazawi, Quasi-optic transmit/receive front ends. IEEE Trans. Microw.
Theory Technol. 46(11), 1964–1975 (1998)
10. N. Gagnon, A. Petosa, D.A. McNamara, Comparison of 2-, 3- and 4-Layer Phase Shifting
Surface Lens Antennas, in 2013 IEEE AP-S International Symposium (APS URSI), (Orlando,
FL, July 7–13, 2013)
218 L. Dussopt

11. M. Imbert, A. Papió, F. De Flaviis, L. Jofre, J. Romeu, Design and performance evaluation of
a dielectric flat lens antenna for millimeter-wave applications. IEEE Antennas Wirel. Propag.
Lett. 13, 342–345 (2014)
12. P. Padilla, A. Muñoz-Acevedo, M. Sierra-Castañer, Passive microstrip transmitarray lens for
Ku band, in 4th European Conference on Antennas and Propagation (EuCAP), (Barcelona,
Spain, 12–16 April 2010)
13. R.H. Phillion, Flat lenses for circularly polarized electromagnetic waves, in PhD Thesis,
(University of Calgary, AL, Canada, Dec 2010)
14. M.K. Taher Al-Nuaimi, W. Hong, Fabrication and experimental validation of high gain lens
antenna for 71–86 GHz band, in 2015 IEEE Inernational Wireless Symposium (IWS),
(Shenzhen, China, 30 March–1 April, 2015)
15. S.B. Yeap, X. Qing, Z.N. Chen, 77-GHz dual-layer transmit-array for automotive radar
applications. IEEE Trans. Antennas Propag. 63(6), 2833–2837 (2015)
16. A. Abbaspour-Tamijani, K. Saranbandi, G.M. Rebeiz, A millimeter-wave bandpass filter-lens
array. IET Microw. Antennas Propag. 1(2), 388–395 (2007)
17. H. Kaouach, L. Dussopt, J. Lantéri, T. Koleck, R. Sauleau, Wideband low-loss linear and
circular polarization transmit-arrays in V-band. IEEE Trans. Antennas Propag. 59(7), 2513–
2523 (2011)
18. E.G. Plaza, G. Leon, S. Loredo, F. Las-Heras, A simple model for analyzing transmitarray
lenses. IEEE Antennas Propag. Mag. 57(2), 131–144 (2015)
19. N. Gagnon, A. Petosa, and D.A. Mc Namara, Electrically thin free-standing phase and
amplitude shifting surface for beam shaping applications. Microw. Opt. Technol. Lett. 54(7),
1566–1571 (2012)
20. A.H. Abdelrahman, P. Nayeri, A.Z. Elsherbeni, F. Yang, Design of single-feed multi-beam
transmitarray antennas, in IEEE AP-S International Symposium (APS URSI) (Memphis, TN,
7–11 July 2014)
21. A.H. Abdelrahman, A.Z. Elsherbeni, F. Yang, Transmission phase limit of multilayer
frequency-selective surfaces for transmitarray designs. IEEE Trans. Antennas Propag. 62(2),
690–697 (2014)
22. H. Nematollahi, J.-J. Laurin, J.E. Page, J.A. Encinar, Design of broadband transmitarray unit
cells with comparative study of different numbers of layers. IEEE Trans. Antennas Propag. 63
(4), 1473–1481 (2015)
23. M.A. Al-Joumayly, N. Behdad, Wideband planar microwave lenses using sub-wavelength
spatial phase shifters. IEEE Trans. Antennas Propag. 59(12), 4542–4552 (2011)
24. M. Li, N. Behdad, Wideband true-time-delay microwave lenses based on metallo-dielectric
and all-dielectric lowpass frequency selective surfaces. IEEE Trans. Antennas Propag. 61(8),
4109–4119 (2013)
25. M. Li, N. Behdad, Broadband true-time-delay microwave lenses based on miniaturized
element frequency selective surfaces. IEEE Trans. Antennas Propag. 61(3), 1166–1179
(2013)
26. S.H. Zainud-Deen, S.M Gaber, H. Malhat, K.H Awadalla, Single feed dual-polarization
dual-band transmitarray for satellite applications, in 30th National Radio Science Conference,
(Cairo, Egypt, 16–18 April 2013)
27. C. Huang, W. Pan, X. Ma, X. Luo, 1 bit reconfigurable circularly polarized transmit-array in
X band. IEEE Antennas Wireless Propag. Lett. 15, 448–451 (2016)
28. L. Di Palma, A. Clemente, L. Dussopt, R. Sauleau, P. Potier, P. Pouliguen,
Circularly-polarized reconfigurable transmitarray in Ka-band with beam scanning and
polarization switching capabilities. IEEE Trans. Antennas Propag. 65(2), 529–540 (2017)
29. W. Pan, C. Huang, X. Ma, X. Luo, A dual linearly-polarized transmitarray element with 1-bit
phase resolution in X-band. IEEE Antennas Wireless Propag. Lett. 14, 167–170 (2015)
30. C. Chih-Chieh, B. Lakshminarayanan, A. Abbaspour-Tamijani, A programmable lens-array
antenna with monolithically integrated MEMS switches. IEEE Trans. Microw. Theory
Technol. 7(8), 1874–1884 (2009)
6 Transmitarray Antennas 219

31. L. Boccia, I. Russo, G. Amendola, G. Di Massa, Multilayer antenna-filter antenna for

beam-steering transmit-array applications. IEEE Trans. Microw. Theory Tech. 60(7), 2287–
2300 (2012)
32. C. Huang, W. Pan, X. Ma, B. Zhao, J. Cui, X. Luo, Using reconfigurable transmitarray to
achieve beam-steering and polarization manipulation applications. IEEE Trans. Antennas
Propag. 63(11), 4801–4810 (2015)
33. A. Iturri-Hinojosa, J.I. Martinez-Lopez, A.E. Martynyuk, Analysis and design of E-plane
scanning grid arrays. IEEE Trans. Antennas Propag. 58(7), 2266–2274 (2010)
34. J.Y. Lau, S.V. Hum, A wideband reconfigurable transmitarray element. IEEE Trans.
Antennas Propag. 60(3), 1303–1311 (2012)
35. J.Y. Lau, S.V. Hum, Reconfigurable transmitarray design approaches for beamforming
applications. IEEE Trans. Antennas Propag. 60(12), 5679–5689 (2012)
36. P. Padilla, A. Muñoz-Acevedo, M. Sierra-Castaner, M. Sierra-Pérez, Electronically recon-
figurable transmitarray at Ku band for microwave applications. IEEE Trans. Antennas Propag.
58(8), 2571–2579 (2010)
37. M. Sazegar, Y. Zheng, C. Kohler, H. Maune, M. Nikfalazar, J.R. Binder, R. Jakoby, Beam
steering transmitarray using tunable frequency selective surface with integrated ferroelectric
varactors. IEEE Trans. Antennas Propag. 60(12), 5690–5699 (2012)
38. T. Chaloun, C. Hillebrand, C. Waldschmidt, W. Menzel, Active transmitarray submodule for
K/Ka band satcom applications, in 2015 German Microwave Conference (GeMIC),
(Nuremberg, Germany, 16–18 March 2015)
39. E. Erdil, K. Topalli, N.S. Esmaeilzad, O. Zorlu, H. Kulah, O.A. Civi, Reconfigurable nested
ring-split ring transmitarray unit cell employing the element rotation method by microfluidics.
IEEE Trans. Antennas Propag. 63(3), 1163–1167 (2015)
40. C.A. Balanis, Antenna theory, analysis and design, 2nd edn. (Wiley, 1997)
41. A.K. Bhattacharyya, Phased Array Antennas: Floquet Analysis, Synthesis, BFNS and Active
Array Systems, (Wiley-Interscience, 2006)
42. R.C. Hall, R. Mittra, K.M. Mitzner, Analysis of multilayered periodic structures using
generalized scattering matrix theory. IEEE Trans. Antennas Propag. 36(4), 511–517 (1988)
43. M. Steer, Microwave and RF Design: A Systems Approach, (SciTech Publishing, 2010)
44. L. Di Palma, A. Clemente, L. Dussopt, R. Sauleau, P. Potier, P. Pouliguen,
Circularly-polarized transmitarray with sequential rotation in Ka-band. IEEE Trans.
Antennas Propag. 63(11), 5118–5124 (2015)
45. A. Moknache, L. Dussopt, J. Säily, A. Lamminen, M. Kaunisto, J. Aurinsalo, T. Bateman,
J. Francey, A switched-beam linearly-polarized transmitarray antenna for V-band backhaul
applications, in 10th European Conference on Antennas and Propagation (EuCAP), (Davos,
Switzerland, 11–15 April 2016)
46. L. Dussopt, A. Moknache, T. Potelon, R. Sauleau, Switched-beam E-band transmitarray
antenna for point-to-point communications, in 11th European Conference on Antennas and
Propagation (EuCAP), (Paris, France, 19–24 March 2017)
47. N. Gagnon, A. Petosa, D.A. McNamara, Thin microwave quasi-transparent phase-shifting
surface (PSS). IEEE Trans. Antennas Propag. 58(4), 1193–1201 (2010)
48. N. Gagnon, A. Petosa, D.A. Mc Namara, Printed hybrid lens antenna. IEEE Trans. Antennas
Propag. 60(5), 2514–2518 (2012)
49. M.K. Taher Al-Nuaimi, W. Hong, On the beam scanning capabilities of discrete dielectric flat
lens at 73.5 GHz and 83.5 GHz, in 2015 IEEE International Wireless Symposium (IWS),
(Shenzhen, China, 30 March–1 April, 2015)
50. E.B. Lima, S.A. Matos, J.R. Costa, C.A. Fernandes, N.J.G. Fonseca, Circular polarization
wide-angle beam steering at Ka-band by in-plane translation of a plate lens antenna. IEEE
Trans. Antennas Propag. 63(12), 5443–5455 (2015)
51. N. Gagnon, A. Petosa, Using rotatable planar phase shifting surfaces to steer a high-gain
beam. IEEE Trans. Antennas Propag. 61(6), 3086–3092 (2013)
52. J. Zevallos Luna, L. Dussopt, A V-band switched-beam transmit-array antenna. Int.
J. Microw. Wireless Technol. 6(1), 51–56 (2014)
220 L. Dussopt

53. P. Nayeri, F. Yang, A.Z. Elsherbeni, Design of multifocal transmitarray antennas for
beamforming applications, in 2013 IEEE AP-S International Symposium (APS URSI),
(Orlando, FL, 7–13 July 2013)
54. A. Petosa, S. Thirakoune, I.V. Minin, O.V. Minin, Array of hexagonal Fresnel zone plate lens
antennas. IEE Electron. Lett. 42(15), 834–836 (2006)
55. L. Di Palma, A. Clemente, L. Dussopt, R. Sauleau, P. Potier, P. Pouliguen, Design and
experimental characterization of a reconfigurable transmitarray with reduced focal distance.
Int. J. Microw. Wireless Technol. 8(03), 447–454 (2016)
56. L. Di Palma, A. Clemente, L. Dussopt, R. Sauleau, P. Potier, P. Pouliguen, Radiation pattern
synthesis for monopulse radar applications with a reconfigurable transmitarray antenna. IEEE
Trans. Antennas Propag. 64(9), 4148–4154 (2016)
57. O. Soykin, A. Artemenko, A. Mozharovskiy, A. Myskov, R. Maslennikov, Millimeter-wave
high gain lens antenna with reduced focus distance, in European Conference on Antennas and
Propagation (EuCAP), (Davos, Switzerland, 10–15 April 2016)
58. J.G. Nicholls, S.V. Hum, Full-space electronic beam-steering transmitarray with integrated
leaky-wave feed. IEEE Trans. Antennas Propag. 64(8), 3410–3422 (2016)
59. C. Pfeiffer, A. Grbic, Millimeter-wave transmitarrays for wavefront and polarization control.
IEEE Trans. Microw. Theory Technol. 61(12), 4407–4417 (2013)
60. C. Pfeiffer, A. Grbic, Planar lens antennas of subwavelength thickness: collimating
leaky-waves with metasurfaces. IEEE Trans. Antennas Propag. 63(7), 3248–3253 (2015)
Chapter 7
Fabry-Perot Cavity Antennas

Alexandros Feresidis, Konstantinos Konstantinidis

and Peter Gardner

Abstract Fabry-Perot Cavity Antennas (FPAs) are a type of highly directive planar
antennas that offer a promising alternative to standard planar microstrip patch arrays
or waveguide slot array antennas. They offer significant advantages in terms of low
fabrication complexity, particularly at mm wave frequencies, high radiation effi-
ciency, and good radiation pattern performance. These advantages, in conjunction
with a renewed interest in periodic surfaces and meta-surfaces, led to a reinvigo-
ration of international research on this antenna type. This chapter reports recent
advances on the design and implementation of FPAs at mm-wave bands. The main
concept of FPAs, their operating principles and analysis approaches are briefly
introduced. The basic types of FPAs are summarized following a historical account
of various implementations until recent years. The main body of this chapter pro-
vides an overview of recent designs with a main focus on mm-wave bands and the
advantages of the reported antennas for high-frequency implementations.
This chapter reports recent advances on the design and implementation of
Fabry-Perot Cavity-type antennas (FPAs) at mm-wave bands. The main concept of
FPAs, their operating principles, and analysis approaches are briefly introduced.
The basic types of Fabry-Perot cavity antennas are summarized following a his-
torical account of various implementations until recent years. The main body of this
chapter provides an overview of recent designs with a main focus on mm-wave
bands and the advantages of the reported antennas for high-frequency
implementations.

A. Feresidis (&)
K. Konstantinidis
P. Gardner

Department of Electronic Electrical and Systems Engineering,
University of Birmingham, Birmingham B15 2TT, UK
e-mail: [email protected]
P. Gardner
e-mail: [email protected]

© Springer International Publishing AG 2018 221

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_7
222 A. Feresidis et al.

7.1 Introduction

7.1.1 Fabry-Perot Cavity Antenna Concept

The term Fabry-Perot Cavity Antenna (FPA) was coined in recent years for a
relatively simple planar antenna structure that yields high directivity and gain,
offering a promising alternative to standard planar microstrip patch array or
waveguide slot array antennas. FPAs offer significant advantages in terms of low
fabrication complexity, particularly at mm-wave frequencies, high radiation effi-
ciency, and good radiation pattern performance [1]. These advantages, in con-
junction with a renewed interest in periodic surfaces and meta-surfaces, led to a
reinvigoration of international research on this antenna type that led to a surge of
journal and conference publications since the early 2000s.
The basic concept of FPAs, although the term was not used at the time, was first
introduced by Trentini [2] in 1956. This work demonstrated how several types of
periodic surfaces, acting as partially reflective surfaces (PRSs), can increase the
directivity of a single waveguide aperture surrounded by a ground plane. In
mid-1980s, the directivity and gain enhancement produced by dielectric super-
strates over printed antennas was reported independently [3, 4]. Around the same
time, Trentini’s work was revisited by [5], producing highly directive antennas with
multiple periodic superstrates. However, it was not until 2001, when the similarities
of the operating principle of these antennas with the Fabry-Perot interferometer
widely used in optics were reported [1] and a study on the effect of the periodic
superstrate (or PRS) characteristics on the antenna bandwidth and directivity was
performed. A large number of papers have been published since, with the term
Fabry-Perot Cavity antenna used in many of those. However, other terms have also
been used for this type of antennas, such as Electromagnetic Band-Gap
(EBG) resonator antennas, high-gain leaky-wave antennas, or PRS antennas.
The main FPA structure is formed by a partially reflective surface (PRS) placed
at a distance of about half-wavelength and in parallel to a ground plane thus
forming an open Fabry-Perot type resonant cavity. An excitation source is typically
included within the cavity, e.g., a dipole, a microstrip patch or a slot in the ground
plane. The gain and bandwidth of the FPA depend on the reflection (amplitude and
phase) from the PRS as well as the distance from the ground plane. The PRS can be
either a passive periodic array, such as Frequency-Selective Surfaces (FSSs) which
are customarily used for filtering electromagnetic waves, or a metasurface with
sub-wavelength and potentially nonuniform unit cells. Various implementations
have been proposed in recent years, most of them employing 2D metallo-dielectric
arrays of conducting elements [1, 2, 5–12] or metallic periodic arrays of apertures in
a conducting sheet [13–15]. However, different configurations have also been
proposed employing one or more dielectric layers [3, 4, 16–19] with different
dielectric constants and thicknesses functioning as the PRS layer. The use of
dielectric superstrates as the PRS may offer some advantages in terms of reduced
fabrication complexity, however it reduces the degrees of freedom in the design.
7 Fabry-Perot Cavity Antennas 223

7.1.2 General Overview of FPA Analysis Techniques

The work of Trentini [2] first employed a geometric optics approximation to

describe mathematically the antenna operation by means of the multiple reflections
of plane waves between the ground plane and the PRS. The analysis was adopted to
describe the paths of transmitting and reflecting rays from the PRS. This approx-
imate method assumes an infinite extent of the structure and thus is not taking into
account diffraction effects while it also ignores higher order mode coupling. The
approach gives an estimation of the directivity performance after calculating the
reflection characteristics of the PRS structures under plane-wave incidence
assuming infinite size structures. The operating frequency is defined by the cavity
distance, which is typically close to half-wavelength in order to achieve con-
structive interference of the waves bounced between the PRS and the ground.
Another analysis technique for FP antennas is based on the principle of
reciprocity. This technique is based on the principle that a source placed at a point
A in the far-field will provoke an electric field strength at a point B inside the
structure equal to the one that would be provoked at point A, from a source at point
B (Fig. 7.1). Thus, the technique reduces the estimation of the far-field radiation
characteristics to the calculation of the fields at an observation point inside the
antenna when illuminated by a plane wave of fixed magnitude and a varying angle
of incidence. By sampling the relative field strength, at the observation point B, for
various angles of incidence the (receiving) radiation pattern of the antenna can be
obtained. This approach is described in [20, 21] for antennas consisting of a single
periodic layer and also employed in [22, 23] for FP antennas with multiple periodic
layers. The technique can be combined with any other method that allows the
calculation of the near field at point B. For more accurate results, the calculation of
the near field at point B can be done using a rigorous full-wave modeling technique,
for example as in [22].
A more rigorous analysis and design of FP antennas can be performed using
leaky-wave theory. This is due to the partial electromagnetic transparency of the
PRS at frequencies near the resonance and the associated leaky-modes that are
excited. The main information about the operation of leaky-wave antennas can be
provided by investigating the complex wavenumber, k = b − ja, of the
leaky-modes. Obtaining the values of the real part of k gives an estimation of the

Fig. 7.1 Schematic diagram A

of reciprocity technique A
Incident
wave

θ
180
30

B Measurement of the electric field strength

224 A. Feresidis et al.

angle of the main radiated beam, while the imaginary part of k gives information
about the beamwidth of the radiated beam for the corresponding leaky mode. In
[24], a technique based on the transverse equivalent network (TEN) method has
been applied to study periodic patches printed on a grounded slab via a dispersion
analysis, which can give great insight into the operation of FP leaky-wave antenna.
The procedure is employed to extract the dispersion equation of the surface and
leaky-wave modes from the plane-wave reflection data. The homogenization pro-
cess leads to an equivalent admittance for the periodic surface, whose poles and
zeros are derived from the reflection coefficients in the fast-wave region.
More recently, the finite-difference time-domain (FDTD) method has been used
to model the unit cell and rigorously derive the dispersion properties for 2-D FP
antennas [25]. Moreover, LW analysis of FP antennas based on a PRS and a HIS
has been carried out using FDTD [26]. The LWAs under investigation are studied
using a periodic FDTD methodology, initially proposed in [27, 28]. According to
this methodology, any periodic leaky-wave structure can be rigorously modeled by
simulating its unit cell within a FDTD computational domain, terminated with
periodic boundary conditions (PBCs). This type of modeling allows for the com-
putationally efficient calculation of the electromagnetically converged complex
propagation constants of the leaky-modes supported by the infinitely long structures
under consideration.

7.2 Basic Types of FPA Antennas

7.2.1 Metallic and Metallo-Dielectric FP Antennas

As mentioned above, the first and most common implementation of an FP antenna

is the one introduced in [1] which has been based on earlier work [2]. It consists of
a doubly periodic array of metallic elements, printed on a dielectric substrate and
placed at approximately half-wavelength distance over a ground plane (Fig. 7.2).
The antenna excitation is a waveguide-fed slot in the ground plane.

(a) (b)
PRS

λ/2 slot z
x y
waveguide Ground plane

Fig. 7.2 Typical implementation of a Fabry-Perot antenna with metallo-dielectric PRS over a
ground plane and a waveguide-fed slot, b front view of a square element unit cell
7 Fabry-Perot Cavity Antennas 225

Fig. 7.3 a Schematic representation of the antenna based on a Fabry-Perot cavity, L = 3.06 mm,
t = 0.14 mm, dx = dy = 5.8 mm, and h = 9 mm, b directivity versus wavelength (taken from [13])

Fig. 7.4 a 3D (left) and lateral (right) view of a single-layer FPC antenna formed by a thick
metallic FSS with periodic circular holes fed by a magnetic current on its ground. b Similar to
(a) but for the dual-layer cavity antenna (taken from [14])
226 A. Feresidis et al.

This idea has been extensively investigated, employing different types of pri-
mary feeds and geometries for the periodic array. Moreover, a lot of work has been
carried out at mm-wave frequencies, in particular at 44 GHz [8], 60 GHz [9, 10]
and 94 GHz [11].
At the same time, all-metallic PRSs have been introduced [13, 14] as the one
shown in Fig. 7.3. This particular design achieves high directivity performance but
with narrow bandwidth operating at Ku-band (Fig. 7.3b). The PRS is formed by an
array of square apertures on a 1 cm-thick metallic sheet and the antenna is fed by a
square patch (Fig. 7.3a). All the dimensions of the structure are shown in the figure.
Another implementation of FP antennas employing all-metallic PRS is presented in
Fig. 7.4 [14]. In this case, the antenna is designed to operate around 60 GHz and it
consists of a PRS with circular apertures.

7.2.2 Dielectric FP Antennas

The use of single [16] or multiple dielectric layers [3, 4, 17–19] as superstrates for
FP antennas has been investigated since the mid-1980s, Fig. 7.5 [3]. Since the

Fig. 7.5 FP antenna structure, consisting of dielectric layers of alternating thickness and material
constants stacked above a ground plane. The structure is fed by a simple dipole source (taken from
[3])
7 Fabry-Perot Cavity Antennas 227

operation of FP antennas is based on the reflection characteristics of the PRS, high

dielectric constant values were chosen in order to achieve high reflectivity values
which in turn leads to high antenna directivity and gain. Moreover, a large number
of layers are necessary in order to obtain comparable directivity values with the
conventional metallo-dielectric PRS antenna. In the design shown in the figure,
dielectric layers of alternating thickness and permittivity values are stacked on top
of each other. The larger the number of layers used, the greater the achieved
directivity.

7.3 Mm-Wave FPA Implementations

In this section, a review of recent Fabry-Perot antenna implementations at

mm-wave frequencies is presented. Mm-wave antenna designs and implementation
are typically limited by the fabrication process employed and the associated tol-
erances. Some of the highest mm-wave frequency implementations of FPAs are
reported in this section. Furthermore, the reduction of the profile of FPAs has been
studied extensively, by replacing the ground plane with an Artificial Magnetic
Conductor (AMC) surface [29, 30], which leads to quarter-wavelength or even
sub-wavelength profile highly directive antennas [22, 31–38]. Moreover, numerous
implementations with multiple PRS layers have been presented for bandwidth
enhancement of FP antennas [23, 37–43]. Finally, different feeding techniques [8,
10, 44–46] have been studied resulting in alternative configurations providing
enhanced radiation performance.

7.3.1 Higher mm-Wave and Submm-Wave FPAs

Recently, a first report of a submm-wave Fabry-Perot cavity antenna has been

presented employing a fully metallized structure [15]. In this case, the choice of
material was made to avoid dielectric losses which may be detrimental at these
frequencies. The complete antenna structure is shown in Fig. 7.6a and b and
consists of six layers. The first three layers form a waveguide H-bend used in order
to facilitate the interconnection of the antenna with the waveguide flange, while the
remaining three layers form the FPA. The antenna has been fabricated using a
micromachining process based on SU-8 photoresist that has been developed at the
University of Birmingham. The process offers the advantages of high-dimensional
accuracy, high achievable structure aspect ratio, and a capability of large-scale
inexpensive production. Figure 7.6c shows the directivity/gain performance along
with the S11 response. It can be observed that high directivity of 23 dBi has been
achieved at 284 GHz with the proposed design.
Another implementation of FP antenna employing dielectric superstrate is
obtained by using woodpile Electromagnetic Band-Gap (EBG) structures.
228 A. Feresidis et al.

Fig. 7.6 a Cross section and b perspective of the structure (the hollow space is illustrated in blue
color while the surrounding conductors were made transparent to offer a clearer view). c Simulated
directivity, realized gain, and return loss of the proposed antenna (taken from [15])

Fig. 7.7 Geometry of

woodpile superstrate antenna
operating around 90 GHz
(taken from [18])

A representative example of this category is the one presented in [18] (Fig. 7.7). In
the particular design, 3D EBG structures are formed by using ceramic material of
high dielectric constant. The band-gap of the EBG is designed to be around the
desired operating frequency, which in this case is *90 GHz. The woodpile
structure is placed at a distance d = k/2 over a ground plane as illustrated in
Fig. 7.6. The response of the antenna depends on the geometrical parameters of the
woodpile structure and can be modified by varying the filament width (w) and
spacing (a). The directivity performance of the antenna for different w and fixed a is
shown in Fig. 7.8a while for fixed w and different a is shown in Fig. 7.8b. From the
figure, it can be observed that the directivity-bandwidth performance is highly
dependent on these two geometrical parameters.
7 Fabry-Perot Cavity Antennas 229

Fig. 7.8 Directivity versus frequency for: a different filament width w with a = 1.6 mm and
b different filament spacing a with w = 0.4 mm (taken from [18])

7.3.2 Reduced Profile FP Antennas Using HIS

High Impedance Surfaces (HIS) are artificial metamaterial structures introduced by

Sievenpiper [29, 30] that exhibit a reflection phase of zero at a specific frequency
acting as Artificial Magnetic Conductors (AMC). They typically employ 2D peri-
odic arrays of elements on a grounded dielectric substrate. This property has been
exploited to design FP antennas with reduced profile.
More specifically, various implementations have been reported, where the
ground plane of the FP antenna has been replaced by an HIS structure. The concept
is based on altering the resonant condition of the cavity due to the zero reflection
phase of the HIS ground. This has been initially proposed in [31, 32], introducing a
high-gain FP antenna with quarter-wavelength profile. The idea has then been
extensively investigated [22, 33–38] to achieve even further profile reduction. For
example, in [36] a k/60 profile FP antenna is proposed based on an HIS ground and
a composite surface as a superstrate, acting as both a PRS and an AMC (Fig. 7.9).
The composite PRS-AMC surface consists from a square patch periodic array and
an inductive grid printed on either sides of a dielectric substrate. The choice of the
dimensions of the HIS ground and the PRS-AMC surface provides the suitable
reflection phases that result in the desired ultrathin cavity. The antenna is designed
to operate at around 10 GHz achieving a directivity of 18.9 dB. However, this
design has the disadvantage of a narrowband performance due to the rapid
reflection phase variation of both surfaces forming the structure.
Moreover, HIS structures have also been employed for producing circularly
polarized (CP) Fabry-Perot cavity antennas as the one presented in [34]. The
antenna incorporates a double-sided PRS placed above an anisotropic HIS ground
plane (Fig. 7.10) and the cavity is excited by a linearly polarized source rotated by
45°. The CP Fabry-Perot antenna operates at 15 GHz with a 21.1 dB gain. The
circular polarization is obtained by controlling the reflection characteristics of the
HIS and the PRS enforcing the resonance condition for both polarizations along
230 A. Feresidis et al.

Fig. 7.9 a Schematic

diagram of the antenna and
measurement setup, b top
view of the HIS ground plane
with the patch antenna
employed to feed the structure
and c perspective view of the
PRS-AMC unit cell (taken
from [36])

Fig. 7.10 a Schematic

diagram of the antenna, b top
view of the HIS array unit
cell, c top and d bottom view
of the PRS unit cell (taken
from [34])
7 Fabry-Perot Cavity Antennas 231

Fig. 7.11 a Gain (simulated and measured) of finite antenna with a patch antenna source, b Axial
ratio (simulated and measured) of finite antenna with a patch antenna source (taken from [34])

Fig. 7.12 Simulated and measured radiation patterns at 15 GHz for a u = 0°, b u = 90° (taken
from [34])

with a 90° phase difference between them. The simulated and measured gain and
axial ration of the antenna is shown in Fig. 7.11a and b respectively. The axial ratio
at 15 GHz is 0.6 dB with the measured 3-dB axial ratio bandwidth being 9%. In
Fig. 7.12, the radiation patterns for the two principal planes are presented. It can be
observed that highly directive patterns are obtained for both the planes, while the
cross polarization level is less than −20 dB in both planes.

7.3.3 Broadband Multi-layer FP Antennas

The implementations presented thus far, demonstrate some of the advantages of FP

antennas, such as high-gain performance, simple design and feeding techniques.
Moreover, low-profile FP antennas have been obtained with the incorporation and
proper design of HIS structures. However, all these configurations suffer from an
inherit narrowband performance. To address this issue, a technique for bandwidth
improvement was introduced [39] and further investigated, based on the coupling
between two [23] or more [40] PRS layers of dissimilar array dimensions. The
concept exploits the different reflectivity values of the PRS layers to obtain a
232 A. Feresidis et al.

Fig. 7.13 Cross-section of a double-layer free-standing dipole PRS leaky-wave antenna and
b unit cell dimensions of the double-layer dipole PRS (taken from [23])

positive reflection phase gradient with frequency that satisfies the resonance con-
dition of the antenna cavity for a wide frequency range [1].
In [23], the proposed technique has been applied, leading to a double-layer PRS
design with half-wavelength separation between the PRSs and the ground plane
(Fig. 7.13), achieving a significant bandwidth enhancement compared to a
single-layer FP PRS antenna [1]. The PRS layers are formed by 2D periodic arrays
of square patches with the same periodicity but different size, printed on dielectric
substrates (Fig. 7.13b). In Fig. 7.14a, the simulated and measured gain and S11 of
the proposed antenna is depicted. It can be observed that high gain is obtained
between 13.4 and 14.1 GHz, with a maximum value of 17.4 dBi and a 3 dB
bandwidth of 6%. Furthermore, the simulated and measured radiation patterns at the
central frequency for H- and E-plane are presented in Fig. 7.14b, showing good
agreement and validating the concept.
Another approach of the same technique has been adopted in [42], where a
positive PRS reflection phase gradient is achieved employing dipoles printed on
both sides of a dielectric substrate. The unit cell and the complete antenna structure
are illustrated in Fig. 7.15a and b respectively. Three designs were presented with
different dipole dimensions, leading to different performances. Depending on the
obtained reflection phase and magnitude of each double-sided structure, a dual-band
antenna (Fig. 7.16a), and two broadband antennas with different bandwidths and
maximum gain (Fig. 7.16b and c) were obtained.
More recently, a new concept for designing broadband, yet sub-wavelength profile
Fabry-Perot-type antennas has been introduced [37]. The idea is based on combining
the bandwidth enhancement technique (using multiple layers) and the profile
reduction technique (using HIS structures as described in the previous section).
7 Fabry-Perot Cavity Antennas 233

Fig. 7.14 Measured and simulated a gain and S11 response and b H-plane (left) and E-plane
(right) radiation patterns at 13.9 GHz (taken from [23])

Fig. 7.15 Perspective view of a the unit cell of the proposed structure and b the complete antenna
with the PRS, the ground plane and the primary source taken from [42]
234 A. Feresidis et al.

Fig. 7.16 Measured gain and directivity for the three antennas (taken from [42])

More particularly, two composite PRS-AMC layers are employed creating two
sub-wavelength profile Fabry-Perot cavities that significantly enhance the bandwidth
performance of the resulting low-profile antenna (Fig. 7.17). The composite layers
are designed to achieve reflection phases that satisfy the resonance condition for

(a)
(d)
20

15
Directivity,dBi

10

(b) (c)
5

0
13.0 13.2 13.4 13.6 13.8 14.0 14.2 14.4
Frequency,GHz

Fig. 7.17 a Schematic diagram of the proposed antenna (dimensions are not to scale), b top (PRS
array) and c bottom (AMC array) view of the unit cell, and d directivity performance of the
proposed antenna (taken from [37])
7 Fabry-Perot Cavity Antennas 235

quarter-wavelength cavities. Thus, they are placed at about quarter-wavelength from

a ground plane and from each other creating two quarter-wave air cavities with a total
profile of about k/2 (Fig. 7.17). The simulated directivity of the proposed antenna is
illustrated in Fig. 7.17d showing a broadband performance. This concept can be
extended to achieve further profile reduction keeping the increased bandwidth per-
formance. This is obtained using multiple composite layers designed to create k/6 air
cavities [38].

7.3.4 Feeding Techniques for FP Antennas

The design of broadband high-gain FP antennas remains an important challenge.

Over the past few years, several feeding techniques have been proposed to improve
the radiation performance of these antennas. For example, in [45] an array of
multiple sources was employed replacing the single feed in the ground plane, in
order to increase the radiating aperture of the antenna. This led to an increase of the
bandwidth for a fixed gain value and therefore an increase of the gain-bandwidth
product.
In [46] a dual-slot feeding technique is proposed for broadband multi-layer
periodic Fabry-Perot cavity antennas achieving enhanced directivity as well as
broadband radiation and broadband input matching. A microstrip line is printed on
a dielectric substrate with the ground plane printed on the other side of the substrate
and the feeding slots etched off it Fig. 7.18a. The S11 response for a single slot
feeding of the antenna is compared with the proposed dual-slot feeding
(Fig. 7.18b). It can be seen that the introduction of the second slot yields a more
broadband input matching that covers all the operational frequency range.
Moreover, the electric near-field distribution for the three-layer antenna with one
and two feeding slots shown in Fig. 7.18c and d demonstrates that the field dis-
tribution is significantly improved.
In [8] a metallo-dielectric FPC antenna fabricated on a quartz disk is presented
operating at 44 GHz. The choice of the substrate has been made to guarantee
mechanical stability which is particularly important to obtain the expected response.
A planar feed is introduced which is depicted in Fig. 7.19a. The feeding structure is
made of three parts; the Back-Gap (BG) structure for ease of probing, the 50 O
CPW line and a magnetic dipole (slot). The simulated and measured reflection
coefficient of a prototype is shown in Fig. 7.19b. A similar feeding technique has
been used in [10] for operation at 60 GHz.
In [6], a plane-parallel Fabry-Perot resonator comprising two uniform and
nonuniform inductive metal meshes excited by a horn antenna is presented
(Fig. 7.20a). The structure performs beam focusing employing the nonuniform
mirror, whose geometry is chosen so that it behaves as a spherical equiphase surface
with a desired radius of curvature. The directivity of the antenna is controlled by the
curvature of the synthesized wavefront. This concept is successfully validated in the
236 A. Feresidis et al.

(a)

w
s2
y l s1
a b
z x
(b) (c)

(d)

Fig. 7.18 a Back and front view of ground plane with the microstrip line and slots. Cross section
of the electric field distribution (YZ plane) for the three layer Fabry-Perot antenna at 14 GHz b one
slot c two slots d S11 of one and two microstrip fed slots of the proposed three-layer Fabry-Perot
antenna taken from [46]

Fig. 7.19 a Feeding structure of the antenna, b the reflection coefficient of the antenna, a
comparison between the simulation and measurement results (taken from [8])

60 GHz band. The radiation patterns of the proposed antenna are Gaussian with
very low sidelobes (Fig. 7.20b).
Recently, a grid-based multiple-layer FPA has been presented where the layers
are separated by air-filled cavities (Fig. 7.21a) [12]. The FP antenna is fed by a
microstrip patch antenna and operates at 60 GHz. The simulations of these
7 Fabry-Perot Cavity Antennas 237

Fig. 7.20 a Geometry of dielectric-loaded plane-parallel Fabry-Perot cavity excited by horn

antenna, b radiation patterns in H-plane at 57.38 GHz (taken from [7])

superstrate designs showed antenna gains as high as 19 dBi (Fig. 7.21b). These
structures were fabricated on sheets of a flexible, low-loss foam, copper laminate
material, and their performances were measured. Due to the nonuniformity in the
thickness of the foam layers, as well as other material and fabrication issues, the
238 A. Feresidis et al.

Fig. 7.21 a Four-layer grid-based superstrate structure with microstrip patch antenna feeding,
b simulation results for an 11  11 grid structure on two foam layers, optimized in CST with
lossless materials, c measured eye pattern of the proposed antenna excited by a 1.25-Gb/s signal.
(taken from [12])

highest gain achieved, was just over 11 dBi which is significantly lower than the
predicted one. This demonstrates the importance of high accuracy fabrication
processes for the implementation of FP cavity-type antennas at mm-waves.
Importantly, the fabricated antenna prototypes were connected to a practical
60 GHz wireless system in order to measure the bit error rate (BER) of the system
in the presence of the proposed antenna [12]. The maximum data rate of the system
was measured and demonstrated by producing the “eye” diagrams (time-domain
response of the system) (Fig. 7.21c). The results showed that the proposed antennas
produce a significant increase in the system gain for gigabit per second signals in
the 60-GHz band, supporting a maximum error free data rate of 3 Gb/s in this
specific example.
7 Fabry-Perot Cavity Antennas 239

7.4 Discussion and Future Trends

FPAs have attracted increasing research interest over the past decade and this trend
seems to continue. Although the research on FPAs started at lower microwave
frequencies, where there are now already applications on real practical systems
[47], there is now significant effort toward developing FPA-type antennas at higher
mm-wave and THz frequencies. These efforts are taking advantage of emerging
microfabrication technologies as well as new materials that are becoming available.
Some of these efforts have been reported in this chapter, ranging from frequencies
of 10 GHz up to approximately 300 GHz.

References

1. A.P. Feresidis, J.C. Vardaxoglou, High-gain planar antenna using optimized partially
reflective surfaces. IEE Proc. Microw. Antennas Propag. 148(6) (2001)
2. G.V. Trentini, Partially reflecting sheet array. IRE Trans. Antennas Propag. AP-4, 666–671
(1956)
3. D.R. Jackson, A.A. Oliner, A. Ip, Leaky-wave propagation and radiation for a narrow-beam
multiple-layer dielectric structure. IEEE Trans. Antennas Propag. 41, 344–348 (1993)
4. D.R. Jackson, N.G. Alexopoulos, Gain enhancement methods for printed circuit antennas.
IEEE Trans. Antennas Propag. 33, 976–987 (1985)
5. J.R James, S.J.A Kinany, P.D. Peel, G. Andrasic, Leaky-wave multiple dichroic beamformers.
Electron. Lett. 25, 1209–1211 (1989)
6. R. Sauleau, P. Coquet, T. Matsui, J.-P. Daniel, A new concept of focusing antennas using
plane-parallel Fabry-Perot cavities with nonuniform mirrors. IEEE Trans. Antennas Propag.
51(11), 3171–3175 (2003)
7. R. Sauleau, Ph Coquet, D. Thouroude, J.-P. Daniel, Beam focusing using 60-GHz
Fabry-Perot resonators with uniform and non-uniform metal grids. Electron. Letters 39(4),
341–342 (2003)
8. S.A. Hosseini, F. Capolino, F.D. Flaviis, A 44 GHz single-feed Fabry-Perot cavity antenna
designed and fabricated on quartz, in IEEE Antennas and Propagation Society (AP-S)
International Symposium, pp. 1285–1288, Spokane, Washington, USA, July 3–8, 2011
9. S.A. Hosseini, F. Capolino, F. De Flaviis, Design of a single feed all-metal 63 GHz
Fabry-Perot cavity antenna using a TL and a wideband circuit model (IEEE Int.
Symp. Antennas Propag., Charleston, SC, 2009)
10. S.A. Hosseini, F. De Flaviis, F. Capolino, A highly-efficient single-feed planar Fabry-Pérot
cavity antenna for 60 GHz technology, in IEEE International Symposium on Antennas
Propagation, Chicago, IL, USA, July 2012
11. K. Konstantinidis, A.P. Feresidis, P.S. Hall, M.J. Lancaster, Design of Fabry-Perot cavity
antenna at 94 GHz, in Antennas and Propagation Conference (LAPC), 2012 Loughborough,
pp. 1–4, 12–13 November 2012
12. S.J. Franson, R.W. Ziolkowski, Gigabit per second data transfer in high-gain metamaterial
structures at 60 GHz. IEEE Trans. Antennas Propag. 57, 2913–2925 (2009)
13. N. Guerin, S. Enoch, G. Tayeb, P. Sabouroux, P. Vincent, H. Legay, A metallic Fabry-Perot
directive antenna. IEEE Trans. Antennas Propag. 54(1), 220–224 (2006)
14. R. Gardelli, M. Albani, F. Capolino, Gain enhancement of a V-band antenna using a
Fabry-Pérot Cavity with a self-sustained all-metal cap with FSS. IEEE Trans. Antennas
Propag. 63(3) (2015)
240 A. Feresidis et al.

15. K. Konstantinidis, A.P. Feresidis, M.J. Lancaster, P.S. Hall, Micromachined terahertz
Fabry-Perot cavity highly directive antennas. IET Microw. Antennas Propag. 9(13), 1436–
1443 (2015)
16. H. Ostner, E. Schmidhammer, J. Detlefsen, D.R. Jackson, Radiation from dielectric
leaky-wave antennas with circular and rectangular apertures. Electromagnetics 17, 505–535
(1997)
17. A.R. Weily, L. Horvath, K.P. Esselle, B.C. Sanders, T.S. Bird, A planar resonator antenna
based on a woodpile EBG material. IEEE Trans. Antennas Propag. 53(1), 216–223 (2005)
18. Y. Lee, X. Lu, Y. Hao, S. Yang, J.R.G. Evans, C.G. Parini, Low-profile directive millimetre
wave antennas using free-formed three-dimensional (3-D) electromagnetic bandgap struc-
tures. IEEE Trans. Antennas Propag. 57(10), 2893–2903 (2009)
19. H.Y. Yang, N.G. Alexopoulos, Gain enhancement methods for printed circuit antennas
through multiple superstrates. IEEE Trans. Antennas Propag. 35, 860–863 (1987)
20. T. Zhao, D.R. Jackson, J.T. Williams, H.-Y.D. Yang, A.A. Oliner, 2-D periodic leaky-wave
antennas-part I: metal patch design. IEEE Trans. Antennas Propag. 53(11), 3505–3514 (2005)
21. T. Zhao, D.R. Jackson, J.T. Williams, H.-Y.D. Yang, A.A. Oliner, 2-D periodic leaky-wave
antennas-part II: slot design. IEEE Trans. Antennas Propag. 53(11), 3515–3524 (2005)
22. C. Mateo-Segura, G. Goussetis, A.P. Feresidis, Sub-wavelength profile 2-D leaky-wave
antennas with two periodic layers. IEEE Trans. Antennas Propag. 59(2), 416–424 (2011)
23. C. Mateo-Segura, A.P. Feresidis, G. Goussetis, Bandwidth enhancement of 2-D leaky-wave
antennas with double-layer periodic surfaces. IEEE Trans. Antennas Propag. 62(2), 586–593
(2014)
24. S. Maci, M. Caiazzo, A. Cucini, M. Casaletti, A pole-zero matching method for EBG surfaces
composed of a dipole FSS printed on a grounded dielectric slab. IEEE Trans. Antennas
Propag. 53(1), 70–81 (2005)
25. P. Kosmas, A.P. Feresidis, G. Goussetis, Periodic FDTD analysis of a 2-D leaky-wave planar
antenna based on dipole frequency selective surfaces. IEEE Trans. Antennas Propag. 55(7),
2006–2012 (2007)
26. J.R. Kelly, T. Kokkinos, A.P. Feresidis, Analysis and design of sub-wavelength resonant
cavity type 2-D leaky-wave antennas. IEEE Trans. Antennas Propag. 56(9), 2817–2825
(2008)
27. T. Kokkinos, C.D. Sarris, G.V. Eleftheriades, Periodic FDTD analysis of leaky-wave
structures and applications to the analysis of negative-refractive-index leaky-wave antennas.
IEEE Trans. Microw. Theory Techn. 54(4), 1619–1630 (2006)
28. T. Kokkinos, C.D. Sarris, G.V. Eleftheriades, Periodic finite-difference time-domain analysis
of loaded transmission-line negative-refractive-index metamaterials. IEEE Trans. Microwave
Theory Tech. 53(4), 1488–1495 (2005)
29. D. Sievenpiper, High-impedance electromagnetic surfaces. Ph.D. dissertation, Dept. Elect.
Eng., Univ. California at Los Angeles, Los Angeles, CA, 1999
30. D. Sievenpiper, L. Zhang, R.F.J. Broas, N.G. Alexopoulos, E. Yablonovitch, High impedance
electromagnetic surfaces with a forbidden frequency band. IEEE Trans. Microw. Theory
Tech. 47(11), 2059–2074 (1999)
31. S. Wang, A.P. Feresidis, G. Goussetis, J.C. Vardaxoglou, Low-profile resonant cavity antenna
with artificial magnetic conductor ground plane. Electron. Lett. 40(7), 405–406 (2004)
32. A.P. Feresidis, G. Goussetis, S. Wang, J.C. Vardaxoglou, Artificial magnetic conductor
surfaces and their application to low-profile high-gain planar antennas. IEEE Trans. Antennas
Propag. 53(1), 209–215 (2005)
33. J.R. Kelly, T. Kokkinos, A.P. Feresidis, Analysis and design of sub-wavelength resonant
cavity type 2-d leaky-wave antennas. IEEE Trans. Antennas Propag. 56(9), 2817–2825
(2008)
34. R. Orr, G. Goussetis, V. Fusco, Design method for circularly polarized Fabry-Perot cavity
antennas. IEEE Trans. Antennas Propag. 62(1), 19–26 (2014)
35. L. Zhou, H. Li, Y. Qin, Z. Wei, C.T. Chan, Directive emissions from subwavelength
metamaterial-based cavities. Appl. Phys. Lett. 86, 1011011 (2005)
7 Fabry-Perot Cavity Antennas 241

36. A. Ourir, A. de Lustrac, and J.-M. Lourtioz, All-metamaterial-based subwavelength cavities

for ultrathin directive antennas. Appl. Phys. Lett. 88(8), 84103-1–84103-3 (2006)
37. K. Konstantinidis, A. Feresidis, P. Hall, Dual subwavelength Fabry-Perot cavities for
broadband highly directive antennas. IEEE Antennas Wirel. Propag. Lett. 13, 1184–1186
(2014)
38. K. Konstantinidis, A.P. Feresidis, P.S. Hall, Broadband sub-wavelength profile high-gain
antennas based on multi-layer metasurfaces. IEEE Trans. Antennas Propag. 63(1), 423–427
(2015)
39. A.P. Feresidis, J.C. Vardaxoglou, A broadband high-gain resonant cavity antenna with single
feed, in Proceedings of EuCAP, Nice, France, 2006
40. K. Konstantinidis, A.P. Feresidis, P.S. Hall, multilayer partially reflective surfaces for
broadband Fabry-Perot cavity antennas. IEEE Trans. Antennas Propag. 62(7), 3474–3481
(2014)
41. A. Zeb, Y. Ge, K.P. Esselle, Z. Sun, M.E. Tobar, A simple dual-band electromagnetic band
gap resonator antenna based on inverted reflection phase gradient. IEEE Trans. Antennas
Propag. 60(10), 4522–4529 (2012)
42. Y. Ge, K.P. Esselle, T.S. Bird, The use of simple thin partially reflective surfaces with
positive reflection phase gradients to design wideband, low-profile EBG resonator antennas.
IEEE Trans. Antennas Propag. 60(2), 743–750 (2012)
43. M. Al-Tarifi, D. Anagnostou, A. Amert, K. Whites, Bandwidth enhancement of the resonant
cavity antenna by using two dielectric superstrates. IEEE Trans. Antennas Propag. 61, 1898–
1908 (2013)
44. R. Sauleau, Ph Coquet, T. Matsui, Low-profile directive quasi-planar antennas based on
millimeter wave Fabry-Perot cavities. IEE Proc. Microw. Antennas Propag. 150(4), 274–278
(2003)
45. R. Gardelli, M. Albani, F. Capolino, Array thinning by using antennas in a Fabry–Perot cavity
for gain enhancement. IEEE Trans. Antennas Propag. 54(7) (2006)
46. K. Konstantinidis, A.P. Feresidis, P.S. Hall, Dual-slot feeding technique for broadband
Fabry–Perot cavity antennas. IET Microw. Antennas Propag. 9(1) (2015)
47. J.C. Iriarte, I. Ederra, R. Gonzalo, Y. Brand, A. Fourmault, Y. Demers, L. Salgetti-Drioli,
P. de Maagt, EBG superstrate array configuration for the WAAS space segment. IEEE Trans.
Antennas Propag. 57(1), 81–93 (2009)
Chapter 8
Near-Field Focusing by Non-diffracting
Bessel Beams

Mauro Ettorre, Santi Concetto Pavone, Massimiliano Casaletti,

Matteo Albani, Agnese Mazzinghi and Angelo Freni

Abstract This chapter illustrates the capabilities of non-diffractive Bessel beams

for near-field focusing. After a brief introduction of the non-diffractive phenomenon
and its origin, the generation of non-diffractive Bessel beams by inward cylindrical
traveling waves is analyzed in detail. A ray interpretation is proposed for such
beams for infinite and finite radiating apertures. Their main radiation capabilities
and limitations in terms of focusing, bandwidth, and operating range are discussed.
In particular, it is shown that inward cylindrical traveling wave aperture field dis-
tributions can generate non-diffractive Bessel beams over a large bandwidth. As a
practical implementation of the discussed theory, a class of launchers based on
radial waveguides loaded by metallic gratings or slots is proposed, for which an
efficient design and optimization technique is described. In addition, a different and
unconventional leaky-wave approach is also adopted for the design of radial
waveguides loaded by metasurfaces. The chapter ends by outlining the future
research activities and possible applications of non-diffractive beams.

8.1 Introduction

Diffraction is an always-present phenomenon limiting the focusing capabilities of

any radiating wave. The quest for possible solutions overcoming diffraction
spreading has been a topic of interest since a long time, with preliminary works

M. Ettorre (&)
Institut d’Electronique et de Télécommunications de Rennes (IETR),
UMR CNRS 6164, Université de Rennes 1, Rennes, France
e-mail: [email protected]
S.C. Pavone
M. Albani
University of Siena, Siena, Italy
M. Casaletti
Université Pierre and Marie Curie (UPMC), Paris, France
A. Mazzinghi
A. Freni
University of Florence, Florence, Italy

© Springer International Publishing AG 2018 243

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_8
244 M. Ettorre et al.

dating back to the beginning of the last century [1, 2]. However, the seminal works
by Durnin in 1987 gave a completely new light and interest to non-diffractive
waves, especially within the optics and physics communities. Durnin demonstrated
theoretically [3] and verified experimentally [4], the existence of a particular class
of solutions to the scalar wave equation. These solutions, called Bessel beams,
remain confined and do not undergo diffractive
 spreading.
They are of the form Eðq; zÞ ¼ J0 kq q ejkz z , where kq2 þ kz2 ¼ k02 . The
free-space propagation constant is denoted by k0 ; ðq; zÞ are the observation point
coordinates in the cylindrical coordinate system, and J0 ð:Þ is the zeroth-order Bessel
function of the first kind. The name Bessel beam stems from the transverse dis-
tribution of the electric field, as clear from the previous expression. This expression
also shows that the field profile does not change with z, and thus does not exhibit
diffractive spreading. In the ideal case, such beams have a finite energy density, but
are not square integrable. Another peculiar feature of such beams is that their spatial
spectrum consists of a single ring (annulus). In other words, an ideal Bessel beam
can be thought as a superposition of plane waves with propagation directions lying
on a cone. The transverse propagation constant of the beam kq controls the


null-to-null beamwidth (NNBW) of the beam NNBW  4:81=kq . In other words,
the beamwidth can be arbitrarily reduced by increasing the transverse propagation
constant. For 0  kq  k0 , the beam is propagating (along z) whereas for kq [ k0 is
evanescent [5]. It is worth mentioning that Bessel beams are not the only
non-diffractive solution of the wave equation. During the years, other invariant
solutions have been proposed and generalized to the polychromatic case [1, 2, 6–8].
However, this chapter will mainly consider the generation of zeroth-order Bessel
beams at micro- and millimeter waves.
The generation of an ideal Bessel beam requires an infinite amount of energy and
unlimited radiating apertures. In practice, Bessel beams have been generated over a
finite area (aperture size) and did not undergo diffraction over a limited range within
the Fresnel zone of the structure [4, 9]. This operating range is called the
non-diffractive range of the Bessel beam launcher.
At optical frequencies, Bessel beams are typically generated by using axicons [4,
9–11] which are conical lenses able to convert an impinging Gaussian beam into a
Bessel one in a bi-conical region close to the axicon. Other methods considered
holograms [12], localized modes [13], Fabry–Perot cavities [14, 15], just to mention
a few. In optics, Bessel beams have been used in a wide range of applications
including laser machining, biosensing, optical trapping of nanoparticles, etc.
[16, 17].
At radio frequencies, Bessel beams have recently received an increased attention
and several experimental implementations of Bessel beam launchers have been
reported from the micro to the Terahertz frequency range. These implementations
include axicons [18], holograms [19], metallic circular gratings [20, 21], parabolic
reflectors [22], planar arrays [23], open circular waveguides [24], metamaterials and
metasurfaces [25–32], near-field plates [33], and radial line slot arrays (RLSA)
[34, 35]. Among the cited works, this chapter will focus the attention on the Bessel
8 Near-Field Focusing by Non-diffracting Bessel Beams 245

launchers using radial waveguides loaded with slots (RLSA) or metallic gratings
and by metasurfaces. This chapter is organized as follows.
Section 8.2 provides an accurate analysis of the non-diffraction radiation gen-
erated by inward cylindrical traveling waves. This analysis completes the general
approach followed in literature where standing cylindrical waves are considered for
the generation of the non-diffractive radiation [1, 2]. The Geometrical Optics
(GO) and Space Wave (SW) contributions to the field radiated by infinite radiating
apertures are given in closed-form expressions (Sect. 8.2.1). The SW contribution is
expressed by using incomplete Hankel functions. It is clearly shown that the GO
contribution creates a non-diffractive Bessel beam close to the axis of symmetry of
the generating aperture, where the SW is negligible. Therefore, non-diffractive
radiation is not only generated by standing cylindrical waves (Bessel-like distri-
butions) thus simplifying the beam generator synthesis. In contrast, outward
cylindrical waves over an infinite radiating aperture cannot provide a
non-diffractive beam. The finite aperture case is then considered in Sect. 8.2.2. The
radiated field is decomposed in its GO and diffractive (D) contributions. The GO
term defines the region in which non-diffractive Bessel beams can be generated,
whereas the nonuniform description of the edge-diffracted field gives a correction to
the GO field and rigorously shows why the non-diffractive radiation is limited by
the non-diffractive range. In addition, it is found that the interface between the GO
and the D contributions produces field amplitude ripples over the longitudinal axis
as described in Sect. 8.2.3.
Section 8.3 considers the practical generation of non-diffractive radiation. An
equivalent current approach is proposed in Sect. 8.3.1 to implement the cylindrical
aperture field distributions discussed in Sect. 8.2 with RLSAs and radial waveg-
uides loaded with metallic gratings. Section 8.3.2 presents the slot layout for RLSA
Bessel beam launchers and its main design parameters. An innovative and fast
optimization tool is described in Sect. 8.3.3 for the synthesis of the required
aperture current distribution with RLSA launchers and metallic gratings. The design
procedure is used in Sect. 8.3.4 to design a circular-polarized RLSA Bessel
launcher operating at 15 GHz. The same procedure is adopted for the Bessel
launcher presented in Sect. 8.3.5 operating at 30 GHz and made by a radial
waveguide loaded with metallic gratings.
Section 8.4 presents the generation of Bessel beams by cylindrical leaky-wave
modes. The motivations behind the proposed design are introduced in Sect. 8.4.1. It
is shown that a radial waveguide loaded with a homogenous impedance sheet can
radiate non-diffractive Bessel beams as long as the structure support leaky-wave
modes. A detailed field analysis is thus presented in Sect. 8.4.2, where the main
design criteria and limitations are provided. In particular, it is shown that the
non-diffractive radiation is achieved as a combination of inward and outward
traveling leaky-wave modes within the radial waveguide. Section 8.4.3 introduces
the dispersion analysis of the proposed launcher and give the main design equa-
tions. Section 8.4.4 presents two prototypes of the proposed launcher operating in
X- and Ka-band. The prototypes are based on a radial waveguide loaded by a
uniform impedance sheet. The impedance sheets are synthesized by uniform
246 M. Ettorre et al.

metasurfaces (please refer to Chap. 11 for more details on metasurfaces). An

innovative leaky-wave launcher based on a modulated metasurface is also presented
in Sect. 8.4.5.
Finally, Sect. 8.5 concludes the chapter by outlining the current applications and
novel research venues of non-diffractive Bessel beams. Particular attention is
devoted to Bessel beam launchers operating over a large bandwidth for the gen-
eration of non-diffractive pulses at millimeter waves.

8.2 Non-diffractive Radiation

8.2.1 Analytical Formulation for Infinite Apertures

The geometry of the problem is shown in Fig. 8.1. The origin of the reference
system is at the aperture center with the z-axis normal to the aperture. For the sake
of convenience, the observation point is expressed either in cylindrical ðq; /; zÞ or
in spherical ðq; h; /Þ coordinates; vectors are bold and a hat denotes a unit vector.
In the following discussion, Transverse Magnetic (TM) modes with respect to the z-
direction will only be considered. However, a similar procedure can be applied to
Transverse Electric (TE) modes. An inward cylindrical wave is assumed for the
magnetic current distribution on the radiating aperture, namely

ð1Þ
Mðq; /; z ¼ 0Þ ¼ Mðq; z ¼ 0Þ^
u ¼ H1 ðkqa qÞ^
u; ð8:1Þ

Fig. 8.1 Schematic view of

the considered configuration
for the generation of a Bessel
beam. The aperture field
distribution at z = 0 is a
cylindrical inward wave with
radial propagation constant
kqa . The non-diffractive zone
is limited by the cone

with

angle ha ¼ arcsin kqa =k ,
where k is the free-space
wavenumber
8 Near-Field Focusing by Non-diffracting Bessel Beams 247

where kqa is the radial propagation constant of the cylindrical inward wave and
ðiÞ
Hn ð:Þ is the n-th order Hankel function of the i-th kind.
A time dependence ejxt ; x ¼ 2pf being the angular frequency, is assumed and
suppressed. The electric field radiated by the aperture is given by [36]

Zþ 1  
1 kq ð2Þ ð2Þ e q Þejkz z kq dkq ;
Eðq; zÞ ¼ H ðkq qÞ^z þ jH1 ðkq qÞ^
q Mðk ð8:2Þ
4p kz 0
1ejp

Zþ 1
e q Þ ¼ 2pj
Mðk Et ðq; z ¼ 0ÞJ1 ðkq qÞqdq; ð8:3Þ
0

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where kq , and kz ¼ k2  kq2 are the transverse and longitudinal spectral variables,
ðiÞ
k is the wavenumber in free space, whereas Jn ð:Þ and Hn ð:Þ are the n-th order
Bessel and Hankel functions of the i-th kind, respectively. Equation (8.3) is the
Hankel transform of the magnetic current distribution over the aperture and, for the
assumed inward cylindrical wave distribution, it is given by

e ðkq Þ ¼ 4kq
M : ð8:4Þ
kqa ðkq2  kqa
2 Þ

The radiated electric field can be exactly expressed as the superposition of the
GO and SW contributions [35], namely

EðrÞ ¼ EGO ðrÞ þ ESW ðrÞ; ð8:5Þ

with
 
kqa jkza z
E GO
ðrÞ ¼ 2 J0 ðkqa qÞ^z þ J1 ðkqa qÞ^
q e Uðha  hÞ
jkza
  ð8:6Þ
kqa ð1Þ ð1Þ jkza z
þ H ðkqa qÞ^z þ H1 ðkqa qÞ^ q e Uðh  ha Þ;
jkza 0

kqa h ð2Þ    jk z

  ð2Þ
i
ESW ðrÞ ¼  sgnðw0 ÞH0 ðkqa q; w0 Þe
za
 H0 ðkqa q; w0þ Þejkza z ^z
jkza
h    jk z i 2  ejkr
 sgnðw
ð2Þ   ð2Þ
þ H1 ðkqa q; w0þ Þejkza z ^ ^z þ cot h^
0 ÞH1 ðkqa q; w0 Þe q  jpk q r
za

qa
2 sin h ejkr
 ^;
jpkqa ðcos h  cos ha Þ r h
2 2

ð8:7Þ
248 M. Ettorre et al.

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
with kza ¼ k2  kqa 2 denoting the Bessel beam normal propagation constant. In

ðiÞ
(8.7), Hn ðX; w0 Þ are the n-th order incomplete Hankel functions of i-th kind with
the second argument w 1 1
0 ¼ tanh cos h  tanh cosha [37], whereas sgnð:Þ and
Uð:Þ are the sign and Heaviside step function, respectively. The GO contribution
exhibits the well-known Bessel beam shape (first term in (8.6)) within the cone
h\ha where Uðh  ha Þ ¼ 1, whereas becomes an inward Hankel beam outside
such a cone where Uðh  ha Þ ¼ 0.
By considering the Bessel function as the superposition of two Hankel functions
ð1Þ ð2Þ
2Jn ¼ Hn þ Hn , GO is interpreted as the superposition of two ray contributions,
namely, an inward Hankel beam conical wave is associated with a ray arising from
a point on the aperture, and an outward Hankel beam conical wave is associated
with a ray arising from a point on the aperture which is diametrically opposite with
respect to the observation point. Indeed, the launched inward conical wave becomes
an outward conical wave beyond the caustic at the z-axis. While the former ray
reaches the observation point regardless its location, the latter exists only when
observing inside the cone h\ha (Fig. 8.2a). At the GO discontinuity cone h ¼ ha ,
the SW contribution (8.7) exhibits an opposite abrupt discontinuity which renders
the total field smooth and continuous. Such a transitional behavior is described by
the sign and incomplete Hankel functions when w 0 ! 0. Outside the transition
region near the discontinuity cone, the SW contribution exhibits a ray-optical
behavior which is derived from the asymptotic expression of incomplete Hankel
functions for large arguments [37] and reported in the last line of (8.7). Indeed, the
SW is a transverse (i.e.,^
h polarized) spherical wave associated with a ray launched
at the aperture center (origin of the reference system in Fig. 8.1), with a radiation
null on the aperture symmetry axis h ¼ 0.
As an example, Fig. 8.3 shows the amplitude of the z (Fig. 8.3a, d, and g) and q
(Fig. 8.3b, e and h) components and the total electric field (Fig. 8.3c, f and i)
radiated by a standard Bessel distribution (Fig. 8.3a, b and c), by an inward
cylindrical wave (Fig. 8.3d, e and f), and by an outward cylindrical wave

Fig. 8.2 Ray interpretation of the GO field for an infinite aperture. a Inward Hankel aperture
distribution: the GO field comprises an inward Hankel beam ray (in red), present throughout the
space, and an outward Hankel beam ray (in green) bounded inside the cone h\ha ; the
superposition of the two rays inside the cone (yellow area) creates a Bessel beam. b Outward
Hankel aperture distribution: the GO field comprises only an outward Hankel beam ray (in green)
bounded outside the cone h\ha [38]
8 Near-Field Focusing by Non-diffracting Bessel Beams 249

 
Fig. 8.3 Electric field radiated by an infinite aperture. jEz j (a, d, g), Eq  (b, e, h) and total electric
field amplitude jE j (c, f, i). Standard Bessel beam reference field (a, b, c), field radiated by an
inward Hankel distribution (d, e, f), and field radiated by an outward Hankel distribution (g, h, i).
The axes are normalized with respect to the wavelength ðkÞ at the operating frequency. The dashed
line marks the GO boundary h ¼ ha

(Fig. 8.3g, h and i), with kqa ¼ 0:6k in the vertical qz plane. As clear from the
previous results, the non-diffractive behavior of the field radiated by an inward
cylindrical wave can be appreciated within a cone with angle ha  37 (dotted line)
where the various components of the electric field (Fig. 8.3d, e, and f) recover the
respective components of a standard Bessel beam (Fig. 8.3a, b, and c). By repeating
the same formulation for the case of an outward traveling wave over the radiating
ð2Þ
aperture Mðq; /; z ¼ 0Þ ¼ H1 ðkqa qÞ^ u, the radiated field is found again as in
Eq. (8.5) with the following expression for the GO contribution:
 
kqa ð2Þ ð2Þ jkza z
E GO
ðrÞ ¼ H ðkqa qÞ^z þ H1 ðkqa qÞ^
q e Uðh  ha Þ; ð8:8Þ
jkza 0
250 M. Ettorre et al.

and a SW contribution which is the negative of that in (8.7). In (8.8), differently

from (8.6), the GO field contribution is constituted by only a single outward Hankel
beam conical wave, which however vanishes within the cone delimited by h\ha .
Its ray interpretation is shown in Fig. 8.2. Again, the SW contribution discontinuity
perfectly matches the GO jump at the shadow boundary cone, thus providing a
smooth continuous total field. Therefore, it is apparent that an outward Hankel
distribution cannot produce a Bessel beam, as also clear from Fig. 8.3g, h, and i.

8.2.2 Nonuniform Asymptotic Field Evaluation for Finite

First Kind Hankel Aperture Distributions

Once established that an infinite inward Hankel current distribution is able to

generate a non-diffractive Bessel beam in the longitudinal component of the electric
field, let us assume a finite equivalent magnetic current distribution of radius a and
directed along the /-axis over a planar aperture. The Hankel transform of such
equivalent magnetic current distribution has the form
 
~ qÞ ¼ 4 kq pa  ð1Þ ð1Þ
Mðk þ kq H1 ðkqa aÞJ0 ðkq aÞ  kqa H0 ðkqa aÞJ1 ðkq aÞ :
kq2  kqa
2 kqa 2j
ð8:9Þ

~ q Þ more suitable for an asymptotic evaluation, it is convenient to

To render Mðk
recast its expression as the sum of three terms

~ qÞ ¼ M
Mðk ~ 1 ðkq Þ þ M
~ þ ðkq Þ þ M
~  ðkq Þ; ð8:10Þ

where

~ 1 ðkq Þ ¼  kq
M
4
; ð8:11Þ
kqa kq2  kqa
2

is the Hankel transform of the magnetic current distribution over an infinite aperture
(refer to (8.4)), whereas

~  ðkq Þ ¼ jpa ð1Þ ð2;1Þ ð1Þ ð2;1Þ

M ½kq H1 ðkqa aÞH0 ðkq aÞ  kqa H0 ðkqa aÞH1 ðkq aÞ ; ð8:12Þ
kq2  kqa
2

in which the Hankel functions of second or first kind apply to the upper or lower
sign, respectively, and arise because of the aperture truncation whose radius is
denoted by a.
By using (8.12), the S scalar potential introduced in the appendix of [38] is split
into three terms, namely S1; þ ; , in which the corresponding Me 1; are used. Such
8 Near-Field Focusing by Non-diffracting Bessel Beams 251

Fig. 8.4 Topology of the

integrand function in (8.4).
The deformation of the initial
integration path (red line) into
the SDP (blue line) is shown
for completeness. The black
lines correspond to the brunch
cuts associated with the
Hankel functions and the
normal propagation
constant kz

integrals can be asymptotically evaluated by using the Steepest Descent Path

(SDP) method. As far as S1 is considered, its spectral integral representation
exhibits a saddle point located at kq ¼ k sin h and pole singularities at kq ¼ kqa ,
which are detoured by the integration contour as in Fig. 8.4. The integration path is
therefore deformed onto the SDP through the saddle point, and the pole residue
contribution has to be taken into account when poles are captured during the path
deformation. Hence, the infinite aperture TM scalar potential asymptotically
reduces to the sum of a Geometrical Optics (GO) and a diffracted (D) rays,

S1 ðrÞ ¼ SGO
1 ðrÞ þ S1 ðrÞ
D
ð8:13Þ

corresponding to the residue

ejkza z h ð2Þ ð1Þ

i
1 ðrÞ ¼
SGO H0 ðkqa qÞUðha  hÞ þ H0 ðkqa qÞ ; ð8:14Þ
jkqa kza
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
with kza ¼ k2  kqa 2 , and the SDP

Z jkz z
1 ~ 1 ðkq ÞH0ð2Þ ðkq qÞ e
1 ðrÞ ¼
SD M dkq ; ð8:15Þ
4p kz
SDP

contributions, respectively. In (8.14), Uð:Þ is the unit step function, which vanishes
for negative argument, i.e., in the GO shadow region h [ ha ¼ arcsinðkqa =kÞ where
the pole at kq ¼ kqa in Fig. 8.4 is not captured.
The diffractive contribution given in (8.15) can be asymptotically evaluated as
the saddle point contribution, leading to a closed-form Geometrical Theory of
Diffraction (GTD) ray description

2 ejkr
1 ðrÞ 
SD
 : ð8:16Þ
jpk3 sin ha sin h  sin ha r
2 2
252 M. Ettorre et al.

Fig. 8.5 Aperture geometry

and map of the GO
contribution to the
z-component of the electric
field radiated by a finite
inward cylindrical traveling
wave aperture distribution.
The Bessel beam is generated
in the yellow region around
the longitudinal z-axis

The above expression reveals that the diffractive contribution of the infinite
aperture S1 asymptotically behaves like a spherical wave (space wave) arising from
the aperture center. According to the spherical wave spreading, SD decays as 1=r,
unlike SGO , whose amplitude does not decay, thus the diffractive contribution
becomes more and more negligible with respect to the GO contribution moving
away from the aperture. In addition, SD1 exhibits an angular pattern with a zero on
the longitudinal axis ðh ¼ 0Þ. Therefore, its contribution close to the axis is par-
ticularly weak. The nonuniform GTD asymptotic expression (8.16) fails at the GO
shadow boundary h ¼ ha . A more sophisticated uniform asymptotics might cure
such a singularity, but it is beyond the purpose of the present chapter.
An analogous procedure can be followed for the calculation of the contributions
to the potentials S corresponding to M ~  . Their spectral integral representation
exhibits a saddle point at kq ¼ k sin h , with h ¼ arctan½ðq  aÞ=z (refer to
Fig. 8.5), and one pole singularity at kq ¼ kqa . By resorting to the path defor-
mation onto the SDP, one obtains

S ðrÞ ¼ SGO
 ðrÞ þ S ðrÞ;
D
ð8:17Þ

in which

1 ð2;1Þ
 ðrÞ ¼
SGO H ðkqa qÞejkza z Uðha
h Þ; ð8:18Þ
jkqa kza 0

where the Hankel functions of second or first kind, respectively, apply to the upper
or lower sign, and
Z jkz z
1 ~  ðkq ÞH ð2Þ ðkq qÞ e
 ðrÞ ¼
SD M 0 dkq : ð8:19Þ
4p kz
SDP

Also in this case, the GTD asymptotic evaluation of (8.19) provides insight
about the ray nature of these wave constituents
8 Near-Field Focusing by Non-diffracting Bessel Beams 253

p pffiffiffi ð1Þ
jej4 aH0 ðkqa aÞ ejkr
SD
 ðrÞ  p ffiffiffiffiffiffiffi
ffi p ffiffiffiffiffiffiffiffi ; ð8:20Þ
2pk k2 sin h ðsin h
sin ha Þ qr
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
with r ¼ ðq  aÞ2 þ z2 . Indeed SD  appears in the form of a cylindrical
edge-diffracted wave, arising from the aperture rim, as can be inferred from the
pffiffiffiffiffi
ejkr = r factor. Such a GTD expression fails at the corresponding GO shadow
boundary ðh ¼ ha Þ and at the edge-diffracted wave caustic on the longitudinal z-
axis ðq ¼ 0Þ, as expected. After gathering all the GO and D terms, the scalar
potential can be finally arranged in the following form:

SðrÞ ¼ SGO ðrÞ þ SD

1 ðrÞ þ S þ ðrÞ þ S ðrÞ
D D
ð8:21Þ

in which the overall GO term is

ejkza z ð2Þ ð1Þ

SGO ðrÞ ¼ fH ðkqa qÞ½Uðha  hÞ  Uðha  h þ Þ þ H0 ðkqa qÞ
jkqa kza 0
Uðh  ha Þg: ð8:22Þ

It is worth noting that in the case of an infinitely large aperture a ! þ 1,

h ! p=2, and in turn Uðha
h Þ ! 0, whence, in (8.18), SGO  ! 0. Also,
pffiffiffi ð1Þ
r ! þ 1 and since in (8.20) aH0 ðkqa aÞ remains limited, it follows that
 ! 0. Therefore, it is easily proved that as a ! þ 1, then the scalar potential
SD
reduces to that of the infinite aperture case reported in [38], namely S ! SGO1 þ S1 ,
D

as expected. Note that such a beam is not a pure Bessel beam throughout the space,
but it comprises a Bessel beam, which is confined within the conical region h\ha ,
and a diffractive space wave component, though it is negligible close to, and at the
z-axis.
From the scalar potential, the GO contribution to the electric field can be exactly
calculated by differentiating the GO contribution of the scalar potential (refer to the
appendix of [38]) as follows:
 
kqa ð2Þ ð2Þ jkza z
EGO ðrÞ ¼ H0 ðkqa qÞ^z þ H1 ðkqa qÞ^
q ½Uðha  hÞ  Uðha  h þ Þ e
jkza
 
kqa ð1Þ ð1Þ jkza z
þ H ðkqa qÞ^z þ H1 ðkqa qÞ^q Uðh  ha Þe :
jkza 0
ð8:23Þ

Focusing the attention on the z-component of the previous expression, it can be

noticed that
254 M. Ettorre et al.

• In the region defined by 0\jhj\ha ; ha \h þ \p=2, and ha \  h \p=2, a

Bessel beam can be established, since the GO contribution to the electric field is
proportional to J0 ðkqa qÞejkza z , as highlighted in Fig. 8.5 (yellow region).
• In the regions 0\jhj\ha ; ha \h þ \p=2; h \ha or 0\h\ha ; h þ
\ha ; ha \  h \p=2, the field is described by an outward cylindrical traveling
ð2Þ
wave H0 ðkqa qÞejkza z , as highlighted in Fig. 8.5 (green regions).
• In the regions delimited by ha \jhj\p=2; ha \h þ \p=2, and ha \  h \p=2,
ð1Þ
the field is described by an inward cylindrical traveling wave H0 ðkqa qÞejkza z ,
as highlighted in Fig. 8.5 (red regions).
The diffractive contributions to the radiated field can be evaluated by differen-
tiating the respective potential terms SD1; þ ; [20] which, in the asymptotic limit,
lead to the space wave ray fields

2j sin h ejkr
1 ðrÞ 
ED
2  ^ ð8:24Þ
pk sin ha sin h  sin ha r h
2

2j sin h ejkr
1 ðrÞ 
HD
2  ^; ð8:25Þ
pkg sin ha sin h  sin2 ha r /

and the edge-diffracted ray fields

p pffiffiffi ð1Þ
ej4 aH0 ðkqa aÞ ejkr
 ðrÞ
ED ¼ pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi ^ ð8:26Þ
j 2pkðsin h
sin ha Þ qr h
p pffiffiffi ð1Þ
ej4 aH0 ðkqa aÞ ejr
 ðrÞ
HD ¼ pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi ^ : ð8:27Þ
j 2pkgðsin h
sin ha Þ qr /

The various ray constituents (Fig. 8.5) of the diffracted fields inherit the same
shape and behavior of the respective contributions of the potential. According to
ray-optical approximation, field polarization is transverse to the ray, i.e., tangent to
the wavefront. The diffracted magnetic field ray contributions HD 1; are all polar-
ized along /^ orthogonally to the picture plane in Fig. 8.5, whereas the diffracted
electric field ray contributions ED
1; are polarized along the local unit vectors ^ h; ^
h
transverse to the ray and parallel to the picture plane in Fig. 8.5.
The above ray format expressions for the various wave constituents provide a
neat physical insight to the Bessel beam generation by an inward cylindrical
traveling wave aperture distribution and clearly show that its capabilities are similar
to those of commonly used Bessel aperture distributions. The above expressions fail
close to the z-axis, which is of particular interest for Bessel beams. In the next
section we provide alternative expressions valid in such region.
8 Near-Field Focusing by Non-diffracting Bessel Beams 255

8.2.3 Evaluation of the Electromagnetic Field Along

the Longitudinal Z-Axis

Let us focus our attention on the z-axis, corresponding to q ¼ 0. Here it is con-

venient to rewrite the TM scalar potential introduced in [38], to calculate the
electromagnetic field radiated by Hankel aperture distributions in terms of Bessel
functions, namely
Z þ1 jkz z
1 ~ q ÞJ0 ðkq qÞ e
SðrÞ ¼ Mðk dkq : ð8:28Þ
2p 0 kz

By differentiating the previous expression [20], the electric field can be evaluated
on the longitudinal z-axis and it is clear that Eq and H/ vanish, as expected due to
the cylindrical symmetry of the radiating aperture, whereas the longitudinal com-
ponent of the electric field can be evaluated as
Z þ1
1 ~ qÞ ejkz z
Eðz^zÞ ¼ kq2 Mðk dkq bz : ð8:29Þ
2p 0 kz

As in Sect. 8.2.2, the Hankel transform of the equivalent magnetic current is split
into the three terms and the resulting three integrals are evaluated by deforming the
integration path onto the SDP. The residue contributions corresponding to the GO
terms are the same as in (8.23) once evaluated at q ¼ 0, thus reducing to

2kqa jkza z
EGO ðz^zÞ ¼ e Uðha  h0 Þ^z; ð8:30Þ
jkza

in which h0 ¼ arctanða=zÞ is the limiting value of h þ and h , when q ! 0 (refer

to Fig. 8.5). The saddle point contribution relevant to the infinite aperture exhibits
now a quadratic zero of the integrand function at the saddle point itself. Therefore,
its asymptotic value is very weak and corresponds to a longitudinally polarized ray
contribution

4k ejkz
1 ðz^
ED zÞ   3
^z; ð8:31Þ
pkqa z2

indeed the leading asymptotic term of the space wave (8.31) vanishes on the axis,
where h ¼ 0 and r ¼ z. The other two saddle point contributions provide the
expression for the edge-diffracted rays at their caustic on the z-axis

ð1Þ
H0 ðkqa aÞ sin2 h0 jkr0
EDþ ðz^zÞ þ ED
 ðz^
zÞ  e ^z; ð8:32Þ
jðsin h0  sin ha Þ
256 M. Ettorre et al.

pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where (8.26) fails. In (8.32), r0 ¼ a2 þ z2 is the limiting value of r þ and r , when
q ! 0 (refer to Fig. 8.5).
The above GTD asymptotic expression is nonuniform with respect to the pole
singularity which might lie in the neighborhood or at the saddle point and fails at
the GO shadow boundary h0 ¼ ha , or equivalently z ¼ zNDR ¼ a cot ha , where
the GO contribution (8.30) abruptly disappears. zNDR is commonly referred as
non-diffractive range.
In conclusion, on the z-axis the total electric field results from the summation of
the three wave constituents, namely GO (8.30), space wave (8.31), and
edge-diffracted wave (8.32). As the observation point moves away from the aper-
ture, the space wave contribution rapidly decays and becomes negligible. Therefore,
in addition to the ideal Bessel beam GO contribution, a spurious diffractive con-
tribution arises from the aperture truncation. This diffractive contribution interferes
with the dominant term and creates an amplitude ripple in the total field. Such an
effect is present also in the field radiated by a standard Bessel distribution and it is
unavoidably associated to the aperture finiteness. It can be mitigated by increasing
the aperture size, as it is apparent from the dependence on a in (8.32).

8.3 Generation of Non-diffractive Radiation

8.3.1 Equivalent Aperture Field Distribution

According to the Schelkunoff theorem, the aperture field distributions discussed in

Sect. 8.2 are equivalently generated in the half-space z [ 0 by a distribution of
magnetic surface currents, radiating on a perfect electric conductor, of the form

M ¼ E  ^zjz¼0 : ð8:33Þ

In particular for the generation of a circularly polarized Bessel beam such current
is given by [34, 39]:


^RH;LH ;
M ¼ E  ^zjz¼0 ¼ A0 J0 kq q p ð8:34Þ

pffiffiffi
where p ^RH;LH ¼ ð^x
j^yÞ 2 for right-hand and left-hand circularly polarized
beams.
In the following sections, it will be discussed the possibility to synthetize such
current distributions or equivalently the associate aperture field by using a radial
waveguide loaded with concentric circular slits or sub-resonant slots. A holographic
approach will be adopted. Holography predicts that a certain antenna aperture
equivalent current distribution is produced by the interference between the field
map generated by the input feeding wave alone and the hologram. In our cases the
latter is the field generated by the array radiating elements [40].
8 Near-Field Focusing by Non-diffracting Bessel Beams 257

The case of sub-resonant slots distribution will be treated in detail in the next
section for the design and optimization of a circularly polarized radial line slot array
(CP-RLSA) antenna [41–43].

8.3.2 Slot Layout for Radial Line Slot Array Bessel

Launchers

A CP-RLSA is usually composed of several hundreds of circular-polarized radi-

ating elements, each one comprising two slots kd =4 apart, making an angle of 45°
with respect to the radial direction (Fig. 8.6). Since the two slots in the pair are
orthogonal to each other and fed in phase quadrature by the feeding wave in the
PPW, each element radiates a CP wave in the broadside direction. The arrangement
of these CP elements on the radiating aperture determines the antenna radiation
pattern.
Since the elements of the array are closely spaced, their radiation pattern is
strongly related to that of the associated continuous aperture current distribution, of
which the array constitutes a spatial sampling. Such an aperture distribution is
chosen specifically to achieve the desired radiation pattern by following one of the
techniques present in the literature. Among them, here the alternate projection
method [44, 45] has been adopted. As it is well known, the radiation pattern of the
slot array will sufficiently approximate the continuous aperture reference pattern if
the following two conditions are verified: the array element excitation corresponds
to the continuous aperture sampling, and, according to the Schelkunoff sampling
criterion, the inter-element distance is sufficiently tightly linked. It is worth pointing
out that an array pattern synthesis technique assuming given element positions
cannot be adopted here. Indeed, it is necessary to resort to a continuous aperture
distribution because the specific slot positions are used to calibrate the radiating
element phase and cannot be given a priori.
According to Bethe theory [46], the radiation from a slot cut into a metallic plane
can be modeled by an equivalent magnetic dipole sitting on a ground plane, with
dipole moment proportional to the magnetic field of the feeding wave in the PPW

Fig. 8.6 Geometry of a slot pair element radiating a RHCP when illuminated by an outward
traveling wave. The distance between the slot centers is set to kd =4, where kd is the wavelength in
the PPW. The LHCP is achieved by mirroring the slots with respect to q ^
258 M. Ettorre et al.

M ¼ am
H inc ; ð8:35Þ

through a dyadic polarizability am . Although Bethe theory also introduces an

equivalent electric dipole, its radiation effect is negligible with respect to that of the
magnetic dipole for slots with a length much greater than their width. Since here we
are dealing with such kind of elongated slots the effect of the equivalent electric
dipoles is not considered.
For a single slot oriented along u ^, the dyadic polarizability is given by
am ¼ am u ^u^, where the amplitude of am grows by increasing the slot length up to the
resonance length. Therefore, the slot length can be adjusted in order to match the
amplitude of the target aperture distribution, while the phase of the equivalent
magnetic dipole of each slot can be controlled by adjusting its position along the
feeding wave propagation direction, thus varying the phase of the incident wave
excitation Hinc .
Since the basic element of the CP-RLSA is a slot pair, it is worthwhile intro-
ducing an analogous description for the whole slot pair. Thus, by summing the
magnetic dipoles relevant to the two slots in the pair, and normalizing them to the
incident magnetic field at the slot pair center, a polarizability apm for the whole slot
pair is obtained. Note that apm depends on the phase progression of the incident
feeding wave in the PPW, which establishes the phase shift between each slot and
the pair center reference point.
In the following, since the Bessel aperture distribution AðqÞ is rotationally
symmetric, we consider only the class of CP-RLSAs with a rotationally symmetric
radiation pattern. Also, without loss of generality, we focus on a right-hand CP
(RHCP) antenna. The left-hand CP (LHCP) can be achieved similarly. Under the
above hypothesis, the ideal target aperture magnetic current will be in the form

pffiffiffi
M0 ðqi Þ ¼ AðqÞ^ ^RH;LH ¼ ð^x
j^yÞ 2 denotes the polarization unit
pRH;LH , where p
vector and the upper/lower sign applies to the RHCP/LHCP.
To match the target aperture distribution, one has to assure that, for each slot pair
element, the equivalent magnetic dipole Mðqi Þ ¼ apm ðqi Þ
Hinc ðqi Þ is parallel to
M0 ðqi Þ, i.e., Mðqi Þ ¼ Cejc M0 ðqi Þ, where qi is the position of the i-th slot pair
element center, and c; C are arbitrary real constants. Such condition is satisfied only
if, for each slot pair, it holds
 p 
a ðqi Þ
Hinc ðqi Þ ¼ CjAðqi Þj; ð8:36Þ
m

and
 
\ M0 ðqi Þ
apm ðqi Þ
Hinc ðqi Þ ¼ c: ð8:37Þ

For a given c only specific positions of the slot pair elements verify (8.37).
Hence, (8.37) is the key equation controlling the correct slot pair positioning which
ensures the target aperture phase distribution. On the other hand, (8.36) imposes the
8 Near-Field Focusing by Non-diffracting Bessel Beams 259

correct amplitude distribution on the radiating aperture which can be achieved by

properly choosing the slot length. In fact, a variation in slot length mainly affects
the associated magnetic moment magnitude, while the phase change is a higher
order effect; similarly, a variation in slot position mainly affects the associated
magnetic moment phase while the magnitude remains almost the same. Such a
quasi-orthogonality property permits independent actions on the two sets of vari-
ables (i.e., the slot pair lengths and positions) to determine the holographic slot
layout.
For example, for the slot pair arrangement in Fig. 8.6, excited by a radial out-
ward feeding wave, by adopting the simple Bethe model and neglecting the slot
coupling, the slot pair polarizability can be estimated as follows:
h i h i
apm ¼ am ejp=4 u ^1 þ ejp=4 u
^1 u ^2 ¼ am ej/ p
^2 u ^LH ^ ^RH ^
p þ jej/ p / : ð8:38Þ

z
By considering a PPW excited by a TM0;0 cylindrical mode, (i.e., Hinc ðqÞ ¼
ð2Þ ð2Þ
H inc H1 ðkd qÞ^
/, where H1 denotes the first-order Hankel function of second kind,
and kd ¼ 2p=kd the wavenumber inside the PPW, (refer to Sect. 9.4.1 in [47] for
modes definition), Eq. (8.36) reduces to

C jAðqÞj
jam ðqÞj ¼  ;
 ð8:39Þ
ð2Þ
H1 ðkd qÞ

which is exploited so as to calculate the slot length profile. Indeed, a parametric

analysis of an isolated slot pair provides am as function of the slot length ‘ [42],
from which a slot length profile ‘ ¼ ‘ðqÞ satisfying (8.39) can be obtained. Note
that the constant C does not affect the aperture amplitude shape, but has to be
chosen to keep the slot length within a proper range, and to calibrate the total
aperture radiating power. Namely, if the value of the C constant is too large, the
power of the feeding wave is entirely radiated by the slots close to the antenna
center, and the slots at the antenna rim are fed too weakly to provide the desired
amplitude, even assuming the maximum resonant length. On the other hand, when
C is too small, the desired aperture shape is easily achieved but a considerable
amount of power is not radiated and remains trapped in the PPW beyond the last
peripheral slots. Although the residual power can be absorbed with an absorbing
rim [48–50], this results in a poor antenna radiation efficiency. Summarizing, C
must be optimized as a tradeoff between fine aperture amplitude shape recon-
struction and high radiation efficiency.
Next, from (8.37), the centers of the slot pairs have to be located on the curve
n o
ð2Þ
\ A ðqÞam ðqÞej/ H1 ðkd qÞ ¼ c: ð8:40Þ
260 M. Ettorre et al.

Fig. 8.7 Slot pair positioning. Slot pair centers qi ¼ ðqi ; /i Þ (turquoise dots) are initially chosen
by sampling the curve q ¼ qð/Þ defined by (8.40), which ensures that the slot pair elements match
the target aperture phase distribution, according to the simplified model based on Bethe theory.
Next, in the optimization process the slot positioning curve q ¼ qð/Þ is represented by a thinned
set of control points Pi ¼ ðPi ; Ui Þ (black dots) [40]

It is worth noting that for a uniform distribution (i.e., AðqÞ ¼ const:Þ, by using
ð2Þ
the asymptotic large argument expansion \H1 ðkd qÞ   kd q þ 3p=4 and
neglecting the phase variation for the polarizability of slots having different lengths,
Eq. (8.40) results in the classic CP-RLSA spiral slot arrangement /  kd q ¼
const[42].
Likewise, in any general CP-RLSA design, Eq. (8.36) allows the calculation of
the slot length profile ‘ðqÞ, and then Eq. (8.37) provides the definition of a curve
where the slot pair centers are to be distributed. Such a curve will be in general
of the form q ¼ qð/Þ (Fig. 8.7). The slot pair centers are then distributed
uniformly along such a curve by choosing the sampling points /i via
/i ¼ /i1 þ Ds=qð/i1 Þ. The sampling step Ds along the curve is arbitrary, and its
value is a design parameter which is constrained between a minimum value, to
avoid slot overlapping, and a maximum one, according to the Schelkunoff sampling
criterion. Given the sampling points /i , slot pair centers will be located at
qi ¼ ðqi ; /i Þ, with qi ¼ qð/i Þ, and the length of the slots in the pair will be
‘i ¼ ‘ðqi Þ. Note that the total number of slot pairs is, in our approach, not a priori
determined. In fact, we keep adding slots until they remain within the assigned
antenna size.
The above-mentioned algorithm completely neglects mutual coupling among
slot pairs, which can be accounted for only with a full-wave analysis. However, it is
sufficient for initializing the optimization process described in the next section.
8 Near-Field Focusing by Non-diffracting Bessel Beams 261

8.3.3 Optimization Procedure

When a not uniform field aperture distribution, such as a Bessel one, is employed or
the geometrical requirements on the antenna radius determine a number of slot pair
elements in the order of several thousands, the use of an optimization algorithm to
synthesize the required distribution is necessary. A fine optimization cannot be
performed by using simplified models as in [51, 52], since they do not describe
accurately the global slot coupling. An accurate but efficient full-wave numerical
method is then mandatory for the optimization of the antenna. A successful example
is the technique described in [53] which reduces the return loss in a linear polarized
RLSA configuration, even if the antenna is recursively optimized ring by ring and
not globally. Moreover, several papers point out that, when a careful antenna design
is applied and the efficiency of the antenna is maximized, the amplitude of higher
modes excited in the PPW by the slots distribution itself is no longer negligible.
Indeed, it results in an appreciable degradation of the radiation pattern and/or of the
reflection coefficient of the antenna. For example, in [54] the authors have recently
proved that the distortion of the rotational symmetry of the radiation pattern and the
unbalanced sidelobe levels typical of a spiral RLSA are related to these higher order
z
modes, and in particular to the spurious TM0;2 mode of the PPW.
The optimization procedure here used is tailored on the particular physical
mechanism that governs the RLSA’s operation, as explained in [40]. Specifically, it
acts on the position and the length of each slot so as to obtain a matching between
the actual antenna aperture distribution, calculated by the full-wave analysis, and its
target profile. However, the optimization cannot act individually on each slot;
otherwise the enormity of the problem would result in an instable procedure that
does not converge at all. Instead, the antenna parameterization, here proposed,
involves some global parameters the tuning of which is performed resorting to
specific fitness functions. Such fitness functions are introduced on the basis of the
physical picture behind the working mechanism of the antenna, and are chosen in
such a way that each fitness function provides the optimization feedback for only
one geometrical parameter. Therefore, the optimization scheme of the large mul-
tivariable problem is reduced to the parallel arrangement of some single-parameter
problems. Such size reduction and quasi-orthogonality property ensure a very fast
convergence.
For a given antenna geometry, by using a full-wave method, the dipole moment
of each slot is easily evaluated. Specifically, M1 ðqi Þ and M2 ðqi Þ denote the mag-
netic dipole moments of the inner and outer slots in the pair, respectively, while
Mðqi Þ ¼ M1 ðqi Þ þ M2 ðqi Þ is the dipole moment of the i-th slots pair of center qi .
For each slot pair a complex fitness functions Fi is then introduced as follows:


^RH
Mðqi Þ A
p
Fi ¼ ; ð8:41Þ
Aðqi Þ M
262 M. Ettorre et al.

P
Np
P
Np

¼
where A  ¼
jAðqi Þj Np and M ^RH
Mðqi Þj Np are the target and the
jp
i¼1 i¼1
realized average co-polar dipole moment amplitude, respectively, with Np denoting
the total number of slot pairs.
Note that \Fi and jFi j indicate the phase and amplitude errors between the target
and the actual dipole moment distribution, which can be compensated by adjusting
the slot pair radial position qi and the slot length ‘i , respectively.
In principle, to minimize the overall error we could act on each slot pair inde-
pendently. However, the strong mutual coupling between the slots renders this way
of operating instable, and after a few steps the slots start breaking the order of the
initial layout and overlapping.
To overcome this drawback a set of N control points P
n ¼ ðPn ; Un Þ is intro-
duced. Since N is much lower than the total slot pair number N Np , the control
point set represents a coarse sampling of the slot pair distribution curve q ¼ qð/Þ
(Fig. 8.7). Analogously, the slot length is described by the thinned set of N length
samples Ln . For a given set of angular control point locations Un , the antenna
geometry is generated from the reduced sets Pn and Ln by calculating slot pair
positions and lengths through a linear interpolation as

X
N
qð/Þ ¼ Pn bn ð/Þ ð8:42Þ
n¼1

X
N
lð/Þ ¼ lðqð/ÞÞ ¼ Ln bn ð/Þ; ð8:43Þ
n¼1

where
8 /Un1
< Un Un1 Un1 \/  Un ;
bn ð/Þ ¼ Un þ 1 /
Un  /\Un þ 1 ; ð8:44Þ
: Un þ 1 Un
0 elsewhere:

Equations (8.42) and (8.43) represent a parameterized version of the position

and length curves and allow the automatic generation of the antenna layout. In
Fig. 8.7 and in the examples provided in Sect. 8.3.4, we have chosen Un ¼ 2np
(i.e., a control point each spiral turn), though different choices are possible.
The initial set of parameters is chosen by sampling the initial design
curves obtained with the simplified approach illustrated in Sect. 8.3.2 as
Pn ¼ qðUn Þ; Ln ¼ ‘ðUn Þ. The antenna layout can next be adjusted by varying both
the values of the control points Pn (thus smoothly moving the slot pairs), and the
values of the samples Ln (thus smoothly controlling the slots length). This smooth
deformation of the layout geometry avoids a significant variation of the distance
between adjacent slots, thus minimizing the risk of overlapping, unlike the scheme
8 Near-Field Focusing by Non-diffracting Bessel Beams 263

where each slot moves independently of its neighbor. The updating law is described
in the following.
Optimization Loop
At each step s of the optimization process, when the antenna layout is parameterized
by the values of the parameters PðnsÞ ; LðnsÞ , a full-wave analysis of the whole antenna
is performed and the equivalent magnetic dipole moments Mðqi Þ are calculated. In
addition, the antenna spillover efficiency

Pacc  Pso
gðsÞ
so ¼ ð8:45Þ
Pacc

is also calculated, where Pacc is the power accepted by the antenna at the feeding
port, whereas Pso is the total power trapped in the PPW beyond the slots distri-
bution. It is worth noting that when losses in the PPW are negligible gso equals the
radiation efficiency. Then, a linearly weighted version of the complex fitness
function (8.41) is defined at each step s as

X
Np
ðsÞ
nðsÞ ¼
F Fi bn ð/i Þ ð8:46Þ
i¼1

which expresses the local average error on the slot pairs in the neighborhood of the
n-th control point. In fact, each n-th value of the fitness function is associated with
the relevant n-th control point, but it averages the local error at each i-th slot pairs in
the sub-spiral around the control point. The minimization at each control point of
the magnitude and the phase of the fitness function leads to the optimum slot pair
length and position, respectively.
Then, the control parameters PðnsÞ ; LðnsÞ are updated according to

Pðs
n
þ 1Þ
¼ PðsÞ  ðsÞ
n  vq \Fn =kd ð8:47Þ

and
 qffiffiffiffiffiffiffi
 ðsÞ 
Lðs þ 1Þ  
¼ 1 þ v‘ 1  Fn  ðsÞ
gso LðsÞ ð8:48Þ
n n

with 0\vq ; v‘  1 suitable damping factors (typically vq ’ 0:5 and v‘ ’ 0:5Þ

which prevent the optimization loop from oscillating. The updating rule (8.47)
corrects the slot position minimizing the local average phase error; indeed slots are
moved inward/outward to compensate for a local aperture illumination phase
delay/advance. The updating rule (8.48) corrects the slot length minimizing the
local amplitude error; indeed, slots are shortened/lengthened to compensate for
excessive/scarce local aperture illumination amplitude. Also, the global factor
264 M. Ettorre et al.

qffiffiffiffiffiffiffi
ðsÞ
gso has been introduced in (8.48) so that the slot length profile of the entire
antenna is adjusted to maximize the radiation efficiency. Indeed, all the slots are
globally shortened/lengthened, without changing the aperture distribution shape, to
guarantee that most of the power injected in the PPW at the feeding point is radiated
by the slots before reaching the antenna rim.
The loop is repeated until the spillover efficiency gso is over a fixed threshold
(usually gso [ 95%), and the average aperture distribution error

Np 
 Np  
1 X  1 X
qRH
Mðqi Þ Aðqi Þ
^ Aðqi Þ
 
eD ¼    
  Np¼  A jFi  1j ð8:49Þ
Np i¼1  M A i¼1

is sufficiently small (e.g., eD \0:01Þ.

Minimization of Spurious Modes Excited by the Slots Arrangement

Another effect, which is neglected in the simplified model used for the first design
of the CP-RLSA, is the perturbation of the incident feeding wave due to the field
scattered in the PPW by the slots etched in the upper plate. Despite the field
scattered inside the PPW by a single-slot pair is relatively weak, the collective
radiation of all the slot pairs generates a higher order of PPW modes, some of which
can be of appreciable amplitude with respect to the feeding. Since the PPW height h
is chosen less than half wavelength ðh\kd =2Þ, only TMn;m z
modes with n ¼ 0 can
propagate (where n and m tag the radially propagating mode eigenvalues along z
z
and /, respectively). When the antenna is fed by an outgoing TM0;0 mode, excited
higher order ðm 6¼ 0Þ modes vanish at the feeding point q ¼ 0, and therefore their
occurrence does not result in an increased antenna reflection loss. Nevertheless,
since they exhibit a phase variation expðjm/Þ, their presence in the PPW creates an
interference with the TM0;0 z
feeding mode resulting in a not uniform total field
impinging on the slots. As a matter of fact, the feeding wave loses the rotational
symmetry and exhibits an amplitude ripple and a phase oscillation along /. In turn,
the realized antenna aperture distribution loses the rotational symmetry with a
degradation of the radiation pattern. Therefore, in the case of Bessel aperture dis-
tribution, this effect results in a circular asymmetry of the pseudo-Bessel beam
around the propagation axis which needs to be kept under control during the design.
To minimize a given undesired TM0;m z
mode one can adjust the distance Dqi
between the two slots in a pair, and their length difference D‘i . In particular, the
inner (tagged by
q ¼ 1) and the outer (tagged by q ¼ 2) slots in the i-th pair will be
placed at qiq ¼ qiq ; /iq , with qiq ¼ qi þ ð1Þq Dqi =2, and their respective length
will be ‘iq ¼ ‘i þ ð1Þq D‘i =2. Note that the azimuthal coordinate of the inner/outer
slot /iq might be displaced with respect to the slot pair center /i to avoid slot
overlapping [41]. To optimize the design parameters Dqi and D‘i , for each slot pair
a complex fitness function
8 Near-Field Focusing by Non-diffracting Bessel Beams 265

ð2Þ0
Hm ðkd qi1 Þejm /i1 /
^
M1 ðqi Þ
m
Fi ¼  ð2Þ0
i1
ð8:50Þ
Hm ðkd qi2 Þejm/i2 /
^
M2 ðqi Þ
i2

z
is introduced, where m tags the undesired mode TM0;m to which the fitness function
ð2Þ0
is related, and Hm denotes the derivative of the m-th order Hankel function of
^ is the azimuthal unit vector at inner/outer slot center qiq .
second kind. In (8.50) /
iq
The fitness function in (8.50) is the negative of the ratio between the amount of
z
TM0;m individually excited by each slot in the pair. When m Fi ¼ 1 the contributions
z
of the two slots to the TM0;m mode perfectly cancel out each other. Hence, the
amplitude j Fi j?1 and the phase \m Fi ?0 can be used to feedback D‘i and Dqi ,
m

respectively. Namely, jm Fi j can be reduced/increased by increasing/reducing D‘i , so

that the scattering contribution of the inner/outer slot is weakened while that of the
outer/inner slot is reinforced. Also, \m Fi can be increased/reduced by
reducing/increasing Dqi , so that the phase delay between the scattering contribution
of the two slots is reduced/increased.
Again, to introduce smooth variation and avoid optimization instability, we
resort to a linear variation of the design parameters

X
N
Dqð/Þ ¼ DPn bn ð/Þ ð8:51Þ
n¼1

X
N
D‘ð/Þ ¼ DLn bn ð/Þ ð8:52Þ
n¼1

which are calculated in terms of a thinned set of samples DPn and DLn defined at the
control points Un . Such parameters are initialized with their nominal values DPn ¼
k=4; DLn ¼ 0 according to the ideal simplified model introduced in Sect. 8.3.2.
Next, they are updated in the optimization loop by resorting to a linearly weighted
version of the complex fitness function (8.50) associated to each control point n

X
Np
ðsÞ
m  ðsÞ
Fn ¼ m
Fi bn ð/i Þ: ð8:53Þ
i¼1

In particular, the updating rules are defined as follows:

m  ðsÞ
DPðs
n
þ 1Þ
¼ DPðsÞ
n  vDP \ Fn =kd ð8:54Þ

and
266 M. Ettorre et al.

h m ðsÞ i
DLðs þ 1Þ
¼ DL ðsÞ
þ v DL 1   F
n  D ð8:55Þ
n n

with 0\vDP ; vDL  1 suitable damping factors (usually vDP ’ 0:1 and vDL ’ 0:1Þ
ðsÞ ðsÞ
and D ¼ max Ln =20; DLn . The ratio between the amplitude of the undesired
z
TM0;m mode in the PPW and the amplitude of the feeding mode can be monitored at
each step s and introduced into the stop criterion of the optimization procedure.
Therefore, also this second fitness function is calculated at each control point and
the minimization of its magnitude and phase, respectively, leads to the optimum slot
length difference inside the pair, and the optimum displacement Dq of each slot in
the pair with respect to the pertinent pair center qi .
Note that the parameters pair Dqi and D‘i might alternatively be adjusted in
order to minimize the cross-polarized radiation of the slot pair. However, the
CP-RLSA spiral arrangement presents a broadside null of the cross-polarized
radiation pattern in the array factor. Therefore, the polarization level of the antenna
combining a CP element and a CP array factor is usually excellent and does not
require a cross-polarization level improvement.

8.3.4 Circular-Polarized Bessel Beam Launcher Made

by a RLSA Structure

The optimization procedure presented in Sect. 8.3.3 converges to the optimum

solution in a very limited number of steps, allowing the use of a full-wave analysis
of the entire antenna [40] that, in turn, guarantees an accurate prediction of the
antenna behavior. In the full-wave analysis, at least five rectangular waveguide type
modes are required to accurately model the shape of the magnetic current along
each slot, because of the strong coupling between the slots [53]. In the present
design, unlike [53], the slots couple significantly both in the internal (PPW) and in
the external (grounded half-space) regions, because the dielectric constant is almost
the same in the two regions.
Consider now the design of a CP-RLSA antenna working at the nominal fre-
quency of 15 GHz for the generation of a Pseudo-Bessel beam characterized by a
half-power beamwidth of less than 12 cm and to need a focalization distance in
between 4 and 8 m.
Hence, the antenna has to generate, on the plane z ¼ 0, an aperture field dis-
tribution AðqÞ ¼ A0 J0 ðkB q0Þ, as specified in Eq. (8.34), where the constant kB is
related to the half-power beamwidth (HPBW) through kB ¼ 2:235=HPBW [3].
Thus, the HPBW requirement is satisfied by choosing kB ¼ 20 m1 . Moreover,
assuming reasonable a depth of field zmax ¼ 9 m, through the relation
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
zmax ¼ R= ðk0 =kB Þ2 1, it is obtained an antenna radius of 57 cm. In the fol-
lowing, by presuming a slight degradation of the actual antenna performances,
8 Near-Field Focusing by Non-diffracting Bessel Beams 267

Fig. 8.8 Evolution of the equivalent dipole moment of each slot pair along the optimization steps:
magnitude (above), and phase (below). Initial design (dotted green line), final design (red line),
and target aperture distribution M0 ðqi Þ (dashed-black line)

an antenna radius of 60 cm will be considered. Due to the large dimension of the

antenna, a hollow PPW cannot be employed. Thus, to guarantee mechanical
robustness and to maintain its thickness constant, the PPW is supposed filled by a
foam substrate (ROHACELL WF51, er ¼ 1:07 and tan d ¼ 0:0035 @10 GHz).
The optimization process converges in 16 steps. The comparison between the
target aperture distribution and the synthesized one is shown in Fig. 8.8, where one
can trace how the equivalent dipole moment of each slot pair Mðqi Þ ¼ Aðqi Þb p
evolves during the optimization process from the first draft design (green dotted
line) and it converges to the target (black dashed line) M0 ðqi Þ, at the last step (red
solid line). It is also noticeable a rippled behavior of the synthetized dipole moment
due to the excitation of spurious higher circumferential modes in the PPW, and
z
especially modes TM0;2 . If no action had been taken at each optimization step
when increasing the efficiency of the antenna (i.e., the length of the slots and
consequently its interaction), the ripple would have become unacceptable.
However, this phenomenon was contrasted by adjusting the relative position Dq of
each slot pair that constitutes the elementary circular-polarized element of the array
through Eq. (8.53). Introducing this degree of freedom in the optimization process,
z
we limit the amplitude of the undesired TM0;2 modes to less than 15% with respect
z
to the amplitude of the feeding TM0;0 mode. Thus, the effect of the ripples is
mitigated.
The discretization of the aperture distribution by slots necessarily introduces a
degradation of the Bessel beam performances as the reduction of the maximum
268 M. Ettorre et al.

(a) (b)

Fig. 8.9 a the axial z-component of the radiated power density sz along the z-axis for the target
continuous distribution (black red line), for the target distribution sampled on the slot pair
positions (gray line), for the synthesized RLSA distribution (black solid line), and for the
measured antenna prototype (black dashed line). b A photograph of the RLSA prototype

radiated power density, but not of the beam shape. In Fig. 8.9 the axial z-compo-
nent of the radiated power density associated to the ideal continuous distribution
(red solid line) is compared to that associated to the ideal distribution (gray line)
sampled on the slot pair positions of the synthesized RLSA (the power accepted by
the antenna is set equal to 1 W). The small reduction of the maximum radiated
power density associated to the synthesized antenna (solid black line), with respect
to the one associated to the sampled ideal distribution, demonstrates the effec-
tiveness of the optimization process. The RLSA is expected to have the maximum
power density at a distance from the radiating surface of 5.1 m instead of 5.2 m of
the continuous aperture, with a −6 dB depth of field of 430k0 instead of 455k0 .
By defining the focusing efficiency as the ratio between the power focalized in
the target spatial xy-spot of 12 cm diameter at the maximum radiated power dis-
tance for the actual antenna and the ideal (continuous) pseudo-Bessel beam dis-
tribution, the expected efficiency for the synthesized RLSA is about 84%.
The point of maximum power density was expected at z = 5.1 m (continuous
red line in Fig. 8.9a). However, it has been measured at z = 5.0 m. The shift of the
maximum power density distance toward the antenna is more evident in Fig. 8.9
where the dashed-black curve represents the measured power density along z. This
deviation is compatible with the manufacturing error associated with the laser
cutting of the slots in the metallic plate of the antenna. However, it can be con-
sidered a very good result for that electrically large antenna (60 k0 at 15 GHz) with
almost ten thousand slots (i.e., 9154), as shown in Fig. 8.9b.
The measured radiated power density (normalized to its relative maximum
value) in the xz-plane is shown in Fig. 8.10a. The −3 dB beam remains confined in
the specified spatial region of diameter 12 cm for more than 2 m in the entire band
of frequency. In Fig. 8.10b, the simulated and measured normalized power densities
8 Near-Field Focusing by Non-diffracting Bessel Beams 269

Fig. 8.10 Power density at 15 GHz, normalized to its maximum value, in the xz-plane (a) and in
the xy-plane for a distance from the radiating surface of 5.0 m (b)

Fig. 8.11 Geometry of the problem: an antenna that produces a Bessel beam is used to detect a
buried metallic plate. Specifically, we are considering a noncontact detection of buried objects at
15 GHz when a circularly polarized antenna is on-board of a vehicle. Practical observation
suggests focalizing with a half-power beamwidth (HPBW) of a few centimeters (<12 cm) and with
a not well-defined focalization distance in between 4 and 8 m (i.e., 200  400k0 Þ

are compared at 5 m distance from the radiating surface. The agreement between
the simulated and measured beam shapes is very good.
The antenna here described has been used for the specific application for non-
contact microwave detection of buried mines, as sketched in Fig. 8.11.
In conventional ground penetrating radar, the detection cannot be performed in
remote conditions, because of the very large clutter contribution that degrades radar
270 M. Ettorre et al.

Fig. 8.12 Echo map relevant to a 10 cm  10 cm metal plate buried 15 cm deep in Sahara dry
sand when a classical pencil beam (a) or a Bessel beam (b) are used

Fig. 8.13 Radial waveguide

loaded with metallic gratings
and fed by a coaxial fed

performance. A sensing architecture based on the concept of localized waves has

been proposed in [34] to dramatically increase the signal-to-noise ratio. The
advantage of using a Bessel beam instead of a classical pencil beam in such kind of
ground penetrating radar is evident in the comparison of the relevant echo maps
shown in Fig. 8.12.

8.3.5 Bessel Launcher Made by a Radial Waveguide Loaded

with Metallic Gratings

The practical implementation of a Bessel beam launcher exploiting an

inward-traveling wave aperture distribution is shown in Fig. 8.13. The structure is
made by a parallel-plate waveguide (PPW) with annular slots etched on the top
plate. A coaxial feed at the center excites an outward cylindrical radial wave inside
the PPW. An in-house Method of Moments (MoM) [20] is used to take into account
8 Near-Field Focusing by Non-diffracting Bessel Beams 271

Table 8.1 Geometrical sizes Slot qi (mm) wi (mm) Slot qi (mm) wi (mm)
of the Bessel beam launcher:
radial positions ðqi Þ and 1 1.5 0.55 7 110 2.3
widths ðwi Þ of the annular 2 29.8 0.94 8 126.5 2.6
slots 3 45.7 1.1 9 143 3
4 61.5 1.5 10 159.5 3.5
5 76.4 1.2 11 175.4 3.8
6 94.1 3 12 192.4 4

Fig. 8.14 Electric field radiated by  the launcher in Fig. 8.13, calculated using COMSOL
Multiphysics. From left to right, jEz j; Eq  and jEj. The axes are normalized with respect to the
wavelength ðkÞ at the operating frequency. The dashed-white lines mark the GO boundaries

the mutual coupling amongst the various slots. The design follows the same steps
outlined in Sect. 8.3.3 to tailor the magnetic current distribution over the top plate
in the shape of an inward cylindrical wave. This is accomplished by controlling the
positions along the radial direction qi and sizes wi of the circular slots (refer to
Fig. 8.13).
Indeed, slots are positioned at those points where the phase of the outward
cylindrical feeding wave inside the PPW matches that of the target inward cylin-
drical aperture distribution, according to a holographic criterion.
The slot width is used to modulate the amplitude of the aperture distribution. In
addition the design procedure assures that the total energy launched by the coaxial
feed in the PPW is radiated, thus avoiding any spurious radiation by the edges of the
structure. The geometrical details of the structure are provided in Table 8.1.
COMSOL multiphysics has been used to analyze the structure. Field maps of the
electric field above the launcher are shown in Fig. 8.14, which are found very similar
to their theoretical counterparts in Sect. 8.2.1. To allow a clearer quantitative com-
parison, we also show in Fig. 8.15 the Ez component of the electric field at the distance
z ¼ 4:667k from the aperture, corresponding to the maximum transverse extension
of the rhomboidal region where the Bessel beam is created (refer to Fig. 8.14),
272 M. Ettorre et al.

Fig. 8.15 Normalized jEz j component of the electric field at z ¼ 4:667k for the structure in
Fig. 8.13 simulated with COMSOL Multiphysics. The field radiated by a truncated inward Hankel
distribution is shown for comparison

compared to the ideal truncated ða ¼ 7kÞ inward Hankel aperture distribution. The
good agreement reveals the accuracy of the launcher design. It is worth noting that the
traveling nature of the synthesized aperture distribution guarantees a wideband
operation as a difference with resonant Bessel designs [9, 40, 55].

8.4 Leaky-Wave Bessel Beam Launchers

8.4.1 The Leaky Radial Waveguide: Motivations

In this section, we consider a propagating transverse magnetic (TM) Bessel beam

[56]. The z-directed (in the direction of the structure’s
surface
 jknormal) electric field
component takes on the following form: Ez ðq; zÞ ¼ J0 kq q e zz
. Since the wave is
TM-polarized, the other components of the electromagnetic field can be expressed
as [57]

E/ ¼ Hz ¼ Hq ¼ 0; ð8:56Þ

kz
 jkz z
Eq ¼ j J1 kq q e ; ð8:57Þ
kq

k0

Hq ¼ j J1 kq q ejkz z ; ð8:58Þ
g0 k q

where g0 is the characteristic impedance of free space and the wavenumbers obey
the separation relation: k02 ¼ kq2 þ kz2 .
8 Near-Field Focusing by Non-diffracting Bessel Beams 273

Fig. 8.16 A leaky radial

waveguide loaded with a
capacitive sheet impedance

Equations (8.56)–(8.58) show that the components of the electric and magnetic
field are Bessel functions, and do not change their profile with respect to z. Since
the Bessel beam is propagating, the field components possess a transverse propa-
gation constant kq smaller than k0 ; the free-space wavenumber. As a result, an open
structure exciting a propagating Bessel beam supports an azimuthally invariant
fast/leaky mode in the transverse direction. Since the structure supports a leaky
mode, the normal wavenumber is complex: kz ¼ bz þ jaz and the structure’s surface
impedance becomes

Eq g

gsurf ¼  ¼  0 bz þ jaz : ð8:59Þ
H/ k0

From (8.59), it is clear that a structure with a capacitive surface impedance is

required to support a TM-polarized, cylindrical leaky mode [58]. A radiating sys-
tem that supports such a leaky mode is the radial waveguide loaded with a
capacitive impedance sheet, depicted in Fig. 8.16. This capacitively loaded radial
waveguide will be studied in more detail in the next sections.
It is worth mentioning that the argument above for TM modes can be also
extended to transverse electric (TE) modes. In the TE case, the z-component of the
magnetic field presents a Bessel profile, and an inductive impedance sheet is
required to generate an azimuthally invariant leaky mode inside the radial
waveguide.

8.4.2 Field Derivation for the Leaky Radial Waveguide

The considered radial waveguide is shown in Fig. 8.16. It consists of a capacitive

impedance sheet at a height h above a metallic ground plane. A /-invariant TM
mode with respect to the z-direction is considered. A magnetic vector potential with
only a z-component (A) is used to derive the field radiated and within the radial
waveguide. In particular, for a radial waveguide filled with a homogenous media
with permittivity er1, the vector potential can be expressed in cylindrical coordinates
as [57]
274 M. Ettorre et al.

h i
ð1Þ
 ð 2Þ

A ¼ cosðkz1 zÞ C1 H0 kq q þ D1 H0 kq q bz ; ð8:60Þ

ð1Þ
where kz1 and kq are the normal and transverse propagation constants, and H0 ð:Þ
ð 2Þ
and H0 ð:Þ are the zero-order Hankel functions of the first and second kind,
respectively. The variables C1 and D1 denote amplitude coefficients stipulated by
the boundary conditions and excitation.
The components of the electric and magnetic field can now be derived as

E/ ¼ Hz ¼ Hq ¼ 0; ð8:61Þ

jg1 @A
Eq ¼ ; ð8:62Þ
k1 @q@z

jg1 @ 2
Ez ¼ þ k1 A;
2
ð8:63Þ
k1 @z2

@A
H/ ¼  ; ð8:64Þ
@q

where A, g1 , and k1 are the z-components of the magnetic potential, characteristic

impedance, and propagation constant within the radial waveguide, respectively. The
fields within the radial waveguide can therefore be expressed as
h i
jg1 kz1 kq ð1Þ
 ð2Þ

Eq ¼ sinðkz1 zÞ C1 H1 kq q þ D1 H1 kq q ; ð8:65Þ
k1
h i
jg1 kq2 ð1Þ
 ð2Þ

Ez ¼ cosðkz1 zÞ C1 H0 kq q þ D1 H0 kq q ; ð8:66Þ
k1
h i
ð1Þ
 ð2Þ

H/ ¼ kq cosðkz1 zÞ C1 H1 kq q þ D1 H1 kq q : ð8:67Þ

Equations (8.65)–(8.67) show that the field components

consist of two cylin-
ð2Þ ð2Þ
drical waves: one traveling outward H0 ð:Þ or H1 ð:Þ , and the other inward

ð1Þ ð1Þ
H0 ð:Þ or H1 ð:Þ . The field component Ez assumes a J0 profile when C1 ¼ D1 :
Such field profiles can be achieved by placing a circular metallic rim at a distance
q ¼ q1 , which corresponds to a null of the J0 function. Therefore, the cancelation of
the z-directed electric field within the radial waveguide leads to

ð2Þ

H0 kq q1
C1 ¼ D1 ð1Þ
: ð8:68Þ
H0 kq q1
8 Near-Field Focusing by Non-diffracting Bessel Beams 275

Fig. 8.17 The leaky radial

waveguide with a metallic rim
at q ¼ q1 : An array of patches
represents the impedance
surface Zs. The launcher is fed
by a coaxial probe at the
center

The corresponding structure is shown in Fig. 8.17.


Assuming a complex wavenumber along q kq ¼ bq  jaq and large rim
ð1Þ;ð2Þ

radius q1 , the Hankel functions can be approximated as H0 kq q 
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi jðk qp=4Þ
2=pk qe q
q . As a result, two conditions can be derived from (8.68) to
ensure C1 ¼ D1 :
p
bq q1 ¼  þ qp; ð8:69Þ
4
aq q1 1; ð8:70Þ

where q is an integer. Equation (8.69) requires that the inward and outward modes
are out of phase at q ¼ q1 , one of the zeros of the zeroth-order Bessel function.
Equation (8.70) states that the amplitudes of the backward and forward leaky-wave
modes are nearly identical. Note that in practical leaky-wave antennas, the
leaky-wave mode is almost zero at the end of the structure, in order to avoid an
additional sidelobe due to backward traveling leaky-wave modes. Here, the struc-
ture is resonant in the q−direction and the backward traveling mode is critical to
establishing the Bessel beam.
Once Eqs. (8.69) and (8.70) are satisfied, the field components (8.65)–(8.67) can
be rewritten as

jg1 kz1 kq

Eq ¼ C1 sinðkz1 zÞJ1 kq q ; ð8:71Þ
k1

jg1 kq2

Ez ¼ C1 cosðkz1 zÞJ0 kq q ; ð8:72Þ
k1


H/ ¼ C1 kq cosðkz1 zÞJ1 kq q : ð8:73Þ
276 M. Ettorre et al.

Fig. 8.18 Non-diffractive range of the leaky radial waveguide

As expected, the z-component of the electric field assumes a J0 profile.

The radius q1 of the Bessel beam launcher (leaky radial waveguide) is also
directly related to the non-diffractive range of the truncated Bessel beam [3] (see
Fig. 8.18):
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
! ffi
u
u k 2
¼ q1 t
0
zmax 1: ð8:74Þ
bq

As shown in Fig. 8.18, the non-diffractive range is defined by the intersection of

the leaky-wave radiation boundaries.
Two characteristics of the Bessel beam radiation become evident from Fig. 8.18
and (8.74): (1) larger non-diffractive ranges can be achieved with leaky-wave
modes pointing closer to broadside; (2) the far-field radiation pattern is conical and


points at h ¼ sin1 bq =k0 . As a result, the non-diffractive range can be located in
the Fresnel zone or near-field of the structure by changing the transverse propa-
gation constant from a value close to zero to k0 . In addition, larger non-diffractive
ranges correspond to Bessel beams with wider null-to-null beamwidths, where the
null-to-null beamwidth is given by NNBW  4:81=bq . Therefore, a tradeoff exists
between beam size and non-diffractive range, where the leaky complex
wavenumber is determined by (8.69), (8.70), and (8.74).
Finally, to completely define the fields within the leaky radial waveguide, the
coefficient D1 in (8.68) needs to be found. This can be done by choosing a
8 Near-Field Focusing by Non-diffracting Bessel Beams 277

particular excitation for the structure. Here, the coaxial excitation shown in
Fig. 8.17 is selected. It is assumed that the total current flowing on the inner
conductor of the coaxial probe’s surface is uniform and equal to I ¼ 2paJs , where a
is the radius of the inner conductor and Js is the surface current density. The
assumption is valid for very thin radial waveguides [57], h k0 ðk0 , free-space
wavelength at the operating frequency f0 Þ, and small values of the impedance sheet
Zs . Imposing the continuity of the /-component of the magnetic field along the
coaxial probe’s surface and for small probes’ radii, the coefficient D1 can be found:

I
D1 ¼ : ð8:75Þ
H0 ð k q q1 Þ
ð 2Þ

4j 1 þ
H0 ðkq q1 Þ
ð 1Þ

8.4.3 Dispersion Analysis of the Leaky Radial Waveguide

In order to design the leaky radial waveguide, closed-form expressions for Zs and h
as a function of the transverse propagation constant are needed. These expressions
can be found using the equivalent transmission line model of the radial waveguide
shown in Fig. 8.19. The dispersion equation of the radial waveguide will be derived
first. The parameters Zs and h will then be expressed in terms of the normal
propagation constant of the leaky-wave modes within the radial waveguide. Using
the transverse resonance technique, the dispersion equation can be expressed as

Yair þ Ys  jY1TM cotðkz1 hÞ ¼ 0; ð8:76Þ

with Yair ¼ 1=Zair ; Ys ¼ 1=Zs and Y1TM ¼ er1 k0 =g0 kz1 denoting the characteristic
admittance for a TM mode in free space. It is worth noting that the expression
provided in [59] can be used for Zair . However, in order to derive closed-form

Fig. 8.19 Equivalent

transmission line model for
the leaky radial waveguide in
Fig. 8.17
278 M. Ettorre et al.

expressions for Zs and h, in the following it is assumed Zs equal to the free-space

characteristic impedance of a TM mode [56].
Considering Zs ¼ jXs , and using the small argument expression for the cotangent
function, (8.76) becomes

k0 1 er1 k0 1
j  ¼ 0: ð8:77Þ
g0 k z Xs g0 kz1 kz1 h

Equation (8.77) can be further simplified for a hollow radial waveguide

ðer1 ¼ 1Þ. Under this assumption, (8.77) simplifies to

1 k2
kz h  j g0 h z þ 1 ¼ 0: ð8:78Þ
Xs k0


Given that the normal wavenumber is complex kz ¼ bz þ jaz for leaky-wave
modes, (8.78) can be separated into real and imaginary parts. After some algebrical
simplifications, Zs and h can be written as a functions of kz :

2az
Xs ¼ g; ð8:79Þ
k0 0

2az
h¼ : ð8:80Þ
þ b2z
a2z

Therefore, a unique solution can be found for Zs and h once the transverse or
normal propagation constant is fixed, assuming the structure is operating in its
fundamental mode. The sheet impedance Zs is found to be capacitive, as needed for
exciting TM-leaky-wave modes [58]. Equations (8.79) and (8.80) show that the
impedance sheet values are determined by the attenuation constant of the
leaky-wave mode, whereas the height h is dependent on both the propagation and
attenuation constants. In particular, lower values of h correspond to faster
leaky-wave modes, at a given frequency of operation. Finally, the expressions for
the general case of dielectric filled higher order (not operating on the principal
mode) leaky-wave radial waveguide can be found in [60]. Reference [60] also
defines the operating bandwidth of Bessel launchers based on radial leaky
waveguides.

8.4.4 Leaky Radial Waveguide: Experimental Results

In this section two prototypes are presented for the leaky-wave Bessel beam
launchers. The final prototypes have been designed using the procedure outlined in
the previous sections. More details about the structures and designs can be found in
[25, 60].
8 Near-Field Focusing by Non-diffracting Bessel Beams 279

Fig. 8.20 Leaky-wave Bessel beam launcher operating in X-band

The first prototype is shown in Fig. 8.20. It operates at f0 ¼ 10 GHz using the
principal leaky-wave mode within the radial waveguide. The launcher is fed by a
coaxial probe at the center of the structure.
The normalized transverse complex wavenumber of the excited leaky-wave
mode is kq =k0 ¼ 0:8  j0:0167: The aperture radius is fixed equal to 88.89 mm
(2.96 k0, k0 is the wavelength at f0). The sheet impedance and radial waveguide
height are given by (8.79) and (8.80): Zs ¼ 16:77j Ohm; h ¼ 0:588 mm. The
required impedance sheets for the two designs shown in this section are realized by
printing periodic square patch elements on both sides of a thin dielectric substrate,
as shown in Fig. 8.20, [25].
The Ez component of the electric field above the prototype was measured using a
short electric probe attached to an automatically controlled 3D translation stage.
The electric probe was approximately k0/10 (k0 is the free-space wavelength at
f0 = 10 GHz). Figure 8.21a and b reports the normalized Ez component of the
electric field of the Bessel beam launcher at z = 0.75 k0. The measured fields are at
f = 9.54 GHz [25]. This frequency shift with respect to the initial operating fre-
quency is attributed to tolerances inherent to the milling and etching fabrication
processes used for the prototype. Close agreement is observed between the simu-
lated and measured results. The cylindrical symmetry of the measured Bessel beam
is evident in the 2D plot of Fig. 8.21b. The dark and light rings correspond to nulls
and maxima of the Bessel beam. The 2D Fourier transform is shown in Fig. 8.21c.
The spectral ring of the Bessel beam is prominent confirming the generation of the
Bessel beam in free space. Figure 8.21d provides the measured Ez component of
the electric field along the xz-plane within the non-diffractive range equal to zmax
2.1 k0. The main beam, zeros, and shape of the field do not change for distances up
to z = 2 k0, further verifying the Bessel beam operation. However, at z = 2 k0
higher sidelobes appear at the edges of the measurements.
280 M. Ettorre et al.

(a) (b)

(c) (d)

Fig. 8.21 Measured and normalized Ez component of the electric field. a Comparison between
measured and simulated results at z = 0.75 k0 from the impedance surface. b 2D plot at z = 0.75
k0 from the impedance surface. c Normalized amplitude of the Fourier transform of the measured
Ez component of the electric field at the operating frequency at z = 0.75 k0 from the impedance
surface. d 2D plot along the xz-plane

Fig. 8.22 Leaky-wave

Bessel beam launcher
operating in Ka-band

The second prototype is shown in Fig. 8.22. The aim of this prototype is to use
higher order leaky-wave modes for the generation of Bessel beams [60]. As a
difference with the structure in Fig. 8.20, the present launcher is suited to Printed
8 Near-Field Focusing by Non-diffracting Bessel Beams 281

(a) (b)

Fig. 8.23 a Measured Ez component of the normalized electric field at f = 38.3 GHz along the
xz-plane. The white-dashed line corresponds to the limit of the non-diffractive range. b Variation
of the HPBW versus frequency for the Bessel beam launcher at z = k0 from the radiating aperture.
The vertical lines mark the extrema of the maximum available bandwidth as defined in [60]

Circuit Board (PCB) technology. The operating frequency is f ¼ 39:3 GHz with a
final radiating aperture of about 44.6 mm and a thickness smaller than 3.5 mm. The
launcher is still fed by a simple coaxial probe at the center of the structure.
Also in this case, the Ez component of the electric field above the prototype was
measured using a short electric probe attached to an automatically controlled 3D
translation stage. Figure 8.23 provides the measured Ez field component along the
xz-plane. As expected, the half-power beamwidth (HPBW) of the launched Bessel
beam is preserved within the non-diffractive range with an upper limit at
z = 16.4 mm (marked with a dashed-white line). Above this distance, diffraction
prevails, and the beam starts to widen.
The variation of the measured HPBW with the frequency at a fixed distance
(z = k0, with k0 is the free-space wavelength at f = 40 GHz) from the radiating
aperture is shown in Fig. 8.23b. Simulation results obtained with Ansys HFSS are
also proved for comparison. Due to the dispersive behavior of leaky-wave mode
generating the beam, both the simulated and the measured HPBWs slowly decrease
within the band 37.3–40.3 GHz. However, the measured HPBW remains almost
constant within the frequency range 38–39.5 GHz. This result proves that the
proposed launcher can generate a Bessel launcher over a relative bandwidth of
about 4%.

8.4.5 Bessel Launcher Using a Modulated Metasurface

This section presents an implementation of Bessel beam launchers using modulated

metasurfaces [32]. The modulated metasurface (surface impedance) is used to
transform a surface wave (SW), launched by a feeder, into a LW mode by using a
locally periodic modulation. In the following, we suppose that the feeder launches a
282 M. Ettorre et al.

cylindrical incident SW propagating along a metasurface placed in the xy-plane. For

a lossless non-modulated inductive scalar surface impedance Zs ¼ jX s [ 0Þ, the
 s ðX
dominant SW mode has a TMz polarization (where z is the normal to the surface,
and TM stands for transverse magnetic). The magnetic field on the surface is
assumed to be of the form

0 TM 0 ð2Þ
aperture ðq Þ ¼ A
Hinc ðq ÞH0 ðkt qÞ^
q; ð8:81Þ

where q0 ¼ x0 x^ þ y0 y^ is a general point on the antenna surface, ATM the amplitude,

and kt is obtained by solving the transverse dispersion problem [61].
The field radiated by the locally modulated metasurface can be expressed as the
field radiated by the equivalent surface magnetic currents, defined as
h i
Mðr0 Þ ¼ 2Eðr0 Þ  ^z ¼ 2 Zs ðq0 Þ^z  Hinc 0
aperture ðq Þ  ^
z
ð2Þ
¼ 2ATM ðq0 ÞZs ðq0 ÞH0 ðkt qÞ^
q: ð8:82Þ

A Bessel beam polarized along z-direction can be obtained by a magnetic current

^. The radial
aperture distribution placed in z ¼ 0 of the form (8.34) polarized along q
wavenumber kq and the radius of the metasurface qmax are directly related to the
non-diffractive range of the Bessel beam. The needed phase behavior can be
obtained from (8.82) using the following surface impedance [32]:
 
s 1 þ M ðq0 Þ cos kt q0 þ 3 p  pv  cos bBq q0  3 p
Zs ðq0 Þ ¼ jX ; ð8:83Þ
4 4

1x0
where vð xÞ ¼ is the step function.
0 x\0
It can be noted that (8.83) can also assure a similar amplitude law with respect to
(8.34) since the Bessel and the Hankel functions have the same amplitude factor in
their asymptotic expansion for large arguments. The two amplitude laws match
exactly only in the case where kq ¼ kt . This latter condition could not be satisfied
since kq \k0 while kt [ k0 . The effect of this small amplitude error (tapering) is
shown in Fig. 8.24a where a circular aperture of radius 3k illuminated with a Bessel
function with kq ¼ 0:6k0 has been considered. As can be seen, the above amplitude
difference gives rise to small differences on the secondary lobes while the focusing
effect is still present. However, it can be noticed that a perfect amplitude matching
could be obtained by locally changing radially the average impedance X s ðqÞ.
An alternative formulation can be obtained by using an inward cylindrical wave
distribution as in (8.1). Also for this case, a tapered distribution law can be obtained
(amplitude and phase) using the following expression [32]:
8 Near-Field Focusing by Non-diffracting Bessel Beams 283

(a) (b)

Fig. 8.24 a Near-field comparison at a distance of 2k from the antenna between an ideal Bessel
beam, the field generated by an aperture described by Eq. (8.34) and its metasurface antenna
implementation (8.83). b Same comparison using the aperture distribution (8.1) and its
metasurface implementation (8.84)

Fig. 8.25 Reactance chart for a Bessel beam polarized along z, generated with a J1 distribution or
ð1Þ
an H1 distribution. The average impedance is Zs = j 290 Ω and modulation index M = 0.17

 n o
s 1 þ M ðq0 ÞRe ejArgðH1 ðkt q0 ÞÞ ejArgðH1 ðbq q0 ÞÞ :
ð 2Þ ð 1Þ
Zs ðq0 Þ ¼ jX
B
ð8:84Þ

Similar consideration on the aperture field amplitudes holds also for this con-
figuration as shown in Fig. 8.24b.
Figure 8.24 shows a comparison between the ideal Bessel beam, the tapered
ideal beam, and the field radiated by the metasurface antenna at 2k from the antenna
obtained with HFSS for both configurations. A good agreement is observed
between theory and full-wave results. Reactance chart examples for both formu-
lations are shown in Fig. 8.25. An operating frequency of 20 GHz is considered in
all cases.
284 M. Ettorre et al.

Fig. 8.26 Normalized near-field scanning in dB for a Bessel beam antenna a using the aperture
distribution (8.34). b Alternative aperture distribution (8.1)

The two presented configurations have been implemented using a circular panel
of radius 3k. The metasurface impedance is realized by printing squared patches
with different dimensions on a squared lattice of side 1.25 mm on a grounded
dielectric slab (Rogers 4350b with a relative permittivity of 3.66 and a thickness of
1.52 mm). A higher modulation index has been used close to the periphery of the
antenna in order to prevent reflections [32]. The parameter for the Bessel functions
has been selected as kq ¼ 0:6k0 ; the corresponding non-diffracting zone is given by
z  4k.
Figure 8.26 represents the map of the near-field generated by the metasurface
launcher. The non-diffractive behavior of the generated beam can be recognized
within the non-diffracting region of the launcher.

8.5 Conclusions and Future Directions

This chapter has briefly introduced the concept of non-diffractive radiation with a
special focus on Bessel beams. The possibility to generate Bessel beams with
inward-traveling waves has been presented. It has been proved that
inward-traveling waves can generate non-diffractive Bessel beams over a rhom-
boidal region in front of the radiating aperture. In addition, the non-diffractive
radiation is achieved over a large bandwidth with respect to classical methods based
on resonant radiating aperture. The wideband feature of such launchers is extremely
appealing for the generation of non-diffractive pulses and in particular of X-waves
[1, 2, 62]. Besides, planar radiating structures are preferred at millimeter waves to
cumbersome axicon lenses adopted at optical frequencies [9, 10].
An efficient design procedure has then been presented for planar Bessel
launchers based on radial waveguides loaded with radiating slots (RLSA antennas)
or metallic gratings. The procedure allows synthetizing a defined aperture field
distribution by controlling the location and size of the slots or metallic gratings of
the radial waveguide. The procedure has been adopted for designing a RLSA Bessel
8 Near-Field Focusing by Non-diffracting Bessel Beams 285

launcher operating in Ku-band with circular polarization and a deep of focus of

several meters. Such launcher can find application in remote sensing for the
detection of buried objects [34] or as a field concentrator for drilling operations in
planetary explorations [63].
The same design procedure has been used for a Bessel launcher using
inward-traveling waves. The launcher is made by a radial waveguide loaded with
metallic gratings. The structure is suited to low-cost printed circuit board fabrication
processes. The wideband behavior, low loss, and low cost of such launcher are
attractive features for imaging applications at millimeter waves [64]. Besides, the
proposed structures may find applications for future short-range wireless links for
next-generation communication networks in D-band (120 GHz) and even beyond
300 GHz [65]. On the other hand, the design procedure proposed here for
non-diffractive launcher can be adopted for the design of antenna systems with an
arbitrary near-field profile [66].
The chapter has also outlined the possibility to generate Bessel beams with
leaky-wave modes. The non-diffractive beam is obtained by the radiation of an
inward and outward leaky-wave mode traveling within a radial waveguide. These
modes are excited by a coaxial probe at the center of the radial waveguide loaded
with a uniform metasurface. The propagation constant of the leaky-wave modes
controls the non-diffractive range and beamwidth of the generated beam. In par-
ticular, it has been shown that resonant apertures are needed for the generation of
the non-diffractive radiation. Despite the resonant nature of the radiating aperture,
Bessel launchers based on leaky-wave modes are a novel paradigm for the gen-
eration of non-diffractive radiation.
The combination of this concept with the recent achievements on
inward-traveling waves presented in this chapter may pave the way to extremely
efficient launcher at millimeter waves. Recent works are already pursuing such
possibility by using metasurfaces [27, 30, 32]. The possible applications of such
launchers include low-profile antennas [67], wireless power transfer systems [68],
microparticle manipulations with tractor beams [69], and the generation of surface
plasmon polaritons [70].

References

1. H.E. Hernandez-Figueroa, M. Zamboni-Rached, E. Recami, Localized Waves (J. Wiley, NY,

2008)
2. H.E. Hernandez-Figueroa, E. Recami, M. Zamboni-Rached, Non-diffractive Waves
(J. Wiley-VCH, Berlin, 2013)
3. J. Durnin, Exact solutions for nondiffracting beams. I. The scalar theory. J. Opt. Soc. Am.
A 4, 651–654 (1987)
4. J. Durnin, J.J. Miceli Jr., J.H. Eberly, Diffraction-free beams. Phys. Rev. Lett. 58, 1499 (1987)
5. S. Ruschin, A. Leizer, Evanescent Bessel beams. J. Opt. Soc. Amer. A 15(5), 1139–1143
(1998)
286 M. Ettorre et al.

6. M.A. Bandres, J.C. Gutierrez-Vega, S. Chàvez-Cerda, Parabolic nondiffracting optical wave

fields. Opt. Lett. 29, 44–46 (2004)
7. J.C. Gutierrez-Vega, M.D. Iturbe-Castillo, S. Chàvez-Cerda, Alternative formulation for
invariant optical fields: mathieu beams. Opt. Lett. 25, 1493–1495 (2000)
8. J. Salo, J. Fagerholm, A.T. Friberg, M.M. Salomaa, Unified description of nondiffracting X
and Y waves. Phys. Rev. A 62, 4261 (2000)
9. M.R. Lapointe, Review of non-diffracting Bessel beam experiments. Opt. Laser Tech. 24(6),
315 (1992)
10. J. Arlt, K. Dholakia, Generation of high-order Bessel-beams by use of an axicon. Opt.
Commun. 177(4), 297–301 (2001)
11. O. Brzobohaty, T. Cizmàr, P. Zemànek, High quality quasi-Bessel beam generated by
round-tip axicon. Opt. Express 16, 12688–12700 (2008)
12. J. Turunen, A. Vasara, A.T. Friberg, Holographic generation of diffraction-free beams. Appl.
Opt. 27(19), 3959–3962 (1988)
13. J. Williams, J.B. Pendry, Generating Bessel beams by use of localized modes. J. Opt. Soc.
Am. A 22(5), 992–997 (2005)
14. G. Indebetouw, Nondiffracting optical fields: some remarks on their analysis and synthesis.
J. Opt. Soc. Am. A. 6(1), 150–152 (1989)
15. A.J. Cox, D.C. Dibble, Nondiffracting beam from a spatially filtered Fabry-Perot resonator.
J. Opt. Soc. Am. A. 9(2), 282–286 (1992)
16. Q. Zhan, Cylindrical vector beams: from mathematical concepts to applications. Adv. Opt.
Photon. 1, 1–57 (2009)
17. D. McGloin, K. Dholakia, Bessel beams: diffraction in a new light. Contemp. Phys. 46, 15–28
(2005)
18. S. Monk, J. Arlt, D.A. Robertson, J. Courtial, M.J. Padgett, The generation of Bessel beams at
millimetre-wave frequencies by use of an axicon. Opt. Commun. 170, 213–215 (2000)
19. J. Salo, J. Meltaus, E. Noponen, J. Westerholm, M.M. Salomaa, A. Lönnqvist, J. Säily,
J. Häkli, J. Ala-Laurinaho, A.V. Räisänen, Millimetre-wave Bessel beams using computer
holograms. Electron. Lett. 37(12), 834–835 (2001)
20. S.C. Pavone, M. Ettorre, M. Albani, Analysis and design of Bessel beam launchers:
longitudinal polarization. IEEE Trans. Antennas Propag. 64, 2311–2318 (2016)
21. Z. Li, K.B. Alici, H. Caglayan, E. Ozbay, Generation of an axially asymmetric Bessel-like
beam from a metallic subwavelength aperture. Phys. Rev. Lett. 102, 143901 (2009)
22. D. Mugnai, A. Ranfagni, R. Ruggeri, Observation of superluminal behaviors in wave
propagation. Phys. Rev. Lett. 84, 4830–4833 (2000)
23. P. Lemaitre-Auger, S. Abielmona, C. Caloz, Generation of Bessel beams by two-dimensional
antenna arrays using sub-sampled distributions. IEEE Trans. Antennas Propag. 61, 2645–
2653 (2012)
24. M.A. Salem, A.H. Kamel, E. Niver, Microwave Bessel beams generation using guided
modes. IEEE Trans. Antennas Propag. 59(6) part 2, 2241–2247 (2011)
25. M. Ettorre, S.M. Rudolph, A. Grbic, Generation of propagating Bessel beams using
leaky-wave modes: experimental validation. IEEE Trans. Antennas Propag. 60, 2645–2653
(2012)
26. C. Pfeiffer, A. Grbic, Metamaterial Huygens’ surfaces: tailoring wave fronts with
reflectionless sheets. Phys. Rev. Lett. 110, 197401 (2013)
27. C. Pfeiffer, A. Grbic, Controlling vector Bessel beams with metasurfaces. Phys. Rev. Appl. 2,
044012 (2014)
28. Y.B. Li, B.G. Cai, X. Wan, T.J. Cui, Diffraction-free surface waves by metasurfaces. Opt.
Lett. 39, 5888–5891 (2014)
29. F. Aieta, P. Genevet, M.A. Kats, N. Yu, R. Blanchard, Z. Gaburro, F. Capasso,
Aberration-free ultrathin flat lenses and axicons at telecom wavelengths based on plasmonic
metasurfaces. Nano Lett. 12, 4932–4936 (2012)
30. B.G. Cai, Y.B. Li, W.X. Jiang, Q. Cheng, T.J. Cui, Generation of spatial Bessel beams using
holographic metasurface. Opt. Express 23, 7953—7601 (2015)
8 Near-Field Focusing by Non-diffracting Bessel Beams 287

31. M.Q. Qi, W.X. Tang, T.J. Cui, A broadband Bessel beam launcher using metamaterial lens.
Sci. Rep. 5, 11732 (2015)
32. M. Casaletti, M. Śmierzchalski, M. Ettorre, R. Sauleau, N. Capet, Polarized beams using
scalar metasurfaces. IEEE Trans. Antennas Propag. AP-64(8), 3391–3400 (2016)
33. M.F. Imani, A. Grbic, Generating evanescent Bessel beams using near-field plates. IEEE
Trans. Antennas Propag. 60, 3155–3164 (2012)
34. A. Mazzinghi, M. Balma, D. Devona, G. Guarnieri, G. Mauriello, M. Albani, A. Freni, Large
depth of field pseudo-Bessel beam generation with a RLSA Antenna. IEEE Trans. Antennas
Propag. 62(15), 3911–3919 (2014)
35. M. Ettorre, S.C. Pavone, M. Casaletti, M. Albani, Experimental validation of Bessel beam
generation using an inward Hankel aperture distribution. IEEE Trans. Antennas Propag. 63
(6), 2539–2544 (2015)
36. L.B. Felsen, N. Marcuvitz, Radiation and Scattering of Waves. Electromagnetic Wave Theory
(IEEE Press, Piscataway, NJ, 1994)
37. R. Cicchetti, A. Faraone, Incomplete Hankel and modified Bessel functions: a class of special
functions for electromagnetics. IEEE Trans. Antennas Propag. 52, 3373–3389 (2004)
38. M. Albani, S.C. Pavone, M. Casaletti, M. Ettorre, Generation of non-diffractive Bessel beams
by inward cylindrical traveling wave aperture distributions. Opt. Express 22, 18354–18364
(2014)
39. S.C. Pavone, M. Ettorre, M. Casaletti, M. Albani, Transverse circular-polarized Bessel beam
generation by inward cylindrical aperture distribution. Opt. Express 24, 11103–11111 (2016)
40. M. Albani, A. Mazzinghi, A. Freni, Automatic design of CP-RLSA antennas. IEEE Trans.
Antennas Propag. 60, 5538–5547 (2012)
41. M. Ando, K. Sakurai, N. Goto, K. Arimura, Y. Ito, A radial line slot antenna for 12 GHz
satellite TV reception. IEEE Trans. Antennas Propag. AP-33(12), 1347–1353 (1985)
42. M. Takahashi, J. Takada, M. Ando, N. Goto, A slot design for uniform aperture field
distribution in single-layered radial line slot antennas. IEEE Trans. Antennas Propag. 39(7),
954–959 (1991)
43. T. Yamamoto, M. Ando, N. Goto, and M. Takahashi, Application of radial line slot antenna to
mil-limiter-wave systems at 60 GHz band, in Proceedings of the 1996 IEEE AP-S
International Symposium, vol. 1, pp. 62–65, 21–26 July 1996
44. H. Elmikati, A. Elsohly, Extension of projection method to nonuniformly linear antenna
arrays. IEEE Trans. Antennas Propagat. 32(5), 507–512 (1984)
45. O.M. Bucci, G. D’Elia, G. Mazzarella, G. Panariello, Antenna pattern synthesis: a new
general approach. Proc. IEEE 82(3), 358–371 (1994)
46. H.A. Bethe, Theory of diffraction by small holes. Phys. Rev. 66(7–8), 163–182 (1944)
47. C.A. Balanis, Advanced Engineering Electromagnetics, Chapter 9 (Wiley, NY, 1989)
48. J. Takada, M. Ando, N. Goto, A reflection canceling slot set in a linearly polarized radial line
slot antenna. IEEE Trans. Antennas Propag. AP-40(4), 433–438 (1992)
49. P.W. Davis, M.E. Bialkowski, Experimental investigation into a linearly polarized radial slot
antenna for DBS TV in Australia. IEEE Trans. Antennas Propag. AP-45(7), 1123–1129
(2007)
50. M.E. Bialkowski, P.W. Davis, Analysis of a circular patch antenna radiating in a parallel-plate
radial guide. IEEE Trans. Antennas Propag. AP-50(2), 180–187 (2002)
51. A. Akiyama, T. Yamamoto, M. Ando, N. Goto, Numerical optimisation of slot parameters for
a concentric array radial line slot antenna. IEE Proc. Microw. Antennas Propag. 145(2), 141–
146 (1998)
52. J.I. Herranz, A. Valero-Nogueira, F. Vico, V.M. Rodrigo, Optimization of beam-tilted linearly
polarized radial-line slot-array antennas. IEEE Antennas Wirel. Propag. Lett. 9, 1165–1168
(2010)
53. J.I. Herranz-Herruzo, A. Valero-Nogueira, M. Ferrando-Bataller, Optimization technique for
linearly polarized radial-line slot-array antennas using the multiple sweep method of
moments. IEEE Trans. Antennas Propag. AP-52(4), 1015–1023 (2004)
288 M. Ettorre et al.

54. H. Ueda, J. Hirokawa, M. Ando, M. Albani, A coaxial feeder with two pairs of parasitic pins
for realizing rotationally symmetric aperture illumination in spiral array radial line slot
antennas. IECIE Trans. Commun. E93-B, 2554–2561 (2010)
55. S. Chàvez-Cerda, A new approach to Bessel beams. J. Mod. Opt. 46, 923–930 (1999)
56. M. Ettorre, A. Grbic, Generation of propagating Bessel beams using leaky-wave modes. IEEE
Trans. Antennas Propag. 60, 3605–3613 (2012)
57. R.F. Harington, Time-Harmonic Electromagnetic Fields (Wiley-IEEE Press, New York,
2001)
58. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill, New York, 1969)
59. N. Marcuvitz, Waveguide Handbook (McGraw-Hill, New York, 1951)
60. W. Fuscaldo, G. Valerio, A. Galli, R. Sauleau, A. Grbic, M. Ettorre, Higher-order leaky-mode
Bessel-Beam launcher. IEEE Trans. Antennas Propag. 60(3), 904–913 (2016)
61. A. Oliner, A. Hessel, Guided waves on sinusoidally-modulated reactance surfaces. IRE Trans.
Antennas Propag. 7(5), 201–208 (1959)
62. W. Fuscaldo, S.C. Pavone, G. Valerio, A. Galli, M. Albani, M. Ettorre, Analysis of
limited-diffractive and limited-dispersive X-waves generated by finite radial waveguides.
J. Appl. Phys. 119, 194903 (2016)
63. S. Costanzo, G. Di Massa, A. Borgia, A. Raffo, T.W. Versloot, L. Summerer, Microwave
Bessel beam launcher for high penetration planetary drilling operations, in 10th European
Conference on Antennas and Propagation, Switzerland (2016)
64. H. Meng, J. Zhang, B. Xiang, W. Dou, Y. Yu, Millimeter Wave Imaging with Bessel Beams
(IEEE MTT-S International Microwave Symposium Digest, USA, 2013)
65. P. Smulders, The road to 100 Gb/s wireless and beyond: basic issues and key directions. IEEE
Commun. Mag. 51(12), 86–91 (2013)
66. M. Ettorre, M. Casaletti, G. Valerio, R. Sauleau, L. Le Coq, S.C. Pavone, M. Albani, On the
near-field shaping and focusing capability of a radial line slot array. IEEE Trans. Antennas
Propag 62, 1991–1999 (2012)
67. C. Pfeiffer, A. Grbic, Planar lens antennas of subwavelength thickness: collimating
leaky-waves with metasuraces. IEEE Trans. Antennas Propag. AP-63(7), 3248–3253 (2015)
68. M. Ettorre, M. Casaletti, A. Grbic, Power link budget for propagating Bessel beams, in IEEE
Antennas and Propagation Society International Symposium (APSURSI) (2013), pp. 960–961
69. C. Pfeiffer, A. Grbic, Generating stable tractor beams with dielectric metasurfaces. Phys. Rev.
B 91, 115408 (2015)
70. J. Lin, J. Dellinger, P. Genevet, B. Cluzel, F. de Fornel, F. Capasso, Cosine-Gauss plasmon
beam: a localized long-range nondiffracting surface wave. Phys. Rev. Lett. 109, 093904
(2012)
Chapter 9
Metasurface Antennas

Gabriele Minatti, Marco Faenzi, Mario Mencagli,

Francesco Caminita, David González Ovejero,
Cristian Della Giovampaola, Alice Benini, Enrica Martini,
Marco Sabbadini and Stefano Maci

Abstract This chapter reports design and analysis methods for planar antennas
based on modulated metasurfaces (MTSs). These antennas transform a surface
wave (SW) into a leaky wave by means of the interaction with a MTS having a
spatially modulated equivalent impedance. The basic concept is that the MTS
imposes the impedance boundary conditions (BCs) seen by the SW, and therefore
the MTS controls amplitude, phase, and polarization of the aperture field. Thus,
MTS antennas are highly customizable in terms of their performances, by simply
changing the MTS and without affecting the overall structure. Several technological
solutions can be adopted to implement the MTS, from sub-wavelength patches
printed on a grounded slab at microwave frequencies, to a bed of nails structure in
the millimetre and sub-millimetre wave range: in any case, the resulting device has
light weight and a low profile. The design of the MTS is based on a generalized
form of the Floquet wave theorem adiabatically applied to curvilinear locally
periodic BCs. The design defines the continuous BCs required for reproducing a

G. Minatti
M. Faenzi
M. Mencagli
F. Caminita
C.D. Giovampaola
A. Benini

E. Martini
S. Maci (&)
Department of Information Engineering and Mathematics, University of Siena,
Via Roma 56, 53100 Siena, Italy
e-mail: [email protected]
G. Minatti
e-mail: [email protected]
M. Faenzi
e-mail: [email protected]
M. Mencagli
e-mail: [email protected]
C.D. Giovampaola
e-mail: [email protected]
A. Benini
e-mail: [email protected]
E. Martini
e-mail: [email protected]

© Springer International Publishing AG 2018 289

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_9
290 G. Minatti et al.

desired aperture field, and it is verified by a fast full-wave solver for impedance
BCs. Next, the continuous BCs are discretized and implemented by a distribution of
electrically small printed metallic elements in a regular lattice, like pixels in an
image. The final layout is composed of tens of thousands of pixels and it is analyzed
by a full-wave solver which makes use of entire domain basis functions combined
with a fast-multipole algorithm. Examples of design and realizations of MTS
antennas are shown, proving the effectiveness of the concept.

9.1 Introduction

Metasurfaces (MTSs) are thin metamaterial layers characterized by unusual reflection

properties of plane waves and/or dispersion properties of surface/guided waves.
Metamaterials are artificial materials that can be formed by periodic arrangements of
many small inclusions in a dielectric host environment, for achieving macroscopic
electromagnetic or optical properties that cannot be found in nature.
MTSs have found several applications in designing antennas and microwave
components and they may be distinguished in penetrable and impenetrable.
A penetrable MTS (sometimes called metafilm) is constituted by a planar distri-
bution of small periodic elements in a very thin host medium [1]. A ground plane is
instead a constitutive element for impenetrable MTSs, which are the ones treated in
this chapter. In the microwave regime, impenetrable MTSs are usually formed by a
dense texture of small elements in a regular lattice, printed on a grounded slab with
or without shorting vias. Impenetrable MTSs can be also simply constituted by a
dense distribution of pins [2] on a ground plane. This solution is usually adopted at
the optical frequencies to overcome the issue of dielectric losses.
The MTS is referred to as uniform when the elements in the periodic lattice are
identical all over the MTS plane [3]. Here, however, we deal with aperiodic ele-
ments having geometrical parameters gradually changing from cell-to-cell: in this
case the MTS is referred to as nonuniform or modulated. Nonuniform MTSs allow
one to change the phase velocity and/or propagation path of the guided wave
sustained by the MTS.

F. Caminita
E. Martini
Wave Up s.r.l, Via Edimburgo 6, 50126 Florence, Italy
e-mail: [email protected]
D. González Ovejero
Institut d’Électronique et de Télécommunications de Rennes - UMR CNRS 6164, 263
Avenue du Général Leclerc, Rennes 35042, France
e-mail: [email protected]
M. Sabbadini
European Space Agency, Keplerlaan 1, 2201 AZ Noordwijk, The Netherlands
e-mail: [email protected]
9 Metasurface Antennas 291

The applications of modulated MTSs can be divided in four classes: (a) MTSs
for aperture antennas [4–12], (b) MTSs for the control of surface-waves (SWs)
wavefront [13–16], (c) MTS screens for the control of field transmission [17–19],
(d) MTSs for EBG [20]. Here, we focus the attention on the first class of
applications.
The subject of surface wave control through MTSs has been treated in several
recent papers ([4–33]), some of them addressing the relevant phenomena in the
framework of Transformation Optics (see [14, 23–29] and Chap. 11 of this book).
On the other hand, the control of leaky-wave (LW) radiation through modulation of
boundary conditions (BCs) has led to a new class of modulated MTS antennas [4–
9] (referred to as MoMetAs in [6], which stands for Modulated Metasurface
Antennas). This kind of antennas transforms a bounded SW into a radiative LW
through the SW interaction with a modulated MTS. In the microwave range, the
MTS antennas we are dealing with are typically constituted by sub-wavelength
patches of various shapes, printed on a grounded dielectric substrate [4–10]. They
exhibit low profile, low weight, reduced complexity, and low cost. Neither external
protruding or backing feed arrangements nor (sub-) reflectors are required, since the
feeding element is embedded in the MTS; this constitutes an advantage with respect
to other types of printed antennas, like reflectarrays, especially for space applica-
tions [6]. Although modulated MTS antennas are based on a LW mechanism, their
behavior offers an enormous flexibility with respect to conventional LW antennas.
The key features of modulated MTS antennas, making them very appealing mostly
for space applications but also for a wide range of ground applications, can be
summarized as follows:
• Unique decoupling of electrical and thermo-mechanical design.
• Unique capability of low-complexity (dynamic) beam control (shaping, point-
ing, and scanning).
• Unique support for late electrical performance refinement and design reusability.
• Complete and simple on-surface control of medium-to-large size apertures.
• Low power losses.
• Low mass and low envelope.
• Implementation amenable to very different technologies.
• Simple low-cost manufacturing.
Some prototypes of MTS antennas are shown in Fig. 9.1. At the frequency of
operation, the elements that implement the BCs are small in terms of the wave-
length (between k/5 and k/10, with k being the free-space wavelength), so that the
interacting SW sees the interface BC as a continuum. These elements behave like
pixels in a black and white printed image, whose gray scale is realized by changing
the dimension of the pixel elements inside a regular lattice. One can also control the
polarization by using asymmetric pixel elements (this can be done, e.g., by adding a
feature like a cut or a slot to a circular patch, or using an elliptical patch), and
rotating them inside the lattice.
292 G. Minatti et al.

Fig. 9.1 Examples of realized prototypes of MTS antennas. The SW is excited by a single probe
(left and right insets) or by 4 probes (central inset). In the three examples, the feeding region is
located at the center of the MTS

Table 9.1 summarizes a number of modulated MTS designs, several of which

have been built and measured, while the most recent ones are yet to be manufac-
tured. Nevertheless, the extremely good accuracy of the design tools in use makes it
possible to rely on the calculated performance figures.
The model of these antennas is inherently multi-scale (see Fig. 9.2). To convey
the idea, we adopt a terminology used in physics (Fig. 9.2), to denote the scales in
terms of (a) “macroscopic scale”, (b) “mesoscopic scale”, (c) “molecular scale”, and
(d) “atomic scale”. The largest dimension of the antenna can be several wavelengths
(e.g., 15k, Fig. 9.2a). The local period of the global modulation is normally of the
order of one wavelength (Fig. 9.2b). Zooming in the scale, one sees the lattice of
the texture, which consists of pixel elements with a size of approximately k/10
(Fig. 9.2c). The specific features added to the patches to control polarization (e.g.,
the slots) can be of the order of k/50, which is approximately the size of the
discretization triangles required in a Method of Moment mesh (Fig. 9.2d). It is,
hence, clear that a conventional global full-wave analysis of the overall structure
would require the solution of a million-unknowns problem. For this reason, a single
model cannot cover all the scales, and the synthesis procedure cannot be solely
based on a direct full-wave solver. Note that the number of unknowns is even larger
if the computational approach is based on other methods such as FEM or FDTD.
Here, we only deal with MoM methods for solving these structures, as the use of
other methods is out of the scope of this chapter. In the following, we introduce a
design process for modulated MTS antennas that makes use of different models for
the different scales, corresponding to different levels of accuracy.
Table 9.1 Spot-beam modulated metasurface antennas—design and performances summary
Configurations
Date 2010 2014 2015 2015 2015 2015 2015 2015 2015 2016 2016
Application – Space exploration – – – Telecom EO – Space exploration Space exp Telecom
Type Disk Disk Disk Disk Disk Disk Disk Disk Disk Disk Disk
Verification Test Test Test Test Analysis Analysis Analysis Analysis Analysis Model Model
9 Metasurface Antennas

Frequency (GHz) 17 8.3 7.1 28 28 13.25 29.5 26.3 25.1 7.1 8.4 7–8 29.5 GHz
Diameter 200 mm 540 mm 300 mm 86 mm 86 mm 120 mm 560 mm 270 mm 270 mm 890 mm 1m 1m
Directivity (dBi) 25 28.5 21.4 23.8 23.3 26 43.0 36.2 24 27.5 32.0 36 48.4
Gain (dBi) 24.5 28.0 21.1 23.5 23.0 – – – – – – 35.5 47.4
Pointing angle (°) 0 0 0 30 0 0 0 0 0 0 0 0
Polarization CP CP CP CP LP Dual-LP CP CP CP CP CP CP
Cross-polar level −13 dB −25 dB −20 dB −25 dB −25 dBi −25 dB −30 dB −30 dB – −25 dB −30 dB −35 dB −35 dB
−3 dB 5.5 4.6 13.3 10.3 8.1 6.0 1 2.7 – 3.5 3.0 2.5 0.7
beamwidth (°)
Bandwidth @ 3 4 6 – – – – – 12.5 3 3 17 5
−1 dB (%)
Bandwidth @ 6 8 15 – – – – – 25 6 6 34 10
−3 dB (%)
Efficiencya (%) 25 32 27 43 34 40 65 70 – – – 75 73
Return loss (dB) −20 −23 −20 −20 −22 – – – – – – −25 −25
Isolation – >60 dB – – – – – – – – – –
Remarks First Split-aperture Low Low Efficiency Bandwidth Dual-band Low Low
design solution accuracy accuracy optimisat. enlargem. solution accuracy accuracy
attempt prototype prototype modeling modeling
a
Efficiency = Aperture efficiency  Spillover efficiency (reduction of aperture illumination w.r.t uniform law  amount of surface wave power converted into radiation)
293
294 G. Minatti et al.

(a) (b) (c) (d)

15λ λ λ / 10 λ / 50

Fig. 9.2 Multiscale features of MTS antennas: “macroscopic” scale (a), “mesoscopic” scale (b),
“molecular” scale (c), “atomic” scale (d)

The chapter is organized as follows. Section 9.2 provides a global picture of the
design process, whose blocks are described more in detail in the following sections.
In particular, Sect. 9.3 introduces the Flat Optics model for the description of fields
and currents on the modulated MTS. Section 9.4 describes the synthesis scheme for
the design of an impedance BC providing a desired radiating aperture field.
Section 9.5 discusses the full-wave analysis method based on a continuous impe-
dance BC model. Section 9.6 presents the technique used for the synthesis of the
BC, based on the extraction of the equivalent impedance of the single pixels.
Section 9.7 illustrates an efficient MoM approach for the full-wave analysis of
textured MTS antennas consisting of elliptical patches. An example of design
process, including all the previously mentioned steps, is presented in Sect. 9.8.
Also, other designs for highly efficient antennas or wide band capability are dis-
cussed in Sect. 9.8. The issue of the efficiency for these antennas is discussed in
Sect. 9.9. Finally, practical examples of MTS antennas are reported in Sect. 9.10,
and conclusions are given in Sect. 9.11.

9.2 Overview of the Design Process

The global synthesis scheme is shown in Fig. 9.3 [12]. It is composed of three main
blocks, each of them working on a different model scale. The first block synthesizes
the continuous impedance surface, which, when excited by a reference SW pro-
duces the objective aperture field. Then, within the second block, the continuous
impedance surface is discretized and implemented through a dense texture of small
patches. We will refer to them as “pixels” since they resemble pixels in a black and
white printed image. Finally, the third block analyzes the textured antenna layout
through a global full-wave solver. Since the number of pixels is quite large (typi-
cally several thousands), the global solver must be engineered with smart numerical
strategies to overcome the computational complexity arising from the huge number
of unknowns of the linear problem.
9 Metasurface Antennas 295

few minutes
DATA− BASE
Objective
(based on a Pixel
Aperture Field periodic modeling
full-wave solver)
few seconds EA Atomic level
(Periodic BC)
Continuous Antenna
Reactance YES Layout Molecular level
NO (SFX)
Synthesis

Mesoscopic X Check Global Final

Level (continuous Full-wave
(continuous BC) check YES
impedance analysis
MoM)
few minutes NO 1 hour

Fig. 9.3 Block diagram of the synthesis process

Fig. 9.4 Geometry of the

problem. The sub-wavelength
2π
printed patches are modeled
X d (ρ) =
by an anisotropic tensor X op K ∇t s ( ρ ) ⋅ ρˆ
reactance, which defines
h εr
through (9.1) discontinuous ρ
z
BCs over a grounded y
dielectric slab (inset on the
left corner). The inset on the
right top corner sketches the
local periodic problem for the ρ
definition of the adiabatic ϕ
Floquet modes x

The design process starts assuming that the radiative components of the objec-
tive aperture field are known. We do not treat here the process to synthesize a
continuous aperture field from the pattern requirements, since this is the subject of a
rich literature, see for instance [34] and the references therein. Within the first
block, the synthesis of the BCs is performed at “mesoscopic” level (area on the left
in Fig. 9.3), namely, assuming that the patterned metallic cladding can be treated as
a continuous, space-dependent reactance tensor X, which models continuous ani-
sotropic BCs related to the discontinuity of the magnetic field (Fig. 9.4). The block
in red is the core of this task, and it is based on the analytical method introduced in
[35, 36] that we denote succinctly as “Flat Optics” (FO) and synthesized in the next
paragraph. The time required to produce X is a few seconds using a good notebook
independently of the antenna size. After the synthesis has been performed,
296 G. Minatti et al.

a full-wave analysis still based on a continuous BC approach, is used to check the

accuracy of the synthesis. This step is performed with an extremely fast formulation
described in [37], which requires less than 10 s for antennas with diameter of 20
wavelengths. If the first check with this solver is successful, one proceeds with the
synthesis of the MTS elements.
The element synthesis (called pixel modeling in Fig. 9.3) is carried out at
“atomic” level, namely, using a full-wave code with RWG basis functions. A local
“micro-periodicity” concept is used in this phase: it is assumed that the local
element is immersed in a periodic environment of equal elements with the peri-
odicity of the lattice. The above assumption legitimizes the use of periodic BCs on
the elemental cell boundaries, and therefore the use of the periodic Green’s func-
tion, thus reducing the computational effort to that of a single periodic cell. The
analysis of the periodic structure, which is inherently extremely fast, is repeated
several times in order to construct a database that relates the different geometries of
the unit cell element with the corresponding entries of the anisotropic reactance
tensor. The database is built before the entire synthesis process, and possibly reused
for other syntheses. Actually, only a few samples of the parameter-space are
directly calculated and interpolation is next used to create a denser database.
The database construction requires few tens of minutes for a 2D database (i.e.,
obtained by varying 2 geometrical parameters) and few hours for a 3D database (3
geometrical parameters). The database is normally constructed at the antenna
operating frequency; its extension around the frequency of operation can be
important when the antenna is required to work over a large band. To this purpose,
the pole-zero matching method [38] or the analytical model presented in [32] can be
used. The output of the “atomic” part of the design is a complete layout of the
antenna, where the dimensions and the orientation of any single pixel-element are
defined.
The final phase of the design process is a detailed analysis of the layout, which
requires a global full-wave integral equation model. This model is implemented at
“molecular level”, namely using few entire domain basis functions per pixel-
element and incorporating them into a fast-multipole algorithm to accelerate the
solution. In case the final check is not successful, a repetition of the continuous
model synthesis may be required. Actually, as the analytic core of the synthesis, i.e.,
the “continuous reactance synthesis” block in Fig. 9.3, is powerful and accurate, the
overall process rarely requires a feedback, which is typically needed only for very
demanding requirements, like large bandwidth or very high efficiency.
The synthesis step that we are going to introduce in the next paragraph is based
on a description of the MTS in terms of “transparent” BC in place of “opaque” BC
(see left-side inset of Fig. 9.4). With the first terminology, introduced in [39], we
denote the relationship between the tangential electric field and the discontinuity of
the magnetic field at the metallic cladding level. “Opaque” BCs relate instead the
tangential electric field to the limit of the magnetic field at the interface between the
MTS and free space. The motivation of this choice is that the transparent BC is less
spatially dispersive than the opaque BC [32,33].
9 Metasurface Antennas 297

In the next section, we present the main issue concerning each of the blocks of
the design scheme in Fig. 9.3. In the last part of the chapter, we summarize the
presented design workflow through an applicative example.

9.3 Adiabatic Floquet Wave Expansion

The synthesis of the continuous reactance (red block in Fig. 9.3) is based on a
theoretical model we have developed to describe currents and fields on a class of
synthesizable reactances that we will refer to as “constant-average non uniform
reactances”. Such reactances describe the transparent locally periodic modulated
BCs formed by the patch cladding on a grounded dielectric slab. We have found
that currents and fields on such a locally periodic environment are well described by
an adiabatic Floquet Wave (FW) expansion matched to the local periodicity.
Let us refer to the configuration shown in Fig. 9.4. The MTS is defined on a
circular surface of radius a, centered at the origin of a reference system described in
cylindrical coordinates ðq; uÞ; with unit vectors ð^ ^ Þ. Accordingly, the observa-
q; u
tion point on the surface at the interface between the metal cladding and free space
will be indicated by q ¼ q cos u^x þ q sin u^y. Vectors will be denoted by bold
characters and tensors by bold characters underlined by a double bar; k and f
identify free-space wavenumber and impedance, respectively. Transverse magnetic
(TM) and transverse electric (TE) modes will be referred to the normal ^z to the
surface. We will restrict our analysis to MTSs composed by a lossless grounded
dielectric slab of relative permittivity er and thickness h, on which perfectly electric
conducting subwavelength patches are printed. We will assume that the SW is
excited by an electric vertical dipole located at the center of the reference system. In
practice, the feeder shall be designed to maximize the amount of SW power and
minimize the amount of power directly radiated in the free-space. Therefore, to
maximize the SW efficiency, some strategies are adopted that can make the feeding
system more complex than a simple small vertical dipole. Despite this, the excited
SW can still be considered azimuthally symmetric, as the one of the vertical dipole,
and this is enough for giving validity to our simplified dipole assumption.

9.3.1 A Class of Synthesizable Continuous Reactances

We model the printed MTS by anisotropic “transparent” BCs [35, 39] identified by


Et ¼ jX
^z  Ht j0 þ Ht j0 ¼ jX
J ð9:1Þ

where X is the “transparent” reactance and J is the electric current flowing in the
metallic homogenized cladding. In practical realizations, the transparent reactance
298 G. Minatti et al.

mainly depends on the patch loading. In absence of losses, X is a Hermitian tensor

[29] (or the impedance jX is anti-Hermitian); the eigenvectors ^e1 ; ^e2 of X are
therefore orthogonal, and the eigenvalues are real.
The class of transparent lossless reactance tensors we would like to synthesize is
given by

X ¼ Xð0Þ þ Xð1Þ þ Xð þ 1Þ ð9:2Þ

q ^q^q þ X
Xð0Þ ¼ X u u
^u^ ð9:3Þ

1 
 
q ^q^q  X
Xð1Þ ¼ ejKs mq X u u
^u  q ð^
^ ejUq þ mu X qu^ þuqÞejUu
^^ ð9:4Þ
2

where Xq and X u are negative (capacitive) quantities and K is a large q-independent
 
constant such that K jrt sðqÞj  rt Uq;u ðqÞ. The above reactance is written in
terms of exponentials, which combined together lead to cosine functions. Therefore,
the above class is constituted by symmetric tensors, namely, a special case of
Hermitian tensors which are representative of MTSs constituted by elements with
two orthogonal symmetry axes. The above conditions identify reactances with
rapidly oscillating entries, whose average tensor possesses principal axes aligned
with ð^q; u
^ Þ. We denote this class as “constant-average MTSs” and X in (9.2) as
“constant-average reactance”. In (9.3), Xð0Þ denotes the q-independent average
reactance, whereas Xð1Þ in (9.4) are the rapidly oscillating contributions which
contain the modulation functions mu;q ðqÞ: Within the entries of Xð1Þ one can
identify three factors: (i) a rapidly varying phase factor expðjKsðqÞÞ, which
provides the

main interaction
 with the exciting SW; (ii) the slowly varying phase
factors exp jUq;u ðqÞ , which mainly control the polarization; (iii) the modulation
indexes mu;q ðqÞ, which mainly control the amplitude of the field. The latter are
considered small enough to avoid local changes of the nature of the transparent
reactance from capacitive to inductive, that could lead to the excitation of a higher
order (TE-dominant) SW mode. This is ensured by the condition mq;u ðqÞ\1:

9.3.2 Analysis by Adiabatic Floquet Waves

One of the main properties of the constant-average reactance X in (9.2) is con-

cerned with the fact that the average current flowing in it is quite similar to the SW-
currents that the dipole source would excite in Xð0Þ . An appropriate modification of
these currents will constitute therefore the 0-index mode of an adiabatic FW
expansion. Since Xð0Þ is q-independent, the TM-dominant SW excited on it by a
monopole is a purely TM cylindrical wave asymptotically given by
9 Metasurface Antennas 299

ð2Þ
J0 ¼ J0 H1 ðbsw qÞ^q ð9:5Þ

ð2Þ
where H1 is the Hankel function of second kind and first order. The propagation
constant bsw is solution of a dispersion equation which is affected by the X q
component only. An excellent closed-form approximation of bsw is given in [32].
In presence of the modulation, the 0-indexed FW mode (simply denoted as
0-mode) is obtained from (9.5) by locally transforming the real unperturbed
wavenumber bsw in the complex radial wavenumber kð0Þ ðqÞ ¼ bsw þ bD ðqÞ  jaðqÞ.
It is, therefore, implicitly assumed that the 0-mode has a cylindrical, attenuated
wavefront, where the local attenuation parameter aðqÞ accounts for the transfer of
energy from the 0-mode to the −1 (leaky) mode during the propagation path. The
parameters bD ðqÞ and aðqÞ are functions of the X entries. Since k ð0Þ ðqÞ is space
dependent, its associated global phase ~kð0Þ ðqÞq from the reference point at the origin is


obtained by integrating the local relation @ ~kð0Þ ðqÞq =@q ¼ k0 ðqÞ; i.e.,

Zq Zq
~ð0Þ 1 ð0Þ 0 1
k ð qÞ ¼ k dq ¼ bsw þ ½bD ðq0 Þ  jaðq0 Þdq0 ð9:6Þ
q q
0 0

This leads to the following global adiabatic FW expansion for the current
X
J JðnÞ ð9:7Þ
n
 
JðnÞ ¼ jðnÞ ejnKsðqÞ H1
ð2Þ ~kð0Þ q ð9:8Þ

where KsðqÞ is the same as in the definition of the reactances (9.4). Although the
adiabatic expansion contains an infinite number of terms, only three terms of the
expansion are sufficient to describe the total current, whenever only the −1 mode
belongs to the visible range. Therefore, we will retain only these three terms for a
ðnÞ
synthesis process. The complex terms jðnÞ ¼ Ju u ^ þ JqðnÞ ^
q in (9.8) denote the
slowly varying part of the n-th current mode and they are the unknowns of the
ð0Þ
problem. The 0-indexed mode has a dominant TM component Jq , whose ampli-
ð0Þ
tude is much larger than that of the TE component Ju , and also much larger than
both the components of the ð1Þ-indexed modes. This dominant behavior is more
evident for small modulation indexes. Indeed, in the limit for vanishing modulation,
ð0Þ
Jq should be the only component and should coincide with J0 in (9.5).
Taking the asymptotic form of the Hankel function (valid for distances
larger than about one wavelength from the source), it is apparent that each mode in
(9.8) has a curvilinear-wavefront, which propagates with local n-indexed FW
wavevector
300 G. Minatti et al.

h i
bðnÞ ¼ Rert ~kð0Þ q þ nKsðqÞ ¼ ðbsw þ bD Þ^
q þ nKrt sðqÞ ð9:9Þ

and with n-independent local attenuation parameter aðqÞ. The curvilinear-wavefront

associated with each FW current mode is given by bsw q þ nKsðqÞ ¼ const. The
adiabatic FW electric fields can be obtained through the spectral Green’s function
(GF) of the grounded slab evaluated at the local wavevector bðnÞ in (9.9), that is
X X
E t ð qÞ ¼ EðnÞ ZðnÞ
GF

Jð n Þ ð9:10Þ
n n

where ZðnÞ
GF
depends on the grounded slab spectral GF (see appendix in [36]). It
can be seen that ZðnÞ
GF
is purely imaginary for any n 6¼ 1 while it is complex for
n = −1. The −1 mode is actually the only mode of interest in order to design the
modulated impedance that provides a given radiation pattern.
The basis on which we have expanded currents and fields does not satisfy
Maxwell’s equations, since it is based on an asymptotic, adiabatic approximation
valid far from the point source. However, this basis locally recovers a Floquet wave
expansion and may be used for obtaining an adiabatic solution. This solution is
found by using (9.10) in the transparent boundary conditions (9.1).

X
2    
ZðnÞ
GF

J ðnÞ
¼ j Xð0Þ
þ X ð1Þ
þ X ð þ 1Þ

J ð1Þ
þ J ð0Þ
þ J ð1Þ
ð9:11Þ
n¼2

This allows for finding an analytical solution and a local q–dependent adiabatic
dispersion equation. To this end, the terms in (9.11) with the same rapid phase
variation are equalized and a set of equations is obtained. The latter is solved by
substitutions, thus leading to a homogeneous system of type jvðqÞ
Jð0Þ ¼ 0: The
h i
latter admits a non trivial solution only for det vðqÞ ¼ 0 which is the local dis-
persion equation; this equation allows for determining the complex value of the 0-
mode local wavenumber k ð0Þ . The final solution is therefore individuated in terms of
ð0Þ ð0Þ
the two TE/TM components of the 0-mode Jð0Þ (namely Ju and Jq ) still to be
determined. While their relative amplitude (i.e., their ratio) is found by using kð0Þ in
ð0Þ
the linear system jvðqÞ
Jð0Þ ¼ 0; the amplitude Jq can be found by some con-
siderations on power conservation [35, 36]. Indeed, within a zero-order approxi-
mation, the average current flowing in the constant-average reactance X can be
ð0Þ
identified with the SW-currents flowing in Xð0Þ in absence of modulation. Thus, Jq
9 Metasurface Antennas 301

is estimated by means of the radial Poynting vector associated with the surface
wave excited on the uniform MTS [35].

9.4 Continuous Reactance Synthesis

Using the theoretical framework introduced in the previous paragraph, it is possible

to set up the synthesis process of the BCs, namely the red block in Fig. 9.3. The aim
of the design is to determine X in (9.2) that creates a far field as the one radiated by
the objective circular aperture field. We describe the latter aperture field by the
rather general form
h i
EA ¼ E0 ejk‘ðqÞ eq ðqÞejcq ðqÞ ^q þ eu ðqÞejcu ðqÞ u
^ UA ð9:12Þ

where UA is a unit step function which is 1 inside a circle of radius a and zero
elsewhere. In (9.12), cq ðqÞ; cu ðqÞ; eq ðqÞ; eu ðqÞ; ‘ðqÞ are real, weakly variable
functions of space.
In order to determine X, we identify EA with Eð1Þ , namely, with the field
associated with the adiabatic leaky-wave field. In doing this, we implicitly assume
that the difference between the objective aperture field EA and the Et field in (9.10)
is a dominantly reactive field contribution, with spectrum mainly concentrated
outside the visible region. This assumption can be verified once the design pro-
cedure is completed and the aperture fields are assessed.
The determination of X involves the iterative process shown in Fig. 9.5 and
thoroughly described in [36]. The process starts by setting the aperture field, the
aperture size, substrate parameters, and operational frequency as input parameters.
An initial guess of the modulation parameters (namely KsðqÞ; mq;u ðqÞ; Uq;u ðqÞÞ is
found taking the 0-mode current and wavenumber as the ones in the uniform MTS
with average reactance Xð0Þ .
Next, this first initial guess of the MTS is analyzed to get the effective values of
the complex displacement bD ðqÞ  jaðqÞ on the modulated MTS. Such displace-
ment is then used to refine the estimation of the modulation parameters in the next
iterative step. When the modulation parameters vary less than a chosen threshold,
then the process is stopped and X is determined.

9.5 Full-Wave Homogenized Impedance Analysis

The continuous reactance obtained in Sect. 9.4 is verified at this step (see the block
diagram in Fig. 9.3) using a full-wave solver for planar apertures that supports a
space-dependent impedance BC. The synthesized apertures are typically large,
302 G. Minatti et al.

Setting the initial values

η k (0) = βsw X ϕ
EA Ks ( ρ )
j(0) = J 0 ρˆ
β sw Xρ

Iterative loop ⎡ mρ ,ϕ ⎤
⎢ ⎥
EA = E ( −1)
⎢Φ ρ ,ϕ ⎥
N ⎢ Ks ⎥
⎣ ⎦
J, X check
Y
Ks βΔ
Solve dispersion
j(0) α

Fig. 9.5 Block diagram for synthesizing the continuous reactance X from the aperture field EA .
The input parameters for setting the initial value are: aperture field, aperture size, substrate
parameter, and frequency

and an efficient analysis is thus necessary to complete the “mesoscopic” (homog-

enized impedance) block of the scheme in Fig. 9.3. We solve the problem of a
transparent impedance BC (9.2) imposed on top of a grounded dielectric slab of
infinite extent in the x-y plane. An equivalent problem can be described only by
horizontal electric currents at the air–dielectric interface. The imposed IBC leads to
the following integral equation
ZZ

0 0 0 0 0
^z  G ðq; q Þ
Jðq Þ dS  jXðq Þ
Jðq Þ
EJ
¼ ^z  Einc ð9:13Þ
S0

where ^q and q^0 are the source and observation points, respectively. G EJ represents the
electric dyadic Green’s function, Einc is the incident field due to the excitation and J is
the unknown surface current density. The motivation for carrying out the MoM
analysis with X instead of an opaque reactance is, on the one hand, to account for the
spectral dispersion due to the thickness of the grounded dielectric slab [31]. Resorting
to layered medium Green’s functions, one can introduce the effect of the grounded
slab in an efficient manner, since the unknowns in (9.13) will be just the electric
currents J: On the other hand, by using X along with the relevant Green’s functions it
is possible to avoid ill-conditioning issues that arise in the formulations based on
opaque reactances for typical surface reactance values in MTS structures [39].
It has been shown in [37] that an appropriate choice of entire domain basis
functions leads to an extremely fast solution of (9.13). Such basis functions,
hereinafter referred to as Gaussian ring basis functions (GRBFs) [37], constitute a
9 Metasurface Antennas 303

particularly convenient choice for representing J on circular apertures, and they can
be written in the space-domain as

1 ðqqm2 Þ2 qm q
f m;n ðq; uÞ ¼ ejnu Wm;n ðqÞ; Wm;n ðqÞ ¼ e 4r !n ð9:14Þ
2r2 2r2

where !n ðzÞ ¼ In ðzÞ ejRefzgj , with In ðzÞ being the modified Bessel function of the
first kind and order n. Equation (9.14) corresponds to a Gaussian-type ring with
linear azimuthal phase. GRBFs peak at qm and their half-maximum beamwidth
pffiffiffiffiffiffiffi
(HMBW) is 4 ln 2 r. Figure 9.6a shows the shape of f m;n ðq; 0Þ for different values
pffiffiffiffiffiffiffi
of r. One can also fix r to a constant value and choose qm ¼ q0 þ m 4 ln 2 r ðm ¼
0; 1; 2; 3. . .Þ for numbering the basis functions in a progressive sequence of GRBFs
that cross each other at half of their maximum (see Fig. 9.7a). Since the dependence
of Wm;n ðqÞ on n (Bessel function’s order) is extremely weak for this particular
choice of qm , changing n maintains the sequence of functions unchanged.
On the other hand, the spectral-domain counterpart of (9.14) presents an ana-
lytical form:

Fm;n ðkq Þ ¼ ejna 2pjn Um;n ðkq Þ; Um;n ðkq Þ ¼ er kq Jn ðqm kq Þ
2 2
ð9:15Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where kq ¼ kx ^x þ ky ^y ¼ kq cos a ^x þ kq sin a ^y, with kq ¼ kx2 þ ky2 and kx , ky
being the spectral variables of the classical Fourier domain. Equation (9.15) can be
used for obtaining closed-form expressions of the MoM impedance matrix ele-
ments. Examples of GRBFs spectra are shown in Figs. 9.6b and 9.7b.
The surface current density on the MTS plane can be expanded using the pro-
posed basis functions as

X
M 1 X
N=2
JðqÞ ¼ ixm;n f m;n
x
ðq Þ ^x þ iym;n f m;n
y
ðq Þ ^y ð9:16Þ
m¼0 n¼N=2

Fig. 9.6 Representation of the GRBF in the space-domain as a function of q=k0 (a) and in the
pffiffiffiffiffiffiffi
spectral-domain as a function of kq =k0 (b), for different values of 4 ln 2 r (qm =k0 ¼ 1; n ¼ 0Þ
304 G. Minatti et al.

Fig. 9.7 Representation of the GRBF in the space-domain as a function of q=k0 (a) and in the
pffiffiffiffiffiffiffi
spectral-domain as a function of kq =k0 (b), for qm ¼ q0 þ m 4 ln 2 r ðm ¼ 0; 1; 2; 3. . .Þ with q0 ¼
pffiffiffiffiffiffiffi
0:5k0 and m ¼ 0; 1; . . .; 4 (r ¼ 0:25k0 =ð4 ln 2Þ; n ¼ 0Þ

where M and N/2 are the number of basis functions along q and the higher order
azimuthal harmonic respectively. In (9.16), x and y-directed components are
required to account for the vector nature of the current.
After testing with the complex conjugate of the basis functions and arranging the
unknowns in the system of equations according to their direction, one arrives to an
algebraic system Z i ¼ v; where Z ¼ ½Zxx ; Zxy ; Zyx ; Zyy , with Zxx being an
   
MðN þ 1Þ  MðN þ 1Þ matrix. Similarly, i ¼ ix ; iy v ¼ vx ; vy with ix and vx
being MðN þ 1Þ  1 vectors. A generic entry of the Zxx sub-matrix can be written
as the sum of
Z Z
0 0 0
Gxx ¼ x;I
f m;n ðq Þ xx ðjq  q jÞ f m0 ;n0 ðq Þ dS dS
GEJ x
ð9:17Þ
S S0

and,
Z
v ¼
xx x;I
f m;n ðq ÞjXxx ðqÞf mx 0 ;n0 ðq ÞdS ð9:18Þ
S

Both quantities depend implicitly on ðm; n; m0 ; n0 Þ: In (9.17), (9.18), I

denotes
the complex conjugate and GEJ EJ
is the xx component of G . The term G involves
xx
xx

the relevant grounded slab Green’s function, whereas vxx is related to the impedance
boundary condition. It can be shown [37] that the integral in (9.17) can be written in
closed-form after combining the spectral-domain approach and rational function
fitting, whereas the use of an asymptotic expansion allows one to solve (9.18)
without resorting to any numerical integration. The other sub-matrices can be found
9 Metasurface Antennas 305

by using formal substitution of superscripts. Thus, the proposed basis functions

allow one to write the MoM impedance matrix entries in a closed-form.
The typical structure of the 2MðN þ 1Þ  2MðN þ 1Þ MoM impedance matrix is
symbolically shown in Fig. 9.8. The dashed lines in Fig. 9.8 limit the ðN þ 1Þ 
ðN þ 1Þ block for m ¼ m0 ¼ 0 and n; n0 2 ½N=2; N=2. The matrix with the
Green’s function contribution, shown in Fig. 9.8a, is composed by square tri-
diagonal blocks of dimension ðN þ 1Þ  ðN þ 1Þ; where all the elements with
jn  n0 j [ 2 are equal to 0. The matrix containing the IBC contribution is shown in
Fig. 9.8b. It is important to notice that here the color-bar is amplified by a factor of
10 with respect to Fig. 9.8a. Therefore, it is obvious that the IBC contribution ðvxx Þ
dominates that of the Green’s function (Gxx ). This is a general feature and suggests
the use of IBCxx as preconditioner for the iterative solution of very large problems
(larger than 30 k0 ). The matrix IBC is dominantly block-diagonal, i.e., only the few
off-diagonal blocks with jm  m0 j
2 are significant. This property, that derives
from the fact that vxx is nonvanishing only for overlapping GRBFs, greatly helps the
design phase when the profile of X should be optimized by successive iterations. It
is also important to notice that the contribution Gxx is independent of the IBC and
should be calculated only once in an optimization procedure. We have shown how
the introduction of the MTS as a transparent impedance BC in the integral equation,
and the closed-form spectra of the GRBFs enables an efficient computation of the
MoM impedance matrix using the spectral-domain approach and an asymptotic
expansion. More importantly, these basis functions represent the global evolution of
the surface current density in an effective manner, which results in a significant
reduction of the number of unknowns, if compared with sub-entire domain basis
functions, like Rao–Wilton–Glisson functions defined on triangular domains.

Fig. 9.8 Magnitude of the MoM impedance matrix elements for a typical structure with M = 48
and N = 16. For simplicity, the first 51 rows and columns are shown for: the matrix with the
Green’s function contribution (a) and the matrix with IBC contribution (b)
306 G. Minatti et al.

Computational cost of MoM IBC

35
slab
MTS
Approximate computation time (s)

Z matrix
30 solve
J
E
25 Total
O(n+n 2/6)

20

15

10

0
5 10 15 20
Antenna diameter ( )

Fig. 9.9 Computational time as a function of the antenna diameter evaluated on a Quad-core
laptop with 8 GB RAM. The overall computational time has a rather mild quadratic dependence
with the size (red dot-dashed line)

The fast analysis framework obtained with this method is particularly important
when used in an optimization procedure of the impedance profile. Figure 9.9 shows
the timing as function of the antenna diameter evaluated on a Quad-core i7 with
8 GB RAM together with a fitting curve that shows a rather mild quadratic
dependence with size. Antennas with diameter of 20 k can be solved in about 35 s.

9.6 Pixel Modeling

At this point of the design process, the continuous reactance X that produces the
objective aperture field is completely known all over the aperture domain. The next
design step consists in sampling it and implementing it by a dense texture of
electrically small patches. Typically, the reactance is distributed on a Cartesian
lattice with square unit cells, whose dimensions range, as a rule of thumb, from k/
10 to k/6. However, in principle, other lattices with non-square cells can be con-
veniently exploited. To emphasize that patches are small in terms of wavelength,
sometimes we denote them and the cell that they occupy as “pixels” in the global
aperture picture.
The periodicity of the lattice is always taken as a constant on the aperture and the
variation of impedance is achieved by changing the geometrical parameters of the
elements. To produce anisotropic impedances, the geometry of the element should
have additional features in order to change the electromagnetic properties for
propagation along different axes. Different geometrical shapes can be used.
9 Metasurface Antennas 307

(a) (b) (c) (d)

ψ ψ ψ ψ

a'
a a
a
a

Fig. 9.10 Patch geometries for anisotropic impedance surface. Each geometry possesses two
specific nondimensional parameters a/a’ and w that are considered for constructing the reactance
database

The solutions of Fig. 9.10b and Fig. 9.10d are the ones suggested in [7, 8],
respectively. The elliptical shape in Fig. 9.10c can be analyzed with the
quasi-analytical method proposed in [30]. All the investigated solutions exhibit two
nondimensional parameters a/a’ and w. Retrieval of the reactance values from the
geometrical parameters is performed by resorting to a local periodicity concept; i.e.,
by periodic boundary conditions applied to a unit cell with parameters a=a0 and w.
This allows one to construct a database, which is next interpolated defining two
continuous functions of the parameters a=a0 and w. The values required in (9.2) are
obtained pixel by pixel by a best fit process. This process is typically very fast after
the database construction.
Figures 9.11 and 9.12 show two examples of databases, namely the impedance
maps, relevant to the patch geometry of Fig. 9.10d and Fig. 9.10c, respectively,
obtained through a periodic full-wave MoM solver. All of the examples are eval-
uated for a periodic square cell with side a k/13 on a dielectric slab with er = 13
and thickness k/23. The maps show the components of the opaque reactance tensor
Xqq ; Xqu ; Xuu normalized to an impedance X s ¼ 300 X (for the Hermitian property
of the reactance tensor it results Xqu ¼ Xuq ).
The database can be visualized in two different ways. The first one is presented
in Fig. 9.11a, c, and e for slotted circular patches and in Fig. 9.11b, d and f for
elliptical patches. The second one is presented in Fig. 9.12 and consists on
isofrequency dispersion curves (IDCs) which represent the locus of the trajectory of
the wavevectors end-point when changing the direction of propagation.
Although a systematic way to build the database is given by full-wave analysis
[30, 38], the alternative approach presented in [33] can be used. It allows for
constructing extensive databases for various shapes, minimizing the number of
parameters to be stored. The formulation is restricted to elements with at least two
symmetry axes and is valid for the dominant TM SW until the limit of the Floquet-
Bloch region [32], which is the region where the higher order Floquet modes
become significant. This generalization permits a closed-form representation of the
isofrequency dispersion curves as a function of two parameters only; these are the
equivalent quasi-static capacitances along the symmetry directions of the geometry.
308 G. Minatti et al.

Fig. 9.11 Example of impedance maps for the patch geometry of Fig. 9.10d ((a) Xqq, (c) Xqu, (e)
Xuu) and for the patch geometry of Fig. 9.10c. ((b) Xqq, (d) Xqu, (f) Xuu). All the maps (except for
s ¼ 300 X and relevant to a cell size of k/13 on a dielectric with
(c) and (d)) are normalized to X
thickness k/23 and er = 13

Once the latter are determined [32], the IDCs can be approximated by using the
“two-circle approximation” [33].
Here, we provide some examples to show the accuracy of the analytical for-
mulation in [33], considering various shapes of the printed elements: elliptical
patch, rectangular patch, elliptical ring, and circular patch with rectangular slot.
9 Metasurface Antennas 309

(a) (b)
d d
h h

(c) (d)
d d

h h

(e) (f)
d
h d
h

(g) (h)
d
d
h
h
kyd

Fig. 9.12 IDCs for several patch geometries, solid lines are relevant to MoM results, dashed lines
are relevant to the analytical formulation, curve labels are in GHz. Elliptical patch with h = d/3 (a)
and with h = d (c). Elliptical ring patch with h = d/3 (b) and with h = d (d). Rectangular patch
with h = d/3 (e) and with h = d (g). Slotted circular patch with h = d/3 (f) and with h = d (h)

The basic periodic cell is always square and with period d ¼ 3 mm: The substrate
has relative permittivity er ¼ 9:8 and two different thicknesses have been consid-
ered (h ¼ 1 mm and h ¼ 3 mmÞ for each configuration. Figure 9.12 shows
the isofrequency dispersion diagrams for the above-mentioned patch geometries.
310 G. Minatti et al.

The two-parameters-dependent analytical formulation (dashed lines) is compared

with the one obtained from a full-wave analysis (solid lines). The agreement is good
up to the Floquet-Bloch frequency, as expected.

9.7 Fast Integral Equation Method for Metasurface

Antennas

The global full-wave analysis of the textured layout of the MTS is the last step of
the design, which serves as “virtual prototyping” of the antenna structure. This is
not a straightforward issue, as the final layout is composed of thousands of elec-
trically small patches with different shapes, thus yielding an ill-conditioned linear
system, consisting of hundreds of thousands or even several millions of unknowns.
Clearly, the solution of this problem cannot be afforded with conventional
techniques.
The use of suitable patch geometries to texture the MTS allows for representing
currents on them with entire domain basis functions, thus greatly reducing the
number of unknowns of the problem. For instance, currents on the elliptical patches
in Fig. 9.13 can be effectively represented using the entire domain basis functions
defined in [30]. We have also found that two entire domain basis functions per
patch are usually sufficient to describe currents with the same accuracy obtained
with more than 100 RWG basis functions. The MoM procedure can be furthermore

Fig. 9.13 MTS antenna realized by texturing of elliptical patches. The current on each patch is
represented by two entire domain basis functions
9 Metasurface Antennas 311

engineered using an algorithm based on the fast multiple method (FMM) to

accelerate the solution.
To set up the simulation, the metallic patches are replaced by unknown currents.
The MoM is then applied to convert the electric field integral equation (EFIE) into a
linear algebraic system of equations, expanding the unknown current distribution in
a set of basis functions. A suitable couple of entire domain basis functions, able to
represent the currents on elliptical patches, have been proposed in [30] where both
spatial and spectral expressions are reported.
The spatial behavior of the two functions is illustrated in the inset of Fig. 9.13.
Applying the Galerkin procedure, the standard EFIE-MoM linear system assumes
the matrix form Z
I ¼ V where the computation of the elements of V depends on
the nature of the source. As in the classical FMM [40, 41], the computational
efficiency is increased by representing the MoM interaction matrix Z as superpo-
sition of a near interaction matrix Znear and a far interaction matrix Zfar . The 2D-
FMM version reported in [40, 41] is used to accelerate the matrix-vector product
relevant to Zfar , while the near interaction matrix is processed using the standard
MoM. In particular, the Znear entries are calculated by using the spectral method of
moment, implementing the efficient procedure in [42].

9.8 Examples of Antenna Design

In this paragraph, we recap the design process while showing some examples of
modulated MTS antenna design. A wide range of aperture field distributions can be
realized by properly shaping the MTS, spanning from the ones radiating a highly
directive beam to the ones producing a shaped beam. In [43] a modulated MTS
antenna has been designed and analyzed, which is capable to handle a broadside
dual polarization. Also, here we will show that it is possible to design a MTS
antenna with an operational bandwidth greatly wider than other periodic leaky-
wave antennas.

9.8.1 Shaped Beam Antenna

As first example, we discuss the steps relevant to the design of a circularly polarized
isoflux pattern antenna. Such shaped beam is of interest for data transmission
antennas on satellite platforms for Earth observation missions. The shaped beam is
conceived to provide a uniform power-flux density (isoflux) over a well-defined
portion of the visible Earth surface, compensating for the rather large differential
path loss between Nadir and grazing incidence on the Earth. The Earth coverage is
normally obtained with conical isoflux beam [8]. To achieve higher gains as
required to transmit higher data rates, the same overall coverage of the axially
312 G. Minatti et al.

symmetric isoflux antenna can be obtained by a sector isoflux beam. An azimuthal

rotation is needed to keep the beam pointing toward the ground station while the
satellite moves along the orbit. For this example, the operational frequency is
chosen at 26.25 GHz inside the Ka-band and the substrate thickness is 0.5 mm,
with relative permittivity of 9.8. The size of the aperture is chosen arbitrarily as 9 k
and it allows to convert 90% of the input SW power into LW power. The target
aperture distribution can be found through standard techniques, as stated in
Sect. 9.2. For this example, we have extended to polarized aperture fields the
method presented in [44]. Figure 9.14 shows phases and amplitudes of the objec-
tive aperture field distribution for both the radial component, ((a) and (b) respec-
tively) and the azimuthal component ((c) and (d)). Once the input parameters for the
continuous reactance synthesis block in Fig. 9.3 have been set, the impedance
surface is found through the iterative process described in Sect. 9.3, assuming the
impedance surface fed by a small vertical electric dipole located at the center of the

Fig. 9.14 Objective aperture field distributions for the isoflux-shaped beam antenna. Amplitude
and phase of the radial component (a, b), and amplitude and phase of the azimuthal component
(c, d)
9 Metasurface Antennas 313

Fig. 9.15 Impedance maps of X ¼ ^q^qXqq þ ð^qu

^ þu
^ ^qÞXqu þ u
^u^ Xuu resulting from the synthe-
sis process. a Xqq, b Xqu, c Xuq, and d Xuu

surface. The entries of X for the concerned example, as a function of the position,
are shown in Fig. 9.15. The impedance surface is analyzed through the continuous
impedance boundary condition full-wave solver (IBC-MoM) [35].
If the design is unsatisfactory, which is something that rarely happens and only
for very demanding requirements (for instance a very high efficiency), then the
synthesis of X is repeated changing some of the design parameters (for instance, for
a highly efficient beam one could act on the average impedance or on the aperture
size).
When the agreement between the target far-field and the one radiated by the
continuous impedance is judged satisfactory, the metasurface is textured with
electrically small patches. In this example, we have chosen the elliptical patch
geometry as the one depicted in Fig. 9.10c, so that we can use the entire domain
basis functions described in [30] to create the database maps and to analyze the final
layout with the global full-wave solver. To create the database, the parameters of
314 G. Minatti et al.

Fig. 9.16 Final layout of the antenna analyzed by the FMM-MoM: global picture and a detail of
the patch texture close to the center of the antenna

the elliptic geometry are changed and the impedance associated with the geometry
is retrieved by using a full-wave solver in which the concerned patch geometry is
embedded into a periodic environment. The database allows for defining a function
that maps a given patch geometry into the relevant tensorial impedance.
The final layout is shown in Fig. 9.16 and it is composed of elliptical patches
characterized by geometrical parameters gradually variable with the position on the
surface. The final structure is composed of 8936 patches and it is next analyzed
using a global full-wave solver. The current on each patch is described by 2 entire
domain basis functions and the MoM procedure is accelerated using a fast multipole
algorithm. Figure 9.17 shows a comparison among the three methods described
here for the normalized radiation diagrams: the Flat Optics solution (Sect. 9.3) the
IBC-MoM solution (Sect. 9.5), which analyzes the continuous impedance BCs, the
FMM-MoM solution (Sect. 9.7) which analyzes the textured layout of the antenna.
Figure 9.18 shows the directive pattern and the spectral u-v plane coming from the
IBC-MoM analysis.

9.8.2 High-Efficiency Antenna

Figure 9.19 shows the directive pattern of a broadside MTS antenna having a radius
of 8 wavelengths whose impedance surface has been designed to produce an
aperture field with very high tapering efficiency. The directivity predicted by MoM
code solving the continuous impedance surface is around 33.3 dBi, corresponding
to a tapering efficiency of about 85%. The tapering efficiency of the aperture
distribution on the MTS, broadly speaking, is mainly controlled by tapering the
modulation indexes which in turn control the leakage rate.
9 Metasurface Antennas 315

FMA-MoM vs IBC-MoM vs Flat Optics

30 z
θ
FO
MoM FMA
x y MoM IBC
20 Target
Directivity (dBi)

15

10

-5

-10
-90 -60 -30 0 30 60 90
Elevation (deg)

Fig. 9.17 Comparison between the copolar (RHCP) directivity diagrams. Green-dashed line is
produced by the objective aperture distribution in Fig. 9.14. Blue solid line is the directivity
diagram produced by the Flat Optics (FO) [35, 36] currents introduced in Sect. 9.3. Red solid line
is resulting from the IBC-MoM analysis of the synthesized continuous impedance surface. Black
solid line is the directive pattern resulting from the FMM-MoM analysis of the textured layout of
the antenna

Fig. 9.18 Far field diagrams in the spectral u-v plane coming from the IBC-MoM analysis of the
synthesized impedance surface, LHCP (a) and RHCP components (b)

The inset of Fig. 9.20 shows the Xqq component of the modulated impedance
surface, whilst the main frame shows the amplitude along the x-axis of the
z-directed component of the Poynting vector, the modulation index mq and the
attenuation parameter a. It is seen that mq (which is the same as mu for this
example) increases with x, thus increasing the leakage rate a and therefore keeping
316 G. Minatti et al.

Fig. 9.19 Directive pattern from a MTS designed for a high tapering efficiency (85%). The results
from the IBC-MoM (red lines) are compared with the target directivity patterns (black lines). Solid
lines are relevant to RHCP components, dotted-dashed lines to the LHCP components. The insets
show the radiation diagrams in the u-v spectral plane from the IBC-MoM

100 0.5
m
90
(right scale)
80 0.4

70

60 0.3

50
2
S [W/m ]
40 (left scale) 0.2

30 -1
[m ]
20 (left scale) 0.1

10

0 0
0 1 2 3 4 5 6 7 8
x/

Fig. 9.20 The inset shows the Xqq component of the modulated impedance for the highly efficient
beam antenna. The main frame reports the following quantities taken along the x-cut of the MTS:
black line, amplitude of the Poynting vector component S along the normal to the surface
(left y-axis), red line, attenuation parameter of the SW on the MTS (left y-axis), modulation index
(q-component) of the impedance surface
9 Metasurface Antennas 317

uniform the amplitude of the aperture distribution (described by S) up to a distance

of 6.5 k from the center of the surface.

9.8.3 Wide Band Antenna

Despite the fact that the LW radiation typically suffers of beam squinting with the
frequency, a proper shaping of the BCs allows to overcome this limitation. The
study for the bandwidth enlargement of modulated MTS antennas is, however,
currently in progress and here we briefly sketch the concept of wideband modulated
MTS antenna supporting it with a numerical example of a wideband, circularly
polarized, broadside antenna. To widen the operational bandwidth of the MTS
antenna we operate on the local periodicity of the impedance modulation, imple-
menting a transition of the period’s length across the aperture according to the
operational frequencies. Such transition is a device to trade the available space with
the operative bandwidth in a similar way as in spirals and log-periodic antennas. At
the same time, we operate on the amplitude of the modulation to keep constant the
peak directivity as the operational frequency change, namely reducing the effective
area of the surface as the wavelength decreases. Figure 9.21 reports the frequency
behavior of a synthesized broadside pointing, broadband aperture resulting from the
IBC-MoM code in [37]. Figure 9.21 also shows some samples of directivity pat-
terns at different frequencies over the wide band. The antenna has a radius of
16.7 cm on a substrate with er = 6.15, thickness 0.635 mm. The low aperture
efficiency is a trade off with the resulting wide bandwidth of about 50%.

9.9 Efficiency of Metasurface Antennas

Till now, we have discussed about the design process, starting from the definition of
the continuous impedance BCs allowing to reproduce a desired aperture field. We
have shown how the continuous impedance BCs are analyzed and implemented,
and finally how the textured layout is verified through a global full-wave analysis.
The effectiveness of the procedure has been discussed through numerical examples,
and will be proved next by the experimental results. However, before going through
the practical realization of MTS antennas, we spend a few comments on the effi-
ciency of these antennas. The definition of the efficiency for MTS antennas involves
several aspects related to the radiation mechanism [45]. When the input port of the
feeder is fed by a power Pin, part of it is directly radiated in free space (Pfeed), while
the remaining part is delivered as SW (Psw). The SW power is partly lost due to
losses (PX), partly radiated as LW (Plw) and the remaining part reaches the
318 G. Minatti et al.

(a) (b) (c)

(d) Frequency (GHz)

18 20 22 24 26 28 30
30

25 Upper scale Realized peak directivity

20
Directivity (dBi)

15
29.88 GHz LHCP
23.94 GHz LHCP
10 18.33 GHz LHCP
29.88 GHz RHCP
23.94 GHz RHCP
18.33 GHz RHCP
5

-5

-10
-90 -60 -30 0 30 60 90
Elevation (deg)

Fig. 9.21 Directive patterns for a broadband modulated MTS antenna obtained from a continuous
impedance boundary conditions (IBC) MoM solver [35]. RHCP patterns are shown in the u-
v spectral plane in at 18.33 GHz (a), at 23.94 GHz (b) and at 29.88 GHz (c). The relevant cuts
along the v = 0 spectral line are in (d). In the same frame, it is reported the peak directivity (top
scale) which has a good stability within the Ka-band in the range 18–30 GHz

aperture rim giving rise to edge effects. All the efficiencies relevant to the intro-
duced wave mechanism can be characterized more in detail as explained next.

9.9.1 Feed Efficiency

The feed efficiency efeed is defined as the ratio between the input power and the
power delivered to the SW, i.e.,
9 Metasurface Antennas 319

efeed ¼ Psw =Pin ¼ 1  Pfeed =Pin ð9:19Þ

Ideally, the feeder of a MTS antenna delivers all the input power to the SW, thus
minimizing Pfeed. The problem of the optimum excitation of surface waves for this
kind of antennas is studied in details in [45] where it is shown that high values of
efeed (around 90%) can be reached with simple SW launcher as a coaxially fed
monopole on a circular patch, provided the size of the patch, the substrate and the
average opaque impedance are properly chosen. More complex feeding solution
can be sought for obtaining higher values of efeed, but the discussion on this topic is
beyond the scope of this book and the reader is referred to [45–53] for further
insights.

9.9.2 Ohmic Efficiency

The ohmic efficiency eX is defined assuming lossless metallic cladding and ground
plane, and a lossy substrate. This is reasonable in the microwave regime where the
dielectric losses dominate on the metal losses. The ohmic efficiency is defined as

Plw =Psw
eX ¼ ð9:20Þ
Plw jPX ¼0 =Psw

where Plw jPX ¼0 is the LW power radiated in absence of losses.

9.9.3 Conversion and Tapering Efficiency

Conversion efficiency econv represents the fraction of radiated LW power with

respect to the SW power. In absence of losses, i.e., when PX ¼ 0; the conversion
efficiency is
.
econv ¼ Plw jPX ¼0 Psw ð9:21Þ

The tapering efficiency etap is related to the directivity loss of a given aperture
illumination with respect to a uniform distribution. For broadside beams, it is
quantified as
R R 2
 
A Et dA
etap ¼ RR 2
ð9:22Þ
A A jEt j dA
320 G. Minatti et al.

1
0.9 ε tap
0.8
ε conv
0.7
0.6
0.5 ε conv ε tap
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3
α0a

Fig. 9.22 Tapering and conversion efficiencies and their product as a function of a0a for
modulation with uniform amplitude

The quantities in (9.21) and (9.22) are controlled by the attenuation parameter a,
which, in turn, is controlled by the amplitude of the modulation [36]. When the
amplitude of the modulation is uniform over the surface, the leakage parameter has
a uniform value a0, which depends on the depth of the modulation. Figure 9.22
shows the values assumed by econv and etap as a function of a0a. Since the modu-
lation is uniform, there is no control on the aperture distribution: the tapering
efficiency is high only for small a0a, where the conversion efficiency is small, and
hence the radiated power is a little amount. The peak of efficiency is reached around
a0a = 0.9, where the product etapeconv is 58%. This limit can be overcome by
shaping a, recurring to a nonuniform amplitude modulation, thus making the
aperture distribution more efficient and at the same time increasing the amount of
radiated power (Fig. 9.19).

9.9.4 Overall Efficiency

Figure 9.23 shows a plot of the overall efficiency for broadside, circularly polar-
ized, MTS antennas with different sizes. The overall efficiency has been obtained
and as the product of all the efficiencies introduced in the previous section. We have
considered several lossy substrates and several values of opaque impedances that
can be realized easily by a printed patch technology. It is seen that the overall
efficiency can easily reach 75% by designs that make use of standard substrates, a
nonuniform modulation and a simple feeder consisting of a vertical electric dipole
with a metallic disk printed on the MTS (see top-left inset of Fig. 9.23). The values
9 Metasurface Antennas 321

80

75

70
%

65
Antenna design parameters
Dielectric permittivity
60 Solid lines : 9.8
Dashed lines: 4.5
Average surface impedance
55 Black lines: 0.75∙ζ
Blue lines : ζ
Red lines: 1.25∙ ζ
50
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Fig. 9.23 Overall efficiency for MTS antennas of several sizes, radiating a broadside, circularly
polarized beam. The substrate has a thickness corresponding to a phase change of 0.2 radians and a
loss tangent of 0.002. The curves are obtained for several substrates and several values of the
opaque impedance

shown in Fig. 9.23 are to be read as typical minimum overall efficiency figures for a
good design. They can be exceeded with a more refined design and under specific
conditions. On the other side, additional requirements on the antenna pattern may
adversely affect the efficiency, like for any other antenna, e.g., a need to further
reduce the side lobes level or the cross-polar radiation.

9.10 Practical Realization

Several prototypes have been built and measured during the recent years to prove
the effectiveness of the concept of modulated MTS antennas. In this section we
recap some of them. The earliest prototypes are all based on a modulation of the
opaque reactance instead of the transparent reactance. On the other hand, the most
recent ones, designed according to the theory presented in Sect. 9.4, are based on
the modulation of the transparent reactance and they are currently under manu-
facturing or measurement.
Figure 9.24 shows the picture of a MTS antenna realized according to the theory
presented in Sect. 9.4 for the synthesis of the transparent reactance. The substrate
on which the patch texture has been printed, has a diameter of 24 k at the opera-
tional frequency of 20 GHz, with er = 6.15 and thickness 0.762 mm. The patch
322 G. Minatti et al.

Fig. 9.24 Picture of the high gain modulated metasurface antenna prototype. The modulation of
the transparent reactance is realized by a texture of elliptic patches on surface with diameter 24k at
the operational frequency of 20 GHz. The feeding system is embedded on the substrate (er = 6.15,
thickness 0.762 mm) and is constituted by a small vertical dipole, top loaded by a slotted circular
patch

texture is composed of elliptical patches and it is shown more in detail in the right
side of Fig. 9.24 together with the feeding system. This latter is a coaxially fed
monopole, top-loaded with a patch slotted with an annular ring. Close to the feeding
region, the patch sizes are progressively reduced to improve the input matching of
the SW with the modulated impedance surface. The antenna radiates a circularly
polarized directive beam with 35.5 dBi of gain tilted at 30° with respect to the
normal to the surface. This prototype is still under measurement, and currently we
can only provide the numerical prediction of the performance. Figure 9.25 com-
pares the results from the flat optics theory (FO) of Sect. 9.3, with the ones coming
from the continuous impedance MoM of Sect. 9.5 (IBC-MoM) and from the
full-wave solver based on the fast multiple method (Sect. 9.7) of the textured layout
(FMM-MoM). The three methods are in good agreement, except for a slight
depointing of the peak resulting from the numerical simulation of the textured
surface (inset of Fig. 9.25).
All the prototypes of modulated MTS antennas presented from here on are
designed by shaping the opaque impedance rather than the transparent one. They
represent the first realizations of such antennas and they were designed according to
the theory described in [6]. One of the earliest prototypes of modulated MTS
antenna is shown in Fig. 9.26. It is realized on a substrate with er = 3.66, thickness
1.524 mm. The panel radius is 10.1 cm, corresponding to 5.7 k at the operational
frequency of 17 GHz. The MTS is formed by a dense texture of small square
9 Metasurface Antennas 323

36
High Gain MTS antenna - 20 GHz
40 35.5

35 RHCP MoM IBC

35 LHCP MoM IBC
34.5
LHCP FMM IBC
dBi

30 34
RHCP FMM IBC
33.5 LHCP Flat optics
25
33

20 32.5
dBi

32
15 28 28.5 29 29.5 30 30.5 31 31.5 32
(deg)

10

-5
-90 -60 -30 0 30 60 90
(deg)

Fig. 9.25 Numerical prediction of the directivity of the antenna presented in Fig. 9.24. Solid lines
are relevant to the copolar RHCP component, dot-dashed lines are relevant to the LHCP cross-
polar components. Blue lines: pattern resulting from the IBC-MoM analysis of the continuous
reactance. Red lines: pattern resulting from the FMM-MoM analysis of the textured surface. Black
line, directivity diagram pattern resulting from the Flat-Optics model of the continuous impedance
surface

Fig. 9.26 Prototype of modulated MTS antenna realized on a substrate with er = 3.66, thickness
1.524 mm and radius 10.1 cm. The MTS is formed by a dense texture of small square patches
arranged in a spiral shape, resulting in an opaque isotropic impedance with sinusoidal modulation

patches arranged in a spiral shape, resulting in an opaque impedance with sinu-

soidal modulation. As the patches do not have geometrical asymmetries, the
resulting modulation is isotropic, meaning that the TM nature of the SW is not
significantly modified while propagating on the MTS. This latter has a radial
324 G. Minatti et al.

(a)
25
LHCP RHCP
10.1 cm
Peak Gains:
20 22.4 dBi @ 16.50 GHz
25.0 dBi @ 17.00 GHz
24.5 dBi @ 17.25 GHz
15

10 17.25 GHz
17.00 GHz
dBi

17.25 GHz 17.00 GHz

5
16.50 GHz 16.50 GHz

-5

-10
-90 -60 -30 0 30 60 -60 -30 0 30 60 90
deg

(b) 0

-5

-10

-15
dB

-20

-25

-30
|S11|
-35
12 13 14 15 16 17 18 19 20
GHz

Fig. 9.27 Gain measurement for the modulated spiral MTS antenna. LCHP and RHCP
components (a). The antenna (top-left inset in (a) is designed at 17 GHz on a panel with radius
10.1 cm. The measured reflection coefficient is in (b) with a detail of the feeding vertical monopole
(bottom-left inset in (b))

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
periodicity pr ¼ k= 1 þ ðX op =fÞ2 synchronized with the wavelength of the SW on
the average opaque impedance X op . Intuitively speaking, the spiral shape of the
modulation makes any two orthogonal sectors radiating at the broadside with phase
quadrature. Figure 9.27a shows the measured gain patterns at several frequencies
around the designed working frequency of 17 GHz, respectively for the copolar
(RHCP) and the cross (LHCP) components. As the implemented impedance surface
is isotropic, the cross-polar levels are not controlled, differently from the prototype
of Figs. 9.24 and 9.25. The antenna is fed through a small vertical monopole,
9 Metasurface Antennas 325

Fig. 9.28 Picture of the proof-of-concept MTS antenna radiating a CP isoflux beam realized for
Earth observation missions on LEO satellites (a). Detail of the feeding system: the circular patch
has the twofold role of adjusting the broadside gain and launching the SW on the modulated MTS
(b). Picture of the feeding circuitry responsible for sequentially exciting four small vertical
monopoles with initial phases of 0°, 90°, 180°, and 270° (c)

coaxially fed, and placed at the center of the device. The measured reflection
coefficient is presented in Fig. 9.27b together with a detail of the feed in the bottom-
left inset. Further details can be found in [6, 13].
A proof-of-concept MTS antenna radiating a CP isoflux beam realized for Earth
observation missions on LEO satellites is shown in Fig. 9.28a. The antenna on the
satellite platform shall radiate a homogenous power density toward the Earth sur-
face, increasing the gain toward the angular regions where the path is longer, to
compensate for the power loss. The realized antenna works in the X-band and the
main frame of Fig. 9.29 shows the measured gain patterns both for LHCP and
RHCP components around 8.6 GHz. The MTS is realized on a dielectric panel
having radius of 27 cm with er = 9.8, and thickness 1.575 mm and composed of
about 12000 patches having a circular shape with a small cut in it. The small cut
within the patch shape allows to realize an anisotropic impedance surface, therefore
allowing for a full control of the aperture field polarization. The patch texture
modulates the opaque impedance sinusoidally along the radial direction, without
variation along the azimuthal direction. The radial periodicity of the modulation is
such that it produces a conical radiation with peak at 60° off-axis, while the feeder
(Fig. 9.28b) has the twofold role of adjusting the broadside gain and providing the
SW on the modulated metasurface. The SW is excited with a rotational phase
326 G. Minatti et al.

(a)
12

9
27 cm
6 8.5 GHz
8.6 GHz
3 8.7 GHz
0

-3
dBi

-6

-9

-12

-15
Solid line: RHCP
-18 Dashed line: LHCP

-21
-180 -150 -120 -90 -60 -30 0 30 60 90 120 150 180
deg

(b) -5

-10

-15
dB

-20

-25

|S11|
-30
8.0 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 9.0
GHz

Fig. 9.29 Measured gain (RHCP vs. LHCP) of the isoflux metasurface antenna prototype (a) and
measured reflection coefficient (b)

distribution, by four small vertical monopoles, sequentially fed with initial phases
of 0°, 90°, 180° and 270° (Fig. 9.28c). The input reflection coefficient is presented
in Fig. 9.29b together with a picture of the input feeding system in the bottom-right
inset of the same figure. Further details about design, analysis, and realization can
be found in [6, 8].
One of the most interesting feature of these antennas is that just by changing the
impedance boundary conditions provided by the MTS, completely different per-
formances can be obtained. This key feature of modulated MTS antenna allows a
unique support for late electrical performance refinement and design re-usability
which is very appealing for space applications. Figure 9.30 shows two broadside
circularly polarized MTS antennas designed on the same dielectric panel used for
9 Metasurface Antennas 327

Fig. 9.30 Photos of the two realized prototypes. Left side Tx antenna. Right side Rx antenna,
Details of the feeders and of the printed elements are shown in the insets

the isoflux antenna prototype. The two antennas can be set up on the same platform
to get a TX-RX system with low coupling at the two ports. They are proof-of-
concept devices in the X-band designed using typical requirements for deep space
science and automated exploration missions.
The main frames of Fig. 9.31a and b present the peak-normalized radiation
patterns relevant, respectively, to the Rx antenna and the Tx antenna. The Rx
antenna exhibits a Gain larger than 21 dBi over a bandwidth of 4%, the Tx antenna
has gain larger than 27 dBi over a bandwidth of 2.4%. It is to notice that, even
without operating on the modulation depth profile, efficiencies of about 35% are
obtained. Also, measurements prove good polarization purity at broadside being the
CX component close to −15 dB below the copolar component level. The measured
reflection coefficients is reported in Fig. 9.32. Further details can be found in [9].
The results shown in this section confirm the attractiveness of modulated MTS
antennas, which preserve all its simplicity and handiness with different far field
performances. All the reported prototypes were realized using standard PCB pro-
cess, ensuring cost-effectiveness together with good performance of gain and
polarization.
328 G. Minatti et al.

(a)
0
LHCP RHCP
15.2 cm Peak Gains:
-5
21.8 dBi @ 7140 MHz
21.8 dBi @ 7165 MHz
-10 21.9 dBi @ 7190 MHz
7140 MHz
7165 MHz
-15

7190 MHz
dB

7140 MHz
7165 MHz
-25
7190 MHz

-30

-35

-40
-90 -60 -30 0 30 60 deg -60 -30 0 30 60 90

(b)
0
LHCP RHCP
27 cm Peak Gains:
-5
27.7 dBi @ 8375 MHz
28.3 dBi @ 8425 MHz
-10 28.3 dBi @ 8475 MHz

-15
dB

8475 MHz
8475 MHz
8425 MHz
-25 8425 MHz
8375 MHz
8375 MHz
-30

-35

-40
-90 -60 -30 0 30 60 deg -60 -30 0 30 60 90

Fig. 9.31 Measured normalized far field patterns relevant to two X-band prototypes designed as
proof-of-concept devices for deep space exploration missions. LHCP and RHCP component of the
Rx antenna at three different frequencies (a). LHCP and RHCP component of the Tx antenna at
three different frequencies (b). The top-left insets show the picture of both the antennas
9 Metasurface Antennas 329

-5
Rx Tx
-10

-15
dB

-20

-25
|S11|
-30
6.50 6.75 7.00 7.25 7.50 7.75 8.00 8.25 8.50 8.75 9.00
GHz

Fig. 9.32 Measured reflection coefficients relevant to two X-band prototypes designed as proof-
of-concept devices for deep space exploration missions. Rx prototype (a) and Tx prototype (b).
Both prototypes are fed by a vertical monopole, coaxially fed and top-loaded with a slotted patch
to ensure the input matching. Pictures in the insets show the details of the feeding points

9.11 Discussion and Future Trends

Modulated metasurface antennas constitute a significant improvement from tradi-

tional antenna solutions. Rather than shaping a fixed boundary condition (mostly
PEC) to achieve the desired radiation characteristics, they offer the revolutionary
possibility to design the boundary condition itself to achieve the desired behavior,
with an antenna having a predetermined global shape. All the designs shown are in
fact obtained using antennas differing only in the metasurface layout and feeding
details.
The full control of the aperture field and hence of the far-field properties (beam
shape, pointing angle, polarization) is strictly related to the capability of accurately
designing the MTS. Here, an effective synthesis method has been introduced which
permits the control of amplitude, phase, and polarization of the aperture field by
adjusting the boundary conditions imposed by the MTS. The complete procedure is
described in [36]. The polarization control is based on the local value of the
anisotropy, the phase is controlled by both the shape and the periodicity of the
modulation, and the amplitude control is obtained by designing the local leakage
attenuation distribution of the leaky wave. The synthesis is based on the adiabatic
Floquet-wave expansion of currents and fields over the surface introduced in [35].
The main advantages of this process with respect to the ones already introduced are
summarized below.
• Introduction of a systematic way to synthesize the aperture field with analytical
formulas and to validate it in the systematic frame of Fig. 9.3.
• Development of new formulas based on the energy conservation, which allow
for controlling the aperture field amplitude—not possible before—and therefore
designing antennas with higher efficiency than before.
330 G. Minatti et al.

• Use of the transparent boundary conditions (in place of the opaque ones used by
the authors in previous schemes) in a rigorous Green’s Function scheme. This
allows for accounting for space dispersion of the grounded slab in practical
implementations.
Although the problem of the losses has not been treated here, small local losses
in the dielectric and in the metal can be easily introduced in the formulation. In our
experience, losses may have an impact when dealing with antennas with more than
40 dBi of gain.
We have also shown that despite the fact that modulated MTS antennas are
leaky-wave antennas, it is possible to overcome the typical beam squinting with
frequency by a proper shaping of the BCs imposed by the MTS. However, the
enlargement of the pattern bandwidth is still under study and the issue requires a
long discussion that will be the subject of a future paper. Concerning the bandwidth
of MTS, in some recent works [54, 55] it has been proposed the use of cells with
glide symmetry for realizing planar MTS lenses with ultra wide band.
Future research trends span from the design of conformal modulated MTS
antennas to the design of a reliable dynamic impedance pattern. Conformal MTSs
will allow to further reduce the encumbrance of these antennas by perfectly
matching them to curved surfaces. For instance, one may use conformal MTSs
matched to the fuselage of airplanes, reducing the impact of the antenna on the
aerodynamics. Dynamic control of the impedance patterns is of importance for all
the tracking applications, requiring to hold on a link between moving platforms. For
instance, radar applications, data connection between satellite and ground station or
Internet connections on board of trains. Also, MTS antennas with multiple opera-
tional frequencies, contoured beams and multi-beam designs are challenging and
interesting topics. However, among all of them, the dynamic control of the MTS is
the one that most attracts the attentions of the telecom-market. It is a research topic
that appears ambitious and feasible at the same: in modulated MTS antennas, the
excitation spectrum and the surface impedance pattern completely control the beam
shape and polarization. Therefore, by dynamically controlling the impedance, vir-
tually all the degrees of freedom of the radiated wave can be addressed, with
conceptually very simple means and in a way that lends to attractive implemen-
tations. Possible solutions for reconfigurable dynamic metasurfaces can involve
mechanical stretching of a proper dielectric material or electronic reconfigurations.
For pencil beams antennas, a mechanically reconfigurable surface seems to be
particularly suitable. At the opposite end of the spectrum of solutions an elec-
tronically reconfigurable surface could offer more potentialities in terms of flexi-
bility of radiative characteristics. Materials having characteristics similar to liquid
crystals at RF or based on arrays of MEMS are also good candidates for adaptive
surfaces, as well as electronic devices like PiN diodes. The final goal is to realize a
device able to radiate a wide range of arbitrarily shaped beams, circularly or linearly
polarized, by simply modifying the surface pattern impedance, without changing
the in-plane feeder.
9 Metasurface Antennas 331

References

1. C.L. Holloway, A. Dienstfrey, E.F. Kuester, J.F. O’Hara, A.K. Azad, A.J. Taylor, A
discussion on the interpretation and characterization of metafilms/metasurfaces: the two
dimensional equivalent of metamaterials. Metamaterials 3, 100–112 (2009)
2. M.G. Silverinha, C.A. Fernandes, J.R. Costa, Electromagnetic charactrerization of textured
surfaces formed by metallic pins. IEEE Trans. Antennas Propag. 56(2), 405, 415 (2008)
3. C.L. Holloway, E.F. Kuester, J.A. Gordon, J. O’Hara, J. Booth, D.R. Smith, An overview of
the theory and applications of metasurfaces: the two-dimensional equivalents of metamate-
rials. IEEE Antennas Propag. Mag. 54(2), 10–35 (2012)
4. G. Minatti, F. Caminita, M. Casaletti, S. Maci, Spiral leaky-wave antennas based on
modulated surface impedance. IEEE Trans. Antennas Propag. 59(12), 4436–4444 (2011)
5. A.M. Patel, A. Grbic, A printed leaky-wave antenna based on a sinusoidally-modulated
reactance surface. IEEE Trans. Antennas Propag. 59(6), 2087–2096 (2011)
6. G. Minatti, M. Faenzi, E. Martini, F. Caminita, P. De Vita, D. Gonzalez-Ovejero,
M. Sabbadini, S. Maci, Modulated metasurface antennas for space: synthesis, analysis and
realizations. IEEE Trans. Antennas Propag. 63(4), 1288–1300 (2015)
7. B.H. Fong, J.S. Colburn, J.J. Ottusch, J.L. Visher, D.F. Sievenpiper, Scalar and tensor
holographic artificial impedance surfaces. IEEE Trans. Antennas Propag. 58(10), 3212–3221
(2010)
8. G. Minatti, S. Maci, P. De Vita, A. Freni, M. Sabbadini, A circularly-polarized isoflux
antenna based on anisotropic metasurface. IEEE Trans. Antennas Propag. 60(11), 4998–5009
(2012)
9. M. Faenzi, F. Caminita, E. Martini, P. De Vita, G. Minatti, M. Sabbadini, S. Maci, Realization
and measurement of broadside beam modulated metasurface antennas, in Antennas Wirel.
Propag. Lett. IEEE 15, 610–613, (2016)
10. S. Pandi, C.A. Balanis, C.R. Birtcher, Design of scalar impedance holographic metasurfaces
for antenna beam formation with desired polarization. IEEE Trans. Antennas Propag. 63(7),
3016–3024 (2015)
11. M. Casaletti, M. Śmierzchalski, M. Ettorre, R. Sauleau, N. Capet, Polarized beams using
scalar metasurfaces. IEEE Trans. Antennas Propag. 64(8), 3391–3400 (2016)
12. M. Sabbadini, G. Minatti, S. Maci, P. De Vita, Method for designing a modulated metasurface
antenna structure. Patent WO 2015090351 A1, 25 June 2015
13. S. Maci, G. Minatti, M. Casaletti, M. Bosiljevac, Metasurfing: addressing waves on
impenetrable metasurfaces. IEEE Antennas Wirel. Propag. Lett. 10, 1499–1502 (2011)
14. E. Martini, S. Maci, Metasurface Transformation Theory, in Transformation
Electromagnetics and Metamaterials, ed. by D.H. Werner, D.H. Know (Springer, London,
2013), pp. 83–116
15. C. Pfeiffer, A. Grbic, A printed, broadband Luneburg lens antenna. IEEE Trans. Antennas
Propag. 58(9), 3055–3059 (2010)
16. M. Bosiljevac, M. Casaletti, F. Caminita, Z. Sipus, S. Maci, Non-uniform metasurface
Luneburg lens antenna design. IEEE Trans. Antennas Propag. 60(9), 4065–4073 (2012)
17. C. Pfeiffer, A. Grbic, Metamaterial Huygens’ surfaces: tailoring wave fronts with
reflectionless sheets. Phys. Rev. Lett. 110(19), 197401 (2013)
18. M. Selvanayagam, G. Eleftheriades, Discontinuous electromagnetic fields using orthogonal
electric and magnetic currents for wavefront manipulation. Opt. Express 21(12),
14409–14429 (2013)
19. N. Yu, P. Genevet, F. Aieta, M.A. Kats, R. Blanchard, G. Aoust, J.-P. Tetienne, Z. Gaburro,
F. Capasso, Flat optics: controlling wavefronts with optical antenna metasurfaces. IEEE J. Sel.
Topics Quantum Electron. 19(3), 4700423–4700423 (2013)
20. P.S. Kildal, E. Alfonso, A. Valero-Nogueira, E. Rajo-Iglesias, Local metamaterial-based
waveguides in gaps between parallel metal plates. IEEE Antennas Wirel. Propag. Lett. 8,
84–87 (2009)
332 G. Minatti et al.

21. A.M. Patel, A. Grbic, Transformation electromagnetics devices based on printed-circuit tensor
impedance surfaces. IEEE Trans. Microwave Theory Tech. 62(5), 1102–1111 (2014)
22. R. Quarfoth, D. Sievenpiper, Surface wave scattering reduction using beam shifters. IEEE
Antennas Wirel. Propag. Lett. 13, 963,966 (2014)
23. A. Vakil, N. Engheta, Transformation optics using graphene. Science 332, 1291–1294 (2011)
24. R. Yang, Y. Hao, An accurate control of the surface wave using transformation optics. Opt.
Express 20(9), 9341–9350 (2012)
25. R. Yang, W. Tang, T. Hao, Wideband beam-steerable flat reflectors via transformation optics.
IEEE Antennas Wirel. Propag. Lett. 10, 99–102 (2011)
26. W. Tang, C. Argyropoulos, E. Kallos, S. Wei, Y. Hao, Discrete coordinate transformation for
designing all-dielectric flat antennas. IEEE Trans. Antennas Propag. 58(12), 3795–3804
(2010)
27. M. Mencagli Jr., E. Martini, D. Gonzàlez-Ovejero, S. Maci, Metasurface transformation
optics. J. Opt. 16, 125106 (2014)
28. M. Mencagli Jr., E Martini, D Gonzàlez-Ovejero, S Maci Metasurfing by Transformation
Electromagnetics. IEEE Antennas Wirel. Propag. 13, 1767, 1770 (2014)
29. E. Martini, M. Mencagli, S. Maci, Metasurface transformation for Surface wave control. Phyl.
Trans. R. Soc. A A373, 20140355 (2015)
30. M. Mencagli, E. Martini, S. Maci, Surface wave dispersion for anisotropic metasurfaces
constituted by elliptical patches. IEEE Trans. Antennas Propag. 63(7), 2992–3003 (2015)
31. A.M. Patel, A. Grbic, Effective surface impedance of a printed-circuit tensor impedance
surface (PCTIS). IEEE Trans. Microwave Theory Tech. 61(4), 1403–1413 (2013)
32. M. Mencagli Jr., E. Martini, S. Maci, Transition functions for closed form representation of
metasurface reactance. IEEE Trans. Antennas Propag. 64(1), 136–145 (2016)
33. M. Mencagli Jr., C. Della Giovampaola, S. Maci, A Closed-form representation of
isofrequency dispersion curve and group velocity for surface waves supported by anisotropic
and spatially dispersive metasurfaces, 64(6), 2319–2327 (2016)
34. O.M Bucci, G Franceschetti, G. Mazzarella, G. Panariello, Intersection approach to array
pattern synthesis. IEE Proc. H Microwaves Antennas Propag. 137(6), 349–357 (1990)
35. G. Minatti, F. Caminita, E. Martini, S. Maci, Flat optics for leaky-waves on modulated
metasurfaces: adiabatic floquet-wave analysis. IEEE Trans. Antennas Propag. 64(9),
3896–3906 (2016)
36. G. Minatti, F. Caminita, E. Martini, M. Sabbadini, S. Maci, Synthesis of modulated-
metasurface antennas with amplitude, phase and polarization control. IEEE Trans. Antennas
Propag. 64(9), 3907–3919 (2016)
37. D. Gonzalez-Ovejero, S. Maci, Gaussian ring basis functions for the analysis of modulated
metasurface antennas. IEEE Trans. Antennas Propag. 63(9), 3982–3993 (2015)
38. S. Maci, M. Caiazzo, A. Cucini, M. Casaletti, A pole-zero matching method for EBG surfaces
composed of a dipole FSS printed on a grounded dielectric slab. IEEE Trans. Antennas
Propag. 53(1), 70–81 (2005)
39. M.A. Francavilla, E. Martini, S. Maci, G. Vecchi, On the numerical simulation of
metasurfaces with impedance boundary condition integral equations. IEEE Trans. Antennas
Propag. 63(5), 2153–2161 (2015)
40. V. Rokhlin, Rapid solution of integral equations of scattering theory in two dimensions.
J. Comput. Phys. 36(2), 414–439 (1990)
41. C.R. Anderson, An implementation of the fast multipole method without multipole. SIAM
J. Sci. Stat. Comput. 13(4), 923–947 (1992)
42. M. Albani, A. Mazzinghi, A. Freni, Asymptotic approximation of mutual admittance involved
in MoM analysis of RLSA antennas. IEEE Trans. Antennas Propag. 57(4) (2009)
43. A. Tellechea, F. Caminita, E. Martini, I. Ederra, J.C. Iriarte, R. Gonzalo, S. Maci, Dual
circularly-polarized broadside beam metasurface antenna. IEEE Trans. Antennas Propag. 64
(7) 2944–2953 (2016)
44. A. Chakraborty, B.N. Das, G.S. Sanyal, Determination of phase functions for a desired
one-dimensional pattern. IEEE Trans. Antennas Propag. 29(3), 502–506 (1981)
9 Metasurface Antennas 333

45. G. Minatti, E. Martini, S. Maci, Efficiency of metasurface antennas. IEEE Trans. Antennas
Propag. 65(4), 1532–1541, (2017)
46. A.L. Cullen, The excitation of plane surface waves. Proc. IEE—Part IV Inst. Monogr. 101(7),
225–234 (1954)
47. A. Kay, F. Zucker, Efficiency of surface wave excitation, in 1958 IRE International
Convention Record (New York, NY, USA, 1955), pp. 1–5
48. A.L. Cullen, A note on the excitation of surface waves. Proc. IEE—Part C Monogr. 104(6),
472–474 (1957)
49. D.A. Hill, J.R. Wait, Excitation of the Zenneck surface wave by a vertical aperture. Radio Sci.
13(6), 969–977 (1978)
50. S. Mahmoud, Y.M.M. Antar, H. Hammad, A. Freundorfer, Theoretical considerations in the
optimization of surface waves on a planar structure. IEEE Trans. Antennas Propag. 52(8),
2057–2063 (2004)
51. H. Hammad, Y.M.M. Antar, A. Freundorfer, S. Mahmoud, Uniplanar CPW-fed slot launchers
for efficient TM0 surface wave excitation. IEEE Trans. Microwave Theory Tech. 51(4),
1234–1240 (2003)
52. B. Friedman, W. Elwyn Williams, Excitation of surface waves. Proc. IEE—Part C Monogr.
105(7), 252–258 (1958)
53. G. Tsandoulas, Excitation of a grounded dielectric slab by a horizontal dipole. IEEE Trans.
Antennas Propag. 17(2), 156–161 (1969)
54. M. Ebrahimpouri, E. Rajo-Iglesias, Z. Sipus, O. Quevedo-Teruel, Low-cost metasurface using
glide symmetry for integrated waveguides, in 2016 10th European Conference on Antennas
and Propagation (EuCAP), Davos (2016), pp. 1–2
55. O. Quevedo-Teruel, M. Ebrahimpouri, M. Ng Mou Kehn, Ultrawideband metasurface lenses
based on off-shifted opposite layers. IEEE Antennas Wirel. Propag. Lett. 15, 484–487 (2016)
Chapter 10
Terahertz Antennas and Feeds

Goutam Chattopadhyay, Maria Alonso-delPino, Nacer Chahat,

David González-Ovejero, Choonsup Lee and Theodore Reck

Abstract Terahertz antennas present a different set of challenges to the antenna

designer typically striving for very high performance while at the very limit of the
chosen fabrication process. Many of the same design techniques used at lower
frequencies are still applied, but fabrication constraints impose significant limita-
tions on the type of structure that can be used, forcing the designer to consider
unique fabrication processes or completely new antenna structures. Through
advances in fabrication and computational techniques, the variety of terahertz
antennas is growing. This chapter presents a range of antennas applied at these
frequencies from 1.9 THz horn antennas to superconducting planar arrays. The
chapter will cover different antenna technologies and feeds such as corrugated horn
antennas, smooth-walled profiled horn antennas, multi-flare angle horn antennas,
lens antennas, microlens leaky wave antennas, metasurface antennas, antenna
arrays, off-chip antennas, and others. It will detail theory, simulation, fabrication
techniques, and state-of-the-art antenna results in all these different technologies at
millimeter and terahertz frequencies. The chapter will also provide details for ter-
ahertz antennas in the context of terahertz systems.

G. Chattopadhyay (&)
M. Alonso-delPino
N. Chahat
D. González-Ovejero
C. Lee

T. Reck
NASA Jet Propulsion Laboratory, California Institute of Technology,
Pasadena, CA, USA
e-mail: [email protected]
D. González-Ovejero
e-mail: [email protected]
D. González-Ovejero
Institut d’Électronique et de Télécommunications de Rennes—UMR CNRS 6164,
263 Avenue du Général Leclerc, Rennes 35042, France

© Springer International Publishing AG 2018 335

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_10
336 G. Chattopadhyay et al.

10.1 Introduction

Terahertz antennas present a different set of challenges to the antenna designer

typically striving for very high performance while at the very limit of the chosen
fabrication process. Many of the same design techniques used at lower frequencies
are still applied, but fabrication constraints impose significant limitations on the
type of structure that can be used, forcing the designer to consider unique fabri-
cation processes or completely new antenna structures. Through advances in fab-
rication and computational techniques, the variety of terahertz antennas is growing.
This chapter presents a range of antennas applied at these frequencies from 1.9 THz
horn antennas to superconducting planar arrays.
Horn antennas dominate terahertz systems because of the low losses and excellent
performance. In particular, corrugated horn antennas have superior radiation charac-
teristics, but are difficult to fabricate at terahertz frequencies, thus one of the most
commonly used horns above 300 GHz is the diagonal horn antenna, despite its high
cross-polarization. In addition to describing multimode horn designs such as the
Pickett-Potter horn, Sect. 10.2 presents two approaches to overcome fabrication
obstacles for achieving high gain and low cross-polarization horns. The first relies on the
stacking of platelets to form a corrugated horn. The second uses extensive optimization
to create profiled horns that has been shown to be manufacturable as high as 1.9 THz.
Lens-coupled antennas have been widely used for decades in Terahertz systems
because they avoid the high tolerances of horns. But their application is limited
because Terahertz systems are often waveguide-based system. Section 10.3
describes the leaky wave fed lens that directly couples to hollow-waveguide. In
addition, the leaky wave system reduces the area required for illumination area,
increasing pixel-spacing for array systems, and enabling a batch-fabrication process
for forming the lenses.
Section 10.4 applies the emerging technique of modulated metasurface antennas,
discussed in detail in Chap. 12, to Terahertz frequencies. These designs result in a
planar antenna which is attractive for space-borne instruments because high gains
can be achieved with less mass and volume than a comparable reflector system.
Planar Terahertz antennas are discussed in Sects. 10.5 and 10.6. To avoid lossy
on-chip antennas when coupling to CMOS components, membrane-suspended
antennas are demonstrated in Sect. 10.5 that use beam-lead technology to couple
directly to the CMOS wire-bond pads. Finally, a slot array antenna is described that
is used on Cosmic Microwave Background experiment that uses superconducting
transmission lines for routing the large-format array.

10.2 Feed Horn Antennas

The most common use for horn antennas [1–4] at submillimeter-waves is to feed
reflectors and/or quasioptical systems [5]. Reflector antennas and Gaussian beam
telescopes produce pencil beams, and to be effective, their gain has to be as high as
possible within an acceptable sidelobe envelope. Therefore, in order to increase the
10 Terahertz Antennas and Feeds 337

spillover and aperture efficiencies, feed horns are required to have equalized
beam-widths in all planes and low sidelobe levels. In addition, the reflector is
normally required to transmit and receive orthogonally polarized waves, so one has
to guarantee a low level of the cross-polar radiation patterns. Finally, for instru-
ments involving complex optics and/or focal plane arrays for multiple beam
operation it is also convenient to have good coupling to the fundamental Gaussian
mode. These are the reasons why, over the years, there has been a need in Terahertz
instruments for horn antennas which provide symmetric patterns in all planes, phase
center coincidence for the E- and H-planes, sidelobe suppression, low
cross-polarized fields, and highly Gaussian beams.
Pyramidal horns have been widely used, with good success, in many applica-
tions. Such radiators, however, possess non-symmetric beamwidths and undesirable
sidelobe levels, especially in the E-plane. Conical horns, operating in the dominant
TE11 mode, have a tapered aperture distribution in the E-plane. Thus, they exhibit
more symmetric E- and H-plane beamwidths and lower sidelobes than the pyra-
midal horns. Nevertheless, conical horns require appropriate transitions to make
them compatible with a rectangular waveguide input. To overcome some of the
limitations of pyramidal and conical horns, and further improve their attractive
characteristics, horn structures with multiple-mode excitations have been suggested
to enhance the beamwidth equalization in all planes, and to reduce both the sidelobe
level and cross-polarized fields. By appropriately combining the fundamental mode
in the waveguide with higher order modes, multimode horns allow one to tailor the
field distribution at the horn aperture. This strategy can be used to generate highly
symmetric beams with low sidelobe levels and low cross-polarized fields.
Diagonal horns are, for its simplicity, one of the most popular feeds at
submillimeter-waves [6, 7]. Figure 10.1 shows the picture of a diagonal horn with a
rectangular WR-1.5 waveguide input. Its flared section consists of a pyramid which,
at the throat of the horn, intersects a rectangular waveguide with its wide and
narrow walls parallel to the diagonals of the pyramid’s cross section. Love [6] used
a circular transition from a TE10 mode, but the direct transition described above
provides good performance with a relatively simple fabrication process.
Figure 10.2a and b shows cross sections through the block at various points along

Fig. 10.1 Pictures of a split-block diagonal horn with WR-1.5 input from Virginia Diodes.
a Assembled blocks including the horn and integrated flange. b Upper and lower blocks, where
one can also observe the dowel pins used for alignment
338 G. Chattopadhyay et al.

Fig. 10.2 a Longitudinal section of the transition from the rectangular waveguide to the diagonal
horn’s flared section. b Transverse cross sections through the block at various points along the
transition

Fig. 10.3 a TE10 and TE01

modes coexisting in a square
waveguide. b Result of
combining the two modes in
the diagonal horn’s aperture

the transition. The field beyond the transition section consists of the superposition
of two orthogonal TE10 modes, having power equally distributed between them, as
shown in Fig. 10.3a. Bearing in mind the geometry in Fig. 10.3b, one can write the
electric field at the aperture of the horn as
h
py
px i
EA ¼ E0 cos ^x þ cos ^y ej kd ð10:1Þ
a a

where k is the free space wavenumber, j xj\a=2, j yj\a=2, and d ¼ ðx2 þ y2 Þ=2Rh ,
with Rh being the slant length of the horn and a the side of the square aperture.
Therefore, for small flare angles, the field structure within the horn aperture is

E0 h
py
pxi
Ev ¼ EA
^v ¼ pffiffiffi cos þ cos ej kd
2 a a
ð10:2Þ
^¼p E0 h
py
pxi
Eh ¼ EA
h ffiffiffi cos  cos ej kd
2 a a

such that the E-field vector is parallel to one of the diagonals (Fig. 10.3b). One can
argue that, strictly speaking, the diagonal horn is not a multimode horn, given that it
does not make use of higher order TE and TM modes. However, it presents the
desirable attributes of the usual multimode horns, such as suppressed sidelobes and
cross-polarized fields in the E- and H-planes and equal beamwidths in the principal
planes. In addition, Johansson [7] showed by using a Gauss-Laguerre expansion of
the aperture fields that the diagonal horn pattern couples to the fundamental
Gaussian mode with an 84.30% efficiency when wA = 0.43a, where wA is the beam
10 Terahertz Antennas and Feeds 339

Fig. 10.4 E-field lines for a TE11 mode and TM11 and how a combination of both in a hybrid
mode leads to a very symmetric beam with very pure polarization

radius at the horn’s aperture. Nonetheless, all the attractive features above are
accomplished at the expense of pairs of cross-polarized lobes in the intercardinal
planes, making such a horn unattractive for applications where a high level of
polarization purity is required. Indeed, the analysis in [7] predicts 10% of the power
going into the cross-polarized component. Regarding their fabrication, diagonal
horns present a geometry which is suited for being milled onto two symmetric
separate pieces (split-blocks), which are then joined. The TE10 mode losses in the
rectangular waveguide are low since the split occurs along the center of the broad
walls of the waveguide. Diagonal horns are also easy to pack, which makes them
attractive for focal plane imaging applications [7, 8].
A second option consists in using multimode axisymmetric horns. An appropriate
combination of the circular TE11 and TM11 modes at the horn’s aperture leads to
far-fields with very good beam symmetry and pure linear polarization, as depicted in
Fig. 10.4. Such a dual-mode horn, also known as Potter horn [9, 10], consists of a
mode converter section, designed to excite the TM11 mode with a prescribed
amplitude, and then a phasing section (the two modes have different cut-off fre-
quencies) to ensure that both modes fulfill the correct relationship at the horn’s
aperture (see Fig. 10.5a). As a consequence of the required amplitude and phase
balance, Potter horns are relatively narrowband, although they can meet the band-
width requirements of some applications. Indeed, the Potter horn achieves complete
beam-width equalization in all planes, complete phase center coincidence, and at
least 30 dB sidelobe suppression in the electric plane over a somewhat narrowband.

Fig. 10.5 a Cross-section of a Potter horn. b Cross-section of a Picket-Potter horn, the TM11 is
generated at the step discontinuity
340 G. Chattopadhyay et al.

A similar strategy is used in Pickett-Potter horns [10, 11], which is a simplification of

the Potter design. Essentially, they consist of a smooth-walled conical horn with a
single step in the throat (Fig. 10.5b). An analysis analogous to the one carried out in
[7] to determine the Gaussicity of the pyramidal horn was performed for dual-mode
horns in [12]. The fraction of power coupled to the fundamental Gaussian mode is
96.3% when wA = 0.59a, with a being the radius of the horn aperture. This efficiency
compares well with the 98.1% of corrugated horns, normally taken as a reference. In
addition, only 1.4% of the radiated power goes into cross-polarized fields. Another
advantage of Potter and Pickett-Potter horns is that, as in the case of diagonal horns,
they are easy to machine using split-blocks.

10.2.1 Corrugated Horns

In the previous section, we have described two different approaches to generate a

hybrid mode at the horn aperture to obtain beam symmetry, low sidelobes, and
highly pure linear polarization. Despite the improvement provided with respect to
pyramidal and conical horns, the high level of cross-polarized fields that pyramidal
horns present in the intercardinal planes, and the narrow bandwidth of Potter horns
may hinder their use in some applications. The corrugated horn [2, 3, 13, 14], also
known as scalar horn, is the feed of choice for high-performance systems in which
the limitations of diagonal and Potter horns cannot be assumed.
The purpose of the corrugated surface is to provide the means to support the
propagation of a hybrid HE11 mode, which results from the combination of the TE11
and TM11 modes. The geometry of a corrugated circular waveguide, including the
definition of the different variables, is shown in Fig. 10.6 and the general expres-
sion of the HE11 mode reads:

  ðX  YÞ  
Ex ¼ A J0 kq q  B J2 kq q cosð2/Þ
kr1
ð10:3Þ
ðX  YÞ  
Ey ¼ B J2 kq q sinð2/Þ
kr1

Fig. 10.6 Corrugated

circular waveguide
10 Terahertz Antennas and Feeds 341
 
where A and B are the amplitude coefficients, Jn kq q are the Bessel functions of
the first kind and order n, and k and kq are the free space and transverse -to-
z wavenumbers. Finally, X and Y are the normalized reactance and admittance at
q ¼ r1 defined as
rffiffiffiffiffiffi rffiffiffiffiffiffi
Z/ E/ 20 Z0 H/ l0
X¼j ¼ j ; Y ¼j ¼ j ð10:4Þ
Z0 Hz l0 Zz Ez 20

One can easily realize from (10.3) that the fields are exclusively polarized in the
x direction when ðX  Y Þ ¼ 0. Moreover, the Ex component will taper radially with
q without any / variation. The ðX  Y Þ ¼ 0 condition is usually referred to as
balanced hybrid condition and it can be fulfilled either when X ¼ Y or when both
X and Y are equal to 0. In a corrugated waveguide, provided that one has several
slots per wavelength (preferably four or more), E/ has to vanish at q ¼ r1 , so X will
be 0. In addition, for narrow corrugation ridges (t\w=10) and narrow slots
(w\k=2), the normalized admittance can be approximated as Y ¼ 1= tanð2pd=kÞ
and, it will be thus 0 when d ¼ nk=4, with n being an odd integer. In other words,
the short circuit is transformed to an open-circuit at q ¼ r1 , which ensures the
absence of axial currents resulting from H/ and implies that Y ¼ 0. In [14] and the
references therein, a more refined set of formulas is provided in which d is modified
by a correction factor resulting from a rigorous formulation of the surface reactance.
The balanced hybrid condition will be fulfilled only at the frequencies in which
d ¼ nk=4, so the absence of cross-polarization will be determined by the electrical
length of d. Nevertheless, (10.3) shows that the cross-polar term decreases as kr1
increases and the larger the diameter of the horn, the wider the bandwidth for a
given cross-polar level. This is a general property of cylindrical hybrid mode feeds,
where large apertures give an inherently better performance than feeds with a small
aperture, although larger apertures will also lead to narrower beam (not suitable for
prime feed focus).
We have succinctly explained in the paragraph above how the corrugated surface
supports the propagation of the HE11 mode. However, the input of the corrugated
horn will typically be the TE11 mode of a circular waveguide. Hence, preceding the
flared section of the horn one has to introduce a mode converter to excite the hybrid
mode. There are several schemes of mode converter, namely:
(a) variable depth slot mode converter [15, 16], which is the most commonly used
mode converter, and provides up to 1.8:1 bandwidths (see Fig. 10.7a);
(b) ring-loaded slot mode converter [17], which provides up to 2.4:1 bandwidths
(see Fig. 10.7b);
(c) and variable pitch-to-width slot mode converter [18], which can deliver up to
2.05:1 bandwidths, although is less commonly used (see Fig. 10.7c).
In practice, the number of slots (NMC) in a mode converter is between 5 and 7
for mode converters (a) and (b), and between 7 and 12 for the mode converter (c).
342 G. Chattopadhyay et al.

Fig. 10.7 Conical corrugated horn with: a variable depth slot mode converter, b ring-loaded-slot
mode converter and c variable pitch-to-width slot mode converter

It is also important to note that corrugated horns are the classical solution for
launching Gaussian beams. The optimum Gaussian coupling efficiency of a conical
corrugated horn has been studied in [5] and it is equal to 98% when wa = 0.644 a0,
where wa is the beam radius at the aperture and a0 is the horn’s aperture radius.
Even if conical corrugated horns (Fig. 10.7) are the most common option, it is
possible to enhance the horn’s performance by giving them a profile other than
linear [19]. For instance, the authors of [20] proposed a Gaussian profiled horn
antenna to improve its electromagnetic performance and/or optimize its geometry
(reduce its length, weight, and/or output radius). Similarly, a horn with dual
sinus-squared and exponential profile was presented in [21]. On the other hand,
Maffei et al. [22] added a flared section at the end of a profiled horn, with the form
heuristically developed to give the desired patterns and very low sidelobes. In
general, the profile of the horn can be described using combination of exponential,
sinusoidal, hyperbolic curves, etc., with control points that are optimized to obtain
the desired far-field patterns using nonlinear least squares, minmax or genetic
algorithms. Figure 10.8a and b show a corrugated horn with spline profile opti-
mized to have a maximum coupling to the fundamental Gaussian mode (see the
magenta line in Fig. 10.8d), while keeping a good match (S11 < 25 dB) across the
490–600 GHz band, as shown in Fig. 10.8c. On the other hand, the dual-profile
(sine-squared/parallel and hyperbolic tangent/parallel) horns reported in [23] are
obtained by exploiting the fact that the coupling from the HE mode in straight
corrugated guide to a fundamental Gaussian mode in free space can be improved by
10 Terahertz Antennas and Feeds 343

Fig. 10.8 a Spline-profiled corrugated horn. b Section of the horn including the rectangular
waveguide to circular waveguide transition. c Magnitude of the reflection coefficient in dB.
d Far-field radiation patterns at the central frequency

exciting higher order HE modes. This analysis, based on higher order HE modes,
provides additional physical insight into the profile choice.
A crucial aspect of corrugated horns consists in how to fabricate the narrow
ridges at Terahertz frequencies. The continuous evolution of computer numerical
control (CNC) milling machines has made it possible to fabricate corrugated horns
at millimeter and submillimeter frequencies by direct machining into two halves of
a split-block, despite the difficulty of constructing azimuthal corrugations at short
wavelengths. However, the length of narrow flare angle feed horns, and the depth of
the matching grooves near the intersection of the circular waveguide, and the
corrugated conical flared sections restrict the tool clearance available for direct
machining operations. Hence, in some cases, a higher cost electroforming process
will be required. The first step involves the high precision machining of the internal
geometry of the circular waveguide and the corrugated conical section on an alu-
minum mandrel. The feed horn is then electroformed by electrochemical deposition
of copper and nickel onto the mandrel surface until the desired wall thickness is
achieved. Afterward, the mandrel is removed by a chemical etching process.
Finishing includes flange soldering and a gold electroplate.
344 G. Chattopadhyay et al.

In modern instruments there is a trend toward increasing the number of detectors

in the focal plane, which enables a significant increase in telescopes or imagers
mapping speeds. Electroformed horns and machined split-blocks are not easy to
adapt to the fabrication of focal planes with a very large pixel count. As an alter-
native, one can fabricate corrugated horns using platelets [24], which are thin metal
sheets [25–27] or silicon wafers [28–30]. In order to create the corrugations for an
array of horns, one has to etch two features in each sheet, namely, through holes of
thickness p and radius aj, and cylindrical steps of thickness w, and radius aj + d,
which serve to create the ridges of the corrugations (see Fig. 10.7). These sheets are
sandwiched together in a stack of many layers either using screws or pins [25, 26,
30] or diffusion bonding them together to make a single piece [24]. With the design
completed, and the photolithographic masks or machine programs created, many
platelets can be reproduced accurately and economically. This technique is esti-
mated to produce good results up to 1 THz.
Figure 10.9a shows the cross section of a platelet conical corrugated horn
including a transition, manufactured using a Deep Reactive-Ion Etching (DRIE)
process of silicon wafers. The first 20 platelets form a conical corrugated horn. An

Fig. 10.9 a Detail of the 20 corrugations in the conical horn. b Different 300-lm-thick-silicon
wafers before assembly. c Picture of the fabricated and assembled platelet antenna consisting of 24
300 lm-thick-silicon wafers. Apart from the silicon compression pins (not visible in the picture),
two plastic screws and dowels pins were included for alignment. d Measured magnitude of the
reflection coefficient in dB. e Measured co-polar radiation patter at 330 GHz, normalized to the
maximum directivity value (21.2 dB)
10 Terahertz Antennas and Feeds 345

additional 4 platelets form a circular-to-rectangular waveguide transition connected

to the metal waveguide block at WR-2.2 band (330–500 GHz). Regarding micro-
fabrication, it requires only two etch depths of 100 and 200 µm from the front- and
back side of a 300 µm thick wafer, respectively. During the etching processes of
100/200 µm, screw holes, trenches for the alignment compression pins, and
alignment marks all etched are simultaneously. A rectangular-to-circular waveguide
transition is made on a separate 4-inch silicon wafer. All 24 silicon platelets are then
coated with 2 µm thick gold using a metal sputtering system. Finally, the 24
gold-coated silicon platelets are precisely aligned and stacked together using a
silicon compression pin technique [31].
The horn presents a return loss around 20 dB across the band (see Fig. 10.9c),
better than 30 dB cross-polarization, and more than 20 dBi directivity over a 20%
fractional bandwidth at 340 GHz. The measured radiation pattern also presents
good beam symmetry, as shown in Fig. 10.9d.
To conclude this section, it is worth mentioning the importance that 3D printing
techniques might have the future development of large focal plane arrays of cor-
rugated horns. In [31, 32], the capability of fabricating W-band corrugated horns
has been investigated.
Summarizing, corrugated horn antennas provide low cross-polarization, excel-
lent beam-circularity, low sidelobe levels, nearly constant beam-width over a large
frequency band, low loss, high Gaussian coupling efficiency, and low VSWR.
However, at millimeter and submillimeter wave frequencies, the decreasing horn
size results in increasing manufacturing difficulties, time, and cost. Therefore,
corrugated horns are hardly compatible with multi-pixel arrays. We have discussed
the design, fabrication, and assembly of platelet corrugated horns and focal plane
arrays, which provide good performance and constitute a cost-effective alternative
to electroformed corrugated horns. However, these techniques have their own
limitations. In the case of platelet horns, the maximum wafer thickness has to be of
the order of k/4 to maintain the hybrid boundary condition on the horn’s wall.
Therefore, long horns (necessary for large f/D, which stands for the focal distance
f over the reflector diameter D) will require stacking a large number of wafers,
which complicates their assembly. Therefore, solutions such as multi-flared angle
[34–36] and spline-profiled horns [37] have been extensively explored.

10.2.2 Multi-flare Angle

Multi-flare horns are a multi-mode horn consisting of a small number of sections

(i.e., typically 2–3) with a linear profile (see Fig. 10.10). The simple interior profile
of makes them much easier to fabricate than spline, diagonal, Potter, and corrugated
horns because is enables a direct-drilling technique that avoids electroforming and
split-block designs. As a result, multi-flare angle horns have been identified as
promising solutions allowing fabricating large arrays out of a single metal block
[34, 35].
346 G. Chattopadhyay et al.

The fabrication steps are straightforward:

(1) A machine tool is fabricated. Its cutting edges are manufactured with the shape
of the optimized horn’s interior.
(2) The horns are drilled out of a metal block one at a time.
(3) For a better surface accuracy, the input waveguide is machined after the horn
steps are drilled.
This approach then shifts the challenge from fabrication to optimization, but
using genetic algorithm optimization excellent designs have been achieved.
Depending on the application, horn performance requirements are set in terms of
gain, beam-width, sidelobe level, cross-polarization, beam-efficiency, and return
loss. The feed needs to comply with f-number (f/D) for a given edge taper. The
directivity D0 of a horn can be determined for a given f-number using [5]:

8p2 w20 feed

D0 ¼ ð10:5Þ
k2

where w0feed is the beam waist radius of the feed given by

f
w0 feed ¼ 0:216
½TeðdBÞ0:5 k ð10:6Þ
D

where f is the focal length, and D is the reflector diameter.

It is necessary to optimize a horn accurately based on the aforementioned
requirement over a given frequency range. We will illustrate the optimization of
four multi-flare angle horns as summarized in Table 10.1.

Table 10.1 Optimized multi-flare angle horn

Frequency band Bandwidth F/D Edge taper Directivity
(GHz) (%) (dB) (dBi)
Horn #1 1700–2100 21 6 −10 31.2
Horn #2 118–183 43 2.1 −12 22.1
Horn #3 230–390 52 2.1 −12 22.1
Horn #4 650–690 6 2.97 −12 25.9

Fig. 10.10 Multi-flare angle horn with three sections

10 Terahertz Antennas and Feeds 347

10.2.2.1 Optimization Method

The multi-flare horn antennas are optimized using an in-house BoR-FDTD

(Body-of-Revolution FDTD) solver combined with real-valued and binary-coded
genetic algorithm (GA) [38]. A detailed description of the BoR-FDTD technique
can be found in [38–40]. The employed GA is associated to linear-combination
selection, crossover, and mutation operator of the chromosomes [41]. The CAD
tool is a combination between a time domain full-wave analysis solver
(BoR-FDTD) and the global search algorithm (GA). This CAD tool is designed to
find the global optimum complying with the desired specifications in radiation (i.e.,
directivity, cross-polarization, and beam circularity) and return loss over a specified
frequency band. It is based on a time domain solver (FDTD) where one simulation
can cover the entire frequency range. Indeed, this can be done using Fourier
transform of the time domain signal. This makes the tool very useful where such
large frequency band needs to be analyzed.
To reduce the computation time during the optimization process, the typical
FDTD mesh size is Dq = Dz = k0/20, where k0 is the free space wavelength. To
check the validity of the staircase approximation, the numerical result is checked
with finer mesh (Dq = Dz = k0/50) after the optimization is completed.
The optimization accounts for the cross-polarization, the antenna directivity, and
beam circularity. In addition, this optimization is effected for multiple frequency
points ranging from the minimum frequency fmin to the maximum frequency fmax.
Hence, we chose a cost function that is minimized during the optimization process
for horns having a given directivity D0, low cross-polarization level (i.e.,
X-pol < −20 dB), high beam circularity, and a maximum length Lmax (i.e.,
L1 + L2 + L3  Lmax). The cost function is given as:

fMax X
X N X
M  
Fitness ¼ W ð f Þ D hi ; / j ð10:7Þ
f ¼fMin i¼1 j¼1

where
8
  < MLow  G ðif G\MLow Þ
D hi ; /j ¼ 0 ðif MLow  G  MUp Þ ð10:8Þ
:
G  MUp ðif G [ MUp Þ

W ð f Þ ¼ 1; 8f : fMin  f  fMax ð10:9Þ

   
In this relation, G hi ; /j is the horn antenna radiation pattern, and MLow hi ; /j
 
and MUp hi ; /j are the desired masks of the pattern models. N and M denote the
number of observation points in h and u directions of the 3D patterns. In this work
N, M, hiand ui equal 361, 3, [0:0.25:90]° and [0:45:90]°, respectively. The
 function

D hi ; /j measures the differences between the horn radiation pattern G hi ; /j and
   
the lower mask MLow hi ; /j and the upper mask MUp hi ; /j (10.8). The fitness
348 G. Chattopadhyay et al.

function is precoded to include the weighting control W(f) for stressing on the
optimizing frequency. In this case, all the frequencies have the same weight, and
therefore the weight coefficients are assigned to 1 (10.9).
To control the cross-polarization level and the maximum length value of the
horn antenna, we use the penalty technique as demonstrated in [38]:
• if the cross-polarization level > −20 dB is detected, the cost function will be
double (Fitness = Fitness  2),
• if the total length of the horn is bigger the maximum length
L1 + L2 + L3  Lmax, the cost function will be triple (Fitness = Fitness  3).
The input waveguide radius R0 is fixed and its value is determined to provide a
good starting point for obtaining a good return loss, with TE11 well above the
cut-off frequency and TM11 below the cut-off frequency at the design frequency. As
such antennas are optimized by combining the TE11 mode with a TM11 mode, it is
important to make sure no TM11 mode can propagate through the input port.
Otherwise, a degenerative TM11 mode could affect the radiation pattern of the
antenna. The value R3 is approximated using a closed-form equation for a standard
linear horn providing directivity D0. The parameters R1, R2, L1, L2, and L3 are then
optimized.

10.2.2.2 1900 GHz Multi-flare Angle Horn Antenna

Design optimization
A feed horn with a f-number f/D of 6 is designed within the 1700–2100 GHz
frequency band. An f/D of 6 translates into a gain of 31.2 dBi for an edge taper Te
[−10 dB] (see Eqs. 10.5 and 10.6). The optimized horn has a gain of 31.2 dBi at
1900 GHz with a total length of 9.45 mm. Its cross-polarization is under −22 dB
and its Gaussicity higher than 97% within the 1700–2100 GHz frequency range.
The optimized dimensions are summarized in Table 10.2.
The radiation patterns obtained using the BoR-FDTD code are shown in
Fig. 10.11 at 1700 GHz and 2100 GHz in the E-, D-, and H-plane. The
cross-polarization level remains below −22 dB across the frequency range and a
good beam-circularity is obtained.

Fabrication
The multi-flare angle horn antenna is machined out of a single metallic block of
oxygen-free copper (Fig. 10.12) using an end-mill that is shaped in accordance with

Table 10.2 Optimized dimensions in micrometer for the 1.9THz multi-flare angle horn
R0 R1 R2 R3 L1 L2 L3
Optimized 74 189 261 1279 237 237 8976
Measured 83 185 253 1262 231 226 8867
10 Terahertz Antennas and Feeds 349

0 0 0

−5 −5 −5

Normalized gain (dBi)

Normalized gain (dBi)
Normalized gain (dBi)

−10 −10 −10

−15 −15 −15

−20 −20 −20

−25 −25 −25

−30 −30 −30

−35 −35 −35

−40 −40 −40

−20 −15 −10 −5 0 5 10 15 20 −20 −15 −10 −5 0 5 10 15 20 −20 −15 −10 −5 0 5 10 15 20
Angle (degree) Angle (degree) Angle (degree)
(a) E-plane (b) D-plane (c) H -plane
0 0 0
−5 −5 −5
Normalized gain (dBi)

Normalized gain (dBi)

Normalized gain (dBi)

−10 −10 −10

−15 −15 −15

−20 −20 −20

−25 −25 −25

−30 −30 −30

−35 −35 −35

−40 −40 −40

−20 −15 −10 −5 0 5 10 15 20 −20 −15 −10 −5 0 5 10 15 20 −20 −15 −10 −5 0 5 10 15 20
Angle (degree) Angle (degree) Angle (degree)
(d) E-plane (e) D-plane (f) H -plane

Fig. 10.11 Calculated radiation patterns with IETR BoR-FDTD at (top) 1700 GHz and (bottom)
2100 GHz

the multi-flare angle horn contour. The fabricated multi-flare angle horn antenna is
shown in Fig. 10.12. Four horns were manufactured and one was cut in half for
dimension inspection (Fig. 10.12). The designed and measured dimensions are
compared in Table 10.2. The measured dimensions are all within 5% of the design
except the throat radius. A better alignment of the input waveguide to the horn was
obtained when the circular input waveguide was milled through the aperture of the
horn, compared to attempting to mill from the backside. However, doing so requires
an extremely high aspect ratio end-mill to reach down the horn. Hence, the tool
tends to bow as it cuts and produced a wider throat.

Fig. 10.12 Fabricated 1.9 THz multi-flare angle horn prototype. a Antenna prototype.
b Cross-section of the antenna prototype
350 G. Chattopadhyay et al.

Fig. 10.13 Radiation pattern measurement set-up using an Infrared Labs HLR-5 bolometer.
a Photograph and b block diagram of the measurement set-up

Measurement
The multi-flare angle horn is measured in transmission mode with an Infrared Labs
HLR-5 bolometer (Fig. 10.13). A 1.9 THz multiplier chain consisting of a Millitech
WR-10 AMC, a JPL GaAs WR-10 power amplifier chain, a JPL designed doubler,
and two JPL triplers are assembled to drive the horn [35]. The output of the 1.9 THz
tripler includes a rectangular-to-circular transition. The output power at the circular
waveguide that feeds the horn is approximately 600 nW. The bolometer is a
cryogenically cooled silicon bolometer with a responsivity of 220,000 V/W
focused by a Winston cone with an f/D of 3.8. The bolometer is insensitive to
polarization so a grating is used to filter either the co- or cross-polarization. The
multiplier chain is mounted on a stepper-motor controlled mount that rotates in two
axes around the approximate phase center of the horn. The synthesizer driving the
multiplier chain is cycled at 50 Hz and a lock-in amplifier (SR-830) is used to
detect the measured power. Finally, a computer is used to record the output signal
from the lock-in with the stage’s position.
10 Terahertz Antennas and Feeds 351

(a) (b)
0 0

−5 −5
Normalized gain (dBi)

Normalized gain (dBi)

−10 −10

−15 Calculated −15 Calculated

Measured Measured
−20 −20

−25 −25

−30 −30

−35 −35

−40 −40
−6 −4 −2 0 2 4 6 −6 −4 −2 0 2 4 6
Angle (degree) Angle (degree)

(c)
0

−5
Normalized gain (dBi)

−10

−15 Calculated
Measured
−20

−25

−30

−35

−40
−6 −4 −2 0 2 4 6
Angle (degree)

Fig. 10.14 Calculated and measured antenna radiation pattern in the a E-, b D-, and c H-planes

The radiation pattern is measured at 1.9 THz in the E-, D-, and H-planes. The
calculated and measured results are shown in Fig. 10.14 are in good agreement. The
directivity is calculated from the normalized intensity measured over a ±6-degree
scan in both elevation and azimuth. The measured directivity of 31.7 dBi is very
close to the design value of 31.6 dBi. To demonstrate the fabrication yield, the three
fabricated horns are measured and only 0.1 dB variation in gain is observed. The
measured cross-polarization remains below −20 dB. Using the Friis formula, the
horn efficiency is assessed to be around 80 ± 15% [35].

10.2.2.3 TWICE Multi-flare Angle Horn Antenna

Three multi-flare angle horns are designed for the mission Tropospheric Water and
Cloud ICE (TWICE) which consists of a radiometer integrated in a 6U CubeSat
measuring water vapor and ice clouds in upper troposphere. First, the 183 GHz
channel and 380 GHz channels are chosen to retrieve water vapor in the tropo-
sphere and upper troposphere/lower stratosphere (UTLS). To constrain the water
352 G. Chattopadhyay et al.

Table 10.3 Optimized dimensions in millimeter for TWICE’s multi-flare angle horns
R0 R1 R2 R3 L1 L2 L3
Horn #2 0.900 1.995 2.518 5.500 6 8.911 11.608
Horn #3 0.450 0.884 1.211 2.746 4.339 4.777 6.141
Horn #4 0.230 0.496 0.613 2.004 0.492 1.346 6.506

vapor retrievals, 118 GHz channel measures tropospheric temperature profile using
O2 absorption line.
Horn #2 and #3 are optimized across broad frequency bands (i.e., band-
width > 40%) and horn #4 is optimized across a narrow frequency band (i.e., band-
width = 6%) (Table 10.1). The optimized dimensions for each antenna are
summarized in Table 10.3. The optimized radiation pattern for each antenna is shown
in Fig. 10.15 at the lower and upper band-edge frequencies. The multi-flare angle horn
antennas clearly demonstrate low cross-polarization levels (i.e., <−19 dB), high beam
circularity, and high Gaussicity.
As this specific mission uses single feed elements, the feed horns are machined
by milling machines, operated under computer numerical control (CNC) and using
a split-block package (Fig. 10.16). This allows to integrate the
rectangular-to-circular waveguide transition in the same block.
Horn #4 was fabricated and measured. The calculated and measured radiation
patterns at 670 GHz (Fig. 10.17) are in excellent agreement.

10.2.3 Smooth-Walled Horn

It is also possible to define the profile of our axisymmetric horn using cubic splines
and then optimize the horn by shaping the splines. Such spline profile is typically
described using a few nodes (two extreme nodes and five inner nodes, as shown in
Fig. 10.18, in order to simplify the optimization process by limiting the number of
parameters. The electromagnetic parameters of the horn are evaluated using modal
matching techniques [3] or body-of-revolution solvers [38–41] at each optimization
step. The initial value of the parameters d1 . . . d6 and L1 . . . L6 are introduced by the
user, and they can be retrieved from the radius and length of a corrugated horn with
the desired beam waist, bearing in mind that spline-profiled horns are, in general,
longer. An alternative staring profile is given by a sin0.75 profile followed by a
Gaussian section. Then, the parameters are optimized to provide the desire radiation
pattern by minimizing a penalty function. The penalty function used is analogous to
the one described in the previous section. Finally, it is important to note that some
constraints must be applied to the radii to force the optimizer to create shapes that
are physically possible to manufacture.
Figure 10.19a shows a block with a spline-profiled horn, whereas in Fig. 10.19b
one can see the two split-blocks and, in the right-hand side block, one can also
10 Terahertz Antennas and Feeds 353

p
0 0
E-plane E-plane
-2 D-plane -2 D-plane
Normalized directivity (dB)

H-plane

Normalized directivity (dB)

H-plane
-4 -4

-6 -6

-8 -8

-10 -10

-12 -12

-14 -14

-16 -16

-18 -18

-20 -20
-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
Angle (degree) Angle (degree)
fmin = 118 GHz fmax = 183 GHz
(a) Horn #2
0 0
E-plane E-plane
-2 D-plane
-2 D-plane
Normalized directivity (dB)

-4
H-plane Normalized directivity (dB) -4
H-plane

-6 -6

-8 -8

-10 -10

-12 -12

-14 -14

-16 -16

-18 -18

-20 -20
-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
Angle (degree) Angle (degree)
fmin = 230 GHz fmax = 390 GHz
(b) Horn #3
0 0
E-plane E-plane
-2 D-plane -2 D-plane
Normalized directivity (dB)

H-plane H-plane
Normalized directivity (dB)

-4 -4
-6 -6
-8 -8

-10 -10
-12 -12
-14 -14
-16 -16
-18 -18

-20 -20
-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
Angle (degree) Angle (degree)
fmin = 650 GHz fmax = 690 GHz
(c) Horn #4

Fig. 10.15 Optimized radiation patterns of each horn

distinguish the dowel pins used for alignment. The horn has been designed for the
300–380 GHz band, where the simulated magnitude of the reflection coefficient is
lower than −30 dB. Figure 10.19c shows the measured co-polar directivity pattern
354 G. Chattopadhyay et al.

Fig. 10.16 Channel 3 multi-flare angle horn for TWICE

(a) E-plane (b) H-plane

0 0
Calculated Calculated
-5 Measured -5 Measured
Normalized directivity (dB)

Normalized directivity (dB)

-10 -10

-15 -15

-20 -20

-25 -25

-30 -30

-35 -35

-40 -40
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
Angle (degree) Angle (degree)

Fig. 10.17 Calculated and measured radiation pattern of the channel 3 multi-flare angle horn at
670 GHz

Fig. 10.18 Example of

spline-profiled smooth-walled
horn, including the variables
used to define the cubic
splines nodes

at the central frequency, while Fig. 10.19d shows the cross-polar one. The mea-
sured directivity is 24.25 dB. The ripples in the main lobe are due diffraction and
standing wave effects in the measurement set-up. In this particular design, beam
symmetry and cross-polar discrimination have been slightly sacrificed in order to
obtain a good impedance matching over a larger frequency band.
10 Terahertz Antennas and Feeds 355

Fig. 10.19 a Block with a spline-profiled horn. b Split-blocks

10.3 Lens Antennas

10.3.1 Introduction and State of the Art

As it has been explained in the previous section, horn antenna can be fabricated
either by split-block machining, drilling a metal block with custom drill tips or
stacking layers of etched holes in silicon to form platelet horns. Although these
techniques can provide good efficiencies at terahertz, higher directivities are diffi-
cult to achieve. The higher the directivity the longer the horn, which increases the
conductor losses and the fabrication tolerances. Thus, lenses may be preferred at
submillimeter-wave frequencies; they can be fabricated with silicon microma-
chining and be easily integrated with the front end.
Dielectric lens antennas are taken from the optical regime, replacing the
waveguide horn for a planar printed antenna on a thick dielectric and a silicon lens
[42]. The advantages of these hybrid antennas compared to waveguide systems are
clear: low loss, easy integration, and low cost of manufacture, as the antennas can
be fabricated using photolithographic techniques and the lenses using laser
micromachining. They are commonly found in focal plane array configurations for
imaging applications as in [43, 44], providing high efficiency over a narrow
bandwidth with double slots antennas. They can also be used as standalone
antennas using broadband printed antenna such as bow-ties, logarithmic spirals as
photo-mixing antennas for spectroscopy applications as in [45–49].
356 G. Chattopadhyay et al.

The use of a dielectric lens is the most practical solution for obtaining high
efficiency from a planar antenna at terahertz frequencies. When planar antennas are
printed on dielectric substrates, they are prone to suffer from power loss due to the
surface wave modes propagating within the substrate. These surface waves are
confined inside the dielectric (they have a propagation constant larger compared to
the free space) and absorb a certain percentage of the radiated power, which con-
stitutes an efficiency loss. If the planar antenna is printed on the back of a high
dielectric lens, the radiation will mostly go to the highest dielectric and by the use
of a lens profile this radiation will be released into free space. This solution has
been extensively used in the literature because of its simplicity and mechanical
robustness compared to the use of substrate membranes, which should be in the
order of 0.04k0 or 0.01k0 for a low loss operation of an elemental slot or a dipole
antenna, respectively [43].
The dielectric lens design that has been most extensively is the extended
hemispherical lens proposed by Rutledge [52]. The extended hemispherical shape is
synthesized from an ellipsoidal shape where the antenna feed is placed at the second
pffiffiffiffi
focus F of the ellipse with an eccentricity e ¼ 1= er in order to achieve maximum
directivity and provide good coupling to an optical system. The rays emerging from
the focus point above the critical angle will radiate outside the lens and will be
perfectly collimated, while the rays that hit below will be reflected in different
directions. The performance is constrained by three effects, all caused by the illu-
mination of the lower part of the lens. The first is the dispersion effect caused by the
multiple reflections that the rays undergo when illuminating the lower portion of the
lens, even with the use of a matching layer (see Fig. 10.20a). The second is the
phase error associated to the difference between an ideal elliptical and an extended
hemispherical lens (see Fig. 10.20b). The last is the coupling loss to the optical
system, i.e., how good the antenna is coupled to a Gaussian beam, defined as
Gaussicity (see Fig. 10.20c). Thus, to improve the performance of these lens
antennas the solution has been to employ directive feeds which focus the radiation
on the upper part of the lens. That is why double slots, which avoid the illumination

Fig. 10.20 Lens illumination scheme representing the reflection losses, loss in Gaussicity and
phase error of extended hemispherical dielectric lenses
10 Terahertz Antennas and Feeds 357

of these areas, are preferred as lens antenna feed compared to a simple dipole or slot
[44].
A solution for this limitation was recently presented in [53, 54]. It consists of the
combination of these simple printed antennas with a leaky wave cavity between the
lens and the feed antenna. This section will be centered on the use of a leaky wave
mechanism for a waveguide type of antenna feed. It is a work that was first started
on [55] and has been continued on [56, 57]. The general design rules and the
fabrication will be covered in this chapter, with a focus on the development and
challenges for these dielectric lens waveguide antennas.

10.3.2 Silicon Microlens Antennas Fed by Leaky Wave

Waveguide Feeds

From the guidelines of Jackson [58], Neto et al. proposed in [59] and then in [53,
54] the use of leaky wave cavities to increase the gain of planar printed antennas at
terahertz frequencies. The cavity increases the directivity of the antenna feed that
excites the dielectric lens in order to diminish the impact of the reflections at the
dielectric air interface.
Instead of using resonant dielectric layer stratifications as in [60], the lens can be
considered an infinite layer of dielectric. By leaving a gap between the antenna feed
and the lens, the directivity of the feed increases. The highest directivity is achieved
at broadside when the cavity thickness is k0/2, being k0 the wavelength in the air,
which leads to a resonance at f0. There are three modes propagating in the cavity:
the dominant TE1 and TM1 modes that point toward broadside, and the non-desired
TM0 mode that radiates toward larger angles. The radiation of the TM0 mode will
not be intercepted by the lens, resulting in a spillover loss. The spillover loss of the
lens is defined as the amount of power transmitted that is not caught by the aperture
of the lens.
For the correct excitation of the leaky wave modes using a waveguide as a
feeding structure, a membrane with a double arc slot iris is required. The pair of
slots is excited in phase and is separated by a distance such that their contributions
cancel each other. This double slot iris configuration cancels the spurious mode
TM0 mode generated inside the leaky wave cavity [61]. By adjusting the length and
the width of the slot, the antenna impedance is matched to the TE10 mode of the
waveguide. An example of a design is shown in Fig. 10.21b, where two double
slots are tuned to operate 550 GHz on a 15% bandwidth for the dimensions
specified. Dual-polarization can be achieved if another set another pair of arc slots
are placed orthogonally from the original (see the inset of Fig. 10.21a) and excite
the waveguide feed with two linear TE10 modes.
Using the dimensions of the provided example, we can examine the radiation
patterns along the infinite silicon dielectric. An example of such patterns for the
leaky wave feed designed at 550 GHz is shown in Fig. 10.22a and b. As it can be
358 G. Chattopadhyay et al.

Fig. 10.21 a Scheme of the leaky wave waveguide feed and an inset of the iris for single
polarization (up) and for double polarization (down). b Reflection coefficient example at 550 GHz
for the dimensions shown in the inset. The air cavity has a thickness of 273 lm and the membrane
is 15-lm thick

seen, the peak directivity in the broadside direction and the radiation patterns are
very symmetric. This pattern makes it very suitable to feed a focusing element such
as a lens with a moderate f-number (F/D). Due to the high directivity of the feed, the
lens surface can be in the near or far-field of the antenna, depending on the lens
aperture one needs to use. An example is shown in Fig. 10.22b, which illustrates
that broader patterns are achieved in the case of the operation in the near-field
compared to the far-field patterns. Thus, special care needs to be taken into account
when optimizing the lens antenna.
The phase center for the leaky wave is not in the plane of the waveguide
aperture, it is below the waveguide aperture. This is clearly perceived in the phase
of the radiation patterns shown in Fig. 10.22 example, which have been calculated
considering the reference plane is at the ground plane of the waveguide feed. This
displacement of the phase center from the waveguide feed can be compensated by
adjusting the height of the lens, as it will be shown next.

Fig. 10.22 Directivity (right axis) and phase (left axis) of the leaky wave waveguide feed inside
the silicon, a at the far-field and b at the near-field distance of 4.5k0
10 Terahertz Antennas and Feeds 359

Fig. 10.23 a Sketch of the silicon microlens geometry. b Example that illustrates the relationship
between the taper angle hf and the wafer thickness W at 550 GHz

The lens geometry, shown in Fig. 10.23a, is characterized by a silicon wafer of

thickness W on top of the cavity and a silicon lens defined by a certain diameter D,
edge angle hf, radius of the hemisphere R, and height of the lens H. All these
parameters need to be optimized to achieve maximum gain and Gaussicity.
In order to do the optimization of the antenna and lens, a Physical Optical
(PO) tool is essential to decrease the computational time that a full 3D electro-
magnetic simulator would need to compute and optimize the whole structure. For
all the optimizations, a combination of a full 3D simulation and a PO tool has been
employed. The full 3D simulator provides accuracy in the computation of the
primary field of the feed (the leaky wave cavity, membrane and iris). This primary
field is used to compute the distribution of the electric and magnetic fields across
the spherical surface of the extended hemispherical lenses. The Fresnel transmission
coefficients are used at the boundary to calculate the reflected and radiated field
values. After the fields have been multiplied by the appropriate transmission
coefficients, the equivalent and magnetic current densities are determined just
outside the spherical surface. From these currents, the far-fields are then calculated
by integrating over the lens surface. For the cases considered in this chapter, the
lens is not placed in the reactive near-field region of the feed, but in the radiative
region, and therefore the PO can still be performed.
The aim is to design the lens for optimal performance in terms of a gain and
optical coupling to a Gaussian beam (Gaussicity) for a certain aperture diameter
D. Thus, the lens needs to have at a certain height, H, and it needs to be placed at a
certain distance from the feed W (see the scheme of Fig. 10.24a). The procedure to
design the dielectric lens antenna for a certain diameter/aperture is the following:
• The waveguide, double slot, and air cavity are designed at the central frequency,
in the example is 550 GHz. The double slot dimensions are optimized with a
full-wave simulator for maximum radiation efficiency. The electric field com-
ponents in the cavity plane are exported from the 3D simulator into a
custom-made program that will perform the optimization of the lens geometry.
360 G. Chattopadhyay et al.

Fig. 10.24 a Sketch of the dependency between the lens height H and the radius of the extended
hemispherical lens. b Gain with and without matching layer as a function of the lens height H.
c Directivity and gain as a function of the frequency with and without matching layer (left axis),
Gaussicity as a function of the frequency (right axis)

• The optimum distance W is determined based on the frequency, the feed antenna
illumination and the diameter of the lens. First, the taper edge angle hf asso-
ciated with the field level at the edge of the lens aperture is determined for
certain q. For example, a taper of 12 dB, Te[12 dB], means that the field level at
the edge of the lens diameter is 12 dB below the field level at the broadside
direction of the lens. The field taper is chosen according to the tradeoff between
the spillover loss and the taper efficiency. The taper efficiency expresses the
uniformity of the field distribution on the aperture of the lens. The optimum
taper is between 10 and 14 dB, similar to reflector antennas.

Thus, we will compute the field from the antenna feed inside the silicon for
certain distances (in the near or far-field depending on the distance) and compute
the taper edge angle hf for a taper field of Te[10 dB], Te[12 dB] and Te[14 dB].
The intersection of these values with the theoretical value of the taper for a fixed
10 Terahertz Antennas and Feeds 361

aperture hf ¼ sin1 ðD=2qÞ will give the optimum distance where the lens should be
placed. An example is shown in Fig. 10.23b, where the dashed lines present the
taper angle obtained from the computation and the solid lines present the theoretical
value of the taper for a diameter of 2.4 mm. Thus, for a lens aperture of
D = 2.4 mm and a taper of Te[12 dB], the lens aperture should be placed at
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
W ¼ q2opt  ðD=2Þ2 ¼ 3:8 mm. This computation is essential when the lens is
placed in the near-field of the primary feed, as it can be noted in Fig. 10.23b, when
we look at the dependency between the distance and the taper angle. When we
move toward the far-field region this dependency fades, and the curves in
Fig. 10.23b get flatten (hf becomes constant with q).
• Once the optimum distance W is set for a certain diameter, the optimum height
of the lens H can be tuned to achieve maximum of directivity or taper efficiency.
This tuning performs the optimization of the phase error associated to the phase
center displacement and the phase error associated to the fact that the extended
hemispherical lens is not quite an ellipse. In general, the larger the diameter the
higher the optimum height. Following the previous example, for a lens of
2.4 mm at 550 GHz, the optimum lens height is H = 181 lm which provides a
directivity of 22.2 dB (see Fig. 10.24b). The gain includes spillover losses and
the reflection losses inside the lens with and without the use of a matching layer.
The reflections due to the use of high dielectric substrates can strongly impact
the performance of the antenna, causing a large power loss. As Fig. 10.24b
shows, the total gain improves by 2 dB because of the use of a matching layer.
pffiffiffiffiffi
The optimum matching layer is designed with a dielectric of em ¼ eSi and has
pffiffiffiffiffi
a thickness of km =4 ¼ k0 =4 em , being k0 the wavelength at the desired central
frequency. Since this antenna has a relatively narrowband, the use of one
matching layer will be sufficient to cover the whole band. The frequency
behavior of the resulting antenna is shown in Fig. 10.24c, where the directivity,
the gain with and without matching layer, and the Gaussicity are presented as a
function of the frequency.

10.3.3 Fabrication and Assembly

As it has been previously mentioned in the introduction, advanced fabrication

processes and techniques are often required when working at terahertz frequencies,
and it is one of the major constraints when designing these antennas. Silicon
dielectric lens antennas have the advantage that the whole antenna can be fabricated
using a silicon micromachining process, which provides excellent precision and
accuracy, as well as a high level of integration with front-end transceivers.
362 G. Chattopadhyay et al.

For instance, the silicon lens fed by the leaky wave waveguide feed can be
completely discredited among silicon wafers, processed with a DRIE silicon
micromachining process and then stacked together. Specifically, it can be assem-
bled in three parts: the first one is the leaky waveguide feed that contains waveg-
uide, membrane and air-cavity; the second one are the blank silicon wafers required
to achieve the correct thickness of the lens, W, and the third one is the lens surface.
The following sections describe in detail the fabrication process for the two main
parts.

10.3.4 Fabrication of the Leaky Wave Waveguide Feed

The three elements of the feed are the waveguide, the membrane with a double slot
and the k0/2 air cavity. All three can be fabricated on the same double SOI wafer,
using a three-step etch process.
The double SOI wafer sketch is presented in Fig. 10.25, and it contains two SiO2
layers that act as etch stops to define the boundaries of the air cavity, the membrane
with the iris and the waveguide length. For the leaky wave feed example shown in
Fig. 10.21, designed at 550 GHz, the SOI wafer scheme consists of: 272 lm of
silicon that provides the depth for the air cavity, a 250 nm layer of SiO2 as the etch
stop, a silicon layer 17 lm thick as the membrane for the iris, another 250 nm layer
of SiO2 as the etch stop and a final silicon layer of 995 lm for the rectangular
waveguide.
The wafer can be patterned using conventional photolithographic techniques
over the photoresist. The etching of the silicon can be performed using DRIE
processes for deep selective silicon trenches which can achieve a selectivity of
300:1. The SiO2 features can be etched using inducted coupled plasmas (ICP). Both
DRIE and ICP allow the fabrication of clean and perfectly controlled silicon
membranes.
After etching, both sides of the SOI wafer are sputtered with gold to metallize
the waveguide and iris. A uniform coverage can be achieved by a good optimization
of the sputtering process, even in structures that contain vertical walls with high
aspect ratios.

Fig. 10.25 Sketch of the etching steps in the SOI wafer for the leaky wave waveguide feed
fabrication
10 Terahertz Antennas and Feeds 363

10.3.5 Fabrication of the Shallow Silicon Lens

The fabrication of the shallow silicon lens consists of a single photolithographic

process that can control the diameter, height and curvature of the lens, based on
control of the shape of the deposited photoresist and the etching selectivity between
the photoresist and the silicon.
The basic idea is to transfer the lens shape into a photoresist lens and transfer the
shape onto the silicon wafer by a DRIE process as Fig. 10.26 shows. The process
consists of four steps:
(1) Deposit a thick layer of photoresist and pattern it with the desired lens aperture
using conventional photolithographic process.
(2) Reflow the photoresist pattern with heat in order to obtain the lens profile. By
coating the photoresist above the glass transition temperature, the surface
tension applied to the photoresist makes the photoresist to reflow in the
resulting spherical shape of the lens.
(3) Etch the photoresist and silicon simultaneously with a DRIE process.
(4) The process finishes when a complete etching is achieved and the lens profile is
transferred to the silicon wafer.

This process can control both the diameter and the curvature of silicon lens. The
diameter of the lens depends on the thickness of the photoresist that can be coated
and patterned. So far, the lens diameter of around 6.5 mm with a height of 400 lm
has been successfully microfabricated as in [63]. Here, we will focus on a lens
fabricated with a diameter of 3 mm that will follow the example at 550 GHz. The
curvature of the shallow lens can be controlled by two process variables:

Fig. 10.26 Sketch of the

silicon lens fabrication
process using
photolithographic techniques
364 G. Chattopadhyay et al.

Fig. 10.27 a Resulting scanning of the fabricated lens profile. b Error Surface between the
fabricated lens profile and a spherical surface

• Amount of photoresist: The pattern and the height of the photoresist applied
onto the silicon wafer will determine the curvature of the resulting lens. The
thicker the photoresist deposited, the thicker the lens can be.
• Etching selectivity applied to the photoresist and silicon: The curvature of the
silicon lens can be controlled by the adjusting the selectivity between of the
silicon and photoresist, which is optimized by controlling the gas ratio between
the CF4 and O2, the bias power and RF power. For example, if the etching
selectivity is 1:3, meaning 1 unit of photoresist etches three times the rate of the
silicon, the curvature of the silicon lens will be three times higher than the
curvature of the photoresist lens. In the lens shown in Fig. 10.27, the original
height of photoresist was 210 lm but because the etching selectivity was 1:3,
the height of the silicon lens was around 280 lm.
The surface roughness achieved with this process on the silicon lens is better
than 0.5 lm which is more than sufficient for terahertz frequencies.
The use of a matching layer on dielectric lenses is essential to reduce the high
reflection losses produced by the use of high permittivity substrates. At terahertz
frequencies the material that is commonly used as a matching layer for silicon
lenses is Parylene. It has a permittivity of 2.69 which is not ideal for silicon (the
optimum permittivity for a matching layer for a silicon lens would be
pffiffiffiffiffiffiffiffiffi
em ¼ 11:9 ¼ 3:45) but close enough to reduce considerably the losses by
reflection. Also, Parylene polymer coatings can be applied using vapor deposition
equipment, which allows a very conformal deposition with high control of the
coating rate and thickness. The dimer (Parylene-C) is vaporized under vacuum and
heat to form a dimeric gas. The gas is pyrolized to transform it into its monomeric
form so it can be deposited as a transparent polymer film.
Characterizing the accuracy of the fabricated lens compared to an ideal hemi-
spherical lens has been the key for understanding and optimizing the fabrication.
The surface of the fabricated lens can be characterized and studied from the
mapping performed by a surface profilometer. In order to study how the fabricated
surface resembles a sphere, we can use the optimization toolbox in Matlab to
10 Terahertz Antennas and Feeds 365

Fig. 10.28 a 2D and b 1D cuts of the radiation pattern of a Gaussian aperture with a field taper of
14 dB which contains the aberrations of the fabricated microlens at 550 GHz

compute the sphere dimensions. The sphere has a certain radius, R, aperture
diameter, D, sphere height, H, that minimizes the error with the fabricated surface.
As for example the fabricated lens shown in Fig. 10.27a, the optimum dimensions
were R = 3.90 mm, D = 2.93 mm and H = 285.6 lm, and the standard deviation
of the error was r = 18.36 lm = 0.11 kSi. From the error surface (see Fig. 10.27b)
we can see that the highest error is contained in the edges of the lens. For instance,
if 100 lm of the edges of the lens are not considered, the standard deviation goes
down to r = 5.4 lm = 0.04 kSi. This adjustment in the illumination can be
achieved by the tuning of the wafer thickness, W, which in this case will be
designed to obtain a field taper of −14 dB at the border of this new effective lens
aperture. Thus, by reducing the wafer thickness from the original W = 3.8 mm to
W = 3.21 mm, we will reduce the effective diameter of the lens from 2.93 to
2.38 mm.
The effect of this aberrations can be understood as a phase distortion in the
aperture field of the lens. Assuming a tapered Gaussian aperture of the diameter of
the lens we can estimate what would be the effect of this phase error over the
radiation pattern. For the error surface shown in Fig. 10.27b, the lens distortion
would correspond to an increase of the secondary lobes in the radiation pattern and
a tilt in the main direction of the main beam, as Fig. 10.28 shows.

10.3.6 Silicon Microlens Antenna Performance

and Limitations

The key advantage of using this silicon dielectric lens antenna at terahertz fre-
quencies is the fact that the lens antenna can be fabricated entirely using a silicon
micromachining process. This means that from the simple processing of one wafer,
over thousands of elements can be obtained. Figure 10.29, shows some examples of
366 G. Chattopadhyay et al.

Fig. 10.29 Examples of lens arrays manufactured using silicon micromachining processes a One
of the first 5  5 lens array of D = 2.5 mm, b 50 arrays of 10  10 lenses of D = 1.5 mm, c 4
linear arrays of 6 and 8 lenses of D = 6.4 mm

silicon lens arrays fabricated with this technology, with diameters of 1 mm, 3 mm,
and 6.5 mm.
Lenses from the arrays (shown in Fig. 10.29) were scanned to check their
accuracy [62]. The agreement to a spherical profile is good particularly for the small
lenses (see Fig. 10.30). From the prototypes, several lens profiles have been studied
with the technique presented before, and even individual antenna lens prototypes
have been fabricated and measured. In this section, we will present one of the
published prototypes in [56] to give an idea of how the assembly and measurements
were performed.
Figure 10.31 shows the lens and the iris prototype at 550 GHz. Both pieces are
assembled using metal pins that provide alignment and screws that provide the
pressure to the wafers. The thickness between the lens and leaky wave is optimized
for the illumination of an effective diameter of 2.39 mm and reduce the aberrations
from the fabricated profile. To achieve the desired thickness, five silicon wafers are
placed in between. These silicon wafers, as well as the lens wafers, need to be of
high resistivity, i.e., of 10 kΩ cm, in order to lower the dielectric loss. In this
prototype, the dielectric losses are calculated to be around 0.06 dB for the entire
silicon stack. Moreover, double-sided polished wafers are employed in order to
have good surface contact between all the wafers, avoiding air gaps.
Radiation pattern measurements were performed of the prototype at 550 GHz.
Figure 10.32 presents the E and H-planes (solid and dashed lines, respectively) of
the measured normalized radiation pattern at 530 GHz, 550 GHz and 590 GHz.
The agreement with the simulations (shown in gray) is very good. Well-focused
beams were achieved, and the difference with the simulations were primarily due to
a small misalignment between the lens and the antenna feed and, as it was portrayed
before, the aberrations of the shallow lens surface that would increase the secondary
lobes and a tilt on the radiation pattern.
The results achieved so far show a great potential to use these dielectric lens
antennas in the development of future focal plane arrays at terahertz frequencies.
Using the leaky wave waveguide feed, we only need a small part of the surface of
the lens, which reduces the reflection losses and phase errors that these types of
10 Terahertz Antennas and Feeds 367

Fig. 10.30 Examples of fabricated lenses with diameters a 1.5 mm, b 2.5 mm and c 6.5 mm, and
their comparison with a perfect hemispherical surface

Fig. 10.31 Photograph of the fabricated prototype at 550 GHz

Fig. 10.32 Measurements of the radiation patterns at 530, 550 and 590 GHz. The solid lines
correspond to the E-plane and dashed line to the H-Plane

lenses suffer. But most of all, it allows the use of photolithographic process when
fabricating the lens. The fabrication of the lenses using photolithographic process
reduces the cost, with the comparable performance achieved with other fabrication
methods, such as laser micromachining. At the moment, it is still challenging to
predict precisely some of the lens parameters in the fabrication, particularly the
curvature, when we millimeter diameter silicon lenses are attempted. However, by
characterizing accurately the fabricated profiles, we can partially compensate the
368 G. Chattopadhyay et al.

aberrations and still obtain acceptable performances while continuing to keep

improve the fabrication process of these lenses.

10.4 Terahertz Modulated Metasurface Antennas

The reader can find a comprehensive description of the analysis and design of
modulated metasurface (MTS) antennas in Chap. 9. These antennas analyzed
consist of planar apertures made of subwavelength patches printed on a grounded
dielectric slab. This class of antenna is, thus, inherently flat and presents a low
mass. Its most attractive feature is the capability of obtaining high- to very high
gains with low-profile structures. Indeed, the aforementioned characteristics make
modulated MTSs particularly appealing for space [64, 65]. Among their other
advantages it is important to note their capability of conforming to curved surfaces
[66], and providing a simple on-surface control of the aperture fields for beam
shaping, pointing and scanning.

10.4.1 Sinusoidally Modulated Inductive Surfaces

Radiation by modulated MTS stems from the interaction of a surface wave (SW),
which can be launched using a coaxial pin, and a modulated inductive plane,
realized in the microwave range by means of printed subwavelength patches [64,
65]. The interaction between the SW and the periodic modulation makes the (−1)
indexed Floquet mode enter the visible region so it becomes a radiating mode. To
illustrate the concept, let us consider a scalar one-dimensional modulation at the
z = 0 plane, invariant in the y direction, and with propagation along x. The sinu-
soidally modulated inductance can be written as
  
2p
Xs ð xÞ ¼ X 1 þ M sin x ð10:10Þ
d

where X  is the average surface reactance, M is the modulation factor and d is the
period of the modulation. This case was studied by Oliner and Hessel in [67].
The solution to the periodic problem consists in an infinite series of Floquet modes,
the transverse-to-z wavenumber for the n-indexed mode is

kt;n ¼ bsw þ bD  ja þ 2pn=d ð10:11Þ

where bD and a are perturbations in the phase and attenuation constants, respec-
 M and d. In turn, bsw stands for the unperturbed value of
tively, which depend on X,
10 Terahertz Antennas and Feeds 369


kt (M = 0), obtained by imposing the transverse resonance condition between j X
and the free space TM impedance

j X þ Z0;TM ¼ 0 ð10:12Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Given that we are looking for SWs bsw [ k and Z0;TM ¼ jf b2sw  k2 =k, with
k and f being the free space wavenumber and impedance, respectively. Solving for
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bsw in (10.12) gives bsw ¼ k 1 þ ðX=fÞ2 . When j<fkt;n gj\k, the corresponding
mode enters in the visible region of the spectrum, and it can be identified as a leaky
wave solution. The n = −1 mode is the dominant leaky-mode, and it radiates in a
direction given by

bsw þ bD  2p=d ¼ k sin h0 ð10:13Þ

where h0 is the angle with respect to the z axis. Consequently, for a single beam
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 and d can be related using bsw ¼ k 1 þ ðX=fÞ2 in
directed at an angle h0 , X
(10.13). For instance, given a free space wavelength k, one can generate a single
forward beam using a period
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
2
d ¼ k= 1 þ ðX=fÞ  sin h0 ð10:14Þ

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
when X=f [ 4 sin h0 ð1 þ sin h0 Þ. In the derivation of (10.14), the effect of bD
(\\bsw ) has been neglected.

10.4.2 Constituent Elements for THz Metasurfaces

As mentioned in Sect. 10.4.1, one can implement modulated MTSs by artificial

surfaces realized with subwavelength patches, which offer a good performance
below 100 GHz. However, there are some aspects which hinder their use in tera-
hertz (THz) antennas. First, the losses in the dielectric become, in most applications,
a limiting factor, so all-metal structures are preferred. Second, it is challenging to
combine a coaxial SW launcher with the rectangular waveguide (RW) output of
solid-state frequency-multiplied continuous-wave sources, which are the most
commonly used sources at THz frequencies [68]. One could, alternatively, use a
graphene sheet transferred onto a back-metalized substrate to realize the modulated
impedance plane [69]. Such modulation is achieved via graphene’s field effect by
applying adequate DC bias voltages to gating pads beneath the graphene sheet.
However, the bias network renders the structure more complex, and graphene
cannot be used to realize anisotropic inductance surfaces, which limits the degree of
control on the aperture fields.
370 G. Chattopadhyay et al.

In the remainder of this section, we will show the suitability of silicon micro-
machined elements for modulated MTS antennas in the THz range. Among the
existing techniques, deep reactive-ion etching (DRIE) is particularly well-adapted
for micromachining subwavelength features at submillimeter wavelengths. Since it
is based on etching, one may argue that it is challenging to maintain straight
sidewalls and uniform depth across the wafer for each depth step. Nevertheless,
these drawbacks can be overcome by extensive process development [33, 70] and a
thorough design. After carrying out the desired number of etching steps, 2 lm of
gold is deposited by sputtering to metallize the structure.
The proposed MTS consists in an array of metallic cylinders of elliptical cross
section (see insets in Fig. 10.33, placed on a ground plane and arranged in a square
lattice with subwavelength unit cell size [71, 72]. Such a structure, which resembles
a Fakir’s bed of nails, has been used in the past for synthesizing artificial surfaces
of inductive nature [73], and can be easily fabricated using DRIE. The surface
reactance corresponding to an elliptical cylinder is, in general, a tensor, which
depends on the transverse wave vector bsw ¼ bxsw ^x þ bysw^y and relates the transverse
electric and magnetic fields (evaluated at the upper interface) as
~
Et jz¼0 þ ¼ j XS
^z  Ht jz¼0 þ , where XS ¼ ½½Xxx Xyx T ½Xxy Xyy T  is defined in
Cartesian coordinates, and ^z is the normal to the MTS plane.
The elliptical cylinder possesses two orthogonal symmetry axes, which principal
directions can be identified, in the low frequency regime, with the principal axes of
the dispersion ellipse. By dispersion ellipse we denote the curve that, for a fixed
frequency, provides the bxsw  bysw pairs corresponding to waves propagating at
such a frequency. The surface reactance tensor at a given frequency can be retrieved
from its isofrequency dispersion ellipse by fitting, in the least square sense, the
simulated data with [74]
   
1 þ Xxx Xyy  Xxy Xyx kkz þ Xxy þ Xyx bxsw bysw
h  i ð10:15Þ
þ Xxx þ Xyy k 2  Xxx ðbysw Þ2  Xyy ðbxsw Þ2 ¼ 0

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where kz ¼ ðbxsw Þ2 þ ðbysw Þ2  k2 . Figure 10.33a shows the isofrequency disper-
sion ellipses for an elliptical cylinder with minor axis equal to 30 lm and axial ratio
equal to 4, in a square unit cell with side a = 138.5 lm. This case presents at
300 GHz an artificial tensor surface with XS ¼ f½½1:04; 0:46T ½0:46; 1:03T . By
changing the height, orientation, and axial ratio of the elliptical cylinders, each
component of the tensor will undergo a different modulation. This feature is
exploited to get the required degrees of freedom to control the aperture fields [65,
75].
In the limiting case of a circular cross section, for electrically small period, the
isofrequency dispersion ellipse becomes a circumference and the surface reactance
is scalar. Figure 10.33b shows the dispersion curves (obtained with the eigenmode
solver of a commercial software) [76] of the fundamental TM-SW modes
10 Terahertz Antennas and Feeds 371

Fig. 10.33 a Isofrequency dispersion contours for the unit cell depicted in the insets. The unit
cell’s side is a = 138.5 lm, the height of the cylinder is h = 180 lm, and the elliptical
cross-section has minor axis equal to 30 lm and axial ratio equal to 4. b Dispersion curves of the
fundamental TM-SW modes propagating on a set of bed of nails. Each curve corresponds to an
infinite array of cylinders of height h, radius r = 17.5 lm, and arranged on a square lattice with
side a = 147 lm. Results obtained using the eigenmode solver in [76]. c Equivalent surface
reactance of an infinite array of cylinders on a ground plane and arranged on a square lattice at
300 GHz, as a function of the height h of the cylinders. The radius of the cylinders is r = 17.5 lm,
and the side of the unit cell is a = 147 lm. The solid line represents the values calculated solving
 after using in (10.12) the bsw obtained for each h with a full-wave eigenmode solver
(10.12) for X,
[76]

propagating on a set of bed of nails. Each curve corresponds to an infinite array of

cylinders with a given height h, radius r = 17.5 lm, and arranged on a square
lattice with side a. The period d has been computed at 300 GHz using (10.14) for
an average impedance X ¼ 0:7f and a forward beam pointing at h0 ¼ 5o , which
for N ¼ 6 implies a ¼ 147lm. The surface reactances retrieved through (10.12)
are shown in Fig. 10.33c. Unfortunately, long cylindrical rods are difficult to realize
with DRIE while keeping a good precision and a simple process. Therefore, upon
choosing X  and N, one has to verify that the required heights satisfy h=r  10
before proceeding with the design. Modulated surface reactances can be realized
using the data in Fig. 10.32c under a local periodicity assumption.
372 G. Chattopadhyay et al.

10.4.3 Design Example

The design below consists in a spirally modulated MTS antenna. The spiral has
been designed at 300 GHz and it provides a broadside pencil beam with circular
polarization. The synthesized surface reactance is
  
2p
Xs ð xÞ ¼ X 1 þ M sin q/ ð10:16Þ
d

where q and / represent the position on the MTS plane in polar coordinates. The
 ¼ 0:7f, M = 0.65, N = 6
reader is referred to [65] for further details. In our case, X
and a ¼ 138:5 lm. The structure is shown in Fig. 10.34a, where each color rep-
resents different height of the metallic cylinders. Figure 10.34b shows a SEM
picture of a portion of the bed of nails fabricated using DRIE. The structure is fed
with an overmoded circular waveguide with only the TM01 mode propagating, this
mode provides an efficient excitation of the TM surface wave on the MTS plane,
while avoiding the use of a coaxial feed. The feeder (not shown here) is placed
underneath the MTS and it transforms the TE10 mode in the input RW to the TM01
in the CW. The magnitude of the S11 in dB obtained with the aforementioned
feeding structure is shown in Fig. 10.34c. The far-field gain patterns have been
computed with HFSS [76] and they are shown in Fig. 10.34d for two orthogonal
planes at 300 GHz. A similar performance has been verified for the patterns in the
rest of the 290–315 GHz band.

10.5 CMOS Antennas

Complementary metal oxide semiconductor (CMOS) technology has recently

made tremendous improvement in millimeter-wave capabilities because of the
aggressive technology scaling during the past decade. As a demonstration of this
rapid progress, full transceivers have been demonstrated in the millimeter-wave
band [77, 78]. More recently, the NASA’s Jet Propulsion Laboratory has exten-
sively studied the applicability of system on chip communication, radar, and
spectrometer terahertz solutions to the exploration of planets and moons in the outer
solar system.
CMOS technology offers a pathway to improve the payload size, weight, and
power of instruments and telecom systems on spacecraft directly through higher
levels of integration and added calibration functionality. In Fig. 10.35, a mm-wave
receiver of an existing radiometer/spectrometer pixel is compared to a receiver fully
implemented in System on Chip (SoC) technology.
In the past few years, NASA’s Jet Propulsion Laboratory in collaboration with
UCLA, has been exploring the use of CMOS SoC in millimeter-wave and terahertz
instruments [79–83]. The severe metal density rules imposed to the CMOS
10 Terahertz Antennas and Feeds 373

Fig. 10.34 a Simulated spiral modulated MTS antenna. b SEM picture of a bed of nails fabricated
by means of DRIE. c Magnitude of the S11 in dB at the input RW port of the antenna.
d Right-handed circular polarized (RHCP) and left-handed circular polarized (LHCP) gain patterns
at 300 GHz plotted with solid and dashed lines, respectively, on two orthogonal cuts

Fig. 10.35 Comparison of size between existing mm-wave receiver of an existing

radiometer/spectrometer pixel and a receiver implemented in System on Chip (SoC) technology

fabrication process, making efficient on-chip antennas almost impossible to

implement, can be overcome using off-chip antennas with beam-lead intercon-
nections [79]. The beam-lead interconnection appears to be an excellent intercon-
nection technique for CMOS circuits at terahertz frequencies that can be applied to
374 G. Chattopadhyay et al.

Fig. 10.36 Example of beam-lead interconnection for a ground-signal (GS) RFIC and
b ground-signal-ground (GSG) pad RFIC

ground-signal (GS) pads (i.e., differential line) and GSG pads (CPW line) as shown
in Fig. 10.36.
Several beam-lead antennas designed as part of CMOS SoCs are discussed here.
The antenna shown in Fig. 10.37 consists of a micromachined silicon dipole
antenna interconnected using the beam-lead technique. A dipole was selected as it
naturally couples to the differential output of the CMOS device. The dipole is
printed on a 1-lm thick Silicon Nitride membrane (Si3N4 with er = 7.5) [79]. The
hatched area in Fig. 10.37 is where the 300-lm-thick silicon wafer material is
present below the Si3N4, while the light gray area surrounding the dipole is only
made of the thin Si3N4 membrane. The cavity thickness was chosen to align the
antenna and the chip at the same height. As the distance between the radiating
element and the PCB ground plane is about 300 lm, it increases the antenna
directivity. In addition, two metal bars are employed in order to reduce backward
radiations toward the chip that is made of lossy dielectric.
The antenna input impedance was matched to ZAnt = Z0* = 29 + j38 Ω, where
Z0 is the impedance of the chip. The fabricated antenna and its CMOS chip are
shown in Fig. 10.38. The antenna was measured at 154 GHz (i.e., center frequency
of the CMOS chip). This is an active measurement as the antenna is measured with
its Tx CMOS chip. A standard pyramidal horn antenna was connected on one head
of an ABmm measurement system and the CMOS chip prototype is connected to a
DSP lock-in amplifier (Model SR830 from SRS, Sunnyvale, CA). Two rotating
stages were employed to move the antenna in elevation and azimuth planes to scan
the far-field. The measured and calculated radiation patterns are in good agreement
as shown in Fig. 10.39. Measurement has shown that the antenna can reach up to
70% efficiency near the lossy CMOS transmitter chip. The overall transmitter
operating at 154 GHz shows excellent performance as its effective isotropic radi-
ated power (EIRP) and total radiated power (TRP) reach 20.6 dBm and 10.2 dBm,
respectively [79].
10 Terahertz Antennas and Feeds 375

Fig. 10.37 Dipole antenna integrated with the Tx CMOS chip [79]. (Top cross section. Bottom
top view

This antenna was also used as a part of a 65 nm CMOS 140 GHz, 27.3 dBm
EIRP transmit array. This transmit-phased array offers beam steering capability
[80]. It uses local phase-locked loop (PLL) reference generation system. Unlike
traditional CMOS phased arrays, this enables the array to be formed over multiple
chips while avoiding the challenges of distributing mm-wave signals between them.
A 2  4 multi-chip array prototype was fabricated and tested using the beam-lead
dipole membrane antenna.
The antenna shown in Fig. 10.40 was built on a high-resistivity silicon substrate
for a full CMOS SoC navigation processor operating at 94 GHz. This navigation
processor can be used to safely guide unmanned aerial vehicles (UAVs) and
unmanned ground vehicles (UGVs) through cluttered and urban environments
while consuming very little payload [83]. The navigation system defines multiple
pathways using mm-wave base-stations called path generators and then uses a
single CMOS SoC containing a receiver, ADC and an FFT processor to detect and
navigate these pathways. The demonstrated confined pathway SoC (CP-SoC)
occupies 5.4 mm2 of silicon area in 65 nm technology, and consumes only
199 mW, making it suitable for lightweight payloads associated with UAVs and
UGVs.
376 G. Chattopadhyay et al.

Fig. 10.38 a Die photo of a proposed SoC Transmitter operating at 150 GHz [79]. b Die photo of
a 140 GHz transmit array for highly scalable multi-chip phase arrays [80]

(a) (b)
30 0

-1
20
-2

-3
10
Theta (degree)

-4

0 -5

-6
-10
-7

-8
-20
-9

-30 -10
-30 -20 -10 0 10 20 30
Phi (degree)

Fig. 10.39 a Measured and b simulated normalized 2D radiation patterns of the CMOS off-chip
dipole antenna at 154 GHz (co-polarization)
10 Terahertz Antennas and Feeds 377

Fig. 10.40 Die photo of the proposed SoC navigation processor combined with the external
antenna module (on a HiRes Si substrate) used to capture incoming radiation [83]

10.6 Slot Antennas in BICEP2

The measurement of the polarization of the cosmic microwave background (CMB)

is the holy grail of present day cosmology. CMB contains the most valuable relic of
the early universe [84]. The details of the structure and evolution of the universe are
believed to be encoded in the anisotropy of the CMB. It is believed that the direct
evidence of the inflationary epoch—the rapid expansion of the universe after the
Big Bang—lies in the traces of gravitational waves in the CMB, which manifests in
the temperature anisotropy and polarization of the CMB. If inflationary gravita-
tional waves do indeed exist, they will be the oldest relic of the universe, created
500,000 years before the CMB was emitted. The measurement of the CMB
polarization—the B-mode polarization in particular—will be the indirect proof of
the existence of the inflationary gravitational waves and will give insight into the
epoch of inflation.
CMB polarization detection instruments require large focal planes with thou-
sands of detectors to have large throughput improving overall sensitivity of the
instrument. To reduce stray radiation pickup, the field of view of these detectors
needs to be restricted. The purpose of BICEP (i.e., Background Imaging of Cosmic
Extragalactic Polarization) is to measure the polarization of the CMB radiation; in
particular, to measure the B-mode of the CMB. In BICEP1, the first generation of
the BICEP instruments, the collimation of the incoming beam is achieved using
metal feed horns (Fig. 10.41a). Even though feed horns have excellent perfor-
mance, their mass, size, and expense make them unsuitable for large arrays with
hundreds and thousands of pixels.
A monolithic array of antenna-coupled detectors on a planar substrate, which can
be fabricated with photolithographic techniques, would solve the aforementioned
problem. However, most planar antenna designs use thin dielectric substrate to
avoid excitation of surface waves, and they produce broad beam patterns. Most
378 G. Chattopadhyay et al.

Fig. 10.41 a BICEP1 antenna using feed horns with different apertures. b BICEP2 antenna using
planar dual-polarized phased arrays

often, they require substrate lenses or micromachined horns for efficient coupling to
the telescope optics. As a significant improvement, the second generation instru-
ment BICEP2 uses a novel dual-polarization planar slot antenna phased array which
produces quite a narrow beam with no additional optical coupling elements. The
output from the antenna array is two superconducting thin-film microstrip lines, one
for each polarization, which can be efficiently coupled to bolometer detectors [6] to
produce a single pixel in an imaging focal plane array. Figure 10.41 illustrates the
BICEP1 and BICEP2 instruments.
The “T” shaped dual-polarized antenna structure chosen for this work is ideal for
low cross-polarization performance. The T-slots were arrayed in a compact archi-
tecture to form the slot array antenna.
The antenna is fabricated on a high dielectric silicon substrate (er = 10.8) with
thickness between ¾ kg and kg. The slots are illuminated through substrate to take
advantage of the stronger antenna response on the dielectric side. A quarter
wavelength thick quartz, whose dielectric constant is optimum for silicon
anti-reflection coating, is used on the silicon substrate to minimize reflection at the
air–substrate interface.
Due to the thick dielectric substrate, excitation of the lossy grating lobes for the
array is a concern. The antenna slots in each detector must be spaced to Nyquist
sample the focal plane surface to avoid generating the grating lobes. The radiation
pattern of each axis of an array is calculated from the N elements per linear
dimension spaced at distance S as follows:

X
ðN1Þ=2 pffiffiffi pffiffiffiffi
j2p
mS er
sinðhÞ sinðNpS er sinðhÞ=k0 Þ
AðhÞ ¼ e k0
¼ pffiffiffiffi
m¼ðN1Þ=2
sinðpS er sinðhÞ=k0 Þ

where ɛr the relative permittivity of the surrounding medium and k0 is the free space
wavelength.
10 Terahertz Antennas and Feeds 379

The key features of the design are the slot length L, slot width W, and the slot
separation S. Due to the relatively thick substrate with no coupling lenses, the
parameter to optimize not to excite grating lobes within the frequency band of
interest, is the slot separation distance S, which must follow:
 
k0;min 1
S  pffiffiffiffi 1 
er N

where k0,min is minimum wavelength of operation. With an operating bandwidth of

150–180 GHz (i.e., k0,min = 1.7 mm), the spacing must satisfy S  460 lm.
Another key design parameter is the length of the slots. Since the design is
constrained by the slot separation distance S, the length of the slots that can used is
also constrained. Ideally, for broadband performance, the slot antennas are used at
their second resonance where the real part of the slot impedance is relatively
constant and its value can be optimized between 20 and 60 Ω, when fed at the
center of the slot. However, for the slots to operate in the second resonance, the
length of the slots needs to be more than 0.3 k0. The grating lobe constraint makes it
impossible to operate the slots in their second resonance. Therefore, the slots for the
BICEP2 antenna are operational in their first resonance. Unfortunately, the impe-
dance bandwidth of a single slot operating in the first resonance is relatively narrow
and the real part of the impedance is very high, in the 100–150 Ω range, depending
on the specific design (caused by a current node at the slot center as shown in
Fig. 10.42). The impedance can be optimized to somewhat lower value (50–60 Ω)
in a “T” slot array configuration. Since the impedance of the detectors used for the
BICEP2 instrument is relatively low, in the 20–25 Ω range, it needs a matching
circuit to match the detector impedance with the antenna impedance when the slots
are fed at the center. One way to overcome this is to feed the slots off-center. The
real part of the impedance of a slot antenna operating at the first resonance
decreases as the feed is moved away from the center, toward the edge of the slot.
However, that breaks the symmetry of the slot array, leading to cross-polar cou-
pling, and poor cross-polarization performance. The conflicting requirements of low
value for the real part of the antenna at first resonance and low cross-polarization
performance was addressed by feeding the slots at two points, toward the edge of
the slots. This made the antenna a symmetric structure and at the same time pro-
viding low value for the real part of the antenna impedance (*22 Ω), as shown in
Fig. 10.43c.

Fig. 10.42 Current and

voltage distribution on a
half-wavelength slot. The
impedance Z = V/I is
maximum in the center of the
slot. The impedance reduces
as the feeding point moves
away from the center
380 G. Chattopadhyay et al.

(a) (b)
(c)

Fig. 10.43 a Layout of BICEP2 slot array. b SEM micrograph of the fabricated slot array
showing the 1-lm-thick microstrip coupling line. c SEM micrograph of microstrip crossover and
shunt capacitor at a sub-antenna slot

Provided that impedances between the microstrip line and slot are well matched,
the coupling effectively transmutes the electric fields across the slot into fields. For
the microstrip feed lines to match the antenna resistance, the thinnest lines that can
be reproducibly fabricated is required (i.e., 1 lm as shown in Fig. 10.43b). The
feed points also have a 5 Ω inductive reactance. It is tuned away using series
capacitors that shunt current to ground (Fig. 10.43c). This inductive reactance is
relative stable across the frequency and that is why a simple capacitor is enough.
The waves from the slots coherently sum in a corporate microstrip feed network.
The network first combines incoming signals across rows and then sums the signals
from each row in a column tree at the side of the detector. The signals sum in
microstrip tee-junctions with proper impedance matching.
Each pixel of the phased array consists of 10  10 slot arrays for each polar-
ization, 100 horizontal slots and 100 vertical slots, over a 10 mm  10 mm
area (see Fig. 10.44). The slots are excited inphase with superconducting microstrip
transmission lines. There are two transition edge sensor (TES) detectors for each
pixel, one for each polarization. The overall slot array is sized to match the f/2.2
camera optics for the BICEP2 instrument.
For characterization of the dual-polarization slot array, the TES outputs were
read out by SQUIDs. To The detector system has an integrated microstrip bandpass
filter to define the passband of the instrument. Two separate chips were fabricated
and tested, one with the bandpass filter and the other without it. This was done to
characterize both the filter and slot antenna designs independently. FTS response of
the antenna with integrated bandpass filter and the measured co-polarized beam
pattern of the antenna are shown in Fig. 10.45.
The slot array is currently integrated into the BICEP2 instrument and is col-
lecting data every day in search of answers about the creation of the Universe, and
all these are dictated by the performance of the dual-polarized “T” slot phased array
antenna.
10 Terahertz Antennas and Feeds 381

Fig. 10.44 Schematic diagram of the slot antenna array showing the slots for one pixel.
a Cross-section of the pixel geometry. b Slots position for one pixel. Each pixel is
10 mm  10 mm, and there are 10  10 pixels in the focal plane

Fig. 10.45 a Measured co-polarized beam pattern. b FTS response of the 145 GHz slot antenna
array

10.7 Conclusion

This chapter discusses the dominant types of antenna used at terahertz frequencies.
For waveguide-based systems where polarization content is not a significant con-
cern, diagonal horns are commercially available, simple to integrate into custom
designs and have sufficient Gaussicity. In planar systems, a substrate integrated lens
fed by a double slot or double-dipole the common choice because of the direct
integration, high efficiencies, and relatively simple assembly required.
Because many applications at terahertz are for science instruments, higher per-
formance is often demanded. Spline-profiled and multi-flare angled horns provide
excellent performance while still being manufacturable with standard CNC milling
processes. Corrugated horns are more challenging to produce, but have been
demonstrated up to 340 GHz, with higher frequencies possible. The leaky wave
coupled lens designs demonstrate a step forward in the lens-coupled designs by
reducing the aperture size of the lens. This enables batch fabrication of the lenses
382 G. Chattopadhyay et al.

and higher array densities. Finally, the slot array described in the final section
shows a high gain, highly polarization sensitive design, enabled by superconducting
transmission lines.

Acknowledgments Part of this work was supported by NASA Jet Propulsion Laboratory (JPL),
CA. and “National Aeronautics and Space Administration” (NASA). Part of this work was also
supported by a Marie Curie International Outgoing Fellowship within the 7th European
Community Framework Programme.

References

1. A. Love, Electromagnetic Horn Antennas (IEEE Press, 1976)

2. P. Clarricoats, A. Olver, Corrugated Horns for Microwave Antennas (P. Peregrinus, 1984)
3. A.D. Olver, P.J. Clarricoats, A.A. Kishk, L. Shafai, Microwave Horns and Feeds (Institution
of Engineering and Technology, 1994)
4. T.S. Bird, C. Granet, Profiled horns and feeds, in Handbook of Reflector Antennas and Feed
Systems: Volume 2—Feed Systems, ed. by L. Shafai et al. (Artech House, 2013)
5. P.F. Goldsmith, Quasioptical Systems: Gaussian Beam, Quasioptical Propagation, and
Applications (IEEE Press/Chapman & Hall, New York, 1998)
6. A.W. Love The diagonal horn antenna. Microw. J. V, 117–122 (1962)
7. J.F. Johansson, N.D. Whyborn, The diagonal horn as a sub-millimeter wave antenna. IEEE
Trans. Microw. Theory Tech. 40(5), 795–800 (1992)
8. T. Reck, C. Jung-Kubiak, J. Siles, C. Lee, R. Lin, G. Chattopadhyay, I. Mehdi, K. Cooper, A
silicon micromachined eight-pixel transceiver array for submillimeter-wave radar. IEEE
Trans. THz Sci. Technol. 5(2), 197–206 (2015)
9. P.D. Potter, A new horn antenna with suppressed sidelobes and equal beamwidths. NASA
Technical Documents, Technical report No. 32-354, 25 Feb 1963
10. P. Kittara, A. Jiralucksanawong, G. Yassin, S. Wangsuya, J. Leech, The design of potter
horns for THz applications using a genetic algorithm. Int. J. Infrared Milli. Waves 28(12),
1103–1114 (2007)
11. H.M. Pickett, J.C. Hardy, J. Farhoomand, Characterization of a dual-mode horn for
submillimeter wavelengths. IEEE Trans. Microw. Theory Tech. 32(8), 936–937 (1984)
12. J.F. Johansson, A Gauss-Laguerre analysis of the dual-mode (‘Potter’) horn, in 4th
International Symposium on Space Terahertz Technology, Apr 1993, p. 134
13. B.M. Thomas, Design of corrugated conical horns. IEEE Trans. Antenna Propag. 26(2), 367–
372 (1978)
14. C. Granet, G.L. James, Design of corrugated horns: a primer. IEEE Antennas Propag. Mag.
47(2), 76–84 (2005)
15. G.L. James, Analysis and design of TE11-to-HE11 corrugated cylindrical waveguide mode
converters. IEEE Trans. Microw. Theory Tech. 29(10), 1059–1066 (1981)
16. G. James, Design of wide-band compact corrugated horns. IEEE Trans. Antennas Propag. 32
(10), 1134–1138 (1984)
17. G.L. James, B.M. Thomas, TE11 to HE11 cylindrical waveguide mode converters using
ring-loaded slots. IEEE Trans. Microw. Theory Tech. 30(3), 278–285 (1982)
18. X. Zhang, Design of conical corrugated feed horns for wide-band high-frequency
applications. IEEE Trans. Microw. Theory Tech. 41(8), 1263–1274 (1993)
19. A.D. Olver, J. Xiang, Design of profiled corrugated horns. IEEE Trans. Antennas Propag. 36
(7), 936–940 (1988)
20. R. Gonzalo, J. Teniente, C. del Rio, Improved radiation pattern performance of Gaussian
profiled horn antennas. IEEE Trans. Antennas Propag. 50(11), 1505–1513 (2002)
10 Terahertz Antennas and Feeds 383

21. G.G. Gentili, R. Nesti, G. Pelosi, V. Natale, Compact dual-profile corrugated circular
waveguide horn. Electron. Lett. 36(6), 486–487 (2000)
22. B. Maffei et al., Planck pre-launch status: HFI beam expectations from the optical optimisation of
the focal plane. A & A 520, A12 (2010)
23. J.E. McKay, D.A. Robertson, P.A.S. Cruickshank, R.I. Hunter, D.R. Bolton, R.J. Wylde, G.
M. Smith, Compact wideband corrugated feedhorns with ultra-low sidelobes for very high
performance antennas and quasi-optical systems. IEEE Trans. Antennas Propag. 61(4), 1714–
1721 (2013)
24. R.W. Haas, D. Brest, H. Mueggenburg, L. Lang, D. Heimlich, Fabrication and performance of
MMW and SMMW platelet horn arrays. Int. J. Infrared Milli. Waves 14(11), 2289–2294
(1993)
25. M. Kangas, K. Copsey, P. Lubin, A modular 100-GHz high-gain scalar corrugated nonbonded
platelet antenna. IEEE Antennas Wirel. Propag. Lett. 4, 89–92 (2005)
26. M.M. Kangas, M. Ansmann, B. Horgan, N. Lemaster, R. Leonardi, A. Levy, P. Lubin,
J. Marvil, P. McCreary, T. Villela, A 31 pixel flared 100-GHz high-gain scalar corrugated
nonbonded platelet antenna array. IEEE Antennas Wirel. Propag. Lett. 4, 245–248 (2005)
27. L. Lucci, R. Nesti, G. Pelosi, S. Selleri, A stackable constant-width corrugated horn design for
high-performance and low-cost feed arrays at millimeter wavelengths. IEEE Antennas Wirel.
Propag. Lett. 11, 1162–1165 (2012)
28. J.W. Britton, J.P. Nibarger, K.W. Yoon, J.A. Beall, D. Becker, H.-M. Cho, G.C. Hilton,
J. Hubmayr, M.D. Niemack, K.D. Irwin, Corrugated silicon platelet feed horn array for CMB
polarimetry at 150 GHz, in Proceedings of SPIE, vol. 7741, 2010, pp. 77 410T–77 410T-11
29. J.P. Nibarger, J.A. Beall, D. Becker, J. Britton, H.-M. Cho, A. Fox, G.C. Hilton, J. Hubmayr, D.
Li, J. McMahon, M.D. Niemack, K.D. Irwin, J. Van Lanen, K.W. Yoon, An 84 pixel all-silicon
corrugated feedhorn for CMB measurements. J. Low Temp. Phys. 167(3), 522–527 (2012)
30. C. Lee, G. Chattopadhyay, E. Decrossas, A. Peralta, I. Mehdi, C.A. Leal-Sevillano, M.A.
Pino, N. Llombart, Terahertz antenna arrays with silicon micromachined-based microlens
antenna and corrugated horns, International Workshop on Antenna Technology (iWAT), Mar
2015, pp. 70–73
31. E. Decrossas, T. Reck, C. Lee, C. Jung-Kubiak, I. Mehdi, G. Chattopadhyay, “Evaluation of
3d printing technology for corrugated horn antenna manufacturing,” in IEEE Int. Symp
Electromagn. Compat. 2, (2016)
32. E. Decrossas, T. Reck, C. Lee, C. Jung-Kubiak, I. Mehdi, and G. Chattopadhyay,
“Development of w-band horn antennas using 3d printing technologies,” in IEEE Int.
Symp. Antennas Propagation Soc. 2, (2016)
33. T. Reck, C. Jung-Kubiak, J. Gill, G. Chattopadhyay, Measurement of silicon micromachined
waveguide components at 500–750 GHz. IEEE Trans. THz Sci. Technol. 4(1), 33–38 (2014)
34. J. Leech, B.K. Tan, G. Yassin, P. Kittara, S. Wangsuya, Experimental Investigation of a
low-cost, high performance focal-plane horn array. IEEE Trans. Terahertz Sci. Technol. 2(1),
61–70 (2012)
35. N. Chahat, T. Reck, C. Jung-Kubiak, T. Nguyen, R. Sauleau, G. Chattopadhyay, 1.9 THz
multi-flare angle horn optimization for space instruments. IEEE Trans. Terahertz Sci.
Technol. 5(6), 914–921 (2015)
36. N. Chahat, R.E. Hodges, J. Sauder, M. Thomson, E. Peral, Y. Rahmat-Samii, CubeSat
Deployable Ka-band mesh reflector antenna development for Earth science missions. IEEE
Trans. Antennas Propag. 64(6), 2083–2093 (2016)
37. C. Granet, G.L. James, R. Bolton, G. Moorey, A smooth-walled spline-profile horn as an
alternative to the corrugated horn for wide band millimeter-wave applications. IEEE Trans.
Antennas Propag. 52(3), 848–854 (2004)
38. A. Rolland, M. Ettorre, M. Drissi, L. Le Coq, R. Sauleau, Optimization of reduced-size
smooth-walled conical horns using BoR-FDTD and genetic algorithm. IEEE Trans. Antennas
Propag. 58(9), 3094–3100 (2010)
39. T.G. Jurgens, J.G. Blaschak, G.W. Saewert, Bodies of revolution, in Computational
Electrodynamics: The Finite-Difference Time-Domain Method, 3rd edn., ch. 12, ed. by A.
Taflove, S.C. Hagness (Artech House, Boston, MA, 2005), pp. 517–552
384 G. Chattopadhyay et al.

40. M. Celuch, W.K. Gwarek, Industrial design of axisymmetrical devices using a customized
solver from RF to optical frequency bands. IEEE Microw. Mag. 9, 150–159 (2008)
41. G. Godi, R. Sauleau, L. Le Coq, D. Thouroude, Design and optimization of three-dimensional
integrated lens antennas with genetic algorithm. IEEE Trans. Antennas Propag. 55(3), 770–
775 (2007)
42. G.M. Rebeiz, D.P. Kasilingam, Y. Guo, P.A. Stimson, D.B. Rutledge, Monolithic
millimeter-wave two-dimensional horn imaging arrays. IEEE Trans. Antennas Propag. 38
(9), 1473–1482 (1990)
43. G.M. Rebeiz, Millimeter-wave and terahertz integrated circuit antennas. Proc. IEEE 80(11),
1748–1770 (1992)
44. D.F. Filipovic, S.S. Gearhart, G.M. Rebeiz, Double-slot antennas on extended hemispherical
and elliptical silicon dielectric lenses. IEEE Trans. Microw. Theory Tech. 41(10), 1738–1749
(1993)
45. T.H. Buttgenbach, An improved solution for integrated array optics in quasi-optical mm and
submm receivers: the hybrid antenna. IEEE Trans. Microw. Theory Tech. 41(10), 1750–1760
(1993)
46. R. Compton, R. McPhedran, Z. Popovic, G. Rebeiz, P. Tong, D. Rutledge, Bow-tie antennas
on a dielectric half-space: theory and experiment. IEEE Trans. Antennas Propag. 35(6),
622–631 (1987)
47. R. DuHamel, D. Isbell, Broadband logarithmically periodic antenna structures, 1958 IRE
International Convention Record, New York, NY, USA, 1957, pp. 119–128
48. J. Dyson, The equiangular spiral antenna. IRE Trans. Antennas Propag. 7(2), 181–187 (1959)
49. T.H. Buttgenbach, R.E. Miller, M.J. Wengler, D.M. Watson, T.G. Phillips, A broad-band
low-noise SIS receiver for submillimeter astronomy. IEEE Trans. Microw. Theory Tech. 36
(12), 1720–1726 (1988)
50. D.F. Filipovic, G.P. Gauthier, S. Raman, G.M. Rebeiz, Off-axis properties of silicon and
quartz dielectric lens antennas. IEEE Trans. Antennas Propag. 45(5), 760–766 (1997)
51. G. Rebeiz, W. Regehr, D. Rutledge, R. Savage, N. Luhmann, Submillimeter-wave antennas
an thin membranes, in Antennas and Propagation Society International Symposium, 1987,
pp. 1194–1197
52. D. Rutledge, M. Muha, Imaging antenna arrays. IEEE Trans. Antennas Propag. 30(4), 535–
540 (1982)
53. N. Llombart, A. Neto, THz time-domain sensing: the antenna dispersion problem and a
possible solution. IEEE Trans. THz Sci. Technol. 2(4), 416–423 (2012)
54. A. Neto, N. Llombart, J.J.A. Baselmans, A. Baryshev, S.J.C. Yates, Demonstration of the
leaky lens antenna at submillimeter wavelengths. IEEE Trans. THz Sci. Technol. 4(1), 26–32
(2014)
55. N. Llombart, G. Chattopadhyay, A. Skalare, I. Mehdi, Novel terahertz antenna based on a
silicon lens fed by a leaky wave enhanced waveguide. IEEE Trans. Antennas Propag. 59(6),
2160–2168 (2011)
56. N. Llombart et al., Silicon micromachined lens antenna for THz integrated heterodyne arrays.
IEEE Trans. THz Sci. Technol. 3(5), 515–523 (2013)
57. M. Alonso-DelPino et al., Design guidelines for a terahertz silicon micro-lens antenna. IEEE
Antennas Wirel. Propag. Lett. 12, 84–87 (2013)
58. D.R. Jackson, A.A. Oliner, A leaky-wave analysis of the high-gain printed antenna
configuration. IEEE Trans. Antennas Propag. 36(7), 905–910 (1988)
59. A. Neto, UWB, non dispersive radiation from the planarly fed leaky lens antenna—Part 1:
Theory and design. IEEE Trans. Antennas Propag. 58(7), 2238–2247 (2010)
60. N. Llombart, A. Neto, G. Gerini, M. Bonnedal, P. De Maagt, Impact of mutual coupling in
leaky wave enhanced imaging arrays. IEEE Trans. Antennas Propag. 56(4), 1201–1206
(2008)
10 Terahertz Antennas and Feeds 385

61. M. Qiu, G.V. Eleftheriades, M. Hickey, A reduced surface-wave twin arc-slot antenna
element on electrically thick substrates, in Proceedings of IEEE AP-S Society International
Symposium, Boston, MA, USA, 2001, vol. 3, pp. 268–271
62. C. Lee, G. Chattopadhyay, K. Cooper, I. Mehdi, Curvature control of silicon microlens for
THz dielectric antenna, in 37th International Conference on Infrared Milli. THz Waves
(IRMMW-THz), Wollongong, NSW, 2012, pp. 1–2
63. C. Lee, G. Chattopadhyay, M. Alonso-DelPino, N. Llombart, “6.4 mm Diameter silicon
micromachined lens for THz dielectric antenna,” 39th Int. Conf. Infrared Millimeter, and THz
waves (IRMMW-THz), Tucson, AZ, 2014, pp. 1–1
64. B. Fong, J. Colburn, J. Ottusch, J. Visher, D. Sievenpiper, Scalar and tensor holographic
artificial impedance surfaces. IEEE Trans. Antennas Propag. 58(10), 3212–3221 (2010)
65. G. Minatti, M. Faenzi, E. Martini, F. Caminita, P. De Vita, D. González-Ovejero, M.
Sabbadini, S. Maci, Modulated metasurface antennas for space: synthesis, analysis and
realizations. IEEE Trans. Antennas Propag. 63(4), 1288–1300 (2015)
66. D. Sievenpiper, J. Colburn, B. Fong, J. Ottusch, J. Visher, Holographic artificial impedance
surfaces for conformal antennas, in Proceedings of IEEE AP-S Society International
Symposium, vol. 1B, 2005, vol. 1B, pp. 256–259
67. A. Oliner, A. Hessel, Guided waves on sinusoidally-modulated reactance surfaces. IRE Trans.
Antennas Propag. 7(5), 201–208 (1959)
68. G. Chattopadhyay, Technology, capabilities, and performance of low power terahertz sources.
IEEE Trans. THz Sci. Technol. 1(1), 33–53 (2011)
69. M. Esquius-Morote, J. Gomez-Diaz, J. Perruisseau-Carrier, Sinusoidally modulated graphene
leaky-wave antenna for electronic beamscanning at THz. IEEE Trans. THz Sci. Technol. 4(1),
116–122 (2014)
70. C. Jung, B. Thomas, C. Lee, A. Peralta, J. Gill, K. Cooper, G. Chattopadhyay, E. Schlecht, R.
Lin, I. Mehdi, Compact submillimeter-wave receivers made with semiconductor nanofabri-
cation technologies, in Proceedings of IEEE International Microwave Symposium (IMS), June
2011, pp. 1–4
71. D. González-Ovejero, T.J. Reck, C.D. Jung-Kubiak, M. Alonso-DelPino, G. Chattopadhyay,
Silicon micromachined modulated metasurface antennas in the terahertz range, in
Proceedings of 10th European Conference on Antennas and Propagation, 2016, pp. 1–4
72. D. González-Ovejero, T.J. Reck, C.D. Jung-Kubiak, M. Alonso-DelPino, G. Chattopadhyay,
A class of silicon micromachined metasurface for the design of high-gain terahertz antennas,
in Proceedings of IEEE AP-S Society International Symposium, 2016, pp. 1–4
73. R. King, D.V. Thiel, K. Park, The synthesis of surface reactance using an artificial dielectric.
IEEE Trans. Antennas Propag. 31(3), 471–476 (1983)
74. H. Bilow, Guided waves on a planar tensor impedance surface. IEEE Trans. Antennas Propag.
51(10), 2788–2792 (2003)
75. D. González-Ovejero, S. Maci, Gaussian ring basis functions for the analysis of modulated
metasurface antennas. IEEE Trans. Antennas Propag. 63(9), 3982–3993 (2015)
76. ANSYS Inc., HFSS, Version 15, Pittsburgh, PA, 2015
77. A. Tang et al., A CMOS D-band transmitter with IF envelope feed-forward pre-distortion and
injection locked frequency tripling synthesizer. IEEE Trans. Microw. Theory Tech. 60(12),
4129–4137 (2012)
78. A. Tang et al., A low overhead self-healing embedded system for ensuring high performance
yield and long-term sustainability of a 60 GHz 4 Gb/s Radio-on-a-Chip, in IEEE
International Solid-State Circuits Conference Digest of Technical Papers, Feb 2012,
pp. 316–318
79. N. Chahat, A. Tang, C. Lee, R. Sauleau, G. Chattopadhyay, Efficient CMOS systems with
beam–lead interconnects for space instruments. IEEE Trans. Terahertz Sci. Technol. 5(4),
637–644 (2015)
80. A. Tang, N. Chahat et al., A 65 nm CMOS 140 GHz 27.3 dBm EIRP transmit array with
membrane antenna for highly scalable multi-chip phase arrays, in Proc. IEEE International
Microwave Symposium (IMS), Tampa, FL, 2014
386 G. Chattopadhyay et al.

81. A. Tang, N. Chahat, CMOS mm-wave transceiver techniques beyond 50 GHz, in

Proceedings of IEEE Global Conference on Signal and Information Processing
(GlobalSIP), Austin, TX, 2013, pp. 715–718
82. A. Tang et al., CMOS (Sub)-mm-wave system-on-chip for exploration of deep space and
outer planetary systems, in Proceedings of IEEE Custom Integrated Circuits Conference, San
Jose, CA, 2014
83. A. Tang et al., A 95 GHz centimeter scale precision confined pathway system-on-chip
navigation processor for autonomous vehicles in 65 nm CMOS, in Proceedings of IEEE
International Microwave Symposium (IMS), Phoenix, AZ, 2015, pp. 1–3
84. R.R. Cladwell, M. Kamionkowski, Echoes from the Big Bang. Sci. Am. 38–43 (2001)
Chapter 11
Transformation Optics Applied
to Antennas and Focusing Systems

Rhiannon C. Mitchell-Thomas and Oscar Quevedo-Teruel

Abstract This chapter describes the way in which transformation optics can be
used to design lens antennas. It begins with a brief introduction to the analytical
method and describes the difficulties faced by practical applications, in terms of
anisotropic material requirements and bandwidth restrictions. Two further variants
of transformation optics are provided to circumvent these issues, which are the
quasiconformal transformation technique, and non-Euclidean transformations for
surface devices. The use of both these methods results in material property
requirements that can be easily implemented with dielectrics. These two techniques
are thoroughly described, with the benefits and drawbacks they present detailed.
The discussion is supported throughout with a number of application examples in
each section, which include planar lenses, bespoke lenses, curved surface wave
lenses, and a surface wave cloak. Examples are chosen to illustrate the variety of
ways in which transformation optics can be applied to antenna systems, and are all
supplemented by simulation or measurement data.

11.1 Introduction to Transformation Optics

The advent of transformation optics [1, 2] has provided electromagnetic engineers

and physicists with a very powerful tool. This mathematical technique allows a user
to predefine the propagation characteristics of electromagnetic waves, and imple-
ment this wave behavior in a device. This is possible by employing complex

R.C. Mitchell-Thomas
Department of Physics and Astronomy, University of Exeter, Physics building,
Stocker road, Exeter EX4 4QL, United Kingdom
e-mail: [email protected]
O. Quevedo-Teruel (&)
School of Electrical Engineering (EES), Electromagnetic Engineering (ETK),
KTH Royal Institute of Technology, Osquldas väg 6, fifth floor,
SE-100 44 Stockholm, Sweden
e-mail: [email protected]
URL: http://www.etk.ee.kth.se/personal/oscarqt/

© Springer International Publishing AG 2018 387

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_11
388 R.C. Mitchell-Thomas and O. Quevedo-Teruel

material properties to direct waves, which are generally spatially dependent and
anisotropic. It can be used to enable waves to perfectly transmit through sub-
wavelength apertures [3], to allow waves to avoid a certain region of space [4], or
for waves to concentrate to a particular point in space [5], illustrating the versatility
of the technique.
Over the last 20 years as the research on this topic has developed and intensified,
the areas in which the method has been used have broadened significantly, and
range from theoretical concepts such as black hole analogies [6], to the applied end
of the spectrum including improvements in solar cell technology [7]. The versatility
and opportunities posed by this technique may appear endless; however, as the field
has progressed, the understanding of the limitations and difficulties has also grown.
Efforts to circumvent these problems have resulted in a number of variants of
transformation optics being developed, two of which are outlined here.
This chapter focusses on the application of transformation optics to a range of
lens antenna systems. This has been a particularly successful area where the use of
transformation optics has provided solutions to difficult engineering problems such
as completely removing the reflection caused by waveguide bends [8] creating
isotropic radiators [9] or designing retroreflectors that do not need infinite refractive
indices [10]. The challenges posed when designing antennas for the mm and
sub-mm range using technology commonly used for the GHz range, such as
waveguides, reflectors, or phased arrays is that they need extremely high precision,
which increases the cost of manufacture considerably. Therefore, different routes
need to be sought, and this chapter proposes the use of transformation options to
overcome these issues.
Here, a description of how transformation optics can be used to design a range of
lenses which are appropriate for increasing the directivity of high-frequency
antennas will be given. These are alternatives to those designed using ray tracing,
for example, the Fresnel lens, the hyperbolic lens, or the Luneburg which either
suffer from bandwidth restrictions or complex geometry. Overcoming these issues
mean that they can be employed in transport and automotive communication sys-
tems, where weight and volume are important issues, or radioastronomy where
ultrawide bandwidths are required. In addition, surface wave devices where cur-
vature plays a crucial role will be shown. For example, curvature can be employed
so that dielectric is either not necessary, or lower permittivities can be used, thus
reducing the losses, which is highly desirable for aerospace applications.

11.2 Analytical Transformation

In its purest and original form, the transformation optics technique relies on an
exact transformation of coordinates. It was pioneered by Prof. Pendry and Prof.
Leonhardt in 2006 [1, 2], when two seminal papers were published in the same
Science journal issue. The process defines the modification of constituent material
parameters required, so that the electromagnetic behavior remains rigorously
11 Transformation Optics Applied to Antennas and Focusing Systems 389

Fig. 11.1 Illustration of a (a) (b)

Cartesian coordinate system
(a), and an example of a
transformed coordinate
system (b)

invariant in both the original and new coordinate systems. An example of two
coordinate systems is given in Fig. 11.1, where in the original Cartesian space,
labeled (x, y), the coordinate lines can be interpreted as rays propagating in straight
lines through a vacuum. The transformed system, labeled by (x′, y′), now exhibits
ray propagation that follows curved paths.
In order to achieve this equivalent behavior, the nature of the chosen coordinate
transformation is subsumed into the material parameters, namely the permeability
and permittivity tensors, which are of the same form and depend on spatial
parameters. Mathematically, this transformation of the permittivity and perme-
ability is calculated as follows:

JeJ T 0 JlJ T
e0 ¼ ;l ¼ ; ð11:1Þ
jJ j jJ j

where J represents the Jacobian transformation tensor that connects the two coor-
dinate systems; e and l represent the permittivity and permeability tensors in the
original space; and e′ and l′ are the same tensors in the transformed space. The
Jacobian transformation tensor takes the form

@x0


@x0 @x0


@x0 @y @z


@y0 @y0

J ¼
@y
@x @y @z
: ð11:2Þ

@z


0 @z0 @z0

@x @y @z

This type of analytical transformation has been employed for a number of

antenna applications [11], such as the redesign of traditional devices such as
polarization rotators [12], wave collimators and beam benders [13], or reduction of
the size of conventional lenses [14, 15]. However, one of the most simple trans-
formations that can be employed is the dilation of one of the axes of a Cartesian
coordinate system [15], where x0 ¼ x, y0 ¼ y, and z0 ¼ gz, and will be used here as
an example. Figure 11.2 displays this transformation in two dimensions, where the
vertical axis can be either x or y.
If we consider the original space to be a vacuum, then e ¼ l ¼ 1, and therefore it
is simple to deduce that the forms of the transformed permeability and permittivity
390 R.C. Mitchell-Thomas and O. Quevedo-Teruel

Fig. 11.2 A representation of

a region of space before, a,
and after, b, a compression in
the z-direction. The arrow
represents a ray propagating
through the space

that are required to reproduce the same effect as a dilation of one of the coordinate
axes are

1


0 0



g
e ¼ l ¼

0
0 0 1
g 0

: ð11:3Þ

0 0 g

These material properties now exhibit uniaxial anisotropy, where two of the
elements are lower than unity, which can only be achieved with metamaterials, and
therefore introduces the narrowband behavior generally associated with them. One
area that this type of analytical transformation has found application is in the
domain of metasurfaces. In this case, advantage can be taken off the inherently 2D
nature, so that a surface impedance tensor can be employed, which has a reduced
number of elements that need to be controlled. For certain cases, the required
anisotropy can be achieved, details of which are covered in Chap. 10 of this book
devoted to metasurface antennas. There is a rich literature in this area, and much
progress has been made in the theoretical description of such anisotropic meta-
surfaces, which is detailed in a number of publications [16–22]. This method has
successfully been used to design beam shifters [19, 22] and beam splitters [17, 20]
for surface waves, in order to reduce the scattering [22]. While this is a thriving area
for transformation optics, the current chapter will focus on methods which do not
rely on anisotropy, and can be simplified to isotropic material requirements.
For most of the bulk lens antenna systems designed using this technique, the
extreme anisotropy required is very difficult to achieve, particularly when combined
with spatial variation in three directions. This means that relatively few have been
experimentally validated, and those that have are only valid for a given polarization.
The following section will present a route that succeeds to escape these limiting
factors.
11 Transformation Optics Applied to Antennas and Focusing Systems 391

11.3 Quasi-conformal Transformation

The ability to determine the path of electromagnetic waves is a concept that holds a
great deal of promise for the design of electromagnetic devices. However, the main
limitation of the analytical transformation method is the necessity to fabricate
anisotropic materials. Even for very simple transformations, such as the dilation of
one of the Cartesian coordinates given above, it is necessary to implement aniso-
tropy in both the permeability and permittivity tensors. For more complex trans-
formations, these material tensors can require non-zero off-diagonal elements and
some of the tensor elements can reach infinity or zero. While the advances of
metamaterials are promising, and some degree of anisotropy can be realized, there
are still strong limitations to what is achievable. Additionally, the resonant nature of
these metamaterials implies that high losses would also be present. These diffi-
culties still remain a barrier to the deployment of transformation optics.
One method developed to overcome these problems is the quasi-conformal
transformation technique [23]. A conformal transformation is one which retains the
angles and shape of the original coordinate grid. In this case, the material tensors
would be spatially dependant but, crucially, are isotropic. In the majority of
problems, it is not possible to analytically find the exact conformal transformation
required, but the quasi-conformal technique uses numerical methods to approximate
a conformal transformation. An example of a quasi-conformal grid is given in
Fig. 11.3, where (a) is the original Cartesian coordinate system, and (b) is the
approximation of a conformal map within a new domain.
Once again, this transformation is effected by a change of the material param-
eters to play the role of the coordinate transformation. To ease the issues with
implementation, only the permittivity is altered, and value of the permittivity for
each individual mesh element is given by the following equation:
e
e0 ¼ : ð11:4Þ
det ðJ Þ

As it can be assumed that the transformation is conformal, and therefore the

angles and shape of each individual element of the mesh are conserved after the

Fig. 11.3 An example of a

quasi-conformal grid on a
square domain, and a
distorted domain
392 R.C. Mitchell-Thomas and O. Quevedo-Teruel

distortion, then the permittivity simply reduces to a ratio of the change of the area of
the cell, given by [24]

DxDy
e0 ¼ ; ð11:5Þ
Dx0 Dy0

where e′ is the new permittivity value for each cell of the coordinate systems. This
technique negates the use of anisotropic materials, but it also introduces its own
problems. After a quasi-conformal transformation is effected, there are usually regions
in the domain which have grown in volume. This means that the index at that position
will reduce as can be appreciated from Eq. (11.5), and therefore values less than unity
exist in the permittivity map. For this reason, care must be taken when employing this
technique, and two considerations must be taken into account. The first is an appro-
priate coordinate system that must always be chosen to ensure that the areas of the
sub-unity permittivity areas are minimized, and the second is to ensure that the mesh
cells are not warped to an extent that the approximation of isotropy is invalid [25].
The quasi-conformal transformation technique has been applied to antenna
systems in a number of ways [25–28], and four examples will be given here. The
first is the possibility to alter the shape of existing lenses, such as a hyperbolic lens
[27]. In this redesign of the lens, the focusing effect caused by the curved shape of
the homogenous dielectric lens can be reproduced by a flat lens with a graded index
after applying the quasi-conformal transformation technique. Figure 11.4a shows
the shape of the original lens, plus the quasi-conformal map where the color gives
an indication of the area of each mesh element. Figure 11.4b displays a photograph

Fig. 11.4 a A hyperbolic lens complete with a quasi-conformal coordinate grid, where the
colored regions relate to the area of each mesh cell, and the arrows emit from the focal point. b A
photograph of the cross section of the graded index lens, where each color corresponds to a
different permittivity of the lens. c and d show experimental measurements of the lens with the
phase illustrated in (c) and the collimated beam amplitude shown in (d). Extracted from [27]
11 Transformation Optics Applied to Antennas and Focusing Systems 393

of a half lens, and now the color refers to the values of permittivity and shows the
internal discretisation of the lens. Figure 11.4c, d shows the measured phase and
magnitude of a beam that has been emitted from a horn antenna and collimated by
the lens. It can be appreciated that this redesigned planar lens has served to increase
the directivity of the horn.
Another way in which the quasi-conformal technique can be applied is to the
redesign of standard graded index lenses so that their shapes are more appropriate
for a specific application. The Luneburg lens in one such lens has a spherically
symmetric refractive index profile which converts a focal point on the surface to a
plane wave on the opposite side. It allows steering of the beam to be achieved
simply by changing the position of the feed source, while the lens itself remains
stationary. However, one of the main drawbacks is that the focal plane is a
hemisphere, and therefore the mechanical actuation required to position the feed is
not ideal. It is possible to implement a quasi-conformal transformation so that the
feed surface becomes a flat plane, and the beam-steering can be achieved much
more simply [25, 29]. This example is shown in Fig. 11.5, where (a) is the

(a) 2 (c)

1.5

1
Y
0.5

0
(d)
-0.5

-1
0 0.5 1 1.5 2 2.5 3 3.5 4
X

(b) 2

1.5 (e)

Y 1

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
X

Fig. 11.5 Transformation of a Luneburg lens to provide a flat feed plane. a The quasi-conformal
map showing the hemispherical Luneburg feed plane. X′, Y′ plotted on the X, Y coordinates.
b The inverse map, X, Y plotted on X′, Y′ coordinates. c–e Full-wave simulations of the
transformed lens, with three different beam angles
394 R.C. Mitchell-Thomas and O. Quevedo-Teruel

Fig. 11.6 Mapping of a circular wavefront to a four-beam lens. a A quasi-conformal map of the
domain terminated by a circular wavefront. b The inverse map of the transformed square grid.
c The permittivity map calculated from the mesh element areas. d A full-wave simulation of a
point source positioned at the center of the lens showing the four planar beams

quasi-conformal map of a domain encompassing the Luneburg lens along with a

region of the surrounding free space, and (b) shows the inverse map. Often in these
designs, some of the cells experience an increase in the area, which would result in
a reduction of the permittivity. If this falls below unity, these regions can be
approximated to free space. In this example, the area of the lens that is implemented
with a permittivity above one is bounded by the white curve in (c–e). Full-wave
simulations are also shown for three different feed positions (illustrated by the black
circle), each producing a different beam angle.
The quasi-conformal technique can also be used to design lenses from scratch,
rather than to alter an existing one, using the phase-front shape of a source as the
original domain, and transforming it to desired phase-front shape, which defines the
new domain. This can be seen in Fig. 11.6 where (a) is a quasi-conformal coor-
dinate map on a circle, where the latter represents the emitted phase front from a
point source; and (b) is the inverse map on the transformed domain, which has the
shape of a square. The result of this transformation is to create a lens which has four
directive beams, due to the planar phase fronts [30]. The permittivity map required
is given in (c) where, as explained earlier, some of the regions are above, and some
below unity. The full-wave simulation shown in (d) includes the entire,
11 Transformation Optics Applied to Antennas and Focusing Systems 395

(a) (b)

(c) (d)

Fig. 11.7 A bespoke lens designed for a waveguide aperture. a The curved line in the left plot
represents an arbitrary shaped wavefront, and the transformation to the flat line in right plot
represents a planar wavefront. b A simple waveguide aperture. c The graded index lens showing a
cross section of the layers of various permittivities. d A full-wave simulation showing the efficient
transformation of the phase fronts from the aperture at the bottom of the lens, to a planar output at
the top of the lens. Extracted from [31]

un-approximated permittivity map, bounded by the white dashed line, and shows
that the four beams are evident emitted from each side of the lens.
The final example is to design a tailored lens for a particular feed type. Rather
than using a standard lens, which is designed using geometrical optics for a perfect
point focus, it is possible to employ this method to use the radiation characteristics
of the feed to design a bespoke graded index lens [31]. This is done by taking a line
of constant phase that has been emitted from the feed, which can be of arbitrary
shape, an example of which is shown in Fig. 11.7a. This curve then defines the
upper boundary of the domain to be transformed. The other three sides of the
domain are chosen appropriately so that the conformality of the map is maximized.
This is then transformed to a domain in which the upper boundary is planar, also
illustrated in Fig. 11.7a. Here, a rectangular waveguide aperture is used as an
antenna feed, as shown in (b). The complete lens design is shown in (c), where the
colors indicate the regions of different values of permittivity and (d) is a full-wave
simulation showing that the phase of the wave emitted from the waveguide aperture
has been effectively transformed to a plane wave output through the use of the
bespoke lens.
From the range of examples given here, it can be appreciated that this technique
is very versatile for lens antenna design. The challenge of spatially varying the
dielectric properties within these lenses is not to be underestimated. Additionally,
396 R.C. Mitchell-Thomas and O. Quevedo-Teruel

the loss presented by materials with a high permittivity is also a necessary con-
sideration. However, many advances have been made in the fabrication of graded
index devices [27], and together with the design method described here, a real
opportunity for a step-change in lens design is possible.
The next section will depart from the volumetric lens designs given above, and
describe a method for designing a whole range of surface wave lens antennas.

11.4 Non-Euclidean Transformation

A non-Euclidean transformation, as the name suggests, deals with the transfor-

mation between geometries which exhibit curvature [32]. The analytical transfor-
mation optics method is a very powerful tool and allows the most diverse range of
geometries to be transformed. However, one restriction it possesses is that all
geometries being considered must be flat, or Euclidean. The method that will be set
out in this section will overcome this limitation, and allow the same propagation
characteristics to be displayed on two two-dimensional surfaces where at least one
of these surfaces is curved [33]. This additional degree of freedom, the inclusion of
curvature in the geometries being transformed, will be shown to allow more flex-
ibility in device design [34]. This type of transformation is most useful for surface
wave applications because all the geometries being considered are necessarily
two-dimensional to allow for curvature to be employed. Additionally, as this takes a
geometrical optics approach, all the devices must be significantly larger than the
wavelength of operation for the approximations to hold. The resultant material
requirements are always isotropic and have values of refractive index above unity,
and are therefore very simple to implement using either dielectric layers or meta-
surfaces [35].
Three different applications of this method will be described in this section,
including surface wave cloaking [34, 35], lenses on curved surfaces [36], and using
curvature to remove the requirement of singular refractive indices in some lenses
[10]. We will begin with the design of a surface wave cloak, which has the simplest
mathematical formulation.
In order to hide an obstruction on a surface so that a surface wave would not be
scattered by it, it is possible to guide the wave up and over the object with the use of
a curved portion of the surface. However, although the scattering from the object
has now been removed, the curvature will leave a signature on the wave because of
the difference in optical path lengths experienced by different parts of the wave. We
want our curved surface to behave as if it were a flat homogenous surface free from
obstruction, and this is made possible if we add a radially dependant refractive
index profile to the curved surface. Consider the two surfaces represented in
Fig. 11.8. The bottom-flat surface has a refractive index of one and it is the
propagation of a wave on this surface that we want to emulate on the curved
surface. The upper curved surface, bounded by a radius, a, and with a maximum
height, b, has an arbitrary rotationally symmetric shape. The refractive index, nðhÞ,
11 Transformation Optics Applied to Antennas and Focusing Systems 397

Fig. 11.8 Diagram of the two surfaces under consideration. The curved surface has two ray paths
labeled as s1 and l1 , and the flat surface has two ray paths labeled s2 and l2 . Extracted from [33]

that we have to cover the curved surface with so that it appears flat can be calculated
if we consider the ray paths labeled as s1 , s2 , l1 , and l2 .
The optical path lengths of l1 and l2 , when equated, give [34]
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
nðhÞ RðhÞ2 þ R0 ðhÞ2 dh ¼ dr; ð11:6Þ

and the optical path lengths of the two circular paths, as s1 and s2 , can be given by

2pRðhÞsinðhÞnðhÞ ¼ 2pr: ð11:7Þ

Dividing Eq. (11.6) by dh gives

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dr
¼ nð hÞ RðhÞ2 þ R0 ðhÞ2 : ð11:8Þ
dh

If we take the derivative with respect to h of Eq. (11.2), we find

dr d
¼ ðRðhÞsinðhÞnðhÞÞ ð11:9Þ
dh dh
dr
¼ R0 ðhÞsinðhÞnðhÞ þ RðhÞcosðhÞnðhÞ þ RðhÞsinðhÞn0 ðhÞ: ð11:10Þ
dh

Equating Eqs. (11.8) and (11.10) gives

n0 ðhÞ R0 ðhÞsinðhÞnðhÞ þ RðhÞcosðhÞnðhÞ þ RðhÞsinðhÞ

¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð11:11Þ
nð hÞ
RðhÞ2 þ R0 ðhÞ2
398 R.C. Mitchell-Thomas and O. Quevedo-Teruel

This equation can then be solved either analytically or numerically to find nðhÞ.
One example that can be analytically calculated is that of a hemispherical bump. In
this case, R is no longer dependent upon h, and the solution is found to be

n0 ðhÞ 1  cosðhÞ
¼ : ð11:12Þ
nð hÞ sinðhÞ

And using an identity it can be shown that

tanðh=2Þ
nð hÞ ¼ : ð11:13Þ
sinðhÞ

Equation (11.13) gives the exact refractive index profile that will make a
hemispherical surface appear as if it were flat to rays propagating across it. This is
shown in Fig. 11.9, where a parallel plate waveguide that follows the shape of the
curvature is employed to illustrate the operation. In Fig. 11.9a, the
hemispherical-shaped waveguide is filled with a homogeneous dielectric with a
refractive index of two, and it can be seen that the curvature changes the behavior of

(a) (b)
n=2 n=3.2

(c) n 2 (d) n 3.2

1 1

Fig. 11.9 Simulations performed using planar waveguides following the curvature of the surface
being considered, with a plane wave source incident from the left. a, c—A hemispherical
waveguide filled with; a homogeneous dielectric (a) and a rotationally symmetric profile (c). b, d
—a cosine-shaped waveguide filled with; a homogeneous dielectric (b) and a rotationally
symmetric profile (d). Extracted from [34]
11 Transformation Optics Applied to Antennas and Focusing Systems 399

-1 1 -π π

Constant thickness (a) (b)

waveguide

z
x

(c) (d) y
Varying thickness
waveguide

Fig. 11.10 Experimental data of the electric field 0.5 mm above the curved dielectric waveguide
surface. a, b—A constant thickness waveguide, c, d—A varying thickness waveguide with the
graded index properties of the cloak. a, c—Electric field component Ez, b, d—Phase data. The
figures on the left show the cross section of the two waveguides. Extracted from [35]

the waves, creating an imperfect focal point. When the hemispherical guide is filled
with the angle-dependent refractive index profile given in Eq. (11.13), the planar
wavefronts that are incident from the left are reformed on the far side of the bump,
demonstrating that the guide appears as if the curved portion were absent
(Fig. 11.9c). However, this choice of guide shape creates reflections at the interface
between the flat and curved regions of the guide. This can be overcome if a
cosine-type function is instead chosen, as is the case in Fig. 11.9 b, d. Once again it
can be seen that the curvature in the guide distorts the shape of the planar wave-
fronts incident from the left (Fig. 11.9b), but when the guide is filled with the
appropriate refractive index profile, the planar nature of the wave is perfectly
restored after propagating over the curved guide, so that it would not be possible to
detect a signature from the curved surface on the wave.
One option to implement the required graded index surfaces shown in Fig. 11.9
for the purpose of creating a surface wave cloak is to employ a varying thickness
dielectric waveguide. This waveguide has a metallic sheet on the underside of the
curved guide and is open at the upper surface. In order to calculate the mode index
of a dielectric waveguide of a given thickness, either numerical simulations or an
analytical solution can be used. For a given frequency, this mode index will alter
depending on the thickness of the guide, and by slowly varying the thickness it is
possible to create a graded mode index medium. One such example is shown in
Fig. 11.10, where a cosine-shaped obstacle is chosen, and both a graded mode
index cloak and a homogeneous mode index comparison are shown. The surface
wave on both samples was imaged using a motorized 3-stage scanner to obtain the
field distribution 0.5 mm above the upper surface of the dielectric waveguide. It can
400 R.C. Mitchell-Thomas and O. Quevedo-Teruel

be seen in the measurement data shown in Fig. 11.10 that the constant thickness
waveguide, which exhibits a homogeneous mode index, causes a disturbance of the
circular wave propagating across it (Fig. 11.10a, b). However, in Fig. 11.10c, d, the
circular nature of the wavefronts has been restored by employing the appropriate
variation in guide thickness to achieve the mode index profile required to remove
the influence of the curvature, and thus act as a surface wave cloak.
This type of surface wave cloak has advantages over other designs that employ
the analytical transformation method due to the fact that the material requirements
are isotropic and therefore it is very simple to fabricate. In addition, the cloaks
designed here are also very thin (the order of k/6) and possess omnidirectional
operation in the plane, as opposed to those designed using the quasi-conformal
transformation method [37].
However, the design of surface wave cloaks is not the only use of this technique.
Another application is to redesign rotationally symmetric lenses, for example, the
Luneburg lens, to operate effectively when applied to a curved surface [36]. As it
has been demonstrated, there is an equivalence between curvature and refractive
index profiles for rotationally symmetric surfaces. It has previously been shown that
it is possible to design lenses using this principle. One well-known example of this
is the Maxwell fish-eye lens, which emulates the propagation of rays on the surface
of a hemisphere using a flat surface with a refractive index profile. Because of the
curvature, the rays emitted from a point on the circumference will diverge until they
meet the mid-point, before converging once again to a point on the opposite side of
the circumference. Although these rays are not changing speed, the curvature serves
to direct the rays on a particular path. This exact same behavior is exhibited by the
index profile given by the Maxwell fish-eye lens. Once again, a point source
positioned on the circumference of the lens emits rays that diverge; however, this
time the rays are being directed by the change in speed created by the index profile.
Just like the propagation on the hemisphere, the rays then converge at a point on the
opposite side of the lens circumference.
The Luneburg lens has been widely used in antenna design since the 60s. This
lens focusses an incident plane wave to a point on the opposite side of the cir-
cumference, and vice versa, using a rotationally symmetric index profile to guide the
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rays. Its index profile has the form nL ¼ 2  ðr=R0 Þ2 , where R0 is the lens radius.
The index profile also has an equivalent homogenous surface which can be used to
remove the requirement of a graded index medium. This provides an appealing
opportunity to create a surface wave Luneburg with a single-mode index surface.
However, the problem with this surface is that when it is applied to a flat surface,
there is a 90-degree transition between the curved surface and the flat. This, once
again, causes unwanted reflections and scattering of the surface wave. To circumvent
this problem, it is possible to calculate the profile of the Luneburg lens that is
appropriate for other surfaces, for example, a cosine-type function that smoothly
attaches to the flat place. In order to calculate this, Eqs. (11.6) and (11.7) must
include an index profile n1 ðr Þ on the flat surface, resulting in the following equation:
11 Transformation Optics Applied to Antennas and Focusing Systems 401

(a) (b) (c)

(d) (e)

Fig. 11.11 a—Required refractive index profiles for a Luneburg lens on each of the four labeled
surfaces. b–e—Electric field distribution in parallel plate curved waveguides, each filled with one
of the refractive index profiles given in (a), which are b—Flat, c—Luneburg shape, d—
Hemispherical shape, and e—Cosine shape. Extracted from [36]

 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n0 ðrÞ
0
n ð hÞ 1 þ r n11 ðrÞ RðhÞ2 þ R0 ðhÞ2  R0 ðhÞsinðhÞ  RðhÞcosðhÞ
¼ : ð11:14Þ
nð hÞ RðhÞsinðhÞ

Using this equation, if the flat Luneburg profile given above is substituted into
n1 ðr Þ, it is possible to calculate the index nðhÞ that will create the same behavior for
any surface represented by the function RðhÞ. In Fig. 11.11, four different surface
shapes are chosen, and the corresponding index profile for the surface to function as
a Luneburg lens is given in Fig. 11.11a. Figure 11.11b–e shows the electric field on
the surface of a curved planar waveguide, each illustrating effective transformation
of the point source positioned at the circumference of each device, to a plane wave
propagating to the left on the far side of the lenses.
In addition to the Luneburg and the Maxwell fish-eye lenses mentioned above,
there is a whole family of graded index lenses that exhibit interesting behavior [33].
Some examples of which are the Eaton lens, which redirects incident rays back in
the direction that they originated from, the monopole lens that allows a point source
of rays to only emit in one direction, and the invisible lens that can only be detected
by a temporal delay of the rays incident on it. The ray tracing for these three
remarkable lenses can be seen in the inset of Fig. 11.12. What these lenses have in
common is that at the center of the lens the refractive index goes to infinity. This is
a major barrier to their implementation, as it is impossible to create a dielectric
medium with a singular index, and it is also practically impossible to fabricate such
a steeply increasing graded index profile. However, using Eq. (11.14), it is possible
to solve for RðhÞ to find the equivalent surfaces [36].
All of the equivalent surfaces of singular lenses exhibit a tip which has the same
influence on the ray trajectories as the infinity in the refractive index. This simple
addition, and sculpting of the surface, means that lenses that were formerly
impossible to implement for practical applications are now achievable. In addition,
402 R.C. Mitchell-Thomas and O. Quevedo-Teruel

15
Eaton Monopole Invisible

10
n

0
0 0.2 0.4 0.6 0.8 1
/a

Fig. 11.12 The index profiles of three lenses with a singularity at the center. Inset—Ray tracing
demonstrating how each of the lenses performs, where the color of the ray trajectory is the speed of
the ray at that position

it is also possible to combine the use of finite, and low contrast index profiles with
curvature to allow for more appropriate surface shapes to be employed, provided
that the tip in the surface remains. One other method for removing the singularities
is called transmutation [38, 39]. However, these designs require highly anisotropic
materials, so once again are very difficult to implement, in contrast to the technique
used here. The limitation common to the two techniques is when the index profile
reaches infinity at a rate equal to or faster than r12 , it is no longer possible to use
either transformation method to redesign the lens. This is the condition for a black
hole, where all rays incident on the profile are unable to escape the device, and
spiral into the center [40].
An example is given in Fig. 11.13, which is of an Eaton lens. This design
requires the modest index contrast of *1.65 (shown in Fig. 11.13c), and the sur-
face smoothly transitions to a flat plane so that a surface wave incident upon it will
not suffer any reflections. It can be seen in Fig. 11.13a that the beam incident at the
bottom left of the lens is rotated around the tip and emitted in the same direction it
was incident from, creating a retroreflecting effect.

11.5 Conclusions

This chapter has introduced the concept of transformation optics and explained how
it can be used to design a range of different devices, particularly in the field of
lenses for antenna engineering. In the mm and sub-mm ranges, these lenses are of
specific interest because the volumes required are no longer prohibitive and the
11 Transformation Optics Applied to Antennas and Focusing Systems 403

(a) (b)

(c)

Fig. 11.13 The Eaton lens utilizing a hybrid shaped surface and a finite refractive index profile.
Left—Electric field magnitude of a beam incident from bottom left being rotated and exiting
bottom right. Top right—Instantaneous electric field viewed from the side. Bottom right—
Refractive index profile. Extracted from [10]

technologies traditionally used at lower frequency need very high manufacturing

tolerances, making them expensive to fabricate. We have described in detail three
different variants of the method, namely the analytical technique, quasi-conformal
transformation optics, and non-Euclidean transformations. Each has its own merits
and limitations, all of which were outlined here. The most general formulation is the
analytical technique; however, even simple devices designed using this technique
require anisotropic materials. While it is possible to fabricate some degrees of
anisotropy using metamaterials, more often than not approximations are required
leading to single polarization devices and/or a decrease in the accuracy of the
device. The quasi-conformal transformation method overcomes this problem,
resulting in purely isotropic devices. It allows the redesign of common electro-
magnetic devices to make them planar in shape, while retaining the behavior of the
original. It can also be employed to create bespoke lenses, designed to alter the
radiation properties of specific antenna feed systems, for example, to improve the
directivity, which is of particular importance for submm-wave antennas. For this
method, care must be taken to choose appropriate coordinate system to ensure the
approximations hold. The final technique allows non-Euclidean coordinate systems
to be considered. This adds a degree of freedom that is not possible with the
previous two techniques, however, and is only applicable for two-dimensional
systems. For this reason, it is invaluable in the design of curved surface wave
devices, and new types of surface wave cloak. It allows for the redesign of lenses
that were previously impossible to fabricate, meaning that retro-reflectors and beam
steerers can be created using either air-filled, or modest ranges of graded
dielectric-filled curved waveguides.
404 R.C. Mitchell-Thomas and O. Quevedo-Teruel

There are still numerous challenges involved in the implementation of the

aforementioned devices. Metasurfaces have proven to be instrumental in the
development of 2D-graded index devices, in that they are able to achieve
high-index contrast by varying the geometric parameters, can attain anisotropy, and
are simply manufactured using printed circuit technology. However, the ability to
fabricate bulk materials with a graded index that varies in all three dimensions is
less well established. The advances in 3D printing technology bring much hope that
this issue may soon be overcome. Research on printable low-loss and variable
permittivity materials is well underway, which will provide a straightforward route
to fabricating complex graded dielectric geometries. Coupled to the printing of
conductive inks, this will also offer a less labor-intensive solution to the manu-
facture of graded metamaterial structures, with exotic material properties such as
anisotropy or chirality.
In summary, the transformation optics tool is a very powerful one and has great
potential to be utilized in the design of electromagnetic lenses, particularly in the
mm-wave region where the size of the lenses is not prohibitive. It is hoped that the
methods described here provide the reader with the essentials to apply these
techniques to their own antenna problems, and allow advantage to be taken from the
opportunities posed by the use of transformation optics.

References

1. U. Leonhardt, Optical conformal mapping. Science 312(5781), 1777–1780 (2006)

2. J.B. Pendry, D. Schurig, D.R. Smith, Controlling electromagnetic fields. Science 312, 1780
(2006)
3. W. Tang, Y. Hao, F. Medina, Broadband extraordinary transmission in a single
sub-wavelength aperture. Opt. Express 18(16), 16946 (2010)
4. D. Schurig, J.J. Mock, B.J. Justice, S.A. Cummer, J.B. Pendry, A.F. Starr, D.R. Smith,
Metamaterial electromagnetic cloak at microwave frequencies. Science 314, 977 (2006)
5. M. Rahm, D. Schurig, D.A. Roberts, S.A. Cummer, D.R. Smith, J.B. Pendry, Design of
electromagnetic cloaks and concentrators using form-invariant coordinate transformations of
Maxwell’s equations. Photon. Nanostruct. Fundam. Appl. 6, 87 (2008)
6. H. Chen, R.-X. Miao, M. Li, Transformation optics that mimics the system outside a
Schwarzschild black hole. Opt. Express 18(14), 15183 (2010)
7. M.F. Schumann, S. Wiesendanger, J.C. Goldschmidt, B. Bläsi, K. Bittkau, U.W. Paetzold, A.
Sprafke, R.B. Wehrspohn, C. Rockstuhl, M. Wegener, Cloaked contact grids on solar cells by
coordinate transformations: designs and prototypes. Optica 2, 850–853 (2015)
8. D.A. Roberts, M. Rahm, J.B. Pendry, D.R. Smith, Transformation-optical design of sharp
waveguide bends and corners. Appl. Phys. Lett. 93, 251111 (2008)
9. P.-H. Tichit, S.N. Burokur, X. Wu, D. Germain, A. de Lustrac, Novel antenna concepts via
coordinate transformation. Adv. Electromagn. 2, 25 (2013)
10. S.A.R. Horsley, I.R. Hooper, R.C. Mitchell-Thomas, O. Quevedo-Teruel, Removing singular
refractive indices with sculpted surfaces. Sci. Rep. 4, 4876 (2014)
11. D. Kwon, D.H. Werner, Transformation electromagnetics: an overview of the theory and
applications. IEEE Antennas Propag. Mag. 52, 24 (2010)
12. D. Kwon, D.H. Werner, Polarization splitter and polarization rotator designs based on
transformation optics. Opt. Express 16(23), 18731 (2008)
11 Transformation Optics Applied to Antennas and Focusing Systems 405

13. D. Kwon, D.H. Werner, Transformation optical designs for wave collimators, flat lenses and
right-angle bends. New J. Phys. 10, 115023 (2008)
14. A. Demetriadou, Y. Hao, Slim luneburg lens for antenna applications. Opt. Express 19(21),
19925 (2011)
15. D.A. Roberts, N. Kundtz, D.R. Smith, Optical lens compression via transformation optics.
Opt. Express 17(19), 16535 (2009)
16. E. Martini, S. Maci, Metasurface transformation theory, in Transformation Electromagnetics
and Metamaterials, ed. by D.H. Werner, D.-H. Kwon (Springer, London, 2014), pp. 83–116
17. E. Martini, M. Mencagli, D. González-Ovejero, S. Maci, Flat Optics for Surface Waves. IEEE
Trans. Antennas Propag. 64, 155–166 (2016)
18. E. Martini, M. Mencagli, S. Maci, Metasurface transformation for surface wave control. Phil.
Trans. R. Soc. A 373, 20140355 (2015)
19. M. Mencagli, E. Martini, D. González-Ovejero, S. Maci, Metasurfing by transformation
electromagnetics. IEEE Antenn. Wirel. Propag. Lett. 13, 1767–1770 (2014)
20. A. Patel, A. Grbic, Transformation electromagnetics devices based on printed-circuit tensor
impedance surfaces. IEEE Trans. Microw. Theory Tech. 62, 1102–1111 (2014)
21. A. Patel, A. Grbic, The effects of spatial dispersion on power flow along a printed-circuit
tensor impedance surface. IEEE Trans. Antennas Propag. 62, 1464–1469 (2014)
22. R. Quarfoth, D. Sienvenpiper, Surface wave scattering reduction using beam shifters. IEEE
Antenn. Wirel. Propag. Lett. 13, 963–966 (2014)
23. J. Li, J.B. Pendry, Hiding under the carpet: a new strategy for cloaking. Phys. Rev. Lett. 101,
203901 (2008)
24. W. Tang, C. Argyropoulos, E. Kallos, W. Song, Y. Hao, Discrete coordinate transformation
for designing all-dielectric flat antennas. IEEE Trans. Antennas Propag. 58(12), 3795–3804
(2010)
25. N. Kundtz, D.R. Smith, Extreme-angle broadband metamaterial lens. Nat. Mater. 9(2), 129
(2010)
26. D.H. Kwon, Quasi-conformal transformation optics lenses for conformal arrays. IEEE
Antennas Wirel. Propag. Lett. 11, 1125–1128 (2012)
27. O. Quevedo-Teruel, W. Tang, R.C. Mitchell-Thomas, A. Dyke, H. Dyke, L. Zhang, S. Haq,
Y. Hao, Transformation optics for antennas: why limit the bandwidth with metamaterials? Sci.
Rep. 3, 1903 (2013)
28. R. Yang, W. Tang, Y. Hao, Wideband beam-steerable flat reflectors via transformation optics.
IEEE Antennas Wirel. Propag. Lett. 10, 1290 (2011)
29. H.F. Ma, T.J. Cui, Three-dimensional broadband and broad-angle transformation-optics lens.
Nat. Commun. 1, 124 (2010)
30. Q. Wu, Z.H. Jiang, O. Quevedo-Teruel, J.P. Turpin, W. Tang, Y. Hao, D.H. Werner,
Transformation optics inspired multibeam lens antennas for broadband directive radiation.
IEEE Trans. Antennas Propag. 61(12), 5910–5922 (2013)
31. M. Ebrahimpouri, O. Quevedo-Teruel, Bespoke Lenses Based on Quasi Conformal
Transformation Optics Technique. IEEE Trans. Antennas Propag. 65(5), 2256–2264 (2017)
32. U. Leonhardt, T.G. Philbin, Geometry and Light: The Science of Invisibility (Dover, New
York, 2010)
33. M. Sarbort, T. Tyc, Spherical media and geodesic lenses in geometrical optics. J. Opt. 14,
075705 (2012)
34. R.C. Mitchell-Thomas, T.M. McManus, O. Quevedo-Teruel, S.A.R. Horsley, Y. Hao, Perfect
surface wave cloaks. Phys. Rev. Lett. 111, 213901 (2013)
35. R.C. Mitchell-Thomas, O. Quevedo-Teruel, J.R. Sambles, A.P. Hibbins, Omnidirectional
surface wave cloak using an isotropic homogeneous dielectric coating. Sci. Rep. 6, 30984
(2016)
36. R.C. Mitchell-Thomas, O. Quevedo-Teruel, T.M. McManus, S.A.R. Horsley, Y. Hao, Lenses
on curved surfaces. Opt. Lett. 39, 3551–3554 (2014)
37. P.A. Huidobro, M.L. Nesterov, L. Martin-Moreno, F.J. Garcia-Vidal, Transformation optics
for plasmonics. Nano Lett. 10, 1985–1990 (2010)
406 R.C. Mitchell-Thomas and O. Quevedo-Teruel

38. Y.G. Ma, C.K. Ong, T. Tyc, U. Leonhardt, An omnidirectional retroreflector based on the
transmutation of dielectric singularities. Nat. Mater. 8(8), 639 (2009)
39. T. Tyc, U. Leonhardt, Transmutation of singularities in optical instruments. New J. Phys. 10,
115038 (2008)
40. E.E. Narimanov, A.V. Kildishev, Optical black hole: broadband omnidirectional light
absorber. Appl. Phys. Lett. 95(4), 041106 (2009)
 Part II
Antenna Measurements
Chapter 12
Antenna Measurements at Millimeter
and Submillimeter Wavelengths

Antti V. Räisänen, Juha Ala-Laurinaho, Thomas Crowe,

Sergiy Pivnenko, Manuel Sierra Castañer and Ville Viikari

Abstract Testing of electrically large, high-gain antennas as well as that of small

integrated antennas at millimeter and submillimeter wavelengths is extremely
challenging. Basically, there are three methods for measuring radiation properties of
an antenna: the far-field method, the near-field method, and the compact antenna
test range (CATR). In case of large antennas, the classical far-field method has two
major obstacles at mm and submm wavelengths: impractically large measurement
distance and high atmospheric loss. The planar near-field scanning method has been
used up to 1 THz. However, the applied near-field methods often give useful
information only on the main beam and its vicinity, because the field-sampling is
typically very sparse. Reflector-based and hologram-based compact antenna test
range (CATR) measurements have been demonstrated up to 500 GHz and
650 GHz, respectively. In the case of small integrated antennas, various techniques
for on-wafer measurements have been developed. This chapter discusses the theory,
techniques and limitations of the various test methods—the far-field method, planar
near-field scanning and CATR as well as on-wafer measurements. Also, antenna
pattern correction techniques are discussed.

A.V. Räisänen (&)
J. Ala-Laurinaho
V. Viikari

School of Electrical Engineering, Department of Electronics and Nanoengineering,
Aalto University, Espoo, Finland
e-mail: antti.raisanen@aalto.fi
T. Crowe
Virginia Diodes, Inc., Charlottesville, VA, USA
S. Pivnenko
Antenna Systems Solutions S.L., Santander, Spain
M.S. Castañer
Dpto. Señales, Sistemas Y Radiocomunicaciones, ETSI Telecomunicación,
Universidad Politécnica de Madrid (UPM), Madrid, Spain

© Springer International Publishing AG 2018 409

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_12
410 A.V. Räisänen et al.

12.1 Introduction

12.1.1 General Overview of Antenna Measurements at mm

and submm Wavelengths

Testing of electrically large, high-gain satellite antennas as well as that of small

integrated antennas at millimeter and submillimeter wavelengths is extremely
challenging. Basically, there are three methods for measuring radiation properties of
an antenna: the far-field method, the near-field method, and the compact antenna
test range (CATR), e.g. [1]. The natural choice for antenna measurements is the
far-field method. However, in the case of electrically large antennas, the classical
far-field method has two major obstacles. The separation between the antenna under
test (AUT) and the source antenna must be so large that the spherical wave front
radiated by the source antenna can be approximated as a plane wave. Typically the
required far-field distance is d = 2D2/k, where D is the diameter of the AUT and k
is the wavelength. For example, the required separation is 9.8 km for a 1.5-m
antenna at 650 GHz. A larger separation may be needed when very low sidelobe
levels are measured. At mm- and submm wavelengths, the large attenuation and the
distortions to the signal due to the atmosphere are the major problems. Variations of
temperature and humidity in time and space cause distortions to the amplitude and
phase of the signal. However, the far-field method is readily applicable for
mm-wave antennas with gain less than 20 dB, because in this case the far-field
measurement can be done in a controlled environment such as an anechoic
chamber.
Near-field range measurements can be done indoors in a relatively small space.
The temperature and humidity can be controlled and thus the distortions due to the
atmosphere become less severe. The amplitude and the phase of the AUT near-field
are sampled with a probe antenna and the far-field radiation characteristics are
calculated using numerical methods. The sampling of the near-field can be
accomplished over a plane, a sphere, or a cylinder. The sampling area has to be
comprehensive so that all the significant energy transmitted by the AUT can be
captured. The sampling has to be dense enough for satisfying the Nyquist sampling
criterion, i.e., the distance between two sampling points has to be smaller than one
half of the wavelength, i.e., <k/2. For example, 1.8 m  1.8 m rectangular sam-
pling grid, which covers a 1.5-m antenna, includes about 60 million sampling
points, if full sampling is carried out at 650 GHz. The sampling rate has to be
extremely high to keep the acquisition time reasonable. The requirements for the
stability of the measurement system are very stringent. Sparser sampling grid can be
used, if only the main beam region and the beam alignment are of interest, e.g. [2].
Major error sources in the near-field scanning are the phase errors in flexing and
bending cables, inaccurate positioning of the probe, and the reflections between the
probe and the AUT.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 411

In a compact antenna test range (CATR), the spherical wave front is transformed
into a plane wave with a collimating element, which can be a reflector, e.g. [3], a
lens, e.g. [4, 5], or a hologram, e.g. [6–8]. Also, an array of antennas may be used to
produce a plane wave at lower frequencies. CATRs enable the antenna measure-
ments indoors in plane wave conditions, i.e., the radiation pattern can be directly
measured by rotating the AUT. The main beam of a high-gain antenna can be
measured without rotating the AUT by scanning the CATR feed transversally.
The AUT is placed into the quiet-zone (QZ), which is the extent of the volume
where the wave front meets the requirements set on the plane wave. Typically, the
maximum amplitude and phase deviations have to be less than ±0.5 dB and ±5°,
respectively. Most of the CATRs are based on reflectors. The major problem is the
very stringent surface accuracy requirement at high frequencies. Usually the root
mean square (rms) surface error should be less than 0.01 k, e.g., 4.6 lm at
650 GHz. The surface accuracy requirements for lenses are not as stringent as for
reflectors, since the lens is a transmission type of an element. In practice, the major
problem is to find suitable material for lenses. If dielectric constant is small, like for
foam, the material is usually not homogeneous. The problem is more severe at high
frequencies.

12.1.2 Specific Aspects Concerning the Instrumentation

at mm and submm Wavelengths

As is clear from the descriptions above, the high test frequency introduces chal-
lenges for the RF instrumentation in any antenna measurement. The minimum
requirement is to have a frequency-stable transmitter with sufficient output power
and a sensitive receiver at the given mm-wave or THz frequency. However, the
frequency range from 0.1 to 10 THz is often called also as the terahertz gap,
because technologies for generating and detecting this radiation are less developed
than those at microwave frequencies or at infrared. Generation of THz waves with
microwave means may take place by using vacuum tube generators such as
backward wave oscillators (BWO) or by multiplying semiconductor oscillator
frequency from microwave or low millimeter wave frequency by a Schottky diode
multiplier, see e.g. [9]. On the other hand, generation of THz radiation with pho-
tonics means may take place using optically pumped gas lasers or photomixers
pumped with two infrared semiconductor lasers [10]. However, power of the THz
signal from a photomixer is typically only microwatts and therefore not really
usable for antenna measurements.
Amplitude (i.e., scalar) measurements may suffice in some applications, and the
phase may also be retrieved by using holographic means as discussed in Sect. 12.5
(Chapter 13 is fully devoted to holographic principles). In such measurements a
sensitive power meter or detector may be used as a receiver. A spectrum analyzer is
an excellent incoherent receiver (when equipped with external down-conversion
412 A.V. Räisänen et al.

units). In most antenna measurements, however, a vector, i.e., both amplitude and
phase, measurement is desired. Today, there are commercially available excellent
vector network analyzer s (VNA) with extension units up to 1 THz based on
frequency multiplication and harmonic mixing.

12.2 Millimeter- and Submillimeter-Wave

Instrumentation

12.2.1 Background and Conceptual Review of VNA

Frequency Extenders

The testing of antenna systems is greatly facilitated by the availability of high

quality test and measurement equipment, such as the vector network analyzers
(VNA) from leading equipment manufacturers such as Keysight Technologies [11],
Anritsu [12] and Rohde & Schwarz [13]. Present VNA models operate to fre-
quencies as high as about 70 GHz and are exceptionally useful tools for general
laboratory measurements; including measurements of passive devices, amplifiers,
and frequency conversion devices (such as mixers), as well as antenna testing.
However, measurements above *70 GHz require the use of external frequency
extender modules, and these devices are the primary subject of this section.
A block diagram for a generic VNA extender module is shown in Fig. 12.1. The
module requires two inputs from the VNA: the RF signal and the local oscillator
(LO) power required to pump the mixers. The VNA must supply these signals with
the appropriate power levels and frequency range, while maintaining the necessary
frequency offset so that the IF signals generated by the module’s mixers are gen-
erated at a precisely controlled frequency. Also, the VNA signals should have low
phase noise. This is particularly important in the case of very high frequency
extension modules, where the high multiplication factor degrades the phase noise.
At measurement frequencies above 500 GHz, the VNA’s synthesizer phase noise
can be a limiting factor in the quality of the measurements.
The extender module consists of the transmitter and the measurement and ref-
erence receivers. The transmitter uses a cascade of amplifiers and frequency mul-
tipliers to increase the signal frequency to the desired band. The exact configuration
of this amplifier-multiplier—chain is selected by the manufacturer based on a
variety of considerations; including their technological capabilities, performance
goals and cost. For example, a multiplier with a high harmonic factor is simplest to
achieve and has the lowest cost, but a cascade of multipliers with lower harmonic
factor may lead to greater transmit power and a more pure test signal. Other features
of the transmitter may include a power attenuator and an isolator. The attenuator is
useful for tests that require less power delivered to the device under test (DUT), for
example measurements of the gain and linearity of an amplifier. The isolator is
almost always beneficial in that it improves the source match of the system.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 413

Fig. 12.1 A generic frequency extender module. The system consists of one signal transmitter,
two receivers (signal and reference) and directional couplers. (1—Isolator, 2—variable attenuator,
3—frequency down-conversion mixers)

However, isolators are difficult to achieve at frequencies above *220 GHz, and
their insertion loss can be significant.
The two receiver systems are essentially identical in design, but serve different
functions. The reference receiver samples the outgoing power. The intermediate
frequency (IF) output from this mixer is coupled back to the VNA and is used to
determine the magnitude and phase of the outgoing wave. The measurement
receiver samples the power that enters the module from the test port. This can be
power from the module’s own transmitter that is reflected by the device under test,
or transmitted power from a second module that is attached to a different port of the
DUT. The IF output from the measurement receiver is coupled to the VNA and is
used to determine the magnitude and phase of the incoming wave. A comparison of
the measured IF to the reference IF allows the compensation of drift in the mag-
nitude and phase of the RF test signal. As with the transmitter, the exact config-
uration of the receiver system is selected by the module manufacturer. The critical
choices of when to split the LO signal and the design of the amplifiers, multipliers,
and mixers have a very significant impact on the sensitivity, stability and manu-
facturing cost of the receivers.
The final major component of the module is the directional coupling system. The
couplers send samples of the outgoing and incoming waves to the mixers. The
simple model of the extender module assumes that each mixer receives only the
intended signal, but leakage of other signals through the couplers complicates the
signal analysis and degrades system performance. The design and manufacture of
414 A.V. Räisänen et al.

the couplers is thus a critical aspect of the development of the module, and each
manufacturer carefully balances the trade-offs between different aspects of system
performance, manufacturing complexity and cost.

12.2.2 Considerations for Frequency Extenders Well Above

110 GHz

The availability of electronic equipment for use in the upper part of the mm-wave
band and into the submillimeter band has grown significantly in the past decade.
This has been made possible by the development of improved simulation tools for
the linear and nonlinear circuit elements, the advancement of microelectronic
fabrication technologies and the development of advanced “terahertz” circuits with
higher levels of integration. Together, these advancements have enabled the
development of frequency multipliers and mixers that operate across full waveguide
bands with excellent performance to frequencies above 1 THz. Most importantly,
these components can now be reliably manufactured at an acceptable cost.
As an example, Fig. 12.2 shows a VDI integrated diode circuit that operates
above 1 THz [14]. The entire chip is designed and simulated using modern CAD
tools; including the diodes, fingers, pads, frequency filters, and waveguide probes.
The wafer is then fabricated and singulated into individual circuits in the cleanroom
facility. For a size reference, the thickness of the substrate is about five microm-
eters. This gives an indication of the extremely small feature sizes that are required
for terahertz components and the quality of the fabrication technology. To complete
the component, the split-block waveguide housing must be co-designed with the
terahertz IC, and then precision manufactured. Finally, the IC is placed and secured
within the housing and additional low-frequency components, including bond
wires, are placed. Upon final assembly of the block, the component is RF tested and
qualified units integrated into the frequency extender module.

Fig. 12.2 Two SEM images of a THz integrated circuit focusing on the anti-parallel diode pair.
The circuit substrate is *5 lm thick. This gives an indication of the extremely small feature sizes
that are required for terahertz components and the quality of the fabrication technology
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 415

A VNA extender module to 1 THz requires of about ten multiplier and mixer
stages, each of which must be fully optimized for performance across the correct
waveguide band using the exact power level that is available from the preceding
component in the chain. Also, these components, and the transmitter and receiver
systems as a whole, must be developed so that they are stable and do not generate
significant power at unwanted frequencies. Thus, the design of very high frequency
extender modules requires not only the development of individual components
covering the frequency bands of interest, but also the development of a system
architecture that yields exceptional performance across the full waveguide band
without spurious signals or oscillations.

12.2.3 Extender Module Configuration, Considerations

for Antenna Test

A complete VNA extender module is depicted in Fig. 12.1. Two of these modules
are required for full S-parameter characterization of a two-port DUT. These mod-
ules connect to a VNA system that supplies the required RF and LO signals, and
processes the four IF outputs from the modules. Additionally, a switch in the VNA
system is used to alternately direct the RF test signal from one extender module to
the other, since only one transmitter should be active at any time. This is commonly
achieved with a four-port VNA or the use of some sort of additional signal con-
troller. The LO distribution is somewhat simpler. Since all four mixers are intended
to be active at all times, it is necessary only to split the power between the modules.
However, it should be considered that any relative drift in the phase of the LO
signal delivered to the two modules will create measurement drift that cannot easily
be calibrated or otherwise compensated.
The chart in Fig. 12.3 shows various configurations of the VNA extender
modules. For antenna measurements the full two-port solution is not required, and a
simpler configuration will not only reduce costs, but can also improve performance.
For the typical case of an antenna range only S21 is required and the “Through”
configuration (TxRef-Rx) reduces the cost by nearly a factor of two. This is because

Fig. 12.3 Various

configurations of VNA
modules and the supported
S-parameter measurements
416 A.V. Räisänen et al.

Fig. 12.4 The measured “enhanced dynamic range” of various VDI VNA extender modules for
antenna test systems [14]

the TxRef includes only the transmitter and reference receiver, and the Rx consists
of only a measurement receiver. Also, one of the two coupler systems is eliminated.
This simplification also allows performance improvements for certain applica-
tions. Considering first the Rx module, the elimination of the coupler and other
components allows all of the power that passes through the DUT to be coupled to
the mixer. This can increase the measurement receiver sensitivity, referred to the
input of the module, by orders of magnitude. Also, the simplification of the
transmitter module (TxRef) can allow a re-optimization of the internal components
to achieve greater test port power. This improvement is not as great, but can still be
significant.
Figure 12.4 shows a graph of the “enhanced dynamic range” of VDI extender
modules that are optimized for antenna test systems in the 75–1100 GHz frequency
range. The data shown represents the ratio of the test port power of the transmitter
to the noise power of the receiver module referred to its input port in a 10 Hz
measurement band. This would equal the system dynamic range provided the power
does not saturate either the module’s mixer or the low-frequency mixers in the
VNA, which is unfortunately not the case for a low loss DUT. However, in the case
of a lossy DUT, such as an antenna test range, the power actually reaching the Rx
module may be low enough to prevent saturation. For example, in the WR-10 band
the enhanced dynamic range is about 150 dB. However, practical considerations,
including the saturation of the extenders’ mixers and the dynamic range limits of
the VNA itself, normally prevent the actual instantaneous dynamic range from
exceeding about 125 dB. But, if this module configuration is used in an antenna
range with 30 dB minimum range loss, then there will be no saturation effects and
the remaining dynamic range for measurements, after the 30 dB loss is overcome,
will be 120 dB. This represents about a 30 dB improvement over the use of two full
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 417

Fig. 12.5 A size comparison of the original modules (11′′ x 5′′ x 3′′) with the new mini-modules
(8.5′′ x 3′′ x 1.5′′). The volume is reduced by 75% and the weight by more than 50%

Fig. 12.6 A complete Rx module for the WR-10 waveguide band, including a two inch test port
and a mounting plate; both are removable. The Rx module measures 2.5″ x 1″ x 0.75″. The LO
and IF connectors are shown, but the DC power connector is not visible

TxRx modules and this increased dynamic range can be used to measure deep nulls
in the antenna pattern and/or to greatly increase the measurement speed by
increasing the resolution bandwidth.
A final benefit of this simplified configuration is the ability to greatly reduce the
size and weight of the Rx module. Recently VDI introduced new “mini-modules”
with reduced size (Fig. 12.5). This introduction has begun with the lower frequency
modules, WR-15 through WR-6.5, and will soon extend to higher bands. This
improvement is made possible through the development of new internal compo-
nents and greater levels of system integration. However, for the Rx modules, it is
possible to achieve an even smaller package. A WR-10 “micro” receiver module is
shown in Fig. 12.6. This is a complete receiver system with LO input and IF output
suitable for connection directly to the VNA. This small, lightweight package is
418 A.V. Räisänen et al.

ideal for use in antenna test systems, especially in the case of on-wafer pattern
measurements where the available space is very limited. An additional benefit of
this integrated module is the elimination of several waveguide flange connections,
which are a primary source of unwanted signal leakage.

12.2.4 Future Perspectives

The advancement of VNA extender technology is made possible through the

continued development of advanced terahertz components such as frequency
mixers, multipliers, and amplifiers. Also, the development of VNAs that are better
optimized for high frequency extenders can have a significant impact. In May of
2015 VDI demonstrated a TxRx-Rx extender system that yielded about 70 dB of
dynamic range (10 Hz) in the 1.1–1.5 THz range [14]. However, this level of
performance required the use of external, ultra-low-noise synthesizers. Given the
demand for higher frequency measurements, it is expected that all major VNA
manufacturers will strive to improve the phase noise performance of their top-level
vector network analyzers. This will lead to improvements in all extender modules
operating above 500 GHz.
Modern vector network analyzers and frequency extender modules allow full
two-port measurements with excellent dynamic range and stability. The extender
modules are now being optimized for specific applications, such as antenna test.
This includes the development of more compact modules and the optimization of
the system to overcome range losses while maintaining excellent dynamic range.
These technologies now enable high quality antenna measurements up to 1.1 THz,
and further advances in performance and frequency range are expected.

12.3 Near-Field Measurements

12.3.1 Planar Near-Field Scanning

Aperture antennas at mm-wave bands usually have very large electrical size, with
the characteristic aperture dimensions ranging from several tens to several hundreds
of wavelengths. For these kinds of antennas, due to their very large directivity and
thus primary radiation in one direction orthogonal to the aperture, specifically
planar near-field (PNF) technique has found wide application.
In the PNF technique, the field is measured by a small probe on a plane in front
of the aperture antenna on a regular grid of points. The most often used mea-
surement grid is a rectilinear grid, in which the probe is moved by two linear
orthogonal slides along two coordinates usually called x and y (Fig. 12.7a). Other
types of regular grids are also used, for example, plane-polar grid, in which the
probe is moved along a linear slide, while the antenna is rotated along the vertical
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 419

Fig. 12.7 Examples of the scanning grids in planar near-field technique: a rectilinear grid,
b plane-polar grid, and c bipolar grid

axis (Fig. 12.7b), and bipolar grid, in which both the probe and the antenna are
rotated (12.7c).
For the rectilinear grid, a rigorous solution exists for the expansion of the
measured field over the plane waves, while for the other two grids, such a solution
is not available and thus the measured field must first be interpolated into the
rectilinear grid, which introduces interpolation errors. On the other hand, flexing
cables following the probe moving on the rectilinear grid and mechanical uncer-
tainties of the slides are among the largest error sources at very high frequencies,
while the errors of the rotary joints used in the polar grids and mechanical uncer-
tainties of the rotating stages introduce smaller effects and are more systematic thus
allowing their more effective compensation.

12.3.2 Plane-Wave Expansion and Far Field

By means of the equivalence principle [15], the far-field of an antenna can uniquely
be determined by means of suitable equivalent currents on an imaginary surface in
front of the antenna; in the case of the PNF measurements, this surface is an
imaginary flat aperture parallel to the antenna aperture, Fig. 12.7a. Thus, instead of
considering the real antenna, we consider an aperture located for convenience in
z = 0 plane of the coordinate system on which the tangential components of the
electric field Et ¼ Ex ^x þ Ey ^y are assumed known.
The principle of near-field antenna measurements is based on expansion of the
fields measured on the chosen surface close to the antenna into a set of free space
modes, which represent a solution of Maxwell’s equations in the chosen coordinate
system [16].
420 A.V. Räisänen et al.

In a source-free and lossless medium, the electric field satisfies the following
equations [15]

r  r  E  x2 leE ¼ 0; ð12:3:1Þ

r
E ¼ 0: ð12:3:2Þ

Using a vector identity r  r  E ¼ rr
E  r2 E and substituting x2 le ¼

k02 the first equation becomes

r2 E þ k02 E ¼ 0: ð12:3:3Þ

In Cartesian coordinates ðx; y; zÞ these two equations are expressed as


@2 @2 @2
þ þ þ k 0 Eðx; y; zÞ ¼ 0;
2
ð12:3:4Þ
@x2 @y2 @z2

@Ex ðx; y; zÞ @Ey ðx; y; zÞ @Ez ðx; y; zÞ

þ þ ¼ 0: ð12:3:5Þ
@x @y @z

The Fourier transform of a two-dimensional function uðx; yÞ (with exp ðjxtÞ

time notation) is
Z Z
  1
U k x ; ky ¼ uðx; yÞ ejðkx x þ ky yÞ dxdy ð12:3:6Þ
1

with the corresponding inverse Fourier transform

Z Z
1 1  
uðx; yÞ ¼ U kx ; ky ejðkx x þ ky yÞ dkx dky : ð12:3:7Þ
4p2 1

Applying the Fourier transform to the Eqs. (12.3.4) and (12.3.5) with respect to
x and y we obtain
    
@ 2 =@z2 þ k02  kx2  ky2 E kx ; ky ; z ¼ 0; ð12:3:8Þ
     
kx Ex kx ; ky ; z þ ky Ey kx ; ky ; z þ j@Ez kx ; ky ; z =@z ¼ 0; ð12:3:9Þ
 
where E kx ; ky ; z is the Fourier transform of the electric field with respect to x and
y.
Now we let k02  kx2  ky2 ¼ kz2 and note that (12.3.8) has only derivative with
respect to z and thus the problem of solving a partial differential equation has been
reduced to solving an ordinary differential equation. The general solution to this
equation is
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 421

     
E kx ; ky ; z ¼ f 1 kx ; ky ejkz z þ f 2 kx ; ky ejkz z ; ð12:3:10Þ
   
where f 1 kx ; ky and f 2 kx ; ky are unknown constants with respect to z. The two
terms in (12.3.10) represent plane waves propagating in positive and negative
directions, respectively, with respect to z. Since in our formulation the antenna
under test (AUT) is located in the hemisphere z < 0 and we consider only waves
propagating outwards
  from the AUT, in the positive z-direction, we shall keep only
the first term f kx ; ky ejkz z omitting the subscript for f.
Substituting the solution into (12.3.9) we obtain

kx fx þ ky fy þ kz fz ¼ k
f ¼ 0; ð12:3:11Þ

which shows that only two components of the vector f are independent.
Using the inverse Fourier transform, the electric field can then be found
Z Z 1
1  
Eðx; y; zÞ ¼ 2 f kx ; ky ejkz z ejðkx x þ ky yÞ dkx dky
4p Z Z1
1  
1
¼ 2 f kx ; ky ejk
r dkx dky : ð12:3:12Þ
4p 1
 
The unknown function f kx ; ky is found by noting that for z = 0 the solution
must be equal to the assumed known tangential electric field Et in the aperture, thus
Z Z
  1
f t kx ; ky ¼ Et ðx; y; 0Þejðkx x þ ky yÞ dxdy: ð12:3:13Þ
1

Using (12.3.13) to find the tangential components f t of the plane-wave spectrum

and then evaluating (12.3.12), the electric field can be found everywhere in the
hemisphere z > 0.
In the far-field of the AUT, the integral in (12.3.12) can be asymptotically
evaluated by the method of stationary phase [15, 17]

jk0 ejk0 r
EðrÞ  cos h f ðk0 sin h cos /; k0 sin h sin /Þ; ð12:3:14Þ
2p r

where kx ¼ k0 sin h cos /; ky ¼ k0 sin h sin / and h; / are the spherical coordinate
angles. Equation (12.3.14) expressed in terms of the spherical components, with ^ h
^ being the spherical unit vectors, becomes
and /

jk0 ejk0 r h    i
EðrÞ  fx cos / þ fy sin / ^h þ cos h fy cos /  fx sin / /
^ : ð12:3:15Þ
2p r
422 A.V. Räisänen et al.

12.3.3 Scattering Matrix Approach of Antenna-Antenna

Interactions, Transmission Formula

The assumption regarding the known, somehow measured, tangential aperture field
is not valid in practical cases, since the aperture field cannot be measured using an
ideal probe—Hertzian dipole of zero physical extent. All practically used probes
have finite dimensions, usually on the order of a wavelength, for example, a
commonly used rectangular open-ended waveguide probe has the aperture size of
some 0.375 k by 0.75 k at the center frequency. Due to these finite dimensions, the
received signal at the probe terminals is not proportional to the field in a point, but
represents a weighted value of the incident field over the probe aperture. Thus the
receiving properties of the probe must be properly compensated in order to find the
correct characteristics of the AUT. In order to get the relation between the received
signal at the probe terminals depending on the probe location in front of the AUT,
and the provided signal at the AUT terminals, we shall use the scattering matrix
formulation for the AUT and probe.
An antenna can be represented by a multi-port network with one-feed port—the
feeding line with a single propagating mode, for example, the TEM mode in a
coaxial waveguide, and with an infinite number of the radiation ports representing
the plane waves radiated, received, and scattered by the antenna. The relation
between the ports can be expressed through the scattering matrix, which contains
the reflection coefficient CAUT , the receiving coefficients Rs, the transmitting
coefficients Ts, and the scattering coefficients Ss,r, see Fig. 12.8.
It is noted that at a given frequency, the reflection coefficient is a complex
number, while the other coefficients represent continuous complex functions
reflecting the property of continuous spectra of the plane waves. The indices s and r
are used to count the two orthogonal components of the plane waves—in our
formulation here, these components will be parallel to the axes x and y of the
antenna coordinate system, see Fig. 12.9.
The complex amplitude of the outgoing wave b0 at the antenna feed port can be
written as

Fig. 12.8 Scattering matrix of an antenna

12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 423

Fig. 12.9 Antenna and probe coordinate systems

Z Z
1 1    
b0 ¼ CAUT a0 þ Rs kx ; ky as kx ; ky dkx dky ð12:3:16Þ
4p2 1

where a0 is the complex amplitude of the incoming wave at the antenna feed port
and as are the complex amplitudes of the incoming plane waves at the radiation
ports. Similarly, the complex amplitudes of the outgoing waves bs at the antenna
radiation ports can be written as
Z Z X2
    1 1    
bs k x ; k y ¼ T s k x ; k y a 0 þ 2 r¼1
Ss;r k x ; k y ; j x ; jy a r jx ; jy djx djy
4p 1
ð12:3:17Þ

In the absence of the incoming plane waves, which is a usual assumption in

near-field antenna measurements, the second term in (12.3.17) is 0.
Similar formulation is applicable for the probe, in its own coordinate system
with the z0 -axis pointing outward from the probe aperture, and using the primed
symbols, to distinguish from the AUT: the complex amplitude of the outgoing wave
b00 at the probe port is written similar to (12.3.16). Assuming for simplicity a
matched load at the probe port, the first term is equal to 0. For the coordinate system
with the z-axis pointing inwards to the probe aperture, the following relations are
valid, for rotation around the y0 -axis: ^x0 ¼ ^x; ^y0 ¼ ^y; ^z0 ¼ ^z; and thus
kx0 ¼ kx ; ky0 ¼ ky . We also need to use a relation between the probe receiving and
transmitting coefficients,
  which  according
 to the reciprocity relation can be
 
expressed as Rs kx ; ky / T s kx ; ky ¼ T 0s kx ; ky .
0 0 0 0 0 0

The probe is moved on a plane in front of the AUT at a fixed distance z0 , but
with arbitrary x0 and y0 , and such that the probe z0 -axis is parallel to the AUT z-axis
424 A.V. Räisänen et al.

(Fig. 12.9). The relation between the incoming waves in the probe coordinate
system and the outgoing waves in the AUT coordinate system is written as
     
a0s kx0 ; ky0 ¼ bs kx ; ky ejkr0 ¼ T s kx ; ky a0 ejkr0 ; ð12:3:18Þ

where ejkr0 ¼ ejðkx x þ ky y þ kz z0 Þ .

Finally, the relation between the signal received at the probe output and the
signal at the AUT input depending on the probe position in the AUT coordinate
system is written as
Z Z
a0 1    
b00 ðr0 Þ ¼ T 0s kx ; ky T s kx ; ky ejkz z0 ejðkx x þ ky yÞ dkx dky : ð12:3:19Þ
4p2 1

This is recognized as a two-dimensional inverse Fourier transform over the

product of the transmission coefficients of the AUT and the probe in their own
coordinate systems. By applying the Fourier transform, the product of the AUT and
probe coefficients is found as
Z Z
      ejkz z0 1
T 0s kx ; ky T s kx ; ky ¼ B kx ; ky ¼ b00 ðr0 Þejðkx x þ ky yÞ dxdy ð12:3:20Þ
a0 1

This  represents
  one
 equation
 with the two unknowns
T s kx ; ky ¼ Tx kx ; ky ^x þ Ty kx ; ky ^y. Another equation is obtained by performing
a second measurement with the probe rotated orthogonally to the first orientation or
using a dual-polarized probe and the signals from the two orthogonal ports. The two
components of the AUT transmission coefficients are then resolved from the two
equations as follows, written in the matrix form:
  0 0

Tx 1 Ty;2 Ty;1 B1
¼ 0 0 : ð12:3:21Þ
Ty 0
Txy Tx;2 Tx;1 B2

The indices 1 and 2 indicate the two probe orientations and

0 0 0 0 0
Txy ¼ Tx;1 Ty;2  Tx;2 Ty;1 . It is noted that in this derivation, for the probe orientation
0 0
1, with its port aligned along the x-axis, Tx;1  Ty;1 , while for the probe orientation
0 0
2, with its port aligned along the y-axis, Tx;2  Ty;2 .

12.3.4 Discretization and Sampling Intervals

In practice, the probe signal is not measured continuously, but in a series of discrete
points both in x and in y directions with the intervals Dx and Dy. The integral in
(12.3.20) is then replaced by a double summation
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 425

Fig. 12.10 Normalized plane-wave spectra of uniform aperture antennas

Z Z 1 X
b00 ðr0 Þejðkx x þ ky yÞ dxdy ¼ DxDy n;m
b00 ðxn ; ym ; z0 Þejðkx xn þ ky ym Þ : ð12:3:22Þ
1

The values of the sampling intervals are found recalling that the spectrum of the
plane waves becomes evanescent for kx ; ky [ k0 ¼ 2p=k. It is then found that with
the sampling intervals Dx ¼ Dy  k=2 the whole propagating part of the plane-wave
spectrum is obtained.
In some cases, especially at the very high frequencies with the AUT aperture
being electrically very large, for example, hundreds of wavelengths, the plane-wave
spectrum falls down very rapidly, reaching negligibly small values already at parts
of k0 . This is illustrated in Fig. 12.10, where the normalized simulated plane-wave
spectra of two uniform aperture antennas are shown. It is seen that for the 300 k
antenna the spectrum drops below −50 dB around kx ¼ 0:17 and thus, provided
that errors at the level of −50 dB are acceptable, the sampling intervals can be
increased 6 times from k=2 to 3k thus significantly decreasing the number of the
measured points and the total measurement time. On the other hand, using the
sampling intervals larger than k=2 will only provide the spectrum in the propor-
tionally smaller part of the k-domain and thus smaller angular sector, according to
the relation Dx ¼ Dy ¼ k=ð2 sinðhmax ÞÞ, and it also results in aliasing of the cal-
culated spectrum thus increasing the uncertainty of the calculated far-field [18].
426 A.V. Räisänen et al.

12.3.5 Measurement Errors and Uncertainties

One inherent feature of the PNF technique is the truncation of the measurement
plane at some finite dimensions, either due to the finite size of the scanner or driven
by the condition of minimizing the measurement time. Similar to the effects of
increasing the sampling interval, this truncation limits the obtained angular sector
and introduces truncation errors into the calculated field. The valid angle for the
calculated field is found from simple geometrical relations of the measurement
setup as follows hv ¼ tan1 ððL  DÞ=2z0 Þ, where L is the size of the measurement
plane in a given direction, for example, along the x-axis, D is the AUT aperture size
in the same direction, and z0 is the distance between the antenna aperture and the
measurement plane—the measurement distance.
Among other errors and uncertainties, the following must be noted due to their
increasing importance at very high frequencies. Since the probe is moved along two
linear paths, the RF cable connected to the probe changes the shape and thus its
electrical characteristics are slightly varied. This results in induced variations on the
received signal and uncertainties of the calculated far-field.
To mitigate the phase errors caused by bending cables, a pilot signal based phase
correction method has been developed [19]. A pilot signal, slightly at different
frequency than the measurement signal, is injected in the bending measurement
cable. A filter is implemented at the end of the cable inducing a reflection at the
pilot-signal frequency while letting the measurement signal pass through. The phase
of the pilot signal transmitted twice through the bending cable is observed and any
changes in the phase indicate also changes at the measurement frequency.
A correction can be applied taking into account the frequency difference. The
correction method is applied at 310 GHz and at 650 GHz for quiet-zone field
verification in hologram compact antenna test ranges [7, 8]. Harmonic mixers were
used as receivers and microwave frequency LO signals were transmitted through
flexing cables. Phase changes at microwave frequencies were translated to the phase
corrections at submillimeter wave test frequencies.
Another important uncertainty factor is nonideal planarity of the measurement
plane. Variations in the probe z position not exceeding k=100 rms are usually
required; this results in equivalent phase variations of ±3.6º, which is considered
acceptable. However, for example, at 650 GHz this converts to the planarity
requirement of ±4.6 µm, which is not that easy to achieve. In order to ensure
accurate planar movement of the probe, granite reference plates are used for
measurements at 550 GHz [20]. Tracking laser interferometers can be used to
measure the planarity of the near-field scanner. The measured planarity data can be
used either for phase correction of the near-field data or for active correction of the
probe movement during the data acquisition using an additional translation stage in
z-direction.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 427

The difficulties in accurate phase measurements have led to the development of

phaseless, amplitude only measurements where the phase is retrieved with
numerical methods [21–23]. However, these methods require measurements at two
different distances or use of reference antennas.
Other uncertainties of the PNF technique, listed here in no particular order, are
receiver nonlinearity, thermal noise, drift, leakage and cross-talk, multiple reflec-
tions between the antenna and probe and reflections from the surroundings, probe x-
and y-position uncertainties, etc. The detailed investigations of the PNF uncer-
tainties can be found in [24, 25].

12.4 Compact Antenna Test Range (CATR)

12.4.1 CATR: Applications and Specifications

CATRs have been widely used for antenna measurements since 1968, when the first
range was designed at Georgia Tech in the United States. Basically, a compact
antenna test range [26] is a hardware setup that transforms the spherical wave
generated by a feeder in a plane wave, to simulate the far-field conditions. Although
there are different setups, the most common are based on reflectors. This chapter
examines the application of reflectors, lens, and holograms.
CATRs are used for two main applications: antenna measurements and radar
cross-section (RCS) measurements. In the first case, the feeder generates a con-
tinuous wave, and after being reflected or modulated by the intermediate hardware,
a plane wave impinges in the volume where the antenna under test (AUT) is placed.
This volume is called quiet-zone. Therefore, the measurements can be performed as
in far-field systems [1]. Figure 12.11a shows the VAST12 antenna measured in the
facilities of Universidad Politécnica de Madrid. In the case of RCS measurement,
the signal (pulsed or continuous wave) impinges in the target, and it is reflected.
The same feeder (in a monostatic configuration), or another horn (in a
quasi-monostatic configuration) receives the field, and this field is compared to the
one reflected by the reference target (sphere, disk, or rectangular plate).
Figure 12.11b shows the measurement of a target in the facilities of INTA, Spanish
National Research Centre for Aerospace techniques.
The maximum dimension of AUT and frequency band depends on the
Quiet-Zone, the volume, where the AUT is located, in which the illuminating field
amplitude and phase differ from those of a plane wave by less than a pre-established
amount. Obviously, the field in the quiet-zone is not perfectly a plane wave; but a
taper and a ripple disturb this field. The taper is the amplitude variation in the
quiet-zone border induced by the feed pattern and path lengths to the reflector
surface. Typical reflector edge illumination is –1.5 to –2 dB and typical taper
specifications are from –0.5 to –1 dB. The ripple includes the field variations
produced by edge reflector diffractions and reflections from the room walls: typical
428 A.V. Räisänen et al.

Fig. 12.11 Compact antenna test ranges: a LEHA-UPM (Universidad Politécnica de Madrid) for
antenna measurements and b INTA (Spain) for radar cross-section measurements

Fig. 12.12 Sources of the

ripple and taper in a CATR

specifications for the ripple are ±0.5 dB in amplitude and ±5º in phase. This taper
and this ripple create some errors and uncertainties in the measurements and are
different for each AUT (depending on the size) and for each frequency bands. The
sources of ripple and taper are shown in Fig. 12.12, and can be summarized in:
• Surface inaccuracy, important for higher frequencies as millimeter waves.
• Edge diffraction, important for lower frequencies.
• Direct feed radiation to quiet-zone.
• Reflections from chamber walls.
• Multiple reflections between AUT and CATR feed.
All these sources of errors create spurious waves those are added with the
desired wave (the one reflected in the parabolic), and disturb the plane wave. In
general, the minimum frequency is limited by the diffraction associated with the
electrical size of the reflectors. For a 30-k reflector (serrations included) the ripple is
of 2 dB peak to peak; therefore a 10 m reflector is required for measuring at 1 GHz.
The maximum frequency is limited by the phase errors associated with the accuracy
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 429

of the reflectors surfaces. It is required a k/100 accuracy at the maximum desired

frequency.
The last important factor to be considered is the cross-polar level. This level
depends on the system used and on the feed cross-polar quality. Single offset
reflectors intrinsically introduce cross-polarization in the antisymmetric plane and
are not adequate for antenna measurements, although the cross-polar level can be
reduced using reflectors with a large focal distance to diameter ratio (F/D). Also,
dual reflector systems can be designed, using Mizuguchi–Dragone condition [27,
28] to cancel this cross-polarization.

12.4.2 Absorbers and Anechoic Chambers

CATRs are located in anechoic chambers. These chambers are closed areas,
shielded, and covered by electromagnetic absorbing material, which simulates free
space propagation conditions, due to the absorption of the radiation absorbing
material (RAM) [29]. These anechoic chambers are not only used for antenna or
RCS measurements, but also for anechoic chambers are used for measurements of
radiated emissions, radiated susceptibility or electromagnetic compatibility in
general. The main advantages of using anechoic chambers in antenna measurements
are the all weather conditions operation, the controlled environment (for instance in
temperature or cleanness), the security aspects, and the absence of interferences.
The main drawback is the large maximum dimension leading to a large, expensive
special space.
Anechoic chambers are covered by electromagnetic absorbers. These absorbers
are fabricated with a material that simulates free space conditions. This material is
fabricated from polystyrene with a dielectric constant close to one in order to avoid
reflections, filled with carbon graphite material to absorb the impinging signal.
Usual absorbing materials have pyramidal shape in order to avoid the reflections
and they should be long enough in terms of wavelengths. Therefore, for low fre-
quencies these pieces of absorbers can be very long and heavy. The signal should
impinge close to normal incidence to maintain good absorption properties; therefore
the chambers cannot be very long and narrow. Table 12.1 presents the specifica-
tions of one of the providers of absorbing materials (AEMI) [30]. In this case they
are the classical pyramidal ones. It can be observed the reflectivity in terms of the
frequency and the height of the material: it can be observed that around 9 wave-
lengths are required for getting 50 dB of reflectivity.
There is also another kind of absorbing materials: in the walls of the CATR,
wedge materials are placed: these materials guide the signal to the back wall of the
chamber, and there, the pyramidal ones absorb the signal. For millimeter wave-
lengths, convoluted materials are used instead of pyramidal ones. For covering
430 A.V. Räisänen et al.

Table 12.1 Specifications of pyramidal absorbers (AEMI-MVG) at RF and microwave

frequencies
AEP-4 AEP-8 AEP-18 AEP-36 AEP-60 AEP-96
Geometry Normal N N N N/T N T
(N)/
Twisted
(T)
Height cm 10.2 20.3 45.7 91.4 152.4 243.8
Pyramids/block 144 64 16 4 1 1
Weight kg/pc. 1.1 2 4.1 8.4 13.6 27.2
Power W/m2 775 775 775 775 775 775
Absorption (dB) at 30 MHz 5 8 15
normal incidence at 125 MHz 15 30 35
500 MHz 30 37 40 45
1.0 GHz 30 40 42 47 50
3.0 GHz 30 37 45 50 50 50
10.0 GHz 40 50 50 50 50 50
18.0 GHz 50 50 50 50 50 50

Table 12.2 Maxima of monostatic reflectivities of absorbers at 310 GHz [31]

Absorber Reflectivity, max Reflectivity, max Transmission
(dB) H-pol (dB) V-pol (dB) V-pol
FIRAM-500 −24
Wedges horizontal −30 −26
Wedges vertical −31 −30
TERASORB-500 −33
Wedges horizontal −29 −41
Wedges vertical −41 −30
TK THz RAM −51 −56 −50
Eccosorb −50 −55 −50
VFX-NRL-2
Carpet #1 −40 −40 −17
Carpet #2 −40 −40 −25
Carpet #3 −21 −20 −5

masts and towers, laminated materials are used, while for the paths, there are special
pieces as walkway ones. During the last years, rubberized materials have been
fabricated to improve the cleanness and durability of the material. Also, there are
some providers that offer weatherproof absorbers for outside measurements.
Table 12.2 shows measurement results of some commercial absorbers at 310 GHz.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 431

12.4.3 Reflector Compact Antenna Test Ranges

Among the different kinds of compact antenna test ranges, reflectors based are the
most common. There are different designs, but in a first division, there are single or
dual reflectors. The single offset reflectors, also known as virtual vertex systems, are
the simplest ones. They use a large focal distance to diameter (F/D), 0.8–1.2, and
low offset angle, in order to improve the cross-polar performance and to equalize
the amplitude tapering. In this case, the focus of the parabolic reflector coincides
with the phase center of the feeder. The main advantages are that they are much less
expensive than dual reflector systems and easier to design and align. To reduce the
scattering produced by the edge of the reflector, two main techniques are used:
serrated and rolled edge reflectors. Figure 12.13 shows both designs (from
ORBIT/FR [30]). The serrations control the direction of the diffracted signals,
sending them in the direction of the peak. The rolled edge controls the direction of
the reflected signal on the edge of the reflector. In both cases, these designs make a
better quiet-zone. In a first approach the diameter of the quiet-zone is 35% of the
diameter of the reflector. The quiet-zone is defined by the taper (typically 1 dB in
amplitude and ±2° in phase), ripple (typically ±0.5 dB in amplitude and ±5° in
phase), and cross-polar level (typically −30 dB for single offset reflectors). The
minimum frequency depends on the dimension of the reflector, and typically the
diameter is around 30 wavelengths.
If the cross-polar level has to be improved, dual reflectors should be designed.
The easiest system is the dual cylindrical reflector configuration, or Vokurka system
[32]. It uses two cylinder parabolic reflectors with single curvature: the subreflector
with the horizontal axis and the main reflector with the vertical axis. This system
collimates in elevation and azimuth the spherical wave from a point source, with a
moderate reflector cost with high accuracy (single curvature reflectors). This system

Fig. 12.13 Serrated and rolled edge reflectors (ORBIT/FR systems)

432 A.V. Räisänen et al.

Fig. 12.14 ESA/ESTEC 100000 cleanroom compensated compact range in the HERTZ facility

gets good linear cross-polarization performance in V and H planes (although bad

performance in diagonal planes ±45º). If a better cross-polar is desired, compen-
sated compact ranges are designed, using Cassegrain or Gregorian configurations.
Cassegrain configuration is composed by a hyperboloid, as subreflector, and a
paraboloid as main reflector. Figure 12.14 shows the ESA/ESTEC [33]
Compensated Compact Range using a Cassegrain configuration. This CATR shares
the room with a near-field antenna measurement range, both planar and spherical.
Another design for improving the cross-polar is the Gregorian configuration, that is,
an ellipsoid and a paraboloid. As in the case of the Cassegrain configuration, one of
the foci of the ellipsoid is located in the phase center of the feeder, and the other
coincides with the focus of the parabolic reflector.

12.4.4 Lens and Hologram CATR

As mentioned in earlier section, reflectors are most commonly used as the CATR
collimating elements. However, the stringent surface accuracy requirements, and
therefore, difficulties in accurate manufacturing and high costs have lead a way to
the research of the use of other collimating elements such as lenses or holograms for
millimeter and submillimeter wavelengths. Figure 12.15 shows schematics of
CATRs based on a dielectric lens and a transmission type of hologram.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 433

Fig. 12.15 Schematic of CATRs based on a a lens and b a transmission type hologram

Lens CATR
The first lens-based CATR was presented in the end of the 1970s [34]. Research on
mm-wave lens CATR have been reported, e.g., in [5] where a CATR based on a
Teflon lens operating at 110 GHz is presented and in [4] where measurements with
a 50 cm lens to facilitate the designing of a 30 cm reflector antenna at 94 GHz are
presented. The lens CATR could include several lenses, e.g., the first one shapes the
illumination of the second collimating lens. The achievable quiet-zone diameter is
about 70…75% of the lens diameter [5, 35].
The surface accuracy requirement for the CATR lenses is not as stringent as for
reflectors because of the transmission type of operation. The accuracy requirement
pffiffiffi pffiffiffiffi 
set for a single CATR reflector is multiplied by a factor of 2= er  1 [5] for
similar performance. This factor takes into account the two surfaces of the lens. The
factor encourages the use of low-permittivity lens. However, low-permittivity
materials are often foam-like materials the homogeneity of which is not easy to
control. Also, the lens thickness increases with the decreasing permittivity of the
lens material. For example, when er ¼ 1:1 and the F/D ratio is 3 the thickness of the
lens is of the same order than the diameter [35]. Also, an integrated lens antenna
could be used, i.e., range feed is in direct contact with the lens. However, also then
the lens becomes large compared to achievable quiet-zone size. The dielectric
constant of the lens affects also the reflection from the lens surface: the larger er is
the more of the power radiated by the feed is reflected from the lens surface.
Matching layers can be used to reduce the reflections.
The edge diffraction has less effect on the performance of the lens CATR than on
that of the reflector CATR [36]. However, for minimizing the edge diffraction,
serrated metal plates may be used with the lens [5] or the surfaces of the lens can be
shaped, analogous to rolled edges in reflectors, in order to achieve an appropriate
aperture distribution.
Hologram CATR
The use of computer-generated holograms as the collimating element of a CATR was
introduced in the beginning of the 1990s [37]. The hologram-based CATR is a
potential antenna measurement method especially for millimeter and submillimeter
wavelengths. Figure 12.15b shows the principle of the CATR based on a hologram in
434 A.V. Räisänen et al.

a transmission type of operation. The planar hologram structure is an interference

pattern of the CATR feed antenna and the desired planar wave front. The hologram
modulates the spherical wave front radiated by a feed horn in such a way that a planar
wave is formed. The transmission-type amplitude hologram is a binarized interfer-
ence pattern on a thin metal sheet supported by a plastic substrate. The hologram
either lets the field go through (slot) or blocks the field (metal). The slots are tapered
towards the edges in order to reduce edge diffraction. The hologram is often designed
in such a way that the plane wave leaves the hologram at an angle of 33° with respect
to the hologram normal. Thus, the diffraction modes, which are propagating to the
hologram normal direction, do not disturb the quiet-zone field [38].
The hologram CATR may also be based on a reflection-type amplitude holo-
gram, or reflection- or transmission-type of phase holograms [39, 40]. The phase
holograms are manufactured by milling a groove structure to a dielectric (trans-
mission type) or metal (reflection type) plate. In transmission type of operation the
surface accuracy requirements are more relaxed compared to the reflection type of
operation. In particular, for the phase holograms, the difference is similar to that
between the reflectors and lenses discussed in the previous section.
So far, transmission type amplitude holograms are used for antenna measure-
ments [8, 41]. The hologram is assembled into a frame, which ensures sufficient
planarity for the hologram. The accuracy requirement for the hologram pattern is
about 0.01–0.02 k, which is of the same order as the surface accuracy for the
reflector. However, the manufacturing of the two-dimensional hologram is easier
than the manufacturing of the three-dimensional reflector surface, thus the manu-
facturing costs are reduced substantially. Manufacturing techniques allow produc-
tion of large holograms up to 3-m diameter with accurate enough pattern resolution
and accuracy up to THz operation. Figure 12.16 shows an artistic view of a
hologram CATR used for 1.5-m reflector antenna tests at 650 GHz [8].

Fig. 12.16 Artistic view of hologram CATR used for 1.5-m test antenna measurements [8]. Some
absorbing material is removed for clarity
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 435

The relation between the hologram and the quiet-zone sizes depends on the
required size of the quiet-zone and frequency in a similar way as in the case of
reflector CATRs. The proportion of the tapering area decreases when the size of the
hologram increases at the given frequency. Therefore the ratio of the quiet-zone and
hologram sizes is larger for larger holograms. For example, at 119 GHz for the
550 mm hologram the quiet-zone diameter is about 43% of the hologram diameter
while for the 2.4 m  2.0 m hologram the quiet-zone diameter is about 70% of the
smaller dimension [6, 42].
A CATR based on a reflector or a lens operates equally well with all polarisa-
tions and frequencies (the edge diffraction and the surface accuracy give limita-
tions). However, the operation of a hologram depends fairly strongly on the
frequency and polarization, because it is a diffractive element. In principle, a
hologram can have an octave bandwidth [43] if the feed location is moved while
changing the operation frequency. However, the substrate thickness of mechani-
cally reasonable substrate may reduce the bandwidth of the hologram CATR.

12.4.5 Antenna Measurements in CATR

Antenna measurements in a CATR are carried in a similar way than in a conven-

tional far-field antenna test range. The AUT is placed in the quiet-zone region
where the planar wave front is created. Antenna is rotated with an antenna posi-
tioner and the radiation pattern is recorded.
Feed scanning, i.e., movement of the CATR feed can be used for radiation
pattern measurements in small angular region. The plane wave steers with respect to
the feed offset from the focal point of the collimating element. The applicable
angular region depends on the quiet-zone field requirements. See also Sect. 12.4.7
for feed scanning antenna pattern comparison [44].
For the antenna gain measurements, a gain comparison method is well appli-
cable. The measured amplitude level of the AUT is compared to the amplitude level
of a known reference antenna. In principle, the gain of the reference should be of
the same order as the AUT gain in order to obtain accurate results. It is recom-
mended, that the quiet-zone of the CATR is verified by probing the field with a
small probe antenna, e.g., an open-ended waveguide or a small horn antenna. If the
probe gain is known, the gain of the AUT can be estimated by comparison against
the probe antenna gain. For electrically large antennas the gain difference may be
large, e.g., 50–60 dB which leads to stringent linearity requirement of the mea-
surement system or use of accurate attenuators.
436 A.V. Räisänen et al.

12.4.6 RCS Measurement in CATR

CATR can be also used for radar cross-section measurement. In this case, the signal
is reflected by the target and received by one probe. The received signal is com-
pared with the one received using a reference target. For this kind of measurements,
it is important to separate the effect of the target from the effect of the surrounding
elements. In this case, the effect of the pylon is really important. Figure 12.11b
shows a knife pylon, designed to reduce the reflection. Other pylon designs are
cylindrical polystyrene ones: in this case the dielectric constant is close to 1, to
reduce the reflected wave. However, other techniques are used to reduce the effect
of the reflections in the surroundings of the target. One of them is the “empty
chamber” method. The measurement is repeated with target and without target, and
the complex difference between both measurements is calculated. The reflectivity of
the target can be calculated with the following equation:
   
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  Etarget  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  Emeastarget  Eemptychamber 
rtarget ¼ rreference
 ¼ rreference

Ereference  Emeasreference  Eemptychamber 
ð12:4:1Þ

For measuring RCS, pulsed systems are usually used, in order to isolate the
transmitted ray from the received wave. However, these systems are expensive, and
not always available. Continuous wave transmitters can be used if a measurement in
a wideband is performed and time filtering is used. The received signal is trans-
formed to time domain, a time filter is applied, and it is transformed again to
frequency domain. In this case, monostatic configurations do not provide enough
isolation between transmitter and receiver, and quasi-monostatic configurations
(with two different probes) are used.

12.4.7 Antenna Pattern Correction and Range Evaluation

Techniques

The quality of the quiet-zone (QZ) field limits the measurement accuracy in a
CATR. The QZ field is ideally a plane wave, but in practice it is distorted by many
unwanted effects. Common sources of distortions include (1) edge diffraction of the
collimating element, (2) surface errors of the collimating element, (3) direct leakage
from the range feed to the QZ, (4) multiple reflections between the AUT and the
range, and (5) range reflections. Figure 12.17 illustrates these error sources. The
level of the distortion signals compared with the level of the desired plane wave
should be much lower than the measured side lobe level of the AUT. Otherwise
distorting signals may cause a significant error the measured results.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 437

Fig. 12.17 Common error

sources in a CATR include
edge diffraction, surface
errors of the collimating
element, direct leakage from
the range feed to the QZ,
range reflections, and multiple
reflections

The presence of distorting signals in CATRs raises many questions, such as:
• How accurately can a given antenna be tested in the CATR?
• How could the CATR be improved by identifying and eliminating distortion
sources?
• As distortion signals are always present, is it possible to reduce computationally
their effect?
Many methods have been developed to address the above questions. Some of the
methods have been primarily developed for correcting measured antenna pattern in a
distorted quiet-zone field, while others have been developed for identifying distortion
sources in the range. Most of the methods, however, are suitable for both purposes.
Antenna pattern correction is always based on some additional information. For
instance, the antenna can be measured multiple times in different locations in the
CATR or the measurements can be performed at different frequencies. Based on the
additional information used, the methods can be categorized into four groups:
1. Methods utilizing the time or frequency response of the CATR
2. Methods utilizing the spatial response of the CATR
3. Methods utilizing the probed quiet-zone field of the CATR
4. Methods utilizing information on the physical size of the AUT.
The following reviews antenna pattern correction techniques in each
group. Some of the reviewed techniques may not have been developed for CATRs,
but are reviewed here because of their educational aspects or because they are
potentially feasible for CATRs. Finally, this section introduces methods that are
solely meant for identifying distortion sources in a CATR.
Time and frequency techniques
Distorting signals often follow other than the wanted signal path from the range feed
to the AUT. In such a case, the distortion and desired signals delay differently during
438 A.V. Räisänen et al.

Fig. 12.18 Spurious signals that arrive after a different time-delay than the desired signal, are
gated out in the time gating method

their propagation through the CATR. Time and frequency techniques utilize the time
or frequency response of the CATR for eliminating the undesired signal contribu-
tions from the measured response. Figure 12.18 illustrates a situation where there are
the desired signal and distortion signals in a CATR. When the range feed transmits a
pulse, the receiver records three differently time-delayed and attenuated copies of it.
The desired signal can be picked out from other signals in the time domain.
Multipath signals can be separated from each other when the pulse length is
smaller than the path length difference. The usable pulse length and thus also the
correction efficiency is limited by the bandwidth of the equipment or the AUT. For
instance, if the relative bandwidth is 10%, multipath signals whose path length
differs by 10% or more from that of the desired signal can be eliminated.
Measurements can be performed either in time or frequency domain, because
signals in both domains contain the same information. Measurements are often
performed with a network analyzer offering the possibility to sweep the frequency.
This possibility also makes the implementation of time and frequency methods
convenient: No additional equipment is needed and the measurement time does not
considerably increase.
There are various time and frequency domain techniques differing in how the
corrected pattern is obtained from the measured time or frequency domain response.
Simple Fourier transform is often used to obtain the time domain response from the
frequency response. An implementation of a Fourier transform based time gating at
microwave frequencies is presented in [45]. Instead of the Fourier transform, other
methods have been used to obtain multipath components, such as the Matrix-pencil
[46] and oversampled Gabor transform [47]. These methods are potentially more
accurate over the Fourier transform.
Depending on the measurement equipment, it is not always possible to preserve
phase coherence across the whole frequency band. In such a case, Fourier
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 439

Fig. 12.19 A compact

antenna test range with a
direct signal and a reflected
signal

transformation cannot be used. A frequency shift technique capable of correcting

the antenna pattern from noncoherent frequency response is described and
demonstrated at 310 GHz in [48].
Spatial methods
Spatial methods utilize the spatial response of the test range. That means that the
AUT is measured several times at different locations in the test range. Let us
consider the test range shown in Fig. 12.19, where there is the desired signal and
one reflected signal arriving at the AUT. In each measurement point, the AUT
receives a signal that is a vector sum of the desired and reflected signal. The phase
of each vector depends on the spatial location of the AUT. For example, if the AUT
is moved transversally to the propagation direction of the direct signal, the phase of
direct signal remains unaltered, whereas the phase of reflected signal changes. This
way signals arriving from different directions can be separated from each other.
The antenna pattern comparison (APC) method was first used for determining
the reflectivity level of the chamber [49], but the method can also be used for
antenna pattern correction. In this method, the antenna pattern of the AUT is
measured several times at different locations in the range. The variations between
the measured patterns are related to the reflectivity level of the range and the
corrected pattern can be obtained as an average of the measured patterns.
Instead of simple averaging of the measured antenna patterns, the corrected
pattern can be obtained in other ways. In the virtual array method the antenna
pattern is measured twice. During the first measurement, the AUT is kept in place,
whereas it is spatially displaced as a function of rotation angle during the second
measurement. The displacement is selected such that when the two patterns are
summed, the desired signals sum in phase and the signals arriving through the main
beam of the antenna sum in the opposite phase cancelling each other.
440 A.V. Räisänen et al.

Somewhat similar approach can also be used even if the AUT is not displaced as
a function of the antenna rotation angle. The measurements performed at different
locations form a synthetic antenna array at each rotation angle of the AUT, provided
that the measurements are coherent in each position. If the precise locations of the
AUT are known, it is possible to synthesize an array that has a high gain towards
the desired signal, and high attenuation towards possible spurious signals.
Examples of such methods have been presented in [50–52].
The aforementioned spatial methods necessitate the possibility to move the
AUT. This may be challenging in particular if the AUT is heavy and if it needs to
be moved with a sub-wavelength precision. It has been shown in [44], that instead
of potentially heavy AUT, the range feed can be moved to obtain a phase difference
between the direct and reflected signals. The method has been demonstrated at
310 GHz in a hologram-based CATR.
Techniques utilizing the measured Quiet-Zone field
If the quiet-zone field differs from the plane wave field, the measured antenna
pattern is erroneous. However, if the quiet-zone field is precisely known, it is, at
least in principle, possible to solve the true antenna pattern from the measured
antenna pattern and the quiet-zone field. There are two ways of measuring the
quiet-zone field. The quiet-zone field can either be probed with a near-field scanner
or be found out by measuring a precisely known reference antenna in it.
A correction method utilizing measured quiet-zone field is presented in [53].
Somewhat similar method was used in a hologram-based CATR at 650 GHz [8].
Here the quiet-zone field was probed and it was calculated how the simulated
antenna pattern should look if it was measured in that quiet-zone field. This result
was then compared with the measured antenna pattern. Although the measured
antenna pattern was not corrected with the method, it was found out that some
unexpected features in the measured antenna pattern were due to quiet-zone field
instead of fabrication errors of the AUT.
Methods utilizing information on the size of the AUT
Antenna diagnostics refers to an operation in which the near-field or surface cur-
rents of the AUT are solved from its measured far-field pattern. Near-field or
surface currents are solved for diagnostic purposes, i.e., to identify possible mal-
functions or manufacturing errors in the antenna structure. This approach offers a
possibility for antenna pattern correction as well. For instance, the near-field may
reveal that part of radiation comes outside the physical volume of the antenna. In
such a case, the field outside the physical volume of the AUT can be filtered out. An
example of such a method is presented in [54].
Other range diagnostic techniques
A quiet-zone field of a CATR is often probed in order to verify that the field meets
certain specifications. In addition to range verification, the probed field can be used
for identifying spurious signals in the range. After identifying the signal sources,
the range can be modified to remove them.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 441

Various methods for locating the spurious signals have been presented. Many of
the methods are reviewed in [55]. One way is to calculate the plane-wave spectrum
of the quiet-zone field. Large contributions outside the desired signals indicate a
potential spurious signal. The plane-wave spectrum has been calculated with a
Fourier transform [7] and superresolution techniques [56, 57]. These methods
provide information on the direction of possible spurious signal sources.
Plane-wave spectrum gives correct information on the direction of spurious
signal sources when the signal source is in the far-field of the probed area. In
CATRs, however, the spurious signal sources are typically in the near-field of the
probed area and, therefore, they do not appear as a peak in the plane-wave spec-
trum. To overcome this challenge, a near-field focusing method is proposed for
locating the sources of spurious signals [55]. In this method, the phase of the probed
field is adjusted to focus at a certain point in the image domain. After repeating this
operation for every point in the image domain, spurious signal sources should
appear as high-intensity points in the image.
The aforementioned methods can also be utilized in time domain provided that
quiet-zone field is probed in time or frequency domain. Time dimension gives addi-
tional information for example on the distance of the source of the spurious signal.

12.5 Special Topics

12.5.1 Holographic Principles Used in Antenna

Measurements at mm and submm Wavelengths

Holographic imaging is discussed in detail in Chap. 13 “Holographic imaging

approach”. It is highly recommended that the reader studies Chap. 13 before going
into details of this subsection.
Microwave holography has been under intensive research since the mid-1960s,
and has been applied, e.g., for antenna design and antenna measurements since
1970s. Holographic principles may be applied directly into the antenna metrology
for phase retrieval when using a reference antenna, or in producing plane wave
conditions for the measurement. In indirect holography, scalar measurements are
sufficient, and the use of rather expensive vector measurement equipment described
in Sect. 12.2 can be avoided.
In the following, we briefly discuss how holographic principles have been used
in antenna metrology in the millimeter wave and THz bands. As we will see,
microwave holography has provided many advantages in antenna metrology at
millimeter and submillimeter wavelengths. In measurement of large radio astron-
omy telescopes, it has become a universal method in determining the reflector
surface accuracy. In many other measurements it may bring more accuracy and
better insight to the unideal antenna radiation performance. A computer-generated
radio hologram is a prospective alternative to large reflectors as a collimating
element in CATR at THz frequencies (see Sect. 12.4.4).
442 A.V. Räisänen et al.

Fig. 12.20 Setup for indirect off-axis holographic imaging in order to be able to detect the
interference pattern of the object field (field scattered by the object when illuminated by the source
field) and the reference field (lhs), and the applied in antenna measurement (rhs)

Fig. 12.21 Hologram

formation geometry (after
[58])

The schematic of indirect holographic imaging of Fig. 13.8 is repeated in

Fig. 12.20 (lhs) and then applied to indirect off-axis holographic antenna mea-
surement in Fig. 12.20 (rhs).
Holographic principles used in determining the surface accuracy of large reflectors
Microwave holographic metrology of large reflector antennas was introduced by
Bennett et al. in 1970s [58], see Fig. 12.21. Since then many major millimeter wave
and THz radio telescopes for radio astronomy have undergone a holographic
measurement during the reflector panel adjustment or as a final check of the
reflector surface quality, see, e.g., [59] (see Fig. 12.22), [60, 61].
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 443

Fig. 12.22 Feed system for

holographic measurements of
the 13.7-m telescope of
University of Massachusetts
at 38 GHz. The MIT Lincoln
Laboratory LES-8
geostationary satellite was
used as the source (after [59])

The antenna under test (AUT), typically a large telescope antenna, is pointed to
a far-field source: a radio astronomical source (maser) or a satellite equipped with a
mm-wave transmitter, or to a far-away terrestrial source. Also a synchronous ref-
erence receiver with a lower-gain antenna is pointed to the same source. Then the
AUT scans the source using its azimuth-elevation rotating mechanism. The phase
difference pattern of the two signals (from AUT and reference receivers) over the
aperture reveals possible deviations of the AUT reflector surface from the ideal
parabolic shape.
Holography for phase retrieval in near-field antenna measurements
A method to determine the AUT aperture field distribution without phase mea-
surement, but with a synchronous reference transmitter, was first studied by Napier
and Bates in early 1970s [62].
This holographic method for phase retrieval in planar near-field antenna mea-
surements at THz frequencies was theoretically studied by Junkin et al. [23, 63].
A small and coherent reference antenna is placed next to the antenna under test
(AUT) and the interference pattern, i.e., hologram, of the ensemble of the AUT and
reference antenna is recorded. The phase of the AUT can be retrieved from the
hologram by numerical means when the radiation characteristics of the reference
antenna are known. Recently Arboleya et al. have demonstrated a method for
broadband antenna characterization from planar phaseless acquisitions [64]. The
method is based on off-axis indirect holography and use of a reference antenna as
the methods above. However, a frequency sweep is performed at every acquisition
point which allows retrieval of the phase at each frequency point and at each spatial
444 A.V. Räisänen et al.

point independently. This allows less dense sampling than required in the con-
ventional off-axis holography.
Holograms used to create plane wave in CATR
If a hologram is synthesized by computational methods; we call it a
computer-generated hologram [65]. Such a hologram is determined by computing
the image pattern that will change an incident beam into a desired beam. Usually
such holograms are designed by assuming that the image pattern modulates either
only the amplitude or the phase of the incident field known as amplitude or phase
holograms. In order to facilitate the production of a hologram, binary coding is
used: the amplitude or phase is described using only ones and zeros. For example, a
binary amplitude hologram fully transmits the incident field at some positions, but
blocks it completely at some other positions.
Such a computer-generated binary hologram is the basis for the holograms that
can be used in antenna testing for creating a plane wave field needed in mea-
surements. If the hologram is printed as a metal pattern on a thin dielectric substrate,
the resulting hologram can be used as a collimating element in a compact antenna
test range (CATR) instead of a lens or a reflector.
At Aalto University, use of radio frequency holograms in CATRs has been
extensively studied. The holograms are planar structures, which is advantageous with
respect to the manufacturing. Demonstrations have been carried out at the frequency
range of 40–650 GHz. The major part of the experiments is carried out using trans-
mission type binary amplitude holograms, which consist of an etched metal structure
on a supporting dielectric film. Large test objects, such as the Odin space telescope
1.1-m antenna and the 1.5-m prototype reflector antenna by EADS Astrium have been
tested at 119 GHz and at 320 GHz and 650 GHz, respectively, in a hologram CATR
[6–8, 41, 42]. The challenges in the use of holograms are the rather narrow operation
band (*10%) and the polarization limitations, especially with amplitude holograms.
Very recently, also a wideband hologram-based CATR has been demonstrated [43].
From the above discussion we can conclude that microwave holography has
provided many advantages in antenna metrology at millimeter and submillimeter
wavelengths. In measurement of large radio astronomy telescopes, it has become a
universal method in determining the reflector surface accuracy. In many other
measurements it may bring more accuracy and better insight to the unideal antenna
radiation performance. A computer-generated radio hologram is a prospective
alternative to large reflectors as a collimating element in CATR at THz frequencies.

12.5.2 Measurement of Small Integrated Antennas

Advancement of integrated circuit technologies has enabled the use of even higher
frequencies in wireless communication and other applications. At the same time, the
antennas having size proportional to the wavelength are getting smaller. The aim is
to integrate the antennas on an IC chip possibly measuring only a few mm.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 445

Fig. 12.23 A probe station and typical space around it. A microstrip antenna (AUT) at the center
of chuck connected to a VNA with a ground-signal-ground (GSG) probe

The measurement techniques presented previously, far-field, near-field, and com-

pact antenna test ranges, may encounter difficulties when measuring integrated
antennas or antennas which can only be accessed with an on-wafer probe. There are
often a lot of scattering objects in the vicinity of the AUT when it is measured on a
probe station. See Fig. 12.23 for a typical probe station for on-wafer measurements.
Also, the rotation of the antenna is impractical because of the probe heads that
are fixed on the probe station. Therefore, in far-field measurements, the measuring
probe antenna is moved along a surface of a sphere centered at the AUT phase
center. The movement of the probe is usually produced using pivoted arms as
shown in Fig. 12.24. The free rotation of the probe antenna around the AUT may be
limited by other blocking objects like microscopes and supporting structures.
Therefore, custom-made mounting fixtures are required for the antennas in order to
enable rotation of the measuring antenna around the AUT. Also, the polarization of
the probe antenna should be correct. Far-field measurements of integrated antennas
are reported, e.g. in [66–70].
Often, the compact antenna test ranges are not practical for measuring small
integrated antennas because there is no possibility to rotate the AUT and, therefore,
the collimating element with its feed should be rotated around the AUT.
Near-field technique for antenna characterization on a probe station has been
demonstrated by Klein et al. [71]. The near-field technique is sensitive to phase
measurement errors because the far-field properties of the AUT are solved from the
complex near-field. The phase errors in the near-field result in an amplitude error in
the antenna pattern. During the probe movement, the bending RF cables easily
change their electrical length resulting in phase errors. The phase errors may be
corrected in some extent using error correction techniques, e.g., pilot signal based
446 A.V. Räisänen et al.

Fig. 12.24 Typical mechanical assembly for movement of the probe on a spherical surface

method described in Sect. 12.3.5. Accurate mechanical assembly of the near-field

scanner may be difficult to accomplish because of the restricted space around the
probe station.
The above-mentioned techniques require a probe antenna which may be difficult
to facilitate on a probe station environment due to size limitations. Calazans et al.
have proposed a technique where there is no need for a probe antenna [72]. A metal
wire is moved in front of the AUT and the antenna radiation pattern is solved from
the measured reflection coefficients. This method enables to solve the antenna
pattern on the plane normal to the wire.
Another antenna measurement method avoiding the use of a probe antenna has
been proposed in Aalto University [73, 74]. In this method, the reflection coefficient
of the antenna is measured when there is a reflective object in the vicinity of the
radiating AUT. The measurement is repeated with many known reflective objects.
The field propagation from the AUT to the reflective object and after the reflection
back to the AUT is modeled with Fourier techniques for each object using an initial
guess for the antenna aperture field. Iterative methods are then used to find such an
antenna aperture field that best explains the modeled propagation and measured
reflection coefficients. The radiation characteristics can be calculated from the
retrieved antenna aperture field. Figure 12.25 shows the concept of the measure-
ment method. A flat metal strip is used as the reflective object and measurements
are carried out with different strip positions. The setup can be arranged in a small
volume and, because it avoids the movement of the probe antenna, the problems
with flexing cables are also avoided. Therefore, the method offers some benefits for
testing antennas on probe stations.
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 447

Translation stage S11

Flat metal strip

Antenna under test On-wafer probe

Antenna port

Fig. 12.25 Potential measurement setup for characterizing an antenna on a probe station. The
translation stage is used to alter the position of the reflective flat metal strip

The method was first demonstrated with a pyramidal horn antenna [73] and then,
in a probe station environment, using a planar antenna array as an AUT [74]. The
limitations of the method are the limited angular range, at maximum the mea-
surement range is the half hemisphere, and multiple reflections which may need to
be mitigated by time gating or other methods. Also, the inversion problem is
complex and the associated iterative numerical methods have to be selected care-
fully in order to ensure reliable convergence.

References

1. G.E. Evans, Antenna Measurement Techniques (Artech House Inc. 1990)

2. D. Slater, P. Stek, R. Cofield, R. Dengler, J. Hardy, R. Jarnot, R. Swindlehurst, A large
aperture 650 GHz near-field measurement system for the earth observing system microwave
limb sounder, in Proceedings of 23rd AMTA, Oct. 2001, Denver, USA, pp. 468–473
3. J. Hartmann, J. Habersack, H.-J. Steiner, J. Lemanczyk, P. de Maagt, Calibration and
verification measurements in compensated compact ranges up to 500 GHz, in Proceedings of
23rd AMTA, Denver, USA, Oct 2001, pp. 377–382
4. W. Menzel, B. Huder, Compact range for millimeter-wave frequencies using a dielectric lens.
Electron. Lett. 20(19), 768–769 (1984)
5. T. Hirvonen, J. Tuovinen, A. Räisänen, Lens-type compact antenna test range at mm-waves,
in 21st European Microwave Conference Proceedings (Microwave Exhibitions and
Publishers Ltd., Kent, England, 1991), pp. 1079–1083
6. J. Ala-Laurinaho, T. Hirvonen, P. Piironen, A. Lehto, J. Tuovinen, A.V. Räisänen, U. Frisk,
Measurement of the Odin telescope at 119 GHz with a hologram type CATR. IEEE Trans.
Antennas Propag. 49(9), 1264–1270 (2001)
7. A. Lönnqvist, T. Koskinen, J. Häkli, J. Säily, J. Ala-Laurinaho, J. Mallat, V. Viikari,
J. Tuovinen, A.V. Räisänen, Hologram-based compact range for submillimeter-wave antenna
testing. IEEE Trans. Antennas Propag. 53(10), 3151–3159 (2005)
8. A. Karttunen, J. Ala-Laurinaho, M. Vaaja, T. Koskinen, J. Häkli, A. Lönnqvist, J. Mallat, A.
Tamminen, V. Viikari, A.V. Räisänen, Antenna tests with a hologram-based CATR at
650 GHz. IEEE Trans. Antennas Propag. 57(3), 711–720 (2009)
9. A.V. Räisänen, Frequency multipliers for millimeter and submillimeter wavelengths. Proc.
IEEE 80(11), 1842–1852 (1992)
10. G. Carpintero, L.E. García Muñoz, H.L. Hartnagel, S. Preu, A.V. Räisänen (eds.)
Semiconductor Terahertz Technology: Devices and Systems at Room Temperature
Operation (Wiley, 2015)
448 A.V. Räisänen et al.

11. Keysight Technologies, http://www.keysight.com/en/pc-1000000457%3Aepsg%3Apgr/

network-analyzer?nid=-536902444.0&cc=US&lc=eng
12. Anritsu, http://www.anritsu.com/en-us/products-solutions/products/ms4640b-series.aspx
13. Rohde & Schwarz. https://www.rohde-schwarz.com/en/products/test-measurement/network-
analyzers/pg_overview_64043.html
14. Virginia Diodes Inc. http://vadiodes.com/index.php/en/about-vdi
15. R.E. Collin, Antennas and Radiowave Propagation (McGraw-Hill, 1985)
16. A.D. Yaghjian, An overview of near-field antenna measurements. IEEE Trans. Antennas
Propag. 34(1), 30–45 (1986)
17. C.A. Balanis, Antenna Theory: Analysis and Design, 3rd edn. (Wiley, 2005)
18. M.H. Francis, Aperture-sampling requirements. IEEE Antennas Propag. Mag. 39(5), 76–81
(1997)
19. J. Säily, P. Eskelinen, A.V. Räisänen, Pilot signal-based real-time measurement and
correction of phase errors caused by microwave cable flexing in planar near-field tests. IEEE
Trans. Antennas Propag. 51(2), 195–200 (2003)
20. D. Slater, “A 550 GHz near-field antenna measurement system for the NASA Submillimeter
Wave Astronomy Satellite,” in Antenna Measurement Techniques Association Conference,
October 3–7, 1994
21. R.G. Yaccarino, Y. Rahmat-Samii, A comparison of conventional and phaseless planar
near-field antenna measurements: the effect of probe position errors, in Proc. of IEEE Int.
Conference on Phased Array Systems and Technology, Dana Point, CA, USA, 21–25 May
2000, pp. 525–528
22. T. Isernia, G. Leone, R. Pierri, Radiation pattern evaluation from near-field intensities on
planes. IEEE Trans. Antennas Propag. 44(5), 701–710 (1996)
23. G. Junkin, T. Huang, J.C. Bennett, Holographic testing of terahertz antennas. IEEE Trans.
Antennas Propag. 48(3), 409–417 (2000)
24. A.C. Newell, Error analysis techniques for planar near-field measurements. IEEE Trans.
Antennas Propag. 36(6), 754–768 (1988)
25. IEEE Recommended Practice for Near-Field Antenna Measurements, IEEE Std 1720-2012
(IEEE, New York, 2012)
26. IEEE Standards Test Procedures for Antennas, IEEE Std. 149-1979 (Wiley, 1979)
27. Y. Mizuguchi, M. Akagawa, H. Yokoi, Offset gregorian antenna. Electron. Commun. Jpn.
61-B(3), 58–66 (1978)
28. C. Dragone, Offset antennas with perfect pattern symmetry and polarization discrimination.
Bell Syst. Tech. J. 57(7), 2663–2684 (1978)
29. L.H. Hemming, Electromagnetic Anechoic Chambers: A Fundamental Design and
Specification Guide (IEEE Press and John Wiley Interscience, 2002)
30. Microwave Vision Group, http://www.microwavevision.com/
31. A. Lönnqvist, A. Tamminen, J. Mallat, A.V. Räisänen, Monostatic reflectivity measurement
of radar absorbing materials at 310 GHz. IEEE Trans. Microw. Theory Tech. 54(9), 3486–
3491 (2006)
32. V.J. Vokurka, New compact range with cylindrical reflectors and high efficiency factor, in
Proceedings of Electronica 76 Conference, Munich, Germany (1976)
33. European Space Agency, http://www.esa.int/Our_Activities/Space_Engineering_Technology/
Antenna_Test_Facilities_and_Electro-Magnetic_Compatibility_Laboratories/
34. A.D. Olver, A.A. Saleeb, Lens-type compact antenna range. Electron. Lett. 15(14), 409–410
(1979)
35. A.A. Saleeb, Theory and design of lens-type compact antenna ranges, p. 279, Doctoral Thesis,
Queen Mary & Westfield College, University of London, 1982
36. A.D. Olver, Compact antenna test ranges, in Proceedings of the Seventh International
Conference on Antennas and Propagation (ICAP), York, UK, (1991), pp. 99–108
37. J. Tuovinen, A. Vasara, A. Räisänen, A new type of compact antenna test range, in
Proceedings of the 22nd European Microwave Conference, Espoo, Finland, vol. 1, (1992),
pp. 503–508
12 Antenna Measurements at Millimeter and Submillimeter Wavelengths 449

38. A. Vasara, J. Turunen, A. Friberg, Realization of nondiffracting beams with

computer-generated holograms. J. Opt. Soc. Am. A 6, 1748–1754 (1989)
39. J. Meltaus, J. Salo, E. Noponen, M.M. Salomaa, V. Viikari, A. Lönnqvist, T. Koskinen,
J. Säily, J. Häkli, J. Ala-Laurinaho, J. Mallat, A.V. Räisänen, Millimeter-wave beam shaping
using holograms. IEEE Trans. Microw. Theory Tech. 51(4), 1274–1280 (2003)
40. A. Tamminen, A. Karttunen, M. Vaaja, E. Noponen, J. Ala-Laurinaho, J. Mallat, A.V.
Räisänen, Reflection-type phase hologram for beam shaping: Experimental results at 310
GHz, in Proceedings of the 30th ESA Antenna Workshop, Noordwijk, The Netherlands, 27–
30 May 2008, pp. 470–473
41. J. Häkli, T. Koskinen, A. Lönnqvist, J. Säily, V. Viikari, J. Mallat, J. Ala-Laurinaho,
J. Tuovinen, A.V. Räisänen, Testing of a 1.5 m reflector antenna at 322 GHz in a CATR
based on a hologram. IEEE Trans. Antennas Propag. 53(10), 3142–3150 (2005)
42. T. Hirvonen, J. Ala-Laurinaho, J. Tuovinen, A.V. Räisänen, A compact antenna test range
based on a hologram. IEEE Trans. Antennas Propag. 45(8), 1270–1276 (1997)
43. Z. Li, J. Ala-Laurinaho, Z. Du, A.V. Räisänen, Realization of wideband hologram
Compact Antenna Test Range by linearly adjusting the feed location. IEEE Trans.
Antennas Propag. 62(11), 5628–5633 (2014)
44. V. Viikari, J. Häkli, J. Ala-Laurinaho, J. Mallat, A.V. Räisänen, A feed scanning based APC
technique for compact antenna test ranges. IEEE Trans. Antennas Propag. 53(10), 3160–3165
(2005)
45. A.M. Predoehl, W.L. Stutzman, Implementation and results of a time-domain gating system for
a far-field range, in Proceedings of the 19th Annual Meeting & Symposium of the Antenna
Measurement Techniques Association (AMTA), Boston, MA, USA, 17–21 Nov 1997, pp. 8–12
46. B. Fourestie, Z. Altman, J. Wiart, A. Azoulay, On the use of the matrix-pencil method to
correlate measurements at different test sites. IEEE Trans. Antennas Propag. 47(10), 1569–
1573 (1999)
47. B. Fourestie, Z. Altman, Gabor schemes for analyzing antenna measurements. IEEE Trans.
Antennas Propag. 49(9), 1245–1253 (2001)
48. V. Viikari, J. Mallat, J. Ala-Laurinaho, J. Häkli, A.V. Räisänen, A frequency shift technique
for pattern correction in hologram-based CATRs. IEEE Trans. Antennas Propag. 54(10),
2963–2968 (2006)
49. J. Appel-Hansen, Reflectivity level of radio anechoic chambers. IEEE Trans. Antennas
Propag. AP-21(4), 490–498 (1973)
50. M.D. Migliore, Filtering environmental reflections in far-field antenna measurement in
semi-anechoic chambers by an adaptive pattern strategy. IEEE Trans. Antennas Propag. 52
(4), 1112–1115 (2004)
51. V. Viikari, V.-M. Kolmonen, J. Salo, A.V. Räisänen, Antenna pattern correction technique
based on an adaptive array algorithm. IEEE Trans. Antennas Propag. 55(8), 2194–2199
(2007)
52. V. Viikari, A.V. Räisänen, Antenna pattern correction technique based on signal to
interference ratio optimization. IEEE Antennas Wirel. Propag. Lett. 6, 267–270 (2007)
53. P.L. Garcia Muller, J.L. Cano, R. Torres, A deconvolution method for correcting antenna
measurement errors in compact antenna test ranges, in Proceedings of the 17th Annual
Antenna Measurement Techniques Association (AMTA) Meeting and Symposium, USA, 1995,
pp. 509–514
54. G. Hindman, A.C. Newell, Mathematical absorber reflection suppression (mars) for anechoic
chamber evaluation and improvement, in Proceeings of the 30th Annual Antenna
Measurement Techniques Association (AMTA) Meeting and Symposium, USA, 2008
55. I.J. Gupta, Stray signal source location in far-field antenna/RCS ranges. IEEE Antennas
Propag. Mag. 46(3), 20–29 (2009)
56. T.P. Delfeld, F.C. Delfeld, Use of the MUSIC algorithm in the analysis of compact range field
probe data, in Proceedings of the 11th Annual Antenna Measurement Techniques Association
(AMTA) Meeting and Symposium, Monterey, CA, USA, 9–13 Oct 1989, pp. 6-3–6-9
450 A.V. Räisänen et al.

57. I.J. Gupta, W.D. Burnside, Imaging the compact range probe data, in Proceedings of the 12th
Annual Antenna Measurement Techniques Association (AMTA) Meeting and Symposium,
Philadelphia, PA, USA, 1990, pp. 14-13–14-17
58. J.C. Bennett, A.P. Anderson, P.A. Mclnnes, A.J.T. Whittaker, Microwave holographic
metrology of large reflector antenna. IEEE Trans. Antennas Propag. AP-24(3), 295–303
(1976)
59. P.F. Goldsmith, N.R. Erickson, Holographic measurements, in Kraus, J.D (ed.) Chapter 6
“Radio-telescope antennas” in the 2nd edition of Radio Astronomy (Cygnus-Quasar Books,
1986), pp. 6-63–6-64
60. J.M. Payne, Millimeter and submillimeter wavelength radio astronomy. Proc. IEEE 77(7),
993–1016 (1989)
61. J.W.M. Baars, R. Lucas, J.G. Mangum, J.A. Lopez-Perez, Near-field radio holography of
large reflector antennas. IEEE Antennas Propag. Mag. 49(5), 1–14 (2007)
62. P.J. Napier, R.H.T. Bates, Antenna aperture distributions from holographic-type radiation
pattern measurement. Proc. IEEE. 120(1), 30–34 (1973)
63. G. Junkin, Phase shifting holography for THz near-field/far-field prediction. Prog.
Electromagnet. Res. Lett. 44, 15–21 (2014)
64. A. Arboleya, J. Laviada, J. Ala-Laurinaho, Y. Álvarez, F. Las-Heras, A.V. Räisänen,
Phaseless characterization of broadband antennas. IEEE Trans. Antennas Propag. 64(2), 484–
495 (2016)
65. W.-H. Lee, Computer-generated holograms: techniques and applications, in E. Wolf (ed.)
Progress in Optics XVI (Elsevier, Amsterdam, 1978) pp. 121–231
66. L. Roy, M. Li, S. Labonte, N.R.S. Simons, Measurement techniques for integrated-circuit slot
antennas. IEEE Trans. Instrum. Meas. 46(4), 1000–1004 (1997)
67. R.N. Simmons, Novel on-wafer radiation pattern measurement technique for MEMS
actuator based reconfigurable patch antennas, in Proceedings of the 24th Annual
Antenna Measurement Techniques Association Meeting and Symposium, Cleveland,
Ohio, 3–8 Nov 2002
68. T. Zwick, C. Baks, U.R. Pfeiffer, D. Liu, B.P. Gaucher, Probe based MMW antenna
measurement setup, in Proceedings of the IEEE AP-S Symposium 2004, Monterey, CA, USA,
vol. 1, 20–25 June 2004, pp. 747–750
69. S. Ranvier, M. Kyrö, C. Icheln, C. Luxey, R. Staraj, P. Vainikainen, Compact 3-D on-wafer
radiation pattern measurement system for 60 GHz antennas. Microw. Opt. Technol. Lett. 51
(2), 319–324 (2009)
70. K. Mohammadpour-Aghdam, S. Brebel, A. Enayati, R. Faraji-Dana, G.A.E. Vandenbosch,
W. Deraedt, RF probe influence study in millimeter-wave antenna pattern measurements. Int.
J. RF Microw. Comput. Aided Eng. 21(4), 413–420 (2011)
71. B. Klein, M. Jenning, P. Seiler, K. Wolf, D. Plettemeier, Verification and demonstration up to
67 GHz of an on-chip antenna pattern measurement setup, in Proceedings of IEEE
Conference on Antenna Measurements and Applications (CAMA), Antibes Juan-les-Pins,
France, 16–19 Nov 2014
72. T. Calazans, H.D. Griffiths, A. L. Cullen, D.E.N. Davies, R. Benjamin, Antenna radiation
pattern measurement using a near-field wire scattering technique. IEEE Proc. Microw.
Antennas Propag. 145(3), 263–267 (1998)
73. Z. Du, V. Viikari, J. Ala-Laurinaho, A. Tamminen, A.V. Räisänen, Antenna pattern retrieval
from reflection coefficient measurements with reflective loads. Prog. Electromagn. Res. 148,
15–22 (2014)
74. V. Viikari, Z. Du, V. Semkin, J. Ala-Laurinaho, A.V. Räisänen, Reflection coefficient method
for characterizing antennas on probe stations, in Proceedings of the 9th European Conference
on Antennas and Propagation, Lisbon, Portugal, 12–17 Apr 2015
Chapter 13
Holographic Imaging Approach

Christian Schildbach and Lorenz-Peter Schmidt

Abstract This chapter provides an overview over common categories of micro-

wave holography systems. Terms like resolution or speckle are defined and design
rules are presented. Performance and cost of the various microwave holography
systems are discussed as well as their applications. A special focus is devoted to
indirect holographic systems at millimeter wave frequencies. A setup using arrays
of planar antenna coupled zero bias Schottky diodes as detectors is presented in
detail.

13.1 Introduction

Before the performance of microwave holographic systems and their specific

applications are discussed, at first the term holography is clarified. The word
holography stems from the Greek words “holos” meaning the “whole” and
“gráphein” which means to write or to draw. The expression signifies that by
holography, one can record the complete wave front of scalar waves, such that these
wave fronts can be reconstructed later. The process is usually called a two-step
process with the two steps: first the recording of the hologram, and second the
reconstruction of the wave front [1]. Therefore, the recorded hologram must contain
both magnitude and phase information completely describing the wave front. This
already implies the existence of a distinct frequency, the wave oscillates at. In short,

C. Schildbach (&)
Am Schillerplatz 4/1, 71522 Backnang, Germany
e-mail: [email protected]
L.-P. Schmidt
Schleifweg 33, 91093 Heßdorf, Germany
e-mail: [email protected]
C. Schildbach
L.-P. Schmidt
Lehrstuhl für Hochfrequenztechnik/Institute of Microwaves and Photonics,
Friedrich-Alexander-Universität Erlangen Nürnberg/Friedrich-Alexander-University
Erlangen Nuremberg, Cauerstr 9, 91058 Erlangen, Germany

© Springer International Publishing AG 2018 451

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1_13
452 C. Schildbach and L.-P. Schmidt

coherence is required. Holography can be applied to many different kinds of waves,

including sound waves, particle waves, and with scalar approximation also to
electromagnetic waves and others. In the following chapter, the focus lies on one
specific part of the electromagnetic spectrum, namely the microwave holography. In
the beginning, holography was not intended for use in the microwave regime, but
its inventor Dennis Gabor searched for a way to improve the resolution of electron
microscopes [2]. The challenge in the recording step is that usually intensities are
recorded rather than complex phasors. Gabor realized in 1948 that to overcome this
difficulty a phase reference is needed for comparison, which he called “coherent
background”. The superposition of both the wave front and the coherent back-
ground is called a hologram. Although he never used holography for its original
purpose, Gabor was able to demonstrate the functionality in the optical domain.
Because in those days, computers were not powerful enough to process or even
store the amount of data contained in holograms, the reconstruction had to be done
by a coherent optical setup. Another problem in both steps was the lack of powerful
coherent sources. Gabor used a high pressure mercury lamp yielding a coherence
length of about 0.1 mm, which was sufficient for about 200 fringes.
Leith and Upatnieks were the first to use holography for microwave imaging
[3, 4]. They found out that the reconstruction of side-looking airborne radar images
can be viewed as the reconstruction of a hologram. Although the data was recorded
in a different range of wavelengths, they still stored it on a photographic film and
processed it with a coherent optical setup [5]. So not only optical, but also
microwave holography benefited from the invention of the laser with its superior
power and coherence properties compared to any other source available. A lot of
improvement has since been made to microwave holography with the most
important one being the computer-based reconstruction. Performance of nowadays
computers makes storage and reconstruction even available in real time.
Today, microwave holography has found a widespread field of applications. In
the field of remote sensing, the special frequency range of microwaves offers some
advantages. In the microwave region, wavelengths have dimensions reaching from
about 1 mm to several centimeters. Structures in these dimensions can be produced
with high precision. On the other hand, the lower limit of the resolution of imaging
systems is more or less given by the wavelength. So the microwave region is a good
compromise between resolution and complex technological processes. Setups in the
microwave region often allow to distinguish the two orthogonal components of
polarization, giving access to the complete polarimetric information about a target.
Material properties differ from the easy to observe optical range, sometimes
yielding important additional information. Microwave radiation is nonionizing and
thus not hazardous. This allows for use in diagnostic medicine and security
applications by visualizing the interior of nonconducting motives. Today’s
Synthetic Aperture Radar systems use holography for geophysical exploration.
Other areas of application are nondestructive testing, production quality control or
automotive radar, just to name a few. New microwave holography applications still
emerge.
13 Holographic Imaging Approach 453

13.2 Resolution

13.2.1 Definition

One of the most important figures of merit of an imaging system is its resolution. In
principal, it is defined as the minimum distance between two small targets such that
those two targets can still be recognized as distinct. Since this definition is rather
vague, several criteria have emerged that give an estimate for the resolving power
of a system. They all use the point spread function (PSF), the image of a point like
object. Because imaging systems are approximately linear space invariant systems,
the PSF is comparable to an impulse response of the system. For linearity, two
arbitrary targets imaged together have to produce an image whose brightness is the
sum of the two images of the single targets. This is usually true for imaging
systems. For space invariance, the PSF may not depend on the location of the target.
This is only approximately true for most imaging systems that cover only a limited
solid angle of the scenery.
The produced image is then the convolution of the real scenery with the PSF.
There are various shapes of PSFs that have been investigated. The earliest one is the
sinc2 function, which was discussed by Rayleigh [6]. He claimed that two sinc2
functions of equal width and magnitude can be resolved if the maximum of the first
sinc2 is situated at the first null of the second one as depicted in Fig. 13.1.
Differently shaped square integrable PSFs like, e.g., Gaussian, Lorentzian or Voigt
profiles may not have any null. Jones et al. compare the most common methods to
calculate the width of a function [7]. They recommend the use of the equivalent
width, the ratio of the area under the profile function, and its peak height.

Fig. 13.1 Two sinc2 1

functions displaced according
to the Rayleigh criterion and
their sum. The valley in the 0.8
middle indicates that both
peaks can clearly be resolved
Magnitude →

0.6

0.4

0.2

0
−4 −2 0 2 4 6
X→
454 C. Schildbach and L.-P. Schmidt

Another method to derive the resolution of any given PSF is to examine the
equivalent width of its spectrum. In the case of the sinc2 function, the spectrum is a
triangular function with an equivalent width of

2p
Dk ¼ ; ð13:1Þ
d

where d is the distance between the sinc’s maximum and its first null. The other
way round, the resolution can be calculated from the spectral width.

2p
d¼ ð13:2Þ
Dk

13.2.2 Lateral Resolution Limited by Diffraction

In order to determine the spectral width of the PSF, a closer look at the imaging
system has to be taken. Therefore, the propagation of the electromagnetic waves
from the point source to the aperture is considered as well as the reconstruction of
the image. The propagation of nonplane waves in free space is covered by
diffraction theory, first discussed by Sommerfeld [8, 9]. Such waves originate from
a radiator or a scattering object. There are many different theories on diffraction,
getting simpler the more approximations are made. In the following, the approxi-
mations that will be used later are summarized. First, electromagnetic waves in the
microwave region are treated as scalar waves. Silver [10] and Goodman [3] show
that this approximation can give very good results if the fields are measured suf-
ficiently far away from the diffracting object and if the diffracting object details are
large compared to the wavelength. The corresponding wave equation then is:

r2 E þ k02 E ¼ 0 ð13:3Þ

with k0 representing the wavenumber in free space and E is the scalar complex field
strength. The second approximation used in the following is called the Fresnel
Approximation. Here, a spherical wave front of radius r is approximated by a
second order Taylor polynomial. The resulting paraboloid wave front describes the
wave over a limited region around the expansion point contained in a rectangle of
size Dx
Dy. For this approximation to be valid, the distance z from the center of
the spherical wave front must be large compared to the distance from the axis of the
paraboloid, as indicated in (13.4).

Dx2 þ Dy2
z ð13:4Þ
2k
13 Holographic Imaging Approach 455

As (13.3) is a linear homogeneous differential equation, all possible solutions

can be constructed as superposition of a set of fundamental solutions. In the fol-
lowing, the two most important solutions are discussed.
First, there is the spherical wave originating from a point source, given by (13.5)

1
E ¼ ejk0 r ; ð13:5Þ
r

where r is the distance from the point source. Its significance for the resolution has
already been discussed. In fact, every source distribution can be interpreted as a
density of point sources, and therefore spherical waves with different source points
are a set of fundamental solutions to the wave equation. Due to its symmetry, the
spherical wave homogeneously propagates in each direction of the far-field.
Second, the plane wave solution to the differential equation only propagates in a

T
single direction. This direction is given by the vector ~ k ¼ kx ; ky ; kz . Then the
field distribution
~T
~
Ej ¼ ejk r
ð13:6Þ

 
 
r ¼ ðx; y; zÞ solves the wave equation if ~
with the coordinate vector ~ k  ¼ k0 .
Evaluating the field along any straight line whose direction is given by the unit
vector ~eu always yields a linearly changing phase. The corresponding wavenumber
is ku .

ku ¼ ~
k T
~
eu ð13:7Þ

This implies that kx ; ky and kz are the wavenumbers observed, when evaluating
the field along lines parallel to the coordinate axes. As the norm of the vector ~
k is
fixed, two components are sufficient to determine the third one.
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r 
kz ¼ k02  kx2  ky2 ð13:8Þ

Plane waves with different directions are also a set of fundamental solutions to
the wave equation. Hence, each possible field distribution can be described by a
superposition of plane waves. Their use offers some advantages compared to
spherical waves. For example, the calculation of the field value in some given
points of interest in a Cartesian coordinate system is easier. In contrast to spherical
waves, the magnitude of plane waves does not change, so only the phase has to be
adjusted. Also this change in phase depends linearly on the change in coordinates,
whereas the phase of a spherical wave changes with its distance from the source
point. In particular, when considering a multitude of source points and points of
field evaluation, the distance for each combination of points has to be calculated
456 C. Schildbach and L.-P. Schmidt

separately. For plane waves on the other hand, the calculation becomes extraor-
dinarily simple if the field is evaluated in points, arranged in a cubic array. For each
plane wave, only the phase in one point of interest and the change to the neigh-
boring point in each direction have to be known. Going n steps further in an
arbitrary direction, the phase has to be turned n times the amount for one step.
As stated above, each field distribution can be described as a superposition of
fundamental solutions. When decomposing a field in plane waves, the result is
called the angular spectrum of plane waves or k-space representation. If the field is
known in each point of a plane, the decomposition becomes a two-dimensional
Fourier transform. For instance, the angular spectrum of a spherical wave is uni-
formly distributed in all directions.
Z Z

 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r  
E ¼ exp j kx x þ ky y þ k0  kx  ky z dkx dky
2 2 2 ð13:9Þ

Reference [11] proves (13.9) to be accurate by using the method of stationary

phase.
Using this and reconstruction methods, the spectral width of the PSF for an
example setup as shown in Fig. 13.2 can be estimated. A rectangular aperture of
width Lx and height Ly receives the signal from a point source. The source point is
situated at a distance z0 in front of the aperture. Reconstruction algorithms give an
estimate E^ of the field E in front of the aperture based on the measured fields. The

Fig. 13.2 Imaging setup

13 Holographic Imaging Approach 457

simplest way to reproduce the fields is called backpropagation or Back Projection

(BP). The measured field in the aperture is decomposed into plane waves by
Fourier analysis. The propagation of plane waves is easy to invert by just turning
back the phase according to the direction and distance as indicated in (13.10).

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
b ¼ FT2D
1
FT2D fE ðx; y; z ¼ 0Þg
ej ðk0 kx ky Þz
2 2 2
E ð13:10Þ

Therefore, the spectral width of the PSF is identical to the spectral response of
the measured field in the aperture to a point source. The same reconstruction can
also be regarded in the space domain. Here, E ^ at each position x; y; z is calculated
as the correlation of the measured field in the aperture with the field of a spherical
wave centered at x; y; z. There exist many different reconstruction algorithms with
different implementations, but the basic idea always is the above described
BP. A detailed analysis of the different algorithms cannot be given here.
Often, the received spectrum of plane waves is approximated by geometrical
optics. The maximum angle of incidence is determined by a straight line going from
the source point to the edge of the aperture. Actually, the angle between the straight
line and the normal vector of the aperture gives the local spatial frequency (the
analog of the instantaneous frequency) at the corresponding point in the aperture.
As Mandel points out, the spectrum of the Fourier decomposition does not have a
one-to-one correspondence to local spatial frequencies [12]. He also illustrates that
the Fourier spectrum can be calculated by analyzing the instantaneous frequency of
the function divided into small intervals, if the amplitude is smooth enough. The
spectrum can then be approximated by the complex superposition of the contri-
butions of the small intervals. To show this, the method of stationary phase can be
applied. To confirm the applicability, the spectral content of the received signal in
the aperture is analyzed by Fourier decomposition in the following. Using the
Fresnel Approximation, the field at an arbitrary point in the aperture can be
calculated:
2 2
 
1 jk0 x 2zþ y x y
E ðx; yÞ  x2 þ y2
ejk0 z0 e 0 rect rect ð13:11Þ
z0 þ 2z0
Lx Ly

Also, the amplitude term only varies slowly and thus its influence can be
neglected compared to the phase term.
   
2 2
jk0 z0 þ x 2zþ y x y
E ðx; yÞ  e 0
rect rect ð13:12Þ
Lx Ly
458 C. Schildbach and L.-P. Schmidt

Ly
 


Lx 2 2
Z2 Z2 jk0 z0 þ x 2zþ y
) U kx ; ky ¼ e 0
ejkx x ejky y dxdy
L2x L
 2y
 
ðk 2 þ k 2 Þ
pz0 j k0 þ x4k0 y z0
¼ j e
2k0
 rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi  rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi
jk0 Lx jz0 jk0 Lx jz0

erf þ kx  erf  þ kx
2z0 2 2k0 2z0 2 2k0
 rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi  rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi
jk0 Ly jz0 jk0 Ly jz0

erf þ ky  erf  þ ky ;
2z0 2 2k0 2z0 2 2k0
ð13:13Þ

where erf denotes the Gauss error function. Apart from the scaling factor and phase,
there are two distinct terms for the x- and y-dependence, respectively. The error
function changes its value rapidly in a region where the argument is close to zero
and is almost constant elsewhere. This is also true for arguments with a phase of
p=4. So if k0 L2i  z0 for i ¼ x; y, the error function can be approximated by the sign
function.
   
k0 þ
ðkx2 þ ky2 Þ
pz0 j z0 z 0 kx z0 ky
Uðx; yÞ  ð13:14Þ
4k0
e rect rect
k0 Lx k0 Ly k0

For illustration, Fig. 13.3 shows the magnitude of the x-dependence of the
k-space support in case that z0 ¼ 0; 175 m, Lx ¼ 0; 1 m and the frequency is
f ¼ 300 GHz. Under these conditions, the estimation using the local spatial

Fig. 13.3 kx -dependence of

k-space support of an aperture
with z0 ¼ 0; 175 m, Lx ¼
0; 1 m and f ¼ 300 GHz
1
Arbitrary Unit →

0.5

0
−6,000 −4,000 −2,000 0 2,000 4,000 6,000
1
kx in m →
13 Holographic Imaging Approach 459

Fig. 13.4 Maximum local spacial frequency introduced to the aperture by a point source at a
distance of z0 in front of a rectangular aperture

frequencies predicts the width of the k-space support to be between 1800 1=m,
agreeing very well with the results shown in Fig. 13.3.
In the above, it has been shown that the resolution in direction i ¼ x; y due to the
PSF of a receiving square aperture is approximately given by (13.15), with c0 being
the speed of light.
c0 z 0
di  ð13:15Þ
fLi

Accepting the approach using local spatial frequencies, the maximum local
frequency in direction i can be calculated by (13.16) as illustrated in Fig. 13.4.

k0
ki;max  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ð13:16Þ
2
4 z0
Li þ1

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
c0 z0
) di  4 þ1 ð13:17Þ
2f Li

The total image spectrum may be much wider, if a multitude of sources in

different directions is considered. For an object of size Lx;Obj
Ly;Obj for instance,
(13.18) holds.

k0
ki;Obj;max  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi ð13:18Þ
2
4 z0
Li þ Li;Obj þ1
460 C. Schildbach and L.-P. Schmidt

In some cases, it is also possible to use a second aperture for illumination. Such
systems are called multistatic. Due to reciprocity, the same rules apply to the
transmitting aperture as to the receiving one. They also create a PSF independently.
The total PSF of the system is then the product of the PSFs of both the transmitting
and the receiving aperture. Of course, the transmitting and receiving aperture may
be close to each other or even overlapping.

13.2.3 Range Resolution

Also the resolution in range direction is determined by (13.2). The bandwidth has
got the strongest impact on Dkz . But even with a single frequency, the range
resolution is finite due to the different angles of incidence on the aperture. So as
comes out of Fig. 13.4, the k-space support and the according resolution are given
by (13.19), (13.20) and (13.21).

2pfmax
kz;max ¼ ð13:19Þ
c0

2pfmin 1
kz;min  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2ffi ð13:20Þ
c0
1 þ 4 Lx þ 4 Lz0y
z0

c0
) dz  2 0 13 ð13:21Þ
6 B C7
26 B
4Df þ fmin @1  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
  2 A5
C7
1 þ 4ðLx Þ þ 4 Ly
z0 2 z0

Moreover, the resolution in range direction is improved in the multistatic case.

13.3 Array Design

In the preceding section, a simple aperture has been assumed. In microwave

holography, the field in the entire aperture has to be measured in magnitude and
phase. This is usually accomplished by an array of small antenna elements, filling
the aperture. That means that the field is not measured homogeneously everywhere
in the aperture but it is rather sampled in discrete points of space. Just like the
spatial sampling, the characteristic of each single element influences the perfor-
mance of the imaging system. In the following, both effects will be examined.
As the single antenna elements are usually about the size of a wavelength or
smaller, they are small compared to the distance to the object. So the object is in the
13 Holographic Imaging Approach 461

far-field of the single element. Thus, the directivity of the single element to a certain
direction yields an extra factor when receiving plane waves from this direction.
Therefore, it is desirable to have an antenna element that covers the complete
scenery as homogeneously as possible. On the other hand, elements in close
proximity to each other can couple considerably, particularly if they are omnidi-
rectional radiators. This must either be prevented by careful antenna design of the
single element or calibrated and corrected later. Both topics will be covered later in
this section.
Mathematically, equidistant sampling can be represented by multiplication with
the two-dimensional Ш-function, an equidistant rectangular grid of Dirac d-impulses.
In the k-space, this corresponds to a convolution with the corresponding
two-dimensional Ш-function. Hence the obtained spectrum is repeated. The relation
between the sample distance d and the repetition in the spectrum kr is given by (13.22).

2p
kr ¼ ð13:22Þ
d

For an alias free sampling, k2r has to be larger than any spatial frequency content
in the aperture. Obviously, the maximum spatial frequency that can be obtained is
k0 , so sampling with a distance of

k
dmin ¼ ð13:23Þ
2

is always small enough. But of course even larger sampling distances suffice if large
angles of incidence umax can be excluded. Then the required sampling distance is:

k
dmax ¼ ð13:24Þ
2 sinðumax Þ

But the total number of antenna elements can be reduced even further. The
reconstruction of a field measured with an antenna array is often compared to
Digital Beam Forming (DBF). In fact, the only difference is that for DBF, the
Fraunhofer far-field of the array is optimized, while in holography the focus is still
in the Fresnel zone. Just like in the above, the performance of DBF depends on the
k-space support of the array. Therefore, the far-field beam width of the array is
also a measure for the resolution of the imaging system, although they are not the
same thing. For beamforming applications, sparse arrays have been widely studied
[13–19]. The array factor AF of a general plane array in the aperture of Fig. 13.2 is
given by (13.25).

X
N
AF ¼ An ejk0 sinð#Þðxn cosðuÞ þ yn sinðuÞÞ ð13:25Þ
n¼1
462 C. Schildbach and L.-P. Schmidt

Here, xn and yn are the coordinates of the n th element, which is weighted with
the complex amplitude An . These complex An can be used either to produce or to
receive beams from various directions. There are many text books on antennas like
[20] that provide formulas for the case of antenna elements on a regular lattice.
Obviously, (13.25) has the form of a Fourier series of the amplitude distribution An
with the fundamental spatial frequencies

kx ¼ k0 sinð#Þ cosðuÞ ð13:26Þ

ky ¼ k0 sinð#Þ sinðuÞ: ð13:27Þ

This shows the close relationship between beamforming and imaging. With a
limited amount of elements and limited space of the aperture, the choice of the
weights An is always a trade-off between a small beam width and a low grating lobe
level [21]. Similarly, in imaging the trade-off occurs between resolution and side
lobe level of the PSF. For sparse arrays, one simply has to set the weights of the left
out antennas to zero to obtain the array factor of the remaining array. This of course
worsens at either resolution or side lobe level. Those side lobes are one cause of
clutter. Clutter is a disturbance in the image that appears random to the viewer and
may have various causes. Further causes of clutter are noise, multiple reflections
that cannot be reconstructed, approximations made for the reconstruction algorithm,
imperfect calibration or coupling between the antenna elements. In [22], Gupta
et al. show how the performance of an array is impaired by inter-element coupling.
Decrease in side lobe level can be achieved by multiplication with taper functions
as described, for example, in [23] or following specific thinning strategies [13, 16–
18]. If clutter is dominated by noise, it might be beneficial to optimize the weights
for maximum signal-to-noise ratio SNR by using the SNR of a single receiver as
weight of the respective signal. Assuming equal noise level for each receiver, the
SNR is proportional to the signal power itself. A drawback can be seen in the fact
that the resulting resolution depends on the imaged scenery. If the clutter is dom-
inated by other processes, an increased resolution may be attainable at the cost of
only very little increase in clutter. According methods have been proposed very
early. References [24] and [25], for example, increase the effective aperture area by
extrapolating the measured signals to the outside of the real aperture. The extrap-
olated values are calculated depending on the measurement results within the
aperture. Therefore, such methods can also be expressed as weighting for the
individual receivers within the aperture. The other way round, setting the An of
missing antenna elements to zero can be interpreted as interpolating the missing
signals by zero. Due to Parseval’s theorem, this will produce the spectrum con-
taining the least amount of energy still explaining the measured data. It is on the
other hand possible to minimize just the part of the spectrum according to the
directions from where no radiation is expected. Solving this problem is covered by
13 Holographic Imaging Approach 463

Fig. 13.5 Principle of the

synthetic aperture radar
(SAR): linear transmit and
receive arrays are swept
Receive
across an aperture Array

Along
Track

Transmit
Array

compressed sensing [26]. Sparse arrays are especially useful in the case of inde-
pendent transmit and receive arrays [27, 28]. As it has been shown earlier, the
far-field characteristics of both arrays multiply. By placing the zeros of the receive
far-field characteristic at the angles of the grating lobes of the transmit array and
vice versa, the total side lobe level of the system can be reduced dramatically. The
combination of both arrays can also be viewed as an effective aperture distribution,
calculated as the convolution of the field distributions of both arrays. A special form
of multistatic imaging using sparse arrays is Synthetic Aperture Radar (SAR).
Linear transmit and receive arrays are swept along a linear trajectory, spanning an
area called synthetic aperture, as shown in Fig. 13.5 [29–31]. The effective aperture
of this setup is the convolution of the transmit and receive arrays and the positions
in along track direction. The length of the effective aperture is bounded by the
beamwidth of the single elements. For comparison, supposing a nonsparse/fully
occupied array, the effective aperture would be the transmit array convoluted with
the receive array. Each one would be the respective linear array convoluted with the
along track positions. So the effective aperture would contain the along track
positions twice. This is not the case, because the measured radiometric information
does not contain all combinations of the transmit and receive positions [32].
464 C. Schildbach and L.-P. Schmidt

13.4 Categories

Holographic imaging setups can be split up into various categories. This section
gives an overview over the most important categories of holography and discusses
their properties. A structure of the described categories is depicted in Fig. 13.6. The
imaging process relies on the knowledge of the fields in the aperture in magnitude
and phase. One of the most important differentiation criteria between systems is the
way how they gain the phase information. In order to detect the phase of a signal,
the time information can be provided by a reference signal. So the detected quantity
is the relative phase between the measured field and the reference signal. Such a
phase difference is a measurable quantity. Direct and indirect holography use dif-
ferent ways to provide the reference signal to the detectors. The characteristics of
direct and indirect holography will be the focus of this section, followed by further
useful categories.

13.4.1 Direct Holography

In direct holography, the illumination and the reference signal are generated
independently and in general at different frequencies. In the following, the fre-
quency used for imaging will be called fRF and the frequency of the reference signal
will be called fLO . Both the transmitted and received signals are measured in
magnitude and phase using the same reference [33]. For the measurement of the
transmitted illumination, usually a small fraction is coupled out from the generator,
such that the remaining power can be used as illumination. The RF signals are
measured by mixing them with the reference signal in a heterodyne mixer. The
resulting signal will be at the intermediate frequency fIF .

fIF ¼ jfRF  fLO j ð13:28Þ

In a heterodyne mixer, the RF and LO currents both flow through a component

with a nonlinear characteristic. Thereby the currents create voltages at frequencies

Fig. 13.6 Overview over described categories of holography

13 Holographic Imaging Approach 465

that are all linear combinations of fRF and fLO . A filter suppresses all frequency
components except fIF . The IF signal is then converted to digital information using
an I/Q demodulator. This way both inphase and quadrature components of the
signals are known and consequently magnitude and phase. The advantage of
directly measuring the phase difference between transmit and receive elements
comes at the cost of high system complexity. Because all these components are
already combined in a vector network analyzer (VNA), these instruments are
commonly used for imaging applications. Unfortunately, VNAs only come with a
few ports and hence antennas are often swept over the aperture.
If an antenna array is used, the signal from each antenna has to be measured.
Because the reference signal will experience different electrical lengths and atten-
uations when traveling to the respective antenna elements, calibration is obligatory.
As stated earlier, calibration can also compensate for mutual coupling (between two
transmit or receive antenna elements, respectively) and cross-coupling (from a
transmit antenna element directly to a receive antenna element) [22]. When cali-
brating a system, objects with known properties, so-called standards, are measured.
From the differences to the obtained measurement values, the systematic mea-
surement errors can be modeled and later corrected. As the receivers translate field
strength linearly to measured data, a linear error model is sufficient to compensate
for magnitude and phase errors. The transmission between the antennas can be
modeled using a signal flow graph. In Fig. 13.7, all transmission paths that involve
at most one coupling term from transmit antenna j to receive antenna i are depicted.
All paths including two or more coupling terms and loops with coupling edges are
neglected, because coupling constants are usually very small. Paths connecting the
transmit antenna i to the receive antenna j directly via the object are called Cij .
Cross-coupling between the same antennas is called Wij . Mutual coupling between
two transmit antennas j and k is described by Tkj and between the two receive
antennas i and l is described by Kil . As a consequence of reciprocity, the mutual
coupling coefficients show the following symmetry:

Transmit
Antennas

Receive
Antennas
Object

Fig. 13.7 Error terms for calibration: cross-coupling and mutual coupling are shown
466 C. Schildbach and L.-P. Schmidt

Tkj ¼ Tjk ð13:29Þ

Kil ¼ Kli ð13:30Þ

To make it short a vector notation is adapted. T j is the column vector of mutual

coupling coefficients between the transmit antenna j and all other transmit antennas,
and C j is the column vector containing the imaging transmission paths from
transmit antenna j to all receive antennas. Similarly, K i and Ci are the row vector of
mutual coupling coefficients between the receive antenna i and all other receive
antennas and the row vector containing the imaging transmission paths from each
transmit antenna to the receive antenna j, respectively. Adding up all relevant paths
and considering possible differences of the reference signal at the respective mixers
due to different transmission line length Cij , the measured signal Mij is given by
(13.31).


Mij ¼ Cij Cij þ Wij þ Ci T j þ K i C j ð13:31Þ

One suitable standard is an absorbing half space. In this case, all Cij are zero. The
corresponding measurement results a Mij then only contain Cij and the
cross-coupling term Wij .
j
a Mi ¼ Cij Wij ð13:32Þ

With this result, (13.31) reduces to (13.33) and further to (13.34).


Mij  a Mij ¼ Cij Cij þ Ci T j þ K i C j ð13:33Þ
0 1

 Cij
) Mij a Mij ¼ Cij CjT @ Cij T A ð13:34Þ
j
Ci
j T
Ci K i

Noise occurring in this absorber measurement will disturb any further mea-
surement. It is therefore recommended to use a low IF bandwidth and accept longer
measurement time for this one measurement. Also the return loss of the absorbing
material is critical.
As the column vector in (13.34) still contains unknown variables, further cali-
bration measurements are needed. Further standards may be implemented like
reflecting sheets and spheres at different distances, because it is rather easy to
calculate the expected fields. Each measurement yields one line generating a system
of linear equations. Suppose that mutual coupling is considered between nT
transmit antennas and nR receive antennas, respectively, then T j 2 CnT 1 and
K i 2 CnR 1 . Thus, there are nT þ nR  1 unknown variables yielding a minimum
for the rank and hence for the number of linear equations. Clever spacing of the
13 Holographic Imaging Approach 467

standards can prevent exact linear dependence of two equations, but because of the
vast amount of combinations of transmit and receive antennas at different fre-
quencies, usually some will occur, where the linear system shows a bad condition
number. Each additional calibration measurement improves this condition number
and unknown coupling constants are computed using Gaussian least squares.
It follows from the above that the required effort for the calibration of a system
for direct holography grows quickly with the number of antenna elements.
Fortunately, the majority of the coupling constants are usually small enough to be
neglected. That is because coupling coefficients between two antenna elements
decay quickly when one antenna element leaves the reactive nearfield of the other.
As a rule of thumb, a coupling attenuation above 35 dB will cause a phase error
smaller than 1 . In some cases, even all mutual coupling coefficients can be ignored.
Then (13.34) becomes a simple single equation and the solution is given by (13.35).

Cij
Cij ¼ ð13:35Þ
Mij a Mij

Averaging this result for several standards again will increase accuracy.
The specific applications for direct holography are widespread. Personal security
scanners in airports represent a quickly growing market. At the moment, four
different commercial products are available, all using frequencies between 25 GHz
and 100 GHz. In contrast to X-ray based systems, the nonionizing nature of these
millimeter waves prevents damage of biological cells. In fact, millimeter waves
hardly penetrate human tissue, while textiles are mostly transparent in this domain.
This allows for detection of threats by concealed objects and weapons. As space-
borne SAR systems improve, they increasingly replace airborne systems. Still
airborne systems are more flexible for surveillance of a nearby region, because
satellite orbits cannot be changed at will, but the orbits of other objects have to be
taken into account. For global operation on the other hand, spaceborne systems are
advantageous. Nowadays spaceborne SAR systems in the X-band achieve resolv-
able distances below 1 m. While resolution performances of the two kinds of SAR
systems are about to level up, the swath width observed by satellites is much larger,
extending up to more than 100 km. Apart from military reconnaissance, many
different areas of science and industries profit from such systems. In hydrology, the
surface coverage with ice, snow and water can be surveilled. Continuously mapping
earth’s surface incorporating interferometric features yields a deeper insight to
tectonics or volcanism. From those abilities, also disaster and crisis management
benefit, as well as archaeology and the search for natural resources. Another geo-
logical application of direct holography is the ground penetrating radar, using
frequencies from 10 MHz to 1 GHz. Microwave mammography is still in devel-
opment. The systems use frequencies around 1 GHz. The required resolution can
only be achieved because of the high permittivity of the tissue, greatly increasing
the wavenumber.
468 C. Schildbach and L.-P. Schmidt

13.4.2 Indirect Holography

In contrast to direct holography, in indirect holography the phase reference is not

independent of the illumination, but one coherent source provides both. Although it
is less popular, indirect holography features some general advantages. First, the
phase of the illuminating source does not have to be measured, because its phase is
inherently identical to the reference. Second, the architecture of a single receiver
can be much simpler. A square law detector for power detection is sufficient. Third,
if mutual coupling between the receive antennas can be neglected, calibration can
be omitted. That is because there is only one transmit antenna, and phase difference
between reference and object wave needs not to be characterized.
Indirect holography setups further divide into off-axis and inline holography. In
the following, those two types are distinguished. Thereafter, receiver design will be
discussed.

13.4.3 Off-Axis Holography

Off-axis holographic setups include a free space reference wave that is also pro-
jected to the receiving aperture, but from a steeper angle of incidence HRef . The
situation is illustrated by Fig. 13.8. A well-collimated reference beam only has a
small k-space bandwidth around a center kRef determined by HRef , whereas the
object wave will exhibit a larger spatial bandwidth DkObj .

kRef ¼ k0 sinðHRef Þ ð13:36Þ

Without loss of generality, let the axis of the reference beam be in the xz-plane as
indicated in Fig. 13.8. Then ERef is described by

ERef ðx; yÞ ¼ ERef;0 ðx; yÞejkRef x ; ð13:37Þ

Fig. 13.8 Principal setup for Coherent Beam

off-axis holography
Source

Reference Object

Receiver
13 Holographic Imaging Approach 469

where ERef;0 has got as small spectral width. The measurement values recorded by a
square law detector are proportional to the intensity iðx; yÞ of the interference signal
by the reference and object waves at the place of the antenna element.
 2  2
iðx; yÞ ERef þ EObj  ¼ jERef j2 þ EObj  þ ERef EObj

þ ERef EObj ð13:38Þ

 n  2 o n o n o
I kx ; ky FT2D jERef j2 þ EObj  þ FT2D ERef EObj

þ FT2D ERef EObj

n  2 o n o
¼ FT2D jERef j2 þ EObj  þ FT2D fERef g  FT2D EObj

n o


þ FT2D ERef;0  UObj kx  kRef ; ky

ð13:39Þ

The first addend in (13.39) comprises the magnitude

of the coherent background
and the object wave. The last term contains UObj kx  kRef ; ky , which is the Fourier
transform of EObj , displaced by kRef in x-direction. The ⊗ symbol means that it is
convoluted with the spectrum of the complex conjugate reference wave, indicated
by the * symbol in the upper index. The second addend includes a so-called n twin o

image. Again, for a well-collimated reference beam, the influence of FT2D ERef;0
on the last addend will be small because of its small bandwidth DkRef compared to
DkObj . A sketch of the complete plane wave spectrum is depicted in Fig. 13.9, for
example, case that the image is about as large as the aperture. In this figure, all three
addends appearing in (13.39) can be seen. From their bandwidth and position, it
follows that they will not overlap if (13.40) holds.

kRef  3DkObj þ DkRef ð13:40Þ

In this case, the last addend can be isolated by filtering the k-space. For the
derivation of the original object field, there are three possibilities. If DkRef is small
enough, the influence of the reference wave can be neglected. For that, the reference

Fig. 13.9 K-space distribution for an off-axis holography setup

470 C. Schildbach and L.-P. Schmidt

wave must be approximately a plane wave, which requires noticeable effort in the
setup. If ERef is approximately described by a fundamental mode Gaussian beam, it
acts as a taper function, affecting the resolution, but at the same time decreasing
clutter by side lobes. Another way of recovering the object wave is deconvolution.
As long as there is no zero of the reference wave in the aperture area, the procedure
is stable. For the best signal-to-noise ratio on the other hand, matched filtering can
be applied. The two latter procedures require the knowledge of the reference wave.
This either has to be measured in magnitude and phase, but a simulation is usually
sufficient. The final result is much more sensitive to the relative phase between two
points in the aperture than to the absolute value of the phase, so small deviations in
the phase center position of the simulated reference wave can be tolerated [34].
The restriction introduced by (13.40) implies that the maximum angle of inci-
dence from the object must be much smaller than HRef , inherently limiting the
resolution in x-direction. The resolution in y-direction is not impaired. This phe-
nomenon has been investigated quantitatively in [35]. The maximum wavenumber
kRef;max ¼ k0 is obtained if HRef ¼ 90 . Using (13.40), DkObj is limited to
DkObj k0 =3.

) dx;min  3k ð13:41Þ

The spacing of an array in the aperture must obey a strict Nyquist criterion for
the highest spatial frequency kRef þ DkObj .
p
) Dx ð13:42Þ
kRef þ DkObj

13.4.4 Inline Holography

For inline holography, the reference and the illumination beam axis are identical. In
this way Gabor produced his first results. This way, the recorded intensity still
contains the three addends as in (13.39), but they are not separated in the spectrum.
So other methods have to be used to isolate the image information. Gabor realized
that the image is in focus at a different depth than the other terms. Those therefore
contribute uniformly to the clutter in the whole image, but with the correct depth,
the image can be reconstructed. Hence depth resolution is crucial to the clutter level
of the image. A step frequency approach increases the bandwidth of the setup,
thereby improving the depth resolution and thus image clutter by background and
twin image are reduced as reported in [36].
On the counterpart of the increased clutter, inline holography requires lower
sampling density and less effort for the setup, because only a single beam has to be
considered. Also, the required coherence length of the illumination is smaller,
because the difference in propagation distance between the reference and object
wave can be made small by proper setup design.
13 Holographic Imaging Approach 471

13.4.5 Modes of Detection

Many of the reported indirect microwave holograms have actually been recorded
using mixers in a network analyzer instead of square law detectors, ignoring the
provided phase information [34–37]. In this paragraph, those two receiving meth-
ods are compared.
First, square law detectors are considered. These also called direct detectors
require an element with nonlinear characteristic, often a Schottky diode [38]. In a
small range around the operating point V0 ; I0 , the characteristic can be approxi-
mated by a second-degree Taylor polynomial.

R0 1
I ðV Þ  ðV  V 0 Þ2 þ ðV  V0 Þ þ I0 ð13:43Þ
Zdif Zdif

The choice of the coefficients R0 and Zdif seems arbitrary at the moment, but
their meaning is discussed in the following. Provided with a sinusoidal voltage
oscillating at an angular frequency x with sufficiently small magnitude V, ^ a
parabolic characteristic will produce three frequency components as shown in
(13.44).

^2
V ^
V V^2
iðtÞ ¼ R0 sinð2xtÞ þ sinðxtÞ þ R0 þ I0
2Zdif Zdif 2Zdif
^
V ð13:44Þ
¼ R0 Pin sinð2xtÞ þ sinðxtÞ þ R0 Pin þ I0
Zdif |fflffl{zfflffl}
Idet

The component at x is determined by the differential impedance Zdif of the

device. The DC component includes the bias current I0 given by the operating point
and additionally the detection signal Idet , which is proportional to R0 and input
power Pin . R0 can therefore be identified as the responsivity.

I 00 ðV Þ
R0 ¼ ð13:45Þ
2I 0 ðV Þ

In order to distinguish Idet from the DC bias I0 , the input signal is usually
chopped and fed into a lock-in amplifier. The third spectral component is located at
a frequency of 2x and is therefore not used. R0 can be identified as the sensitivity of
the detector, which is hence determined by the first and second derivative of the
characteristic. The dynamic range of a detector reaches from the noise equivalent
power NEP to the power, where second-order Taylor approximation is no longer
valid. Obviously, the NEP depends not only on the sensitivity of the detector, but
also on the noise level. The three most relevant sources of noise for a diode detector
are thermal noise, shot noise introduced by a bias current and 1/f-noise. Depending
on the chopping frequency, the latter may be predominant. Otherwise, the output
472 C. Schildbach and L.-P. Schmidt

noise voltage of the detector is proportional to the square root of bandwidth.

Therefore, the NEP is given in the unit pWffiffiffiffiffiffi. Accordingly, the detected SNR is
Hz
given by (13.46).

P2RF
SNRdet ð13:46Þ
Df

Unfortunately, diodes always come with parasitic components. A common

equivalent circuit diagram of the parasitic components is shown in Fig. 13.10. The
cutoff frequency xc of the diode is defined as (13.47).

1
xc ¼ ð13:47Þ
Rs CJ

Beyond this frequency, only a fraction of the signal power reaches the junction
and so sensitivity of the detector drops drastically.
Resistive mixers on the other hand often also use Schottky diodes, suffering from
the same kind of cutoff frequency. The local oscillator (LO) current changes the
operating point of all diodes in the mixer periodically, modulating the differential
impedance and hence the radio frequency (RF) current. The created voltage com-
ponent at the intermediate frequency (IF) is proportional to the RF component.
A mixer can hence be considered a linear component. The IF noise power is again
proportional to the bandwidth. The resulting SNR differs from the detector funda-
mentally in that the former is only proportional to the input power PRF and the latter
to P2RF .

PRF
SNRmix ð13:48Þ
Df

As a consequence, the dynamical range of a mixer measured in dB is roundabout

double the dynamical range of the detector. On the other hand, the setup of a mixer
is much more complicated than the one of a square law detector. There are two
signal components at high frequency, namely the RF and the LO. The device needs
to be matched for both. All three ports need to be decoupled, which can be
accomplished by filtering or the introduction of symmetries, coming at the cost of

Fig. 13.10 Equivalent circuit

diagram of a diode including
parasitic components
13 Holographic Imaging Approach 473

more diodes with identical characteristics, as far as possible. Summarizing, mixers

are costly and require considerable space impeding a dense array of receivers. To
circumvent the space problem, the signals can be fed from the antenna to the mixer
via a waveguide, introducing additional loss. A detector setup on the other hand is
simple. Only one port carries a high frequency signal that can require a matching
network. The frequency difference to the bias and converted signal is large and so
filtering effort is low. In research, increasing attention is devoted to zero bias
Schottky diodes that exhibit high responsivities of over 4 WA without the need for a
bias circuit [39–44]. Such diodes are available with cutoff frequencies in the tera-
hertz domain. Still, the differential impedance of these diodes is in the order of
1 kX.

13.4.6 Planar Antennas for Detector Arrays

The majority of antenna elements for an array are either horn antennas or planar
antennas. For flip chip mounted zero bias Schottky diodes, a planar antenna design
is appropriate. The topology of the antenna depends on the bandwidth of the
application. Self-complementary antennas feature a frequency independent source
impedance of about 190 X. Such broadband antennas are used in [41, 43] and yield
a flat frequency response, while resonant antennas as in [42, 45] provide better
source point matching for a small bandwidth. If the antenna elements are well
matched to a high impedance, mutual coupling is reduced by an effect similar to
preamplifier decoupling in MRI [46]. For high bandwidths like one waveguide
band and bearing in mind the restrictions on space consumption, a good compro-
mise is achieved by using limited size bow tie antennas [47–49]. The source point
impedance of a bow tie antenna is determined by its bow angle. The smaller the
bow angle the higher the input impedance. For impedances in the kX regime, the
bow angle is too small to be considered much different from a strip line. The DC
feed would be just as important to the antenna characteristic as the antenna layout
itself. In order to preserve a small opening angle near the antenna feed and at the
same time having a low RF coupling to the DC connections, the antenna element
shown in Fig. 13.11 is considered. A zero bias Schottky diode is mounted in the
center of a bow tie like structure on a quartz substrate. As depicted in Fig. 13.12,
the symmetric shape is determined by two ellipses. The antenna is optimized for a
frequency band from 230 to 325 GHz. The parameters for optimization include the
dimensions shown in Fig. 13.12 as well as the length and width of the rectangular
defected ground structure on the rear side. The latter increases the antenna source
point impedance. The goal of the optimization is to increase the minimum received
signal power for all angles of incidence with H\40 . A possible concept for a
474 C. Schildbach and L.-P. Schmidt

Fig. 13.11 Planar antenna

element for use in 300 GHz
indirect holographic imaging
system

Fig. 13.12 Scheme of the

antenna structure. The
symmetric shape is
determined by the large and
half axis of the ellipses
a1 ; b1 ; a2 and b2

receiver array discussed earlier is to sweep a line array across the aperture area for
instance in the x-direction as indicated in Fig. 13.13. With two sparse line arrays
that are shifted by n times k=2 in x-direction and k=2 in y-direction, the distance
between two neighboring antenna elements can be enlarged, diminishing mutual
coupling. After n steps in y-direction, the second array will be at the x-position of
the first array, filling up the array and fulfilling the k=2 sample criterion. Another
potential improvement would be a higher level of integration. Integrating the
detector diodes, the planar antennas as well as low noise amplifiers on a semi-
conductor waver level would allow for comparatively cheap and precise production
of a large number of identical receiver units.
In the following, measurement results of a holographic setup as shown in
Fig. 13.14 are discussed. The major difference to the setup in Fig. 13.8 is the
position of the receive array behind the object, so transmission is recorded instead
of reflection. The operating frequency is chosen to be 312 GHz. The angle HRef is
40° and the step size between two measurement positions is about 400 lm. The
results have been recorded using the described bow tie antenna. Figure 13.15 and
13 Holographic Imaging Approach 475

Fig. 13.13 Possible spacing

of linear array of bow tie
structures, fulfilling the
sampling criterion and
minimizing cross-coupling

Fig. 13.14 Setup for direct

holography in the millimeter
wave regime

Fig. 13.16 show the recorded intensity of the reference beam and the object beam of
free space, respectively. A Gaussian distribution can be noticed. Their interference
pattern is shown in Fig. 13.17 and the magnitude of the respective k-space in
Fig. 13.18. An object like the ones photographed in Figs. 13.19 and 13.20 perturb
the received field (Figs. 13.21 and 13.22). The magnitude of the k-space is given in
Figs. 13.23 and 13.24. For the imaging algorithm, a region around kx ¼ 4270 m 1
,
ky ¼ 0 m of width Dkx ¼ 2830 m and height Dky ¼ 5590 m is processed. One
1 1 1
476 C. Schildbach and L.-P. Schmidt

Fig. 13.15 Measured

intensity of the reference
beam in the aperture

Fig. 13.16 Measured

intensity of the empty space
object beam in the aperture

Fig. 13.17 Interference

pattern of the intensities of
reference beam (Fig. 13.15)
and object beam (Fig. 13.16)

Fig. 13.18 K-space

representation of Fig. 13.17
13 Holographic Imaging Approach 477

Fig. 13.19 Photograph of the

first measurement object

Fig. 13.20 Photograph of the

second measurement object

resolution cell measures about consequently dx ¼ 2:2 mm in width and dy ¼

1:1 mm in hight. The reconstructed images are shown in Figs. 13.25 and 13.26.
Despite considerable speckle, the prediction of the resolution is confirmed by the
images.
478 C. Schildbach and L.-P. Schmidt

Fig. 13.21 Measured

interference pattern for the
first object

Fig. 13.22 Measured

interference pattern for the
second object

Fig. 13.23 K-space

representation of Fig. 13.21

Fig. 13.24 K-space

representation of Fig. 13.22
13 Holographic Imaging Approach 479

Fig. 13.25 Reconstructed

image of the first object

Fig. 13.26 Reconstructed

image of the second object

13.5 Summary

In this chapter, different categories of holography have been compared and their
specialties have been investigated. Mostly, the system effort is determined by the
chosen category. For instance, a direct holographic setup using mixers will be more
expensive in design and components than an indirect holographic setup using
simple detector diodes. The latter on the other hand will suffer from a lower
dynamic range. For off-axis holography also the smallest obtainable resolution is
impaired. Finally, a specific example case has been discussed, including an off-axis
indirect holographic setup. The planar antenna design has been presented and
measurement data have been shown. The predicted performance agrees well with
the obtained data.

References

1. D. Gabor, Holography—the reconstruction of wavefronts. Electron. Power 12(7), 230–234

(1966)
2. D. Gabor, Holography, 1948–1971. Proc. IEEE 60(6), 655–668 (1972)
3. J.W. Goodman, Introduction to Fourier Optics, 2nd edn. (McGraw-Hill, 1996)
480 C. Schildbach and L.-P. Schmidt

4. E.N. Leith, Quasi-holographic techniques in the microwave region. Proc. IEEE 59(9),
1305–1318 (1971)
5. E.N. Leith, Optical processing techniques for simultaneous pulse compression and beam
sharpening. IEEE Trans. Aerosp. Electron. Syst. AES-4(6), 879–885 (1968)
6. L. Rayleigh, XXXI. Investigations in optics, with special reference to the spectroscope. Phil.
Mag. Ser. 5 8(49), 261–274 (1879)
7. A.W. Jones, J. Bland-Hawthorn, P.L. Shopbell, Towards a general definition for spectro-
scopic resolution, in ASP Conference on Astronomical Data Analysis Software and Systems
IV (1995), pp. 503
8. A. Sommerfeld, Electrodynamics: Lectures on Theoretical Physics (Academic Press, 2013)
9. J.A. Stratton, L.J. Chu, Diffraction theory of electromagnetic waves. Phys. Rev. 56(1),
99–107 (1939)
10. S. Silver, Microwave aperture antennas and diffraction theory. J. Opt. Soc. Am. 52(2),
131–139 (1962)
11. S. Gu, C. Li, X. Gao, Z. Sun, G. Fang, Three-dimensional image reconstruction of targets
under the illumination of terahertz Gaussian beam—theory and experiment. IEEE Trans.
Geosci. Remote Sens. 51(4), 2241–2249 (2013)
12. L. Mandel, Interpretation of instantaneous frequencies. Am. J. Phys. 42(10), 840–846 (1974),
http://scitation.aip.org/content/aapt/journal/ajp/42/10/10.1119/1.1987876
13. R.L. Haupt, Thinned arrays using genetic algorithms. IEEE Trans. Antennas Propag. 42(7),
993–999 (1994)
14. D.G. Leeper, Isophoric arrays-massively thinned phased arrays with well-controlled
sidelobes. IEEE Trans. Antennas Propag. 47(12), 1825–1835 (1999)
15. Y.T. Lo, S.W. Lee, Antenna Handbook: Theory, Applications, and Design (Springer Science
& Business Media, 2013)
16. S. Caorsi, A. Lommi, A. Massa, M. Pastorino, Peak sidelobe level reduction with a hybrid
approach based on GAs and difference sets. IEEE Trans. Antennas Propag. 52(4), 1116–1121
(2004)
17. B.D. Steinberg, The peak sidelobe of the phased array having randomly located elements.
IEEE Trans. Antennas Propag. 20(2), 129–136 (1972)
18. B.D. Steinberg, Comparison between the peak sidelobe of the random array and
algorithmically designed aperiodic arrays. IEEE Trans. Antennas Propag. 21(3), 366–370
(1973)
19. K.C. Kerby, J.T. Bernhard, Sidelobe level and wideband behavior of arrays of random
subarrays. IEEE Trans. Antennas Propag. 54(8), 2253–2262 (2006)
20. C.A. Balanis, Antenna Theory: Analysis and Design (Wiley, 2005)
21. R.M. Leahy, B.D. Jeffs, On the design of maximally sparse beamforming arrays. IEEE Trans.
Antennas Propag. 39(8), 1178–1187 (1991)
22. I.J. Gupta, A.A. Ksienski, Effect of mutual coupling on the performance of adaptive arrays.
IEEE Trans. Antennas Propag. 31(5), 785–791 (1983)
23. P. Kabal, Time windows for linear prediction of speech. Technical Report (Telecomunications
& Signal Processing Laboratory, Electrical & Computer Engineering, McGill University, Jan.
2003)
24. G. Tricoles, E.L. Rope, R. Hayward, Improved resolution in microwave holographic images.
IEEE Trans. Antennas Propag. 29(2), 320–326 (1981)
25. R.E. Abdel-Aal, Expansion of two-dimensional imaging apertures for resolution improvement
in long-wavelength holography. IEE Proc. I (Commun. Speech Vision) 137(3), 157–162
(1990)
26. M.D. Migliore, A simple introduction to compressed sensing/sparse recovery with
applications in antenna measurements. IEEE Antennas Propag. Mag. 56(2), 14–26 (2014)
27. G.R. Lockwood, J.R. Talman, S.S. Brunke, Real-time 3-D ultrasound imaging using sparse
synthetic aperture beamforming. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 45(4), 980–
988 (1998)
13 Holographic Imaging Approach 481

28. S.S. Ahmed, A. Schiessl, L. Schmidt, A novel fully electronic active real-time imager based
on a planar multistatic sparse array. IEEE Trans. Microw. Theory Tech. 59(12), 3567–3576
(2011)
29. H. Shigesawa, K. Takiyama, T. Toyonaga, O. Hirao, Microwave holography by synthetic
aperture. Proc. IEEE 60(1), 137–139 (1972)
30. W.E. Kock, Side-looking radar, holography, and doppler-free coherent radar. Proc. IEEE 56
(2), 238–239 (1968)
31. Y.K. Chan, V.C. Koo, An introduction to synthetic aperture radar (SAR). Prog. Electromagn.
Res. B 2, 27–60 (2008)
32. N. Gebert, G. Krieger, A. Moreira, Digital beamforming on receive: techniques and
optimization strategies for high-resolution wide-swath SAR imaging. IEEE Trans.
Aerosp. Electron. Syst. 45(2), 564–592 (2009)
33. R.W. Larson, E.L. Johansen, J.S. Zelenka, Microwave holography. Proc. IEEE 57(12), 2162–
2164 (1969)
34. A. Tamminen, J. Ala-Laurinaho, A.V. Raisanen, Indirect holographic imaging at 310 GHz, in
2008 European Radar Conference, EuRAD-2008 (2008), pp. 168–171
35. J. Adametz, F. Gumbmann, L. Schmidt, Inherent resolution limit analysis for millimeter-wave
indirect holographic imaging, in 2011 German Microwave Conference (GeMIC) (2011),
pp. 177–182
36. X. Gao, C. Li, Z. Sun, G. Fang, Implementation of step-frequency continuous-wave scheme
in millimeter-wave inline holography for interferences elimination. IEEE Antennas Wirel.
Propag. Lett. 12, 1176–1179 (2013)
37. A. Enayati, A. Tamminen, J. Ala-Laurinaho, A.V. Raisanen, G.A.E. Vandenbosch, W. de
Raedt, THz holographic imaging: a spatial-domain technique for phase retrieval and image
reconstruction, in 2012 IEEE MTT-S International Microwave Symposium Digest (MTT)
(2012), pp. 199–201
38. B.L. Sharma, Metal-semiconductor Schottky Barrier Junctions and Their Applications
(Springer Science & Business Media, 2013)
39. J.L. Hesler, T.W. Crowe, NEP and responsivity of THz zero-bias Schottky diode detectors, in
15th IEEE International Conference on Terahertz Electronics (IEEE, 2007), pp. 844–845
40. M. Hoefle, A. Penirschke, O. Cojocari, R. Jakoby, Advanced RF characterization of new
planar high sensitive zero-bias Schottky diodes, in 2011 European Microwave Integrated
Circuits Conference (EuMIC) (2011), pp. 89–92
41. L. Liu, JL. Hesler, H. Xu, A.W. Lichtenberger, R.M. Weikle, A broadband quasi-optical
terahertz detector utilizing a zero bias Schottky diode. IEEE Microwave Wirel. Compon. Lett.
20(9), 504–506 (2010)
42. V.I. Shashkin, Y.A. Drjagin, V.R. Zakamov, S.V. Krivov, L.M. Kukin, A.V. Murel, Y.I.
Chechenin, Millimeter-wave detectors based on antenna-coupled low-barrier schottky diodes.
Int. J. Infrared Millimeter Waves 28(11), 945–952 (2007)
43. A. Semenov, O. Cojocari, H.-W. Hübers, F. Song, A. Klushin, A.-S. Müller, Application of
zero-bias quasi-optical Schottky-diode detectors for monitoring short-pulse and weak
terahertz radiation. IEEE Electron Device Lett. 31(7), 674–676 (2010)
44. X. Fan, X. Pei, X. Xiong, Zero bias Schottky diodes use in high performance detection
circuits, in 2011 International Conference on Electronics and Optoelectronics (ICEOE)
(2011), pp. V3-27–V3-30
45. J. Montero-de-Paz, E. Ugarte-Munoz, L.E. Garcia-Munoz, D. Segovia-Vargas,
D. Schoenherr, I. Oprea, A. Amrhein, O. Cojocari, H.L. Hartnagel, Millimeter-wave receiver
based on a folded dipole antenna and Schottky diode for maximum power transfer, in 2012
6th European Conference on Antennas and Propagation (EUCAP) (2012), pp. 1259–1262
46. P.B. Roemer, W.A. Edelstein, C.E. Hayes, S.P. Souza, O.M. Mueller, The NMR phased
array. Magn. Reson. Med. 16(2), 192–225 (1990)
47. D.B. Rutledge, M.S. Muha, Imaging antenna arrays. IEEE Trans. Antennas Propag. 30(4),
535–540 (1982)
482 C. Schildbach and L.-P. Schmidt

48. C. Schildbach, J. Schür, L.-P. Schmidt, Broadband detector array concept for 3D holographic
imaging at THz frequencies, in 2013 European Microwave Conference (EuMC) (2013),
pp. 1243–1246
49. V.A. Petriakov, F.F. Sizov, O. Golenkov, S.A. Dvoretskii, D.S. Krasilnikov, Direct detection
MM-wave linear antenna array on the base of hemispherical lens antenna elements coupled
with narrow-gap hot-carrier bolometers, in 2013 International Kharkov Symposium on
Physics and Engineering of Microwaves, Millimeter and Submillimeter Waves (MSMW),
(2013), pp. 79–81
Index

A B
Aberrations, 365 Back-Gap (BG) structure, 235
Addition theorems, 50 Back projection, 457
Additive manufacturing, 96, 127, 137 Backward leaky wave, 128, 133
Adiabatic Floquet Wave (FW), 297, 300 Balanced hybrid, 341
Adiabatic FW expansion, 298 Balanced hybrid condition, 99
Air-cavity, 362 Bandwidth, 8, 21, 97, 145, 150, 184, 330
Air-filled cavities, 236 Bandwidth enhancement, 227
Air-filled feeding waveguide, 101 Bandwidth enlargement, 317
Alternate projection method, 257 Bandwidth improvement, 231
Amplitude hologram, 434 Base station, 20
Analysis, 310 Beamforming, 8
Analytical transformation, 388 Beamforming network, 27
Anechoic chamber, 429 Beam-lead, 374
Angular spectrum of plane waves, 456 Beams, 4
Anisotropic, 229 Beam scanning, 12, 57
Anisotropy, 16 Beam shape, 26
Annular, 134 Beam-steering, 212
Antenna array, 147 Beam-steering ILAs, 29
Antenna elements, 473 Beam-switching, 212
Antenna Pattern Comparison (APC) method, Beamwaist, 91
439 Bespoke lens, 395
Antenna Under Test (AUT), 443 Bessel beam, 124, 243, 249, 250, 254
Aperture efficiency, 23, 116, 162, 201, 202 Bessel beam launcher, 271, 276, 281
Aperture field, 314 Bessel distribution, 248
Aperture illumination efficiency, 80 Bessel functions, 255, 341
Array, 235 Bessel launcher, 270
Array factor, 160–163, 461 Biasing network, 179
Artificial magnetic conductor, 227 BICEP, 377
Atomic, 294 Bipolar grid, 419
Automotive radar, 8, 29, 30, 57, 194 B-mode, 377
Automotive radar systems, 26 Body scanner, 88
Axial and radial corrugated feed horn, 88 Bolometer, 350
Axial and radial corrugations, 87 Bolometer detectors, 14
Axial beam symmetry, 76 BoR-FDTD, 347
Axial corrugations, 87 Boundary integral equation (BIE) methods, 8

© Springer International Publishing AG 2018 483

A. Boriskin and R. Sauleau (eds.), Aperture Antennas for Millimeter
and Sub-Millimeter Wave Applications, Signals and Communication Technology,
DOI 10.1007/978-3-319-62773-1
484 Index

Bow tie antennas, 473 Current distribution, 152, 175

Broadband antennas, 232 Curved planar waveguide, 401
Broadband ILAs, 21 Cutoff frequency, 115
Broadside, 324, 326, 357 Cylindrical lens antennas, 38
Broadside beam, 131 Cylindrical Luneberg lenses, 59
Bull’s-eye, 116, 120 Cylindrical traveling wave, 243

C D
Calibration, 465 Database, 307
Canonical geometries, 46 Debye potentials, 40
Cassegrain, 151 Debye’s asymptotic formulas, 46
Categories of holography, 464 Deep Reactive-Ion Etching (DRIE), 344, 362
Cavities, 357 3D electromagnetic bandgap, 98
Cheaper to manufacture, 94 Design, 326
Circular-cylindrical dielectric lenses, 59 Design process, 295, 296
Circularly polarized, 229 Design re-usability, 326
Circularly polarized Bessel beam, 256 Detection of threats, 467
Circularly polarized radial line slot array, 257 Diagonal horns, 337
Circular Polarization (CP), 158, 166, 169, 181, Dielcore horns, 100
205 Dielectric FP antennas, 226
Circular-polarized Bessel beam launcher, 266 Dielectric lens, 4, 355
Cloud, 351 Dielectric material, 15
Clutter, 462 Diffraction, 112, 243, 454
CMB, 377 Diffraction effects, 10
CMOS, 336, 372 Digital beam forming, 461
CNC, 343 Direct detectors, 471
Coherent background, 452 Direct holography, 464
Collimator, 122 Discrete lens, 192
Communications satellite, 97 Dispersion, 370
Communication systems, 184 3D periodic, 104
Compact, 117 2D periodic arrays, 229
Compact Antenna Test Range (CATR), 410 3D-printed, 126
Compact-size dielectric lenses, 9 3D printing, 345
Compensated compact range, 432 Double-layer PRS, 232
Complex Huygens source, 56 Double Lens (DL) system, 31
Compound eye, 8 Double-shell dielectric lens, 9
Conical horn, 103 Double-shell shaped dielectric lens, 21
Constant flux illumination, 20 Double slot, 359
Constant-n lens, 58 Dual-band, 23
Constant phase-shifters, 202 Dual-band antenna, 232
Conversion efficiency, 319 Dual-bandlow-profile, 119
Corrugated antennas, 120 Dual-polarized reflect array, 157
Corrugated horn, 76, 340 Dual-slot feeding technique, 235
Corrugation depth, 82
Corrugation period, 82 E
Corrugation tooth width, The, 83 Eaton lens, 402
Coupled-layers, 205 Edge amplitude taper, 202
Coupling coefficient, 81 Edge illumination, 7, 56
Courtesy of ANTERAL S.L., 96, 98 Effective normal, 148
Coverage, 20 Effects of lens truncation, 65
Critical angle, 6 Efficiency, 199, 317, 319, 320
Cross-coupling, 465, 466 EH1n mode, 78
Cross-polarization, 14, 231 Electrodeposition, 92
CubeSat, 124 Electroforming, 90, 345
Index 485

Electronic reconfiguration, 193 G

Electroplating, 92 GaAs, 350
Elliptical lens, 5, 11 Gain, 162
Equidistant sampling, 461 Gaunt coefficients, 50
Equivalent Circuit Model (ECM), 172 Gaussian, 338
Equivalent current distribution, 49 Gaussian-like mode, 80
Evanescent coupling, 105 Gaussian modes, 80
Extender module, 415 Gaussian ring basis functions, 302
Extraordinary optical transmission, 112 Gaussicity, 12, 22, 26, 340
Gauss-Laguerre, 338
F Generalized reflection and transmission
Fabrication, 57, 227 coefficient, 46
Fabrication accuracies, 18 Generalized Scattering Matrix (GSM), 154,
Fabrication tolerance, 87 160, 161, 170
Fabry-Perot cavity, 132 Genetic algorithms, 11, 342
Fabry-Perot cavity antenna, 222 Geometrical optics, 8, 245, 251
Far-field method, 410 Geometrical theory of diffraction, 251
FDTD, 8 Global full-wave, 310
Feed, 318 Gold, 370
Feed blockage, 163 Gradient-index dielectric flat lenses, 209
Feed directivity and position, 57 Graphene, 166, 185
Feeding techniques, 227, 235 Grating, 133
Feed position, 54 Grating lobes, 155, 379
Feeds, 13 Green’s function approach, 39
Feed scanning, 435 Green’s function in the spectral domain, 42
Feed selection, 57 Green’s functions for layered media, 39
Ferroelectric films, 166 Gregorian, 151
Field of view, 7 Grooves, 113, 134
Fields, 300 Ground penetrating radar, 467
Flare angle, 343 Ground-Signal-Ground (GSG) probe, 445
Flare region, 83 Guided-wave approach, 180
Flat lens, 192 Guided-wave cells, 168
Floquet, 368 Guided-wave unit cell, 157
Floquet analysis, 149, 204 Guidelines, 132
Floquet mode, 153, 155, 174
Floquet series, 153 H
FMM, 311 Half-bowtie resonances, 9
Focal arc, 7, 30 Half Maxwell fisheye lens, 47
Focal distance, 214 Hankel aperture distributions, 255
Focal source, 197, 201 Hankel beam, 250
Focusing, 8 Hankel transform, 250
Focusing system, 31 HD TV Display, 30
Folded reflect array, 151 HE11 mode, 78
Folded transmit array, 215 Heat harvesting, 136
Forward leaky wave, 128, 133 Hemiellipse, 6
Fourier transform, 420 Hemisphere, 6
Frequency bandwidth, 202 Hemispherical lens with a ground plane, 47
Frequency-selective surfaces, 222 High-data-rate communication, 3
Frequency-shifting, 135 High frequencies, 94
Full-widthhalf-maximum, 88 High frequency techniques, 8
Fundamental mode, 77 High Impedance Surfaces, 229
Future research, 330 High-precision manufacturing, 93
486 Index

High resistivity, 366 Leaky radial waveguide, 273, 276

Hole arrays, 114 Leaky wave, 115, 336
Hologram CATR, 433 Leaky-wave antennas, 275
Holographic imaging, 441 Leaky-wave modes, 245, 275
Holographic setup, 474 Leaky-wave theory, 223
Holography, 256, 451, 468 Lens antenna, 3, 388
Homogeneous cylindrical lenses, 59 Lens CATR, 433
Horn, 116 Lens fabrication, 17
Horn antennas, 81 Limb sounder, 89
Horn-like structures, 104 Linear error model, 465
Huygens source, 48 Linear polarized modes, 99
Hybrid mode horns, 99 Liquid crystal, 166, 168, 180, 184, 185, 197
Hybrid modes, 77 Local and global coordinate systems, 49, 50
Hyperbolic lens, 392 Local Multi-Point Distribution (LMDS), 184
Longitudinal, 121
I Longitudinal resonant modes, 64
IBC, 302 Losses, 319
ICP, 362 Love’s equivalence theorem, 41
Illumination, 360 Low cross-polarization, 76
Illumination efficiency, 199 Low Earth Orbit satellite, 29
Imaging, 4, 194 Low return loss, 84
Impedance, 6, 312 Low sidelobes, 76, 236
Improper waves, 128 Low spillover, 88
Indirect holography, 468 Luneburg, 47
Infinite array, 152 Luneburg lens, 47, 51, 53, 393
Inline holography, 470
In-phase, 133 M
Input radius, 82 Macroscopic, 294
Integral equation, 302 Magnetic dipole, 235
Integrated lens antenna, 3 Mandrel, 92, 343
Intelligent transport, 3 Mass-production, 19
Inverse Fourier transform, 420 Matching layers, 6, 24
Inward cylindrical traveling wave, 254 Matlab, 364
Inward cylindrical wave, 248 Maxwell fisheye variation of index of
Inward cylindrical wave distribution, 247 refraction, 47
Isofrequency, 370 Measurements, 327
Isofrequency dispersion curves, 307 Mechanical beam-steering, 209
Isotropic, 323 Mechanical scanning ILA, 29
Membrane, 362
J MEMS switches, 196
JPL, 350 MEMS varactors, 196
Mesoscopic, 294
K Metallic and metallo-dielectric FP antennas,
Key features, 291 224
K/Ka-band, 97 Metallic walls, 98
K-space, 456 Metallo-dielectric FPC antenna, 235
Metal meshes, 235
L Metamaterial, 113, 290
Lab prototyping, 19 Metamaterial based feeds, 98
Laser sintering, 96 Metasurface, 59, 100, 243, 283, 290, 313, 325
Lathe, 87 Metasurface antenna, 216
Lattice, 306 Metasurface feed, 215
Layout, 314 Method of moments (MoM), 155, 176, 311
Leakage constant, 128
Index 487

Micro Electro Mechanical Systems (MEMS), N

166, 168, 182 Narrowband performance, 231
Microfluidics, 196 Near-field method, 410
Microlens, 136, 357 Noise, 471
Micromachining, 227 Non-diffraction radiation, 245
Microstrip, 145 Non-diffractive range, 276
Microstrip patch, 149 Non-Euclidean transformation, 396
Microwave holography, 441 Non-homogeneity, 16
Microwave holography applications, 452 Numerical control milling, 17
Microwave mammography, 467 Nyquist, 378
Microwave range, 120 Nyquist sampling criterion, 410
Mie series, 38
Millimeter, 125 O
Millimeter-wave mobile access, 194 Object field, 469
Minimum cylindrical lens height, 61 Objective lens, 30
Minimum scan loss, 31 Off-axis, 126, 468
Mixed-mode generalized scattering matrix, 170 Off-axis beaming, 136
Mm-wave frequencies, 88 Offset, 133
Mobile broadband systems, 19 Offset feed, 163
Model, 292 Offset reflectors, 151, 161
Modelling of feed antennas, 47 One-dimensional structures, 39
Mode matching techniques, 85 Optically transparent, 184
Modulated metasurface, 281, 368 Optimization, 135, 263
Modulated metasurface antennas, 329 Optimization algorithm, 9
Modulation, 323 Optimization of the profiles, 80
Molecular, 294 Optimization tool, 245
MoM impedance matrix, 305 Optimized individually, 86
Moulding, 17 Optimum directivity of feeding antenna, 56
MTS antenna design, 311 Outward cylindrical traveling wave, 254
MTS antennas, 321, 368 Overall efficiency, 320
Multi-beam, 7 Oxygen-free copper, 348
Multi-beam lens antenna, 25
Multibeam reflector antennas, 97 P
Multibeam single reflector system, 23 Paraboloidal reflector, 146
Multi-flare, 93 Parallel-plate waveguide, 59
Multi-flare angle feed horns, 94 Parasite, 117
Multi-flare horns, 345 Parseval theorem, 46
Multilayer lenses, 19 Parylene, 364
Multilayer lens-like dome, 24 Parylene-C, 364
Multi-layer periodic Fabry-Perot cavity Passive transmitarrays, 207
antennas, 235 Patch, 125, 314
Multi-mode, 345 Path length, 147
Multipactor, 125 Pattern symmetry, 76
Multi-pixel arrays, 345 Pencil beam, 146, 150
Multiple beams, 57 Periodic, 129
Multi-shell lens antennas, 12 Periodic Boundary Conditions (PBCs), 154
Multisource, 215 Periodic environment, 149, 153
Multistatic, 460 Periodic structure, 113
Mushroom structure, 103 Periodic surfaces, 222
Mutual coupling, 465, 466 Permittivity measurement, 15
488 Index

Phase center, 4, 7, 12, 22 Q

Phase constant, 128 Quantum cascade lasers, 122
Phase curve, 149, 156 Quasi-conformal transformation, 391
Phased arrays, 144, 164 Quiet-zone, 427
Phase distribution, 199
Phase error, 150 R
Phase error efficiency, 163 RADAR, 163, 183
Phase gradient, 147 Radar Cross-Section (RCS) enhancers, 57
Phase hologram, 434 Radially corrugated, 82
Phase quantization, 167, 194 Radial waveguide, 245, 273
Phase range, 194 Radiation absorbing material, 429
Phase resolution, 194 Radiation efficiency, 5, 30
Phase retrieval, 443 Radiation mechanism, 317
Phase-shifters, 196 Radiation patterns, 89
Phase-shifting surface, 192 Radio Frequency (RF) switches, 166
Photoconductive materials, 166 Ray optics, 146
Photoresist, 363 Reactance, 298
Physical optics, 8 Reconfigurable, 330
PIN diode, 166, 168, 180, 196, 213 Reconfigurable focal array, 193
Pixel, 306, 380 Reconfigurable FoV, 26
Planar, 356 Reconfigurable functionality, 146
Planar antenna, 14, 104 Reconfigurable radiation pattern, 8
Planar lens, 192, 393 Reconfigurable transmit array, 212
Planar Near-Field (PNF) technique, 418 Reconstruction, 456
Planar technology, 125 Rectangular to circular transition, 91
Plane-parallelFabry-Perot resonator, 235 Rectilinear grid, 419
Plane-polar grid, 419 Reflection coefficient, 83
Plane wave, 455 Reflection type amplitude hologram, 434
Plasmonic regime, 185 Reflective surfaces, 222
Plastic, 127 Reflector antenna, 144
Point spread function, 453 Reflector compact antenna test range, 431
Point-to-point and point-to-multi-point Reflector feeds, 21
communications, 7 Renewable energies, 136
Point-to-point links, 193 Resistive mixers, 472
Polarization-twisting, 151 Resolution, 453
Polymer, 364 Resolution in range direction, 460
Potter horns, 339 Resonant cavity, 222
Potter-type, 94 Responsivity, 350
Potter-type horns, 81 RLSA, 245, 268
Power combining, 169
Power transfer, 12 S
Principle of reciprocity, 223 Saddle point, 251
Principle of superposition, 160 SAR systems, 467
Printed antenna, 14 Satellite communications (SATCOM), 57, 194,
Profiled horn, 342 195
Profilometer, 364 Satellite payload, 97
Propagator matrix approach, 41 Satellite systems, 163
Proper waves, 128 Scalability, 135
Prototype, 321, 322, 327 Scalar, 340
Pseudo-Bessel beam, 268 Scan loss, 25
Pyramidal horn, 102, 121 Scanning, 7
Pyrolized, 364 Scanning performance, 105
Index 489

Sectoral horn, 105 Surface wave cloaking, 396

Sector isoflux beam, 312 Surface wave launcher, 125
Sensing, 4 Surface wave lens antennas, 396
Sensing devices, 136 Switched-beam transmit arrays, 211
Sequential rotation, 181 Symmetrical patterns, 102
Sequential rotation cells, 169 Synthesis, 294, 296
Sequential rotation technique, 158 Synthesis process, 13, 301
Serration, 431 Synthesis techniques, 150
Shaped-beam, 7 Synthetic aperture radar, 463
Shaped radiation patterns, 4 System on Chip (SoC), 372
Shared aperture, 57
Side lobe level, 462 T
Silicon, 361 Tapered aperture, 100
Silicon Nitride membrane, 374 Taper efficiency, 162, 202
Silicon platelets, 345 Taper functions, 462
Single-layer FP PRS, 232 Tapering, 319
Singular index, 401 Tapering efficiency, 314
Sinusoidal, 125 Tapering of the horn, 82
Sinusoidally, 368 Terahertz, 120, 335
Slot, 116, 377 Terahertz gap, 411
Smallest size, 86 85% TE11 and 15 % TM11, 81
Smooth-walled, 354 Thinning strategies, 462
Smooth-walled feed horn, 93 Three-dimensional printing, 19
Snell’s law, 10, 147 Throat region, 83
Soft-surface, 125 Time gating, 438
Space applications, 183 Timing, 306
Space harmonic, 130 Transformation optics, 291, 387
S-parameter, 415 Transmission type an amplitude hologram, 434
Spatial sampling, 460 Transmit array, 191
Spectrum of plane waves, 457 Transmit array unit cells, 205
Specular reflection, 163 Transmitting and reflecting rays, 223
Spherical, 38 Transparent, 298
Spherical mode representation of a far-field Transversal, 121
radiation pattern, 49 Transverse equivalent network, 224
Spherical wave, 455 Traveling-wave, 128
Spill-over efficiency, 162, 201, 202 Tropospheric, 351
Spline profile, 352 True Time Delay (TTD), 150, 157, 185, 202
Spline profiled, 93 \“T\” slot, 379
Splines, 11 Tuned-resonator approach, 156, 180
Split-block, 339, 345 Turrin-type horn, 95
Square law detectors, 471 Two-dimensional (2D) Fourier transformation,
Steepest descent path, 251 40
Stereolithography, 126 Two-layer lens, 58
Stop-band, 130 Two-step process, 451
Strip loaded, 100
Sublambda, 112 U
Submillimeter (mm)-wave, 97, 336 Uniform, 129
Substrate lens, 30 Unit cell, 197
Superposition of forward and backward
propagating waves, 46 V
Surface, 312 Varactor diode, 178, 196, 213
Surface current density, 303 Vector Legendre transformation, 42
Surface impedance, 101 Vector Network Analyzer (VNA), 412
Surface plasmon, 112
490 Index

W Wireless indoor transmission of uncompressed

Wave equation, 454 HD video signal, 29
Waveguide feed, 13, 215 Woodpile Electromagnetic Band-Gap
Waveguides, 215 (EBG) structures, 227
Waveguide walls, 105 Woodpile structure, 105
Wavenumber, 455
Whispering-gallery modes, 59 Z
Wideband, 317 Zero bias schottky diodes, 473