Metasurface Antennas For Synthetic Aperture Radar

Metasurface Antennas for
Synthetic Aperture Radar

by
Michael Boyarsky
Department of Electrical and Computer Engineering
Duke University
Approved:
David R. Smith, Advisor
Michael E. Gehm
Willie J. Padilla
Henry D. Pfister
Matthew S. Reynolds
Dissertation submitted in partial fulfillment of the requirements for the degree of

Doctor of Philosophy in the Department of Electrical and Computer Engineering
in the Graduate School of Duke University
2019
Abstract
Metasurface Antennas for
by
Michael Boyarsky
Department of Electrical and Computer Engineering
Duke University
Date:
Approved:
David R. Smith, Advisor
Michael E. Gehm
Willie J. Padilla
Henry D. Pfister
Matthew S. Reynolds
An abstract of a dissertation submitted in partial fulfillment of the requirements for

the degree of Doctor of Philosophy in the Department of Electrical and Computer
Engineering in the Graduate School of Duke University
2019
Copyright
c 2019 by Michael Boyarsky
All rights reserved
Abstract
Synthetic aperture radar offers unparalleled satellite imaging capabilities for plan-
etary observation. Future systems will realize high resolution with near-real-time
revisit rates by using coordinated satellites, but their development has been slowed
by the high cost, high power draw, and substantial weight associated with existing
antenna technology. Metasurface antennas — a lightweight, low cost, and planar
alternative — can address these challenges to make large scale, multi-satellite sys-
tems practical. In this work, an electronically steered metasurface antenna prototype
is developed for synthetic aperture imaging. A cohesive approach to modeling and
design led to a Nyquist sampled layout which minimizes inter-element coupling and
suppresses grating lobes. Experimental measurements validate its ability to steer a
beam in 2D across a wide bandwidth. Robust performance and favorable hardware
characteristics have poised metasurface antennas to affect many microwave indus-
tries and to facilitate multi-satellite constellations for spaceborne synthetic aperture
radar.
iv
Contents
Abstract iv
List of Tables viii
List of Figures ix
Acknowledgements xvi
1 Introduction 1
2 Synthetic Aperture Radar 11
2.1 Traditional SAR Modalities . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.1 Stripmap SAR . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.2 Spotlight SAR . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Traditional SAR Hardware . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.1 TerraSAR-X Phased Array . . . . . . . . . . . . . . . . . . . . 15
3 Metasurface Apertures 21
3.1 Metasurface Architectures . . . . . . . . . . . . . . . . . . . . . . . . 22
3.1.1 Passive Metasurface Apertures . . . . . . . . . . . . . . . . . . 22
3.1.2 Dynamic Metasurface Apertures . . . . . . . . . . . . . . . . . 24
3.1.3 Holographic Antennas . . . . . . . . . . . . . . . . . . . . . . 26
3.1.4 Numerical Modeling of Metasurface Apertures . . . . . . . . . 29
3.2 Computational Imaging Framework . . . . . . . . . . . . . . . . . . . 31
3.2.1 Mathematical Description . . . . . . . . . . . . . . . . . . . . 32
v
3.2.2 Measurement Matrix Modeling . . . . . . . . . . . . . . . . . 33
3.3 Experimental Imaging with a Dynamic Metasurface . . . . . . . . . . 34
3.3.1 1D Dynamic Metasurface Aperture . . . . . . . . . . . . . . . 35
3.3.2 Stationary Imaging . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.3 Single-Frequency Imaging . . . . . . . . . . . . . . . . . . . . 39
3.3.4 Experimental SAR with Metasurfaces . . . . . . . . . . . . . . 43
3.3.5 3D SAR Imaging . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3.6 Single Frequency SAR . . . . . . . . . . . . . . . . . . . . . . 50
4 Simulated Metasurfaces for SAR 71
4.1 Simulated SAR with Metasurface Antennas . . . . . . . . . . . . . . 72
4.1.1 Simulated Metasurface Architecture . . . . . . . . . . . . . . . 72
4.1.2 Beamforming Performance . . . . . . . . . . . . . . . . . . . . 75
4.1.3 Stripmap and Spotlight SAR Imaging . . . . . . . . . . . . . . 77
4.1.4 Alternative SAR Modalities . . . . . . . . . . . . . . . . . . . 80
4.1.5 Noise Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.2 Metasurface Tuning Strategies . . . . . . . . . . . . . . . . . . . . . . 89
4.2.1 Tuning Strategy Definitions . . . . . . . . . . . . . . . . . . . 91
4.2.2 Simulation Methodology . . . . . . . . . . . . . . . . . . . . . 95
4.2.3 Simulated Beamforming Results . . . . . . . . . . . . . . . . . 98
4.2.4 Scaled Euclidean Modulation . . . . . . . . . . . . . . . . . . 99
4.3 Large Scale Simulated Metasurface . . . . . . . . . . . . . . . . . . . 105
4.3.1 Initial Spaceborne Metasurface Design . . . . . . . . . . . . . 105
4.3.2 Final Large Metasurface Design . . . . . . . . . . . . . . . . . 109
4.4 Grating Lobe Suppression . . . . . . . . . . . . . . . . . . . . . . . . 113
4.4.1 Metasurface Antenna Architecture . . . . . . . . . . . . . . . 114
vi
4.4.2 Grating Lobe Suppression . . . . . . . . . . . . . . . . . . . . 121
5 Metasurface Antenna Prototype 131
5.1 Individual Element Design . . . . . . . . . . . . . . . . . . . . . . . . 131
5.2 Antenna Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.2.1 Challenges with Metasurface Antennas . . . . . . . . . . . . . 136
5.2.2 Metasurface Antenna Design . . . . . . . . . . . . . . . . . . . 137
5.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
5.3.1 Element Characterization . . . . . . . . . . . . . . . . . . . . 142
5.3.2 Experimental Beamforming Results . . . . . . . . . . . . . . . 148
6 Conclusion 160
Bibliography 162
vii
List of Tables
2.1 TerraSAR-X antenna parameters. . . . . . . . . . . . . . . . . . . . . 18
3.1 Resolution of single-frequency and bandwidth imaging cases. . . . . . 62
4.1 Antenna parameters for the metasurface aperture used in this section
to simulate SAR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
to simulate beamforming. . . . . . . . . . . . . . . . . . . . . . . . . 96
4.3 Broadside gain for three different tuning strategies applied to a meta-
surface antenna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.4 Antenna parameters for the initial large scale spaceborne metasurface
antenna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.5 Antenna parameters for the final large scale spaceborne metasurface
antenna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
to simulate grating lobe suppression in metasurface antennas. . . . . 118
5.1 Antenna parameters for the prototype metasurface antenna. . . . . . 138
5.2 Layers of the characterization board and the metasurface antenna pro-
totype. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.3 Antenna performance metrics targeted for the metasurface antenna

prototype. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5.4 Realized antenna performance metrics targeted for the metasurface

antenna prototype. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.1 Comparison between phased arrays, reflector dish, and metasurface

antennas as antenna solutions for spaceborne SAR. . . . . . . . . . . 161
viii
List of Figures
1.1 Depiction of the TerraSAR-X satellite synthetic aperture radar system. 2
1.2 Example microwave satellite images taken with the TerraSAR-X system. 4
1.3 Stripmap and spotlight synthetic aperture radar. . . . . . . . . . . . . 4
1.4 Metasurface computational imaging system for security screening. . . 6
1.5 Metasurface antenna prototype developed for synthetic aperture radar. 8
2.1 Illustration of conventional stripmap and spotlight SAR modalities. . 13
2.2 Examples of traditional spaceborne SAR systems. . . . . . . . . . . . 14
2.3 Schematic of the layout of TerraSAR-X. . . . . . . . . . . . . . . . . 16
2.4 TerraSAR-X operating modes, phased array antenna, full satellite,

and sample SAR image. . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.5 Azimuth steering with TerraSAR-X. . . . . . . . . . . . . . . . . . . . 19
2.6 Elevation steering with TerraSAR-X. . . . . . . . . . . . . . . . . . . 20
3.1 Initial passive metasurface aperture developed for microwave compu-

tational imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Individually-tunable metamaterial element for creating a dynamic meta-

surface aperture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Initial dynamic metasurface aperture developed for microwave com-

putational imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4 Diagram of a generic holographic metasurface antenna. . . . . . . . . 26
3.5 A sample metasurface antenna illustrating holographic antenna oper-

ation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.6 Numerical modeling procedure for 1D metasurface apertures. . . . . . 32
ix
3.7 Elements comprising an experimental dynamic metasurface aperture
used for computational imaging . . . . . . . . . . . . . . . . . . . . . 36
3.8 Diagram of the feed locations of a 1D metasurface aperture. . . . . . 36
3.9 Computational imaging environment using a dynamic metasurface

aperture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.10 Point spread function analysis for experimental stationary computa-

tional imaging with a dynamic metasurface aperture. . . . . . . . . . 37
3.11 Images of two metallic cylinders, reconstructed with matched filter

and GMRES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.12 Comparison between traditional microwave imaging and single fre-

quency multistatic imaging. . . . . . . . . . . . . . . . . . . . . . . . 40
3.13 Aperture configuration for single-frequency stationary computational

imaging with dynamic metasurfaces. . . . . . . . . . . . . . . . . . . 42
3.14 Single-frequency stationary computational imaging results. . . . . . . 44
3.15 Experimentally-optimized beam-like radiation patterns generated by

a dynamic metasurface aperture. . . . . . . . . . . . . . . . . . . . . 45
3.16 Schematic view of stripmap, spotlight, and diverse pattern stripmap,

conducted with a metasurface aperture. . . . . . . . . . . . . . . . . . 47
3.17 Experimental point spread function conducted with a metasurface

aperture in various SAR modalities. . . . . . . . . . . . . . . . . . . . 48
3.18 Experimental imaging of a line of point scatterers conducted with a

metasurface aperture in various SAR modalities. . . . . . . . . . . . . 48
3.19 Experimental 3D images measured with a dynamic metasurface aper-

ture operating in diverse pattern stripmap mode. . . . . . . . . . . . 49
3.20 A single frequency synthetic aperture imaging system. . . . . . . . . . 51
3.21 Illumination strategy for a single frequency SAR imaging system. . . 52
3.22 Experimental setup for single frequency SAR imaging. . . . . . . . . 59
3.23 Simulated 3D image of a point scatterer. . . . . . . . . . . . . . . . . 60
3.24 Experimental 3D image of a point scatterer. . . . . . . . . . . . . . . 61
x
3.25 Experimental PSF with 64 transmit masks, 64 receive masks, and a
single frequency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.26 Experimental PSF with 16 transmit masks, 16 receive masks, and 16

frequency point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.27 Experimental PSF with 64 transmit masks, 64 receive masks, and 16

frequency points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.28 Reconstructed 3D image of four copper cylinders with approximate

diameter 2 cm and length of 10 cm. . . . . . . . . . . . . . . . . . . . 65
3.29 Reconstructed 3D image of a cylinder of diameter 15 cm and height

70 cm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.30 Single-frequency SAR images of a large cylinder at 5 different fre-

quency points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.31 Experimental 3D image of four metallic cylinders taken with different

numbers of measurements. . . . . . . . . . . . . . . . . . . . . . . . . 70
4.1 Unit cell geometry used to generate a simulated metasurface antenna

for beamforming. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2 Schematic view of a simulated metasurface antenna. . . . . . . . . . . 75
4.3 Broadside beams generated with the simulated metasurface antenna. 77
4.4 Steered beams generated with the simulated metasurface antenna. . . 78
4.5 Simulated stripmap PSF conducted with a metasurface antenna. . . . 79
4.6 Simulated spotlight PSF conducted with a metasurface antenna. . . . 80
4.7 Line of point scatterers imaged with a simulated metasurface antenna

in stripmap and spotlight mode SAR. . . . . . . . . . . . . . . . . . . 81
4.8 Schematic view of enhanced resolution stripmap. . . . . . . . . . . . . 83
4.9 Simulated images conducted with a metasurface antenna in enhanced

resolution stripmap mode. . . . . . . . . . . . . . . . . . . . . . . . . 84
4.10 Schematic view of diverse pattern stripmap. . . . . . . . . . . . . . . 85
4.11 Radiation patterns optimized for diverse pattern stripmap. . . . . . . 85
4.12 Simulated images conducted with a metasurface antenna in diverse

pattern stripmap mode. . . . . . . . . . . . . . . . . . . . . . . . . . 87
xi
4.13 Noise study conducted by calculating the resulting SNR from different
SAR modalities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.14 A visualization of the phase hologram tuning strategy. . . . . . . . . 93
4.15 A visualization of the Lorentzian-constrained modulation tuning strat-

egy plotted as described in Figure 4.14. . . . . . . . . . . . . . . . . . 94
4.16 A visualization of the Euclidean modulation tuning strategy plotted

as described in Figure 4.14. . . . . . . . . . . . . . . . . . . . . . . . 94
4.17 Metamaterial element model and results from full-wave simulation. . 95
4.18 Comparison of full-wave simulation with an approximate model ap-

plied to a sample metasurface antenna. . . . . . . . . . . . . . . . . . 97
4.19 Steered, directive beams generated with different tuning strategies. . 98
4.20 The decay of the guided wave (and dipole moments) associated with
different tuning strategies applied to a metasurface antenna. . . . . . 100
4.21 A visualization of the scaled Euclidean modulation tuning strategy for

different values of A, plotted as described in Figure 4.14. . . . . . . . 102
4.22 Broadside beams generated with scaled Euclidean modulation for A =

1, 1.5, and 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.23 Broadside gain as a function of scale factor and antenna size. . . . . . 104
4.24 Large scale metasurface antenna initial design. . . . . . . . . . . . . . 106
4.25 Analytically modeled metamaterial element response for large aper-

ture beamforming simulations. . . . . . . . . . . . . . . . . . . . . . . 107
4.26 Azimuth steering capabilities of the initial large metasurface antenna

design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.27 Elevation steering capabilities of the initial large metasurface antenna

design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.28 Final large metasurface design for spaceborne SAR. . . . . . . . . . . 109
4.29 Azimuth steering capabilities of the final large metasurface antenna

design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.30 Elevation steering capabilities of the final large metasurface antenna

design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
xii
4.31 2D beam patterns for the final large scale metasurface design. . . . . 112
4.32 Diagram of an example metasurface antenna acting as a holographic

antenna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.33 A sample metamaterial element’s response, modeled with an analytic

polarizability based on a Lorentzian response described in Equation (3.2)116
4.34 Complex mapping of Lorentzian constrained modulation. . . . . . . . 119
4.35 Antenna weights without a feed layer. . . . . . . . . . . . . . . . . . . 119
4.36 Farfield plot of a broadside beam calculated from a metasurface an-

tenna without a feed layer. . . . . . . . . . . . . . . . . . . . . . . . . 120
4.37 Farfield plot of a steered beam calculated from a metasurface antenna

without a feed layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.38 Metasurface antenna design incorporating a waveguide feed layer. . . 124
4.39 Antenna weights for the metasurface antenna with a feed layer. . . . 125
4.40 Farfield plot of a broadside beam calculated from a metasurface an-

tenna with a feed layer. . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.41 Farfield plot of a steered beam calculated from a metasurface antenna

with a feed layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.42 Farfield patterns from a metasurface antenna tuned at 10.0 GHz, but
operating at other frequencies, with and without the waveguide feed
layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.43 Farfield patterns from a metasurface antenna tuned for different fre-
quencies, with and without the waveguide feed layer. . . . . . . . . . 129
4.44 Farfield pattern from the metasurface antenna for different tuning
strategies, both with and without a waveguide feed layer. . . . . . . . 130
5.1 Varactor tuned metamaterial element design. . . . . . . . . . . . . . . 133
5.2 Grayscale resonance tunable metamaterial element design in a full-

wave simulation environment. . . . . . . . . . . . . . . . . . . . . . . 134
5.3 Power radiated by the metamaterial element shown in Figure 5.2 as a

function of tuning from 0 - 5 V in full-wave simulation. . . . . . . . . 135
5.4 Metasurface antenna prototype design. . . . . . . . . . . . . . . . . . 139
xiii
5.5 Schematic drawing of each layer of the metasurface antenna prototype. 140
5.6 Feed transitions used to launch a wave into the SIW within a PCB
antenna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.7 Pictures of the metasurface antenna prototype. . . . . . . . . . . . . . 142
5.8 Picture of the metasurface antenna’s feed structure. . . . . . . . . . . 143
5.9 Characterization board for characterizing the individual metamaterial

response and waveguide characteristics experimentally. . . . . . . . . 144
5.10 Power radiated by a single metamaterial element for different tuning

states. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.11 Comparison between the simulated, measured, and corrected waveg-

uide propagation of the fabricated substrate integrated waveguide. . . 146
5.12 Experimentally characterized metamaterial element response. . . . . . 147
5.13 Tunability, in resonant frequency and phase control, of a fabricated

metamaterial element. . . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.14 Power radiated consistency among different instances of a fabricated

metamaterial element as a function of tuning. . . . . . . . . . . . . . 149
5.15 S parameter consistency among different instances of a fabricated

metamaterial element as a function of tuning. . . . . . . . . . . . . . 149
5.16 Broadside beam generated with the metasurface antenna prototype. . 151
5.17 An isometric view of the broadside beam generated with the metasur-
face antenna prototype, shown with a linear colormap. . . . . . . . . 151
5.18 Metasurface antenna prototype being measured at the Wireless Re-

search Center (WRC). . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.19 Comparison between simulated and experimental beamforming with

the metasurface antenna prototype. . . . . . . . . . . . . . . . . . . . 152
5.20 Azimuth steering capabilities of the metasurface antenna prototype. . 153
5.21 Elevation steering capabilities of the metasurface antenna prototype. 154
5.22 Steering accuracy of the metasurface antenna in azimuth and elevation.154
5.23 Steering the metasurface antenna prototype in azimuth and elevation. 155
xiv
5.24 Metasurface antenna prototype forming a broadside beam across a
narrow bandwidth of operation (9.75-10.25 GHz). . . . . . . . . . . . 156
5.25 Metasurface antenna prototype forming a broadside beam across a

wide bandwidth of operation (9-11 GHz). . . . . . . . . . . . . . . . . 156
5.26 Metasurface antenna generating two simultaneous beams in azimuth

(left) and elevation (right). . . . . . . . . . . . . . . . . . . . . . . . . 157
5.27 Experimental investigation of scaled Euclidean modulation with the

prototype metasurface antenna. . . . . . . . . . . . . . . . . . . . . . 158
xv
Acknowledgements
I would like thank my advisor, David Smith, for his support and guidance throughout
my studies. I enjoyed the collaborative nature of his lab and found it to be a pro-
ductive environment. Thanks also to Duke University, the Electrical and Computer
Engineering department, and the Graduate School for support and for opportunities
to travel. Thanks to Willie Padilla for advising me during my undergraduate studies
and for recommending Duke University’s doctoral program.
I would also like to acknowledge many of my colleagues. First and foremost,
thanks to Timothy Sleasman for his friendship, support, and knowledge. Thanks
also to Laura Pulido-Mancera for creating a positive office environment. Thanks
also to Jonah Gollub, Mohammadreza F. Imani, Thomas Fromenteze, Aaron Diebold,
Parker Trofatter, Daniel Marks, Claire Watts, Daniel Arnitz, Andreas Pedross-Engel,
and Matthew Reynolds. I have enjoyed collaborating with all of you.
Last, I would like to thank David Smith, Michael Gehm, Willie Padilla, Henry
Pfister, and Matthew Reynolds for serving on my committee. I have thoroughly
enjoyed discussing my research with you and I found our conversations helpful and
stimulating.
xvi
For my parents, Karen and Bruce Boyarsky
xvii
1
Introduction
Satellite imaging is an essential capability for many Earth observation applications.

Disaster management, agricultural maintenance, city/infrastructure planning, and
reconnaissance demand frequent coverage over a wide area [1–9]. As launch costs be-
come more economical, satellite systems with a focus on sensing and communication
are proliferating. Among satellite imaging systems, those which can use multiple
satellites cooperatively offer the most potential, including the ability to provide high
resolution 3D images with near-real-time revisit rates [10, 11].
Various satellite imaging systems have been deployed to address these needs
across the electromagnetic spectrum, including optical, infrared, and microwave sys-
tems [12–16]. While resolution is high, optical and infrared systems suffer from
limited availability due the need for clear skies and sunlight. Meanwhile, microwave
imaging systems can better penetrate inclement weather to offer added capabilities
(such as 3D imaging) and continuous availability. In such systems, a high perfor-
mance, electronically steered antenna is paramount to best performance. Existing
antenna hardware, however, is expensive, bulky, and power hungry — characteristics
which can be avoided with the use of metasurface antennas [17–20]. In this work,
1
Figure 1.1: Depiction of the TerraSAR-X satellite synthetic aperture radar system,
traveling along its synthetic aperture path [24].
metasurface antennas are developed as an alternative hardware for microwave satel-

lite imaging systems, with their favorable hardware characteristics able to render
multi-satellite systems practical.
For any imaging system, resolution can be determined by the size of the aperture
(relative to the wavelength) and the distance to the scene [3, 21–23]. In the optical
and infrared regimes, small wavelengths lead to aperture sizes which can fit within
practical device sizes (< 1 m) for imaging distances up to hundreds of kilometers.
However, at microwave frequencies, longer wavelengths would require aperture sizes
on the order of kilometers in order to realize high resolution from low earth orbit
(LEO), making it impossible to construct and deploy such large devices. To circum-
vent this challenge, synthesized apertures, created by moving a small antenna, have
been employed to realize large aperture sizes at microwave frequencies.
Synthetic aperture radar (SAR) systems move an antenna along a path to create
2
a large virtual aperture [3]. This operation enables a small antenna to represent
a large aperture with size equal to the distance traveled, not the physical antenna
size. Some SAR systems do so in order to realize a large aperture size which would
be impractical for a stationary system [25]. Other SAR systems use mechanical
motion to reduce the number of independent antennas in order to reduce cost (one
common example is airport scanners). Existing microwave satellite SAR systems
have proven to be capable of forming high quality images for many applications,
with one example system being TerraSAR-X, which was deployed by the German
Aerospace Center (DLR) to operate in LEO and is depicted in Figure 1.1 [18,25–28].
Three specific example images taken with TerraSAR-X are shown in Figure 1.2:
deforestation detection [29], 3D mapping [30], and oil spill detection [31]. In this
work, I will focus on spaceborne SAR systems, where a virtual aperture is used to
realize an impractically large aperture size in LEO.
When using a SAR system, there are two prominent modes of operation: stripmap
and spotlight. Stripmap mode involves radiating/receiving from an antenna with a
fixed radiation pattern. Spotlight mode involves continuously steering a beam to-
wards a fixed spot on the ground as the vehicle travels along the SAR path. “Fo-
cusing” the beam on a specific region in this way lengthens the effective aperture
size relative to the targeted region, increasing resolution. Stripmap mode can im-
age across a large scene size with lower resolution, while spotlight mode can realize
higher resolution, but for a smaller scene size. Both modes are illustrated in Figure
1.3. For both stripmap and spotlight modes, a highly directive antenna is required
for spaceborne operation in order to return sufficient signal strength from backscat-
tered measurements. For spotlight mode specifically, an antenna which can steer a
directive beam is an essential capability [3].
Traditional antenna hardware for satellite SAR systems takes the form of phased
array architectures or reflector dish systems [26,33]. These antenna types have facil-
3
Deforestation Detection 3D Mapping Oil Spill Detection
Before
After
Figure 1.2: Example capabilities for microwave satellite imaging. The left pane
shows deforestation which has been detected in the Amazonian rainforest with in-
terferometry; smooth areas show where trees have been systematically removed [29].
The center pane shows 3D imaging of a volcano in Chile; the false color corresponds
with height [30]. The right pane shows oil detection after the Deepwater Horizon
spill; the polarimetric signature is used to detect regions covered in oil (red) [31].
Images taken with the TerraSAR-X system [18, 26].
Stripmap Spotlight
Figure 1.3: Stripmap and spotlight synthetic aperture radar [32].
4
itated the deployment of single-satellite SAR systems, but their high cost and large
footprint have slowed the development of next generation spaceborne SAR systems
which will use many satellites collaboratively. Using numerous satellites enables
multiple input, multiple output (MIMO) operation, which can improve resolution,
revisit time, and scene size, while also offering added capabilities such as 3D imag-
ing, decluttering, and polarization determination. MIMO SAR systems have been
sought after for many years, but the unfavorable hardware characteristics associated
with phased array antennas, along with high launch costs, have severely limited their
deployment.
In this work, I propose metasurface antennas as an alternative hardware for
spaceborne SAR antennas. Metasurfaces offer favorable hardware characteristics rel-
evant to satellite antennas, including that they require little power, are low cost,
lightweight, and planar, and can be easily fabricated. Metasurface apertures have
demonstrated promise for many microwave applications from satellite communica-
tions to biomedical imaging [20, 34, 35]. Metasurface apertures can be easily fabri-
cated promise to be a lightweight, low cost, and planar hardware for shaping mi-
crowave radiation, with spatially diverse patterns and directive beams among the
potential radiation patterns. Recent demonstrations have shown high performance
from metasurface aperture computational imaging systems, which exhibit a compact,
streamlined architecture [36–38].
Metasurface apertures use a waveguide structure to excite a series of metamaterial
elements, each of which selectively leak energy from the waveguide into free space as
radiation [35, 39]. In this way, they act as holographic antennas, where the guided
wave acts as the reference, the metamaterial elements form the hologram, and the
interference between the reference and the hologram forms the radiation pattern. By
changing the hologram (metamaterial elements) or the reference wave (frequency),
the radiation of a metasurface aperture can be modulated [36].
5
Front Top
Side
Figure 1.4: Metasurface computational imaging system for security screening. This
system was developed for airport security and uses frequency diverse, passive meta-
surface apertures to illuminate a scene and collect measurements in MIMO fashion.
This system can generate human-sized 3D images at near video rates [38].
One system based on metasurface aperture technology described in [38] (and

shown in Figure 1.4), has shown the ability to realize video rate, human-scale imag-
ing for airport security screening. The metasurface apertures in this system built on
early metasurface aperture work [36] and are frequency modulated such that chang-
ing the input frequency changes their radiation pattern. By conducting a frequency
sweep, a series of spatially diverse radiation patterns are generated which can encode
scene information into a measurement set. Computational imaging algorithms can
then mathematically reconstruct the measurements to form an image. While these
apertures work well for computational imaging, they cannot generate arbitrary, pre-
scribed radiation patterns at will, limiting their range of applications.
Building on the framework of their passive counterparts, dynamic metasurface
apertures have been devised to enhance control over the radiation from a metasurface
aperture [40, 41]. Dynamic metasurface apertures use tunable components, embed-
ded within each element, to control radiation. The ability to change the radiation
pattern with tuning, rather than with frequency, reduced the required bandwidth
for computational imaging systems using this technology, even allowing for single
6
frequency operation. These systems showed many benefits associated with tunable
metasurface apertures, but were still limited in their ability to form arbitrary ra-
diation patterns. These limitations stem from challenges associated with modeling
early metasurface apertures and incomplete control over the radiating elements. If
these challenges could be addressed, their favorable hardware characteristics would
poise dynamic metasurface apertures to be an excellent candidate to replace existing
hardware for many microwave applications.
For spaceborne SAR applications, an antenna must be able to form a steerable,
highly directive beam which can be rapidly reconfigured. Existing dynamic metasur-
face apertures have focused on applications like computational imaging [37] and satel-
lite communication [34]. These systems have shown great promise in their respective
arenas, but would face challenges if applied directly to SAR. Metasurface antennas
for satellite communication can form arbitrary radiation patterns but suffer from slow
switching speeds and high loss. Metasurface antennas for computational imaging suf-
fer from incomplete element control and difficulty with mathematical modeling. One
method to address these limitations is to opt for a metasurface which relies on var-
actor diodes for element control; using semiconductors enables high switching speed,
while using varactors provides grayscale phase/magnitude control. A prototype of
such a metasurface antenna has been developed (see Figure 1.5) and is described in
detail later in this work [42, 43].
In any metasurface antenna design, the incomplete control afforded by metama-
terial elements can limit antenna performance. Incomplete element control is the
cost necessary for improving hardware characteristics, but can lead to strong grating
lobes if not accounted for during the design and implementation process. Existing
metasurface antenna designs have turned towards extremely dense element spacing
(< λ/4) along with high dielectrics to address this challenge, but this approach leads
to difficulty with mathematical modeling and high loss. In this work, I describe an
7
Figure 1.5: Metasurface antenna prototype developed for synthetic aperture radar.
The antenna is capable of electronically beamsteering in azimuth and elevation and
can be fabricated with printed circuit board technology.
alternative method where the antenna’s feed structure is used to suppress grating
lobes. With high dielectrics and dense element spacing not required, a wider range
of antenna layouts are unlocked, creating a path towards Nyquist sampled, efficient
metasurface antennas [44].
Many metasurface apertures are difficult to model in full wave simulation due to
their electrically large size and subwavelength variations. Instead, a numerical model
is often used which approximates each metamaterial element as a dipole [39, 45–47].
This model, however, assumes that inter-element coupling is negligible, which can
be a poor assumption in the case of densely sampled metasurface apertures. Since
grating lobe suppression is addressed with the feed structure, element spacing close
to λ/2 is possible, increasing the fidelity of this mathematical model. Also, element
placement is offset from the center of the waveguide and alternates sides, providing
additional distance between adjacent elements to further reduce coupling to the
guided wave and decreasing inter-element coupling [43].
A third challenge that faces metasurface antennas is determining the appropriate
8
tuning strategy. In a phased array, there is explicit, independent control over the
phase and magnitude of each radiating element. To form a beam in a given direction,
a prescribed phase profile is applied exactly, while amplifiers apply a desired magni-
tude taper (often for sidelobe suppression) [48]. Since tuning metamaterial elements
affects both their magnitude and phase response simultaneously, phased array tun-
ing methods can cause unintended consequences which limit performance [49]. In
this work, I also describe a metasurface-specific tuning strategy, called scaled Eu-
clidean modulation, which jointly considers the magnitude and phase response of
the constituent metamaterial elements. Scaled Euclidean modulation also provides
a means for controlling the decay of the guided wave, which can balance efficiency
and effective aperture size to optimize antenna performance [50].
In summary, the antenna prototype presented in this work addresses three main
challenges associated with metasurface antenna design: mathematical modeling,
grating lobe suppression, and tuning strategy determination. These three challenges
are addressed concurrently, with the grating lobe suppression method unlocking a
wide range of layout possibilities, the antenna layout facilitating a straightforward
mathematical model, and the tuning strategy optimizing antenna controlling the de-
cay for the specific layout at hand. The prototype metasurface antenna provides an
experimental demonstration of the solutions proposed to address these challenges.
Independently verified experimental results have shown that this antenna can form
steerable, directive beams and has exceeded all targeted antenna performance met-
rics. In azimuth, the antenna can steer continuously through ±50◦ . In elevation,
the antenna can cover ±70◦ . The antenna can operate across a frequency range of
9.25 - 10.75 GHz. Further, though not targeted during design, the antenna exhibits
reasonable performance in terms of efficiency and polarization isolation. In total, the
demonstrated metasurface antenna prototype validates the proposal of using meta-
surface antennas for SAR applications. The design approach used in this work could
9
readily be modified to accommodate many of the strict requirements for spaceborne
SAR antennas while maintaining excellent hardware characteristics.
The remainder of this document is organized as follows. First, synthetic aperture
radar is introduced, with traditional modalities and hardware discussed. Second,
metasurface apertures are described, with a discussion of their history, operation,
and use in computational imaging systems. Experimental demonstrations of tradi-
tional and alternative SAR modalities with metasurface apertures are also included.
Next, several simulated metasurface antennas are discussed, with a focus on their
application to SAR, comparison with traditional antenna hardware, and grating lobe
suppression. Finally, the metasurface antenna prototype is described in detail, with
its goals, design, operation, and results presented.
Contribution
Chapter 2 is primarily background on synthetic aperture radar, available in pub-

lished literature. I conducted all simulations to reproduce beamforms generated by
TerraSAR-X as described in Section 2.2. Sections 3.1 and 3.2 provide a background
of published literature on metasurface apertures for computational imaging, some
of which I co-authored. The dynamic metasurface aperture used for computational
imaging (in Section 3.3) has been published across various works. I contributed to-
wards its development and experiments, which were chiefly led by Timothy Sleasman.
I led the single frequency SAR imaging work, including the mathematical descrip-
tion, reconstruction algorithm derivation, and experimental measurements discussed
in Section 3.3.6. I conducted all simulations and led all mathematical derivations
presented in Chapter 4. I led the simulation, design, and experimental measurements
associated with the metasurface antenna prototype outlined in Chapter 5.
10
2
Synthetic aperture radar (SAR) is an imaging technique which uses mechanical mo-
tion to increase the resolution of an imaging system. By translating an antenna, it
is often possible to realize an aperture size which is much larger than the physical
antenna, often reaching sizes that would be impossible with a stationary antenna.
SAR has been practiced since the 1950’s, with early systems being implemented with
aircraft-mounted systems for reconnaissance [51–54]. The most popular modern sys-
tems are the personnel security screening devices at airports. In environments like
Earth orbit, SAR is particularly useful due to the fact that stationary apertures
would otherwise need to be impractically large in order to achieve reasonable resolu-
tion when operating at microwave frequencies [11,55]. In this chapter, I will describe
traditional SAR operation, including stripmap and spotlight modalities. Then I will
briefly describe existing spaceborne SAR hardware, TerraSAR-X, to facilitate later
comparison with metasurface antenna performance.
11
2.1 Traditional SAR Modalities
2.1.1 Stripmap SAR
Initial airborne SAR imaging systems conducted stripmap mode SAR. Stripmap
mode uses an antenna fixed to a vehicle, as shown in Figure 2.1 [3]. As the vehicle
moves, the antenna emits a fixed radiation pattern. In this setting, 2D images of a
scene on the ground can be reconstructed, with the cross range information extracted
from the motion and the range information extracted from the projection of the
frequency sweep onto the scene [3]. For many radar and microwave imaging scenarios,
including stripmap and spotlight SAR, range resolution is set by the bandwidth of
operation (B) and the speed of light (c) [21].
c
δrange = (2.1)
2B
Stripmap was a popular initial modality due to its ease of implementation and
since measurements taken this way are compatible with analog processing methods.
In addition, since the antenna is fixed relative to the vehicle during stripmap op-
eration, the scene size is not theoretically limited in the cross range direction. In
practice, however, transmitting/recording measurements for a long period of time
in spaceborne environments can overheat an antenna, depending on the antenna’s
efficiency. Cross range resolution, for stripmap SAR, is defined as a function of the
minimum wavelength (λmin ) and the beamwidth (φ) of the antenna [3].
λmin
δcross range = (2.2)
4 sin φ2
From Equation (2.2) this equation, it can be seen that in order to increase resolution,
the beamwidth must grow. However, a large beamwidth incurs substantial noise,
leading most practical systems to require a narrow beamwidth in order to achieve an
12
Stripmap Spotlight
𝐿
𝜃𝑜
ϕ
𝑑
Figure 2.1: Illustration of conventional stripmap and spotlight SAR modalities [32].
adequate signal-to-noise ratio (SNR) [55]. It should also be noted that in order for
Equation (2.2) to be valid with respect to a given spot on the ground, the antenna’s
beam pattern must fully move from illuminating the ground spot with one edge of the
beam pattern, through the center, and to the other edge of the beam pattern, where
the edges of the beam pattern are defined as the half-power beamwidth (HPBW).
2.1.2 Spotlight SAR
Another popular SAR modality is spotlight mode. In this modality, the antenna is
continuously steered towards a constant spot on the ground, as shown in Figure 2.1.
By continually steering, a strong signal can be measured from the targeted region
throughout a larger portion of the synthetic aperture path. Spotlight mode imaging
follows a cross range resolution defined as a function of the distance from the synthetic
aperture path to the scene (d), the minimum steered angle (θ0 ), and the length of
the synthetic aperture path (L) [3].
dλmin
δcross range = (2.3)
2L cos θ0
The steering approach thus circumvents the tradeoff between SNR and resolution
13
a) b) c)
Figure 2.2: Examples of traditional spaceborne SAR systems. a) shows NASA’s

soil moisture active passive (SMAP) system, which uses a reflector dish [61]. b) and
c) show NASA’s SIR-C/X-SAR system, which uses a phased array antenna [56]. One
can appreciate the size of the antenna by comparing to the space shuttle’s cargo bay.
experienced when using stripmap mode. Owing to this operation, resolution becomes
a function of the steering limits of the antenna, not the beamwidth, providing a
method to increase resolution from levels possible with strimpap. However, since the
antenna must be continually steered towards the target ground spot, the scene size
is limited to the antenna’s projected spot size on the ground.
Implementing stripmap and spotlight SAR from a spaceborne platform requires
a high-powered radio frequency (RF) source in order to achieve an adequate SNR.
Additionally, these systems must be broadband in order to achieve sufficient range
resolution which can be on par with the cross range resolution. In order to realize
higher resolution, there has been a recent shift towards higher frequencies, including
X band (8-12 GHz) [28, 56], K band (18-27 GHz) [57], W band (75-110 GHz) [58],
or higher [59]. These higher frequencies must be balanced with atmospheric trans-
mission, however, leading to a tradeoff between, power, scene size, resolution, and
computational demands [60].
14
2.2 Traditional SAR Hardware
Traditional hardware for spaceborne SAR operation has taken the form of phased
array antennas and reflector dish systems, as shown in Figure 2.2 [26, 33]. The
antenna must be able to form a directive beam across the operational bandwidth
and be compatible with high power, chirped pulse RF sources. While reflector dish
systems can be lightweight and low-cost, they require complex deployment and either
need a mechanical gimbal or a small phased array feed antenna in order to steer a
beam for spotlight mode imaging. As a result, phased array antennas have become
a more popular solution for high performance systems. Examples of both systems
are shown in Figure 2.2.
2.2.1 TerraSAR-X Phased Array
Many spaceborne SAR platforms have been deployed over the past few decades.
In this section, I describe TerraSAR-X, a high resolution spaceborne SAR platform
launched in 2007. It was launched by the German aerospace center (DLR) for gov-
ernment, commercial, and public use. TerraSAR-X is a useful system for reference
due to the extensive, unclassified information available about its construction and
performance. TerraSAR-X is described in Figure 2.3 and 2.4, which also includes a
sample SAR image of Antarctica. Throughout this document, the performance and
capabilities of TerraSAR-X are used for comparison to various metasurface architec-
tures.
Figure 2.4 shows TerraSAR-X, while Table 2.1 highlights its various metrics.
TerraSAR-X operates at X band, with a center frequency of 9.65 GHz and a de-
fault bandwidth of 150 MHz (expandable up to 300 MHz). X band was chosen due
to its high resolution coupled with operating within a transmission band to avoid
atmospheric attenuation. TerraSAR-X operates from an altitude of 515 km in sun-
15
Figure 2.3: Schematic of the layout of TerraSAR-X; there are 32 radiating slotted
waveguide antennas in the elevation (vertical) direction. The key shown here is used
throughout the document.
synchronous orbit, allowing access to almost any terrain within 2.5 days. It is capable
of operating in various SAR modes, including stripmap, spotlight, and ScanSAR – a
method which uses continual elevation steering to extend scene size [27].
TerraSAR-X utilizes an active phased array comprising center-fed slotted waveg-
uide antennas (SWAs), each fitted with two phased shifters, an amplifier, and a 6 W
independent RF source. The SWAs are 2.2 cm wide and 40 cm long and are oriented
with their longer dimension in azimuth. 32 SWAs are stacked together to form a
panel and the overall antenna comprises 12 such panels, as outlined in Figure 2.3.
The antenna is 4.8 m (azimuth) by 0.7 m (elevation). This array architecture leads
to 12 degrees of freedom (DOF) in azimuth, sampled at 13λ, covering an azimuth
steering range of ±0.75◦ . In elevation, there are 32 DOF, sampled at 32 λ, covering an
elevation steering range of ±20◦ . If steered past either of these limits, grating lobes
appear which quickly rival the main lobe in prominence.
TerraSAR-X, like many phased array antennas, faced a tradeoff in design between
cost and performance. In a phased array antenna, a significant portion of the cost
comes from the number of RF sources/phase shifters. For TerraSAR-X, the antenna
16
a) b)
c) d)
Figure 2.4: TerraSAR-X modes of operation (a) [27], phased array antenna (b) [62],
full satellite [63], and sample SAR images, showing the Larsen ice shelf breaking off
from Antarctica [64].
cost was reduced by using long SWAs, which span 40 cm. If the SWAs had been made
smaller, more RF sources/phase shifters would have been required, driving the cost
up. With more control points, however, the finer sampling in azimuth would lead
to a larger steering range, increasing performance. This tradeoff between number of
phase shifters and steering limits is fundamental to phased array architectures and
requires a compromise between price and performance.
One of the foremost claims made throughout this document is that metasurface
antennas are a suitable replacement for phased array antennas in spaceborne SAR
contexts. To validate this claim, TerraSAR-X has been simulated for comparison
with metasurface designs. Array factor calculations were used to reproduce the
beamforms exhibits by TerraSAR-X, following Equation (2.4) [67].
17
Table 2.1: TerraSAR-X antenna parameters [25, 27, 63, 65, 66].
Launch date June 15, 2007

Antenna type Active phased array
Radiator type Slotted waveguide
Antenna mass 400 kg
Antenna size 4.8 m (azimuth) x 0.7 m (elevation)
Number of waveguides 12 (azimuth) x 32 (elevation)
Azimuth steering limits ±0.75◦
Elevation steering limits ±20.0◦
Gain 46 dB
Center frequency 9.65 GHz
Bandwidth 150 MHz/300 MHz
Spotlight resolution 1m
Spotlight scene size 10 km x 5 km
Stripmap resolution 3m
Stripmap scene size 30 km x 50 km
DC control power 800 W
Orbit period 94.76 minutes
Altitude 515 km
Revisit time 2.5 days
M X
X N
AF (θ, φ) = amn ejk(xm sin θ cos φ+yn sin θ sin φ) (2.4)
m n
Here, AF is the array factor calculation, amn is the antenna weight, k is the free space
wavenumber, xm and yn are the position of each element, θ is the azimuth direction,
and φ is the elevation direction. To use Equation (2.4) for TerraSAR-X, the guided
wave emanating from the center of each SWA is calculated (following H = H0 e−jβx )
at the location of each slot. A phase shift (contained within H0 ) is applied to each
SWA in order to beamform, according to the desired steered angle. For the phase
shift applied to each H0 , there is no limit to the phase value — full 2π control is
available. From these two calculations (the phase shift and guided wave sampling),
a set of amn is created which represents the entire antenna.
It should also be noted that in practice, TerraSAR-X employs a Hamming win-
18
Azimuth Steering
-0.75o 0o 0.75o
9.50 GHz
9.65 GHz
9.80 GHz
Figure 2.5: Azimuth steering with TerraSAR-X. Steering is limited to ±0.75◦ ;

steering beyond this limit leads to strong grating lobes.
dow (α = 0.6) to suppress sidelobes, depending on the environment [68]. For the
analyses presented here, window functions have not been applied such that a better
comparison between antenna architectures can be conducted. Figure 2.5 shows the
antenna steering to its limits in azimuth; Figure 2.6 show the antenna steering to its
limits in elevation. The performance and capabilities of TerraSAR-X, as described
in this chapter, are used as target metrics for simulated metasurfaces presented later
in this document.
19
Elevation Steering
-20o 0o 20o
9.50 GHz
9.65 GHz
9.80 GHz
Figure 2.6: Elevation steering with TerraSAR-X. Steering is limited to ±20◦ ; steer-
ing beyond this limit leads to strong grating lobes.
20
3
Metasurface Apertures
Metasurface apertures have been demonstrated for a variety of applications from

computational imaging [20, 37, 38, 69, 70] to satellite communications [34]. Their
recent development has been motivated by their ability to shape radiation from
a platform with favorable hardware characteristics. A variety of architectures have
been demonstrated, including mode-mixing cavities [71], multiplexing media [72–74],
spatial light modulators [75], and waveguide-fed metasurface apertures [36–38]. Fur-
ther, metasurface architectures can often leverage existing manufacturing technolo-
gies such as printed circuit board (PCB) or liquid crystal display (LCD), promising
easy fabrication with low long-term cost.
In this chapter, various metasurface architectures are presented, including a brief
description of the development of waveguide-fed metasurface apertures, which are
the focus of this document. The operation of passive and dynamic metasurfaces as
holographic antennas is discussed, before moving into the numerical method used to
model a dynamic metasurface aperture. A mathematical framework for computa-
tional imaging (CI) is provided, followed by experimental CI results with a dynamic
metasurface aperture. Last, other imaging modes are demonstrated, including sin-
21
gle frequency imaging, traditional and alternative SAR modes, and single frequency
SAR. The different imaging modes possible with one dynamic metasurface aperture
highlight the dexterity associated with metasurface antenna technology.
3.1 Metasurface Architectures
Metasurface apertures were developed over the past decade as devices for shaping
electromagnetic (EM) radiation, with early systems frequently targeting computa-
tional imaging as an application [20,36]. Metasurface apertures are electrically-large
antennas composed of a waveguide loaded with subwavelength metamaterial ele-
ments [20, 36, 76]. As the guided wave passes the metamaterial elements, they leak
energy into free space as radiation [69].
3.1.1 Passive Metasurface Apertures
Metasurface apertures can be classified as either passive metasurface apertures or

dynamic metasurface apertures. Passive apertures can generate a collection of radi-
ated waveforms as a function of frequency, while dynamic apertures do this by tuning
the radiating elements. Metasurface apertures have also been demonstrated through
various portions of the EM spectrum [38, 75]; in this document I focus on microwave
metasurface apertures. One of the first microwave metasurface apertures took the
form of a microstrip transmission line with radiating metamaterials embedded within
the upper conductor [36]. In this aperture, each element’s resonant frequency was
randomly distributed within the operational bandwidth, such that as the incident
frequency changed, a new collection of elements were excited. This operation results
in the radiated waveform varying rapidly with frequency, leading to the ability to
generate a collection of diverse radiation patterns to probe a scene. After measuring
the signal backscattered from each of these patterns, computational techniques were
able to recover scene information (see Section 3.2 for more detail of the mathematical
22
f=18.5 GHz
f=21.8 GHz
Figure 3.1: Initial passive metasurface aperture developed for microwave compu-
tational imaging. The left shows an illustration of the aperture; the right shows the
radiation patterns for two different frequencies [36].
post-processing).
In an alternative approach, mode-mixing cavities can leverage the variation among
cavity modes, rather than among the resonant frequencies of each metamaterial ele-
ment [71]. In this case, the radiating elements can be circular/slotted irises [70]. As
the input frequency changes, the cavity mode correspondingly changes, with irregular
shapes/features able to amplify this effect [38, 40]. This design approach can enable
a greater set of accessible, distinct radiation patterns [70, 77]. The ideas developed
in these works were recently extended to small PCB planar cavity designs which
enabled a massive MIMO system to demonstrate human-sized 3D security screening
imaging at video rates (see Figure 1.4) [38].
Passive metasurface apertures are a great solution when a specific radiation pat-
tern is not required or when control circuitry/control is limited. Owing to the fact
that passive metasurface apertures can readily generate sets of spatially diverse ra-
diation patterns, they are compatible with computational imaging applications, but
other applications may require more from the radiated patterns. For applications
like communications and SAR, where a specific beam pattern is desired, the limited
23
control afforded by passive metasurface apertures may be insufficient, which has led
to the deveopment of dynamic metasurface apertures.
3.1.2 Dynamic Metasurface Apertures
In addition to passive metasurface antennas, dynamic metasurface antennas have

emerged as an alternative approach for generating different radiated waveforms. In
this style, individual elements can be electronically tuned, by semiconductors [78,79]
or liquid crystal [80,81], to adjust the aperture’s radiation pattern. Tunability allows
different radiation patterns to be generated by applying a different tuning state to the
aperture. While added control circuitry is required to operate dynamic metasurface
apertures, their benefits include a reduced bandwidth requirement and the ability
to generate a larger collection of distinct radiation patterns. Dynamic metasurface
apertures provide more deliberate control over the radiated fields, which leads to
an improved ability to tailor radiation patterns to specific applications [82]. More
specifically, by leveraging the individual control over the radiating elements, dynamic
metasurface apertures can deliberately craft arbitrary waveforms, including directive
beams necessary for SAR [81]. The remainder of this document will exclusively focus
on and investigate the operation and performance of dynamic metasurface apertures
for SAR systems.
Towards the development of dynamic metasurface apertures, an individually-
addressable radiating metamaterial element was designed in [83] and shown in Fig-
ure 3.2. This radiator followed a complementary electric inductive-capacitive (cELC)
geometry [84–87], in which the element could be switched on or off by a PIN diode
which connected the radiator to the surrounding microstrip [83]. The elements could
be individually addressed by means of a via which connected each cELC element to
a tuning circuitry layer below the ground plane. An initial implementations of a dy-
namic metasurface aperture involved creating an array of these elements [79]. From
24
Figure 3.2: Individually-tunable metamaterial element for creating a dynamic
metasurface aperture. Biasing PIN diodes connecting an element to the surrounding
microstrip allows elements to be tuned from a radiating to a non-radiating state [83].
measurements taken with a CI system featuring this dynamic metasurface aperture,

similar to in the passive metasurface aperture case, computational post-processing
techniques were able to reconstruct high-fidelity 2D images [37].
When considering the design of a dynamic metasurface aperture, there are two
styles for tuning the individual elements of a metasurface: binary and resonance [47].
Binary tuning enables elements to be tuned from a radiating (on) state to a non-
radiating (off ) state, for example with PIN diodes [83]. Resonance tuning allows
for a shift in the resonance to be introduced, which can provide finer, grayscale
control of the magnitude and phase of each radiating element, for example with
liquid crystal [80,88] or varactor diodes [89,90]. In the case of a metasurface antenna
applied to SAR imaging, the added control granted by grayscale resonance tuning is
used to generate steerable beams for stripmap and spotlight SAR, as will be described
in further detail in Chapter 4 [91, 92].
25
Figure 3.3: Initial dynamic metasurface aperture developed for microwave compu-
tational imaging. The left shows an illustration of the aperture in an imaging envi-
ronment; the right shows the radiation patterns for three different tuning states [41].
Figure 3.4: Illustration of a holographic metasurface antenna. Here, the incident

field, H, interferes with the radiating elements embedded within the upper conductor
of the waveguide to form the radiation pattern of the aperture [35].
3.1.3 Holographic Antennas
To better grasp the operation of a metasurface aperture, it is helpful to think of them

as holographic antennas. In this paradigm, the guided wave acts as the reference,
the metamaterial elements form the hologram, and the interference between them
forms the radiation pattern [69, 93, 94]. This process is shown visually in Figure 3.4
[35,95]. To change the radiation pattern, the guided wave can be altered (by changing
frequency) or the hologram can be modulated (by changing/tuning the elements).
26
Feed Wave/Magnetic Field Tuning State
Effective Radiators Radiated Power
Figure 3.5: A sample metasurface antenna. The feed wave traverses the waveguide,
exciting each individually tuned metamaterial in the array. The excited metamate-
rial elements radiate, with steerable, directive beams among the available radiation
patterns.
To produce a desired radiation pattern from a holographic metasurface aperture,

the desired farfield radiation pattern is projected onto the aperture plane to form
a set of desired effective sources (or magnetic dipole moments). By interfering the
set of desired dipole moments with the guided wave, the response required from the
individual elements can be calculated. Tuning the elements to match the desired
response will then allow the metasurface aperture to generate the guided wave when
the incident (guided) wave is applied [96–101]. This process follows backwards from
the antenna shown in Figure 3.5, which shows a metasurface antenna generating a
steered beam.
As the incident field travels through the waveguide, it becomes attenuated by a
27
combination of radiation losses, dielectric losses, and Ohmic losses [47]. This process
results in elements far from the feed receiving less energy than those close to the feed.
For some applications, a gradual tapering of the guided wave is a desired behavior.
However, other applications may desire either a particular decay rate or a more
uniform illumination in the aperture. With a more uniform illumination, the elements
can more readily radiate with a similar magnitude, leading to the possibility of a more
uniformly radiating aperture. To more deliberately engineer the guided wave’s decay
different approaches may be used. To reduce losses, the element geometry can be
designed to minimize Ohmic losses (for example with wider metal trace widths).
Additionally, one may opt for a low-loss dielectric substrate, though this may incur
added cost.
Many antenna applications desire a uniform illumination, in spite of the various
loss mechanisms associated with metasurface apertures. To achieve a more uniform
aperture illumination, the radiating element characteristics and usage must be specif-
ically engineered. The largest contributing factor to the attenuation is the radiation
losses, so using elements which do not couple strongly to the guided wave can reduce
attenuation. Minimizing this coupling strength can be achieved in a few different
ways, including modifying the element geometry, offsetting the elements within the
waveguide, or adjusting elements’ tuning states. To achieve the most uniform radia-
tion, the elements close to the feed must couple weakly to the guided wave, so as to
retain sufficient energy to reach elements far from the feed, and the elements far from
the feed must couple stronger to the guided wave, to ensure that they radiate with
sufficient magnitude. While metasurface apertures are lossy structures, considering
the guided wave’s attenuation in design can help to realize a desired excitation of a
metasurface aperture when applying holographic beamforming techniques.
28
3.1.4 Numerical Modeling of Metasurface Apertures
Metasurface apertures are electrically large devices with subwavelength variations,

rendering them computationally difficult to simulate with full-wave solvers. As a
result, approximate numerical models must be used in order to predict their re-
sponse/performance. In this document, a numerical approximation which treats
each metamaterial element in a metasurface as a frequency-dependent, infinitesi-
mal dipole is used in order to facilitate a rapid estimation of a given metasurface
aperture’s performance. Similar to [39], the dipole moment associated with each
metamaterial element can be characterized by its polarizability, α, which is the com-
plex ratio between the induced magnetic dipole moment, η, and the incident field,
H. From H and α, the dipole moment for a given element can be calculated [102].
η = Hα. (3.1)
Metamaterial elements are resonant structures and their polarizability is a function

of geometry and material parameters. The polarizability of a metamaterial element
follows a Lorentzian response as a function of the incident frequency (ω) and the
resonance frequency (ω0 ).
F ω2
α= . (3.2)
ω02 − ω 2 + jΓω
Here, F is the coupling factor and Γ is the damping term [103]. In practice, many
of these parameters are difficult to estimate analytically. As a result, full-wave sim-
ulations of single metamaterial elements may be used to determine α for a given
metamaterial geometry. More detail on this process is provided in Section 4.1.
Once a model of α has been determined, the magnetic field, H, incident on each
unit cell must be modeled. The magnetic field of the traveling wave at a distance
29
r from the feeding point can be written as e−jβr , where the propagation constant
within the waveguide, β = βreal + iβimag , is a function of frequency, the material
properties, and the waveguide geometry [102]. As the guided wave passes each radi-
ating element, a portion of the energy is coupled to free space from the waveguide as
radiation, a smaller fraction is lost in Ohmic and dielectric losses, and the remainder
is transferred to subsequent elements. These three different fractions can be calcu-
lated using the scattering (S) parameters obtained from the single-element full-wave
simulation. Specifically, the wave is attenuated by S21,n as it passes through the nth
element, while dielectric losses are applied as a function of distance from the feed
location. It should be noted that while some reflections may occur due to impedance
mismatch, this effect is insignificant in the context of this work due to a low |S11 |
over the operational bandwidth and is therefore neglected in the present formulation.
With all of these factors taken into consideration, a final formulation for the exciting
field H, at element m, for a given incident field, H0 , can be written as
Y
Hm = H0 e−jβrm S21,n . (3.3)
n<m
This calculation of the magnetic field is then used to simulate a guided wave intro-
duced from any feed location in a metasurface aperture. If multiple feed locations
are introduced, the linearity of the system can accommodate the superposition of
the various feed waves to calculate the overall field incident upon each element.
Once the magnetic field incident on each element is calculated, the magnetic
dipole corresponding to each element is obtained using Equation (3.1). Thus, the
metasurface aperture is reduced to an array of magnetic dipoles with known phases
and magnitudes. The radiation of the dipole array at any location in the scene
can be calculated using conventional Green’s functions along with the superposition
principle
30
X e−ik|r−rm |
E(r) = A0 ηm , (3.4)
m
|r − rm |
where r is the position in the scene, rm is the position of the mth resonator, k is
the free space wavenumber, and A0 is a proportionality constant with units Ω/m2
[39]. Either the transmitted (Et ) or received fields (Er ) from a metasurface can
be calculated this way. The overall modeling method is summarized visually in
Figure 3.6 [47].
It should be noted that the framework for modeling metasurface apertures pre-
sented in this section applies equally well to passive and dynamic metasurface aper-
tures. Since α and H are both a function of frequency, all essential phenomena
associated with passive metasurfaces are naturally captured. Meanwhile, α can be
modeled with an additional input for tunability. Either by incorporating tunability
analytically or through simulation, a library of available αtunable can be created, for
modeling dynamic metasurface apertures. Such a library of αtunable can represent
different types of tunable elements, including binary and resonance tunability.
3.2 Computational Imaging Framework
Many of the first demonstrations of metasurface apertures were developed for sta-
tionary computational imaging [36, 37]. In order to describe specific details of how
metasurface apertures were used in this way, a description of the framework of com-
putational imaging is provided [104]. By illustrating the principles of computational
imaging, the capabilities of metasurfaces and the ways in which a specific architecture
affects performance can be better highlighted.
31
Approximate dipole response from S parameters
Calculate the overall

Full wave simulation radiation pattern of the
|α|
of unit cells Calculate the equivalent dipole array
dipole for each element
Frequency
|S2,1|
Frequency
Model the guided wave with S parameters
Figure 3.6: Numerical modeling procedure for 1D metasurface apertures. This

method extends the full-wave simulation of one unit cell to model an entire aperture
and predict the radiated fields [47].
3.2.1 Mathematical Description
In many computational imaging settings, spatially-uncorrelated radiation patterns

are generated to illuminate a scene and multiplex its spatial content into a collec-
tion of backscattered measurements. A receiving antenna collects the backscattered
signals which are then post-processed with computational techniques to reconstruct
images of the scene’s reflectivity [36, 37]. Mathematically, the imaging process for
such a system can be represented by the matrix equation [105]
g = Hσ, (3.5)
where g is the signal collected by the receiving aperture, H is the measurement

matrix, and σ represents the scene’s reflectivity. By assuming the first Born approx-
imation (weak scattering) [106], the entries of the measurement matrix are given
by
32
−
→− − →−
Hij = Eri (→
rj ) · Eti (→
rj ). (3.6)
−
→ −
→
Here, Eti and Eri are the fields generated by the transmitting and receiving apertures,
respectively, at each point in the scene. More explicitly, the ith row of H (referred
to as a measurement mode) corresponds to one measurement of the scene locations,
→
−
rj , at a given frequency and tuning state of the aperture [37, 104].
To reconstruct a collection of measurements, the inverse problem corresponding
with Equation (3.5) must be solved to retrieve an estimate of the scene’s reflectivity,
σest , from a given H and g. Various algorithms have been demonstrated to solve this
problem, including matched filter (carried out as σest = H† g, where †
is the conju-
gate transpose operator) [107], GMRES (an iterative, least-squares solver [108]), and
TwIST [109]. Since the mathematical framework presented in this section is agnos-
tic to the imaging system at hand, it can be applied to a variety of computational
imaging systems, including SAR systems, 3D imaging devices, and single frequency
imaging systems.
3.2.2 Measurement Matrix Modeling
Predicting the performance of any computational imaging system requires knowl-

edge of the measurement matrix, H. For many well-behaved antennas, an analytic
calculation may be used to determine H by propagating the fields from the transmit
and receive apertures into a defined scene. Other antennas can use full wave simu-
lation to determine H. As mentioned in Section 3.1.4, metasurface apertures often
require an approximate model to determine their radiated fields (and thereby H).
Using this approximate model provides a means of predicting the performance of a
CI system which uses metasurface hardware without requiring a fabricated sample.
This process can rapidly generate performance estimates for a variety of metasurface
designs.
33
While performance predictions for metasurface CI systems can be obtained us-
ing the approximate model outlined in Section 3.1.4, approximate models of H are
often insufficient to reconstruct experimental measurements. Inaccuracies can be in-
troduced from fabrication tolerances, semiconductor component inconsistencies, and
mutual coupling among elements. While these effects do not significantly change
the performance of a metasurface CI system, these discrepancies render experimen-
tal reconstructions impossible for the metasurfaces discussed in the next section. In
this scenario, experimental characterization can be used to build H in place of an-
alytic/simulated methods. To experimentally characterize an antenna, a near field
scan (NFS) may be used. Near field scanning uses a reference antenna (typically an
open ended waveguide or a horn antenna) to measure the radiated fields directly in
front of the antenna under test (AUT). Measuring the radiated fields of the antenna
this way forms a set of effective sources which may then be propagated further into
the near field or into the farfield in order to populate H. This experimental method
for characterizing an antenna thus includes all fabrication defects and other features
which may be difficult to capture analytically.
3.3 Experimental Imaging with a Dynamic Metasurface
A dynamic metasurface aperture was developed for computational imaging and is

used in the remainder of this chapter for various experimental demonstrations. First,
stationary computational imaging will be described and demonstrated, followed by
single frequency computational imaging. Next, stripmap, spotlight, and diverse pat-
tern stripmap SAR imaging will be presented. Last, the principles, mathematical
derivation, and experimental results for a single frequency SAR system will be de-
scribed.
34
3.3.1 1D Dynamic Metasurface Aperture
Throughout this chapter, the experimental work described has been conducted with
a dynamic metasurface aperture detailed in [37] and briefly described here. This
aperture consists of 112 cELC elements embedded within a microstrip. The ele-
ments are spaced 3.33 mm apart (∼ λ0 /4), spanning a total size of 38 cm. There
are two alternating element geometries; the first was designed to have a resonant
frequency of 19.0 GHz and the second was designed to have a resonant frequency
of 20.5 GHz. Using elements with different resonant frequencies extends the band-
width of operation. The element characteristics are plotted in Figure 3.7. Note that
fabrication tolerances, variation among material parameters, and element-element
interactions have led to a downward shift in the resonant frequencies realized by the
elements in the fabricated samples.
This dynamic metasurface may be fed from any of three feeds: two end-launch
feeds are available at each end of the aperture and a coaxial feed is in the center
(see Figure 3.8). This feed architecture leads to four propagating waves within the
aperture (each with equal power when the appropriate power splitter configuration is
used), as shown in Figure 3.9. Using multiple feed locations simultaneously ensures
that the full extent of the aperture radiates [37, 47]. This aperture was designed
to generate spatially distinct radiation patterns by applying different tuning states
to the overall aperture, termed masks. This aperture uses elements which couple
somewhat strongly to the guided wave, causing the guided wave to decay quickly
when many elements are turned on at once. To account for the decay of the guided
wave and ensure a more uniform aperture illumination, the most frequently used
masks each had 28 elements on at a time (14 on either side of the center feed). In
this manner, each of the four feed waves only needs to excite approximately seven
elements. A schematic view of different radiation patterns is shown in Figure 3.9.
35
PIN diode
Figure 3.7: The left side shows an illustration of the unit cell geometry comprising
a dynamic metasurface aperture; the top right shows the simulated/designed element
response for the two alternate geometries. The bottom right shows a picture of the
fabricated sample [37].
Coaxial feed Metamaterial radiators
End launch feeds
Figure 3.8: Diagram of the metasurface aperture. The aperture has three feeds,
which generate a total of four propagating waveguides within the microstrip, shown
with the blue arrows.
3.3.2 Stationary Imaging
Stationary computational imaging was conducting using the dynamic metasurface

aperture as the transmitter and three open-ended rectangular waveguide probes as
the receivers (see Figure 3.9). These low-gain receive antennas were spaced across
30 cm and located approximately 5 cm above the dynamic metasurface. In total,
measurement sets taken in this imaging configuration result in a number of measure-
ments equal to the multiplication of the number of probes, the number of frequency
points, and the number of masks.
36
Figure 3.9: The left side shows a schematic view of the dynamic metasurface
aperture’s operation. As a new electronic tuning state is applied, a new pattern is
radiated. The right side shows the experimental setup used to conduct computational
imaging experiments, with the metasurface transmitting, while three low gain probes
receive [37].
Figure 3.10: PSF analysis for stationary computational imaging with a dynamic
metasurface aperture. a), b), and c) show reconstructions with Matched Filter,
GMRES, and TwIST, respectively [37].
The first experimental results using the imaging configuration pictured in Fig-
ure 3.9 determined the resolution of the imaging system. Note the coordinate system
shown in Figure 3.9; this coordinate system is used throughout this document. To
characterize the resolution of an imaging system and reveal possible aliasing, a point
spread function (PSF) analysis may be used. A PSF is the resulting reconstructed
image of a subresolution point scatterer. A PSF is shown in Figure 3.10, which was
reconstructed using 25 masks, with a bandwidth of 17.5-21.1 GHz, sampled at 180
MHz, and reconstructed using matched filter, GMRES, and TwIST. The cross sec-
37
Figure 3.11: Images of two metallic cylinders. a) shows the target, while b), c), d),
and e) show reconstructed images. b) and c) are reconstructed with matched filter
and d) and e) are reconstructed with GMRES. b) and d) are simulations; c) and e)
are experimental [37].
tion plots highlight that the system achieved near-diffraction-limited resolution, at

3.2 cm in cross range and 7.6 cm in range.
A more complex target, consisting of two cylinders was also imaged, as shown
in Figure 3.11. In this study, two different reconstruction methods, matched filter
and GMRES, were used. This study shows the high fidelity of simulations used to
estimate imaging performance of a metasurface computational imaging system which
has been modeled following Section 3.1.4. This study, and the PSF analysis above,
serve as a brief demonstration of the stationary imaging capabilities for the described
metasurface hardware.
38
3.3.3 Single-Frequency Imaging
For many microwave imaging systems, a large operational bandwidth is a severely

inhibiting aspect of the hardware requirements. Near-field imaging has recently
shown promise in addressing this challenge by enabling the reconstruction of 2D
objects from a 1D platform with only a single frequency point. Initial demonstrations
of this imaging paradigm involved translating horn antennas along a large aperture
extent; this operation measured each combination of positions between a transmit
synthetic aperture and a receive synthetic aperture path, creating a fully bistatic
measurement set. Probing the scene from such a range of angles allowed for high-
fidelity reconstructed images without bandwidth, but suffered long measurement
times and moving parts [110].
To better understand single frequency imaging, it is helpful to think about an
imaging system in terms of spatial frequencies (or k components). A monostatic
SAR system takes a frequency sweep from a transceiver at a series of SAR positions
to cover an area in k space as a function of position (rs ) and frequency (f ) [2, 3].
As an example, consider a monostatic system with 12 SAR locations, evenly spaced
along the y direction, spanning 2 m, and with 6 frequency points covering a 20%
fractional bandwidth. At each SAR location, the angle to the target, located at
(1 m, 0 m), is calculated and the plane wave probing the target is projected onto
the x and y directions to determine the illuminating wavenumber, (kxt , kyt ). Here,
the illuminating waves can be approximated as plane waves by assuming that the
target is an infinitesimal point in the far field of the transmitter [110]. The signal
is collected by the receive antenna with a (kxr , kyr ) pair associated with the same
incidence angle as the transmitter. These k values are added together to determine
the overall (kx , ky ) probed by the imaging system as kx = kxt +kxr and ky = kyt +kyr .
This process is repeated for each SAR location and frequency to determine the full k
39
a) Monostatic SAR
At each SAR position (rs) and frequency (f),
the system measures one (kx,ky)
Transmit Receive ∝B
[rs,f] [rs,f] SAR path
1 2 3 4 5 6 7 8 9 10 11 12
b) Multistatic Array
For Tx #11  Rx #5, the system
measures one (kx,ky) pair
Receive Transmit
[kxr,kyr] [kxt,kyt]
1 2 3 4 5 6 7 8 9 10 11 12
Figure 3.12: Schematic description of two styles of microwave imaging systems.

a) shows a monostatic SAR system, which covers an area in k space function of
SAR position and bandwidth. b) shows a multistatic array, which covers an area in
k space as a function of the transmit and receive aperture size, while only using a
single frequency point [35].
space coverage of the imaging system, as shown in Figure 3.12a. The k space coverage
shows sampling across a 2D area, stemming from the bandwidth of the system and the
motion of the antenna, leading to the ability to reconstruct 2D images. It is important
to note that due to the monostatic SAR operation, the transmitting incidence angle
and the receiving incidence angle will always be the same. Additionally, although
this analysis is performed with respect to an infinitesimal point target, a similar
analysis could be applied to a range of targets with similar results.
In contrast to a monostatic SAR system, a multistatic array involves distributing
antennas throughout a large aperture extent and using different combinations of
40
antennas (as transmitters and receivers) to form a measurement set. To describe how
a multistatic linear array can image with measurements taken at a single frequency,
a k coverage map was calculated (in the same way as the monostatic SAR system
coverage map described in the previous paragraph). Here, the SAR locations from
the monostatic SAR case are used as the antenna locations for a multistatic uniform
linear array (ULA) in order to ensure a fair comparison. In the case of a multistatic
linear array, one antenna is selected as the transmitter and another as the receiver.
The incidence angle from the transmitter to the target determines the unique plane
wave emanating from the transmitter with a given (kxt , kyt ). Since there is a distinct
receiver antenna, due to the multistatic operation, the (kxr , kyr ) is different from that
of the transmitter. Once again, adding the k space pairs associated with the plane
waves from the transmitter and receiver determines the overall (kx , ky ) probed by
one measurement mode. This process is repeated for the exhaustive combination of
each transmitter with each receiver. In this way, an area in k space is covered not
by using a large frequency bandwidth, but by the measurement paths created by the
different combinations of antennas as transmitters and receivers. The resulting k
coverage map for a multistatic ULA is shown in Figure 3.12b, where it is clear that
a sufficient 2D area in k space can be covered by measurements taken at a single
frequency point.
Practically demonstrating single-frequency imaging, as described in Figure 3.12b,
can result in prohibitively time-consuming mechanical raster scanning as in [110], or
a costly, complex array of antennas. In [111], these practical shortcomings were over-
come by utilizing electrically-large, singly-fed metasurface apertures. By radiating
different spatial patterns from the dynamic metasurface aperture described above,
different combinations of (kx , ky ) pairs can be simultaneously measured because these
apertures exhibit spatial variation along the range and cross range directions [111].
This illumination strategy is depicted in Figure 3.13. Metasurface antennas can
41
a) b)
c)
Figure 3.13: Aperture configuration for single-frequency stationary computational

imaging with dynamic metasurfaces. a) shows a schematic view of the transmitted
and received wavefronts, b) shows the control circuitry, and c) shows a picture of the
experimental setup with two dynamic metasurface apertures [111].
generate spatially-distinct radiation patterns at a rapid rate, each of which probes

a superposition of (kx , ky ) pairs. While these apertures multiplex the information
from many incidence angles, post-processing techniques can isolate the contributions
of each transmit and receive element and virtually achieve the multistatic operation
described in Figure 3.12b.
To use dynamic metasurface apertures for single-frequency imaging, a large aper-
ture extent must be realized and there must be spatial diversity among the radiated
waveforms. In this manner, performance benefits from using a transmit and receive
aperture which can independently probe the scene with diverse radiation patterns.
To craft such an imaging configuration, two dynamic metasurfaces were deployed,
one acting as the transmitter and one as the receiver. These apertures were sep-
arated by 20 cm, resulting in an aperture which spanned 100 cm. In this case,
different masks were applied to the transmitter and receiver, resulting in a set of
Nm,T x Nm,Rx measurements, where Nm,T x and Nm,Rx are the number of transmit and
receive masks, respectively.
In the case of single-frequency, panel-to-panel imaging, reconstructions were fa-
42
cilitated by an adapted version of the range migration algorithm (RMA). The RMA
was developed for processing SAR data and has been demonstrated as an efficient
method for reconstructing large data sets [112]. An adapted version of the RMA was
initially developed in [113] to facilitate fast reconstructions of metasurface compu-
tational imaging data. This method has been further extended to single-frequency
configurations in [111]. Since the range resolution of this imaging system is not set
by the bandwidth, analytic calculations or simulations must be used to estimate the
spatially-varying resolution across a given scene. To this end, a spatial frequency
analysis can be used. The k space of an imaging system determines the angles from
which the scene is illuminated and considers them in the spatial frequency domain.
Each radiation pattern can be decomposed into a set of plane waves such that dif-
ferent spatial frequencies are being generated in the imaging domain. The extent
of k space covered by an imaging system sets the resolution, while gaps in k space
coverage can lead to aliasing [114]. Resolution expectations can be extracted from
analyzing the k space coverage relative to each point in the scene. To experimen-
tally characterize the performance of this system, a PSF analysis was conducted and
the results are shown in Figure 3.14 [115]. Here, the target is well-resolved in both
cross range and range, despite the lack of bandwidth, validating the proposed imag-
ing paradigm. The experimentally achieved PSF matches well with the simulation
predicted PSF, as shown in Figure 3.14. Additionally, a more complex target, con-
sisting of 5 scatterers, has been reconstructed with the adapted RMA as shown in
Figure 3.14c.
3.3.4 Experimental SAR with Metasurfaces
In addition to conducting stationary computational imaging, various SAR modalities

were also demonstrated. These experiments were carried out with the dynamic meta-
surface aperture described in Section 3.3.1. This aperture was designed to generate
43
a) b) c)
Figure 3.14: Single-frequency stationary computational imaging results. a) shows

a simulated PSF, b) shows an experimental PSF, and c) shows a complex target con-
sisting of five scatterers. All reconstructions have been conducted with the adapted
RMA [111].
pseduo-orthogonal spatial waveforms, leading to suboptimal beamforming abilities

and limited performance. Nevertheless, these experiments serve as a sufficient proof-
of-concept demonstration of traditional and alternative SAR modalities.
To experimentally demonstrate SAR with a metasurface platform, the aperture
described in Section 3.3.2 was used. Since there is not a sufficient predictive model,
an empirical, experimental investigation was used to generate directive beams from
this aperture. This involved measuring the response of the aperture for a variety of
tuning states and then mining those measurements for those which exhibited beam-
like radiation patterns. The first set of attempted masks was a collection of square-
wave masks, which each possess a certain periodicity. In addition, a large variety of
pseudorandom masks have been measured in an effort to generate additional steered
beams. By sorting through many measurements of random masks, several radiation
patterns matching desired characteristics can be found.
From the empirically-mined set of tuning masks, directive beams spanning ±23◦
have been generated. The beams have a mean width of 10◦ and all have sidelobes
of less than -4 dB. Figure 3.15a shows a collection of these beams at 19.3 GHz;
Figure 3.15b shows random radiation patterns for comparison. As shown in Fig-
44
Figure 3.15: Experimentally-optimized beam-like radiation patterns generated by
a dynamic metasurface aperture. a) shows optimized directive beams and b) shows
pseudorandom radiation patterns; both are shown for 19.3 GHz. c) shows the cover-
age in frequency and angle where beams with sidelobes below -4 dB were found [82].
ure 3.15c, the dynamic aperture masks can be selected such that a directive beam is
steered over angles within ±23◦ across a bandwidth of 2.88 GHz (17.50-20.38 GHz).
A few angle-frequency pairs have not been realized, but this does not severely hinder
the performance in this work and could be corrected with further exploration of ad-
ditional masks. The experimentally-determined beams are sufficient for conducting
both stripmap and spotlight imaging.
The process for finding beams from this metasurface aperture was completed
through experimental near field scanning and was a cumbersome approach to the
problem. With the 112 elements all individually addressable, an enormous variety
(2112 ) of masks are possible and allow for the creation of the desired radiation pat-
terns. The set of possible masks left many possibilities untested, but a sufficient
quantity of random masks was tested to produce satisfactory beams to fill the de-
sired frequency/angle coverage, as highlighted in Figure 3.15c. While this aperture is
sufficient for initial experimental results, its design could be better tailored towards
the goal of beam synthesis by applying techniques such as grayscale tuning, denser
element placement, and tapered element design.
45
After directive beams were experimentally determined from the dynamic meta-
surface aperture, stripmap and spotlight mode imaging were demonstrated. For the
experiments described in this section, the bandwidth of available beams stretches
from 17.50 GHz to 20.38 GHz and this range is used for each imaging case, with
the bandwidth sampled at intervals of 90 MHz. Imaging was conducted in an ane-
choic chamber; measurements were formulated as described in Section 3.2 and re-
constructed with matched filter. The synthetic aperture spanned 60 cm, sampled at
intervals of ∼ λ/3 (∆ys = 5 mm). Stripmap mode was first conducted, with a broad-
side beam radiated from each synthetic aperture position. A new tuning mask must
be used for each frequency to maintain a given radiation pattern across a frequency
sweep. For practical reasons associated with the experimental setup, measurements
were taken in an inverse SAR mode, where the scene was translated and the antenna
remained stationary.
In addition to conducting stripmap and spotlight imaging, an alternative modal-
ity, termed diverse pattern stripmap, can be also realized using metasurface antennas.
In this method, which is inspired by computational imaging systems [75,107,116], di-
verse radiation patterns are used in place of traditional, directive beams. These beam
profiles do not necessarily exhibit a distinct main lobe, but instead include multiple
lobes. While it may seem as though this type of radiation pattern may exhibit low
directivity, the majority of the radiated energy can be contained within a small num-
ber of lobes such that there is still a high concentration of energy probing different
parts of the scene. The overall patterns can span a wide collective beamwidth, which
allows for high resolution by probing a wide range of k components. This method
is described visually in Figure 3.16. Similar to stripmap, diverse pattern stripmap
does not focus on one particular region on the ground, eliminating any limitations on
the size of the ROI. Diverse pattern stripmap stands to provide gains as compared
to traditional stripmap by providing high resolution while maintaining the ability to
46
Stripmap
Spotlight
Diverse Pattern Stripmap
Figure 3.16: Schematic view of stripmap, spotlight, and diverse pattern stripmap,
conducted with a metasurface aperture [82].
image a large ROI. Furthermore, since this modality has less stringent requirements
regarding beam characteristics, a larger bandwidth may be used due to the fact that
distortion of the radiated beam profiles does not degrade image quality.
To assess the performance of SAR imaging using a dynamic metasurface aperture,
first a PSF was reconstructed. These results are plotted in Figure 3.17 and highlight
the improvement in resolution of both of these modes as compared to stationary
imaging. From these images and as expected, spotlight imaging results in better
cross range resolution as compared to stripmap. The achieved cross range resolution
for stripmap is 1.6 cm, while for spotlight, it is 1.1 cm.
In addition to characterizing the resolution of these two modalities, an analysis
of the scene size accessible to them is also provided. To highlight the ability of each
47
Figure 3.17: Experimental PSF conducted with a metasurface aperture in various
SAR modalities. Stripmap, spotlight, and diverse pattern stripmap have been shown
[82].
Figure 3.18: Experimental imaging of a line of point scatterers conducted with a

metasurface aperture in various SAR modalities. Stripmap, spotlight, and diverse
pattern stripmap have been shown [82].
of these modes image across a given scene, a line of spheres was imaged. In total, the
target was 6 metallic spheres (with diameter = 1.5 cm), spaced at 3 cm, spanning
15 cm. The reconstructed images are plotted in Figure 3.18. These images serve to
highlight the tradeoff associated with stripmap and spotlight; stripmap can image
a larger scene size, spotlight can achieve higher resolution, and diverse pattern can
realize high resolution across a large scene.
48
3.3.5 3D SAR Imaging
In addition to demonstrating diverse pattern stripmap in this manner, other mo-

tion paths were also explored. Up to this point, the aperture was translated in the
along-aperture direction, which yields improved cross range resolution as compared
to stationary imaging. In other implementations, the aperture can move vertically
rendering it possible to reconstruct volumetric images. This imaging paradigm ex-
tracts cross range information from the aperture’s spatial diversity, range information
from spectral bandwidth, and height information from mechanical motion. In one
implementation of this method, the synthetic aperture path was oriented in the z di-
rection, spanning 32.25 cm, sampled at 7.5 mm. Measurements taken this way were
reconstructed with matched filter. Reconstructions of measurements taken this way
are shown in Figure 3.19a and 3.19b, which can clearly resolve objects distributed in
all three directions.
In an attempt to maximize the overall performance of this imaging system, fi-
nal experiments in this style were conducted with a meandering synthetic aperture
path. This path spanned two dimensions, with along-aperture (y direction) motion
a) b) c)
Figure 3.19: Experimental volumetric images measured with a dynamic metasur-

face aperture operating in diverse pattern stripmap mode. In a) and b), the synthetic
aperture path was one-dimensional, with motion along z. In c) motion in both y and
z create the synthetic aperture area. The targets in a) and b) were comprised of a
collection of point scatterers; the target in c) was individual wireframe letters [82].
49
improving cross range resolution and height motion (z direction) providing height
information from the scene. In these experiments, the synthetic aperture path was
extended to cover 24 cm in the z direction (sampled at 10 mm) and 20 cm in the
y direction (sampled at 20 mm). Results of this method are shown in Figure 3.19c.
Here, the word “DUKE” has been spelled in a metallic wireframe, with each letter
imaged individually. Ultimately, this imaging method yields the highest resolution
in all directions for this system.
3.3.6 Single Frequency SAR
Combining the principles of single frequency imaging and SAR, single frequency
SAR has been demonstrated. Such operation enables fully 3D imaging with single
frequency measurements by obtaining cross range resolution from aperture extent,
range resolution from multistatic operation, and height resolution from a synthetic
aperture. Numerous analyses of the dynamic metasurface aperture described in Sec-
tion 3.3.1 have been documented [37,82,111], but this section details its use for single
frequency SAR [35]. With this application at hand, a detailed mathematical deriva-
tion of the transmitted/measured signal is provided, along with the corresponding
reconstruction method (an adapted form of the range migration algorithm).
Concept and Method
Building on the 2D stationary, single-frequency metasurface imaging system [111],

the single-frequency approach is extended to form a synthetic aperture for full 3D
imaging. To form the effective aperture, a pair of metasurface antennas is translated
over an area as shown in Figure 3.20. This system can extract range (x) and cross
range (y) information from the electrically large aperture extent and multistatic
operation, while movement of the system (along the height direction) provides height
(z) information. To probe a scene, the metasurface antennas apply collections of
50
Rx
Tx
Z
Y
X
Figure 3.20: Conceptual diagram of a single-frequency imaging system using meta-
surface antennas. One metasurface antenna serves as the transmitter while another
receives the backscattered signal. At each synthetic aperture position, measurements
are taken by cycling through combinations of spatially-diverse radiation patterns on
the transmitter and receiver [35].
pseudorandom tuning states to generate radiation patterns that exhibit low spatial
correlation to illuminate a scene as shown in Figure 3.21 [36–38]. This illumination
strategy, coupled with the movement associated with aperture synthesis, leads to an
imaging system that can be conceptualized as producing sets of plane waves from
various angles of incidence. The superposition of plane waves thereby interrogates
multiple spatial frequency components, in x, y, and z, simultaneously. This operation
enables 3D imaging without relying on a wide bandwidth for range information.
Single-frequency microwave imaging with metasurfaces can be supplemented by
incorporating mechanical motion along the z direction (as in the coordinate system
51
shown in Figure 3.20) to provide height information. In this manner, the system
performs stripmap synthetic aperture radar (SAR), but with an unconventional illu-
mination strategy in the direction perpendicular to the SAR path [3]. Note that the
virtual multistatic operation is maintained with respect to the range and cross range
directions, but the imaging system is operating in a monostatic manner with respect
to the height direction. In this mode of operation, a virtual 2D multistatic aperture is
formed, with each of the transmit and receive apertures covering independent areas.
Similar imaging configurations, including different metasurface antenna placement
or different synthetic aperture geometries may be considered with a similar analy-
sis. Different synthetic aperture paths could help to realize denser k space coverage,
which could help to reduce aliasing effects. Since the geometry shown in Figure 3.20
probes the scene from the maximum and minimum incidence angles availed to the
synthetic aperture footprint, different synthetic aperture paths would not extend
T1
T2 X
T3
Scene
Y
Tx Rx
Figure 3.21: Diagram of the illumination strategy for a single-frequency, meta-

surface antenna imaging system. The metasurfaces measure the scene by cycling
through combinations of transmit and receive radiation patterns (each aperture is
set with a tuning state, Tn ). This process is repeated at each synthetic aperture
position, as shown in Figure 3.13. Note that the transmitter and receiver do not
necessarily apply the same tuning state during a given measurement.
52
the limits of the k space coverage unless the overall footprint of the synthetic aper-
ture was expanded. In this work, due to experimental configuration restrictions, the
analysis is limited to the system shown in Figure 3.20.
To better illustrate the utility of the proposed imaging system, it is briefly com-
pared with a more conventional architecture. In this system, two 1D metasurface
apertures are used, each spanning 25λ, which are positioned side-by-side. If the trans-
mit and receive metasurfaces were replaced with arrays of independent antennas, 50
independent antennas would be required for transmission and reception, assuming
Nyquist (λ/2) sampling. Thus, the metasurface hardware allows for a reduction from
100 independent antennas to 2 independent antennas. If the full area covered by the
mechanical translation of the metasurface antennas was instead replaced with a solid
state array of antennas, 5,000 independent antennas would be required.
Incorporating motion along the height dimension results in a different set of
relations governing the k space coverage of the imaging system. To account for
this behavior, the different k space contributions must be accounted for within the
reconstruction method, which can be accomplished by deriving the signal in Fourier
space. Ultimately, full 3D information can be obtained with this operation, in which
range and cross range information is extracted from the aperture size and height
information is extracted from mechanical motion.
Signal Description
Due to the large data sets that result from single-frequency SAR operation, the range
migration algorithm (RMA) is used for post-processing. To implement the RMA,
the spatial frequency components probed by the system must be known and the
nontraditional illumination strategy (highlighted in Figure 3.21) must be accounted
for [112,113,117–119]. In this section, these two considerations are addressed, leading
to an implementation of the RMA which is specific to the imaging system described
53
in this section.
To understand the pre-processing step required for measurements taken with
metasurface apertures, a brief description of these measurements is provided here.
Measurements taken with a metasurface-based, single-frequency SAR imaging sys-
tem are a function of transmit tuning state, receive tuning state, and synthetic
aperture position [82]. Each metamaterial radiator may be modeled as an infinites-
imal effective dipole which may be considered as an independent source/receiver
[39, 69, 95, 120–122]. To facilitate reconstruction of measurements taken this way
with the RMA, a pre-processing step must be used to isolate the independent con-
tributions of each virtual source [113]. The transmitted fields can be written in
decomposed matrix form as Et = ΦG, where Φ is a matrix of the fields at the sur-
face of the aperture and G is the propagation matrix of those fields to the scene
(and analogously for the received fields). Through a matrix decomposition of Φ,
the multiplexed measurements taken as a function of transmit/receive tuning state
can be transformed into the equivalent measurements taken as a function of inde-
pendent transmit/receive elements [123]. This pre-processing step thus generates a
measurement for each combination of transmit/receive element [113]. In summary,
the pre-processing step transforms the signal in the following manner
S(Tt , Tr , zs ) → ST (yt , yr , zs ). (3.7)
Here, Tt and Tr represent the tuning state of the transmitter and receiver, respec-
tively, yt and yr represent the positions of the transmitting and receiving virtual
sources, and zs is the synthetic aperture position.
In order to carry out the described pre-processing step, the fields radiated by each
aperture must be known for each tuning state. A near field scan is used to obtain the
response of each aperture [124]. A set of pseudorandom tuning states is measured for
54
both the transmit and receive metasurfaces, with a subset of the measured tuning
states used for each experiment [37]. The measured fields are used to convert the
measured signals to those of effective, independent sources characterized by the near
field scan [36].
After applying the pre-processing step described in Equation (3.7), the signal as
represented by the transmit and receive dipoles can be reconstructed. To properly
reconstruct signals measured by such an imaging system using RMA formulation,
the k space relations of the imaging system must be properly described—this is done
through the dispersion relation of the system. To determine the dispersion relation
of the system, a derivation of the signal in Fourier space must be conducted (in a
similar manner to the derivations described in [118] and [113]). For a signal which
is a function of the transmit element locations (yt ), receive element locations (yr ),
and synthetic aperture positions (zs ), the phase term of the transmitted field, Et ,
can be isolated according to the propagation by the Green’s function as e−jkRt (and
similarly for the received field, Er ). Here, k is the free space wavenumber, defined
p
as k = 2πf /c and also k = kx2 + ky2 + kz2 . By combining Et and Er (and neglecting
the amplitude term) along with the scene reflectivity (σ), the measured signal can
be written as
Z
ST = σ(x, y, z)e−jkRt e−jkRr dV (3.8)
V
where
p
Rt = x2 + (y − yt )2 + (z − zt )2
p (3.9)
Rr = x2 + (y − yr )2 + (z − zr )2 .
Here, zt and zr are the position in the z direction of the transmitter and receiver,
respectively. The Fourier transform of the transformed signal (ST ), in the zt , zr , yt ,
55
and yr dimensions, can then be calculated as
Z Z Z
ŜT = σ e−jkRt e−jkyt yt e−jkzt zt
V yt zt
Z Z (3.10)
−jkRr −jkyr yr −jkzr zr
e e e dzr dyr dzt dyt dV.
yr zr
Note that here, since the transmitting locations and receive locations have been
distinctly defined, the contributions of the transmitter and receiver are separable.
To solve this equation, first the contribution of the transmitter is addressed, which
is
Z Z
e−j(kRt +kyt yt +kzt zt ) dyt dzt . (3.11)
yt zt
This integral can be solved by applying the method of stationary phase, which yields
the stationary points yt,s and zt,s .
kyt x
yt,s = y ± q (3.12)
2 2
k 2 − kyt − kzt
kzt x
zt,s = z ± q . (3.13)
2 2 2
k − kyt − kzt
Thus, the solution to the integral shown in Equation (3.11) can be written by sub-
stituting Equation (3.12) and Equation (3.13) to obtain
√2 2 2
e−j( k −kyt −kzt x+kyt y+kzt z) . (3.14)
In an analogous manner, the contribution of the received signal can be solved to be
√2 2 2
e−j( k −kyr −kzr x+kyr y+kzr z) . (3.15)
56
Using the solutions from Equation (3.14) and Equation (3.15), Equation (3.10)
can then be written as
Z √ 2 −k 2 x+k y+k z)
k2 −kyt
ŜT = σe−j( zt yt zt
V
(3.16)
√
k2 −kyr
2 −k 2 x+k y+k z)
e−j( zr yr zr
dV.
From this solution, the dispersion relations of the imaging system can be regrouped
to write Equation (3.16) as
Z
ŜT = σe−j(kx x+ky y+kz z) dV, (3.17)
V
where kx , ky , and kz are
q q
2 2
kx = kxt + kxr = k − kyt − kzt + k 2 − kyr
2 2 − k2
zr (3.18)
ky = kyt + kyr (3.19)
kz = kzt + kzr . (3.20)
At this point, the relations governing the k space sampling of the imaging sys-
tem have been identified. Properly modeling the dispersion relation of the system is
paramount to successfully reconstructing SAR measurements [113,118]. In this imag-
ing configuration, kyt = kyr = [−π/∆ya , π/∆ya ], which is determined by the spacing
between the neighboring virtual dipole sources representing the metasurface. Mean-
while, kzt = kzr = [−π/2∆z , π/2∆z ], with ∆z determined by the synthetic aperture
step size. In this work, the transmitter and receiver are moved concurrently, meaning
that measurements are only taken when zt = zr . Once these dispersion equations
57
have been defined, Stolt interpolation can be performed within the range migration
processing to map the signal from the kyt , kyr , and kz domain to the kx , ky , and kz
domain, following Equation (3.18), Equation (3.19), and Equation (3.20).
Note that the signal must also be spatially demodulated to account for the center
position of the transmitter and receiver according to their positions, following [110,
111] and mathematically implemented as
S∗ =Sej(kx xc,t +ky yc,t +kz zc,t )

(3.21)
S∗ =Sej(kx xc,r +ky yc,r +kz zc,r )
The position of the center of the transmitter and receiver are (xc,t , yc,t , zc,t ) and
(xc,r , yc,r , zc,r ), respectively.
To summarize the specific RMA implementation used in this work, the following
steps are performed. First, a matrix decomposition is applied to pre-process the
measured signal, following Equation (3.7). Then a Fourier transform of the signal
is taken to move the data into the wavenumber domain. Once in the wavenumber
domain, the signal is mapped into the kx , ky , and kz directions using the dispersion
relation described in Equation (3.18), Equation (3.19), and Equation (3.20). After
that, the signal is spatially demodulated and matched filter is applied. Next, Stolt
interpolation is applied to map the signal onto a regular grid in kx , ky , and kz . Last,
a 3D inverse Fourier transform of the signal is taken to form an image [113, 117, 118,
121]. This process represents an implementation of the RMA which is specific to
the imaging system at hand and can efficiently reconstruct images taken with the
single-frequency, metasurface-based SAR imaging system.
Experimental Results
To experimentally demonstrate single-frequency, 3D synthetic aperture imaging, two

dynamic metasurfaces were deployed, one as a transmitter and one as a receiver, as
58
Tx Rx
Figure 3.22: Experimental setup consisting of two metasurface apertures. One is
used as a transmitter (labeled Tx) and the other as a receiver (labeled Rx).
shown in Figure 3.22. A detailed description of the specific dynamic metasurface

used can be found in Section 3.3.1 [37]. The experiments conducted in this sec-
tion involve two metasurfaces oriented along the same axis (the y axis as shown in
Figure 3.20), with edge-to-edge separation of 20 cm. This arrangement leads to an
overall aperture size of 100 cm. The imaging system is controlled by a PC which uses
Matlab to coordinate the two Arduino microcontrollers (one for each metasurface),
the vector network analyzer used as the RF source/detector (Keysight E8364C), and
the mechanical stage. Each measurement is acquired by setting the position of the
synthetic aperture (by way of a linear stage), applying a new tuning state to the
transmit and receive metasurfaces (by way of the microcontrollers), and measuring
the signal between the transmitter and receiver. In this manner, the collection of
signals as a function of transmit tuning state, receive tuning state, and synthetic
aperture position can be obtained.
Measurements taken for this work involved cycling through a number of transmit
and receive tuning states at each position on the synthetic aperture. In this case,
due to physical constraints, the apertures remained stationary, while the target(s)
moved, consistent with inverse SAR (ISAR) operation. The synthetic aperture path
spanned 40 cm and was sampled at 7.5 mm (∼ λ/2), leading to 54 locations. With
25 tuning states for the transmitter and receiver, a total of 25 × 25 × 54 = 33,750
measurements were taken. For each of these tuning states, a random set of 24
elements were radiating simultaneously. This choice ensures that sufficient energy
59
Figure 3.23: 3D image reconstruction of simulated measurements of a point scat-
terer. The target is well-resolved in all three directions with measurements at a single
frequency.
can reach all portions of the metasurface, while radiating efficiently to maintain an
adequate SNR [37].
The first feature of the system experimentally investigated was resolution. Since
the operation is based in the near field, the resolution of the system is spatially-
varying and does not follow traditionally-modeled equations [111]. As result, an
expectation for the resolution performance of the system was determined from a
simulation of a PSF.
The signal for a simulated PSF was calculated by propagating the fields radiated
by the transmitter (as measured by near field scan) to the scene, applying the scat-
tering off of a point, and propagating the scattered fields back to the receive aperture
to form a simulated signal. The simulated signal for a point object located at (x =
0.35 m, y = 0 m, z = 0 m) was reconstructed with the range migration algorithm, as
adapted for this system. Here, 65 transmit and receive tuning states were used. From
the simulated PSF, the expected resolution in x, y, and z was calculated from the
full-width at half-max (FWHM) to be 3.5 cm, 1.6 cm, 0.8 cm, respectively. Using the
60
Figure 3.24: 3D image reconstruction of experimental measurements of a point
scatterer. Here, a metallic marble of diameter 1.75 cm is used as target and is
well-resolved in all three directions with measurements at a single frequency.
same imaging parameters as in simulation, an experimental PSF was also performed,

with the resulting resolution (as determined by the FWHM of the reconstruction)
in x, y, and z found to be 3.8 cm, 2.7 cm, and 3.0 cm, respectively. Images of the
simulated and experimental PSF are shown in Figs. 3.23 and 3.24, respectively. It
should be noted that the metallic sphere used to experimentally characterize the
PSF was 1.75 cm in diameter, which is slightly above the expected resolution in the
y and z directions. Here, the object is well-resolved in range, cross range, and height,
which validates the idea that using a metasurface antennas in a SAR context enables
single-frequency 3D imaging.
Single-frequency SAR imaging is then directly compared with bandwidth imag-
ing. To conduct similar studies with bandwidth, single-frequency measurements are
reconstructed. Then reconstructions of measurements taken with different frequen-
cies are coherently summed. In this study, introducing multiple frequency points
across a bandwidth increases the number of total measurements. Without compen-
sating for the total number of measurements, the bandwidth case could have an unfair
61
Table 3.1: Resolution of single-frequency and bandwidth imaging cases. NT is the
number of transmit tuning states, NR is the number of receive tuning states, and Nf
is the number of frequency points. For each bandwidth case, the bandwidth used is
17.68-19.03 GHz.
NT × N R × N f δX (cm) δY (cm) δZ (cm)
64 × 64 × 1 3.8 2.7 3.0
16 × 16 × 16 3.5 2.9 2.8
64 × 64 × 16 4.2 2.5 3.8
advantage, since using more measurements can suppress noise. With this consider-
ation in mind, test cases are shown with different numbers of transmit and receive
tuning states at each synthetic aperture position. The three cases are: a single-
frequency result which uses 64 transmit states, 64 receive states, and one frequency
point (4,096 total measurement modes per synthetic aperture location), a bandwidth
result which uses 16 transmit states, 16 receive states, and 16 frequency points (4,096
total measurement modes per synthetic aperture location), and a bandwidth result
which uses 64 transmit states, 64 receive states, and 16 frequency points (65,536 total
measurement modes per synthetic aperture location). The third case is included to
provide a direct comparison from single-frequency to bandwidth when the number of
measurement modes per frequency is the same. Each of the bandwidth cases uses a
frequency sweep from 17.68 GHz to 19.03 GHz, sampled at 90 MHz. This bandwidth
is chosen because the resolution in certain directions is determined by the maximum
frequency — using the single-frequency operating point as the upper bound of the
bandwidth ensures a fair comparison.
Each of the described cases was used to reconstruct an experimental PSF. The
resulting images are plotted in Figs. 3.25, 3.26, and 3.27, which show a 3D image
and the corresponding cross sections at the target’s location. For these studies, the
target used was the same as in Figure 3.24. From the cross section plots in Figs. 3.25,
3.26, and 3.27, the resolution of the three cases can be extracted by measuring the
FWHM in each direction — these results are shown in Table 3.1. From these results,
62
Figure 3.25: Experimental PSF with 64 transmit masks, 64 receive masks, and a
single frequency (19.03 GHz). Cross sections are plotted with a linear colormap.
Figure 3.26: Experimental PSF with 16 transmit masks, 16 receive masks, and
16 frequency points, spanning 17.68 - 19.03 GHz. Cross sections are plotted with a
linear colormap.
63
Figure 3.27: Experimental PSF with 64 transmit masks, 64 receive masks, and
16 frequency points, spanning 17.68 - 19.03 GHz. Cross sections are plotted with a
linear colormap.
it can be seen that the resolution of the bandwidth cases remains consistent with
the single-frequency case, indicating that operating at a single frequency does not
degrade resolving power. It is expected that if the standoff distance were increased
substantially, the bandwidth case would begin to demonstrate superior resolution,
specifically in the X direction. At standoff distances comparable to the length of the
aperture, however, the single-frequency imaging system can match the performance
of bandwidth systems. The similar resolution stems from the fact that within the
near field, the resolution in both range and cross range are dominated by the geom-
etry of the system, rather than by the bandwidth [111]. Additionally, Figure 3.27
shows suppressed background noise compared to the other cases; this effect can be
attributed to a larger number of overall measurement modes.
Returning to single-frequency imaging demonstrations, a more complex target
was also measured, comprising four copper tubes. These tubes were oriented along
64
Figure 3.28: Reconstructed 3D image of four copper cylinders with approximate
diameter 2 cm and length of 10 cm.
the y and z directions, as shown in Figure 3.28 and had a diameter of approximately
2 cm and a length of approximately 10 cm. This target highlights the ability of this
imaging system to resolve multiple targets in all three directions; the experimentally-
reconstructed image is shown in Figure 3.28. The number of tuning states and
synthetic aperture positions remained consistent with those used for the experimental
PSF shown in Figure 3.24. The reconstructed image shows the ability of this system
to image an object spanning a large extent in y and z.
In addition to the two objects described above, a larger, extended object was
also imaged. The target in this case was a metallic cylinder, with diameter = 15 cm
and height = 70 cm. The resulting image is shown in Figure 3.29, which highlights
the ability of this imaging system to image objects which span a larger extent in z.
Furthermore, when using such a large object, the signal returned is much greater,
leading to higher SNR. As a result, this study allowed us to reduce the number to
65
Figure 3.29: Reconstructed 3D image of a cylinder of diameter 15 cm and height
70 cm.
transmit and receive tuning states from 65 to 25 to demonstrate the ability of the
imaging system to reconstruct using fewer measurements.
Another study performed involved investigating the particular choice of frequency
used. For any imaging system performing single-frequency imaging, choosing the
optimal frequency is paramount to achieving the best performance. Different factors
can affect the optimal frequency choice, most of which are specific to the imaging
hardware being used. As a result, the frequency choice study performed here only
applies to the specific metasurface antennas discussed in this work. While these
specific results only apply to the metasurface antenna used throughout this work,
the concepts and procedure can be readily applied to other imaging hardware. For all
single-frequency imaging systems, choosing a higher frequency will result in higher
resolution.
The metasurfaces used in this work (as described in Section 3.3.1 and detailed
66
17.50 GHz 18.04 GHz 19.03 GHz 20.02 GHz 21.01 GHz
Figure 3.30: Single-frequency SAR images of a large cylinder at 5 different fre-

quency points.
in [37]) have two element geometries, resonant at 17.8 and 18.9 GHz, respectively.
Each element radiates more strongly close to these frequencies. When the elements
radiate strongly, there is a higher contrast ratio between the on/off states, but the
strong radiation from each element results in a quick attenuation of the guided wave.
When the guided wave is attenuated quickly, there is not sufficient energy left to
excite elements far from the feeds, leading to portions of the aperture which do not
radiate. With such a nonuniform illumination of the aperture, aliasing can become
significant. To ensure that sufficient energy can reach all elements in the aperture,
the operational frequency can be chosen to be off-resonance, either above or below
the resonant frequencies of the two elements. When choosing such an operating
frequency, the guided wave attenuates slower, allowing the full extent of the aperture
to radiate. If the frequency choice is too far away from the resonances of the two
element designs, however, the contrast between the on and off states becomes small,
leading to more correlation among radiation patterns and the elements radiate too
little energy, reducing the efficiency of the overall aperture. These behaviors can be
seen by analyzing the power radiated spectrum shown in [37]. All of these factors
(radiation efficiency, illumination uniformity, and contrast ratio) must be balanced
by the frequency choice.
To balance the factors described above, best performance is predicted at a fre-
quency slightly above 18.9 GHz (the higher resonant frequency among the two ele-
67
ments in the metasurface). This prediction was tested by imaging the target from
Figure 3.29 with different frequency points. From the results shown in Figure 3.30,
19.03 GHz (the frequency used throughout this section) yields the best reconstructed
image. Operating at 17.5 GHz results in the elements coupling too strongly to the
guided wave, such that the metasurfaces only radiate from elements close to the feed.
Working at 17.5 GHz leads to high radiation efficiency (as evidenced by the low noise
in the reconstructed image), but causes the aperture to only radiate near the feeds.
Since the metasurface has nonuniform illumination, spatial aliasing occurs, leading
to strong sidelobes in the reconstructed image. Operating too far above the reso-
nance of each element (at 20.02 or 21.01 GHz) produces uniform illumination, which
can be seen from the small sidelobes. However, this frequency choice also leads to
low radiation efficiency and correlation among radiation patterns, which contribute
to the noisy image. When these considerations are properly balanced, at 19.03 GHz,
the aperture demonstrates adequate radiation efficiency, uniform aperture illumina-
tion, and low correlation among measurement modes, resulting in a high fidelity
reconstruction.
To further investigate the behavior of this single-frequency SAR imaging system,
an analysis of the number of transmit/receive tuning states was performed. If fewer
states can be used on the transmit/receive metasurfaces, the acquisition time can be
significantly reduced. In this case, the four copper tubes (shown in Figure 3.28) were
used as the target, while the number of transmit/receive states was varied from 25
to 65. There are two major consequences of the number of transmit/receive states
(and thereby the number of measurements). First, with more measurements, noise
in the system is better averaged across the measurement set. Second, a larger set of
measurements results in a larger matrix representing the effective dipolar sources that
characterize the radiation of a metasurface antenna. As the number of measurements
increases, the corresponding pseudo-inverse of this ill-conditioned matrix will yield a
68
better transformation from the measured signal in terms of transmit/receive states to
the effective independent dipolar sources (following Equation (3.7)). On top of these
effects, more measurements may mean that the system is more likely to span the
entire scene’s basis. Selecting a set of tailored transmit/receive states that samples
this basis with a reduced set of measurements is the focus of compressive sampling
techniques, which is discussed in other works (see [125], for example). The goal
here is to demonstrate the viability of the proposed imaging configuration; further
analysis and optimization of tuning states are left for future works.
The consequences of the number of measurements are highlighted by the results in
Figure 3.31, which show a reconstruction for 25, 45, and 65 tuning states. From these
images, using 65 states results in a cleaner and higher-fidelity image as compared to
using 45 states. With 25 states, the image has significantly degraded.
69
Figure 3.31: Reconstructions of four metallic cylinders. Each image uses a different
number of transmit/receive tuning states, NT , resulting in NT2 Ns total measurements
(Ns is the number of synthetic aperture positions).
70
4
Simulated Metasurfaces for SAR
The premise of this document is to demonstrate metasurface antennas as a suitable

alternative to traditional hardware. To demonstrate that, simulated studies and
experimental demonstrations are provided in the next two chapters. This chapter
details four main simulated studies which highlight the utility of metasurface anten-
nas in a SAR context. Chapter 5 then uses these studies to inform the design of a
metasurface antenna prototype and to present experimental results.
This chapter covers an initial metasurface antenna design and its simulated re-
sults conducting SAR. Next, tuning strategies specific to metasurface antennas are
described. Then a large scale metasurface antenna for spaceborne SAR is presented,
including a comparison with TerraSAR-X. Last, grating lobe suppression is discussed
in the context of a simulated metasurface antenna and the source of metasurface grat-
ing lobes is mathematically derived and addressed. To help distinguish the different
antenna designs being simulated, a parameter table is presented any time a new an-
tenna design is introduced. Note that each section analyzes a different metasurface
design.
71
Table 4.1: Antenna parameters for the metasurface aperture used in this section to
simulate SAR.
Parameter Value
# of elements 100
Spacing 5 mm
Waveguide width 20 mm
Bandwidth 9.2-9.7 GHz
4.1 Simulated SAR with Metasurface Antennas
The first goal of simulating metasurfaces for SAR is to show rudimentary demon-
strations of stripmap and spotlight SAR with a metasurface antenna. In order to
do so, a metasurface antenna must be able to generate a steerable, directive beam.
The modeling method described in Section 3.1.4 is applied to simulate a sample
metasurface antenna. Once the antenna has been modeled numerically, it is used to
demonstrate SAR imaging in various modalities.
4.1.1 Simulated Metasurface Architecture
Since the main goal of this section is an initial demonstration of SAR with a meta-
surface antenna, the metasurface design has not been thoroughly optimized. Still,
it proves sufficient for initial demonstrations and sheds light on various aspects of
metasurface antenna design for future iterations. The simulated metasurface an-
tenna, illustrated in Figure 4.2, consists of 100 elements (spaced at 5 mm) embedded
within a rectangular waveguide (width = 20 mm and height = 5 mm), spanning 50
cm. Here, λ is the free space wavelength at the center of the operational bandwidth,
9.2-9.7 GHz. The operational bandwidth is in the X band, as it is commonly used
for earth observation [63, 126]. The elements are fed by a guided wave emanating
from one end of the waveguide. This aperture is one-dimensional and therefore only
generates beams which are directive in one direction. Its parameters are summarized
in Table 4.1.
72
The individual metamaterial element used in the simulated metasurface antenna
is a complementary electric inductive-capacitive (cELC) resonator [84–87], similar
to those in previous dynamic metasurface designs [37] and pictured in Figure 4.1.
Varactors, introduced into each element, allow for an applied voltage to change the
capacitance across the gap between the cELC element and the surrounding con-
ductor. Changing this capacitance results in grayscale tunability of the resonant
frequency of the radiator. These elements may be biased using vias which extend
through the bottom layer to connect to a control layer underneath the waveguide,
following the design described in [83]. A single cELC element was modeled in a
rectangular waveguide in a full wave solver and the S parameters were extracted as
a function of frequency and tuning state. From the S parameters, the polarizabiltiy
(α) can be extracted [35, 127].
jab
α= (1 + S11 − S21 ) (4.1)
β
Here, a and b are the waveguide dimensions and β is the propagation constant. The
result of these simulations is a grayscale-tunable, resonant metamaterial radiator,
with a response characterized by a library of extracted αtunable values. This behavior
is illustrated in Figure 4.1, where the S parameters of this element are depicted for
several tuning values. The cELC has been offset from the center of the waveguide to
minimize the coupling with the guided wave. It should be noted that the unit cell
and waveguide have not been optimized for any particular application, but rather
this element serves as an example radiator that can demonstrate the principles of
beamforming and SAR with a metasurface aperture.
By using α, as extracted from full-wave simulations, the simulated antenna can be
used for beamforming. Following the principle of operation outlined in Section 3.1.3
and visually described in Figure 3.5, a set of αdesired can be calculated from the de-
73
2
a b
Z
Y X
Figure 4.1: Unit cell geometry used to generate a simulated metasurface antenna
for beamforming. 1) and 2) represent the waveguide port locations for full-wave
simulations. The S parameters plotted are show for four different tuning states, Tn ,
which correspond to different capacitance values [35].
sired radiation pattern (a beam in a given direction) and the guided wave. While
traditional phased array antennas would directly match the αdesired to the αtunable
with the closest phase, for metasurface apertures, Euclidean modulation is used, a
technique which matches these quantities by minimizing the Euclidean norm be-
tween them (min |αdesired − αtunable |). This operation allows the tuning method to
consider the magnitude effects of metamaterial elements and is further elucidated
in Section 4.2.1. A schematic view of the resulting aperture is shown in Figure 4.2.
These dipoles, corresponding with locations rm relative to the center of the physical
metasurface antenna, are determined by Equation (3.1), but resemble the following
equation in more explicit terms
ηm (f, rm , T ) = Hm (f, rm , T )α(f, T ), (4.2)
where Hm is the magnetic field in the waveguide as calculated in Equation (3.3) and T
is the optimized tuning state, which is a function of frequency and aperture position.
It is worth noting that Hm in this case (and correspondingly ηm ) is a function of the
tuning state of the aperture due to the fact that it is predicated upon the previous
74
Figure 4.2: Schematic view of a simulated metasurface antenna. The arrows rep-
resent the array of simulated dipole moments comprising the aperture [35].
elements’ tuning states as a function of their S21 . By propagating these effective

dipoles into the scene with the scalar free space Green’s function, the transmitted
electric field, ET , along the transverse direction, can be calculated for each synthetic
aperture position, rl [128]
X Y e−jk|r−rm −rl |
ET (f, rl , r) = A0 [H0 e−jβrm S21,m ]αm . (4.3)
m n<m
|r − rm − rl |
Here, rl is the position of synthetic aperture location, k = 2πf /c is the free space
wavenumber, and A0 is a proportionality constant. The received field, ER , can be
calculated similarly. When using a metasurface antenna in a monostatic setting, the
backscattered signal, S, can be calculated as
Z
S(f, rl ) = ET (f, rl , r)σ(r)ER (f, rl , r)dr, (4.4)
V
where σ(r) is the reflectivity of the scene and V is the scene’s volume.
4.1.2 Beamforming Performance
When calculating the beams generated by the simulated metasurface antenna in a

SAR setting, careful consideration must be given to the absolute power radiated. In
waveguide-fed structures with resonant elements, losses can be significant and differ-
ent tuning states can result in different efficiencies. To properly assess the directive
75
beams generated by the simulated metasurface antenna, they were compared with
an ideal isotropic dipole, ηisotropic , modeled as
ab
ηisotropic = H0 αmax = H0 . (4.5)
β
Here, H0 is the the incident magnetic field and |αmax | is the maximum possible po-
larizability, which represents the perfectly efficient conversion of energy in the guided
wave into radiation. The radiation patterns from the metasurface antenna were then
divided by the radiation from the isotropic source to determine the directivity for
these radiation patterns. This methodology allows for proper consideration to be
given to the absolute energy radiated by one tuning state for a constant energy in-
put to the waveguide. Doing so enables meaningful SNR comparisons, as described
in detail in Section 4.1.5. In this case, a lack of optimized feed mechanism, waveg-
uide properties, and element geometry limits the maximum realized gain from the
simulated metasurface to 8 dBi. Based on the antenna size, a linear phased array of
the same size would have an expected gain of 15 dBi [129].
Metasurfaces are dispersive by nature and small changes in input frequency result
in changes to the resulting radiation pattern [20, 130, 131]. To address this, a new
tuning state was optimized for each frequency point. These different tuning states
may be implemented in the context of a chirped RF pulse. The broadside beams
generated with the simulated metasurface antenna, each of which uses a new tuning
state, are shown in Figure 4.3.
The simulated metasurface antenna can also steer a directive beam to a col-
lection of desired angles. To do this, the same method used to create broadside
beams with the simulated metasurface antenna was used to optimize a unique tun-
ing state for each frequency point and steered angle, which spanned ±30◦ at incre-
ments of 0.5◦ . As a result of the metasurface’s dispersion, the different frequencies
76
Figure 4.3: Broadside beams generated with the simulated metasurface antenna.
Four different frequencies are shown, with a new tuning state applied at each fre-
quency [35].
and steered angles resulted in beams with variation in beam characteristics. For
example, the beamwidth varied from 7.2◦ to 10.8◦ as the steered angle/operating
frequency changed; the mean beamwidth is 9.2◦ . In Figure 4.4, steered beams are
shown for various frequencies.
4.1.3 Stripmap and Spotlight SAR Imaging
To demonstrate how metasurface antennas can be used to perform SAR imaging,

simulations of a metasurface antenna translated along a synthetic aperture path
have been conducted. To ease computational demands, a small scale system was
considered, with a synthetic aperture length of L = 30 m, sampled at 1.5 cm (∼ λ/2),
imaging at a range of d = 100 m. The bandwidth of operation (B = 9.2-9.7 GHz)
is sampled with 26 frequency points, resulting in a maximum swath size of 7.8 m
[132–135]. Additive white Gaussian noise was included in the simulations, with SNR
= -30 dB, an empirically chosen level where background noise begins to appear.
To properly account for the variation in beamwidth associated with a metasurface
antenna, Equation (2.2) should be used with φ set to be the maximum beamwidth
77
Figure 4.4: Steered beams generated with the simulated metasurface antenna. Four
different frequencies are shown, with a new tuning state applied at each frequency,
for each steered angle [35].
among frequency-dependent broadside beams produced by the metasurface antenna.

To reconstruct images from the simulated measurements, the range migration
algorithm (RMA) has been used. The RMA involves first taking the Fourier trans-
form of data measured as a function of frequency and synthetic aperture position.
Once the data is in the Fourier domain, the free space dispersion relation, in con-
junction with Stolt Interpolation, can be used to extract 2D scene information from
frequency-swept measurements [22, 112, 113, 136–138]. Since this method operates in
the Fourier domain, the speed of fast Fourier transforms (FFTs) allows for efficient
processing of large SAR data sets [136, 139].
Initial imaging simulations of stripmap mode SAR were performed with the beams
shown in Figure 4.3. The target for the first simulations was a single point scatterer
located at (100, 0, 0 m), consistent with a PSF analysis [107]. For comparison with a
78
Figure 4.5: Stripmap PSF conducted with a simulated metasurface antenna. An
analytic beam pattern is simulated for comparison with the beamwidth set to the
a) minimum, b) mean, and c) maximum beamwidth among the frequency-varying
metasurface-generated beam profiles. Here, DMA refers to simulated dynamic meta-
surface antenna [35].
conventional antenna, an analytic beam pattern (modeled as a sinc(x)=sin(x)/x func-

tion) was used [140]. In order to make a fair comparison, three different beamwidths
were simulated for the stripmap case: the minimum, maximum, and mean beamwidth
among the frequency-varying beamwidths generated by the dynamic metasurface. As
shown in Figure 4.5, the achieved resolution for the metasurface antenna (10.4 cm)
is on par with the analytically simulated beam at its maximum beamwidth (10.2
cm), thereby demonstrating the proposal that a metasurface antenna can perform
competitively.
A similar simulation was performed in spotlight mode. In this case, the synthetic
aperture size was the same as in the stripmap simulations (30 m). The resulting PSFs
are shown in Figure 4.6. In this case, the beamwidth for the analytically-modeled
beam was set to the mean of the metasurface antenna beams. Here, the cross range
resolution achieved with the metasurface antenna (4.9 cm) is nearly identical that
79
Figure 4.6: Spotlight PSF conducted with a simulated metasurface antenna. An
analytic beam pattern is simulated for comparison with the beamwidth set to the
mean beamwidth among the frequency-varying metasurface-generated beam profiles.
Here, DMA refers to simulated dynamic metasurface antenna [35].
achieved with the analytic beam profile (4.8 cm). It is important to note that for both
the stripmap and spotlight case, the metasurface architecture achieves comparable
performance as compared to a traditional beam.
Agnostic to the antenna used, there is a trade-off between resolution and field-of-
view in all SAR imaging systems. Stripmap mode can image a large region of interest
(ROI) with lower resolution, while spotlight can focus on a smaller ROI to achieve
higher resolution. To highlight this trade-off in the context of a metasurface antenna,
a collection of twelve point scatterers, spaced 1.1 m apart, was used as the target. The
synthetic aperture length, synthetic aperture spacing, and bandwidth were the same
as in the previous simulations, but the range was set to 50 m. Figure 4.7 shows that
stripmap mode is able to clearly resolve all twelve scatterers. Meanwhile, spotlight
mode incurs a dropoff in the fidelity of the reconstructed images as a function of
distance from the center of the ROI. Both modalities maintain the resolution similar
to that from the PSF analyses.
4.1.4 Alternative SAR Modalities
In this section, alternative SAR modalities are described and demonstrated. These
modalities, enhanced resolution stripmap and diverse pattern stripmap, stand to
80
Figure 4.7: Simulated line of point scatterers imaged with a simulated metasurface
antenna in spotlight and stripmap mode SAR. These images are not to scale [35].
provide added imaging flexibility as compared to traditional stripmap and spotlight.

They promise the ability to image a large ROI in high resolution, in a manner not
possible with traditional modalities. To validate these modalities, simulated imaging
results are presented. While these modalities are subject to their own drawbacks in
terms of reconstruction complexity and SNR, they can provide further means of opti-
mizing the usage of an antenna in a SAR setting. Metasurface antennas also possess
particular characteristics which make them uniquely well-suited to perform these al-
ternative modalities as compared to conventional hardware. It is worth emphasizing
that the following modalities, and the traditional ones discussed in previous sections,
may all be conducted by applying different tunings to the same metasurface antenna.
Enhanced Resolution Stripmap
The first alternative modality examined here is enhanced resolution stripmap, which
uses a sweep of steered angles at every synthetic aperture location. This approach
allows each point in the scene to be probed with a wider range of k-components,
thereby increasing resolution. This is in contrast to spotlight in that this operation
does not involve focusing the steered angles towards a particular spot on the ground.
81
As a result, there is no restriction on the ROI size, leading to an imaging modality
which can increase the resolution without sacrificing ROI.
In this modality, the range of steered angles used can vary. If the range of steered
angles matches that used for spotlight imaging, the resolution will also match that
of spotlight imaging. Meanwhile, if the range of steered angles is smaller, it can
more closely approximate stripmap mode imaging, ultimately availing an imaging
modality which is more flexible than either stripmap or spotlight. This method is
illustrated schematically in Figure 4.8. Furthermore, a multitude of approaches for
choosing the steered angles may considered. A regular sweep of available steered
angles as a function of frequency may be used, or a random steered angel could be
chosen per frequency point; a more in-depth investigation into the optimal method
for applying the steered angles for enhanced resolution would be required to achieve
the best performance.
Simulated studies conducted with the metasurface antenna described in Sec-
tion 4.1.2 are used to evidence the benefits of enhanced resolution stripmap. In this
scenario, a different steered angle was used for each frequency in order to preserve
the same number of measurements as other modality simulations. Other simulation
parameters including the bandwidth/frequency sampling, synthetic aperture path,
etc. remained consistent with the simulations from Section 4.1.2. A PSF analysis
was performed and it can be seen that enhanced resolution stripmap can match spot-
light in terms of cross range resolution, as shown in Figure 4.9a (see Figure 4.6 for
comparison). Meanwhile, imaging a line of targets shows that enhanced resolution
stripmap can achieve a similar ROI to stripmap mode.
Diverse Pattern Stripmap
In addition to enhanced resolution stripmap, another alternative modality, termed

diverse pattern stripmap, can be realized using DMAs. In this method, which is in-
82
Figure 4.8: Schematic view of enhanced resolution stripmap. Here, each color
represents a different steered angle, where each angle is probed from each synthetic
aperture location [35].
spired by computational imaging systems [75,107,116], diverse radiation patterns are

used in place of traditional, directive beams. These beam profiles do not necessarily
exhibit a distinct main lobe, but instead include multiple lobes. The overall patterns
span a wide collective beamwidth, which allows for high resolution by probing a
wide range of k-components. This method is described visually in Fig. 4.10. Similar
to enhanced resolution stripmap, diverse pattern stripmap stands to provide gains
as compared to traditional stripmap by providing high resolution while maintaining
the ability to image a large ROI. Furthermore, since this modality has less stringent
requirements regarding beam characteristics, a larger bandwidth may be used due to
the fact that distortion of the radiated beam profiles does not degrade image quality.
When implementing diverse pattern stripmap, the nature of the diverse radia-
tion patterns must be considered. Ultimately, the desired patterns would span a
83
Figure 4.9: Simulated images conducted with a metasurface antenna in enhanced
resolution stripmap mode. a) and b) show a PSF, while c) shows imaging of a line
of point scatterers. c) is not shown to scale [35].
wide overall beam envelope and the patterns would be as orthogonal in space as
possible. Further, the patterns would also exhibit moderate directivity among the
lobes comprising each pattern in order to maintain a high SNR. Due to losses in
the waveguide and variation among radiation efficiency as a function of tuning state,
different tuning styles can lead to radiation patterns which do not effectively couple
energy out of the waveguide, which leads to low SNR. To generate more deliberate
diverse patterns, beam profiles may be crafted with multiple lobes, each directed at
a random angle among the range of steered angles used for spotlight imaging.
The simulated metasurface antenna described in Section 4.1.1 was used to verify
the utility of diverse pattern stripmap. The diverse radiation patterns used for these
studies were generated by creating radiation patterns with three main lobes, each
at a random angle chosen within the ±30◦ range used for spotlight imaging. Three
84
Figure 4.10: Schematic view of diverse pattern stripmap. Radiation patterns used
here exhibit three main lobes, each at a random angle [35].
Figure 4.11: Three sample radiation patterns optimized for diverse pattern
stripmap. Each has been generated by performing Euclidean modulation as a means
of generating beam profiles with three main lobes within the range of steered angles
described in Figure 4.4. The gray dashed line shows a broadside beam for refer-
ence [35].
85
sample radiation patterns in this style are shown in Figure 4.11. To generate beams in
this style, Euclidean modulation was used, with the desired surface current set equal
to the average of the surface currents corresponding to the three individual steered
beams. At each synthetic aperture position, a different diverse pattern was used
per frequency, cycling through a set of unique diverse patterns across the frequency
sweep.
In a similar manner to enhanced resolution stripmap, diverse pattern stripmap
can also enable high resolution imaging across a large ROI. To support this claim,
a PSF analysis was performed and the resulting reconstruction is shown in Figure
4.12a. Here, the cross range resolution achieved by diverse pattern stripmap matches
spotlight imaging (see Figure 4.6 for comparison). Further, diverse pattern stripmap
can also image across a large ROI, as evidenced by the ability of this method to
reconstruct a line of point scatterers in a manner consistent with stripmap mode
imaging (see Figure 4.7 for comparison). Overall, like enhanced resolution stripmap,
diverse pattern stripmap can achieve high resolution across a large ROI.
Both enhanced resolution stripmap and diverse pattern stripmap have demon-
strated the ability to image a large ROI in high resolution. Depending on the ap-
plication or environment, however, these two modes present different benefits which
may guide the decision as to which among these methods would be preferable. En-
hanced resolution stripmap can reduce hardware requirements by scanning through a
range of steered angles as a function of frequency. This reduces the required switching
speeds and memory required for the system. Meanwhile, by probing a larger portion
of the scene with each measurement, diverse pattern stripmap is more likely to enable
the reconstruction of high-quality images from a smaller set of measurements, in a
manner consistent with many compressed imaging systems [75].
86
Figure 4.12: Simulated images conducted with a metasurface antenna in diverse
pattern stripmap mode. a) and b) show a PSF, while c) shows imaging of a line of
point scatterers. c) is not shown to scale [35].
4.1.5 Noise Analysis
Enhanced resolution stripmap and diverse pattern stripmap can both provide added
imaging flexibility, but can be subject to drawbacks due to their nontraditional il-
lumination strategies. These nontraditional illumination strategies probe different
portions of the scene from different synthetic aperture positions as compared to con-
ventional spotlight and stripmap. Particularly in the case of diverse pattern stripmap
(with no main lobe), there are questions as to how robust these alternative imag-
ing modalities are to noise. Further questions regarding SNR stem from the way
that the tuning state of a metasurface antenna affects its efficiency. In the context
of a metasurface, beamforming is entirely achieved by implementing an optimized
tuning state to each element in the aperture. For these tuning states, a complex
polarizability is set by the frequency and tuning state, which varies in magnitude
87
between one element and the next. As a result, when the exciting field H travels
through the waveguide, the decay is determined by the particular tuning state of
the overall aperture, meaning that different tuning states can incur more or less loss
in the waveguide. Metasurface antennas therefore radiate a different portion of the
feed energy as a function of tuning state leading to a different magnitude of returned
signal, as calculated in Equation (4.4).
In order to investigate this aspect of these systems, simulated imaging studies
with a constant noise level were performed. In these simulations, the target was
the line of scatterers described in Section 4.1.4. Additive white Gaussian noise,
as in the previous simulations, was included, but with higher noise power which
remained constant across simulations; this is in contrast to previous results, which
used a constant SNR. Since the noise power remained constant, irrespective of signal
power, an analysis of the relative impact of the noise on the signal can be performed
to compare the different imaging modalities. From the simulated signal resulting
from each modality, SNR is calculated as
PS |Sµ (f, rl )|2

SN R = 10log10 = 10log10 , (4.6)
PN PN
where PS is the average signal power, PN is the average noise power, and Sµ (f, rl ) is
the mean signal.
The resulting reconstructed images are shown in Figure 4.13, which also includes
the SNR levels calculated from Equation (4.6) for each modality. From these images,
it can be seen that spotlight is the most robust to noise, as expected, achieving
SNR = -26.6 dB. Enhanced resolution stripmap (SNR = -33.0 dB) results in similar
robustness to noise as compared to traditional stripmap (SNR = -32.4 dB), while
diverse pattern stripmap incurs slightly worse SNR (SNR = -35.1 dB). It is important
to note that these SNR values are target dependent, so a different environment/scene
88
Figure 4.13: Noise study conducted by calculating the resulting SNR from different
SAR modalities. a), b), c), and d) correspond with stripmap, spotlight, enhanced
resolution stripmap, and diverse pattern stripmap, respectively. For each simulation,
the target and noise level remained constant. Not to scale [35].
could alter these results. This phenomenon results from the fact that the received
noise power is typically constant throughout a measurement set, but the signal power
is a function of the average reflectivity of the scene. This study further highlights the
difficulty associated with optimizing metasurface performance and that by availing
additional control knobs in the form of alternative modalities, metasurface antennas
stand to offer added imaging flexibility as a result of their agile nature.
4.2 Metasurface Tuning Strategies
When beamforming with a metasurface antenna, applying traditional tuning meth-

ods can limit performance due to the differences between traditional radiating ele-
ments and metamaterial elements. The most significant differences are that meta-
89
material elements exhibit less than half of the phase control of traditional antenna
elements and that they have an inherently coupled magnitude and phase response. To
compensate for these differences, Lorentzian-constrained modulation and Euclidean
modulation have recently been proposed as tuning strategies which consider the na-
ture of metamaterial radiators to improve performance. These methods have demon-
strated excellent performance from analytically modeled metasurface antennas, but
can substantially limit performance in a practical setting by invoking strong losses
in the guided wave [50, 103].
To overcome the limited analysis possible with analytic modeling, full-wave sim-
ulations are used to accurately model losses associated with metamaterial radiators.
Chiefly, there are two types of loss that are important to capture from full-wave
simulation results: Ohmic and lumped element losses directly associated with each
metamaterial element and the guided wave losses resulting from the scattering off
of each element. Both of these loss mechanisms can be captured by the full wave
simulation of one metamaterial element. This approach does not require the full
wave simulation of an entire metasurface antenna, but nonetheless captures the loss
behavior more accurately than analytic estimation, following Equation 3.3, which
was described in detail in Section 3.1.4.
To properly apply the scattering losses, a perturbative guided wave model is
used and an iterative tuning state optimization technique to accurately capture the
phenomena inside the waveguide. Applying various tuning methods with this more
accurate model reveals a different estimation of performance from previous modeling
methods, often showing a lower maximum gain due to the rapid decay of the guided
wave. Previous analyses considered a lossless or bulk loss model [39,103]. By analyz-
ing these methods in the context of an attenuating guided wave, the effect of method
on loss is highlighted in the context of a waveguide-fed metasurface antenna. From
this analysis, these different tuning methods affect the guided wave differently, with
90
a traditional phase hologram tuning strategy liable to cause poor efficiency, while
Lorentzian-constrained modulation can lead to a small effective aperture size.
Following the results described above, a modified version of Euclidean modula-
tion, scaled Euclidean modulation, can grant control over the guided wave necessary
to realize high gain from a metasurface antenna. Scaled Euclidean modulation can
be tailored to a specific metasurface antenna’s parameters to optimize performance.
Through numeric simulations, scaled Euclidean modulation realizes high efficiency
and a large effective aperture, ultimately improving the overall antenna performance
as compared to other methods.
4.2.1 Tuning Strategy Definitions
Among the responses achievable with a Lorentzian element, the range of available
phase values is −π < φ < 0. As compared to using phase shifters, this is only
half of the control. Additionally, shifting the resonance simultaneously shifts the
magnitude and phase response, which further limits the control available to meta-
material elements. These differences between metamaterial radiators and traditional
antenna elements lead to challenges when beamforming with a metasurface antenna.
To address these challenges, a traditional holographic antenna tuning strategy—a
phase hologram— is described before delving into two metasurface-specific methods:
Lorentzian-constrained modulation and Euclidean modulation. While the methods
presented in this section are applied to form a steerable, directive beam, these ap-
proaches could also be formulated to craft arbitrary radiation patterns. It should
also be noted that these methods have been derived in the context of a lossless guided
wave, following H = H0 e−jβxm , though they will later be applied in simulation with
a lossy, perturbative model of the guided wave. These methods have been derived
assuming an analytically modeled α, following Equation 3.2.
91
Phase Hologram
Applying a desired phase profile across an aperture allows phased array antennas to
generate arbitrary radiation patterns. If the elements exhibit a phase profile following
ejkx sin θs , a plane wave will be generated at steered angle θs . Interfering this desired
profile with the incident wave yields the desired set of polarizabilities. To apply the
phase hologram tuning strategy to a metasurface, the tuning state of each element is
selected to minimize the phase difference between the set of available polarizabilities
and the calculated desired polarizability.
αd = ejxm (β+k sin θs )

(4.7)
min |6 αt − 6 αd |
This minimization is repeated for each element within the metasurface to create a
tuning state for each element in the aperture. A visualization of this mapping in
the complex plane can be seen in Figure 4.14, which shows how αd (blue) is mapped
to/approximated by αt (red). It should be noted since a Lorentzian polarizability is
restricted to lie within −π < 6 α < 0, half of the elements cannot reach 6 αd . When
6 αd cannot be reached, the closest points in αt tend to be those with the lowest
|α|, sending the magnitude of those elements to near zero. This phenomenon results
in more than half of the elements radiating weakly, leading to low efficiency, and
introduces an unwanted, periodic magnitude profile — this in turn introduces strong
grating lobes.
Lorentzian-Constrained Modulation
Lorentzian-Constrained modulation is a metasurface-specific mapping which jointly

considers the magnitude and phase response of each element in the metasurface.
Lorentzian-Constrained modulation was developed by combining antenna theory
92
1 1
d
t
| |
0.5 0.5
Imag( )
0
0
2
-0.5 0
-2
-1
-1 -0.5 0 0.5 1 0 0.5 1 1.5 2 2.5 3 3.5
Real( ) g
Figure 4.14: A visualization of the phase hologram tuning strategy. The left plot
shows a discretized version where each point in the waveguide has a calculated αd
(blue dot), which is mapped to αt (red dot) in correspondence with the black arrows.
The magnitude and phase of αd and αt are plotted continuously on the right.
with an analytic definition of a metamaterial response while attempting to elimi-

nate possible grating lobes.
xm
αd = ej 2
(β+k sin θs )
(4.8)
min |6 αt − αd | 6
Unlike the phase hologram approach, which incurs unwanted magnitude effects,
Lorentzian-constrained modulation can optimize the phase and magnitude response
together to suppress unwanted radiation and concentrate more energy into the main
lobe. Applying Lorentzian-constrained modulation, following Equation (4.8), leads
to the mapping in the complex plane shown in Figure 4.15. While the agreement
between 6 αd and 6 αt is worse than the phase hologram approach, the magnitude
fluctuations are less pronounced and better efficiency can be realized.
Euclidean Modulation
Euclidean modulation is an approach which maps from the desired polarizability val-
ues to the available values by minimizing their Euclidean norm. Euclidean modula-
tion thereby also jointly considers the magnitude and phase response of metamaterial
93
1 1
d
t
| |
0.5 0.5
Imag( )
0
0
2
-0.5 0
-2
-1
-1 -0.5 0 0.5 1 0 0.5 1 1.5 2 2.5 3 3.5
Real( ) g
Figure 4.15: A visualization of the Lorentzian-constrained modulation tuning

strategy plotted as described in Figure 4.14.
elements.

(4.9)
min |αt − αd |
Following this mapping results in a compromise between reducing the amount of

non-radiating elements and achieving a close approximation to the desired phase
profile in order to beamform. This method is plotted visually in Figure 4.16, which
shows this mapping between the αd and αt .
1 1
d
t
| |
0.5 0.5
Imag( )
0
0
2
-0.5 0
-2
-1
-1 -0.5 0 0.5 1 0 0.5 1 1.5 2 2.5 3 3.5
Real( ) g
Figure 4.16: A visualization of the Euclidean modulation tuning strategy plotted

as described in Figure 4.14.
94
4.2.2 Simulation Methodology
While metamaterial elements follow a Lorentzian polarizability, practical details such

as specific geometry, losses, and waveguide characteristics set the physically realiz-
able polarizability exhibited by a given element. To capture the effects of these
characteristics, full-wave simulations of an individual metamaterial element may be
conducted to extract the complex S parameters, which lead to an estimate of an
element’s polarizability as described in Equation (4.1).
In this work, a simulated metasurface antenna is considered which comprises of
101 metamaterial elements, spaced at 7.5 mm (∼ λ/4), embedded within a dielectric-
filled (r = 2.2) rectangular waveguide. The metamaterial element design (shown
in Figure 4.17) uses varactor diodes to provide grayscale control which can shift
the resonance of an element with a voltage bias. The response of the simulated
metamaterial element is plotted in Figure 4.17, which shows the S parameters and
extracted polarizability for different tuning states at a constant frequency, f = 10.3
GHz. The antenna parameters are summarized in Table 4.2.
In Section 4.2.1, each of the tuning strategies has been derived in the context of
an unperturbed guided wave, with a constant β. Assuming that β is a constant is
common practice for many traveling wave antennas, but scattering from each element
Figure 4.17: Metamaterial element geometry modeled with a full wave solver. The
right side plots show the S parameters and the extracted polarizability as a function
of resonant frequency/tuning.
95
simulate beamforming.
Parameter Value
# of elements 101
Spacing 7.5 mm
Frequency 10.3 GHz
perturbs the guided wave and introduces local phase shifts to the field incident upon
each element. In some metasurface architectures, this effect can be particularly pro-
nounced when each element is set to a different tuning state and scatters differently
according to its complex S21 .
Here, the perturbative model of the guided wave described in Equation (3.3) is
used which considers the S parameters of each element in the antenna as a function
of frequency and tuning state. Following this definition of H, each of the tuning
strategies described in Section 4.2.1 must be applied sequentially. The tuning state
for one element must be calculated, using the incident Hm , and then the field incident
upon the next element, Hm+1 must be calculated from Hm , along with the complex
S21 as a function of the tuning state applied to element m.
Model Verification
As a brief aside, validation of the described modeling method is provided using full-
wave simulation. As mentioned above, the primary method for modeling metasurface
antennas used in this work involves conducting a full-wave simulation of a single
element, extracting a library of element responses, then mathematically constructing
an antenna by patterning a series of elements. This method constitutes an efficient
platform for simulating metasurface antenna designs with different element spacing,
tuning states, and operating frequencies. However, certain phenomena not considered
by the modeling method have the capacity to diminish its accuracy. For example,
96
20
Matlab
CST
Matlab - lossless
10
Gain (dB)
0
-10
-20
-90 -60 -30 0 30 60 90
(deg.)
Figure 4.18: Comparison between the farfield response of a metasurface antenna
simulated with a full wave solver and the numerical modeling method described in
Section 3.1.4.
inter-element interactions are not considered, which can be substantial if elements

couple strongly. The expectation is that these effects will not be significant due to
the element geometry at hand being weakly coupled.
As mentioned earlier, simulating full-sized metasurface antennas is impractical
due to the computational demands. Nevertheless, a simulation of a small, sample
antenna was simulated as a means of validating the modeling method. The sample
antenna operates at 10.3 GHz and has 11 elements, spaced at 7.5 mm. To compare
the mathematical model to full-wave simulations, the sample antenna was first sim-
ulated in Matlab, using the mathematical model described above. This simulation
included optimizing the tuning state using Euclidean modulation, calculating the
guided wave profile, and determining the far field radiation pattern of the antenna.
Meanwhile, the optimized tuning state determined from the Matlab calculation was
loaded into CST Microwave Studio and the antenna was simulated to determine the
far field pattern. The far field pattern for each of these calculations are plotted below
in Figure 4.18, where good agreement can be seen between the two methods. The
97
Figure 4.19: Steered, directive beams generated with different tuning strategies.
Each method is able to form a beam steered to the corresponding desired θS .
strong agreement between the two simulation methods indicates that the assump-
tions used for the modeling method are valid and that results from the mathematical
simulations constitute good predictions. Figure 4.18 also includes a Matlab simu-
lation which does not include the perturbative loss model described above. From
the disagreement between the lossless result and the CST simulation result, an ac-
curate loss model, as described in Equation (3.3), is essential to obtaining realistic
performance predictions.
4.2.3 Simulated Beamforming Results
To highlight their unique advantages and disadvantages, each tuning strategy de-
scribed in Section 4.2.1 was applied to the simulated antenna mentioned in Sec-
tion 4.2.2 to form steered, directive beams. This process involves sequentially calcu-
lating the guided wave at each element, interfering that field with the desired dipole
moment (ηd ), and determining a tuning state from the resulting αd according to the
tuning strategy being used. The realized dipole moments are then propagated to
determine the farfield pattern. Figure 4.19 shows that each method is capable of
forming a directive beam steered to θs = −15◦ , 0◦ , and 15◦ .
In Figure 4.19, it is worth noting that there appears a large grating lobe when the
beam is steered away from broadside with a phase hologram due to periodic magni-
98
tude fluctuations. This phenomenon occurs strongly for the phase hologram tuning
strategy due to the fact that this method does not consider the magnitude effects
of tuning a metasurface. Meanwhile, Lorentzian-constrained modulation considers
the coupled tuning associated with metasurface antennas and therefore can minimize
this effect. Euclidean modulation, whose mapping is in between these two methods,
therefore exhibits smaller grating lobes than the phase hologram method.
4.2.4 Scaled Euclidean Modulation

Effective Aperture Size
When applying the tuning methods described in Section 4.2.1, losses in the waveg-
uide can severely limit the effective aperture size and the uniformity of illumination
associated with the radiating elements. Since a larger effective aperture size results
in higher directivity, properly controlling the decay of the guided wave can unlock
higher performance. These effects are governed by the scattering of each metamate-
rial element, with lower |S21 | corresponding with higher |α|. If the average |α| set by
the tuning state is too high, the guided wave decays too quickly, resulting in a small
effective aperture. To highlight how each tuning strategy affects the gain realized
from the simulated metasurface antenna, the peak broadside gain for each method
is summarized in Table 4.3.
From the mapping plots in Figure 4.14, 4.15, and 4.16, each tuning method tends
towards a different average coupling strength, which is directly related to |α|. The
coupling strength of each of these methods is somewhat captured by the mean value
of α, which is included in Table 4.3. The methods which couple stronger decay the
guided wave more quickly due to the relationship between α and S21 , in which high
|α| values correspond with low |S21 | values. To explicitly plot this effect, |H| and
|η| are plotted for each method in Figure 4.20. In the case of Lorentzian-constrained
modulation, the guided wave decays quickly, dropping below 1% of its initial energy
99
1
Phase
0.8 Lorentzian-Constrained
Euclidean
0.6
|H|
0.4
0.2
0.2
0.15
| |
0.1
0.05
0
0 0.2 0.4 0.6
Position (m)
Figure 4.20: The top plot shows the decay of the guided wave along the waveg-
uide for each tuning method. The bottom plot shows the resulting dipole moments
(analogous to the antenna weights).
after traveling approximately 0.3 m, leaving more than half of the antenna unused.
Meanwhile, the phase hologram approach has a lower average |α| and therefore can
realize a larger effective aperture size, which can be seen by the maximum gain
realized in the plots shown in Figure 4.19. Furthermore, if the guided wave decays
too quickly, a nonuniform illumination of the elements in the aperture results, which
can be seen from the magnitude of each dipole moment, plotted in Figure 4.20.
100
Table 4.3: Broadside gain for three different tuning strategies applied to a metasur-
face antenna.
Tuning Method Gain (dBi) Mean(|α|)

Phase 12.6 0.39
Lorentzian-Constrained 7.9 0.65
Euclidean 10.3 0.53
Scale Factor Definition
If the decay of the guided wave can be reduced, an antenna’s available aperture size
can be better used to realize high gain. Introducing a scaling coefficient within the
Euclidean modulation algorithm allows for control over the average coupling of the
metamaterial elements to the guided wave. Dividing αd from Equation (4.9) by a
scaling coefficient (A) can grant added control over the tuning states selected when
mapping from αd to αt .

αd (4.10)
min αt −

A
By increasing A above 1, the tuning states selected tend towards the lower magnitude
states, leading to a higher average |S21 |, which in turn leads to a larger effective
aperture size. The mapping realized by applying a scaling coefficient can be seen in
Figure 4.21, which shows the mapping from αd to αt for different values of A.
To see the benefit of applying the scaling factor to Euclidean modulation, the
aperture described in Section 4.2.2 was simulated for different values of A. The
radiation pattern for generating a broadside beam with A = 1, 1.5, and 3 is shown
in Figure 4.22. Applying a scaling coefficient of 1.5 provides a modest increase in
effective aperture size as compared to A = 1. Using A = 3, however, allows for a
much larger effective aperture, which increases from approximately 0.3 m to 0.75
101
1 1
a) d
| |
0.5 t 0.5
Imag( )
0
0
-0.5 0
-1 -
1 1
b) d
| |
0.5 t 0.5
Imag( )
0
0
-0.5 0
-1 -
1 1
c) d
| |
0.5 t 0.5
Imag( )
0
0
-0.5 0
-1 -
-1 -0.5 0 0.5 1 0 1 2 3
Real( ) guided
Figure 4.21: A visualization of the scaled Euclidean modulation tuning strategy

for different values of A, plotted as described in Figure 4.14.
m as A goes from 1 to 3. The effect of the larger aperture size is apparent from
Figure 4.22, which shows the higher gain from A = 3. The broadside gain of the
metasurface antenna is 14.8 dB when scaled Euclidean modulation is applied (A = 3),
while it is only 10.2 dB when unscaled, representing an improvement of 4.6 dB.
Scale Factor Optimization
When choosing A, if the scaling is too low, the guided wave will decay before reach-
ing the end of the aperture, reducing the effective aperture size and resulting in a
102
15 A=1
A=1.5
10 A=3
Gain (dB)
5
0
-5
-10
-15
-60 -30 0 30 60
(deg.)
Figure 4.22: Broadside beams generated with scaled Euclidean modulation for
A = 1, 1.5, and 3.
nonuniform illumination pattern. On the other hand, if the scaling coefficient is too
high, the phase states available to the optimization algorithm are limited. Addition-
ally, a high scaling coefficient can cause energy to be left over when the guided wave
reaches the end of the antenna, which will hurt the antenna’s efficiency.
A must be optimized to realize the best performance, with the specific meta-
surface antenna architecture determining the optimal value of A. In particular, the
length of the metasurface antenna, in conjunction with the element behavior and
spacing, can determine the optimal value for the scaling coefficient. To investigate
the consequences of changing A, a series of values was simulated versus antenna
length, with the peak broadside gain as the primary performance metric. The re-
sults of this study are plotted in Figure 4.23. From this result, one can see how
the optimal scaling coefficient changes with antenna length. As the antenna length
increases, so too does the optimal value of the scaling coefficient. This trend results
103
Max. Gain (dB)
20
2
15
Antenna Length (m) 1.5
10
1 5
0
0.5
-5
-10
2 4 6 8 10
Scale Factor
Figure 4.23: Broadside gain from a metasurface antenna tuned with scaled Eu-
clidean modulation for different lengths and scale factors (A). The blue star shows
the gain for Euclidean modulation; the red star shows the increase in broadside gain
associated with applying the optimal scale factor.
from the fact that as the antenna gets larger, the guided wave must retain more
energy in order to excite elements far from the feed. From this plot, the optimal
value of A can be quickly determined. For example, if an antenna of length 1 m is
considered, using A = 3.45 (the optimal value from Figure 4.23), the gain improves
from 12.6 dB to 15.9 dB by using scaled Euclidean modulation instead of the phase
tuning strategy (which yielded the best results initially according to Table 4.3). This
is consistent with a drop in the average |α|, which drops to 0.2, half of that which
resulted from the phase tuning method used above. It should also be noted that this
optimization technique is useful for designing both small and large antennas; if a
particular size antenna is desired, this method can assist in balancing efficiency and
antenna size.
104
4.3 Large Scale Simulated Metasurface
In this section, a large scale metasurface antenna for spaceborne SAR is demon-
strated. The goal of these studies is to determine whether a metasurface antenna
design can match the performance of a spaceborne SAR phased array. To this end,
the overall size and operating bandwidth match that of TerraSAR-X. Through a se-
ries of simulated studies, the metasurface design proves to be equally capable in some
ways and superior to TerraSAR-X in other ways. These studies have been conducted
with array factor calculations, following Section 2.2.1.
4.3.1 Initial Spaceborne Metasurface Design
To start designing metasurfaces for spaceborne SAR, a large area must be covered.
In Section 2.2.1, the TerraSAR-X phased array is 4.8 m by 0.7 m, dimensions which
will be used for the large scale metasurface designs discussed in this section. The
initial metasurface design used to fill this space uses a series of 1D waveguides. Each
waveguide is 0.03 m by 0.7 m, spanning the full size of the antenna in elevation.
160 antennas, arrayed in the azimuth direction, fill the full antenna size, as shown
visually in Figure 4.24. The waveguides are fed from the center and the 70 elements
are spaced at 10 mm. These parameters are summarized in Table 4.4.
Table 4.4: Antenna parameters for the initial large scale spaceborne metasurface
antenna.
Parameter Value
Antenna size 4.8 m by 0.7 m
# of feed layers 0
Elements per waveguide 70
# of waveguides 160
# of RF sources 160
Total tunable elements 11,200
Element spacing 10 mm
105
10 mm
Figure 4.24: Large scale metasurface antenna initial design.
In this design, the metamaterial elements are assumed to be dynamically tunable

with grayscale control. Varactor diodes are one method for realizing this control, but
the analysis here remains agnostic the specific element geometry. Since the operating
bandwidth is 9.5 - 9.8 GHz, that range is used as the tuning range, as outlined in
Figure 4.25. To tune the metamaterial elements, the holographic procedure outlined
in Section 3.1.3 is followed, leading to a set of αdesired . Then Euclidean modulation
(as outlined in Section 4.2) is used to determine the tuning state of each element.
The beamsteering performance of the initial metasurface antenna design are
shown in Figure 4.26 and Figure 4.27. The results, which are compared with
TerraSAR-X, show that the metasurface antenna is capable of steering in azimuth
and elevation. In azimuth, the metasurface design exhibits improved performance as
compared to TerraSAR-X, due to the lower grating lobes in the metasurface radia-
tion patterns. For TerraSAR-X, when steering to the quoted limits (±0.75◦ ), grating
lobes appear, even reaching as high as 36 dB. These grating lobes appear due to the
fact that TerraSAR-X only has phased shifters every 40 cm.
106
Figure 4.25: Analytically modeled metamaterial element response for large aper-
ture beamforming simulations.
TerraSAR-X
Metasurface
Figure 4.26: Azimuth steering capabilities of the initial large metasurface antenna
design.
107
TerraSAR-X
Metasurface
Figure 4.27: Elevation steering capabilities of the initial large metasurface antenna
design.
In elevation, however, the performance of TerraSAR-X is improved, while the

metasurface antenna suffers from strong grating lobes. Several grating lobes appear,
especially as the beam is steered to 20◦ , reaching as high as 34 dB. These grating
lobes in the metasurface design do not appear due to a sparsely sampled layout,
but instead appear due to the incomplete element control afforded by metamaterial
elements.
108
4.3.2 Final Large Metasurface Design
In this section, the focus is on developing the metasurface antenna design further
to consider several practical features and to reduce the grating levels identified in
the initial metamaterial design approach. One method identified for reducing the
grating lobes was to use a denser array of elements. By reducing the separation
between elements, better sampling of the guided wave can be realized to achieve a
better approximation of a desired phase profile across the aperture. Dense sampling
causes an increase in cost (due to more semiconductors being required), but can
improve performance by reducing grating lobes and improving efficiency.
In addition to moving towards denser element spacing, other details associated
with large scale antenna design are also considered. Specifically, the question of how
to feed each waveguide/element in the antenna array is addressed with two waveguide
feed layers. These two waveguide feed layers, shown visually in Figure 4.28, provide a
Figure 4.28: Final large metasurface design for spaceborne SAR.
109
Table 4.5: Antenna parameters for the final large scale spaceborne metasurface an-
tenna.
Parameter Value
Antenna size 4.8 m by 0.7 m
# of feed layers 2
# of waveguides (elevation) 23
# of waveguides (azimuth) 160
# of RF sources 0
Total tunable elements 22,080
means of distributing energy to each element without using so many independent RF

sources. In this architecture, the radiating waveguides are separated into 23 divisions,
each being 3 cm long. Each of these radiating waveguides holds 6 metamaterial
elements, each spaced at 5 mm (λ/6), and is fed from the center. The lower feed
structure distributes energy in the elevation direction. The middle feed layer takes
energy from the lower feed structure and distributes along the azimuth direction.
Using a feed structure provides a means of reducing the number of independent RF
sources, reducing cost, and can provide structural rigidity in the process.
The beamforming results for the final metasurface antenna design are shown in
Figure 4.29 and Figure 4.30. From these results, the metasurface demonstrates su-
perior performance in azimuth and elevation. When steering to the quoted limit in
azimuth, 0.75◦ , TerraSAR-X has a large grating lobe, while the metasurface does
not. Further, if steered beyond 0.75◦ , the metasurface continues to exhibit good
beamforming performance, while the grating lobe from TerraSAR-X becomes debil-
itating, reaching as high as 40 dB. If steered substantially further, the metasurface
would eventually realize a grating lobe in azimuth due to the fact that this design
uses a waveguide width of 3 cm (∼ λ), meaning that the independent control points
110
TerraSAR-X
Metasurface
Figure 4.29: Azimuth steering capabilities of the final large metasurface antenna
design.
TerraSAR-X
Metasurface
Figure 4.30: Elevation steering capabilities of the final large metasurface antenna
design.
111
Figure 4.31: 2D beam patterns for the final large scale metasurface design.
in azimuth are sampled above λ/2. In elevation, the metasurface antenna also ex-
ceeds the performance of TerraSAR-X. Unlike the initial design, this metasurface
antenna does not exhibit strong grating lobes. Steering beyond the quoted limits
(±20◦ ), TerraSAR-X begins to exhibit strong grating lobes, reaching nearly 40 dB
112
when steering to 30◦ . Meanwhile, the metasurface, which is sampled at λ/6 in el-
evation, does not exhibit grating lobes when steered to 30◦ . The final metasurface
design performance is further highlighted in Figure 4.31, which shows 2D beam pat-
terns generated with the final large scale metasurface design.
In this section, the final metasurface antenna design has shown the ability to
match the performance of TerraSAR-X. With the opportunity to exhibit improved
hardware characteristics, metasurface antennas are poised to act as a suitable alter-
native to phased array antennas. In addition to validating their use in spaceborne
SAR systems, the studies presented in this section also serve to highlight a major
distinction between metasurface antennas and phased array antennas. In a phased
array antenna, as described in Section 2.2.1, phased array antennas face a trade-
off between cost and performance when choosing the number of independent RF
sources/phase shifters. In TerraSAR-X, this led to a design with phase shifters every
40 cm in azimuth, limiting the steering range to ±0.75◦ . If the steering range were
to be increased, more phase shifters would be required, driving up the cost of the
antenna. With a metasurface architecture, this tradeoff does not exist — metamate-
rial elements will always provide dense control through individual control over each
radiator. Avoiding this tradeoff affords metasurface antennas the possibility to steer
to extreme angles without a corresponding cost increase.
4.4 Grating Lobe Suppression
Metamaterial elements exhibit incomplete magnitude/phase control. Control is lim-

ited in phase to cover only half of the states available to a phase shifter (π rather than
2π). Further, tuning the resonance of a metamaterial element affects both the mag-
nitude and phase response, following Equation (3.2). Section 4.2.1 describes some of
the limitations associated with applying traditional tuning strategies to metasurface
antennas. When tuning the phase of a series of metamaterial elements, unwanted,
113
periodic magnitude oscillations can arise, leading to strong grating lobes. In some
metasurface antenna architectures, grating lobes can be avoided by using high di-
electrics and dense element spacing, but these approaches limit efficiency. It would
be more desirable to avoid these requirements because high dielectrics are costly and
incur high loss, while dense element spacing complicates modeling methods. In this
section, an alternative strategy for grating lobe suppression — using phase diversity
in the feed — is described and demonstrated.
4.4.1 Metasurface Antenna Architecture
The waveguide-fed metasurface illustrated in Figure 4.32a consists of an array of

metamaterial irises patterned into one side of a waveguide [35,103]. The metamaterial
elements couple energy from the waveguide mode to free space as radiation [46].
The overall radiation pattern is then the superposition of each element’s radiation,
whose complex amplitude is a function of the element’s response and the input
frequency. By altering each metamaterial element’s properties, or by changing the
driving frequency, different radiation patterns can be generated from the aperture. To
realize an electrically large 2D antenna, 1D waveguide-fed metasurface apertures can
be patterned side-by-side to form a 2D metasurface antenna array as in Figure 4.32c.
To analyze a 2D metasurface antenna analytically, first the constituent metama-
terial elements are modeled. In this work, the elements are assumed to be weakly
coupled, such that inter-element coupling from the elements within the waveguide
can be ignored. Other modeling methods can account for these effects, but are not
used in this work [122]. The simplified model used here facilitates an array factor
analysis which provides sufficient insight into the grating lobe problem associated
with metasurface antennas.
Each metamaterial element can be modeled as a point dipole with a response
dictated by the incident magnetic field and the element’s complex polarizability,
114
a)
b) c)
y
x
Figure 4.32: Diagram of an example metasurface antenna architecture. a) shows

the operation of a holographic metasurface antenna, with the reference wave (shown
in blue) interfering with the metamaterial elements which form the hologram [35].
b) shows a sample metamaterial element and the surrounding conductor (shown in
gold), along with the tunable component (shown in grey). c) shows a sample antenna
layout consisting of eight individually fed rectangular waveguides, each with twenty
radiating elements.
following Equation (3.2). While polarizability is a tensor quantity, these elements

are assumed to be only polarizable in the x direction and thus use a scalar approx-
imation [45, 122]. For this analysis, η follows Equation 3.1 and H is modeled as
Hn = H0 e−jβyn , where β is the waveguide constant and yn is the position measured
from the origin, as shown in Figure 4.32. To keep the analysis general, each of
the metamaterial elements are modeled with an analytic, Lorentzian polarizability,
following Equation (3.2) [103]. A sample Lorentzian element response is shown in
Figure 4.33. Tuning an element is often accomplished by changing ω0 , leading to
115
1 1
0.75
| |
0.5 0.5
0.25
Imag( )
0 0
0,1
0,2
- /2 0,3 -0.5
- -1
0.9 0.95 1 1.05 1.1 -1 -0.5 0 0.5 1
Real( )
Figure 4.33: A sample metamaterial element’s response, modeled with an analytic
polarizability based on a Lorentzian response described in Equation (3.2)
shifts in the magnitude and phase of the polarizability. Since the polarizability is a
complex analytic function of the resonant frequency, the magnitude and phase are
inextricably linked, as can be expressed by the relationship [103]
Fω
|α| = | cos ψ|, (4.11)
Γ
where ψ denotes the phase advance introduced by the element.

As expressed in Equation (4.11) and depicted in Figure 4.33, the Lorentzian
response has a range of phase advance restricted to −180◦ < φ < 0 —less than
half of the control range of active phase shifters. Moreover, the magnitude varies
over this phase range, falling to zero at either extreme such that the actual useful
phase range of the element is even smaller than 180◦ . Using a subwavelength spacing
of elements (λ/4 or denser, where λ is the free space wavelength) and leveraging
the phase accumulation of the guided wave can help to regain some of the reduced
element control, but these approaches can be costly and challenging, especially at
higher frequencies or with large apertures [91, 94, 141, 142]. When these approaches
are not possible, however, the strong magnitude modulation of the element weights
116
(which accompany a typical desired phase distribution for beamforming) will produce
grating lobes. For 1D antennas there is no simple way to avoid these grating lobes;
for 2D arrays formed by tiling a set of 1D antennas, the additional degrees of freedom
can be used to cancel out the grating lobes, providing alternative design approaches
for larger waveguide-fed metasurface arrays.
To illustrate the grating lobe problem associated with metasurface antennas,
radiation patterns are computed using array factor calculations, with the dipole
moments from Equation (3.1) as the antenna weights [19, 143]
N X
X M
AF (θ, φ) = ηn,m e−jk(xm sin θ sin φ+yn sin θ cos φ) (4.12)
n=1 m=1
where k is the free space wavenumber, N is the number of elements (in the y direc-
tion), M is the number of waveguides (in the x direction), θ is the elevation angle,
and φ is the azimuth angle. Ideally, the polarizability of each element could be set
to counteract the phase of the guided wave and form a beam steered to (θs , φs ),
following
αn = ej(βyn +kxm sin θs sin φs +kyn sin θs cos φs ) . (4.13)
However, due to the Lorentzian-constrained nature of metamaterial elements, the

polarizability profile given by Equation (4.13) cannot be fully realized. In [103],
Lorentzian-constrained modulation (LCM) was derived, which optimizes the phase
and magnitude of each metamaterial element simultaneously in order to approximate
Equation (4.13).
j − ej(βyn +kxm sin θs sin φs +kyn sin θs cos φs )

αn = (4.14)
2
Inserting Equation (4.14) into Equation (4.12), the array factor becomes
117
" N M
H0 X X −j(βyn +kyn sin θ cos φ+kxm sin θ sin φ)
AF (θ, φ) = je
2 n=1 m=1
(4.15)
N X
X M #
− e−jk yn (sin θ cos φ−sin θs cos φs )+xm (sin θ sin φ−sin θs sin φs )
n=1 m=1
With the constrained Lorentzian scheme for choosing the polarizabilities, the radia-
tion pattern consists of the main beam and potentially only one diffracted order.
Array factor calculations were used to model the beamforming performance of
a metasurface antenna array shown in Figure 4.32, which comprises eight air-filled
rectangular waveguides, each with twenty elements. Each waveguide has a width
ar = 2.0 cm and is fed with the same phase. The antenna operates at 10 GHz and
the elements within each waveguide are spaced at 0.5 cm (λ/6). Each element is
analytically modeled, according Equation (3.2), with the resonant frequency tunable
from 9.8 to 10.2 GHz. The analysis remains agnostic to any specific element design
and tuning mechanism and follow the analytic model described in Equation (3.2)
and shown in Figure 4.33. The antenna parameters are summarized in Table 4.6.
Using the array factor described in Equation (4.15), a broadside beam was gen-
erated with the sample metasurface antenna. Figure 4.34 shows the comparison
between the ideal polarizability expressed in Equation (4.13) (shown with the gray
dashed line) and the realized polarizability expressed in Equation (4.14) (shown with
simulate grating lobe suppression in metasurface antennas.
Parameter Value
# of elements 20
# of waveguides 8
Spacing 5 mm
Frequency 10.0 GHz
118
1 1
| |
| |
0 0
0 0
- -
0.5 1 1.5 2 0.5 1 1.5 2
guided guided
Figure 4.34: Mapping from desired to available α and corresponding η. The gray,
dashed line shows the desired response, while the blue, solid line shows the achievable
response. The red dots indicate the discrete locations of the metamaterial elements.
||
1
2
(y)
1.5
guided
0.5 0
1
0.5
0 -
2 4 6 8 2 4 6 8
Waveguide # (x) Waveguide # (x)
Figure 4.35: Antenna weights for the sample metasurface antenna shown in Fig-
ure 4.32. In the case of metamaterial radiators, the weights are given by the dipole
moments, η, for which the magnitude and phase are shown.
the blue line, with metamaterial locations at the red markers); these plots show the
response of the elements in one of the 1D waveguide-fed metasurfaces in the antenna
array shown in Figure 4.32. Note that to achieve the best approximation to the linear
phase advance required for a steered beam, the magnitude ends up with a periodic
modulation rather than the ideally flat profile. Figure 4.35 shows the metamaterial
elements’ dipole moments, which are analogous to the antenna weights of traditional
radiators. Figure 4.36 shows that a broadside beam is created, but a large grating
lobe appears. The source of the grating lobe can be traced back to the oscillating
119
90 0
60 -5
Elevation (deg.)
30 -10
0 -15
-30 -20
-60 -25
-90 -30
-90 -60 -30 0 30 60 90
Azimuth (deg.)
0°
°
Azimuth -30 30 °
Elevation
-60° 60 °
-90° 90 °
-30 -20 -10 0
Directivity (dB)
Figure 4.36: Normalized farfield pattern resulting from the simulated metasurface
antenna array, which follows the mapping shown in Figure 4.34 and 4.35. The beam
is generated towards broadside, with the target location indicated by the black star
and the grating lobe indicated by the red star.
magnitude profile in Figure 4.34 and Figure 4.35 and the resulting coarse effective
element spacing in the y direction. The grating lobe remains present when a beam
is generated which is steered to θ = −20◦ , φ = 20◦ , as shown in Figure 4.37.
120
90 0
60 -5
Elevation (deg.)
30 -10
0 -15
-30 -20
-60 -25
-90 -30
-90 -60 -30 0 30 60 90
Azimuth (deg.)
0°
°
Azimuth -30 30 °
Elevation
-60° 60 °
-90° 90 °
-30 -20 -10 0
Directivity (dB)
antenna array, forming a beam steered to θ = −20◦ , φ = 20◦ (the black star shows
the target location while the red star shows the grating lobe).
4.4.2 Grating Lobe Suppression

Grating Lobe Derivation
To better illustrate the source of the grating lobes, the grating lobe term is mathe-
matically isolated from the array factor. By examining Equation (4.12), the second
term can be seen as the beamsteering term, in which the phase of the guided wave
121
has been counteracted and the steering phase term has been applied. Meanwhile,
the first term shows the grating lobe term, where the grating lobe exists along the
φ = 0 direction. Substituting φ = 0 into the first term of Equation (4.12) leads to
N
X
e−jyn (β+k sin θ) . (4.16)
n=1
From this equation, the grating lobe will appear at θg = arcsin (β/k), φg = 0◦ . For
the simulated metasurface, this equation predicts a grating lobe at θg = 41◦ , which
is consistent with Figure 4.36 and Figure 4.37. It should be noted that a sufficiently
high dielectric may avoid this phenomenon, but this often invokes additional loss.
Waveguide Feed Layer
In this section, varying the phase exciting each radiating waveguide in a metasurface
antenna array is proposed as a means of suppressing the grating lobes. This proposal
can be illustrated mathematically by updating the definition of H to include an
arbitrary phase term applied to each waveguide-fed metasurface.
Hn,m = H0 e−jβyn +jγm (4.17)
Here γm is the phase applied to the feed of the mth waveguide in the array. The
optimal polarizability, as determined by Lorentzian-constrained modulation, then
becomes
j − ej(βyn +kxm sin θs sin φs +kyn sin θs cos φs −γm )

αn,m = . (4.18)
2
Combining Equation (4.12) with Equation (4.18), while incorporating Equa-

tion (4.17), leads to a new array factor
122
" N M
H0 X X −jyn (β+k sin θ cos φ) −j(kxm sin θ sin φ−γm )
AF (θ, φ) = j e e
2 n=1 m=1
# (4.19)
N X
X M
− e−jkyn (sin θ cos φ−sin θs cos φs ) e−jkxm (sin θ sin φ−sin θs sin φs )
n=1 m=1
where M is the number of waveguides. Equation (4.19) can be separated into two
terms: the first term is the grating lobe term; the second term is the beamsteer-
ing term. The grating lobe term can be separated into the multiplication of two
summations as
N M
H0 j X −jyn (β+k sin θ cos φ) X −j(kxm sin θ sin φ−γm )
e e (4.20)
2 n=1 m=1
From Figs. 4.36 and 4.37, the grating lobe from the metasurface behavior occurs
along the θ direction, in the φ = 0 plane. To analyze the grating lobe more explicitly,
substitute φ = 0 into Equation (4.20).
N M
H0 j X −jyn (β+k sin θ) X jγm
e e (4.21)
2 n=1 m=1
In order to cancel the grating lobe term, the summation of ejγm from m = 1 to M
must equal 0. The most straightforward method for canceling the grating lobe then
is select γm as γm = ±m(2π/M ), such that the term, ejγm , is evenly spaced in the
complex plane.
The optimal values for γm could be realized with phase shifters (passive or active)
or with a waveguide feed layer. Waveguide feed layers offer a small form factor, low
loss, and have been used in previous works to excite antenna arrays [144–146]. If
a feed layer were used to implement γm , it could be placed underneath the array
123
Figure 4.38: Diagram of a waveguide feed layer located at the bottom of the
metasurface antenna shown in Figure 4.32. The waveguide feed layer simplifies the
feed structure of the antenna to being a one port device and provides the phase
diversity required to suppress grating lobes. Note that here, the waveguide feed
layer is pictured on the same layer as the radiating waveguides, but the waveguide
feed could also be located beneath the radiating waveguides.
and coupling irises could be etched at each radiating waveguide’s location [147, 148].
To suppress the grating lobes, the phase accumulation of the waveguide feed must
match the desired γm as
2π
ejγm = e−jm M = e−jβf xm . (4.22)
Here, βf is the propagation constant of the waveguide feed and xm is the position
along the feed waveguide. When the feed layer samples the radiating waveguides
at a spacing equal to the radiating waveguide width (ar ), this condition can be
124
mathematically represented as βf ar = 2π/M , which is equivalent to
s
π2 2π
ar 0 µ0 ω 2 − 2
= . (4.23)
af M
Here, af is the width of the waveguide feed. If this condition is satisfied, the waveg-
uide feed layer will provide the optimal phase shift to each waveguide in the array
and cancel the grating lobe. Rearranging the terms, Equation (4.23) can be written
as a design equation which determines the waveguide width of the feed (af ).
M ar λ
af = p (4.24)
2 M 2 a2r − λ2
Equation (4.24) optimally suppresses the grating lobe, but can result in forcing the
waveguide to operate close to cutoff if M is too large. In such cases, substituting M
with an M 0 which is a factor of M (but greater than 1) will equivalently suppress
grating lobes.
The result from Equation (4.24) was used in an array factor calculation to create
a feed layer to excite each waveguide in the antenna shown in Figure 4.32. In this
particular metasurface antenna array, M 0 = 4 is used, since using M = 8 would
result in operation close to the cutoff frequency. Figure 4.39 shows the updated
||
1
2
(y)
1.5
guided
0.5 0
1
0.5
0 -
2 4 6 8 2 4 6 8
Waveguide # (x) Waveguide # (x)
Figure 4.39: Antenna weights for the metasurface antenna with a feed layer.
125
90 0
60 -5
Elevation (deg.)
30 -10
0 -15
-30 -20
-60 -25
-90 -30
-90 -60 -30 0 30 60 90
Azimuth (deg.)
0°
°
Azimuth -30 30 °
Elevation
-60° 60 °
-90° 90 °
-30 -20 -10 0
Directivity (dB)
is generated towards broadside, with the target location indicated by the black star.
antenna weights, which demonstrate the phase offset provided by the feed layer.
The corresponding radiation pattern is shown in Figure 4.40 which clearly shows the
grating lobe has been suppressed. Figure 4.41 shows that when a beam is steered in
both azimuth and elevation, the grating lobe remains suppressed.
126
90 0
60 -5
Elevation (deg.)
30 -10
0 -15
-30 -20
-60 -25
-90 -30
-90 -60 -30 0 30 60 90
Azimuth (deg.)
0°
°
Azimuth -30 30 °
Elevation
-60° 60 °
-90° 90 °
-30 -20 -10 0
Directivity (dB)
is generated towards broadside, with the target location indicated by the black star.
Frequency Dependence
To further analyze the effectiveness of the waveguide feed layer, a frequency study
is conducted. First, the metasurface antenna’s behavior is examined when tuned to
the center frequency and operating at other frequencies. For this study, a broadside
beam is generated with and without the feed layer, at 9.8, 10.0, and 10.2 GHz. From
127
a) b)
0° 0°
f=9.8 GHz °
-30
f=10.0 GHz 30° -30° 30°
f=10.2 GHz
-60° 60° -60° 60°
-90° 90° -90° 90°

-40 -30 -20 -10 0 -40 -30 -20 -10 0
Azimuth (deg.) Directivity (dB) Azimuth (deg.) Directivity (dB)
c) d)
0° 0°
-30° 30° -30° 30°
-60° 60° -60° 60°
-90° 90° -90° 90°

-40 -30 -20 -10 0 -40 -30 -20 -10 0
Elevation (deg.) Directivity (dB) Elevation (deg.) Directivity (dB)
Figure 4.42: Normalized farfield patterns for 9.8, 10.0, and 10.2 GHz. a) shows
the azimuth radiation pattern without the feed layer. b) shows the azimuth pattern
with the feed layer. c) shows the elevation radiation pattern without the feed layer.
d) shows the elevation pattern with the feed layer. In all cases, each element has
been tuned for operation at 10.0 GHz.
the elevation radiation patterns in Figure 4.42, the grating lobe is problematic at the
operating frequency (10.0 GHz) and fully eclipses the steered beam when operating
away from the target frequency. With the feed layer present, the grating lobe is well
suppressed at the center frequency and remains mostly suppressed when operating
away from the center frequency. It is also worth noting that when operating in this
manner, the azimuth direction shows squinting behavior as a function of frequency,
similar to a leaky wave antenna’s operation, due to the existence of the feed layer.
Metasurface antennas can also consider another method of operation if integrated
with components that enable high switching speeds. In addition to tuning the ele-
ments to the center frequency, it can be possible to update the tuning state of each
metamaterial element as the frequency changes. In this context, a new tuning state
is optimized and applied at each frequency. Here, the study above is repeated, but
128
a) b)
0° 0°
f=9.8 GHz °
-30
f=10.0 GHz 30° -30° 30°
f=10.2 GHz
-60° 60° -60° 60°
-90° 90° -90° 90°

-40 -30 -20 -10 0 -40 -30 -20 -10 0
Azimuth (deg.) Directivity (dB) Azimuth (deg.) Directivity (dB)
c) d)
0° 0°
-30° 30° -30° 30°
-60° 60° -60° 60°
-90° 90° -90° 90°

-40 -30 -20 -10 0 -40 -30 -20 -10 0
Elevation (deg.) Directivity (dB) Elevation (deg.) Directivity (dB)
Figure 4.43: Normalized farfield patterns for 9.8, 10.0, and 10.2 GHz. a) shows
the azimuth radiation pattern without the feed layer. b) shows the azimuth pattern
with the feed layer. c) shows the elevation radiation pattern without the feed layer.
d) shows the elevation pattern with the feed layer. For each operating frequency, a
new tuning state has been optimized for each element.
while updating the tuning state of each metamaterial element for each frequency.
From the elevation radiation patterns in Figure 4.43, the grating lobe dominates the
radiation pattern when the feed layer is not present. In addition, the azimuth fre-
quency squinting seen in Figure 4.42b is avoided by re-optimizing the tuning state at
each frequency. Further, tuning the antenna at each frequency shows improved per-
formance, as seen by comparing the grating lobes in Figure 4.42d and Figure 4.43d.
Different Tuning Strategies
Note that the need for a grating lobe suppression method is inherent to metamaterial
elements, irrespective of the tuning strategy used. In addition to using Lorentzian-
constrained modulation, the same metasurface antenna is examined with direct phase
tuning and with Euclidean modulation. Both of these methods involve tuning the
129
polarizability of each element to match the polarizability prescribed for beamforming
in Equation (4.13). These methods are described in Section 4.2.1. Figure 4.44 shows
radiation patterns from the simulated metasurface antenna using these methods. In
both Figure 4.44a and b, using the waveguide designed according to Equation (4.24)
suppresses the grating lobes, independent of tuning strategy.
a) °
b) °
With Feed
0 0 Without Feed
-30° 30 ° -30° 30 °
-60° 60 ° -60° 60 °
-90° 90 ° -90° 90 °
-30 -20 -10 0 -30 -20 -10 0
Directivity (dB) Directivity (dB)
Figure 4.44: Normalized farfield pattern of the metasurface antenna array with
and without a feed layer in the elevation direction. a) uses a direct phase matching
approach to tuning the metamaterial elements. b) uses Euclidean modulation to
tune the metamaterial elements.
Metasurface antennas stand to offer substantial hardware benefits as compared

to existing electronically reconfigurable antennas, but the nature of metamaterial
elements presents challenges with realizing the full potential of metasurface antennas.
Without considering the differences in element behavior between traditional radiators
and metamaterial elements, strong grating lobes can appear. A waveguide feed layer
obviates the need for extremely dense element spacing or high dielectrics to avoid
grating lobes and retains a metasurface antenna’s hardware benefits in terms of cost
and manufacturability.
130
5
Metasurface Antenna Prototype
In this chapter, the design and performance of a metasurface antenna prototype are
presented. The metasurface antenna design is based on the principles discussed in
Chapter 2 and Chapter 3 and is heavily influenced by the results of the various studies
detailed in Chapter 4. The antenna design presented in this section verifies several
principles of metasurface antennas and shows that they can realize high performance
electronic beamforming from a favorable physical platform.
This chapter starts with a discussion of the individual element design and perfor-
mance. Then the metasurface antenna prototype layout is described in detail. Next,
the experimental characterization of a single metamaterial element is demonstrated,
including experimental polarizability extraction. Last, beamforming results with the
metasurface antenna prototype are presented and the realized antenna performance
metrics are described.
5.1 Individual Element Design
The behavior of the constituent metamaterial elements directly affects the perfor-
mance of a metasurface antenna. To that end, one of the first steps towards designing
131
a metasurface antenna is the design of an element which can exhibit adequate con-
trol for beamforming. Section 3.3 describes how dynamic metasurface apertures with
binary control work well for producing spatially diverse radiation patterns, but can
lead to challenges associated with synthesizing specific radiation patterns. Grayscale
phase control can better facilitate realizing a phase profile across the antenna, which
is essential for beamsteering. Meanwhile, grayscale magnitude control allows for the
ability to control the guided wave’s decay, which can improve efficiency and realize
desired tapers for sidelobe suppression. Though a metamaterial element’s magnitude
and phase response are linked through their resonance, metasurface-specific tuning
strategies (as described in Section 4.2) can provide a means of jointly optimizing the
response of each element in a metasurface antenna. In this section, a metamaterial
element design (and its integration within a waveguide structure) is described that
uses a varactor diode to provide grayscale resonance tuning.
Beamforming with metasurface antennas, as described in Section 3.1.3, leads
to a set of αdesired which must be approximated with an αtunable . Paramount to
this process is the ability of the constituent metamaterial radiators to realize as
close to an arbitrary complex response as possible. Metamaterial elements exhibit
a Lorentzian response, as described in Equation (3.2). A Lorentzian response can
be tuned either by decreasing the strength of the resonance (magnitude tuning –
tuning F ) or by shifting the resonant frequency (resonance tuning – tuning ω0 ). To
grant the necessary control for beamforming, grayscale resonance tuning is used in
the forthcoming metamaterial element design.
As mentioned above, cELC metamaterial elements can be tuned by integrating a
semiconductor within the geometry and biasing the semiconductor to tune the ele-
ment response. A diagram of the design of the element used in this section is shown
in Figure 5.1; the element in a full-wave simulation environment is shown in Fig-
ure 5.2. The element uses varactors placed across the gap between the metamaterial
132
Metamaterial Varactor Control Via
Radiator
Waveguide
SIW Vias Control line

Figure 5.1: Varactor tuned metamaterial element design.
element geometry and the surrounding upper conductor of a substrate integrated

waveguide (SIW), which acts as a rectangular waveguide. Biasing the element with
different voltages changes the capacitance across the gap and thus shifts the resonant
frequency of the element. In order to individually address each element, a bias line
must connect to each element without contacting the walls of the waveguide. In the
design shown above, a control via extends through the waveguide structure to lower
layers used for control circuitry.
One aspect of metamaterial element design which must be considered is the self-
resonance of the varactor. The parasitic inductance of the varactor, combined with
its capacitance, leads to a self-resonant frequency. If the operating frequency of the
metamaterial is above the self-resonant frequency of the varactor, the varactor will
not be capacitive and radiation from the metamaterial can be suppressed. Many
commercially available varactors exhibit a high series inductance and thus cannot
operate at frequencies as high as 10 GHz. For this metamaterial design, a MACOM
varactor (MAVR-011020-1141) has been selected due to its parasitic inductance of
0.3 nH. When combined with its tunable capacitance range (0.24 - 0.065 pF), its self
resonant frequency can be estimated to move from 19 - 36 GHz as the varactor is
tuned, remaining above the operating frequency of 10 GHz.
133
Figure 5.2: Grayscale resonance tunable metamaterial element design in a full-wave
simulation environment.
The metamaterial element shown in Figure 5.2 follows the geometry of a cELC
and responds as a magnetic dipole. The element is 3.65 mm by 3.65 mm, with the
tracewidths being 0.5 mm, and the gap between the metamaterial and the surround-
ing conductor being 0.2 mm. Other benefits of this element design include that it can
have a reasonably sharp resonance. While a sharp resonance can sometimes cause
loss, it also means that an external bias can tune the response from being below
resonance, to on resonance, to above resonance, providing a wide range of phase
control. For the design presented here, biasing the varactors from 0 - 5 V tunes the
resonance from 8.2 - 11.0 GHz, as shown in Figure 5.3. Further, it is a practical
choice for metamaterial element design due to its ease of integration with tunable
semiconductors and various waveguide structures.
When using the dipole model presented in Section 3.1.4 to represent a series of
metamaterial elements, certain assumptions must be made, including that the ele-
ments do not experience mutual coupling. Many metasurface designs that include
elements being placed a subwavelength distance apart (λ/4 or denser) are difficult
to model accurately this way due to strong mutual coupling. By offsetting the ele-
ment from the center, the element exhibits weaker coupling to the guided wave (and
to nearby elements). In addition, offsetting the element from the center also en-
134
Power Radiated (normalized)
8 9 10 11 12
Frequency (GHz)
Figure 5.3: Power radiated by the metamaterial element shown in Figure 5.2 as a
function of tuning from 0 - 5 V in full-wave simulation.
ables elements to alternate sides of the waveguide, creating additional space between
elements.
Furthermore, when envisioning a metasurface antenna comprising many of such
elements along a waveguide, a weaker coupled element also avoids depleting the
energy in the guided wave too quickly. If each element radiates a substantial portion
of the guided wave’s energy, the guided wave can be depleted quickly, before it
reaches elements far from the feed; this phenomenon can limit the effective aperture
size of a metasurface antenna. In addition, by moving the element’s position away
from the center of the waveguide, the control via can be located very close to the
waveguide’s sidewalls. Moving the control via as far away from the center as possible
also decreases reflections in the waveguide, allowing more energy that is not radiated
to pass along subsequent elements.
135
5.2 Antenna Layout
5.2.1 Challenges with Metasurface Antennas
There are three major challenges associated with the design and implementation
of a metasurface antenna which are addressed in this section: modeling the an-
tenna, suppressing grating lobes, and tuning the antenna. Through the simulations
in Chapter 4, a better understanding of metasurface antenna operation led to a co-
hesive antenna design which addresses these challenges simultaneously. The design
presented in this section led to high performance beamforming from a lightweight,
low cost, and planar architecture.
First, the Lorentzian response associated with metamaterial elements can lead to
strong grating lobes. Grating lobes do not occur due to the inter-element spacing,
but instead from the lack of independent magnitude and phase control. As a result,
the process of trying to match the phase of each metamaterial element to a specified
phase unintentionally creates an oscillating magnitude profile. The periodicity of
this magnitude profile leads to a phenomenon where the effective element spacing
becomes coarse, leading to metasurface-specific grating lobes. Here, a phase diverse
feed layer, as detailed in Section 4.4, is used as a means of suppressing grating lobes
in the metasurface antenna design. Section 4.4 elucidates the metasurface grating
lobe problem and suggests a phase diverse feed layer as a solution, which is used
here.
Second, modeling metasurface antennas presents another challenge due to the dif-
ficulty associated with modeling electrically-large structures that contain subwave-
length variations. The relative size of features present in metamaterials often forces
the use of approximate models. Any approximate model is tasked with accurately
predicting the guided wave and in order to calculate a set of effective dipoles that rep-
resents each element in the antenna. An accurate model of the antenna is paramount
136
to realizing high performance due to the fact that the model is directly used to de-
termine the tuning state for a desired radiation pattern. For the antenna described
in this chapter, the method described in Section 3.1.4 is used, incorporating experi-
mentally characterized H, α, and S21 .
The third challenge associated with demonstrating a high performance metasur-
face antenna is how to tune the constituent elements. With traditional phased array
antennas, the full and independent control afforded by phase shifters and amplifiers
trivializes this challenge; the exact phase profile is applied with the phase shifters in
order to generate the required beam and the amplifiers can be tuned to apply the
desired taper. As described in Section 4.2, however, the control over metamaterials
is limited in two ways: the maximum achievable phase shift is only π (instead of 2π)
and the magnitude and phase response are linked (tuning the phase also tunes the
magnitude and vice versa). As a result, metasurface-specific tuning strategies are
required in order to optimize the performance of a metasurface antenna.
Any metasurface method is thereby tasked with mapping from a set of αdesired to
a set of αtunable . To address this challenge, a compromise between pure phase tuning
and pure magnitude tuning must be realized, often taking the form of metasurface-
specific tuning strategies described in Section 4.2.
5.2.2 Metasurface Antenna Design
A metasurface antenna prototype was designed, consisting of an array of 1D waveg-

uides, shown in Figure 5.4. The 12 elements along each waveguide are spaced 11 mm
apart and are offset from the center by 4.75 mm. The elements are placed on alter-
nating sides of the waveguide to increase the amount of space between elements in
order to reduce inter-element coupling. The elements occupy a region with an area
of approximately 12 cm by 12 cm, leading to a theoretically achievable 22.4 dBi of
gain with 100% efficiency. It should be noted that using a progressively decreasing
137
Table 5.1: Antenna parameters for the prototype metasurface antenna.
Parameter Value
Total antenna size 15 cm by 22.5 cm
Radiating area size 12 cm by 12 cm
# of waveguides 8
Total tunable elements 96
# of terminators 8
# of end launch connectors 16
# of passive phase shifters 12
# of DACs 12
# of varactors 192
offset from the center could possibly counteract some of the natural decay of the
guided wave. Since counteracting this magnitude effect purely with tuning comes at
the cost of some phase control, this approach could be one way to retain more phase
control while realizing a desired magnitude profile.
Each waveguide takes the form of a substrate integrated waveguides, which behave
as rectangular waveguides but are easy to fabricate within a PCB. Each SIW is
14 mm wide, with the SIWs separated by 15 mm to provide space for the via fences.
In order to launch a wave within an SIW, end launch connectors are used. A detailed
look at an end launch connector is provided in Figure 5.4. End launch connectors
are commercially available components which can launch a wave into a grounded
coplanar waveguide (CPW). Similar to the characterization board, after the wave
is launched into the CPW, there is a CPW-to-SIW transition, which is shown in
Figure 5.6. At the end of each waveguide is a 50Ω terminator. The fabricated
antenna prototype is shown in Figure 5.7. The parameters of the metasurface antenna
prototype are highlighted in Table 5.1.
Towards the goal of low-cost fabrication and a planar footprint, a printed circuit
138
End Launch Connectors Metamaterial Elements
Figure 5.4: Metasurface antenna prototype design [43, 149].
board (PCB) antenna design was used. The antenna is fabricated on a 4 (metal)
layer PCB with a low loss Rogers 4003C substrate. The top layer contains the
metamaterial elements and the varactors. Between the top two layers of the PCB is
the waveguide structure. The third layer contains various control lines. The bottom
layer contains radial stubs, control lines, and all control components. The PCB layer
139
Table 5.2: Layers of the characterization board and the metasurface antenna proto-
type.
Layer Material Thickness Contents Components

1 Copper 0.0175 mm Metamaterial elements Varactors
Feed transitions
Waveguide top
2 Rogers 4003C 30 mil Waveguide interior
3 Copper 0.0175 mm Waveguide bottom
4 Rogers 4003C 15 mil
5 Copper 0.0175 mm Control lines
6 Rogers 4003C 15 mil
7 Copper 0.0175 mm Control lines DACs, resistors
Radial stubs buffers, header
Layer 1 (Top) Layer 2
Layer 3 Layer 4 (Bottom)
Radiating area
Figure 5.5: Schematic drawing of each layer of the metasurface antenna prototype.
The insets, clockwise from the top left, show a single metamaterial element, the
placement of a DAC and its corresponding control lines, and a radial stub.
stackup is described in Table 5.2. Each layer can be seen schematically in Figure 5.5.
All vias — both those used for control and those used to create the waveguide walls
— extend through the all layers of the board. The antenna consists of 8 waveguides,
each of which contain 12 cELC metamaterial elements, for a total of 96 elements.
Each element is controlled by one output of an 8 bit, 8 channel DAC (LTC1665),
which provides a means of biasing each element individually, from 0 to 5V.
140
Figure 5.6: Feed transitions used to launch a wave into the SIW within a PCB
antenna. An end launch, connected to the left side of the board, launches a wave
into a CPW, which then transitions to an SIW [150].
As mentioned above, this antenna uses phase diversity at the feed point to sup-
press grating lobes. The principles of this operation are described in detail in Sec-
tion 4.4. While a more streamlined design could be realized with a waveguide feed
layer, for this prototype, passive phase shifters were used to approximate the re-
sponse of a waveguide feed layer. Here, an 8-way power divider is connected to a
series of phase shifters to create the requisite phase diversity. From top to bottom,
the outputs of the power divider are connected to 3, 2, 1, 0, 0, 1, 2, and 3 passive
phase shifters, respectively. Each phase shifter has been manually set to provide
approximately a 90◦ phase shift at 10 GHz. It should be noted that since the phase
shifters have been set for a target frequency of 10 GHz, performance at other fre-
quencies is understated. A more elegantly designed waveguide feed structure would
be able to operate over a larger bandwidth. The feed structure described here is
shown in Figure 5.8.
141
Figure 5.7: Pictures of the metasurface antenna prototype.
5.3 Experimental Results

5.3.1 Element Characterization
Various practical effects mandate experimental characterization to accurately deter-

mine the response of an individual metamaterial element. Factors including fabrica-
tion tolerances and semiconductor parameter variation can lead to inaccuracies if a
purely analytic/simulated model of a metamaterial element is used when modeling
142
Figure 5.8: Picture of the metasurface antenna’s feed structure. A power divider
connects to a series of passive phase shifters which provide phase diversity at the
feed point of each waveguide in the prototype antenna.
a metasurface antenna. To address this challenge, experimental characterization of

a metamaterial element is conducted and the measured response is loaded into the
model of the overall metasurface antenna prototype.
To experimentally characterize the response of the element, a characterization
board was designed and fabricated. This board contains various measurement chan-
nels, including an empty waveguide, a single element in a waveguide, and a thru-
reflect-line calibration channel set. The first test conducted with the characterization
board was to measure the power radiated by the metamaterial element as a func-
tion of tuning state. Figure 5.10 shows the power radiated as a function of tuning
state, showing that biasing the element from 0 - 5 V moves the resonance from 8.5
143
Figure 5.9: Characterization board for characterizing the individual metamaterial
response and waveguide characteristics experimentally.
- 10.7 GHz. This tuning range closely matches that predicted in simulation (8.2-
11.0 GHz), indicating that there is no shift up or down in frequency, but a slight,
symmetric reduction from either side of the tuning range. Note that the power ra-
diated curves appear different from those in Figure 5.3 due to the presence of end
launch connectors and the feed transitions, which radiate in addition to the meta-
material element.
In order to model each metamaterial element experimentally, the metamaterial
element’s complex S parameters and polarizability must be measured for each tuning
state and frequency. To do this in simulation, the S parameters can be de-embedded
to the location of the element along the waveguide, isolating the S parameters of the
element independent of the surrounding waveguide/feed structure. The S parameters
can then be used to calculate the element’s polarizability, following Equation (4.1).
To de-embed the S parameters, experimentally, a thru-reflect-line (TRL) cali-
bration can be conducted, which is a robust calibration method and requires only
144
Power Radiated (normalized)
8 9 10 11 12
Frequency (GHz)
Figure 5.10: Power radiated by a single metamaterial element for different tuning
states. From these results, biasing the element from 0 - 5 V tunes the resonant
frequency from 8.5-10.7 GHz.
three standards [151, 152]. The TRL circuit set was included as a series of adja-
cent waveguides within the test board, specifically in the bottom three channels
shown in Figure 5.9. Counting from the top waveguide, channels 6, 7, and 8 are
the thru, reflect, and line channels, respectively. Using the VNA’s TRL calibration
protocol, de-embedded S parameter measurements were taken for each frequency and
tuning state for the metamaterial element (channels 3 and 4 from the top). The de-
embedding process moves the measurement ports from the end of the physical cable
to virtual planes located + and - 5 mm from the center of the waveguide. The first
de-embedded measurements taken were of channel 6; measuring the empty waveg-
uide this way provides a means of determining the waveguide propagation constant
across a 1 cm section of the waveguide. The results of that measurement, shown
in Figure 5.11, provide a means of correcting the simulated waveguide propagation
constant (previously modeled from the geometry of the waveguide and the dielectric
145
Figure 5.11: Comparison between the simulated, measured, and corrected waveg-
uide propagation of the fabricated substrate integrated waveguide. Material param-
eters (such as dielectric constant) and via width can lead to small discrepancies
between the simulated and fabricated waveguide propagation constant.
constant of the circuit board). In addition, this measurement can be divided from
measurements of a single metamaterial element (measured from either channel 3 or
4 on the characterization board) to truly isolate its response to the instantaneous
location of the center of the metamaterial element.
Once the de-embedded S parameters were measured, a library of available element
responses (αtunable ) was created, which is a function of frequency from 8-12 GHz and
of bias voltage from 0-5V. The tuning range has been restricted to 0-5V for the sake
of simplicity in the biasing circuitry; a larger tuning is possible with more advanced
circuitry. The de-embedded S parameters are shown as a function of tuning voltage
for four different frequencies in Figure 5.12. From these results, a clear resonance
exists within the tuning range. At 9.25 and 10 GHz, tuning from 0-5V goes across
the full resonance. At 8.5 GHz, tuning the element moves the resonance from the
peak to the right tail; at 10.5 GHz, tuning the element moves the resonance from the
peak to the left tail. Figure 5.12 also shows the extracted polarizability as a function
of bias voltage for several frequencies.
146
Figure 5.12: Experimentally characterized metamaterial element response. The
extracted α was calculated from Equation (4.1), using de-embedded S parameters.
One important aspect of the element response which can be seen from the ex-
tracted polarizability plots is the realized phase tuning range. In theory, a Lorentzian
resonator can only realize phase shift (180◦ as opposed to 360◦ for a phase shifter),
with the magnitude falling off away from the resonance peak. From the curves in
Figure 5.12, the experimentally realized phase shift nearly matches the theoretical
180◦ phase shift. When operating from 9.00 – 10.25 GHz, a 150◦ phase shift is re-
alizable by tuning the element. It should also be noted that the realized magnitude
ratio is 4.5:1 when tuning from the highest to lowest magnitude at 10 GHz. These
results are shown in Figure 5.13.
Last, consistency among different instance of the metamaterial element behavior
was measured. Since the numerical model of the full metasurface antenna proto-
type assumes that each element behaves identically, it is important that fabrication
147
Figure 5.13: Tunability, in resonant frequency and phase control, of a fabricated
metamaterial element. From 9.0 - 10.4 GHz, the element achieves at least 150◦ of
phase control.
tolerances and semiconductor component specifications do not vary among different

instances of the metamaterial elements. To test this, channels 3 and 4 of the charac-
terization board provide a means of comparing the response between two instances
of one metamaterial element design. The results in Figure 5.14 and Figure 5.15 show
nearly identical performance, indicating that there is not significant variation among
different instances of the same metamaterial element.
5.3.2 Experimental Beamforming Results
The main goal for the prototype metasurface antenna was to demonstrate the gen-
eration of a beam and to show steering in azimuth and elevation across a defined
bandwidth. For the demonstrations in this section, the target metrics are summa-
rized in Table 5.3. If met, these goals serve to validate the metasurface antenna
prototype as a capable electronic beamforming antenna. Other metrics, such as po-
larization isolation and efficiency, are also important for antenna performance, but
were not targeted as goal metrics during design.
Once a fabricated sample was on hand, several characterization steps had to be
conducted before beamforming. First, the feed structure had to be implemented and
148
Figure 5.14: Power radiated consistency among different instances of a fabricated
metamaterial element as a function of tuning. Channel 3 and channel 4 are identical
element geometries.
Figure 5.15: S parameter consistency among different instances of a fabricated

metamaterial element as a function of tuning. Channel 3 and channel 4 are identical
element geometries.
149
Table 5.3: Antenna performance metrics targeted for the metasurface antenna pro-
totype.
Metric Target
Azimuth Steering ±20◦
Elevation Steering ±20◦
measured. As described above, the desired phase shifts were [270◦ 180◦ 90◦ 0◦ 0◦
90◦ 180◦ 270◦ ] for each of the outputs of the power divider/phase shifters, which are
the inputs to each of the feed waveguides – this structure is shown in Figure 5.8.
After the feed network was implemented, each channel was measured with a vector
network analyzer (VNA) to determine the initial phase of each feed wave and the
measurements were incorporated within the numerical model of the antenna.
After the element and feed structure were measured, those measurements were
incorporated into the mathematical model of the antenna. The model was then used
to determine the tuning state for each element for each experimental demonstration.
To characterize the radiation pattern of the antenna, near field scan measurements
were taken and the measurements were propagated to the farfield.
The first demonstration with the metasurface antenna prototype was to generate
a broadside beam. Using scaled Euclidean modulation, a tuning state was determined
for an operating frequency of 10 GHz. The measured farfield pattern is shown in
Figure 5.16, plotted with a normalized, logarithmic colormap. The antenna is able
to generate a beam steered in azimuth and elevation to (0◦ , 0◦ ) accurately. A 3D,
isometric view of the radiation pattern, plotted with a normalized linear colormap is
shown in Figure 5.17. The beamwidth is 12.3◦ in azimuth and 12.8◦ in elevation.
Initial measurements were taken at Duke University, using an open ended waveg-
uide as the reference antenna for NFS measurements. In this case, NFS measure-
ments were taken approximately 5 cm from the aperture, performing a raster scan
which covers approximately a 22 cm by 22 cm area around the radiating area of the
150
Figure 5.16: Broadside beam generated with the metasurface antenna prototype.
Figure 5.17: An isometric view of the broadside beam generated with the meta-
surface antenna prototype, shown with a linear colormap.
antenna (approximately 12 cm by 12 cm). To independently verify the performance

of the antenna, a set of measurements was taken at the Wireless Research Center
(WRC) in Wake Forest, NC. The anechoic chamber at the WRC has the capability
to take dual polarized measurements across a wider range of angles, increasing mea-
surement accuracy; the metasurface antenna prototype is pictured at their facility
151
Figure 5.18: Metasurface antenna prototype being measured at the Wireless Re-
search Center (WRC).
Gain (dB)
Figure 5.19: Comparison between simulated and experimental beamforming with

the metasurface antenna prototype.
in Figure 5.18. Figure 5.19 shows a comparison between a broadside beam in sim-
ulation, measured at Duke, and measured at the WRC. For the remainder of this
document, results presented were measured at the WRC.
It should be noted that all beamforms produced with the metasurface antenna
are based on the mathematical model presented in Section 3.1.4 — no numeric op-
timization has been applied. Numeric optimization could potentially improve the
beamforming results due to the built-in assumptions associated with the approxi-
mate model used here. Exploration of such methods, however, is beyond the scope
152
of this work.
After demonstrating broadside beamforming, beamsteering performance was in-
vestigated. As mentioned above, the target beamsteering limits were ±20◦ in both
azimuth and elevation. Figure 5.20 shows cross sections of a beam steered contin-
uously in azimuth from −60◦ to 60◦ in 7.5◦ increments. Figure 5.21 shows cross
sections of a beam steered continuously in elevation from −67.5◦ to 67.5◦ in 7.5◦
increments. In addition, 2D beamforms are shown for steering in azimuth/elevation
to ±15◦ in Figure 5.23.
Figure 5.20: Azimuth steering capabilities of the metasurface antenna prototype.
In addition to exploring the steering limits in azimuth and elevation, the accuracy
of the beamforming in those directions is also explored. Those results are summarized
in the two plots shown in Figure 5.22, which show the target angle compared to the
realized steered angle in both azimuth and elevation, respectively. In both directions,
there is little error between the targeted angle and the realized angle, validating the
accuracy of the modeling method and tuning strategy used to set the state of each
element. From these plots and those in Figures 5.20 and 5.21, the antenna can
153
Figure 5.21: Elevation steering capabilities of the metasurface antenna prototype.
realize a steering range of ±50◦ in azimuth and ±70◦ in elevation. Both of these
values exceed the target metrics of ±20◦ .
To explore the operational bandwidth of the antenna, a broadside beam was gen-
erated at various frequencies from 8 – 12 GHz. The target bandwidth of operation
was 9.6-10 GHz. Since the passive phase shifters were mechanically tuned for 10 GHz,
Azimuth Elevation
Figure 5.22: Steering accuracy of the metasurface antenna in azimuth and eleva-
tion.
154
Figure 5.23: Steering the metasurface antenna prototype in azimuth and elevation.
The steered angles are 15◦ away from broadside.
initial experiments used 9.75 - 10.25 GHz as the test bandwidth. It is also worth
noting that performance at frequencies away from 10 GHz may be degraded due to
the fact that the feed structure has been set for 10 GHz. Figure 5.24 shows that the
antenna is able to generate a broadside beam with minimal degradation in perfor-
mance from 9.75 - 10.25 GHz. The metamaterial element is able shift its resonance
from 8.5 GHz to 10.7 GHz with the external bias applied by electronic tuning. This
155
Figure 5.24: Metasurface antenna prototype forming a broadside beam across a
narrow bandwidth of operation (9.75-10.25 GHz).
Figure 5.25: Metasurface antenna prototype forming a broadside beam across a

wide bandwidth of operation (9-11 GHz).
resonance shift translates into as much as 150◦ of phase tuning at 10 GHz, but the
phase tuning range is more limited at other frequencies. Despite these challenges as-
sociated with operating away from 10 GHz, broadside beams were able to be formed
across a wide bandwidth, as shown in Figure 5.25. For these beamforms, the gain
is reduced, but other metrics such as sidelobe level and beamwidth are comparable
with the metrics realized at 10 GHz.
To further explore the capabilities of the antenna, multiple beams were gener-
ated simultaneously. Among these efforts, two beams were formed in azimuth and
then two beams in elevation. This experiment yielded the beam patterns shown in
156
Figure 5.26: Metasurface antenna generating two simultaneous beams in azimuth
(left) and elevation (right).
Figure 5.26, which shows distinct, steered beams.

For all results presented above, scaled Euclidean modulation was the chosen tun-
ing strategy, described in detail in Section 4.2. Figure 5.27 shows the antenna’s
response as a function of scale factor to show how this tuning strategy unlock high
performance. From these results, one can see how this tuning method is able to slow
the decay of the guided wave to realize a more uniform set of antenna weights, which
are analogous to the dipole moments (η). The broadside gain values shown indicate
that if the scale factor is too low, the directivity is limited (limiting the broadside
gain) and if the scale factor is too high, the efficiency is low (also limiting the broad-
side gain). For the metasurface antenna prototype, the experimentally determined
optimal scale factor was found to be 3.5.
As mentioned above, the primary target metrics were steering in azimuth and
elevation and beamforming over a specified bandwidth. Therefore, efficiency and
polarization isolation were not metrics that the antenna was designed to optimize.
Despite the fact that these characteristics were not targeted during design, they were
characterized during experimentation. The polarization isolation was measured to be
20 dB when forming a broadside beam. The efficiency of the antenna was measured
to be 11.1%. While this number may be low, many of the sources of loss would be
avoidable in future designs. Several measurements were conducted in an attempt to
157
1
2
3
4
5
Figure 5.27: The left plot shows the dipole moment for Broadside gain as a function
of scale factor while using scaled Euclidean modulation with the metasurface antenna
prototype.
isolate the sources of loss and estimate the efficiency of the antenna structure alone.
The losses in the feed structure (including the power divider, passive phase shifters,
and the cables) were measured to be 1.31 dB. Energy lost in terminators at the end
of the waveguides was measured to be 0.24 dB. The combination of the dielectric
losses and the end launch-CPW-SIW feed transition losses was measured to be 1.84
dB. Last, 3.08 dB was lost due to feed transitions. Many of these sources of loss
could be avoided in future designs, particularly air-filled waveguide designs. If these
sources of loss were removed, there is a clear path towards efficiencies near 50%.
A metasurface antenna which can electronically beamsteer in azimuth and eleva-
tion has been demonstrated. The antenna is lightweight, low-cost, and planar and
operates over a large bandwidth. In contrast to other metasurface antennas, this
design comprises Nyquist sampled elements – a design which facilitates rapid and
accurate modeling for tuning the antenna. The antenna prototype was independently
measured and exceeded all target metrics, as summarized in Table 5.4.
158
Table 5.4: Realized antenna performance metrics targeted for the metasurface an-
tenna prototype.
Metric Target Realized

Bandwidth 9.6-10.0 GHz 9.25 - 10.75 GHz
Azimuth Steering ±20◦ ±50◦
Elevation Steering ±20◦ ±70◦
159
6
Conclusion
In conclusion, metasurface antennas are a suitable alternative to many existing mi-

crowave antenna architectures. Through a series of simulations, metasurface an-
tennas were able to demonstrate similar or improved performance when compared
with existing hardware for spaceborne SAR. Simulations of various metasurface an-
tenna designs led to an prototype design which experimentally validates the claims
made throughout this document. The fabricated metasurface antenna exceeded all
performance goals and proved to be a capable platform for electronic beamforming
from a lightweight, low-cost, and planar physical platform. Their comparative ad-
vantages are highlighted below in Table 6.1. The high performance of metasurface
antennas, coupled with their favorable physical characteristics, has poised them to be
a competitive technology for microwave applications, especially in spaceborne SAR
systems.
160
Characteristic Phased Array Reflector Dish Metasurface Antenna
Cost High Low Low
Weight High Low Low
DC power High Low Low
Footprint Moderate Large Small
Switching speed Moderate Moderate Fast
Moving parts No Yes No
Deployment No Yes Possible
Beam tailoring Optimal No Yes
Table 6.1: Comparison between phased arrays, reflector dish, and metasurface an-
tennas as antenna solutions for spaceborne SAR.
161
Bibliography
[1] W. M. Brown and L. J. Porcello, “An introduction to synthetic-aperture

radar,” IEEE spectrum 6, 52–62 (1969).
[2] J. C. Curlander and R. N. McDonough, Synthetic aperture radar (John Wiley

& Sons, 1991).
[3] M. Soumekh, Synthetic aperture radar signal processing (New York: Wiley,
1999).
[4] G. Franceschetti and R. Lanari, Synthetic aperture radar processing (CRC

press, 1999).
[5] A. W. Doerry, D. F. Dubbert, M. Thompson, and V. D. Gutierrez, “A portfolio

of fine resolution Ka-band SAR images: part I,” in “Defense and Security,”
(International Society for Optics and Photonics, 2005), pp. 13–24.
[6] T. P. Ager, “An introduction to synthetic aperture radar imaging,” Oceanog-

raphy 26, 20–33 (2013).
[7] H. A. Zebker and R. M. Goldstein, “Topographic mapping from interferometric

synthetic aperture radar observations,” Journal of Geophysical Research: Solid
Earth (1978–2012) 91, 4993–4999 (1986).
[8] M. Soumekh, “A system model and inversion for synthetic aperture radar imag-
ing,” IEEE Transactions on Image Processing 1, 64–76 (1992).
[9] R. N. Treuhaft, S. N. Madsen, M. Moghaddam, and J. J. Zyl, “Vegetation

characteristics and underlying topography from interferometric radar,” Radio
Science 31, 1449–1485 (1996).
[10] W.-Q. Wang, “Space–time coding mimo-ofdm sar for high-resolution imaging,”
IEEE Transactions on Geoscience and Remote Sensing 49, 3094–3104 (2011).
162
[11] G. Krieger, “Mimo-sar: Opportunities and pitfalls,” IEEE Transactions on
Geoscience and Remote Sensing 52, 2628–2645 (2013).
[12] T. Lewis and H. Hutchins, “A synthetic aperture at optical frequencies,” Pro-

ceedings of the IEEE 58, 587–588 (1970).
[13] S. M. Beck, J. R. Buck, W. F. Buell, R. P. Dickinson, D. A. Kozlowski, N. J.

Marechal, and T. J. Wright, “Synthetic-aperture imaging laser radar: labo-
ratory demonstration and signal processing,” Applied optics 44, 7621–7629
(2005).
[14] B. Walker, G. Sander, M. Thompson, B. Burns, R. Fellerhoff, and D. Dubbert,

“A high-resolution, four-band sar testbed with real-time image formation,”
in “Geoscience and Remote Sensing Symposium, 1996. IGARSS’96.’Remote
Sensing for a Sustainable Future.’, International,” , vol. 3 (IEEE, 1996), vol. 3,
pp. 1881–1885.
[15] M. S. Moran, A. Vidal, D. Troufleau, Y. Inoue, and T. A. Mitchell, “Ku-

and c-band sar for discriminating agricultural crop and soil conditions,” IEEE
Transactions on Geoscience and Remote Sensing 36, 265–272 (1998).
[16] H. Cantalloube and P. Dubois-Fernandez, “Airborne X-band SAR imaging

with 10 cm resolution: Technical challenge and preliminary results,” IEE
Proceedings-Radar, Sonar and Navigation 153, 163–176 (2006).
[17] J. Way and E. A. Smith, “The evolution of synthetic aperture radar systems
and their progression to the EOS SAR,” IEEE transactions on geoscience and
remote sensing 29, 962–985 (1991).
[18] S. Ochs and W. Pitz, “The TerraSAR-X and Tandem-X satellites,” in “2007 3rd
International Conference on Recent Advances in Space Technologies,” (IEEE,
2007), pp. 294–298.
[19] R. C. Hansen, Phased array antennas, vol. 213 (John Wiley & Sons, 2009).
[20] C. L. Holloway, E. F. Kuester, J. A. Gordon, J. O’Hara, J. Booth, and D. R.

Smith, “An overview of the theory and applications of metasurfaces: The two-
dimensional equivalents of metamaterials,” IEEE Antennas and Propagation
Magazine 54, 10–35 (2012).
[21] D. M. Sheen, D. L. McMakin, and T. E. Hall, “Three-dimensional millimeter-

wave imaging for concealed weapon detection,” IEEE Trans. Microw. Theory
Techn. 49, 1581–1592 (2001).
163
[22] M. Soumekh, Fourier array imaging (Prentice-Hall, Inc., 1994).
[23] A. J. Devaney, Mathematical foundations of imaging, tomography and wavefield

inversion (Cambridge University Press, 2012).
[24] R. Linck, J. Fassbinder, and S. Buckreuss, “Integrated geophysical prospection

by high-resolution optical satellite images, synthetic aperture radar and mag-
netometry at the example of the unesco world heritage site of palmyra (syria),”
(2012).
[25] W. Pitz and D. Miller, “The terrasar-x satellite,” IEEE Transactions on Geo-
science and Remote Sensing 48, 615–622 (2010).
[26] M. Stangl, R. Werninghaus, and R. Zahn, “The TerraSAR-X active phased

array antenna,” in “Phased Array Systems and Technology, 2003. IEEE Inter-
national Symposium on,” (IEEE, 2003), pp. 70–75.
[27] Airbus, “TerraSAR-X Image Product Guide,” (https://www.intelligence-

airbusds.com/files/pmedia/public/r459 9 20171004 tsxx-airbusds-ma-
0009 tsx-productguide i2.01.pdf).
[28] G. Krieger, A. Moreira, H. Fiedler, I. Hajnsek, M. Werner, M. Younis, and

M. Zink, “TanDEM-X: A satellite formation for high-resolution SAR interfer-
ometry,” IEEE Transactions on Geoscience and Remote Sensing 45, 3317–3341
(2007).
[29] German Aerospace Center (DLR) (https://www.dlr.de/dlr/en

/desktopdefault.aspx/tabid-10680/#gallery/9106).
[30] German Aerospace Center (DLR) (https://www.dlr.de/dlr/en

/desktopdefault.aspx/tabid-10680/).
[31] E. Ramsey III, A. Rangoonwala, Y. Suzuoki, and C. E. Jones, “Oil detection

in a coastal marsh with polarimetric synthetic aperture radar (sar),” Remote
Sensing 3, 2630–2662 (2011).
[32] C. Wolff, “Radar basics,” (http://www.radartutorial.eu/20.airborne/

ab08.en.html).
[33] P. A. Rosen, S. Hensley, S. Shaffer, L. Veilleux, M. Chakraborty, T. Misra,

R. Bhan, V. R. Sagi, and R. Satish, “The nasa-isro sar mission-an interna-
tional space partnership for science and societal benefit,” in “2015 IEEE Radar
Conference (RadarCon),” (IEEE, 2015), pp. 1610–1613.
164
[34] M. C. Johnson, S. L. Brunton, N. B. Kundtz, and J. N. Kutz, “Sidelobe cancel-
ing for reconfigurable holographic metamaterial antenna,” IEEE Transactions
on Antennas and Propagation 63, 1881–1886 (2015).
[35] M. Boyarsky, T. Sleasman, L. Pulido-Mancera, T. Fromenteze, A. Pedross-

Engel, C. M. Watts, M. F. Imani, M. S. Reynolds, and D. R. Smith, “Synthetic
aperture radar with dynamic metasurface antennas: a conceptual develop-
ment,” J. Opt. Soc. Am. A 34, A22–A36 (2017).
[36] J. Hunt, T. Driscoll, A. Mrozack, G. Lipworth, M. Reynolds, D. Brady, and

D. R. Smith, “Metamaterial apertures for computational imaging,” Science
339, 310–313 (2013).
[37] T. Sleasman, M. Boyarsky, M. Imani, J. Gollub, and D. Smith, “Design

considerations for a dynamic metamaterial aperture for computational imaging
at microwave frequencies,” JOSA B 33, 1098–1111 (2016).
[38] J. N. Gollub, O. Yurduseven, K. P. Trofatter, D. Arnitz, M. F. Imani, T. Sleas-

man, M. Boyarsky, A. Rose, A. Pedross-Engel, H. Odabasi, T. Zvolensky,
G. Lipworth, D. Brady, D. L. Marks, M. S. Reynolds, and D. R. Smith, “Large
metasurface aperture for millimeter wave computational imaging at the human-
scale,” Scientific reports 7 (2017).
[39] G. Lipworth, A. Rose, O. Yurduseven, V. R. Gowda, M. F. Imani, H. Odabasi,

P. Trofatter, J. Gollub, and D. R. Smith, “Comprehensive simulation platform
for a metamaterial imaging system,” Appl. Opt. 54, 9343–9353 (2015).
[40] T. Sleasman, M. F. Imani, J. N. Gollub, and D. R. Smith, “Microwave imag-

ing using a disordered cavity with a dynamically tunable impedance surface,”
Physical Review Applied 6, 054019 (2016).
[41] T. Sleasman, M. F. Imani, J. N. Gollub, and D. R. Smith, “Dynamic metama-

terial aperture for microwave imaging,” Appl. Phys. Lett. 107, 204104 (2015).
[42] M. Boyarsky, T. Sleasman, I. Yoo, M. F. Imani, J. Gollub, and D. R.

Smith, “Resonance tuned metamaterial element for metasurface antennas,”
In Progress (2019).
[43] M. Boyarsky, T. Sleasman, M. F. Imani, J. Gollub, and D. R. Smith,

“Nyquist metasurface antenna,” In Progress (2019).
[44] M. Boyarsky, M. F. Imani, and D. R. Smith, “Grating lobe suppression

in metasurface antenna arrays with a waveguide feed layer,” arXiv preprint
arXiv:1905.09846 (2019).
165
[45] G. Lipworth, A. Mrozack, J. Hunt, D. L. Marks, T. Driscoll, D. Brady, and
D. R. Smith, “Metamaterial apertures for coherent computational imaging on
the physical layer,” JOSA A 30, 1603–1612 (2013).
[46] J. Hunt, T. Driscoll, A. Mrozack, G. Lipworth, M. Reynolds, D. Brady, and

D. R. Smith, “Metamaterial apertures for computational imaging,” Science
339, 310–313 (2013).
[47] M. Boyarsky, T. Sleasman, M. Imani, J. Gollub, and D. Smith, “Analysis of

1D dynamic metasurfaces for microwave computational imaging,” In Progress
(2019).
[48] M. H. Ackroyd and F. Ghani, “Optimum mismatched filters for sidelobe sup-
pression,” IEEE Transactions on Aerospace and Electronic Systems pp. 214–
218 (1973).
[49] P. T. Bowen, M. Boyarsky, L. Pulido-Mancera, and D. R. Smith, “Optimizing

beam performance of metasurface antennas with Euclidean modulation,” In
Progress (2019).
[50] M. Boyarsky, T. Sleasman, L. Pulido-Mancera, P. T. Bowen, and D. R.

Smith, “Beamforming with waveguide-fed metasurface antennas,” In Progress
(2019).
[51] C. W. Sherwin, J. Ruina, and R. Rawcliffe, “Some early developments in syn-

thetic aperture radar systems,” IRE Transactions on Military Electronics pp.
111–115 (1962).
[52] M. I. Skolnik, “Introduction to radar,” Radar handbook 2, 21 (1962).
[53] J. A. Develet, “Performance of a synthetic-aperture mapping radar system,”

IEEE Transactions on Aerospace and Navigational Electronics pp. 173–179
(1964).
[54] R. Ottolini, “Synthetic aperture radar data processing,” SEP report pp. 203–21
(1956).
[55] G. Krieger, M. Younis, N. Gebert, S. Huber, F. Bordoni, A. Patyuchenko, and

A. Moreira, “Advanced concepts for high-resolution wide-swath SAR imaging,”
in “Synthetic Aperture Radar (EUSAR), 2010 8th European Conference on,”
(VDE, 2010), pp. 1–4.
166
[56] NASA, “Spaceborne imaging radar,” (https://www.jpl.nasa.gov/missions/
spaceborne-imaging-radar-c-x-band-synthetic-aperture-radar-sir-c-x-sar/).
[57] C. M. Watts, P. Lancaster, A. Pedross-Engel, J. R. Smith, and M. S. Reynolds,

“2D and 3D millimeter-wave synthetic aperture radar imaging on a PR2 plat-
form,” in “Intelligent Robots and Systems (IROS), 2016 IEEE/RSJ Interna-
tional Conference on,” (IEEE, 2016), pp. 4304–4310.
[58] K. B. Cooper, S. L. Durden, C. J. Cochrane, R. R. Monje, R. J. Dengler, and

C. Baldi, “Using fmcw doppler radar to detect targets up to the maximum
unambiguous range,” IEEE Geoscience and Remote Sensing Letters (2017).
[59] K. Cooper, C. Baldi, G. Chattopadhyay, M. Choukroun, C. Cochrane, R. Den-

gler, S. Durden, T. El Bouayadi, D. Gonzalez, R. Monje et al., “A combination
millimeter-wave doppler radar and thz spectrometer for planetary science,”
in “Microwave Conference (EuMC), 2016 46th European,” (IEEE, 2016), pp.
1537–1540.
[60] D. Atlas and R. K. Moore, “The measurement of precipitation with synthetic

aperture radar,” Journal of atmospheric and oceanic technology 4, 368–376
(1987).
[61] NASA, “Soil moisture active passive,” (https://www.jpl.nasa.gov/news/

news.php?release=2014-444).
[62] German Aerospace Center (DLR), “Terrasar-x spacecraft bus in construction.”

(https://earth.esa.int/web/eoportal/satellite-missions/t/terrasar-x).
[63] R. Werninghaus and S. Buckreuss, “The TerraSAR-X mission and system de-
sign,” IEEE Transactions on Geoscience and Remote Sensing 48, 606–614
(2010).
[64] German Aerospace Center (DLR), “Terrasar-x image of larsen ice

shelf, antarctica.” (https://www.intelligence-airbusds.com/en/5750-image-
gallery-results?search=gallery&keyword=TerraSAR+Staring+Spotlight).
[65] A. Moreira, G. Krieger, I. Hajnsek, D. Hounam, M. Werner, S. Riegger, and

E. Settelmeyer, “TanDEM-X: a TerraSAR-X add-on satellite for single-pass
SAR interferometry,” in “Geoscience and Remote Sensing Symposium, 2004.
IGARSS’04. Proceedings. 2004 IEEE International,” , vol. 2 (IEEE, 2004),
vol. 2, pp. 1000–1003.
167
[66] A. Moreira, P. Prats-Iraola, M. Younis, G. Krieger, I. Hajnsek, and K. P. Pap-
athanassiou, “A tutorial on synthetic aperture radar,” Geoscience and Remote
Sensing Magazine, IEEE 1, 6–43 (2013).
[67] S. J. Orfanidis, Electromagnetic waves and antennas (Rutgers University New

Brunswick, NJ, 2002).
[68] J. Mittermayer, S. Wollstadt, P. Prats, R. Scheiber, and W. Koppe, “Staring

spotlight imaging with terrasar-x,” in “2012 IEEE International Geoscience
and Remote Sensing Symposium,” (IEEE, 2012), pp. 1606–1609.
[69] J. Hunt, J. Gollub, T. Driscoll, G. Lipworth, A. Mrozack, M. S. Reynolds,

D. J. Brady, and D. R. Smith, “Metamaterial microwave holographic imaging
system,” J. Opt. Soc. Amer. A 31, 2109–2119 (2014).
[70] T. Fromenteze, X. Liu, M. Boyarsky, J. Gollub, and D. R. Smith, “Phase-

less computational imaging with a radiating metasurface,” Optics Express 24,
16760–16776 (2016).
[71] T. Fromenteze, O. Yurduseven, M. F. Imani, J. Gollub, C. Decroze, D. Carse-

nat, and D. R. Smith, “Computational imaging using a mode-mixing cavity at
microwave frequencies,” Appl. Phys. Lett. 106, 194104 (2015).
[72] G. Lerosey, J. De Rosny, A. Tourin, and M. Fink, “Focusing beyond the diffrac-
tion limit with far-field time reversal,” Science 315, 1120–1122 (2007).
[73] T. Fromenteze, C. Decroze, and D. Carsenat, “Waveform coding for passive

multiplexing: Application to microwave imaging,” Antennas and Propagation,
IEEE Transactions on 63, 593–600 (2015).
[74] A. Porat, E. R. Andresen, H. Rigneault, D. Oron, S. Gigan, and

O. Katz, “Widefield lensless endoscopy via speckle-correlations,” arXiv
preprint arXiv:1601.01518 (2016).
[75] C. M. Watts, D. Shrekenhamer, J. Montoya, G. Lipworth, J. Hunt, T. Sleas-

man, S. Krishna, D. R. Smith, and W. J. Padilla, “Terahertz compressive imag-
ing with metamaterial spatial light modulators,” Nature Photonics (2014).
[76] N. Landy, J. Hunt, and D. R. Smith, “Homogenization analysis of complemen-

tary waveguide metamaterials,” Photonics and Nanostructures-Fundamentals
and Applications 11, 453–467 (2013).
168
[77] O. Yurduseven, V. R. Gowda, J. N. Gollub, and D. R. Smith, “Printed aperi-
odic cavity for computational and microwave imaging,” IEEE Microwave and
Wireless Components Letters 26, 367–369 (2016).
[78] I. Gil, J. Garcia-Garcia, J. Bonache, F. Martin, M. Sorolla, and R. Marques,

“Varactor-loaded split ring resonators for tunable notch filters at microwave
frequencies,” Electronics Letters 40, 1347–1348 (2004).
[79] T. Sleasman, M. F. Imani, J. N. Gollub, and D. R. Smith, “Dynamic metama-

terial aperture for microwave imaging,” Applied Physics Letters 107, 204104
(2015).
[80] D. Shrekenhamer, W.-C. Chen, and W. J. Padilla, “Liquid crystal tunable

metamaterial absorber,” Physical review letters 110, 177403 (2013).
[81] C. M. Watts, A. Pedross-Engel, D. R. Smith, and M. S. Reynolds, “X-band

SAR imaging with a liquid-crystal-based dynamic metasurface antenna,” JOSA
B 34, 300–306 (2017).
[82] T. Sleasman, M. Boyarsky, L. Pulido-Mancera, T. Fromenteze, M. F. Imani,

M. Reynolds, and D. R. Smith, “Experimental synthetic aperture radar with
dynamic metasurfaces,” arXiv preprint arXiv:1703.00072 (2017).
[83] T. Sleasman, M. F. Imani, W. Xu, J. Hunt, T. Driscoll, M. S. Reynolds,

and D. R. Smith, “Waveguide-fed tunable metamaterial element for dynamic
apertures,” IEEE Antennas and Wireless Propagation Letters 15, 606–609
(2016).
[84] H.-T. Chen, J. F. O’Hara, A. J. Taylor, R. D. Averitt, C. Highstrete, M. Lee,

and W. J. Padilla, “Complementary planar terahertz metamaterials,” Optics
Express 15, 1084–1095 (2007).
[85] T. H. Hand, J. Gollub, S. Sajuyigbe, D. R. Smith, and S. A. Cummer, “Charac-

terization of complementary electric field coupled resonant surfaces,” Applied
Physics Letters 93, 212504 (2008).
[86] N. Landy, J. Hunt, and D. R. Smith, “Homogenization analysis of complemen-

tary waveguide metamaterials,” Photonics and Nanostructures-Fundamentals
and Applications 11, 453–467 (2013).
[87] H. Odabasi, F. Teixeira, and D. Guney, “Electrically small, complementary

electric-field-coupled resonator antennas,” J. Appl. Phys 113, 084903 (2013).
169
[88] J. A. Bossard, X. Liang, L. Li, S. Yun, D. H. Werner, B. Weiner, T. S. Mayer,
P. F. Cristman, A. Diaz, and I. Khoo, “Tunable frequency selective surfaces
and negative-zero-positive index metamaterials based on liquid crystals,” IEEE
Transactions on Antennas and Propagation 56, 1308–1320 (2008).
[89] S. Lim, C. Caloz, and T. Itoh, “Metamaterial-based electronically controlled

transmission-line structure as a novel leaky-wave antenna with tunable radia-
tion angle and beamwidth,” Microwave Theory and Techniques, IEEE Trans-
actions on 52, 2678–2690 (2004).
[90] T. Sleasman, M. F. Imani, J. N. Gollub, and D. R. Smith, “Microwave imag-

ing using a disordered cavity with a dynamically tunable impedance surface,”
Physical Review Applied (2016).
[91] D. F. Sievenpiper, J. H. Schaffner, H. J. Song, R. Y. Loo, and G. Tangonan,

“Two-dimensional beam steering using an electrically tunable impedance sur-
face,” IEEE Transactions on Antennas and Propagation 51, 2713–2722 (2003).
[92] M. C. Johnson, S. L. Brunton, N. B. Kundtz, and J. N. Kutz, “Sidelobe can-

celing for reconfigurable holographic metamaterial antenna,” Antennas and
Propagation, IEEE Transactions on 63, 1881–1886 (2015).
[93] D. Gabor, “A new microscopic principle,” Nature 161, 777–778 (1948).
[94] A. J. Martinez-Ros, J. L. Gómez-Tornero, and G. Goussetis, “Holographic

pattern synthesis with modulated substrate integrated waveguide line-source
leaky-wave antennas,” IEEE Transactions on Antennas and Propagation 61,
3466–3474 (2013).
[95] G. Lipworth, A. Mrozack, J. Hunt, D. L. Marks, T. Driscoll, D. Brady, and

D. R. Smith, “Metamaterial apertures for coherent computational imaging on
the physical layer,” J. Opt. Soc. Amer. A 30, 1603–1612 (2013).
[96] J. Colbum, D. Sievenpiper, B. Fong, J. Ottusch, J. Visher, and P. Herz, “Ad-

vances in artificial impedance surface conformal antennas,” in “2007 IEEE
Antennas and Propagation Society International Symposium,” (IEEE, 2007),
pp. 3820–3823.
[97] G. Minatti, F. Caminita, M. Casaletti, and S. Maci, “Spiral leaky-wave anten-

nas based on modulated surface impedance,” IEEE Transactions on Antennas
and Propagation 59, 4436–4444 (2011).
170
[98] R. Quarfoth and D. Sievenpiper, “Artificial tensor impedance surface waveg-
uides,” IEEE Transactions on Antennas and Propagation 61, 3597–3606
(2013).
[99] A. M. Patel and A. Grbic, “Effective surface impedance of a printed-circuit

tensor impedance surface (PCTIS),” IEEE Transactions on Microwave Theory
and Techniques 61, 1403–1413 (2013).
[100] A. M. Patel and A. Grbic, “Modeling and analysis of printed-circuit tensor

impedance surfaces,” IEEE Transactions on Antennas and Propagation 61,
211–220 (2013).
[101] G. Lipworth, N. W. Caira, S. Larouche, and D. R. Smith, “Phase and mag-

nitude constrained metasurface holography at W-band frequencies,” Optics
Express 24, 19372–19387 (2016).
[102] C. A. Balanis, “Advanced engineering electromagnetics,” (2012).
[103] D. R. Smith, V. R. Gowda, O. Yurduseven, S. Larouche, G. Lipworth,

Y. Urzhumov, and M. S. Reynolds, “An analysis of beamed wireless power
transfer in the fresnel zone using a dynamic, metasurface aperture,” Journal
of Applied Physics 121, 014901 (2017).
[104] D. J. Brady, Optical imaging and spectroscopy (John Wiley & Sons, 2009).
[105] D. J. Brady, Optical imaging and spectroscopy (John Wiley & Sons, 2009).
[106] F. Lin and M. A. Fiddy, “Image estimation from scattered field data,” Inter-
national Journal of Imaging Systems and Technology 2, 76–95 (1990).
[107] O. Yurduseven, M. F. Imani, H. Odabasi, J. Gollub, G. Lipworth, A. Rose,

and D. R. Smith, “Resolution of the frequency diverse metamaterial aperture
imager,” Progress In Electromagnetics Research 150, 97–107 (2015).
[108] Y. Saad and M. H. Schultz, “GMRES: A generalized minimal residual algo-

rithm for solving nonsymmetric linear systems,” SIAM Journal on scientific
and statistical computing 7, 856–869 (1986).
[109] J. M. Bioucas-Dias and M. A. Figueiredo, “A new twist: two-step iterative

shrinkage/thresholding algorithms for image restoration,” Image Processing,
IEEE Transactions on 16, 2992–3004 (2007).
171
[110] T. Fromenteze, M. Boyarsky, J. Gollub, T. Sleasman, M. F. Imani, and
D. Smith, “Single-frequency near-field MIMO imaging,” in “European Confer-
ence on Antennas and Propagation,” (EurAAP, 2017).
[111] T. Sleasman, M. Boyarsky, M. Imani, T. Fromenteze, J. Gollub, and

D. Smith, “Single-frequency microwave imaging with dynamic metasurface
apertures,” JOSA B (2017 - Submitted).
[112] C. Cafforio, C. Prati, and F. Rocca, “SAR data focusing using seismic migra-
tion techniques,” IEEE transactions on aerospace and electronic systems 27,
194–207 (1991).
[113] L. Pulido-Mancera, T. Fromenteze, T. Sleasman, M. Boyarsky, M. F. Imani,

M. Reynolds, and D. Smith, “Application of range migration algorithms to
imaging with a dynamic metasurface antenna,” JOSA B 33, 2082–2092 (2016).
[114] S. S. Ahmed, “Electronic microwave imaging with planar multistatic arrays,”

(2014).
[115] O. Yurduseven, M. F. Imani, H. Odabasi, J. Gollub, G. Lipworth, A. Rose,

and D. R. Smith, “Resolution of the frequency diverse metamaterial aperture
imager,” Progress In Electromagnetics Research 150, 97–107 (2015).
[116] D. J. Brady, D. L. Marks, K. P. MacCabe, and J. A. O’Sullivan, “Coded

apertures for x-ray scatter imaging,” Applied optics 52, 7745–7754 (2013).
[117] X. Zhuge and A. G. Yarovoy, “A sparse aperture MIMO-SAR-based UWB

imaging system for concealed weapon detection,” IEEE Transactions on Geo-
[118] X. Zhuge and A. Yarovoy, “Three-dimensional near-field MIMO array imaging

using range migration techniques,” IEEE Transactions on Image Processing
21, 3026–3033 (2012).
[119] A. V. Diebold, L. Pulido-Mancera, T. Sleasman, M. Boyarsky, M. F. Imani,

and D. R. Smith, “Generalized range migration algorithm for synthetic aper-
ture radar image reconstruction of metasurface antenna measurements,” JOSA
B 34, 2610–2623 (2017).
[120] L. M. Pulido-Mancera, T. Zvolensky, M. F. Imani, P. T. Bowen, M. Valayil, and

D. R. Smith, “Discrete dipole approximation applied to highly directive slotted
waveguide antennas,” IEEE Antennas and Wireless Propagation Letters 15,
1823–1826 (2016).
172
[121] L. Pulido-Mancera, M. F. Imani, and D. R. Smith, “Discrete dipole approxima-
tion for simulation of unusually tapered leaky wave antennas,” in “2017 IEEE
MTT-S International Microwave Symposium (IMS),” (2017), pp. 409–412.
[122] L. Pulido-Mancera, P. T. Bowen, N. Kundtz, M. F. Imani, and D. Smith,

“Polarizability extraction for waveguide-fed metasurfaces,” arXiv preprint
arXiv:1708.05061 (2017).
[123] T. Fromenteze, E. L. Kpré, D. Carsenat, C. Decroze, and T. Sakamoto, “Single-

shot compressive multiple-inputs multiple-outputs radar imaging using a two-
port passive device,” IEEE Access 4, 1050–1060 (2016).
[124] A. D. Yaghjian, “An overview of near-field antenna measurements,” IEEE

Trans. Antennas Propag. 34, 30–45 (1986).
[125] M. Liang, Y. Li, H. Meng, M. Neifeld, and H. Xin, “Reconfigurable array de-
sign to realize principal component analysis (pca)-based microwave compressive
sensing imaging system,” Antennas and Wireless Propagation Letters, IEEE
14, 1039–1042 (2015).
[126] F. Ulaby and D. B. Lang, “A strategy for active remote sensing amid increased
demand for spectrum,” in “Radio Science Meeting (Joint with AP-S Sympo-
sium), 2015 USNC-URSI,” (IEEE, 2015), pp. 263–263.
[127] A. D. Scher and E. F. Kuester, “Extracting the bulk effective parameters of

a metamaterial via the scattering from a single planar array of particles,”
Metamaterials 3, 44–55 (2009).
[128] R. E. Collin, Field theory of guided waves (McGraw-Hill, 1960).
[129] C. A. Balanis, Antenna theory: analysis and design, vol. 1 (John Wiley & Sons,
2005).
[130] Q. Feng, M. Pu, C. Hu, and X. Luo, “Engineering the dispersion of metamate-
rial surface for broadband infrared absorption,” Optics letters 37, 2133–2135
(2012).
[131] S. Enoch, G. Tayeb, P. Sabouroux, N. Guérin, and P. Vincent, “A metamaterial

for directive emission,” Physical Review Letters 89, 213902 (2002).
[132] M. Eineder, T. Fritz, J. Mittermayer, A. Roth, E. Boerner, and H. Breit,

“Terrasar-x ground segment, basic product specification document,” Tech. rep.,
DTIC Document (2008).
173
[133] B. R. Mahafza, Radar systems analysis and design using MATLAB (CRC
press, 2002).
[134] R. L. Jordan, “The Seasat-A synthetic aperture radar system,” IEEE Journal
of Oceanic Engineering 5, 154–164 (1980).
[135] R. L. Jordan, B. L. Huneycutt, and M. Werner, “The SIR-C/X-SAR synthetic

aperture radar system,” IEEE Transactions on Geoscience and Remote Sensing
33, 829–839 (1995).
[136] X. Zhuge and A. G. Yarovoy, “A sparse aperture MIMO-SAR-based UWB

imaging system for concealed weapon detection,” IEEE Transactions on Geo-
[137] J. M. Lopez-Sanchez and J. Fortuny-Guasch, “3-d radar imaging using range

migration techniques,” IEEE Transactions on antennas and propagation 48,
728–737 (2000).
[138] S. Guillaso, A. Reigber, L. Ferro-Famil, and E. Pottier, “Range resolution

improvement of airborne SAR images,” IEEE Geoscience and Remote Sensing
Letters 3, 135–139 (2006).
[139] M. Soumekh, Synthetic aperture radar signal processing (New York: Wiley,
1999).
[140] P. T. Gough and D. W. Hawkins, “Unified framework for modern synthetic

aperture imaging algorithms,” International Journal of Imaging Systems and
Technology 8, 343–358 (1997).
[141] S. Lim, C. Caloz, and T. Itoh, “Metamaterial-based electronically controlled

transmission-line structure as a novel leaky-wave antenna with tunable radi-
ation angle and beamwidth,” IEEE Transactions on Microwave Theory and
Techniques 52, 2678–2690 (2004).
[142] A. A. Oliner and D. R. Jackson, “Leaky-wave antennas,” Antenna engineering

handbook 4 (2007).
[143] C. A. Balanis, Antenna theory: analysis and design (John wiley & sons, 2016).
[144] J. Hirokawa, T. Miyagawa, M. Ando, and N. Goto, “A waveguide-fed parallel

plate slot array antenna,” in “Antennas and Propagation Society International
Symposium, 1992. AP-S. 1992 Digest. Held in Conjuction with: URSI Radio
174
Science Meeting and Nuclear EMP Meeting., IEEE,” (IEEE, 1992), pp. 1471–
1474.
[145] W. Wu, J. Yin, and N. Yuan, “Design of an efficient x-band waveguide-fed

microstrip patch array,” IEEE Transactions on Antennas and Propagation 55,
1933–1939 (2007).
[146] Y. Li and K.-M. Luk, “Low-cost high-gain and broadband substrate-integrated-

waveguide-fed patch antenna array for 60-ghz band,” IEEE Transactions on
Antennas and Propagation 62, 5531–5538 (2014).
[147] S. R. Rengarajan, “Characteristics of a longitudinal/transverse coupling slot

in crossed rectangular waveguides,” IEEE transactions on microwave theory
and techniques 37, 1171–1177 (1989).
[148] S. R. Rengarajan and G. Shaw, “Accurate characterization of coupling junc-

tions in waveguide-fed planar slot arrays,” IEEE transactions on microwave
theory and techniques 42, 2239–2248 (1994).
[149] S. Microwave, “Optimizing Test Boards for 50 GHz End Launch Con-
nectors,” (https://mpd.southwestmicrowave.com/wp-content/uploads/
2018/07/Optimizing-Test-Boards-for-50-GHz-End-Launch-Connectors.pdf).
[150] F. Taringou, D. Dousset, J. Bornemann, and K. Wu, “Broadband cpw feed

for millimeter-wave siw-based antipodal linearly tapered slot antennas,” IEEE
Transactions on Antennas and Propagation 61, 1756–1762 (2012).
[151] G. F. Engen and C. A. Hoer, “Thru-reflect-line: An improved technique for

calibrating the dual six-port automatic network analyzer,” IEEE transactions
on microwave theory and techniques 27, 987–993 (1979).
[152] H.-J. Eul and B. Schiek, “Thru-match-reflect: One result of a rigorous theory
for de-embedding and network analyzer calibration,” in “1988 18th European
Microwave Conference,” (IEEE, 1988), pp. 909–914.
175

Metasurface Antennas For Synthetic Aperture Radar

Uploaded by

Copyright:

Available Formats

Metasurface Antennas For Synthetic Aperture Radar

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Metasurface Antennas For Synthetic Aperture Radar

Uploaded by

Copyright:

Available Formats

Metasurface Antennas for

Synthetic Aperture Radar

David R. Smith, Advisor

Dissertation submitted in partial fulfillment of the requirements for the degree of

David R. Smith, Advisor

An abstract of a dissertation submitted in partial fulfillment of the requirements for

List of Tables viii

2 Synthetic Aperture Radar 11

2.1 Traditional SAR Modalities . . . . . . . . . . . . . . . . . . . . . . . 12

2.1.1 Stripmap SAR . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.1.2 Spotlight SAR . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2 Traditional SAR Hardware . . . . . . . . . . . . . . . . . . . . . . . . 15

2.2.1 TerraSAR-X Phased Array . . . . . . . . . . . . . . . . . . . . 15

3.1 Metasurface Architectures . . . . . . . . . . . . . . . . . . . . . . . . 22

3.1.1 Passive Metasurface Apertures . . . . . . . . . . . . . . . . . . 22

3.1.2 Dynamic Metasurface Apertures . . . . . . . . . . . . . . . . . 24

3.1.3 Holographic Antennas . . . . . . . . . . . . . . . . . . . . . . 26

3.1.4 Numerical Modeling of Metasurface Apertures . . . . . . . . . 29

3.2 Computational Imaging Framework . . . . . . . . . . . . . . . . . . . 31

3.2.1 Mathematical Description . . . . . . . . . . . . . . . . . . . . 32

3.3 Experimental Imaging with a Dynamic Metasurface . . . . . . . . . . 34

3.3.1 1D Dynamic Metasurface Aperture . . . . . . . . . . . . . . . 35

3.3.2 Stationary Imaging . . . . . . . . . . . . . . . . . . . . . . . . 36

3.3.3 Single-Frequency Imaging . . . . . . . . . . . . . . . . . . . . 39

3.3.4 Experimental SAR with Metasurfaces . . . . . . . . . . . . . . 43

3.3.5 3D SAR Imaging . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.3.6 Single Frequency SAR . . . . . . . . . . . . . . . . . . . . . . 50

4 Simulated Metasurfaces for SAR 71

4.1 Simulated SAR with Metasurface Antennas . . . . . . . . . . . . . . 72

4.1.1 Simulated Metasurface Architecture . . . . . . . . . . . . . . . 72

4.1.2 Beamforming Performance . . . . . . . . . . . . . . . . . . . . 75

4.1.3 Stripmap and Spotlight SAR Imaging . . . . . . . . . . . . . . 77

4.1.4 Alternative SAR Modalities . . . . . . . . . . . . . . . . . . . 80

4.1.5 Noise Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.2 Metasurface Tuning Strategies . . . . . . . . . . . . . . . . . . . . . . 89

4.2.1 Tuning Strategy Definitions . . . . . . . . . . . . . . . . . . . 91

4.2.2 Simulation Methodology . . . . . . . . . . . . . . . . . . . . . 95

4.2.3 Simulated Beamforming Results . . . . . . . . . . . . . . . . . 98

4.2.4 Scaled Euclidean Modulation . . . . . . . . . . . . . . . . . . 99

4.3 Large Scale Simulated Metasurface . . . . . . . . . . . . . . . . . . . 105

4.3.1 Initial Spaceborne Metasurface Design . . . . . . . . . . . . . 105

4.3.2 Final Large Metasurface Design . . . . . . . . . . . . . . . . . 109

4.4 Grating Lobe Suppression . . . . . . . . . . . . . . . . . . . . . . . . 113

4.4.1 Metasurface Antenna Architecture . . . . . . . . . . . . . . . 114

5 Metasurface Antenna Prototype 131

5.1 Individual Element Design . . . . . . . . . . . . . . . . . . . . . . . . 131

5.2 Antenna Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

5.2.1 Challenges with Metasurface Antennas . . . . . . . . . . . . . 136

5.2.2 Metasurface Antenna Design . . . . . . . . . . . . . . . . . . . 137

5.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

5.3.1 Element Characterization . . . . . . . . . . . . . . . . . . . . 142

5.3.2 Experimental Beamforming Results . . . . . . . . . . . . . . . 148

2.1 TerraSAR-X antenna parameters. . . . . . . . . . . . . . . . . . . . . 18

3.1 Resolution of single-frequency and bandwidth imaging cases. . . . . . 62

5.1 Antenna parameters for the prototype metasurface antenna. . . . . . 138

5.3 Antenna performance metrics targeted for the metasurface antenna

5.4 Realized antenna performance metrics targeted for the metasurface

6.1 Comparison between phased arrays, reflector dish, and metasurface

1.1 Depiction of the TerraSAR-X satellite synthetic aperture radar system. 2