MOHAMED-RABIGH KHODJA
IMAGING AND RADIATION ENHANCEMENTS FROM
METAMATERIALS
PHD DISSERTATION
NORTHEASTERN UNIVERSITY BOSTON, MASSACHUSETTS
2008
IMAGING AND RADIATION ENHANCEMENTS FROM METAMATERIALS
A Dissertation Presented
by
Mohamed-Rabigh Khodja
to
The Department of Electrical and Computer Engineering
in partial fulfillment of the requirements
for the degree of
Doctor of Philosophy
in
Electrical Engineering
Northeastern University Boston, Massachusetts
September 2008
c° Copyright 2008 by Mohamed-Rabigh Khodja
All Rights Reserved
Vita
Mohamed-Rabigh Khodja was born in Biskra, Algeria. He obtained a Diplôme
d’Études Supérieurs (DES) in Theoretical Physics from the University of Constan-
tine (now Mentouri University of Constantine), Constantine, Algeria, in 1994, and
a Master’s degree in Physics from King Fahd University of Petroleum and Minerals,
Dhahran, Saudi Arabia, in 1998. (The title of his Master’s thesis was “Radiative and
Nonleptonic Decays of Bottom Baryons in the Quark Model”). He came to North-
eastern University in 2004 to do a PhD in Physics then moved to the Department
of Electrical and Computer Engineering in 2005. Since then he has been engaged
in research, supervised by Professor Edwin A. Marengo, on electromagnetic field
theory. Prior to coming to Northeastern, he was a Lecturer in Physics at King Fahd
University of Petroleum and Minerals, Dhahran, Saudi Arabia. He is a member
of the American Physical Society (APS), the Institute of Electrical and Electronics
Engineers (IEEE), and the American Association for the Advancement of Science
(AAAS).
iv
Publication Record
To Date
1. M. R. Khodja and E. A. Marengo. Radiation enhancement due to metama-
terial substrates from an inverse source theory, Phys. Rev. E, 77:046605,
2008.
2. E. A. Marengo, M. R. Khodja, and A. Boucherif. Inverse source problem in
non-homogeneous background media. Part II: Vector formulation and antenna
substrate performance characterization, SIAM J. App. Math., 69(1):81-110,
2008.
3. M. R. Khodja and E. A. Marengo. Comparative study of radiation enhance-
ment due to metamaterials, invited paper, Radio Sci., doi:10.1029/2007RS003803,
in press, 2008.
4. E. A. Marengo and M. R. Khodja. Generalized power-spectrum Larmor for-
mula for extended charged particle embedded in a harmonic oscillator, Phys.
Rev. E, 74:036611, 2006.
5. P. Abdel-Jalil, M. R. Khodja, and A. Al-Suwayyan. A Physical model for the
distribution of ions and electrons in laminar premixed hydrocarbon flames,
Arab. J. Sc. Eng., 26(2A):127-135, 2001.
6. M. R. Khodja, Riazuddin, and Fayyazuddin. Two-body nonleptonic decays of
bottom baryons in the quark model, Phys. Rev. D, 60:053005, 1999.
v
7. A. Aksoy and M. R. Khodja. Calibration of natural gamma-ray spectrometer
at the ERL, Arab. J. Sc. Eng., 24(1A):73-79, 1999.
8. E. A. Marengo and M. R. Khodja. Source synthesis in substrate media: Fun-
damental bounds, International Symposium on Electromagnetic Theory, URSI
Commision B, Ottawa, Ontario, Canada, July 26-28, 2007.
9. E. A. Marengo and M. R. Khodja. Source inversion in metamaterial sub-
strate media: Fundamental limits and applications, URSI/CNC/USNC North
American Radio Science Conference, Ottawa, Ontario, Canada, July 22-26,
2007.
Projected
1. M. R. Khodja and E. A. Marengo. The Casimir effect in the presence of
metamaterials, to be submitted to Phys. Rev. Lett., 2008.
2. M. R. Khodja and E. A. Marengo. The Casimir effect and the dispersive
nature of metamaterials, to be submitted to Phys. Rev. E, 2008.
3. M. R. Khodja and E. A. Marengo. Inverse scattering and imaging resolution
enhancements in metamaterial backgrounds, to be submitted to J. Opt. Soc.
Am., 2008.
4. R. Hernandez, H. Lev-Ari, A. M. Stankovic, M. R. Khodja, and E. A. Marengo.
Fundamental performance limits in lossy polyphase systems: Apparent power
and optimal compensation, to be submitted to IEEE Trans. on Circuits and
Systems, 2008.
5. E. A. Marengo, M. R. Khodja, and A. Widom. Possibilities and limitations of
employing the Aharonov-Bohm effect for remote sensing and communications,
in preparation.
vi
Dedication
I dedicate this dissertation to the spirit of my late father, Abdelhamid, and to mybeloved mother, Saida, for all their sacrifices to advance my education and to getme to this point in my life.
I also dedicate this dissertation to my wife, Fatima-Zohra, for sharing withme the ups and downs of my journey, and to my parents-in-law Abdelkader andNadjia for all they have done to see me succeed, and they have done so much.
Last but not least, I dedicate this dissertation to the lights of my world, my sons
Abdelhamid (Didu) and Belkacem-Eltaher (Weewer), for enriching my lifein so many ways.
vii
Acknowledgments
I wish to express my deepest gratitude to my mentor and friend Dr. Edwin A.
Marengo for his guidance and support over the past three years. Throughout these
years, I have greatly enjoyed his dedication, creativity, and untiring enthusiasm. His
trust and unflagging pursuit of excellence have made this dissertation what it is. I
could not have hoped for a better dissertation supervisor.
I am privileged to have had Drs. Anthony J. Devaney and Philip E. Serafim
as teachers and members in my dissertation committee. I thank them for their
continuous encouragement and for all the wonderful things I have learned from
them. My gratitude goes to my father-in-law Dr. Abdelkader Boucherif for his
wholehearted support and for introducing me to functional analysis. I am also
grateful to Dr. Stephen W. McKnight for his kind help in the transfer process
from the Department of Physics to the Department of Electrical and Computer
Engineering when he was Chairman of the latter.
All of my family and my wife’s family supported me throughout the years, despite
the great distances between us. I especially thank my brothers, Noureddine-Omar,
Karim-Abdelaziz, and Tarek-Abdelhafidh, my sisters, Nadia, Linda, and Merzaka,
my brothers-in-law, Ammar and Bachir, and my sisters-in-law, Nabila and Dalal. I
also thank my wife’s grand parents, Youcef and Halima, for always remembering us
in their prayers.
I thank my friends and my family’s friends for being such an important part of
our lives. They are too many to mention by name. My family and I are especially
indebted to Mr. Belkacem Naidjate and his family, Dr. Jamal Lebeche and his
family, Dr. Amine Nehari-Talet and his family, Dr. Ghassen Benbrahim, and Mr.
viii
Abubakr Naidjate, for their unconditional kindness.
I cannot thank my wife, Fatima-Zohra, enough for her emotional support, pa-
tience, and unwavering faith during the hard times.
Finally, I gratefully acknowledge financial support provided by Northeastern
University in the form of a Dissertation Writing Fellowship during the summer of
2008. The work embodied in this dissertation was also supported in part by the
United States Air Force Office of Scientific Research under Grant No. FA9550-06-
01-0013, and by the National Science Foundation under Grant No. 0746310.
ix
Abstract
This dissertation is divided into two parts. The first part solves the full-vector, elec-
tromagnetic inverse source problem of synthesizing an unknown source embedded in
a given substrate medium of volume V and radiating a prescribed exterior field. Im-
portantly, the derived formulation is non-antenna-specific. It comprises forward, or
radiation, characterization as well as inverse-theoretic characterization. The forward
characterization is focused on the singular value spectrum while the inverse-theoretic
characterization is performed via the “minimum-energy” sources. Particular atten-
tion is given to the case of two nested spheres made up of materials with oppositely
signed constitutive parameters. We find that, for a given antenna radiating at a
prescribed frequency, the singular values spectra exhibit resonances that correspond
to maximum radiation enhancements. These resonances are primarily due to the
presence of polaritons which, in turn, correspond to sets of constitutive parameters
that maximize the radiated fields. We also find that for electrically small antenna
systems made up of materials with oppositely signed constitutive parameters the
emergence of resonances depends on the ratio of the two radii rather than on the
overall size of the system.
The second part of this dissertation explores the effects that the presence of
electromagnetic metamaterials would have on the Casimir forces. Along with their
companion van der Waals forces, the Casimir forces, are identified as the primary
cause for the collapse of microelectromechanical systems (MEMS) and nanoelectro-
mechanical systems (NEMS). Hence, it is crucial for the future development of nan-
otechnology to investigate the possibility of shielding these potentially destructive
forces. One such avenue is explored by means of electromagnetic metamaterials. A
x
surface modes sum is performed to derive an analytic expression for the Casimir force
in a dispersive four-region system with planar geometry. In the case of two perfectly
conducting plates separated by two media, the numerical simulations demonstrate
that the Casimir force is characterized by a new singular behavior in the presence of
DPS-DNG combinations. The possible utilization of such systems for the shielding
of MEMS and NEMS and for the realization of quantum levitation is discussed.
The theoretical studies are accompanied by numerical illustrations.
xi
Contents
Vita iv
Publication Record v
Dedication vii
Acknowledgments viii
Abstract x
1 Introduction 1
1.1 Electromagnetic Metamaterials . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Physical Constraints on the Response Functions . . . . . . . 2
1.1.2 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Inverse Source Problem in Substrate Media . . . . . . . . . . . . . . 3
1.3 The Casimir Effect and Electromagnetic Metamaterials . . . . . . . 8
2 Inverse Source Problem in Non-homogeneous Background Media 12
2.1 The Forward Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.1 Electromagnetic Generalities . . . . . . . . . . . . . . . . . . 14
2.1.2 Source-to-Multipole-Moment Mapping . . . . . . . . . . . . . 16
2.2 Inverse Source Theory Based on Constrained Optimization . . . . . 19
2.2.1 Minimum Energy Solution by Constrained Optimization . . . 19
2.2.2 Minimum Energy Source Having Zero Reactive Power . . . . 22
2.3 Computer Simulation Study . . . . . . . . . . . . . . . . . . . . . . . 32
xii
2.3.1 Minimum Energy Sources . . . . . . . . . . . . . . . . . . . . 33
2.3.2 Tuned Minimum Energy Sources: Additional Zero Reactive
Power Constraint . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3 Radiation Enhancement due to Metamaterial Substrates: Core-Shell System 543.1 The Radiation Problem . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Inverse Source Theory Based on Constrained Optimization . . . . . 58
3.3 Numerical Results and Case Studies . . . . . . . . . . . . . . . . . . 59
3.3.1 Lossless Substrates . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3.2 Lossy Substrates . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4 Comparative Study of Radiation Enhancement due to Metamate-rials 754.1 Reference Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2 Numerical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5 The Casimir Effect in the Presence of Metamaterials 825.1 General Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.1.1 Eigenfrequencies . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.1.2 The Casimir Force . . . . . . . . . . . . . . . . . . . . . . . . 86
5.2 Particular Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.2.1 One Slab of Material Separating Two Half-Spaces . . . . . . 88
5.2.2 Two Slabs of Material Separating Two Perfectly Conducting
Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.2.3 The One-Slab Case and Casimir’s Classical Result . . . . . . 91
5.3 Numerical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.3.1 Cavities with DPS-DPS and DNG-DNG Slab Combinations . 93
5.3.2 Cavities with DPS-DNG Slab Combinations . . . . . . . . . . 93
5.3.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
xiii
6 Future Directions 1006.1 Extensions to the Inverse Source Problem in Antenna Substrate Media100
6.2 Extensions to the Inverse Scattering Problem in Substrate Media . . 101
6.3 Electrodynamics of Metamaterials . . . . . . . . . . . . . . . . . . . 101
A Definition and Properties of Vector Spherical Harmonics 103
B WavefunctionsB(j)l,m for Piecewise-Constant Radially-Symmetric Back-
grounds 106
C Wavefunctions B(j)l,m for a System of Two Nested Spheres 113
D Calculations of the Fréchet Derivatives of the Objective Functionaland the Constraints 115
E Wavefunctions B(j)l,m for Spherically Symmetric Backgrounds 117
F Establishment of The Dispersion Relations (5.2) and (5.3) 119
G Application of the Argument Principle to the Derivation of Eq.(5.20) 123
Bibliography 127
xiv
List of Figures
2.1 Schematic of a general antenna whose driving points and material
structure are confined within a spherical volume V of radius a. . . . 13
2.2 Free-space singular valueshσ(1)l (x = x0, r = +1 = µr)
i2versus l for
a few representative values of x0 ≡ k0a/π. (The unit of the singular
values is V 2m/A2.) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Normalized singular valuesh(1)l
i2versus l for x0 = 1/4 (quarter-wave
case), r = +1 and a few representative values of x. . . . . . . . . . 35
2.4 Normalized singular valuesh(1)l
i2versus l for x0 = 10 (resonant or
electrically-large antenna), r = +1 and a few representative values
of x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.5 Normalized singular valuesh(2)l
i2versus l for x0 = 1/4 (quarter-wave
case), r = +1 and a few representative values of x. . . . . . . . . . . 37
2.6 Normalized singular valuesh(2)l
i2versus l for x0 = 10 (resonant or
electrically-large antenna), r = +1 and a few representative values
of x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.7 Logarithmic mesh plot of the source energy E(j=1=l)ME versus x0 and x
for a double-positive material with r = +1. . . . . . . . . . . . . . . 41
2.8 Logarithmic plot of the source energy E(j=1=l)ME versus x for r = +1
and some representative values of x0 for a double-positive medium. . 42
2.9 Logarithmic plot of the source energy E(j=1=l)ME versus x for r = −1and some representative values of x0 for a double-negative medium. . 43
2.10 Logarithmic plot of the source energy E(j=1=l)ME versus x for µr = 1
and some representative values of x0. . . . . . . . . . . . . . . . . . . 44
xv
2.11 Logarithmic plot of the source energy E(j=1=l)ME (x = x0) versus x = x0
for r = +1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.12 Logarithmic plot of the gainG, versus x for r = +1 and some selected
values of x0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.13 Plot of the normalized reactive power g(1)1 versus χ for x0 = 1/4 = x
and r = +1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.14 Plot of the source energy E(j=1=l)EP versus χ for for x0 = 1/4 = x and
r = +1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.15 Plot of the normalized reactive power g(1)1 versus χ for x0 = 1/4 = −x
and r = −1 (“anti-vacuum.”) . . . . . . . . . . . . . . . . . . . . . . 48
2.16 Plot of the source energy E(j=1=l)EP versus χ for x0 = 1/4 = −x and
r = −1 (“anti-vacuum.”). . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Geometry of the three-region system under consideration. The driving
points and material structure of the antenna are confined within a
spherical volume V of radius a. The inner sphere of radius a has rel-
ative permittivity a and relative permeability µa. This inner sphere
is surrounded by a spherical shell of inner radius a and outer radius
b and has relative permittivity b and relative permeability µb. The
core-shell system is immersed in the vacuum. . . . . . . . . . . . . . 55
3.2 Logarithmic plot of the normalized electric singular valuesh(1)l
i2for a λ/4 antenna versus the radii ratio d ≡ b/a. The inner sphere
is assumed not to contain any material (i.e., a = 1 = µa) and the
surrounding shell is assumed to be a lossless DPS material with µb = 1
and xb = 50 m−1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3 Logarithmic plot of the normalized electric singular valuesh(1)l
i2for a λ/4 antenna versus the radii ratio d ≡ b/a. The inner sphere
is assumed not to contain any material (i.e., a = 1 = µa) and the
surrounding shell is assumed to be a lossless DPS material with µb = 1
and xb = 150 m−1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
xvi
3.4 Logarithmic plot of the normalized electric singular valuesh(1)l
i2for a λ/4 antenna versus the radii ratio d ≡ b/a. The inner sphere
is assumed not to contain any material (i.e., a = 1 = µa) and the
surrounding shell is assumed to be a lossless DPS material with µb = 1
and xb = 500 m−1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.5 Logarithmic plot of the normalized singular valuesh(1)l
i2for a λ/400
antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed
not to contain any material (i.e., a = 1 = µa) and the surrounding
shell is assumed to be a lossless DPS material with µb = 1 and xb =
150 m−1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.6 Logarithmic plot of the normalized singular valuesh(1)l
i2for a λ/4
antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed
not to contain any material (i.e., a = 1 = µa) and the surrounding
shell is assumed to be a lossless DNG material with µb = −1 andxb = −50 m−1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.7 Logarithmic plot of the normalized electric singular valuesh(1)l
i2for
a λ/40 antenna versus the radii ratio d ≡ b/a. The inner sphere
is assumed not to contain any material (i.e., a = 1 = µa) and the
surrounding shell is assumed to be a lossless DNG material with µb =
−1 and xb = −50 m−1. . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.8 Logarithmic plot of the normalized electric singular valuesh(1)l
i2for
a λ/400 antenna versus the radii ratio d ≡ b/a. The inner sphere
is assumed not to contain any material (i.e., a = 1 = µa) and the
surrounding shell is assumed to be a lossless DNG material with µb =
−1 and xb = −50 m−1. . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.9 Logarithmic plot of the normalized singular valuesh(1)l
i2for a λ/400
antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed
not to contain any material (i.e., a = 1 = µa) and the surrounding
shell is assumed to be an ENG material with µb = 1 and xb = i50 m−1. 69
xvii
3.10 Logarithmic plot of the normalized electric singular valuesh(1)l
i2for
a λ/400 antenna versus the radii ratio d ≡ b/a. The inner sphere
is assumed not to contain any material (i.e., a = 1 = µa) and the
surrounding shell is assumed to be a lossless DNG material (black
curves), a lossy DNG shell with a loss tangent set to 1/20 (blue
curves), and a lossy DNG material with a loss tangent set to 1/60
(red curves). In all three cases µb = −1, Re[xb] = −150 m−1. . . . . 70
3.11 Logarithmic plot of the gainG for a λ/4-electric-dipole antenna versus
the radii ratio d ≡ b/a. The surrounding shell is assumed to be a DNG
material with µb = −1. . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.12 Logarithmic plot of the gain G for a λ/40-electric-dipole antenna
versus the radii ratio d ≡ b/a. The surrounding shell is assumed to
be a DNG material with µb = −1. . . . . . . . . . . . . . . . . . . . 72
3.13 Logarithmic plot of the gain G for a λ/400-electric-dipole antenna
versus the radii ratio d ≡ b/a. The surrounding shell is assumed to
be a DNG material with µb = −1. . . . . . . . . . . . . . . . . . . . 73
4.1 Logarithmic plot of the enhancementh(1)l
i2for a quarter-wavelength
antenna with radius a = 3.75 mm versus the radii ratio b/a. The shell
is a lossless DNG material with b = −4 and µb = −1. The referenceantenna is assumed to be RA1 antenna. . . . . . . . . . . . . . . . . 77
4.2 Logarithmic plot of the enhancementh(1)l
i2for a quarter-wavelength
antenna with radius a = 3.75 mm versus the radii ratio b/a. The shell
is a lossless DNG material with b = −4 and µb = −1. The referenceantenna is assumed to be an RA2 antenna. . . . . . . . . . . . . . . 78
4.3 Logarithmic plot of the enhancementh(1)l
i2for a quarter-wavelength
antenna with radius a = 3.75 mm versus the radii ratio b/a. The shell
is a lossless DNG material with b = −4 and µb = −1. The referenceantenna is assumed to be an RA3 antenna. . . . . . . . . . . . . . . 79
4.4 Logarithmic plot of the enhancementh(1)l
i2for a full—wavelength
antenna with radius a = 1.5 cm versus the radii ratio b/a. The shell
is a lossless DNG material with b = −4 and µb = −1. The referenceantenna is assumed to be an RA1 antenna. . . . . . . . . . . . . . . 80
xviii
4.5 Logarithmic plot of the enhancementh(1)l
i2for a full-wavelength
antenna with radius a = 1.5 cm versus the radii ratio b/a. The shell
is a lossless DNG material with b = −4 and µb = −1. The referenceantenna is assumed to be an RA2 antenna. . . . . . . . . . . . . . . 80
4.6 Logarithmic plot of the enhancementh(1)l
i2for a full-wavelength
antenna with radius a = 1.5 cm versus the radii ratio b/a. The shell
is a lossless DNG material with b = −4 and µb = −1. The referenceantenna is assumed to be an RA3 antenna. . . . . . . . . . . . . . . 81
5.1 Planar geometry under consideration. Two slabs of thicknesses a and
b and relative constitutive parameters a, µa and b, µb, respec-tively, are sandwiched between two half-spaces with relative consti-
tutive parameters 1, µ1 and 2, µ2 . . . . . . . . . . . . . . . . . 83
5.2 A rectangular cavity of dimensions L×L×d is constructed by puttingtwo perfectly conducting plates with dimensions L×L× τ (LÀ dÀτ) in the vacuum parallel to one another and separated by a distance
d. Two slabs of isotropic, homogeneous, lossless material are inserted
between the two plates. The first slab has dimensions L × L × a
and relative constitutive parameters a, µa; the second slab hasdimensions L × L × b and relative constitutive parameters b, µb .(a+ b = d). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.3 Plot of the normalized Casimir force f for DPS-DPS and DNG-
DNG slab combinations as a function of the ratio of thicknesses
r ≡ a/b. The constitutive parameters of the DPS-DPS combina-
tion are a, µa = 2, 1 and b, µb = 2.5, 1.5 . The constitutiveparameters of the DNG—DNG combination are a, µa = −2,−1and b, µb = −2.5,−1.5 . For both combinations the thickness ofthe second slab is set to b = 0.001 mm. . . . . . . . . . . . . . . . . . 94
xix
5.4 Plot of the normalized Casimir force f for a DPS-DNG slab com-
bination as a function of the ratio of thicknesses r ≡ a/b. For this
combination we have b/ a = −a/b = −r = µb/µa.The constitutive
parameters of the DPS slab are a, µa = 2, 1 and those of theDNG slab are b, µb = −2r,−r . The thicknesses of the DNGslabs were set to the arbitrary values 0.001 mm, 0.1 mm and 10 mm.
Because the curves coincide exactly with one another, a purely arti-
ficial shift of magnitude −10 has been introduced in the plot of thenormalized force that corresponds to b = 0.1 mm and a similar arti-
ficial shift of magnitude −20 has been introduced in the plot of thenormalized force that corresponds to b = 10 mm. . . . . . . . . . . . 95
5.5 Plot of the normalized Casimir force f for a DPS-DNG slab com-
bination as a function of the ratio of thicknesses 1/r ≡ b/a. The
constitutive parameters of the DPS slab are a, µa = 2, 1 andthose of the DNG slab are b, µb = −2.5,−1.5 . The thickness ofthe DPS slab is set to a = 0.001 mm. . . . . . . . . . . . . . . . . . . 96
5.6 Plot of the normalized Casimir force f for a DPS-DNG slab combi-
nation as a function of the ratio of thicknesses r ≡ a/b. The consti-
tutive parameters of the DPS slab are a, µa = 2, 1 and those ofthe DNG slab are b, µb = −2.5,−1.5 . The thickness of the DNGslab is set to a = 0.001 mm. . . . . . . . . . . . . . . . . . . . . . . . 97
5.7 Plot of the normalized Casimir force f described in Fig.( 5.6) for
different values of r. . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
G.1 The contour, C, used in Eqs.(G.4,G.5); it is defined by the imaginary
axis and the semicircle of infinite radius directed to the right. The
points ω1, ..., ωN represent the N solutions of the dispersion relations
g (ω) = 0 or gµ (ω) = 0. . . . . . . . . . . . . . . . . . . . . . . . . . 125
xx
List of Tables
2.1 Results of the numerical study for the constrained quarter-wave an-
tenna embedded in a double-positive material with r = +1. (The
unit of the source energies is A2/m.) . . . . . . . . . . . . . . . . . . 51
2.2 Results of the numerical study for the constrained half-wave antenna
embedded in a double-positive material with r = +1. (The unit of
the source energies is A2/m.) . . . . . . . . . . . . . . . . . . . . . . 51
2.3 Results of the numerical study for the constrained quarter-wave an-
tenna embedded in a double-negative metamaterial with r = −1.(The unit of the source energies is A2/m). . . . . . . . . . . . . . . . 51
2.4 Results of the numerical study for the constrained half-wave antenna
embedded in a double-negative metamaterial with r = −1. (Theunit of the source energies is A2/m.) . . . . . . . . . . . . . . . . . . 52
xxi
Chapter 1
Introduction
1.1 Electromagnetic Metamaterials
An electromagnetic wave with wavelength λ much larger than a certain object does
not really “see” that object. Consequently, a wave propagating in a homogeneous
substrate with small inclusions that are, let us say, ∼ λ/10 will interact effectively
not with the individual inclusions but with the system substrate,inclusions as a
whole. The wave, thus, sees a new effective medium having electromagnetic proper-
ties different from those of its components. This suggests that effective media with
properly chosen inclusions may exhibit electromagnetic responses not found in na-
ture. Metamaterials are such media. They are engineered materials whose effective
constitutive parameters can, in principle, have any value, even negative.
The onset of the “metamaterials era” is frequently attributed to the landmark
theoretical article by Veselago [1], though the study of complex media has a long
history that dates back, at least, to the late part of the nineteenth century [2, 3, 4].
Yet, until just a few years ago the study of these materials was largely ignored by
scientists and engineers alike. Then, at the turn of the century, Smith et al. [5]
announced the fabrication of a metamaterial with negative index of refraction in
the microwave regime. This discovery instigated several studies that resulted in
the identification of several novel and exciting applications of metamaterials. They
include perfect lenses [6], subwavelength cavities [7], highly efficient electrically-
small antennas [8], subwavelength waveguides [9], cloaking devices [10], ultrathin
1
laser cavities [11], etc.
1.1.1 Physical Constraints on the Response Functions
Despite the fact that electromagnetic field theory in metamaterial media relaxes the
constraint on the response functions range for these media, it still requires them to
satisfy certain conditions for them to correspond to physically acceptable materials.
Below we merely list these physical constraints. For more detail see, for instance,
[12].
Dissipation and Passive Media
Let f be the response function of a certain isotropic medium. f could either be the
electric permittivity or the magnetic permeability µ. We have, in the most general
case,
f = Re [f ] + i Im [f ] , (1.1)
where i2 = −1, Re [·] stands for the real part, and Im [·] stands for the imaginarypart. The appearance of an imaginary part for f is an indication of the lossy nature
of the medium in question.
For metamaterials, the real part of the response function f is not necessarily
positive, it is a number that could take on any real value, be it positive or negative,
that is,
Re [f ] ∈ R. (1.2)
However, for passive media, i.e., media within which electromagnetic energy is es-
sentially dissipated not created, we must have
Im [f ] > 0. (1.3)
Positive-Definiteness of Electric and Magnetic Energies
The positive-definiteness of the electric and magnetic energies also imposes the fol-
lowing constraints on the response functions
∂ (ωf)
∂ω≥ 0 (1.4)
2
Causality and the Kramers-Kronig Relations
If the medium is to be causal, i.e., if the cause is to precede the effect in such
a medium, then the real and imaginary parts of the response functions must be
Hilbert transforms of each other. It can be shown that this is equivalent to the
following Kramers-Kronig dispersion relations
Re [f (ω)] =2
πP
Z ∞
0
Im [f (ω0)]ω02 − ω2
ω0dω0 (1.5)
and
Im [f (ω)] = −2ωπP
Z ∞
0
Re [f (ω0)]ω02 − ω2
dω0, (1.6)
wherein P stands for the Cauchy principal value. Clearly, according to our current
understanding, metamaterials cannot be nondispersive nor nondissipative. Never-
theless, such assumptions are routinely made in the literature for convenience.
1.1.2 Terminology
There exist different terminologies to describe the different types of metamaterials,
here we adopt the following one (due to Richard Ziolkowski): When Re [ ] > 0 and
Re [µ] > 0 the material is said to be double-positive (DPS); when Re [ ] < 0 and
Re [µ] < 0 the material is said to be double-negative (DNG); and when Re [ ] Re [µ] <
0 the material is said to be single-negative (SNG).1 These particular choices for the
signs of the constitutive parameters are required if wave propagation in the medium
is to be causal [13].
1.2 Inverse Source Problem in Substrate Media
In the first part of this dissertation (Chapters 2-4) the full-vector, electromagnetic
inverse source problem is used to reconstruct an unknown source (antenna) that is
embedded, within a source region V , in a given material or metamaterial substrate,
and that radiates a given exterior field outside V . The derived formulation and
1A more detailed terminology labels materials for which Re [ α] < 0 as ENG or -negativemedia, and materials for which Re [µα] < 0 as MNG or µ-negative media.
3
results on this inverse source problem in substrate media generalize, within the full-
vector formulation, previous work on the inverse source problem in free space (cf.
[14] and the references therein for review and applications), as well as previous work
on the scalar version of the problem for non-homogeneous backgrounds [15, 16],
particularly [17]. Here, as in [17], the inverse problem is addressed in the context
of the Helmholtz operator with emphasis on piecewise-constant radially symmetric
backgrounds. The formulation is based on constrained optimization and, unlike the
vast majority of previous presentations, can be used to implement different kinds
of constraints. Two such possibilities are emphasized in the sequel, in particular,
the minimizing of the source L2-norm or functional energy characterizing the “cur-
rent level”, with and without tuning to resonance, the former case corresponding
to zero source reactive power. Fundamental radiation limits related to the realiz-
ability of given fields or radiation performance with given source resources (antenna
size, current level as measured by the source energy, reactive power, and so on)
or, alternatively, of the minimal resources needed for a given performance, are also
elucidated as a by-product of the derived inverse source theory.
Motivation for this research is provided by the possibility of embedding an an-
tenna in a substrate of a given size, where the original antenna plus the substrate
are treated as the total antenna, so as to generate a given field or performance level
which could not be achieved under the same physical constraints by another antenna
in free space (i.e., without the substrate medium). This possibility has attracted in-
terest from time to time in the antenna community; of particular interest have been
a variety of antenna-embedding materials, including plasmas [18], non-magnetic di-
electrics [19, 20, 21, 22, 23, 24], magneto-dielectrics [25, 26, 27, 28, 29, 30], and,
more recently, double-negative and single-negative metamaterials which are receiv-
ing much attention as antenna performance-enhancing substrates by a number of
groups [8, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49]. The
envisaged property is miniaturization of antennas by controlling electric size, via
larger wavenumber, but other effects are involved, particularly when metamaterials
are used. (A review of the pertinent state-of-the-art can be found in [50]).
For instance, it is well known [51, 52] that in the free-space case the source
energy increases exponentially, for a given radiation pattern, with decreasing k0a,
where k0 is the free-space wavenumber of the field. This increase occurs below a
4
critical point determined by the fine detail that is desired in the radiation pattern,
specifically, the antenna directivity. The question then is whether the critical source
size in question can be made smaller by embedding the source in a properly se-
lected substrate which becomes integral to the antenna. For small antennas (whose
dimensions are, for instance, smaller than about 1/3 of the wavelength [53, 54])2
one is particularly interested in achieving radiation of an elemental dipolar mode,
using minimal resources. Can antenna substrates help toward this goal? Alterna-
tively, in certain applications using larger, resonant antennas whose dimensions are
comparable to or larger than the wavelength one can dispose of some “extra space”
to accommodate a substrate, and the question is: Does antenna embedding yield
enhancement of antenna directivity? Which values of the constitutive parameters
give better performance?
Contrary to previous presentations, which focused on particular devices, we em-
phasize in the present work a non-device-specific analysis that aims at understanding
the practical possibilities opened by antenna-embedding substrates. The source in-
version is approached by including solution constraints yielding minimum energy
sources generating a given radiation pattern. Formal tractability as well as engi-
neering applications both dictate the particular choices made. All the results are
derived for time-harmonic fields, and thus the values of the constitutive parameters,
which generally vary with frequency, are considered in this work for a given central
frequency only. We also employ effective media theory considerations and ignore
material dissipation as well as the general bi-anisotropic nature of metamaterials
[55, 56] whose explicit consideration within the inverse theory is left for the future.
Thus in this dissertation we treat all materials and metamaterials as having zero
conductivity, and generally real-valued scalar permittivity and permeability, (r)
and µ(r), respectively, at the relevant frequency.
As far as we know, there are only three treatments of the inverse source problem
for the non-free-space case [15, 16, 17]. They consider only the scalar version of the
problem. Reference [15] emphasizes the minimum energy solution within an integral
equation framework pertinent to holography (as is also shown in [57]). Reference
[16] generalizes [15] to lossy media. Reference [17] uses an optimization-theoretic
framework which is a simpler, scalar counterpart of the full-vector treatment derived
2Cf. definition given in Section 3.3.
5
in the present sequel.
Electromagnetically the ultimate sources of radiation are: (1) an impressed cur-
rent distribution which under a suppressed time dependence e−iωt is described byits space-dependent part J(r) and which can be controlled at antenna terminals;
and (2) an induced current distribution Jind(r) appearing over the antenna ma-
terial structure upon the presence of the former primary source (as schematically
illustrated in the same figure). The superposition of the two, i.e., the total current
distribution Jtot(r) = J(r) + Jind(r), generates, via the linear total source-to-field
mapping dictated by the free-space outgoing wave dyadic Green function [58, 59],
the respective total electric field E(r). This point of view is relevant to the formula-
tion of an inverse source problem in free space, consisting of deducing an unknown
total source Jtot(r), of known support, say, the entire volume V , that is consistent
with a measured exterior electric field E(r). It is this formulation in free space that
has been the subject of the vast majority of investigations on the inverse source
problem (cf. [51, 52, 14] and the references therein.) Its solutions are electric cur-
rent distributions that are equivalent to the true total antenna currents in that they
generate the same field outside V .
However, for antennas embedded in material substrates or if a number of can-
didate background media are known a priori as part of the antenna analysis and
design, then such a characterization in terms of total currents in free space is very
inadequate. In particular, the true currents are of a certain form dictated by the
wave propagation in the relevant substrate medium that is not taken into account in
the inversion of the total source in free space. In so doing, the thus-found “equiva-
lent sources in free space” can even be non-physically realizable with the particular
driving and material components of the antenna. In other words, source inversion
in free space is unsuited in the presence of antenna-embedding media. It may even
give false leads when addressing the fundamental physical limits of the far-fields, as
determined, e.g., from the Picard conditions [60].
Consequently, a better approach is to formulate an inverse source problem in
substrate media, whose objective is to deduce an unknown primary current density
J(r) that is contained, along with the substrate, in the spherical volume V , and that
generates a prescribed exterior field. After all, from a practical point of view, it is
only this primary source J(r) that one has control of at antenna terminals. Thus in
6
this approach the background medium is treated not as an equivalent source which
could even be non-realizable with the allowable driving excitations and antenna
material, but instead as a propagating medium that is integral to the antenna and
within which the impressed source radiates.
The sought-after source is generally non-unique. This is due to the presence,
within the source region, of nonradiating sources [61, 52, 62, 63] whose generated
fields vanish identically outside the source region. But as constraints are imposed,
one manages to arrive at a unique source generating a given exterior field, such as
the familiar minimum energy solution to the inverse source problem [51, 52, 14].
The most commonly adopted constraint is that of minimizing the square of the L2
norm of the source (as defined in Eq.(2.9)), usually termed “the source energy” in
the inverse problems literature. It is at the heart of the Picard conditions defining
the range of the source-to-field linear mapping from L2 sources to L2 far fields. It
has also been used, recently, in addressing the realizability of electromagnetic pulsed
beams or wavelet fields launchable from finite-size sources [64, 65]. The solution to
the inverse source problem that minimizes the source energy is usually termed “the
minimum (source) energy solution”. It is related to the real image field generated
by a point-reference hologram of the field recorded on a closed surface completely
surrounding the source volume [57, 66, 67]. The ability of an antenna to radiate
a prescribed power with reduced current levels as characterized by this norm is
an indication of efficiency which has been adopted as constraint in the antenna
synthesis problem [68, 69, 70, 71, 72]. Reduction of the source energy also accounts
for reduction of ohmic losses in driving metallic elements, thus this constraint is of
both mathematical convenience as well as physical importance.
A minimized source energy would indicate that the resources of the antenna
generating the given field pattern are optimally used within the prescribed volume
of the source. Furthermore, comparison of the required minimum source energies
for different substrate configurations enables quantification of the enhancement due
to such structures. Upon solving the inverse source problem in these media for the
prescribed exterior fields, one can proceed to tackle the comparison of the required
resources, embodied in the source functional energy in the present case, that are
needed for the launching of the given fields. Substrate configurations for which the
required source energy is lower are then more optimal than alternative configura-
7
tions which require higher source energy for the launching of the same fields. As
mentioned earlier this mathematical framework to characterize substrate enhance-
ment is non-device-specific. In particular, one is then comparing the “best” source,
which minimizes the required source energy for the launching of the given field via a
given substrate, versus the “best” source, which minimizes the source energy for the
launching of the same field but at a different substrate (including the “no-substrate”
or free space-case).
Also, of importance are constraints related to the reactive near fields. High
reactive energy as characterized by the quality factor Q [73, 74, 75, 76, 77] reduces
system bandwidth, and thus it is important to require that the designed antenna
have tolerable reactive near fields. (Some treatments using this constraint can be
found in [78, 79, 80].) Another consideration of practical importance is the tuning
of the antenna to resonance so that its reactive power is zero. In this dissertation
we consider the source energy constraint with and without this tuning to resonance
by means of a generalization to sources in substrates of a constrained optimization
approach to the electromagnetic inverse problem introduced for free space in [14].
1.3 The Casimir Effect and Electromagnetic Metama-terials
The second part of this dissertation (Chapter 5) explores the effects that the pres-
ence of electromagnetic metamaterials would have on the Casimir forces. Quantum
field theory predicts the existence of fluctuations in the ground, or vacuum, state of
all quantum fields [81]. These fluctuations yield an infinite energy for the vacuum
state which ultimately leads to disastrous consequences for the theory [82]. Yet, the
existence of these so-called zero-point fluctuations is necessary for the consistency
of quantum field theory [83]. In spite of the inherently infinite nature of the ener-
gies attributed by quantum field theory to the vacuum, the presence of boundary
conditions alters these energies in a way that makes it possible to extract finite
quantities from the otherwise nonsensical infinities. In 1948, Casimir did just that:
he calculated the change in the electromagnetic vacuum energy due to the pres-
ence of two perfectly conducting parallel plates [84, 85]. He found out that the two
8
plates would attract each other by a force per unit area that is proportional to d−4,where d is the distance between them. This force has now been measured for this
geometry [86] as well as for other geometries [87, 88, 89, 90, 91] with an excellent
agreement between theory and experiment. Nevertheless, the interpretation of the
Casimir effect as proof for the reality of the vacuum quantum fluctuations is still
a debatable issue since Casimir’s result could also be obtained without recurring
to zero-point fluctuations [92, 83, 93]. The debate also extends to the question of
what constitutes a correct theoretical formulation for a given physical problem as
well (see, for instance, [94, 95]).
This being said, the technological impact of the Casimir effect is certainly real.
Besides the experimental vindications mentioned earlier, the Casimir forces, along
with their companion van de Waals forces, are now identified as the primary cause for
the collapse of microelectromechanical systems (MEMS) and nanoelectromechanical
systems (NEMS) [96, 97]. Hence, it is very important for the future development of
the emerging field of nanotechnology to investigate the possibility of shielding these
potentially destructive forces. The nature of the phenomenon allows for several
approaches. The Casimir forces depend strongly on the geometry of the system
and they are known to change sign and become repulsive for certain geometries
[85], although the possibility of geometry-based repulsive forces has been excluded
for a large class of geometries. The Casimir forces also depend on the constitutive
parameters of the media involved. In the context of the Dzyaloshinskii-Lifshitz-
Pitaevskii theory [98, 99] it has been shown that repulsive Casimir-Lifshitz forces
are achievable when the electric permittivity of a nonmagnetic slab sandwiched
between two nonmagnetic half-spaces is itself sandwiched between the permittivities
of the two half-spaces [100], i.e., ε1 (iξ) < εslab (iξ) < ε2 (iξ) , where ξ ≡ −iω is theimaginary frequency.
Another situation in which repulsive Casimir forces may arise involves meta-
materials [101, 102]. Metamaterials are electromagnetically engineered materials
whose constitutive parameters could in principle have any real value [103, 12]. In
particular, it has been suggested [102] that a slab of metamaterial with an index of
refraction n = −1 (also known as “perfect lens”) sandwiched between two perfectlyconducting plates should cause the Casimir force between the two plates to reverse
its sign.
9
Intrigued by the possibility of using metamaterials to produce repulsive Casimir
forces independently of the topology and effectively controlling the magnitude of
these forces we investigate in this dissertation the Casimir effect in the presence of
metamaterials. In particular we focus our attention on the case of two metamaterial
slabs having oppositely signed constitutive parameters sandwiched between two per-
fectly conducting plates. Several authors have considered the multilayered planar
geometry with and without metamaterials [104, 105, 106, 107, 94, 101, 108, 109, 102,
110, 111]. However, to our knowledge, no study has ever investigated the proposed
system. The relevance of this system stems from the realization that the most dra-
matic improvements attributed to metamaterials are often generated in situations
where two or more oppositely signed constitutive parameters are involved [50]. For
instance, the first slab would be made up of a double-positive, or DPS, medium,
i.e., a medium for which both the permittivity and the permeability are positive,
and the second slab would be made up of a double-negative, or DNG, medium, i.e.,
a medium for which both the permittivity and the permeability are negative.
In choosing this system we were particularly inspired by the subwavelength cavity
resonators described in [7]. The idea is to examine the possibility that for such
a cavity the peculiar properties of metamaterials would permit the existence of
electromagnetic vacuum fluctuation modes with wavelengths much larger than the
transverse dimension of the cavity. This would effectively cause an increase in the
outward pressure produced by these modes within the cavity, thus, reversing the
sign of the Casimir force between the two perfectly conducting plates and turning
it into a repulsive force.
The remainder of this dissertation is organized as follows. Chapter 2 develops
the essential theoretical tools needed for the quantitative study of piecewise-constant
radially symmetric backgrounds with particular focus on the lossless homogeneous
sphere. Chapter 3 generalizes the mathematical formulation of Chapter 2 to the
lossy core-shell system. Chapter 4 presents a comparative numerical study whose
aim is to understand the effect of reference antennas on enhancement level estimates.
Chapter 5 explores the effects that the presence of electromagnetic metamaterials
would have on the Casimir forces. Chapters 2, 3, and 5 are, each, supplemented by a
computer simulation study that illustrates the relevant theoretical results. Chapter
6 outlines some future research directions of our interest. Appendix A is a glossary
10
of important results related to spherical multipole theory. Appendices B, C, D, F, G
provide technical details needed in Chapters 2—5. Appendix E presents an important
result that shows how the current study of antenna radiation for antennas embedded
in piecewise-constant backgrounds may be extended to the more general case of a
radially symmetric background. All equations are in the SI system.
11
Chapter 2
Inverse Source Problem inNon-homogeneous BackgroundMedia
In this chapter we solve analytically and illustrate numerically the full-vector, elec-
tromagnetic inverse source problem of synthesizing an unknown source embedded
in a given substrate medium of volume V and radiating a prescribed exterior field.
The derived formulation and results generalize previous work on the scalar version
of the problem, especially [17]. Emphasis is put on substrates having constant con-
stitutive properties within the source volume V , which, for formal tractability, is
taken to be of spherical shape. The adopted approach is one of constrained opti-
mization which also relies on spherical wavefunction theory. The derived theory and
associated implications for antenna substrates are illustrated numerically.
We pay particular attention to lossless piecewise-constant radially symmetric
backgrounds having electric permittivity s and permeability µs, in particular, the
total permittivity distribution is of the form
(r) = sΘ(a− r) + 0Θ(r − a) (2.1)
where Θ denotes Heaviside’s unit step function (Θ(x) = 1, for x > 1, otherwise
12
Figure 2.1: Schematic of a general antenna whose driving points and material struc-ture are confined within a spherical volume V of radius a.
Θ(x) = 0), and the total permeability distribution is of the form
µ(r) = µsΘ(a− r) + µ0Θ(r − a). (2.2)
Generalization of the analysis for distributions of the constitutive parameters which
are spherically symmetric within the source support V and take the free-space values
outside V is also outlined1. Our results, published in [112], reveal the performance
improvements due to antenna-embedding substrates from an inverse antenna theory
point of view which is different than and complementary to efforts by other groups
in this fruitful area (e.g., [8, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 37, 38]; an expanded
bibliography can be found in [50]).
Referring to the schematic given in Fig. 2.1, we consider a general antenna which
is composed of a driving structure and a material structure. The driving structure
could be constituted by information-carrying currents, voltages (e.g., as in a dipole
antenna) and/or equivalent primary fields (e.g., as in a driving horn). The material
structure, on the other hand, could be a reflecting body (e.g., a parabolic reflector),
1See Appendix E.
13
an antenna substrate (e.g., as in ceramic-embedded antennas for mobile telephone
handsets [23]), etc. The antenna radiates at angular oscillation frequency ω. The
smallest spherical volume V within which the entire antenna resides is assumed to
have radius a, that is, V ≡ ©r ∈ R3 : r ≡ |r| ≤ aª.
2.1 The Forward Problem
2.1.1 Electromagnetic Generalities
Our starting point is provided by the frequency-domain Maxwell equations for a
generally lossless, non-homogeneous medium, in particular [58, 59],
∇×E(r) = iωµ(r)H(r) (2.3)
∇×H(r) = J(r)− iω (r)E(r), (2.4)
where J(r) represents an impressed current density (i.e., the source) confined within
the spherical volume V , and E(r) and H(r) are, respectively, the electric and mag-
netic fields it generates. (These fields are subject to Sommerfeld’s radiation condi-
tion [113].) Substituting H(r), from Eq. (2.3), into Eq. (2.4) yields the vector wave
equation
∇×µ∇×E(r)
µ(r)
¶− ω2 (r)E(r) = iωJ(r). (2.5)
The partial differential operator in Eq. (2.5) admits an outgoing-wave dyadic
Green’s function G(r, r0) which, along with Sommerfeld’s radiation condition, obeys
∇×µ∇× G(r, r0)
µ(r)
¶− ω2 (r)G(r, r0) = iωδ(r− r0)I, (2.6)
where I denotes the identity dyadic and δ the Dirac delta.
For future convenience we define the weighted inner product
(f , f 0) =Z
drM(r)f∗(r) · f 0(r), (2.7)
where f and f 0 are any two functions of position and the asterisk ∗ denotes thecomplex conjugate; M(r) is a characteristic (indicator or masking) function defined
14
as
M(r) =
(1 ; r ∈ V
0 ; r /∈ V.(2.8)
Using this inner product, we express the source energy E as
E ≡ (J,J), (2.9)
and the complex interaction power P (cf. [58]) as
P = −12(J, eGJ) = −1
2(J ,E) , (2.10)
where we have introduced the linear mapping eG defined by
[ eGJ](r) ≡ Z dr0G(r, r0) · J(r0). (2.11)
The real part of P, i.e., Re [P], represents the radiated power. This radiatedpower is determined by the radiated field or, equivalently, by the multipole moments
a(j)l,m, in the form of an incoherent sum of the multipole contributions [114, 115, 116],
namely
Re [P] = 1
2η0
2Xj=1
∞Xl=1
lXm=−l
l(l + 1)|a(j)l,m|2, (2.12)
where η0 =pµ0/ 0 is the free-space wave impedance.
On the other hand, the imaginary part of P, i.e., Im [P], corresponds to theenergy-storage reactive power [58]. It can have a prescribed value, say, zero (as was
shown in [14, 52]), which corresponds to a tuned antenna, and is one of the solution
constraints to be employed in the formulation to follow. We note that the reactive
power can be expressed as
Im [P] = −12
ZVdrJ∗(r) ·
ZVdr0GS(r, r
0) · J(r0)
≡ −12(J, eGSJ), (2.13)
15
where
GS(r, r0) ≡ Im £G(r, r0)¤ = 1
2i
£G(r, r0)− G∗(r, r0)¤ , (2.14)
and where we have introduced the linear mapping eGS defined by Eq. (2.11) after
the substitutions eG→ eGS and G→ GS .
2.1.2 Source-to-Multipole-Moment Mapping
To formulate the inverse problem for the cases described in Eqs. (2.1) and (2.2) (as
well as cases described by Eqs. (3.1,3.2) and Eqs. (E.1,E.2)) it is necessary to first
have at our disposal the solution of the associated forward or radiation problem. To
accomplish this, we note that, for these cases, the electric field E(r) generated by the
most general source of support V can be represented, outside V , by the multipole
expansion [114]
E(r) =2X
j=1
∞Xl=1
lXm=−l
a(j)l,mΛ
(j)l,m(r), r /∈ V, (2.15)
where the complex-valued expansion coefficients a(j)l,m are the multipole moments of
the field, and where the multipole fields are
Λ(j)l,m(r) =
⎧⎪⎨⎪⎩∇× [h(+)l (k0r)Yl,m(r)] ; j = 1
ik0h(+)l (k0r)Yl,m(r) ; j = 2,
(2.16)
where r ≡ r/r , h(+)l denotes the spherical Hankel function of the first kind and
order l (as defined in [117]), corresponding to outgoing spherical waves in the far
zone, Yl,m is the vector spherical harmonic of degree l and order m (as defined in
[114], Eqs. (4.7) and (4.8))2, and j = 1 and j = 2 correspond to electric (TMr) and
magnetic (TEr) multipole fields, respectively. For the convenience of the reader we
are outlining the key properties of the vector spherical harmonics in Appendix A.
Note that in (2.15) the summation over l starts from l = 1 because there are no
vector spherical harmonics of zero degree [114]. On the other hand, the magnetic
2Physically, the index l characterizes the so-called multipolarity or modal order of the field;thus l = 1 corresponds to 21-pole (dipole) radiation, l = 2 corresponds to 22-pole (quadrupole)radiation, l = 3 corresponds to 23-pole (octupole) radiation, and so on.
16
field outside the source region is defined by substituting from these results into Eq.
(2.3) with the free-space substitution µ(r)→ µ0.
The electric and magnetic multipole moments, a(1)l,m and a(2)l,m, respectively, are
related to the current distribution J by
a(j)l,m = (B
(j)l,m,J), j = 1, 2, (2.17)
i.e., they are the projections of the current distribution J onto the set of source-free
vector fields B(j)l,m which need to be determined for the particular antenna back-
ground medium. For the special free-space case where µ(r)/µ0 = 1 = (r)/ 0 the
latter fields are the familiar source-free multipole fields, in particular (cf. [74] and
[59]),
B(j)l,m(r) ≡
⎧⎪⎪⎨⎪⎪⎩− η0
l(l+1)∇× [jl(k0r)Yl,m(r)] ; j = 1
−i k0η0l(l+1)jl(k0r)Yl,m(r) ; j = 2,
(2.18)
where jl is the spherical Bessel function of the first kind and order l (as defined in
[117], for instance). On the other hand, it is shown in Appendix B that for piecewise-
constant radially symmetric backgrounds whose permittivity and permeability are
given by Eqs. (2.1) and (2.2)
B(j)l,m(r) ≡
⎧⎪⎪⎨⎪⎪⎩−η0l(l+1)F
∗(1)l ∇× [jl(k∗r)Yl,m(r)] ; j = 1
−ik0η0l(l+1) F
∗(2)l jl(k
∗r)Yl,m(r) ; j = 2,
(2.19)
where the substrate wavenumber k = ω√µs s, the relative permittivity r ≡ s/ 0,
the relative permeability µr ≡ µs/µ0, and where we have defined the complex am-
plitudes
F(j)l ≡
⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩i/(k0ka2)
( r/µr)1/2jl(ka)Vl(k0a)−h(+)l (k0a)Ul(ka)
; j = 1
iµr/(k0ka2)(µr/ r)
1/2jl(ka)Vl(k0a)−h(+)l (k0a)Ul(ka); j = 2,
(2.20)
where
Ul (λa) ≡ Ul (λr)|r=a ≡∙djl(λr)
d (λr)+
jl(λr)
λr
¸¯r=a
, (2.21)
17
and
Vl (λa) ≡ Vl (λr)|r=a ≡"dh(+)l (λr)
d (λr)+
h(+)l (λr)
λr
#¯¯r=a
. (2.22)
Note that F (1)l and F(2)l are functions of k0a, ka, r, and µr.
Because of the self-imposed restriction to the study of lossless substrates in
this chapter, the relative constitutive parameters µr and r admit only real values.
Consequently, the wavenumber k can assume only real values (positive for DPS
materials and negative for DNG metamaterials) or purely imaginary values (for
single-negative metamaterials). When k is purely imaginary, i.e., k = iα, α ∈ R,the arguments of the spherical Bessel functions involving k in Eqs. (2.20)-(2.22)
are, accordingly, purely imaginary. In this case one notes that the regular spherical
Bessel functions jl and h(+)l are replaced, respectively, with the modified spherical
Bessel functions il and kl such that (see, for instance, [117])
jl (ka) ≡ ilil(αa), (2.23)
and
h(+)l (ka) ≡ −i−lkl(αa). (2.24)
(There shall be no confusion between the modified spherical Bessel functions il and
kl and the imaginary unit i and the wavenumber k since these latter ones do not
carry a subscript.)
We draw the attention of reader to the fact that F (j)l , j = 1, 2, represent the
Mie amplitudes due to the scattering of a plane electromagnetic wave off a sphere
of radius a and wavenumber k embedded in an infinite homogeneous medium of
wavenumber k0; F(1)l being the amplitudes of the electric oscillations and F (2)l those
of the magnetic oscillations. This should not come as a surprise in view of the
physics as well as the formulation itself of the problem.3
It is not hard to show that the amplitudes F (j)l −→r,µr→1
1, as expected, since, in
that case, Eqs. (2.19) and (2.20) reduce to the free-space result (2.18). The corre-
sponding results for more general spherically-symmetric backgrounds are outlined
in Appendix E. Substitution of the associated results into Eq. (2.17) completes
3See Section 2.3 and Appendix B.
18
the description of the forward problem. Armed with these developments, we are in
position to formulate next the corresponding inverse source problem.
2.2 Inverse Source Theory Based on Constrained Opti-mization
The inverse source problem of deducing the source J(r), confined within V from
knowledge of the exterior field E(r) is seen from Eqs. (2.15) and (B.4)4 to be equiv-
alent to that of determining the source from knowledge of the multipole moments,
i.e., to that of inverting Eq. (2.17). The respective inversion is addressed next via
a generalization of the free-space optimization theory in [14] to non-homogeneous
backgrounds. Emphasis is given to the particular case of piecewise-constant radially
symmetric backgrounds, but the derived expressions apply to more general cases as
long as one uses the appropriate projective wavefunctions B(j)l,m which vary from a
medium type to another.5
2.2.1 Minimum Energy Solution by Constrained Optimization
We start by addressing the problem of determining the minimum energy source JME
embedded in a substrate of volume V with fixed constitutive parameters r, µr and
generating a given exterior field. The problem can be cast as
minJ∈S
E (J) , (2.25)
where
S ≡nJ ∈ L2
¡V ;C3
¢: a(j)l,m − (B(j)
l,m ,J) = 0o. (2.26)
Note that the constraint set S is convex; also, the objective functional E is coerciveand strictly convex. The convexity of E along with its continuity at some pointguarantee its continuity on the whole space L2
¡V ;C3
¢[118].
If a minimizer JME exists, then its uniqueness and global minimality are insured
4This can also be derived form Eqs.(B.16)-(B.19)). See Appendix B for more detail.5An important result related to the spherically symmetric substrate case is derived in Appendix
E.
19
by the strict convexity of E and the convexity of S [119]. But what guarantees theexistence of at least one such minimizer? We address this question next. First, we
note that since E is a continuous and convex functional on a Hilbert space then itis also weakly sequentially lower semi-continuous [119]. Given this property of E ,the fact that it is coercive, and the fact that S is a closed and convex subset of a
reflexive Banach space (e.g., a Hilbert space), we can, now, assert that there exists
at least one point JME that minimizes E over S [120]. Hence, problem (2.25,2.26)
admits only one global solution.
Due to the Fréchet differentiability of the objective functional E and the con-straints
ha(j)l,m − (B(j)
l,m, ·)i, the continuity of their Fréchet derivatives6, and the fact
that ∇Jha(j)l,m − (B(j)
l,m ,J)i¯J=JME
maps L2¡V ;C3
¢onto C, there exist [120] La-
grange multipliers c(j)l,m ∈ C such that the generalized Lagrangian
L³J, c
(j)l,m
´≡ E + 2Re
⎡⎣ 2Xj=1
∞Xl=1
lXm=−l
c(j)l,m
³a(j)l,m − (B(j)
l,m , J)´⎤⎦ (2.27)
is stationary at JME.
To compute the solution we require that
δL = 2Re⎡⎣⎛⎝δJ , J−
2Xj=1
∞Xl=1
lXm=−l
c(j)∗l,mB
(j)l,m
⎞⎠⎤⎦ = 0. (2.28)
From the Dubois-Raymond lemma, Eq. (2.28), and the forward mapping relations
(2.17,2.19,2.20,2.21,2.22), one finds that the minimum-energy source is given by
JME(r) =2X
j=1
∞Xl=1
lXm=−l
a(j)l,mh
σ(j)l
i2B(j)l,m(r), (2.29)
where we have introduced the positive-definite “singular values”
[σ(j)l ]
2 ≡ (B(j)l,m,B
(j)l,m), j = 1, 2, (2.30)
6See Appendix D.
20
specifically, hσ(j)l
i2= |F (j)l |2
hκ(j)l
i2, j = 1, 2, (2.31)
where hκ(j)l
i2 ≡⎧⎪⎪⎨⎪⎪⎩
η20R a0 dr
h|jl(kr)|2 + |kr|2
l(l+1) |Ul(kr)|2i
; j = 1
η20k02
l(l+1)
R a0 dr r2 |jl(kr)|2 ; j = 2.
(2.32)
For real k2 the integral associated with the j = 2 case is calculable through the use of
the second Lommel integral (see , for instance, [117]) and the recurrence relations of
the Bessel functions along lines similar to those employed in [51] to evaluate similar
inner products. Afterwards, the recurrence relations are also used to express the
integral associated with the j = 1 case in terms of the calculated integral associated
with the j = 2 case. Consequently, Eqs. (2.32) reduce to,
hκ(j)l
i2=
⎧⎪⎪⎨⎪⎪⎩η20a|ka|2
l(l+1)(2l+1)
£(l + 1)γ2l−1(ka) + lγ2l+1(ka)
¤; j = 1
η20a(k0a)2
l(l+1) γ2l (ka) ; j = 2,
(2.33)
where we have introduced the unitless quantity (cf. [51], Eq. (17))
γ2l (ka) ≡1
a3
Z a
0dr r2j2l (kr)
=1
2
£j2l (ka)− jl−1(ka)jl+1(ka)
¤. (2.34)
For k = iα, α ∈ R as is the case for single-negative metamaterials, one uses definition(2.23) to express Eqs. (2.33) and (2.34) in terms of il.7
Furthermore, the minimum source energy
EME ≡ (JME,JME) =2X
j=1
∞Xl=1
lXm=−l
|a(j)l,m|2
[σ(j)l ]
2. (2.35)
7Note that the lone appearance of the size parameter a in (2.33), i.e., its appearance decoupledfrom the wavenumbers, is a direct consequence of the fact that the multipole moments a(j)l,m aredimensionful quantities. It is, as well, a reminder of the boundedness of the enclosing volume V ,i.e., of the embedding sphere of substrate material.
21
As expected, these developments reduce, for r = 1 = µr, to the free-space result
(Eqs. (13), and (14) in [14]) since F (j)l (k0a, ka, r, µr) = 1; that is, the free-space
minimum-energy solution is given by Eq. (2.29) with B(j)l,m given by Eq. (2.18) and
[σ(j)l ]
2 substituted byhκ(j)l (k0a = ka)
i2.
2.2.2 Minimum Energy Source Having Zero Reactive Power
We consider next the constrained optimization problem of minimizing the functional
energy of the source subject to the additional constraint that the reactive power of
the source has a prescribed value. The results for this problem will be elaborated
next for the particular and important case of zero reactive power, i.e., Im [P] = 0.This corresponds to the minimizing of the antenna currents (the physical resources)
while simultaneously enforcing perfect antenna reactance tuning inside the antenna.
The problem can be cast as
minJ∈X
E (J) , (2.36)
where
X ≡nJ ∈ L2
¡V ;C3
¢: a(j)l,m − (B(j)
l,m ,J) = 0, (J, eGSJ) = 0o. (2.37)
The constraint set X is closed, unbounded, and nonconvex. Its nonconvexity stems
from that of the newly introduced constraint (J, eGSJ) = 0. The set X is assumed to
be nonempty. (If it turns out to be empty this would mean that it is not possible for
an antenna having a substrate medium of constitutive parameters r, µr to produce
the prescribed external field and at the same time have a vanishing reactive power.)
It is clear that problem (2.36,2.37) is an inherently difficult nonconvex program-
ming problem. Not only do we seek to minimize an objective functional under
nonconvex functional constraints but, also, we have to do that on an unbounded
set. Proving, for instance, the existence of a solution to problem (2.36,2.37) would
have been easier if X were convex, but it is straightforward to show that the only
way for X to become convex is to have Reh(J1, eGSJ2)
i≤ 0, ∀J1,J2 ∈ L2
¡V ;C3
¢.
This would amount to imposing a new constraint which appears not to correspond
to anything meaningful, physically speaking.
Now let us address the issue of the existence of a solution to problem (2.36,2.37)
22
in the absence of the convexity and boundedness of the constraint set X. Since X is
a closed subset of a normed vector space and since E is a coercive functional, thenthere exist [120] J0 ∈ X and Γ > 0 such that
infJ∈X
E (J) = infnE (J) : J ∈ X ∩BΓ (J0)
o, (2.38)
where BΓ (J0) is the closed (and bounded) ball of radius Γ and center J0. This is
a powerful result. What this tells us is that minimizing E over the unbounded setX can be reduced to the minimizing of E over a bounded subset in X that could
be much smaller than X. All that remains to complete the proof of existence of a
solution to problem (2.36,2.37) is to demonstrate the existence of a solution to the
auxiliary problem
minJ∈X∩BΓ(J0)
E (J) . (2.39)
A useful variant of the generalized Weierstrass theorem stipulates that for a weakly
sequentially lower semi-continuous functional defined on a weakly sequentially com-
pact subset of a Hilbert space there exists, at least, one solution to the minimization
problem [119]. But we have already shown that E is a weakly sequentially lower semi-continuous functional (see the discussion of problem (2.25,2.26)). Consequently, the
existence of a solution to problem (2.39,2.37) depends entirely on the demonstra-
tion that X ∩BΓ (J0) is a weakly sequentially compact subset. But this, too, is true
because any bounded subset of a reflexive Banach space (e.g., a Hilbert space) is
also weakly sequentially compact [121]. Hence, assuming that X ∩ BΓ (J0) is non-
empty, we are, from the preceding discussion, in position to affirm the existence of at
least one global minimizer JE,P ∈ X ∩BΓ (J0) for the auxiliary problem (2.39,2.37).
However, by virtue of (2.38), this point JE,P is also the sought solution of problem(2.36,2.37), which completes our proof.
Unfortunately, though, we have yet to guarantee the uniqueness of this solution
or even write down a minimality condition that would yield this solution. We shall
now focus on writing down a necessary minimality condition whose solution would
yield the minimizer JE,P .
Let JE,P be a minimizer. In view of the Fréchet differentiability of the objective
23
functional and the constraints, the continuity of their Fréchet derivatives,8 and the
fact that ∇Jha(j)l,m − (B(j)
l,m ,J)i¯J=JE,P
is surjective and the range of ∇J³J, eGSJ
´¯J=JE,P
is closed, there exist Lagrange multipliers χ ∈ R, and c(j)l,m ∈ C such that [120]
Reh³∇JL
³JE,P , χ, c
(j)l,m
´, J− JE,P
´i≥ 0, ∀J ∈ L2
¡V ;C3
¢, (2.40)
where the generalized Lagrangian functional is given by
L³J, χ, c
(j)l,m
´≡ E (J) + χ
³J, eGSJ
´+ 2Re
⎧⎨⎩2X
j=1
∞Xl=1
lXm=−l
c(j)l,m
ha(j)l,m − (B(j)
l,m,J)i⎫⎬⎭ . (2.41)
Condition (2.40,2.41) reduces to
Re
⎡⎣⎛⎝JE,P + χ eGSJE,P −2X
j=1
∞Xl=1
lXm=−l
c(j)l,mB
(j)l,m , J− JE,P
⎞⎠⎤⎦ ≥ 0, ∀J ∈ L2¡V ;C3
¢.
(2.42)
According to Eq. (2.42), to determine JE,P one needs to solve an infinite numberof equations with infinite number of unknowns. That, of course, is not the case in
practical situations. For any real problem the radiation emitted by the source has a
maximum multipolarity lmax ∼ ka (<∞). Thus, for real problems one would needto solve 2lmax (lmax + 2) + 4 integral equations with 2lmax (lmax + 2) + 4 unknowns.
By all standards this is a tedious task, even for small values of lmax. One should
try to find a more clever way of determining what the solution is. For instance, one
could resort to numerical techniques and algorithms available in the literature (see,
e.g., [122] and the references therein). In the sequel we plan on adopting a similar
approach that combines analytical and numerical methods.
We shall assume that X 6= ∅ and adopt partly-analytical partly-numerical
strategies to find a minimizer JE,P , which we proved that it existed, without havingto solve a large number of complicated equations. The “hybrid” approach below is
very much in line with the spirit of those adopted for this kind of problems. We
8See Appendix D.
24
shall also explore some of the properties of the solution. Once a feasible point JE,Pis found by means of the technique below one would substitute it in the derived
minimality conditions to check if it satisfies these conditions.
Let, now, L be the generalized Lagrangian defined as
L³J, χ, c
(j)l,m
´≡ E (J) + χ
n(J, eGSJ) + 2 Im[P]
o+ 2Re
⎧⎨⎩2X
j=1
∞Xl=1
lXm=−l
c(j)l,m
ha(j)l,m − (B(j)
l,m,J)i⎫⎬⎭ , (2.43)
wherein the constraint on the reactive power is now written in such a way that it
permits the latter to have an arbitrary value Im[P ] that is not necessarily zero.
The first variation of the last term in Eq. (2.43) is found from Eqs. (2.13) and
(2.14) to be
χδ(J, eGSJ) = 2Rehχ(δJ, eGSJ)
i. (2.44)
It follows from Eqs. (2.27), (2.28), and (2.44) that the first variation of the La-
grangian in Eq. (2.43) is
δL = 2Re⎡⎣(δJ,J) + χ(δJ, eGSJ)−
2Xj=1
∞Xl=1
lXm=−l
c(j)∗l,m (δJ,B
(j)l,m)
⎤⎦ . (2.45)
By equating the variation in Eq. (2.45) to zero one deduces that the sought
solution, to be denoted as JE,P(r), must obey, within its support V , the relation
JE,P(r) + χ eGSJE,P(r) =2X
j=1
∞Xl=1
lXm=−l
c(j)∗l,mB
(j)l,m(r). (2.46)
If χ = 0 then this approach coincides with the one given earlier, leading to the
minimum energy source in Eq. (2.29) (in such a situation, that source generates
zero reactive power) while for the more general case χ 6= 0 the two formulations
(and their solutions) differ. However, we note that for certain peculiar constitutive-
parameter values the constraint is not active and therefore χ = 0. In that peculiar
case the minimum energy sources are intrinsically resonant.
By letting the vector wave equation operator (∇×∇×− (k∗)2) = (∇×∇×−k2)
25
(the equality stems from the requirement that the substrate be lossless) act on both
sides of Eq. (2.46) and with the aid of the fact that the fields B(j)l,m are solutions
of the homogeneous wave equation associated to the same operator, one concludes
that the source JE,P(r) obeys the homogeneous wave equation
∇×∇× JE,P(r)−K2JE,P(r) = 0, (2.47)
in the interior of the source region V ; the quantity K which appears in Eq. (2.47)
is a modified wavenumber defined by
K2 ≡ k2 − χµsω. (2.48)
(Note that K quickly becomes purely imaginary as χ becomes large and positive.)
Now, the most general source that is confined within the spherical source volume
V and is a solution of Eq. (2.47) in the interior of V must admit the representation
JE,P(r) =2X
j=1
∞Xl=1
lXm=−l
v(j)l,mD
(j)l,m(r), (2.49)
where v(j)l,m are expansion coefficients that need to be determined (for the constraints
of the problem) and where
D(j)l,m(r) =
⎧⎪⎪⎨⎪⎪⎩− η0
l(l+1)∇× [jl(Kr)Yl,m(r)] ; j = 1
− iη0Kl(l+1)jl(Kr)Yl,m(r) ; j = 2.
(2.50)
From the formal similarity of B(j)l,m and D
(j)l,m (cf. Eq. (2.19)) it follows at once from
Eqs. (2.30), (2.31), (2.33), and (2.34) that the inner product³D(j)l,m,D
(j)l,m
´= p(j)
hκ(j)l (k0a,Ka)
i2, (2.51)
26
where
p(j) =
⎧⎪⎨⎪⎩1 ; j = 1
|K|2/k20 ; j = 2.
(2.52)
By substituting from Eqs. (2.49) and (2.50) into Eq. (2.17) while using well-
known orthogonality properties9 of the vector spherical harmonics Yl,m(r) and the
associated vector functions r×Yl,m(r) one obtains the form of the desired solution,
in particular,
JE,P(r) =2X
j=1
∞Xl=1
lXm=−l
a(j)l,m
(B(j)l,m,D
(j)l,m)
D(j)l,m(r), (2.53)
where
(B(j)l,m,D
(j)l,m) =
⎧⎪⎪⎨⎪⎪⎩η20F
(1)l
R a0 dr
hjl(kr)jl(Kr) + kKr2
l(l+1)Ul (kr)Ul (Kr)i
; j = 1
η20F(2)l
k0Kl(l+1)
R a0 drr2jl(kr)jl(Kr) ; j = 2.
.
(2.54)
Similarly to the integrals in (2.32), the integral associated with the j = 2 case
in (2.54) is calculable through the use of the first Lommel integral (cf., for instance,
[117].) The above inner product takes on the form
(B(j)l,m,D
(j)l,m) =
⎧⎪⎪⎪⎨⎪⎪⎪⎩η20kKa3F
(1)l
l(l+1)(2l+1)
£(l + 1)ψl−1 (ka,Ka) + lψl+1 (ka,Ka)
¤; j = 1
η20k0Ka3F(2)l
l(l+1) ψl (ka,Ka) ; j = 2,
(2.55)
where we have introduced the unitless quantity
ψl (ka,Ka) ≡ 1
a3
Z a
0drr2jl(kr)jl(Kr)
=1
a (k2 −K2)[Kjl(ka)jl−1(Ka)− kjl−1(ka)jl(Ka)] . (2.56)
(Note that (2.56) is valid only for k 6= K, i.e., for χ 6= 0. The case k = K, i.e., for
χ = 0, has already been discussed.)9See Appendix A.
27
The source energy corresponding to (2.53) is of the form10
EE,P ≡ (JE,P ,JE,P) =2X
j=1
∞Xl=1
lXm=−l
(D(j)l,m,D
(j)l,m)
|(B(j)l,m,D
(j)l,m)|2
|a(j)l,m|2. (2.57)
We need to incorporate the reactive power constraint, i.e., Eq. (2.13), which
defines the value of the remaining Lagrange multiplier χ. Since the desired reactive
power is specified to be zero, the problem now is to find an expression for the reactive
power in terms of χ from which one can deduce the value of χ which minimizes the
source energy under the constraint Im [P] = 0. This value of χ will be called χ0.
A number of partly analytical, partly numerical strategies can be implemented to
accomplish this step.
One such approach, which generalizes the development for the free-space case
in [14], consists of determining the field E(r) generated by the source JE,P(r) inthe interior of the source region V . In particular, after evaluating the field, one
can compute the interaction power via Eqs. (2.10) and (2.11) and require that its
imaginary part vanish. In particular, plotting Im [P] and EE,P versus χ one can
finally select the value of χ which yields minimum EE,P out of all values of χ forwhich Im [P] = 0. We adopt this approach next.
By rewriting Eq. (2.47) as
¡∇×∇×−k2¢ [JE,P(r)− iχE(r)] = 0, (2.58)
where we have borrowed from Eq. (2.5), one concludes that the field E(r) must
admit in the interior of the source region V an expansion of the form
E(r) =1
iχ
⎡⎣JE,P(r) + 2Xj=1
∞Xl=1
lXm=−l
u(j)l,mB
(j)l,m(r)
⎤⎦ , r ∈ V, (2.59)
where the expansion coefficients u(j)l,m need to be determined taking into account
continuity of the tangential components of the field on the boundary ∂V ≡ r ∈R3 : r = a of V . Continuing on this idea, it is not hard to show from these
10Note that Eqs.(2.53) and (2.57) do not assume any particular value for Im [P].
28
developments, and by straightforward generalization of the discussion of the free-
space version of the problem in [14], Eqs. (30)-(42), that the complex interaction
power can be expressed as
P =2X
j=1
∞Xl=1
lXm=−l
q(j)l |a(j)l,m|2, (2.60)
where
q(j)l =
i
2χ
⎡⎢⎣³D(j)l,m,D
(j)l,m
´¯³B(j)l,m,D
(j)l,m
´¯2 + u(j)l,m
a(j)l,m
⎤⎥⎦ , (2.61)
where the quantity u(j)l,m/a(j)l,m is given by
u(j)l,m
a(j)l,m
=
⎧⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎩
1
F∗(1)l k∗Ul(k∗a)
"−ik0η0χl(l + 1)Vl(k0a)−
KUl(Ka)³B(1)l,m,D
(1)l,m
´#
; j = 1
1
F∗(2)l k0jl(k∗a)
"−ik0η0χl(l + 1)h
(+)l (k0a)− Kjl(Ka)³
B(2)l,m,D
(2)l,m
´#
; j = 2,
(2.62)
where the radial functions Ul and Vl have already been defined in Eq. (2.21) and
Eq. (2.22), respectively.
Thus the reactive power of the source JE,P , is given by
Im [P] =2X
j=1
∞Xl=1
lXm=−l
g(j)l |a(j)l,m|2, (2.63)
where
g(j)l =
1
2χ
⎧⎪⎨⎪⎩³D(j)l,m,D
(j)l,m
´¯³B(j)l,m,D
(j)l,m
´¯2 +Re⎡⎣u(j)l,ma(j)l,m
⎤⎦⎫⎪⎬⎪⎭ . (2.64)
By taking the real part of the complex interaction power, as given by Eqs. (2.60)-
(2.62), one also recovers Eq. (2.12) which is the well-known expression for the
radiated power in terms of the multipole moments.
Eqs. (2.62)-(2.64) relate χ directly to Im [P], as desired. For a certain problem,
29
where a(j)l,m and Im [P] are given, one can compute the values of χ for which Im [P] = 0by using these expressions, and pick, out of those values, the one which minimizes
the functional energy in Eq. (2.57). By substituting that value of χ (i.e., χ0) into
Eqs. (2.48), (2.50), and (2.53) one arrives at the desired solution.
Let
Ξ ≡ χ ∈ R : Im [P (χ)] = 0 . (2.65)
It is found11, numerically, that the minimum source energy is achieved for the value
of χ that is closest to χ = 0, i.e.,
|χ0| = infχ∈Ξ
|χ| . (2.66)
It appears only natural to assume that an increase in the source energy from EME
should correspond to χ0 (and any other value of χ ∈ Ξ for that matter). Thiswould be understood, intuitively, as a cost that one would have to pay to realize
a tuned antenna. The numerical simulations suggest that this, in fact, is the case:
substituting any nonzero value χ ∈ Ξ in the expression for EE,P yields a value thatis larger than EE,P |χ=0 = EME. Does this mean that EME is a lower bound of EE,Pand that EE,P (χ0) is a global minimum? Before we examine this question we notethat the above observations remind us of Eq. (2.38). Indeed, a convenient way of
viewing these observations is to think of the origin of the sphere BΓ (J0) 3 JE,P asthe point J0 = JME and to think of its radius as Γ ≥ |χ0|.
Supposing that EE,P (χ0) corresponds to a feasible point, let us derive a conditionfor it to be a global minimum. By definition EE,P (χ0) is said to be a global minimumwhen
EE,P (χ) ≥ EE,P (χ0) , ∀χ ∈ Ξ. (2.67)
If the inequality is strict then the global minimum is also unique.
11See Section 2.3.
30
It follows from Eqs. (2.17), (2.13), and (2.46) that
EE,P − 2χ Im[P] =2X
j=1
∞Xl=1
lXm=−l
c(j)∗l,m (JE,P ,B
(j)l,m)
=2X
j=1
∞Xl=1
lXm=−l
c(j)∗l,m a
(j)∗l,m . (2.68)
Thus if we require Im [P (χ)] = 0, Eq. (2.68) yields
EE,P =2X
j=1
∞Xl=1
lXm=−l
c(j)∗l,m a
(j)∗l,m . (2.69)
Furthermore, by projecting both sides of Eq. (2.46) onto the functions B(j)l,m while
recalling Eqs. (2.17) and (2.30) one obtains
a(j)l,m + χ(B
(j)l,m,
eGSJE,P) = c(j)∗l,m [σ
(j)l ]
2. (2.70)
By solving for c(j)∗l,m then substituting into Eq. (2.69) one obtains
EE,P = EME + χ2X
j=1
∞Xl=1
lXm=−l
(B(j)l,m,
eGSJE,P)a(j)∗l,m
[σ(j)l ]
2, (2.71)
where χ ∈ Ξ. Upon substituting JE,P(r), from Eq. (2.53), into Eq. (2.71) and usingstandard orthogonality properties of the spherical harmonics, one obtains
EE,P = EME + χ2X
j=1
∞Xl=1
lXm=−l
(B(j)l,m,
eGSD(j)l,m (χ))
(B(j)l,m,D
(j)l,m (χ))
|a(j)l,m|2
[σ(j)l ]
2. (2.72)
Expression (2.72) for the source energy directly assumes that Im [P] = 0, while
expression (2.57) holds for any value of the reactive power Im [P].
It follows from (2.67) and (2.72) that the condition for EE,P (χ0) to be a global
31
minimum is given by
2Xj=1
∞Xl=1
lXm=−l
⎧⎨⎩χ(B
(j)l,m,
eGSD(j)l,m (χ))
(B(j)l,m,D
(j)l,m (χ))
− χ0(B
(j)l,m,
eGSD(j)l,m (χ0))
(B(j)l,m,D
(j)l,m (χ0))
⎫⎬⎭ |a(j)l,m|2
[σ(j)l ]
2≥ 0 (2.73)
for any value of χ ∈ Ξ. Condition (2.73) is a necessary and sufficient condition forEE,P (χ0) to be a global minimum. The way it should be used is as follows. Fora given substrate, solve Im [P (χ)] = 0 for χ (where Im [P (χ)] is given by (2.62)-(2.64)). If condition (2.73) is satisfied for all values of χ ∈ Ξ then EE,P (χ0) is aglobal minimum. If it is not satisfied for at least one value of χ ∈ Ξ then EE,P (χ0)is not a global minimum (but it may still be a local minimum).
Condition (2.73) was written down based on the presumption that EE,P (χ0 6= 0)was a global minimum. For EE,P (χ0 = 0) = EME, condition (2.73) reduces to
χ2X
j=1
∞Xl=1
lXm=−l
(B(j)l,m,
eGSD(j)l,m)
(B(j)l,m,D
(j)l,m)
|a(j)l,m|2
[σ(j)l ]
2≥ 0, ∀χ ∈ Ξ. (2.74)
The preceding developments have emphasized the case of piecewise-constant ra-
dially symmetric backgrounds. Generalization to more general spherically-symmetric
backgrounds follows the same overall approach but then the formulation is based
on the respective forward solver sketched in Appendix E. We shall not dwell on this
here, and instead address numerical illustration of the preceding theory next.
2.3 Computer Simulation Study
The previous theory and algorithms are applied next to elucidate the effect of the
antenna-embedding medium on radiation performance for two classes of antennas:
electrically small, and larger (resonant) antennas. Within each class, we consider
both the minimum energy solution without tuning constraint, and the minimum
energy solution subject to the additional zero reactive power or tuning constraint.
The goal is to gain an understanding of the effect of the antenna substrate on the
minimum source energy for a given radiation pattern. Other related considerations
are also discussed. To present the results with focus, more attention is given in the
32
following to DPS and DNG materials.12
2.3.1 Minimum Energy Sources
It follows from Eq. (2.35) that, generally, the larger the singular valueshσ(j)l
i2, the
smaller the minimum source energy EME required for the launching of a given radi-
ation pattern with a source of a given size. The singular valueshσ(j)l ( r = 1 = µr)
i2correspond to the source in free space, that is, without the substrate. Thus, the
larger the singular valueshσ(j)l
i2for a given substrate wavenumber ka relative to
the corresponding free-space values, the greater the associated enhancement, due to
the substrate, of radiation of the lth multipole order field with given resources. It
is thus important to understand the dependence of the singular valueshσ(j)l
i2on
k0a, ka, r, µr and l, for both the electric (j = 1) and the magnetic (j = 2) cases.
Large singular values, such as resonances or peaks in the plots of the singular values
versus these variables, will indicate enhanced radiation for such operational modes
or conditions, with the given resources. This aspect is investigated numerically next.
Before engaging in the numerical illustrations we make some remarks: (1) the
multipolarity l is handled in the plots as a continuous variable to facilitate under-
standing of the curves, yet the meaningful results correspond solely to the discrete
values of l; (2) in the simulations the size parameter (radius) a of the antenna in-
cluding the substrate has been set to unity, i.e., a = 1 meter; and (3) in the plots
and associated discussion we consider the normalized wavenumbers defined by
x ≡ ka/π (2.75)
and
x0 ≡ k0a/π. (2.76)
The normalized wavenumber x represents the wavenumber of the field in the mater-
ial, hence, the effective electric size in the material, while the normalized wavenum-
ber x0 measures the respective size in free space.
12Having r > 0 and µr > 0 does not necessarily imply that the material is a conventionalone. Indeed, naturally occurring materials do not usually exhibit positive permeabilities and per-mittivities smaller that those of the vacuum. This makes manufacturing a material whose relativepermittivity is equal to 0 < r < 1 still require the use of metamaterials.
33
Behavior of the Singular Valueshσ(j)l
i2Fig. 2.2 shows, for different antenna sizes, the free-space singular values
hσ(1)l ( r = 1 = µr)
i2and
hσ(2)l ( r = 1 = µr)
i2, respectively. No local maxima or resonances are seen for
the free-space cases, in particular, in those cases the singular value spectrum decays
exponentially. Fig. 2.3 shows, for an antenna whose size corresponds to that of a
quarter-wave antenna in free space (x0 = 1/4), plots of the normalized electric singu-
lar valuesh(1)l (x0, x, r, µr)
i2 ≡ hσ(1)l (x0 = 1/4, x, r = 1)i2/hσ(1)l (x0 = 1/4 = x, r = 1 = µr)
i2versus l, parameterized by the normalized wavenumber in the substrate, x. From
now on, the normalized singular valuesh(j)l (x0, x, r, µr)
i2defined as
h(j)l (x0, x, r, µr)
i2 ≡hσ(j)l (x0, x, r, µr)
i2hσ(j)l (free-space case)
i2=
hσ(j)l (x0, x, r, µr)
i2hσ(2)l ( r = 1 = µr)
i2 (2.77)
will be referred to, simply, as singular values, unless otherwise specified. The singular
value spectrum plots for the larger x values considered (x = 5 and 10) reveal well-
defined resonances (local peaks).
The dominant resonances for these larger x values occur around l ∼ π. In fact,
the resonances in question appear to arise only when x>∼ 1. Overall, it is seen that
as the material becomes electromagnetically denser, i.e., as the substrate normalized
wavenumber x increases, the magnitudes of the singular values become accordingly
larger. Since electrically small antennas such as the one considered here can effec-
tively radiate only the lowest multipole orders (such as the dipolar mode), then of
particular interest for small antenna applications is the antenna substrate-induced
enhancement for low multipolarity l. The plots reveal that the dipolar-mode (l = 1)
singular values can be significantly higher for the embedding substrate case than for
the free-space case. The improvement for x = 5 and 10 relative to the free-space case
is of more than 3 orders of magnitude (decades). This means that the magnitude of
34
1 3 5 7 9 11 13 15
l10−15
10−11
10−7
10−3
10
1051 3 5 7 9 11 13 15
10−15
10−11
10−7
10−3
10
105
x=10
x=5
x=1
x=0.5
x=0.25
Figure 2.2: Free-space singular valueshσ(1)l (x = x0, r = +1 = µr)
i2versus l for
a few representative values of x0 ≡ k0a/π. (The unit of the singular values isV 2m/A2.)
1 3 5 7 9 11 13 15
l
1
102
104
106
108
1 3 5 7 9 11 13 15
1
102
104
106
108
x=10
x=5
x=1
x=0.5
x=0.25
Figure 2.3: Normalized singular valuesh(1)l
i2versus l for x0 = 1/4 (quarter-wave
case), r = +1 and a few representative values of x.
35
1 5 10 15 20 25 30 35 40 45
l
10−2
1
102
104
106
1 5 10 15 20 25 30 35 40 45
10−2
1
102
104
106
x=20
x=10
x=5
x=1
x=0.5
x=0.25
Figure 2.4: Normalized singular valuesh(1)l
i2versus l for x0 = 10 (resonant or
electrically-large antenna), r = +1 and a few representative values of x.
the exciting current or source required for launching of the given dipolar field can
be made correspondingly smaller than in free space by embedding the antenna in
a high wavenumber or electromagnetically dense substrate. Alternatively, for fixed
source energy, the antenna size parameter a can be reduced relative to its value
without the embedding substrate. The improvement for l = 2 and 3 associated to
the larger wavenumber cases (x = 5 and 10) is also noticeable.
The respective plot for the case of a resonant or electrically-large x0 = 10 antenna
is shown in Fig. 2.4. The respective magnetic singular value spectra are shown in
Figs. 2.5 and 2.6. Many of the key features outlined above while explaining the
particular electric quarter-wave antenna case also arise for these other cases. Yet
other aspects become salient. A summary of the main results is given next, along
with some of the former observations, as general conclusions learned from these
simulations as a whole.
It is seen that, for sufficiently large multipolarity l (i.e., for l >∼ 6), and for thevalues of x0 considered which comprise both small and large or resonant antennas,
the singular values are consistently higher for the denser substrates (larger x) than
for the less dense substrates including the free-space (x = x0) case. This is true
36
1 3 5 7 9 11 13 15
l
1
102
104
106
108
1 3 5 7 9 11 13 15
1
102
104
106
108
x=10
x=5
x=1
x=0.5
x=0.25
Figure 2.5: Normalized singular valuesh(2)l
i2versus l for x0 = 1/4 (quarter-wave
case), r = +1 and a few representative values of x.
1 5 10 15 20 25 30 35 40 45
l
10−8
10−6
10−4
10−2
1
102
104
106
108
1 5 10 15 20 25 30 35 40 45
10−8
10−6
10−4
10−2
1
102
104
106
108
x=20
x=10
x=5
x=1
x=0.5
x=0.25
Figure 2.6: Normalized singular valuesh(2)l
i2versus l for x0 = 10 (resonant or
electrically-large antenna), r = +1 and a few representative values of x.
37
for both electric (j = 1) and magnetic (j = 2) modes. As we had indicated for the
particular electric quarter-wave antenna case, generally for x = x0 (no embedding
medium or free-space case), the singular value spectrum decays exponentially with
l, i.e., without resonances. This decay is more or less exponential for the smaller
antenna cases. For larger antennas the singular values remain more or less within a
given order of magnitude until about the cutoff l ∼ k0a,13 but this cutoff is clearly
higher, i.e., includes higher order multipoles, for the large wavenumber cases.14 This
further shows performance enhancement via larger wavenumber or electromagneti-
cally denser substrates since higher multipoles represent higher antenna directivity,
i.e., higher level of details or narrower width in the radiation pattern. It is also im-
portant to note that the enhancement in the singular values due to larger substrate
wavenumber k holds for both small and large multipolarities l.
Having shown some of the radiation enhancing possibilities offered by electro-
magnetically denser substrates, we discuss next the question of local optimal se-
lection of the wavenumber x. Consider, for example, a half-wave antenna (so
that x0 = 1/2) embedded in a substrate with r = 1 and launching purely mag-
netic modes (j = 2). Local maxima of the respective normalized singular valuesh(2)l (x0 = 1/2, x, r = 1, µr)
i2for l = 1, 2, and 3 were found to occur as follows: For
the emission of dipole radiation (l = 1) at x ' 1.430, with an enhancement or gainh(2)l (x0 = 1/2, x = 1.430, r = 1, µr)
i2 ' 3.110× 105, relative to free space; for theemission of quadrupole radiation (l = 2) at x ' 1.833, with a gain relative to freespace of 1.925×107; and for the emission of octupole radiation (l = 3) at x ' 2.224,with a gain of 1010. For antennas embedded in denser substrates the numerical
study indicates, however, that the improvement attained is comparatively marginal.
For example, the gain associated to going from the aforementioned values of x to the
local maxima at x ∼ 10 is only 44.03, 23.86, and 5.55 for the dipole, quadrupole andoctupole radiation cases, respectively. Conversely, a half-wave antenna radiating
purely electric modes instead displays a significantly different behavior in this re-
13This value, l ∼ k0a, approximately corresponds to the inflection point in the singular valuespectrum curve for the free-space case.
14Note that in order to see this visually one would need to plot not the normalized singular
valuesh
(j)l
i2but the singular values
hσ(j)l
i2themselves. Yet a careful comparison of normalized
values in Figs. 2.3-2.6 and the illustrative free-space values in Fig. 2.2 leads to the same conclusion.
38
gard, and the overall improvements of the substrate are also more significant. Thus
for modest values of x, a locally maximum improvement in the radiation ability of
the half-wave antenna can be attained for the following values: For electric dipole
radiation at x ' 0.946 with a gainh(1)1 (x0 = 1/2, x = 0.946, r, µr)
i2 ' 5.36; for
quadrupole radiation at x ' 1.362 with a gain relative to free space of 163.6; for
octupole radiation at x ' 1.800 with a gain of 2.952 × 104. For denser materialsthe enhancement relative to free space can be significantly larger. Thus numerical
maximization ofh(1)l (x0, x, r, µr)
i2yields the following gains associated to going
from the aforementioned values of x to the local maxima at x ∼ 10 : 126, 70.81, and33.56 for the dipole, quadrupole and octupole radiation cases, respectively. The first
two of those numbers are relatively significant enhancements, yet for much denser
materials the enhancements are less dramatic, though still meaningful. These con-
siderations are reinforced from an alternative point of view in the next subsection
which revisits the topic of locally optimal antenna substrate wavenumber for a broad
range of antenna sizes under electric dipole radiation.
A legitimate question arises as to the physical reason behind the appearance
of these resonances in the spectra of the non-free-space singular values. As noted
earlier a careful examination of the quantities F (j)l defined in (2.20) shows that
these quantities are essentially the amplitudes of the internal electromagnetic Mie
fields due to the scattering of a plane wave by a sphere of radius a and propagation
constant k embedded in an infinite homogeneous medium of propagation constant
k0; F(1)l being the amplitudes of the electric modes and F
(2)l those of the magnetic
modes [123, 113]. The question that arises now is: how does the presence of these
amplitudes affect the behavior of the singular valuesh(j)l (x0, x, r, µr)
i2, or, more
precisely, are those resonant peaks, which correspond to local maximum enhance-
ment, related to Mie resonances? Before answering this question we review very
briefly the features of Mie resonance that are most relevant to our results. Mie
resonances of the abovementioned sphere are characterized by the vanishing of the
denominators of the amplitudes F (j)l , or, more realistically, by the requirement that
those denominators be minimum [113]. Thus the resonance conditions can be cast
39
in the form of approximate transcendental equations, viz.,
√rVl(x0)
h(+)l (x0)
' õrUl(x)
jl(x), (2.78)
for the electric modes, and
õr
Vl(x0)
h(+)l (x0)
' √ rUl(x)
jl(x), (2.79)
for the magnetic modes, where Ul and Vl are the functions defined in (2.21) and
(2.22). (These rather loosely stated conditions could certainly be made more rigor-
ous but this is enough for our purposes.) Because of the presence of Bessel functions,
Eqs. (3.8) and (3.9) admit a discrete, albeit infinite, set of solutions. These solutions
correspond to the so-called Mie resonances.
Now, we can go back to the question of how the observed resonant peaks
which correspond to local maximum enhancement relate to Mie resonances. Sin-
gular valuesh(j)l
i2, defined by Eqs. (2.77), (2.31), and (2.32), are composed
not only of the quantities¯F(j)l
¯2, defined in (2.20), but also of another term,
viz.,hκ(j)l (x0, x)
i2, defined in (2.32), and unless these latter quantities are suf-
ficiently well-behaved one cannot conclude anything as to the relationship of the
resonant values ofh(j)l (x0, x, r, µr)
i2to Mie resonances. Incidentally, the quan-
titieshκ(j)l (x0, x)
i2where x ∈ R are essentially non-pathological combinations of
the spherical Bessel functions jl (λa) which are well-behaved for all integer values
of l and λ ∈ R [124] (which represent the most general cases considered in this
work). Hence, one can confidently claim that the observed peaks in the spectrum
of the singular values are primarily due to the phenomenon of Mie resonance and
maximum enhancement conditions are effectively summarized by the two conditions
(3.8) and (3.9). Therefore, for a given antenna radiating at a prescribed frequency,
the discrete set of solutions x corresponds to a set of constitutive parameters s and
µs that maximize the radiated electromagnetic fields. As their amplitudes increase
these radiated fields draw energy from the embedding medium. But because this
medium is of finite extent the energy extraction process saturates, ultimately, and
40
02
46
810 x
01
23
45
x0
1. × 10−13
1. × 10−11
1. × 10−9
1. × 10−7
02
46
8 x
12
34
5
Figure 2.7: Logarithmic mesh plot of the source energy E(j=1=l)ME versus x0 and x fora double-positive material with r = +1.
as a result of this saturation the fields fall short of effectively “blowing up.”
Further Details: Electric Dipole Radiation
This part examines in greater detail the fundamental electric dipole radiation case, in
particular, the multipole moment a(j)l,m = 1 if j = 1 = l and m = 0, and a(j)l,m = 0 oth-
erwise. The minimum source energy reduces in this case to E(j=1=l)ME (x0, x, r, µr) =hσ(1)1 (x0, x, r, µr)
i−2. Fig. 2.7 shows a mesh plot of the minimum source energy
E(j=1=l)ME versus the normalized wavenumbers x0 and x for a DPS substrate material
with r = 1. For a DNG substrate material having r = −1, the numerical studyshows that the minimum source energy displays a very similar, though not com-
pletely symmetrical, behavior when x changes sign, for a given x0. Consequently,
source energy E(j=1=l)ME is not an even function of x, and hence distinguishes between
DPS and DNG embedding substrates. Fig. 2.8 shows slices or cross-sections of the
41
0 1 2 3 4 5 6 7 8 9 10 11x
10−8
10−6
10−4
10−2
Source
Energy
0 1 2 3 4 5 6 7 8 9 10 11
10−8
10−6
10−4
10−2
x0=10
x0=5
x0=1
x0=0.5
x0=0.25
Figure 2.8: Logarithmic plot of the source energy E(j=1=l)ME versus x for r = +1 andsome representative values of x0 for a double-positive medium.
mesh plot in Fig. 2.7 for particular values of the free-space normalized wavenumber
x0, and Fig. 2.9 shows a slice similar to 2.8 but for negative x. Fig. 2.7 also shows
that, in general terms, source energy tends to decrease as the size of the antenna
increases, this is also true when x is negative. Thus as the antenna size increases
it tends to be easier to distribute the source currents in a more efficient way. As
shown in Figs. 2.8 and 2.9, for small antennas the source energy exhibits its first
local minima at |x| ∼ 1. In particular, for x0 = 1/4 (quarter-wave antenna case)
and x0 = 1/2 (half-wave case) the first local minimum of E(j=1=l)ME appears for pos-
itive x at x ' 0.960 and x ' 0.946, respectively, and for negative x at x = −0.760and x = −0.860, respectively. For x0 = 1 (full-wave antenna case) the first lo-
cal minimum of E(j=1=l)ME appears for positive x at x ' 1.155 and for negative x atx ' −1.200. However, for large antennas a slightly more subtle behavior is observed.If |x| < x0 the local minima of E(j=1=l)ME appear at |x| ∼ (2n+ 1) /2, n = 1, 2, 3, ...,while if |x| > x0 the minima appear at |x| ∼ n, n = 1, 2, 3, ..., with the least minimum
still belonging to the smallest antenna (cf. Figs. 2.8 and 2.9). These rules-of-thumb
depend on the particular combination of constitutive parameters r and µr under in-
vestigation. To illustrate this we display in Fig. 2.10 the logarithmic plot of E(j=1=l)ME
42
0−1−2−3−4−5−6−7−8−9−10−11x
10−8
10−6
10−4
10−2
Source
Energy
0−1−2−3−4−5−6−7−8−9−10−11
10−8
10−6
10−4
10−2
x0=10
x0=5
x0=1
x0=0.5
x0=0.25
Figure 2.9: Logarithmic plot of the source energy E(j=1=l)ME versus x for r = −1 andsome representative values of x0 for a double-negative medium.
versus x for µr = 1. In this figure one clearly sees that the rules are interchanged,
i.e., now, if x < x0 the local minima of E(j=1=l)ME appear at x ∼ n, n = 1, 2, 3, ..., while
if x > x0 the minima appear at x ∼ (2n+ 1) /2, n = 1, 2, 3, ...(This, of course, is
rather expected as the positions of the minima depend on the particular combination
of constitutive parameters r and µr.) Another feature which is worth noting is that
the local minima of the functional energy keep decreasing for increasing values of |x|at an even slower rate, though the numerical study does not seem to conclusively
indicate whether for |x| À 10 there is a limiting value.
A fundamental aspect is presented in Fig. 2.11 which shows a plot of the free-
space case source energy E(j=1=l)ME (x = x0, r = 1 = µr) versus x0. This plot illus-
trates the well-known behavior that the source functional energy increases extremely
fast as x0 decreases below x0 ∼ 1/2: apparently it increases exponentially. The en-ergy reaches a mildly oscillating valley at x0 ∼ 1/2, but the oscillations die out veryrapidly, while the energy approaches a certain limiting value as ka increases. On the
one hand, these observations imply that launching the dipolar mode costs signifi-
cantly more energy if the size of the antenna is smaller than, say, half a wavelength,
and, on the other hand, that increasing the size of the antenna beyond this threshold
43
0 1 2 3 4 5 6 7 8 9 10 11x
10−8
10−6
10−4
10−2Source
Energy
0 1 2 3 4 5 6 7 8 9 10 11
10−8
10−6
10−4
10−2
x0=10
x0=5
x0=1
x0=0.5
x0=0.25
Figure 2.10: Logarithmic plot of the source energy E(j=1=l)ME versus x for µr = 1 andsome representative values of x0.
0 1 2 3 4 5 6x0=x
10−5
10−4
10−3
10−2
10−1
Source
Energy
0 1 2 3 4 5 6
10−5
10−4
10−3
10−2
10−1
Figure 2.11: Logarithmic plot of the source energy E(j=1=l)ME (x = x0) versus x = x0for r = +1.
44
0 2 4 6 8 10 12 14 16 18 20x
10−2
1
102
104
Gain
0 2 4 6 8 10 12 14 16 18 20
10−2
1
102
104
x0=10
x0=0.5
x0=0.25
Figure 2.12: Logarithmic plot of the gain G, versus x for r = +1 and some selectedvalues of x0.
has little effect on the minimization of the source functional energy associated to
this particular mode. In Fig. 2.12 the gain, defined as
G ≡ E(j=1=l)ME (x = x0, r = 1 = µr)
E(j=1=l)ME
=h(1)1
i2, (2.80)
has been plotted as a function of the normalized substrate wavenumber x. The
information depicted in this plot is equivalent to that in Fig. 2.8, but the latter figure
highlights more clearly the reduction in required source energy for the radiation of
the dipolar mode for small and large antennas. For the small antenna cases, the
enhancement is seen to be of more than 4 orders of magnitude for the resonant x
values. Similar results (not shown) were found pertinent to radiation enhancement
for higher order multipoles in the large antenna cases.
Finally, to further illustrate the possibility of reducing radiator size while achiev-
ing a given radiation pattern with prescribed source resources, specifically, source
energy, we considered the free-space wavenumber k0 = π/4, and sought for values
45
of the size parameter a for which the minimum source energy of a source embedded
in a medium having k = 10π renders the same source energy as a unit-valued a
embedded in free space, for which k = k0 = π/4. For an embedding substrate with
r = 1 the first such values of a are 0.098, 0.101, 0.196, 0.204, ..., (units of meters)
which are seen to occur in pairs around 0.1, 0.2, 0.3, etc. This is not surprising in
light of the formula introduced earlier, in particular, the locally optimal values of ka
are ka ∼ nπ, n = 1, 2, 3, ..., that is, a ∼ 1/10, 2/10, 3/10, .... The values of the sizeparameter a for which the source energy in question coincides with the free-space
case source energy for a larger source having unit-valued radius then occur in pairs
around these optimal values, which completes the picture.
2.3.2 Tuned Minimum Energy Sources: Additional Zero ReactivePower Constraint
Next we consider minimum energy sources subjected to the additional zero reactive
power constraint. In particular, we require the reactive power to vanish, that is,
Im [P] = 0. As in the preceding subsection, the focus is the fundamental case
of an electric dipole radiator (specifically, a(j)l,m = 1 if j = 1 = l and m = 0,
and a(j)l,m = 0 otherwise). Particular attention is given to the quarter-wave and the
half-wave antenna cases, though some results related to larger antennas are also
presented.
As in [14], we define the normalized reactive power
g(1)1 ≡
g(1)1
Re [P]= η0g
(1)1 , (2.81)
where Re [P] = 1/2η0 is the radiated power and where the free-space wave
impedance η0 =pµ0/ 0 ' 120π Ω. We define χ0 as the Lagrange multiplier value
χ which annuls the normalized reactive power g(1)1 , i.e., g(1)1 (χ)
¯χ=χ0
= 0 and for
which the resulting source energy is minimal among all such zero reactive power
Lagrange multiplier values. The value in question was consistently found to occur
in the vicinity of χ = 0. This is not surprising since the absolute or unconstrained
minimum energy source and its energy E(j=1=l)ME correspond to χ = 0, that is, the
46
-4 -2 0 2 4Lagrange Multiplier
-4
-2
0
2
4
ReactivePower
Figure 2.13: Plot of the normalized reactive power g(1)1 versus χ for x0 = 1/4 = xand r = +1.
minimum energy source is min E(j=1=l)EP ≡ limχ→0E
(j=1=l)EP (x0, x, r, µr, χ) = E(j=1=l)ME
(see section 3).
Fig. 2.13 is a plot of the normalized reactive power g(1)1 versus the Lagrange
multiplier χ for a quarter-wavelength antenna, embedded in substrates with r = 1
and x = 1/4. The Lagrange multiplier value χ0 is sought for which the respective
source energy (showed in Fig. 2.14) is minimized among all χ values rendering zero
reactive power. Figs. 2.15 and 2.16 show the respective plots for r = −1, andTables 2.1, 2.2, 2.3, and 2.4 summarize the values of χ0, source energy for χ = χ0,
i.e., E(j=1=l)E,P , and absolute minimum energy E(j=1=l)ME , for the cases addressed in
these plots, as well as for other cases (other values of x).
One notes from these results that the minimum energy solution J(j=1=l)ME yields
minimum source energy E(j=1=l)ME or current level but its reactive power is comparable
to the maximum, saturated value corresponding to χ À 1 for the DPS materials,
and χ ¿ −1 for DNG materials (cf. Figs. 2.13 and 2.15). On the other hand,
the new solution J(j=1=l)E,P corresponding to χ0 yields zero reactive power at the
expense of a raised source energy or current level (cf. Tables 2.1, 2.2, 2.3, and 2.4).
The difference between the source energies E(j=1=l)EP and E(j=1=l)ME of the two sources
47
-4 -2 0 2 4Lagrange Multiplier
10−3
10−1
10
103
105SourceEnergy
10−3
10−1
10
103
105
Figure 2.14: Plot of the source energy E(j=1=l)EP versus χ for for x0 = 1/4 = x andr = +1.
-4 -2 0 2 4
Lagrange Multiplier
-10
-5
0
5
10
Reactive
Power
Figure 2.15: Plot of the normalized reactive power g(1)1 versus χ for x0 = 1/4 = −xand r = −1 (“anti-vacuum.”)
48
-4 -2 0 2 4Lagrange Multiplier
10−4
10−2
1
102
104
106
SourceEnergy
10−4
10−2
1
102
104
106
Figure 2.16: Plot of the source energy E(j=1=l)EP versus χ for x0 = 1/4 = −x andr = −1 (“anti-vacuum.”).
JE,P and JME , respectively, is the source energy of the additional nonradiating part
contained in JE,P whose role in the new source is to counteract the reactive power ofthe minimum energy source alone. It decreases as the electromagnetic density of the
substrate increases, this being true for both DPS and DNG substrates. We found
that, for x0 = 1/4 and 1/2, performances better than those of the free-space cases
(i.e., for which k = k0 and r = 1 = µr) can be achieved.15 Superior performance
can also be obtained by means of a judicious choice of the substrate constitutive
properties, as we explain below. In addition to this, we note that the minimum of the
energy decreases as the electromagnetic density of the substrate increases, whether
the substrate is DPS or DNG. In Figs. 2.14 and 2.16 it is clear that as χ → 0 the
source energy E(j=1=l)E,P reaches an absolute minimum min E(j=1=l)EP , as expected. This
minimum is not the same for DPS materials and DNG metamaterials (cf. Tables
2.1, 2.2, 2.3, and 2.4), as we discussed earlier. Interestingly, the cancellation of the
reactive power is not always possible. For instance, for a quarter-wave antenna, and
for r = 1, the Eq. g(1)1 (χ)
¯χ=χ0
= 0 admits no solutions if x = −1/4 or −1/2, as
15Cf. Tables 2.1, 2.2, 2.3, and 2.4, though for the sake of space the energy differenceE(j=1=l)EP (χ0)−min E(j=1=l)EP is not explicitly displayed in the tables.
49
illustrated in Fig. 2.15 for the x = −1/4 case.16Furthermore, it also follows that if one allows the electromagnetic properties of
the embedding substrate (i.e., r and µr) to vary then one could make the reactive
power vanish for χ0 = 0, this being a matching condition under which the minimum
energy sources are not only of local minimum energy (see below) but also self-
matched to resonance. Let us illustrate this for a quarter-wavelength antenna. For
a given positive relative electric permittivity, for instance r = 1, we find that the
matching condition mentioned above is satisfied for x ' 0.511, i.e., in this case
r = 1 and µr ' 4.18. Now, for a given negative relative electric permittivity, for
instance for r = −1, we find that the matching condition is satisfied for x ' −1.338,i.e., in this case r = −1 and µr = −28.64. A word of caution is necessary at thispoint. From the definition itself of the above-mentioned matching one obtains for
χ0 = 0 : E(j=1=l)E,P = min E(j=1=l)E,P (cf. Tables 2.1 and 2.3.) Yet one must not be
lured into thinking that the substrate constitutive properties r and µr associated
to the matching cases must correspond to global minima for E(j=1=l)E,P , i.e., that they
represent the best substrate values. This is very clearly illustrated in Tables 2.1 and
2.3 where for x = 1, 5 and 10 in Table 2.1 and for x = −1,−5 and −10 in Table2.3 one has E(j=1=l)E,P
¯χ0 6=0
< E(j=1=l)E,P¯χ0=0
. In other words a quarter-wavelength
antenna embedded in substrates having those values of x as their electromagnetic
densities exhibit source energies smaller than those exhibited by the antenna when
it is embedded in a substrate whose constitutive parameters satisfy the matching
condition.
16This is also true for x0 = 1/2, and 1 though the figures are not presented here.
50
Table 2.1: Results of the numerical study for the constrained quarter-wave antennaembedded in a double-positive material with r = +1. (The unit of the sourceenergies is A2/m.)
x χ0[10−4] E(j=1=l)EP (χ0) min E(j=1=l)EP
1/4 −421.2 4.637× 10−4 1.161× 10−41/2 −5.803 8.031× 10−5 8.014× 10−50.511 0 7.809× 10−5 7.809× 10−51 −44.52 2.090× 10−5 2.206× 10−65 −8.039 4.782× 10−6 7.094× 10−810 −4.008 2.486× 10−6 1.763× 10−8
Table 2.2: Results of the numerical study for the constrained half-wave antennaembedded in a double-positive material with r = +1. (The unit of the sourceenergies is A2/m.)
x χ0[10−4] E(j=1=l)EP (χ0) min E(j=1=l)EP
1/4 −899.5 3.087× 10−4 5.854× 10−51/2 −80.37 4.857× 10−5 4.191× 10−51 −35.08 1.513× 10−5 8.824× 10−65 −13.51 3.690× 10−6 2.838× 10−710 −7.234 2.081× 10−6 7.051× 10−8
Table 2.3: Results of the numerical study for the constrained quarter-wave antennaembedded in a double-negative metamaterial with r = −1. (The unit of the sourceenergies is A2/m).
x χ0[10−4] E(j=1=l)EP (χ0) min E(j=1=l)EP
-1/4 · · 3.071× 10−5-1/2 · · 1.462× 10−5-1.338 0 7.980× 10−5 7.980× 10−5-1 24.27 6.288× 10−5 2.540× 10−5-5 7.218 1.708× 10−5 8.167× 10−7-10 3.757 9.392× 10−6 2.029× 10−7
51
Table 2.4: Results of the numerical study for the constrained half-wave antennaembedded in a double-negative metamaterial with r = −1. (The unit of the sourceenergies is A2/m.)
x χ0[10−4] E(j=1=l)EP (χ0) min E(j=1=l)EP
-1/4 · · 3.718× 10−5-1/2 · · 2.554× 10−5-1 37.88 2.696× 10−5 1.462× 10−5-5 13.74 6.821× 10−6 4.702× 10−7-10 7.304 3.825× 10−6 1.168× 10−7
2.4 Conclusion
In this chapter we presented a mathematical theory of the full-vector, electromag-
netic inverse source problem which is applicable to sources embedded in substrates,
with applications to the analysis, source-synthesis and characterization of anten-
nas embedded in substrates. The present work completes the research program on
the inverse source problem in non-homogeneous background media initiated in [17].
The developments considered sources in a spherical volume of radius a, so that the
results and overall conclusions fundamentally hold for rather general sources of a
given maximal dimension (2a), and this also conveniently enabled us to treat the
relevant source-to-field mappings in the spherical coordinate system or multipole
wavefunction domain.
A key objective was to gain understanding from first principles of potential radi-
ation enhancements (reduction of required antenna resources (physical size, current
levels, level of tuning, and so on) for a given far field) due to such substrates. This
problem was treated in the present work within a general and non-device-specific
framework whose predictions (such as performance bounds) under normalized re-
sources are fundamental. The derived theory and the associated numerical illustra-
tions also yielded fundamental insight about the interplay of the variables involved,
as well as an idea of good values for antenna design parameters.
The results were discussed addressing separately the cases of small versus large
or resonant antennas, with the overall conclusion that for small antennas one can
significantly enhance the radiated power or compress source size via the substrates
52
under normalized antenna resources, while for larger antennas the use of substrates
can significantly enhance both radiated power and directivity (related to the num-
ber of essentially independent field modes that can be radiated effectively) under
the given resources. Our analysis thus formally explains, from a first principles,
non-device-specific source-inversion point of view, similar findings by other groups
working in the general area of substrate-enhanced antennas.
53
Chapter 3
Radiation Enhancement due toMetamaterial Substrates:Core-Shell System
To present the non-antenna-specific theory of substrate enhancement developed in
Chapter 2 in a context that relates to important work done by other groups in
this area, we generalize, in this chapter,1 the formalism developed in Chapter 2
to sources that are embedded within two nested spheres2 made up of arbitrary
lossy substrates. This particular configuration is of great practical importance for
both antenna radiation and scattering and hence imaging resolution enhancement.
It enables enhancements that are more dramatic than those due to homogeneous
substrates. Particular emphasis is given to the special case when the two nested
spheres are made up of materials with oppositely signed constitutive parameters.
The radiative as well as the scattering properties of a system of two nested
spheres of ordinary materials associated with a dipole have been considered by
several authors [126, 127, 128, 129]. These studies have now been extended to
cases where metamaterials are present. For instance, Gao and Huang [130] have
calculated the extinction efficiency of the core-shell system. Following the steps of
Aden and Kerker [131], Alù and Engheta [132, 133] have looked at the resonant
1This study has been published in [125].2Other terminologies used for this system are: Core-shell system or three-region system.
54
a b
, aa
b , b
Vacuum
Figure 3.1: Geometry of the three-region system under consideration. The drivingpoints and material structure of the antenna are confined within a spherical volumeV of radius a. The inner sphere of radius a has relative permittivity a and relativepermeability µa. This inner sphere is surrounded by a spherical shell of inner radiusa and outer radius b and has relative permittivity b and relative permeability µb.The core-shell system is immersed in the vacuum.
scattering that arises when the two spheres are constructed by combining a pair
of materials with oppositely signed constitutive parameters. Ziolkowski and Kipple
[40] have established the reciprocity of the peculiar scattering properties described
by Alù and Engheta [132] and the enhanced radiation power they realized would
occur when an electrically small dipole antenna is surrounded by a metamaterial
shell [8].
The geometry of the system to be investigated in the following is depicted in
Fig.(3.1). The inner sphere, of radius a, has relative electric permittivity a ≡sphere/ 0 and relative magnetic permeability µa ≡ µsphere/µ0. This inner sphere
constitutes the core of the system and is the smallest spherical volume V that cir-
cumscribes the largest physical dimension of the original antenna which is treated
next, under a suppressed time dependence e−iωt, as a primary, or impressed, currentdensity J(r). The core is surrounded by a spherical shell, of inner radius a and outer
radius b. The relative constitutive parameters of the shell are relative electric per-
mittivity b ≡ shell/ 0 and relative magnetic permeability µb ≡ µshell/µ0. Thus the
resulting three-region system may be characterized by a total electric permittivity
55
distribution of the form
(r)
0= aΘ(a− r) + b [Θ(r − a)Θ(b− r)] +Θ(r − b) (3.1)
and a total magnetic permeability distribution of the form
µ(r)
µ0= µaΘ(a− r) + µb [Θ(r − a)Θ(b− r)] +Θ(r − b), (3.2)
where Θ denotes Heaviside’s unit step function (Θ(x) = 1, for x > 1, otherwise
Θ(x) = 0), and 0 and µ0 are, respectively, the electric permittivity and magnetic
permeability of the vacuum.
The core and the surrounding shell, being assumed to be generally lossy, are as-
signed relative constitutive parameters that are generally complex. These constitu-
tive parameters are, thus, assumed to have the generic forms: α = Re [ α]+i Im [ α]
and µα = Re [µα] + i Im [µα], α = a, b (where a is for the inner sphere and b is for
the outer shell.) Note that the losses are indicated by the presence of non-negative
imaginary parts of the permittivity and the permeability.
The approach adopted next is to follow the same steps as in Chapter 2 to formu-
late an inverse source problem in substrate media, whose objective is to deduce an
unknown primary current density J(r) that is contained, along with the substrate,
in the spherical volume V , and that generates a prescribed exterior field for |r| > b.
It is important to point out that the present formulation of the inverse source
problem in the two-nested-spheres configuration is also relevant to that of the com-
panion inverse scattering problem. In fact, it is mathematically equivalent to that
of an inverse scattering problem in which a single incident field is used as excita-
tion. Hence, the results on the inverse source problem and on the possibility of
extracting higher spatial frequency information about the unknown object thanks
to the presence of the embedding medium [17] also point out the possibility of
similarly enhancing imaging resolution in the associated inverse scattering problem
with helper substrate media. In addressing the inverse source problem in these me-
dia one is automatically paving the way for inverse scattering formulations in such
media. This is particularly pertinent in the modern context of so-called qualita-
tive imaging methods based on support estimation of induced sources and which
56
are non-iterative (e.g., linear sampling, factorization method, time-reversal MUSIC,
and so on) [134, 135, 136]. Those inverse scattering methods are of the so-called
“inverse-source” type, in other words, they are based on a companion inverse source
problem.
3.1 The Radiation Problem
The electric and magnetic multipole moments, a(1)l,m and a(2)l,m, respectively, are related
to the current distribution J by Eq.(2.17). It is shown in Appendix C that for
backgrounds whose permittivity and permeability are given by Eqs.(3.1,3.2) one
has3
B(j)l,m ≡
⎧⎪⎪⎨⎪⎪⎩−η0l(l+1)F
∗(1)l ∇× [jl(k∗ar)Yl,m(r)] ; j = 1
−ik0η0l(l+1) F
∗(2)l jl(k
∗ar)Yl,m(r) ; j = 2,
(3.3)
where ka = ω√
a 0µaµ0 is the inner sphere substrate wavenumber, and where we
have defined4
F(j)l ≡
⎧⎪⎨⎪⎩− b
∆1k0kba2b2; j = 1
−µaµb∆2k0kba2b2
; j = 2,
(3.4)
where (cf. Eqs. (26,27) in [131], and Eqs.(8,9) in [132, 133])
∆1 =
¯¯¯
0 kbUl(kbb) kbVl(kbb) −k0Vl(k0b)0 bjl(kbb) bh
(+)l (kbb) −h(+)l (k0b)
kaUl(kaa) −kbUl(kba) −kbVl(kba) 0
ajl(kaa) − bjl(kba) − bh(+)(kba) 0
¯¯¯ (3.5)
and
∆2 =
¯¯¯
0 jl(kbb) h(+)l (kbb) −h(+)l (k0b)
0 kbµbUl(kbb)
kbµbVl(kbb) −k0Vl(k0b)
jl(kaa) −jl(kba) −h(+)l (kba) 0kaµaUl(kaa) − kb
µbUl(kba) − kb
µbVl(kba) 0
¯¯¯ . (3.6)
3Cf. Eq.(2.19).4Cf. Eq.(3.4).
57
In (3.5) and (3.6) the quantities Ul and Vl are defined such that by Eqs.(2.21) and
(2.22).
Note that in Eqs.(3.4,3.5,3.6) k0 = ω√
0µ0 and kb = ω√
b 0µbµ0 are the prop-
agation constants in the vacuum and in the shell, respectively. Also it is clear from
Eqs.(3.4,3.5,3.6,2.21,2.22) that the parameters F(j)l , j = 1, 2, are not constants but
in fact functions of several parameters.
It is not hard to show that the terms F(j)l reduce to the Mie amplitudes F (j)l ,
as defined in (2.20) , when a = b, as expected.5 Consequently, they also reduce to
unity in the free-space case, i.e., when a = b, a = 1 = b, and µa = 1 = µb, causing
Eqs.(3.3) to reduce to the free-space case equations [59, 74].
3.2 Inverse Source Theory Based on Constrained Opti-mization
Following the same procedure as in Chapter 2 one finds that the minimum-energy
source is given by (2.29) and that the corresponding minimum source energy is (2.35)
where the positive-definite singular values are nowhσ(j)l
i2 ≡ (B(j)l,m,B
(j)l,m) ≡ |F(j)l |2
hκ(j)l
i2, (3.7)
where are given by Eqs.(3.4,3.5,3.6,2.21,2.22) andhκ(j)l
i2are given by (2.32). In
Chapter 2, analytical expressions where derived for the integrals appearing in (2.32),
for k2a ∈ R, by means of Lommel’s second integral. In the general case, however, thecomplex argument of the Bessel functions makes it necessary to recur to numerical
methods of integration.
The dimensionless positive-definite normalized singular valuesh(j)l
i2have been
defined in Chapter 2 (cf. Eq.(2.77)). In that definition the quantitieshσ(j)l ( r = 1 = µr)
i2referred to what we may call “the free-space case.” In the case of a homogeneous
sphere this was easy to define: Setting r = 1 = µr was equivalent to defining a
reference antenna of length 2a and radiating in the vacuum, with respect to which
radiation performance was evaluated. For the core-shell system investigated in this
5See also [17, 137, 112].
58
chapter, however, the free-space case is not as simple to define. Here, it is defined
as the case in which the original antenna, as defined in Section I, radiates in the
vacuum. Quantitatively this corresponds to the case: a = b, a = b = µa = µb = 1.
In other words, the reference antenna with respect to which the comparisons are
carried out is the original antenna without the shell. This may sound as an unfair
comparison, after all the new antenna, i.e., the core-shell system, represents a totally
new antenna with its own new dimensions and induced currents due to the addition
of the shell to the original antenna. What is more, we allowed, in the numerical
simulations, the dimensions of the outer shell to be of comparable size to the core.
Consequently, this definition may not sound as the best definition for a reference or
standard antenna to measure the enhancement with respect to. Nevertheless, this
definition is underlain by a simple, if not naive, answer to the question of how the
addition of a metamaterial shell would affect the performance of an existing antenna,
or, similarly, how the embedding of an existing antenna in a given core-shell system
with oppositely signed constitutive parameters would affect the performance of the
antenna. A more detailed investigation of the effect of reference antennas on en-
hancement level estimates and the issue of fairness in antenna radiation performance
is presented in Chapter 4.6
3.3 Numerical Results and Case Studies
In this section we turn to the application of the theory exposed above to the elu-
cidation of the effect of embedding media on antenna radiation performance. The
goal is to gain an understanding of the effect of the antenna substrate on the mini-
mum source energy for a given radiation pattern. Because of the dependence of the
problem on so many parameters we limit ourselves to a few illustrative cases. Three
classes of antennas are investigated: a quarter-wavelength antenna (i.e., 2a = λ/4), a
λ/40 antenna (i.e., 2a = λ/40) and, a λ/400 antenna (i.e., 2a = λ/400). The driving
frequency of the antenna is set to f = 3.75 GHz. This corresponds to a = 1 cm, 0.1
cm, 0.01 cm, for the λ/4 antenna, λ/40 antenna, and λ/400 antenna, respectively.
Needless to say that these particular choices of the numerical values of f and a are
arbitrary. However, they lie well within the range of values used in the scientific and
6See also [138].
59
engineering literatures [8, 40, 41, 132, 133]. Particular attention is paid to electri-
cally small antennas. This is in view of the exciting properties that these antennas
exhibit in the subwavelength limit when embedded in a pair of oppositely signed
materials. Here we adopt the definition [139] according to which an electrically
small antenna in the vacuum is defined as an antenna for which k0a = 2πa/λ ≤ 0.5,where a is the radius of the sphere that encompasses the entire original antenna.
Hence, a more detailed investigation is carried out for the λ/400 multipolar and
dipolar antennas. At this point it is appropriate to give one more definition. In the
plots and associated discussion we consider the normalized wavenumber defined by
xb ≡ kb/π. The normalized wavenumber xb represents the wavenumber of the field
in the shell.
It follows from Eqs.(2.35,2.77) that, generally, the larger the singular valuesh(j)l
i2the smaller the minimum source energy EME required for the launching
of a given radiation pattern with a source of a given size. Therefore the larger the
singular valuesh(j)l
i2the greater the associated enhancement, due to the associated
substrates, of radiation of the lth multipole order field with given resources. It is
thus important to understand the dependence of the singular valuesh(j)l
i2on the
several parameters it depends on for both the electric (j = 1) and the magnetic
(j = 2) cases. Large singular values, such as resonances or peaks in the plots
of the singular values versus these variables, will indicate enhanced radiation for
such operational modes or conditions, with the given resources. This aspect is
investigated numerically next. For the sake of conciseness, however, and because of
the noted similarity (see next section) between the behavior of the electric singular
values and the magnetic singular values we concentrate our attention in what follows
on the study of the electric singular values.
As noted above there is a tight relationship between the local behavior of the
singular valuesh(j)l
i2and the launching ability of the antenna: any resonant peaks
in the spectra ofh(j)l
i2would indicate the presence of local enhancements in the
launching ability of the antenna. Nevertheless, it is well known [132, 133, 40] that the
core-shell system does possess a resonant behavior (resonant scattering and resonant
radiation) that can be traced back to the presence of the natural modes (polaritons)
in the system. Thus we should anticipate the occurrence of such resonant behavior
60
in our case too. The question that arises then is: should they appear, can we ascribe
the resonant peaks in the spectra of the singular valuesh(j)l
i2to the presence of
polaritons? The conditions for the existence of polaritons in the core-shell system
are summarized by their dispersion relations [131, 132, 133]
∆1 = 0 (3.8)
for the electric modes, where ∆1 has been defined in (3.5), and
∆2 = 0 (3.9)
for the magnetic modes, where ∆2 has been defined in (3.6). But the singular valuesh(j)l
i2are given by Eqs.(2.77,2.32,3.7), i.e., they are composed of two quantities:¯
F(j)l
¯2, defined in (3.4), and
hκ(j)l
i2, defined in (2.32). It is true that (3.8) and (3.9)
when substituted in the definition of amplitudes¯F(j)l
¯2(i.e., in Eq.(3.4)) would
provide the resonance conditions (3.8) and (3.9) with a very strong effect on the
behavior of the singular valuesh(j)l
i2. Yet, in order for us to be able to confidently
attribute the resonant peaks to the presence of polaritons we have to show that the
quantitieshκ(j)l
i2do not exhibit a resonant behavior similar to that of
¯F(j)l
¯2that
would potentially shift or even kill the peaks created by the resonance conditions
of the polaritons. Actually the quantitieshκ(j)l
i2are essentially non-pathological
combinations of the spherical Bessel functions of the first kind jl and their derivatives
which are sufficiently well-behaved for all integer values of l and complex values of the
argument [124]. Hence, we can confidently claim that the spectrum of the singular
valuesh(j)l
i2will indeed exhibit resonant peaks, and thus maximum enhancements,
and that these peaks will be primarily due to the presence of polaritons as stipulated
by the resonance conditions (3.8) and (3.9).
Finally, we point out that the MATHEMATICA code used for the numerical
simulations has been validated against some well-known cases such as the free-space
case [14] and the single spherical substrate case.7
7See Section 2.3.
61
1.5 2 2.5 3 3.5 4
d10−2
1
102
104
Singular
Values
10−2
1
102
104
l=5
l=4
l=3
l=2
l=1
Figure 3.2: Logarithmic plot of the normalized electric singular valuesh(1)l
i2for
a λ/4 antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not tocontain any material (i.e., a = 1 = µa) and the surrounding shell is assumed to bea lossless DPS material with µb = 1 and xb = 50 m−1.
3.3.1 Lossless Substrates
Vacuum—Core—DPS—Shell System
We focus on the case of a spherical shell of non-magnetic lossless DPS material
( b > 1, µb = 1) surrounding an inner sphere with no substrate material in it ( a =
1 = µa), a system that we will refer to as a vacuum-core-DPS-shell system.
In Figs.3.2-3.4 we plot the electric singular valuesh(1)l
i2versus the radii ratio
d ≡ b/a, for a quarter-wavelength antenna of maximum length 2a = 2 cm operating
at a the frequency f = 3.75 GHz. The plots show that the singular valuesh(1)l
i2exhibit a cyclic behavior with resonant peaks, i.e., local enhancements, appearing at
specified values of the radii ratio d. (The same behavior is exhibited by the magnetic
singular valuesh(2)l
i2, though the plots are not shown here.) As mentioned above
these resonant peaks correspond to an enhancement in the launching ability of the
antenna for such operational modes with the given resources.
The plots also show that for the smaller values of the radii ratio d (by “smaller
values” we mean d ≤ r (xb), where r (xb) is a value that decreases as xb increases)
62
1.5 2 2.5 3 3.5 4
d10−2
1
102
104
106
108
Singular
Values
10−2
1
102
104
106
108
l=5
l=4
l=3
l=2
l=1
Figure 3.3: Logarithmic plot of the normalized electric singular valuesh(1)l
i2for
a λ/4 antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not tocontain any material (i.e., a = 1 = µa) and the surrounding shell is assumed to bea lossless DPS material with µb = 1 and xb = 150 m−1.
the best local enhancements, though not always resonant, are observed for the lower
multipole modes starting with the dipolar modes. As the radii ratio increases the
best enhancements shift to the modes with higher multipolarities: quadrupole, then
octupole, and so on. The explanation of this observation is that the higher the
multipolarity of the mode the more “intricate” is its structure such that exciting
higher multipolarity modes in an efficient way requires thicker shells i.e., shells that
possess a “richer” charge structure. The peaks become sharper and more packed as
the electromagnetic density (i.e., the wavenumber) of the shell material increases.
This occurs because as the electromagnetic density of the material increases, i.e., as
b increases (since µb = 1, in this case), the ability of a given thickness of the material
to support more natural modes of oscillation (i.e., polaritons) also increases. For
very electromagnetically-dense materials, however, the heights of the peaks saturate
indicating a saturation in the launching enhancement levels and the peaks pile up at
almost the same values of the radii ratio d, which are now closely packed (cf. Fig.3.4.)
These closely-packed peaks indicate, on the one hand, that an enhancement in the
launching ability of the antenna occurs, for high values of the wavenumber xb, at
63
1 1.2 1.4 1.6 1.8 2
d10−4
10−2
1
102
104
106
Singular
Values
10−4
10−2
1
102
104
106
l=5
l=4
l=3
l=2
l=1
Figure 3.4: Logarithmic plot of the normalized electric singular valuesh(1)l
i2for
a λ/4 antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not tocontain any material (i.e., a = 1 = µa) and the surrounding shell is assumed to bea lossless DPS material with µb = 1 and xb = 500 m−1.
almost the same radii ratios for the electromagnetic multipolar modes with a pile up
of the resonant peaks at particular values of the radii ratio that is less pronounced for
the magnetic modes. On the other hand, this also indicates that the enhancement
in the launching ability of the antenna occurs for the two types of fields, i.e., electric
and magnetic, at roughly the same values of the radii ratio. Moreover, a closer
examination of the plots in Fig.3.4 and its magnetic counterpart (figure not shown)
reveals that for the electric dipolar mode and its magnetic counterpart, i.e., the
magnetic dipolar mode, in particular the local enhancement peaks appear now at
almost the same values of the radii ratio d ' 1.2, 1.4, 1.6, etc.In Fig.3.5 we plot the normalized electric singular values
h(1)l
i2for a λ/400
antenna versus the radii ratio d. The surrounding shell is assumed to be a lossless
non-magnetic DPS medium with xb = 150 m−1 (i.e., µb = 1 and b = 36.) Fig.3.5
clearly shows that for an electrically small antenna the resonant peaks disappear
over the same range of d values that had been considered for a quarter-wavelength
antenna and that had in several enhancement peaks present in it in that case.
Furthermore, the numerical simulations show that increasing the wavenumber of
64
1 1.2 1.4 1.6 1.8 2
d10−4
10−2
1
102
Singular
Values
10−4
10−2
1
102
l=5
l=4
l=3
l=2
l=1
Figure 3.5: Logarithmic plot of the normalized singular valuesh(1)l
i2for a λ/400
antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not to containany material (i.e., a = 1 = µa) and the surrounding shell is assumed to be a losslessDPS material with µb = 1 and xb = 150 m−1.
the shell medium not only does not restore the peaks but may make things even
worse in terms of the launching ability of the antenna with respect to the free-
space case. This is in perfect agreement with the fact that the actual total physical
dimensions of a resonating cavity made up of ordinary material, in particular a+ b
in this case, is the determining factor when it comes to which modes are supportable
by the cavity, and not just the radii ratio.
65
1 1.1 1.2 1.3 1.4
d10−2
1
102
104
106
108
1010
Singular
Values
10−2
1
102
104
106
108
1010
l=5
l=4
l=3
l=2
l=1
Figure 3.6: Logarithmic plot of the normalized singular valuesh(1)l
i2for a λ/4
antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not to containany material (i.e., a = 1 = µa) and the surrounding shell is assumed to be a losslessDNG material with µb = −1 and xb = −50 m−1.
Vacuum—Core—DNG—Shell System
The next lossless system we wish to investigate is a vacuum-core-DNG-shell system,
i.e., a = 1 = µa, and b < 0, µb < 0 (in fact in all what follows the relative electric
permittivity of the surrounding DNG shells is b = −4 and its relative magneticpermeability is µb = −1.) The driving frequency is as before set to f = 3.75 GHz.
In Fig.3.6 we plot the electric singular valuesh(1)l
i2versus the radii ratio d, for
a quarter-wavelength antenna of maximum length 2a = 2 cm; in Fig.3.7 we ploth(1)l
i2versus d, for a λ/40 antenna of maximum length 2a = 0.2 cm; and in Fig.3.8
we ploth(1)l
i2versus d, for a λ/400 antenna of maximum length 2a = 0.02 cm.
The simulations show that (cf. Figs.3.6-3.8) as the ratio of the length of the
antenna to the wavelength of the radiation in the vacuum, viz., 2a/λ, decreases the
resonant peaks, which correspond to a local enhancement in the launching ability
of the antenna for different modes, appear at certain fixed values of the radii ratio
d. This indicates that for a small enough k0a the enhancement for all the modes
appears at certain specified values of d regardless of the total physical dimensions
66
1 1.1 1.2 1.3 1.4
d10−2
1
102
104
106
108
1010
Singular
Values
10−2
1
102
104
106
108
1010
l=5
l=4
l=3
l=2
l=1
Figure 3.7: Logarithmic plot of the normalized electric singular valuesh(1)l
i2for a
λ/40 antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not tocontain any material (i.e., a = 1 = µa) and the surrounding shell is assumed to bea lossless DNG material with µb = −1 and xb = −50 m−1.
of the antenna. This is in total agreement with the reported subwavelength res-
onator concept [132, 133, 40] where the determining parameter for the existence of
a natural mode (polariton), and thus the occurrence of a local enhancement in the
launching ability of the antenna in this case, turns out to be the ratio of the two
radii rather than the total physical size of the antenna itself as would be the case
in the presence of only ordinary media. This clearly shows that encompassing a
subwavelength antenna in a judiciously chosen DNG metamaterial shell makes it
possible to distribute the resources of the antenna in a fashion that is as efficient as
that made possible only through the use of a much larger volume in free space.
A natural continuation to our previous investigation of the radiation efficiency
of an electrically small antenna embedded in a metamaterial substrate is displayed
in Fig.3.9. In this figure the normalized singular valuesh(1)l
i2for a λ/400 antenna
have been plotted versus the radii ratio d. The shell circumscribing the antenna
is assumed to be an ENG material with µb = 1 and xb = i50 m−1 (i.e., b =
−4). A comparison of Fig.3.9 and Fig.3.8 shows that these figures are in fact the
same though they describe two totally different systems. This clearly demonstrates
67
1 1.1 1.2 1.3 1.4
d10−2
1
102
104
106
108
1010
Singular
Values
10−2
1
102
104
106
108
1010
l=5
l=4
l=3
l=2
l=1
Figure 3.8: Logarithmic plot of the normalized electric singular valuesh(1)l
i2for a
λ/400 antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not tocontain any material (i.e., a = 1 = µa) and the surrounding shell is assumed to bea lossless DNG material with µb = −1 and xb = −50 m−1.
the fact that one can attain the same level of performance achieved through the
utilization of a DNG shell by using an SNG (in this case an ENG) shell. This, too,
is in total agreement with the results reported in the literature [8, 40, 132, 133, 41]
which stipulate that the use of DNG media is not really necessary in order to achieve
high performance levels and that similar performance levels could be achieved by
pairing two materials that possess oppositely signed values of at least one of the
constitutive parameter (In our case we had on one hand a DPS medium, i.e., the
vacuum core, and on the other hand the ENG shell such that µb = 1 = µa while
b = −4 = −4 a). The problem of pairing other types of substrates, such as an
MNG core and an ENG shell, has also been considered and the obtained results are
consistent with the published literature. Attaining high performance levels, such
as high radiation enhancement, through the utilization of an ENG medium is an
interesting possibility since such media exist in nature (plasmonic materials such as
silver, etc.)
68
1 1.1 1.2 1.3 1.4
d10−2
1
102
104
106
108
1010
Singular
Values
10−2
1
102
104
106
108
1010
l=5
l=4
l=3
l=2
l=1
Figure 3.9: Logarithmic plot of the normalized singular valuesh(1)l
i2for a λ/400
antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not to containany material (i.e., a = 1 = µa) and the surrounding shell is assumed to be an ENGmaterial with µb = 1 and xb = i50 m−1.
3.3.2 Lossy Substrates
The case of lossy substrates is illustrated in Fig.3.10. In this figure the normalized
electric singular valuesh(1)l
i2have been plotted versus the radii ratio d for a λ/400
antenna embedded in a vacuum-core-DNG-shell system. The surrounding DNG
shell is assumed to have a magnetic permeability µb = −1 and Re[xb] = −150m−1 (i.e., Re [ b] = −36.) The investigated cases are: 1) lossless case (loss tangentIm [ b] /Re [ b] = 0), 2) DNG shell with loss tangent Im [ b] /Re [ b] ' 1/60, and 3)DNG shell with loss tangent Im [ b] /Re [ b] ' 1/20. Fig.3.10 clearly shows that theinclusion of losses simply reduces the heights of the peaks but does not make the
peaks disappear. Also the decrease in the height of the resonant peaks becomes
larger relative to the lossless cases as the loss tangent of the shell increases. These
findings are not surprising and are in agreement with the results reported in the
literature [40, 132, 133, 41].
69
1 1.02 1.04 1.06 1.08 1.1
d10−2
1
102
104
106
108
Singular
Values
10−2
1
102
104
106
108
l=5
l=4
l=3
l=2
l=1
Figure 3.10: Logarithmic plot of the normalized electric singular valuesh(1)l
i2for
a λ/400 antenna versus the radii ratio d ≡ b/a. The inner sphere is assumed not tocontain any material (i.e., a = 1 = µa) and the surrounding shell is assumed to bea lossless DNG material (black curves), a lossy DNG shell with a loss tangent set to1/20 (blue curves), and a lossy DNG material with a loss tangent set to 1/60 (redcurves). In all three cases µb = −1, Re[xb] = −150 m−1.
Further Look at the Electric Dipole Case
We now initiate an investigation focused on the electric dipole case, i.e., in this case
j = 1 = l . We define the electric dipole antenna gain as8
G ≡ E(j=1=l)ME (free-space case)
E(j=1=l)ME
=h(1)1
i2. (3.10)
This quantity is plotted next versus the radii ratio d for some representative systems
and some selected values of the shell wavenumber xb. The systems considered here
are electric dipoles of different physical sizes embedded in vacuum—core-DNG-shell
systems. As explained above this means that in all the cases the inner sphere is
assumed not to contain any material, i.e., a = 1 = µa, while the outer shell is made
8Cf. Eq.(2.80).
70
1 1.2 1.4 1.6 1.8 2
d
10−2
1
102
Gain
10−2
1
102
xb=−500
xb=−150
xb=−100
xb=−50
Figure 3.11: Logarithmic plot of the gain G for a λ/4-electric-dipole antenna versusthe radii ratio d ≡ b/a. The surrounding shell is assumed to be a DNG materialwith µb = −1.
up of a DNG material with µb = −1. The driving frequency f is still set to 3.75
GHz.
In Figs.3.11-3.13 the gain G has been plotted versus the radii ratio d for a λ/4-
electric-dipole antenna (a = 1 cm), a λ/40-electric-dipole antenna (a = 0.1 cm), and
a λ/400-electric-dipole antenna (a = 0.01 cm). The aim is to study the effect of the
physical size a on the performance of the antenna. From Figs.3.11-3.13 we notice
that there is something that is counterintuitive here. It appears that as the physical
size of the dipole antenna becomes smaller the antenna’s ability to optimize the
utilization of its resources to radiate the dipolar field efficiently increases. This is
counterintuitive because what one would expect is that as the volume encompassing
the antenna decreases it becomes more difficult to distribute the resources of the
antenna so as to allow the antenna to radiate efficiently [51, 52, 17, 137, 112]. The
explanation of this seemingly counterintuitive situation lies in the physical interpre-
tation of the resonant peaks. As established above the resonant peaks correspond
to the presence of polaritons. These polaritons have a certain dispersion relation,
viz., Eq.(3.8) for the electric modes and Eq.(3.9) for the magnetic modes. These dis-
persion relations, or resonance conditions, establish a certain relationship between
71
1 1.2 1.4 1.6 1.8 2
d10−6
10−4
10−2
1
102
104
106
108
1010
Gain
10−6
10−4
10−2
1
102
104
106
108
1010
xb=−500
xb=−150
xb=−100
xb=−50
Figure 3.12: Logarithmic plot of the gainG for a λ/40-electric-dipole antenna versusthe radii ratio d ≡ b/a. The surrounding shell is assumed to be a DNG materialwith µb = −1.
the different parameters relevant to the problem. When all the parameters are fixed
except for the physical size of the antenna, as is the situation in this case, one should
be able, at least numerically, to solve for the optimum value of the physical size that
would satisfy the resonance condition. This optimal value of the physical size of the
antenna is what we are dealing with in this case. But if this is true then values on
both sides of this optimal size should cause a reduction in the ability of the antenna
to radiate the dipolar field which is no the case. The simulations show that as the
physical size of the antenna is reduced further the resonant peaks remain at the
same location. This objection may be explained away by invoking the concept of
subwavelength resonator [132, 133, 40]. Indeed, if the optimal value of the physical
size turns out to satisfy the subwavelength resonator conditions [132, 133, 40], that
is, if the size of the core-shell system turns out to be smaller than the wavelength
in all three regions then further reducing the physical size of the antenna will not
affect the radiation performance of the antenna, as discussed above.
72
1 1.2 1.4 1.6 1.8 2
d10−6
10−4
10−2
1
102
104
106
108
1010
Gain
10−6
10−4
10−2
1
102
104
106
108
1010
xb=−500
xb=−150
xb=−100
xb=−50
Figure 3.13: Logarithmic plot of the gain G for a λ/400-electric-dipole antennaversus the radii ratio d ≡ b/a. The surrounding shell is assumed to be a DNGmaterial with µb = −1.
3.4 Conclusion
To conclude, we have investigated, both analytically and numerically, the effects
that the presence of metamaterials would have on the performance of a general
antenna embedded in a generally lossy system of two nested spheres (core-shell
system) in terms of the efficiency with which the available resources of the antenna
could be distributed within a prescribed volume so as to generate a given radiated
field. The derived developments constitute a fundamental inverse-source-theoretic
framework for analysis and design of different substrate structures. This framework
also complements in analytical and computational tools and insight the pioneering
work by some of the leading authors in this area.
The adoption of the inverse-source-theoretic approach is aimed at enabling in-
trinsic, i.e., non-antenna-specific, and fair characterization of different substrate con-
figurations by comparing optimal radiation in either configuration (i.e., the “best”
in each one). This characterization is governed by a formally tractable source-energy
cost function that is physically motivated by ohmic loss control. Via analytical and
numerical examples we have explained and illustrated important enhancements due
73
to the presence of metamaterials in the context of the two-nested-spheres configu-
ration, in particular for media with oppositely signed constitutive parameters.
74
Chapter 4
Comparative Study ofRadiation Enhancement due toMetamaterials
In this chapter we present a comparative (numerical) study of radiation enhancement
due to metamaterial substrates. This study, published in [138], complements our
investigations of the homogeneous substrate case presented in Chapter 21 and the
core-shell system case presented in Chapter 3.2 As in Chapter 3 we consider the
core-shell system but contrary to the study carried out in that chapter where we
studied only the effects that adding a shell would have on the performance of an
existing antenna, we shall explore in this chapter the effects that different definitions
of the reference antenna would have on enhancement level estimates.
We shall limit ourselves to a few illustrative examples. We focus on the case
when the core does not contain any material, i.e., a = 1 = µa, and the shell is
a lossless double-negative medium (DNG). In particular, we shall set b = −4 andµb = −1. Two types of antennas are investigated: A quarter-wavelength antenna
(i.e., 2a/λ = 1/4) and a full-wavelength antenna (i.e., 2a/λ = 1). The driving
frequency of the antennas is set to f = 10 GHz. This corresponds to a = 3.75
mm for the quarter-wavelength antenna and a = 1.5 cm for the full-wavelength
1See also [112] and [137].2See also [125].
75
antenna. These particular choices of the numerical values of f and a are, of course,
arbitrary. However, they lie well within the range of values used in the scientific
and engineering literatures [8, 40, 41, 132, 133].
4.1 Reference Antennas
In Chapter 2 we defined the normalized singular valuesh(j)l
i2(cf. Eq.(2.77)). In
this chapter we shall refer to them as enhancements. Three “reasonable” definitions
of the reference antenna are considered: (1) it is an antenna which resides within
a sphere of radius a, (2) it is an antenna which resides within a sphere of radius b
, or, (3) it is an antenna which resides within a sphere of radius bmax, where bmax
is the outer radius of the shell that maximizes the singular valuehσ(1)l
i2. These
three classes of reference antennas will subsequently be referred to as RA1, RA2,
and RA3, respectively. The idea behind the first definition is a simple answer to the
question of how the addition of a metamaterial shell would affect the performance
of an existing antenna radiating in free space. This definition proved to be adequate
when the antenna substrate was merely a homogeneous sphere of given radius.3 The
second definition is motivated by the realization that adding a shell to an existing
antenna creates, in fact, a new antenna with new dimensions. As for the third
definition the idea behind it is to compare optimal radiation in either configuration,
i.e., to compare different antennas when they operate at their best [8, 40, 41].
4.2 Numerical Study
Because of the noted similarity between the behavior of the electric enhancements
and the magnetic enhancements we concentrate our attention in what follows on the
study of the electric enhancements.4
In Figs.4.1-4.3 we plot the enhancementsh(1)l
i2for a quarter-wavelength an-
tenna with radius a = 3.75 mm versus the radii ratio b/a. In Fig.4.1 the reference
antenna is assumed to be an antenna radiating in free space and circumscribed by
a sphere of radius a; in Fig.4.2 the reference antenna is assumed to be an antenna
3See Section 2.3.4See Sections 2.3 and 3.3.
76
1 1.2 1.4 1.6 1.8 2
bêa10−2
1
102
104
106
108
1010
Enhancement
10−2
1
102
104
106
108
1010
l=5
l=4
l=3
l=2
l=1
Figure 4.1: Logarithmic plot of the enhancementh(1)l
i2for a quarter-wavelength
antenna with radius a = 3.75 mm versus the radii ratio b/a. The shell is a losslessDNG material with b = −4 and µb = −1. The reference antenna is assumed to beRA1 antenna.
radiating in free space and circumscribed by a sphere of radius b; in Fig.4.3 the refer-
ence antenna is assumed to be an antenna radiating in free-space and circumscribed
by a sphere of radius bmax, where bmax is the outer radius of the shell that maximizes
the singular valueshσ(1)l
i2. The maxima occur at bmax/a = 1.40, 1.29, 1.17, 1.13, 1.1,
for l = 1, 2, 3, 4, 5, respectively. In Figs.4.4-4.6 we show plots similar to those of the
quarter-wavelength antenna case. The outer radii ratios that maximize the singular
valueshσ(1)l
i2are in this case bmax/a = 1.35, 1.16, 1.16, 1.18, 1.13, for l = 1, 2, 3, 4, 5.
Reference antennas that belong to RA2, i.e., reference antennas for the radius of
the circumscribing volume is equal to b, tend to shift the positions of the resonant
peaks. For instance, the first electric dipole peak appears for the singular valuehσ(1)1
i2at bmax/a = 1.40, as mentioned earlier, but when an RA2 antenna is adopted
the position of the peak shifts slightly: it appears at b/a = 1.32 forh(1)1
i2. However,
no relative shifts in the positions of the peaks are produced by the adoption of RA1
or RA3 antennas. This is due to the fact that with these antennas the enhancements
77
1 1.2 1.4 1.6 1.8 2
bêa10−2
1
102
104
106
108
1010
Enhancement
10−2
1
102
104
106
108
1010
l=5
l=4
l=3
l=2
l=1
Figure 4.2: Logarithmic plot of the enhancementh(1)l
i2for a quarter-wavelength
antenna with radius a = 3.75 mm versus the radii ratio b/a. The shell is a losslessDNG material with b = −4 and µb = −1. The reference antenna is assumed to bean RA2 antenna.
are nothing but the singular valueshσ(1)l
i2scaled by a constant.
Adopting an RA1 reference antenna, i.e., a reference antenna for which the cir-
cumscribing sphere has radius a, yields relatively higher estimates of the performance
levels with respect to the performance levels obtained by means of the remaining
two reference antenna classes, cf. Figs.4.1-4.6. This is quite understandable because
the size of an RA1 reference antenna is by definition smaller that the sizes of the
corresponding RA2 and RA3 antennas.
The simulations show also that the discrepancies between the estimates yielded
by the three reference antenna classes are not significant in the immediate vicinity of
the resonant peak which corresponds to bmax: the ratios between any two estimates,
in particular those obtained by means of RA1 and RA3 antennas on one hand and
RA2 antennas on the other hand, are less than one order of magnitude in favor of
the estimates obtained by means of RA1 and RA3.5 It is only when one attempts
to compare estimates yielded by the three reference antenna classes far away from
5Note that RA2 and RA3 antennas yield identical enhancements at b/a = bmax/a.
78
1 1.2 1.4 1.6 1.8 2
bêa10−2
1
102
104
106
108
1010
Enhancement
10−2
1
102
104
106
108
1010
l=5
l=4
l=3
l=2
l=1
Figure 4.3: Logarithmic plot of the enhancementh(1)l
i2for a quarter-wavelength
antenna with radius a = 3.75 mm versus the radii ratio b/a. The shell is a losslessDNG material with b = −4 and µb = −1. The reference antenna is assumed to bean RA3 antenna.
this resonant peak that the discrepancies become significant (the ratios between
the estimates obtained by means of the RA2 and RA3 antennas, for instance, are
about one to two orders of magnitude larger, in favor of RA3 estimates, for radii
ratios b/a > 1.8). Note also that the discrepancies are more pronounced for higher
multipole modes (cf. Figs.4.1-4.6). Thus for RA3 reference antennas, i.e., for those
which have a circumscribing volume of radius bmax , to yield a fair comparison they
have to be used locally. In other words it is not suitable to use a value of bmax which
corresponds to a maximum enhancement that appears for a specific range of shell
thickness values as an absolute reference for enhancement level characterization.
The discrepancies would be even larger if bmax corresponded to an infimum of the
set of shell thickness values at which resonant peaks appear in the singular values
spectrum.
79
1 1.2 1.4 1.6 1.8 2
bêa10−2
1
102
104
Enhancement
10−2
1
102
104
l=5
l=4
l=3
l=2
l=1
Figure 4.4: Logarithmic plot of the enhancementh(1)l
i2for a full—wavelength an-
tenna with radius a = 1.5 cm versus the radii ratio b/a. The shell is a lossless DNGmaterial with b = −4 and µb = −1. The reference antenna is assumed to be anRA1 antenna.
1 1.2 1.4 1.6 1.8 2
bêa10−2
1
102
104
Enhancement
10−2
1
102
104
l=5
l=4
l=3
l=2
l=1
Figure 4.5: Logarithmic plot of the enhancementh(1)l
i2for a full-wavelength an-
tenna with radius a = 1.5 cm versus the radii ratio b/a. The shell is a lossless DNGmaterial with b = −4 and µb = −1. The reference antenna is assumed to be anRA2 antenna.
80
1 1.2 1.4 1.6 1.8 2
bêa10−2
1
102
104
Enhancement
10−2
1
102
104
l=5
l=4
l=3
l=2
l=1
Figure 4.6: Logarithmic plot of the enhancementh(1)l
i2for a full-wavelength an-
tenna with radius a = 1.5 cm versus the radii ratio b/a. The shell is a lossless DNGmaterial with b = −4 and µb = −1. The reference antenna is assumed to be anRA3 antenna.
4.3 Conclusion
These observations are true for all multipolar modes and for small (quarter-wavelength)
and large (full-wavelength) antennas alike. (In the simulations we also investigated
5λ-antennas but the results are not shown here.) They lead us to the conclusion that
RA2 antennas, i.e., the antennas radiating in free space and having a circumscribing
volume of radius b, are more convenient as reference antennas, especially when one
is interested in performance characterization over a wide range of shell thicknesses,
though the adoption of an RA2 reference antenna may result in a slight shift of the
resonant peaks position.
81
Chapter 5
The Casimir Effect in thePresence of Metamaterials
In this chapter we investigate the Casimir effect in the presence of electromagnetic
metamaterials. We consider in particular the planar geometry sketched in Fig.5.1:
we consider two slabs of thicknesses a and b and constitutive parameters a, µa and b, µb, respectively, sandwiched between two half-spaces of constitutive parame-ters 1, µ1 and 2, µ2 . The four media are all assumed to be isotropic, dispersivemedia (although dissipation shall be assumed to be negligible in this case)1. In par-
ticular, this means that all the constitutive parameters are in general real functions
of the frequency.
It is true that several authors have considered the multilayered planar geometry
with and without metamaterials [104, 105, 106, 107, 94, 101, 108, 109, 102, 110, 111].
However, to our knowledge, all the published studies concentrate on the case of two
stacks of materials separated by a single medium. The concentration on such systems
is, of course, justified by the need for a better understanding of the multilayered
system which is ubiquitous in experimental research on the Casimir effect. In effect,
the mirrors used in the measurements of the Casimir force are practically always
made up of substrate media coated with one, two, or possibly even more layers of
other materials.
After deriving the formula that gives the Casimir force between the two half-
1See Appendix G for more detail.
82
2,µ2
a,µaa
b,µbb
1,µ1
z
x
Figure 5.1: Planar geometry under consideration. Two slabs of thicknesses a andb and relative constitutive parameters a, µa and b, µb, respectively, are sand-wiched between two half-spaces with relative constitutive parameters 1, µ1 and 2, µ2 .
spaces (see Fig.5.1), we validate this formula against some well-known results then
use it to derive the force between two perfectly conducting plates separated by two
slabs of material. Subsequently, a numerical study is performed on the Casimir
force in this case. As we mentioned in the Introduction, in choosing to focus on this
system we are particularly inspired by the subwavelength cavity resonators described
in [7]. The idea is to examine the possibility that, for similar cavities, the peculiar
properties of metamaterials would give rise to some interesting phenomena.
5.1 General Formulation
5.1.1 Eigenfrequencies
By definition, the vacuum energy is given by [83]
U ≡Xψ
1
2~ωψ, (5.1)
83
where ~ is Planck’s constant and ψ stands for all the parameters upon which the
modes depend. To calculate this energy we first have to determine the allowed
electromagnetic vacuum modes for the system under consideration. To do this we
shall adopt the semiclassical surface modes approach initiated in [140]. These modes
correspond to waves propagating parallel to the interfaces and form a complete set
of solutions [141]. The calculation given below is a straightforward generalization of
[140]. Here, we shall follow closely some well-known presentations given, in partic-
ular, in [83, 85].
After imposing the relevant boundary conditions at the interfaces between the
slabs and requiring that the solutions to Maxwell’s equations be nontrivial, one can
show that the allowed wave frequencies in the four-region system sketched in Fig.5.1
are the ω solutions to the dispersion relations2
g (ω) =(K1 a +Ka 1) (K2 b +Kb 2)
(K1 a −Ka 1) (K2 b −Kb 2)e2(Kaa+Kbb)
+(Kb a −Ka b) (K1 a +Ka 1)
(Kb a +Ka b) (K1 a −Ka 1)e2Kaa
+(Kb a −Ka b) (Kb 2 +K2 b)
(Kb a +Ka b) (Kb 2 −K2 b)e2Kbb − 1
= 0 (5.2)
gµ (ω) =(K1µa +Kaµ1) (K2µb +Kbµ2)
(K1µa −Kaµ1) (K2µb −Kbµ2)e2(Kaa+Kbb)
+(Kbµa −Kaµb) (K1µa +Kaµ1)
(Kbµa +Kaµb) (K1µa −Kaµ1)e2Kaa
+(Kbµa −Kaµb) (Kbµ2 +K2µb)
(Kbµa +Kaµb) (Kbµ2 −K2µb)e2Kbb − 1
= 0, (5.3)
2See Appendix F for more detail.
84
wherein
K2j ≡ k2x + k2y − j (ω)µj (ω)
ω2
c2
≡ κ2 − j (ω)µj (ω)ω2
c2, j = 1, 2, a, b. (5.4)
These dispersion relations, (5.2) and (5.3), correspond to two different polarizations
of the electric field. On the one hand, we have the waves that obey the dispersion
relation (5.2). These are the waves for which the polarization lies in the plane formed
by κ ≡ k⊥ and the z-axis (i.e., they have ey = 0). We shall call these waves “ -type”modes.3 On the other hand, we have the waves that obey the dispersion relation
(5.3). These are the waves for which the polarization is perpendicular to the plane
formed by κ ≡ k⊥ and the z-axis (i.e., they have ez = 0). We shall call these waves“µ-type” modes.4 We also note that κ and ω are independent variables and unless
ω/κ ≤ c the waves are evanescent.
To simplify these expressions we define the following quantities
A ≡ (K1 a +Ka 1) (K2 b +Kb 2)
(K1 a −Ka 1) (K2 b −Kb 2), (5.5)
B ≡ (Kb a −Ka b) (K1 a +Ka 1)
(Kb a +Ka b) (K1 a −Ka 1), (5.6)
C ≡ (Kb a −Ka b) (Kb 2 +K2 b)
(Kb a +Ka b) (Kb 2 −K2 b), (5.7)
and
Aµ ≡ (K1µa +Kaµ1) (K2µb +Kbµ2)
(K1µa −Kaµ1) (K2µb −Kbµ2), (5.8)
Bµ ≡ (Kbµa −Kaµb) (K1µa +Kaµ1)
(Kbµa +Kaµb) (K1µa −Kaµ1), (5.9)
Cµ ≡ (Kbµa −Kaµb) (Kbµ2 +K2µb)
(Kbµa +Kaµb) (Kbµ2 −K2µb). (5.10)
3They are similar to TMz (or E) waveguide modes.4They are similar to TEz (or H) waveguide modes.
85
With these quantities (5.2) takes on the simple form
g (ω) = A e2(Kaa+Kbb) +B e2Kaa + C e2Kbb − 1 = 0 (5.11)
and similarly (5.3) takes on the form
gµ (ω) = Aµe2(Kaa+Kbb) +Bµe
2Kaa +Cµe2Kbb − 1 = 0. (5.12)
5.1.2 The Casimir Force
In view of the fact that two types of modes—“ -type” modes and “µ-type” modes—
exist for the system under consideration, Eq.(5.1) transforms into
U =Xψ
1
2~ω(ψ) +
Xψ
1
2~ω(ψ)µ (5.13)
where, now, ψ labels all the parameters on which the -type and µ-type modes de-
pend. After substituting the standard cavity quantum electrodynamics prescription
Xψ
→Z
L
2πdkx
ZL
2πdky
XN
=
µL
2π
¶2 Z ∞
0κdκ
Z π
−πdφXN
, (5.14)
in (5.13), the energy takes on the form
U =~L2
4π
Z ∞
0κdκ
XN
h~ω(N) (κ) + ~ω(N)µ (κ)
i, (5.15)
where N stands for the solutions of (5.2) or (5.3).
From (5.15) we see that the evaluation of the vacuum energy has reduced to a
sum over the zeros of the two functions g (ω) and gµ (ω). If these functions are
viewed as complex functions, it is clear from (5.2) and (5.3) that g (ω) and gµ (ω)
are analytic except for poles (i.e., they are meromorphic functions [142]). Thus,
to evaluate the right-hand side of (5.15) we recur to the theory of meromorphic
functions. We introduce a new variable ξ such that
ξ ≡ −iω. (5.16)
86
With this definition, Eq.(5.4) becomes
K2j = κ2 + j (iξ)µj (iξ)
ξ2
c2, j = 1, 2, a, b. (5.17)
We also define the two new functions
G (ξ) ≡ g (iξ) (5.18)
and
Gµ (ξ) ≡ gµ (iξ) . (5.19)
Consequently, the argument principle for meromorphic functions allows us to write5
U =~L2
8π2
Z ∞
0κdκ
Z ∞
−∞[lnG (ξ) + lnGµ (ξ)] dξ. (5.20)
It can be shown that writing the vacuum energy in this form is equivalent to renor-
malizing it [85].
The Casimir force per unit area between the two half-spaces separated by the
distance d ≡ a+ b is defined as
F ≡ − ∂
∂d
U
L2
= −12
µ∂
∂a+
∂
∂b
¶U
L2, (5.21)
where U is given by Eq.(5.20). Substituting U from (5.20) into (5.21) yields
F = − ~8π2
Z ∞
−∞
∙Ka +Kb +Kb
1−B e2Kaa
G (ξ)+Ka
1− C e2Kbb
G (ξ)
¸dξ
− ~8π2
Z ∞
−∞
∙Ka +Kb +Kb
1−Bµe2Kaa
Gµ (ξ)+Ka
1− Cµe2Kbb
Gµ (ξ)
¸dξ.
(5.22)
Note that the terms that do not contain G (ξ) or Gµ (ξ) are independent of the two
5See Appendix G.
87
half-spaces. They do not correspond to any force between them. Hence
F = − ~4π2
Z ∞
0κdκ
Z ∞
0
"Kb
¡1−B e2Kaa
¢+Ka
¡1− C e2Kbb
¢G (ξ)
#dξ
− ~4π2
Z ∞
0κdκ
Z ∞
0
"Kb
¡1−Bµe
2Kaa¢+Ka
¡1− Cµe
2Kbb¢
Gµ (ξ)
#dξ.
(5.23)
This is the general formula for the Casimir force between two half-spaces separated
by two slabs of material.
5.2 Particular Cases
We shall now show that (5.23) reduces in some special cases to results published in
the literature.
5.2.1 One Slab of Material Separating Two Half-Spaces
Our aim here is to compare our general formula (5.23) to the results published by
several authors on the case of two half-spaces separated by only one slab [98, 99, 92].
If only one slab is assumed to separate the two half-spaces one has B = 0 = C
and Bµ = 0 = Cµ. Consequently
F = − ~2π2
Z ∞
0κdκ
Z ∞
0
£G−1 (ξ) +G−1µ (ξ)
¤Kdξ. (5.24)
where K = Ka = Kb. Now let
κ2 ≡ aµaξ2
c2¡p2 − 1¢ , p ∈ [1,∞). (5.25)
This allows us to rewrite K2j as
K2j ≡ a (iξ)µa (iξ)
ξ2
c2s2j j = 1, 2, a, b, (5.26)
88
where
s2j ≡ p2 − 1 + jµj
aµa. (5.27)
It is straightforward to show that
F = − ~2π2c3
Z ∞
1p2dp
Z ∞
0
∙(s1 + p 1) (s2 + p 2)
(s1 − p 1) (s2 − p 2)e2ξ
√µpd/c − 1
¸−1( µ)3/2 ξ3dξ
− ~2π2c3
Z ∞
1p2dp
Z ∞
0
∙(s1µ+ pµ1) (s2µ+ pµ2)
(s1µ− pµ1) (s2µ− pµ2)e2ξ
√µpd/c − 1
¸−1( µ)3/2 ξ3dξ
(5.28)
where d ≡ 2a = 2b, ≡ a = b, µ ≡ µa = µb. If the slab separating the two media
is nonmagnetic, i.e., if µ = 1, this expression reduces to the result obtained within
the Dzyaloshinskii-Lifshitz-Pitaevskii theory [98, 99] regarded by many authors as
the standard theory of the Casimir effect in dielectric media.
5.2.2 Two Slabs of Material Separating Two Perfectly ConductingPlates
The result (5.23) can be used to derive a formula for the Casimir force between two
perfectly conducting plates separated by two media (see Fig.5.2). In the remainder
of this chapter we shall focus on the abovementioned system.
The case of two perfectly conducting plates is recovered by taking the limits
1 →∞ and 2 →∞ in (5.23). We obtain
F = − ~4π2c3
Z ∞
1p2dp
Z ∞
0
1− γ (p) eβ(p)ξ
e[α(p)+β(p)]ξ − γ (p)£eα(p)ξ − eβ(p)ξ
¤− 1ξ3 ( aµa)3/2 dξ
− ~4π2c3
Z ∞
1p sb (p) dp
Z ∞
0
1 + γ (p) eα(p)ξ
e[α(p)+β(p)]ξ − γ (p)£eα(p)ξ − eβ(p)ξ
¤− 1ξ3 ( aµa)3/2 dξ
− ~4π2c3
Z ∞
1p2dp
Z ∞
0
1− γµ (p) eβ(p)ξ
e[α(p)+β(p)]ξ + γµ (p)£eα(p)ξ − eβ(p)ξ
¤− 1ξ3 ( aµa)3/2 dξ
− ~4π2c3
Z ∞
1p sb (p) dp
Z ∞
0
1 + γµeα(p)ξ
e[α(p)+β(p)]ξ + γµ (p)£eα(p)ξ − eβ(p)ξ
¤− 1ξ3 ( aµa)3/2 dξ,
(5.29)
89
z
a,µaa
b,µbb
x
Figure 5.2: A rectangular cavity of dimensions L×L×d is constructed by putting twoperfectly conducting plates with dimensions L×L× τ (LÀ dÀ τ) in the vacuumparallel to one another and separated by a distance d. Two slabs of isotropic,homogeneous, lossless material are inserted between the two plates. The first slabhas dimensions L× L× a and relative constitutive parameters a, µa; the secondslab has dimensions L×L×b and relative constitutive parameters b, µb . (a+b =d).
wherein
α (p) ≡ 2ac
√aµap ≡ 2
a
cnap, (5.30)
β (p) ≡ 2b
c
√aµasb (p)
≡ 2b
c
√aµa
rp2 − 1 + bµb
aµa
≡ 2b
cna
sp2 − 1 +
µnbna
¶2, (5.31)
γ (p) ≡ sb (p) a − p b
sb (p) a + p b, (5.32)
and
γµ (p) ≡sb (p) µa − pµbsb (p) µa + pµb
. (5.33)
Thus, the Casimir force between two perfectly conducting plates separated by two
material slabs is given by (5.29).
90
5.2.3 The One-Slab Case and Casimir’s Classical Result
Two simple cases that can be studied analytically emerge from (5.29). They are the
cases for which either
γ (p) = 0 = γµ (p) (5.34)
or
α (p) = β (p) . (5.35)
In the first case, i.e., when γ (p) = 0 = γµ (p) is satisfied, it is straightforward to
show that (5.34) is equivalent to ⎧⎪⎨⎪⎩a = b
and
µa = µb,
(5.36)
that is, condition (5.34) corresponds to one slab of total thickness d ≡ a+ b sand-
wiched between two perfectly conducting plates. If n is the index of refraction of
the slab, one obtains in this case
F = − ~4π2c3
Z ∞
1p2dp
Z ∞
0ξ3 ( aµa)
3/2 4
e2√
aµaξd/c − 1dξ (5.37)
= − ~c16π2d4
Z ∞
0
1
n
x3
ex − 1dx. (5.38)
If the slab is nondispersive the formula reduces further to
F = − ~cπ2
240d41
n, (5.39)
which is the same result that one obtains by means of zeta-functional regularization,
for instance. If now a = 1 = b and µa = 1 = µb, that is, when n = 1 and vacuum
is what separates the two plates, this result, (5.39), reduces to Casimir’s classical
result. Note also that when n = −1 in (5.39), one obtains an expression for theforce identical to the one derived by Leonhardt and Philbin [102].
It is worth noting, at this point, that the case of a single slab separating the two
conducting plates cannot be recovered by setting the thickness of one of the slabs
91
to zero. This is due to the fact that the very formulation of the problem assumes
that ab 6= 0. It is also worth noting that the semiclassical nature of the formulationimposes a lower limit on the values of the slabs’ thicknesses: in order for the classical
response functions to be physically meaningful the slabs’ thicknesses must be much
larger than the Bohr radius, i.e., a (or b)À a0 ∼ 10−10 meter.The second special case, i.e., when α (p) = β (p), leads to an interesting new
result. Eq.(5.35) is equivalent to the two new conditions⎧⎪⎨⎪⎩√
a µa = −p | b| |µb|
and
a = b.
(5.40)
where we have taken a and µa to be positive, for definiteness. In other words, (5.40)
corresponds to the situation when the plates are separated by two slabs having the
same thickness and oppositely signed refraction indices√
a µa = −p | b| |µb| ≡ |n|.
Note that (5.40) necessarily implies that a = − b and µa = −µb. In this case too
one arrives at a very simple formula reminiscent of (5.39), namely
F = − ~cπ2
240d41
|n| (5.41)
In other words a DPS-DNG combination used such that nDPS/nDNG = −1 anda = b produces the same attractive Casimir force caused by the presence of one
slab of material having thickness d = 2a and index of refraction nDPS . This makes
sense as in the case of a DPS-DNG combination the ratio a/b is expected to play
an important role.
92
5.3 Numerical Study
Here we wish to investigate further the behavior of the Casimir force in the case
of two slabs sandwiched between two perfectly conducting plates. In view of our
inability to integrate analytically the right hand side of (5.29), we recur to numerical
simulations. The MATHEMATICA code used for this simulation was validated
against Casimir’s classical result and (5.39).6
We define the normalized (unitless) Casimir force
f ≡ F
(−~cπ2/240d4) ; (5.42)
it is the force normalized to that associated with Casimir’s original cavity (i.e., two
plates separated by vacuum), for a total separation d = a+ b. The normalized force
f , thus defined, is an estimate of the effect that the use of materials between the
two plates would have on the Casimir force. Note that, because of (5.42), negative
values in the plots of f indicate the presence of a repulsive force.
5.3.1 Cavities with DPS-DPS and DNG-DNG Slab Combinations
The first question that we would like to address is whether there is any advantage
in putting two slabs instead of one between the plates. In Fig.5.3 we plot the
normalized Casimir force f for a DPS-DPS slab combination and a DNG-DNG slab
combination as a function of the ratio of thicknesses r ≡ a/b. Fig.5.3 shows that
even in those cases the Casimir force changes its sign, clearly demonstrating that
there is a fundamental difference between one-media systems and two-media systems
even when the two media are both DPS or DNG.
5.3.2 Cavities with DPS-DNG Slab Combinations
Now, we turn to systems with DPS-DNG combinations. In Fig.5.4 we plot the
normalized Casimir force for a particular type of DPS-DNG slab combinations as a
6 In the code the usual procedure of converting an improper integral into a proper one byrescaling the variable of integration has been used.
93
0 1 2 3 4 5 6r
-20
-10
0
10
20
NormalizedCasimirForce
DNG−DNG
DPS−DPS
Figure 5.3: Plot of the normalized Casimir force f for DPS-DPS and DNG-DNG slabcombinations as a function of the ratio of thicknesses r ≡ a/b. The constitutive para-meters of the DPS-DPS combination are a, µa = 2, 1 and b, µb = 2.5, 1.5 .The constitutive parameters of the DNG—DNG combination are a, µa = −2,−1and b, µb = −2.5,−1.5 . For both combinations the thickness of the second slabis set to b = 0.001 mm.
function of the ratio of thicknesses r. For this type of combination we require that
b
a= −a
b= −r = µb
µa. (5.43)
This particular choice is considered because the corresponding normalized Casimir
force exhibits a simple profile in the plots. Fig.5.4 clearly demonstrates three impor-
tant features of the force: (1) the use of metamaterials in a DPS-DNG combination
does change the sign of the Casimir force and turns it from an attractive force to
a repulsive force, (2) the magnitude of the repulsive force could be made orders of
magnitude larger than the magnitude of the attractive Casimir force in the vacuum,
and (3) the Casimir force depends on the ratio of thicknesses r in an essential way.
In fact, the numerical simulations suggest that the Casimir force F has, in this case,
the generic profile
f ∼ R³ab
´⇒ F ∼ 1
(a+ b)4R³ab
´(5.44)
94
0 0.5 1 1.5 2 2.5 3r
-200
-180
-160
-140
-120
-100
-80
-60
NormalizedCasimirForce
b=10 mm
b=0.1 mm
b=0.001 mm
Figure 5.4: Plot of the normalized Casimir force f for a DPS-DNG slab combi-nation as a function of the ratio of thicknesses r ≡ a/b. For this combination wehave b/ a = −a/b = −r = µb/µa.The constitutive parameters of the DPS slab are a, µa = 2, 1 and those of the DNG slab are b, µb = −2r,−r . The thick-nesses of the DNG slabs were set to the arbitrary values 0.001 mm, 0.1 mm and10 mm. Because the curves coincide exactly with one another, a purely artificialshift of magnitude −10 has been introduced in the plot of the normalized force thatcorresponds to b = 0.1 mm and a similar artificial shift of magnitude −20 has beenintroduced in the plot of the normalized force that corresponds to b = 10 mm.
where R (·) is a function of the ratio of thicknesses only. These features have beenconfirmed for different DPS-DNG combinations. They appear to be characteristic
of DPS-DNG-based systems not just some peculiar behavior due to the particular
values of the constitutive parameters.
The numerical study also shows that the DPS-DNG combination exhibits some
other interesting features. Fixing the constitutive parameters for a certain combi-
nation and varying the individual thicknesses yields wildly oscillating profiles of the
normalized force with many separation values at which the force assumes a doublet
structure, as shown in Fig.5.5 and Fig.5.6. Then, when the total distance between
the two plates reaches a certain threshold, an “oscillation threshold,” the magnitude
of the normalized Casimir force starts decaying. However, depending on which slab
thickness is being varied the decay is either to f = 0+ or f = 0−. If the thicknessof the DPS slab is held fixed and the thickness of the DNG slab is varied the force
95
40 42 44 46 48 501êr
-1×108
-5×107
0
5×107
1×108
NormalizedCasimirForce
Figure 5.5: Plot of the normalized Casimir force f for a DPS-DNG slab combinationas a function of the ratio of thicknesses 1/r ≡ b/a. The constitutive parametersof the DPS slab are a, µa = 2, 1 and those of the DNG slab are b, µb =−2.5,−1.5 . The thickness of the DPS slab is set to a = 0.001 mm.
is repulsive, beyond the “oscillation threshold,” and its magnitude decays to 0−, ascan be seen in Fig.5.5. While if the thickness of the DNG slab is held fixed and the
thickness of the DPS slab is varied, instead, the force is attractive and its magnitude
decays to 0+, as can be seen in Fig.5.6.
The wildly oscillating profiles exhibited in Figs.5.5 and 5.6 are an indication
of instability in the corresponding systems. Visibly, the slightest variation in the
ratio of thicknesses r around an inflection point causes the systems to collapse. An
inspection of Figs.5.5 and 5.6 beyond the “oscillation threshold” also shows that
the smoother behavior of the force manifested there does not actually prevent the
collapse of the system.
5.3.3 Applications
The behavior of the Casimir force noted above is not the kind of behavior that would
permit the shielding of MEMS and NEMS components or quantum levitation. We
need situations characterized by the opposite behavior. Ideally, we would like to
have a system for which the Casimir force would vanish for a certain separation d0,
96
40 50 60 70 80r
-1×108
-5×107
0
5×107
1×108
NormalizedCasimirForce
Figure 5.6: Plot of the normalized Casimir force f for a DPS-DNG slab combinationas a function of the ratio of thicknesses r ≡ a/b. The constitutive parametersof the DPS slab are a, µa = 2, 1 and those of the DNG slab are b, µb =−2.5,−1.5 . The thickness of the DNG slab is set to a = 0.001 mm.
is attractive and increases for increasing separations, and is repulsive but increasing
for decreasing separations. Intuitively, we seek a situation in which the Casimir force
behaves somewhat like the elastic force of a loosely wound spring. Such a situation
is achievable, as illustrated in Fig.5.7. In this figure we focus our attention on the
profile of the force in the interval 37 < r . 40.6, for instance. For the most part,
the Casimir force is repulsive in this interval. But it exhibits an inflection point at
r ' 38.8, vanishes at r ' 40.3 then becomes an attractive force afterwards. This
exactly is the sought behavior. In effect, inside this interval of thicknesses ratio,
37 < r . 40.6, the force exhibits a behavior reminiscent of that of a loosely woundspring. A direct consequence of this is that the upper plate 7 would not only levitate
but also remain confined to a certain height.
It has been suggested that quantum levitation is achievable even in the presence
of only one DNG metamaterial slab. In particular, it has been speculated [102]
that a “perfect lens” of fixed thickness sandwiched between two conducting plates
would produce a repulsive Casimir force (whose magnitude is obtained by setting
n = −1 in (5.39)) and that this force would be able to levitate a metallic foil. Our7Cf. Fig.5.2.
97
30 35 40 45 50r
-1×108
-5×107
0
5×107
1×108
NormalizedCasimirForce
Figure 5.7: Plot of the normalized Casimir force f described in Fig.( 5.6) for differentvalues of r.
results suggest that this is an intricate issue. Because the perfect lens has a fixed
thickness, levitating the metallic foil would necessarily introduce a second slab of
DPS medium, say, vacuum, over the perfect lens which would then bring in regions
of instability similar to those discussed above. Quantum levitation, as described in
[102], entails a hidden assumption. It assumes that, as the foil levitates, no gap is
created between the foil and the perfect lens. But that, of course, is not the case.8
We also note that (5.39) suggests another situation in which one could obtain
large repulsive Casimir forces. Actually, the inverse dependence of the Casimir
force on the index of refraction in this case suggests the possibility of dramatically
increasing the magnitude of the repulsive Casimir force by sandwiching a metama-
terial with n ∼ 0− between two conducting plates. To our knowledge, this proposalhas not been suggested in the literature.
8See Fig.2 in [102]. There exist, however, one situation in which it would be theoreticallypossible to realize quantum levitation in the manner described in [102]. The space separating thefoil and the lower mirror would have to be continuously occupied by some kind of a "metamaterialfluid" having a refraction index n = −1. That, obviously, is not the case considered by the authors.
98
5.4 Conclusion
The numerical simulations described above establish the fact that, in the context of
the Casimir force between two perfectly conducting plates, major differences exist
between the situations that involve two-slab combinations separating the plates,
and the situations that involve only one slab. We found that (1) the presence of
two slabs greatly affects the behavior of the Casimir force causing it to turn into
a repulsive force, even when the response functions of the two slabs have the same
sign, (2) the force exhibits a profile determined in an essential way by the ratio of
the slabs’ thicknesses, (3) the force could be made orders of magnitude larger by
means of carefully selected response functions and slab separations, and (4) the force
exhibits a singular behavior at particular plate separations when the two slabs have
oppositely signed response functions. Evidently, clear advantages may result from
this behavior of the force depending on the application. Most notably, based on
these results, it appears that quantum levitation is indeed possible. It also appears
that, surprisingly, the energy of the vacuum could be used as a source for both the
actuation and shielding of MEMS and NEMS.
These are very interesting results. They could open up a whole new horizon
in technology. Yet, for all intents and purposes the system illustrated in Fig.5.2
is a highly idealized one. The force-decreasing skin-depth effects due to finite con-
ductivity are ignored; the force-increasing roughness effects of the metallic surfaces
are ignored; nonzero-temperature effects are ignored; the anisotropic nature of cur-
rent metamaterials is ignored; and, most importantly, the dispersive and dissipative
nature of metamaterials are also ignored.9 This, certainly, is the most question-
able assumption we have made as nondispersive or nondissipative metamaterials are
acausal (causality being epitomized by the Kramers-Kronig relations [115]). More-
over, the Casimir effect depends in a nontrivial way on the essentially dispersive
nature of metamaterials [110]. However, we believe that the simplifying assump-
tions made here do not diminish the insight that this study provides, as our primary
intention is to gain a rudimentary understanding of the class of systems proposed.
The effects of dispersion and dissipation shall be addressed elsewhere.
9As emphasized at the onset, the theory is valid for dispersive media (see Eq.(5.23)) and justi-fications exist for its extension to dissipative media as well [143].
99
Chapter 6
Future Directions
6.1 Extensions to the Inverse Source Problem in An-tenna Substrate Media
The theory developed in the first part of this dissertation holds for a given fre-
quency. I am interested in the development of a broadband inverse theory, which
can be derived as a stepped-frequency approach or as a theory directly in the time
domain. Extension of the published scalar formulation to the full-vector context and
dispersive embedding media is an important open problem of my interest. Another
future research avenue is the extension of the theory to prolate and oblate spheroidal
support regions so as to more tightly characterize dipole and planar antennas. Po-
tential future directions also include the development of a general multiport antenna
theory for small antennas formed by a number of independently addressable feeding
ports which may benefit, e.g., from nanotechnology. This approach is expected to
facilitate physical synthesis of near-optimal, wave-like current distributions in given
source volumes.
100
6.2 Extensions to the Inverse Scattering Problem inSubstrate Media
At the moment, I am also working on the scalar inverse scattering counterpart of the
abovementioned inverse source problem in metamaterial backgrounds. In particu-
lar, my goal is to investigate possible enhancements in the imaging resolution of an
object embedded in a metamaterial background. Imaging resolution enhancement
is expected to be possible in this case due to multiple scattering interactions of the
object with a helper substrate that acts as a near-field agent, i.e., a “re-transmitting
station,” that facilitates communication to the far field of evanescent field informa-
tion about the object. In fact, the achieving of super-resolution thanks to multiple
scattering, and the re-evaluation of the so-called “diffraction limit” in imaging, is an
area that has been receiving much attention in recent years, and is closely connected
to the developments in radiation and scattering enhancements due to metamaterials.
Once the current study of metamaterial-engendered imaging super-resolution is
finished, I am interested in expanding it to the full-vector inverse scattering problem
in metamaterial substrates including multiple scattering. In addition to this, I would
like to quantitatively characterize the “enhancement” in imaging in the presence
of noise via the fundamental Cramer-Rao bound. Another interesting open area
for further exploration is the utilization of metamaterials in optical imaging with
phaseless data, particularly in the imaging with far-field data.
6.3 Electrodynamics of Metamaterials
On account of the identified exciting potential applications of metamaterials, a bet-
ter understanding of their electrodynamics is in order. For instance, one major
problem that plagues the manufacturing of metamaterials is the ubiquitous lossy
nature of the meta-cells. I am very motivated to explore the effects that replacing
ordinary conductors as a base material for the meta-cells with superconducting ones
would have on the overall response of these materials. Another area that is still al-
most completely open and where I believe I can make a difference is the development
of a quantum electrodynamical theory in metamaterial backgrounds.
Because of the promising interplay between metamaterial research and nanotech-
101
nology, there is another important topic related to the study of the electrodynamics
of metamaterials that I find very interesting. That is nanoelectromagnetics. As a
matter of fact, new theoretical models supplemented by numerical studies are essen-
tial for the understanding of electromagnetic-wave propagation in media embedded
with nanostructures. At a more fundamental level, but no less important and rele-
vant, one may need to recur to quantum electrodynamics not just electromagnetics
to elucidate some of the physical properties of these media. This already is an ac-
tive line of research pursued in connection with the study of carbon nanotubes and
graphene sheets and I am interested in exploring it in the context of metamaterials.
I am planning to further investigate the Casimir effect in the presence of meta-
materials. The next obvious step would be to take into account the dispersive and
dissipative nature of metamaterials. The practical possibility of quantum levitation
suggested by the results of the second part of this dissertation is an exciting avenue
that I wish to pursue. I also would like to extend my study to other Casimir-type
effects such as the dynamical Casimir effect, where the boundary conditions are no
longer static but dynamic.
102
Appendix A
Definition and Properties ofVector Spherical Harmonics
The scalar spherical harmonics are defined as (cf., e.g., [117])
Yl,m(r) ≡ Yl,m(θ, ϕ)
≡s(2l + 1) (l −m)!
4π (l +m)!Pml (cos θ)e
imϕ, (A.1)
where the associated Legendre polynomials are given by the Rodriguez formula
Pml (x) =
(−1)m2ll!
¡1− x2
¢m/2µ
d
dx
¶l+m ¡x2 − 1¢l , (A.2)
l is a positive integer and |m| 6 l. Note that the Condon-Shortley phase factor
(−1)m has been absorbed in the definition of the Pml ’s.
The vector spherical harmonics are defined as [114]
Yl,m(r) ≡ Yl,m(θ, ϕ)
≡ LYl,m(θ, ϕ), (A.3)
103
L being the orbital angular momentum operator defined as
L ≡ −ir×∇
≡ −iÃ−bθsin θ
∂
∂ϕ+ bϕ ∂
∂θ
!. (A.4)
From the above definitions it can be shown that the scalar spherical harmonics
Yl,m(r) and the vector spherical harmonics Yl,m(r) satisfy the analytic continuation
properties
Y ∗l,m(r) = (−1)m Yl,−m(r), (A.5)
and
Y∗l,m(r) = (−1)m+1Yl,−m(r), (A.6)
along with the orthogonality conditions [114]ZYl,m(r)Y
∗l0,m0(r) dr = δll0δmm0 , (A.7)
ZYl,m(r) ·Y∗l0,m0(r) dr = l (l + 1) δll0δmm0 , (A.8)Z
r×Yl,m(r) · r×Y∗l0,m0(r) dr = l (l + 1) δll0δmm0 , (A.9)ZY∗l,m(r) · r×Yl0,m0(r) dr = 0, (A.10)
where the δxx0 ’s are the Kronecker deltas, and dr ≡ sin θdϕdθ.The following formulae [144] were useful in the derivation of some of the results
∇× [rφl(r)Yl,m(r)] =−irφl(r)Yl,m(r), (A.11)
∇× [φl(r)Yl,m(r)] = ril(l + 1)
rφl(r)Yl,m(r) +
1
r
d
dr[rφl(r)]r×Yl,m(r), (A.12)
∇× [φl(r)r×Yl,m(r)] = −1r
d
dr[rφl(r)]Yl,m(r), (A.13)
∇×∇× [φl(r)Yl,m(r)] = Yl,m(r)
∙l (l + 1)
r2−∇2
¸φl(r). (A.14)
104
A concise source on vector spherical harmonics and their properties is [145].
Note, though, that the notation and normalization adopted in this reference are
slightly different from the ones above.
105
Appendix B
Wavefunctions B(j)l,m forPiecewise-ConstantRadially-SymmetricBackgrounds
The aim of this appendix is to show that the multipole moments a(j)l,m are given by
Eq.(2.17) with the source-free wavefunctions B(j)l,m (r) given by Eqs.(2.19,2.20). A
straightforward way of arriving at these results would be to use the dyadic Green’s
function that governs the propagation of electromagnetic radiation between the
source-enclosing inner sphere and the surrounding vacuum. The spectral-domain
electromagnetic Green’s function linking the different layers of a spherically multi-
layered medium has been calculated by Li et al. [146].1 Afterwards, one calculates
the external electric field by means of Eq.(2) in [146]. Finally one uses Eqs.(2.15)
and (2.16), in this paper, to arrive at the desired results, i.e., Eqs.(2.17,2.19,2.20).
Another way of arriving at Eqs.(2.17,2.19,2.20) is to invoke Lorentz’s reciprocity
theorem and the concept of reaction (also called coupling). Here we adopt this
approach. One reason behind this choice is that this latter approach attests to the
1For the homogeneous sphere the Green function is given by Eqs.(25a) and (26b) along with therelevant definitions while for the three-region geometry of the core-shell system the Green functionis given in [146] by Eqs.(14), (29a)-(29d), and the relevant definitions.
106
visible similarity between the mathematics of the radiation problem at hand and its
scattering counterpart.
The reaction RE→J0 of a field E(r) produced by a source J(r) on another sourceJ0(r), is defined as the integral
RdrE(r) · J0(r). The reciprocity theorem can be
stated as follows (see, for instance, [58]): The reaction of the field E(r) produced
by a source J(r) on another source J0(r) is equal to the reaction of the field E0(r)
produced by the source J0(r) on the source J(r), i.e.,
RE→J0 = RE0→J. (B.1)
We focus, first, on the two-region (homogeneous sphere) case. The calculation
below outlines also the general approach that will be used in the case of the three-
region (core-shell) system. The details of that calculation are presented in Appendix
C.
To evaluate the field due to a current distribution J(r) that is embedded in the
piecewise-constant radially symmetric background of interest, we consider, without
loss of generality, the following two classes of canonical sources:hJ(1)l,m
i0(r) = δ(r −R)r×Yl,m(r) (B.2)
and hJ(2)l,m
i0(r) = δ(r −R)Yl,m(r), (B.3)
where in both expressions R > a represents the radius of the helper source cen-
tered around the origin. (Ultimately, the multipole moments a(j)l,m are expected to,
and in fact will, turn out to be independent of R.) The justification for the above
considerations relies on two results:
• The transverse component of an arbitrary vector field on the spherical sur-face of radius R > a centered about the origin is uniquely characterized by
its expansion in terms of the vector spherical harmonics Yl,m(r) and their
associated vector functions r×Yl,m(r), and
• the multipole moments characterizing any electric field outside the support ofthe emitting sources is uniquely determined by the tangential component of
107
the field on any such spherical surface. Indeed, it can easily be shown that if
R > a is the radius of a sphere centered about the origin, then
a(j)l,m =
⎧⎪⎪⎨⎪⎪⎩−i
l(l+1)k0h(+)l (k0R)
RY∗l,m(r) ·E(Rr)dr ; j = 1
1l(l+1)k0Vl(k0R)
Rr×Y∗l,m(r) ·E(Rr)dr ; j = 2.
(B.4)
The fieldhE(j)l,m
iincthat would be produced by the source
hJ(j)l,m
i0in Eq. (B.3)
if it were in free space (this will be the incident field in the following) is given byhE(j)l,m
iinc=
Zdr0G0(r, r
0) ·hJ(j)l,m
i0
¡r0¢, (B.5)
where G0(r, r0) is the multipole representation of the free-space electric dyadic Green
function [147], viz.,
G0(r, r0) =
∞Xl=1
lXm=−l
−ωµ0k0l (l + 1)
nk20 [jl(k0r<)Yl,m(r<)]
hh(+)l (k0r>)Y
∗l,m(r>)
i+ ∇× [jl(k0r<)Yl,m(r<)]∇×
hh(+)l (k0r>)Y
∗l,m(r>)
io+
i
ω 0rrδ
¡r− r0¢ . (B.6)
The < (>) subscript designates the smaller (larger) of r and r0.
For r < R the fieldhE(1)l,m
iincis found to be given by
hE(1)l,m
iinc(r) = τ l(k0, R)∇× [jl(k0r)Yl,m(r)], (B.7)
where we have defined
τ l(k0, R) ≡ −η0k0R2Vl (k0R) . (B.8)
Along analogous lines, the (incident) fieldhE(2)l,m
iincproduced by the source
hJ(1)l,m
i0
108
in Eq. (B.2) in free space is found, for r < R, to be given byhE(2)l,m
iinc(r) = ζ l(k0R)jl(k0r)Yl,m(r) (B.9)
where we have introduced
ζ l(k0R) ≡ −η0 (k0R)2 h(+)l (k0R). (B.10)
The obtainment of the above results requires the use of orthogonality properties
of the vector spherical harmonics Yl,m(r) and the associated vector functions r ×Yl,m(r).2
Introducing the index of refraction n =õs s/
õ0 0 =
õr r, the evaluation of
the total fieldshE(j)l,m
i0, j = 1, 2, associated with these sources for r < R is seen from
Eqs. (2.1), (2.2), (2.3)-(2.5) to correspond to the solution of the forward scattering
problem associated for r < R with the equation
£∇2 + k20Θ(r − a) + n2k20Θ(a− r)¤ hE(j)l,m
i0(r) = 0 (B.11)
upon excitation by the incident fieldshE(j)l,m
iinc.
The total (incident plus scattered) fieldhE(1)l,m
i0must be, due to considerations
of causality (in the scattered field) and well-behavedness of the interior field for
r < a, of the form
hE(1)l,m
i0(r) =
⎧⎪⎪⎨⎪⎪⎩∇×
hτ l(k0, R)jl(k0r)Yl,m(r) +D1h
(+)l (k0r)Yl,m(r)
i; r > a
A1∇× [jl(kr)Yl,m(r)]; r ≤ a,
(B.12)
where k = nk0 is the wavenumber of the field in the background material confined
within the source volume V and A1 and D1 are coefficients that are to be determined
by imposing continuity of the tangential components of the electric and magnetic
fields on the boundary ∂V ≡ r ∈ R3 : r = a. Analogously, the total field
2See Appendix A.
109
hE(2)l,m
i0must be of the form
hE(2)l,m
i0(r) =
⎧⎪⎪⎨⎪⎪⎩hζ l(k0R)jl(k0r) +D2h
(+)l (k0r)
i×Yl,m(r); r > a
A2jl(kr)Yl,m(r); r ≤ a,
(B.13)
where A2 and D2 are coefficients that need to be determined from the boundary
conditions on ∂V.
By imposing the continuity requirements on the boundary ∂V and using the
Wronskian relation for spherical Bessel functions [117], one obtains (for j = 1)
A1τ l(k0, R)
=i/ (k0a) (ka)
( r/µr)1/2 jl(ka)Vl(k0a)− h
(+)l (k0a)Ul(ka)
≡ F(1)l (k0a, ka, r, µr) , (B.14)
where Ul and Vl have already been defined in Eqs. (2.21) and (2.22). The coefficient
A2 associated with the fieldhE(2)l,m
i0can be obtained by an analogous procedure
which yields
A2ζ l(k0R)
=iµr/ (k0a) (ka)
(µr/ r)1/2 jl(ka)Vl(k0a)− h
(+)l (k0a)Ul(ka)
≡ F(2)l (k0a, ka, r, µr) . (B.15)
Along with Eqs. (B.12) and (B.13) the above results define the fieldshE(1)l,m
i0andh
E(2)l,m
i0.
By applying the reciprocity relation Eq. (B.1) to the preceding results (in par-
ticular, Eqs. (B.2), (B.3), (B.8), (B.10), (B.12), (B.13), (B.14), (B.15)), one finds
that the multipole moments a(j)l,m are indeed independent of R and given by Eq.
(2.17) with the source-free wavefunctions B(j)l,m(r) given by Eqs. (2.19)-(2.22). In
obtaining these results we have also recalled the multipole expansion for the electric
field E(r), i.e., (2.15) and (2.16), along with the orthogonality and analytic con-
tinuation properties of the vector spherical harmonics, cf. Appendix A., and the
analytic continuation property of the spherical Bessel functions of the first kind,
110
viz., j∗l (ka) = jl(k∗a). The derived results (2.19,2.20) reduce to the free-space result
Eq. (2.18) as r → 1 and µr → 1, as expected.
An alternative way of arriving at Eq. (2.17) and Eqs. (2.19)-(2.22) would be
to follow the same procedure as above but to use the far-field expressions for E(r)
and H(r) instead of the multipole expansion given by Eqs. (2.15) and (2.16). The
far-field expressions are obtained from Eqs. (2.15) and (2.16) by using the large-
argument approximation for the spherical Hankel function h(+)l . This yields the
far-field approximations (cf. also [51], Eqs. (4)-(6))
E(r r) ∼k0rÀ1
eik0r
rfe(r) =
η0eik0r
rr× fm(r) (B.16)
H(r r) ∼k0rÀ1
eik0r
η0rr× fe(r) = eik0r
rfm(r), (B.17)
where η0 =pµ0/ 0 is the free-space wave impedance. The two vector quantities
fe(r) and fm(r) are, respectively, the far electric field radiation pattern and the far
magnetic field radiation pattern. They are given as a function of the observation
direction r by
fe(r) =∞Xl=1
lXm=−l
(−i)lha(1)l,mr×Yl,m(r) + a
(2)l,mYl,m(r)
i(B.18)
and
fm(r) =1
η0
∞Xl=1
lXm=−l
(−i)lh−a(1)l,mYl,m(r) + a
(2)l,mr×Yl,m(r)
i. (B.19)
It follows from Eqs.(B.18,B.19) and the orthogonality of the vector spherical har-
monics Yl,m(r) and the associated vector functions r×Yl,m(r) that the multipole
moments a(j)l,m are uniquely defined by projections of the far-field radiation patterns
onto the orthogonal set of functions Yl,m(r) and r×Yl,m(r) (see, e.g., [51], Eq.(7)
and the associated discussion). Thus either the far fields or the multipole moments
uniquely define each other as well as the exterior field everywhere outside the source
volume V (via Eq.(2.15)).
In order to arrive at Eq.(2.17) and Eqs.(2.19,2.20) we follow the same procedure
as above but instead of the multipole expansion (2.15,2.16) we use Eqs.(B.16,B.17,B.18,B.19).
111
We then require that k0R À 1 and use the large-argument approximation for the
spherical Hankel function, in particular, h(+)l (k0R) ∼ (−i)l+1eik0R/(k0R) (see, forinstance, [117]). Afterwards one lets k0R→∞. Whether one uses this approach orthe previous one, the final results are the same.
112
Appendix C
Wavefunctions B(j)l,m for aSystem of Two Nested Spheres
To determine the source-free fields B(j)l,m for the spherical core-shell system we follow
the same procedure used in appendix B. However, we note that the radius R of the
helper source centered around the origin is, now, R > b.
The total fieldhE(1)l,m
i0must be, due to considerations of causality in the scattered
field and well-behavedness of the interior field for r < b, of the form
hE(1)l,m
i0(r) =
⎧⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎩
∇×nh
τ l(k0, R)jl(k0r) +D1h(+)l (k0r)
iYl,m(r)
o; r > b
∇×nh
B1jl(kbr) + C1h(+)l (kbr)
iYl,m(r)
o; a < r ≤ b
A1∇× [jl(kar)Yl,m(r)]; r ≤ a,
(C.1)
where A1, B1, C1 and D1 are coefficients that are to be determined by imposing con-
tinuity of the tangential components of the electric and magnetic fields on the inner
and outer boundaries of the spherical shell. Analogously, the total fieldhE(2)l,m
i0must
113
be of the form
hE(2)l,m
i0(r) =
⎧⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎩
hζ l(k0R)jl(k0r) +D2h
(+)l (k0r)
iYl,m(r); r > b
hB2jl(kbr) + C2h
(+)l (kbr)
iYl,m(r); a < r ≤ b
A2jl(kar)Yl,m(r); r ≤ a,
(C.2)
where A2, B2, C2 and D2 are coefficients that need to be determined from the bound-
ary conditions.
Imposing the abovementioned continuity requirements on the inner and outer
boundaries of the spherical shell, i.e., on the spheres of radii a and b, respectively,
yields two systems (one for j = 1 and one for j = 2) of four equations each linear in
the unknown coefficients.
Upon solving the two linear systems of equations and using the Wronskian re-
lation for spherical Bessel functions, specifically jl(x)h0l(+)(x) − j0l(x)h
(+)l (x) = i
x2
[117], we find that for j = 1
A1τ l(k0, R)
=b
∆1k0kba2b2
≡ F(1)l , (C.3)
where ∆1 is the determinant given by Eq.(3.5). Similarly for j = 2 one obtains
A2ζ l(k0R)
=1
µb∆2k0kba2b2
≡ F(2)l , (C.4)
where∆2 is the determinant given by (3.6). The remaining constantsB1, C1,D1, B2, C2,
and D2 are straightforwardly obtained; here they are omitted because of their irrel-
evance to the rest of the problem. Along with Eqs.(C.1,C.2), Eqs. (C.3,C.4) define
the fieldshE(1)l,m
i0and
hE(2)l,m
i0for the spherical core-shell system.
114
Appendix D
Calculations of the FréchetDerivatives of the ObjectiveFunctional and the Constraints
Let f : L2¡V ;C3
¢ → C be a functional. The Fréchet derivative of f at J0 ∈L2¡V ;C3
¢is defined as the gradient ∇Jf (J0) : L2
¡V ;C3
¢→ C such that [119]
|f (J0 + d)− f (J0)− (∇Jf (J0) ,d)|CkdkL2(V ;C3)
−→kdkL2(V ;C3)→0
0, ∀d ∈L2 ¡V ;C3¢ . (D.1)
However, it is well-known [119] that difficulties related to the definition of linearity
of the Fréchet differentiation operator would be encountered when f maps a complex
Banach space (in our case L2¡V ;C3
¢) into a real Banach space (in our case R). This
is in particular the case for E and³J, eGSJ
´. In such cases one considers L2
¡V ;C3
¢as a Hilbert space over R instead of C [119]. Consequently, the definition of the
Fréchet derivative given by (D.1) takes on the slightly modified form [119]
[f (J0 + d)− f (J0)−< (∇Jf (J0) ,d)]kdkL2(V ;C3)
−→kdkL2(V ;C3)→0
0, ∀d ∈L2 ¡V ;C3¢ . (D.2)115
Given these definitions, one can show that
< (∇JE (J0) ,d) = < (2J0 , d) , (D.3)
whereby one identifies ∇JE (J0) with 2J0. Similarly, one obtains for the first groupof (convex) constraints³
∇Jha(j)l,m − (B(j)
l,m,J0)i,d´=³−B∗(j)l,m ,d
´, (D.4)
whereby one identifies ∇J³a(j)l,m − (B(j)
l,m,J0)´with −B(j)
l,m. Finally, for the sec-
ond group of (nonconvex) constraints, one also obtains (after using the symmetry
property GS(r, r0) = GS(r
0, r))
<³∇J(J0, eGSJ0),d
´= <
³2 eGSJ0, d
´. (D.5)
In this case, one identifies ∇J(J0, eGSJ0) with 2eGSJ0.
116
Appendix E
Wavefunctions B(j)l,m forSpherically SymmetricBackgrounds
Another important and formally tractable case is that of a spherically symmetric
background for which
(r) = s(r)Θ(a− r) + 0Θ(r − a) (E.1)
µ(r) = µs(r)Θ(a− r) + µ0Θ(r − a), (E.2)
In this case the key step is the determination of the eigenfunctions of the homoge-
neous equation
∇×∇×∙E(r)
µs(r)
¸− ω2 s(r)E(r) = 0 (E.3)
which follows from equations (E.1), (E.2) and (2.5).
Fundamental for the dealing of this case, is the verification that equation (E.3)
admits solutions of the separable form E(2)l,m(r) = φl(r)Yl,m(r). Using well-known
properties of the vector spherical harmonics Yl,m(r) and the associated vector func-
tions r ×Yl,m(r) as well as the useful results (A.12)-(A.14) one readily finds that
the postulated solution E(2)l,m(r) = φl(r)Yl,m(r) obeys equation (E.3) as long as the
117
radial eigenfunctions φl(r) obey
µ−1s (r)∙l(l + 1)
r2−ω2 s(r)− 1
r2d
dr
µr2
d
dr
¶¸φl(r)=
1
r
d
dr
∙rφl(r)
dµ−1s (r)dr
¸. (E.4)
This is a key result, thanks to which one can readily carry out the forward mapping
and subsequent source inversion in this kind of spherically symmetric background.
In particular, following a procedure similar to that employed in appendix B for
piecewise-constant radially-symmetric backgrounds, but with the new radial eigen-
functions φl(r) playing for r < a the role of the spherical Bessel functions jl(kr) of
that appendix, one finds that
B(2)l,m(r) =
−η0Yl,m(r)
l(l + 1)a2
⎧⎨⎩ µr (r)φl(r)
µr (r) k0φl(a)Vl(k0a)− h(+)l (k0a)
hφ0l(a) +
φl(a)a
i⎫⎬⎭∗
.
(E.5)
The relations governing the associated electric source-free wavefunctions B(1)l,m(r)
can be derived similarly.
118
Appendix F
Establishment of The DispersionRelations (5.2) and (5.3)
The aim of this appendix is to establish the dispersion relations (5.2) and (5.3). As
mentioned in Chapter 5, we shall follow closely the simple presentations given in
[83, 85].
The proper modes for the system depicted in Fig.(5.1) are obtained by solving
Maxwell’s equations. For isotropic sourceless media, as is our case, the problem is
equivalent to solving the sourceless Helmholtz equations
£∇2 + k2j (ω)¤Ej (r) = 0, (F.1)
£∇2 + k2j (ω)¤Bj (r) = 0, (F.2)
where j = 1, 2, a, b and
k2j (ω) =ω2 j (ω)µj (ω)
c2, (F.3)
subject to the following static Dirichlet boundary conditions
z ·Dj ≡ D⊥j is continuous, (F.4)
z×Ej ≡ Ekj is continuous, (F.5)
z ·Bj ≡ Bkj is continuous, (F.6)
119
and
z×Hj ≡ Hkj is continuous. (F.7)
which must be satisfied at the interfaces between the media, i.e., at z = 0, z = a,
and z = a+ b.
We consider solutions of the form
Ej (r) = ej (z) eiκ·r⊥
= [ejx (z) bx+ ejy (z) by + ejz (z)bz] ei(kxx+kyy) (F.8)
and
Bj (r) = bj (z) eiκ·r⊥
= [bjx (z) bx+ bjy (z) by + bjz (z)bz] ei(kxx+kyy). (F.9)
It follows thatdejνdz−K2
j ejν = 0 (F.10)
anddbjν (z)
dz−K2
j bjν = 0, (F.11)
where the Kj ’s are defined as1
K2j ≡ k2x + k2y − j (ω)µj (ω)
ω2
c2
≡ κ2 − j (ω)µj (ω)ω2
c2, (F.12)
with ν = x, y, z. The solutions to Maxwell’s equations that satisfy (F.10) and (F.11)
form a complete set of solutions [141]. They are called surface modes because they
correspond to waves propagating parallel to the interfaces and decaying exponen-
tially for z < 0 and z > a+ b.
1Cf. with (5.4).
120
It is easy to show that all the boundary conditions (F.4-F.7) are satisfied if2⎧⎪⎨⎪⎩j (ω) ejz (z) is continuous
anddejz(z)dz is continuous,
(F.13)
and ⎧⎪⎪⎨⎪⎪⎩ejy (z) is continuous
and1
µj(ω)dejz(z)dz is continuous.
(F.14)
where j = 1, 2, a, b and ν = x, y, z.
Ignoring exponentially growing solutions to Eqs.(F.10) leaves us only with the
following physically acceptable ones
ejν (z) =
⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩Aνe
K1z , z < 0
BνeKaz + Cνe
−Kaz , 0 < z < a
DνeKbz +Eνe
−Kbz , a < z < a+ b
Fνe−K2z , z > a+ b,
(F.15)
where j = 1, 2, a, b as before but where the values of ν are now restricted to y and
z. These solutions along with the boundary conditions (F.13) and (F.14) lead to
two systems of linear algebraic equations for the coefficient Aν , Bν , Cν ,Dν , Eν , and
Fν . Imposing nontriviality conditions on the solutions of these systems yields the
following dispersion relations
g (ω) = 0 (F.16)
and
gµ (ω) = 0, (F.17)
2One can arrive at (F.13), for instance, in the following way. Demanding that (F.4) be satisfiednecessitates the continuity of j (ω) ejz (z) . Also, requiring that ∇ ·Dj = 0 and taking into accountthat ∇ ×Ej = (iω/c)Bj , automatically guarantees that ∇ ·Bj = 0 . Now, (F.5) is guaranteed ifejx is continuous. But this in turn is guaranteed if dejx/dz is continuous. Hence, (F.13). Conditions(F.14) are established in a similar fashion (See [83] for more detail.)
121
where
g (ω) ≡
¯¯¯¯
1 − a − a 0 0 0
0 aeKaa
ae−Kaa − be
Kba − be−Kba 0
0 0 0 beKb(a+b)
be−Kb(a+b) − 2e
−K2(a+b)
K1 −Ka Ka 0 0 0
0 KaeKaa −Kae
−Kaa −KbeKba Kbe
−Kba 0
0 0 0 KbeKb(a+b) −Kbe
−Kb(a+b) K2e−K2(a+b)
¯¯¯¯
(F.18)
and
gµ (ω) ≡
¯¯¯¯
1 −1 −1 0 0 0
0 eKaa e−Kaa −eKba −e−Kba 0
0 0 0 eKb(a+b) e−Kb(a+b) −e−K2(a+b)
K1µ1
−Kaµ1
Kaµa
0 0 0
0 KaµaeKaa −Ka
µae−Kaa −Kb
µbeKba Kb
µbe−Kba 0
0 0 0 KbµbeKb(a+b) −Kb
µbe−Kb(a+b) K2
µ2e−K2(a+b)
¯¯¯¯.
(F.19)
After some straightforward algebraic maniplulations one can show that (F.16) and
(F.17) do in fact reduce to (5.2) and (5.3), respectively.
122
Appendix G
Application of the ArgumentPrinciple to the Derivation ofEq. (5.20)
For the convenience of the reader we present here some technical details skipped
in Chapter 5 in the establishment of the Casimir energy formula (5.20). The idea
is to convert the sums on the right-hand side of (5.15) into contour integrals of
complex functions. This is a well-known technique [148]. We follow closely the
simple presentation given in [83].
A useful variant of the argument principle for meromorphic functions [142] stip-
ulates that
1
2πi
IC
f 0 (z)f (z)
zdz =
⎛⎝Xj
zj
⎞⎠f(zj)=0
−⎛⎝X
j
zj
⎞⎠f(zj)→∞
, (G.1)
i.e., the difference between the sum of the zeros of a meromorphic function f (z)
inside a contour C and the sum of its poles inside the same contour is given by the
integral on the left-hand side of (G.1).
Now, we want to apply (G.1) to the evaluation of the right hand side of (5.15).
123
Treating ω as a complex variable we haveXN
ω(N) (κ) = sum of zeros of g (ωj) (G.2)
and XN
ω(N)µ (κ) = sum of zeros of gµ (ωj) . (G.3)
The poles of gα (ωj) and gµ (ωj) are independent of d because d appears only in the
exponentials. Therefore,
1
2πi
IC
g0(ω)
g (ω)ωdω =
XN
ω(N) (κ)− a quantity independent of d (G.4)
and similarly
1
2πi
IC
g0µ (ω)
gµ (ω)ωdω =
XN
ω(N)µ (κ)− a quantity independent of d, (G.5)
where C is the contour depicted in Fig.(G.1). Implicit in this definition of C is
the fact that all physically acceptable solutions ω to the equations g (ω) = 0 and
gµ (ω) = 0 are positive real numbers. This is guaranteed if the constitutive parame-
ters are real functions of ω. Consequently, we shall restrict our selves to dissipation-
less backgrounds.1
Obviously the quantities that are independent of d do not contribute to the force.
1However, one must mention that when the technique is applied “blindly” to the case of complex-valued constitutive parameters it does reproduce the exact same results obtained by means of moresophisticated methods. An intuitive justification for the unanticipated success of the technique inthe presence of the dissipation is given in [143].
124
Im[ ]
C
1… N Re[ ]
Figure G.1: The contour, C, used in Eqs.(G.4,G.5); it is defined by the imaginaryaxis and the semicircle of infinite radius directed to the right. The points ω1, ..., ωNrepresent the N solutions of the dispersion relations g (ω) = 0 or gµ (ω) = 0.
125
Hence, we can ignore this term and write that
U =~L2
4π
1
2πi
Z ∞
0κdκ
⎡⎣IC
g0(ω)
g (ω)ωdω +
IC
g0µ (ω)
gµ (ω)ωdω
⎤⎦=
~L2
8π2i
Z ∞
0κdκ
Z −∞
∞
∙1
g (iξ)
∂g (iξ)
∂ (iξ)(iξ) +
1
gµ (iξ)
∂gµ (iξ)
∂ (iξ)(iξ)
¸d (iξ)
+~L2
8π2i
Z ∞
0κdκ
ZSemicircle
∙1
g (iξ)
∂g (iξ)
∂ (iξ)(iξ) +
1
gµ (iξ)
∂gµ (iξ)
∂ (iξ)(iξ)
¸d (iξ)
=~L2
8π2i
Z ∞
0κdκ
Z −∞
∞
∙1
g (iξ)
∂g (iξ)
∂ (iξ)(iξ) +
1
gµ (iξ)
∂gµ (iξ)
∂ (iξ)(iξ)
¸d (iξ) +Υ,
(G.6)
where Υ is a quantity that is independent of d in the limit of infinite semicircle
radius. Ignoring the d-independent term and substituting G (ξ) and Gµ (ξ) for
gµ (iξ) and g (iξ), respectively, yields
U =~L2
8π2
Z ∞
0κdκ
Z ∞
−∞
∙1
G (ξ)
∂G (ξ)
∂ξξ +
1
Gµ (ξ)
∂Gµ (ξ)
∂ξξ
¸dξ
=~L2
8π2
Z ∞
0κdκ
Z ∞
−∞[lnG (ξ) + lnGµ (ξ)] dξ, (G.7)
which establishes Eq.(5.20).
126
Bibliography
[1] V. G. Veselago. Electrodynamics of substances with simultaneously negative
electrical and magnetic permeabilities. Sov. Phys. Usp., 10:5—13, 1968.
[2] R. W. Ziolkowski and N. Engheta. Introduction, history and selected topics
in fundamental theories of metamaterials. In R. W. Ziolkowski and N. En-
gheta, editors, Metamaterials: Physics and Engineering Explorations, chap-
ter 1, pages 5—41. IEEE Press/Wiley, 2006.
[3] V. G. Veselago and E. E. Narimanov. The left hand of brightness: Past,
present and future of negative index materials. Nature Materials, 5:759—762,
October 2006.
[4] M. Lapine and S. Tretyakov. Contemporary notes on metamaterials. IET
Microw. Antennas and Propagation, 1(1):3—11, 2007.
[5] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz.
Composite medium with simultaneously negative permeability and permittiv-
ity. Phys. Rev. Lett., 84:4184—4187, May 2000.
[6] J. B. Pendry. Negative refraction makes a perfect lens. Phys. Rev. Lett.,
85:3966, 2000.
[7] N. Engheta. An idea for thin subwavelength cavity resonators using meta-
materials with negative permittivity and permeability. IEEE Antennas and
Wireless Propagation Lett., 1:10—13, 2002.
127
[8] R. W. Ziolkowski and A. D. Kipple. Application of double negative materials
to increase the power radiated by electrically small antennas. IEEE Trans.
Antennas and Propagation, 51:2626—2640, Oct. 2003.
[9] A. Alù and N. Engheta. Guided modes in a waveguide filled with a pair of
single-negative (SNG), double-negative (DNG), and/or double positive (DPS)
layers. IEEE Trans. Microw. Theory and Tech., MTT-52:199—210, 2004.
[10] A. Alù and N. Engheta. Achieving transparency with plasmonic and meta-
material coatings. Phys. Rev. E., 72:016623, 2005.
[11] R. W. Ziolkowski. Ultrathin, metamaterial-based laser cavities. J. Opt. Soc.
Am. B, 23(3):451—460, March 2006.
[12] R. W. Ziolkowski and N. Engheta. Metamaterials: Physics and Engineering
Explorations. IEEE Press/Wiley, 2006.
[13] R. W. Ziolkowski and E. Heyman. Wave propagation in media having negative
permittivity and permeability. Phys. Rev. E, 64:056625—1(15), 2001.
[14] E. A. Marengo, A. J. Devaney, and F. K. Gruber. Inverse source problem with
reactive power constraint. IEEE Trans. Antennas and Propagation, 52:1586—
1595, June 2004.
[15] A. J. Devaney and R. P. Porter. Holography and the inverse source problem:
Part II: Inhomogeneous media. J. Opt. Soc. Am. A, 2:2006—2011, 1985.
[16] L. Tsang, A. Ishimaru, R. P. Porter, and D. Rouseff. Holography and the
inverse source problem. III. inhomogeneous attenuative media. J. Opt. Soc.
Am. A, 4:1783—1787, 1987.
[17] A. J. Devaney, E. A. Marengo, and Mei Li. Inverse source problem in non-
homogeneous background media. SIAM J. Appl. Math., 67:1353—1378, 2007.
[18] H. R. Raemer. Radiation from linear electric or magnetic antennas surrounded
by a spherical plasma shell. IRE Trans. Antennas and Propagation, 10:69—78,
January 1962.
128
[19] D. Lamensdorf. An experimental investigation of dielectric-coated antennas.
IEEE Trans. Antennas and Propagation, 15:767—771, Nov. 1967.
[20] D. Lamensdorf and C.-Y. Ting. An experimental and theoretical study of
the monopole embedded in a cylinder of anisotropic dielectric. IEEE Trans.
Antennas and Propagation, 16:342—349, May 1968.
[21] S. A. Long, M. W. McAllister, and L. C. Shen. The resonant cylindrical
dielectric cavity antenna. IEEE Trans. Antennas and Propagation, 31:406—
412, May 1983.
[22] K. W. Leung. Complex resonance and radiation of hemispherical dielectric-
resonator antenna with a concentric conductor. IEEE Trans. Microw. Theory
and Tech., 49:524—531, March 2001.
[23] J. R. James and J. C. Vardaxoglou. Investigation of properties of electrically-
small spherical ceramic antennas. Electronics Lett., 38:1160—1162, 2002.
[24] E. E. Altshuler. Electrically small genetic antennas immersed in a dielectric.
Proc. IEEE Antennas and Propagation Soc. Symp., 3:2317—2320, June 2004.
[25] R. C. Hansen and M. Burke. Antennas with magneto-dielectrics. Microw.
Opt. Technol. Lett., 26:75—78, July 2000.
[26] K. Buell, H. Mosallaei, and K. Sarabandi. A substrate for small patch antennas
providing tunable miniaturization factors. IEEE Trans. Microw. Theory and
Tech., 54:135—146, Jan. 2006.
[27] H. Mosallaei and K. Sarabandi. Magneto-dielectrics in electromagnetics: Con-
cept and applications. IEEE Trans. Antennas and Propagation, 52:1558—1567,
June 2004.
[28] Y. Mano and S. Bae. A small meander antenna by magneto-dielectric material.
Proc. IEEE Intl. Symp. Microw. Antenn. Propagation and EMC Tech. for
Wireless Comm., 1:63—66, Aug. 2005.
129
[29] P. M. T. Ikonen, K. N. Rozanov, A. V. Osipov, P. Alitalo, and S. A. Tretyakov.
Magnetodielectric substrates in antenna miniaturization: Potential and limita-
tions. IEEE Trans. Antennas and Propagation, 54(11):3391—3399, November
2006.
[30] H. Mosallaei and K. Sarabandi. Design and modeling of patch antenna printed
on magneto-dielectric embedded-circuit metasubstrate. IEEE Trans. Anten-
nas and Propagation, 55(1):45—52, January 2007.
[31] M. Thèvenot, C. Cheype, A. Reineix, and B. Jecko. Directive photonic-
bandgap antennas. IEEE Trans. Microw. Theory and Tech., 47:2115—2122,
Nov. 1999.
[32] B. Temelkuran, M. Bayindir, E. Ozbay, R. Biswas, M. M. Sigalas, G. Tuttle,
and K. M. Ho. Photonic crystal-based resonant antenna with a very high
directivity. J. Appl. Phys., 87:603—605, 2000.
[33] S. Enoch, G. Tayeb, P. Sabouroux, N. Guérin, and P. Vincent. A metamaterial
for directive emission. Phys. Rev. Lett., 89:213902—1(4), 2002.
[34] A. Alú and N. Engheta. Radiation from a traveling-wave current sheet at the
interface between a conventional material and a metamaterial with negative
permittivity and permeability. Microw. Opt. Technol. Lett., 35:460—463, 2002.
[35] D. Tonn and R. Bansal. Practical considerations for increasing radiated power
from an electrically small antenna by application of a double negative meta-
material. In IEEE Internat. Antennas and Propagation Symp., volume 2A,
pages 602—605, July 2005.
[36] A. Erentok, P. L. Luljak, and R. W. Ziolkowski. Characterization of a volu-
metric metamaterial realization of an artificial magnetic conductor for antenna
applications. IEEE Trans. Antennas and Propagation, 53:160—172, Jan. 2005.
[37] G. Lovat, P. P. Burghignoli, F. Capolino, D. R. Jackson, and D. R. Wilton.
Analysis of directive radiation from a line source in a metamaterial slab with
low permittivity. IEEE Trans. Antennas and Propagation, 54:1017—1030,
March 2006.
130
[38] B.-I. Wu, W. Wang, J. Pacheco, X. Chen, T. Grzegorczyk, and J.A. Kong.
A study of using metamaterials as antenna substrate to enhance gain. Prog.
Electromag. Res., 51:295—328, 2005.
[39] S. F. Mahmoud. A new miniaturized annular ring patch resonator partially
loaded by a metamaterial ring with negative permeability. IEEE Antennas
and Wireless Propagation Lett., 3:19—22, 2004.
[40] Richard W. Ziolkowski and Allison D. Kipple. Reciprocity between the effects
of resonant scattering and enhanced radiated power by electrically small an-
tennas in the presence of nested metamaterials. Phys. Rev. E, 72:036602—1—5,
September 2005.
[41] R. W. Ziolkowski and A. Erentok. Metamaterial-based efficient electrically
small antennas. IEEE Trans. Antennas and Propagation, AP-54:2113—2130,
July 2006.
[42] S.A. Tretyakov, S.I. Maslovski, A.A. Sochava, and C.R. Simovski. The influ-
ence of complex material coverings on the quality factor of simple radiating
systems. IEEE Trans. Antennas and Propagation, 53(3):965— 970, March 2005.
[43] S. A. Tretyakov and M. Ermutlu. Modeling of patch antennas partially loaded
with dispersive backward-wave materials. IEEE Antennas and Wireless Prop-
agation Lett., 4:266—269, 2005.
[44] A. D. Yaghjian and S. R. Best. Impedance, bandwidth, and q of antennas.
IEEE Trans. Antennas and Propagation, 53(4):1298— 1324, April 2005.
[45] R. W. Ziolkowski and A. Erentok. At and below the Chu limit: Passive and
active broad bandwidth metamaterial-based electrically small antennas. IET
Microw. Antennas and Propagation, 1(1):116—128, February 2007.
[46] A. Erentok and R. W. Ziolkowski. A hybrid optimization method to analyze
metamaterial-based electrically small antennas. IEEE Trans. Antennas and
Propagation, 55(3):731—741, March 2007.
131
[47] A. Alù, F. Bilotti, and N. Engheta. Subwavelength, compact, resonant patch
antennas loaded with metamaterials. IEEE Trans. Antennas and Propagation,
55(1):13—25, January 2007.
[48] P.M.T. Ikonen, P. Alitalo, and S.A. Tretyakov. On impedance bandwidth of
resonant patch antennas implemented using structures with engineered dis-
persion. Antennas and Wireless Propagation Lett., 6:186—190, 2007.
[49] E. Rodes, L. Moustapha, L. Mercier, M. Thévenot, T. Mondeniere, and
B. Jecko. On the use of metamaterial surfaces for EBG antenna improvements.
In Antennas and Propagation Conference, 2007. LAPC 2007. Loughborough,
Loughborough, UK, 2-3 April 2007.
[50] N. Engheta, A. Alù, R. W. Ziolkowski, and A. Erentok. Fundamentals of
waveguide and antenna applications involving DNG and SNG metamaterials.
In N. Engheta and RichardWalter Ziolkowski, editors,Metamaterials: Physics
and Engineering Explorations, chapter 2, pages 43—85. IEEE Press/Wiley,
2006.
[51] E. A. Marengo and A. J. Devaney. The inverse source problem of electro-
magnetics: Linear inversion formulation and minimum energy solution. IEEE
Trans. Antennas and Propagation, 47(2):410—412, Feb. 1999.
[52] E. A. Marengo and R. W. Ziolkowski. Nonradiating and minimum energy
sources and their fields: Generalized source inversion theory and applications.
IEEE Trans. Antennas and Propagation, 48:1553—1562, Oct. 2000.
[53] H. A. Wheeler. Fundamental limitations of small antennas. Proc. IRE,
35:1479—1484, Dec. 1947.
[54] H. A. Wheeler. Small antennas. IEEE Trans. Antennas and Propagation,
AP-23:462—469, July 1975.
[55] J.A. Kong. Electromagnetic Wave Theory. EMW Publishing, Cambridge,
MA, 2005.
132
[56] X. Chen, T. M. Grzegorczyk, and J. A. Kong. Optimization approach to
the retrieval of the constitutive parameters of a slab of general bianisotropic
medium. Prog. Electromag. Res., 60:1—18, 2006.
[57] R. P. Porter and A. J. Devaney. Holography and the inverse source problem.
J. Opt. Soc. Am., 72(3):327—330, 1982.
[58] C. A. Balanis. Advanced Engineering Electromagnetics. Wiley, New York,
1989.
[59] R. E. Collin. Field Theory of Guided Waves. IEEE Press, New York, 1991.
[60] M. Bertero. Linear inverse and ill-posed problems. In P. W. Hawkes, editor,
Adv. Electronics and Electron Phys., volume 75, pages 1—120. Academic Press,
San Diego, 1989.
[61] A. J. Devaney and G. C. Sherman. Nonuniqueness in inverse source and
scattering problems. IEEE Trans. Antennas and Propagation, 30:1034—1037,
Sept. 1982.
[62] A. J. Devaney and E. Wolf. Radiating and nonradiating classical current
distributions and the fields they generate. Phys. Rev. D, 8:1044—1047, 1973.
[63] N. Bleistein and J. K. Cohen. Nonuniqueness in the inverse source problem
in acoustics and electromagnetics. J. Math. Phys., 18:194—201, 1977.
[64] G. Kaiser. Making electromagnetic wavelets II. spheroidal shell antennas. J.
Phys. A: Math. Gen., 38:495—508, 2005.
[65] A. J. Devaney, G. Kaiser, E. A. Marengo, R. Albanese, and G. Erdmann.
The inverse source problem for wavelet fields. to appear in IEEE Trans. on
Antennas and Propagation, 2008.
[66] R. P. Porter and A. J. Devaney. Generalized holography and computational
solutions to inverse source problems. J. Opt. Soc. Am., 72:1707—1713, 1982.
[67] K. J. Langenberg. Applied inverse problems for acoustic, electromagnetic and
elastic wave scattering. In P. C. Sabatier, editor, Basic Methods of Tomography
and Inverse Problems, pages 128—467. Adam Hilger, Bristol, 1987.
133
[68] T. S. Angell, A. Kirsch, and R. E. Kleinman. Antenna control and optimiza-
tion. Proc. IEEE, 79:1559—1568, Oct. 1991.
[69] D. Liu, R. J. Garbacz, and D. M. Pozar. Antenna synthesis and solution
of inverse problems by regularization methods. IEEE Trans. Antennas and
Propagation, 38:862—868, June 1990.
[70] O. M. Bucci, G. D’Elia, G. Mazzarella, and G. Panariello. Antenna pattern
synthesis: A new general approach. Proc. IEEE, 82:358—371, March 1994.
[71] G. A. Deschamps and H. S. Cabayan. Antenna synthesis and solution of inverse
problems by regularization methods. IEEE Trans. Antennas and Propagation,
20:268—274, May 1972.
[72] Y. T. Lo, S. W. Lee, and Q. W. Lee. Optimization of directivity and signal-
to-noise ratio of an arbitrary antenna array. Proc. IEEE, 54:1033—1045, Aug.
1966.
[73] L. J. Chu. Physical limitations of omni-directional antennas. J. Appl. Phys.,
19:1163—1175, 1948.
[74] R. F. Harrington. On the gain and beamwidth of directional antennas. IRE
Trans. Antennas and Propagation, 6:219—225, 1958.
[75] R. L. Fante. Quality factor of general ideal antennas. IEEE Trans. Antennas
and Propagation, AP-17:151—155, March 1969.
[76] J. S. McLean. A re-examination of the fundamental limits on the radiation q
of electrically small antennas. IEEE Trans. Antennas and Propagation, AP-
44:672—676, May 1996.
[77] G. A. Thiele, P. L. Detweiler, and R. P. Penno. On the lower bound of
the radiation q for electrically small antennas. IEEE Trans. Antennas and
Propagation, 51:1263—1269, June 2003.
[78] H. P. Chang, T. K. Sarkar, O. Maroja, and C. Pereira-Filho. Antenna pat-
tern synthesis utilizing spherical bessel functions. IEEE Trans. Antennas and
Propagation, 48:853—859, June 2000.
134
[79] J. M. Jarem. Strip and line source synthesis with a quality factor constraint.
IEEE Trans. Antennas and Propagation, AP-28:36—41, January 1980.
[80] D. R. Rhodes. On the quality factor of strip and line source antennas and its
relationship to superdirectivity ratio. IEEE Trans. Antennas and Propagation,
AP-20:318—325, May 1972.
[81] S. Weinberg. The Quantum Theory of Fields, volume 1. Cambridge University
Press, Cambridge, UK, 1995.
[82] S. Weinberg. Cosmology. Oxford University Press, Oxford, UK, 2008.
[83] P. W. Milonni. The Quantum Vacuum: An Introduction to Quantum Electro-
dynamics. Academic Press, San Diego, CA, 1994.
[84] H. B. G. Casimir. On the attraction between two perfectly conducting plates.
Proc. K. Ned. Akad. Wet., 51:793—795, 1948.
[85] M. Bordag, U. Mohideen, and V. M. Mostepanenko. New developments in the
Casimir effect. Phys. Rep., 353:1—205, 2001.
[86] G. Bressi, G. Carugno, R. Onofrio, and G. Ruoso. Measurement of the Casimir
force between parallel metallic surfaces. Phys. Rev. Lett., 88:041804, 2002.
[87] S. K. Lamoreaux. Demonstration of the Casimir force in the 0.6 to 6 µm
range. Phys. Rev. Lett., 78:5—8, 1997.
[88] U. Mohideen and A. Roy. Precision measurement of the Casimir force from
0.1 to 0.9 µm. Phys. Rev. Lett., 81:4549—4552, 1998.
[89] T. Ederth. Template-stripped gold surfaces with 0.4-nm rms roughness suit-
able for force measurements: Application to the Casimir force in the 20-100
nm range. Phys. Rev. A, 62:062104, 2000.
[90] H. B. Chan, V. A. Aksyuk, R. N. Kleinman, D. J. Bishop, and F. Ca-
passo. Quantum mechnical actuation of microelectromechanical systems by
the Casimir forces. Science, 291:1941—1944, 2001.
135
[91] R. S. Decca, D. Lopez, E. Fischbach, and D. E. Krause. Measurement of the
Casimir force between dissimilar metals. Phys. Rev. Lett., 91:050402, 2003.
[92] J. Schwinger, L. L. DeRaad, and K. A. Milton. Casimir effect in dielectrics.
Ann. Phys. (N.Y.), 115:1—23, 1978.
[93] R. L. Jaffe. The Casimir effect and the quantum vacuum. Phys. Rev. D,
72:021301(R), 2005.
[94] C. Raabe and D.-G. Welsch. Casimir force acting on magnetodielectric bodies
embedded in media. Phys. Rev. A, 71:013814, 2005.
[95] L. P. Pitaevskii. Comment on "Casimir force acting on magnetodielectric
bodies embedded in media". Phys. Rev. A, 73:047801, 2006.
[96] M. Serry, D. Walliser, and J. Maclay. The role of the Casimir effect in the
static deflection and stiction of membrane strips in microelectromechanical
systems MEMS. J. App. Phys., 84:2501—2506, 1998.
[97] F. Capasso, J. N. Munday, D. Ianuzzi, and H. B. Chan. Casimir forces and
quantum electrodynamical torques: Physics and nanomechanics. IEEE J.
Selected Topics in Quantum Electronics, 13(2):400—414, March/April 2007.
[98] E. M. Lifshitz. The theory of molecular attractive forces between solids. Sov.
Phys. JETP, 2:73—83, 1956.
[99] I. E. Dzyaloshinskii, E. M. Lifshitz, and L. P. Pitaevskii. The general theory
of van der waals forces. Adv. Phys., 10:309—318, 1961.
[100] J. Mahanty and B. W. Ninham. Dispersion Forces. Academic Press, London,
UK, 1976.
[101] C. Henkel and K. Joulain. Casimir force between designed materials: What
is possible and what not. Europhys. Lett., 72(6):929—935, December 2005.
[102] U. Leonhardt and T. G. Philbin. Quantum levitation by left-handed metama-
terials. New J. Phys., 9:254, 2007.
136
[103] S. A. Ramakrishna. Physics of negative refractive index materials. Rep. Prog.
Phys., 68:449—460, 2005.
[104] M. S. Tomaš. Casimir forces in absorbing multilayers. Phys. Rev. A.,
66:052103, 2002.
[105] W. L. Mochán, C. Villareal, and R. Esquivel-Sirvent. On Casimir forces for
media with arbitrary dielectric properties. Rev. Mex. Fís., 48(4):339—342,
2002.
[106] C. Raabe, L. Knöll, and D.-G. Welsh. Three-dimensional Casimir force be-
tween absorbing multilayer dielectrics. Phys. Rev. A, 68:033810, 2003.
[107] C. Raabe, L. Knöll, and D.-G. Welsh. Erratum: Three-dimensional Casimir
force between absorbing multilayer dielectrics. Phys. Rev. A, 69:019901, 2004.
[108] S. A. Ellingsen. Casimir attraction in multilayered plane parallel magnetodi-
electric systems. J. Phys. A: Math. Theor., 40:1951—1961, 2007.
[109] S. A. Ellingsen and I. Brevic. Casimir force on real materials—the slab and
cavity geometry. J. Phys. A: Math. Theor., 40:3643—3664, 2007.
[110] F. S. S. Rosa, D. A. R. Dalvit, and P. W. Milonni. Casimir-lifshitz theory and
metamaterials. Phys. Rev. Lett., 100:183602, 2008.
[111] Y. Yang, R. Zeng, J. Xu, and S. Liu. Casimir force between left-handed-
material slabs. Phys. Rev. A, 77:015803, 2008.
[112] E. A. Marengo, M. R. Khodja, and A. Boucherif. Inverse source problem in
non-homogeneous background media. Part II: Vector formulation and antenna
substrate performance characterization. SIAM J. Appl. Math, 69(1):81—110,
2008.
[113] J. A. Stratton. Electromagnetic Theory. McGraw Hill, New York, 1941.
[114] A. J. Devaney and E. Wolf. Multipole expansions and plane wave representa-
tions of the electromagnetic field. J. Math. Phys., 15:234—244, 1974.
[115] J. D. Jackson. Classical Electrodynamics. Wiley, New York, 1999.
137
[116] E. A. Marengo and A. J. Devaney. Time-dependent plane wave and multipole
expansions of the electromagnetic field. J. Math. Phys., 39:3643—3660, 1998.
[117] G. B. Arfken and H. J. Weber.Mathematical Methods for Physicists. Academic
Press, 2005.
[118] J. Jahn. Introduction to the Theory of Nonlinear Optimization. Springer-
Verlag, New York, 2007.
[119] T. S Angell and A. Kirsch. Optimization Methods in Electromagnetic Radia-
tion. Springer-Verlag, New York, 2004.
[120] A. J. Kurdilla and M. Zabarankin. Convex Functional Analysis. Birkhauser,
Basel, 2005.
[121] M. S. Berger. Nonlinear Functional Analysis. Academic Press, New York,
1977.
[122] H. Tuy and N. V. Thuong. On the global minimization of a convex function
under general nonconvex constraints. Appl. Math. Optim., 18:119—142, 1988.
[123] G. Mie. Beiträge zur optik trüber medien, speziell kolloidaler metallösungen.
Ann. Phys., 25:377—445, 1908.
[124] M. Abramowitz and I. A. Stegun. Hankbook of Mathematical Functions. Dover,
New York, 1964.
[125] M. R. Khodja and E. A. Marengo. Radiation enhancement due to metama-
terial substrates from an inverse source theory. Phys. Rev. E., 77(4):046605,
April 2008.
[126] H. B. Keller and J. B. Keller. Reflection and transmission of electromagnetic
waves by a spherical shell. J. Appl. Phys., 20:393—396, April 1949.
[127] M. G. Andreason. Radiation from a radial dipole through a thin dielectric
shell. IRE Trans. Antennas and Propagation, 5:337—342, Oct. 1957.
138
[128] R. V. Row. Radiation efficiency of electric and magnetic dipole antennas
surrounded by a small spherical shell of lossy dielectric. IEEE Trans. Antennas
and Propagation, 12:646—647, Sept. 1964.
[129] D. E. Burrick. Spheres. In G. T. Ruck, D. E. Burrick, W. D. Stuart, and
C. K. Krichbaum, editors, Radar Cross Section Handbook, volume 1, chapter 3.
Plenum Press, New York, 1970.
[130] L. Gao and Y. Huang. Extinction properties of a coated sphere containing a
left-handed material. Optics Communications, 239:25—31, 2004.
[131] A. L. Aden and M. Kerker. Scattering of electromagnetic waves from two
concentric spheres. J. Appl. Phys., 22(10):1242—1246, October 1951.
[132] A. Alù and N. Engheta. Polarizabilities and effective parameters for collec-
tions of spherical nanoparticles formed by pairs of concentric double negative,
single negative and/or double positive metamaterial layers. J. Appl. Phys.,
97:094310, 2005.
[133] A. Alù and N. Engheta. Erratum: Polarizabilities and effective parameters
for collections of spherical nanoparticles formed by pairs of concentric double
negative, single negative and/or double positive metamaterial layers. J. Appl.
Phys., 99:069901, 2006.
[134] F. Cakoni and D. Colton. Qualitative Methods in Inverse Scattering Theory.
Springer, Berlin, 2006.
[135] R. Potthast. Point Sources and Multipoles in Inverse Scattering Theory. Chap-
man and Hall / CRC, Boca Raton, Florida, 2001.
[136] E. A. Marengo, F. K. Gruber, and F. Simonetti. Time-reversal MUSIC imag-
ing of extended targets. IEEE Trans. Image Processing, 16:1967—1984, 2007.
[137] E. A. Marengo and M. R. Khodja. Source synthesis in substrate media: Funda-
mental bounds. International Symposium on Electromagnetic Theory, URSI
Commision B, Ottawa, Ontario, Canada, 26-28 July, 2007.
139
[138] M. R. Khodja and E. A. Marengo. Comparative study of radiation enhance-
ment due to metamaterials. Radio Sci., 2008. doi:10.1029/2007RS003803, in
press.
[139] S. R. Best. A discussion on the properties of electrically small self-resonant
wire antennas. IEEE Antennas and Propagation Mag., 46(6):9—22, 2004.
[140] N. G. Van Kampen, B. R. A. Nijboer, and K. Schram. On the macroscopic
theory of van der Waals forces. Phys. Lett., 26A:307, 1968.
[141] I. Bialynicki-Birula and J. B. Brojan. Completeness of evanescent waves. Phys.
Rev. D, 5:485—487, 1972.
[142] S. G. Krantz. Handbook of Complex Variables. Birkhäuser, Boston, MA, 1999.
[143] S. K. Lamoreaux. The Casimir force: Background, experiments, and applica-
tions. Rep. Prog. Phys., 68:201—236, 2005.
[144] E. A. Marengo and R. W. Ziolkowski. A new procedure for specifying non-
radiating current distributions and the fields they produce. J. Math. Phys.,
41:845—866, 2000.
[145] E. L. Hill. The theory of vector spherical harmonics. Am. J. Phys., 22:211—214,
1954.
[146] Le-Wei Li, Pang-Shyan Kooi, Mook-Seng Leong, and Tat-Soon Yeo. Electro-
magnetic dyadic Green’s function in spherically multilayered media. IEEE
Trans. Microw. Theory and Tech., 42(12):2302—2310, December 1994.
[147] C. T. Tai. Dyadic Green’s Functions in Electromagnetic Theory. Intext Edu-
cational, Scranton, PA, 1971.
[148] P. M. Morse and H. Feshbach. Methods of Theoretical Physics, volume 1.
McGraw-Hill, New York, 1953.
140