by
ARVIN REZA
A thesis submitted to the Department of Physics, Engineering Physics, and Astronomy in conformity with the requirements for the degree of Master of Science
Abstract
Metamaterials, articially engineered structures with negative average relative permittivity and permeability, provide a route to creating potential devices with exciting electromagnetic properties that cannot be obtained with natural materials. One particularly interesting metamaterial device, is a planar metamaterial waveguide structure (MWS) that has potentially exciting applications. In this thesis, the properties and potential applications of metamaterial waveguide structures are explored. In particular, we examine the properties of metamaterial waveguides when the limitations arise from fabrication techniques and physical principles are taken into account. First, we study the basic properties of dispersion curves of an idealized (without loss and dispersion) metamaterial waveguide structure. We show that there are a
rich variety of modes, such as the bound modes, the surface polariton modes, and the complex leaky modes, that are supported in MWS and have entirely dierent properties than the modes of a conventional waveguide structure. These novel modes provide more control over the electromagnetic elds and consequently lead to potential applications ranging from waveguide miniaturization to the slowing down of the light. Next, we study the eects of dispersion and loss, which are the inherent features of metamaterials, on the properties of MWS. We numerically show that the characteristic modes of the MWS are signicantly changed particularly near the slow-light-modes i
when the intrinsic loss is introduced into the system. In particular we show that the stopped-light-modes disappear even in the presence of an arbitrarily small amount of loss. Moreover, we nd several novel properties such as a splitting of complex leaky modes in a lossy MWS.
ii
Co-Authorship
Some of the results in Chapter 3 have been published in Nature [49]. created by the author unless otherwise cited.
Figures are
iii
Acknowledgments
The author would like to express his sincere gratitude hereby to his parents and his supervisors, Professor Marc Dignam and Professor Stephen Hughes, who have provided him the best of support, supervision and insight.
iv
Contents
Abstract
Co-Authorship
iii
Acknowledgments
iv
Contents
List of Tables
viii
List of Figures
ix
Chapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1 4 7 11 14
Thesis overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17 18 24
Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46 48 52
Energy density in a dispersive and absorptive media . . . . . . . . . . Power ux, group velocity, and energy velocity in a lossy and dispersive metamaterial waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57 66 73
3.5 3.6
. . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Transfer matrix formalism for lossy metamaterial multilayer structures The Attenuated Total Reection technique (ATR) . . . . . . . . . . . Anomalous Goos-Hanchen shift due to surface mode excitation . . . .
75 76 82 94
vi
4.5
Concluding remarks
. . . . . . . . . . . . . . . . . . . . . . . . . . .
96
. . . . . . .
98
Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparison between
and
approaches . . . . . . . . . . . . . . . . . . . . . . . . .
Bibliography
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
119
vii
List of Tables
1.1
. . . . . . . . . . . .
viii
List of Figures
2 4 6
Schematic geometry of a metamaterial waveguide structure . . . . . . Graphical solution of the TE modes . . . . . . . . . . . . . . . . . . . Dispersion curves of a dispersionless and lossless waveguide . . . . . . The
18 27 33 35 36 40 41 42 43
and
. .
. . .
The surface polariton mode coupling with the higher order mode . . . The
T E1
and
T E5
modes coupling
. . . . . . . . . . . . . . . . . . .
The bending back of the dispersion curves near Permeability near the resonance frequency
fres
. . . . . . . . . .
. . . . . . . . . . . . . . .
2.10 Dispersion curves of the lossy dispersive metamaterial waveguide structure for the TM modes . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.1 3.2
. . . . . . .
47
Power ux, energy density and energy velocity of the dispersionless waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
ix
3.3
The energy velocity and the power ux for the backward propagating
T E5
3.4
and
T E1
coupling mode . . . . . . . . . . . . . . . . . . . . . . .
61
The energy velocity and the power ux for the forward propagating
T E5
3.5 3.6
and
T E1
coupling mode . . . . . . . . . . . . . . . . . . . . . . .
63 64
T E2
mode
. . . . . . . . .
(a) The normalized energy velocity and (b) the propagation loss for the backward-propagation
T E2
. . . . . . . . . . . . .
65 67
3.7 3.8
= 0.01c
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
3.9
Dispersion curves of the proposed waveguide structure in Fig. 3.7(b) for TM modes with
= 0.01c .
. . . . . . . . . . . . . . . . . . . . .
70
3.10 The energy velocity and propagation loss of the proposed waveguide structure for
T M2
mode
. . . . . . . . . . . . . . . . . . . . . . . . .
71
3.11 The energy velocity of the proposed waveguide structure with and
= 0.01
73
= 0.0001
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . .
77 78 84 87 89
Dispersion curves and prism coupling . . . . . . . . . . . . . . . . . . Eects of the air gap width, d, on the reectance. . . . . . . . . . . . . . . .
Comp arisen between the solutions obtained from the ART with the exact solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 91
4.7
4.8 4.9
nprism
for the
T E1
mode . . . . . . . . . . . . .
92
Schematic of the positive and negative lateral beam shift (Goos-Hanchen shift). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 96
4.10 The lateral Goos-Hanchen (GH) shift for the reected beam
5.1
Dispersion curves of the dispersionless and lossless metamaterial slab in the complex- and in the complex- approaches . . . . . . . . . . 100
5.2
The dispersion curves of the dispersive, lossless metamaterial waveguide in the complex- and in the complex- approaches . . . . . . . . 101
5.3
The dispersion curves of the dispersive, lossy metamaterial waveguide for the
T E2
mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
103
5.4
Group velocity and loss propagation of the metamaterial waveguide for the
T E2
104
5.5
Comparison between the imaginary part of the exact obtained from Eq. 5.5
with the
106 108
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.6 5.7
Reectance versus the incident angle for the metamaterial waveguide with
= 0.1
GHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
111 112
5.8 5.9
xi
Chapter 1
Introduction
their constituent materials but by man-made structures that are smaller than the electromagnetic wavelength involved. These structures that form the metamaterial act like articial atoms and can be coupled to both the electric and magnetic eld components of the electromagnetic waves, leading novel optical properties, such as a negative index of refraction. In this chapter, we provide a brief history of the
eld, a simple introduction to the fundamental properties and fabrication methods of metamaterials and an overview of the recent theoretical and experimental work in 1
CHAPTER 1.
INTRODUCTION
Figure 1.1: Lycurgus Cup (British Museum; AD fourth century) is made of ruby glass, the rst optical metamaterials. The cup has a greenish color when light reected from the glass; however, it has a reddish color when light transmitted through the glass. [source: Nature Photonics 1, 207-208 (2007).]
the eld, emphasizing the particular area of metamaterial waveguide structures. It is probably surprising to learn that metamaterials have a long history; for example, medieval ruby glass might be considered as the rst optical metamaterial [7]. Ruby glass contains nano-scale gold droplets (5-60 nm) that give an unusual
color to the glass. Actually, a resonance of the surface plasmon on the gold droplets causes the extraordinarily color of the glass. Ruby glass appears neither golden nor transparent, but depending on the size of the droplets and the direction of the light beam, looks red or green. In Fig. 1.1, Lycurgus cup, the Roman cup made of ruby glass, is shown. It has a greenish color when viewed in reected light, but it has a reddish color when light is transmitted through the glass. We can name plenty of metamaterials in the last century; for example, the work of
CHAPTER 1.
INTRODUCTION
Bose on the rotation of the plane of polarization by articial twisted structures in 1898 [8] or articial dielectrics for microwave antenna lenses by Kock in 1945 [9]. However, what we today call metamaterials was born at the beginning of this century (2000) in a lab at the University of California at San Diego [10] when D.R.Smith and his team presented evidence of both a negative electric permittivity and a negative magnetic permeability at a specic microwave range of frequencies in their experiment. The idea of metamaterials with negative permittivity and permeability was rst proposed by Victor Veselago in 1968 [12], who termed such media left-handed material (LHM) because the electric and magnetic elds and the wave vector form a lefthanded set of vectors rather than the usual right-hand set. In his study, Veselago
showed that the Poynting vector of an electromagnetic wave is anti-parallel to the wave vector in a left-handed material; thus, light propagates in the opposite direction to the energy ux. This leads to number of unusual properties in metamaterials such as negative-index refraction, the reversal of Cherenkov radiation, the reversal of the Doppler shift, and the reversal of Snell's law. These revolutionary properties give
remarkable control over electromagnetic elds with incredible possible applications such as invisibility devices. As a further consequence of negative-refraction, metamaterials produce unusual optical eects. For example, a bulb located in front of a metamaterial slab, would appear to be in front of the slab; a sh swimming in water made of metamaterial would appear to be moving above the water surface (See Fig 1.2). Negative-refraction could even lead to negative space [45]. For example, if a distance between two
objects is properly lled with a material with negative-refraction and a conventional material, the two objects would appear to be beside each other.
CHAPTER 1.
INTRODUCTION
Figure 1.2: The schematic of how a sh observed in water and in the water made of metamaterials. [source: http://www.berkeley.edu/news/media/download/]
work of Veselago [12, 13, 14] is generally considered the rst comprehensive investigation of metamaterials and even it did not receive signicant attention in the scientic community until the rst modern metamaterial was fabricated in 2000 [10]. In fabricating the rst metamaterial with negative permeability and permittivity, Smith and his coworkers followed the idea of Sir John Pendry [11]. They designed a structure, composed of metallic rings and gaps, for creating negative permeability in a specic range of frequency. The proposed structure by Pendry, the so-called split ring resonator (SRR), is magnetically active though it is made of non-magnetic materials. To achieve simultaneously negative electric permittivity, he arranged metallic wires in periodic lattices with SRRs. By using a periodic arrangement of metallic wires and
CHAPTER 1.
INTRODUCTION
rings, Smith and his coworkers showed evidence of negative-index refraction in 2001 [15]. The rst metamaterials were demonstrated in microwave frequencies; however most potential applications are in the optical domain. It is a very challenging problem to extend metamaterials to optical frequency because this requires scaling the articial atoms from the millimeter size down to the nanometer size. However, the progress in nanotechnology and the thrust for optical metamaterials, made this goal possible, at least at infrared frequencies (780 nm) only ve years after the creation of the rst metamaterial [16]. Of course, some critical problems still remain unsolved. The key one is that most potential applications of metamaterials cannot be achieved if there are signicant losses in the metamaterials. [19, 20]. The high losses originate from the design required to create metamaterials and particularly are a critical problem at the infrared and optical frequencies since the metal required for fabrication departs from a perfect conductor in the optical domain. The losses of the best designs to date (see table 1.1) are too high for most of the potential applications to become practical (e.g. see [49]). Another problem in the fabrication of metamaterials is the development of truly large-area bulk 3D isotropic metamaterial structures. Although recently Valentine et al. reported to have successfully fabricated a 3D metamaterial structure at infrared frequencies [46], the proposed metamaterial structure is not a true 3D metamaterials because it only works for light coming from a narrow range of directions and with a single polarization [47]. Fig. 1.3 gives a brief history of the development of metamaterials. The early metamaterials were built from double split ring resonators and parallel wires. Changing the double SRR to a single one and reducing the size of the structure, the metamaterial frequency value reaches about 200 THz. However, by this approach, one cannot
CHAPTER 1.
INTRODUCTION
Double SRR
Single SRR
U-shape SRR
Plate pairs
Double -fishnet
go beyond this frequency because the composition of SRR and parallel metallic wires are so sensitive at nano scale and it is very dicult to measure transmission and refraction along the direction parallel to the SRR plane; moreover, the metal of the SRR strongly departs from an ideal conductor at and above this frequency. Therefore, in 2005, several groups employed alternative designs to create metamaterials at higher frequencies [17, 18]. The new designs contained pairs of metal wires or plates which were separated by a dielectric spacer. Wire pairs can behave like SRRs and give a negative permeability,
and
are negative with only wire pairs, is dicult. Therefore, a new design,
the so-called double-shnet, which contains a pair of metal shnets, was proposed in 2005 [19]. Using this structure, several groups reported negative-index metamaterials near the visible spectrum [16, 19, 20]; however, the proposed metamaterials suer from high losses that originate from the metal in the structure. To provide a
CHAPTER 1.
INTRODUCTION
Structure used
2005:
Purdue
April 13 (2005) arXiv:physics/0504091 Opt. Lett. (2005) April 28 (2005) arXiv:physics/0504208 Phys. Rev. Lett. (2005)
0.3 2
1.5 m 2.0 m
0.1
0.5
2006:
UNM & Columbia Karlsruhe & ISU Karlsruhe & ISU
J. of OSA B (2006) OL. (2006) OL (2007) OL (2006)
Purdue
OL (2007)
Nano-fishnet
picture of the status of metamaterial at present, we summarize the recent successful observations of negative-index refraction with corresponding gure of merit by active groups in this eld in table 1.1. The gure of merit (FOM) is a suitable measure for the loss dened as the ratio of the real to the imaginary part of the refractive index. So far, the largest FOM for infrared frequencies is around 1 and for THz frequencies is around 3.
conrming this property in metamaterials (e.g. [15]); moreover, several groups have
CHAPTER 1.
INTRODUCTION
mathematically shown that the index of refraction should be negative in metamaterials (e.g. [27]). In this section, for the sake of completeness, we show why the index of refraction is negative in metamaterials. We consider a plane wave propagating in a metamaterial, with electric and magnetic elds
(1.1)
k E = i H, c
where
k H = i E, c
(1.2)
and
spectively and
k2 =
2 . c2 k
(1.3)
E, H,
and
are simultaneously negative, and form a right-handed set of vectors when and
only
Next, we will
show that the phase velocity of a wave (represented by the wave vector) is opposite to the movement of the energy ux (represented by the time-averaged Poynting vector) of the wave in a metamaterial medium. We consider an electric eld polarized in the
p direction, E = p.
Using Maxwell's
CHAPTER 1.
INTRODUCTION
1 S = Re(E H ), 2 = c p (k p), 2 ck . 2
(1.4)
The result shows that the time-averaged Poynting vector is in the opposite direction of the wave vector if electric eld when
is negative.
and
, the phase velocity of a wave is always opposite to the movement of the energy ux
of the wave. It is not evident from
n2 =
that having negative
k2 = 2 k0
(1.5)
and
Both negative
and positive index of refraction solutions satisfy Maxwell's equations and boundary conditions for refraction of light. However, the negative solution for the index of
refraction is selected in metamaterials to satisfy the requirement that in the refracted beam the energy ows away from the interface. The choice of negative refractive
index in metamaterials is discussed in detail in the literature (e.g. [27]). Here, we provide a simple argument to demonstrate negative index property in a passive lossy metamaterial.
CHAPTER 1.
INTRODUCTION
10
In a lossy metamaterial,
and
= + i = + i ,
where
(1.6)
< 0, < 0
and
> 0, > 0
medium is absorbing. Using Eq. (1.5), the index of refraction for a passive medium with small absorption is
n = = ( ) + i( + )
+ [ + i 2 ].
The (1.7)
< 0, < 0
and
> 0, > 0
conditions lead to
+ < 0. 2
(1.8)
Since the imaginary component of refraction index is positive in a lossy medium, Eq. (1.8) demands that the negative sign of the square root in Eq. (1.7) be chosen. Therefore, the real component of index of refraction must be negative if both
and
are negative. In the reminder of this chapter, we provide a brief overview of some of the po-
tential applications of metamaterials proposed in the literature, emphasizing those applications related to metamaterial waveguide structures and the multilayer systems. It is not our aim to give a picture of all of the applications of metamaterials, but with our examples from the recent literature, we will provide a background to better understand the results of the following chapters.
CHAPTER 1.
INTRODUCTION
11
for example, the metal waveguide can guide only transverse magnetic, TM, waves while the metamaterial slab can guide both polarizations of light (the evidence of TE surface polariton modes was found in 2005 by Smith [22]). Thus, the existence of surface polariton modes in metamaterials opened the door for further applications [23]. Ruppin's work, drew more attention to metamaterial waveguide structures [23, 24]. Further study showed that not only can the metamaterial waveguide support surface polariton modes but also there are new types of modes (complex surface or leaky modes) with immediate applications to leaky antennas [26]. Moreover, it was shown that the properties of the normal guided modes of the metamaterial waveguide are completely dierent than in a conventional waveguide. For example, Nader Engheta and Andrea Alu showed that there is no cut-o thickness for the rst mode of metamaterials waveguide. Thus, this feature provides a solution to the problem of energy transmission with lateral cross section below the diraction limits [5]. Moreover, Yuri Kivshar and his coworkers showed that the fundamental mode (the mode with no node in the modal prole) does not exist in a metamaterial waveguide. By using
this feature, we can introduce the concept of surface wave suppression which leads to enhance radiation eciency for microstrip antennas [26] or introduce a one dimensional periodic waveguide with a complete band gap [28]. Usually, we require 3D
CHAPTER 1.
INTRODUCTION
12
photonic crystals (PhC) to have complete band gap; however, Kivshar and his team theoretically showed that 1D PhC containing matamaterials have this property. A multilayer system containing metamaterials has a very important property which was discovered by Li et al. in 2003 [29]. The proposed structure supports
new types of transmission band gaps , so-called zero-n gaps, with unique properties. The novel gap is completely dierent from the conventional Bragg reection gap, as it is insensitive to the direction or polarization of the beam [30]. Another interesting problem in the 1D PhC containing matamaterials, is the investigation of the spontaneous emission of an atom embedded in the structure [31]. It is shown theoretically that the spontaneous emission of an atom can be completely suppressed (modied Purcell eect) because metamaterials provide the possibility of vanishing optical path length between two points [32]. It is not an exaggeration to state that the vortex-like energy ux structure [24] (see Fig. 3.1) is the most important property of a metamaterial waveguide structure. Almost all unusual properties of metamaterial waveguides such as eld localization [25], super-waveguiding (a theoretical design which can transmit light with high-power density [33, 34]) and trapped-rainbows (an idealized structure which could stop light over broadband range of frequency [35]) come from this feature. This phenomena
happens at the interfaces of the metamaterial waveguide because the energy in a metamaterial waveguide core ows in the opposite direction to that in the cladding. One can potentially slow down or even stop the light passing through a metamaterial waveguide structure due to the vortex-like energy structure. This concept was rst discussed by Y.Kivshar and coworkers [24] in 2003 and was investigated in detail by He et al. [38]. However, the idea lay dormant until Tsakmakidis et al. proposed
CHAPTER 1.
INTRODUCTION
13
in 2007 a theoretical wedge-shaped metamaterial waveguide structure which could stop light over a broadband range of frequencies at room temperature producing a so-called trapped-rainbow [35]. It is interesting to note that Tsakamakidis et al.
provided another approach to explain the stopped-light mode phenomenon in the metamaterial waveguide rather than vortex-like energy ux approach. They showed that by controlling the thickness of the waveguide, the optical path of the light could be set to zero because of the negative Goos-Hanchen shift in metamaterials [36, 37]. Because of the exotic properties of the theoretical structure proposed in [35], considerable interest arose in the literature, and several groups have designed dierent theoretical structures to create trapped-rainbows based on the work of Tsakmakidis
et al.; for example, the trapped-rainbows are proposed in a hybrid metamaterial-
photonic crystal structure [43], in a surface plasmon in a graded metallic structure [39], in graphene for an electronic system [44], and in an anisotropic metamaterial waveguide structure [41, 42]. We should mention here that the idea of slow light in an anisotropic metamaterial waveguide was rst demonstrated by Narimanov and Alekseyev one year before the idea of trapped-rainbows [40]. They showed not only can light be slowed down in an anisotropic metamaterial waveguide but the velocity of a particular slow-light mode is almost constant over a broadband range of frequencies. All papers mentioned up to this point, astonishingly, neglect the eect of material loss, which is an inherent feature of metamaterials (see table 1.1), in their proposed structures. As we have previously discussed in this chapter, all currently methods
to create negative-index metamaterials are based on the use of metallic structures and all real metals introduce some losses particularly at optical frequencies (see section 1.2). One might propose to use the other methods rather than to use lossy
CHAPTER 1.
INTRODUCTION
14
metallic structures to create negative-index metamaterials. One could employ a photonic crystal in its negative-index regime or introduce chirality or optically amplifying materials (gain mechanism) to remove losses in negative-index metamaterials. However, recently, Stockman showed that loss with dispersion are the inherent features of negative-index materials and it is impossible to compensate loss completely over a broadband rang of frequencies [48]. To ensure a positive energy density (see Chapter 3) and to satisfy causality, any negative-index material must have dispersion. Thus, to satisfy Kramers-Kronig relations, it should have nite loss at some frequencies.
of the eect of the negative-index refraction of metamaterials on dispersion, and to present normalization and orthogonality relations. In particular, it is shown that the properties of the dispersion curves of metamaterial waveguide are quite dierent from those of a conventional waveguide. In this chapter, we also study the role of dispersion and absorption on the modes of a metamaterial waveguide structure. We show that the dispersion curves are signicantly changed in the presence of intrinsic loss. The purpose of Chapter 3 is to study the slow-light modes in a metamaterial
CHAPTER 1.
INTRODUCTION
15
this chapter is to provide the correct forms of power ux, energy density, and energy velocity in a dispersive and lossy medium. As discussed above, the topic of the slowlight mode in metamaterials has received much attention in the literature after the idea of trapped-rainbows proposed in [35]. The trapped-rainbows claim is examined by performing a simple and clear numerical analysis in a realistic lossy metamaterial waveguide structure. Chapter 4 is concerned with addressing the possible methods for the excitation of the guiding modes. The contribution of the rst part of this chapter is to introduce the transfer matrix formalism in a multilayer containing metamaterials with loss. Next, we provide an understanding of the attenuated total reection method (ATR) for exciting the modes. We then discuss the limitations of the ATR method and the conditions needed to optimize the accuracy of the ATR method. The lateral GoosHanchen shifts and their relation with the guiding modes are also covered in this chapter. Chapter 5 is concerned with addressing the modal characteristics of a metamaterial waveguide structure using a complex-frequency approach with a real propagation constant. In this chapter, we make the connection between the complex frequency approach with the complex propagation constants approach. We indicate that the
dispersion curves of the metamaterial waveguide structure are dierent in these two approaches. The other topic covered in this chapter includes perturbation theory. We provide a comparison between the exact solutions with the solutions calculated by perturbation theory.
CHAPTER 1.
INTRODUCTION
16
In Chapter 6, we summarize previous chapters main points and give some suggestions for future works.
Chapter 2
2.1 Introduction
In the recent years, many investigations have been performed on metamaterial waveguide structures (e.g. [23, 24, 53, 57]) due to their unique properties, such as slow-light modes, backward propagation electromagnetic waves, and surface polariton modes. These phenomena are potentially important for devices; however, the role of intrinsic loss which is an inherent feature of negative-index metamaterials [48], has been largely neglected in previous studies. In this chapter, we derive the dispersion equations
and electromagnetic eld distributions in planar metamaterial waveguide structures. First, we consider a symmetric waveguide and then extend our results to asymmetrical waveguide structures. Although we only study planar-geometry waveguide structures, the presented results can be straightforwardly generalized to waveguides with dierent geometries [63, 64]. In section 2.3, we derive the orthogonality relations between 17
CHAPTER 2.
18
By exploiting
the orthogonality relations, we discuss the normalization of electromagnetic eld amplitude in metamaterial waveguides. We then compare the novel properties of the
proposed metamaterial structures with that of conventional waveguide structures in section 2.4. Finally, we investigate the behavior of the bulk and complex modes using three dierent models of the material in the structure: i) dispersionless and lossless, ii) dispersive and lossless, and iii) dispersive and lossy models. Our computational calculations indicate that material loss modies signicantly the dispersion curves.
-d
Region(I) Region(II)
d
Region(I)
2d
Figure 2.1: Schematic geometry of (a) symmetric and (b) asymmetric metamaterial waveguide structure. An innite metamaterial slab of thickness 2d and constitutive parameters
2 < 0
and
2 < 0
Cartesian coordinate system is employed. The light propagates along the the structures are independent of
axis , and z
axis.
CHAPTER 2.
19
metamaterials. First, we derive the electric and magnetic elds distributions of metamaterial waveguide structures by using Maxwell's equations and boundary conditions [50, 51, 52] for transverse electric (TE ) and transverse magnetic (TM ) modes. We then obtain the dispersion relations for the proposed structures. Finally, we discuss the physical and unphysical solutions of the dispersion equation. The physical so-
lutions satisfy guidance conditions. But the unphysical solutions do not. Although the form of the dispersion equations are the same in metamaterial waveguide and conventional waveguide structures, the physical solutions in these two structures are dierent. We consider an isotropic homogeneous slab waveguide in a planar geometry for two dierent structures: i) a symmetric waveguide structure with the negative-index core located in free space [Fig. 2.1(a)] and ii) an asymmetric waveguide structure [Fig. 2.1(b)]. The core, with a thickness of
(2.1)
E = 0, H = 0,
where c is the speed of light in vacuum and
E = i H, c H = i E, c
and
(2.2)
meability, respectively. We also assume electromagnetic elds that are independent of the y-component in our proposed waveguide structure and that the propagating
CHAPTER 2.
20
modes can be represented as either transverse electric modes (TE ), or transverse magnetic modes (TM ). Thus, the y-component of the TE modes,
E y, can be expressed as
E y(x, z) = E(x)exp[i(z)], z,
and
where
is the angular frequency of the eld. The time-harmonic wave equation can
be straightforwardly obtained by using Eq. (2.2). From the time-harmonic electromagnetic wave equation,
1/(r)
{1/(r)
(2.3)
the propagating TE modes of the electric eld are described by the following scalar equation for the proposed waveguide structure
(2.4)
E(x, z) = y E0 exp[iz]
A exp[k1 (x d)]
(x > d)
(2.5)
where
k1 =
2 1 1 2 /c2
k2 =
2 2 2 /c2 2
determined by the boundary conditions. As the propagating modes in this waveguide structure may be split into even and odd modes, the electric eld of the even and odd
TE modes are
CHAPTER 2.
21
cos(k d) exp[k (x d))] (x > d) 2 1 cos(k2 x) cos(k d) exp[k (x + d)] 2 1 sin(k d) exp[k (x d)] 2 1 (d < x < d) (x < d), (x > d)
(2.6)
(2.7)
H,
Ey
E = i H, c
as
even
cE0 e[iz] 1 (x, z) = x z 2 [ cos(k2 x) + ik2 sin(k2 x)] 1 [ik z + x] cos(k d) exp[k (x + d)],
1 1 2 1
(2.8)
Hodd (x, z) =
cE0 e[iz]
1 [ik z + x] sin(k d) exp[k (x d)] (x > d) 1 1 2 1 1 x z (d < x < d) 2 [ sin(k2 x) + ik2 cos(k2 x)] 1 [ik z + x] sin(k d) exp[k (x + d)], (x < d) 1 2 1 1
(2.9)
Employing the boundary conditions, the tangential components of the electric, magnetic,
E, and
H,
x = d
and
x = d,
CHAPTER 2.
22
relations for the even and odd TE modes [50, 51, 52]
k2 d k1 = tan(k2 ) 1 2 2
(2.10)
k1 k2 d = cot(k2 ). 1 2 2
Finally, the results for the magnetic transverse wave, TM, can be found from duality by changing
and
E H .
k1 k2 d = tan(k2 ) 1 2 2
(2.11)
k1 k2 d = cot(k2 ). 1 2 2
These dispersion equations are valid for any complex or negative value of
and
Hence, they also work for the lossy double-negative index planar geometry waveguide structure. We can easily generalize the results to the asymmetric waveguide of Fig. 2.1(b) [51]. The electromagnetic elds of the asymmetric waveguide have the form,
Ey (x, z)
E0 exp[iz]
sin( 2k d) exp[k (x 2d)] (x > 2d) 2 1 sin( + k2 x) sin() exp[k x], 3 (0 < x < 2d) (x < 0)
and
CHAPTER 2.
23
H(x, z) =
cE0 e[iz] 1 x z 2 [ sin( + k2 x) + ik2 cos( + k2 x)] 1 [ik z + x] sin() exp[k x],
3 3 3 2 2 1 1 k0
1 [ik z + x] sin( 2k d) exp[k (x 2d)] (x > 2d) 1 1 2 1 (0 < x < 2d) (x < 0)
(2.12)
where
k1 =
and
k3 = k2 =
2 2 3 3 k0
are the wave numbers in the two is the wave number in the core,
2 2 2 k0 2
k0 = /c
can be obtained
tan( k2 d) = k2 /k1 .
(2.13)
The dispersion equation for the asymmetric waveguide of the Fig. 2.1(b) can readily be put in the form
2 2d 2 2 k0 2 = arctan(
2 1
2 2 1 1 k 0 2 2 2 k0 2
) + arctan(
2 3
2 2 3 3 k0
2 2 2 k 0 2
) + m/2,
(2.14)
2 2d 2 2 k0 2 = arctan(
2 1
2 2 1 1 k 0 2 2 2 k0 2
) + arctan(
2 3
2 2 3 3 k0 2 2 2 k0 2
) + m/2,
(2.15)
m = 0, 1, 2, ... E y (x)
represents the number of the nodes in the The roots of the above transcendental
eld.
CHAPTER 2.
24
due to the intrinsic loss in the structure or the light leakage below the total internal reection angle (see section 2.5.3). transverse wave number equations, e.g. The double-value of the square root in the
ki = k2 ;
2 2 i i k0 ,
problem. It is not important which sign is chosen for the propagation constant, for the transverse wave number in the core,
or
wave number in the cladding has signicant physical meaning. If the real part of it is negative, this leads to a violation of the guidance condition, which requires that the eld amplitudes vanish in the transverse direction at
x .
E0 ,
which can be obtained from eld normalization. Before we normalize the elds, we need to derive the orthogonality relations in the metamaterial waveguide structures. First, we assume the permeability and permittivity are real (lossless model). We note that the negative value of
[60] does not work to nd orthogonality relations in metamaterial structures. We use the Generalized Hermitian Hamiltonian technique [58, 59] to nd the orthogonality relations in a metamaterial waveguide structure. The mathematical details are given in appendix A. The orthogonality relation is
1 2
(2.16)
CHAPTER 2.
25
where N is the normalization parameter. We can interpret it as a total power ow carried by a specic mode when
= .
now can derive the electric eld amplitude of the proposed waveguide of Fig. 2.1(a) ,
E0 ,
sym E0 =
2 c d +
2
2 1 k1
2 2 k1 +k2 2 k 2 +2 k 2 2 1 1 2
(2.17)
k1 and k2
and the core respectively. We note that the normalized eld amplitudes of the metamaterial waveguide structure may diverge for particular sets of parameters. How-
ever, this unphysical divergence disappears when we consider a realistic metamaterial waveguide structure with material loss. In some recent work (e.g. [35]), the Authors insert the absolute value of
malized eld amplitude; however, we lose orthogonality relations if we normalize the eld amplitude with respect to
The normalization and orthogonality relations derived in this section are not valid for a lossy metamaterial structure with complex
and
metamaterial waveguide structure. However, we will use the orthogonality relations for a lossless waveguide to derive the approximate dispersion curves of the lossy metamaterial waveguide by using perturbation technique (see Chapter 5).
CHAPTER 2.
26
k2 d k1 = tan(k2 m ) 1 2 2 2
2 2 k1 + k2 =
(2.18)
2 d2 (2 2 1 1 ) R2 . c2
(2.19)
We exploit the graphical method to determine the solutions of the transcendental equation. Fig. 2.2 shows the Riemann sheet for the planar waveguide structure. The vertical axis represents the real part of the wave number in the cladding,
k1 ,
and the
horizontal axis shows the real and imaginary parts of the wave number in the core,
k2 .The
intersection of the solid curves, Eq. (2.18), and the dotted curves, Eq. (2.19),
determine the solution of the guided modes. The roots located in the half top of the Riemann sheet satisfy the guidance condition since their elds exponentially decay in the cladding due to positive value of the
k1
2.2(a)). However, the roots in the half bottom of the Riemann sheet are improper
CHAPTER 2.
46
CHAPTER 2.
27
Re[k1]
2
TE0
TE1
0 2 2 0 2
(C) Re[k2]
Img[k2] (A)
2
2 0
(B)
0 2 2 2 2 0 2 0 2
2 2
2 0
TE2
0
i/2
/2 Re[k1]
2
TE2 (B)
1 2
(A)
0 2
0 2 2 0 2
Img[k2]
2
Re[k2] (C)
0 2
0 2 2
TE0 i/2
0
TE1
/2
Figure 2.2: Graphical solution of the TE modes for (a) the dielectric waveguide with
==4
and
The intersection of the solid curves, Eq. proles of the guided modes. (2.18), and the dotted curves, Eq. (2.19), Insets show the transverse determines the solution of the guided modes. Insets show the transverse proles of the guided modes for
d = Figure 2.2: Graphicalthe metamaterial waveguide dielectric waveguide with. 2 cm , and (b) solution of the TE modes for (a) the with = = 4
==4
and
d = 2cm
= = 4.
E y (x)
eld.
solutions because their eld amplitudes exponentially grow in the cladding (e.g. see the inset (A and B) in Fig. 2.2(a)). The right half of the Riemann sheet determines the oscillatory modes and the left half represents the evanescent or surface polariton modes. By looking at the transcendental equation, we see that the improper roots of a dielectric waveguide, which are lying on the bottom of the Riemann sheet, are proper solutions for metamaterial waveguides since the negative sign of
in Eq.
(2.18)
leads to a reection of the Riemann sheet about the horizontal axis, (compare Fig.
CHAPTER 2.
28
2.2(a) and (b)); therefore, the guided modes of the metamaterial waveguide behave as the improper modes of the conventional waveguide. In the past, the properties of the improper conventional waveguide modes, usually termed leaky modes, have been studied because of their useful applications such as leaky wave antenna, and leaky wave lters. Several new properties may be observed in the metamaterial waveguide structure [23, 24]. First of all, the metamaterial waveguide supports the surface polariton modes for both the TE and TM polarizations. By looking at the Riemann sheet, we notice that there is no solution in the second quadrant for the dielectric waveguide structure; therefore, the conventional waveguide cannot support surface polariton modes. At
the metallic interface, only the magnetic transverse, TM, modes can be supported; however, a metamaterial waveguide can support both TM and TE surface polariton modes; moreover, it supports complex modes. The complex modes are the complex solution of the dispersion equation above the cut-o frequencies. wave number, The propagation
We can assume that the complex mode represents the wave propagation below the angle of total internal reection in the core; therefore, the wave leaks out of the waveguide and the imaginary parts of the complex modes indicate the damping factor of the leakage. Analytically, with a straightforward calculation, it can be shown that the complex modes are always improper solutions of the dispersion equation in a conventional waveguide [52]; however, a lossless metamaterial waveguide structure always supports proper complex modes [56]. By using the Eq. (2.10) and separating
CHAPTER 2.
29
the real part and imaginary part of the wave numbers, we have
r k1 =
1 y tanh(y)[1 + tan2 (x)] + x tan(x)[1 tanh2 (y)] 2 tan2 (x) + tanh2 (y)
(2.20)
i k1 =
1 x tanh(y)[1 + tan (x)] y tan(x)[1 tanh (y)] , 2 tan2 (x) + tanh2 (y)
and
where
r i k1 = k1 + ik1
k2 = x + iy.
is a real
r k1
a positive value (the dielectric waveguide case), the real part of the wave number in the cladding is always negative which leads to an unphysical solution; and if
<0
(the metamaterial waveguide case), the real part of the wave number in the cladding is always positive and all complex mode solutions are on the top Riemann sheet. Second, it is possible for two modes to have the same number of nodes since one of them propagates in the same direction of the group velocity, (forward propagating mode), and the other propagate in the opposite direction of the group velocity, (backward propagating mode); hence, both forward and backward modes are supported by the metamaterial slab. Third, the rst bound mode,
metamaterial waveguide structures. This property gives an opportunity to introduce one dimensional periodic waveguides with a complete band gap [28], introduce the concept of surface wave suppression which leads to enhance radiation eciency for microstrip antennas [26], and introduce the slow-light modes under the single mode operation [54]. However, in the realistic metamaterial waveguide, which should be
dispersive and lossy, and also by considering the role of the complex modes, we cannot see this property anymore.
CHAPTER 2.
30
Finally, not only does the metamaterial waveguide support evanescent and oscillatory eld distributions, but also supports linear eld distribution. The point (A) on the Fig. 2.2(b), where the root is located on the top vertical axes of the Riemann sheet,
k2 = 0,
T E1
or
cent modes. The electric eld distributions for the TE mode, considering boundary condition in the Eq. (2.5), is
2 k1 d 1 exp[k1 (x d)] (x > d) Ey (x, z) = E0 exp[iz] 2 k1 x (d < x < d) 1 2 k1 d exp[k1 (x + d)] (x < d), 1
which leads to constant longitudinal magnetic eld for these modes.
(2.21)
The constant
longitudinal magnetic eld gives more degrees of freedom for potential applications in a metamaterial waveguide. [61]
In this section, we numerically study the behavior of dispersion curves for three different material models of metamaterial waveguide structures: i) lossless and dispersionless ii) lossless and dispersive iii) lossy and dispersive models. We know the rst
CHAPTER 2.
31
two models are unrealistic since these models violate the causality principle and positive energy density [48]. However, we investigate these two models for two reasons. First, almost all previous studies in the literature employ one of these two models. Thus, we can compare our results with others and also compare the properties of the dispersion curves between metamaterial and conventional waveguide with the same material models. Second, by comparing results from these models with the realistic model, lossy and dispersive model, we can highlight the eects of loss on the dispersion curves. Our analysis indicates that loss signicantly modies the dispersion curves and some astonishing properties in unrealistic metamaterial waveguide, such as a stopped-light mode, are destroyed or disappear in the realistic material model [49]. We begin by investigating the properties of the dispersion curves for the lossless and disprsionless metamaterial waveguide. We consider the symmetric metamaterial waveguide structure, illustrated in Fig. 2.1(a), with a thickness of free space, Fig.
d = 2 cm
and
in the
1 = 1, 1 = 1,
2 = 4,
2 = 4.
2.3(b) indicates the dispersion curves for the three lowest transverse electric,
TE, modes for the proposed structure. To compare the properties of the dispersion
curves for the metamaterial waveguide with a conventional waveguide, we plot in Fig. 2.3(a) the dispersion curves of a conventional waveguide with
2 = 4,
and
2 = 4.
The rest of the parameters are chosen to be the same as the proposed metamaterial waveguide. Although the dispersion curves of the metamaterial and conventional
waveguide are the solutions of the same transcendental equations (2.10,2.11), Fig. 2.3 clearly indicates that the dispersion curve characteristics are dramatically dierent for the metamaterial and dielectric waveguide structures . As mentioned before, this
CHAPTER 2.
32
dierence comes from the fact that the physical dispersion equation roots for the metamaterial waveguide are the unphysical roots for the conventional waveguide. From our calculations, we can nd several signicant dierences between the conventional and metamaterial waveguide structures. First, the metamaterial waveguide structure supports new types of modes: so-called surface polariton modes and complex leaky modes. Hence, as shown in Fig. 2.3(b), a rich variety of modes (bound, surface polariton, and complex modes) exist in the metamaterial waveguide structure. The bound modes are restricted between the vacuum light-line and metamaterial lightline. We plot
T E1 , T E2 ,
and
T E3
T E1
T E1
T E2 ,
and
T E3
modes behave like a bound mode when the propagation constant of the dispersion curve is purely real and behave like a complex leaky mode when the propagation constant is complex. Although the proposed structure is lossless, the propagation
constant has a complex value for complex modes. In the complex modes, the energy is not conned in the core and leaks out through the cladding. In Fig. 2.3(b), the dispersion curve of a complex
T E2
the cut-o frequency where the bound mode curve is at, e.g. at f = 2.8 GHz, and goes to lower frequencies, and the dispersion curve of a complex mode with a imaginary propagation constant, shown by the thin curve in the Fig. monotonically by decreasing the frequencies. 2.3(b), increases
energy in the lossless metamaterial waveguide structure [52], they do not play an important role in the structure. Second, we have domains of frequencies where there are
CHAPTER 2.
33
no bound modes; for instance, there is no bound mode in the range of 2 and 3 GHz in Fig. 2.3(b). Third, in contrast with the dielectric waveguide modes, the modes in the metamaterial waveguide are not restricted to lie between the vacuum light-line and the metamaterial light-line. For example, the complex modes go above the vacuum light-line and the
T E1
note that the dispersionless and lossless metamaterial is a non-realistic model since it violates causality principle; for example, in our dispersion curve model, one can nd the points where the group velocity exceeds the speed of light in vacuum and even goes to innity! We discuss this issue in the next chapter in detail, where we consider
CHAPTER 2. MODAL CHARACTERISTICS OF MWS
47
(a)Light line
Frequency (GHz) TE
2
(b)Light line
5
TE3
Frequency (GHz)
TE1
2
TE2
TE0
TE1
0 0
0.5
1.5
(1/cm)
2.5
3.5
0 0
0.5
1.5
(1/cm)
2.5
3.5
Figure 2.3: The dispersion curves for the dielectric and metamaterial frequency independent waveguide structure with the thickness of with
=4
and
= 4.
thick and thin curves correspond to the real and imaginary components of
= 4. .
The
Figure 2.3: The dispersion curves for the dielectric and metamaterial frequency inde-
CHAPTER 2.
34
2.5.2
We next consider the symmetric waveguide of Fig. dispersive, lossless metamaterial core.
dispersion curves for this model, we describe the complex frequency-dependent electric permittivity and magnetic permeability by using the standard Drude and Lorentz Models [62]. These are given respectively by
() = 1
2 p , 2 + ie
(2.22)
() = 1
F 2 , 2 2 0 + im
(2.23)
where F is the fractional area of the unit cell occupied by the interior of the split-ring,
and
are respectively the electric and magnetic broadening constants associated with loss. We plot the permeability and permittivity in Fig. rameters: 2.4 with the following pawhich are
chosen to be in agreement with recent experiments [62]. Fig. 2.4 indicates that in the range of frequencies between 4 and 6 GHz, the proposed model behaves as a doublenegative-material (DNG) with both negative permeability and permittivity. Below 4 GHz frequency and above 6 GHz Frequency, it behaves as a single-negative-material (ENG) or plasmonic media. Although the model provides a negative-index refraction condition between 4 and 6 GHz, only the range of frequencies between 4 and 5.2 GHz satises the guiding condition waveguide structure with
ncladding = 1.
CHAPTER 2.
35
3 2 1 0 1 2 3 4 ENG 5 4
= 0.1 =0
fsm
fsp
ENG
8 9
f(GHz)
Figure 2.4: The permeability and permittivity of the metamaterial slab as a funcp = 10 GHz, 0 = tion of frequency using Lorentz-Drude models with F = 0.56, 2 2 4 GHz, and = 0.1 GHz. The solid curves show the real part of and , and the dashed curves show their imaginary part. Im[] and Im[] are scaled up by a factor of 20. If the metamaterial slab is in the free space,0 polariton frequency, and 7.03 GHz, respectively.
= 0 = 1, the magnetic surface fsm , and the plasmon surface polariton frequency, fsp , are at 4.71
and
2 ,
lowest-order TE modes of the proposed metamaterial waveguide structure without loss in Fig. 2.5(a). When we compare the dispersion curves of Fig. 2.3(b) and Fig. 2.5(a), we note that the dispersion curves qualitatively rotate 180 degree when we include dispersion in our model. The dispersion curves have a cuto frequency that is an upper bound. However, the dispersion curves of the dispersionless waveguide have a cuto frequency that is a lower bound. Also, in the dispersive case higher-order
CHAPTER 2.
36
modes appear at lower frequencies rather than higher frequencies. Fig. 2.5(a) shows that for higher-order modes, the frequency depends only weakly on the propagation constant,
and
f () fres = 4
frequency.
CHAPTER 2.
5.2
TE0 TE1
(a)
frequency (GHz)
4.7
4.4
4 0
fres
TE0
(b)
frequency (GHz)
TE1
4.7
4.4
4 0 1 2 3 4 5 6
fres
Figure 2.5: The(b) the lossy dispersive metamaterial waveguide structure with the parameters given lossless and dispersion curves for the rst four TE modes for (a) the
in Fig.4 for 2 and 2 . The 2-cm-thickness metamaterial slab is in (b) the lossy dispersive metamaterial waveguide structure withthe freeparameters given the space
as shown in Fig.1 (a). The thick and thin curves represent the real and imaginary
2 of the 2 The 2-cm-thickness metamaterial slab is in free part and propagation constant, ; and the dashed curves indicate the light-line and
metamaterial light-line. 2.1(a). The thick
space as
and thin curves represent the real and imaginary and the dashed curves indicate the light-line and
As in the non-dispersive structure, this structure does not support the nodeless
CHAPTER 2.
37
bound mode. But, the nodeless surface polariton and complex modes,
observed. In Fig. 2.5, the nodeless surface polariton mode starts from where
ncore (fn ) = 1, and goes to innity at the magnetic surface polariton frequency,
4.714 GHz, where
fms =
(fms ) + 0 = 0. T E0
mode. As the propa-
The
T E1
gation constant,
, increases, the T E1
polariton mode. There is a transition point on the metamaterial light-line between the
T E1
distribution is linear or the magnetic eld is constant in the core of the waveguide structure. Thus, the
T E1
bound mode, complex mode, surface polariton mode, at dierent Although the shape of waveguide's core, the
().
T E0
T E1
Nevertheless, it is important to note that in principle there is no limit on the reduction of the slab thickness [66]: all the modes are preserved by changing the thickness of the slab. This feature, which is contrary to what we have in a conventional waveguide, leads to possibility to design ultra-thin slab waveguides. When we decrease
the slab thickness, the dispersion curves shift to lower frequencies and concentrate near resonance frequency region. However, in the dispersionless waveguide structure, modes shift to higher frequencies with decreasing slab thickness . In summary, our simulations indicate that many of the exciting properties of dispersionless and lossless metamaterial waveguide structure are preserved for the dispersive metamaterial waveguide structure although the dispersion changes the dispersion curves of the bound modes.
CHAPTER 2.
38
2.5.3
We now study the dispersion curves behavior of realistic metamaterial waveguide structures including both dispersion and loss. We nd several new properties in a
metamaterial waveguide structure when loss is included, such as a splitting of the complex modes into two curves, coupling the surface polariton modes with higher order modes, and continuing dispersion curves into frequencies below the resonance frequency . We study these properties in detail in the following paragraphs. We now consider the proposed structure in the previous subsection but with a loss parameter of
= m = e = 0.1 GHz.
Fig. 2.5(b) shows the dispersion curves The thick curve corresponds
for the rst ve TE modes in the presence of loss. to the real parts of the propagation constant, the imaginary part of
splits the complex modes into the two modes: a complex backward propagation mode and a complex forward propagation mode. We cannot ignore the importance of the complex modes in a lossy waveguide structure since the complex modes carry energy. The average energy ux in the guidance direction is always zero for the complex modes of a lossless waveguide structure [52]. The imaginary part of imaginary part of
for the backward mode has a negative value while the We note that the In
absolute values of the imaginary parts of the modes is plotted in our graphs.
a lossless waveguide structure, the real parts of the forward and backward complex modes overlap with each other and the imaginary part of them have the same value with dierent sign. But, the loss breaks this symmetry as shown in Fig. 2.5(b). The forward and backward complex curves of the particular mode have approximately the
CHAPTER 2.
39
same behavior except near the magnetic surface polariton frequency region. Second, the surface polariton modes,
instead they join to higher order modes with nite wave number, For example, Fig. 2.6 shows the coupling of
T E4
mode.
In Fig.
T E0
fms = 4.71 T E4
GHz when we insert loss in the structure and couples with the complex
modes. The thick and thin light blue curves represent the real and imaginary
T E4
mode for the lossless structure, and the thick and thin red
structure respectively in Fig. 2.6. We note that below 4.75 GHz and above 4.68 GHz, the inuence of material loss on mode structure is highly non-perturbative; however, the red curve approaches the blue curves above 4.75 GHz and below 4.68 GHz. To have a better insight into this phenomenon, we plot the normalized electric eld at several points on the
T E5
and
T E1
to its amplitude in the core. The rst point at 4.05 GHz is on the bound
T E5
mode.
As predicted, the electric eld has ve nodes inside the waveguide core at this point. The second point is chosen on the complex
T E5
has ve dips instead of nodes at this point. When frequency increases, the amplitude of these dips gradually decrease. Finally, four of them disappear at
fms = 4.71
GHz.
At this frequency, the electric eld is sharply conned at the waveguide surfaces as the normalized electric eld intensity approaches the electric eld at 4.85 GHz on the
2 E 2 /E0 = 104 !
Finally, we plot
T E1
curves.
has only one node at this frequency. We conclude that the backward complex
T E5
CHAPTER 2.
40
polariton frequency region. However, there is no specic boundary between the and
T E1
modes.
4.78
TESP 0
fms
4.66 0 2 4
(1/cm)
10
Figure 2.6: The surface polariton mode coupling with the higher order mode. The blue curves shows the
T E0
The dashed lines represent light-line and metamaterial light-line. Note: The forward
T E4
and
T E0
Third, the dispersion curves go below the resonance frequency when we introduce material loss in the waveguide structure. The modes of the lossless waveguide touch at
fres = 4
GHz as
T E4
However, the TE modes of the lossy waveguide bend near resonance frequency and go below it until
Re[] 0.
CHAPTER 2.
41
5.2 5
1 0.5 0 2 2 1 0 2 0 2 x 10
4
4.8 f(GHz)
2
4.6
1
4.4 4.2
0 2
0.5 0 2
TEB
1
4 0
TEB
Figure 2.7: The eld intensities,
3 (1/cm)
T E1 and T E5 2 2 Ey (x)/E0 T E4
resonance frequency although their slope is positive. This result comes from the fact that the group velocity in the lossy waveguide is not equal with energy velocity. We study this issue in detail in Chapter 3. We can explain the penetration of the modes below the resonance frequency by looking at Fig. 2.9. It shows magnetic permeability near resonance frequency. The blue and red curves represent permeability without and with material loss respectively. For the lossless Lorentz model, permeability becomes innite,
( ) at the
resonance frequency. Since in the presence of material loss, the permeability has a
CHAPTER 2.
42
4.05
4.03
frequency (GHz)
TE4
4.01
frs
3.99
TEB 4
3.97
3.95 0
(1/cm)
10
15
20
Figure 2.8: The blue curve indicates that the TE mode extends below the resonance frequency in the presence of loss. The red curve represents the backward imaginary parts of the
T E4
mode
of the lossless waveguide and the thick and thin blue curves represents the real and
T E4
nite value at the resonance frequency (see Fig. 2.9), the TE modes can penetrate into the single negative-index media, where the Fig. 2.8 for
T E4
mode, but the real part of the wave number rapidly goes to zero.
Since the damping factor enters in the denominator of the characteristic LorentzDrude equation, both imaginary and real part of
and
CHAPTER 2.
43
permeability
= 0.1
50 Re[] 100 150 3.97 3.98 3.99 4 4.01 4.02 4.03
=0
frequency (GHz)
Figure 2.9: 2 2 The permeability near the resonance frequency with
The solid and dashed curves represents the real and imaginary parts of the permeability with
(f ) = 1 = 0 .
2.6 The
terial waveguide
In this section, we calculate the rst ve TM modes of the lossy, dispersive metamaterial waveguide structure with the parameters used in section 2.5. for The TM modes can be obtained from duality by changing
E H .
Fig. 2.10(a) indicates that the dispersion curves of the TM modes behave like the
TE modes.
fms = 4.71
GHz, where
(fms ) + 0 = 0.
CHAPTER 2.
44
fsp = 7.07
GHz, where
(fsp ) + 0 = 0.
double-negative-index region, we do not plot surface plasmon region in Fig. 2.10(a). If we extend Fig. 2.10(a) to higher frequencies, we see that the TM surface polariton modes,
T M0
and
T M1 ,
touch at
GHz as
as the TE modes at
fms .
53
5.2
TM0
(a)
5.2
TE0
(b)
frequency (GHz)
TM1
4.7
frequency (GHz)
TE1
4.7
fms TM2
4.4
4.4
TM3 TM4
4 0 1 2 3 4
fres
5 6
4 0 1 2 3 4 5 6
fres
Figure 2.10: (a) The dispersion curves of the lossy dispersive metamaterial waveguide structure for the TM modes with the same parameters as in the Fig. 2.5. The curves of the TM modes in contrast with the TE modes (b) are insensitive to the magnetic surface polariton frequency,
fms .
There are two types of surface modes in the metamaterial waveguide structure. One is called a surface plasmon mode. It is a result of the coupling of photons and electrons. Its frequency is determined by
Figure the cladding. The other is called2.10: (a) The dispersion curves ofplasmon mode. It is a result of the a magnetic surface the lossy dispersive metamaterial waveguide the TM modes in contrast with the TE modes (b) are insensitive to the magnetic
Figure 2.10: (a)structure for the TM modes with the same parameters in the Fig. 5. The curves of The dispersion curves of the lossy dispersive metamaterial waveguide structure for the TM modes with the same parameters in the Fig. 5. The curves of the TM modes in contrast with the TE modes (b) are insensitive to the magnetic
surface polariton Its frequency is determined by coupling of photon and electron spins. frequency, fms .
(fms )+0 = 0,
where
surface polariton frequency, fms . is the permeability of the cladding. If we were to set the optical parameters
fsp = fms ,
CHAPTER 2.
45
that the properties of the metamaterial waveguide structure are completely dierent from those of a conventional waveguide system. It was shown that the metamaterial waveguide structure supports a rich variety of modes with unusual properties. These properties provide more control over electromagnetic elds and open a door to exciting potential applications, such as miniaturizing devices and systems with slow-light modes. In the last section of this chapter, we have examined these properties by considering the practical limitations. We have explicitly studied the roll of dispersion and loss, which are inherent features in metamaterials, on the dispersion curves. We have found that the properties of modes can change greatly when loss comes into play. In the following chapter, we will study the eect of intrinsic loss on the one of the most important potential properties of the metamaterial waveguide structure, slow-light modes, in more detail.
Chapter 3
3.1 Introduction
Metamaterials can change the electromagnetic properties of many dierent physical structures. Even a slab waveguide containing a metamaterial shows exotic features that may have important potential applications, such as a vortex-like energy ux structure[24], and stopped-light modes [35]. In 2003, Y.Kivshar and coworkers showed that a double vortex-like energy ux pattern forms at the two interfaces of a waveguide with the core made of metamaterial and claddings made of positive-indexof-refraction materials (see Fig 3.1). The vortex-like energy structure forms because the propagating energy in a metamaterial waveguide core ows in the opposite direction to that in the cladding [24]. This property plays a key role in controlling light propagation in a metamaterial waveguide [25]. The vortex-like energy ux structure property leads to a slowing down or even a 46
CHAPTER 3.
47
Figure 3.1: The schematic of the double vortex-like energy ux structure in a metamaterial waveguide (based on Fig.4 in Ref. [24]).
stopping of the light traveling through the metamaterial waveguide structure. Such an anomalous feature has aroused strong interest recently due to its potential applications in data storage, optical buers, and optical sensing [5-10]. Moreover, the vortex-like surface wave property may lead to an extremely high power density in a metamaterial waveguide, which may nd potential applications in medical treatments, and industry [33, 34]. In this chapter, we investigate the fascinating features of slow-light and highpower density in a realistic metamaterial waveguide, by including both dispersion and loss. model. First, we consider a dispersionless and lossless metamaterial waveguide
We calculate the energy density, energy velocity, and the group velocity Our
for a particular mode for this model over a broadband range of frequencies.
numerical calculations show that the group velocity can exceed the velocity of light in vacuum, and the total energy density can take negative values in some frequency ranges, which of course contradicts the principle of causality and the second law of
CHAPTER 3.
48
thermodynamics [97, 98]. Thus, we repeat our calculations for a lossy and dispersive metamaterial waveguide. It is well-known that the group velocity is not equal to the energy velocity, where the latter one is the correct measure of energy transport for a lossy medium. To nd the energy velocity, we need to derive the correct form of the energy density and the total power ux. In section 3.3, we derive the appropriate energy density formulation in a lossy metamaterial medium for the Lorentz medium model and the Drude-Lorentz medium model. We derive the closed form of the
total power ux and the energy velocity for a symmetric metamaterial waveguide in section 3.4. Then, we numerically investigate the energy velocity and the power density for various types of modes and structures. Finally, we study the role of
intrinsic loss on the slow-light modes of the metamaterial waveguide structures. It has been proposed recently [35], that it could be possible to slow down or even stop light in a metamaterial waveguide structure. Because of the exotic properties of the theoretical structure proposed in [35], there is considerable interest in the literature (e.g. [39, 41, 44]). However, the intrinsic loss, which is the inherent feature of any realistic metamaterial, is completely neglected in the proposed structure. At the end of this chapter, we critically re-examine the structure proposed in [35] in the presence of material loss.
CHAPTER 3.
49
slab with
1 = 4
and
1 = 4
The dispersion curves of this structure have been studied in Chapter 2; from Fig. 2.3, we know that the proposed structure supports both complex modes and normal modes. In this section, we only investigate the group velocity of the particular normal mode,
T E2 ,
medium.
T E2
mode in Fig.
range of frequencies for which the backward and forward propagation modes coexist. In Chapter 2, we showed that the dispersion curve of a metamaterial waveguide, in contrast to a dielectric one, is not monotonic at all frequencies and the transcendental equation may have two proper solutions at some frequencies. These two possible
solutions have two dierent power distributions. In one solution, (the red curve in Fig. 3.2) the averaged power ux in the core is greater than that in the cladding
(forward propagating mode) while in the other solution, the averaged power ux in the core is less that that in the cladding (backward propagating mode) . We dene the normalized total power ux per unit length (in
y axis)
as
total Sz
1 = Re 2
dx (E H )z .
(3.1)
We can calculate the total power ux for the proposed structure by substituting and
from Eqs.
(3.1).
2 E0
dened previously in
Chapter 2, are plotted in Fig. 3.2(c). As shown in Fig. 3.2(c), the total power ux can be positive (forward propagating mode), negative (backward propagating mode) or even zero when the averaged power ux in the core is completely canceled by that in the cladding (here, at 2.85 GHz). Thus, light can be stopped at that critical frequency
CHAPTER 3.
50
point as shown in Fig. 3.2(d). We plot the slope of the proposed mode (group velocity) in Fig. 3.2(d) to highlight the critical point with zero group velocity. Fig. 3.2(d)
demonstrates that the group velocity behaves anomalously versus frequency in the proposed structure. Although we know that the energy velocity should be equal to the group velocity for the proposed lossless structure, we calculate the energy velocity of the proposed mode to nd an answer for the unusual behavior of the velocity. Our results conrm that the slope of the dispersion curves of the proposed mode, which indicates the group velocity velocity, density.
vg =
vE = S / w
and
is
w =
1 dx ( |E|2 + |H|2 ). 4
(3.2)
We calculate the total energy density over the range of the frequencies in Fig. 3.2(b) for the proposed structure. Astonishingly, Fig. 3.2(b) indicates that
takes neg-
ative values at certain frequencies or goes to zero at a critical frequency; however, classical physics ticular point,
restricts the energy density to have a positive value. At this parthe energy velocity diverges and changes sign suddenly. Fig. at 3.45 GHz, which obviously contradicts the causal-
w = 0,
vE
ity principle. To address this unphysical result we next investigate, more carefully, the electromagnetic energy density in metamaterials. It is clear from Eq. (3.2) that energy density is always negative in a dispersionless metamaterial structure with both negative
and
1 Unlike classical physics, quantum physics does not restrict the energy density to have a positive
value [75].
CHAPTER 3.
51
(a)
f (GHz)
<w> = 0
f (GHz)
(b)
3.5
3.5 3 2.5 0 4
<w> = 0
3 2.5 20 4
<Sz> = 0
1
(1/cm)
<w>/|E0|2
20
40
(c)
f (GHz) f (GHz)
(d)
3.5 3 2.5 10
3.5 3 2.5 10
vE
<Sz> = 0
0 10 20
<Sz total>/|E0|2
0
vE/c
10
Figure 3.2: (a)The dispersion curve of the dispersionless metamaterial waveguide for the
T E2
mode with
this graph: a point with zero total power ux, which leads to zero energy velocity, and a point with zero total energy density, which leads to innite energy velocity. The red and blue curves indicate the mode with negative and positive total power 2 ux, respectively. (b) Total energy density (Normalized with respect to |E0 |) versus frequency. (c) The average total energy ux versus frequency. (d) The normalized energy velocity versus frequency. It shows that
1 = 4
and
1 = 4.
vE
where
w = 0.
dispersive, the average energy density for time harmonic waves assumes the form [67]
w=
1 4
(3.3)
and
are
()
and
()
CHAPTER 3.
52
required in order to secure positive energy density. However, the energy density expression with Eq. (3.3) is inappropriate in a lossy metamaterial. Ziolkowski numerically showed that the above expression yields negative energy density value in a high-loss region near the resonance frequency [73]. After he pointed out this issue, several approaches have been presented to nd an appropriate expression for the energy density in a dispersive and lossy metamaterial [68, 69, 70]. In the next section, we investigate this important energy density problem in detail.
S+
we wm + = 0, t t
(3.4)
CHAPTER 3.
53
where
we
and
wm
are the
electric and magnetic energy densities, which can be generally dened as [72]
D we = E, t t
wm B = H . t t
(3.5)
Now, we derive energy densities in the Lorentz medium model. This simple classical model consists of a collection of non-interacting electrons of displacement coupled to the external electric eld according to the equation of motion
r,
m(
2r r 2 + e + eo r) = qE, 2 t t eo
is the resonance frequency,
(3.6)
where
and
are the
eective mass and charge of the electrons. We can rewrite Eq. (3.6) in terms of the electric polarization vector,
P = qr ,
as follows
2P P 2 2 + e + eo P = 0 ep E, 2 t t
where
(3.7)
ep q 2 /m0
teraction between the electron oscillators and the electric eld. By using Eq. (3.5) and the relationship between the electric displacement, namely
D,
E,
D = 0 E + P ,
we have
we D E P = E = 0 E+ E . t t t t
(3.8)
CHAPTER 3.
54
Substituting
1 1 we = 0 [E]2 + 2 t t 2 20 ep
Hence, the electric energy density is
P t
2 eo
[P ]
e P + 2 0 ep t
(3.9)
1 1 we = 0 [E]2 + 2 2 20 ep
P t
+
2 eo
[P ]
P e dt 2 0 ep t
,
(3.10)
we + we .
We can separate the electric energy density in Eq. (3.10) into two parts. The second part,
we ,
represents the heat energy due to loss; as this is not energy in the eld, we
neglect this part from now. The rst part represents the stored electric eld energy density, which has the form
1 1 we = 0 [E]2 + 2 2 20 ep
P t
2 2 + eo [P ]2 .
(3.11)
We can simplify Eq. (3.11) by adopting a time harmonic electromagnetic eld. Hence, the time-averaged stored electric energy density that follows directly from Eq. (3.11), is
1 1 2 we = 0 |E|2 + 2 + eo |P |2 . 2 4 40 ep
We then express the time-harmonic polarization,
(3.12)
P,
using Eq. (3.7). Finally, the electric energy density in terms of electric eld for the
CHAPTER 3.
55
we =
2 2 ( 2 + eo ) ep 0 1+ |E|2 . 2 2 )2 + 2 2 4 (eo e
(3.13)
From Eq. (3.13), it is obvious that the electric energy density always takes a positive value in the lossy, dispersive Lorentz model even though the permittivity may be negative. Similarly, we can write the equation of motion of the magnetic polarization. It has the form
M 2M 2 2 + m + mo M = F mp H, 2 t t
where
(3.14)
mo
mp
is
After some
algebraic manipulations, we can derive the time average magnetic energy density for the Lorentz medium model with
2 F mp . () = 1 + m = 1 + 2 mo 2 im
The stored magnetic energy density is
(3.15)
2 2 F ( 2 + mo ) mp 0 wm = 1+ |H|2 . 2 2 )2 + 2 2 4 (mo m
From Eq.
(3.16)
all frequencies. Hence, the total energy density is always positive in a metamaterial structure that obeys the Lorentz medium model. Next, we derive the energy density
CHAPTER 3.
56
() = 1 +
F 2 , 2 mo 2 im
2 ep
(3.17)
() = 1
2 + ie
.
This means
that the problem of nding the energy density should be solved separately for each model. However, we can always dene the energy density properly by using the
Poynting's theorem. The energy density must be positive even in a double-negativeindex material according to the second principle of thermodynamics. It has been
shown [69] that the time-average electric and magnetic energy densities for the DrudeLorentz medium model dened by Eq. (3.17) are
2 ep 0 we = 1+ 2 |E|2 , 2 4 + e
(3.18)
wm =
Again, the total energy density remains positive at all frequencies, even near resonance frequencies or high loss regions. If the losses are negligible,
e = m = 0,
Eqs.
(3.13,3.16,3.18) reduce to the well-known energy density formulation, namely Eq. (3.3) in a dispersive medium for a narrow-band electromagnetic waves [67].
CHAPTER 3.
57
3.4 Power ux, group velocity, and energy velocity in a lossy and dispersive metamaterial waveguide
In this section, we rst derive the closed-form of the total power ux in a symmetric metamaterial waveguide structure. We then show that the total power ux is always zero in a lossless metamaterial waveguide for the complex modes. Second, employing the total energy ux and energy density from the previous section, we derive a correct measure of transport, the energy velocity, in a lossy metamaterial waveguide. Finally, we analyze the behavior of the metamaterial waveguide structure, previously studied in Chapter 2, by calculating the power ux and the energy velocity of the proposed structure for some particular modes . By denition, the total power ux of the waveguide structure is the spatial average of the Poynting vector over the guide's cross section, which has the following normalized (per length in
y axis)
form [72]
total = 1 Re Sz 2
dx (E H )z .
(3.19)
The total power ux of a waveguide structure can be expressed as the sum of the power ux in the core and in the cladding,
For a
symmetric waveguide with the thickness of 2d, the total power ux can be expressed as
1 total Sz = 2 Re
+d
d
dx (E H )z + 2
+
+d
dx (E H )z ,
(3.20)
where the rst term is the power ux in the core and the second term is that in the cladding. Substituting Eqs. (2.6-2.9) into Eq. (3.20), the closed-form of the total
CHAPTER 3.
58
power ux is
total Sz
2 c |E0 | e 2
2 i z
Re
where
cladding, respectively. The rst two terms in Eq. (3.21) corresponds to the energy ux in the core and the last two terms corresponds to the energy ux in the cladding. Since the sign of the permeability in the core and in the cladding are dierent, the energy ux in the core is in the opposite direction of that in the cladding. Thus,
the total energy ux can become negative or even zero in a metamaterial waveguide. It can be analytically shown that the total energy ux is always zero,
total = 0, Sz
for the complex modes of the lossless metamaterial waveguide. After some algebraic manipulations and separating the real and imaginary parts of wave numbers in the transcendental equation, Eq. (2.10), for complex mode, we obtain
r k2 =
r r i i 1 sinh (2k1 d) /k1 + sin (2k1 d) /k1 . r i 2 cosh (2k1 d) + cos (2k1 d)
(3.22)
total Sz = 0.
in a lossless metamaterial waveguide never carry energy, although the normal modes do. However, we note that
complex modes carry energy as we show later in our numerical calculations (see e.g. Fig. 3.4 or Fig. 3.5).
CHAPTER 3.
59
Finally, in this subsection, we nd the closed-form of the energy velocity for the proposed structure. Dening the energy velocity,
of the Poynting vector divided by the average energy density [70], we have
+ dx Sz Sz . = + vE w dx (we + wm )
Using Eqs. (3.18,3.21), and introducing
(3.23)
P1 Re
P2 Re
0 1+ 2 , 4 + 2 e
2 p
2 2 F 2 [o (3o 2 ) + 2 2 ] 0 m 1+ , 2 2 4 o (o 2 )2 + 2 2 m
vE = w + w
c 2 ( k1 )c 1
2
P1 P2 + 1 2 P1 + 2 + 2 ( + k2 )c 2
2
. P2
(3.25)
Since the energy velocity measures the transport of the electromagnetic energy, it must be always smaller than the speed of light in vacuum. In the following, we show that
CHAPTER 3.
60
this condition is preserved in the proposed metamaterial waveguide at all frequencies even near the resonance frequency or high-loss regions. Though, it can be negative for the backward propagation modes. It is important to note that the energy velocity is dierent from the group velocity in general. Since the energy velocity the group velocity,
vE
is equal to
vg d/d ,
be considered as a correct measure of transport in a lossless medium. However, the group velocity can be a misleading concept in lossy media. For example, from Fig. 2.5(b), we can conclude that the group velocity goes to innity near the resonance frequency or the magnetic surface polariton frequency. Thus, if the group velocity
shows the energy transport, the causality principle would certainly be violated in a lossy metamaterial waveguide. In some recent publications (e.g. [42, 76]), the group velocity is incorrectly used as the measure of transport in lossy metamaterial media; however, it is well-established in the community that the energy velocity, correct measure of transport in lossy media [67].
vE ,
is the
3.4.1
Numerical analysis
() = 1
0.56 2 2 16 + i0.016
&
() = 1
100 , 2 + i0.016
(3.26)
with a free-space cladding as shown in Fig. 2.1(a). In Chapter 2, we have previously studied the modal characteristics of the proposed structure for the TE and the TM modes. Now, we investigate the energy velocity,
of this proposed structure . In Chapter 2, we have also calculated the dispersion curves of this structure for the rst ve TE modes (see Fig. 2.5(b)). Here, we study the
CHAPTER 3.
61
T E1 , T E2 ,
and
T E5
5.5
5.5
(a)
5
f (GHz)
(b)
f (GHz)
TEB 1
TEB 5
(1/cm)
10
10
10
10
vE/c
(c)
5
f (GHz)
(d)
5
f (GHz)
4.5 4 3.5
4 2 0 2
4.5 4 3.5
4 2 0 2
10
10
<Sz
10
>/|E0|
2
10
10
10
10
10
total
<Score>/|E0|2 z
Figure 3.3: (a) The blue curve indicates the real part of propagation constant for the backward propagating
T E5
and
T E1
a thickness of 2 cm, is in free-space and we use the standard Lorentz-Drude model p = 10 GHz, 0 = 4 GHz, = 0.1 GHz to calculate its permittivity with F = 0.56, 2 2 and permeability. (b) the energy velocity versus frequency. (c) The normalized total energy ux and (d) The normalized energy ux in the core versus frequency.
In Fig. 3.3, we investigated the total power ux and the energy velocity of the backward-propagating
T E1
and
T E5
B T E1 ,
B T E5
modes in the presence of the material loss; hence, instead of two backward
propagating
T E1
and
T E5
Also shown in Fig. 3.3, are the computed (b) normalized energy velocity, (c) total power ux, and (d) power ux in the core versus frequency. Fig. 3.3(b) indicates that
CHAPTER 3.
62
the normalized energy velocity never exceeds the speed of light in vacuum and never goes to zero; the
vE
near the surface magnetic plasmon frequency (at 4.71 GHz) has the minimum value. We do not show the loss number here, but by looking at Fig. 2.5, we can conclude that the imaginary parts of the propagation constant is large whenever
vE
is small;
thus, the slow-light modes decay very fast before they can propagate an appreciable distance. From Fig. 3.3(d), we can see that the power density in the core is anoma-
lously large near the magnetic surface polariton frequency. The dimensionless power ux in the core has a maximum peak at the magnetic surface polariton frequency,
which
10, 000
fms .
is considerably smaller at this point because of the vortex-like property of a metamaterial waveguide. The power density in the core is partially canceled by the power density in the cladding; hence, the total power ux is much smaller than the power ux in the core. Consequently, the energy velocity has a small value at
fms
although
the energy propagates very rapidly in the core or in the cladding. The high-power density property comes from the phase matching condition ,
core + cladding = 0,
at
fms .
so-called super waveguide, proposed in Ref. [33, 34] . In Fig. propagating 3.4, we study the energy velocity and the power ux for the forward-
T E1 and T E5 modes.
The behavior of these modes is completely dierent First, we note that no coupling occurs
F T E1
and
F T E5
modes, which are shown with the red and blue curves,
CHAPTER 3.
63
5.5
5.5
(a)
f (GHz)
4.5
f (GHz)
TEF 1
(b)
5 4.5 4 4 10 5.5
TEF 5
4 0 5.5 1
(1/cm)
10
10
10
10
vE/c
(c)
5
f (GHz) f (GHz)
(d)
5 4.5 4
<Sz total>/|E0|2
<Score>/|E0|2 z
Figure 3.4: (a) The dispersion curves of the lossy, dispersive metamaterial waveguide for the forward propagating parameters used in Fig. respectively. 3.
T E1
(b).
T E5
(c) and (d) indicate the normalized total power ux and power ux in the core,
respectively, in Fig. 3.4. Second, the high-power transmission property disappears in these types of modes. Fig. 3.4(d) indicates that the maximum power density
in the core for the forward-propagation modes at the magnetic surface polariton frequency is at least 1000 times smaller than the backward-propagating modes. From the previous chapter, we know that the forward-propagating modes have the lower cuto frequency, where the dispersion curve intersects with the light-line. At this point, the energy velocity is equal to the speed of light in the cladding, and monotonically decreases with increasing frequencies as shown in Fig. 3.4(b). However, the energy velocity is never zero.
CHAPTER 3.
64
Next, we calculate the energy velocity and the total power ux of the In Fig. 3.5(a), we plot the real part of the propagation constant, mode.
T E2
mode.
Re[],
for the
T E2
The normalized power ux with respect to the eld intensity is shown in
Fig. 3.5(c). As expected, the total power ux has a positive value for the forwardpropagation mode and a negative value for the backward-propagating mode at all frequencies. Fig. 3.5(d) indicates that the power density in the core has a peak at the magnetic surface polariton frequency. However, the magnitude is small in comparison with the power density in the core for the coupled
T E5
and
T E1
(a) 5
f (GHz)
(b) 5
f (GHz)
4.5 4 0
TEB 2 TE2
F
4.5 4
(1/cm)
10 (c)
f (GHz)
|vE|/c
10
10
5
f (GHz)
5 4.5 4 3
(d)
4.5 4 0.5
0.5
Stotal/E2 z 0
1
Score/E2 z 0
B Figure 3.5: (a) The dispersion curves of the backward propagating T E2 mode (blue) F and the forward propagating T E2 mode (red). (b) log-plot of the normalized energy velocity versus frequency. (c) The normalized total power ux and (d) the power ux
in the core for the
T E2
T E2
CHAPTER 3.
65
modes (Fig.
3.5(b)).
cannot go to zero in a lossy metamaterial waveguide, the energy velocity is extremely small near the resonance frequency (smaller than
106
frequency, the propagation loss is massive. To have a better insight, we re-plot the energy velocity for the backward-propagation
T E2
versus frequency in Fig. 3.6. As can be seen, when the group velocity is larger than 0.01 c, the loss [20Im[]/ log(10)] is smaller than 0.01 dB/cm. However, the loss is greater than 150 dB/cm at the resonance frequency. Thus, at resonance, we have an ultra-slow-light mode with extremely high loss!
10
(a) |vE|/c
10
2
10
10
4.5
Loss (dB/cm)
(b)
4.5
5.5
f (GHz)
Figure 3.6: (a) The normalized energy velocity and (b) the propagation loss for the backward-propagation
T E2
CHAPTER 3.
66
direction inside the metamaterial layer is opposite to the one in the dielectric layers; thus, when the power ux in the core is completely canceled by the power ux in the cladding, light can be stopped. Previously, in our dispersion curves calculations for a lossless metamaterial waveguide with a xed thickness (see e.g. Fig. 2.5(a))
we have shown that at some critical frequencies the TE or TM modes have zero group velocity. We could also plot the propagation constant,
a xed frequency. Hence, we can nd the zero-group velocity modes by changing the thickness of a waveguide. Using this approach, Tsakmakidis et al. have proposed
a wedge-shaped metamaterial structure, which can stop various frequencies of light altogether at dierent points, forming a trapped rainbow. It is interesting to note
CHAPTER 3.
67
that the relation between the frequency of a light beam and the thickness of the dispersionless waveguide is dierent than in the dispersive waveguide, as waves with larger frequencies are stopped at larger core thickness in a dispersive metamaterial waveguide. We show the schematic of the wedge-shaped waveguide in Fig. 3.7(a). The green light is stopped at larger core thickness while the yellow light, with smaller frequency, is stopped at smaller core thickness.
(a) (a)
-d
Region(I) Region(II)
(b)
d
Region(I)
2d
Figure 3.7: (a) The schematic of the wedge-shaped metamaterial waveguide proposed in [35]. The green light with higher frequency is stopped at higher core thickness while the yellow light is stopped at smaller core thickness. (b) The schematic of the metamaterial waveguide structure used in our calculation.
To avoid complications, we consider an asymmetric waveguide structure with a xed thickness rather than a wedge-shape waveguide, in which a lossy metamaterial layer is surrounded by two regular dielectrics as shown in Fig. 3.7(b). The relative permittivity and permeability of the three waveguide layers (top to bottom) are and
i ()
i (),
where
i = {1, 2, 3}.
As in Ref. [35], the permittivity and permeability of are lossless and dispersionless,
while for the lossy metamaterial we use the standard Drude and Lorentz medium models, Eqs. (2.22,2.23). In all calculations, we use and we take the metamaterial slab thickness to be
(1 = 2.56 3 = 2.25 1 = 3 = 1) 1
F = 0.6, p =
6c , o =
9c ,
d/2 = 0.55c/c ;
these parameters
CHAPTER 3.
68
have been chosen so as to yield identical results to those presented in [35] for when
= c
= 0.
= 0.01c .
We investigate the rst three TM modes, and calculate the complex waveguide dispersion curves in the presence of metamaterial loss and dispersion by using transcendental equations (2.15,2.16) for the proposed asymmetric waveguide structure. In Fig. 3.8, we plot (a) the real part and (b) the imaginary part of the normalized propagation constant versus the normalized frequency with respect to
c . It is impor-
tant to note that we mainly use dimensionless units in Fig. 3.8; therefore, converting to any frequency is trivial, and our essential results applies to all frequencies. The characteristic model of an asymmetric waveguide and a symmetric waveguide structure are similar. However, some minor dierence can be observed between them; for example, in comparison with a symmetric waveguide (e.g. see Fig. 2.5(b) and Fig. 8), there are two light-lines, which correspond to
and
3 ,
the TM modes are perturbed at two frequencies, which correspond to the magnetic surface polariton frequency for
and
3 . The T M1
cause of its surface polariton characteristic, distinctly changes near these two points,
changed at the magnetic surface frequencies because the phase matching condition,
Since
1.14 c ,
we re-plot the dispersion curves of the proposed structure for the frequency
range of Fig. 3.9. In light of dispersion curves of Fig. 3.9, we can conclude that, rst, it is impossible to have single-mode operation ; second, there is no point with zero group velocity at any frequency. Therefore, the claims in Ref. [35] are not valid when
CHAPTER 3.
69
loss is included. To obtain better insight, we highlight the critical frequency region for
T M2
modes in Fig. 3.9, and compute the lossless model (purple curve) for
T M2
mode in this region. Using lossless model, we recover the key result of [35] and obtain a point of zero group velocity for
= c . c
in two and the normal modes pulls away from its lossless counterpart; rather than having a normal mode and a complex mode, one obtains complex forward and backward propagation modes. The inuence of material loss on mode structure is highly non-perturbative near
c ,
changes in the dispersion such that the group velocity is never zero.
1.4 1.3
TM1
1.2 1.1
TM2
30 45 60
1 0 15
2Re[]c/c
2Im[]c/c
Figure 3.8: Dispersion curves of the proposed waveguide structure in Fig. 3.7(b) with
= 0.01c
Re[],
Im[].
CHAPTER 3.
70
1.02 1.13 1.09 1.05 1 TM3 0.95 6 12 18 24 0.99 12 metamaterial light line TM2 light line TM1 1.01 =0 =0.01c
/c
/c
1 vE = 0 18 24
2c/c
2c/c
Figure 3.9: Dispersion curves of the proposed waveguide structure in Fig. 3.7(b) for TM modes with the
= 0.01c .
light-line. The right plot shows the dispersion curves near the critical frequency for with
T M2 mode for the lossless model (purple curve) and the lossy model (green curves) = 0.01c .
In Fig. 3.10(a), we rotate Fig. 3.9 and plot frequency versus normalized propagation constant near
c .
does not go to zero; but, as we mentioned before, the energy velocity is the correct measure of transport in the lossy media, not the group velocity; therefore, we plot the energy velocity for
= 0.01 c
in marked contrast to the lossless case, is not zero at any frequency. Although the energy velocity goes as low as
is large whenever the energy velocity is small, which leads to very high propagation losses. In Fig. 3.9(b), we also plot the propagation loss (=
20 Im[]/ ln[10])
in prac-
= 385 THz).
vE < 0.01c,
CHAPTER 3.
71
in proper context, if this waveguide were used as a delay line, then for a delay, equal to 1000 times the fundamental period of the eld (e.g., for
D ,
c /2 = 385 THz,
D = 2.6 ps) the loss would be at least 200 dB. It is interesting to note that high losses
are also observed in slow-light band-edge modes in photonic crystal waveguides when there is even a tiny amount of surface-roughness [74] . In both cases, the large loss is a direct consequence of the large time that the slow light spends near the potential absorbers and/or scatterers.
30
TMB 2 vE = 0
(a)
2c/
25 20 15
(b)
10
6
10
10
10
10
0.98
1.02
1.04
10
10 0 10
10
|vE|/c
T M2
mode
velocity and propagation loss versus frequency for the loss model with = 0.01 c B F for T M2 (blue) and T M2 (green) modes. The propagation loss is calculated for a critical frequency in the optical (c /2
= 385 THz).
and conclusions scale to GHz and THz. For example, at a frequency of 1 THz, the loss is approximately 1,000 dB/cm or greater; this is completely impractical as a delay element, given the corresponding energy velocity.
= 0.01 c ,
CHAPTER 3.
72
has been experimentally achieved to date near optical frequencies and comparable to state-of-the-art at GHz frequencies [19, 20]. Therefore, it is expected that results would be even worse in a real metamaterial structure due to a higher loss number. We consider again the proposed metamaterial waveguide structure of Fig. 3.10; but this time with velocity with 3.11(a).
= 0.0001 c . The blue and red curves indicate the normalized energy
and
= 0.01 c
= 0.0001 c c ,
for the
T M2
for the both cases; however, the group velocity is 100 times smaller in the latter one above the critical frequency. The essential point is that the energy velocity is never zero even if we introduce very small amount of loss into the metamaterial structure. Therefore, the assumption that the metamaterial metallic elements can be treated as perfect, lossless, conductors in the GHz domain is, of course, misleading. The neglect of loss is certainly not acceptable in a regime of slow light in metamaterials. In Fig. 3.11(b), we plot the loss propagation and the energy velocity with
= 0.01 c
and
= 0.0001 c
for the
Re[]/Im[]
velocity is smaller than 0.01 c. Fig. 3.11(b) indicates that the slow-light modes are extremely lossy even in the presence of small amount of loss
( = 0.0001 c ).
Regarding the possibility of compensating for the loss by introducing gain into the metamaterial: it has been rigorously shown in [48] that is impossible to eliminate loss in a negative-index material (except possibly at a single frequency). This is the result of the basic requirement of causality. Thus, we conclude that it is not possible to stop light over a rainbow of frequencies even in an active metamaterial waveguide structure.
CHAPTER 3.
73
(a) 10 10 |v |/c 10 10 10
E
0 1 2 3 4
TMF
2
~ = 0.01 ~ = 0.0001
TMB 2
TMB 2
TMF 2
0.01
= 0.0001
= /c .
=
(b)
The solid curves indicate loss propagation versus the energy velocity and the dashed curves indicate
r /i
CHAPTER 3.
74
Next, we have derived the correct form of the energy density and energy velocity in a dispersive and lossy metamaterials. We have shown that the intrinsic loss nonperturbatively aects the slow-light modes. In conclusion, our calculation for various modes, TE and TM modes, and structures, a symmetric and asymmetric structure, indicates that it is impossible to stop light in metamaterial waveguides such as those proposed in [35].
Chapter 4
4.1 Introduction
In the previous chapters, we studied the properties of the modes of a metamaterial waveguide structure. Since the modes are almost all below the light-line, excitation of the modes is dicult unless a special technique for phase-matching is employed. One of the well-known techniques is the attenuated total reection (ATR) method, which has been commonly used to couple surface polariton modes in metals with an incident beam using a prism (see Fig. 4.1). In 2000, when Ruppin investigated the properties of surface polariton modes in metamaterials, he used the ATR method in his numerical analysis [21]. Then Smith et al., based on the Ruppin idea, employed the ATR method to excite surface polariton modes in a metamaterial slab [22]; and Park
et al. extended Ruppin's work to the bulk and surface modes [78]. However, these
authors only made the connection between lossless dispersion curves of a metamaterial 75
CHAPTER 4.
76
waveguide and the minimum reectivity of the ATR method with the angle xed while varying the frequency. In this chapter, we develop the ATR method to study a lossy waveguide structure at a xed frequency while varying the angle. Also, we study the limitations of the ATR method and suggest how the accuracy of the ATR method might be improved by optimizing the air gap and the refraction index of the prism. However, before we study the ATR method, we rst derive the generalized transfer matrix formalism [84, 85, 86], which is applicable to a lossy multilayer system with a negative-index slab. The chapter closes with a brief look at the eect of the excited modes on the lateral Goos-Hanchen beam shift [87, 88].
CHAPTER 4.
77
Surface wave
n2 < 0
metamaterial
Figure 4.1: Schematic geometry of the ATR method. A plane wave of frequency is incident on the boundary of a prism and air, with the angle of incident, dip in the reection at certain angles.
varied,
which couples with a surface wave on the metamaterial slab. Hence, it can lead to a
4.2.1
Theory
(t , t ) regions are semi-innite, and a plane wave is incident from the rst region with wave vector k and the angle
plane of incidence parallel to the x-z plane; thus, all eld vectors are independent of
y. We separate the Maxwell equations in each region into those for the TE and TM
modes. From Eq. (2.3), we know that the wave equation for the electric eld has the
(fig 1). The region (t) is semi-infinite . the permittivity and permeability in each region are m & m , the plane wave is incident from region 0 and has the plane of incident parallel to the y-z plane. All field vectors are dependant on y and z only and independent of x. since = 0 , the Maxwell equations in any region can be separated into TE and x TM components governed by Em and Hm.
CHAPTER 4. EXCITATION ON THE MODES OF MWS 78
0 , 0
1, 1
2 , 2
3 , 3
4 , 4
l 1 , l 1
l ,l
z x
d1 d2 d3 d4 dm dl
2 2 + 2 + 2 m m 1/(r)0 , {1/(r) E(r)} = 2 /c2 E(r). ) Emx = y 2 z 1 H my = Emx We assume that all the regions of the aforementioned structure are i m z (
the electric eldE each region for the modes is described by in H =
mz mx
(4.1)
isotropic; hence,
1 im y
(4.2)
E m = i m H m . c
(4.3)
CHAPTER 4.
79
Using Eqs. (4.2) and (4.3), the electric and magnetic elds in each region for the TE modes are
Hm z = Hm x =
+ Em e+ikm x ),
(4.4)
c iz e (Em eikm x m
+ + Em e+ikm x ),
H m y = E m x = E m z = 0,
where the amplitude
+ Em
and
Em
direction, respectively.
From
is
x, k m ,
km =
m m
2 2 = 2 c c
m m 0 0 sin2 ,
(4.5)
which satises the phase matching conditions and the dispersion relation.
Finally,
using the fact that the tangential components of the electric and magnetic eld should be continuous at
x = dm+1 ,
+ + Em eikm dm+1 + Em e+ikm dm+1 = Em+1 eikm+1 dm+1 + Em+1 e+ikm+1 dm+1 ,
+ Em e+ikm dm+1 ) =
(4.6)
CHAPTER 4.
80
(4.7)
1 M m,m+1 = 1
k m m km m
1 x
to
1
km+1 m+1
where
x,
1
k m+1 m+1
(4.8)
x = x dm+1 ,
and obtain
+ Em+2 Em+2
;
(4.9)
hence,
+ Em+2 Em+2
e =
ikm+1 dm+1
0 e
ikm+1 dm+1
+ Em+2 Em+2
.
(4.10)
where
+ Em Em
= T m,m+1
+ Em+1 Em+1
,
(4.11)
T m,m+1 = 1 (1 + 2 (1
(1 (1 +
.
(4.12)
The transfer matrix formalism can be used to determine the electromagnetic eld amplitudes in any region in terms of those in any other region. In particular, the
CHAPTER 4.
81
transmission, t, and reectance, r, coecients of the proposed l-layer structure in Fig. 4.1, with respect to a plane wave amplitude, can be calculated with
(4.13)
1 t = T 1,l , r 0
where
(4.14)
Sz = 1/2 Re[E H ] z ,
the
power of the incident, reected and transmitted plane wave can be expressed as
Pi = Re
k0 c 2 E , 20 0
PR = Re
k0 c 2 2 r E0 , 20
PT = Re
kl c 2 2 t E0 . 2l
(4.15)
We normalized the incident power to unity; thus, the conservation of the power yields
r2 + Re
kl 0 k0 l
t2 = 1 Ploss ,
(4.16)
where
sion coecient of the aforementioned structure for the TE modes can be expressed as
k T = Re l 0 k0 l
where is
1 T11
(4.17)
T11
R=
T21 T11
(4.18)
CHAPTER 4.
82
containing metamaterial with intrinsic loss for the TE modes. We also note that the angle of incident
km .
The transfer matrix for the TM modes can be obtained from duality by changing
T M modes
is
T = Re
k l 0 k 0 l
1 T11
(4.19)
R=0
4.3.1
Basic principles
In the previous chapters, we have shown that the dispersion curves of a metamaterial waveguide structure for the normal and surface modes are always below the light-line
CHAPTER 4.
83
(for example, see Fig. 2.5). Thus, the wave vector of a beam of light, k, is always smaller than the propagation constant,
phase-matching cannot be achieved even at normal incidence ( if we add a prism into the system (see Fig.
> k sin ).
However,
kz =
prism k sin ,
< kz .
Thus, the
prism light-line can be coupled with a light beam. In Fig. 4.3, we show schematically the allowed region in a (, f )-plane. The dark and light blue lines represent the lightline and prism light-line with
prism = 9.
right or left by changing the index refection of the prism; thus, we can excite the normal and surface modes with any propagation constants with an appropriate refraction index. In the ATR method, two forms of experimental scan are possible: an angle scan with the frequency xed, and a frequency scan with the angle xed. The trajectory of the former is a straight-line parallel to the
by choosing a prism
a line between the air light-line and the prism light line as shown schematically with the red dashed lines in Fig. 4.3. In both forms, the mode is excited where the scan line crosses the dispersion curve; however, the dispersion curves measured by angle scan and frequency scan are dierent. The angle scan measures the dispersion curve in the real- and complex- space while the frequency scan measures the dispersion curves in the real- and complex- space. This is discussed in more detail in Chapter 5. Fig. 4.1 shows the schematic geometry of the ATR method with the angle scan. A beam of light is incident through one side of a prism with
where
is
CHAPTER 4.
84
frequency (GHz)
4.7
fms
4 0
air = 1 1
fres
(1/cm)
Figure 4.3: Dispersion curves and prism coupling. Only the modes, with the propagation constant between the air light-line and the prism light-line, can be excited by the ATR method. The blue line shows the prism light-line with prism light-line can be shifted to higher
prism = 9.
The
prism. Two forms of experimental scan (frequency xed and angle xed) in the ATR are shown schematically with red dashed-lines.
the permittivity. The beam passes through the prism and is incident on the second interface at an angle
is larger than the critical angle for , the beam undergoes total internal
> c = sin1
air prism
reection and exists at the third side of the prism. But, because of the presence of the metamaterial slab near the second interface, the beam is not totally reected. The decaying evanescent wave of the light beam on the air side of the prism can couple to the normal and surface modes of the metamaterial waveguide; thus, the reectivity is reduced below unity, and a dip is observed in the reectivity at the angle at which the parallel component of the wave vector of the light beam matches with the propagation constants
because the waveguide structure is perturbed by the presence of the prism. However,
CHAPTER 4.
85
by controlling the air gap and the index of the prism, we can minimize the error. In the following section, we numerically nd the optimum conditions for the ATR method.
4.3.2
Using the ATR method, we can excite the normal and surface modes of a metamaterial waveguide structure; however, the main challenge in the ATR method is to nd an optimal distance between the prism and the slab. If the air gap is too small, the
system is over-coupled and the prism massively perturbs the waveguide system; thus, the excited mode will not be observed. On the other hand, if the air gap, d, is
too large, the system is under-coupled; thus, the mode is only weakly excited and only a small reectivity dip can be observed. We show numerically the eects of
the air gap width on the reectivity in Fig. 4.4. In our numerical results presented in Fig. 4.4, we consider a lossy metamaterial slab with the following parameters: in the Lorentz-Drude model. The
prism = 6
(the schematic
geometry of the system is shown in Fig. 4.1). We present the calculation dispersion curves for the
T E1
is calculated in the complex- space, we use the angle scan with the frequency xed at 4.8 GHz in the ATR method. The scan line, the red line parallel to the crosses the forward and backward propagating
axis,
T E1
modes at
= 27
and
= 70 ,
respectively. Hence, we expected dips appear in the reectivity at these angles. In Fig. 4.4(b), we show the calculated reectivity of the proposed system by using the transfer matrix method for four dierent air gap widths,
where
CHAPTER 4.
86
= 30/4.8 = 6.25
because the
cm). When
d = /2,
= 27
forward propagating mode is only weakly coupled with the beam, and the air gap is too large to excite the backward propagating mode. the air gap, However, when we reduce
d = /4,
zero reectance at the minimum for the forward propagating mode; moreover, the backward propagating mode now can very weakly couple with the light beam. The very shallow dip can be observed near
= 65 ,
d = /8,
the dip
of the backward propagating mode becomes deeper and the width of the dip becomes smaller; however, the dip of the forward propagating mode disappears. It is important to note that the minimum reection seen at total internal reection,
= 23
the width of the dip becomes broader and tends to disappear; moreover, the place of the dip slightly shifts to the left. This means that the dierence between the
angle of minimum reection and that obtained from the dispersion curve starts to increase when the air gap reduces more than a specic value. In a later section, we attempt to nd numerically the optimal value for the air gap; but, as the rule of thumb
d /2 /4
d /4 /8
for the
backward propagating mode. With the simple numerical analysis used to produce Fig. 4.4, we have shown
that the choice of the optimal air gap is a challenging part of the ATR method. In the literature, several schemes have been developed to improve the accuracy of the
CHAPTER 4.
87
4.9
(a)
4.88 4.86 0.5 0
d=/2
(b)
Reflectance |R|
0.5 0
d=/4
F (GHz)
4.84 4.82
TE
d=/8
0.5 0
= 70
4.8 4.78
= 27
0.5 0 1
d = / 16
4.76
(1/cm)
20
30
40
(degree)
50
60
70
80
Figure 4.4: Eects of the air gap width, d, on the reectance. (a) The
T E1 mode of the
waveguide proposed in 2.5. At 4.8 GHz, the forward and backward propagating T E1 modes are excited with the ATR technique at = 27 and = 70 , respectively (red lines). The prism we used has respectively, where
= 6.25
cm.
ATR method under the condition where there is a zero refraction in respect with the air gap, is studied with the exact solutions. Using the zero-reection condition, the theoretical expression for the optimal air gap has already been derived in the literature (see for example [81]); however, in this section, we numerically analyzed the optimal air gap for the aforementioned structure to give physical insight about the relationship between the optimal air gap, optimal refraction index of a prism,
CHAPTER 4.
88
intrinsic loss and the modes. Again, we consider the waveguide structure with a prism, which was previously studied in Fig. 4.4, and calculate the reectivity at 4.8 GHz under the zero-reection condition. The optimal air gap and the minimum refraction angle can be obtained by using the transfer matrix method and zero-refraction condition. The optimal
air gaps for the forward and backward propagating modes are 1.89 cm and 0.83 cm, respectively. However, Fig. 4.5 indicates that the dips corresponding to the backward and forward propagating modes are located at smaller angles than obtained from the dispersion relation. The dierence between the exact value and that determined
and
7 ,
respectively.
= 0.1
GHz in the proposed structure, and the index of the prism are the To study the eect of the intrinsic loss, we recalculate
the reectivity under the zero-refraction condition for the aforementioned structure with the material loss a hundred times smaller than the previous calculations. The results are shown with the red curves in Fig. 4.5. As shown, in this case, there is a very good agreement between the solutions obtained from the ATR method and the dispersion equation. Therefore, we can conclude that the accuracy of the ATR method is reduced by the intrinsic loss even under the zero-reection conditions. In a very lossy waveguide structure or for the complex modes, this dierence becomes so signicant that the ATR method under the zero-reection condition cannot be used to calculate the dispersion curves. To obtain better insight, we calculate the dispersion curves of the waveguide structure for the
T E1
mode by using the ATR method for two dierent material loss pa-
rameter values in Fig. 4.6. As shown in Fig. 4.6(a), there is an excellent agreement
CHAPTER 4.
89
a
|R|
0
1 0.5
b
|R|
(degree)
Figure 4.5: Reectance with an optimal air gap for dierent material loss. (a) The optimal air gap for the forward propagating cm with
T E1
mode is
d = 1.89
cm and
d = 6.05
= 0.1
GHz and
= 0.001
are the same as for Fig. 4.4. (b) The optimal air gap for the backward propagating
T E1
mode is
d = 0.83
cm and
d = 1.76
= 0.1
GHz and
= 0.001
GHz,
respectively.
T E1
method under the zero-reection condition (red curves) and those calculated directly (blue curves) when the intrinsic loss is small (
stant above the prism light-line. Nevertheless, the dierence between them for the complex modes becomes larger when the frequency increases. Hence, we can conclude that the ATR method only works when the imaginary component of the propagation constant
is small.
In other words, the ATR method fails for the complex mode
at higher frequency even in the very low loss waveguide structure because at that
CHAPTER 4.
90
persion curves by using the ATR method for the waveguide with the moderate loss (
= 0.1GHz)
and calculated dispersion curves are signicant in this case as we cannot observe the splitting of the complex mode in the dispersion curves calculated by the ATR method. Hence, the ATR technique needs to be improved in order to be applicable for all the modes in lossy waveguide structures.
TE1
4.85
TE1
4.8
4.8
1.5
2 2.5 (1/cm)
3.5
1.5
2 2.5 (1/cm)
3.5
Figure 4.6: The dispersion curves of the system proposed in Fig. 4.4 calculated by the dispersion equation (blue solid curves) and by the ATR method with an optimal air gap (red dashed curves). Although (a) there is a good agreement between the solutions of the two methods for the small loss, between them are signicant for
= 0.001
= 0.1
GHz.
Fig. 4.6 The dispersion curves of the system proposed in Fig. 4.4 calculated by the dispersion equation (blue solid curves) and by an ATR method with an optimal of the prism can be controlled and using a prism with the appropriate index of refracair gap (red dashed curves). Although (a) there is a good agreement between the tion is an integral part of the experiment. If prism is = 0.001 GHz, (b) the dierence solutions of the two methods for the small loss, small, only a small propagation between them are signicant for = 0.1 GHz. constant range can be excited by the prism. On the other hand, if is too large,
prism
the waveguide structure is considerably perturbed by the prism with a high index
In the ATR experiment, not only the prism air gap, but also the index of refraction
CHAPTER 4.
91
refraction and the properties of the waveguide structure destroyed. In the following, we try to numerically nd the optimal air gap and refraction index of a prism to yield agreement between the dispersion curves calculated by the ATR method and the exact solutions.
4.88
prism = 3 lightline
prism = 12 lightline
4.84
4.82
4.8
TE1
4.78
4.76
1.5
2.5 (1/cm)
3.5
Figure 4.7: The dispersion curves of the waveguide structure proposed in Fig. indicate those with
4.4
using the ATR method with an optimal air gap. The green and red dashed curves
In Fig. 4.7, we show the dispersion curves using the ATR method under the zerorefraction condition with two dierent prisms. The prism with low-index,
prism = 3,
(the green curve in Fig. 4.7) is suitable to use for exciting the forward propagating
T E1
CHAPTER 4.
92
prism = 12,
agating
(the red curve) is more suitable to use for exciting the backward propmode. We notice that neither the low-index nor the high-index prisms
T E1
could correctly demonstrate the complex mode properties, as the splitting property of the lossy complex modes cannot be seen with the dispersion curves obtained from the ATR method. Since the imaginary component of the propagation constant
of the complex mode is considerably larger than the normal modes, the dispersion curves obtained by the ATR method depart from the exact solution more rapidly in the complex modes.
4.9
4.9
(a)
4.85 4.85
(b)
f (GHz)
TEF 1
4.8
TEB 1
f (GHz)
4.8 3 4 4.75 0
4.75
(1/cm)
4.9
d (cm)
(c)
4.85
f (GHz)
4.8
4.75
10
nprism
Figure 4.8: Optimum air gap and
nprism
for the
T E1
in Fig. 4.4. (a) The red and blue curves show the forward and backward propagating
T E1
mode. (b) The optimum air gap and (c) The optimum index of refraction of a
CHAPTER 4.
93
Finally, we derive the optimal air gap and the refraction index of the prism to yield agreement between the dispersion curves calculated by the ATR method and the exact solutions (see Fig. 4.8). To nd the optimal air gap and the refraction
equation, into the transfer matrix formalism, and solve the zero-refraction equation for the air gap and the refraction index of the prism. The optimal air gap and index of the prism are shown in Fig. 4.8(b) and Fig. 4.8(c), respectively. It can be seen that the range of the optimal air gap is between 0.5 cm and 3.5 cm (
/12 /2,where
1
is valid
= 6.25 cm), and it is larger for the forward propagating mode than for the backward
propagating mode. It seems that the simple rule mentioned previously,
here. However, the result for the optimal refraction index is more interesting. For the forward propagating mode, the optimal index starts just a little above the vacuum index refraction (1) and slowly increases with increasing frequency.
We can explain
this result by the fact that the prism perturbs the waveguide structure. If the index of the prism is close to that of the waveguide cladding, then the eect of the prism perturbation can be minimized. However, to satisfy exciting the mode with propagation constant as the
/c < nprism ,
of the mode is increased. This is the reason why the backward propagating
modes need a high-index prism to be excited. The point that is not clear for us in Fig. 4.8(c) is why is the optimal index of refraction of the prism is so large for the complex modes. For future work, the excitation of the complex modes with the ATR method needs to be studied further.
1d
/2/4 for the forward propagating mode and d /4/8 for the backward propagating
CHAPTER 4.
94
the Goos-Hanchen (GH) shift [72] and happens because the interference of dierent transverse wave vectors are undergoing dierent phase changes. The GH shift was rst predicted by Newton; however, it took two centuries before Goos and Hanchen could experimentally proved this lateral beam shift. In a dielectric medium the GoosHanchen shift is usually positive; however, it could be negative in a lossy multilayer dielectric structure. In 2002, Berman showed that the GH shift is negative if one of the interfaces is a metamaterial [36]. Actually, the negative sign of the permeability and permittivity in metamaterials lead to the negative GH shift. After Berman's
paper, much attention was drawn to the GH shift in metamaterials (for example [37, 87, 88]). Chen et al. showed that the GH shift can be negative as well as positive in a metamaterial slab and Kiveshar et al. [37] predicted that the GH shift can be giant
Usually, the GH shift is much less than the beam width and cannot be easily
3 It is comparable or even larger than the wavelength of a beam (see Fig. 4.10).
CHAPTER 4.
95
Incident beam
Reflected beam
Reflected beam
n1 = 1
GH shift
air gap z x
n1 = 1
GH shift
air gap z x
n2 < 0
metamaterial
n2 < 0
metamaterial
Figure 4.9: Schematic of the positive and negative lateral beam shift (Goos-Hanchen shift).
measured in practice; however, a large GH shift can be seen in a multilayer structure whenever surface modes of a structure are excited. Since the excited mode can transfer energy along the interface, the lateral GH beam shift signicantly intensies. In the following, we calculate the GH shift for the structure determined in Fig. 4.4. We have
T E1
=6
= 25 .
GH shift in respect to the wavelength of the light beam versus the angle between 20 to 30 degree in Fig. 4.10. The Goos-Hanchen shift can analytically be calculated as [87]
=
where
, 2
(4.20)
reection coecient.
/ 19,
CHAPTER 4.
96
= 25 .
= 24 ,
which corresponds to
the critical total internal reection angle. It is important to note that the GH shift is sensitive to the prism air gap. In our calculation, we derived the optimal air gap using the zero-reection condition, which is 1.89 cm for the forward propagating mode.
T E1
5 c 0
5 GH/ |R|
10
0.5
15
0 20
40 60 (degree)
80
20 20
22
24 26 (degree)
28
30
Figure 4.10: The lateral Goos-Hanchen (GH) shift for the reected beam of the system described in Fig. 4.4. Inset shows the reectance with the optimal air gap (d cm).
= 1.89
CHAPTER 4.
97
integral part of the ATR method and it can be found by using the zero-reection condition. Although the modes obtained from the ATR method under the zero-
reection condition give a good agreement with the exact modes in a structure with a small intrinsic loss, the modes obtained by the ATR method considerably depart from the exact modes in a waveguide structure with a moderate loss. We have suggested that in a lossy waveguide structure and for the complex modes, the index of refraction of the prism also needs to be optimized. Finally, we have shown that a large GoosHanchen shift can be seen at the excited mode angle as the sign and magnitude of the GH shift is sensitive to the air gap and index of refraction of the prism.
Chapter 5
5.1 Introduction
The dispersion curves of a waveguide structure can be obtained from the dispersion equation,
D(, ) = 0,
where
The modes, in general when intrinsic loss is introduced in a structure, are stated with complex
and complex
curves of a metamaterial waveguide structure by looking for the modes with a complex and
and a real
( ).
complex
real, or vice-versa, or even to make both of them complex [89, 90, 91, 92]. In 98
CHAPTER 5.
99
this chapter, we calculate the dispersion curves of a metamaterial waveguide for real
and complex
( )
for several metamaterial waveguide structures; and compare subspace. We nd a signicant dierence between the dis-
The dierence is understandable since these two approaches solve fundamentally different problems. The real frequency approach
( )
wave decays in space, while the real wave vector approach tions that decay away in time.
( )
( )
experimentally when the structure is excited with monochromatic light, and can be computed using frequency-domain methods, such as the vectorial eigenmode expansion technique [89]. However, the real wave vector approach
( )
can be realized
when the structure is excited with highly-collimated ultra-short pulse, and can be calculated using frequency-dependent time-domain method computationally. Next, we show that the dispersion curves of both
and
approaches can be
reproduced by using the ATR method, but with dierent experimental conditions. For the
xed and the angle varied, while for the latter approach, we should measure the ATR reectivity with the angle xed and the frequency varied [79]. Thus, one can measure the dispersion curves of both types of modes experimentally. Finally, we calculate the dispersion curves of a lossy metamaterial waveguide structure employing perturbation theory for both these approaches. We show that there is some disagreement between the solutions obtained by perturbation theory with the exact solutions near the stopped-light mode regions in the
approach; however,
the perturbation theory can work well in the other regions. As we have mentioned in
CHAPTER 5.
100
the previous chapters, the inuence of the material loss on mode structure is highly non-perturbative near the stopped-light mode regions, so this disagreement is not surprising.
In this section, we numerically investigate the dispersion curves of the systems that have been studied in 2.5, both in complex- and complex- approaches
( and ).
Light line
5
TE3
(a)
Frequency (GHz)
TE
5
Light line
(b)
Frequency (GHz)
TE2
TE2 TE
TE1
0 0
0.5
1.5
(1/cm)
2.5
3.5
0 0
0.5
1.5
(1/cm)
2.5
3.5
Figure 5.1:
= = 4 (a) for the complex- approach ( ), and (b) for the complex- approach ( ) . The thick and the thin curves indicate the real and imaginary parts of for (a) and frequency for (b), respectively.
First, we consider a 2-cm-thickness metamaterial slab with
= = 4
in free
space. Fig. 5.1 indicates the calculated normal and complex modes of the proposed structure for (a) complex- and (b) complex- states. Since the proposed model is lossless, the calculated normal modes of the metamaterial waveguide structure for the
Figure 5.1: Dispersion curves of the dispersionless and lossless metamaterial slab in the free space with imaginary parts of the complex- approach
= = 4 (a) for the complex- approach ( ), and (b) for ( ) . The thick and the thin curves indicate the real and for (a) and frequency for (b), respectively.
CHAPTER 5.
101
that the complex modes are completely dierent between the 5.1(a) shows that the real part of the wave vector,
and
states. Fig.
the cut-o frequency, the critical point where the frequency has its minimum value and the group velocity is zero, and monotonically decreases. On the other hand, in Fig. 5.1(b) the real part of the frequency for the complex modes starts at the cut-o propagation constant, the critical point where the energy density is zero and and goes to zero when
vg ,
0.
almost above the light-line in the complex- state, and look like the leaky modes in a conventional waveguide, while the complex modes in the complex- state are not restricted above the light-line. However, the complex modes of a dispersionless metamaterial waveguide in both approaches are the proper solution of the dispersion equation as we have previously shown for
CHAPTER 5. DISPERSION CURVES OF MWS WITH COMPLEX-
approach.
117
0.35 4.7 Re[f] (GHz) 4.6 4.5 4.4 4.3 4.2 4.1 0 1 TEproper 2 TEImproper
2
(a)
(b)
0.05
2 (1/cm)
4.1
Figure 5.2: The dispersion curves of the dispersive, lossless metamaterial waveguide p for the T E2 mode with F = 0.56, = 10 GHz, and 0 = 4 GHz for the Lorentz2 2 Drude model of the metamaterial slab and with air claddings (a) for a real- , complex-
and (b) a real- , complex- approach. The blue and red curves indicate the proper
Im[f] (GHz)
0.25
CHAPTER 5.
102
T E2
mode in the
T E2
mode, the unphysical solution of the dispersion equation which violates the guidance condition (see section 2.4), in Fig. 5.2 (red curves). As expected, we have the identical dispersion curves for the normal
mode in the real- and the complex- states; however, compared with the previous case, (the dispersionless waveguide structure), we nd that the dierence between the complex- and the complex- states in the complex
T E2
striking. While analytically we have previously shown that the complex modes are always the proper solutions in the complex- state, Fig. 5.2(b) indicates that the
complex modes are the improper solutions in the complex- state. In other words, the complex modes of a dispersive metamaterial waveguide surprisingly disappear in the complex- approach. Finally, we compare the dispersion curves of a lossy and dispersive metamaterial waveguide in the complex- and complex- states. In Fig. 5.3, we show the
T E2
in
= 0.1 GHz
and
the intrinsic loss modies considerably the dispersion curves in the complex- state; for example, the curves that bend back near the resonance frequency or stopped-lightmode regions are completely destroyed in the presence of material loss. However, the dispersion curves, astonishingly, remain almost unchanged by introducing the material loss for the complex- approach. Fig. 5.3(c) clearly indicates that the real part of dispersion curves of the lossy metamaterial waveguide with
CHAPTER 5.
103
(a)
frequency (GHz)
complex approach
4.4
frequency (GHz)
4.7
4.7
4.4
4 0 2 4 Re[] (1/cm) 6 7
4 0
3
4 Im[] (1/cm)
8.5
x 10
(c)
4.7 Re[f] (GHz) Im[f] (GHz)
4.4
complex approach
4 0 2 4 (1/cm) 6 7
10.5
4 (1/cm)
Figure 5.3: The dispersion curves of the dispersive, lossy metamaterial waveguide for
= 0.1 GHz. The red curve indicates (a) the real and (b) the imaginary part of in a subspace. The blue curve indicates (c) the real and (d) the imaginary part of frequency in a
the mode with the parameters given in the Fig 5.2, and with subspace.
T E2
those of the lossless metamaterial waveguide. We have repeated our calculation for the proposed waveguide with the higher loss in the complex- approach, and highlight that the real part of the dispersion curves are evidently insensitive to the material loss in this approach. Hence, the point with zero slope survives in the presence of intrinsic loss in the complex- approach; however, the mode is very lossy at this critical point (see Fig. 5.3(d)). It is important to note that the conventional denition of the power ux, energy density, and energy velocity are not correct in the complex- approach;
CHAPTER 5.
104
hence, the derivation of their correct forms are indeed an interesting problem for future work.
10
10
10 Loss (dB/cm)
10
1
10
10 10
0
10
4.3
4.32
4.4
10
Figure 5.4: The group velocity (blue) and the loss propagation (red) of the metamaterial waveguide proposed in Fig. 5.2 for the critical frequency,
T E2 mode in the complex- approach. At 4.4118 GHz, the group velocity becomes zero while loss is innitive.
We calculate the group velocity and the loss propagation of the aforementioned lossy metamaterial waveguide for the bound
T E2
|vg|/c
over a broadband range of frequencies in Fig. 5.4. At the critical frequency point,
f=
4.4118
GHz, the group velocity is zero; however, the loss propagation is enormously
f = 4.4118
or
GHz. It is well
the small or even zero group velocity value in the complex- approach is, indeed, unphysical, and it does not show the correct measure of transport. We now derive the relationship between the dispersion curves in the complex-
CHAPTER 5.
105
and the complex- approaches. We follow a similar analysis as in [89] . In general, both
and
D(r + ii , r + ii ) = 0.
(5.1)
and
D(, ) = D(0 , 0 ) + ( 0 )
So, using Eq. (5.2), we have
0 ,0
+ ( 0 )
.
0 ,0
(5.2)
D(0 , 0 + r + ii ) = D(0 , 0 ) + (r + ii )
D D
= 0,
0 ,0
(5.3)
D(0 + r + ii , 0 ) = D(0 , 0 ) + (r + ii )
= 0.
0 ,0
Hence,
vg
=
0 ,0
D D /
=
0 ,0
r + ii . r + ii
(5.4)
If the dispersion curves in the complex- and complex- coincide at a point (0 , 0 ), we have
i =
i . vg
and
(5.5)
Eq. (5.5) gives us the relationship between the imaginary part of this equation is only valid when the intrinsic loss is small. In Fig. 5.5, we show the imaginary part of
However,
CHAPTER 5.
106
4.4
4.35 4.41 Frequency (GHz) 4.3 4.4 4.25 4.39 4.2 0.1 4.15 0.2 0.3
4.1
0.1
0.2
0.4
Figure 5.5: The imaginary part of the propagation constant, space for the
in the complex-
T E2
mode of the waveguide proposed in Fig. 5.2. The solid blue curve
represents the exact solutions of the imaginary part of represents the imaginary part of highlighted in the inset.
light mode region), where there is some disagreement between the two solutions, is
for the
T E2
sion equation (blue curve). Next, we derive the imaginary part of group velocity and the imaginary part of
by inserting the
by the red dashed curve). Fig. 5.5 demonstrates that the relationship in Eq. (5.5) holds well except near the critical frequency region where
vg 0.
velocity has a small value and the intrinsic loss has non-perturbative inuence near the critical frequency region, it is expected that Eq. (5.5) will fail. It is interesting to note that the relationship in Eq. (5.5) is not valid in a photonic crystal medium near the band edge, where the group velocity also has a small value [89].
CHAPTER 5.
107
the real part of propagation constant can be found from the momentum conservation equation
(5.6)
We have assumed that the incident frequency is xed and that the incident angle is varied in the experiment. Since the minimum of the reectance is associated with
excitation of the modes, the dispersion curves in the complex- space can be found from Eq. ( 5.6). However, if we are interested in reproducing the dispersion curves in the complex- space by using the ATR method, we have to change the initial condition in the ATR experiment. This time, we x the incident angle and vary the incident frequency; hence, the
(5.7)
It is interesting to note that this issue was discussed extensively in 1970's. Though calculated dispersion curves of a lossy metallic waveguide demonstrated that the surface plasmon modes bend back at dispersion curve did not bend back at
fms , fms .
Otto [80] found that the experimental However, Arakawa et al. reported that
the curve did bend back in their experiment [93]. Several years after Otto's paper, Kovener et al. [79] found that the two diering results of Otto's experiment and
CHAPTER 5.
108
Arakawa's experiment are not contradictory, and both of them satisfy the Fresnel equations. However, Otto's method gives the dispersion curves in the complex-
space while the Arakawa's method gives the dispersion curves in the complex- space.
4.8 TE0
= 58 = 70
1 (a) 0.8 |R| 0.6 0.4 Akawaras method 40 f = 4.715 B f = 4.725 A 60 (deg) 80 100 (b)
0.2 20 1
0.8
C B
np = 8 (c) np = 7
0.6
np=6 np=7 np=8
4.65 0
4 6 (1/cm)
T E0 surface polariton mode of the metamaterial waveguide proRe[] bends back near fms in the complex- space (blue curve) while Re[] at fms in the complex- space (red curve). (b) Reectance versus the incident angle with nprism = 6 and dgap = 0.5 cm. As expected from (a), min 70 . (c) Reectance versus frequency with dgap = 0.5 cm and = 70 . The labeled letters, (A,B,C), represent the resonance frequencies for nprism = 6, 7, 8, reFigure 5.6: (a) The posed in Fig 5.2. The spectively.
The same scenario arises in a lossy metamaterial waveguide. The surface polariton modes bend back near
fms
in the complex- approach while they do not bend back In Fig. 5.6, we show this phenomenon by using the
and absorption with air claddings. The blue curve in Fig. 5.6(a) indicates the
T E0
CHAPTER 5.
109
T E0
magnetic surface polariton frequency, 4.71 GHz, and couples with the However, in the complex- approach, the
T E4
mode.
T E0
at fms . First, we simulate Arakawa's experiment with frequency xed and the
incident angle varied. We need a prism with an extremely high index of refraction to excite surface polariton modes at a region where the dispersion curve bends back. The dashed lines in Fig. 5.6(a) are the prism light lines at indices of refractions. excite the As shown in Fig.
= 70
nprism < 6
T E0
>6
(1/cm). Thus, we set a prism 0.5 cm above We then calculate the reectance for
nprism = 6.
the frequency xed at 4.4725 GHz. The minimum reectance is at is in good agreement with the calculated dispersion curve,
68 ,
which
exact = 70 ,
(where the
prism light-line and blue dispersion curve cross each other in Fig. 5.6(a)). We repeat our calculation for the frequency xed at 4.4715 GHz. The minimum reectance is now at
53 ,
5.6(b)).
exact = 58 ,
surface polariton mode at this frequency. However, this ATR technique veries that the dispersion curve bends back because the location of the dip goes to lower angles by decreasing frequency. Next, we simulate the ATR technique proposed by Otto to measure the
for the
aforementioned metamaterial waveguide structure. The distance between the prism and the waveguide is again xed to be 0.5 cm, and the beam propagates from the prism with the incident angle xed, indices of refractions,
x = 70 .
nprism = 6, 7, 8;
CHAPTER 5.
110
As expected, the resonance frequencies decrease by increasing the index of refraction (see Fig. 5.6(a) and (c)). If the dispersion curve bents back, we should not observed the minimum reectance for
nprism = 7 or 8.
that the Otto's ATR technique does not work well for the dispersion curves of a lossy waveguide in the complex- approach if the loss is too high. By looking at Fig. 5.6(c), we nd that the magnitude and the width of the dips are changed by modifying the index of refraction. Since the propagation loss (e.g. see Fig. 2.5(b)) signicantly
increases from point A to point C, the width of the dip labeled C is greater than the dip labeled A. Finally, by the same method, we can show that the bound modes bend back at the resonance frequency in the complex- approach while they do not bend back in the complex- approach. In Fig. 5.7, we show the role of material loss on the reectance spectrum. It
is well-known that we cannot use the ATR method to measure the dispersion curve of a lossless waveguide structure. If a waveguide is lossless, the reectance would The
be unity because no energy could be coupled to the surface or bound mode. insets in Fig. 5.7 show the with
T E2
= 0.1
= 0.01
the critical frequency region, (stopped-light mode region), in both complex- and complex- approaches. We have previously shown that the dispersion curves in the presence of material loss drastically pull away from the lossless counterpart near the critical frequency in the complex- approach while they are almost unchanged in the complex- approach. incident frequency xed, In Fig. 5.7(a), we calculate the reectance with the GHz, by using Akawara's ATR method for the GHz, (blue), and
f = 4.4118
= 0.1
= 0.01
CHAPTER 5.
111
1 0.9 0.8
(a)
4.41
0.9 0.8
(b)
4.42
0.6
4.4
Re[f]
0.7
0.7 0.6
4.4 1 1.5 2
|R|
|R|
1.5
Re[]
complex approach
0.2 0.1
complex approach
(deg)
60
70
80
0 4.38
4.4
4.42
4.44
f (GHz)
Figure 5.7: (a) Reectance versus the incident angle for the metamaterial waveguide
= 0.1 GHz, (blue solid curve), and = 0.01 GHz, (red dashed curve) with prism = 7, and dgap = 1.8 cm at f = 4.4118 GHz. The inset shows the dispersion curves of the proposed structure near the critical frequency region for the T E2 mode in the complex- approach. (b) Reectance versus frequency with the parameters used in (a) at = 37 . The inset indicates that the dispersion curves are almost insensitive to material loss in the complex- approach.
with 5.7(a) indicates that the value of
reectance minima. The two reectance minima move toward each other when the value of
decreases.
aects the magnitude and the width of the reectance dip. This result is acceptable because Otto's ATR method gives the dispersion curves in the complex- approach, and we know that the dispersion curves in this approach is insensitive to the material loss. Hence, Otto's method is better to use when we are interested in reproducing the dispersion curves of a lossless waveguide.
CHAPTER 5.
112
5.2 5 Re[f] (GHz) 4.8 4.6 4.4 4.2 4 1 4 2 (1/cm) 3 4 2 3 1 = 45 0 (a) Reflectance |R|
1 (b) 4 3 0.5 2
= 26 = 39
f = 4.4 B
0.5
B F
2 (1/cm)
0 20
40
60 (deg)
80
Figure 5.8: The dispersion curves of the metamaterial waveguide structure proposed in Fig. 5.2 for
TE
approach. (b) Reectance versus frequency with the following parameters: prism = 9, dgap = 1 cm, and = 45 . The marked numbers represent the resonance frequencies for
T E0 , ..., T E4
modes. (d) Reectance versus the incident angle at 4.4 GHz for mode, respectively.
T E2
mode. The B and F marks represent the resonance for the backward and forward propagation
T E2
To obtain better insight into the dierence between Otto's ATR method and Arakawa's ATR method, we plot the complete dispersion curve spectrum of the aforementioned lossy metamaterial waveguide with
= 0.1
complex- (Fig. 5.8(a)) and complex- (Fig. 5.8(c)) approaches. Although we have not optimized the air gap, a distance between the prism and the waveguide, and the prism index of refraction, there is a good agreement between the calculated curves in Fig. 5.8(a) and the resonance frequencies in Fig. 5.8(b). The marked numbers in Fig. 5.8(b) represent the order of TE modes. The small unmarked dips on the right corner
CHAPTER 5.
113
of the Fig. 5.8(b) correspond to the higher order TE modes, which are not plotted in Fig. 5.8(a); however, they are dense near the resonance frequency. It is important to highlight that the resonance frequencies obtained by the Otto's ATR method correspond to the dispersion curves of a lossy structure in the complex- approach and not to the dispersion curves of a lossless structure in the complex- approach. Since the dispersion curves of a lossy structure in the complex- approaches are similar to the dispersion curves of a lossless structure in the complex- approach (compare Fig. 5.8(a) with Fig. 2.5(a)), this lead to a misinterpretation of Otto's ATR method in the literature [21, 78, 79]. We note that under no conditions can the complex modes be detected with Otto's ATR method. This is acceptable because the complex modes are not supported in the complex- approach. However, if Otto's ATR method corresponds to the dispersion curves of a lossless waveguide in the complex- approach, the complex modes should be observed. As expected, the complex modes can be detected by using Arakawa's ATR method; however, the index of refraction of the prism must be high to detect them. We have discussed this issue previously in Chapter 4.
When we introduce the small amount of loss into a system, the permeability and permittivity will be slightly changed (
+ , + ).
CHAPTER 5.
114
(A
(0)
(0)
+ A) = (B
(0)
+ B) ,
(5.8)
where
(1)
+ ...
and
(0) (0) (1) (0) (0) (0) (0) (1) A + A = (1) B + (0) B + (0) B .
(5.9)
By using the Hermitian property of the operators and taking the inner product on both side of Eq. (5.9), the generalized form of the perturbation theory is
(0)
(1)
(0)
(0) (0)
(0)
(0)
(5.10)
and , z 0
we get
B = 0
and
A =
z +
(
t ) c
0
c t
z +
t )
(5.11)
Inserting Eq. (5.11) and Eq. (A.14) into Eq. (5.10), the perturbation equation for the TE modes is
(1)
= 4c
(5.12)
In Fig. 5.9, the blue curves show the real and the imaginary components of the
CHAPTER 5.
CHAPTER 5.
115
(a)
f (GHz)
4.4
(b)
4.3 4.2 4.1
3 4 Re[]
0.1
0.4
Figure 5.9: The blue and red curves represent the exact and perturbation solutions of the waveguide proposed in Fig. 5.3 for (a) the real part and (b) the imaginary part of the
T E2
T E2
mode in the complex- space, and the dashed red curves show the dispersion curves of that structure calculated by perturbation theory (Eq. (5.12)). Fig. 5.9 indicates that the dispersion curves calculated by perturbation theory agree with the exact dispersion curves solutions except near the critical frequency region. Moreover, the complex modes cannot be derived by perturbation theory because the denominator of Eq. (5.12) is always zero for the complex modes. From the standard perturbation theory [51, 58, 95], it can be shown that the perturbation equation in the complex- space is
(1)
(0) = 2
H 2 (x)]dx + z
(5.13)
Figure 5.9: The blue and red curves represent the exact and perturbation sol of the waveguide proposed in Fig. 5.3 for the
T E2
CHAPTER 5.
116
However, Eq.
Finding an ap-
propriate perturbation equation for a dispersive waveguide structure is an interesting problem for future work.
in general are not the same because these approaches fundamentally solve dierent problems. We have also shown that the dispersion curves of MWS in a complex
propagation constant approach and a complex frequency approach can be reproduced by using the ATR method but with dierent experimental conditions.
Chapter 6
In this thesis, we have investigated the optical properties of metamaterial waveguide structures. We have shown that the negative-index of refraction of a metamaterial leads to unusual optical properties, which cannot be seen in a conventional waveguide structure. We have calculated the dispersion curves of the metamaterial waveguide structure and demonstrated a rich variety of modes, such as surface polariton modes, complex modes, backward propagating modes, and even stopped-light modes that can be supported by metamaterial waveguide systems. Most of the unique properties of the metamaterial waveguide come from the fact that the energy ux in the core ows in the opposite direction to that in the cladding. Subsequently, we have examined these unusual electromagnetic properties by considering the role of dispersion and loss, which are inherent features of metamaterials. Our results demonstrated
that the intrinsic loss signicantly modies the dispersion curves and consequently the properties of the metamaterial waveguide structures. 117 Moreover we have found
CHAPTER 6.
several novel properties in the presence of loss, such as a splitting of the complex modes, bending back of the modes near the resonance frequency, and disappearance of the stopped-light modes. One of our more useful and important result is that we have explicitly shown that light cannot be slowed down in a realistic metamaterial waveguide structure. In summary, the dispersion curves of a lossless metamaterial
waveguide structure near the slow-light-mode domains are strongly modied when arbitrarily small amounts of loss are introduced into the system. We also studied the ATR method to excite the modes of the metamaterial waveguide structure. Although the ATR method is a useful tool to excite the normal modes in a waveguide with a low loss, this technique requires satisfying some non-trivial conditions in order to be applicable for complex modes or the normal modes in a lossy metamaterial waveguide. We also have shown that the excitation of the modes leads to anomalous lateral beam phase shifts. In summary, the ATR method is only a
useful tool to obtain the dispersion curves of a lossy metamaterial waveguide in the complex- approach and those of a metamaterial waveguide with very small loss in the complex- approach. Finally, we have shown that one can study the modes of a lossy waveguide using two approaches: a complex frequency approach and a complex propagation constant approach. The results of these two approaches are not the same; thus, one needs to obtain the dispersion curves of a waveguide system in both these approaches to fully study the optical properties of the system. Future work on metamaterial waveguide structures could include examining the eects of anisotropy and gain and further investigation of the complex- modes.
Bibliography
[1]
J. B. Pendry, Negative refraction makes a perfect lens, Phys. Rev. Lett. 85, 3966 (2000).
[2]
A. Alu, N. Engheta, Achieving transparency with plasmonic and metamaterial coatings, Phys. Rev. E 72, 016623 (2005).
[3]
[4]
[5]
A. Alu, N. Engheta, Guided modes in a waveguide lled with a pair of singlenagative (SNG), double-negative (DNG), and/or double-positive (DPS) layers,
IEEE Trans. Microwave theory Technol. 52, 199 (2004).
[6]
M. C. K. Wiltshire, J. B. Pendry, I. R. Young, D. J. Larkman, D. J. Gilderdale, J. V. Hajnal, Microstructured magnetic materials for RF ux Guides in magnetic resonance imaging, Science 291, 849, (2001).
[7]
119
BIBLIOGRAPHY
120
[8]
J. C. Bose, On the rotation of plane of polarization of electric waves by a twisted structure, Proc. R. Soc. 63, 146 (1898).
[9]
W. E. Kock, Mettalic delay lenses, Bell Syst. Tech. J. 27, 58, (1948).
[10] D. R .Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, S. Schultz, Composite medium with simultaneously negative permeability and permittivity, Phys.
Rev. Lett. 84, 4184 (2000).
[11] J. B. Pendry, A. J. Holden, D. J. Robbins, W. J. Stewart, Magnetism from conductors and enhanced nonlinear phenomena, IEEE Trans. Microw. Theory
Tech. 47, No.11, 2075 (1999).
and
[15] R. A. Shelby, D. R. Smith, S. Schultz, Experimental verication of a negative index of refrection, Science 292, 77 (2001).
[16] G. Dolling, C. Enkrich, M. Wegener, C. M. Soukoulis, and S. Linden, Negativeindex material at 780 nm wavelength, Opt. Lett. 32, 53 (2007).
[17] V. M. Shalaev, W. Cai, U. K. Chettiar, H. Yuan, A. K. Sarychev, V. P. Drachev, and A. V. Kildishev, Negative index of refraction in optical metamaterials, Opt.
Lett. 30, 3356 (2005).
BIBLIOGRAPHY
121
[18] S. Zhang, W. Fan, N. C. Panoiu, K. J. Malloy, R. M. Osgood, Experimental Demonstration of Near-Infrared Negative-Index Metamaterials, Phys. Rev. Lett.
95, 137404 (2005).
[20] C. M. Soukoulis, S. Linden, M. Wegener, Negative refractive index at optical wavelengths, Science 315, 47 (2007).
277,
61
[22] J. Gollub,D. Smith,D. Vier,T. Perram,J. J. Mock, Experimental characterization of magnetic surface plasmons on metamaterials with negative permeability, Phys.
Lett. B 71,195402 (2005).
[23] I. Wu,T. M. Grzegorczyk, Y. Zhang, J. A. Kong, Guided modes with imaginary transverse wave number in a slab waveguide with negative permittivity and permeability, J. Appl. Phys. 93,9386 (2003).
[24] I. V. Shadrivov, A. A. Sukhorukov, Y. S. Kivshar, Guided modes in negativereftactive-index waveguides, Phys. Rev. E 67,057602 (2003).
[25] L. Zhou, C. Chan, Vortex-like surface wave and its role on the transient phenomena of metamaterial focusing, Appl. Phys. Lett. 86,101104(2005).
[26] P. Baccarelli, P. Burghignoli, F. Frezza, A. Galli, G. Lovat, S. Paulotto, Fundamental modal properties of surface waves on metamaterial grounded slab, IEEE
Trans. Microwave Theory Tech. 53, 1431-1442 (2005).
BIBLIOGRAPHY
122
[28] I. V. Shadrivov, A. A. Sukhorukov, Y. S. Kivshar, Complete Band Gaps in OneDimensional Left-Handed Periodic Structures, Phys. Rev. Lett. 95,193903(2005).
[29] J. Li, L. Zhou, C. T. Chan, and P. Sheng, Photonic Band Gap from a Stack of Positive and Negative Index Materials,Phys. Rev. Lett. 90, 083901 (2003).
[30] H. Jiang, H. Chen, H. Li, Y. Zhang, S. Zhu , Omnidirectional gap and defect mode of one-dimensional photonic crystals containing negative-index materials,
Appl. Phys. Lett. 83, 5386 (2003).
[31] J. P. Xu, Y. P. Yang, H. Chen, S. Y. Zhu, Spontaneous decay process of a two-level atom embedded in a one-dimensional structure containing left-handed material, Phys. Rev. A 76, 063813 (2007).
[32] J. Kastel, M. Fleischhauer, Suppression of spontaneous emission and superradiance over macroscopic distances in media with negative refraction, Phys. Rev. A
71, 011804 (2005).
[33] Q. Cheng, T. J. Cui, High-power generation and transmission through a lefthanded material, Phys. Rev. B 72, 113112 (2005).
[34] Y. H. Yao, T. J. Ciu, Q. Cheng, R. Liu, D. Huang, D. R. Smith, Realization of a super waveguide for high-power-density generation and transmission using right and left-handed transmission-light circuits, Phys. Rev. E 76, 036602 (2007).
[35] K. L. Tsakmakidis, A. D. Boardman, O. Hess, Trapped rainbow storage of light in metamaterials, Nature 450, 397 (2007).
BIBLIOGRAPHY
123
[36] P. R. Berman, Goos-hanchen shift in negative index media,Phys. Rev. E 66, 067603 (2002).
[37] I. V. Shadrivov, A. A. Zharov, Y. S. Kivshar, Giant Goos-Hanchen eect at the reection from Left-handed metamaterials, Appl. Phys. Lett. 83, 2713 (2003).
[38] J. He, S. He, Slow propagation of electromagnetic waves in a dielectric slab waveguide with a left-handed material substrate, IEEE microwave and wireless
components letters, Vol. 16, No. 2, (2005).
[39] Q. Gan, Z. Fu, Y. J. Ding, F. J. Bartoli, Ultrawide-bandwidth slow-light system based on THz plasmonic graded metallic grating structures, Phys. Rev. Lett.
100, 256803 (2008).
[40] L. V. Alekseyev, E. Narimanov , Slow light and 3D imaging with non-magnetic negative index systems, Optics Express, Vol. 14, Issue 23, pp. 11184-11193 (2006).
[41] Y. J. Huang, W. T. Lu, S. Sridher, Nanowire waveguide made from extremely anisotropic metamaterials, Phys. Rev. A 77, 063836 (2008).
[42] T. Jiang, Y. J. Feng, Slow and frozen waves in a planar air waveguide with anisotropic metamaterial cladding, CMMT 450775 (2008).
[43] J. He, Y. Jin, Z. Hong, S. He , Slow light in a dielectric waveguide with negativerefractive-index photonic crystal cladding, Optics Express, Vol. 16, Issue 15, pp. 11077-11082 (2008).
BIBLIOGRAPHY
124
[45] J. B. Pendry and D. R. Smith, Reversing Light With Negative Refraction, Phys.
Today 57(6), 37 (2004).
[46] J. Valentine, S. Zhang, T. Zentgraf, E. Ulin-Avila, D. A. Genov, G. Bartal and X. Zhang, Three-dimensional optical metamaterial with a negative refractive index,
Nature 455, 376 (2008).
[47] A. Cho, Bizarre 'Metamaterials' for Visible Light in Sight?, Science 321, 900 (2008).
[48] M. I. Stockman, Criterion for negative refraction with low losses from a fundamental principle of causality, Phys.Rev.Lett. 98, 177404 (2007).
[49] A. Reza, M. M. Dignam, S. Hughes, Can light be stopped in realistic metamaterials?, Nature 455, E-10 (2008).
[53] A. Peacock, N. Broderick, Guided modes in channel waveguides with a negative index of refraction, Optics Express, Vol. 11, No. 20, 2502 (2003).
[54] K. L. Tsakmakidis, A. Klaedtke, D. Aryal, O. Hess, Single-mode operation in the slow-light regime using oscillatory waves in generalized left-handed hetrostructures, J. Appl. Phys. 89, 201103 (2006).
BIBLIOGRAPHY
125
[55] W. Shu, J. M. Song, Complete mode spectrum of a grounded dielectric slab with metamaterials, PIER 65, 103-123 (2006).
[56] Q. Sui, F. Li, Complex guided wave soloutions of grounded dielectric slab made of metamaterials, PIER 51, 187-195 (2005).
[57] R. Chern, C. C. Chang, Surface and bulk modes for periodic structures of negative index materials, Phys. Rev. B 74, 155101(2006).
[58] M. Skorobogatiy, M. Ibanescu, S. G. Johnson, O. Weisberg, T. D. Engeness, M. Soljacic, S. A. Jacobs, Y. Fink, Analysis of general geometric scaling fundamental connection between
[59] S. G. Johnson, M. Ibanescu, M. A. Skrobogatiy, O. Weisberg,T. D. Engeness, M. Soljacic, S. A. Jacobs, J. D. Joannopoulos, Y. Fink, Low-loss asymptotically single-mode propagation in large-core Omniguide brs, Opt. Express 9, 748 (2001).
[61] J.l u, B. Wu, J. A. Kong, Guided modes with a linearly varying transverse eld inside a left-handed dielectric slab, J. Electro. Waves and Appl. 20, 689-697 (2006).
BIBLIOGRAPHY
126
[63] K. Halteman, S. Feng, P. L. Overfelt, Guided modes of elliptical metamaterials waveguides, Phys. Rev. A 76, 013834 (2007).
[64] A. Govyadinov, V. Podolskiy, Gain-assisted slow to superluminal group velocity management in nano-waveguides, Phys. Rev. Lett. 97, 223902 (2006).
[65] I. V. Shadrivov, A. A. Sukhorukov, Y. S. Kivshar, Nonlinear surface waves in left-handed materials, Phys. Rev. E 69, 016617 (2004).
[67] P. W. Milonni, Fast light, slow light and left-handed light, (IOP publishing, London, 2005).
[68] T. J. Cui, J. A. Kong, Time-domain electromagnetic energy in a frequencydispersive left-handed medium, Phys. Rev. B 70, 205106 (2004).
[70] R. Ruppin, Electromagnetic energy density in a dispersive and absorptive material, Phys. Lett. A 299, 309 (2002).
[71] J. W. Dong, H. Z. Wang, Slow electromagnetic propagation with low group velocity dispersion in an all-metamaterial-based waveguide, Appl. Phys. Lett.
91, 111909 (2007).
BIBLIOGRAPHY
127
[73] R. W. Ziolkowski, Superluminal transmission of information through an electromagnetic metamaterial, Phys. Rev. E 63, 046604, (2001).
[74] S. Hughes, L. Ramunno,J. Young, J. E. Sipe, Extrinsic optical scattering loss in photonic crystal waveguides: role of fabrication disorder and photon group velocity, Phys. Rev. Lett. 94, 033903 (2005).
[75] A. Borde, L. H. Ford, T. A. Roman, Constraints on spatial distributions of negative energy, Phys. Rev. D 65, 084002 (2002).
[76] P. Yao, Z. Liang, X. Jiang, Limitation of the electromagnetic cloak with dispersive material, Appl. Phys. Lett. 92, 031111 (2008).
[77] R. Ruppin, Surface polaritons of a left-handed material slab, J. Phys. Condens.Matter 13,1811 (2001).
[78] K. Park,B. J. Lee,C. Fu,Z. M. Zhang, Study of the surface and bulk polaritons with a negative index metamaterials, J. Opt. Soc. Am. B /Vol. 22,No. 5 (2005).
[79] G. S. Kovener,R. W. Alexander,R. J. Bell, Surface electromagnetic waves with damping. I. Isotropic media, Phys. Rev. B 14, 1458-1464 (1976).
[80] A. Otto, Excitation of nonradiative surface plasma waves in silver by the method of frustrated total relection, Z. Phys. 216,398-410 (1968).
[81] H. Kitajima, K. Hieda,Y. Suematsu, Optimum condition in the attenuated total reection technique, Appl. Opt. 20, 1005-1009 (1981).
BIBLIOGRAPHY
128
[82] M. O. Petersen,B. S. Zhu,E. Dalsgaard,Extreme attenuation of total internal reection used for determination of optical properties of metals, J. Opt. Soc.
Am. A / Vol. 4,No. 9 (1987).
[83] H. Kitajima, K. Fujita, H. Cizmic, Zero reection from a dielectric lm on metal substrate at oblique angles of incidence, Appl. Opt. 23, 1937 (1984).
[84] C. J. Fu,Z. M. Zhang,D. B. Tanner, Energy transmission by photon tunneling in multilayer structures, J. Heat. Transfer 1046 / Vol.127 (2005).
[85] L. Gao, C. J. Tang, Near-eld imaging by amulti-layer structure consisting of alternate right-handed and leeft-handed materials, Phys. Lett. A 322, 390 (2004).
[86] J. Gerardin, A. Lakhtakia, Negative index of refraction and distributed bragg reections, Mic. Opt. Technol. Lett. 34, 409, (2002).
[87] L. G. Wang, S. Y. Zhu, Large positive and negative Goos-Hanchen shifts from a weakly absorbing Left-handed slab, J. Appl. Phys. 98, 043522 (2005).
[88] X. Chen, C. F. Li, Lateral shift of the transmitted light beam through a lefthanded slab, Phys. Rev. E 69, 066617 (2004).
[89] K. C. Huang, E. Lidorikis, X. Jiang, J. D. Joannopoulos, K. A. Nelson, Nature of lossy Bloch states in polaritonic photonic crystals, Phys. Rev. B 69, 195111 (2004).
[90] N. F. Declercq, J. Degrieck, Diraction of complex harmonic plane waves and the stimulation of transient leaky Rayleigh waves, J. Appl. Phys. 98, 113521 (2005).
BIBLIOGRAPHY
129
[91] A. Bernard, M. Deschamps, Comparison between the dispersion curves calculated in complex frequency and the minima of the reection coecients for an embedded layer, J. Acoust. Soc. Am. 107, 793 (2000).
[92] P. A. Sturrock, Kinematics of growing waves, Phys. Rev. 112, 1488 (1958).
[94] L. C. Andreani, M. Agio, Intrinsic diraction losses in photonic crystal waveguides with line defects, Appl. Phys. Lett. 82, 2011 (2003).
[95] C. Kottke, A. Farjadpour, S. G. Johnson, Perturbation theory for anisotropic dielectric interfaces, and application to subpixel smoothing of discretized numerical methods, Phys. Rev. E 77, 036611 (2008).
http://www.nanohub.org/courses/metamaterials (2008).
Appendix A
Orthogonality relations
In this appendix, we derive the orthogonality and normalization relations for a lossless metamaterial waveguide structure by using generalized Hermitian Hamiltonian formulation [58, 95]. The time-harmonic electromagnetic wave equations are
1 (r) 1 (r)
where
1 (r) 1 (r)
H(r) = E(r) =
2 H(r) c2 2 E(r), c2
We assume they are real.
(A.1)
and
The
LH H(r) LE E(r)
where
1 (r) 1 (r)
1 (r) 1 (r)
H(r) = E(r) =
2 H(r) c2 2 E(r), c2
(A.2)
LH
and
LE
130
APPENDIX A.
ORTHOGONALITY RELATIONS
131
not orthogonal. We introduce the following set of eigenfunctions and Hermitian operators which satisfy Eq. (A.1) [60]
(r)H(r) (r)E(r)
(A.3)
(r)
1 (r) 1 (r)
1 QH (r)
(r)
(A.4)
1 (r)
1 QE (r) (r)
QE (r)
are orthogonal:
(A.5)
HH
and
HE
and
and
, the above operators are not Hermitian and Eq. (A.5) is no longer
valid. Therefore, we need to derive generalized Hermitian Hamiltonian formulation to nd the eigenmodes of the waveguide [58, 59]. We start with Maxwell's equations for the time-harmonic electromagnetic elds
E = 0, H = 0,
E = i H, c H = i E. c
(A.6)
APPENDIX A.
ORTHOGONALITY RELATIONS
132
By introducing transverse and longitudinal components of the elds, Maxwell's equations take the form
t t
E t = E z , z
tE z tH z
E t z H t z
+ z( + z(
t t
E t ) = i ( H t + H z ), z c H t ) = i ( E t + E z ). z c
(A.7)
H t = H z , z
After eliminating the longitudinal components of the elds and moving all of the z derivatives to one side in Eq. (A.7), we arrive at
z 0
1 z
(
t ) c
0
c t
1 z
t )
Et = Ht
0 E t z i z . z 0 Ht
Next, we dene operators notation,
(A.8)
and
B,
| ,as
z 0
1 z
(
t ) c
0
c
t z
1 z
t )
(A.9)
B 0
z , z 0
(A.10)
APPENDIX A.
ORTHOGONALITY RELATIONS
133
(A.11)
Et | . Ht
The following eigenproblem can be obtained
A | = i B | . z
(A.12)
We assume the waveguide is uniform along z; therefore, the electromagnetic elds has the form
eiz
Finally, by inserting
exp(iz) |
A = B A B
(A.13)
In this form,
and
and
values. They satisfy the properties of the eigenproblems including the eigenvalues , , are real and satisfy the orthogonality relationship
= ,
[58].
Now, we can obtain the orthogonality and normalization relationship for the TE polarization elds as follows
APPENDIX A.
ORTHOGONALITY RELATIONS
134
dx
Et Ht
0 z
z 0
Et Ht
+ 1 Re[ 2
+ = 1 Re[ 2
(x)dx]
= P , ,
where
is the power ux. Eq. (A.14) shows that every mode plays its own part in
the power ux independent of the other modes. It is clear that in the lossy waveguide structure, since the energy is absorbed in the media, the power carried in each mode is not independent from the powers in the other modes anymore. We can also derive the generalized Hermitian Hamiltonian eigenproblem in terms of the frequency,
relations. By introducing as
D t = E t ,
and
B t = H t ,the
APPENDIX A.
ORTHOGONALITY RELATIONS
135
Light line
5
Light line
Frequency (GHz)
Frequency (GHz)
2
TE2
1 0
TE2
TE1 TE0
TE1 TE0
0 0
0.5
1.5
(1/cm)
2.5
3.5
0 0
0.5
1.5
(1/cm)
2.5
3.5
Fig. A.1 The orthogonality relations (a) in terms of frequency and (b) in terms of wave number. with The blue curves are the dispersion curves of a dielectric waveguide
=4
and
=4
D t = Dz , z B t = B z , z
1 t Dz
1 D t z
+ z(
1 D t ) = i ( B t + B z ), z c
1 t Bz
1 B t z
+ z(
1 B t ) = i ( D t + D z ). z c
(A.15)
| (x, z, t) =
D t (x) B t (x)
exp(iz it),
(A.16)
1 t
0
1 1 t t
Dt Bt
0 z z 0
Dt Bt
(A.17)
In the Eq. (A.17), the operators are Hermitian and thus have real eigenvalues,
, and
APPENDIX A.
ORTHOGONALITY RELATIONS
136
(A.18)
is the energy density. One should notice that the above equation only works
for non-dispersive media. Next, we normalize the amplitude of the electric eld,
E0 ,
for an asymmetric
metamaterial waveguide structure. By using the orthogonality relationship in terms of wave number, Eq. (A.14), we have
(A.19)
P = 1.
(A.19), we
APPENDIX A.
ORTHOGONALITY RELATIONS
137
1 2
=1 2d
0
2 cE0 2
dx dx
1 3
sin2 () exp[2k3 x]
dx
1 2
sin2 ( + k2 x)
(A.20)
2d
1 1
asy E0 =
2 c d +
2 3 k3
2 2 k3 +k2 2 2 2 k3 +2 k2 2 3
E0
in terms of
frequency. By substituting Eq. (2.12) into Eq. (A.18), the normalized electric eld has the form
asy E0 =
2 d2 +
1 k1
2 2 2 2 k1 + 1 1 k2 2 2 2 k1 +2 k2 2 1
3 k3
2 2 2 2 k3 + 3 3 k2 2 2 2 k3 +2 k2 2 3
(A.21)
sym E0 =
2 c d +
2
2 1 k1
2 2 k1 +k2 2 2 2 k1 +2 k2 2 1
,
(A.22)
sym E0 =
1 d2 +
1 k1
2 2 2 2 k1 + 1 1 k2 2 2 2 k1 +2 k2 2 1
Finally, the results for the transverse magnetic modes, TM, can be found from duality
APPENDIX A.
ORTHOGONALITY RELATIONS
138
by changing
and
E H .