cond-mat0504152/mw.tex
1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,groupedaddress,pre]{revtex4}
2: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3: 
4: % Some other (several out of many) possibilities
5: %\documentclass[preprint,aps]{revtex4}
6: %\documentclass[preprint,aps,draft]{revtex4}
7: %\documentclass[prb]{revtex4}% Physical Review B
8: 
9: \usepackage{graphicx}% Include figure files
10: \usepackage{dcolumn}% Align table columns on decimal point
11: \usepackage{bm}% bold math
12: \usepackage{epsfig}
13: 
14: \def\e{\begin{equation}}
15: \def\f{\end{equation}}
16: \def\=#1{\overline{\overline #1}}
17: \def\s{\strut\displaystyle}
18: \def\-#1{{\bf #1}}
19: \def\.{\cdot}
20: \def\l#1{\label{eq:#1}}
21: \def\r#1{(\ref{eq:#1})}
22: \def\vec#1{{\bf #1}}
23: 
24: \begin{document}
25: 
26: \title{Subwavelength metallic waveguides loaded by uniaxial resonant scatterers}
27: 
28: \author{Pavel A. Belov}
29: \affiliation{Mobile Communication Division, Telecommunication Network Business, Samsung Electronics Co., Ltd., \\
30: 94-1, Imsoo-Dong, Gumi-City, Gyeong-Buk, 730-350, Korea}
31: 
32: \email{belov@rain.ifmo.ru}
33: 
34: \author{Constantin R. Simovski}
35: \affiliation{Photonics and Optoinformatics Department, St.
36: Petersburg State University of Information Technologies, Mechanics
37: and Optics, Sablinskaya 14, 197101, St. Petersburg, Russia}
38: 
39: \date{\today}
40: 
41: 
42: \begin{abstract}
43: The dispersion properties of rectangular metallic waveguides
44: periodically loaded by uniaxial resonant scatterers are studied
45: with help of an analytical theory based on the local field
46: approach, the dipole approximation and the method of images. 
47: The cases of both magnetic and electric uniaxial scatterers with both longitudinal and
48: transverse orientations with respect to the waveguide axis are considered.
49: It is shown that in all considered cases waveguides support propagating modes 
50: below cutoff of the hollow waveguide within some frequency bands 
51: near the resonant frequency of the individual scatterers.
52: The modes are forward ones except the case of transversely oriented magnetic scatterers
53: when the mode turns out to be backward. The described effects can be applied for 
54: the miniaturization of the guiding structures.
55: \end{abstract}
56: 
57: \pacs{41.20.Jb, 42.70.Qs, 42.25.Fx
58: % 41.20.Jb Electromagnetic wave propagation; radiowave propagation
59: % 42.70.Qs,   Photonic bandgap materials
60: % 42.25.Fx  Diffraction and scattering
61: }
62: \maketitle
63: 
64: \section{Introduction}
65: 
66: Recently, a very unusual waveguide was proposed by R. Marques in \cite{Marqueswaveguide} and 
67: then extensively studied by S. Hrabar in \cite{Hrabarwaveguide}. It is a rectangular metallic waveguide
68: periodically loaded by resonant magnetic scatterers, so-called split-ring-resonators (SRR:s) \cite{PendrySRR,MarquesSRR},
69: which are also used as components of a realization of the left-handed medium (LHM) \cite{SmithWSRR}, composite
70: with negative permittivity and permeability \cite{Veselago,Metaspecial}. 
71: The geometry of the Marques waveguide (MW) is presented in Fig.\ref{geometry}.
72: The SRR:s in MW are oriented so that their magnetic moments are orthogonal to the waveguide axis and to one of
73: the walls. 
74: \begin{figure}[h]
75: \centering \epsfig{file=geometry.eps, width=8.5cm}
76: \caption{Geometry of a subwavelength Split-Ring-Resonator-loaded
77: metallic waveguide} \label{geometry}
78: \end{figure}
79: 
80: The MW support a propagating mode within a frequency band near the resonance of SRR:s even if it is located 
81: below cutoff frequency of the hollow waveguide \cite{Marqueswaveguide,Hrabarwaveguide}.
82: The transversal dimensions of the waveguide happen to be much smaller than the wavelength in free space. 
83: Thus, loading by SRR:s makes waveguide subwavelength and provide unique method for miniaturization of guiding structures.
84: The mode of MW is backward wave (the group velocity is negative). 
85: This effect was interpreted in \cite{Marqueswaveguide} in terms of the
86: effective LHM to which such the loaded waveguide is apparently equivalent.
87: The empty waveguide was considered as an artificial electric plasma with negative permittivity and the array of
88: magnetic scatterers as a magnetic material with negative permeability. 
89: This interpretation is not completely adequate because it cannot explain why the effect disappears in
90: the case of loading by isotropic magnetic with negative permeability. 
91: Really, it is clear that the hollow waveguide filled by 
92: isotropic magnetic with negative permeability does not support guiding modes. Meanwhile,
93: the doubly-negative medium having the isotropic negative permittivity and permeability 
94: would support propagating backward waves \cite{Lind}.
95: Also, the LHM interpretation is not instructive in our opinion since it does not allow to notice possibilities 
96: to obtain propagation below the cutoff frequency of hollow waveguide with help of the other loadings that SRR:s.
97: 
98: The goal of the present study is to give an adequate explanation for the extraordinary propagation effect in MW 
99: and to suggest another loadings which would lead to the similar effects.
100: In this paper it is shown that the propagation below the cutoff frequency of hollow waveguide can be achieved 
101: with magnetic scatterers oriented longitudinally with respect to waveguide axis, as well as with electric scatterers
102: oriented either longitudinally or transversally. This demonstrates that the miniaturization of the
103: rectangular waveguide at a fixed frequency using loading by the resonant scatterers is not restricted by the case 
104: when the scatterers are magnetic and transversally oriented.
105: The miniaturization is possible with help of either magnetic or electric resonant scatterers
106: with either transversal or longitudinal orientation with respect to the waveguide axis. 
107: Of course, the miniaturization can be also reached using periodically located capacitive posts, however the
108: loading by resonant scatterers is a qualitatively different effect.
109: In \cite{Hrabarwaveguide} it was pointed out that the miniaturization obtained in this way refers also to the
110: longitudinal size of the waveguide since the period of the loads
111: in incomparably smaller than the wavelength in free space, unlike
112: the propagation in a capacitively loaded waveguide, where the period of the posts is of the order of $\lambda/2$. 
113: 
114: The mini pass band below the cutoff frequency of a rectangular waveguide loaded by resonant scatterers is caused
115: by the properties of the periodical one-dimensional array (chain) of resonant dipoles and has nothing to do with doubly-negative
116: media. The backward wave appears in a special case of transverse orientation of magnetic scatterers and is not a necessary attribute of such mini-band. It is known that a chain of the resonant scatterers in a homogeneous
117: matrix supports guided modes. In the optical frequency range it refers to chains of metallic nanoparticles
118: \cite{Brongersma,Maier1,Maier2,Weber}. At microwaves it refers to so-called magneto-inductive waveguides (chains of SRR:s) \cite{Shamonina1,Shamonina2,Shamonina3,Shamonina4} or to chains of inductively loaded electric dipoles \cite{Tretlines}. 
119: The metallic walls of loaded waveguide perturb the dispersion of the
120: guided mode in a chain but do not cancel the propagation. This is because the
121: wavelength of the guided mode in the chain of resonant scatterers
122: is dramatically shortened compared to that in the matrix. As a
123: result, the energy of a guided mode is concentrated in a narrow
124: domain around the chain, and the interaction between the chain and
125: the waveguide walls turns out to be not critical for the existence
126: of the guided mode.
127: 
128: The paper is organized as follows. In the Section II the dispersion
129: properties of the chains of resonant scatterers located in free space are considered.
130: The known results, obtained in the work \cite{Weber} by
131: numerical simulations, are reproduced with help of analytical theory based on local field method. 
132: This is a necessary part of the work in the view of comparison with the case of the loaded waveguide.
133: The coincidence with known results can be considered as a validation of our approach. 
134: In the Section III the dispersion properties 
135: of the chains located inside of the rectangular waveguide are considered using two approaches:
136: an accurate method of local field (as in Section II) and an effective medium filling approximation. 
137: The Section IV is devoted to comparison between properties of the chains and loaded waveguides.
138: The Section V contains concluding remarks.The details of the local field theory are given in Appendix.
139: 
140: In the present paper we consider both magnetic and electric
141: resonant uniaxial scatterers. As an example of a magnetic
142: scatterer we have chosen the SRR:s
143: \cite{PendrySRR,SmithWSRR,Shelbyscience} (see Fig. \ref{scat}.a). The
144: electric dipoles are represented in our work by the short
145: inductively loaded wires (ILW) \cite{LWD} (see Fig. \ref{scat}.b). Any
146: individual scatterer can be characterized by polarizability
147: relating the dipole moment (magnetic or electric) with
148: the local field (magnetic or electric external field acting to the
149: scatterer). This polarizability is scalar since the only possible
150: direction of the induced dipole moment is possible for a scatterer
151: with given orientation. The details concerning calculation of the polarizabilities
152: for SRR:s and ILW:s are presented in Appendix A.
153: 
154: \begin{figure}[h]
155: \centering \epsfig{file=scat.eps, width=7cm} \caption{Geometries of
156: resonant scatterers: a) Split-Ring-Resonator, b) inductively loaded wire dipole.} \label{scat}
157: \end{figure}
158: 
159: The inverse values of the polarizabilities $\alpha(\omega)$ and $\alpha_e(\omega)$
160: (see Appendix A, formulae \r{alpha} and \r{alpe}) of SRR:s and ILW:s 
161: have the same dependencies on frequency within the resonant band: 
162: \e
163: {\rm Re} \{\alpha^{-1}(\omega)\}=A^{-1}\left(\frac{\omega_0^2}{\omega^2}-1\right).
164: \l{inva} 
165: \f 
166: Here $A$ is amplitude and $\omega_0$ is resonant frequency, the parameters determined
167: by the geometry of the scatterer. Notice, that the result \r{inva} is
168: also valid for a silver nanosphere in the vicinity of its plasmon
169: resonance \cite{Weber}. Thus, it is clear that there is no
170: principal difference between the dispersion properties of the
171: chain of SRR:s and ILW:s (at microwaves) or silver nanospheres (in the optical range).
172: 
173: 
174: \section{Chains of uniaxial resonant scatterers}
175: 
176: Let us study dispersion properties of the linear chains with period
177: $a$ formed by resonant scatterers. We will consider only two
178: typical orientations of scatterers: longitudinal and transverse.
179: The geometries of the structures are presented in Fig. \ref{chains}. 
180: The case of longitudinal orientation was analyzed in \cite{Tretlines}, and the both 
181: longitudinal and transverse orientations were studied in \cite{Weber}. 
182: In the present section we reproduce the main results of these works with help of the local field approach.
183: \begin{figure}[h]
184: \centering \epsfig{file=chains.eps, width=8cm}
185: \caption{Chains of resonant scatterers. a) Longitudinal orientation. b) Transverse orientation.}
186: \label{chains}
187: \end{figure}
188: 
189: \subsection{Basic theory}
190: 
191: Without loss of generality we can restrict consideration by the case of the chain of magnetic scatterers (SRR:s).
192: The chains of electric scatterers are dual structures to the considered ones and 
193: have completely the same dispersion properties.
194: 
195: The spatial distribution of dipole moments of SRR:s corresponding to an eigenmode of a chain 
196: is determined by a propagation constant $q$: $M_n=M e^{-jqan}$.
197: Following to the local field approach the dipole moment $M$ of a
198: reference (zeroth) scatterer can be expressed in terms of the magnetic field
199: $\-H_{\rm loc.}$ acting to it: $M=\alpha H^d_{\rm loc.}$, where
200: $H^d_{\rm loc.}=(\-H_{\rm loc.}\.\-d)$ is the projection of the
201: field on the direction of the scatterer ($\vec d=\vec x_0$ for longitudinal
202: orientation of scatterers and $\vec d=\vec y_0$ for transverse one). 
203: This local field is a sum of partial magnetic fields
204: $\-H_{m}$ produced at the coordinate origin by all other
205: scatterers with indexes $m\ne 0$: $ \-H_{\rm loc.}=\sum\limits_{m
206: \ne 0} \-H_{m}$.
207: 
208: The magnetic field produced by a single scatterer with dipole moment $\-M_m$
209: at a point with radius vector $\vec R$ is given by dyadic Green's function $\=G(\-R)$:
210: \e
211: \-H_m(\vec R)=\mu_0^{-1}\=G(\-R)\-M_m,
212: \l{Ggen}
213: \f
214: where
215: \e
216: \=G(\-R)=\left(k^2\=I+\nabla\nabla\right)\frac{e^{-jkR}}{4\pi R}.
217: \l{G}
218: \f
219: 
220: Since all dipole moments of the chain are oriented along $\-d$ it
221: is enough to use only the $\-d\-d$ component of dyadic Green's
222: function. So, we replace \r{Ggen} by the scalar expression: \e
223: H_m^d(\vec R)=\mu_0^{-1}G_{dd}(\-R)M_m, \l{Green} \f where \e
224: G_{dd}(\-R)=\left(k^2+\frac{\partial^2}{\partial
225: d^2}\right)\frac{e^{-jkR}}{4\pi R}, \l{gre}\f and $d$ means $x$
226: for the longitudinal case and $y$ for the transverse case.
227: 
228: Finally we obtain the expression for the field acting to the
229: reference scatterer in the form: \e H^d_{\rm
230: loc.}=\sum\limits_{m\ne 0} G_{dd}(am\vec x_0) e^{-jqam} M. \f
231: 
232: It allows to get dispersion equation for the chains under
233: consideration: 
234: \e 
235: \mu_0\alpha(\omega)^{-1}=C_{d}(\omega,q,a),
236: \l{displine} \f where
237: $$
238: C_{d}(\omega,q,a)=\sum\limits_{m\ne 0} G_{dd}(am\vec x_0) e^{-jqam}.
239: $$
240: In the Appendix B we provide expressions \r{Cx} and \r{Cy} 
241: which we use for effective numerical calculations of interaction constants $C_x$ and $C_y$,
242: corresponding to transverse and longitudinal orientations of scatterers, respectively.
243: 
244: \subsection{Analysis of dispersion properties}
245: 
246: The dispersion diagram for guided modes can be obtained solving
247: transcendental dispersion equation \r{displine} with interaction
248: constants given by expressions \r{Cx} and \r{Cy}. Geometrically,
249: dispersion curves correspond to the lines of level where the
250: surface plot of function ${\rm Re} \{C_{x,y}(\omega,q)\}$ is
251: crossed by $\mu_0 {\rm Re}\{\alpha^{-1}(\omega)\}$. Note, that
252: only the solutions with $k<q<2\pi/a-k$ for $k<\pi/a$ correspond to
253: guided modes: For $|q|<k$, ${\rm Im} \{ C_{x,y}(\omega,q)-
254: \mu_0\alpha^{-1}\} \ne 0$ and dispersion equation \r{displine} has
255: complex solutions corresponding to leaky modes (see details in Appendix B).
256: 
257: \begin{figure}[h]
258: \centering \epsfig{file=cxbw.eps, width=8.5cm}
259: \caption{Dependence of ${\rm Re}\{C_x a^3\}$ on normalized frequency $ka/\pi$ and propagation factor $qa/\pi$}
260: \label{cx}
261: \end{figure}
262: \begin{figure}[h]
263: \centering \epsfig{file=cybw.eps, width=8.5cm}
264: \caption{Dependence of ${\rm Re}\{C_y a^3\}$ on normalized frequency $ka/\pi$ and propagation factor $qa/\pi$}
265: \label{cy}
266: \end{figure}
267: 
268: The dependencies of $C_x$ and $C_y$ on on normalized frequency
269: $ka/\pi$ and propagation factor $qa/\pi$ are shown in Fig.
270: \ref{cx} and \ref{cy}, respectively. The interaction constants 
271: vary in $[-1,1]$ range except the case of $C_y$ with $q$ close
272: to $k$, which has logarithmic singularity at the light line $q=k$.
273: The function $\mu_0 {\rm Re}\{\alpha^{-1}(\omega)\}$ decreases
274: very rapidly within $[-1,1]$ range of values near the resonant
275: frequency $\omega_0$. It means, that guided modes exist only
276: within a narrow bands near the resonant frequency of scatterers.
277: The behavior of dispersion curves can be easily predicted from the
278: plots in Fig. \ref{cx} and \ref{cy}. If at a fixed frequency the
279: interaction constant decays when propagation factor increases then
280: the dispersion curve grows and the eigenmode is forward wave (the group velocity $v_g=\frac{d\omega}{dq}$ is positive),
281: but if the interaction constant grows then the dispersion curve decays
282: and the eigenmode is backward wave (the group velocity $v_g=\frac{d\omega}{dq}$ is negative).
283: From Fig. \ref{cx} it is clear
284: that for any resonant frequencies satisfying to the evident
285: condition $\omega_0<\pi/(a\sqrt{\varepsilon_0\mu_0})$ 
286: (corresponding to the propagation below cutoff of the hollow waveguide )
287: the longitudinal mode is forward wave because $C_x$ decays versus $q$.
288: In the case of transverse modes the situation is different. 
289: While $k_0a<0.5\pi$ ($k_0=\omega_0\sqrt{\varepsilon_0\mu_0}$) a two mode regime holds.
290: The interaction constant $C_y$ decays while $q$ is close to $k$,
291: but from a certain $q$ it starts to grow. It means that one of the
292: transverse eigenmodes is forward (with $q\approx k $) and the other one is
293: backward. If the resonant frequency is high enough
294: ($0.5\pi<k_0a<\pi$) the two mode regime disappears and only the
295: forward wave remains.
296: 
297: \begin{figure}[h]
298: \centering \epsfig{file=displine2.eps, width=8.5cm}
299: \caption{Dispersion diagram for chains of resonant scatterers: transverse (thick line) and longitudinal (thin line)
300: orientations}
301: \label{displine}
302: \end{figure}
303: 
304: The typical dispersion diagram for both longitudinal and
305: transverse modes is presented in Fig. \ref{displine} for the case
306: of scatterers with $A=0.1\mu_0a^3$ and $\omega_0
307: a=1/\sqrt{\varepsilon_0\mu_0}$. The similar dispersion diagram was
308: obtained in \cite{Weber} (see Fig. 3 of this work) by a numerical
309: simulation. Though in \cite{Weber} the electric scatterers in the optical range were considered, 
310: but we consider the magnetic scatterers in the microwave range, the use of duality principle, and normalized frequency
311: $ka/\pi$ and wave vector $qa/\pi$ eliminates this difference. 
312: The polarizability of silver nanospheres, for which Fig. 3 from \cite{Weber} was obtained, obeys to
313: expression (8) of \cite{Weber} which is identical to our formula \r{inva}).
314: 
315: As it was predicted, the longitudinal mode is forward wave and there is
316: a two mode regime for transverse modes. The dispersion curve for
317: transverse waves has the asymptote $q=k$ and both leaky ($q<k$) and guided ($q>k$)
318: modes exist at very low frequencies where they have almost equal
319: wave vectors. This fact indicates the good matching between the
320: radiated wave and the guided mode. Within the band $0.995<ka<1$
321: there are two guiding modes at every frequency. The solution corresponding
322: to the backward wave is close to the Bragg's mode ($qa\approx
323: \pi$) whose group velocity is close to zero. The field of this
324: mode is concentrated near the chain within the spatial region
325: $r=\sqrt{z^2+y^2}<1/\sqrt{q^2-k^2}\sim a$. The same concerns the
326: longitudinal mode within the band $1.015<ka<1.020$. If the period
327: of the chain is much smaller than wavelength in free space the
328: waveguide is sub-wavelength (the field of the guided mode is
329: concentrated within a cylindrical domain whose diameter is much
330: smaller than $\lambda$). Thus, the chains of resonant scatterers
331: (electric or magnetic, it does not matter) form sub-wavelength
332: waveguides which can support either forward or backward waves
333: \cite{Brongersma,Maier1,Maier2,Weber,Tretlines,Shamonina1,Shamonina2,Shamonina3,Shamonina4}.
334: 
335: \section{Loaded waveguides}
336: 
337: \subsection{Basic theory}
338: 
339: Let us study the eigenmodes of the rectangular metallic waveguides
340: periodically loaded by resonant uniaxial scatterers. Such
341: structures can be effectively considered as linear chains located
342: inside of the metallic waveguides. The geometries of the four
343: waveguides considered in the present paper are shown in Fig.
344: \ref{geometry} (left sides of subplots). The chains with period
345: $c$ along the waveguide axis are located at the center of rectangular metallic waveguides with
346: dimensions $a\times b$. The structures in Figs. \ref{geometry}.a-d differ by orientation of
347: scatterers (longitudinal or transverse) and their type (electric or magnetic). 
348: The first structure (with transversely oriented magnetic scatterers) is the sub-wavelength waveguide (see Fig.
349: \ref{geometry}) suggested by R. Marques \cite{Marqueswaveguide,Hrabarwaveguide}. The other ones are
350: considered in order to show three other possible ways to obtain miniaturization of rectangular waveguides.
351: 
352: Note, that the chains of electric scatterers are no more dual to the chains of magnetic scatterers 
353: (in contrast to the chains in free space)
354: due to the different interaction of electric and magnetic dipoles with metallic walls.
355: 
356: \begin{figure}[h]
357: \centering \epsfig{file=image.eps, width=8.5cm}
358: \caption{Transformation of the waveguide problem to the lattice
359: one with the use of the image principle} \label{image}
360: \end{figure}
361: 
362: The dispersion equation for the chains keeps the same form as \r{displine}, but
363: free-space dyadic Green's function $\=G(\vec R)$ \r{G} should be replaced by Green's
364: function of the waveguide which takes into account the metallic walls.
365: This Green's function can be determined with help of the image principle. 
366: This approach transforms the eigenmode problem for the loaded waveguide 
367: to the problem of the eigenwave propagation in a three-dimensional
368: electromagnetic lattice formed by same scatterers. The details of
369: the transformation are illustrated by Fig. \ref{image} (right
370: parts of all subplots). The electromagnetic crystals obtained in
371: such a way have orthorhombic elementary cell $a\times b\times c$
372: and their dispersion properties was studied in \cite{Belovhomo}
373: using local field approach. Thus, we can apply the theory of the
374: electromagnetic interaction in dipole crystals presented in
375: \cite{Belovhomo} in order to study dispersion properties of 
376: waveguides under consideration.
377: 
378: In the coordinate system associated to the axes of the crystal
379: the center of a scatterer with indexes $(m,n,l)$ has coordinates $\vec R_{m,n,l}=(am,bn,cl)^T$. 
380: From Fig. \ref{image} it is clear, that distribution of dipole moments 
381: in the lattice has the following form:
382: $$
383: \vec M_{m,n,l}=(-1)^mMe^{-jqcl}\vec x_0,
384: $$
385: for the case of transverse orientation of magnetic scatterers (Fig. \ref{image}.a);
386: $$
387: \vec M_{m,n,l}=Me^{-jqcl}\vec z_0,
388: $$
389: for the case of longitudinal orientation of magnetic scatterers (Fig. \ref{image}.b);
390: $$
391: \vec P_{m,n,l}=(-1)^nPe^{-jqcl}\vec x_0,
392: $$
393: for the case of transverse orientation of electric scatterers (Fig. \ref{image}.c); and
394: $$
395: \vec P_{m,n,l}=(-1)^{m+n}Pe^{-jqcl}\vec z_0,
396: $$
397: for the case of longitudinal orientation of electric scatterers (Fig. \ref{image}.d).
398: 
399: Any of these distributions can be rewritten in terms of a
400: wavevector $\vec q$ as $e^{-j(\vec q\. \vec R_{m,n,l})}$, where
401: the wavevector $\vec q$ for four cases considered above has the
402: form $(\pi/a,0,q)^T$, $(0,0,q)^T$, $(0,\pi/b,q)^T$ and
403: $(\pi/a,\pi/b,q)^T$, respectively. This notation finally makes
404: clear that the waveguide dispersion problems reduce to those of the
405: three-dimensional lattices in the special cases of certain propagation directions.
406: 
407: The dispersion equation for three-dimensional electromagnetic
408: crystal formed by magnetic scatterers oriented along $x$-axis has the form \cite{Belovhomo}:
409: \e 
410: \mu_0\alpha^{-1}(\omega)-C(k,\-q)=0, \l{disper} 
411: \f
412: where \e C(k,\-q,a,b,c)=\sum\limits_{(m,n,l)\ne(0,0,0)}
413: G(\-R_{m,n,l}) e^{-j(q_xam+q_ybn+q_zcl)}. \l{C} \f
414: We call $C(k,\-q,a,b,c)$ as the dynamic interaction constant of
415: the lattice using the analogy with the classical interaction
416: constant from the theory of artificial dielectrics and magnetics
417: \cite{Collin}. The explicit expression for $C$ for the general
418: case was derived in \cite{Belovhomo} and it is given in Appendix C by formula \r{cfinal}.
419: 
420: The dispersion equation for waveguides with transverse orientation of scatterers
421: can be directly obtained from equation \r{disper} 
422: by substitution $\vec q=(\pi/a,0,q)^T$ and $\vec q=(0,\pi/b,q)^T$ for magnetic and electric scatterers, respectively.
423: Also, in the case of electric scatterers following duality principle $\mu_0\alpha^{-1}(\omega)$
424: should be replaced by $\varepsilon_0\alpha_e^{-1}(\omega)$.
425: The similar operation for case of longitudinal orientation happens to be possible only after
426: rotation of coordinate axes: $z \rightarrow x'$, $x \rightarrow y'$, $y \rightarrow z'$, since equation \r{disper} requires scatterers to be directed along $x$-axis, but for longitudinal orientation they are directed along $z$-axis.
427: After such manipulation substitution of $\vec q=(q,0,0)^T$ and $\vec q=(q,\pi/a,\pi/b)^T$ 
428: (in the new coordinate axes $(x',y',z')$) into equation \r{disper} provide desired dispersion equations
429: for waveguides with transverse orientation of magnetic and electric scatterers, respectively.
430: 
431: This way we obtain the following dispersion equations for all loaded waveguides under consideration: \e
432: \mu_0\alpha^{-1}(\omega)-C(k,(\pi/a,0,q)^T,a,b,c)=0, \l{dispmt} \f
433: for transverse orientation of magnetic scatterers; \e
434: \mu_0\alpha^{-1}(\omega)-C(k,(q,0,0)^T,c,a,b)=0, \l{dispml} \f for
435: longitudinal orientation of magnetic scatterers; \e
436: \varepsilon_0\alpha_e^{-1}(\omega)-C(k,(0,\pi/b,q)^T,a,b,c)=0,
437: \l{disppt} \f for transverse orientation of electric scatterers;
438: and \e
439: \varepsilon_0\alpha_e^{-1}(\omega)-C(k,(q,\pi/a,\pi/b)^T,c,a,b)=0,
440: \l{disppl} \f for longitudinal orientation of electric scatterers.
441: 
442: The obtained dispersion equations are real valued ones, because the imaginary parts of
443: its components cancels out. It can be clearly seen from Sipe-Kronendonk condition \r{sipe}
444: and following expression proved in \cite{Belovhomo}:
445: \e
446: {\rm Im}(C)=\frac{k^3}{6\pi}.
447: \l{imc} \f
448: 
449: \subsection{Effective medium filling approximation}
450: 
451: The chain of SRR:s with transverse orientation located in the waveguide has
452: been interpreted in the literature as a piece of an uniaxial magnetic medium \cite{Marqueswaveguide,Hrabarwaveguide}. 
453: We call this approach as effective medium filling approximation.
454: It can be applied practically to every waveguide considered in this paper except the case 
455: of longitudinal orientation of magnetic scatterers 
456: since the uniaxial magnetic model does not describe longitudinal modes.
457: This approach provide qualitatively acceptable results which are compared with exact ones in the next subsection.
458: 
459: Let us start from the case of transversely oriented magnetic
460: scatterers (Fig. \ref{image}.a) and consider a chain of parallel uniaxial magnetic
461: scatterers as a piece of infinite resonant uniaxial magnetic. The
462: permeability of such a magnetic is a tensor (dyadic) of the form:
463: $$
464: \=\mu=\mu \-x_0\-x_0+\mu_0(\-y_0\-y_0+\-z_0\-z_0).
465: $$
466: 
467: The permeability $\mu$ along the anisotropy axis $x$, can be
468: calculated though the individual polarizability of a single
469: scatterer using the Clausius-Mossotti formula: 
470: \e
471: \mu=\mu_0\left(1+\frac{\alpha(\omega)/(\mu_0V)}{1-C_s(a,b,c)
472: \alpha (\omega)/\mu_0}\right), \l{CM} 
473: \f where $V=abc$ is a volume of the elementary cell of an infinite three-dimensional lattice 
474: and $C_s(a,b,c)$ is the known static interaction constant of the lattice
475: \cite{Collin,Belovhomo}. In the case of a simple cubic lattice
476: $a=b=c$ the interaction constant is equal to the classical value
477: $C_s=1/(3V)$.
478: 
479: Notice, that we should skip the radiation losses contribution in
480: expression \r{invalph} while substituting into formula \r{CM}.
481: This makes permeability purely real number as it should be for
482: lossless material. This manipulation is based on the fact that the
483: far-field radiation of the single scatterer is compensated by the
484: electromagnetic interaction in a regular three-dimensional array,
485: so that there were no radiation losses for the wave propagating in
486: the lattice \cite{Sipe,Belovhomo}.
487: 
488: The dispersion equation for the uniaxial magnetic medium has the
489: following form (see e.g. \cite{BelovMOTL}): 
490: \e 
491: \mu_0 (q_y^2+q_z^2)=\mu (k^2-q_x^2). \l{dispunis} 
492: \f 
493: To solve the waveguide dispersion problem is to solve the dispersion problem
494: \r{dispunis} for a special case $\vec q=(\pi/a,0,q)^T$. The
495: substitution of $\vec q=(\pi/a,0,q)^T$ into \r{dispunis} gives: 
496: \e
497: q^2=\frac{\mu}{\mu_0}\left[k^2-\left(\frac{\pi}{a}\right)^2\right].
498: \l{qq} \f
499: 
500: For the frequencies below cutoff of the hollow waveguide the expression in the brackets of \r{qq} is
501: negative and for positive $\mu$ there is no propagation. However,
502: if $\mu$ is negative (it happens in accordance with \r{alpha} and
503: \r{CM} within a narrow frequency range near the resonance of the
504: scatterers, just above the resonant frequency of the media) $q$
505: becomes real. This mini pass-band can be located much lower than the
506: cutoff frequency of the empty waveguide with help of reduction of resonant frequency of the scatterers.
507: It is easy to see from \r{qq} that the mode is backward wave, i.e. $\frac{dq}{d\omega}<0$.
508: This follows from the basic inequality $\frac{d\mu}{d\omega}>0$ (Foster's theorem).
509: 
510: So, for transverse magnetic scatterers the effective medium filling model 
511: gives (at least qualitatively) a correct result. Namely, it predicts a mini-band within the resonance
512: band of SRRs and the backward wave propagating within it. 
513: However, this model is not accurate. The reason of this inaccuracy is simple. 
514: In spite of the low frequency of operation 
515: (the waveguide dimensions are small compared to the wavelength {\it in
516: free space}) the effective magnetic medium should operate in the
517: regime when its period is comparable with the wavelength {\it in
518: the effective medium} because $q_x=\pi/a$. Such regime for an electromagnetic
519: crystal can not be described with help of homogenization and requires
520: taking into account spatial resonances of the lattice \cite{Belovhomo}.
521: 
522: In the case of transversely oriented electric scatterers (Fig. \ref{image}.c) the
523: effective medium filling model implies that the wave propagates in a
524: uniaxial dielectric with permittivity tensor:
525: $$
526: \=\varepsilon=\mu \-x_0\-x_0+\varepsilon_0(\-y_0\-y_0+\-z_0\-z_0).
527: $$
528: 
529: The permittivity $\varepsilon$ along the anisotropy axis $x$ is
530: given by the Clausius-Mossotti formula: \e
531: \varepsilon=\varepsilon_0\left(1+\frac{\alpha_e(\omega)/(\varepsilon_0V)}{1-C_s(a,b,c)
532: \alpha_e (\omega)/\varepsilon_0}\right). \l{CMe} \f 
533: The dispersion equation for such uniaxial dielectric reads \cite{BelovMOTL}: 
534: \e
535: \varepsilon_0 (q_y^2+q_z^2)=\varepsilon (k^2-q_x^2). \l{dispunise}
536: \f 
537: Solution of waveguide dispersion problem corresponds to the
538: case when $\vec q=(0,\pi/b,q)^T$: \e
539: q^2=\frac{\varepsilon}{\varepsilon_0}k^2-\left(\frac{\pi}{b}\right)^2.
540: \l{qqe} 
541: \f 
542: This mode propagates only at the frequencies when permittivity takes high positive values
543: $\varepsilon>\varepsilon_0[\pi/(kb)]^2$. In our case of resonant
544: dielectric it happens within a mini-band just {\it below} the resonance of the medium.
545: It is clear from \r{qqe} that the mode is forward wave, i.e.
546: $\frac{dq}{d\omega}>0$. This follows from the Foster's theorem $\frac{d\varepsilon}{d\omega}>0$.
547: 
548: In the case of longitudinally oriented electric scatterers (Fig.
549: \ref{image}.d) the solution of the waveguide dispersion problem
550: can be obtained from the dispersion equation \r{dispunise} with
551: $\vec q=(q,\pi/a,\pi/b)^T$ (the axis were transformed in order to
552: have x-axis along dipoles) in the next form: \e
553: q^2=k^2-\frac{\varepsilon_0}{\varepsilon}
554: \left[\left(\frac{\pi}{a}\right)^2 +
555: \left(\frac{\pi}{b}\right)^2\right]. \l{qqe2} \f 
556: The mode is propagating at the frequencies when permittivity is positive and rather high, or negative.
557: This mode is forward wave in the same manner as \r{qqe} since $\frac{d\varepsilon}{d\omega}>0$.
558: 
559: 
560: \subsection{Analysis of dispersion properties}
561: 
562: For numerical calculation of dispersion curves using \r{dispmt}, \r{dispml},
563: \r{disppt}, \r{disppl} and  \r{cfinal} we have chosen square
564: waveguides ($a=b=c$) loaded by scatterers with the same parameters
565: which were used for studies of chains: 
566: $\omega_0=1/(a\sqrt{\varepsilon_0\mu_0})$, and $A=0.1\mu_0a^3$ for magnetic scatterers,
567: and $A_e=0.1\epsilon_0a^3$ for electric ones.
568: 
569: \begin{figure}[h]
570: \centering \epsfig{file=surfm01bw.eps, width=8.5cm}
571: \caption{Dependence of the real part of normalized interaction
572: constant $C(k,\vec q)a^3$ with $\vec q=(\pi/a,0,q)^T$
573: (corresponding to transverse orientation of magnetic scatterers)
574:  on normalized frequency $ka/\pi$ and propagation constant $qa/\pi$.
575: } \label{surfm01}
576: \end{figure}
577: 
578: The dependence of the real part of normalized interaction constant
579: $C(k,\vec q)$ with $\vec q=(\pi/a,0,q)^T$ on the normalized frequency
580: $ka/\pi$ and on the propagation constant $qa/\pi$ is presented in
581: Fig. \ref{surfm01}. This interaction constant corresponds to dispersion equation
582: \r{dispmt} for transverse orientation of magnetic scatterers. The
583: value of $\mbox{Re} (C)a^3$ varies within $[-2,0.5]$ interval
584: while the normalized frequency $ka/\pi$ is bounded by unity (which
585: corresponds to the cutoff of hollow waveguide). If a value of the normalized frequency
586: is fixed then the real part of the interaction constant is a monotonously growing function of $qa/\pi$.
587: The dependence of the real part of the interaction constant on
588: frequency is quite weak as compared with rapidly decreasing
589: $\alpha^{-1}(\omega)$ as follows from \r{invalph}. The function
590: $\mu_0 \alpha^{-1}(\omega)$ takes values within $[-2,0.5]$
591: interval at the frequencies close to resonant frequency $\omega_0$.
592: Therefore, dispersion equation \r{dispmt} has a real solution for $qa/\pi$
593: within a mini-band of frequencies near the resonant frequency of
594: the scatterers $\omega_0$, and this solution is a decaying function of frequency 
595: which corresponds to the backward wave (the group velocity $v_g=\frac{d\omega}{dq}$ is negative).
596: The obtained result, of course, confirms existence of backward wave below cutoff
597: of the hollow waveguide predicted and experimentally demonstrated in \cite{Marqueswaveguide,Hrabarwaveguide}.
598: 
599: 
600: \begin{figure}[h]
601: \centering \epsfig{file=surfm00bw.eps, width=8.5cm}
602: \caption{Dependence of the real part of normalized interaction
603: constant $C(k,\vec q)a^3$ with $\vec q=(q,0,0)^T$
604: (corresponding to longitudinal orientation of magnetic scatterers)
605:  on normalized frequency $ka/\pi$ and propagation constant $qa/\pi$.
606: } \label{surfm00}
607: \end{figure}
608: 
609: 
610: In contrast to Fig. \ref{surfm01}, the dependencies of normalized
611: interaction constants $C(k,\vec q)$ with $\vec q=(q,0,0)^T$, $\vec
612: q=(0,\pi/a,q)^T$ and $\vec q=(q,\pi/a,\pi/a)^T$ are decaying
613: functions of $q$ for the fixed values of $k<\pi/a$. It means that
614: solutions of dispersion equations \r{dispml}, \r{disppt} and
615: \r{disppl} are forward waves for any resonant frequency of the
616: scatterer below $\pi/(a\sqrt{\varepsilon_0\mu_0})$.
617: The Fig. \ref{surfm00} shows dependence of the real part of normalized interaction constant
618: $C(k,\vec q)$ with $\vec q=(q,0,0)^T$ on the normalized frequency
619: $ka/\pi$ and on the propagation constant $qa/\pi$.
620: The dependencies for the cases $\vec q=(0,\pi/a,q)^T$ and $\vec q=(q,\pi/a,\pi/a)^T$
621: are not shown in order to reduce size of the paper.
622: 
623: \begin{figure}[h]
624: \centering \epsfig{file=dispcompm01.eps, width=8.5cm}
625: \caption{Dispersion curves for metallic waveguides loaded by
626: magnetic scatterers: exact solution (thick line) and
627: effective medium filling approximation (dashed line) for transverse
628: orientation, and exact solution (thin line) for longitudinal
629: orientation. The effective medium filling model for the longitudinal case is
630: not applicable.} \label{dispm01}
631: \end{figure}
632: 
633: The dispersion curves for the case of magnetic scatterers are
634: presented in Fig. \ref{dispm01}. The thick solid line represents
635: the dispersion curve for the transverse mode. It is obtained by
636: numerical solution of transcendental dispersion equation
637: \r{dispmt}. The dashed line shows the result predicted by the
638: model of effective medium filling \r{qq}. The significant frequency shift
639: between the exact and approximate solutions is observed. Also, the
640: effective medium model gives wittingly wrong results with
641: $q>\pi/a$ in the region $ka<1.0055$, and incorrectly describes
642: group velocity for $q>\pi/(2a)$ (for example,
643: it does not describe the Bragg mode with zero group velocity at
644: the point $qa=\pi$). The dispersion curve for the
645: longitudinal mode obtained by numerical solution of equation \r{dispml}
646: is represented by thin line in Fig. \ref{dispm01}. 
647: As it was mentioned above, the effective medium filling model
648: can not be applied for description of this mode.
649: 
650: \begin{figure}[h]
651: \centering \epsfig{file=dispcompp01.eps, width=8.5cm}
652: \caption{Dispersion curves for metallic waveguides loaded by
653: electric scatterers: exact solution (thick line) and
654: effective medium filling approximation (dashed line) for transverse
655: orientation, and exact solution (thin line) and effective medium filling
656: approximation (thin dashed line) for longitudinal orientation. }
657: \label{dispp01}
658: \end{figure}
659: The dispersion curves for the case of electric scatterers are
660: presented in Fig. \ref{dispp01}. The thick and thin solid lines shows dispersion curves
661: for transverse and longitudinal modes, obtained by numerical solution
662: of dispersion equations \r{disppt}  and \r{disppl}, respectively. 
663: The dispersion curves provided by effective medium models for transverse and longitudinal modes 
664: (formulae \r{qqe} and  \r{qqe2})  are plotted by thick and thin dashed lines, respectively. 
665: The comparison of exact and approximate solutions shows that the model of effective medium
666: filling gives qualitatively right prediction of dispersion curves behavior in the case of electric scatterers
667: as well as in the case of transverse magnetic scatterers.
668: The drawbacks are also the same: wrong group velocity $q>\pi/(2a)$ and wittingly wrong results with
669: $q>\pi/a$ at some frequencies.
670: 
671: The Figs. \ref{dispm01} and \ref{dispp01} demonstrate that
672: the waveguides loaded by electric and magnetic resonant scatterers support
673: modes within the mini-bands below the cutoff frequency of the hollow waveguide.
674: The modes are forward waves, except the case of transverse magnetic scatterers
675: when the mode is forward wave. The bandwidth in the case of transverse electric scatterers
676: is of the same order with bandwidth for magnetic scatterers with the same parameters obtained with help of duality principle,
677: but the bandwidths for the longitudinal modes are significantly narrower than for the transverse ones.
678: 
679: 
680: \section{Discussion}
681: 
682: In the papers \cite{Marqueswaveguide,Hrabarwaveguide} the term `subwavelength waveguide' was applied
683: to a rectangular waveguide with small transversal dimensions as
684: compared to the wavelength in free space. However, there is a lot
685: of other works \cite{Brongersma,Maier1,Maier2,Weber,Tretlines,Shamonina1,Shamonina2,Shamonina3,Shamonina4}. 
686: in which the term `subwavelength waveguiding' means the
687: propagation of a wave along the chain of electrically small
688: nearly-resonant particles below the diffraction limit.
689: In this case the transversal size of the spatial domain, in which the field of the guided mode is concentrated, 
690: is much smaller that the wavelength in free space. Therefore, this mechanism of the wave transmission
691: is considered as prospective for subwavelength imaging.
692: 
693: Since the field of the mode guided along the chain of resonant
694: scatterers is concentrated within a subwavelength cross section,
695: the presence of the metal walls even at the rather small distance
696: turns out to be not crucial for the existence of the guided wave.
697: Thus, any waveguide periodically loaded by the scatterers can be considered as
698: a subwavelength waveguide formed by the chain of nearly resonant scatterers, whose
699: dispersive properties are perturbed by the metal walls. These
700: walls can be described in terms of the image chains forming an
701: infinite lattice. However, the wave propagates along the same
702: direction in every image chain, and the transversal wave
703: numbers $q_x=\pi/a$ or $q_y=\pi/b$ describe the transversal phase
704: distribution of the wave propagating along $z$ but not the energy
705: transport across $z$. The main question is how the image chains of
706: scatterers interact with the real chain, and how this interaction
707: influences its dispersive properties.
708: 
709: For that purpose let us compare  Fig. \ref{displine} with Figs. \ref{dispm01}
710: and \ref{dispp01}. We can conclude that the presence of metal
711: walls around the chain of resonant scatterers produces the
712: following effects: 
713: \begin{itemize}
714: \item It decreases the group velocity and the frequency band of the
715: guided mode which corresponds to the longitudinal orientation of
716: dipoles. 
717: \item It cancels the two-mode regime for the transverse
718: orientation of dipoles, so that the dispersion branch becomes
719: backward for magnetic scatterers, and forward for electric ones.
720: \end{itemize}
721: 
722: Finally, we would like to emphasize, that the width of the pass band for the waveguide loaded by 
723: transversal electric scatterers has the same order as in the case of transversal magnetic scatterers 
724: (Marques waveguide \cite{Marqueswaveguide,Hrabarwaveguide}) if the loading scatterers have parameters
725: obtained using duality principle from each other.
726: So, the loading by electric scatterers could be an alternative and even more appropriate solution for the
727: waveguide miniaturization than the design suggested in \cite{Hrabarwaveguide} for this purpose.
728: 
729: 
730: \section{Conclusion}
731: 
732: The dispersion properties of rectangular waveguides loaded by
733: resonant scatterers (magnetic and electric ones) have been
734: studied. The waveguide problem has been transformed using the
735: image theory into the eigenmode problem of an auxiliary
736: three-dimensional electromagnetic crystal. The dispersion
737: properties of such electromagnetic crystal have been modelled
738: using the local field approach. It has been revealed that not only
739: magnetic but also electric resonant scatterers allow to obtain 
740: mini pass band below cutoff frequency of the hollow waveguide.
741: The corresponding mini-band turns out to be
742: of the same order as for magnetic scatterers with same individual frequency
743: dispersion. So, the electric scatterers (inductively loaded short
744: wires) could be also prospective for the waveguide miniaturization as well as split-ring-resonators.
745: It has been shown that the loading by scatterers with longitudinal orientation of dipole moments
746: also allows to obtain the mini-band of propagation below the cutoff frequency of the hollow waveguide,
747: but the width of this band is significnatly narrower as compared to the case of transverse orientation.
748: The observed effects are explained in terms of the subwavelength guiding properties of the single chains of scatterers.
749: This explanation is supported by comparison of dispersion properties of the loaded waveguides
750: and the chains of the resonant scatterers in free space.
751: Results of our theory are in good agreement with the known literature data. 
752: For the chains of resonant dipoles the results from \cite{Weber} are reproduced. 
753: For the rectangular waveguide loaded by split-ring-resonators
754: the same result as in \cite{Marqueswaveguide} has been obtained.
755: 
756: \bibliography{MW}% Produces the bibliography via BibTeX.
757: 
758: \appendix
759: 
760: \section{Polarizabilities of resonant scatterers}
761: 
762: \subsection{Split-Ring Resonators}
763: 
764: The SRR considered in \cite{PendrySRR,SmithWSRR,Shelbyscience} is
765: a pair of two coplanar broken metal rings (see Fig.\ref{scat}.b).
766: Since the two loops of an SRR are not identical the analytical
767: models of it are rather cumbersome \cite{MarquesSRR,SimSRR}. In
768: fact, such SRR can not be described as a purely magnetic
769: scatterer, because it exhibits bianisotropic properties and has
770: resonant electric polarizability \cite{MarquesSRR,SimSRR} (see
771: also discussion in \cite{APSWSRR}). However, the electric
772: polarizability and bianisotropy of SRR is out of the scope of this
773: paper. We neglect these effects and consider an ordinary SRR as a
774: magnetic scatterer. The analytical expressions for the magnetic
775: polarizability $\alpha(\omega)$ of SRRs with geometry plotted in
776: Fig.\ref{scat}.a were derived and validated in \cite{SimSRR}. The
777: final result reads as follows: \e
778: \alpha(\omega)=\frac{A\omega^2}{\omega_0^2-\omega^2+j\omega\Gamma},
779: \qquad A=\frac{\mu_0^2\pi^2r^4}{L+M}, \l{alpha} \f where
780: $\omega_0$ is the resonant frequency of magnetic polarizability:
781: $$
782: \omega_0^2=\frac{1}{(L+M)C_r},
783: $$
784: $L$ is inductance of the ring (we assume that both rings have the
785: same inductance):
786: $$
787: L=\mu_0 r\left[\log\left(\frac{32R}{w}\right)-2\right],
788: $$
789: $M$ is mutual inductance of the two rings:
790: $$
791: M=\mu_0 r\left[(1-\xi)\log\left(\frac{4}{\xi}\right)-2+\xi\right],
792: \qquad \xi=\frac{w+d}{2r},
793: $$
794: $C_r$ is the effective capacitance of the SRR:
795: $$
796: C_r=\varepsilon_0 \frac{r}{\pi} {\rm arccosh}
797: \left(\frac{2w}{d}\right),
798: $$
799: $\Gamma$ is the radiation reaction factor:
800: $$
801: \Gamma=\frac{A\omega k^3}{6\pi\mu_0},
802: $$
803: $r$ is the inner radius of the inner ring, $w$ is the width of the
804: rings, $d$ is distance between the edges of the rings (see
805: Fig.\ref{scat}.a), $\varepsilon_0$ and $\mu_0$ are permittivity and
806: permeability of the host media, and
807: $k=\omega\sqrt{\varepsilon_0\mu_0}$ is the wave number of the host
808: medium. The presented formulae are valid within the frame of the
809: following approximations: $w,d\ll r$ and the splits of the rings
810: are large enough compared to $d$. Also, we assume that SRR is
811: formed by ideally conducting rings (no  dissipation losses).
812: 
813: The magnetic polarizability \r{alpha} takes into account the
814: radiation losses and satisfies to the basic Sipe-Kranendonk
815: condition \cite{Sipe,Belovcond,Belovnonres} which in the present
816: case has the following form: 
817: \e 
818: {\rm Im}\left\{\alpha^{-1}(\omega)\right\}=\frac{k^3}{6\pi\mu_0}.
819: \l{sipe} \f
820: 
821: In our analysis we operate with the inverse
822: polarizability $\alpha^{-1}(\omega)$, thus, we rewrite \r{alpha}
823: in the following form: \e
824: \alpha^{-1}(\omega)=A^{-1}\left(\frac{\omega_0^2}{\omega^2}-1\right)+j\frac{k^3}{6\pi\mu_0}.
825: \l{invalph} \f
826: 
827: \subsection{Inductively Loaded Short Wires}
828: 
829: An inductively loaded short wire is shown in Fig. \ref{scat}.b.
830: The electric polarizability $\alpha_e$ of an inductively loaded
831: wire following the known model \cite{LWD} has the form: \e
832: \alpha_e^{-1}= \frac{3}{l^2 C_{\rm wire}}
833: \left(\frac{1-\omega^2/\omega_0^2}{4-\omega^2/\omega_0^2}\right)+
834: j\frac{k^3}{6\pi\varepsilon_0} \l{alpe} \f where
835: $C_{\mbox{wire}}=\pi l\varepsilon_0/\log(2l/r_0)$ is the
836: capacitance of the wire, $\omega_0=\sqrt{L C_{\rm wire}}$ is the
837: resonant frequency,  $L$ is the inductance of the load, $l$ is the
838: half length of the wire and $r_0$ is the wire radius.
839: 
840: It is clear, that at the frequencies near the resonance the
841: polarizability of LSW has the form \e \alpha_e^{-1}(\omega)\approx
842: A_e^{-1}\left(\frac{\omega_0^2}{\omega^2}-1\right)+j\frac{k^3}{6\pi\varepsilon_0},
843: \l{invalphe} \f with $A_e=l^2 C_{\rm wire}$, which is similar to
844: \r{invalph}. Moreover, if $A_e/\varepsilon_0=A/\mu_0$ then using
845: duality principle the magnetic dipole with polarizability $\alpha$
846: \r{invalph} can be transformed to the electric dipole with
847: polarizability $\alpha_e$ \r{invalph}, and vice versa. This means
848: that it is enough to consider only one type of resonant
849: scatterers. In the present paper we have chosen magnetic ones to
850: be principal. The case of electric scatterers was obtained using
851: the duality principle with $A=\mu_0A_e/\varepsilon_0$.
852: 
853: \section{Interaction constants of the chains}
854: 
855: The initial expressions for the interaction constants $C_{x,y}$ entering
856: \r{displine} follow from \r{Green} and \r{gre} and read as follows:
857: \e 
858: C_x=\sum\limits_{m\ne 0}\frac{1+jka|m|}{2\pi
859: a^3|m|^3}e^{-j(k|m|+qm)a} \l{Cx} 
860: \f
861: $$
862: =\frac{1}{\pi a^3}\sum\limits_{m=1}^{+\infty}
863: \left[\frac{1}{m^3}+\frac{jka}{m^2}\right]e^{-jkam}\cos(qam),
864: $$
865: for the longitudinal polarization, and
866: \e C_y=\sum\limits_{m\ne 0} \frac{k^2a^2m^2-jka|m|-1}{4\pi
867: a^3|m|^3}e^{-j(k|m|+qm)a} \l{Cy} \f
868: $$
869: =\frac{1}{2\pi a^3}\sum\limits_{m=1}^{+\infty}
870: \left[\frac{k^2a^2}{m}-\frac{jka}{m^2}-\frac{1}{m^3}\right]e^{-jkam}\cos(qam)
871: $$
872: for the transverse one (see e.g. \cite{Weber}).
873: 
874: Note, that $C_x$ includes only near-field terms (of the order
875: $1/R^2$ and $1/R^3$). In contrast to $C_x$, the transverse
876: interaction constant $C_y$ includes also the wave terms (of the order $1/R$) which
877: corresponds to the slowly converging series. It makes the direct
878: numerical summation of \r{Cy} to be not efficient. The series in
879: \r{Cx} have better convergence, but it is also not enough for rapid calculations.
880: 
881: The application of acceleration technique done in \cite{Belovhomo}
882: offers a more cumbersome expression for $C_x(k,q,a)$ than \r{Cx},
883: but it is better for numerical calculations since the series
884: converges very rapidly:
885: $$
886: C_x=\frac{1}{4\pi a^3} \left[ 4\sum\limits_{m=1}^{+\infty}
887: \frac{(2jka+3)m+2}{m^3(m+1)(m+2)}e^{-jkam}\cos(qam) \right.
888: $$
889: \e -(jka+1)\left(t_+^2\log t^++t_-^2\log t^-+2e^{jka}\cos
890: (qa)\right) \l{c3} \f
891: $$
892: \left. -2jka\left(t_+\log t^++t_-\log t^-\right) +(7jka+3)
893: \vphantom{\sum\limits_{m=1}^{+\infty}
894: \frac{(2jka+3)m+2}{m^3(m+1)(m+2)}} \right],
895: $$
896: where
897: $$
898: t^+=1-e^{-j(k+q)a}, \qquad t^-=1-e^{-j(k-q)a},
899: $$
900: $$
901: t_+=1-e^{j(k+q)a}, \qquad t_-=1-e^{j(k-q)a}.
902: $$
903: 
904: As to $C_y$, the series of the order $1/m$ in \r{Cy} can be
905: obtained in the closed form using the tabulated formula (see
906: \cite{Collin}, Appendix): \e \sum\limits_{m=1}^{+\infty} \frac{e^{
907: -j\gamma m}}{m}= -\log \left(1-e^{-j\gamma}\right) \l{sumn} \f
908: $$=-\left(\log \left|2\sin \frac{s}{2}\right|+j\frac{\pi
909: -\gamma'}{2}\right),
910: $$ where $\gamma'=2\pi\{\gamma/(2\pi)\}$ and we
911: use notation $\{x\}$  for fractional part of variable $x$. The
912: rest part of the expression \r{Cy} is simply proportional to
913: $C_x$. Thus, $C_y$ can be evaluated as follows:
914: $$
915: C_y(k,q,a)=-\frac{k^2}{4\pi a} \log \left|2 \left( \cos ka- \cos
916: qa \right)\right|
917: $$
918: \e -
919: j\frac{k^2}{4a}\left(1-\left\{\frac{(k+q)a}{2\pi}\right\}-\left\{\frac{(k-q)a}{2\pi}\right\}\right)
920: -C_x(k,q,a)/2. \l{Cyx} \f
921: 
922: In the works \cite{Belovhomo,Tretlines} it was shown that \e {\rm
923: Im} (C_x)=\frac{k^3}{6\pi}+\frac{1}{4a}\sum\limits_{|q_m|<k}
924: \left(q_m^2-k^2\right), \l{imcx} \f where
925: $$
926: q_m=q+\frac{2\pi m}{a}.
927: $$
928: 
929: From the formula \r{Cy} using some algebra  it follows that \e
930: {\rm Im} (C_y)=\frac{k^3}{6\pi}-\frac{1}{8a}\sum\limits_{|q_m|<k}
931: \left(q_m^2+k^2\right). \l{imcy} \f
932: 
933: The expressions \r{imcx} and \r{imcy} demonstrate energy
934: transformations happening in the chains. A single scatterer
935: radiates cylindrical wave and that is why its polarizability has
936: radiation losses which can be described by Sipe-Kronendonk
937: condition \r{sipe}. Being arranged into the regular chains the
938: scatterers acquire effective polarizability (with respect to the external field) of the form: 
939: \e
940: \alpha_{x,y}=\left[\alpha^{-1}-\mu_0^{-1}C_{x,y}\right]^{-1}, 
941: \f
942: where indices $x$ and $y$ correspond to longitudinal and
943: transverse orientations of scatterers in the chain, respectively.
944: In accordance to \r{imcx} and \r{imcy} one can formulate the following
945: analogues of Sipe-Kronendonk condition for effective
946: polarizabilities of the scatterers in the chains: \e {\rm Im}
947: \left\{\alpha_x^{-1}\right\}=\frac{1}{4\mu_0a}\sum\limits_{|q_m|<k}
948: \left(k^2-q_m^2\right), \l{imax} \f \e {\rm Im}
949: \left\{\alpha_y^{-1}\right\}=\frac{1}{8\mu_0a}\sum\limits_{|q_m|<k}
950: \left(q_m^2+k^2\right). \l{imay} \f
951: 
952: Note, that terms $k^3/(6\pi)$ are canceled. It is clear, that if
953: there are no such index $m$ that $|q_m|<k$ (like it happens for
954: example if $k<q<2\pi/a-k$ for $k<\pi/a$) then effective
955: polarizabilities turn out to be purely real and the chain itself
956: does not radiate. This regime corresponds to the case of guiding
957: modes and it makes dispersion equation \r{displine} real valued
958: one. If there are some indices $m$ that $|q_m|<k$ then the
959: effective polarizabilites acquire non-zero imaginary part which
960: give evidence that the chain radiates cylindrical waves. The
961: number of such waves corresponds to the number of indices $m$
962: fulfilling to the relation $|q_m|<k$. If $|q|<k<\pi/a$ then the
963: chain radiates only one cylindrical wave which can be treated as
964: the main diffraction lobe of this periodical array. For the
965: higher frequencies the grating lobes appears and all of them make
966: its contribution to expressions \r{imax} and \r{imay}.
967: 
968: \section{Interaction constant of an orthorhombic lattice}
969: 
970: For effective numerical calculation of interaction constant $C(k,\-q,a,b,c)$ defined by \r{C} 
971: we use the following formula deducted in \cite{Belovhomo}:
972: $$
973: C(k,\-q,a,b,c)= -\sum\limits_{n=1}^{+\infty}\sum\limits_{{\rm
974: Re}(p_m)\ne 0} \frac{p_m^2}{\pi a}K_0\left(p_mbn\right)\cos(q_ybn)
975: $$
976: \e
977: +\sum\limits_{m=-\infty}^{+\infty}\sum\limits_{n=-\infty}^{+\infty}
978: \frac{p_m^2}{2jab k_z^{(mn)}} \frac{e^{-j k_z^{(mn)}c}-\cos
979: q_zc}{\cos k_z^{(mn)}c-\cos q_zc} \l{cfinal} \f
980: $$
981: -\sum\limits_{{\rm Re}(p_m)=0} \frac{p_m^2}{2ab}
982: \left(\frac{1}{jk_z^{(m0)}}+ \sum\limits_{n=1}^{+\infty}
983: \left[\frac{1}{jk_z^{(m,n)}}+\frac{1}{jk_z^{(m,-n)}}\right.
984: \right.
985: $$
986: $$
987: \left.\left. -\frac{b}{\pi n}-\frac{l_mb^3}{8\pi^3
988: n^3}\right]+1.202 \frac{l_mb^3}{8\pi^3}+ \frac{b}{\pi} \left(\log
989: \frac{b|p_m|}{4\pi}+\gamma\right)+j\frac{b}{2}
990: \vphantom{\sum\limits_{n\ne 0}
991: \left[\frac{1}{jk_z^{(mn)}}-\frac{b}{2\pi |n|}\right]}\right)
992: $$
993: $$
994: +C_x(k,q_x,a),
995: $$
996: where $C_x$ is given by \r{c3} and
997: $$
998: k_x^{(m)}=q_x+\frac{2\pi m}{a},\qquad k_y^{(n)}=q_y+\frac{2\pi
999: n}{b},
1000: $$
1001: $$
1002: p_m=\sqrt{\left(k_x^{(m)}\right)^2-k^2},\qquad l_m=2q_y^2-p_m^2,
1003: $$
1004: $$
1005: k_z^{(mn)}=-j\sqrt{\left(k_x^{(m)}\right)^2+\left(k_y^{(n)}\right)^2-k^2}.
1006: $$
1007: \end{document}
1008: