cond-mat0208211/PRB.tex
1: \documentstyle[aps,prb,epsfig]{revtex}
2: \begin{document}
3: \twocolumn 
4: \wideabs{ 
5: \title{Plasmon excitations and 1D -
6: 2D dimensional crossover in quantum crossbars} 
7: \author{I. Kuzmenko, S.
8: Gredeskul, K. Kikoin, Y. Avishai} 
9: \address{Department of Physics,
10: Ben-Gurion University of the Negev, Beer-Sheva} 
11: \date{\today}
12: \maketitle
13: \begin{abstract}
14: Spectrum of boson fields and two-point correlators are analyzed in
15: quantum crossbars (QCBs, a superlattice formed by $m$ crossed
16: interacting arrays of quantum wires), with short range inter-wire
17: capacitive interaction.  Spectral and correlation properties of double
18: ($m=2$) and triple ($m-3$) QCBs are studied.  It is shown that the
19: standard bosonization procedure is valid, and the system behaves as a
20: sliding Luttinger liquid in the infrared limit, but the high frequency
21: spectral and correlation characteristics have either 1D or 2D nature
22: depending on the direction of the wave vector in the 2D elementary
23: cell of reciprocal lattice.  As a result, the crossover from 1D to 2D
24: regime may be experimentally observed. It manifests itself as
25: appearance of additional peaks of optical absorption, non-zero
26: transverse space correlators and periodic energy transfer between
27: arrays ("Rabi oscillations").
28: \end{abstract}
29: } \vspace{\baselineskip}
30: \section{Introduction}\label{sec:Intro}
31: The behavior of electrons in arrays of $1D$ quantum wires was
32: recognized a challenging problem soon after the consistent theory
33: of elementary excitations and correlations in a Luttinger liquid (LL)
34: of interacting electrons in one dimension was formulated (see
35: \cite{Voit} for a review).  One of fascinating targets
36: is a search for LL features in higher dimensions
37: \cite{Anders}.  Although the Fermi liquid state seems to be rather
38: robust for $D>1$, the possible way to retain some $1D$ excitation
39: modes in $2D$ and even $3D$ systems is to consider highly anisotropic
40: objects, in which the electron motion is spatially confined in major
41: part of the real space (e.g., it is confined to separate linear
42: regions by potential relief).  One may hope that in this case weak
43: enough interaction does not violate the generic long-wave properties
44: of the LL state.  Arrays of interacting quantum wires may be formed in
45: organic materials and in striped phases of doped transition metal
46: oxides.  Artificially fabricated structures with controllable
47: configurations of arrays and variable interactions are available now
48: due to recent achievements in nanotechnology (see, e.g.,
49: Refs.\onlinecite{Rueckes,Dai}).\\
50: 
51: The simplest $2D$ ensemble of $1D$ nanoobjects is an array of 
52: parallel quantum wires.  The conventional LL regime in a single $1D$ 
53: quantum wire is characterised by bosonic fields describing charge and 
54: spin modes.  We confine our discussion to the charge sector (LL in the 
55: spin-gapped phase).  The Hamiltonian of an isolated quantum wire may 
56: then be represented in a canonical form
57: \begin{equation}
58: H = \frac{\hbar v}{2}\int\limits_{-L/2}^{L/2} {dx}
59: \left\{g{\pi}^{2}(x)+
60: \frac{1}{g}({\partial}_{x}\theta^2(x))\right\}.
61: \label{D1}
62: \end{equation}
63: Here $L$ is the wire length, $v$ is the Fermi velocity, $\theta,\pi$
64: are the conventional canonically conjugated boson fields and $g$ is the
65: dimensionless parameter which describes the strength of the interaction
66: within the chain (see, e.g., \cite{Voit,Delft}).
67: The interwire interaction may transform the LL state existing in
68: isolated quantum wires into various phases of $2D$ quantum liquid. The 
69: most
70: drastic transformation is caused by the {\it interwire} tunneling 
71: $t_{\perp}$
72: in arrays of quantum wires with {\it intrawire} Coulomb repulsion.
73: This coupling constant rescales towards higher values for strong 
74: interaction
75: ($g<1/2$), and the electrons in array transform into $2D$ Fermi liquid
76: \cite{Wen}.
77: The reason for this instability is the orthogonality catastrophe, i.e. 
78: the
79: infrared divergence in the low-energy excitation spectrum that 
80: accompanies
81: the interwire hopping processes.\\
82: 
83: Unlike interwire tunneling, the density-density or current-current
84: interwire interactions do not modify the low-energy behavior of quantum 
85: arrays
86: under certain conditions. In particular, it was shown recently
87: \cite{Luba00,Vica01,Luba01}
88: that an interaction of the type $W(n-n')$, which depends on the
89: distance between the wires $n$ and $n'$ but does not contain current
90: coordinates $x,x',$ imparts the properties of a {\it sliding phase} to 2D
91: array of 1D quantum wires. In this state an additional interwire coupling
92: leaves the  fixed-point action invariant under the "sliding" 
93: transformation
94: $\theta_n\to \theta_n+\alpha_n$ and $\pi_n \to \pi_n+\alpha^\prime_n$.
95: The contribution of interwire coupling reduces to a renormalization of 
96: the
97: parameters $v\to v(q_\perp)$, $g\to g(q_\perp)$ in the LL Hamiltonian
98: (\ref{D1}), where $q_\perp$ is a momentum perpendicular to the chain
99: orientation. Such LL structure can be interpreted as a quantum
100: analog of classical sliding phases of coupled $XY$
101: chains\cite{Hern}. Recently, it was found \cite{Sond} that a hierarchy of
102: quantum Hall states emerges in sliding phases when a quantizing magnetic 
103: field is applied to an array. \\  
104: 
105: In the present paper we concentrate on another aspect of the problem
106: of interacting quantum wires.  Instead of studying the conditions
107: under which the LL behavior is preserved in spite of interwire
108: interaction, we consider situations where the {\it dimensional
109: crossover} from $1D$ to $2D$ occurs.  Dimensional crossover is quite
110: well studied e.g. in thin semiconducting or superconducting films
111: where the film thickness is the control parameter that rules the
112: crossover (see e.g. Ref.\onlinecite {Buch}.  It occurs in strongly
113: anisotropic systems like quasi-one-dimensional organic
114: conductors\cite{sault} or layered metals\cite{metals}.  In the latter
115: cases temperature serves as a control parameter and crossover
116: manifests itself in interlayer transport.  In metals the layers appear
117: ``isolated'' at high temperature, but become connected at low
118: temperatures to manifest $3D$ conducting properties.  Here we intend
119: to study another type of dimensional crossover, i.e. a {\it
120: geometrical} crossover, where the phase variable serves as a control
121: parameter, and the excitations in quantum array demonstrate either
122: $1D$ or $2D$ behavior in different parts of reciprocal space.\\
123: 
124: The most promising type of artificial structures where this effect is 
125: expected is a periodic $2D$ system of $m$ crossing arrays of parallel 
126: quantum wires.  We call it "quantum crossbars" (QCB).  The square 
127: grids of this type consisting of $2$ arrays were considered in various 
128: physical context in early papers \cite{Avr,Avi,Guinea,Castro,Kuzm}.  
129: In Refs.\onlinecite {Guinea,Castro} the fragility of the LL state against 
130: interwire tunneling in the crossing areas of QCB was studied.  It was 
131: found that a new periodicity imposed by the interwire hopping term 
132: results in the appearance of a low-energy cutoff $\Delta_l\sim \hbar 
133: v/a$ where $a$ is a period of the quantum grid.  Below this energy, 
134: the system is "frozen" in its lowest one-electron state.  As a result, 
135: the LL state remains robust against orthogonality catastrophe, and the 
136: Fermi surface conserves its 1D character in the corresponding parts of 
137: the $2D$ Brilllouin zone (BZ).  This cutoff energy tends to zero at 
138: the points where the one-electron energies for two perpendicular 
139: arrays $\epsilon_{k_{1}}$ and $\epsilon_{k_{2}}$ become degenerate.  
140: As a result, a dimensional crossover from $1D$ to $2D$ Fermi surface 
141: (or from LL to FL behavior) arises around the points 
142: $\epsilon_{F_{1}}=\epsilon_{F_{2}}$.\\
143: 
144: We study this dimensional crossover for Bose excitations (plasmons)
145: described by canonical variables $\theta,\pi$ in QCB. In order to
146: unravel the pertinent physics we consider a grid with {\it short-range
147: capacitive inter-wire interaction}.  This approximation seems natural
148: for $2D$ grids of carbon nanotubes \cite{Rueckes}, or artificially
149: fabricated bars of quantum wires with grid periods which exceed the
150: lattice spacing of a single wire or the diameter of a nanotube.  It
151: will be shown below that this interaction can be made effectively
152: weak.  Therefore, QCB retains the $1D$ LL character for motion along
153: the wires similarly to the case considered in Ref.\onlinecite{Luba01}.  At
154: the same time, the boson mode propagation along some resonant
155: directions is also feasible.  This is essentially a $2D$ process in
156: the $2D$ BZ (or in the elementary cell of the reciprocal
157: lattice).  \\
158: 
159: We start the studies of QCB with a double QCB $m=2$ (section 
160: \ref{sec:Double}).  In the first two subsections \ref{subsec:Notions} 
161: and \ref{subsec:Hamilt} we introduce basic notions and construct the 
162: Hamiltonian of the QCB. The main approximations are discussed in 
163: subsection \ref{subsec:Approx}.  Here we substantiate the used method 
164: (separable interaction approximation) and show that interaction 
165: between arrays in QCB is weak.  The energy spectra for square QCB and 
166: tilted QCB are described in detail in two parts 
167: \ref{subsubsec:Square} and \ref{subsubsec:Tilted} of subsection 
168: \ref{subsec:Spectr}.  Various correlation functions and related 
169: experimentally observable quantities (optical absorption, space 
170: correlators) are discussed in the last subsection \ref{subsec:Correl}.  
171: We predict here effect of peculiar ``Rabi oscillations'' - periodic 
172: energy transfer from one of the QCB array to another.\\
173: 
174: Triple QCB ($m=3$) formed by three arrays lying in parallel planes are
175: studied in Section \ref{sec:Triple}.  Such hexagonal grids may be
176: useful for three-terminal nanoelectronic devices \cite{Luo}.  The
177: plasmon spectra of triple QCB possess some specific features in
178: comparison with double QCB. We introduce the main notions and
179: construct the Hamiltonian of symmetric triple QCB (subsection
180: \ref{subsec:NotHam}), analyse the peculiarities of the frequency
181: spectrum (subsection \ref{subsec:SpecTriple}), and illustrate them by
182: description of triple Rabi oscillations - periodic energy transfer
183: between all three arrays (subsection \ref{subsec:Observ}).  The
184: results are summarized in Conclusion.  All technical details are
185: placed in Appendices A - E.\\
186: 
187: \section{Double QCB} \label{sec:Double}
188: \subsection{Basic notions}\label{subsec:Notions}
189: Double QCB is a $2D$ periodic grid, which is formed by two 
190: periodically crossed arrays of $1D$ quantum wires.  In experimentally 
191: realizable setups \cite{Rueckes} these are cross-structures of 
192: suspended single-wall carbon nanotubes placed in two parallel planes 
193: separated by an inter-plane distance $d$.  However, some generic 
194: properties of QCB may be described in assumption that QCB is a genuine 
195: $2D$ system.  We assume that all wires of $j$-th array, $j=1,2,$ are 
196: identical.  They have the same length $L_j,$ Fermi velocity $v_{j}$ 
197: and Luttinger parameter $g_{j}.$ The arrays are oriented along the 
198: unit vectors ${\bf e}_{1,2}$ with an angle $\varphi$ between them.  
199: The periods of a crossbars along these directions are $a_{1}$ and 
200: $a_{2},$ and corresponding basic vectors are ${\bf a}_j=a_j{\bf e}_j$ 
201: (Fig.\ref{Bar3}).\\
202: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
203: \begin{figure}[htb]
204: \centering
205: \includegraphics[width=75mm,height=40mm,angle=0,]{Bar3.eps}
206: \epsfxsize=70mm \caption{$2D$ crossbars formed by two
207: interacting arrays of parallel quantum wires. Here ${\bf e}_{1}, {\bf
208: e}_{2}$ are the unit vectors of the superlattice, ${a_{1}},{a_{2}}$ 
209: are the superlattice periods and $d$ is the vertical interarray 
210: distance}
211: \label{Bar3}
212: \end{figure}
213: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
214: The interaction between the
215: excitations in different wires is assumed to be concentrated around the
216: crossing points with coordinates $n_1{\bf a}_1+n_2{\bf
217: a}_2\equiv(n_1a_1,n_2a_2)$. The integers $n_j$ enumerate the wires
218: within the $j$-th array.  Such interaction imposes a superperiodicity
219: on the energy spectrum of initially one dimensional quantum wires, and
220: the eigenstates of this superlattice are characterized by a $2D$
221: quasimomentum ${\bf q}=q_1{\bf g}_1+q_2{\bf g}_2 \equiv(q_1,q_2)$. 
222: Here ${\bf g}_{1,2}$ are the unit vectors of the reciprocal
223: superlattice satisfying the standard orthogonality relations $({\bf
224: e}_i\cdot {\bf g}_j)=\delta_{ij}$.  The corresponding basic vectors of
225: the reciprocal superlattice have the form $m_1Q_1{\bf g}_1 +
226: m_2Q_2{\bf g}_2 $, where $Q_j=2\pi/a_j$ and $m_{1,2}$ are integers. \\
227: 
228: However the crossbars kinematics differs from that of a standard 2D periodic 
229: system. In conventional $2D$ systems, forbidden states in the inverse 
230: space arise due to Bragg diffraction in a $2D$ periodic potential, 
231: whereas the whole plane is allowed for wave propagation in real space, 
232: at least until the periodic potential is weak enough. A 
233: Brillouin zone is bounded by the Bragg lines. It coincides with a 
234: Wigner-Seitz cell of reciprocal lattice.  In sharply anisotropic QCB 
235: most of the real space is forbidden for electron and plasmon 
236: propagation.  The Bragg conditions for the wave vectors 
237: are modulated by a periodic potential  unlike  
238: those in conventional $2D$ plane.
239: These conditions are essentially one-dimensional.  Corresponding BZ is not a 
240: Wigner-Seitz cell of a reciprocal lattice but the elementary cell 
241: containing a site in its center.\\
242: 
243: Indeed, the excitation motion in QCB is one-dimensional in major part 
244: of the $2D$ plane.  The anisotropy in real space imposes restrictions 
245: on the possible values of $2D$ coordinates $x_{1},x_{2}$ (${\bf 
246: r}=x_{1}{\bf e}_{1}+x_{2}{\bf e}_{2}$).  At least one of them, e.g., 
247: $x_2$ ($x_{1}$) should be an integer multiple of the corresponding 
248: array period $a_2$ ($a_{1}$), so that the vector ${\bf r}=(x_1,n_2 
249: a_2)$ (${\bf r}=(n_1a_1,x_2)$) characterizes the point with the $1D$ 
250: coordinate $x_1$ ($x_2$) lying at the $n_2$-th ($n_1$-th) wire of the 
251: first (second) array.  As a result, one cannot resort to the standard 
252: basis of $2D$ plane waves when constructing the eigenstate with a 
253: given wave vector ${\bf k}$. Even in {\it non-interacting} arrays of 
254: quantum wires (empty superlattice) the $2D$ basis is formed as a 
255: superposition of two sets of $1D$ waves.  The first of them is a set 
256: of $1D$ excitations propagating along {\it each} wire of the first 
257: array characterized by a unit vector $k_1{\bf g}_1$ with a phase shift 
258: $a_2k_2$ between adjacent wires.  The second set is the similar 
259: manifold of excitations propagating along the wires of the second 
260: array with the wave vector $k_2{\bf g}_2$ and the phase shift 
261: $a_1k_1$. The dispersion law of these excitations has the form
262: \begin{equation}
263: 	\omega^{0}({\bf k})=\omega_{1}(k_{1})+\omega_{2}(k_{2}).\nonumber 
264: %	\label{noninerdisp}
265: \end{equation}
266: The states of equal energy obtained by means of this 
267: procedure form straight lines in the $2D$ reciprocal space.  For 
268: example, the Fermi surface of QCB developed from the points $\pm 
269: k_{F1,2}$ for individual quantum wire consists of two sets of lines 
270: $|k_{1,2}|=k_{F1,2}$.  Respectively, the Fermi sea is not a circle 
271: with radius $k_F$ like in the case of free $2D$ gas, but a cross in 
272: the $k$ plane bounded by these four lines \cite{Guinea} (see 
273: Fig. \ref{FS3}). Finally, the Bragg conditions read
274: \begin{eqnarray*}
275: 	\omega_{1}(k_{1}) & - & omega_{1}(k_{1}+ m_{1}Q_{1})\\
276: 	& + & \omega_{2}(k_{2})-\omega_{2}(k_{2}+ m_{2}Q_{2})=0.
277: %	\label{Bragg}
278: \end{eqnarray*}
279: and the lines $k_{1}=0$, $|k_{2}|=Q_{2}/2$, and $|k_{1}|=Q_{1}/2$, 
280: $k_{2}=0,$ satisfying these conditions, form a $2D$ BZ of double 
281: QCB.\\
282: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
283: \begin{figure}[htb]
284: \centering
285: \includegraphics[width=75mm,height=70mm,angle=0,]{FS3.eps}
286: \epsfxsize=70mm \caption{Fermi surface of $2D$ metallic quantum
287: bars in the absence of charge transfer between wires.  ${\bf
288: g}_{1}, {\bf g}_{2}$ are the unit vectors of the reciprocal
289: superlattice} \label{FS3}
290: \end{figure}
291: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
292: 
293: Due to the inter-wire interaction, the excitations of QCB (see 
294: Figs.\ref{BZ2},\ref{BZ3} below) acquire genuine two-dimensionality 
295: characterized by the quasimomentum ${\bf q}=(q_1,q_2)$.  However, in 
296: case of weak interaction the $2D$ waves constructed from the $1D$ 
297: plane waves in accordance with the above procedure form an appropriate 
298: basis for the description of elementary excitations in QCB in close 
299: analogy with the nearly free electron approximation in conventional 
300: crystalline lattices. It is easily foreknown that a weak inter-wire 
301: interaction does not completely destroy the above quasimomentum 
302: classification of eigenstates, and the $2D$ reconstruction of the 
303: spectrum may be described in terms of wave mixing similarly to the 
304: standard Bragg diffraction in a weak periodic potential.  Moreover, 
305: the classification of eigenstates of empty superlattice may be 
306: effectively used for the classification of energy bands in a real QCB 
307: superlattice where the superperiodicity is imposed by interaction.\\
308: 
309: Complete kinematics of an empty superchain (wave functions, dispersion 
310: laws, relations between quasiparticle second quantization operators) 
311: is developed in Appendix A. In terms of these $1D$ Bloch functions 
312: (see Eqs.  (\ref{WaveFunc}), (\ref{WaveFunc1}) of Appendix A) we 
313: construct the $2D$ basis of Bloch functions for an empty superlattice
314: \begin{equation}
315: {\Psi}_{s,s',{\bf q}}({\bf r})={\psi}_{1,s,q_{1}}(x_{1})
316:             {\psi}_{2,s',q_{2}}(x_{2}).
317:             \label{Psi}
318: \end{equation}
319: Here $s,s'=1,2,\ldots,$ are the band numbers, and the
320: $2D$ quasimomentum ${\bf q}=(q_{1}, q_{2})$ belongs to the first BZ,
321: $|q_{j}|\leq Q_{j}/2.$ The corresponding eigenfrequencies are
322: \begin{equation}
323:     \omega_{ss'}({\bf q})=
324:     \omega_{1,s}({\bf q})+\omega_{2,s'}({\bf q}).\nonumber 
325: %    \label{Omega}
326: \end{equation}
327: Here
328: $$\omega_{j,s}({\bf q})\equiv \omega_{j,s}(q_{j}),$$
329: and $\omega_{j,s}(q_{j})$ is defined by Eqs. (\ref{Disp1}) below. 
330: We will use this basis in the next subsection when constructing the
331: excitation spectrum of QCB within the reduced band scheme.\\
332: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
333: \subsection{Hamiltonian}\label{subsec:Hamilt}
334: When turning to description of interaction in a QCB, one should refer 
335: to a real geometry of crossbars, and recollect the important fact that 
336: the equilibrium distance between two arrays is finite and large enough 
337: to supress direct electron tunneling 
338: cite{Rueckes}.  We neglect also the elastic and 
339: van der-Waals components of intertaction between real nanotubes, 
340: because these interactions are not involved in formulation of 
341: collective excitations in QCB. Then the full Hamiltonian of the QCB 
342: is
343: \begin{equation}
344:   H =  {H}_{1} + {H}_{2} +  H_{{int}},
345:   \label{TotHam1}
346: \end{equation}
347: where ${H}_{j}$ describes the $1D$ boson field
348: in the $j$-th array
349: \begin{eqnarray*}
350: {H}_{1} & = &
351:          \frac{\hbar{v_1}}{2}\sum_{{n}_{2}}
352:               \int\limits_{-L_1/2}^{L_1/2} {dx}_{1}
353:               \biggl\{
354:                    {g_1}{\pi}_{1}^{2} \left( x_1,
355:                    {n}_{2}a_2\right)
356:                    \\
357:                &+ &    \frac{1}{g_1}
358:                    \left(
359:                         {\partial}_{{x}_{1}}
360:                         {\theta}_{1}
361:                         \left(
362:                              {x}_{1},{n}_{2}a_2
363:                         \right)
364:                    \right)^2
365:          \biggr\},
366: \end{eqnarray*}
367: \begin{eqnarray*}
368: {H}_{2}& = &
369:      \frac{\hbar{v_2}}{2}\sum_{{n}_{1}}
370:          \int\limits_{-L_2/2}^{L_2/2}{dx}_{2}
371:          \biggl\{
372:           g_2{\pi}_{2}^{2}\left({n}_{1}a_1,{x}_{2}\right)
373:               \\
374:              &+ &\frac{1}{g_2}
375:               \left(
376:                    {\partial}_{x_2}
377:                    {\theta}_{2}
378:                    \left(
379:                         {n}_{1}a_1,{x}_{2}
380:                    \right)
381:               \right)^2
382:          \biggr\},
383: \end{eqnarray*}
384: and $\theta_j,\pi_j$
385: are the conventional canonically conjugated boson fields (see,
386: e.g., Ref.\onlinecite {Delft}).\\
387: 
388: \noindent
389:  The interwire interaction results from a short--range
390: contact capacitive coupling in the crosses of bars,
391: \begin{eqnarray*}
392: H_{{int}} & = & \sum\limits_{{n}_{1},{n}_{2}}
393:             \int dx_1
394:             dx_2
395:             V(x_1-n_1a_1,n_2a_2-x_2) \\
396:            & \times & {\rho}_{1}({x}_{1},n_2a_2)
397:             {\rho}_{2}(n_1a_1,{x}_{2}),
398: \end{eqnarray*}
399: where the integration is restricted by the area $-L_j/2\leq
400: x_j\leq L_j/2$. Here $\rho_i({\bf r})$ are density operators, and
401: $V({\bf r}_{1}-{\bf r}_{2})$ is a short-range interwire
402: interaction. Physically, it represents the Coulomb interaction
403: between charge fluctuations 
404: \begin{equation}
405: 	e\zeta\left(\frac{x_j-n_{j}a_{j}}{r_{j}}\right), \ \ 
406: 	\zeta(\xi)=\zeta(-\xi), \ \ 
407:     \zeta(0)=1,
408: 	\label{charge}
409: \end{equation}
410: around the crossing point $(n_{1}a_{1},n_{2}a_{2}).$
411: The size of the fluctuation on the wire of the $j$-th array, 
412: is determined by the screening radius $r_{j}$ within the wire.
413: One may neglect the inter-wire tunneling and restrict oneself by
414: the capacitive interaction only, provided the vertical distance
415: between the wires $d$ is substantially larger than the screening
416: radiuses $r_{j}.$ Therefore the interaction has the form,
417: \begin{eqnarray*}
418:     V({\bf r})=\frac{V_{0}}{2}
419:         \Phi\left
420:         (\frac{x_1}{r_{1}},\frac{x_2}{r_{2}}
421:         \right),
422:         %\label{V}
423: \end{eqnarray*}
424: where the function $\Phi(\xi_{1},\xi_{2})$ is
425: \begin{equation}
426:     \Phi(\xi_{1},\xi_{2})=
427:         \frac
428:     {
429:     \displaystyle{
430:     \zeta_1(\xi_{1})
431:     \zeta_2(\xi_{2})
432:     }}
433:     {\displaystyle{
434:     \sqrt{1+\frac{|{\bf r}_{12}|^2}{d^2}
435:     }}
436:     },
437:         \label{F}
438: \end{equation}
439: $$
440:     {\bf r}_{12}= r_{1}\xi_{1}{\bf e}_{1}-
441:     r_{2}\xi_{2}{\bf e}_{2}.
442: $$
443: It is seen from these equations that $\Phi(\xi_{1},\xi_{2})$
444: vanishes for $|\xi_{1,2}|\ge 1$ and is normalized by condition
445: ${\Phi}(0,0)=1$. The effective coupling strength is
446: \begin{equation}
447:     V_0=\frac{2e^{2}}{d}. 
448:     \label{strength}
449: \end{equation}
450: In terms of boson field operators
451: ${\theta}_{i}$, the interaction is written as
452: \begin{eqnarray*}
453:     H_{{int}} & = & V_{0}\sum\limits_{{n}_{1},{n}_{2}}
454:             \int dx_1 dx_2
455:             \Phi\left
456:         (\frac{x_1-n_1a_1}{r_{1}},\frac{n_2a_2-x_2}{r_{2}}
457:         \right) {}\\
458:         &\times & {} \partial_{x_1}\theta_1(x_1,n_2a_2)
459:          \partial_{x_2}\theta_2(n_1a_1,x_2).
460: \end{eqnarray*}
461: 
462: In the quasimomentum representation
463: (\ref{Psi})
464: the full Hamiltonian (\ref{TotHam1}) acquires the form,
465: \begin{eqnarray}
466: H & = & \frac{{\hbar}{v}{g}}{2}\displaystyle{
467:                       \sum_{j=1}^{2}
468:                       \sum_{s,{\bf q}}
469:                       }
470:                       {\pi}_{js{\bf q}}^{\dagger}
471:                       {\pi}_{js{\bf q}}
472:                       +\nonumber\\
473:           & &\frac{\hbar}{2vg}\displaystyle{
474:           \sum_{jj'=1}^{2}
475:                  \sum_{s,s',{\bf q}}
476:                  }
477:                  {W}_{jsj's'{\bf q}}
478:                  {\theta}_{j s {\bf q}}^{\dagger}
479:                  {\theta}_{j's'{\bf q}},
480:                  \label{TotHam2}
481: \end{eqnarray}
482: where $\sqrt{vg/v_{j}g_{j}}{\theta}_{js{\bf q}}$ and 
483: $\sqrt{v_{j}g_{j}/vg}{\pi}_{js{\bf q}}$ are the Fourier components of 
484: the boson fields ${\theta}_{j}$ and ${\pi}_{j}$, and effective 
485: velocity and coupling are $v=\sqrt{v_1v_2}$, $g=\sqrt{g_1g_2}$ 
486: respectively.\\
487: 
488: The matrix elements for interwire coupling are given by:
489: \begin{eqnarray*}
490: {W}_{jsj's'{\bf{q}}} & = &
491:     {\omega}_{j s }(q_{j} )
492:     {\omega}_{j's'}(q_{j'})
493:     \left[
494:          {\delta}_{jj'}{\delta}_{ss'}+
495:          {\phi}_{jsj's'{\bf{q}}}
496:          \left(
497:               1-{\delta}_{jj'}
498:          \right)
499:     \right].
500: \end{eqnarray*}
501: Here
502: \begin{equation}
503: \omega_{js}(q_{j})=v_j
504:   \left(
505:        \left[\frac{s}{2}\right]Q_j+
506:        \left(-1\right)^{s-1}
507:        \left\vert{q}_{j}\right\vert
508:   \right),\nonumber
509: %  \label{dispersion}
510: \end{equation}
511: are eigenfrequencies of the ``unperturbed'' $1D$ mode (see Eq.  
512: (\ref{Disp1}) of Appendix A), pertaining to an array $j$, band $s$ and 
513: quasimomentum ${\bf q}=q_{j}{\bf g}_{j}.$ The coefficients
514: \begin{eqnarray*}
515: {\phi}_{1s2s'{\bf{q}}} = \phi(-1)^{s+s'}
516:                        \mbox{sign}{(q_{1}q_{2})}
517:                        {\Phi}_{1s2s'{\bf{q}}},\\
518:                        \phi = 
519:                        \frac{gV_{0}r_{0}^{2}}{{\hbar}va},\ \ \
520: 		       r_0=\sqrt{r_{1}r_{2}}, \ \ \
521: 		       a=\sqrt{a_1a_2},
522:                       % \label{phi}
523: \end{eqnarray*}
524: are proportional to the dimensionless Fourier component of the
525: interaction strengths
526: \begin{eqnarray*}
527: {\Phi}_{1s2s'{\bf{q}}}& =&
528:   \int d{\xi}_{1}d{\xi}_{2}{\Phi}({\xi}_{1},{\xi}_{2})
529:   e^{-i(r_{1}q_1\xi_1+r_{2}q_2\xi_2)}\\
530:   & \times &
531:   u_{1,s, q_1}^{*}(r_{1}\xi_1)
532:   u_{2,s',q_2}^{*}(r_{2}\xi_2)={\Phi}^{*}_{2s'1s
533:   {\bf{q}}}.
534: %  \label{Phi}
535: \end{eqnarray*}
536: 
537: The Hamiltonian (\ref{TotHam2}) describes a system of coupled
538: harmonic oscillators, which can be {\em exactly} diagonalized with
539: the help of a certain canonical linear transformation (note that
540: it is already diagonal with respect to the quasimomentum ${\bf
541: q}$).  The diagonalization procedure is, nevertheless, rather
542: cumbersome due to the mixing of states belonging to different
543: bands and arrays. However, it will be shown below that provided
544: $d\gg{r}_{1,2}$, a separable potential
545: approximation is applicable, that shortens calculations noticeably.\\
546: \subsection{Approximations}\label{subsec:Approx}
547: As it was already mentioned, we consider the rarefied
548: QCB with short range capacitive interaction.  In the case of QCB
549: formed by nanotubes, this is a Coulomb interaction screened at a
550: distance of the order of the nanotube radius\cite{Sasaki} $R_{0},$
551: therefore $r_{0}\sim R_{0}.$ The minimal radius of a single-wall
552: carbon nanotube is about $R_{0}=0.35\div 0.4 nm$ (see
553: Ref.\onlinecite{Louie}).  The intertube vertical distance $d$ in
554: artificially produced nanotube networks is estimated as $d\approx
555: 2$nm (see Ref.\onlinecite{Rueckes})  Therefore the ratio
556: $r_{0}^{2}/d^{2}\approx{0.04}$ is really small and {\it the
557: dimensionless interaction} $\Phi(\xi_{1},\xi_{2})$ (\ref{F}) {\it in the 
558: main approximation is separable}
559: \begin{equation}
560:   \Phi(\xi_{1},\xi_{2})\approx\Phi_{0}(\xi_{1},\xi_{2})= 
561:   \zeta_1(\xi_1)\zeta_2(\xi_2).
562:   \label{separ}
563: \end{equation}
564: It should be noted that the interaction in this form is
565: an even function of its arguments, and the odd correction to the
566: $\Phi_{0}$ is of order $r_0^2/d^2$, whereas $\Phi_{0}$ is of order of 
567: $1$.\\
568: 
569: To diagonalize the Hamiltonian (\ref{TotHam2}), one should solve the 
570: system of equations of motion for the field operators.  Generalized 
571: coordinates $\theta$ satisfy the equations
572: \begin{eqnarray}
573:   \left[\omega_{1s}^{2}(q_1)-\omega^{2}\right]
574:   \theta_{1s{\bf{q}}}+
575:   \sqrt{\varepsilon}\phi_{1s}(q_1)\omega_{1s}(q_1)
576:   \nonumber\\
577:   \times\frac{r_0}{a}\sum\limits_{s'}
578:   \phi_{2s'}(q_2)\omega_{2s'}(q_2)
579:   \theta_{2s'{\bf{q}}}=0,\nonumber\\
580:   s=1,2,\ldots ,
581:   \label{Euler-Lagr}
582: \end{eqnarray}
583: and the similar equations obtained by permutation 
584: $1\leftrightarrow 2$.  Here
585: \begin{equation}
586:   \phi_{js}(q)=(-1)^s\mbox{sign}(q)\int d\xi
587:   \zeta_{j}(\xi) e^{ir_{0}q\xi}u_{jsq}(r_{0}\xi),
588:   \label{FC-2}
589: \end{equation}
590: Bloch amplitudes $u_{jsq}(r_{0}\xi)$ are defined by Eqs. 
591: (\ref{WaveFunc1}) of Appendix A, and
592: \begin{equation}
593:     \varepsilon=\left(\phi\frac{a}{r_0}\right)^{2}=
594:     \left(\frac{gV_0r_0}{{\hbar}v}\right)^{2}.
595:     \label{epsilon}
596: \end{equation}
597: Due to separability of the interaction, equations of motion
598: (\ref{Euler-Lagr}) can be solved exactly.  Corresponding 
599: square eigenfrequencies are determined by the characteristic equation
600: \begin{equation}
601:  F_{1q_1}(\omega^2)F_{2q_2}(\omega^2)=\frac{1}{\varepsilon},
602:  \label{secul-eq}
603: \end{equation}
604: where
605: \begin{equation}
606:  F_{jq}(\omega^2)=\frac{r_{j}}{a_{j}}\sum\limits_{s}
607:  \frac{\phi_{js}^{2}(q)\omega_{js}^{2}(q)}
608:       {\omega_{js}^{2}(q)-\omega^2}.
609:  \label{F_j}
610: \end{equation}
611: The function $F_{jq}(\omega^2)$ has a set of poles at 
612: $\omega^2=\omega_{js}^{2}(q)$, $s=1,2,3,\ldots$ .  For squared 
613: frequency smaller than all squared initial eigenfrequencies 
614: $\omega_{js}^{2}(q)$, i.e.  within the interval $[0,\omega_{j1}^{2}]$, 
615: this is a positive and growing function.  Its minimal value $F_{j}$ on 
616: the interval is reached at $\omega^2=0,$ and it does not depend on 
617: quasimomentum $q$
618: \begin{equation}
619: F_{jq}(0)=\frac{r_{j}}{a_{j}}\sum\limits_{s}
620:  \phi_{js}^{2}(q)=\int d\xi \zeta_j^2(\xi)\equiv F_{j} 
621: \label{F_j0}
622: \end{equation}
623: (here Eqs.  (\ref{F_j}) and (\ref{FC-2}) are used). If parameter
624: $\varepsilon$ is smaller than its critical value
625: \begin{equation}
626:   \varepsilon_c=
627:   \frac{1}{F_{1}F_{2}},\nonumber 
628: %  \label{eps-cr}
629: \end{equation}
630: then all solutions $\omega^{2}$ of the characteristic equation are 
631: positive.  When $\varepsilon$ increases, the lowest QCB mode softens 
632: and its square frequency vanishes \textit{in a whole BZ} at 
633: $\varepsilon=\varepsilon_{c}$.  For exponential charge density 
634: distribution $\zeta(\xi)=\exp(-|\xi|),$ one obtaines 
635: $\varepsilon_c\approx 1$.\\
636: 
637: In our model the dimensionless interaction $\varepsilon$ in
638: Eq.(\ref{epsilon}) can be written as
639: \begin{equation}
640:     \varepsilon=\left(\frac{2R_{0}}{d}\frac{ge^{2}}{\hbar v}\right)^{2}.
641:     \label{epsilon1}
642: \end{equation}
643: For nanotube QCB, the first factor within parentheses is about 
644: $0.35.$ The second one which is nothing but the corresponding QCB 
645: ``fine structure'' constant, can be estimated as $0.9$ (we used the 
646: values of $g=1/3$ and $v=8\times 10^{7}$cm/sec, see Ref.\onlinecite 
647: {Egger}).  Therefore $\varepsilon$ approximately equals $0.1,$ so this 
648: parameter is really small. Thus the considered system is stable, its 
649: spectrum is described by Eqs.(\ref{secul-eq}), (\ref{F_j}) with a 
650: \emph{small} parameter $\varepsilon$.\\
651: 
652: The general Eq.(\ref{secul-eq}) reduces in infrared limit ${\bf q}, 
653: \omega \to 0$ to an equation describing the spectrum of two coupled 
654: sliding phases.  i.e.  $1:1$ arrays in accordance with classification 
655: , offered in Ref.  \onlinecite{Luba01}.  Equation (3.13) of this paper 
656: is a long wave limit of our equation (\ref{omega-12}) derived in 
657: Appendix B. Therefore the general analysis of stability of the LL 
658: fixed point is appicable in our approach.  
659: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
660: \subsection{Spectrum}\label{subsec:Spectr}
661: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
662: Due to the smallness of interaction, the systematics of unperturbed 
663: levels and states is grossly conserved, at least in the low energy 
664: region corresponding to the first few energy bands.  This means that 
665: perturbed eigenstates could be described by the same quantum numbers 
666: (array number, band number and quasimomentum) as the unperturbed ones.  
667: Such a description fails in two specific regions of reciprocal space.  
668: The first of them is the vicinity of lines $q_{j}=nQ_{j}/2$ with $n$ 
669: integer.  Indeed, as it follows from the equations of motion 
670: (\ref{Euler-Lagr}), around these lines the interband mixing is 
671: significant.  These lines with $n=\pm 1$ include BZ boundaries.  
672: Because of this BZ which is, generally speaking, non relevant, and 
673: in this subsection we refer mostly to BZ.\\
674: 
675: The second region is the vicinity of the lines 
676: where the resonance conditions are fulfilled
677: \begin{equation}
678: 	\omega^{2}_{1s}(q_{1})=\omega^{2}_{2s'}(q_{2}).
679: 	\label{res}
680: \end{equation}
681: Here inter-array mixing within the same energy band ($s=s'$) or 
682: between neighboring bands ($s\neq s'$) is significant. In what 
683: follows we will pay attention first 
684: of all to these two regions because in the rest of the BZ the initial
685: systematics of the energy spectrum can be successfully used.\\
686: 
687: Equations (\ref{Euler-Lagr}), (\ref{secul-eq}), describing the wave 
688: functions 
689:  and the dispersion laws  are analysed in 
690: Appendix B. We describe below some of these dispersion 
691: qurves for two types of QCB basing on this analysis.\\
692: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
693: \subsubsection{Square QCB}\label{subsubsec:Square}
694: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
695: We start with the simplest case of square QCB formed by identical 
696: wires.  This means that all parameters (wire length, space period, 
697: Fermi velocity, LL parameter, screening radius) are the same for both 
698: arrays.  The corresponding BZ and is also a 
699: square (see Fig.\ref{BZ2}). Resonant lines are the diagonals of BZ.\\
700: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
701: \begin{figure}[htb]
702: \centering
703: \includegraphics[width=75mm,height=70mm,angle=0,]{BZ2.eps}
704: \epsfxsize=70mm 
705: \caption{Two dimensional BZ of square QCB.} 
706: \label{BZ2}
707: \end{figure}
708: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
709: 
710: In the major part of the BZ, for quasimomenta ${\bf{q}}$ lying far
711: from the diagonals, each eigenstate mostly conserves its initial
712: systematics, i.e. belongs to a given array, and mostly depends on a
713: given quasimomentum component.  Corresponding dispersion laws remain
714: linear being slightly modified near the BZ boundaries only.  The main
715: change is therefore the renormalization of the plasmon
716: velocity.\\
717: %%%%%%%%%%%%%%%%%%
718: \begin{figure}[htb]
719: \centering
720: \includegraphics[width=75mm,height=55mm,angle=0,]{sp-sq.eps}
721: \epsfxsize=70mm 
722: \caption{The energy spectrum of QCB (solid
723: lines) and noninteracting arrays (dashed lines) for quasimomenta at
724: the lines $OA,$ $FC,$ and $OC$ of BZ} 
725: \label{sp-sq}
726: \end{figure}
727: %%%%%%%%%%%%%
728: 
729: In the left part of Fig.\ref{sp-sq} we displayed dispersion curves
730: corresponding to quasimomenta belonging to a generic $OA$ line in
731: BZ.  In what follows we use $(j,s)$ notations for the
732: unperturbed boson propagating along the $j$-th array in the $s$-th
733: band.  Then the lowest curve in this part of Fig.\ref{sp-sq} is, in
734: fact, the slightly renormalized dispersion of a $(2,1)$ boson. The
735: middle curve describes $(1,1)$ boson, and the upper curve is the
736: dispersion of a $(1,2)$ boson.  The fourth frequency corresponding to
737: a $(2,2)$ boson, is far above and is not displayed in the figure.  It
738: is seen that the dispersion remains linear along the whole line $OA$
739: except a nearest vicinity of the BZ boundary (point $A$ in
740: Fig.\ref{BZ2}).\\
741: 
742: Dispersion curves corresponding to quasi momenta lying at the BZ 
743: boundary $q_{1}=Q/2,$ $0\leq q_{2}\leq Q/2$ (line $FC$ in 
744: Fig.\ref{BZ2}) are displayed in the central part in Fig.\ref{sp-sq}).  
745: The characteristic feature of this boundary is the intra-band 
746: degeneracy in one of two arrays.  Indeed, in zero approximation, two 
747: modes $(1,s),$ $s=1,2,$ propagating along the first array are 
748: degenerate with unperturbed frequency $\omega=0.5.$ The interaction 
749: lifts the degeneracy.  This interaction occurs to be repulsive at the 
750: BZ boundaries.  As a result the lowest of two middle curves in 
751: Fig.\ref{sp-sq} corresponds to $(1,u)$ boson, and upper of them 
752: describes $(1,g)$ boson.  Here the indices $g,u$ denote a boson parity 
753: with respect to the transposition of the band numbers.  Note that 
754: $(1,g)$ boson exactly conserves its unperturbed frequency 
755: $\omega=0.5.$ The latter fact is related to the square symmetry of the 
756: QCB.\\
757: 
758: Two others curves correspond to almost non pertubed bosons of the 
759: second array.  The lowest curve describes the dispersion of the 
760: $(2,1)$ wave.  Its counterpart in the second band $(2,2)$ is described 
761: by the highest curve in the figure.  Their dispersion laws are 
762: nearly linear, and deviations from linearity are observed only near 
763: the corner of the BZ (point $C$ in Fig.\ref{BZ2}).\\
764: 
765: Consider now dispersion relations of modes with quasi--momenta on the diagonal 
766: $OC$ of BZ and start with ${\bf q}$ not too close to the BZ corner 
767: $C$ ($q_{1}=q_{2}=Q/2$).  This diagonal is actually one of the 
768: resonance lines.  Two modes in the first band coressponding to 
769: different arrays are strongly mixed.  They mostly have a definite 
770: $j$-parity with respect to transposition of array numbers $j=1,2$.  
771: Interaction between these modes occurs to be attractive (repulsive) 
772: for $q_{1}q_{2}>0$ ($q_{1}q_{2}<0$).  Therefore the odd modes $(u,s),$ 
773: at the BZ diagonal $OC$ $s=1,2,$ correspond to lower frequencies and 
774: the even modes $(g,s)$ correspond to higher ones. The corresponding 
775: dispersion curves are displayed in the right part of 
776: Fig.\ref{sp-sq}.\\
777: 
778:  At the BZ corner $q_{1}=q_{2}=Q/2$ (point $C$ in Fig.\ref{BZ2}) all 
779:  four initial modes $j,s=1,2$ are degenerate in the lowest  
780:  approximation.  This four-fold degeneracy results from the square 
781:  symmetry of BZ (the resonant lines are diagonals of the Z).  Weak 
782:  inter-wire interaction partially lifts the degeneracy, however the 
783:  split modes have a definite $s$-parity with respect to transposition 
784:  of band numbers $s=1,2$.  The lowest frequency corresponds mostly to 
785:  $(g,u)$ boson, symmetric with respect to transposition of array 
786:  numbers, but antisymmetric with respect to the transposition of band 
787:  numbers.  The upper curve describes a $(u,u)$ boson with odd both 
788:  $j$-parity and $s$-parity.  The two middle modes with even band 
789:  parity, $(g,g)$ and $(u,g)$ bosons, remain degenerate and their 
790:  frequencies conserve the unperturbed value $\omega=0.5$.  This also 
791:  results from the square symmetry of QCB (\ref{F}).  \\
792: %%%%%%%%%%%%
793: \begin{figure}[htb]
794: \centering
795: \includegraphics[width=65mm,height=65mm,angle=0,]{IsoEn2.eps}
796: \epsfxsize=80mm 
797: \caption{Lines of equal frequency of the lowest
798: mode for QCB (solid lines) and for noninteracting arrays (dashed
799: lines).  The lines $1,2,3$ correspond to the frequencies
800: $\omega_{1}=0.1,$ $\omega_{2}=0.25,$ $\omega_{3}=0.4$}
801: \label{IsoEn}
802: \end{figure}
803: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
804: All these results show that the quantum states of the $2D$ QCB
805: conserve the quasi $1D$ character of the Luttinger--like liquid in
806: major part of momentum space, and that $2D$ effects can be
807: successfully calculated within the framework of perturbation theory. 
808: However, bosons with quasimomenta close to the resonant line (diagonal
809: $OC$) of the BZ are strongly mixed bare $1D$ bosons.  These
810: excitations are essentially two-dimensional, and therefore the lines
811: of equal energy in this part of the BZ are modified by the $2D$
812: interaction (see Fig.\ref{IsoEn}).  It is clearly seen that deviations
813: from linearity occur only in a small part of the BZ. The crossover
814: from LL to FL behavior around isolated points of the BZ due to a
815: single-particle hybridization (tunneling) for Fermi excitations was
816: noticed in Refs.  \onlinecite{Guinea,Castro}, where a mesh of
817: horizontal and vertical stripes in superconducting cuprates was
818: studied.\\
819: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
820: \subsubsection{Tilted QCB}\label{subsubsec:Tilted}
821: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
822: Now we consider the spectrum of a generic double QCB. In this case all
823: parameters (wire length, space period, Fermi velocity, LL parameter,
824: screening radius) depend, generally speaking, on the array index $j.$
825: In what follows we refer to such a QCB as a tilted QCB. Now the
826: resonance condition (\ref{res}) is fulfilled not at the BZ diagonal
827: but at the resonant polygonal line.  Its part $ODE,$ lying in the
828: first quarter of the BZ, is displayed in Fig.\ref{BZ3} (all figures of
829: this subsection correspond to specific values
830: $v_{2}Q_{2}=1,\phantom{aa}v_{1}Q_{1}=1.4$).This results in
831: qualitative changes of the spectrum that are related first of all to
832: the appearance of two points $D$ and $E$ of the three-fold degeneracy for a
833: titled QCB (Fig.\ref{BZ3}) instead of a single point $C$ of four-fold
834: degeneracy for a square QCB (Fig.\ref{BZ2}).\\
835: %%%%%%%%%%%%%%%%%%%%%
836: \begin{figure}[htb]
837: \centering
838: \includegraphics[width=75mm,height=75mm,angle=0,]{BZ3a.eps}
839: \epsfxsize=70mm 
840: \caption{BZ of a titled QCB} 
841: \label{BZ3}
842: \end{figure}
843: %%%%%%%%%%%%%%%%%%%
844: 
845: We start with the resonant line $ODE$ (Fig.\ref{sp-res}).  The
846: dispersion curves at its $OD$ part and the symmetry properties of the 
847: corresponding eigenstates are similar to those at the $OC$ resonant line
848: for the square QCB (Fig.\ref{BZ2}).  The only difference is that instead
849: of the four-fold degeneracy at the BZ corner $C$ of the square QCB,
850: there is a three-fold degeneracy at the point $D$ lying at the BZ
851: boundary. A completely new situation takes place at the $DE$ line, where
852: two other modes $(1,1)$ and $(2,2),$ corresponding to different arrays
853: and different bands, are degenerate.  The interaction lifts this
854: degeneracy and the two middle lines in Fig.\ref{sp-res} describe even
855: $(g)$ and odd $(u)$ combinations of these modes.  The even mode
856: corresponds to the lowest frequency and the odd mode corresponds to the
857: higher one.  At the point $E$ one meets another type of a three-fold
858: degeneracy described in more detail in the next paragraph..\\
859: 
860: Dispersion curves corresponding to quasi momenta lying at the BZ
861: boundary $q_{1}=Q_1/2,$ $0\leq{q}_{2}\leq{Q}_{2}/2$ ($FC$ line in
862: Fig.\ref{BZ3}) and $q_{2}=Q_2/2,$ $0\leq{q}_{1}\leq{Q}_{1}/2$ ($CF'$
863: line in Fig.\ref{BZ3}), are displayed in Fig.\ref{sp-bound}.  The
864: lowest and the highest curves in the $FE$ part of the latter figure,
865: describe two waves propagating along the second array.  They are
866: nearly linear, and deviations from linearity are observed only near
867: the point $E$ where the interaction has a resonant character.  Two
868: modes propagating along the first array, in zero approximation, are
869: degenerate with an unperturbed frequency $\omega=0.7.$ The interaction
870: lifts the degeneracy.  The lowest of the two middle curves corresponds to
871: $(1,u)$ boson, and the upper of one describes $(1,g)$ boson.  Note that
872: $(1,g)$ boson conserves its unperturbed frequency
873: $\omega=0.7$.  The latter fact is related to the symmetry
874: $\zeta_j(\xi)=\zeta_j(-\xi)$ of the separable interaction (\ref{charge}). 
875: At the point $E$, the two modes propagating along the first array and
876: the mode propagating along the second array in the second band are
877: degenerate.  Interactions lifts the degeneracy, and, as a result,
878: the $(1,u)$ and $(2,2)$ waves are strongly mixed and the eigenmodes are
879: their even (highest frequency) and odd (lowest frequency)
880: combinations, and the $(1,g)$ mode (middle level).\\
881: %%%%%%%%%%%%%%%%%%%%%%%%
882: \begin{figure}[htb]
883: \centering
884: \includegraphics[width=70mm,height=60mm,angle=0,]{sp-res_a.eps}
885: \epsfxsize=80mm 
886: \caption{The energy spectrum of a tilted QCB (solid lines) and
887: noninteracting arrays (dashed lines) for quasimomenta on the resonant
888: line of the BZ (line $ODE$ in Fig.\ref{BZ3})}
889: \label{sp-res}
890: \end{figure}
891: %%%%%%%%%%%%%%%%%%%%%%%%%%
892: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
893: \begin{figure}[htb]
894: \centering
895: \includegraphics[width=70mm,height=60mm,angle=0]{sp-bound_a.eps}
896: \epsfxsize=80mm 
897: \caption{Energy spectrum of a tilted QCB (solid lines)
898: and noninteracting arrays (dashed lines) for quasimomenta at the
899:  BZ boundary (line $FCF'$ in the Fig.\ref{BZ3})}
900: \label{sp-bound}
901: \end{figure}
902: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
903: 
904: There are two separate degeneracies within each array at the corner 
905: $C$ of a titled QCB BZ. Both of them are related to interband 
906: mixing conserving array index.  The spectral behavior along the $CF'$ 
907: boundary of the BZ is similar to that considered above but in the vicinity 
908: of the point $D$ of three-fold degeneracy.  Here, two modes 
909: propagating along the second array in the ßseparable potential 
910: approximation (\ref{separ}) remain degenerate.  This degeneracy is 
911: lifted only if deviation from separability is accounted for.\\
912: %%%%%%%%%%%%%%%%%%%%%%%%%
913: \begin{figure}[htb]
914: \centering
915: \includegraphics[width=70mm,height=65mm,angle=0,]{sp-diag_a.eps}
916: \epsfxsize=80mm 
917: \caption{The energy spectrum of a titled QCB (solid lines) and
918: noninteracting arrays (dashed lines) for quasimomenta on the BZ diagonal
919: (line $OC$ in Fig.\ref{BZ3})}
920: \label{sp-diag}
921: \end{figure}
922: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
923: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
924: \begin{figure}[htb]
925: \centering
926: \includegraphics[width=85mm,height=75mm,angle=0,]{IsoEn31.eps}
927: \epsfxsize=80mm 
928: \caption{Lines of equal frequency for a tilted
929: QCB (solid lines) and noninteracting arrays (dashed lines). Lines
930: $1,2,3$ correspond to frequencies $\omega_{1}=0.1$,
931: $\omega_{2}=0.25$, $\omega_{3}=0.45$.}
932: \label{IsoEn1}
933: \end{figure}
934: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
935: 
936: The diagonal $OC$ of a tilted QCB BZ represents a new type of
937: generic line, that crosses a resonant line (Fig.\ref{sp-diag}).  Here
938: the spectrum mostly conserves its initial systematics, i.e. belongs to
939: a given array, and mostly depends on a given quasimomentum component. 
940: However, at the crossing point $B$, the modes $(1,1)$ and $(2,2),$
941: corresponding to both different arrays and bands, become degenerate
942: (two middle dashed lines in Fig.\ref{sp-diag}).  Interaction between
943: the wires lifts the degeneracy.  The eigenstates of QCB have a definite
944: parity with respect to transposition of these two modes.  The lowest
945: and upper of two middle lines corresponds to even ($g$) and odd ($u$)
946: mode, respectively.\\
947: 
948: Like in square QCB, bosons with quasimomenta close to the resonant 
949: lines are strongly mixed bare $1D$ bosons.  These excitations are 
950: essentially two-dimensional, and therefore lines of equal energy 
951: in the vicinity of the resonant lines are modified by the $2D$ 
952: interaction (see Figs.\ref{IsoEn1},\ref{IsoEn2}).  Deviations from 
953: $1D$ behaviour occur only in this small part of the BZ. For $\omega < 
954: 0.5 v_2Q_2$ the lines of equal energy within BZ consist of closed line 
955: around the BZ center and four open lines (within the extendend bands 
956: scheme these lines are certainly closed) around the BZ corners (lines 
957: 1, 2, 3 in Fig.\ref{IsoEn1}). At the line $OD$ in BZ, the modes of 
958: QCB are strongly coupled bare bosons propagating along both arrays in 
959: the first band.\\
960: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
961: \begin{figure}[htb]
962: \centering
963: \includegraphics[width=85mm,height=75mm,angle=0,]{IsoEn32.eps}
964: \epsfxsize=80mm \caption{Lines of equal frequency for a tilted
965: QCB (solid lines) and noninteracting arrays (dashed lines). Lines 
966: $4,5$ in the lower panel correspond to frequencies
967: $\omega_{4}=0.55$, $\omega_{5}=0.65$}
968: \label{IsoEn2}
969: \end{figure}
970: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
971: 
972: For $0.5 v_2Q_2
973: <\omega < 0.5 v_1Q_1$ (lines 4, 5 in Fig.\ref{IsoEn2}) the topology of
974: lines of equal energy is modified.  In this case lines of
975: equal energy within the BZ consist of four open lines.  The
976: splitting of lines at the direction $DE$ corresponds to strong
977: coupling of modes propagating along the first array in the first band
978: with those propagating along the second array in the second band.\\ 
979: 
980: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
981: \subsection{Correlations and Observables}\label{subsec:Correl}
982: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
983: The structure of the energy spectrum analyzed above predetermines
984: optical and transport properties of QCB. We consider here three 
985: types of correlation functions manifesting dimensional crossover in 
986: QCB.\\
987: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
988: \subsubsection{Optical Absorption}\label{subsubsec:OptAbs}
989: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
990: We start with 
991: {\it ac} conductivity 
992: \begin{equation}
993:  {\sigma}_{jj'}({\bf{q}},\omega)=  
994: {\sigma}'_{jj'}({\bf{q}},\omega)+i{\sigma}''_{jj'}({\bf{q}},\omega).
995: \nonumber
996: %\label{accond}
997: \end{equation}
998: The real part ${\sigma}'_{jj'}({\bf{q}},\omega)$ determines an
999: optical absorption.  The spectral properties of \emph{ac} conductivity
1000: are given by a current--current correlator
1001: \begin{equation}
1002:       {\sigma}_{jj'}({\bf{q}},\omega)=\frac{1}{\omega}
1003:       \int\limits_{0}^{\infty}dt{e}^{i{\omega}t}
1004:       \left\langle \left[
1005:           {J}_{j1{\bf{q}}}(t),{J}_{j'1{\bf{q}}}^{\dag}(0)
1006:       \right] \right\rangle.
1007: 	\label{CurrCorr}
1008: \end{equation}
1009: Here ${J}_{js{\bf{q}}}=\sqrt{2}vg{\pi}_{js{\bf{q}}}$ is a current
1010: operator for the $j$-th array (we restrict ourselves
1011: to the first band, for the sake of simplicity).\\
1012: 
1013: The current-current correlator for non-interacting wires is reduced to 
1014: the conventional LL expression \cite{Voit},
1015: \begin{eqnarray*}
1016: \left\langle \left[
1017:      {J}_{j1{\bf{q}}}(t),{J}_{j'1{\bf{q}}}^{\dag}(0)
1018: \right] \right\rangle_0=
1019:   -2ivg{\omega}_{j1{\bf{q}}}
1020:   \sin({\omega}_{j1{\bf{q}}}t)
1021:   {\delta}_{jj'}
1022: \end{eqnarray*}
1023: with metallic-like peak
1024: \begin{equation}
1025: {\sigma}'_{jj'}({\bf{q}},\omega>0)=
1026: {\pi}vg
1027: \delta({\omega}-{\omega}_{j1{\bf{q}}})
1028: {\delta}_{jj'}.
1029: \label{Drude_peak}
1030: \end{equation}
1031: For QCB this correlator is calculated in Appendix C. Its analysis 
1032: leads to the following results.\\
1033: 
1034: The longitudinal absorption 
1035: \begin{equation}
1036: 	\sigma'_{11}({\bf q},\omega)\propto 
1037: 	(1-\phi^{2}_{1{\bf q}})
1038: 	\delta(\omega-\tilde\omega_{1{\bf q}})+
1039: 	\phi^{2}_{1{\bf q}}
1040: 	\delta(\omega-\tilde\omega_{2{\bf q}})\nonumber
1041: %	\label{eq:LongOut}
1042: \end{equation}
1043: contains well pronounced peak on the modified first array frequency 
1044: and weak peak at the second array frequency (the parameter $\phi_{1{\bf 
1045: q}},$ defined by Eq.  (\ref{phi1})of Appendix B, is small).  The 
1046: modified frequencies $\tilde\omega_{1{\bf q}}$ and 
1047: $\tilde\omega_{2{\bf q}}$ coincide with the eigenfrequencies 
1048: $\omega_{+1{\bf q}}$ and $\omega_{-2{\bf q}}$ respectively, if 
1049: $\omega_{1{\bf q}}>\omega_{2{\bf q}}.$ In the opposite case the signs 
1050: $+,-$ should be changede to the opposite ones.\\
1051: 
1052: The transverse absorption 
1053: component contains two weak peaks
1054: \begin{equation}
1055: 	\sigma'_{12}({\bf q},\omega)\propto 
1056: 	\phi_{1{\bf q}}\left[
1057: 	\delta(\omega-\tilde\omega_{1{\bf q}})+
1058: 	\delta(\omega-\tilde\omega_{2{\bf q}})
1059: 	\right].\nonumber
1060: %	\label{eq:TransOut}
1061: \end{equation}
1062: 
1063: At the resonant line, the results change drastically. Both longitudinal 
1064: and transverse components of the optical 
1065: absorption contain two well pronounced peaks corresponding to 
1066: slightly split modified frequencies
1067: \begin{equation}
1068: 	\sigma'_{11}({\bf q},\omega)\propto\frac{1}{2} 
1069: 	\left[
1070: 	\delta(\omega-\tilde\omega_{1{\bf q}})+
1071: 	\delta(\omega-\tilde\omega_{2{\bf q}})
1072: 	\right].\nonumber
1073: %	\label{eq:LongAt}
1074: \end{equation}
1075: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1076: \subsubsection{Space Perturbation}\label{subsubsec:Space}
1077: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1078: One of the main effects specific for QCB is the appearance of
1079: non-zero transverse momentum--momentum correlation function. In
1080: space-time coordinates $({\bf{x}},t)$ its representation reads,
1081: \begin{equation}
1082: G_{12}({\bf x},t) =
1083:            i\left\langle \left[
1084:                              {\pi}_{1}(x_1,0;t),
1085:                              {\pi}_{2}(0,x_2;0)
1086:            \right] \right\rangle.\nonumber
1087: \end{equation}
1088: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1089: \begin{figure}[htb]
1090: \centering
1091: \includegraphics[width=75mm,height=75mm,angle=0,]{GF12.eps}
1092: \epsfxsize=80mm
1093: \caption{The transverse correlation function $G_{12}(x_1,x_2;t)$
1094: for $r_0=1$ and $vt=10$}
1095: \label{GF12}
1096: \end{figure}
1097: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1098: This function describes the momentum response at the point $(0,x_{2})$
1099: of the second array for time $t$ caused by an initial ($t=0$)
1100: perturbation localized in coordinate space at the point $(x_{1},0)$
1101: of the first array.  Standard calculations similar to those described
1102: above, lead to the following expression,
1103: \begin{eqnarray*}
1104: & {G}_{12}({\bf{x}};t) =
1105:                      \displaystyle{
1106:                      \frac{V_0r_0^2}{4{\pi}^2{\hbar}}
1107:                      \int\limits_{-\infty}^{\infty} dk_1 dk_2
1108:                      }
1109:                      {\phi}_1({k_1}){\phi}_2({k_2})k_1k_2
1110:                       &
1111:                      \\
1112:                      & \times
1113:                      \sin(k_1x_1)
1114:                      \sin(k_2x_2)
1115:                      \displaystyle{
1116:                      \frac{v_2k_2\sin(v_2k_2t)-
1117:                            v_1k_1\sin(v_1k_1t)}
1118:                           {v_2^2k_2^2-v_1^2k_1^2},
1119:                      } &
1120: %              \label{spat_trans}
1121: \end{eqnarray*}
1122: where ${\phi}_{j}(k)$ is the form-factor (\ref{FC-2}) written in the
1123: extended BZ. This correlator is shown in Fig.\ref{GF12}.  It is mostly
1124: localized at the line determined by the obvious kinematic condition
1125: $$\frac{|x_{1}|}{v_{1}}+\frac{|x_{2}|}{v_{2}}=t.$$
1126: The time $t$ in the r.h.s.  is thea total time of plasmon propagation 
1127: from the starting point $(x_{1},0)$ to the final point $(0,x_{2})$ or 
1128: vice versa, along any of the shortest ways compatible with a 
1129: restricted geometry of the $2D$ grid. The finiteness of the interaction 
1130: radius slightly spreads this peak and modifies its profile.\\
1131: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1132: \subsubsection{Rabi Oscillations}\label{subsubsec:Rabi}
1133: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1134: Further manifestation of the 2D character of QCB system is related to 
1135: the possibility of periodic energy transfer between the two arrays.  
1136: Consider an initial perturbation which excites a plane wave with 
1137: amplitude $\theta_{0}$ within the 
1138: first array in the system of {\it non}-interacting arrays,
1139: \begin{eqnarray*}
1140:   {\theta}_{1}(x_1,n_2a_2;t)
1141:   & = &
1142:   \theta_{0}
1143:   \sin(q_1x_1+q_2n_2a_2-v_1|q_1|t).
1144: \end{eqnarray*}
1145: If the wave vector
1146: ${\bf q},$ satisfying the condition $|{\bf{q}}|<<Q_{1,2}/2,$ is
1147: not close to the resonant line of the BZ, weak interarray
1148: interaction $\phi=\varepsilon{r}_{0}/{a}$ slightly changes the
1149: $\theta_{1}$ component and leads to the appearance
1150: of a small $\theta_{2}\sim\phi$ component.  But
1151: for ${\bf q}$ lying on the resonant line
1152: ($v_1|q_1|=v_2|q_2|\equiv\omega_{\bf{q}}$), both components within
1153: the main approximation have the same order of magnitude,
1154: \begin{eqnarray*}
1155:    {\theta}_{1}(x_1,n_2a_{2};t) & = &
1156:    {\theta}_{0}
1157:    \cos\left(
1158:                  \frac{1}{2}
1159:                  {\phi}_{1{\bf{q}}}
1160:                  {\omega}_{{\bf{q}}}t
1161:             \right)\\
1162:    & \times &\sin(q_1x_1+q_2n_2a_2-{\omega}_{{\bf{q}}}t),
1163: %   \label{theta1d}
1164: \end{eqnarray*}
1165: \begin{eqnarray*}
1166:    {\theta}_{2}(n_1a_1,x_2;t) & = &
1167:    {\theta}_{0}
1168:    \sin\left(
1169:             \frac{1}{2}
1170:             {\phi}_{1{\bf{q}}}
1171:             {\omega}_{{\bf{q}}}t
1172:    \right)\\
1173:    & \times &
1174:    \cos(q_1n_1a_1+q_2x_2-{\omega}_{{\bf{q}}}t).
1175: %   \label{theta2d}
1176: \end{eqnarray*}
1177: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1178: \begin{figure}[htb]
1179: \centering
1180: \includegraphics[width=70mm,height=60mm,angle=0,]{RO3.eps}
1181: \epsfxsize=80mm 
1182: \caption{Periodic energy exchange between arrays (Rabi oscillations)}
1183: \label{RO3}
1184: \end{figure}
1185: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1186: This corresponds to 2D propagation of a plane wave with wave vector 
1187: ${\bf q },$ {\it modulated} by a ``slow'' frequency $\sim\phi\omega.$ 
1188: As a result, beating arises due to periodic energy transfer from one 
1189: array to another during a long period $T\sim (\phi\omega)^{-1}$ (see 
1190: Fig.\ref{RO3}).  These peculiar ``Rabi oscillations'' may be 
1191: considered as one of the fingerprints of the physics exposed in QCB 
1192: systems.\\
1193: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1194: \section{Triple QCB}\label{sec:Triple}
1195: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1196: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1197: \subsection{Notions and Hamiltonian}\label{subsec:NotHam}
1198: Triple quantum bars is $2D$ periodic grid with $m=3$, formed by three 
1199: periodically crossed arrays $j=1,2,3$ of $1D$ quantum wires.  In fact 
1200: these arrays are placed on three planes parallel to $XY$ plane and 
1201: separated by an inter-plane distances $d.$ The upper and the lower 
1202: arrays correspond to $j=1,2,$ while the middle array has number $j=3.$ 
1203: All wires in all arrays are identical.  They have the same length $L,$ 
1204: Fermi velocity $v$ and Luttinger parameter $g.$ The arrays are 
1205: oriented along the $2D$ unit vectors
1206: \begin{equation}
1207: 	{\bf{e}}_{1}=\left(\frac{1}{2}, \frac{\sqrt{3}}{2}\right),
1208: 	\ \ \ \ {\bf{e}}_2=(1,0),\ \ \ \ 
1209: 	{\bf{e}}_{3}={\bf{e}}_2-{\bf{e}}_{1}.\nonumber
1210: %	\label{basis}
1211: \end{equation}
1212: The periods of QCB along these directions are equal, $a_{j}=a,$ so we 
1213: deal with a regular triangular lattice.  In what follows we choose 
1214: ${\bf a}_{1,2}=a{\bf e}_{1,2}$ as the basic vectors of a superlattice 
1215: (see Fig.\ref{Bar4}).\\
1216: 
1217: The wires within the $j$-th array are enumerated with the integers
1218: $n_j.$ Define $2D$ coordinates along the $n_{j}$-th wire ${\bf r}_{j}$
1219: as ${\bf r}_{j}=x_{j}{\bf{e}}_{j}+n_{j}a{\bf{e}}_{3}$ for upper and
1220: lower arrays ($j=1,2$) and ${\bf
1221: r}_{3}=x_{3}{\bf{e}}_{3}+n_{3}a{\bf{e}}_{1}$ for the middle array. 
1222: Here $x_{j}$ are $1D$ continuous coorinates along the wire. The 
1223: system of three non-interacting arrays is described by the Hamiltonian
1224: \begin{equation}
1225: 	H_{0}=H_{1}+H_{2}+H_{3},\nonumber
1226: %	\label{H_0}
1227: \end{equation}
1228: where
1229: \begin{eqnarray}
1230:  H_1 & = & \frac{{\hbar}v}{2}\sum\limits_{n_1}\int dx_1
1231:  \left[
1232:       g{\pi}_1^2(x_1{\bf{e}}_1+n_1a{\bf{e}}_3)\right.\nonumber\\
1233:       &&+\left.\frac{1}{g}
1234:       \left(
1235:            \partial_{x_1}{\theta}_1(x_1{\bf{e}}_1+n_1a{\bf{e}}_3)
1236:       \right)^2
1237:  \right],
1238:  \label{H1}
1239:  \\
1240:  H_2 & = & \frac{{\hbar}v}{2}\sum\limits_{n_2}\int dx_2
1241:  \left[
1242:       g{\pi}_2^2(x_2{\bf{e}}_2+n_2a{\bf{e}}_3)\right.\nonumber\\
1243:       &&+\left.\frac{1}{g}
1244:       \left(
1245:            \partial_{x_2}{\theta}_2(x_2{\bf{e}}_2+n_2a{\bf{e}}_3)
1246:       \right)^2
1247:  \right],
1248:  \label{H2}
1249:  \\
1250:  H_3 & = & \frac{{\hbar}v}{2}\sum\limits_{n_3}\int dx_3
1251:  \left[
1252:       g{\pi}_3^2(x_3{\bf{e}}_3+n_3a{\bf{e}}_1)\right.\nonumber\\
1253:       &&+\left.\frac{1}{g}
1254:       \left(
1255:            \partial_{x_3}{\theta}_3(x_3{\bf{e}}_3+n_3a{\bf{e}}_1)
1256:       \right)^2
1257:  \right],
1258:  \label{H3}
1259: \end{eqnarray}
1260: and $\pi_{j}$ and $\partial_{x_{j}}\theta_{j}$ are canonically
1261: conjugated fields describing LL within the $j$-th array.\\
1262: %%%%%%%%%%%%%%%%%%%%%
1263: \begin{figure}[htb]
1264: \centering
1265: \includegraphics[width=75mm,height=68mm,angle=0,]{Bar4.eps}
1266: \epsfxsize=70mm 
1267: \caption{Triple QCB} 
1268: \label{Bar4}
1269: \end{figure}
1270: %%%%%%%%%%%%%%%%%%%
1271: Interaction between the excitations in different wires of adjacent
1272: arrays $j,j'$ is concentrated near the crossing points with
1273: coordinates $n_j{\bf a}_j+n_{j'}{\bf a}_{j'}$.  It is actually Coulomb
1274: interaction screened on a distance $r_0$ along each wire which is 
1275: described by Hamiltonian
1276: \begin{equation}
1277: 	H_{int}=H_{13}+H_{23},\nonumber
1278: %	\label{Int}
1279: \end{equation}
1280: where
1281: \begin{eqnarray}
1282:  &\displaystyle{\frac{H_{13}}{V_0}} =  \sum\limits_{n_1,n_3}\int dx_1dx_3
1283:  \Phi\left(
1284:           \frac{x_1-n_3a}{r_0}{\bf{e}}_1-
1285:           \frac{x_3-n_1a}{r_0}{\bf{e}}_3
1286:  \right)
1287:  \nonumber\\  
1288:  &\times
1289:  \partial_{x_1}{\theta}_1(x_1{\bf{e}}_1+n_1a{\bf{e}}_3)
1290:  \partial_{x_3}{\theta}_3(n_3a{\bf{e}}_1+x_3{\bf{e}}_3),
1291:  \label{H13}
1292: \end{eqnarray}
1293: \begin{eqnarray}
1294:  &\displaystyle{\frac{H_{23}}{V_0}}  = \sum\limits_{n_2,n_3}\int dx_2dx_3
1295: \Phi\left(
1296:           \frac{x_2-n_3a}{r_0}{\bf{e}}_2-
1297:           \frac{x_3-n_2a}{r_0}{\bf{e}}_3
1298:  \right)
1299:  \nonumber\\  
1300:  &\times
1301:  \partial_{x_2}{\theta}_2(x_2{\bf{e}}_2+n_2a{\bf{e}}_3)
1302:  \partial_{x_3}{\theta}_3(n_3a{\bf{e}}_2+x_3{\bf{e}}_3).
1303:  \label{H23}
1304: \end{eqnarray}
1305: Here the effective coupling strength $V_{0}$ is defined by 
1306: Eq.(\ref{strength}), the  
1307: dimensionless interaction $\Phi$ is separable
1308: \begin{equation}
1309:  \Phi(\xi_{j}{\bf{e}}_{j}+\xi_{3}{\bf{e}}_{3})=
1310:  \zeta(\xi_{j})\zeta(\xi_{3}),
1311:  \ \ \ \ \ j=1,2,
1312:  \label{separab}
1313: \end{equation}
1314: and $\zeta(\xi)$ is dimensionless charge fluctuation in the $j$-th 
1315: wire (see Eq. (\ref{charge})).\\
1316: 
1317: Such interaction imposes a superperiodicity on the energy spectrum of
1318: initially one dimensional quantum wires, and the eigenstates of this
1319: superlattice are characterized by a $2D$ quasimomentum ${\bf
1320: q}=q_1{\bf g}_1+q_2{\bf g}_2 \equiv(q_1,q_2)$.  Here ${\bf g}_{1,2}$
1321: are the unit vectors of the reciprocal superlattice satisfying the
1322: standard orthogonality relations $({\bf e}_i\cdot {\bf
1323: g}_j)=\delta_{ij}, \ \ j=1,2.$ The corresponding basic vectors of the
1324: reciprocal superlattice have the form $Q(m_1{\bf g}_1 + m_2{\bf g}_2
1325: $, where $Q=2\pi/a$ and $m_{1,2}$ are integers. In Fig.\ref{BZ4} 
1326: elementary cell $BIJL$ of the reciprocal lattice is displayed together 
1327: with the hexagon of the Wigner-Seitz cell that we choose as the  
1328: BZ of the triple QCB.\\
1329: %%%%%%%%%%%%%%%%%%%%%
1330: \begin{figure}[htb]
1331: \centering
1332: \includegraphics[width=75mm,height=80mm,angle=0,]{BZ4.eps}
1333: \epsfxsize=70mm \caption{Elementary cell $BIJL$ of the reciprocal
1334: lattice and the BZ hexagon of the triple QCB}
1335: \label{BZ4}
1336: \end{figure}
1337: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1338: 
1339: To study the energy spectrum and the eigenstates of the total Hamiltonian
1340: \begin{equation}
1341:  	H=H_{0}+H_{int},
1342:  	\label{H}
1343:  \end{equation} 
1344: we define the Fourier components of the field operators 
1345: %Eq.(7-9)
1346: \begin{eqnarray}
1347:  &{\theta}_1(x_1{\bf{e}}_1+n_1a{\bf{e}}_3) =\nonumber\\
1348:  &(NL)^{-1/2}\sum\limits_{{s},{\bf{q}}}
1349:  \theta_{1s{\bf{q}}}e^{i(q_1x_1+q_3n_1a)}
1350:  u_{s,q_1}(x_1),
1351:  \label{Fuorier1}
1352:  \\
1353:  &{\theta}_2(x_2{\bf{e}}_2+n_2a{\bf{e}}_3) =\nonumber\\
1354:  &(NL)^{-1/2}\sum\limits_{{s},{\bf{q}}}
1355:  \theta_{2s{\bf{q}}}e^{i(q_2x_2+q_3n_2a)}
1356:  u_{s,q_2}(x_2),
1357:  \label{Fuorier2}
1358:  \\
1359:  &{\theta}_3(x_3{\bf{e}}_3+n_3a{\bf{e}}_1) =\nonumber\\
1360:  &(NL)^{-1/2}\sum\limits_{{s},{\bf{q}}}
1361:  \theta_{3s{\bf{q}}}e^{i(q_3x_3+q_1n_3a)}
1362:  u_{s,q_3}(x_3).
1363:  \label{Fuorier3}
1364: \end{eqnarray}
1365: Here 
1366: $${\bf q}=q_{1}{\bf e}_{1}+q_{2} {\bf e}_{2},\ \ \ 
1367: q_{3}=q_{2}-q_{1},$$
1368: and $N=L/a$ is the dimensionless length of a wire.  In the ${\bf q}$ 
1369: representation, the Hamiltonians $H_{j}$ (Eqs. (\ref{H1})-(\ref{H3}))  
1370: and $H_{j3}$ (Eqs. (\ref{H13}), (\ref{H23})) can be written as
1371: \begin{eqnarray*}
1372:  H_j  =  \frac{{\hbar}{v}{g}}{2}\sum\limits_{{s},{\bf{q}}}
1373:            \pi_{js{\bf{q}}}^{+}\pi_{js{\bf{q}}}
1374: 		   \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 
1375: 		   \\
1376:            +\frac{\hbar}{{2}{v}{g}}\sum\limits_{{s},{\bf{q}}}
1377:            {\omega}_{s}^{2}(q_j)
1378:            \theta_{js{\bf{q}}}^{+}\theta_{js{\bf{q}}},
1379:  \ \ \ \ \ \ \ j=1,2,3,
1380:  \\
1381:  H_{j3}  =  \frac{V_0r_0^2}{{2}{v}{g}}\sum\limits_{s,s',{\bf{q}}}
1382:  \phi_{s}(q_3)\phi_{s'}(q_j)\omega_{s}(q_3)\omega_{s'}(q_j)
1383:  \\
1384:  \times\left[
1385:       \theta_{3s{\bf{q}}}^{+}\theta_{js'{\bf{q}}}+h.c.
1386:  \right], \ \ \ \ j=1,2,
1387: \end{eqnarray*}
1388: where 
1389: $$
1390: \omega_s(q)=v\left(\left[\frac{s}{2}\right]Q+
1391: \left(-1\right)^{s-1}|q|\right).
1392:  \ \ \ \ \ Q=\frac{2{\pi}}{a}
1393: $$
1394: Thus the total Hamiltonian (\ref{H}) describes a system of coupled 
1395: harmonic oscillators, and can be diagonalized exactly like in the case 
1396: of double QCB.\\
1397: 
1398: %%%%%%%%%%%%%%%%%%%%%
1399: \subsection{Spectrum}\label{subsec:SpecTriple}
1400: %%%%%%%%%%%%%%%%%%%%%
1401: Separability of the interaction (\ref{separab}) allows one to 
1402: derive analytical equations for the spectrum of the total 
1403: Hamiltonian (\ref{H}) (see Appendix D).  Here we 
1404: describe the behavior of the spectrum and the states along some 
1405: specific lines of the reciprocal space.\\
1406: 
1407: The high symmetry of the triple QCB leads to a number of lines where
1408: interarray or interband resonant interaction occurs: \emph{all} lines
1409: in Fig. \ref{BZ4} posess some resonant properties. These lines 
1410: may be classified as follows:
1411: 
1412: On the Bragg lines where one of three array wavenumbers $q_{j}$ is a
1413: multiple integer of $Q/2,$ there is a strong intraband mixing of modes
1414: of the $j$-th array.  In Fig.\ref{BZ4}, these lines are the boundaries
1415: of the elementary cell of the reciprocal lattice $IJLB,$ axes $q_{1}$
1416: and $q_{2}$, lines $OB$ and $EH.$ In particular, along the lines
1417: $OA$ ($q_{2}=0$) and $OB$ ($q_{3}=0$) two modes corresponding to $2$-d
1418: and $3$-d bands and to the second ($OA$) or third ($OB$) array are
1419: mixed.  Along the line $AB$ ($q_{1}=Q/2$) the same mixing happens
1420: between $(1,1)$ and $(1,2)$ modes.  Moreover, the resonant mixing of
1421: different arrays within the same band occurs along the medians $OA,$
1422: $OB,$ etc..  There are two types of such a resonance.  The first of
1423: one (e.g. $OA$ line) is the resonance between neighboring arrays
1424: ($q_{1}=-q_{3}$) and therefore it is of the main order with respect to
1425: interaction.  The second one (e.g. $OB$ line) is the resonance between
1426: remote arrays ($q_{1}=q_{2}$) and is one order smaller.\\
1427: 
1428: The second family consists of resonant lines formed by the BZ hexagon
1429: boundaries and diagonals.  Thus, the diagonal $OC$ realizes a first
1430: order resonance between the first and the third arrays $q_{1}=q_{3},$
1431: and the BZ boundaries $HD$ and $AN$ correspond to the same resonance
1432: up to an umklapp process ($q_{1}=q_{3}-Q$ and $q_{1}=q_{3}+Q$
1433: respectively).  Along the diagonal $OD$ and the BZ boundary $NC$ a
1434: second order resonance takes place with resonance conditions
1435: $q_{2}=-q_{1}$ and $q_{2}=-q_{1}+Q$ respectively.\\
1436:  
1437: In the reciprocal space of the triple QCB there are four different
1438: types of crossing points.  Two of them include the bases of BZ medians
1439: (e.g. points $A,$ $B,$ $E$ and so on).  Here one deals with the
1440: four-fold degeneracy of the modes corresponding to the first order
1441: resonance between the neighboring arrays (e.g. point $A,$
1442: $\omega_{1,s}=\omega_{3,s'},$ $s,s'=1,2$), or to the second order
1443: resonance between remote arrays (like point $B,$
1444: $\omega_{1,s}=\omega_{2,s'},$ $s,s'=1,2$).  One more family consists
1445: of crossing points of the BZ diagonals and the lines connecting the
1446: bases of its medians (points $M,$ $F,$ $G$ and so on).  Here one deals
1447: with three types of two-fold degeneracy simultaneously.  For
1448: example, at the point $M$ two separate pairs of modes corresponding to
1449: neighboring arrays $(2,1),$ $(3,1),$ and $(2,2),$ $(3,2),$ are
1450: degenerate, as well as two modes corresponding to the first array,
1451: $(1,1),$ $(1,2).$ Finally the BZ hexagon vertices form the most
1452: interesting group of points where the three-fold degeneracy between
1453: modes corresponding to all three arrays takes place.  The typical
1454: example of such a point is the vertex $C$ where the resonance condition
1455: $q_{1}=-q_{2}+Q=q_{3}=Q/3$ is satisfied.\\
1456: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1457: \begin{figure}[htb]
1458: \centering
1459: \includegraphics[width=75mm,height=60mm,angle=0,]{TriangSp1.eps}
1460: \epsfxsize=70mm 
1461: \caption{Dispersion curves at the $OAMBO$ polygon of BZ} 
1462: \label{Tri1}
1463: \end{figure}
1464: %%%%%%%%%%%%%%%%%%%
1465: 
1466: Almost all these peculiarities of the triple QCB spectrum can be
1467: illustrated in Fig.\ref{Tri1} where the dispersion curves along the
1468: closed line $OABO$ are displayed.  We emphasize once more that in the
1469: infrared limit $\omega,{\bf q}\to 0$ triple QCB like double QCB
1470: preserves the characteristic LL properties of the initial arrays.\\
1471: %%%%%%%%%%%%%%%%%%%
1472: \subsection{Observables}\label{subsec:Observ}
1473: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1474: The structure of the energy spectrum analyzed above strongly 
1475: influences optical and transport properties of the triple QCB. As in 
1476: the case of double QCB (subsection \ref{subsec:Correl}), one expects to 
1477: observe four peaks of the optical absorption near the points $A,B,E,H$ 
1478: of the four-fold degeneracy.  Then, specific features of space correlators 
1479: like those considered in \ref{subsubsec:Space} can be observed.  But the 
1480: most pronounced manifestation of a triangular symmetry of the triple 
1481: QCB are its Rabi oscillations.\\
1482: 
1483: Consider the vicinity of the point $C$ of three-fold degeneracy mixing 
1484: all three arrays.  Appropriate initial conditions lead (see Appendix E 
1485: for details) to the following time dependence of the field operators 
1486: in the coordinate origin in real space
1487: \begin{eqnarray}
1488:   \theta_1(0,0;t) &=& \theta_0
1489:   \sin(\omega_0t)
1490:   \cos^2\left(\frac{\sqrt{2\varepsilon}\phi^2}{4}\omega_0t\right),
1491:   \nonumber\\
1492:   \theta_2(0,0;t) &=&\theta_0
1493:   \cos(\omega_0t)
1494:   \sin^2\left(\frac{\sqrt{2\varepsilon}\phi^2}{4}\omega_0t\right),
1495:   \nonumber\\
1496:   \theta_3(0,0;t) &=& \theta_0
1497:   \sin(\omega_0t)
1498:   \cos\left(\frac{\sqrt{2\varepsilon}\phi^2}{2}\omega_0t\right).
1499:   \label{sol3}
1500: \end{eqnarray}
1501: The field operators of all three arrays demonstrate fast oscillations
1502: with the resonant frequency $\omega_{0}$ modulated by a slow
1503: frequency.  It is the same for the two remote arrays, and doubled for
1504: the intermediate array.  These beatings are synchronized in a sense
1505: that zero intensity on the intermediate array always coincides with
1506: the same intensity on one of the remote arrays.  At these moments all
1507: the energy is concentrated solely within one of the remote arrays. 
1508: These peculiar Rabi oscillations are displayed in Fig.\ref{ROTr}.\\
1509: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1510: \begin{figure}[htb]
1511: \centering
1512: \includegraphics[width=75mm,height=105mm,angle=0,]{RO-Tr.eps}
1513: \epsfxsize=70mm 
1514: \caption{Periodic energy transfer between three arrays at the triple 
1515: resonant point $C$ of the BZ} 
1516: \label{ROTr}
1517: \end{figure}
1518: %%%%%%%%%%%%%%%%%%%
1519: 
1520: \section{Conclusion}\label{sec:Concl}
1521: We discussed in this paper the kinematics and dynamics of plasmon
1522: spectrum in QCB. These nanostructures may be fabricated from
1523: single-wall carbon nanotubes \cite{Rueckes,Diehl}.  On the one hand,
1524: QCB is promised to become an important component of future molecular
1525: electronics \cite{Rueckes,Tseng}.  On the other hand, the spectrum of
1526: elementary excitations (plasmons) in these grids possesses the
1527: features of both 1D and 2D electron liquid.  As is shown in Refs. 
1528: \onlinecite{Luba00,Luba01} and confirmed in the present study, the energy
1529: spectrum of QCB preserves the characteristic properties of LL at
1530: $|{\bf q}|,\omega\to 0,$.  At finite ${\bf q},\omega$ the density and
1531: momentum waves in double and triple QCB may have either $1D$ or $2D$
1532: character depending on the direction of the wave vector.  Due to
1533: interarray interaction, unperturbed states, propagating along arrays
1534: are always mixed, and transverse components of correlation functions
1535: do not vanish.  For quasi-momentum lying on the resonant lines of the BZ,
1536: such mixing is strong and transverse correlators have the same order
1537: of magnitude as the longitudinal ones.  Periodic energy transfer
1538: betweem arrays (``Rabi oscillations'') is predicted.\\
1539: 
1540: The crossover from 1D to 2D regime may be experimentally observed. 
1541: One of the experimental manifestations, i.e. the crossover from
1542: isotropic to anisotropic (spatially nonuniform) conductivity was
1543: pointed out in Ref.  \onlinecite {Luba01}.  The current may be
1544: inserted in QCB at a point on an array $j$ and extracted from another
1545: array $i$ at a distance $r$.  Then a temperature dependent length
1546: scale $l(T)$ arises, so that for $r\gg l$ the resistance is dominated
1547: by small $q$ and therefore, the current is isotropic.  In the opposite
1548: limit $r<l$ the dependence of the current on the points of
1549: injection/extraction may be detected.  At $T=0$ the length $l$ becomes
1550: infinite, and current can only be carried along the wires.  These
1551: effects are in fact manifestations of the LL behavior of the QCB in
1552: the infrared limit.\\
1553: 
1554: To observe the crossover at finite $\{\omega, {\bf q}\}$, one should
1555: find a way of exciting the corresponding plasmon modes.  Then,
1556: scanning the $\omega(q_1,q_2)$ surfaces, one may in principle detect
1557: the crossover from quasi 1D to 2D behavior in accordance with the
1558: properties of the energy spectra presented in Sections II and III.
1559: Plasmons in QCB may be excited either by means of microwave resonators
1560: or by means of interaction with surface plasmons.  In the latter case one
1561: should prepare the grid on a corresponding semiconductor substrate and
1562: measure, e.g., the plasmon loss spectra.  The theory of these plasmon
1563: losses will be presented in a forthcoming publication.  \\
1564: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1565: \section*{Acknowledgements}
1566: This research is supported in part by grants from the Israel
1567: Science foundations, the DIP German-Israel cooperation program, and the
1568: USA-Israel BSF program.\\
1569: %%%%%%%%%%%%%%%%%%%%%
1570: \section*{Appendix A. Empty Superchain}\label{sec:Empty}
1571: Here we construct eigenfunctions, spectrum, and quasi-particle operators 
1572: for an ``empty superchain'' - quantum wire in an infinitely weak 
1573: periodic potential with period $a$. Excitations in an initial wire are
1574: described as plane waves $L^{-1/2}\exp (ikx)$ with wave number 
1575: $k=2\pi n/L,$ with integer $n,$ and
1576: dispersion law $\omega(k)=v|k|$ (the array number is
1577: temporarily omitted). The following orthogonality relations are valid
1578: \begin{eqnarray*}    
1579:     \int_{-L/2}^{L/2}\psi_{k}^{*}(x)\psi_{k'}(x)dx & = & \delta_{k,k'},
1580:     \\
1581: 	\sum_{k}\psi_{k}^{*}(x)\psi_{k}(x') & = & \delta_{L}(x-x'),
1582: %	\label{Orthog}
1583: \end{eqnarray*}
1584: where $\delta_{L}$ stands for  periodic delta-function
1585: $$\delta_{L}(x-x') \equiv \sum_{n}\delta (x-x'-nL).$$ 
1586: 
1587: ``Empty superchain'' is characterized by a space period $a$ and 
1588: corresponding reciprocal lattice wave number $Q=2\pi/a.$ Each 
1589: excitation in such a superchain is described by its quasi-wavenumber 
1590: $q$ and a band number $s$ ($s=1,2,\ldots$) that are related to the 
1591: corresponding wave number $k$ by the following relation,
1592: \begin{equation}
1593: 	k=q+iQx(-1)^{s-1}\left[\frac{s}{2}\right]
1594:     {\mbox{ sign }}q.\nonumber
1595: %	\label{ext}
1596: \end{equation}
1597: The corresponding wave 
1598: function $\psi_{s,q}(x)$ has the Bloch-type structure,
1599: \begin{equation}
1600:     \psi_{s,q}(x)=\frac{1}{\sqrt{L}}
1601:     e^{iqx}
1602:     u_{s,q}(x),
1603:     \label{WaveFunc}
1604: \end{equation}
1605: and satisfies the orthogonality relations
1606: \begin{eqnarray*} 
1607: 	\int_{-L/2}^{L/2}\psi_{s,q}^{*}(x)\psi_{s',q'}(x)dx & = &\delta_{s,s'}
1608: 	\delta_{Q;q,q'}, 
1609: 	\\ 
1610: 	\sum_{s,q}\psi_{s,q}^{*}(x)\psi_{s,q}(x') & = & \delta_{L}(x-x'),
1611: %	\label{Orthog2}
1612: \end{eqnarray*}
1613: where
1614: $$\delta_{Q;q,q'}=\sum_{n}\delta_{q+nQ,q'}.$$
1615: Within the first BZ, $-Q/2\leq q<Q/2,$ Bloch amplitude and dispersion 
1616: law $\omega_{s}$ have the following form
1617: \begin{eqnarray}
1618:     u_{s,q}(x) & = &
1619:     \exp
1620:     \left\{
1621:     iQx(-1)^{s-1}\left[\frac{s}{2}\right]
1622:     {\mbox{ sign }}q
1623:     \right \},
1624:     \label{WaveFunc1}\\
1625:     \omega_{s}(q) & = &  vQ\left(
1626:     \left[\frac{s}{2}\right]
1627:     +(-1)^{s-1}\frac{|q|}{Q}
1628:     \right).
1629:     \label{Disp1}
1630: \end{eqnarray}
1631: Here square brackets denote an integral part of a number. 
1632: Taking into account that both Bloch amplitude $u_{s,q}(x)$ and 
1633: dispersion law $\omega_{s}(q)$ are periodic functions of $q$ with 
1634: period $Q,$ one obtaines general equations for the Bloch amplitude
1635: \begin{eqnarray*}
1636: 	    u_{s,q}(x) & = &
1637:     \displaystyle{
1638:     \sum_{n=-\infty}^{\infty}
1639:     \frac{\sin\xi_{n}}{\xi_{n}}}\\
1640:     && \times\cos[(2s-1)\xi_{n}]
1641:     \exp\left(-4i\xi_{n}\frac{q}{Q}\right),\\
1642:     4\xi_{n} & = & Q(x-na).
1643: \end{eqnarray*}
1644: and dispersion law $\omega_{s}(q)$
1645: \begin{eqnarray*}
1646: 	(vQ)^{-1}\omega_{s}(q) & = &
1647:     \frac{2s-1}{4}\\
1648:     & + &
1649:     \displaystyle{\sum_{n=1}^{\infty}}
1650:     \frac{2 (-1)^{s}}{\pi^{2}(2n+1)^{2}}
1651:         \cos \frac{2\pi (2n+1)q}{Q}.
1652: %	\label{Disp2}
1653: \end{eqnarray*}
1654: The relations between quasiparticle operators for a free wire, 
1655: $c_{k},$ for momentum $k\neq nQ/2$ with $n$ integer, and those for an 
1656: empty superchain, $C_{s,q},$ for quasimomentum $q$ from the first BZ, 
1657: $-Q/2<q<Q/2,$ look as
1658: \begin{eqnarray*}
1659: 	c_{k}=C_{s,q}{\mbox{sign}}k,\\
1660: 	s=1+\left[\frac{2|k|}{Q}\right], \ \ \ 
1661: 	q=Q\left(\left\{\frac{k}{Q}
1662: 	+\frac{1}{2}\right\}-\frac{1}{2}\right)
1663: %	\label{oper1}
1664: \end{eqnarray*}
1665: \begin{eqnarray*}
1666:     C_{s,q}=(-1)^{\nu}c_{k},\\
1667: 	k=q+(-1)^{\nu}Q
1668: 	\left[\frac{s}{2}\right], \phantom{aaa} 
1669: 	\nu=s+1+\left[\frac{2q}{Q}\right],
1670: %	\label{oper2}
1671: \end{eqnarray*}
1672: where curely brackets denote a fractional part of a number. For  
1673: obtaining these relations we used the folowing expression
1674: \begin{eqnarray*}
1675: 	\int_{-L/2}^{L/2}\psi_{k}^{*}(x)\psi_{s,q}(x)dx=
1676: 	\delta_{s,s(q)}\delta_{Q;q,k}{\mbox{sign}}k,\\
1677: 	s(q)=1+\left[\frac{2|q|}{Q}\right],
1678: %	\label{TransAmpl}
1679: \end{eqnarray*}
1680: for the transition amplitude $\langle k|s,q\rangle $.
1681: In case when $k=nQ/2$ with $n$ integer, hybridization of the 
1682: neighboring bands should be taken into account. This modifies the 
1683: above relations by the following way
1684:  \begin{eqnarray*}
1685:  	c_{nQ/2} =\theta(n)\left[\alpha_{n}C_{n,q_{n}}+
1686:  	\beta_{n}C_{n+1,q_{n}}\right]\\
1687:   +\theta(-n)\left[\beta^{*}_{-n}C_{-n,q_{n}}-
1688:  	\alpha^{*}_{-n}C_{-n+1,q_{n}}\right],\\ 	
1689:  	q_{n}=Q\left(\left\{\frac{n+1}{2}\right\}-
1690: 	\frac{1}{2}\right); 
1691:  \end{eqnarray*}
1692:  \begin{eqnarray*}
1693:  	C_{s,q_{s}} =\alpha^{*}_{s}c_{sQ/2}+
1694:  	\beta_{s}c_{-sQ/2},\\
1695:     C_{s+1,q_{s}} =\beta^{*}_{s}c_{sQ/2}-
1696:  	\alpha_{s}c_{-sQ/2}, 
1697: % 	\label{oper4}
1698:  \end{eqnarray*}
1699:  where $\alpha$, $\beta$ are hybridization coefficients. 
1700:  Corresponding relations between wave functions follow immediately
1701:  from these formulas.\\
1702:   
1703: To write down any of these formulas for a specific array, one should 
1704: add the array index $j$ to the wave function $\psi$, Bloch amplitude 
1705: $u$, coordinate $x$, quasimomentum $q$, and to the periods $a$ and $Q$ 
1706: of the superchain in real and reciprocal space.\\
1707: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1708: \section*{Appendix B. Double QCB Spectrum}
1709: %%%%%%%%%%%%%%%%%%%%%%%%
1710: Here we obtain analytical expressions for dispersion laws and wave 
1711: functions of QCB. For quasimomenta far from the BZ boundaries, the 
1712: energy spectrum of the first band can be calculated explicitly.  
1713: Assuming that $\omega^2\ll\omega_{js}^2(q_j)$, $s=2,3,4,\ldots$, we 
1714: omit $\omega^{2}$ in all terms in the r.h.s.  of Eq. (\ref{F_j}) 
1715: except  
1716: the first one, $s=1$.  As a result the secular equation (17) reads
1717: \begin{equation}
1718:   \prod\limits_{j=1}^{2}
1719:   \left(
1720:        \frac{\varphi_j^2(q_j)\omega^2}
1721:             {\omega_j^2(q_j)-\omega^2}+
1722:        F_j
1723:   \right)\nonumber
1724:   =\frac{1}{\varepsilon},
1725: %  \label{sq-eq}
1726: \end{equation}
1727: where
1728: $$
1729:   \varphi_j^2(q)=\frac{r_{j}}{a_j}{\phi}_{j1}^2(q),
1730:   \ \ \ \ \
1731:   {\omega_j^2(q)}={\omega_{j1}^2(q_j)}.
1732: $$
1733: The solutions of this equation have the form:
1734: 2\begin{eqnarray}
1735:   {\omega}_{{\nu}1{\bf{q}}}^{2} & = &
1736:   \tilde\omega_1^{2}({\bf q})+
1737:         \tilde\omega_2^{2}({\bf q})
1738:   \nonumber\\ &\pm &
1739:        \sqrt{
1740:        \left(
1741:        \tilde\omega_1^{2}({\bf q})-
1742:              \tilde\omega_2^{2}({\bf q})
1743:        \right)^{2}+
1744:        4\varepsilon\varphi^2_{\bf {q}}
1745:        \omega_1^2(q_1)\omega_2^2(q_2)}.
1746:   \label{omega-12}
1747: \end{eqnarray}
1748: Here 
1749: $$
1750:   \varphi_{\bf{q}}={\varphi}_{1}(q_1)
1751:                 {\varphi}_{2}(q_2),
1752: $$
1753: $\nu=+,-$ is the branch number, $\tilde\omega_{j}({\bf q})$ is 
1754: determined as
1755: \begin{equation}
1756: 	\tilde\omega_{1}^2({\bf q})=
1757:               \omega_{1}^{2}(q_{1})
1758:               \frac{1-\varepsilon{F}_1
1759:                                  (F_2-{\varphi}_{2}^{2}(q_{2}))}
1760:                    {1-\varepsilon(F_1-{\varphi}_{1}^{2}(q_{1}))
1761:                                  (F_2-{\varphi}_{2}^{2}(q_{2}))}
1762: 	\label{RenFr}
1763: \end{equation}
1764: for $j=1.$ Expression for $\tilde\omega_{2}^2({\bf q})$ can be
1765: obtained by permutation $1\leftrightarrow 2.$ Parentheses on the
1766: r.h.s. of Eq.  (\ref{RenFr}) describe the contributions to $F_{j}$
1767: from higher bands.  Therefore $\tilde\omega_{j}^2({\bf q})$ is the
1768: $j$-th array frequency renormalized by the interaction with higher
1769: bands.  In principle, contribution of higher bands may turn the
1770: interaction to be strong.  However for specific case of carbon
1771: nanotubes, one stays far from the critical value $\varepsilon_{c}$
1772: (see estimates at the end of subsection \ref{subsec:Approx}). 
1773: Therefore the interaction with higher bands is weak almost in all the
1774: BZ except its boundaries.\\
1775: 
1776: The resonance line equation modified by interaction with higher bands 
1777: is 
1778: \begin{equation}
1779: 	\tilde \omega^{2}_{1}({\bf q})=\tilde \omega^{2}_{2}({\bf q}).
1780: 	\nonumber
1781: %	\label{ResMod}
1782: \end{equation}
1783: Out of this line the branch number is in fact the array number and 
1784: the
1785: renormalized frequencies are frequencies of a boson propagating
1786: along one of the arrays slightly modified by interactions with
1787: the complementary array. In case when 
1788: $\omega_{1}(q_{1})>\omega_{2}(q_{2}),$ one obtains
1789: \begin{equation}
1790:   {\omega}_{+,1{\bf{q}}}^{2}\approx
1791:   \omega_1^2(q_1)
1792:   \left(1-\varepsilon F_2 \varphi_1^2(q_1)\right).
1793:   \label{omega-SecOrd}
1794: \end{equation}
1795: In the opposite case one should replace indices $1\leftrightarrow 2,$
1796:  $-\leftrightarrow+$.\\
1797: 
1798: Consider the frequency correction in the latter equation in more 
1799: details.  The correction term can be approximately estimated as 
1800: ${\omega}_{1}^{2}(q_{1})S(q_{1})$ with
1801: \begin{equation}
1802:    S(q_{1}) =
1803:    \varepsilon{F}_{2}\varphi_1^{2}(q_1) =
1804:    \varepsilon
1805:    \frac{R_0}{a}
1806:    \phi_{11}^{2}(q_1)
1807:    \int d\xi\zeta_2^2(\xi).
1808:     \label{Sq}
1809: \end{equation}
1810: Due to the short-range character of the interaction, the matrix
1811: elements ${\phi}_{11}(q_1)\sim{1}$ vary slowly with the
1812: quasimomentum ${q}_{1}\le{Q}_{1}$. Therefore, the r.h.s. in
1813: Eq.(\ref{Sq}) can be roughly estimated as
1814: \begin{equation}
1815:     S(q_{1})\sim \varepsilon\frac{R_{0}}{a}=0.1
1816:     \frac{R_{0}}{a} \ll 1.
1817:     \label{est1}
1818: \end{equation}
1819: One should also remember that the energy spectrum of nanotube
1820: remains one-dimensional only for frequencies smaller than some
1821: $\omega_{m}.$ Therefore, an external cutoff arises at $s=ak_{m}$
1822: where $k_{m}\sim\omega_{m}/v$. As a results one gets an estimate
1823: \begin{equation}
1824:     S(q_{1})\sim \varepsilon \frac{R_{0}}{a}k_{m}R_{0}.
1825:     \label{est2}
1826: \end{equation}
1827: Hence, one could hope to gain additional power of the small 
1828: interaction radius. However, for nanotubes, $k_{m}$ is of the order of 
1829: $1/R_{0}$ (see Refs.\onlinecite{Ando,Egger}) and both estimates 
1830: coincide.  For quasimomenta close to the BZ center, the coefficient 
1831: $S(q_{1})$ can be calculated exactly.  For exponential form of 
1832: $\zeta(\xi)\propto\exp(-|\xi|)$, one obtains instead of the 
1833: preliminary estimate (\ref{est1}),
1834: $$
1835:      S(0) = 0.14 \frac{R_{0}}{a}.
1836: $$
1837: Thus, the correction term in Eq.(\ref{omega-SecOrd}) is really small.\\
1838: 
1839: The eigenstates of the system are described by renormalized field 
1840: operators. Within the first band they have the form
1841: \begin{eqnarray}
1842:  \tilde\theta_{11{\bf{q}}} & = &
1843:  \left(
1844:       1-\frac{1}{2}\beta_{1{\bf{q}}}
1845:  \right)
1846:  \left(
1847:       u_{\bf{q}}\theta_{11{\bf{q}}}-
1848:       v_{\bf{q}}\theta_{21{\bf{q}}}
1849:  \right)
1850:  \nonumber\\
1851:  & - &
1852:  \sum\limits_{s=2}^{\infty}
1853:  \left(
1854:       \phi_{1s{\bf{q}}}u_{\bf{q}}\theta_{2s{\bf{q}}}+
1855:       \phi_{2s{\bf{q}}}v_{\bf{q}}\theta_{1s{\bf{q}}}
1856:  \right),
1857:  \label{tilde-teta1}
1858:  \\
1859:  \tilde\theta_{21{\bf{q}}} & = &
1860:  \left(
1861:       1-\frac{1}{2}\beta_{2{\bf{q}}}
1862:  \right)
1863:  \left(
1864:       v_{\bf{q}}\theta_{11{\bf{q}}}+
1865:       u_{\bf{q}}\theta_{21{\bf{q}}}
1866:  \right)
1867:  \nonumber\\
1868:  & - &
1869:  \sum\limits_{s=2}^{\infty}
1870:  \left(
1871:       \phi_{1s{\bf{q}}}v_{\bf{q}}\theta_{2s{\bf{q}}}+
1872:       \phi_{2s{\bf{q}}}u_{\bf{q}}\theta_{1s{\bf{q}}}
1873:  \right).
1874:  \label{tilde-teta2}
1875: \end{eqnarray}
1876: Here the coefficients $u_{\bf{q}}$ and $v_{\bf{q}}$ describe mixing 
1877: between the modes with different array indices, within the first band,
1878: \begin{equation}
1879:  u_{\bf{q}}=
1880:  \sqrt{
1881:       \frac{
1882:             \sqrt{\Delta_{\bf{q}}^{2}+\phi_{1{\bf{q}}}^{2}}+
1883:             \Delta_{\bf{q}}
1884:            }
1885:            {2\sqrt{\Delta_{\bf{q}}^{2}+\phi_{1{\bf{q}}}^{2}}}
1886:       },
1887:  \label{u}
1888: \end{equation}
1889: \begin{equation}
1890:  v_{\bf{q}}=
1891:  \sqrt{
1892:       \frac{
1893:             \sqrt{\Delta_{\bf{q}}^{2}+\phi_{1{\bf{q}}}^{2}}-
1894:             \Delta_{\bf{q}}
1895:            }
1896:            {2\sqrt{\Delta_{\bf{q}}^{2}+\phi_{1{\bf{q}}}^{2}}}
1897:       },
1898:  \label{v}
1899: \end{equation}
1900: and
1901: \begin{eqnarray}
1902: \Delta_{\bf{q}}&=&
1903: \frac{\omega_{21}^{2}(q_2)-\omega_{11}^{2}(q_1)}{2},\nonumber\\
1904:  \phi_{1{\bf{q}}}&=&\sqrt{\varepsilon}
1905:  \varphi_{\bf{q}}\omega_{11}(q_1)\omega_{21}(q_2).
1906: 	\label{phi1}
1907: \end{eqnarray}
1908: The parameters $\phi_{1s{\bf{q}}}$, 
1909: $\phi_{2s{\bf{q}}},$ $s=2,3,\ldots,$ in 
1910: Eqs.  (\ref{tilde-teta1}), (\ref{tilde-teta2}) correspond to 
1911: inter-band mixing
1912: \begin{equation}
1913: 	\phi_{1s{\bf{q}}}=\sqrt{\varepsilon}
1914:  \frac{r_0}{a}\phi_{11}(q_1)\phi_{2s}(q_2)
1915:  \frac{\omega_{11}(q_1)}{\omega_{2s}(q_2)},\nonumber
1916: %	\label{Tildephi}
1917: \end{equation}
1918: and the coefficients $\beta_{1{\bf q}},$  $\beta_{2{\bf q}},$ take into 
1919: account corrections from the higher bands
1920:  \begin{equation}
1921:  	\beta_{1{\bf{q}}}=
1922:  \sum\limits_{s=2}^{\infty}
1923:      \left(
1924:           \phi_{1s{\bf{q}}}^{2}u_{\bf{q}}^{2}+
1925:           \phi_{2s{\bf{q}}}^{2}v_{\bf{q}}^{2}
1926:      \right).\nonumber
1927: % 	\label{Beta}
1928:  \end{equation}
1929: Expressions for $\phi_{2s{\bf{q}}}$ and $\beta_{2{\bf{q}}}$ can be 
1930: obtained by permutation ${1}\leftrightarrow{2}$.\\
1931: 
1932: Equations (\ref{omega-12}), (\ref{tilde-teta1}) and
1933: (\ref{tilde-teta2}) solve the problem of QCB energy spectrum away from
1934: the BZ boundaries.  However, due to smalness of the interaction, the 
1935: general expressions (\ref{tilde-teta1}) and (\ref{tilde-teta2}) can be
1936: simplified.  For quasimomenta far from the resonant coupling line, the
1937: expressions for the renormalized field operators of the first array look
1938: like
1939: \begin{equation}
1940:  {\tilde\theta}_{11{\bf{q}}} =
1941:  \left(1-\frac{1}{2}\tilde\beta_{1{\bf{q}}}\right)
1942:    {\theta}_{11{\bf{q}}}+
1943:  \sum\limits_{s=1}^{\infty}
1944:    {\phi}_{1s{\bf{q}}}
1945:    {\theta}_{2s{\bf{q}}},\nonumber
1946: %   \label{theta1off}
1947: \end{equation}
1948: where
1949: $$
1950:  \phi_{11{\bf{q}}}=\sqrt{\varepsilon}
1951:  \frac{r_0}{a}\phi_{11}(q_1)\phi_{21}(q_2)
1952:  \frac{\omega_{11}(q_1)\omega_{21}(q_2)}
1953:       {\omega_{21}^2(q_2)-\omega_{11}^2(q_1)},
1954: $$
1955: and 
1956: \begin{equation}
1957:    \tilde\beta_{1{\bf q}}=\sum\limits_{s=1}^{\infty}
1958:    {\phi}_{1s{\bf{q}}}^{2}.\nonumber
1959: %   \label{beta}
1960: \end{equation}
1961: The corresponding formulas for the second array are obtained by
1962: replacing $1s\rightarrow 2s$.\\
1963:   
1964: Another simplification is made for modes with quasi--momenta on the
1965: resonance line.  Consider for simplicity a square QCB (in this case BZ
1966: coincides with the elementary cell of the reciprocal lattice, and the
1967: resonance line coincides with the BZ diagonal $OC$ in
1968: Fig.\ref{BZ2}) and assume that ${\bf q}$ is not too close to the BZ
1969: corner $C$.  The initial frequencies of modes belonging to the same
1970: band coincide,
1971: \begin{equation}
1972: 	{\omega}_{1s{\bf q}}= {\omega}_{2s{\bf
1973: q}}\equiv {\omega}_{s{\bf q}}.\nonumber
1974: %	\label{res1}
1975: \end{equation}
1976: Therefore renormalization strongly mixes the initial variables
1977: \begin{eqnarray*}
1978: {\tilde\theta}_{gs{\bf{q}}} & = & \frac{1}{\sqrt{2}}
1979:   \left(
1980:        1-\frac{1}{2}
1981:        \beta_{s{\bf q}}
1982:   \right)
1983:   ({\theta}_{2s{\bf{q}}}+
1984:        {\theta}_{1s{\bf{q}}})
1985: \\
1986: & - &
1987:    \frac{1}{\sqrt{2}}
1988:    \sum\limits_{s'\neq s}  
1989:       ({\phi}_{s's{\bf{q}}}
1990:       {\theta}_{1s'{\bf{q}}}-
1991:       {\phi}_{ss'{\bf{q}}}
1992:       {\theta}_{2s'{\bf{q}}}),
1993: %     \label{theta1on}
1994: \end{eqnarray*}
1995: \begin{eqnarray*}
1996: {\tilde\theta}_{usþ{\bf{q}}} & = & \frac{1}{\sqrt{2}}
1997:   \left(
1998:        1-\frac{1}{2}
1999:        {\beta}_{s{\bf q}}
2000:   \right)
2001:   ({\theta}_{2s{\bf{q}}}
2002:       -{\theta}_{1s{\bf{q}}})\\
2003: & - & 
2004:    \frac{1}{\sqrt{2}}
2005:    \sum\limits_{s'\neq s}
2006:    ({\phi}_{s's{\bf{q}}}
2007:       {\theta}_{1s'{\bf{q}}}+
2008:       {\phi}_{ss'{\bf{q}}}
2009:       {\theta}_{2s'{\bf{q}}}),
2010: % \label{theta2on}
2011: \end{eqnarray*}
2012: and the corresponding eigenfrequencies are shifted from their bare
2013: values
2014: \begin{eqnarray*}
2015: {\omega}_{gs{\bf{q}}}^{2} & \approx &
2016:       {\omega}_{s{\bf{q}}}^{2}
2017:        \left(
2018:             1+
2019:             {\phi}_{1s2s{\bf{q}}}\right),
2020: %            \label{even}
2021:   \\
2022: {\omega}_{us{\bf{q}}}^{2} & \approx &
2023:       {\omega}_{s{\bf{q}}}^{2}
2024:        \left(
2025:             1-
2026:             {\phi}_{1s2s{\bf{q}}}\right).
2027: %            \label{odd}
2028: \end{eqnarray*}
2029: 
2030: For the first band $s=1$ these formulas look like
2031: \begin{equation}
2032:   {\omega}_{\pm ,1{\bf{q}}}^{2}\approx
2033:   \omega_1^2(q_1)
2034:   \left(1\pm\sqrt{\varepsilon}\varphi_{\bf{q}}\right).\nonumber
2035: %  \label{FZ}
2036: \end{equation}
2037: Note that in the resonance case the splitting of the degenerate modes 
2038: is of the order of $\sqrt{\varepsilon}$ that essentially exceeds the 
2039: shift 
2040: of eigenfrequencies in the non-resonant case (\ref{omega-SecOrd}).\\
2041: 
2042: The interband mixing becomes significant near the BZ boundaries.  
2043: Not very close to the crossing points of these boundaries with the 
2044: resonant lines, this mixing is accounted for by a standard way.  As a 
2045: result we find that the interband hybridization gap for the bosons 
2046: propagating along the first array can be estimated as
2047: \begin{eqnarray*}
2048:   {\Delta\omega}_{12}\sim
2049:  {vQ}\varepsilon \frac{r_{0}}{a}.
2050: \end{eqnarray*}
2051: Similar gaps exist near the boundary of the BZ for each pair of odd and 
2052: next even energy bands, as well as for each even and next odd band near 
2053: the lines $q_1=0$ or $q_2=0$.  The energy gap between the $s$-th and 
2054: $(s+1)$-th bands is estimated as
2055: \begin{eqnarray*}
2056:   {\Delta\omega}_{s,s+1}\sim
2057:   {vQ}\varepsilon\frac{r_{0}}{a}{\it o}(s^{-1}).
2058: \end{eqnarray*}
2059: For large enough band number $s,$ the interaction is 
2060: effectively suppressed, 
2061: ${\phi}_{1s2s'}\to{0},$ and the gaps vanish.\\
2062: 
2063: The spectral behavior in the vicinity of the crossing points of a 
2064: resonance line and the BZ boundary needs more detailed calculations.  
2065: Nevertheless it can also be analyzed in a similar way.  The results of 
2066: such an alalysis are discussed in subsubsection \ref{subsubsec:Square}.\\
2067: 
2068: %%%%%%%%%%%%%%%%%%%%%%%%%
2069: \section*{Appendix C. {\it ac} conductivity}
2070: %%%%%%%%%%%%%%%%%%%%%%%%%%
2071: For interacting wires, where ${\phi}_{js}({q}_j)\neq{0},$ the 
2072: correlator (\ref{CurrCorr}) may be easily calculated after 
2073: diagonalization of the Hamiltonian (\ref{TotHam2}) by means of the 
2074: transformations (\ref{tilde-teta1}) and (\ref{tilde-teta2}).  As a 
2075: result, one has:
2076: \begin{eqnarray*}
2077: \left\langle\left[
2078:                  {J}_{11{ \bf{q}}}(t),
2079:                  {J}_{11{\bf{q}}}^{\dag}(0)
2080:   \right]\right\rangle = \\
2081:       -2ivg\left(u_{\bf{q}}^{2}
2082:       {\omega}_{+,1{\bf{q}}}
2083:       \sin({\omega}_{+,1{\bf{q}}}t)
2084:       +v_{\bf{q}}^{2}
2085:       {\omega}_{-,1{\bf{q}}}
2086:       \sin({\omega}_{-,1{\bf{q}}}t)\right),
2087: \end{eqnarray*}
2088: \begin{eqnarray*}
2089:   &
2090: \left\langle\left[
2091:                  {J}_{11{ \bf{q}}}(t),
2092:                  {J}_{21{\bf{q}}}^{\dag}(0)
2093: \right]\right\rangle  =
2094:   -2ivgu_{\bf{q}}v_{\bf{q}}
2095:   & \\
2096:   & \times \left(
2097:      {\omega}_{-,1{\bf{q}}}
2098:      \sin({\omega}_{-,1{\bf{q}}}t)-
2099:      {\omega}_{+,1{\bf{q}}}
2100:      \sin({\omega}_{+,1{\bf{q}}}t)
2101:  \right), &
2102: \end{eqnarray*}
2103: where $u_{\bf{q}}$ and $v_{\bf{q}}$ are defined in Eqs.(\ref{u}),
2104: (\ref{v}). Then, for the optical absorption ${\sigma}'$ one obtains
2105: \begin{eqnarray}
2106: & {\sigma}'_{11}({\bf{q}},\omega)  =
2107:     {\pi}{v}{g}
2108:     \Bigl[
2109:          u_{\bf{q}}^{2}
2110:          \delta
2111:          \left(
2112:               {\omega}-
2113:               {\tilde\omega}_{+,1{\bf{q}}}
2114:          \right)+
2115:          v_{\bf{q}}^{2}
2116:          \delta
2117:          \left(
2118:               {\omega}-
2119:               {\tilde\omega}_{-,1{\bf{q}}}
2120:          \right)
2121:     \Bigr]
2122:     \label{opt_l}
2123: \end{eqnarray}
2124: \begin{equation}
2125:    {\sigma}'_{12}({\bf{q}},\omega)  =
2126:     {\pi}{v}{g}
2127:     u_{\bf{q}}v_{\bf{q}}
2128:     \Bigl[
2129:          \delta
2130:          \left(
2131:               {\omega}-
2132:               {\tilde\omega}_{-,1{\bf{q}}}
2133:          \right)-
2134:          \delta
2135:          \left(
2136:               {\omega}-
2137:               {\tilde\omega}_{+,1{\bf{q}}}
2138:          \right)
2139:     \Bigr].
2140:     \label{opt_t}
2141: \end{equation}
2142: For quasimomentum ${\bf{q}}$ away from the resonant coupling line,
2143: $u_{\bf{q}}^{2}\approx{1}$ and $v_{\bf{q}}^{2}\sim\phi_{\bf{q}}^{2}$
2144: for $\Delta_{\bf{q}}>0$ ($v_{\bf{q}}^{2}\approx{1}$ and
2145: $u_{\bf{q}}^{2}\sim\phi_{\bf{q}}^{2}$ for $\Delta_{\bf{q}}<0$).  Then
2146: the longitudinal optical absorption (\ref{opt_l}) (i.e. the absorption
2147: within a given set of wires) has its main peak at the frequency
2148: ${\omega}_{+,1{\bf{q}}}\approx{v\vert{q_1}\vert}$ for
2149: $\Delta_{\bf{q}}>0$ (or
2150: ${\omega}_{-,1{\bf{q}}}\approx{v\vert{q_1}\vert}$ for
2151: $\Delta_{\bf{q}}<0$), corresponding to the first band of the pertinent
2152: array, and an additional weak peak at the frequency
2153: ${\omega}_{-,1{\bf{q}}}\approx{v\vert{q_2}\vert}$, corresponding to
2154: the first band of a complementary array.  It contains also a set of
2155: weak peaks at frequencies $\omega_{2,s{\bf{q}}}\approx [s/2]vQ$
2156: ($s=2,3,\ldots$) corresponding to the contribution from higher bands
2157: of the complementary array (in Eq.(\ref{opt_l}) these peaks are
2158: omitted).  At the same time, a second observable becomes relevant,
2159: namely, the transverse optical absorption (\ref{opt_t}).  It is
2160: proportional to the (small) interaction strength and has two peaks at
2161: frequencies ${\omega}_{+,1{\bf{q}}}$ and ${\omega}_{-,1{\bf{q}}}$ in
2162: the first bands of both sets of wires.\\
2163: 
2164: If the quasimomentum ${\bf{q}}$ belongs to the
2165: resonant coupling line $\Delta_{\bf{q}}=0$, then 
2166: $u_{\bf{q}}^{2}=v_{\bf{q}}^{2}=1/2$. In this case the longitudinal 
2167: optical
2168: absorption (\ref{opt_l}) has a split double peak at frequencies
2169: ${\omega}_{+,1{\bf{q}}}$ and ${\omega}_{-,1{\bf{q}}}$,
2170: instead of a single main peak. The transverse optical absorption
2171: (\ref{opt_t}), similarly to the non-resonant case (\ref{opt_t}),
2172: has a split double peak at frequencies
2173: ${\omega}_{+,1{\bf{q}}}$ and ${\omega}_{-,1{\bf{q}}}$, but 
2174: its amlitude is now of the order of unity.
2175: For $\left\vert{\bf{q}}\right\vert\to 0$ Eq.(\ref{opt_l}) reduces
2176: to that for an array of noninteracting wires (\ref{Drude_peak}),
2177: and the transverse optical conductivity (\ref{opt_t}) vanishes.\\
2178: 
2179: The imaginary part of the {\it ac} conductivity
2180: ${\sigma}''_{jj'}({\bf{q}},\omega)$ is calculated within the same
2181: approach. Its longitudinal component equals
2182: \begin{eqnarray*}
2183: &{\sigma}''_{11}({\bf{q}},\omega)=
2184:             \displaystyle{\frac{2vg}{\omega}
2185:             \left[
2186:                  \frac{u_{\bf{q}}^{2}
2187:                        {\omega}_{+,1{\bf{q}}}^{2}}
2188:                       {{\omega}_{+,1{\bf{q}}}^{2}-
2189:                        \omega^2}+
2190:             \frac{v_{\bf{q}}^{2}
2191:                       {\omega}_{-,1{\bf{q}}}^{2}}
2192:                       {{\omega}_{-,1{\bf{q}}}^{2}-
2193:                        \omega^2}
2194:             \right]}.
2195:         %\label{real_ln}
2196: \end{eqnarray*}
2197: Beside the standard pole at zero frequency, the imaginary part has
2198: poles at the resonance frequencies $\omega_{+,1{\bf q}}$,
2199: $\omega_{-,1{\bf q}}$, and an additional series of high band
2200: satellites (omitted here).  For quasimomenta far from the resonant
2201: lines, only the first pole is well pronounced while amplitude of the
2202: second one as well as amplitudes of all other sattelites is small.  At
2203: the resonant lines, amplitudes of both poles mentioned above are equal. 
2204: The corresponding expression for ${\sigma}'_{22}({\bf{q}},\omega)$ can be
2205: obtained by replacement $1\leftrightarrow 2$.  \\
2206: 
2207: The transverse component of the
2208: imaginary part of the {\it ac} conductivity has the form:
2209: \begin{eqnarray*}
2210: {\sigma}'_{12}({\bf{q}},\omega)=
2211:         \displaystyle{
2212:         \frac{2vg}{\omega}u_{{\bf q}}v_{{\bf q}}
2213:         \left[
2214:              \frac{{\omega}_{-,1{\bf{q}}}^{2}}
2215:                   {{\omega}^{2}-{\omega}_{-,1{\bf{q}}}^{2}}-
2216:              \frac{{\omega}_{+,1{\bf{q}}}^{2}}
2217:                   {{\omega}^{2}-{\omega}_{+,1{\bf{q}}}^{2}}
2218:         \right]}.
2219:         %\label{real_t}
2220: \end{eqnarray*}
2221: It always contains two poles and vanishes for noninteracting wires. 
2222: For quasimomenta far from the resonance lines the transverse component is 
2223: small while at these lines its amplitude is of the order of unity.\\
2224: %%%%%%%%%%%%%%%%%%%%%%%%
2225: \section*{Appendix D. Triple QCB Spectrum}
2226: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2227: To diagonalize the Hamiltonian(\ref{H}), we write down equations of motion
2228: %Eq.(12-14)
2229: \begin{eqnarray}
2230:  \left[
2231:       \omega_s^2(q_{j})-\omega^2
2232:  \right]\theta_{js{\bf{q}}}\nonumber\\
2233:  +\sqrt{\varepsilon}\phi_s(q_j)\omega_s(q_j)
2234:  \frac{r_0}{a}
2235:  \sum\limits_{s'}
2236:  \phi_{s'}(q_3)\omega_{s'}(q_3)
2237:  \theta_{3s'{\bf{q}}}
2238:  &=& 0,
2239:  \label{EqMotion1-2}
2240:  \\
2241:  \left[
2242:       \omega_s^2(q_{3})-\omega^2
2243:  \right]\theta_{3s{\bf{q}}}\nonumber\\
2244:  +\sqrt{\varepsilon}\phi_s(q_3)\omega_s(q_3)
2245:  \frac{r_0}{a}
2246:  \sum\limits_{j,s'}
2247:  \phi_{s'}(q_j)\omega_{s'}(q_j)
2248:  \theta_{js'{\bf{q}}}
2249:  &=& 0.
2250:  \label{EqMotion3}
2251: \end{eqnarray}
2252: Here $j=1,2$, and $\varepsilon$ is defined by Eq.(\ref{epsilon1}).
2253: The solutions of the set of equations
2254: (\ref{EqMotion1-2}) - (\ref{EqMotion3}) have the form:
2255: %Eq.(13)
2256: \begin{eqnarray}
2257:  \theta_{js{\bf{q}}} &=& A_j
2258:  \frac{\phi_s(q_j)\omega_s(q_j)}
2259:       {\omega_s^2(q_j)-\omega^2}.
2260:  \ \ \ \ \ j=1,2,3,\nonumber
2261: % \label{SolEqMot}
2262: \end{eqnarray}
2263: Substituting
2264: this equation into Eqs.(\ref{EqMotion1-2}) and
2265: (\ref{EqMotion3}), we have three equations for constants $A_j$:
2266: %Eq.(14-16)
2267: \begin{eqnarray*}
2268:  A_1+A_3{\sqrt{\varepsilon}}F_{q_3}(\omega^2) &=& 0,\\
2269: % \label{A1-Eq}
2270:  A_2+A_3{\sqrt{\varepsilon}}F_{q_3}(\omega^2) &=& 0,\\
2271: % \label{A2-Eq}
2272:  A_3+\sum\limits_{j=1,2}A_j{\sqrt{\varepsilon}}F_{q_j}(\omega^2) &=& 0,
2273: % \label{A3-Eq}
2274: \end{eqnarray*}
2275: where
2276: %Eq.(17)
2277: \begin{equation}
2278:  F_q(\omega^2)=\frac{r_0}{a}\sum\limits_{s}
2279:  \frac{\phi_s^2(q)\omega_s^2(q)}{\omega_s^2(q)-\omega^2}.\nonumber
2280: % \label{F_q}
2281: \end{equation}
2282: Dispersion relations can be obtained from the solvability condition 
2283: for this set of equations
2284: \begin{equation}
2285:  \varepsilon F_{q_3}(\omega^2)
2286:  \left(
2287:       F_{q_1}(\omega^2)+
2288:       F_{q_2}(\omega^2)
2289:  \right)=1.\nonumber
2290: % \label{SecularEq}
2291: \end{equation}
2292: The function $F_{q_{s}}(\omega^2)$ has a set of poles at
2293: $\omega^2=\omega_{s}^{2}(q)$, $s=1,2,3,\ldots$ .  For 
2294: $\omega^{2}<\omega_{s}^{2}(q)$, i.e. within the interval
2295: $[0,\omega_{1}^{2}(q)]$,  $F_{q_{s}}(\omega^2)$ is positive 
2296: increasing function.
2297: Its minimal value $F$ on the interval is reached at $\omega^2=0$
2298: and does not depend on quasi-momentum $q$
2299: \begin{equation}
2300:  F_q(0) =\frac{r_0}{a}\sum\limits_{s}
2301:  \phi_{s}^{2}(q)=\int d\xi \zeta_j^2(\xi)\equiv F.\nonumber
2302: % \label{F(0)}
2303: \end{equation}
2304: If the parameter $\varepsilon\equiv\eta^{2}$ is smaller than the critical 
2305:  value
2306: \begin{equation}
2307:   \varepsilon_c=
2308:   \frac{1}{2F^{2}},\nonumber
2309: %  \label{eps-crit}
2310: \end{equation}
2311: then all solutions $\omega^{2}$ of the characteristic equation are 
2312: positive.  When $\varepsilon$ increases, the lowest QCB mode softens 
2313: and its square frequency vanishes \textit{in whole BZ} at 
2314: $\varepsilon=\varepsilon_{c}$.  For exponential model 
2315: $\zeta(\xi)=\exp(-|\xi|),$ one obtains $\varepsilon_c\approx 1$.\\
2316: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2317: \section*{Appendix E. Triple Rabi Oscillations}
2318: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2319: The point $C(Q/3,2Q/3)$  of the BZ is the point of
2320: three-fold degeneracy, 
2321: $$
2322: q_{1}=q_{3}=-q_{2}+Q=\frac{Q}{3},
2323: $$
2324: $$
2325:   \omega_{11}(Q/3)=\omega_{21}(2Q/3)=\omega_{31}(Q/3)\equiv\omega_0.
2326: $$
2327: Equations of motion at this point in the resonance approximation read
2328: \begin{eqnarray*}
2329:   &&\left[\frac{d^2}{dt^2}+\omega_{0}^{2}\right]
2330:   \theta_{1}+
2331:   \sqrt{\varepsilon}\phi^2\omega_{0}^2\theta_{3} = 0,\\
2332:   &&\left[\frac{d^2}{dt^2}+\omega_{0}^{2}\right]
2333:   \theta_{2}+
2334:   \sqrt{\varepsilon}\phi^2\omega_{0}^2\theta_{3} = 0,\\
2335:   &&\left[\frac{d^2}{dt^2}+\omega_{0}^{2}\right]
2336:   \theta_{3}+
2337:   \sqrt{\varepsilon}\phi^2\omega_{0}^2
2338:   \left(\theta_{1}+\theta_{2}\right) = 0,
2339: %  \label{Euler-Lagr3}
2340: \end{eqnarray*}
2341: where $\theta_{j}\equiv\theta_{j{\bf q}}.$
2342: General solution of this system looks as
2343: \begin{eqnarray*}
2344: 	\left(\begin{array}{c}\theta_{1}(t)\\
2345: 	\theta_{2}(t)\\
2346: 	\theta_{3}(t)\end{array}\right)=
2347: 	\theta_{0}\left(\begin{array}{c}
2348: 	1\\
2349: 	-1\\
2350: 	0\end{array}\right)e^{i\omega_{0}t}+\ \ \ \ \ \ \ \ \ \ \ \ 
2351: 	\\
2352: 	\theta_{+}\left(\begin{array}{c}
2353: 	1\\
2354: 	1\\
2355: 	\sqrt{2}\end{array}\right)e^{i\omega_{+}t}+
2356: 	\theta_{-}\left(\begin{array}{c}
2357: 	1\\
2358: 	1\\
2359: 	-\sqrt{2}\end{array}\right)e^{i\omega_{-}t},
2360: %	\label{solu}
2361: \end{eqnarray*}
2362: where one of the eigenfrequencies coincides with $\omega_{0},$ while
2363: the two others are
2364: \begin{equation}
2365: 	\omega_{\pm}=\sqrt{1\pm\sqrt{2}\phi^{2}},\nonumber
2366: %	\label{omega}
2367: \end{equation}
2368: and $\theta_{0,\pm}$ are the corresponding amplitudes.\\
2369: 
2370: Choosing initial conditions
2371: \begin{eqnarray*}
2372:   &&{\theta}_{1}(0)=i\theta_0, \ \ \ \ \ \ \
2373:   {\dot{\theta}}_{1}(0)=\omega_0\theta_0,\\
2374:   &&{\theta}_{2}(0)=0,
2375:   \ \ \ \ \ \ \ \
2376:   {\dot{\theta}}_{2}(0)=0,\\
2377:   &&{\theta}_{3}(0)=0,
2378:   \ \ \ \ \ \ \ \
2379:   {\dot{\theta}}_{3}(0)=0,
2380: %  \label{condit3}
2381: \end{eqnarray*}
2382: we obtain for the field amplitudes at the coordinate origin
2383: \begin{eqnarray*}
2384:   \theta_1(0,0;t) &=& \frac{\theta_0}{4}
2385:   \left[
2386:        \frac{\omega_0}{\omega_+}\sin(\omega_+t)+
2387:        \frac{\omega_0}{\omega_-}\sin(\omega_-t)
2388:   \right]
2389:   \\&&+
2390:   \frac{\theta_0}{2}\sin(\omega_0t),
2391:   \nonumber\\
2392:   \theta_2(0,0;t) &=& \frac{\theta_0}{4}
2393:   \left[
2394:        \frac{\omega_0}{\omega_+}\sin(\omega_+t)+
2395:        \frac{\omega_0}{\omega_-}\sin(\omega_-t)
2396:   \right]
2397:   \\&&-
2398:   \frac{\theta_0}{2}\sin(\omega_0t),
2399:   \nonumber\\
2400:   \theta_3(0,0;t) &=& \frac{\theta_0}{2\sqrt{2}}
2401:   \left[
2402:        \frac{\omega_0}{\omega_+}\sin(\omega_+t)-
2403:        \frac{\omega_0}{\omega_-}\sin(\omega_-t)
2404:   \right].
2405: %  \label{solution3}
2406: \end{eqnarray*}
2407: In the limiting case $\varepsilon\ll 1$ these formulas coincide
2408: with Eqs.(\ref{sol3}) in subsection {\bf III.C}.
2409: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2410: \begin{thebibliography}{99}
2411: \bibitem{Voit} J. Voit, Rep. Prog. Phys. {\bf 58}, 977 (1994).
2412: \bibitem{Anders} P. W. Anderson, Science {\bf 235}, 1196 (1987).
2413: \bibitem{Rueckes} T. Rueckes, K. Kim, E. Joselevich, G. Y. Tseng, C. L.
2414: Cheung, and C. M. Lieber, Science {\bf 289}, 94 (2000).
2415: \bibitem{Dai} H. Dai, Surface Sci, {\bf 500}, 218 (2002). 
2416: \bibitem{Delft} J. von Delft, and H. Schoeller,
2417: Ann. der Physik {\bf 7}, 225 (1998).
2418: \bibitem{Wen} X. G. Wen, Phys. Rev. B {\bf 42}, 6623 (1990);
2419: H. J. Schultz, Int. J. Mod. Phys. {\bf 1/2}. 57 (1991).
2420: \bibitem{Luba00} V.J.  Emery, E. Fradkin, S. A. Kivelson, and T.C.
2421: Lubensky,
2422: Phys. Rev. Lett. {\bf 85}, 2160 (2000).
2423: \bibitem{Vica01} A. Vishwanath and D. Carpentier,
2424: Phys. Rev. Lett. {\bf 86}, 676 (2001).
2425: \bibitem{Luba01} R. Mukhopadhyay, C. L. Kane, and T. C. Lubensky,
2426: Phys. Rev. B {\bf 63}, 081103(R) (2001); Phys. Rev. B {\bf 64},
2427: 045120 (2001).
2428: \bibitem{Hern} C. S. Hern, T. C. Lubensky, and J. Toner, Phys. Rev.
2429: Lett. {\bf 83}, 2745 (1999).
2430: \bibitem{Sond} S. L. Sondhi, and K. Yang, Phys. Rev. B {\bf 63} 054430
2431: (2001);
2432: C. L. Kane, R. Mukhopadhyay, and T. C. Lubensky,
2433: Phys. Rev. Lett. {\bf 88}, 036401 (2002).
2434: \bibitem{Buch} E. Buchstab, A.V.Butenko, and V. V. Pilipenko,
2435: Sov. J. Low Temp. Phys. {\bf 11},357 (1985); E. Buchstab, A.V.Butenko, 
2436: N. Ya. Fogel, V. G. Cherkasova, and R. L. Rosenbaum, 
2437: Phys.Rev.B {\bf 50}, 10063 (1994).
2438: \bibitem{sault}G. Mihaly, I. Kezsmarki, F. Zambroszky, and L. Forro,
2439: Phys. Rev. Lett. {\bf 84}, 2670 (2000).
2440: \bibitem{metals} Y. Maeno \textit{et al}, Nature {\bf 372}, 532 (1994);
2441: I. Terasaki, Y. Sasago, and K. Uchinokura, Rev. B {\bf 63} RR12685
2442: (1997); 
2443: S.M. Loureiro \textit{et al},  Rev. B {\bf 63} 094109 (2001);
2444: I, Tsukada \textit{et al}, cond-mat/0012395 (2000);
2445: T. Valla  \textit{et al}, Nature, Nature {\bf 417}, 627 (2002). 
2446: \bibitem{Avr} J. E. Avron, A. Raveh, and B. Zur, Rev. Mod.Phys. {\bf
2447: 60}, 873 (1988).
2448: \bibitem{Avi} Y. Avishai, and J. M. Luck, Phys. Rev. B{\bf 45}, 1074
2449: (1992).
2450: \bibitem{Guinea} F. Guinea, and G. Zimanyi, Phys. Rev. B {\bf 47},
2451: 501 (1993).
2452: \bibitem{Castro} A. H. Castro Neto, and F. Guinea, Phys. Rev. Lett.
2453: {\bf 80}, 4040 (1998).
2454: \bibitem{Kuzm} I. Kuzmenko, S. Gredeskul, K. Kikoin and Y. Avishai,
2455: Fiz. Nizkikh Temp., {\bf 28}, 752 (2002).
2456: \bibitem{Luo} Y. Luo, C. P. Collier, J.O. Jeppesen, K.A. Nielsen,
2457: E. Delonno, G. Ho, J. Perkins, H-R. Tseng, T. Yamamoto, J.F. Stoddardt,
2458: J.R. Heath, ChemPhysChem {\bf3}, 519 (2002). 
2459: \bibitem{Sasaki} K. Sasaki, cond-mat/0112178.
2460: \bibitem{Louie} S.G. Louie in {\it Carbon Nanotubes}, M.S.
2461: Dresselhaus, G. Dresselhaus, Ph. Avouris (Eds.), Topics Appl.
2462: Phys. {\bf 80}, 113 (2001), Springer, Berlin 2001.
2463: \bibitem{Egger} R. Egger, A. Bachtold, M.S. Fuhrer, M. Bockrath, D.H.
2464: Cobden, and P.L. McEuen, in {\it Interacting Electrons in 
2465: Nanostructures}, p. 125, R. Haug, and H. Schoeller (Eds.), Springer 
2466: (2001).
2467: \bibitem{Diehl} M.R. Diehl, S.N. Yaliraki, R.A. Beckman, M. Barahona, and
2468: J.R. Heath, Angew. Chem. Int. Ed. {\bf41}, 353 (2002).
2469: \bibitem{Tseng} G.Y. Tseng and J.C. Ellenbogen, Science {\bf 294}, 1293 (2001). 
2470: \bibitem{Ando} H. Ajiki, and T. Ando, J. Phys. Soc. Jpn. {\bf 65},
2471: 505 (1996).
2472: \end{thebibliography}
2473: \end{document}
2474: