1:
2: %\documentclass [aps,prb,preprint,groupedaddress,showpacs] {revtex4}
3: \documentclass [aps,prb,twocolumn,groupedaddress,showpacs] {revtex4}
4: %\documentclass [aps,prl,preprint,groupedaddress,showpacs] {revtex4}
5: %\documentclass [aps,prl,twocolumn,groupedaddress,showpacs] {revtex4}
6:
7: \usepackage {graphicx,amssymb,amsbsy,color}
8: \bibliographystyle {apsrev}
9:
10: %--------------------------------------------------------------
11: \textheight 9.2 in % Simulate the actual PRB textheight.
12: \topmargin -15 mm
13: \oddsidemargin -0.3 true in
14: %--------------------------------------------------------------
15:
16: \begin {document}
17:
18: \title {
19: Localized states in the continuum in low-dimensional systems
20: }
21: \author {Khee-Kyun Voo$^*$ and C. S. Chu}
22: \affiliation {Department of Electrophysics, National Chiao Tung
23: University, Hsinchu 30010, Taiwan, Republic of China}
24:
25: \date {\today}
26:
27: \begin {abstract}
28:
29: It is shown in this paper that for open systems, states which are
30: localized in space, discrete in energy, and embedded in the continuum of
31: extended states, can be sustained by low-dimensional and channeled leads.
32: These states have an origin different from that of analogous states
33: discussed by J. von Neumann and E. Wigner [Phys. Z. {\bf 30}, 465 (1929)].
34: A few representative systems are discussed. These states cause, for
35: example, infinitely sharp Fano resonance in transport when they are
36: marginally destroyed.
37:
38: \end {abstract}
39:
40: \pacs {73.22.Dj, 03.65.Ge, 73.23.Ad, 73.63.-b}
41: \maketitle
42:
43: \newpage
44:
45: \section {Introduction}
46:
47: Shortly after the discovery of quantum mechanics, von Neumann and Wigner
48: pointed \cite {NW29,SH75} out that potentials defining closed
49: boundary conditions were not the only cause of discrete and localized
50: (normalizable) states. They pointed out that localized states could also
51: be due to the destructive interference in the Bragg scattering from
52: certain long-ranged wiggling potentials defining open boundary conditions.
53: These states are embedded in the continuum, decay in space with a power
54: dependence, and have been studied in atomic and molecular
55: systems \cite {SH75,FW85,CFR03} and superlattices. \cite {CSF92} In this
56: paper, we show that analogous states can also be found in open systems
57: with low-dimensional leads. They decay exponentially in space in contrast
58: to those discussed by von Neumann and Wigner, \cite {NW29,SH75} and also
59: have a different origin. Moreover, they are shown to be related to the
60: infinitely sharp Fano resonance \cite {Fan35} in transport.
61:
62:
63: To illustrate the properties of such states in low-dimensional systems,
64: three representative model systems will be discussed --- A tight-binding
65: (TB) molecular system, a quantum graph with doubly-connected
66: one-dimensional (1D) channels, and a waveguide in a two-dimensional (2D)
67: space. All three are open systems. The first two are simple enough for
68: analytic analyses, where some generic properties can be studied
69: rigorously. The third one is to illustrate the presence of these states
70: in more realistic systems, but unfortunately, it allows only a numerical
71: analysis.
72:
73:
74: \section {Models and Discussions}
75:
76: \subsection {A tight-binding molecular system}
77: \label {tbmol}
78:
79: First we consider the TB system shown in Fig.~\ref {mol}(a) which is
80: defined by the Hamiltonian
81: \begin {eqnarray}
82: H &=& H_{\rm mol} + H_{\rm lead} + H_{\rm mol-lead},
83: \nonumber\\
84: H_{\rm mol} &\equiv& \sum_{i=1}^4 c_i^\dagger V_i^{~} c_i^{~} +
85: \sum_{i=1}^4 \sum_{j>i} \left[ c_i^\dagger h_{ij}^{~} c_j^{~} + {\rm H.c.}
86: \right], \nonumber\\
87: H_{\rm lead} &\equiv& - t \sum_{\eta = {\rm I, II}}
88: \sum_{i=0} ^\infty
89: c_{(i+1)_\eta}^\dagger c_{i_\eta}^{~} + {\rm H.c.}, \nonumber\\
90: H_{\rm mol-lead} &\equiv& \sum_{\eta = {\rm I, II}} \sum_{i=1}^4
91: c_{0_\eta} ^\dagger h_{0_\eta i}^{~} c_i^{~} + {\rm H.c.},
92: \end {eqnarray}
93: where $c_\lambda$, $\lambda \in \{ 1 \sim 4, 0_{\rm I} \sim \infty_{\rm
94: I}, 0_{\rm II} \sim \infty_{\rm II} \}$, is the annihilation operator of a
95: spinless particle on site $\lambda$, $t$ and $V_i$ are real, and $h_{ij}$
96: and $h_{0_\eta i}$ are complex in general. Sites $1 \sim 4$ are in a
97: ``molecule'', and sites $0_\eta \sim \infty_\eta$ are in lead $\eta$,
98: $\eta=$ I and II. This Hamiltonian may describe a nanoscopic molecule or a
99: mesoscopic cluster of quantum dots connected to two leads.
100:
101: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
102: \begin {figure}
103:
104: %\special {psfile=lsc_fig1.EPS hoffset=10cm voffset=-350cm hscale=120cm
105: %vscale=120cm}
106: %\vspace {11cm}
107:
108: \special {psfile=lsc_fig1.EPS hoffset=-10cm voffset=-230cm hscale=70cm
109: vscale=70cm}
110: \vspace {7cm}
111:
112: \caption {
113: (Color online)
114: (a) The considered open TB system, which consists of a molecule with four
115: sites (labeled by 1 $\sim$ 4) connected to two leads (labeled by I and II)
116: with serially connected sites (labeled by $i_\eta$ for lead $\eta$, where
117: $i= 0 \sim \infty$ and $\eta=$ I and II). The hoppings are denoted by
118: bonds.
119: (b) The transmission probability $T$ at different energies for $V_2=0$
120: (solid line), where there is a LSC (located by a dotted line); and $V_2 =
121: 0.05 t$ (dash-dotted line) and $0.3 t$ (dash-dot-dotted line), where
122: there are no LSC. Those not-mentioned system parameters are refered to the
123: text.
124: }
125:
126: \label {mol}
127: \end {figure}
128: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
129:
130:
131: Using a basis set $\{ | \lambda \rangle \}$ defined by $| \lambda
132: \rangle \equiv c_\lambda ^\dagger | 0 \rangle$ and $c_\lambda | 0 \rangle
133: \equiv 0$, one can write the time-independent Schr$\ddot {\rm o}$dinger
134: equation (TISE) $H | \psi \rangle = E | \psi \rangle$, where $E$ is the
135: energy of the particle, into a set of simultaneous finite-difference (FD)
136: equations. In a lead, the FD equations read $t \psi_{j-1}^\eta + E \psi_j
137: ^\eta + t \psi_{j+1}^\eta = 0$, for $j=1 \sim \infty$ and $\eta = $ I or
138: II, where $\psi _{j} ^\eta \equiv \langle j_\eta | \psi \rangle$. It has
139: an analytic solution
140: \begin {eqnarray}
141: \psi_j^\eta = A_\eta e^{-ikj} + B_\eta e^{ikj},
142: \label {wflead}
143: \end {eqnarray}
144: where $A_\eta$ and $B_\eta$ are arbitrary complex numbers, and $k \equiv
145: |k| \equiv {\rm cos}^{-1} [-E/(2t)] \equiv k (E)$. Since $k>0$, the
146: ingoing or inward propagating wave amplitudes are $A_{\rm I}$ and $A_{\rm
147: II}$.
148:
149:
150: There remains six FD equations not solved by Eq.~(\ref {wflead}).
151: Replacing $\psi_0^\eta$ and $\psi_1^\eta$ by $A_\eta$ and $B_\eta$ using
152: Eq.~(\ref {wflead}), and writing $\psi _{j} \equiv \langle j | \psi
153: \rangle$ for $j = 1 \sim 4$, the six equations can be written as a matrix
154: equation,
155: \begin {widetext}
156: \begin {eqnarray}
157: \left[
158: \begin {array} {cccccc}
159: - t e^{ik (E)} - E & h_{0_{\rm I} 1}^{~} & h_{0_{\rm I} 2}^{~} & h_{0_{\rm
160: I} 3}^{~} & h_{0_{\rm I} 4}^{~} & 0 \\
161: h_{0_{\rm I} 1}^* & V_1-E & h_{12}^{~} & h_{13}^{~} & h_{14}^{~} &
162: h_{0_{\rm II} 1}^* \\
163: h_{0_{\rm I} 2}^* & h_{12}^* & V_2-E & h_{23}^{~} & h_{24}^{~} & h_{0_{\rm
164: II} 2}^* \\
165: h_{0_{\rm I} 3}^* & h_{13}^* & h_{23}^* & V_3-E & h_{34}^{~} & h_{0_{\rm
166: II} 3}^* \\
167: h_{0_{\rm I} 4}^* & h_{14}^* & h_{24}^* & h_{34}^* & V_4-E & h_{0_{\rm II}
168: 4}^* \\
169: 0 & h_{0_{\rm II} 1}^{~} & h_{0_{\rm II} 2}^{~} & h_{0_{\rm II} 3}^{~} &
170: h_{0_{\rm II} 4}^{~} & - t e^{ik (E)} - E \\
171: \end {array}
172: \right]
173: \left[
174: \begin {array} {c}
175: B_{\rm I} \\
176: \psi_1 \\
177: \psi_2 \\
178: \psi_3 \\
179: \psi_4 \\
180: B_{\rm II} \\
181: \end {array}
182: \right]
183: =
184: \left[
185: \begin {array} {c}
186: \left( E + t e^{-ik(E)} \right) A_{\rm I} \\
187: - h_{0_{\rm I} 1}^* A_{\rm I} - h_{0_{\rm II} 1}^* A_{\rm II} \\
188: - h_{0_{\rm I} 2}^* A_{\rm I} - h_{0_{\rm II} 2}^* A_{\rm II} \\
189: - h_{0_{\rm I} 3}^* A_{\rm I} - h_{0_{\rm II} 3}^* A_{\rm II} \\
190: - h_{0_{\rm I} 4}^* A_{\rm I} - h_{0_{\rm II} 4}^* A_{\rm II} \\
191: \left( E + t e^{-ik(E)} \right) A_{\rm II} \\
192: \end {array}
193: \right],
194: \label {moleq}
195: \end {eqnarray}
196: \end {widetext}
197: where the components from top to bottom are respectively the FD equations
198: centered at sites $0_{\rm I}$, 1, 2, 3, 4, and $0_{\rm II}$. When $A_{\rm
199: I}$ and $A_{\rm II}$ are given, there are six unknowns ($B_{\rm I}$,
200: $\psi_1$, $\psi_2$, $\psi_3$, $\psi_4$, and $B_{\rm II}$) to be found.
201: The unknowns can be found by a straightforward matrix inversion when the
202: square matrix has a nonzero determinant. Notably, the determinant actually
203: can {\em vanish} at certain energies for some system configurations, and
204: imply a nontrivial solution at $A_{\rm I} = A_{\rm II} = 0$ (no ingoing
205: waves). The solution must be localized within the molecule, since the
206: outgoing wave amplitudes $B_{\rm I}$ and $B_{\rm II}$ are necessarily
207: vanishing due to unitarity.
208:
209:
210: The determinant of the square matrix in Eq.~(\ref {moleq}) vanishes
211: whenever the rows or columns of the matrix are not linearly independent.
212: For instance, consider the configuration $h_{0_{\rm I} 2}^{~} = h_{0_{\rm
213: I} 3}^{~}$, $h_{0_{\rm II} 2}^{~} = h_{0_{\rm II} 3}^{~}$, $h_{12}^{~} =
214: h_{13}^{~}$, $h_{24}^{~} = h_{34}^{~}$, $h_{23}^{~} = h_{23}^*$, and $V_2
215: = V_3$, at the energy $E = V_2 - h_{23}$. In this occasion, the third and
216: fourth rows of the matrix are seen to be identical,
217: which means that the determinant of the matrix vanishes, and the solution
218: to the problem is not unique. Note that the system in this configuration
219: is not really ``symmetric'' in the usual sense.
220:
221:
222: A complete solution $\Psi (\lambda)$ for Eq.~(\ref {moleq}), in the case
223: of a simple symmetric system in which $h_{12} = h_{13} = h_{24} = h_{34}
224: \equiv \Delta$, $h_{0_{\rm I} 1}^{~} = h_{0_{\rm II} 4}^{~} \equiv
225: \Gamma$, and $h_{14}^{~} = h_{23}^{~} = h_{0_{\rm I} 2}^{~} = h_{0_{\rm I}
226: 3}^{~} = h_{0_{\rm I} 4}^{~} = h_{0_{\rm II} 1}^{~} = h_{0_{\rm II} 2}^{~}
227: = h_{0_{\rm II} 3}^{~} = V_1 = V_2 = V_3 = V_4 = 0$ (which is similar to
228: the model considered in Ref.~\onlinecite {ZCP02}), at $E=0$ (i.e.,
229: $k=\pi/2$), where the determinant vanishes, is found to be
230: \begin {widetext}
231: \begin {eqnarray}
232: \Psi (\lambda) &=& \psi_{\rm ext} (\lambda) + \beta \psi_{\rm loc}
233: (\lambda), \nonumber\\
234: \psi_{\rm ext} (\lambda) &=& { {t (A_{\rm I} - A_{\rm II})} \over {i
235: \Gamma} } (\Delta _{\lambda,1} - \Delta _{\lambda,4} ) - { {\Gamma} \over
236: {\Delta} } (A_{\rm I} \Delta _{\lambda,2} + A_{\rm II} \Delta_{\lambda,3}
237: ) \nonumber\\
238: && + \sum_{j=0}^\infty \left[ (i^{-j} A_{\rm I} + i^{j} A_{\rm II}) \Delta
239: _{\lambda, j_{\rm I}} + (i^{-j} A_{\rm II} + i^{j} A_{\rm I}) \Delta
240: _{\lambda, j_{\rm II}} \right], \nonumber\\
241: \psi_{\rm loc} (\lambda) &=& \Delta _{\lambda,2} - \Delta _{\lambda,3},
242: \label {psi0}
243: \end {eqnarray}
244: \end {widetext}
245: where $\beta$ is an arbitrary complex number, $i \equiv \sqrt {-1}$, and
246: $\Delta _{\lambda,\lambda'} = 1$ (0) when $\lambda = \lambda'$ ($\lambda
247: \neq \lambda'$). $\Psi$ is seen to be a superposition of an extended
248: state $\psi_{\rm ext}$ and a localized state $\psi_{\rm loc}$. When
249: $A_{\rm I} = A_{\rm II} = 0$ or $|\beta| \rightarrow \infty$, $\Psi
250: \rightarrow \psi_{\rm loc}$ and it is a localized state in the continuum
251: (LSC).
252:
253:
254: The origin of the LSCs in TB systems can also be understood in a more
255: direct manner, given an insight from the observation that $\psi_{\rm
256: loc}$ vanishes at the sites in the molecule in contact with the leads [see
257: Eq.~(\ref {psi0})].
258: In general, if $\psi_A$ and $\psi_B$ are respectively the stationary
259: states at an energy $E$ in two isolated clusters of sites, labeled by $A$
260: and $B$, and $\psi_A^{~} (j_A^0) = \psi_B^{~} (j_B^0) = 0$, where $j_A^0$
261: ($j_B^0$) is a site on cluster $A$ ($B$), then the direct product $\psi_A
262: \otimes \psi_B$ is a stationary state at $E$ in a system where clusters
263: $A$ and $B$ are coupled by $t_{AB}^{~} c^\dagger_{j_A^0} c^{~}_{j_B^0} +
264: {\rm H.c.}$ This is because though the FD equations for the coupled
265: clusters contain the additional terms $t_{AB}^{~} \psi (j_B^0)$ and
266: $t_{AB}^* \psi (j_A^0)$, the wave function $\psi_A \otimes \psi_B$ is
267: still a solution of the FD equations since these terms vanish due to
268: $\psi_A^{~} (j_A^0) = \psi_B^{~} (j_B^0) = 0$.
269: If one of the clusters, say cluster $A$, is infinitely large or open and
270: $\psi_A$ is trivial, whereas cluster $B$ is finite sized and $\psi_B$ is
271: nontrivial, then $\psi_A \otimes \psi_B$ is a LSC.
272: The generalization of the above argument to the case with more
273: than two clusters is straightforward. An example of such case is
274: the LSC in Eq.~(\ref {psi0}), which can be constructed from three
275: clusters, where two of them (the two leads) have infinitely large
276: sizes and trivial stationary states.
277:
278:
279: For $A_{\rm I} = 1$ and $A_{\rm II} = 0$, the transmission probability $T$
280: defined by $T \equiv |B_{\rm II}|^2$ is plotted versus the energy $E$ for
281: $\Delta = 0.2 t$ and $\Gamma = 0.4 t$ in Fig.~\ref {mol}(b). In the same
282: figure, $T$ is also plotted for the same system parameters but $V_2 =$
283: $0.05 t$ and $0.3 t$, where the LSC is destroyed and has become an
284: almost-localized state. It is seen that $T$ can reflect the LSC only when
285: the LSC is destroyed by a perturbation and a Fano resonance appears. \cite
286: {VC06z} The blue shifts of the resonances from the energy of the LSC is
287: due to the increase of the energies of the almost-localized states by
288: $V_2$.
289:
290:
291: Therefore a comprehensive understanding of the problem may be stated as
292: the following. For problems of open systems, whenever the determinant of
293: the matrix to be inverted vanishes at an energy, there is a localized
294: state at that energy. If the energy is in a continuum of extended
295: scattering states, the localized state is a LSC, and a complete solution
296: is a superposition of the degenerate LSC and extended states. As the
297: states are decoupled, the transport which is related only to the extended
298: states does not reflect the presence of the LSC. When a LSC is destroyed,
299: or a previously localized state is coupled with the extended states, the
300: passing of a particle from one lead to the other through the molecule can
301: take place via two routes --- the extended states spanning the leads and
302: the molecule or the almost-localized state in the molecule. That results
303: in a nonresonant and a resonant transmission amplitudes, and the
304: interference results in a Fano resonance. \cite {Fan35} When the
305: almost-localized state is on the verge of the decoupling from the
306: continuum and acquiring an infinitely long lifetime, the resonance is
307: infinitely sharp. \cite {VC06c}
308:
309:
310: \subsection {A quantum graph}
311: \label {qgsec}
312:
313: The second example is a quantum graph. The quantum graphs are
314: multiply-connected 1D systems, which are meant to be effective models of
315: multiply-connected quasi-one-dimensional (Q1D) systems at low energies.
316: They are defined by the following conditions. Away from the junctions, a
317: particle is governed by a 1D Schr$\ddot {\rm o}$dinger equation. At a
318: junction, the wave functions on different branches are connected by a
319: chosen connecting scheme. \cite {VCT06} For a junction of three branches,
320: we have chosen a scheme defined by the equations (1) $\psi_1 = \psi_2 =
321: \psi_3$ and (2) $\nu \psi _1 + \sum_{i = 1} ^3 \partial \psi_i / \partial
322: x_i = 0$, where $\psi_i$ and $x_i$ are respectively the wave function and
323: coordinate defined on branch $i$. The coordinates are directed away from
324: the junction, and $\nu$ is a given real parameter with a dimension of
325: 1/length.
326:
327:
328: We consider a 1D ring connected to two 1D leads as shown in Fig.~\ref
329: {qgraph}(a). A potential everywhere equal to zero is assumed, and
330: the wave function at energy $E$ on the branch labeled by $\eta$ ($\eta = $
331: I, II, III, and IV) is $\psi ^\eta (x_\eta) = A_\eta e ^ {ikx_\eta} +
332: B_\eta e ^{-ikx_\eta}$, where $A_\eta$ and $B_\eta$ are arbitrary complex
333: numbers and $k \equiv \sqrt {2mE} / \hbar$. Applying the mentioned
334: connecting scheme at the two junctions, we obtain six simultaneous
335: equations or a matrix equation,
336: \begin {widetext}
337: \begin {eqnarray}
338: \left[
339: \begin {array} {cccccc}
340: -1 & 1 & 1 & 0 & 0 & 0 \\
341: -1 & 0 & 0 & 1 & 1 & 0 \\
342: 1-i\nu/k & 1 & -1 & 1 & -1 & 0 \\
343: 0 & e^{ikL_{\rm II}} & e^{-ikL_{\rm II}} & 0 & 0 & -1 \\
344: 0 & 0 & 0 & e^{ikL_{\rm III}} & e^{-ikL_{\rm III}} & -1 \\
345: 0 & -e^{ikL_{\rm II}} & e^{-ikL_{\rm II}} & -e^{ikL_{\rm III}} &
346: e^{-ikL_{\rm III}} & 1-i\nu/k \\
347: \end {array}
348: \right]
349: \left[
350: \begin {array} {c}
351: B_{\rm I} \\
352: A_{\rm II} \\
353: B_{\rm II} \\
354: A_{\rm III} \\
355: B_{\rm III} \\
356: B_{\rm IV} \\
357: \end {array}
358: \right]
359: =
360: \left[
361: \begin {array} {c}
362: A_{\rm I} \\
363: A_{\rm I} \\
364: A_{\rm I} (1-i\nu/k) \\
365: A_{\rm IV} \\
366: A_{\rm IV} \\
367: A_{\rm IV} (1-i\nu/k) \\
368: \end {array}
369: \right],
370: \label {qgrapheq}
371: \end {eqnarray}
372: \end {widetext}
373: where $L_{\rm II}$ and $L_{\rm III}$ are respectively the lengths of
374: branches II and III.
375: When the ingoing wave amplitudes $A_{\rm I}$ and $A_{\rm IV}$ are
376: specified, there are six unknowns to be found ($B_{\rm I}$, $A_{\rm II}$,
377: $B_{\rm II}$, $A_{\rm III}$, $B_{\rm III}$, and $B_{\rm IV}$).
378:
379:
380: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
381: \begin {figure}
382:
383: %\special {psfile=lsc_fig2.EPS hoffset=80cm voffset=-380cm hscale=160cm
384: %vscale=160cm}
385: %\vspace {12cm}
386:
387: \special {psfile=lsc_fig2.EPS hoffset=20cm voffset=-250cm hscale=100cm
388: vscale=100cm}
389: \vspace {8cm}
390:
391: \caption {
392: (Color online)
393: (a) The considered open quantum graph, which consists of a 1D ring (formed
394: by the branches labeled by II and III) connected to two leads (labeled by
395: I and IV). A coordinate $x_\eta$ with a positive direction indicated by an
396: arrow is defined on branch $\eta$ ($\eta =$ I, II, III, and IV).
397: (b) The transmission probability $T$ is plotted versus a dimensionless
398: wave number $\kappa$ defined by $\kappa \equiv k (L_{\rm II} + L_{\rm
399: III}) / (2\pi)$, for $\nu=0$, and $L_{\rm II} : L_{\rm III} = 1:3$ (solid
400: line) and $1:3.4$ (dash-dotted line).
401: There is a LSC for the $1:3$ case (indicated by a dotted line), but not
402: for the $1:3.4$ case in this energy range.
403: }
404:
405: \label {qgraph}
406: \end {figure}
407: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
408:
409:
410: From Eq.~(\ref {qgrapheq}), a LSC is seen at $k = k_n \equiv n n_0 \pi /
411: L_0$, when $L_{\rm II} : L_{\rm III}= n_{\rm II} : n_{\rm III}$, $n_0
412: \equiv {\rm min} (n_{\rm II}, n_{\rm III})$, $L_0 \equiv {\rm min} (L_{\rm
413: II}, L_{\rm III})$, $n$, $n_{\rm II}$, and $n_{\rm III}$ are
414: integers, and $n_{\rm II} + n_{\rm III}$ is even. The above condition
415: results in $e ^{ik_n L_{\rm II}} = e ^{ik_n L_{\rm III}} = (-1)^{n n_0}
416: \equiv \zeta_n$, and for the square matrix in Eq.~(\ref {qgrapheq}), the
417: differences between the corresponding elements in the first and second
418: rows are identical to that between the fourth and fifth rows. Therefore
419: the rows are not linearly independent and the determinant of the matrix
420: vanishes. A complete solution $\Psi_n (x)$, where $x \in \{ x_\eta | ~
421: \eta = {\rm I}, {\rm II}, {\rm III}, {\rm IV} \}$, is found to be
422: \begin {widetext}
423: \begin {eqnarray}
424: \Psi_n (x) &=& \psi_n ^{\rm ext} (x) + \beta \psi_n ^{\rm loc} (x) ,
425: \nonumber\\
426: \psi_n ^{\rm ext} (x) &=& \sum _{\eta = {\rm I, IV}} \left[ A_{\eta} e ^
427: {i k_n x_{\eta}} + \Lambda_n ( \bar {A}_{\eta}, A_{\eta} ) e ^ {-i k_n
428: x_{\eta}} \right] \Delta_{x, x_{\eta}} \nonumber\\
429: && + \sum _{\eta = {\rm II, III} } \left[ \Omega_n (A_{\rm I}, A_{\rm IV})
430: e ^{ik_n x_\eta} + \Omega_n (A_{\rm IV}, A_{\rm I}) e ^{ik_n (L_\eta -
431: x_\eta)} \right] \Delta_{x, x_\eta} , \nonumber\\
432: \psi_n ^{\rm loc} (x) &=& {\rm sin} ( k_n x_{\rm II} ) \Delta_{x, x_{\rm
433: II}} - {\rm sin} ( k_n x_{\rm III} ) \Delta_{x, x_{\rm III}}, \nonumber\\
434: \Lambda_n (X,Y) &\equiv& { { \zeta_n X - Y \nu_n } \over { 1 + \nu_n } },
435: ~~~~~~ \Omega_n (X,Y) \equiv { { X (3 + \nu_n) + \zeta_n Y ( 1 - \nu_n ) }
436: \over { 4 (1 + \nu_n) } } , \nonumber\\
437: \bar {A}_{\rm I} &\equiv& A_{\rm IV}, ~~~~~~ \bar {A}_{\rm IV} \equiv
438: A_{\rm I}, \nonumber\\
439: \nu_n &\equiv& { \nu \over { i k_n} } ,
440: \label {qgsol}
441: \end {eqnarray}
442: \end {widetext}
443: where $\beta$ is an arbitrary complex number, and $\Delta_{x, x_\eta} =
444: 1$ (0) when $x = x_\eta$ ($x \neq x_\eta$). The solution $\Psi_n$ is a
445: superposition of a LSC $\psi_n ^{\rm loc}$ and an extended state
446: $\psi_n ^{\rm ext}$. Like in the case of a TB model discussed in
447: Sec.~\ref {tbmol}, $\psi_n ^{\rm loc}$ vanishes at the point in contact
448: with the leads
449:
450:
451: Since the LSC is decoupled from the extended states, it will not
452: be revealed in those spectral properties due to the scattering. For
453: quantum graphs, Texier \cite {Tex02} has pointed out that the Friedel sum
454: rule, which is related to the phases of the eigenvalues of the scattering
455: matrix, fails to count the number of states in a scattering region in
456: such a situation.
457: Experimentally, it has also been found \cite {SS92} that a 1D lead does
458: not couple to a 2D wave function when the lead is located at a node
459: of the wave functions.
460:
461:
462: The LSCs in the defined quantum graph can also be understood directly
463: from the connecting equations at the junctions.
464: If a stationary state at energy $E$ in a graph labeled by $A$, has a node
465: at a point $P$ on one of the branches, then labeling the segments of the
466: branch on the two sides of $P$ by 1 and 2, defining coordinates $x_1$ and
467: $x_2$ respectively, with positive directions directed away from $P$, and
468: denoting the stationary wave functions on segments 1 and 2 by $\psi_1
469: (x_1)$ and $\psi_2 (x_2)$ respectively, one has $\psi_1 = \psi_2$ (both
470: are vanishing) and $\partial \psi_1 / \partial x_1 + \partial \psi_2 /
471: \partial x_2 = 0$ at $P$, and the stationary wave function can be written
472: in the form of the direct product $\psi_1^{~} \otimes \psi_2^{~} \otimes
473: \psi_{\rm others} ^A$, where $\psi _{\rm others} ^A$ is the direct
474: product of the wave functions on the other branches in the graph.
475: Similarly, for a graph labeled by $B$ containing a branch with an open end
476: labeled by 3, which has also a stationary state at $E$, the stationary
477: wave function can be written into the form $\psi_3^{~} \otimes \psi _{\rm
478: others} ^B$, where $\psi_3$ is the wave function on branch 3, and $\psi
479: _{\rm others} ^B$ is the direct product of the wave functions on the other
480: branches in the graph.
481: If $\psi_3$ is trivial, when the open end of branch 3 is attached to $P$
482: by demanding the connecting equations $\psi_1 = \psi_2 = \psi_3$ and $\nu
483: \psi _1 + \sum_{i = 1} ^3 \partial \psi_i / \partial x_i = 0$ to be
484: fulfilled at $P$, the direct product $\psi_1^{~} \otimes \psi_2^{~}
485: \otimes \psi_3^{~} \otimes \psi _{\rm others} ^A \otimes \psi _{\rm
486: others} ^B$ is a stationary solution at $E$ in the coupled graphs, since
487: the connecting equations are automatically fulfilled.
488: If graph $A$ is finite and $\psi_1^{~} \otimes \psi_2^{~} \otimes \psi
489: _{\rm others} ^A$ is nontrivial, whereas graph $B$ is infinitely large and
490: $\psi_3^{~} \otimes \psi _{\rm others} ^B$ is trivial, the stationary
491: state in the coupled graphs is a LSC.
492: Note that the above arguments can also be straightforwardly generalized to
493: the case of more than two clusters, and the LSC in Eq.~(\ref {qgsol}) is
494: an example of this.
495:
496:
497: For $A_{\rm I}=1$ and $A_{\rm IV}=0$, the transmission probability $T$
498: defined by $T=|B_{\rm IV}|^2$. In Fig.~\ref {qgraph}(b), $T$ is plotted
499: versus the wave number for $\nu = 0$, $L_{\rm II} : L_{\rm III} = 1:3$ and
500: $1 : 3.4$, at the vicinity of a LSC or almost-localized state.
501: For an isolated ring with an uniform potential on it, the nodes of the
502: stationary standing wave states are equally spaced, and hence
503: ``commensurate'' branch lengths (such as 1:3) in the open graph give rise
504: to LSCs. Note the Fano resonance when the LSC is destroyed.
505:
506:
507: \subsection {A two-dimensional waveguide}
508:
509: Our third example is a waveguide in a 2D continuous space as shown in
510: Fig.~\ref {waveguide}(a). The waveguide has a width of $W$, and the
511: potential in the waveguide is set at zero besides a square region that is
512: set at $V_{\rm G}$. We will call the infinitely extended regions on both
513: sides the leads, and the central finite sized region a resonating cavity.
514: This system may model a mesoscopic fabricated structure.
515:
516:
517: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
518: \begin {figure}
519:
520: %\special {psfile=lsc_fig3.EPS hoffset=40cm voffset=-350cm hscale=150cm
521: %vscale=150cm}
522: %\vspace {11cm}
523:
524: \special {psfile=lsc_fig3.EPS hoffset=0cm voffset=-230cm hscale=100cm
525: vscale=100cm}
526: \vspace {7cm}
527:
528: \caption {
529: (Color online)
530: (a) The considered open 2D waveguide. Two leads (labeled by I and II) of
531: width $W$ are attached oppositely to a square cavity (shaded region). The
532: potential in the leads are kept at zero, and that in the cavity is kept at
533: $V_{\rm G}$. A coordinate system $(x,y)$ is defined in the cavity as
534: shown.
535: (b) The transmission probability $T$ is plotted versus a dimensionless
536: wave number $\kappa$ defined by $\kappa \equiv \sqrt {2mE} W / (\pi
537: \hbar)$, for the case shown in (a) (solid line) and the case with an
538: additional $\delta$-potential $V(x,y) = v_0 \delta (x-W/6) \delta
539: (y-W/6)$ (dashed-dotted line).
540: In this energy range, LSCs (indicated by dotted lines) are found only in
541: the case of no $\delta$-potential. The pairs of integers in the form
542: $[m_x,m_y]$ indicate the profiles of the LSCs.
543: }
544:
545: \label {waveguide}
546: \end {figure}
547: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
548:
549:
550: We discretize the continuous space into a square lattice, and the TISE
551: becomes a set of simultaneous FD equations that read \cite {Dat95,FG97} $-
552: t (\psi _{{\bf i}-\hat x} + \psi _{{\bf i}+\hat x} + \psi _{{\bf i}-\hat
553: y} + \psi _{{\bf i}+\hat y} ) + (V_{\bf i}-E+2t) \psi _{\bf i} = 0$, where
554: $E$ is the energy, $t \equiv \hbar^2/(2ma^2)$ ($a$ is the distance between
555: two nearest sites), $\hat x$ and $\hat y$ are respectively the unit
556: vectors along the $x$ and $y$ directions, ${\bf i} \equiv (i_x,i_y)$
557: (where $i_x$ and $i_y$ are integers), and $\psi_{\bf i}$ and $V_{\bf i}$
558: are respectively the wave function and potential at ${\bf i} a$.
559: The FD equations are solved like in the first example.
560:
561:
562: In a lead labeled by $\eta$ ($\eta=$ I or II), taking the integer
563: $i_x^\eta$ ($i_y^\eta$) as the longitudinal (transverse) coordinate, the
564: wave function at energy $E$ for $V_{\bf i}=0$ is \cite {Dat95,FG97}
565: \begin {eqnarray}
566: \psi ^\eta (i_x^\eta, i_y^\eta) &=& \sum_{m=1}^{N_ \eta} \left( A_m ^\eta
567: e ^{ik ^\eta _m i_x^\eta a} + B_m ^\eta e ^{-ik ^\eta _m i_x^\eta a}
568: \right) \nonumber\\
569: && \times {1 \over {\sqrt {N _\eta + 1}}} ~ {\rm sin} \left[ { {m\pi}
570: \over {N_\eta + 1} } (i_y^\eta+1) \right], ~~~
571: \end {eqnarray}
572: where $A_m ^\eta$ and $B_m ^\eta$ are arbitrary complex numbers, $i_y^\eta
573: = 0 \sim N_\eta-1$ [where $W \equiv (N_\eta + 1) a$], and $k^\eta _m$ is
574: defined by $E \equiv -2t \{ 2 - {\rm cos} (k ^\eta _m a ) - {\rm cos}
575: [{m\pi} / (N _\eta + 1)] \} $, where the cosine is defined by ${\cos } ~
576: \xi \equiv (e ^ {i\xi} + e ^{-i\xi} ) / 2$ for a complex $\xi$, and the
577: phase of $k_m ^\eta$ is chosen such that $A_m ^\eta$ is the amplitude of a
578: wave propagating or exponentially decaying inward.
579: The multiple transverse modes for the transverse coordinate $i_y^\eta$ is
580: a consequence of the quasi-one-dimensionality.
581:
582:
583: Taking the ingoing amplitudes $\{ A_m ^\eta \}$ as the input, the unknowns
584: will be $\{ B _m ^\eta \}$ (amplitudes of the waves propagating or
585: exponentially decaying outward) and the point-wise wave function in the
586: cavity $\psi _{\rm cavity}$.
587: The FD equations centered at the sites within the cavity, and the
588: uniqueness requirement of the wave function at the interfaces between the
589: cavity and the leads results in a matrix equation of the form
590: \begin {eqnarray}
591: M (E) \cdot | \{ B_m^\eta \}, \psi _{\rm cavity} \rangle = | \{ A_m^\eta
592: \} \rangle,
593: \label {2dguide}
594: \end {eqnarray}
595: where $| \{ A_m^\eta \} \rangle$ and $| \{ B_m^\eta \}, \psi _{\rm cavity}
596: \rangle$ are respectively the known and unknown column matrices, and
597: $M(E)$ is a square matrix whose determinant ${\rm det} M(E)$ may vanish
598: and imply LSCs in the system.
599:
600:
601: When there are no ingoing waves (i.e., $A_m ^\eta = 0$ for all $\eta$ and
602: $m$) and hence $| \{ A_m^\eta \} \rangle = 0$, a nontrivial solution for
603: $| \{ B_m^\eta \}, \psi _{\rm cavity} \rangle$ can be obtained if ${\rm
604: det} M(E) = 0$. This is necessarily a localized state since
605: the $B _m ^\eta$'s for the outward propagating waves necessarily vanish
606: due to unitarity, leaving only the possibility of nonvanishing $B _m
607: ^\eta$'s for the outward exponentially decaying waves. If ${\rm det} M(E)
608: = 0$ occurs at an energy $E_0$ in the continuum, a complete solution at
609: $E_0$ is a superposition of the localized state found by ${\rm det} M(E) =
610: 0$, and the extended states found by inverting $M(E)$ at $E_0+\delta$,
611: $\delta \rightarrow 0$.
612:
613:
614: Note that to find a nontrivial solution for the column vector $| \phi
615: \rangle$ in an equation $S | \phi \rangle = 0$, where $S$ is a square
616: matrix, is to find an eigenvector of $S$ that corresponds to a vanishing
617: eigenvalue (since the equation is just $S | \phi \rangle = {0 \cdot
618: | \phi \rangle }$). The eigenproblem can be solved by a numerical package
619: such as EISPACK (http://www.netlib.org/).
620: In general, there can be simultaneously more than one eigenvectors having
621: vanishing eigenvalues, and these eigenvectors are the degenerate localized
622: states in the original problem.
623:
624:
625: Figure \ref {waveguide}(b) shows the transmission probability $T$ for a
626: particle with an energy $E$, injected from the first subband in lead I,
627: and passed to the first subband in lead II. Letting $A_m^\eta =
628: \Delta_{\eta,{\rm I}} \Delta_{m,1}$, $T$ is given by $T \equiv |B_1^{\rm
629: II}|^2$.
630: We choose $V_{\rm G} = - 15 \hbar^2 / (m W^2)$, and consider the
631: cases with and without an additional perturbing $\delta$-potential $V
632: (x,y) = v_0 \delta (x-W/6) \delta (y-W/6)$ in the cavity, \cite {VC06h}
633: where $v_0 = 5 \hbar^2 / m$, and $x$ and $y$ are the coordinates defined
634: in Fig.~\ref {waveguide}(a).
635:
636:
637: In the considered energy range in Fig.~\ref {waveguide}(b), zeroes of
638: ${\rm det} M(E)$ \cite {VC06a} or LSCs are found only for the case without
639: a perturbing $\delta$-potential. Note that LSCs can also exist at energies
640: beyond the first subband (at $\kappa > 2$).
641: We have used $N_\eta = 17$ in Fig.~\ref {waveguide}(b), and since we find
642: the LSCs qualitatively the same as those in the calculation using $N_\eta
643: = 8$, \cite {VC06} we believe they will survive in the continuous space
644: limit.
645: As usual, Fano resonances appear when the LSCs are destroyed.
646: The locations of the resonances depend on the details of the perturbing
647: potential, but when the perturbing potentials are vanishingly small, the
648: resonances are always found with infinitesimal widths on the locations of
649: the LSCs. A noteworthy point is the $\delta$-potential does not affect a
650: LSC when it is on a node of it.
651:
652:
653: The LSCs in Fig.~\ref {waveguide}(b) have profiles with exponential tails
654: in the leads, and resemble the fictitious standing waves ${\rm sin} (m_x
655: \pi x / W) \cdot {\rm sin} (m_y \pi y / W)$ in the cavity, where $m_x$ and
656: $m_y$ are integers. Therefore we will use $[m_x,m_y]$ to label the LSCs
657: for the convenience in our discussion. The four LSCs in Fig.~\ref
658: {waveguide}(b), from the lowest to the highest energy, resemble standing
659: waves with $[m_x,m_y] = [1,2]$, [2,2], [1,3], and [2,3] respectively.
660: The probability densities or squares of the absolute values of the wave
661: functions of two of the LSCs are shown in Fig.~\ref {psi2}. The energy of
662: a LSC is found to be always red shifted from the energy of the
663: corresponding fictitious standing wave in the cavity $E_{\rm SW}
664: (m_x,m_y)$, where $E_{\rm SW} (m_x,m_y) = V_{\rm G} + (m_x^2 + m_y^2)
665: \pi^2 \hbar^2 / (2m W^2)$.
666: Moreover, it seems that the larger the exponential tails, the more the
667: red shift [comparing Figs.~\ref {psi2}(a) and \ref {psi2}(b)]. This is
668: conceivable since the tails lead to a lowering of the kinetic energies.
669:
670: % a comparison of the amounts of redshift is dangerous, since the
671: % energies are not close!
672:
673: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
674: \begin {figure}
675:
676: %\special {psfile=lsc_fig4.EPS hoffset=10cm voffset=-380cm hscale=150cm
677: %vscale=150cm}
678: %\vspace {12cm}
679:
680: \special {psfile=lsc_fig4.EPS hoffset=-50cm voffset=-310cm hscale=120cm
681: vscale=120cm}
682: \vspace {9cm}
683:
684: \caption {
685: The probability densities of two of the LSCs in Fig.~\ref {waveguide}(b)
686: are shown in a section of the waveguide with the cavity at the center. The
687: length of the section is three times of the width.
688: (a) The [2,2] LSC at $\kappa \simeq 1.74$ has sizable exponential tails in
689: the leads. Correspondingly, it has also a considerable red shift from
690: $\kappa _{\rm SW} (2,2) \simeq 2.23$, where $\kappa _{\rm SW} (m_x,m_y)$
691: is defined by $\kappa_{\rm SW} (m_x,m_y) \equiv \sqrt {2mE_{\rm SW}
692: (m_x,m_y)} W / (\pi \hbar)$.
693: (b) The [1,3] LSC at $\kappa \simeq 2.51$ has smaller exponential tails in
694: the leads, and also a lesser red shift from $\kappa_{\rm SW} (1,3) \simeq
695: 2.64$.
696: }
697:
698: \label {psi2}
699: \end {figure}
700: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
701:
702:
703: The presence of the above LSCs can be understood by an intuitive picture.
704: Notice that the first two LSCs which resemble the $[m_x,m_y]=$ [1,2] and
705: [2,2] standing waves, are embedded only in the first subband. Since their
706: transverse wave functions resemble ${\rm sin} (2 \pi y / W)$ and are
707: orthogonal to the first transverse modes in the leads ${\rm sin} (\pi y /
708: W)$, the standing waves are trapped in the cavity.
709: With this picture, the absence of a LSC with $m_y=1$, such as a LSC with
710: $[m_x,m_y] = [2,1]$ or [3,1] is conceivable.
711: The two higher energy LSCs which resemble the standing waves with
712: $[m_x,m_y] = [1,3]$ and [2,3] are embedded in both the first and second
713: subband. Likewise, the trapping can be understood by the observation that
714: their transverse wave functions with $m_y=3$ are orthogonal to the
715: transverse modes in the leads with $m_y=1$ and 2. In this energy range we
716: do not find LSCs with $m_y=1$ and 2, such as [3,1] and [3,2]. \cite
717: {VC06f,SRW89,KSJ99,SBR06}
718:
719:
720: The LSCs as the eigenstates of $M(E)$ are orthogonal to each
721: other, though the $M(E)$ in Eq.~(\ref {2dguide}) is non-Hermitian in
722: general. The orthogonality can be argued from the fact that the LSCs are
723: localized and are not affected by a truncation of the leads at distances
724: far away from the cavity. Since the LSCs form a subset of the set of
725: eigenenergy states of the Hermitian Hamiltonian matrix for the truncated
726: (closed) system, they are orthogonal.
727:
728:
729: \section {Concluding Remarks}
730:
731: The behaviors of the LSCs in open low-dimensional systems have been
732: illustrated by examples of various kinds. These LSCs are obtained by
733: studying the zeroes of the determinant of the matrix to be inverted in a
734: considered problem. A zero corresponds to at least one localized state.
735: The crucial factor in the formation of these LSCs is the
736: low-dimensionality of the leads, and the ``symmetricity'' in the systems
737: is not a necessary condition. \cite {VC06d}
738:
739:
740: For the TB molecule and quantum graph, the one-dimensionality of the leads
741: enables them to attach just at the nodes of a LSC in the scattering
742: region and thereby leaving the LSC intact.
743: The same argument also holds for the case of higher dimensional leads with
744: 1D constrictions at the ends joining the scattering region.
745: For the case of Q1D leads, the delocalization of a standing wave in the
746: cavity can be prohibited by the non-overlapping of its transverse wave
747: function and the transverse modes in the leads, and the standing wave is
748: turned into a LSC. \cite {VC06b}
749:
750:
751: In view of the possibility of such LSCs not only in the case of idealized
752: 1D leads but also in the case of Q1D leads, such LSCs may exist or may be
753: realizable in, e.g., mesoscopic structures \cite {GGH00} and nanobridges
754: or molecular junctions, \cite {SRG05} and may not of academic interest
755: only.
756:
757:
758: \indent {\bf Acknowledgments -}
759: This work is supported by the National Science Council of Taiwan under
760: Grant No. 94-2112-M-009-017, and we also thank S.W. Chung for helpful
761: discussions.
762:
763:
764: \begin {thebibliography} {99}
765:
766: \bibitem [*] {coraut} Corresponding author (Email: kkvoo@cc.nctu.edu.tw).
767:
768: \bibitem {NW29} J. von Neumann and E. Wigner, Phys. Z. {\bf 30}, 465
769: (1929).
770:
771: \bibitem {SH75} F. H. Stillinger and D. R. Herrick, Phys. Rev. A {\bf
772: 11}, 446 (1975)
773:
774: \bibitem {FW85} H. Friedrich and D. Wintgen, Phys. Rev. A {\bf 31}, 3964
775: (1985).
776:
777: \bibitem {CFR03} L. S. Cederbaum, R. S. Friedman, V. M. Ryaboy, and N.
778: Moiseyev, Phys. Rev. Lett. {\bf 90}, 13001 (2003); and the references
779: therein.
780: % {CFR03} conical intersections and bound molecular states embedded in the
781: % continuum.
782:
783: \bibitem {CSF92} F. Capasso, C. Sirtori, J. Faist, D. L. Sivco, S.-N. G.
784: Chu, and A. Y. Cho, Nature {\bf 358}, 565 (1992); and the references
785: therein.
786:
787: \bibitem {Fan35} U. Fano, Nuovo cimento {\bf 12}, 156 (1935); Phys. Rev.
788: {\bf 124}, 1866 (1961).
789: %\bibitem {Fan61} U. Fano, Phys. Rev. {\bf 124}, 1866 (1961).
790:
791: \bibitem {ZCP02} Z. Y. Zeng, F. Claro, and A. Perez, Phys. Rev. B {\bf
792: 65}, 85308 (2002).
793:
794: \bibitem {VC06z} A numerical inversion blindly implemented by a computer
795: on a matrix with a vanishing determinant can be stable. This may be due to
796: the finite precisionness. For instance, a $V_2=0$ calculation where the
797: determinant can vanish, may be implemented as, say, a $V_2 \sim 10^{-8}t$
798: calculation where the determinant does not vanish, and the resulting fake
799: Fano resonance is too sharp to be noticeable.
800:
801: \bibitem {VC06c} Such Fano resonance has been seen in the theoretical
802: study of many systems (e.g., see Ref. \onlinecite {ZCP02}), but an
803: explicit solution in the form of Eq.~\ref {psi0} has not been reported.
804:
805: %\bibitem {Kuh49} H. Kuhn, Helv. Chim. Acta {\bf 32}, 2247 (1949).
806: %% the first to use the "griffith" scheme.
807: %
808: %\bibitem {Gri53} J. Stanley Griffith, Trans. Faraday Soc. {\bf 49}, 345
809: %(1953); {\it ibid}., {\bf 49}, 650 (1953).
810: %
811: %\bibitem {RS53} K. Ruedenberg and C. W. Scherr, J. Chem. Phys. {\bf 21},
812: %1565 (1953).
813: %% free-electron network model for conjugated systems.
814: %
815: %\bibitem {KS99} T. Kottos and U. Smilansky, Ann. of Phys. {\bf 274}, 76
816: %(1999).
817: %% periodic orbit theory and spectral statistics for quantum graphs.
818: %
819: %\bibitem {ES89} P. Exner and P. Seba, Rep. Math. Phys. {\bf 28}, 7
820: %(1989).
821: %% free quantum motion on a branching graph.
822:
823: %\bibitem {VC05} K.-K. Voo and C. S. Chu, Phys. Rev. B {\bf 72}, 165307
824: %(2005).
825:
826: \bibitem {VCT06} K.-K. Voo, S.-C. Chen, C.-S. Tang, and C.-S. Chu, Phys.
827: Rev. B {\bf 73}, 35307 (2006); and the references therein.
828:
829: \bibitem {Tex02} C. Texier, J. Phys. A {\bf 35}, 3389 (2002).
830:
831: \bibitem {SS92} J. Stein and H.-J. St$\ddot {\rm o}$ckmann, Phys. Rev.
832: Lett. {\bf 68}, 2867 (1992).
833:
834: \bibitem {Dat95} S. Datta, {\it Electronic Transport in Mesoscopic
835: Systems}, 1st ed. (Cambridge University Press, Cambridge, 1995).
836:
837: \bibitem {FG97} D. K. Ferry and S. M. Goodnick, {\it Transport in
838: Nanostructures}, 1st ed. (Cambridge University Press, Cambridge, 1997).
839:
840: \bibitem {VC06h} A $\delta$-potential in a continuous space $V_0 a^2
841: \delta (x - n_x a) \delta (y - n_y a)$ is modeled by a potential $V_0
842: \Delta _{i_x, n_x} \Delta _{i_y, n_y}$ in the discretized space.
843:
844: %When the sites are chosen dense, the size of the matrix to be inverted in
845: %a straightforward formulation is too large to be manageable by a
846: %computer. The formulation is replaced by a formulation where the
847: %``internal'' sites are eliminated by transfer matrices, and the resulting
848: %matrix to be inverted has a more manageable size.
849:
850: %\bibitem {VC06e} ${\rm det} M(E) = 0$ may also occur at energies outside
851: %the continuum, e.g., below the propagation threshold or in a gap. ${\rm
852: %det} M(E)$ is real in this regime and its zeroes correspond to the usual
853: %bound states.
854:
855: \bibitem {VC06a} Since ${\rm det} M(E)$ is complex when $E$ is in the
856: continuum, a zero here means only a simultaneous vanishing of the real and
857: imaginary parts within numerical precision. For Fig.~\ref {waveguide}(b),
858: the precision is a width $\delta \kappa < 10^{-3}$ on the $\kappa$-axis.
859: By the same token, any structure in $T$ sharper than $\delta \kappa$ will
860: not be seen.
861:
862: \bibitem {VC06} K.-K. Voo and C. S. Chu (unpublished).
863:
864: \bibitem {VC06f} The understanding discussed in this paragraph may be
865: related to the reports in Refs.~\onlinecite {SRW89,KSJ99,SBR06}.
866:
867: \bibitem {SRW89} R. L. Schult, D. G. Ravenhall, and H. W. Wyld, Phys. Rev.
868: B {\bf 39}, 5476 (1989).
869: % quantum bound states in a classically unbound system of crossed wires.
870:
871: \bibitem {KSJ99} C. S. Kim, A. M. Satanin, Y. S. Joe, and R. M. Cosby,
872: Phys. Rev. B {\bf 60}, 10962 (1999).
873: % transport thru a quasi-1D channel with a finite-size attractive well.
874:
875: \bibitem {SBR06} A. F. Sadreev, E. N. Bulgakov, and I.
876: Rotter, Phys. Rev. B {\bf 73}, 235342 (2006).
877: % BICs in open quantum billiards with a variable shape.
878:
879: \bibitem {VC06d} A discussion of LSCs in asymmetric Q1D waveguides will be
880: published elsewhere.
881:
882: \bibitem {VC06b} Similar phenomena may also appear in other wave systems.
883: Such as electromagnetic, acoustic, and water waves in waveguides, as the
884: transverse confinements also lead to transverse modes.
885:
886: \bibitem {GGH00} J. G$\ddot {\rm o}$res, D. Goldhaber-Gordon,
887: S. Heemeyer, M. A. Kastner, H. Shtrikman, D. Mahalu, and U. Meirav, Phys.
888: Rev. B {\bf 62}, 2188 (2000); and the references therein.
889: % fano resonances in electronic transport thru a SET. the first fano
890: % resonance in mesoscopic structures.
891:
892: \bibitem {SRG05} N. Sergueev, D. Roubtsov, and H. Guo, Phys.
893: Rev. Lett. {\bf 95}, 146803 (2005); and the references therein.
894:
895: \end {thebibliography}
896:
897: \end{document}
898:
899: