1: \documentclass[aps,twocolumn]{revtex4}

2: 

3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

5: 

6: \usepackage{graphicx}

7: \usepackage{color}

8: \usepackage{amsmath}

9: \usepackage{bm}

10: 

11: \begin{document}

12: \newcommand{\bea}{\begin{eqnarray}}

13: \newcommand{\eea}{\end{eqnarray}}

14: \newcommand{\be}{\begin{equation}}

15: \newcommand{\ee}{\end{equation}}

16: \newcommand{\bi}{\bibitem}

17: \newcommand{\D}{\mbox{$\Delta$}}

18: \newcommand{\mitD}{\mbox{${\mit \Delta}$}}

19: \renewcommand{\d}{\mbox{$\delta$}}

20: \renewcommand{\t}{\mbox{$\theta$}}

21: \renewcommand{\o}{\mbox{$\omega$}}

22: \renewcommand{\a}{\mbox{$\alpha$}}

23: \newcommand{\s}{\sigma}

24: \newcommand{\g}{\mbox{$\gamma$}}

25: \newcommand{\ep}{\mbox{$\epsilon$}}

26: \newcommand{\p}{\mbox{$\psi$}}

27: \newcommand{\bfk}{\mbox{${\bf k}$}}

28: \newcommand{\bfR}{\mbox{${{\bf R}}$}}

29: \newcommand{\bfr}{\mbox{${\bf r}$}}

30: \newcommand{\naR}{\mbox{$\nabla_{\bf R}$}}

31: \newcommand{\xipa}{\mbox{$\xi_{\parallel}$}}

32: \newcommand{\xipe}{\mbox{$\xi_{\perp}$}}

33: \newcommand{\xia}{\mbox{$\xi_{ani}$}}

34: \newcommand{\rr}{{\bf r}}

35: \newcommand{\kk}{{\bf k}}

36: \newcommand{\pp}{{\bf p}}

37: \newcommand{\qq}{{\bf q}}

38: \newcommand{\BSCCO}{{Bi$_2$Sr$_2$CaCu$_2$O$_8$ }}

39: \newcommand{\YBCO}{{YBa$_2$Cu$_3$O$_{7-\delta}$ }}

40: \newcommand{\nn}{\nonumber}

41: \newcommand{\G}{{\cal G}}

42: \newcommand{\bom}{{\bar\omega}}

43: \newcommand{\KK}{{\cal K} }

44: \newcommand{\EE}{{\cal E} }

45: \newcommand{\sgn}{{\rm sgn}}

46: \newcommand{\uG}{\underline{G}}

47: \newcommand{\ucG}{\underline{{\cal G}}}

48: \newcommand{\uGo}{\underline{G}^0}

49: \newcommand{\ug}{\underline{g}}

50: \newcommand{\uL}{\underline{L}}

51: \newcommand{\uM}{\underline{M}}

52: \newcommand{\uT}{\underline{T}}

53: \newcommand{\ut}{\underline{t}}

54: \newcommand{\uV}{\underline{V}}

55: \newcommand{\uVm}{\underline{V_1}}

56: \newcommand{\uU}{\underline{U}}

57: \newcommand{\uR}{\underline{R}}

58: \def\k{{\bf k}}

59: \def\r{{\bf r}}

60: \def\R{{\bf R}}

61: \def\p{{\bf p}}

62: \def\v{{\bf v}}

63: \def\D{{\cal D}}

64: %\twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse%

65: %\endcsname

66: 

67: \title{Electronic structure of $d$-wave superconducting quantum wires }

68: 

69: \author{A. M. Bobkov$^{1,2}$, L.-Y. Zhu$^{3}$, S.-W. Tsai$^{3, 4}$,

70: T. S. Nunner$^{3}$, Yu. S. Barash$^{1,2}$ and P. J. Hirschfeld$^{3}$\\~}

71: 

72: \affiliation{$^1$Lebedev Physical Institute, Leninsky Prospect 53,

73: Moscow 119991, Russia\\$^2$Institute of Solid State Physics,

74: Chernogolovka, Moscow reg., 142432 Russia \\$^3$Department of

75: Physics, University of Florida, Gainesville, FL

76: 32611\\$^4$Department of Physics, Boston University, Boston, MA

77: 02215}

78: 

79: \date{\today}

80: 

81: \begin{abstract}

82: We present analytical and numerical results for the electronic

83: spectra of wires of a d-wave superconductor on a square lattice.

84: The spectra of Andreev and other quasiparticle states, as well as

85: the spatial and particle-hole structures of their wave functions,

86: depend on interference effects caused by the presence of the

87: surfaces and are qualitatively different for half-filled wires

88: with even or odd number of chains. For half-filled  wires with an

89: odd number of chains $N$  at (110) orientation, spectra consist of

90: $N$ doubly degenerate branches. By contrast, for even $N$ wires,

91: these levels are split, and all quasiparticle states, even the

92: ones lying above the maximal gap, have the characteristic

93: properties of Andreev bound states. These Andreev states above the

94: gap can be interpreted as a consequence of

95: an infinite sequence of Andreev reflections experienced by

96: quasiparticles along their trajectories bounded by the surfaces of

97: the wire. Our  microscopic results for the local density of states

98: display atomic-scale Friedel oscillations due to the presence of

99: the surfaces, which should be observable by scanning tunneling

100: microscopy. For narrow wires the self-consistent treatment of the

101: order parameter is found to play a crucial role.  In particular,

102: we find that for small wire widths the finite geometry may drive

103: strong fluctuations or even stablilize exotic quasi-1D pair states

104: with spin triplet character.

105: \end{abstract}

106: 

107: %\pacs{74.25.Bt,74.25.Jb,74.40.+k}

108: 

109: \maketitle

110: 

111: \section{Introduction}

112: Since the discovery of high temperature superconductivity (HTS), the

113: origin of the pairing phenomenon in these materials has been the

114: subject of intense debate, and is still not clarified. Part of the

115: unusual nature of HTS which has hindered theoretical analysis is the

116: short coherence length, which allows short-wavelength fluctuations of

117: various types of local order to coexist with superconductivity. Most

118: probes of the nature of the superconducting state have been

119: restricted, until fairly recently, to measurements of bulk properties

120: and tunnelling through relatively large areas. Following the

121: pioneering work of Hess et al.\cite{Hess91}, it was realized that

122: scanning tunneling microscopy (STM) could provide an atomic-scale

123: picture of the superconducting state, particularly useful when

124: applied to inhomogeneous situations like the vortex lattice.

125: Measurements of this type were subsequently performed on

126: high-temperature superconductors \cite{Fischer95}. In the past few

127: years, scanning tunnelling microscopy on the surface of HTS have

128: compiled a novel and fascinating picture of the local electronic

129: structure of a few of these materials

130: \cite{yazdani,davisnative,davisZn,cren,davisinhom1,delozanne,howald,davisinhom2}.

131: In the Ba$_2$Sr$_2$CaCu$_2$O$_8$ system (BSCCO), one dramatic

132: implication of these experiments is that even relatively high

133: quality single crystals display inhomogeneous electronic structure

134: at the nanoscale \cite{davisinhom1,howald,delozanne,davisinhom2}.

135: 

136: In parallel to studies of the HTS materials, point contact

137: spectroscopy has been used to study the electronic structure of

138: ultrasmall  conventional superconducting

139: islands \cite{vonDelftreview}. Among many fascinating consequences

140: of the nanoscale geometry are number parity effects, in which the

141: qualitative electronic structure depends sensitively on whether

142: the number of electrons on the island are even or odd. More

143: recently,  superconducting wires of widths tens of

144: nanometers \cite{Tinkham} have also been fabricated.  Although

145: they have not yet been studied by STM or similar methods, this

146: should be technically feasible.

147: 

148: Isolated nanoscale grains and wires of HTS material have not been

149: fabricated to our knowledge. While this may prove technically

150: quite difficult to achieve due to the complexity of the crystal structure,

151: there seems to be no fundamental obstacle in the

152: long run.  This applies as well to other superconductors thought

153: to manifest unconventional superconducting order, where

154: effects of finite geometry should be easier to see since

155: coherence lengths tend to be larger.

156: 

157: It is our purpose in this paper to study how $d$-wave (e.g., HTS)

158: and other unconventional symmetry superconductors behave in finite

159: geometry at the atomic scale. Ziegler et al.\cite{zieglerwire}

160: began the study of this problem in the case of (100) $d$-wave

161: quantum wires, pointing out the dependence of the Fermi level

162: density of states on the parity of the wire width.   This is a

163: natural consequence of the discretization of the electronic energy

164: levels due to the finite wire width in the $d$-wave state.

165: While in the $s$-wave case all the interesting physics is tuned by

166: the level discretization, one expects a priori one  fundamental

167: difference in the $d$-wave case: for any geometry with surfaces

168: making an arbitrary angle with the crystal axes, pair-breaking

169: processes take place on a scale of the coherence length.

170: The most important consequence for the electronic structure should

171: be the formation of Andreev surface states.

172: 

173: However, little is known about how Andreev surface states behave

174: when the size of the superconductor becomes comparable to $\xi_0$.

175: The zero-energy states form on surfaces with orientations different

176: from the antinodal directions of a $d$-wave superconductor, due to

177: the sign change of the order parameter. In high-temperature superconductors,

178: such states manifest themselves as the zero-bias conductance peak

179: (ZBCP) in tunneling spectroscopy in the ab-plane \cite{geerk88,hu94,tan95,%

180: mats95,buch95,tan952,cov96,xu96,fog97,bbs97,cov97,alf97,ek97,ap98,alf981,%

181: alf982,sin98,wei98,apr99,deutsch99,Ting110,cov00,pairor02,greene02,wumou03,%

182: greene03}, the anomalous temperature behavior of the Josephson critical

183: current~\cite{tan96,bbr96,ilichev01,blamire03} and the upturn in the

184: temperature dependence of the magnetic penetration depth~\cite{walter98,%

185: bkk00,carr01} (see also review articles \cite{tan00,wendin01}).

186: The conventional description of Andreev surface states, as well as

187: the Andreev reflection itself, is based on the quasiclassical

188: approximation, a powerful tool in the theoretical study of various

189: properties of inhomogeneous superconducting systems.

190: 

191: The quasiclassical theory of superconductivity gives a so-called

192: coarse-grained description of the phenomena, averaged over

193: interatomic distances. This has been used, for example, to

194: calculate the local quasiparticle density of states (LDOS).

195: However, these coarse-grained averaged results are not adequate to

196: analyze atomic resolution measurements by STM and some other

197: contemporary experimental techniques (e.g. atomic force microscopy

198: \cite{giessibl}). To obtain this type of information, a fully

199: quantum mechanical atomic-scale approach going beyond the

200: quasiclassical approximation in describing inhomogeneous states of

201: superconductors is required. We address this problem in the

202: present paper using a tight-binding BCS-like model of $d$-wave

203: superconductor on a square lattice.

204: 

205: As a first step towards understanding the effect of constrained

206: geometry, we study the simplest case of $d$-wave superconducting

207: wires  consisting of $N$ parallel chains as the system size is

208: reduced. The  quasiparticle spectrum of such systems is described

209: both analytically, with an assumed spatially homogeneous order

210: parameter, and also numerically with a fully self-consistent

211: approach. In the limit $N\gg 1$ the usual surface Andreev states

212: in lattice models \cite{Ting110,tan00,pairor02}, as well as the

213: surface states known in continuous models \cite{bbs97}, can be

214: recovered at each surface of the wire. However, for sufficiently

215: narrow wires, when the transverse wire dimension is the order of

216: the superconducting coherence length, the Andreev states strongly

217: interfere and give rise to qualitatively new effects. We show

218: below that only those effects which occur for bands sufficiently

219: far from half-filling and relatively wide wires  can be described

220: with the quasiclassical theory of superconductivity. In addition,

221: we demonstrate how and under what  conditions one can recover

222: earlier quasiclassical results for Andreev states in $d$-wave

223: superconducting films \cite{nagai} from our microscopic approach.

224: 

225: The microscopic method adopted in this paper to constrain the

226:  geometry involves introducing lines of impurities of

227: potential strength taken to infinity to bound the wire.

228:  We  show that earlier microscopic results on (110) surfaces

229:  \cite{bbs97,Ting110,pairor02,Walker,tan00} and (100) wires

230: \cite{zieglerwire} can be reproduced by this technique, and then

231:  extend it to calculate new results on  wires with other

232:  orientations.  We find that the results for electronic spectra

233:  are very sensitive to the number parity of the wire width, and

234:  that true zero-energy Andreev states can only exist in wires with

235:  odd numbers of chains.  In even wires, the Andreev states are

236:  split,

237:  pushed away from the Fermi energy, and can have either surface or

238:  standing wave character.   Finally, we show that for smaller

239:  wires self-consistency effects become important and can even,

240:  within mean field theory, lead to condensation of a fully gapped

241:  spin triplet state instead of $d$-wave order.

242: 

243: Our results may also have some qualitative relevance for the

244: related problem of $d$-wave superconducting grains, where the

245: geometry is constrained in 2 dimensions.  In addition to

246: artificially fabricated islands, some authors have proposed that

247: the BSCCO-2212 samples which display nanoscale inhomogeneity

248: should be thought of as a collection of weakly coupled $d$-wave

249: grains of roughly the size of the superconducting coherence length

250: $\xi_0$, or $d$-wave grains coupled to grains of another

251: electronic phase \cite{davisinhom1,howald,davisinhom2,Joglekar}.

252: In fact, the structure of a general, possibly irregular small

253: grain of $d$-wave superconductor has not been studied to our

254: knowledge. Understanding how the local density of states (LDOS) of

255: these wires depend on the wire width and the orientation, as well

256: as on the deviation from half-filling, could provide important

257: intuition for the question of the electronic structure of the

258: small irregular grains possibly present in BSCCO samples.

259: 

260: The outline of the paper is as follows. In Sec.\ref{sec:Tmatrix},

261: we introduce the formalism for the problem.  In

262: Sec.\ref{sec:surface}, we discuss three special semi-infinite

263: surface orientations: $(110)$, $(210)$ and $(100)$. In Sec.\ref

264: {sec:swires}, we study the $(110)$ superconducting wires with even

265: and odd width and make an comparison with the discrete states in

266: normal metal wires (Sec.\ref{sec:nwires}) to try to identify the

267: nature of the true Andreev states. In Sec.\ref{sec:finitemu}, we

268: study the effects of deviations from the half-filling. In

269: Sec.\ref{sec:self} the results of fully self-consistent

270: calculations are presented. In particular, we allow the order

271: parameter to vary spatially and comment on the differences in our

272: results. In the case of narrow wires the self-consistent study

273: shows the appearance of some peculiar types of superconducting

274: pairing. Finally, in Sec.\ref{sec:conclusion}, we present our

275: conclusions.

276: 

277: \section{Model description and formalism}

278: \label{sec:Tmatrix}

279: The Hamiltonian for a pure singlet superconductor

280: can be written as:

281: \begin{eqnarray}

282: \cal{H} &=& - t \sum_{\langle i,j\rangle, \sigma}

283: c^\dagger_{i\sigma} c_{j\sigma} - \sum_{i,\sigma} [\mu - U_i]

284: c^\dagger_{i\sigma} c_{i\sigma} \nonumber \\ &+& \sum_{\langle

285: i,j\rangle} \{ \Delta_{ij} c^\dagger_{i\uparrow}

286: c^\dagger_{j\downarrow} + h.c. \}, \label{ham}

287: \end{eqnarray}

288: where we have chosen a  nearest neighbor tight-binding band for

289: simplicity; $\mu$ is the chemical potential.

290: A superconducting pairing is defined for nearest neighbors

291: $\Delta_{ij} = -V \langle c_{j\downarrow}c_{i\uparrow}

292: \rangle$ on the bond $\{i,j\}$. The parameter $t$ is of order 150

293: meV for high $T_c$ materials, and we consider both the

294: particle-hole symmetric model $\mu=0$ and the more realistic case

295: of finite $\mu$. In non-self-consistent calculations the OP has

296: the familiar $k$-space form $\Delta_\k = \Delta_0 [ \cos(k_aa) -

297: \cos(k_ba) ]$, where $ \Delta_0 = \frac{1}{2}\sum_\pm (

298: \Delta_{i\,i\pm r_a} - \Delta_{i\,i\pm r_b} ) $ is independent of

299: $i$, and  is taken to be $0.2t$. The lattice constant is denoted by

300: $a$. The maximum gap is $\Delta_{max}=2\Delta_0$. We also present

301: self-consistent calculations, in which case $V$ is chosen to yield

302: this same value of $\Delta_0$ far from wire edges.

303: 

304: It is possible to constrain the geometry underlying Eq.

305: (\ref{ham}) in several different ways. We present results here for

306: a method discussed, for example, in Refs.\

307: \onlinecite{MorrDemler,Ting110} in which the on-site potentials

308: $U_i$ are chosen to lie on the boundary and their value is taken

309: to infinity to cut off electron transport through the boundary.

310: This technique has the virtue that the strength of the barrier can

311: in principle be lowered to allow different degrees of transparency

312: and the study of tunneling phenomena. In this work we restrict

313: ourselves to impurity configurations and strengths for which the

314: constrained system is completely isolated from its environment.

315: This is equivalent to the assumption of open boundaries, when no

316: hopping and no pairing take place outside the region. No disorder

317: is introduced in the system and we consider $U_i$ as ``impurity''

318: potentials only so as to form the surfaces of the superconducting

319: region.

320: 

321: The Hamiltonian of the surface term is taken to be \be U

322: =U_0\sum_{\ell} c_\ell^\dagger c_\ell, \ee where the set of sites

323: $\ell$ is determined exclusively by the boundaries of the desired

324: system (see below). The full Fourier space Green's function for

325: the system in the presence of these impurities is quite generally:

326: \begin{eqnarray}

327: \check{G}(\k,\k^{\prime},\omega) &=& \check{G}^{(0)}(\k,\omega)

328: \delta_{\k,\k^{\prime}} +\nonumber\\

329: &+&\check{G}^{(0)}(\k,\omega)

330: \check{T}(\k,\k^{\prime},\omega) \check{G}^{(0)}(\k^{\prime},\omega)

331: \enspace ,

332: \label{greens}

333: \end{eqnarray}

334: where the $T$-matrix can be found from the following equations:

335: \begin{eqnarray}

336: \check{T}(\k,\k^{\prime},\omega)&=&\check{U}(\k,\k^{\prime})+\nonumber\\

337: &+&\sum_{\k^{''}} \check{U}(\k,\k^{''})

338: \check{G}^{(0)}(\k^{''},\omega)

339: \check{T}(\k^{''},\k^{\prime},\omega) . \label{tmat}

340: \end{eqnarray}

341: Here $\check{G}$ and $\check{T}$ take $4\times4$ matrix form in

342: the four-dimensional product space of particle-hole and spin

343: variables. If we choose nonmagnetic on-site potentials and

344: consider singlet superconductors, the problem reduces to $2\times

345: 2$ matrices in Nambu space. The Nambu retarded propagator for the

346: pure $d$-wave superconductor is

347: 

348: \begin{equation}

349: \hat{G}^{(0)}(\k,\omega)={\omega \hat{\tau}_0 +\xi_\k \hat{\tau}_3

350: + \Delta_\k\hat{\tau}_1 \over (\omega+i0)^2 -

351: \xi_\k^2-\Delta_\k^2} \enspace ,

352: \end{equation}

353: where the $\tau_\alpha$ are the Pauli matrices and

354: $\xi_{\bf k}=-2t[\cos(k_aa)+\cos(k_ba)]-\mu$.

355: Calculating the  $T$-matrix allows us to  obtain the

356: eigenenergies of the system from the condition

357: ${\rm det }\hat{T}^{-1}(\omega)=0$.

358: 

359: The local spin-resolved quasiparticle density of states is given as

360: \begin{eqnarray}

361:     \rho_{\uparrow(\downarrow)}({\bf r},\omega) &=& -\pi^{-1}\mbox{Im }

362:     G_{11,\uparrow\uparrow(\downarrow\downarrow)}({\bfr},\r,

363: \omega) \enspace .

364: \end{eqnarray}

365: After integration of the LDOS over energy, we should

366: obtain the total number of quasiparticle states per one site.

367: Since each site on the lattice possesses 2 states with opposite

368: spins, the spin-resolved LDOS normalization is

369: 

370: \begin{equation}

371: \int_{-\infty}^\infty d\omega \rho_{\uparrow (\downarrow)}(\r,\omega) = 1.

372: \end{equation}

373: 

374: 

375: \section{surface case }

376: 

377: \label{sec:surface}

378: 

379: Upon a conventional reflection on the (110) surface of a $d$-wave

380: superconductor, the order parameter always changes sign as the

381: direction of the quasiparticle momentum $\k$ is varied. This leads

382: to  Andreev reflection and, eventually, to the formation of the

383: dispersionless zero-energy surface Andreev bound states. For other

384: surface orientations the sign change does not take place for all

385: incoming momentum directions.

386:  It is important to notice that the number of consecutive

387: impurity lines that are needed in order to cut the system depends

388: on the orientation of the surface one wants to consider. For (100)

389: and (110) surfaces and nearest neighbor hopping, one line of

390: impurities is sufficient to cut communication between the two

391: sides. For a (210) surface, the nearest neighbor hopping and

392: pairing terms can still connect this particle with another across

393: a single impurity line, so for a simple tight-binding band,  two

394: consecutive lines are needed to close the system (Fig.

395: \ref{fig:surf210}). Alternatively, if one includes a next-nearest

396: neighbor hopping $t'$, even a (110) surface is not closed by a

397: single line of infinite impurities; the system considered in Ref.

398: \cite{MorrDemler} is therefore not a closed (impenetrable)

399: surface. For a ($hl0$) surface in general, max({$h,l$}) lines are

400: needed for a model that includes nearest-neighbor terms only, and

401: $h+l$ lines are needed for a model that includes

402: next-nearest-neighbor terms. Clearly, the technique becomes

403: cumbersome for arbitrary angles, but one can nevertheless learn a

404: good deal by considering special cases.

405: 

406: 

407: \begin{figure}[tbh]

408: \begin{center}

409: \leavevmode

410: \includegraphics[width=.6\columnwidth]{surface1.eps}

411: \caption{(210) surface or wire. Two lines of impurities (closed

412: squares) are needed to isolate the lattice sites (closed circles)

413: with  nearest neighbor hopping.} \label{fig:surf210}

414: \end{center}

415: \end{figure}

416: 

417: 

418: 

419:  In the presence of a surface of arbitrary orientation, the

420: simplest way of applying Bloch's theorem to this discrete system

421: is by using a surface-adapted Brillouin zone\cite{pairor02,Walker}.

422: We define new coordinates ($\hat{x}$, $\hat{y}$), rotated with

423: respect to the crystal axes ($\hat{a}$, $\hat{b}$), where

424: $\hat{x}$ is the direction normal to the surface and $\hat{y}$ is

425: the direction along the surface. The system is periodic

426: along the $y$-direction and the crystal momentum component $k_y$

427: of a quasiparticle is conserved. Instead of the usual

428: square Brillouin zone $k_a = [-\pi, \pi]$, $k_b = [-\pi, \pi]$

429: (for unit lattice constant $a=1$) we now use the surface-adapted

430: Brillouin zone given by $k_x =[-\pi/d, \pi/d]$ and

431: $k_y = [-\pi d, \pi d]$. Here $d =1/\sqrt{h^2+l^2}$ is the distance

432: between the nearest chains (layers) aligned along the surfaces.

433: The momenta in the two coordinate systems are simply related

434: through rotation of an angle $\theta = \tan^{-1} h/l$.

435: 

436: Now we turn to the solution of the equation for the $T$-matrix

437: (Eq. \ref{tmat}) for the case of one line of impurities. We start

438: with the ansatz

439: \begin{eqnarray}

440: \hat{T} = T_0 \hat{\tau}_0 + T_1 \hat{\tau}_1 + T_3 \hat{\tau}_3

441: \end{eqnarray}

442: and find, for arbitrary strength of impurity potential,

443: \bea T_0(k_y, \omega) &=& {{G}_0^{(0)}(0, k_y, \omega) \over

444: D_1} \ , \\

445: T_1(k_y, \omega) &=& {-{G}_1^{(0)}(0, k_y, \omega) \over D_1} \ , \\

446: T_3(k_y, \omega) &=& {c-{G}_3^{(0)}(0, k_y, \omega) \over D_1}

447: \eea

448: where $D_1(k_y, \omega) = [c-{G}_3^{(0)}(0, k_y, \omega)]^2 -

449: {G}_0^{(0)}(0, k_y, \omega)^2 + {G}_1^{(0)}(0, k_y, \omega)^2$, $c

450: = 1/V_0$  and ${G}_i^{(0)}(x, k_y, \omega)$ is the Fourier

451: transform with respect to $k_x$ of the $i$-th Nambu component of

452: the bare Green's function $G_i^{(0)}(\k, \omega)$

453: \bea

454: {G}_i^{(0)}(n, k_y, \omega) =\frac{d}{2\pi}

455: \!\!\int\limits_{k_x=-\pi/d}^{\pi/d}\!\! G^{(0)}_i (k_x, k_y, \omega) e^{i

456: k_x n d} dk_x\ .

457: \label{gtilde}

458: \eea

459: The site index $n$

460: corresponds to $x$-coordinate $x= nd$. For the case of infinitely

461: strong impurity potential considered here, $c = 0$. In this case

462: the expression for ${\hat T}$ can be written in the compact form \bea

463: \hat{T}(k_y, \omega) = - \left[\hat{{G}}^{(0)}(0, k_y, \omega)

464: \right]^{-1} \label{t} \eea and the poles of the $T$-matrix

465: correspond to zeros of the determinant of the Green's function

466: $\hat{{G}}^{(0)}(0, k_y, \omega)$.

467: 

468: The Fourier transform with respect to $k_x$ of Eq. \ref{greens} is

469: \bea &&\hat{{G}}(n, n^{\prime}, k_y, \omega) = \hat{{G}}^{(0)} (n

470: - n^{\prime}, k_y, \omega) -

471: \\ \nonumber

472: &&\ \ \ \ \ \ \  \ \ \hat{{G}}^{(0)}(n, k_y, \omega)

473: \left[\hat{{G}}^{(0)}(0, k_y, \omega)\right]^{-1}

474: \hat{{G}}^{(0)}(-n^{\prime}, k_y, \omega) \label{gn} \eea Due to

475: periodicity of the system along the $y$-direction, calculation

476: of the LDOS at site $n$ will simply involve a sum of ${G}(n,n,

477: k_y, \omega)$ over all values of $k_y$ within the surface-adapted

478: Brillouin zone and over two spin directions:

479: \begin{equation}

480: \rho(\r,\omega)=\rho(n,\omega)= -\frac{2}{\pi}{\rm Im} \int_{-\pi

481: d}^{\pi d}\frac{dk_y}{2\pi d} {G}_{11}(n,n,k_y,\omega)

482: \label{11}

483: \end{equation}

484: 

485: \subsection{(100) surface}

486: 

487: Since (100) surfaces are not pairbreaking in $d$-wave

488: superconductors, we do not {\it a priori} expect to see

489: interesting physics arising from Andreev states. On the other

490: hand, the mere presence of a surface can induce surface Tamm bands

491: \cite{Tamm}, decaying in the bulk on the atomic scale and experiencing

492: the Friedel-like oscillations of the LDOS. In our model Tamm states

493: have nothing to do with the superconductivity, although the

494: possibility for pairing of electrons occupying these surface

495: states is not excluded \cite{ginzburg}. While it is not our intent

496: to study these in detail, we present some results to show that

497: such states can be seen by the STM even in situations when surface

498: Andreev states are absent.  Fig.\ref{fig:surf100}  shows on upper

499: and lower panels the  LDOS for various distances from the surface

500: in the normal metal and the superconducting

501: states, respectively.

502: 

503: \begin{figure}[tbh]

504: \begin{center}

505: \leavevmode

506: \includegraphics[clip=true,width=.9\columnwidth]{surface100_N.eps}

507: \includegraphics[clip=true,width=0.9\columnwidth]{surface100_S.eps}

508: \caption{ Local density of states $\rho(x,\omega)$ for a (100)

509: surface. Normal state (upper panel) and superconducting state (lower panel),

510: $\rho(x,\omega)$ vs. $\omega/\Delta_{max}$ for various

511: distances $n=x/d$ from surface; $\Delta_{max}=0.4t$, $\mu=0$.}

512: \label{fig:surf100}

513: \end{center}

514: \end{figure}

515: 

516: 

517: \subsection{(110) surface}

518: 

519: Results for a (110) surface are expected to be qualitatively

520: different from the ones obtained for a (100) surface. The bulk

521: order parameter in the coordinate system of the crystal axes $\hat

522: a$ and $\hat b$ is $\Delta_\k =\Delta_0(\cos k_x a -\cos k_y a)$,

523: but in the coordinates of the surface $\hat x$ (perpendicular to

524: surface) and $\hat y$ (parallel to surface) becomes $\Delta_\k = 2

525: \Delta_0 \sin k_x d \sin k_y d$, with $d=a/\sqrt{2}$.

526:  In this

527: case, an incident particle with any non-zero $k_y$ experiences a

528: sign change in the order parameter as it reflects from the

529: surface.

530: 

531: \begin{figure}[tbh]

532: \begin{center}

533: \leavevmode

534: \includegraphics[clip=true,width=.9\columnwidth]{surface110_S.eps}

535: \caption{Local density of states $\rho(x,\omega)$ in

536: superconducting state versus $\omega/\Delta_{max}$ for the case of

537: one (110) surface, $\Delta_0=0.2t$, $\mu=0$. Left: chains an {\it

538: odd} distance $n=x/d$ from surface at $x=0$. Right: chains an {\it

539: even} distance from $x=0$. } \label{fig:surf110}

540: \end{center}

541: \end{figure}

542: 

543: Calculation of the bare Green's function $\hat{{G}}^{(0)}(n, k_y,

544: \omega)$ from Eq. \ref{gtilde} gives at $\mu=0$, $G_{1,3}(n, k_y,

545: \omega)=0$ for even $n$ and $G_{0}(n, k_y, \omega) = 0$ for odd

546: $n$. Explicitly,

547: \begin{eqnarray}

548: \hat{{G}}^{(0)}(n=2m,k_y,\omega)=\qquad\qquad\qquad\qquad\qquad\qquad

549: &\nonumber

550: \\

551: \qquad\qquad -\frac{i|\omega|\exp\left\{-i|n|z\

552: \mathrm{sgn}\left[\omega(q^2-\Delta^2)\right]\right\}}

553: {\sqrt{(\omega^2-\Delta^2)(q^2-\omega^2)}}\hat{\tau_0}

554: \label{geven}

555: \end{eqnarray}

556: and \bea \hat{{G}}^{(0)}(n=2m+1,k_y,\omega)=\qquad \qquad

557: \qquad \qquad \qquad \qquad\nonumber \\

558: =\frac{i\Delta \exp\left[-i|n|z\

559: \mathrm{sgn}\left(\omega(q^2-\Delta^2)\right)\right]\mathrm{sgn}

560: \left(n(q^2-\Delta^2)\right)}

561: {\sqrt{(q^2-\Delta^2)(\omega^2-\Delta^2)}} \hat{\tau}_1\nn\\

562: +\frac{iq \exp\left[-i|n|z\

563: \mathrm{sgn}\left(\omega(q^2-

564: \Delta^2)\right)\right]\mathrm{sgn}(\omega)}{\sqrt{(q^2-\Delta^2)

565: (q^2-\omega^2)}} \hat{\tau}_3 \quad , \qquad \label{godd} \eea

566: 

567: where

568: 

569: \bea

570: z(k_y) &=& \tan^{-1}\sqrt{{q^2-\omega^2 \over \omega^2- \Delta^2 }}

571: \label{zomega} \enspace ,\\

572: \Delta(k_y) &=& 2 \Delta_0 \sin(k_y d) \label{dky} \enspace ,\\

573: q(k_y) &=& 4 t \cos(k_yd) \ , \label{qky} \eea and $\Delta(k_y)$

574: and $q(k_y)$ are the maximum gap $\Delta_\k$ and single-particle

575: spectrum $\xi_\k$ for fixed $k_y$ in 110 geometry. Note that here

576: the square root function takes positive values for positive

577: arguments, i.e. under the conditions

578: $|\Delta(k_y)|<|\omega|<|q(k_y)|$ or

579: $|q(k_y)|<|\omega|<|\Delta(k_y)|$. For $A^2-\omega^2<0$,\,the

580: branch $\sqrt{A^2-\omega^2}\to -i\ \mathrm{sgn}\omega\

581: \sqrt{\omega^2-A^2}$, for either $A=\Delta, q$.  In Fig.

582: \ref{fig:surf110} we show the LDOS spectra on the different

583: layers. Each layer is defined as an array of sites parallel to the

584: surface. Its index indicates its position; layer $n$ corresponds

585: to sites at a distance $nd$ away from the surface.

586: 

587: The Andreev bound states are manifested as zero-energy peaks in

588: the LDOS. For $\mu=0$, such  states are found in all odd layers,

589: and are absent on even ones, as seen in Figure \ref{fig:Andreev}.

590: This even-odd effect can be easily understood from the form of the

591: $T$-matrix and Green's functions Eqs. (\ref{t}-\ref{godd}).

592: \begin{figure}[tbh]

593: \begin{center}

594: \leavevmode

595: \includegraphics[clip=true,width=.8\columnwidth]{Andreev.eps}

596: \caption{Amplitude of Andreev state vs. distance $n=x/d$ from 110

597: surface for $\Delta_0=0.2t$, $\mu=0$ .} \label{fig:Andreev}

598: \end{center}

599: \end{figure}

600: For even $n$, the bare Green's function ${G}^{(0)}(n,k_y,\omega)$

601: is proportional to $\omega$ for low-frequencies. Then the

602: $T$-matrix $T \sim 1/\omega$ for small $\omega$.  Even though the

603: $T$-matrix has a pole at $\omega=0$, the product of two Green's

604: functions in the analog of Eq. \ref{greens}  decreases faster,

605: resulting in zero LDOS at zero frequency. As seen from the right

606: panel of Fig. \ref{fig:surf110}, the low-energy density of states

607: on even layers near the $(110)$ surface is substantially less than

608: in the bulk, where it varies linearly with sufficiently low

609: energy. Furthermore, the amplitude of the gap features in the LDOS

610: is noticeably suppressed with decreasing distance from the

611: surface. These features of the LDOS can be understood based on a

612: simple relation between the Green's function ${\hat G}(n'=n, k_y,

613: \omega)$ of the half-space and the bare Green's function ${\hat

614: G}^{(0)}(n-n=0, k_y, \omega)$, which takes place for even $n=2m$ :

615: \begin{eqnarray}

616: &&\hat{{G}}(n'=n, k_y, \omega)=\Bigl\{1-\qquad \qquad \qquad

617: \qquad \qquad

618: \qquad \nonumber \\

619: &&-\exp\left[-2i|n|z\

620: \mathrm{sgn}\left(\omega(q^2-\Delta^2)\right)\right]\Bigr\}

621: {\hat G}^{(0)}(0, k_y, \omega) . \label{g2m}

622: \end{eqnarray}

623: The factor in the braces in Eq.(\ref{g2m}) controls the deviation

624: from the bulk behavior and diminishes with decreasing even $|n|$.

625: At low energies only narrow regions of $k_y$ near the center and

626: the edge of the Brillouin zone contribute to the LDOS. As a

627: result, for even layers the density of states goes as $|\omega|^3$

628: at low energies, with the main contributions arising from $k_y$

629: near the edge of the Brillouin zone.

630: 

631: For odd layers, the $\tau_1$ and $\tau_3$ components of

632: $\hat{{G}}^{(0)}(n=2m+1,k_y,\omega)$ are the ones that are

633: non-zero, and they approach a constant value as $\omega$ goes to

634: zero. So the pole in the $T$-matrix generates the peak in the

635: LDOS, associated with the zero-energy Andreev surface states. For

636: Andreev states, $z$ is an imaginary quantity. The size of the peak

637: decreases as the distance to the surface increases due to the

638: $e^{-|n {\mathrm Im} z|}$ factor in Eq.(\ref{godd}). For large

639: $|n|$, small $k_y$ dominates the integration over $k_y$ in the

640: LDOS and we obtain the following zero-energy asymptotic behavior

641: of the LDOS:

642: $\rho(\omega)=\dfrac{2t}{\pi\Delta_0n^2}\delta(\omega)$. The size

643: of the zero-energy peak in the LDOS $\propto n^{-2}$.

644: 

645: 

646: \subsection{$(210)$ surface}

647: It is useful to study a case intermediate between the standard

648: $(100)$ and $(110)$ surfaces to see what qualitatively new features

649: arise. From the usual quasiclassical viewpoint, the weight of the

650: zero-energy Andreev $(210)$-surface states should be finite, but

651: smaller than for the $(110)$ surface because the phase space for

652: which the reflecting quasiparticle experiences a sign change of the

653: order parameter is reduced. The tight-binding model leads

654: to a more complicated dependence of the weight of the zero-energy

655: states on the surface orientations relative to the crystal axes.

656: Thus, for the $(210)$ surface the model shows at half-filling

657: no zero-energy Andreev states at all. We associate this discrepancy,

658: in particular, with the difference between reflection channels

659: incorporated in the two approaches.

660: 

661: Standard quasiclassical considerations imply that parallel to a

662: smooth surface the momentum component $k_y$ is conserved in a reflection

663: event, and only conventional specular reflection takes place. A

664: tight-binding model shows that, generally speaking, this is not the

665: case, since the {\it crystal momentum} component $k_y$ can also change in a

666: reflection process by a reciprocal crystal vector along the surface.

667: Due to a difference between reciprocal crystal vectors at the surface

668: and in the bulk, the momentum acquired by a quasiparticle in a

669: reflection event can be physically distinguished in the bulk from

670: that of specularly reflecting quasiparticle. Hence, specific crystal

671: periodicity along a particularly oriented surface can result in

672: additional channels for quasiparticle reflection, if there is a reflected

673: state $k_{x}(k_y),\, k_y$ on the Fermi surface corresponding to the

674: $k_y$ surface Umklapp process \cite{gantmakher}.

675: 

676: In this case the Fermi surface, considered as a part of the surface

677: adapted Brillouin zone, should exhibit multiple values of outgoing

678: $k_{F,x}$ for a fixed value of $k_{F,y}$. This situation is realized

679: for the comparatively complicated multisheet structure of the Fermi

680: surface in the surface-adapted Brillouin zone for the (210) surface,

681: as obtained in Fig.9 of Ref.\onlinecite{pairor02}. A strong dependence

682: of the particular shape of the surface-adapted Brillouin zone, as well

683: as the Fermi surface in the zone, on surface orientation is an important

684: characteristic feature of the tight-binding models \cite{Walker,pairor02}.

685: Within the quasiclassical theory, the shape of the Fermi surface is usually

686: considered as independent of surface orientations relative to the

687: crystal axes. We note, that the zero-energy surface states, as a rule,

688: do not exist within the quasiclassical approach,

689: if multiple channels for reflection of quasiparticles from an impenetrable

690: surface are assumed. In this subsection, we now study the LDOS for

691: quasiparticle spectra obtained with the tight-binding model at half-filling

692: for (210) surface.

693: 

694: Technically, we now need  to solve for the Green's function in the

695: presence of two impurity lines (Fig. \ref{fig:surf210}).

696: In the general case of two parallel impurity lines, we cut the

697: crystal at an orientation given by ($hl0$) and introduce an

698: impurity potential

699: \bea

700: U(\r) = U_0 \sum_j \delta(\r- \R_j) \enspace ,

701: \eea

702: where $\R_j$ are the points on the two impurity lines, at the

703: location of the boundaries. The first boundary is defined to be

704: located at $x=0$ with impurity sites $\R_j = j d^{-1} \hat{y}$

705: and the other is parallel and located at $x=(N+1)d$ with

706: $\R_{j} = (N+1)d \hat{x} + (c+jd^{-1})\hat{y}$, giving a total

707: of $N$ free chains. Here $c$ is a shift along $y$-axis of sites

708: on the $(N+1)$-th chain relative to the sites on the $0$-th chain.

709: For the special case of the $(210)$ surface, we need two

710: adjacent lines at $x=0$ and $x=-d$, so formally this corresponds

711: to the case of $N=-2$.

712: 

713: The equation for the $T$-matrix (Eq. \ref{tmat}) for this impurity

714: potential can be solved by choosing the ansatz

715: \bea

716: \hat T(k_x,k_x^{\prime}, k_y) &=& \hat t_0 + \hat t_1 e^{i (N+1)d k_x} +

717: \hat t_2 e^{-i(N+1)d k_x^{\prime}} \nonumber \\

718: &&+ \hat t_3 e^{i(N+1)d (k_x-k_x^{\prime})} \enspace .

719: \eea

720: In the limit of infinitely strong impurity potential this gives

721: \bea

722: \hat t_0 &=& {-\hat{G}^{(0)}(0)^{-1} \over 1 - \hat{ G}^{(0)}(0)^{-1} \hat{

723: G}^{(0)}(-N-1)\hat{ G}^{(0)}(0)^{-1} \hat{ G}^{(0)}(N+1)} \nonumber \\

724: \hat t_1 &=& {\hat{ G}^{(0)}(0)^{-1} \hat{ G}^{(0)}(N+1) \hat{

725: G}^{(0)}(0)^{-1} \over 1 - \hat{ G}^{(0)}(0)^{-1} \hat{

726: G}^{(0)}(-N-1)\hat{ G}^{(0)}(0)^{-1} \hat{ G}^{(0)}(N+1)}\nonumber \\

727: \hat t_3 &=& - {\hat{ G}^{(0)}(0)^{-1} \over  1 - \hat{

728: G}^{(0)}(0)^{-1} \hat{ G}^{(0)}(N+1) \hat{ G}^{(0)}(0)^{-1} \hat{

729: G}^{(0)}(-N-1)} \nonumber \\

730: \hat t_2 &=& {\hat{ G}^{(0)}(0)^{-1} \hat{ G}^{(0)}(-N-1) \hat{

731: G}^{(0)}(0)^{-1} \over 1 - \hat{ G}^{(0)}(0)^{-1} \hat{

732: G}^{(0)}(-N-1) \hat{ G}^{(0)}(0)^{-1} \hat{ G}^{(0)}(N+1)}

733: \nonumber

734: \eea

735: The Green's function in this case can then be

736: written as:

737: $$ \hat{ G}\!(n, n^{\prime}) = \hat{G}^{(0)}\!(n-n^{\prime}) -

738: \left(\hat{ G}^{(0)}\!(n) \ \ \ \hat{

739: G}^{(0)}\!(n-N-1) \right)\times$$

740: \bea

741: \!\!\left(\!\begin{array}{cc}\! \hat{ G}^{(0)}\!(0) & \hat{ G}^{(0)}\!(

742: \!- N\!-\!1)\\

743: \hat{ G}^{(0)}\!( N\!+\!1) & \hat{ G}^{(0)}\!(0)

744: \end{array} \!\right)^{\!-1}\!

745: \!\left(\!\begin{array}{c} \hat{ G}^{(0)}\!(-n^{\prime})\\

746: \hat{ G}^{(0)}\!(\! N\!+\!1\!-\! n^{\prime})

747: \end{array}\!\right)\enspace ,

748: \label{greenwire}

749: \eea

750: where we have not written the

751: dependence on $k_y$ and $\omega$ explicitly.

752: 

753: 

754: \begin{figure}[tbh!]

755: \begin{center}

756: \leavevmode

757: \includegraphics[width=.9\columnwidth]{surface210_S.eps}

758: \caption{Local density of states $\rho(x,\omega)$ versus

759: $\omega/\Delta_{max}$ for a closed(210) non-transparent surface at

760: half-filling model and $\Delta_0=0.2t$, $\mu=0$. Each curve

761: corresponds to a chain located at a distance $x/d$ to the

762: surface.} \label{fig:LDOS210}

763: \end{center}

764: \end{figure}

765: 

766: 

767: The equation for the bound state on (210) surface is

768: \bea

769: {\rm det} \left(\begin{array}{cc} \hat{{G}}^{(0)}(0) &

770: \hat{{G}}^{(0)}(1)\\

771: \hat{{G}}^{(0)}(-1) & \hat{{G}}^{(0)}(0) \end{array} \right) = 0

772: \enspace .

773: \label{det210}

774: \eea

775: 

776: 

777: Fig \ref{fig:LDOS210} displays the LDOS spectra calculated on

778: several layers for (210) surface.  There is no zero-energy peak

779: observed on any chain at all, for the reasons explained in the

780: beginning of this subsection. The peaks in the LDOS, seen close to

781: the (210) surface at energies $\pm 0.5\Delta_{max}$, originate

782: from the gap features taken for the momentum along the surface

783: normal: $\Delta(k_{F,x}, k_{F,y}=0)=0.5\Delta$. The peak at

784: $\omega=\Delta(k_{F,x},0)$ arises for the homogeneous model of

785: the order parameter, while for the self-consistent spatially

786: dependent order parameter it lies slightly below $\Delta(k_{F,x},0)$.

787: It is associated with the

788: surface Andreev states and decays in the depth of the

789: superconductor. These peaks have been theoretically found first in

790: Ref.\ \onlinecite{bbs97} with a continuous model and then also for

791: the conductance with a lattice model \cite{pairor02}. We notice

792: that the conductance spectrum shown in Fig.14 of

793: Ref.\onlinecite{pairor02}, is in agreement with the LDOS on the

794: first chain (at $x=d$) in Fig.\ref{fig:LDOS210}. The variations of

795: the LDOS from chain to chain, which accompany a large-scale

796: behavior, are the Friedel-like oscillations.

797: 

798: \section{Half-filled Wires}

799: 

800: We now consider wires where the second line of impurities confines

801: the system to a finite width, i.e. we restrict ourselves to the

802: cases with surface normal along the $(100)$ and $(110)$

803: directions. Semiclassically, a quasiparticle will go through

804: multiple scatterings, bouncing back and forth between the two

805: walls. Hence we expect that the interplay between Andreev

806: reflection, taking place due to sign change of the order

807: parameter, and the energy discretization, due to finite wire

808: width, to yield novel features. We again model the surfaces by

809: introducing impurities on the appropriate sites to completely

810: isolate the wires. The corresponding Green's function is given by

811: Eqs.(\ref{greenwire}). The bound state energies are determined by

812: the equation \bea {\rm det} \left(\begin{array}{ll}

813: \hat{{G}}^{(0)}(0) &

814: \hat{{G}}^{(0)}(-N-1)\\

815: \hat{{G}}^{(0)}(N+1) & \hat{{G}}^{(0)}(0) \end{array} \right) = 0

816: \enspace .

817: \label{det}

818: \eea

819: 

820: 

821: \subsection{(100) wires}

822: 

823: 

824: Ziegler et al.\cite{zieglerwire}  investigated the problem of

825: (100) $d$-wave quantum wires in some detail, and discovered the

826: existence of a number-parity effect as a function of the width $N$

827: of a mesoscopic $d$-wave wire: a finite total DOS is found at the

828: Fermi level for odd $N$ and zero DOS (with a full gap in the

829: excitation spectrum, not a $d$-wave like gap) is found for even

830: $N$, at least for a simple tight-binding band at half-filling. The

831: differences between even and odd $N$ were shown to survive for

832: more general bands, as well.   For completeness, we reproduce,

833: using our approach, some of their LDOS spectra in Fig.

834: \ref{fig:wires100}.

835: 

836: \begin{figure}[tbh]

837: \begin{center}

838: \leavevmode

839: \includegraphics[clip=true,width=.7\columnwidth,angle=90]{strips_100_S.eps}

840: \caption{Local density of states $\rho(x,\omega)$ vs.

841: $\omega/\Delta_{max}$ for a $(100)$ wire of width $N=4$ (left panels)

842: and $N=5$ (right panels), using $\Delta_0=0.2t$, $\mu=0$. }

843: \label{fig:wires100}

844: \end{center}

845: \end{figure}

846: 

847: 

848: \subsection{(110) wires}

849: 

850: \subsubsection{Normal metal wires}

851: \label{sec:nwires} Based on intuition from the surface case, we

852: would expect that Andreev states play an important role  in (110)

853: wires, with geometry shown in Figure \ref{fig:strip110}. A crucial

854: question which arises in the following discussion is, how does one

855: identify a subgap state of true Andreev character? By merely

856: measuring the LDOS with an STM, for example, one may see several

857: peak structures, not all of which will be related to Andreev

858: reflections at the surfaces. One set of candidate states which

859: needs to be investigated first is the set of discrete dispersive

860: (with respect to $k_y$) levels which arise simply because of the

861: finite wire width.  These are of course present already in the

862: normal state wire.

863: 

864: The $k_x$-integrated bare normal state Green's function for an

865: infinite lattice above $T_c$ takes the form : \bea

866: {G}^{(0)}_{11}(n,k_y,\omega)&=&\frac{d}{2\pi}\int^{\pi/d}_{-

867: \pi/d}\frac{e^{ik_xnd}dk_x} {(\omega+i0)-\xi_\k}\nonumber\\

868: &=& - {i\exp [i|n| \arccos (-\omega/q(k_y))]\over

869: \sqrt{q^2(k_y)-\omega^2}}\, ,\eea  where $ q(k_y)$ is defined in

870: Eq.(\ref{qky}). The full Green's function for a (110) wire may

871: then be obtained by solving the $T$-matrix equation as

872: \be

873: G_{11}(n,n';k_y,\omega)=\frac{2\sin[n_{min}z]\sin[(n_{max}-N-

874: 1)z]}

875: {\sqrt{q^2(k_y)-\omega^2}\sin[(N+1)z]}

876: \label{gnm}

877: \ee

878: where $n,n'=1,2,...,N$, $z\equiv z(k_y,\omega)=\cos^{-1}(-\omega/q(k_y))$,

879: $n_{min}={\rm min}(n,n')$,  and $n_{max}={\rm max}(n,n')$.

880: 

881: \begin{figure}[tbh]

882: \begin{center}

883: \leavevmode

884: \includegraphics[width=.7\columnwidth]{strip110.eps}

885: \caption{Geometry of 110 wire (sites are filled circles) of width

886: $N$ bounded by 2 impurity lines (filled squares).  Wire is

887: infinite in both $y$ and $-y$ directions.} \label{fig:strip110}

888: \end{center}

889: \end{figure}

890: 

891: 

892: For every value of $k_y$, there generally exists a series of

893: eigenvalues which are the poles of the Green's function. They may

894: be obtained by simply solving

895: \begin{equation}

896: \sin[(N+1)z(k_y,\omega)]=0 \enspace ,

897: \label{eigennormal}

898: \end{equation}

899: which yields:

900: \begin{equation}

901: \omega_\nu(k_y)=-q(k_y)\cos\frac{\pi \nu}{N+1},~~

902: \nu=0,1,...,N+1 \ .

903: \label{modefreqnormal}

904: \end{equation}

905: This gives $N$ branches of solutions for $\nu = 1, ..., N$,

906: distributed symmetrically with respect to the Fermi level, and two

907: special solutions ($\nu=0$ and $N+1$) with $\omega=\pm q(k_y)$

908: which we refer to as defining the "effective band edge". One

909: interesting feature is the existence in the normal metal wire of

910: dispersionless zero-energy  quasiparticle states (ZES) which form

911: the branch $\nu=(N+1)/2$ when the width $N$ of the wire is odd.

912: Since the group velocity vanishes for dispersionless states, they

913: are always localized. We note that the case $N=1$ has only the

914: obvious dispersionless ZES since no transport is allowed with only

915: nearest neighbor hopping. In the general $N$=odd case, it appears

916: that there is always exactly one such localized state (doubly

917: degenerate in  particle-hole space) for any fixed $k_y$. A

918: deviation from half filling shifts the zero-energy states to

919: $-\mu$. Thus, for positive $\mu$ they are the hole states, while for

920: $\mu<0$ - electron states.

921: 

922: \begin{figure}[tbh]

923: \begin{center}

924: \leavevmode

925: \includegraphics[clip=true,width=.9\columnwidth]{strip_110_5_NS.eps}

926: \caption{Local density of states  for a (110)  wire of width

927: $N=5$, with $\Delta_0=0.2t$, $\mu=0$. Left: normal state,

928: $\rho(x,\omega)$ vs. $\omega/\Delta_{max}$; right: superconducting

929: state, $\rho(x,\omega)$ vs. $\omega/\Delta_{max}$.}

930: \label{fig:N5wires}

931: \end{center}

932: \end{figure}

933: 

934: The contribution of each of these states to the LDOS can be

935: estimated by examining the residue near the pole, where the

936: Green's function can be approximated as:

937: \begin{equation}

938: G(n,n^{\prime};k_y,\omega)

939: \approx\frac{Q_\nu(n,n^{\prime})}{\omega-\omega_\nu(k_y)},

940: \end{equation}

941: with the residue given by:

942: \begin{equation}

943: Q_\nu(n,n')=\frac{2\sin\frac{n\pi \nu}{N+1}\sin\frac{n^{\prime}\pi

944: \nu}{N+1}}{N+1},\label{weightn0}

945: \end{equation}

946: where $n$ is the index of the layers. Note from

947: (\ref{weightn0}) and the spectral representation of the Green's

948: function near a pole, it is easy to read off the quasiparticle

949: wavefunctions as $\psi_\nu(n) =\sqrt{2/(N+1)}\sin[n\pi\nu/(N+1)]$,

950: i.e. just the wave functions of a free particle confined to a box

951: of width $(N+1)d$.

952: 

953: \begin{figure}[tbh]

954: \begin{center}

955: \leavevmode

956: \includegraphics[clip,width=.9\columnwidth]{strip_110_4_NS.eps}

957: \caption{LDOS for a (110)  wire of width $N=4$ with

958: $\Delta_0=0.2t$, $\mu=0$. Left: normal state, $\rho(x,\omega)$ vs.

959: $\omega/\Delta_{max}$; right: superconducting state,

960: $\rho(x,\omega)$ vs. $\omega/\Delta_{max}$.} \label{fig:N4wires}

961: \end{center}

962: \end{figure}

963: 

964: It is easy to check that for any $n$, $n'$ the residue $Q_\nu$

965: vanishes everywhere for the "states" at the effective bandwidth $\nu=0$ or

966: $N+1$; we therefore do not discuss them further, but focus on the

967: $N$ branches with finite residue. For these states the

968: residue can vanish locally on chains with number $n$, for

969: which the quantity $n\nu/(N+1)$ is an integer. The odd $N$ ZES,

970: for example, has a residue

971: \begin{equation}

972: Q_{\frac{(N+1)}{2}}(n,n)=\frac{2(\sin\frac{n\pi }{2})^2}{N+1}

973: \enspace .

974: \label{weightnn}

975: \end{equation}

976: As is seen from Eq.(\ref{weightnn}), the probability density of

977: the ZES oscillates with a period $2d$, taking finite values only

978: for odd $n$ and vanishing on all nearest neighbor sites (~where $n$

979: is even~). This is also valid for the states with energies $\pm\mu$

980: in the case of the deviation from half-filling and ensures no transport

981: with nearest neighbor hopping for quasiparticles with energies $\pm\mu$.

982: 

983: All these states are not surface bound states, as one can easily

984: check that their amplitude does not decay across the wire. The

985: quasiparticle spectrum is discrete (for fixed $k_y$) due to the

986: finite width of the wire and transforms into conventional

987: continuous quasiparticle spectrum in the massive normal metal in

988: the limit $N\to\infty$. The momentum resolved LDOS in the normal

989: metal wire with discrete dispersive states takes the form

990: \begin{equation}

991: \rho(n,k_y,\omega)=\sum\limits_{\nu=1}^{N}Q_{\nu}(n=n')\delta(\omega-

992: \omega_{\nu}(k_y)) \enspace .

993: \end{equation}

994: The integration over $k_y$ gives the LDOS:

995: \begin{equation}

996: \rho(n,\omega)=\frac{1}{2\pi d}\sum\limits_{\omega_{\nu}(k_y)=\omega}

997: \dfrac{Q_{\nu}(n,n)}{\left|\dfrac{d\omega_{\nu}}{dk_y}\right|} \enspace ,

998: \label{rhon}

999: \end{equation}

1000: where the sum is taken over those $k_y$ and $\nu$, which satisfy

1001: the equation $\omega_{\nu}(k_y)=\omega$.

1002: 

1003: The position of peaks in LDOS are then determined by extrema of

1004: the dispersive branches (\ref{modefreqnormal}), taking place at

1005: $k_y=0$: \be \omega_{\nu,peaks}=-4t\cos\frac{\pi \nu}{N+1},~~

1006: \nu=1,...,N \ . \label{energypeaks}

1007: \end{equation}

1008: 

1009: The peaks corresponding to the $N$ normal metal wire states are

1010: seen clearly in Figs. \ref{fig:N5wires} and \ref{fig:N4wires} at

1011: the eigenfrequencies given by Eq. (\ref{energypeaks}). It is easy

1012: to check that the weights agree with Eq. (\ref{weightn0}). The

1013: LDOS for the normal metal wires qualitatively differs from the

1014: LDOS for bulk normal metals with the nearest neighbor hopping on

1015: the square. The zero-energy peak (the Van Hove singularity) in the

1016: bulk metal is symmetric as a function of the energy and its log

1017: singularity is much broader than the  $\delta$-like peaks we find

1018: here for wires.

1019: 

1020: \subsubsection{Superconducting wires}

1021: \label{sec:swires}

1022: 

1023: The $T$-matrix and Green's function equations are necessarily more

1024: complicated in the presence of superconductivity, but they are

1025: still tractable in the $(110)$ case. The bare Green's functions

1026: $\hat{G}^{(0)}$ are given by Eqs. (\ref{geven}) and (\ref{godd}),

1027: as before. One must then solve Eq. (\ref{det}), which applies to

1028: any situation which requires two lines of impurities, for the

1029: eigenenergies $\omega$. Similar to the normal metal case, there are

1030: special solutions of Eq. (\ref{det}): $\omega=\pm\Delta(k_y)$,

1031: $\omega=\pm q(k_y)$ for any $N$ and $\omega=0$ for even $N$. They

1032: are  poles of the $T$-matrix, but do not correspond to the poles

1033: of the full Green's function.

1034: 

1035: The behavior of subgap surface states on the narrow

1036: superconducting wire differs qualitatively from the case of the

1037: superconducting halfspace due to the interference of the wave

1038: functions of the states on both surfaces. Since the zero-energy

1039: peak in the LDOS for the half-filled surface vanishes on each even

1040: chain (~see Figs.(\ref{fig:surf110}), (\ref{fig:Andreev})), the

1041: spectrum of  Andreev states on the $(110)$ wires becomes strongly

1042: dependent on the parity of the number $N$ of chains in the

1043: half-filled wire. This effect is quite pronounced for wires whose

1044: width is the order of or less than the superconducting coherence

1045: length. As we demonstrate below, some new qualitative features

1046: arising in the superconducting state of the $(110)$ wires in the

1047: quasiparticle spectrum above $\Delta(k_y)$ (~see Eq. (\ref{dky}))

1048: can also be strongly dependent on the parity of $N$.

1049: 

1050: \vskip .2cm

1051: 

1052: {\it Odd $N$.} For a wire with odd $N$, only the Green's functions

1053: (\ref{geven}) with even arguments  $n=0, \pm(N+1)$ enter

1054: Eq.(\ref{det}).  Since at small frequencies these Green's function

1055: are $\propto \omega$, it is straightforward to show that the only

1056: subgap state is a dispersionless ZES $\omega=0$. Dispersive modes

1057: exist as well and take the form

1058: 

1059: \begin{eqnarray} \omega_{\nu}^2(k_y)=q^2(k_y)\cos^2\frac{\pi

1060: {\nu}}{N+1}+\Delta^2(k_y)\sin^2\frac{\pi

1061: {\nu}}{N+1}~, \nonumber\\

1062: ~~{\nu}=1,...,\frac{N-1}{2} \ .

1063: \label{dm}

1064: \end{eqnarray}

1065: 

1066: This exactly coincides with the quasiparticle spectrum in a bulk

1067: two-dimensional superconductor $\omega^2(k_y, k_x)=\xi^2(k_y, k_x)

1068: +\Delta^2(k_y,k_x)$, if the discrete values of momentum component

1069: across the wire $k_{x,\nu}=\pi{\nu}/(N+1)d$ are introduced. We

1070: note that the dispersionless zero-energy Andreev states are the

1071: only true subgap states in the spectrum, since the energy

1072: $|\omega_{\nu}(k_y)|$ of the dispersive states (\ref{dm}) lie

1073: above the respective value $|\Delta(k_y,k_x)|$ of the order

1074: parameter, for any $\nu$ and $k_y$.

1075: 

1076: \begin{figure}[htb]

1077: \begin{center}

1078: \leavevmode

1079: \includegraphics[width=.9\columnwidth]{N11_combi.eps}

1080: \caption{ Quasiparticle spectra and the LDOS for superconducting

1081: half-filled (110) wire with $N=11$.\ (a) Dispersive modes ($k_yd$

1082: vs. $\omega/t$) for $N=11$, $\Delta_0=0.2t$, $\mu=0$. (b) The

1083: blowup of the dispersive spectra near the edge of the Brillouin

1084: zone.\ (c) The LDOS for energies less than $\Delta_{max}$: n=2

1085: (solid line), n=1 (dashed line), where a broadening $\delta=.005t$

1086: has been used. }

1087: \label{figN11}

1088: \end{center}

1089: \end{figure}

1090: 

1091: There are, as in the normal state case, $N-1$ dispersive modes in

1092: addition to the ZES. They are doubly degenerate due to the

1093: particle-hole symmetry. This simple result can be understood as

1094: follows. For a fixed $k_y$ the problem reduces to a

1095: one-dimensional problem. The corresponding two-component

1096: Bogoliubov-DeGennes wave function takes its values on $N$ sites,

1097: resulting in $2N$ degrees of freedom in the system. With this

1098: point of view it appears natural that for fixed $k_y$ the total

1099: number of levels, which are twice degenerate, is $N$. The set of

1100: levels with positive energies for $N=11$ is represented in panel

1101: (a) of Fig.\ref{figN11}.

1102: 

1103: The LDOS for the superconducting wires with N=5 is shown in the

1104: right panel of Fig.~\ref{fig:N5wires}. The dispersionless ZES

1105: results in a pronounced peak at zero energy on odd layers.

1106: Furthermore, each extremum of the dispersive mode $\omega(k_y)$

1107: results in the peak in the LDOS at the energy

1108: $\omega_{k_{y,extr}}$. One series of  peaks is associated with the

1109: extrema at $k_y=0$. Since $\Delta(k_y=0)=0$, the peaks lie at the

1110: same positions as in the normal metal wire (\ref{energypeaks}).

1111: Although they are irrelevant to the superconducting properties of

1112: the wire, some of these peaks can lie at finite energies below the

1113: maximum of the gap function $\Delta_{max}=2\Delta_0$. For

1114: instance, the lowest position at finite energies of the peaks of

1115: this series is $4t\sin\frac{\pi}{N+1}$. This can be both above or

1116: below $\Delta_{max}$, depending on the ratio $\Delta_{max}/t$ and

1117: the wire width $N$. In contrast with the normal metal wires, the

1118: dispersive quasiparticle modes Eq.(\ref{dm}) in the

1119: superconducting wires have extrema also at the edge of the

1120: surface-adapted Brillouin zone $k_y=\pm \pi/(2d)$ \cite{ebz}. They

1121: are shown in panel (b) of Fig.\ref{figN11} for the wire with

1122: $N=11$. This leads to additional series of $(N-1)$ quasiparticle

1123: peaks in the LDOS for the superconducting wires, which lie below

1124: $\Delta_{max}$:

1125: 

1126: \begin{equation} \omega_{peaks}=\pm \Delta_{max}\sin\frac{\pi \nu}{N+1},~~

1127: \nu=1,...,\frac{N-1}{2} \ .

1128: \label{energypeaksodd}

1129: \end{equation}

1130: Panel (c) of Fig.\ref{figN11} displays the series of peaks in the LDOS for

1131: $N=11$.

1132: 

1133: The  dispersive states forming the nonzero low-energy peaks

1134: (\ref{energypeaksodd}) in LDOS, are {\it not} Andreev states. They lie

1135: below $\Delta(k_y)$ for $k_y$ near the edges of Brillouin zone, but they

1136: are situated above the bulk gap function $\Delta(k_x(\nu),k_y)$. The wave

1137: function for any of these states possesses the finite current of

1138: the probability density, while for Andreev states the total

1139: probability current vanishes.

1140: 

1141: While positions of the peaks are determined by the

1142: extrema of the dispersive energies, their weights in the LDOS

1143: are controlled also by quasiparticle wave functions, which form the

1144: quantity $Q(n,n)$ in Eq.(\ref{rhon}).

1145: For this reason the weights of the peaks can substantially

1146: differ on different layers and may vanish on some of them, quite

1147: analogously to the normal metal wires (see Eq.(\ref{weightn0})).

1148: Due to an interference from two surfaces, the wave function, taken

1149: on even layers, turns out to coincide with the respective wave

1150: function in the normal metal case (apart from a normalization

1151: constant). On the other hand, the wave function on odd layers is a

1152: superposition of electron-like and hole-like Bogoliubov

1153: quasiparticles on the wire.

1154: 

1155: \vskip .5cm {\it Even $N$.}

1156: In the case of even $N$ half-filled wires, the quasiparticle

1157: spectra become more complicated. The Green's functions

1158: (\ref{geven}), (\ref{godd}) with even ($n=0$) and odd

1159: ($n=\pm(N+1)$) arguments enter Eq.(\ref{det}), which can be

1160: reduced to  the following form

1161: \be q(k_y)\tan[(N+1) z]=

1162: \alpha\Delta(k_y)\tan z \enspace ,

1163: \label{z}

1164: \ee

1165: where $\alpha=\pm

1166: 1$, $0\le z\le\pi/2$. Solutions $z_{\nu,\alpha}(k_y)$ of

1167: Eq.\eqref{z} are directly associated with $\omega$, in accordance

1168: with Eq.\eqref{zomega}:

1169: \begin{equation}

1170: \omega^2_{\nu,\alpha}(k_y)=q^2(k_y)\cos^2z_{\nu,\alpha}(k_y)+

1171: \Delta^2(k_y)\sin^2z_{\nu,\alpha}(k_y) \enspace .

1172: \label{evenspectr}

1173: \end{equation}

1174: Here ${\nu}=1,...,\frac{N}{2}$ and $\alpha$ are the indices of the

1175: solution. Comparing Eqs.\eqref{evenspectr}, \eqref{dm} for the

1176: spectra of odd and even wires, one can see that $z/d$ plays the

1177: role of effective discrete values of the momentum component $k_x$

1178: (at fixed $k_y$) across the wire. Eq.\eqref{z} can be transformed

1179: to a polynomial equation in $\tan z$ of the $N$-th degree, if one

1180: excludes the special solutions $\omega=\pm\Delta(k_y), \pm q(k_y)$

1181: mentioned above. Hence, for a fixed $\alpha$ there are exactly

1182: $N/2$ positive and $N/2$ negative solutions for $\omega$,

1183: describing $N$ dispersive branches of the quasiparticle spectrum.

1184: Explicit analytical form of the spectra can be easily found from

1185: Eqs.\eqref{evenspectr}, \eqref{z} in the particular cases $N=2$

1186: \be

1187: \omega_{\alpha}=\pm\dfrac{1}{2}\Bigl(q(k_y)+\alpha\Delta(k_y)\Bigr),

1188: \label{2} \ee and $N=4$ ($\nu=\pm$, $\alpha=\pm$)

1189: \begin{eqnarray}

1190: &\omega^2_{\nu,\alpha}(k_y)=\Delta^2(k_y)+\dfrac{1}{8}q(k_y)\Bigl(q(k_y)+

1191: \alpha\Delta(k_y)\Bigr)\Biggl[3- \nonumber \\

1192: & -5\alpha\dfrac{\Delta(k_y)}{q(k_y)}

1193: +\nu\,\sqrt{5-6\alpha\dfrac{\Delta(k_y)}{q(k_y)}+

1194: 5\dfrac{\Delta^2(k_y)}{q^2(k_y)}}\,\Biggr] . \label{4}

1195: \end{eqnarray}

1196: Eq.\eqref{4} is defined for all $k_y$ in the Brillouin zone for

1197: which $\omega^2_{\nu,\alpha}(k_y)$ is positive. Eqs.\eqref{2},

1198: \eqref{4} describe the quasiparticle spectra for

1199: non-self-consistent wires with small numbers of chains. As we will

1200: show in Sec. \ref{sec:self}, the self-consistent treatment of the

1201: problem can lead to important modifications of the results, at

1202: least if the width of the wire is less than or of  order  the

1203: superconducting coherence length.

1204: 

1205: \begin{figure}[tbh]

1206: \begin{center}

1207: \leavevmode

1208: \includegraphics[width=.9\columnwidth]{N10_combi.eps}

1209: \caption{Quasiparticle spectra and the LDOS for superconducting

1210: half-filled (110) wire with $N=10$.\ (a) Dispersive modes ($k_yd$

1211: vs. $\omega/t$) for $N=10$, $\Delta_0=0.2t$, $\mu=0$.\ (b) The

1212: blowup of the spectra  near the edge of the Brillouin zone.\ (c)

1213: The LDOS for energies less or equal to $\Delta_{max}$ : n=2 (solid

1214: line), n=1 (dashed line), where a broadening $\delta=.005t$

1215: has been used. }

1216: \label{figN10}

1217: \end{center}

1218: \end{figure}

1219: 

1220: Whereas in the normal metal state, $\nu$ is the only index of the

1221: solution (for a given $k_y$), in the superconducting even-$N$ wire

1222: each branch with fixed $\nu$ splits into two, corresponding to two

1223: values of the index $\alpha$, giving a  total  of $2N$ distinct

1224: branches. Since the order parameter vanishes for $k_y=0$, the

1225: splitting is absent in this particular case. The splitting is

1226: associated with the symmetry breaking of the eigenstates with

1227: respect to the sign reversal of $k_y$. In the normal  even-$N$

1228: wires, as well as in the odd $N$ wires (both in the normal metal

1229: and in the superconducting states), each separate branch of the

1230: spectra $\omega_{\nu}(k_y)$ is an even function of $k_y$. This is

1231: not the case, however, for the even $N$ superconducting $(110)$

1232: wires. As directly seen from Eq.\eqref{z} and the relations

1233: $q(-k_y)=q(k_y)$, $\Delta(-k_y)=-\Delta(k_y)$, the solutions of

1234: Eq.\eqref{z} $z_{\nu,\alpha}(k_y)$ with fixed $\alpha$ are not odd

1235: or even functions of $k_y$, due to the nonzero right hand side of

1236: Eq.\eqref{z} in the superconducting state. Evidently, each

1237: separate branch of the spectra $\omega_{\nu,\alpha}(k_y)$ possesses

1238: the same property. The whole spectrum, however, remains symmetric

1239: with respect to $k_y\to-k_y$, since the branches with positive and

1240: negative $\alpha$ simply interchange with each other under this

1241: transformation. Spectra of the half-filled $N=10$ superconducting wire

1242: with positive energies are shown in panel (a) of

1243: Fig.~\ref{figN10}. They agree with the above discussion. For

1244: branches with higher energies, the splitting is less than for lower

1245: branches. For small splittings, $\delta\omega\propto \Delta/(N+1)$.

1246: 

1247: The splitting manifests itself by slightly shifting the positions

1248: of peaks in the LDOS, due to respective shifts of extrema of

1249: dispersive quasiparticle energies. In contrast with the double

1250: number of  extrema, the total number of the peaks in the LDOS does

1251: not change. Any peak in the LDOS originates from the two extrema,

1252: situated symmetrically with respect to the sign of $k_y$. Each

1253: pair of extrema belongs to two spectral branches, which differ

1254: only by their index $\alpha=\pm1$. If we disregard the splitting

1255: of the quasiparticle spectra, the positions of the peaks in the

1256: LDOS coming, for example, from the edge of the Brillouin zone, are

1257: as follows

1258: \begin{equation} \omega_{peaks}=\pm \Delta_{max}\cos\frac{\pi \nu}{N+1},~~

1259: \nu=1,...,\frac{N}{2} \ .

1260: \label{energypeakseven}

1261: \end{equation}

1262: The splitting shifts the peak positions towards slightly lower

1263: energies $|\omega|$, as compared with those in Eq.\eqref{energypeakseven}.

1264: The respective extrema of dispersive energies are shifted towards lower

1265: values of $|k_y|$ from $k_y=\pm\pi/(2d)$. These low-energy peaks in the

1266: LDOS are shown in panel (c) of Fig.~\ref{figN10}.

1267: 

1268: The other peaks in the LDOS come, neglecting the splitting, from

1269: the center of the Brillouin zone. Since the order parameter

1270: vanishes in the center of the zone, the positions of the peaks are

1271: still the same as in Eq.\eqref{energypeaks} for the normal metal

1272: state. The splitting, taking place in the superconducting state,

1273: slightly shifts the extrema of the spectra from the center of the

1274: zone, so that the peaks move to a larger energy values $|\omega|$,

1275: as compared with those in Eq.\eqref{energypeaks}.

1276: 

1277: The LDOS for the superconducting wire with $N=4$ is shown in the

1278: right panel of Fig.\ref{fig:N4wires}. In addition to the peaks at

1279: finite energies, discussed above, there is also a well pronounced

1280: zero-energy peak there (see also panel (c) of Fig.~\ref{figN10}).

1281: This peak originates from the extrema of two dispersive branches

1282: of Andreev states, which take place at $\pm k_{y,0}d=\pm

1283: \tan^{-1}(2t/\Delta_0)$, where the relation

1284: $q(k_{y,0})=\Delta(k_{y,0})$ holds. Indeed, as  follows from

1285: Eqs.(\ref{z}), (\ref{zomega}) and, in particular, directly seen

1286: from Eq.(\ref{4}), the energy and its derivative equal zero for

1287: $k_y=\pm k_{y,0}$. It turns out that the multiplicity of zeroes

1288: of the lowest dispersive branches of states at the extrema $k_y=\pm k_{y,0}$ is $N/2$,

1289: i.e. $\omega\propto |k_y-k_{y,0}|^{N/2}$ in the close vicinity of $k_{y,0}$.

1290: This follows directly, for instance, from Eq.(\ref{omegaa}) given below.

1291: We notice, that the lowest dispersive curve in panel (b) of Fig.~\ref{figN10}

1292: manifests very slow change of energy when it touches the zero-energy line.

1293: This agrees with our expectation $\omega\propto |k_y-k_{y,0}|^{5}$

1294: for the (110) wire with $N=10$. The respective peak in the LDOS of wires with

1295: large even $N$ diverges at $\omega=0$, but is not a delta-like peak as it is

1296: for wires with odd $N$.

1297: 

1298: The identification of the Andreev nature of the quasiparticle bound

1299: states in confined geometries turns out to be a nontrivial

1300: problem, at least in the case of half-filled (110) wires with even

1301: $N$, where there is a symmetry breaking with respect to

1302: $k_y\to-k_y$. As seen from Eqs.(\ref{z})-(\ref{4}), the asymmetry

1303: can be associated with the order parameter behavior

1304: $\Delta(-k_y)=-\Delta(k_y)$ and with the sensitivity of the

1305: quasiparticle energies to the $\pi$-shift of the order parameter

1306: phase. A dependence of the dispersive energy curve on the order parameter

1307: phase is usually an intrinsic feature of Andreev bound states only.

1308: We consider the vanishing of  the probability density current as an

1309: defining property of  Andreev bound states. These states should

1310: be able to carry finite electric current, however. In odd $N$ wires

1311: only zero-energy dispersionless quasiparticle states satisfy the

1312: above requirements. In even $N$ $(110)$ wires the zero-energy

1313: dispersionless states do not arise at all. The interference of

1314: wave functions located near two surfaces of the even $N$ wire

1315: induces dispersive Andreev branches $\pm\omega_A(k_y)$, which

1316: transform into the zero-energy states in the limit of infinitely

1317: large $N$. Moreover, all quasiparticle states in the half-filled

1318: $(110)$ superconducting wire with even $N$, even those lying above

1319: the gap, turn out to satisfy the above mentioned conditions, i.e.

1320: possess the properties of Andreev bound states. As an example, we

1321: describe below in detail the structure of the lowest dispersive

1322: quasiparticle branch, which we denote as Andreev branch

1323: $\omega_A(k_y)$, although its energy varies with $k_y$  over a

1324: large range, including both the subgap and the supergap regions

1325: (see, for example, Fig.~\ref{figN10}).

1326: 

1327: For those $k_y$ where ${\rm

1328: min}\{|\Delta(k_y)|,q(k_y)\}<|\omega_A(k_y)| <{\rm

1329: max}\{|\Delta(k_y)|,q(k_y)\}$, the wave function of the state can

1330: be written on odd layers ($n=2m+1$) as: \be \left(\!

1331: \begin{array}{c} u\\v \end{array}\!\right)= C{\rm sgn}\omega_A

1332: \sin[(\!N\!+\!1\!-\!n\!)z] {1\choose {i{\rm sgn}k_y}}

1333: \label{aodd}

1334: \ee

1335: and on even layers $n=2m$

1336: \be

1337: \left(

1338: \begin{array}{c} u\\v\end{array}\right)= C(-1)^{N/2}\sin(nz)

1339: {1\choose {-i {\rm sgn}k_y}} \enspace .

1340: \label{aeven}

1341: \ee

1342: 

1343: In the range of $k_y$ for which

1344: $|\omega_A(k_y)|<{\rm min}\{|\Delta(k_y)|, q(k_y)\}$,

1345: the quantity $z$, entering Eq.\eqref{evenspectr}, becomes imaginary.

1346: Under the condition $|\omega_A(k_y)|<|\Delta(k_y)|<q(k_y)$

1347: the wave function takes the following form on odd layers

1348: ($n=2m+1$):

1349: \be

1350: \left(\! \begin{array}{c} u\\v \end{array}\!\right)=

1351: C_1 (-1)^m {\rm sgn}\omega_A

1352: \sinh[(\!N\!+\!1\!-\!n\!)z_1] {1\choose {i{\rm sgn}k_y}}

1353: \label{aodd1}

1354: \ee

1355: and on even layers $n=2m$

1356: \be

1357: \left( \begin{array}{c} u\\v\end{array}\right)=

1358: C_1(-1)^m\sinh(nz_1) {1\choose {-i {\rm sgn}k_y}} \enspace .

1359: \label{aeven1}

1360: \ee

1361: Analogously, if $|\omega_A(k_y)|<q(k_y)<|\Delta(k_y)|$, the wave function

1362: on odd layers ($n=2m+1$)

1363: \be

1364: \left(\! \begin{array}{c} u\\v \end{array}\!\right)=

1365: C_2 {\rm sgn}\omega_A

1366: \sinh[(\!N\!+\!1\!-\!n\!)z_1] {1\choose {i{\rm sgn}k_y}}

1367: \label{aodd2}

1368: \ee

1369: and on even layers $n=2m$

1370: \be

1371: \left( \begin{array}{c} u\\v\end{array}\right)=

1372: C_2 \sinh(nz_1) {1\choose {-i {\rm sgn}k_y}} \enspace .

1373: \label{aeven2}

1374: \ee

1375: Here $C$, $C_1$ and $C_2$ are normalization constants,

1376: $z_1=|{\rm Im}z(\omega_A(k_y),k_y)|$, $z$ is defined in

1377: Eq.\eqref{zomega} and taken at $\omega=\omega_A(k_y)$.

1378: 

1379: The condition $|u(n,k_y)|=|v(n,k_y)|$, which is valid for all

1380: solutions Eqs.~\eqref{aodd} -- \eqref{aeven2},  results in

1381: zero total probability current density, while the electric current

1382: does not vanish for the given branch. This ensures the Andreev

1383: character, as defined above, of all the states in even $N$ wires,

1384: regardless of whether their energies lie above or below the gap.

1385: 

1386: The quasiparticle states which belong to the same dispersive

1387: branch $\omega_A(k_y)$ can have their wavefunctions both

1388: symmetrically decaying in the depth of the wire

1389: (Eqs.~\eqref{aodd1} -- \eqref{aeven2}) or oscillating and forming

1390: standing waves across the wire (Eqs.~\eqref{aodd}, \eqref{aeven}).

1391: In particular, the amplitude of the wave functions in

1392: Eqs.~\eqref{aodd1} -- \eqref{aeven2} takes its maximum value on

1393: layers $n=1$ and $n= N$, and manifests  Friedel-like oscillations,

1394: as the layer index changes from odd to a neighbor even value or

1395: vice versa. On a larger scale the amplitude decays in the bulk of

1396: the wire symmetrically with respect to two surfaces.

1397: 

1398: It is instructive to follow how the above results transform into

1399: the well known dispersionless zero-energy Andreev surface states

1400: in the limit of large $N$. For sufficiently large width of the

1401: wire compared with the coherence length, and for those $k_y$ where

1402: the wavefunction decays in the depth of the wire, the dispersive

1403: energy $\omega_A(k_y)$ takes a relatively simple form:

1404: \begin{eqnarray}

1405: \omega_A(k_y)=\pm\dfrac{2\Delta(k_y)

1406: q(k_y)}{\sqrt{|q^2(k_y)-\Delta^2(k_y)|}}\times

1407: \qquad \qquad \qquad \quad \nonumber\\

1408: \times\exp\left[-(N+1){\sinh}^{-1}\dfrac{{\rm min}\{|\Delta(k_y)|,

1409: q(k_y)\}}{\sqrt{|q^2(k_y)-\Delta^2(k_y)|}}\right] . \label{omegaa}

1410: \end{eqnarray}

1411: It follows from Eq.\eqref{omegaa}, that the energy of the Andreev

1412: states is exponentially small and vanishes in the limit of

1413: infinitely large $N$. With decreasing energy, the range of $k_y$

1414: where the wave function oscillates (see Eqs. \eqref{aodd},

1415: \eqref{aeven}), converges to the center and to the edges of the

1416: Brillouin zone and finally collapses to the respective points.

1417: Hence, in the limit of very large $N$ the amplitude of the wave

1418: function decays inside the wire for practically all values of

1419: $k_y$. This means that for $N\to\infty$ the dispersive branch

1420: $\omega_A(k_y)$ transforms into the zero-energy dispersionless

1421: surface states situated near the two surfaces.

1422: 

1423: 

1424: \begin{figure}[tbh]

1425: \begin{center}

1426: \leavevmode

1427: \includegraphics[clip=true,width=.9\columnwidth]{WeightZES.eps}

1428: \caption{The weight of the zero-energy peak in the LDOS, taken at the first layer,

1429: as a function on odd (open circles) and even (filled circles) $N$.}

1430: \label{figweights}

1431: \end{center}

1432: \end{figure}

1433: 

1434: Fig.~\ref{figweights} shows $N$-dependence of the weight of the

1435: zero-energy peak in the LDOS. For odd $N$, the weight diminishes

1436: with increasing $N$, while for even $N$ it increases. In the limit of

1437: large $N$ the two curves will converge to the weight for the

1438: zero-energy surface states when the surfaces are infinitely far

1439: apart. The odd-even effects become negligibly small only for

1440: $Nd\gg \xi_0$. One could try to recover the quasiclassical

1441: results for wide wires $Nd\gg a$ after averaging over odd and even

1442: film widths. The boundary conditions for

1443: quasiclassical propagators are taken somewhere on a distance $l$

1444: from the surface and imply some uncertainty about the boundary

1445: positions, as well as the film (wire) thickness ($a\ll l\ll \xi_0$).

1446: As seen from Fig.~\ref{figweights}, averaging of the weights of the

1447: peaks over odd and even $N$ will strongly reduce their width

1448: dependence. This kind of averaging is much closer (although not

1449: identical) to the quasiclassical results on the width dependence of

1450: the LDOS for the $d$-wave superconducting film \cite{nagai}.  Very

1451: recent quasiclassical results \cite{fominov}, treating various wire

1452: orientations, demonstrate the appearance of energy bands of

1453: quasiparticle states, in particular, for $(110)$ wires. Our

1454: microscopic model for high quality half-filled wires with fixed

1455: number of chains gives, however, only a couple of branches

1456: Eq.\eqref{omegaa} for even $N$ wires and the dispersionless

1457: zero-energy states for odd

1458: $N$ wires. We associate the difference between the microscopic and

1459: the quasiclassical results with the particular condition of

1460: half-filling. The quasiclassical approach implies no singular

1461: behavior of the LDOS in the normal metal state near the Fermi

1462: surface, whereas the Van Hove singularity takes place in the LDOS

1463: on the Fermi surface for the normal metal state of half-filled

1464: infinite square lattice. An agreement of our microscopic results with

1465: the quasiclassical ones arises in the presence of deviations from

1466: half-filling (see below Eqs.\eqref{omegaamu}),\eqref{phi}).

1467: 

1468: As already mentioned, in the even $N$ wires the particle-hole

1469: structure of any quasiparticle states with broken $k_y\to-k_y$

1470: symmetry satisfies the condition $|u(n,k_y)|=|v(n,k_y)|$. Our

1471: picture is that for the states above the gap this Andreev

1472: particle-hole structure is generated by the infinite sequence of

1473: ``overbarrier'' (overgapped) Andreev reflections, induced by a

1474: sign reversal of the order parameter, which the quasiparticles

1475: experience along their trajectories being bounded inside the wire

1476: with impenetrable surfaces. This unconventional feature does not

1477: take place for the states above the gap in the odd $N$ wires,

1478: since the two surfaces always result in the standing waves across

1479: the wire with no important interference effects in this case. For

1480: negligibly small splitting one should consider a superposition of

1481: two wave functions, describing the two split states. Then the

1482: Andreev structure of initially nondegenerate wave functions is

1483: lost, since the moduli of particle and hole amplitudes can easily

1484: differ from each other. The Andreev structure of quasiparticle

1485: wavefunctions can be lost also in the presence of deviations from

1486: the half-filling, if $\mu$ is larger or of the same order as the

1487: splitting. Since the splitting vanishes for $k_y=0$, one can

1488: expect that for sufficiently small $k_y$ the Andreev structure of

1489: the wave functions will be destroyed even for small $\mu$. In the

1490: next section some other consequences of deviations from

1491: half-filling are considered.

1492: 

1493: \subsubsection{Deviations from  half-filling}

1494: 

1495: The shape of the Fermi surface depends on $\mu$ and has a strong

1496: influence on the low-energy quasiparticle spectrum. Consider, for

1497: example, low-energy quasiparticle states under the conditions

1498: $|\omega|,\ |\Delta(k_y)|\ll q(k_y)$. This ensures that the

1499: quasiparticle energies lie  close to the Fermi surface and one can

1500: take their energies in the normal metal state to be in the linear

1501: form $\propto \v_F\cdot(\k-\k_{F})$. Under this approximation,

1502: effects of the particle-hole asymmetry are small and one can use a

1503: quasiclassical approximation, which is valid for quasiparticles

1504: close to the Fermi surface. Then only the order parameter on the

1505: Fermi surface $\Delta(k_{F,x}, k_{F,y})$ enters the equations. For

1506: the wire geometry the momentum component $k_{F,y}$ is an

1507: independent parameter, while $k_{F,x}(k_{F,y})$, and

1508: $\Delta(k_{F,x}(k_{F,y}), k_{F,y})$ are actually functions of

1509: $k_{F,y}$. For $\mu=0$ the Fermi surface for the $(110)$ wire is a

1510: square with sides parallel to $x$ or $y$ axes. Hence, $k_{F,x}=\pm

1511: \pi/(2d)$ actually does not depend on $k_{F,y}$ in this case and

1512: $\Delta(\pm\pi/(2d), k_{F,y})=\pm\Delta(k_{F,y})$, in accordance

1513: with Eq.\eqref{dky}. For finite $\mu$ we find

1514: $\Delta(k_{F,x},k_{F,y})=

1515: \dfrac{\Delta(k_{F,y})}{q(k_{F,y})}\sqrt{q^2(k_{F,y})-\mu^2}$.

1516: 

1517: In the case of superconducting wires the equation for

1518: quasiparticle subgap energies near the Fermi surface

1519: $|\omega|<|\Delta(k_{F,x},k_{F,y})|$ takes the form

1520: \begin{eqnarray} \omega^2\left(

1521: \sinh^2\left[(N+1) d\dfrac{\sqrt{\Delta^2(k_{F,x},

1522: k_{F,y})-\omega^2}}{

1523: |v_{x,f}(k_{F,y})|}\right] \right.\nonumber\\

1524: +\sin^2\phi(k_{F,y})\Biggr)= \Delta^2(k_{F,x},

1525: k_{F,y})\sin^2\phi(k_{F,y})~,\label{spectrumassym}

1526: \end{eqnarray}

1527: where $\phi(k_{F,y}) \equiv k_{F,x}d(N+1)$.

1528:  The lowest branches of quasiparticle spectra, which follow

1529: from Eq. \eqref{spectrumassym} for $N=24,25,26$, are shown in

1530: Fig.\ref{fig:mu02spectrum}. The solution of

1531: Eq.~\eqref{spectrumassym} reduces to a simple form in the limit of

1532: large $N$:

1533: 

1534: \bea \omega_A(k_{F,y})=\pm 2\Delta(k_{F,y})\sin\phi(k_{F,y})\times

1535: \qquad \qquad \qquad

1536: \nonumber\\

1537: \exp\left[-(N+1)d\dfrac{|\Delta(k_{F,y})|}{|v_{x,f}(k_{F,y})|}\right]

1538: \enspace . \label{omegaamu} \eea

1539: 

1540: \begin{figure}[!tbh]

1541: \begin{center}

1542: \leavevmode

1543: \includegraphics[clip=true,width=.9\columnwidth]{StripN24-25-26Mu02.eps}

1544: \caption{The lowest energy branches for N=24, 25, 26. The parameters

1545: are $t=2.5$, $\mu=0.2t=0.5$, $\Delta_0=0.2t=0.5$, $\Delta_{max}=2\Delta_0=1$.}

1546: \label{fig:mu02spectrum}

1547: \end{center}

1548: \end{figure}

1549: 

1550: For half-filled wires, when $\mu=0$, the phase $\phi(k_{F,y})$

1551: does not depend on $k_{F,y}$, being equal to $\phi_{\rm odd}=m\pi$

1552: for odd $N$ wires and $\phi_{\rm even}=(m+1/2)\pi$ for even $N$

1553: wires. This difference between the phases $\phi_{\rm odd}$ and

1554: $\phi_{\rm even}$ plays an important role in forming well

1555: pronounced odd-even effects in the spectra of wires with odd and

1556: even numbers of layers. Indeed, for half-filled odd $N$ wires

1557: Eq.~\eqref{omegaamu} reduces to the zero-energy dispersionless

1558: Andreev states. At the same time, for half-filled even $N$ wires

1559: Eq.~\eqref{omegaamu} describes a dispersive branch of

1560: quasiparticle energies, which coincides with Eq.~\eqref{omegaa}

1561: under the condition $|\Delta(k_y)|\ll q(k_y)$.

1562: 

1563: \begin{figure}[t]

1564: \begin{center}

1565: \leavevmode

1566: \includegraphics[clip=true,width=.9\columnwidth]{mu110.eps}

1567: \caption{Local density of states for a (110) wire

1568: with N=10 (left column) and N=11 (right column); $\mu=0$ (upper panels),

1569: $\mu=0.2t$ (lower panels). On each panel the different curves represent

1570: various chains.}

1571: \label{fig:mu110}

1572: \end{center}

1573: \end{figure}

1574: 

1575: For $\mu\ne0$ the phase $\phi(k_{F,y})$ noticeably depends on

1576: $k_{F,y}$: \be \phi(k_{F,y})=k_{F,x}d(N+1)=

1577: (N+1)\cos^{-1}\left(-\dfrac{\mu}{q(k_{F,y})}\right) \enspace .

1578: \label{phi}\ee

1579: Qualitative deviations of low-energy quasiparticle spectra,

1580: shown in Fig.~\ref{fig:mu02spectrum}, from the respective

1581: spectra of half-filled wires (see

1582: Figs.~\ref{figN11},~\ref{figN10}) are associated with the behavior

1583: of the phase $\phi(k_y)$.

1584: The odd-even effect in the spectra of $(110)$ wires

1585: becomes less pronounced in the case of finite $\mu$, as found by

1586: Ziegler et al. \cite{zieglerwire}. For some values of $\mu$ the

1587: spectra of odd $N$ and even $N$ wires may have no qualitative

1588: differences at all. As it follows from Eq.~\eqref{omegaamu}

1589: and, in a more general case, from Eq.~\eqref{spectrumassym},

1590: for $\mu\ne 0$ additional and  strong dispersion of the spectra

1591: comes from the $k_y$-dependence of the phase $\phi(k_{F,y})$. The

1592: larger the $N$, more oscillations of $\sin\phi(k_y)$ take place

1593: with varying the $k_{F,y}$. Hence, more extrema of

1594: $\omega_A(k_{F,y})$ arise. This results in additional peaks in the

1595: LDOS, which appear only in the presence of finite $\mu$.

1596: As seen from Eq.~\eqref{phi}, the phase $\phi$ can considerably vary,

1597: when the film (wire) thickness  varies from $Nd$ to $Nd-l$ and

1598: $a\ll l\ll\xi_0$, $l\ll Nd$. Thus, in averaging the spectrum over the film

1599: thickness, a large number of respective

1600: additional peaks arises, filling the whole low-energy band

1601: in the $(110)$ wire with edges described by Eq.~\eqref{omegaamu}. This

1602: is in agreement with the quasiclassical results \cite{nagai,fominov}.

1603: 

1604: The particle-hole asymmetry is another important feature of the

1605: spectra. It can be well pronounced for finite $\mu$, but lies

1606: beyond the quasiclassical approximation. Fig.\ref{fig:mu110}

1607: displays the asymmetric LDOS, calculated with Eq.(\ref{11}) for

1608: $(110)$ superconducting wires with $N=10$ and $N=11$ in the case

1609: $\mu=0.2t$. For  comparison, the respective LDOS for $\mu=0$ is

1610: also shown. A well-pronounced peak at $\omega=-\mu$ arises with a

1611: deviation from half-filling in the LDOS for the odd $N$ wires. We

1612: remind the reader that in the half-filled normal state $(110)$

1613: wires with odd $N$ the dispersionless zero-energy quasiparticle

1614: states have been found in Sec.\ref{sec:nwires} (see

1615: Eq.(\ref{modefreqnormal}) with $\nu=(N+1)/2$). The wave function

1616: of these states, as well as the residue of the pole-like term in

1617: the Green's function Eq.(\ref{weightnn}), is a standing wave

1618: across the wire, taking zero values on alternating sites.

1619: In the superconducting state the zero-energy standing wave

1620: disappears and the dispersionless zero-energy Andreev surface

1621: states arise. Their weight exponentially decays in the bulk of

1622: wide odd $N$ wires. In the presence of a deviation from

1623: half-filling the energy of the quasiparticle states in the normal

1624: state odd $N$ wires shifts to $-\mu$. These dispersionless states

1625: with finite energy keep the character of standing waves. Further,

1626: in the superconducting wires with finite $\mu$ the low-energy

1627: states become dispersive both for odd and even $N$

1628: wires. As seen from Fig.(\ref{fig:mu02spectrum})

1629: for the wire with $N=25$, the branch with lowest energy has in

1630: this case two extrema. The maximal value of $|\omega|$ for the

1631: states forming this branch, lies at $k_y=0$ and contributes to the

1632: peak at $\omega=-\mu$ associated with the hole contribution, if

1633: $\mu>0$. Since the order parameter vanishes at $k_y=0$, the

1634: respective quasiparticle wave function is a standing wave and the

1635: peak position coincides with that in the normal metal state of the

1636: wire. The minimal value of $|\omega|$ is zero and contributes to

1637: the zero-energy peak. The zero-energy peak is associated with

1638: comparatively large value of $k_y$, comparable with the size of

1639: the Brillouin zone, and the order parameter at this value of $k_y$

1640: is of order  $\Delta_{max}$. Thus, the zero-energy peak is

1641: associated with surface Andreev states, whose quasiparticle

1642: wave function decays in the bulk of wide odd $N$ wires. For narrow

1643: wires, the self-consistency condition becomes important. As shown

1644: in the next section, at finite $\mu$ the self-consistency

1645: condition can lead to more important consequences as compared with

1646: the case $\mu=0$.

1647: 

1648: 

1649: \label{sec:finitemu}

1650: 

1651: 

1652: 

1653: \section{Self-consistent treatment of $\mathbf{(110)}$-wires}

1654: \label{sec:self}

1655: 

1656: 

1657: In previous sections, the order parameter was assumed constant

1658: over the whole width of the wire in order to allow for analytical

1659: solutions. Even for a single $(110)$-surface, however, a

1660: self-consistent treatment of the order parameter gives rise to

1661: interesting effects: the $d_{x^2-y^2}$--wave order parameter is

1662: strongly suppressed near the surface and a complex $is$--admixture

1663: (or some other time-reversal symmetry-breaking state) is possible, which

1664: leads to a splitting of the zero-energy Andreev bound state~\cite{mats95,%

1665: buch95,cov97,fog97,Ting110,deutsch99,greene03,kos01,sigrist,ohashi99,%

1666: vanHarlingen02,deutsch02,deutsch03}.

1667: Even larger effects are therefore to be expected for the wire limited

1668: by two $(110)$-surfaces.

1669: Indeed, our self-consistent evaluation indicates that for very

1670: narrow wires a quasi one-dimensional triplet superconducting state

1671: can replace the conventional $d_{x^2-y^2}+is$--state. For finite

1672: chemical potential

1673: $\mu$ the normal metal state can become energetically favorable

1674: as the ground state for narrow wires, while superconductivity recovers

1675: with increasing wire width. Under special

1676: conditions, even a mixture of singlet- and triplet-pairing can

1677: occur.

1678: It is important to recall at this point that one

1679: expects mean field theory to break down as the 1D limit is

1680: approached even at $T=0$.   Thus the predictions for various kinds

1681: of superconducting order mentioned below are to be treated with

1682: some skepticism as regards quantitative predictions.  Nevertheless

1683: we view our results as presenting intriguing evidence that when

1684: surface energies begin to become comparable to the energy

1685: differences between $d$--wave and other bulk pair channels, strong

1686: fluctuations with symmetries optimal for quasi-1D system,

1687: including spin triplet pair fluctuations, will result.

1688: 

1689: 

1690: 

1691: \subsection{Order parameter}

1692: 

1693: Self-consistent solutions to Hamiltonian~(\ref{ham}) were

1694: obtained by solving the Bogoliubov-de Gennes equations for

1695: $(110)$-wires,

1696: \begin{eqnarray}

1697: \left ( \begin{array}{cc} \xi_{k_y} & \Delta_{k_y} \\

1698:                           \Delta_{-k_y}^* & -\xi_{k_y}

1699:         \end{array} \right)

1700: \left ( u_{k_y\lambda} \atop{v_{k_y\lambda}} \right ) =

1701: E_{k_y\lambda} \left ( u_{k_y\lambda} \atop{v_{k_y\lambda}} \right)

1702: \end{eqnarray}

1703: where $E_{k_y\lambda}$ with $\lambda = 1 \ldots N$

1704: are the eigenvalues of the Bogoliubov-de Gennes

1705: equations and its eigenvectors $u_{k_y\lambda}(n)$ and

1706: $v_{k_y\lambda}(n)$ are the

1707: coefficients of the Bogoliubov transformation:

1708: \begin{eqnarray}

1709: c_{k_yn\uparrow}&=&\sum_\lambda \left \{

1710:                  \gamma_{k_y \lambda \uparrow} u_{k_y \lambda} (n)

1711:            - \gamma_{k_y \lambda \downarrow}^\dag v^*_{k_y \lambda} (n)

1712:         \right \}

1713: \nonumber \\

1714: c_{-k_yn\downarrow}^\dag&=&\sum_\lambda \left \{

1715:                  \gamma_{k_y \lambda \uparrow} v_{k_y \lambda} (n)

1716:            + \gamma_{k_y \lambda \downarrow}^\dag u^*_{k_y \lambda} (n)

1717:         \right \}

1718: \end{eqnarray}

1719: Furthermore, $\xi_{k_y}$ and $\Delta_{k_y}$ are matrices in the $N$

1720: layers of the $(110)$-wire, i.e. the layers in $x$-direction, and are

1721: given by (for notation see Fig.~\ref{fig:strip110}):

1722: \begin{eqnarray}

1723: (\xi_{k_y})_{nn'} &=& -2t \cos k_y \left(

1724: \delta_{n'n+1}+\delta_{n'n-1}

1725:                                           \right)  -\mu \delta_{n'n}

1726: \nonumber \\

1727: \left (\Delta_{k_y} \right)_{nn'} &=&

1728:       \left( \Delta_{nn'}^+e^{-ik_y}+\Delta_{nn'}^-e^{ik_y} \right)

1729:                                                          \delta_{n'n+1}

1730: \nonumber \\

1731: &+&   \left( \Delta_{nn'}^-e^{-ik_y}+\Delta_{nn'}^+e^{ik_y}

1732: \right)

1733:                                                          \delta_{n'n-1}

1734: \end{eqnarray}

1735: The gap values are determined by the following self-consistency

1736: equations:

1737: \begin{eqnarray}

1738: \Delta_{nn+1}^\pm &=& - V \frac{1}{N} \sum_{k_y} e^{\pm ik_y}

1739:           \langle c_{-k_y n+1 \downarrow} c_{k_y n \uparrow} \rangle

1740: \\

1741:                &=& V \frac{1}{N} \sum_{k_y} \sum_\lambda e^{\pm ik_y}

1742:           u_{k_y\lambda}(n+1) v_{k_y\lambda}^*(n)

1743: \nonumber \label{eq:GapSC}

1744: \end{eqnarray}

1745: 

1746: 

1747: 

1748: To simplify the numerical evaluation of the Bogoliubov-de Gennes

1749: equations we consider isolated wires here. For a Hamiltonian on a

1750: discrete lattice like~(\ref{ham}) an isolated wire is equivalent

1751: to a wire limited by lines of unitary impurities, which is the

1752: boundary condition used for the analytical calculations in the

1753: previous sections.

1754: 

1755: \begin{figure}[t]

1756: \begin{center}

1757: \includegraphics*[width=.99\columnwidth]{DdsmaxD01.eps}

1758: \\[.5cm]

1759: \includegraphics*[width=.99\columnwidth]{PhaseDiagD01.eps}

1760: \end{center}

1761: \caption{Narrow wires with a nearest neighbor interaction strength of

1762: $V$=1.1575$t$, which gives rise to a gap value of $\Delta_0$=0.2$t$ for

1763: the bulk system at $\mu$=0.

1764: Lower panels: phase diagram of the wires with even width

1765: (c) and with odd width (d). White space

1766: denotes $\Delta=0$, i.e., the normal state.

1767: Upper panels: corresponding amplitudes of the $d_{x^2-y^2}$ and $is$-order

1768: parameters displaying the value in the center of the wire for

1769: the $d_{x^2-y^2}$--wave case and the value at the edge of the

1770: wire in the $s$--wave case. These are the positions where the largest

1771: values of the respective order parameters are expected in the

1772: usual $d_{x^2-y^2}+is$-state.}

1773: \label{fig:PhaseDiagD01}

1774: \end{figure}

1775: 

1776: 

1777: 

1778: For narrow wires we find a variety of different phases.

1779: The resulting phase diagram for a nearest neighbor

1780: interaction strength of $V$=1.1575$t$, which gives rise to a gap

1781: of $\Delta_0$=0.2$t$ in a bulk system at $\mu$=0, is displayed in

1782: Fig.~\ref{fig:PhaseDiagD01}. For wires with widths up to $N$=9 we

1783: find a new phase over a wide range of chemical potentials

1784: characterized by $\Delta_{ij}=-\Delta_{ji}$ (stars in

1785: Fig.~\ref{fig:PhaseDiagD01}(c), (d)), which is a signature of

1786: triplet-pairing with $S_z=0$, the only triplet component

1787: compatible with Hamiltonian~(\ref{ham}). For $\mu=0$ the amplitude of

1788: $\Delta_{ij}$ oscillates across the wire between zero and its

1789: maximum value (see Fig.~\ref{fig:DensN08}(a)), indicating a

1790: one-dimensional nature of these new triplet-superconducting

1791: correlations (see Fig.~\ref{fig:DensN08}(b)).

1792: Although the oscillating behavior of $\Delta_{ij}$

1793: remains for finite $\mu$, its amplitude no longer vanishes exactly on

1794: alternating layers. Thus the strict one-dimensionality of the superconducting

1795: correlations seems to be a feature peculiar to $\mu=0$.

1796: This new triplet superconducting phase

1797: will be discussed in more detail below.

1798: At first we will focus on the singlet superconducting phase with possible

1799: $d_{x^2-y^2}$-- and $s$--wave order parameters.

1800: 

1801: 

1802: \begin{figure}[t]

1803: \begin{center}

1804: \begin{minipage}{.49\columnwidth}

1805: \includegraphics*[width=.98\columnwidth,clip=true]{TripletStateN6.eps}

1806: \\[.3cm]

1807: \includegraphics*[width=.98\columnwidth,clip=true]{DensN08p2.eps}

1808: \end{minipage}

1809: \begin{minipage}{.49\columnwidth}

1810: \includegraphics*[width=.98\columnwidth,clip=true]{DensN08p1.eps}

1811: \end{minipage}

1812: \end{center}

1813: \caption{Quasi one-dimensional triplet superconducting state for $N$=8 and

1814: $\mu$=0. (a) Bond gap values $\Delta^\pm_{nn+1}$ (for definition see

1815: Eq.~(\ref{eq:GapSC})) across the wire.

1816: (b) Illustration of the one-dimensional structure of the

1817: superconducting correlations for an $N$=6-(110) wire at $\mu=0$.

1818: (c) Density of states starting from the outermost

1819: layer of the wire up to the middle layer of the wire, where for

1820: illustrational purposes each

1821: layer has been shifted by an additional offset of 0.3$t$.}

1822: \label{fig:DensN08}

1823: \end{figure}

1824: 

1825: 

1826: Although we do not understand all details of the variability of

1827: the phase diagram, a few general trends seem clear.

1828: The larger the

1829: width of the wire the more of the usual

1830: $d_{x^2-y^2}+is$--phase (filled circles in

1831: Fig.~\ref{fig:PhaseDiagD01}(c),(d)) is

1832: recovered. The corresponding amplitudes of

1833: the $d_{x^2-y^2}$-- and $is$--components of the order

1834: parameter are displayed in the upper panels of

1835: Fig.~\ref{fig:PhaseDiagD01}. For narrow wires the amplitude of the

1836: $d_{x^2-y^2}$--wave order parameter is finite only for small chemical

1837: potentials $\mu$ and a pure $s$--wave phase

1838: is favorable for large $\mu$. These two phases are separated by a

1839: normal state region.

1840: For larger wire widths the range of the $d_{x^2-y^2}$--wave phase

1841: and the amplitude of the $d_{x^2-y^2}$--wave order parameter

1842: increase. The amplitude of the $s$--wave order parameter, on the other

1843: hand, decreases and the $s$--wave phase moves

1844: towards smaller chemical potentials until it merges with the

1845: $d_{x^2-y^2}$--wave phase, thereby giving rise to a finite

1846: $is$--admixture near the edges of the wire, as expected in

1847: analogy to a single (110)-surface.

1848: 

1849: 

1850: The dependence of the amplitude of the $d_{x^2-y^2}$--wave order

1851: parameter on the width $N$ of the wire is shown in

1852: Fig.~\ref{fig:D01D02}. The upper panels refer to an interaction

1853: strength  of $V$=1.1575$t$, which gives

1854: rise to a gap of $\Delta_0$=0.2$t$ for the bulk system at $\mu$=0.

1855: For $\mu$=0 we observe an even-odd

1856: oscillation in the amplitude of the order parameter (see

1857: Fig.~\ref{fig:D01D02}(a)), which disappears for

1858: larger chemical potentials. The effect of finite $\mu$ on the suppression

1859: of the $d_{x^2-y^2}$--wave order parameter is

1860: considerably stronger in narrow wires than in the bulk system.

1861: For $\mu/t$=0.4 the $d_{x^2-y^2}$--wave order parameter vanishes for

1862: $N<10$ (squares in Fig.~\ref{fig:D01D02}(a)), although the bulk gap

1863: is only reduced by a factor of approx. 0.8 (see

1864: Fig.~\ref{fig:D01D02}(c)).

1865: The dramatic suppression of the $d_{x^2-y^2}$--wave order parameter

1866: for finite $\mu$ is reduced upon consideration of a larger nearest neighbor interaction

1867: strengths, as can be seen in the lower panels of Fig.~\ref{fig:D01D02}.

1868: For a larger interaction strength of $V$=1.7682$t$, which gives rise to

1869: a bulk gap value of

1870: $\Delta_0$=0.4$t$ at $\mu=0$, a pronounced even-odd effect remains at

1871: $\mu/t$=0.4 (see Fig.~\ref{fig:D01D02}(d)) and the

1872: amplitude in the center of an $N$=12-wire is already very close to the bulk

1873: value (see Fig.~\ref{fig:D01D02}(f)) contrary to the smaller

1874: interaction strength, where the amplitude in the center of an

1875: $N$=12-wire is still suppressed by a factor of more than three with

1876: respect to the bulk value (see Fig.~\ref{fig:D01D02}(c)).

1877: 

1878: 

1879: \begin{figure}[t]

1880: \begin{center}

1881: \includegraphics*[width=.99\columnwidth]{D01D02.eps}

1882: \end{center}

1883: \caption{Evolution of the $d_{x^2-y^2}$-wave order parameter with

1884: increasing wire width for two different interaction strengths

1885: $V$=1.1575$t$ (upper panels) and $V$=1.7682$t$ (lower panels), which

1886: give rise to a gap value of $\Delta_0$=0.2$t$ and $\Delta_0$=0.4$t$  for the

1887: bulk system at $\mu=0$, respectively. The first column displays

1888: the amplitude of the $d_{x^2-y^2}$-order parameter as a function of the

1889: wire width for two different chemical potentials $\mu/t$=0 and $\mu/t=0.4$.

1890: The other two columns show the variation of the $d_{x^2-y^2}$- and

1891: $is$-component of the order parameter across the wire for $\mu/t$=0

1892: (center column) and $\mu/t$=0.4 (right column). For comparison the

1893: reduced magnitude of the bulk order parameter at $\mu/t$=0.4

1894: is displayed with dashed lines in the right panels.}

1895: \label{fig:D01D02}

1896: \end{figure}

1897: 

1898: 

1899: 

1900: 

1901: The emergence of a new triplet-superconducting phase for very

1902: narrow wires can be most easily understood by considering  the

1903: smallest wire, i.e., the $N$=2 wire. Although the effect of

1904: fluctuations will be larger for smaller wires, we focus only on

1905: possible solutions of the BCS mean-field Hamiltonian~(\ref{ham})

1906: in the present paper. Inspection of the phase diagram of

1907: Fig.~\ref{fig:PhaseDiagD01} shows that the $d$-wave order

1908: parameter vanishes for the $N$=2-wire. To investigate alternative

1909: ways in which the $N$=2 wire could lower its ground state energy,

1910: we map it to a 1D-chain (see Fig.~\ref{fig:wire1d}). Although the

1911: gap values $\Delta_{ij}$ and $\Delta_{ji}$ could in principle

1912: differ in amplitude and by a phase factor $\phi$, we restrict our

1913: treatment to $\phi=0$ and $\phi=\pi$. Note that $\phi=0$

1914: corresponds to a singlet pairing state, whereas $\phi=\pi$ is a

1915: triplet pairing state with $S_z=0$\cite{tripletfootnote}. To

1916: investigate the possibility of triplet superconductivity in these

1917: systems in more generality, one should retain pairing correlations

1918: with $S_z=\pm 1$ in the mean-field Hamiltonian as well. For now,

1919: however, we are satisfied with the observation that even for the

1920: simple nearest-neighbor pairing Hamiltonian~(\ref{ham}) a triplet

1921: order parameter can be favored over a singlet order parameter in

1922: narrow geometries.

1923: 

1924: \begin{figure}[h]

1925: \begin{center}

1926: \includegraphics*[width=.99\columnwidth]{strip1d.eps}

1927: \end{center}

1928: \caption{Mapping of the $N$=2 wire with $(110)$ orientation to

1929: the 1D-chain with lattice constant $d$. Note that

1930: $\Delta_2=-\Delta_1$ corresponds to a triplet pairing state with

1931: $S_z$=0.} \label{fig:wire1d}

1932: \end{figure}

1933: 

1934: 

1935: After Fourier transformation Hamiltonian~(\ref{ham}) for the

1936: 1D-chain reads

1937: \begin{eqnarray}

1938: H_{\rm MF}&=&-\sum_{k \sigma} (2t \cos kd + \mu)

1939:                             c_{k \sigma}^\dag c_{k \sigma}

1940: \\

1941:        & &    +\sum_k \left \{ (\Delta_1 e^{ikd}+\Delta_2 e^{-ikd})

1942:                             c^\dagger_{-k \downarrow} c^\dagger_{k \uparrow}

1943:                             + {\rm h.c.} \right\}\,, \nonumber

1944: \label{eq:Ham1d}

1945: \end{eqnarray}

1946: where $\Delta_{ii+1}=-V\langle c_{i+1\downarrow}

1947: c_{i\uparrow}\rangle=\Delta_1$ and $\Delta_{i+1i}=-V\langle

1948: c_{i\downarrow} c_{i+1\uparrow}\rangle=\Delta_2$. It can be easily

1949: diagonalized using the

1950: Bogoliubov-transformation

1951: to give a quasiparticle dispersion of:

1952: \begin{eqnarray}

1953: &E_k&=\sqrt{\epsilon_k^2+ \Delta_k^2} \\

1954: &{\rm with}&

1955: \epsilon_k = \frac{q(k)}{2}+\mu=2t \cos kd + \mu

1956: \nonumber\\

1957: & &\!\!\Delta_k^2 =

1958: \Delta_1^2+\Delta_2^2 +2 \Delta_1 \Delta_2 \cos2kd \,, \nonumber

1959: \end{eqnarray}

1960: where the gap values are to be determined from the following

1961: self-consistency  equations:

1962: \begin{eqnarray}

1963: \Delta_1 &=& - V \frac{1}{4\pi d} \int_{-\pi d}^{\pi d} dk

1964:                     \frac{\Delta_1 + \Delta_2 \cos 2kd}{E_k}

1965: \nonumber \\

1966: \Delta_2 &=& - V \frac{1}{4\pi d} \int_{-\pi d}^{\pi d} dk

1967:                     \frac{\Delta_1 \cos 2kd + \Delta_2}{E_k} \,.

1968: \end{eqnarray}

1969: Four different phases emerge from this model by variation of the

1970: chemical potential $\mu$ and the nearest neighbor interaction

1971: strength $V$ (see Fig.~\ref{fig:PhaseDiag1d}). Below a critical

1972: interaction strength the order parameter vanishes and

1973: the normal state is the ground state of the 1D chain. For

1974: intermediate interaction strengths and small chemical potentials

1975: we find $\Delta_1=-\Delta_2$, i.e., a pure triplet superconducting

1976: state, whereas for large chemical potentials the ground state is

1977: characterized by $\Delta_1=\Delta_2$, i.e., a singlet extended

1978: $s$-wave state. In the singlet state the gap function is $\Delta_k

1979: \sim \cos kd$, i.e., it has nodes at $kd=\pm \pi/2$ whereas the

1980: triplet state gap function is $\Delta_k \sim \sin kd$ and therefore

1981: is maximum at $kd=\pm \pi/2$. This explains why the triplet state

1982: is favored over the singlet state for $\mu=0$, where the Fermi-surface

1983: of the nearest neighbor tight-binding model is at $kd=\pm \pi/2$.

1984: For larger chemical potentials the situation is reversed and the

1985: singlet extended $s$-wave state becomes more favorable than the

1986: triplet state.

1987: For large interaction strengths, we find different magnitudes for

1988: $\Delta_1$ and $\Delta_2$, which corresponds to a mixture of

1989: singlet and triplet pairing. In the limit $V \to \infty$, either

1990: of the gap values $\Delta_1$ and $\Delta_2$ approaches zero

1991: whereas the other goes to infinity, corresponding to an admixture

1992: of triplet and singlet order parameters with equal amplitudes.

1993: 

1994: \begin{figure}[t!]

1995: \begin{center}

1996: \includegraphics*[width=.85\columnwidth,clip=true]{PhaseDiag1d.eps}

1997: \end{center}

1998: \caption{Phase diagram for 1D-chain as a function of chemical

1999: potential $\mu/t$ and nearest neighbor interaction $V/t$.}

2000: \label{fig:PhaseDiag1d}

2001: \end{figure}

2002: 

2003: 

2004: With the phase diagram of the 1D-chain in mind we are now able to

2005: better understand the phase diagram of narrow wires as displayed

2006: in Fig.~\ref{fig:PhaseDiagD01}. For small $\mu$ the new phase with

2007: $\Delta_{ij}=-\Delta_{ji}$ simply arises from the formation of

2008: quasi one-dimensional triplet-superconducting correlations along $N$=2

2009: wires (see Fig.~\ref{fig:DensN08}(a),(b)).

2010: This also explains why this new phase is more favorable

2011: for wires with even width than with odd width (compare left and right panels in

2012: Fig.~\ref{fig:PhaseDiagD01}), as only the former can be

2013: divided evenly into $N$=2 wires.

2014: Note that also for the narrow wires,

2015: a critical coupling strength of similar magnitude as in the

2016: 1D-chain is necessary to induce the triplet superconducting state,

2017: which is, however, smaller than the

2018: interaction strength considered in this paper.

2019: At larger chemical potentials we find an extended $s$-wave order

2020: parameter analogously to the one-dimensional chain. More difficult

2021: to reconcile with the phase diagram of the one-dimensional chain,

2022: however, are  the seemingly arbitrarily

2023: distributed mixtures of singlet- and triplet-superconducting order

2024: parameters (crosses in Fig.~\ref{fig:PhaseDiagD01}).

2025: The normal state regions, which occur for narrow wires and

2026: intermediate chemical potentials in Fig.~\ref{fig:PhaseDiagD01},

2027: reflect the fact that superconductivity becomes less favorable in

2028: finite geometries and a critical coupling strength is necessary to

2029: induce it \cite{nagai}.

2030: 

2031: 

2032: 

2033: \subsection{Density of states}

2034: 

2035: 

2036: 

2037: 

2038: 

2039: 

2040: \begin{figure}[t]

2041: \begin{center}

2042: \includegraphics*[width=.99\columnwidth,clip=true]{DensN10N11.eps}

2043: \end{center}

2044: \caption{Density of states for the $N$=10-wire (left panels) and

2045: the $N$=11-wire (right panels). Upper panels show the

2046: non-selfconsistent results assuming a constant $d_{x^2-y^2}$-order

2047: parameter of $\Delta_0$=0.2$t$, whereas the lower panels display

2048: the density of states resulting from the self-consistently

2049: determined values of the $d$-wave order parameter, whose variation

2050: throughout the wire is depicted in the insets. All panels show the

2051: density of states starting from the outermost layer of the wire up

2052: to the middle layer of the wire, where for each layer towards the

2053: center of the wire an additional offset of 0.3$t$ has been used.}

2054: \label{fig:DensN10N11}

2055: \end{figure}

2056: 

2057: 

2058: 

2059: 

2060: What does the phase diagram in Fig.~\ref{fig:PhaseDiagD01} imply

2061: about the possible existence of Andreev bound states in narrow

2062: wires? In Fig.~\ref{fig:DensN08}(c) the local density of states and

2063: the variation of the order parameter throughout the wire is

2064: depicted for the quasi one-dimensional triplet superconducting

2065: state using the $N$=8-wire at $\mu$=0 as an example. Obviously,

2066: the density of states in this state is fully gapped, in analogy to

2067: the one-dimensional chain, and there are no Andreev bound states.

2068: 

2069: Close to half-filling, for wider wire widths, the results are more

2070: conventional and the effects of self-consistency much simpler.  In

2071: Fig.~\ref{fig:DensN10N11} the local density of states is displayed

2072: for the $N$=10- and the $N$=11-wire at $\mu=0$, which are

2073: characterized by a pure $d_{x^2-y^2}$-order parameter. Here, the

2074: main difference between the self-consistent (lower panels of

2075: Fig.~\ref{fig:DensN10N11}) and the non-self consistent results

2076: (upper panels of Fig.~\ref{fig:DensN10N11}) is the suppression of

2077: the magnitude of the $d$-wave order parameter especially towards

2078: the edges of the wire (see also insets of

2079: Fig.~\ref{fig:DensN10N11}). Whereas the  weight of the zero-energy

2080: peak is reduced in the even-width ($N$=10) wire the weight of

2081: the zero-energy state in the odd-width wire ($N$=11)

2082: is hardly affected.

2083: 

2084: 

2085: 

2086: 

2087: 

2088: 

2089: 

2090: 

2091: 

2092: 

2093: 

2094: 

2095: 

2096: \section{conclusion}

2097: 

2098: \label{sec:conclusion}

2099: 

2100: Motivated by recent scanning tunneling experiments on the

2101: Ba$_2$Sr$_2$CaCu$_2$O$_8$ systems which reveal inhomogeneous

2102: electronic structure on the nanoscale, we have analyzed in detail

2103: the electronic structure of $d$-wave quantum wires. These wires

2104: exemplify the effects of a constrained geometry while still being

2105: simple enough to allow for analytical solutions. To impose a

2106: restricted geometry we use lines of impurities with infinite

2107: scattering strength, a method which allows to cut arbitrarily

2108: shaped objects out of the two-dimensional plane. In principle, it

2109: is straightforward to extend this method to investigate the

2110: effects of tunneling between neighboring grains by reducing the

2111: scattering strength of the impurities and thus lowering the

2112: potential barrier between neighboring grains.

2113: 

2114: New and interesting physics arises from interference effects

2115: between the two surfaces of the wire when its width is of the

2116: order of the superconducting coherence length. Contrary to

2117: $s$-wave superconductors in finite geometries, the surface pair

2118: breaking plays an important role. In this respect, the existence

2119: and nature of Andreev bound states in constricted geometry is of

2120: particular interest. In order to single out new effects peculiar

2121: to quantum wires and arising from the interference of the two

2122: surfaces, we have first addressed the case of a single surface in

2123: the first part of the paper, concentrating on surfaces with

2124: $(100)$-, $(210)$- and $(110)$-orientations in a nearest neighbor

2125: tight-binding model at half-filling. Andreev bound states can form

2126: on surfaces with orientations deviating from the $(100)$ direction

2127: due to the sign change of the $d$-wave order parameter. However, in

2128: the presence of several channels for reflection of quasiparticles

2129: from the surface, the zero-energy Andreev states may not exist, as

2130:  is the case for the $(210)$ surface of the square lattice. Here

2131: our results are in qualitative agreement with earlier work based

2132: either on the quasiclassical approximation or Bogoliubov-de Gennes

2133: equations. Only for the $(110)$-surface, which involves the

2134: strongest pair-breaking, do we find a zero-energy Andreev bound

2135: state, whose amplitude decreases with the square of the inverse

2136: distance from the surface and vanishes on even layers.

2137: 

2138: In the main part of this work, we  focussed on quantum wires with

2139: $(110)$-orientation, which

2140: display a pronounced ``width parity" effect.  The special case of

2141: electrons hopping on a square lattice with half-filled

2142: tight-binding band was treated most extensively.  For wires of

2143: this type with an odd number of chains $N$ only one subgap state,

2144: a dispersionless zero-energy Andreev-bound state, exists. As in

2145: the single surface case, the amplitude of the ZES vanishes on even

2146: layers and decays towards the center of the wire. In addition to

2147: the zero-energy state (ZES), there are $N$-1 dispersive modes,

2148: which are doubly degenerate due to particle-hole symmetry and are

2149: not Andreev states.

2150: 

2151: In the case of even $N$ half-filled wires, a splitting of the

2152: branches occurs which is associated with a symmetry breaking $k_y

2153: \to - k_y$, and a total of 2$N$ dispersive modes exist. Although

2154: there is no dispersionless ZES,  all quasiparticle states in the

2155: even width wire are of Andreev character in the sense that the

2156: current of the probability density vanishes due to opposite

2157: contributions from particle and hole excitations.  These

2158: quasiparticles occupy either conventional Andreev-type surface

2159: states, or a new type of Andreev standing wave, according to their

2160: momentum $k_y$ parallel to the wire. In the 2D limit $N \to

2161: \infty$,  the lowest energy dispersive state was shown to

2162: transform into the usual zero-energy Andreev bound state at the

2163: impenetrable surface, while the standing wave states evolve

2164: into either the continuous spectrum or the surface states. With

2165: increasing deviation from  half-filling, odd-even effects in

2166: the wires become less pronounced. The evolution of the Fermi surface

2167: shape with these deviations can result in additional extrema in

2168: the quasiparticle dispersive modes and, hence, new peaks in the

2169: LDOS. Large-scale faceting of the surfaces with characteristic

2170: scales larger than the wire thickness will not influence our

2171: results significantly. However, small-scale inhomogeneities like point-like

2172: defects and impurities can substantially change the effects of

2173: interference induced by wire surfaces, even if the phase breaking

2174: length is large.

2175: 

2176: The small coherence length of high-temperature superconductors

2177: leads to further restrictions on the applicability of the

2178: quasiclassical results to superconducting wires or films, whose width is

2179: less than or comparable to the coherence length. For narrow wires of

2180: $(110)$-orientation, the self-consistent treatment of the

2181: order-parameter is found to have a large effect. For wires with

2182: widths less than the superconducting coherence length (up to $N$=9

2183: for the particular parameters used in our numerical calculations),

2184: especially for the even width wires, a new phase characterized by

2185: quasi one-dimensional triplet pairing is found in the mean-field

2186: phase diagram. This new phase is fully gapped and characterized by

2187: the absence of Andreev bound states. The larger the width of the

2188: strip, the more the $d+is$-state, which is expected for a single

2189: surface with $(110)$-orientation, dominates the phase diagram near

2190: half filling. With respect to the density of states at

2191: half-filling, where the order parameter has only a $d$-wave

2192: component, the main effect of the self-consistent treatment is the

2193: suppression of the magnitude of the $d$-wave order parameter,

2194: especially near the surfaces of the wire.

2195: 

2196: It is interesting to end this discussion with some speculations on

2197: the role of bound quasiparticle states and edge effects of this

2198: type on the spectra of weakly coupled superconducting grains as

2199: apparently observed in STM experiments. Such irregular grains

2200: should contain nanoscale ``facets" at all possible angles, so

2201: presumably the most general situation with sizeable particle-hole

2202: asymmetry and mixture of even- and odd-$N$ boundary conditions

2203: will apply. If we first assume that the pair interaction and grain

2204: size are such that one may ignore the triplet states found in the

2205: self-consistent treatment, we expect the spectra of weakly coupled

2206: grains to be dominated by the zero-dimensional analogs of the

2207: dispersive subgap states discussed here, i.e. there should be a

2208: wide distribution of bound state energies depending on local

2209: geometry of the grain, and visible in the LDOS as measured by STM.

2210: In this sense we question whether the weakly-coupled grain picture

2211: is in fact applicable to the experiments in question, which appear

2212: to see a very {\it homogeneous} spectrum  at low energies in the

2213: superconducting state.

2214: 

2215: A second remark is based on our observation that in nanoscale

2216: confined geometry, spin triplet fluctuations may become more

2217: favorable. Such time-reversal symmetry breaking fluctuations will

2218: clearly lead to local spontaneous currents, an issue which has

2219: recently been raised again in angle-resolved photoemission

2220: studies\cite{CampuzanoTbreaking}. Future studies of small grains

2221: are planned to address these issues.

2222: 

2223: \section{Acknowledgments}

2224: We thank I.~Bobkova, Ya.~Fominov, T.~Kopp, J.~Mannhart and K.~Ziegler for useful

2225: discussions. This work was supported, in part, by grant RFBR 02-02-16643

2226: (A.M.B. and Yu.S.B.) and NSF-INT-0340536 (P.J.H. and Yu.S.B). A.M.B.

2227: acknowledges the support of Dynasty Foundation and T.S.N.

2228: the support of the Alexander von Humboldt Foundation.

2229: 

2230: 

2231: \begin{thebibliography}{stuffstuffstuff}

2232: 

2233: \bibitem{Hess91} H.F. Hess, R.B. Robinson, R.C. Dynes, J.M. Valles,

2234: and J.V. Waszak, Phys. Rev. Lett. {\bf 62}, 214 (1989).

2235: 

2236: \bibitem{Fischer95} I. Maggio-Aprile, Ch. Renner, A. Erb, E. Walker,

2237: and \O. Fischer, Phys. Rev. Lett. {\bf 75}, 2754 (1995).

2238: 

2239: \bibitem{yazdani} Ali Yazdani, C. M. Howald, C. P. Lutz,

2240: A. Kapitulnik, and D. M. Eigler, Phys. Rev. Lett.  {\bf 83}, 176

2241: (1999).

2242: 

2243: \bibitem{davisnative} E.W. Hudson, S.H. Pan, A.K. Gupta, K-W Ng, and J.C.

2244: Davis, Science {\bf 285}, 88 (1999)

2245: 

2246: \bibitem{davisZn}

2247: S. H. Pan, E. W. Hudson, K. M. Lang, H. Eisaki, S. Uchida, J. C. Davis,

2248: Nature, 403, 746 (2000).

2249: 

2250: 

2251: \bibitem{cren} T. Cren {\em et al.}, Phys. Rev. Lett. {\bf 84}, 147 (2000).

2252: 

2253: \bibitem{davisinhom1} S.-H. Pan et al., Nature

2254: {\bf 413}, 282 (2001).

2255: 

2256: \bibitem{delozanne} D. J. Derro, E. W. Hudson, K. M. Lang, S. H. Pan,

2257: J. C. Davis, J. T. Markert, and A. L. de Lozanne, Phys. Rev. Lett.

2258: 88, 097002 (2002).

2259: 

2260: \bibitem{howald} C. Howald et al., Phys. Rev. B 64, 100504 (2001).

2261: 

2262: \bibitem{davisinhom2} K.M. Lang et al., Nature 415, 412 (2002).

2263: 

2264: \bibitem{vonDelftreview}

2265: J. von Delft, D. C. Ralph, Phys. Rep. 345, 61 (2001).

2266: 

2267: \bibitem{Tinkham}

2268: C. N. Lau, N. Markovic, M. Bockrath, A. Bezryadin, and M. Tinkham,

2269: Phys. Rev. Lett. 87, 217003 (2001).

2270: 

2271: \bibitem{zieglerwire} K. Ziegler, W.A. Atkinson and P.J.

2272: Hirschfeld, Phys. Rev. B 64, 054512 (2001).

2273: %

2274: \bibitem{geerk88}

2275: J.~Geerk, X.~Xi and G.~Linker,

2276: Z.~Phys. B {\bf 73}, 329 (1988).

2277: %

2278: \bibitem{hu94}

2279: C.-R.~Hu, Phys. Rev. Lett. {\bf 72}, 1526 (1994).

2280: %

2281: \bibitem{tan95}

2282: Y.~Tanaka and S.~Kashiwaya,

2283: Phys.~Rev.~Lett. {\bf 74}, 3451 (1995).

2284: %

2285: \bibitem{mats95}

2286: M.~Matsumoto and H.~Shiba,

2287: J.~Phys.~Soc.~Jpn. {\bf 64}, 1703 (1995);

2288: {\bf 64}, 3384 (1995); {\bf 64}, 4867 (1995); {\bf 65}, 2194 (1996).

2289: %

2290: \bibitem{buch95}

2291: L.~J.~Buchholtz, M.~Palumbo, D.~Rainer, and J.~A.~Sauls,

2292: J. Low Temp. Phys. {\bf 101}, 1079 (1995); {\bf 101}, 1099 (1995).

2293: %

2294: \bibitem{tan952}

2295: S.~Kashiwaya, Y.~Tanaka, M.~Koyanagi, H.~Takashima and K.~Kajumura,

2296: Phys. Rev. B {\bf 51}, 1350 (1995).

2297: %

2298: \bibitem{cov96}

2299: M.~Covington, R.~Scheuerer, K.~Bloom, and L.~H.~Greene,

2300: Appl.~Phys.~Lett. {\bf 68}, 1717 (1996).

2301: %

2302: \bibitem{xu96}

2303: J.~H.~Xu, J.~H.~Miller, Jr., and C.~S.~Ting,

2304: Phys. Rev. B {\bf 53}, 3604 (1996).

2305: %

2306: \bibitem{fog97}

2307: M.~Fogelstr\"om, D.~Rainer, and J.~A.~Sauls,

2308: Phys.~Rev.~Lett. {\bf 79}, 281 (1997).

2309: %

2310: \bibitem{bbs97} Yu.~Barash, H.~Burkhardt, A.~Svidzinsky,

2311: Phys. Rev. B {\bf 55}, 15282 (1997).

2312: %

2313: \bibitem{cov97}

2314: M.~Covington, M.~Aprili, E.~Paraoanu, L.~H.~Greene, F.~Xu, J.~Zhu, and

2315: C.~A.~Mirkin, Phys.~Rev.~Lett. {\bf 79}, 277 (1997).

2316: %

2317: \bibitem{alf97}

2318: L.~Alff, H.~Takashima, S.~Kashiwaya, N.~Terada, H.~Ihara, Y.~Tanaka,

2319: M.~Koyanagi, and K.~Kajimura,  Phys. Rev. B {\bf 55}, R14757 (1997).

2320: %

2321: \bibitem{ek97}

2322: J.~W.~Ekin, Y.~Xu, S.~Mao, T.~Venkatesan, D.~W.~Face, M.~Eddy,

2323: and S.~A.~Wolf,

2324: Phys.~Rev.~B {\bf 56}, 13746 (1997).

2325: %

2326: \bibitem{ap98}

2327: M.~Aprili, M.~Covington, E.~Paraoanu, B.~Niedermeier, and L.~H.~Greene,

2328: Phys.~Rev.~B {\bf 57}, R8139 (1998).

2329: %

2330: \bibitem{alf981}

2331: L.~Alff, S.~Kleefisch, U.Schoop, M.~Zittartz, T.~Kemen, T.~Bauch,

2332: A.~Marx , and R.~Gross,

2333: Eur.~Phys.~J. B {\bf 5}, 423 (1998).

2334: %

2335: \bibitem{alf982}

2336: L.~Alff, A.~Beck, R.~Gross, A.~Marx, S.~Kleefisch, Th.~Bauch, H.~Sato,

2337: M.~Naito, and G.~Koren,

2338: Phys.~Rev.~B {\bf 58}, 11197 (1998).

2339: %

2340: \bibitem{sin98}

2341: S.~Sinha and K.-W.~Ng,

2342: Phys.\,Rev.\,Lett. {\bf 80}, 1296 (1998).

2343: %

2344: \bibitem{wei98}

2345: J.~Y.~T.~Wei, N.-C.~Yeh, D.~F.~Garrigus, and M.~Strasik,

2346: Phys.~Rev.~Lett. {\bf 81}, 2542 (1998).

2347: %

2348: \bibitem{apr99}

2349: M.~Aprili, E.~Badica, and L.~H.~Green,

2350: Phys.~Rev.~Lett. {\bf 83}, 4630 (1999).

2351: %

2352: \bibitem{deutsch99}

2353: R.~Krupke and G.~Deutscher,

2354: Phys.~Rev.~Lett. {\bf 83}, 4634 (1999).

2355: %

2356: \bibitem{Ting110} J.-X. Zhu, B. Friedman, and C.S. Ting, Phys.

2357: Rev. B 59, 3353 (1999).

2358: %

2359: \bibitem{cov00}

2360: M.~Covington and L.~H.~Green,

2361: Phys.~Rev.~B {\bf 62}, 12440 (2000).

2362: %

2363: \bibitem{pairor02}

2364: P.~Pairor and M.~B.~Walker,

2365: Phys.~Rev.~B {\bf 65}, 064507 (2002).

2366: %

2367: \bibitem{greene02}

2368: H.~Aubin, L.H.~Greene, Sha Jian and D.G.~Hinks,

2369: Phys.~Rev.~Lett. {\bf 89}, 177001 (2002).

2370: %

2371: \bibitem{wumou03}

2372: Shin-Tza Wu and Chung-Yu Mou,

2373: Phys. Rev.~B {\bf 67}, 024503 (2003).

2374: %

2375: \bibitem{greene03}

2376: L.~H.~Green, P.~Hentges, H.~Aubin, M.~Aprili, E.~Badica,

2377: M.~Covington, M.~M.~Pafford, G.~Westwood, W.~G.~Klemperer,

2378: S.~Jian, D.~G.~Hinks,

2379: Physica C {\bf 387}, 162 (2003).

2380: %

2381: \bibitem{tan96}

2382: Y.~Tanaka and S.~Kashiwaya,

2383: Phys.~Rev.~B {\bf 53}, 11957 (1996).

2384: %

2385: \bibitem{bbr96}

2386: Yu.~S.~Barash, H.~Burkhardt, and D.~Rainer,

2387: Phys.~Rev.~Lett. {\bf 77}, 4070 (1996).

2388: %

2389: \bibitem{ilichev01}

2390: E. Il'ichev et al.,

2391: Phys.~Rev.~Lett. {\bf 86}, 5369 (2001).

2392: %

2393: \bibitem{blamire03}

2394: G.~Testa, A.~Monaco, E.~Esposito, E.~Sarnelli, D.-J.~Kang, E.~J.~Tarte,

2395: S.~H.~Mennema, and M.~G.~Blamire,

2396: cond-mat/0310727

2397: %

2398: \bibitem{walter98}

2399: H.~Walter, W.~Prusseit, R.~Semerad, H.~Kinder, W.~Assmann, H.~Huber,

2400: H.~Burkhardt, D.~Rainer, and J.~A.~Sauls,

2401: Phys.~Rev.~Lett. {\bf 80}, 3598 (1998).

2402: %

2403: \bibitem{bkk00}

2404: Yu.~S.~Barash, M.~S.~Kalenkov and J.~Kurkij\"arvi,

2405: Phys.~Rev.~B {\bf 62}, 6665 (2000).

2406: %

2407: \bibitem{carr01}

2408: A.~Carrington, F.~Manzano, R.~Prozorov, R.~W.~Giannetta, N.~Kameda,

2409: and T.~Tamegai,

2410: Phys.~Rev.~Lett. {\bf 86}, 1074 (2001).

2411: %

2412: \bibitem{tan00}

2413: S.~Kashiwaya and Y.~Tanaka,

2414: Rep.~Prog.~Phys. {\bf 63}, 1641 (2000).

2415: %

2416: \bibitem{wendin01}

2417: T.~L\"ofwander, V.~S.~Shumeiko and G.~Wendin,

2418: Supercond.~Sci.~Technol. {\bf 14}, R53 (2001).

2419: %

2420: \bibitem{giessibl}F.J. Giessibl, Rev. Mod. Phys. 75, 949 (2003).

2421: %

2422: \bibitem{nagai}

2423: Y.~Nagato and K.~Nagai, Phys. Rev. B {\bf 51}, 16254 (1995).

2424: %

2425: \bibitem{Walker}M. B. Walker and P. Pairor, Phys. Rev. B {\bf 59}, 1421

2426: (1999).

2427: %

2428: \bibitem{Joglekar} Y.N. Joglekar, A.H. Castro Neto, and A.V.

2429: Balatsky, Phys. Rev. Lett. {\bf 92}, 037004 (2004).

2430: %

2431: \bibitem{MorrDemler} D. K. Morr and E. Demler, cond-mat/0010460

2432: %

2433: \bibitem{Tamm}I. Tamm, Phys. Z. Sowjetunion 1, 733 (1932).

2434: %

2435: \bibitem{ginzburg} V.L. Ginzburg, D.A. Kirzhnits, Zh.~\'Eksp. Teor. Fiz. {\bf

2436: 46}, 397 (1964) [Sov.~Phys.~JETP {\bf 19}, 269 (1964)].

2437: %

2438: \bibitem{gantmakher}

2439: V.F. Gantmakher, Y.B. Levinson, {\it Carrier Scattering in Metals and

2440: Semiconductors (Modern Problems in Condensed Matter Sciences, vol. 19)},

2441: Elsevier Science Ltd., 1987.

2442: %

2443: \bibitem{ebz} In the presence of half-filling, surfaces with $(110)$

2444: orientations represent a special case for the square lattice, since

2445: the parallel component of the Fermi momentum coincides in this case with

2446: the edge of the surface-adapted Brillouin zone. In the problems we study,

2447: the role of the edge of the Brillouin zone becomes quite important, in

2448: particular, for this reason. In a more general case one can expect, that

2449: extrema of the quasiparticle spectra would arise at other values of the

2450: momentum component parallel to the surface, which are comparable with

2451: the scale of the Brillouin zone.

2452: %

2453: \bibitem{fominov}

2454: Ya.~V.~Fominov and A.~A.~Golubov, cond-mat/0402381.

2455: %

2456: \bibitem{kos01} S. Kos, Phys. Rev. B {\bf 63}, 214506 (2001).

2457: %

2458: \bibitem{sigrist} C.~Honerkamp, K.~Wakabayashi and M.~Sigrist,

2459: Europhys. Lett. {\bf 50}, 368 (2000).

2460: %

2461: \bibitem{ohashi99} Y. Ohashi, Phys. Rev. B {\bf 60}, 15388 (1999).

2462: %

2463: \bibitem{vanHarlingen02} W.K.~Neils and D.J.~Van~Harlingen,

2464: Phys.~Rev.~Lett. {\bf 88}, 047001 (2002).

2465: %

2466: \bibitem{deutsch02} A.~Sharoni, O.~Millo, A.~Koren, Y.~Dagan,

2467: R.~Beck, G.~Deutscher, G.~Koren, Phys.~Rev.~B {\bf 65}, 134526 (2002).

2468: %

2469: \bibitem{deutsch03} A.~Kohen, G.~Leibovitch, and G.~Deutscher,

2470: Phys.~Rev.~Lett. {\bf 90}, 207005 (2003).

2471: %

2472: \bibitem{tripletfootnote} In terms of standard representation of singlet and triplet pairings

2473: $\hat{\Delta}_s=\Delta i\hat{\sigma}_2$, $\hat{\Delta}_t=(\bm{d}\cdot\hat{\bm{\sigma}})i\hat{\sigma}_2$

2474: one finds for the order parameters we introduced $\Delta_1=\Delta+d_z$, $\Delta_2=\Delta-d_z$,\, $d_x=d_y=0$.

2475: Hence, in the purely singlet case $\Delta_1=\Delta_2$, while for the triplet pairing $\Delta_1=-\Delta_2$.

2476: Otherwise there is a mixture of singlet and triplet pairings.

2477: %

2478: \bibitem{CampuzanoTbreaking} A. Kaminski, S. Rosenknanz, H.M. Fretwell, J.C. Campuzano, Z. Li, H. Raffy,

2479: W.G.  Cullen, H. You, C.G.  Olson, and C.M. Varma, Nature {\bf 416}, 610

2480: (2002).

2481: %

2482: \end{thebibliography}

2483: \end{document}

2484: