cond-mat0703540/fsr.tex
1: % "VERSION 6: last correction:  August 29, 2007 by PK"
2: % 
3: \tolerance = 10000
4: % final prl-format (4 pages, for preprint server):
5: \documentclass[twocolumn,showpacs,prl,amsmath,amssymb,floatfix]{revtex4}
6: %
7: % preprint format (for submission)
8: %\documentclass[preprint,showpacs,prl,amsmath,amssymb,floatfix]{revtex4}
9: %
10: % one-sided prl format:
11: %\documentclass[galley,showpacs,prl,amsmath,amssymb]{revtex4}
12: %
13: % take this for final PRB format
14: %\documentclass[twocolumn,showpacs,prb,amsmath,amssymb,floatfix,eqsecnum]{revtex4}
15: %
16: %\documentclass[twocolumn,showpacs,superscriptaddress,prb,amsmath,amssymb,floatfix,eqsecnum]{revtex4}
17: %
18: %\usepackage{dcolumn}
19: \usepackage{amsmath,amssymb}
20: %
21: %\usepackage{showlabels}
22: %\usepackage{drftcite}
23: %
24: \usepackage{bm}
25: \usepackage{epsfig}
26: \usepackage{psfrag}
27: 
28: % set \bd to \bf or \bm
29: \newcommand{\bd}{\bm}
30: \newcommand{\qq}{\frac{q^2}{2m}}
31: \newcommand{\on}{\tilde\omega_{n}}
32: \newcommand{\xq}{\frac{q'q}{2m}}
33: 
34: \begin{document}
35: 
36: \title{Confined coherence in quasi-one-dimensional metals} 
37: 
38: 
39: \author{Sascha Ledowski and Peter Kopietz}
40:   
41: \affiliation{Institut f\"{u}r Theoretische Physik, Universit\"{a}t
42:   Frankfurt,  Max-von-Laue Strasse 1, 60438 Frankfurt, Germany}
43: 
44: %\date{\today}
45: \date{March 28, 2007}
46: 
47: 
48: 
49: 
50:  \begin{abstract}
51: 
52: We present a functional renormalization group calculation
53: of the effect of strong interactions on the
54: shape of the Fermi surface  of weakly coupled  metallic chains.
55: In the regime where the bare interchain hopping is small, 
56: we show that scattering processes involving large momentum
57: transfers perpendicular to the chains
58: can completely destroy  the warping of the true Fermi surface, 
59: leading to a confined state where
60: the renormalized interchain hopping vanishes and a
61: coherent motion perpendicular to the chains is impossible.
62: %In the confined state the system
63: %is a non-Fermi liquid with large anomalous dimension.
64: % anomalous dimension of the Fermi fields 
65: % is larger than
66: % unity, so that the momentum distribution is analytic.  
67: %We also comment on the
68: %truncation scenario  advanced by Rice and co-authors, according to which
69: %only certain sectors of the FS survive in the presence
70: %of strong interactions.
71: 
72: 
73: 
74: 
75: \end{abstract}
76: 
77: \pacs{71.10.Pm, 71.27.+a,71.10.Hf}
78: 
79: \maketitle
80: 
81: 
82: {\it{Introduction.}} 
83:  Electron-electron interactions can strongly modify the
84: Fermi surface (FS) of a metal.  A well known example is the
85: Pomeranchuk transition, where
86: the symmetry of the FS is spontaneously broken due
87: to strong interactions in the zero-sound channel~\cite{Pomeranchuk58}.
88: However, there are other quantum phase transitions 
89: associated with the geometry or the topology
90: of the FS without symmetry breaking, such as the
91: Lifshitz transition \cite{Lifshitz60,Quintanilla06} 
92: or the truncation transition \cite{Furukawa98,Ferraz03}, where certain sectors of the FS
93: are washed out by interactions, while others remain intact.
94: % The latter scenario has been studied  
95: % in Refs.~[\onlinecite{Furukawa98,Ferraz03}]
96: % in order to understand the 
97: % non-Fermi liquid state of the slightly doped cuprates.
98: Another example is the interaction-induced confinement transition, 
99: which has been  proposed by Clarke, Strong, and Anderson more than 
100: ten years ago~\cite{Clarke94}:  they considered
101: metallic chains with small interchain hopping $t_{\bot , 0}$.
102: For weak interactions, the FS 
103: consists then of two disconnected weakly curved sheets as shown
104: Fig.~\ref{fig:FS}.
105: %
106: %
107: %Recently quantum phase transitions involving the
108: %  geometry or the topology of the FS~\cite{Quintanilla06}. Other examples are the  
109: %  Pomeranchuk transition\cite{Pomeranchuk58} (where the symmetry of the FS is
110: %  spontaneously broken), the
111: %  Lifshitz transition \cite{Lifshitz60} (which is associated with a change
112: %  of the topology of the FS without symmetry breaking), or the
113: %  truncation transition~\cite{Furukawa98,Ferraz03}
114: %  (where certain sectors of the FS are destroyed by interactions).
115: %
116: %
117: \begin{figure}[tb]
118: % \begin{center}
119:   \centering
120: %  \psfrag{t}{$\theta$}
121:   \epsfig{file=fig1.eps,width=55mm}
122: %   \end{center}
123:   \vspace{-4mm}
124:   \caption{%
125: (Color online) 
126: FS of a two-dimensional array of weakly coupled metallic chains.
127: The dashed lines mark the boundary of the first
128: Brillouin zone in the direction perpendicular to the chains.
129: }
130:   \label{fig:FS}
131: \end{figure}
132: %
133: %
134: The amplitude of the warping of the FS
135: is proportional to the renormalized interchain hopping $t_{\bot}$.
136: Clarke {\it{et al.}} \cite{Clarke94} suggested that, at least for 
137: sufficiently strong interactions and small $t_{\bot,0}$, 
138: the renormalized $t_{\bot}$ vanishes. 
139: The true FS is then completely flat, so that a coherent motion
140: of the electrons in the direction perpendicular to the chains 
141: is not possible (confined coherence).
142: 
143: In the past decade the confinement problem in weakly coupled metallic chains
144: has been studied by many 
145: authors \cite{Clarke94,Wen90,Kopietz94,Boies95,Arrigoni98}, 
146: but the results have not converged due to a lack of controlled methods.
147: A simple one-loop calculation \cite{Wen90,Boies95} suggests that
148: the renormalized interchain hopping vanishes  if the anomalous dimension
149: $\eta$ of the Luttinger liquid state without interchain hopping is
150: larger than unity. However, this argument does not take the renormalization of $\eta$
151: by interchain hopping into account.
152: Indeed, more refined calculations by Arrigoni \cite{Arrigoni98}
153: suggest that higher order corrections are important and possibly lead to a finite
154: $t_{\bot}$ even for $\eta > 1$.
155: In this Letter we shall re-examine this problem using a novel
156: functional renormalization group (RG) approach involving
157: both fermionic and bosonic fields~\cite{Schuetz05,Ledowski07}.
158: Our main result is that 
159: the regime of confined coherence proposed in Ref.~\cite{Clarke94} indeed exists, so that
160: strong interactions can give rise to a non-Fermi liquid
161: normal state in quasi-one-dimensional metals.  
162: 
163: 
164: 
165: {\it{Model.}} 
166: We start from an effective low energy model 
167: for spinless fermions with linearized energy dispersion and
168: density-density interactions. The  Euclidean action is
169:  \begin{eqnarray}
170:  S [ \bar{\psi} , \psi ] & = & \sum_{\alpha } \int_K \bigl[ - i \omega + \alpha v_F   \delta k_{\parallel}  + \mu_0 ( k_{\bot} ) \bigr]
171:  \bar{\psi}_{  K \alpha }  {\psi}_{  K \alpha } 
172:  \nonumber
173: \\
174:  & + & \frac{1}{2} \sum_{\alpha \alpha^{\prime}} \int_{\bar{K}} 
175: f_{  \alpha \alpha^{\prime} }
176:  \bar{\rho}_{ \bar{K} \alpha} \rho_{ \bar{K} \alpha^{\prime}}
177:  \; ,
178:  \label{eq:Sdef}
179:  \end{eqnarray} 
180: where
181: $ \mu_0 ( k_{\bot} ) = - \Sigma ( {\bd{k}}_F , i 0 )$ 
182: is a counter-term involving  the exact
183: self-energy at the true FS ${\bd{k}} = {\bd{k}}_F$ and
184: zero frequency, 
185: and $\delta k_{\parallel} = k_{\parallel} - \alpha k_F ( k_{\bot} )$.
186: Here $k_{\parallel}$ is the component of the two-dimensional
187: lattice momentum $\bd{k}$ parallel to the chain direction,
188: and
189: $k_{\bot}$ is the corresponding  perpendicular component.
190: The FS $\bd{k}_F$
191: can then be parameterized  as
192: $ k_{\parallel} = \alpha k_F ( k_{\bot})$, where
193: $\alpha = \pm 1$ labels the two disconnected sheets of the FS,
194: see Fig.~\ref{fig:FS}.
195: We neglect the $k_{\bot}$-dependence of the Fermi velocity $v_F$.
196: The chiral fields $ \psi_{ K  \alpha}$ are defined in terms of
197: the usual Fermi fields $\psi_{ k_{\parallel} , k_{\bot} , i \omega }$
198: via
199: $ \psi_{ K  \alpha} = \psi_{  \alpha k_{ F} ( k_{\bot} ) + \delta k_{\parallel} ,  
200: k_{\bot} , i \omega }$, and the
201: chiral densities $\rho_{ \bar{K}  \alpha}$
202: are $
203: \rho_{ \bar{K}  \alpha} = \int_K 
204:  \bar{\psi}_{K \alpha} \psi_{ K + \bar{K} \alpha}
205:  $.
206: We use collective labels
207: $K = ( \delta k_{\parallel} , k_{\bot} , i \omega )$ for fermionic    
208: and $\bar{K} = (  \bar{k}_{\parallel} , \bar{k}_{\bot} , i \bar{\omega} )$ for
209: bosonic fields, 
210: where $\omega$ and $\bar{\omega}$ are Matsubara frequencies.
211: The integration symbols are
212:  $
213:  \int_K =  \int_{k_{\bot}}
214:  \int_{ - \Lambda_\parallel}^{\Lambda_\parallel}  
215:  \frac{d \delta k_{\parallel}}{2 \pi}  \int  \frac{d \omega}{2 \pi}
216: $
217: and
218:  $
219:  \int_{\bar{K}} =   \int_{\bar{k}_{\bot}}
220:  \int_{ - \bar{\Lambda}_\parallel}^{\bar{\Lambda}_\parallel}  
221: \frac{d  \bar{k}_{\parallel}}{2 \pi}  \int  \frac{d \bar{\omega}}{2 \pi }
222: $ where for later convenience we have introduced the notation
223: $
224: \int_{ k_{\bot} }  =   \int_{ - \Lambda_{\bot} }^{  \Lambda_{\bot}}
225:  \frac{ d k_{\bot}}{ 2  \Lambda_\bot }$ and
226: $
227: \int_{ \bar{k}_{\bot} }  =  \int_{ - \bar{\Lambda}_{\bot}}^{ \bar{\Lambda}_{\bot} }
228:  \frac{ d \bar{k}_{\bot}}{  2 \bar{\Lambda}_{\bot} }
229: $.
230: Here $\Lambda_{\parallel}$ is a bandwidth cutoff,
231: $\Lambda_{\bot} = \pi / a_{\bot}$ is the width of the Brillouin zone
232: in transverse direction, and
233: $\bar{\Lambda}_{\bot}$ and $\bar{\Lambda}_\parallel$ 
234: restrict the momentum transfered by the interaction
235: in the directions parallel and perpendicular to the chains.
236: We assume that $ \bar{\Lambda}_{\parallel} \ll {\rm min} \{ k_F ( k_{\bot} )\}$,
237: so that the  interaction 
238: $f_{ \alpha \alpha^{\prime} }$ in Eq.~(\ref{eq:Sdef}) 
239: does not transfer momentum between the two disconnected sheets of the FS. 
240: However, the transverse 
241: momentum transfer cutoff $\bar{\Lambda}_{\bot}$ can be of the order of the 
242: transverse width  $\Lambda_{\bot} = \pi  / a_{\bot}$
243: of the Brillouin zone, so that
244: transverse Umklapp scattering is possible.
245: For simplicity we set $\bar{\Lambda}_{\bot} = \Lambda_{\bot}$ and call
246: $ \bar{\Lambda}_{\parallel} = \Lambda_0 $.
247: 
248: 
249: 
250: {\it{Self-consistent perturbation theory.}} 
251: To begin with, let us calculate the FS within second order 
252: self-consistent perturbation theory.  Using the procedure
253: outlined in Ref.~[\onlinecite{Neumayr03}], we obtain 
254: the following  integral equation for the
255: difference $\delta {k}_F ( k_{\bot} ) = k_{F} ( k_{\bot} ) - 
256: k_{F,0 } ( k_\bot )$ between the true Fermi momentum $k_F ( k_{\bot} )$ 
257: and the corresponding $k_{F,0} ( k_{\bot} )$ without interactions at the same density, 
258:  \begin{eqnarray}
259:  \frac{\delta {k}_F ( k_{\bot} )}{ \Lambda_0 } & = & 
260:     \left[ - {g}_4 + \frac{ {g}_4^2 + {g}_2^2}{2} \right] 
261:  \int_{ \bar{k}_{\bot} }  \tilde{\Delta} ( k_{\bot} , \bar{k}_{\bot} ) 
262: \nonumber
263:  \\
264:   &  & \hspace{-10mm} -  {g}_2^2 
265:  \int_{ \bar{k}_{\bot} }  \int_{ k_{\bot}^{\prime} }
266: J ( \tilde{\Delta} ( k_{\bot} , \bar{k}_{\bot} );  
267: \tilde{\Delta} ( k_{\bot}^{\prime} , \bar{k}_{\bot} ) )
268: \; ,
269:  \label{eq:IE}
270:  \end{eqnarray}
271: where
272:  $ \tilde{\Delta} ( k_{\bot} , \bar{k}_{\bot} ) = 
273:   [ k_F ( k_{\bot} ) - k_F ( k_{\bot} + \bar{k}_{\bot} )]/ {\Lambda}_{0}$, and
274:  \begin{equation}
275:  J ( \tilde{\Delta} ; \tilde{\Delta}^{\prime} ) =  
276:  \frac{ \tilde{\Delta} + \tilde{\Delta}^{\prime}}{4} \ln \left[
277:  \frac{ 4  - ( \tilde{\Delta} - \tilde{\Delta}^{\prime} )^2 }{ 
278:  ( \tilde{\Delta} + \tilde{\Delta}^{\prime} )^2} \right]
279:  \; .
280:  \label{eq:Ydef}
281: \end{equation}
282: The dimensionless couplings $g_2$ and $g_4$ are defined via
283: $2 \pi g_4 = \nu_0 f_{++} = \nu_0 f_{--}$ and
284: $2 \pi g_2 = \nu_0 f_{+-} = \nu_0 f_{-+}$, where 
285: the factor $\nu_0 = \Lambda_{\bot} ( \pi v_F )^{-1} = ( a_{\bot} v_F )^{-1}$ is introduced
286: for convenience.
287: We have solved Eq.~(\ref{eq:IE}) numerically, but for small $t_{\bot}$
288: we can also obtain an approximate analytic solution
289: using the fact that in this case
290: the dominant renormalization of the FS is due to
291: the logarithmic term in Eq.~(\ref{eq:Ydef}). 
292: Suppose that the bare FS is of the form
293:  $
294:  k_{F,0} ( k_{\bot} ) = \bar{k}_F + t_0 \cos ( k_{\bot} a_{\bot} )
295:  $
296: where $t_0 = 2 t_{\bot,0} / v_F \ll \Lambda_0$ 
297: and the average $\bar{k}_F$ is fixed by the total density. 
298: The renormalized FS is then given by
299:  $
300:  k_F ( k_{\bot} ) = \bar{k}_F + t  \cos ( k_{\bot} a_{\bot} ) + \ldots
301:  $, where $t$ is proportional to the renormalized nearest neighbor
302: interchain hopping,
303: and the ellipsis denotes higher harmonics corresponding to longer range hoppings.
304: From the  numerical solution of the integral equation
305: (\ref{eq:IE}) we find that  for $g_2 , g_4 \ll 1$ the higher harmonics
306: are indeed small. Then Eq.~(\ref{eq:IE}) can be reduced
307: to a transcendental equation for $t$, which to 
308: leading logarithmic order in $t/ \Lambda_0$ 
309: can be written as
310: $t / t_0 = [1 + R ( t  ) ]^{-1}$,
311: with
312:  \begin{equation}
313:  R ( t )  \approx   \frac{g_4}{2} -   \frac{{g}_4^2}{4}
314: % \nonumber
315: % \\
316: % &  &  \hspace{-10mm}   
317:     +  \frac{{g}_2^2}{2} 
318:  \ln (  {\Lambda}_0   / | t|  )   
319: \; .
320:  \label{eq:Rdef}
321: \end{equation}  
322: A similar relation has been obtained
323: previously \cite{Fabrizio93,Ledowski07} for the
324: difference between the Fermi momenta associated with the bonding and the
325: anti-bonding band in two coupled spinless chains. 
326: Note that to first order in the bare interaction
327: $ R ( t ) \propto {g}_4$, so that a repulsive interaction ${g}_4 >0$
328: reduces
329: the warping of the FS, while for ${g}_4 < 0$ the warping
330: of the FS is enhanced. However, for  $t_{\bot} \rightarrow 0$ the
331: logarithmic term proportional to  ${g}_2^2$ always dominates and
332: predicts an interaction-induced  reduction of the FS warping,
333: irrespective of  the sign of the interaction.
334: 
335: 
336: {\it{Functional RG approach.}}
337: % Because, the confinement transition is a strong-coupling phenomenon
338: % the usual weak coupling truncation of the formally exact hierarchy
339: % of flow equations for the irreducible vertices~\cite{Kopietz01} is not sufficient.
340: % However, sensible strong-coupling extrapolations can be obtained
341: % by decoupling certain interaction channels by means of 
342: % Hubbard-Stratonovich fields and considering then the FRG flow of the
343: % fixed Fermi-Bose theory~\cite{Schuetz05}.
344: % In Ref.~\cite{Ledowski07}
345: % we have successfully used this strategy to 
346: % to calculate  the renormalized FS of two metallic chains
347: % in the regime where the relevant dimensionless coupling
348: % constant is of the order of unity.
349: We now generalize the RG approach developed in Ref.~[\onlinecite{Ledowski07}]
350: in the context of a simplified two-chain model  
351: to the more interesting two-dimensional case considered here.
352: The method has been  described in detail previously~\cite{Ledowski07}, 
353: so that we will be rather brief here.
354: In the momentum transfer cutoff scheme~\cite{Schuetz05} 
355: we decouple the density-density interaction by means of a
356: bosonic Hubbard-Stratonovich  transformation
357: and then use the maximal momentum carried by the  boson field
358: as flow parameter $\Lambda$ of the RG.
359: Our initial cutoff is thus
360: $\Lambda = \Lambda_0 = \bar{\Lambda}_{\parallel}$.
361: Eliminating boson fields with momenta in the range
362: $\Lambda < | \bar{k}_{\parallel} | < \Lambda_0$ 
363: we obtain a  new effective action, whose vertices
364: are determined by
365: a formally exact hierarchy of  functional RG flow equations.
366: To calculate the true FS, we need the flow equation
367: for the relevant part $r_l ( k_{\bot} )$
368: of the irreducible fermionic self-energy $\Sigma_{\Lambda} ( K , \alpha )$,
369: which is defined via
370: $r_l ( k_{\bot} ) = Z_l ( k_{\bot} )  [
371: \Sigma_{\Lambda} ( {\bd{k}}_F , i0, \alpha ) 
372: + \mu_0 ( k_{\bot} ) ]/ v_F \Lambda $.  
373: Here $l = \ln ( \Lambda_0 / \Lambda )$ and
374:  $Z_l ( k_{\bot})$ is the flowing wave-function
375: renormalization. The functional RG flow equation for
376: $r_l ( k_{\bot} )$ is of the form
377: \begin{equation}
378:  \partial_l r_l ( k_{\bot} ) = [ 1 - \eta_l ( k_{\bot} ) ] r_l ( k_{\bot} )
379:  + \dot{\Gamma}_l ( k_{\bot} )
380:  \; ,
381:  \label{eq:flowr}
382:  \end{equation}
383: where $\eta_l ( k_{\bot} ) = - \partial_l \ln Z_l ( k_{\bot} )$
384: is the flowing anomalous dimension. 
385: An approximate expression
386: for the inhomogeneity $\dot{\Gamma}_l ( k_{\bot} )$ 
387: is given in Eq.~(\ref{eq:dotGamma}) below.
388: As long as $\eta_{\infty} ( k_{\bot} ) < 1$, the FS is well defined.
389: The shift $\delta k_F ( k_{\bot} )$ of the FS due to interactions can then be
390: obtained from the requirement that the initial value $r_0 ( k_{\bot} )$
391: should be fine tuned so that the relevant coupling
392: $r_{l} ( k_{\bot} )$ flows into a fixed point \cite{Kopietz01}.
393: This leads to the following exact integral equation for 
394: the FS,
395:  \begin{equation}
396:   \frac{ \delta k_F ( k_{\bot} ) }{\Lambda_0}
397:  = r_0 ( k_{\bot} ) =
398:  - \int_0^{\infty} d l e^{ -l + \int_0^l d t \eta_t ( k_{\bot} ) }
399:  \dot{\Gamma}_l ( k_{\bot} )
400:  \; .
401:  \label{eq:r0flow}
402:  \end{equation} 
403: Using the same truncation strategy as
404: in Ref.~[\onlinecite{Ledowski07}],
405: we approximate
406:  \begin{eqnarray}
407:   \dot{\Gamma}_l ( k_{\bot} )
408:  & = &
409: -   \int_{ \bar{k}_{\bot}} \int \frac{ d \bar{q} d \bar{\epsilon}}{(2 \pi )^2 }
410: % \nonumber
411: % \\
412: % &  & \hspace{-25mm} \times 
413: \frac{  \delta ( | \bar{q} | -1 )   [ {\bf{F}}_l ( \bar{q} , i \bar{\epsilon} , 
414:  \bar{k}_{\bot}) ]_{\alpha \alpha}  e^{ i \bar{\epsilon} 0} 
415:  }{ i  \bar{\epsilon} 
416:  - \alpha  \bar{q} - \tilde{\Delta}_l ( k_{\bot} , \bar{k}_{\bot}) }
417:  \nonumber
418: \\
419:  &   & \times  
420: \gamma_l (  k_{\bot} , \bar{k}_{\bot} )
421: \gamma_l ( k_{\bot} + \bar{k}_{\bot}, - \bar{k}_{\bot} )
422:  \; ,
423:  \label{eq:dotGamma}
424:  \end{eqnarray}
425: where
426:  $
427:  \tilde{\Delta}_l ( k_{\bot} , \bar{k}_{\bot})  =  
428: \tilde{k}_{F,l} ( k_{\bot} ) - \tilde{k}_{F,l} ( k_{\bot} + \bar{k}_{\bot} )$
429: with $\tilde{k}_{F,l} ( k_{\bot} ) = k_F ( k_{\bot} ) / \Lambda - r_l ( k_{\bot} )$.
430: The inverse of the $2 \times 2$-matrix 
431: ${\bf{F}}_l ( \bar{q} , i \bar{\epsilon} , \bar{k}_{\bot} )$
432: is defined via
433: %(we write $\bar{Q} = ( \bar{q} , i \bar{\epsilon} )$),
434:  \begin{equation}
435:  [ {\bf{F}}_l ( \bar{q} , i \bar{\epsilon},
436: \bar{k}_{\bot} )]^{-1}_{\alpha \alpha^{\prime} }
437:  = [ \nu_0 {\bf{f}} ]^{-1}_{ \alpha \alpha^{\prime}} +
438:  \delta_{ \alpha \alpha^{\prime}} \tilde{\Pi}_l ( \bar{q} , i \bar{\epsilon},
439: \bar{k}_{\bot} , \alpha)
440:  \; ,
441:  \end{equation}
442: where $\bf{f}$ is a matrix in chirality space with
443: elements 
444: $[ {\bf{f}} ]_{\alpha \alpha^{\prime} } = f_{\alpha \alpha^{\prime} }$, and
445: $\tilde{\Pi}_l ( \bar{q} , i
446:  \bar{\epsilon} , \bar{k}_{\bot} , \alpha)$ is the rescaled polarization associated with
447: fermions of chirality $\alpha$, for which we use the
448: adiabatic approximation~\cite{Ledowski07}
449:  \begin{eqnarray}
450: \tilde{\Pi}_l ( \bar{q} , i \bar{\epsilon}, \bar{k}_{\bot}, \alpha ) & = & \frac{1}{2 \pi } 
451:  \int_{ k_{\bot}} 
452:  \frac{ \tilde{\Delta}_l ( k_{\bot} , \bar{k}_{\bot} ) + \alpha \bar{q} }{ \tilde{\Delta}_l ( k_{\bot} , 
453:  \bar{k}_{\bot} )   + \alpha \bar{q} - i \bar{\epsilon}    }
454: \nonumber
455:  \\\
456:  &   \times & \gamma_l (
457:  k_{\bot}  , \bar{k}_{\bot} )   \gamma_l (  
458:  k_{\bot} + \bar{k}_{\bot}  , - \bar{k}_{\bot} )
459:  \; .
460:  \label{eq:poladiabat}
461: \end{eqnarray}
462: The  anomalous dimension is in this approximation
463:  \begin{eqnarray}
464:  \eta_l ( k_{\bot} ) & = &
465: -   \int_{ \bar{k}_{\bot}} \int \frac{ d \bar{q} d \bar{\epsilon}}{(2 \pi )^2 }
466: \frac{  \delta ( | \bar{q} | -1 )  [ {\bf{F}}_l ( \bar{q} , i \bar{\epsilon} , 
467:  \bar{k}_{\bot}) ]_{\alpha \alpha} }{ [ i  \bar{\epsilon}
468:  - \alpha  \bar{q}  - \tilde{\Delta}_l ( k_{\bot} , \bar{k}_{\bot})]^2 }
469:  \nonumber
470:  \\
471:  &  &  \times 
472: \gamma_l (  k_{\bot} , \bar{k}_{\bot} )
473: \gamma_l ( k_{\bot} + \bar{k}_{\bot}, - \bar{k}_{\bot} )
474:  \; .
475:  \label{eq:etaflow}
476:  \end{eqnarray}
477: Finally, the dimensionless vertex 
478: $\gamma_l (  k_{\bot} , \bar{k}_{\bot} )$ 
479: with one bosonic and two fermionic external legs
480: (where $k_{\bot}$ labels the incoming fermion and 
481: $\bar{k}_{\bot}$ labels the boson) satisfies
482: the flow equation 
483:  \begin{eqnarray}
484:  \partial_l  \gamma_l ( k_{\bot} , \bar{k}_{\bot} )
485:  & = & - \frac{1}{2} [  \eta_l ( k_{\bot} ) + \eta_l ( k_{\bot} + \bar{k}_{\bot}) ]
486: \gamma_l (  k_{\bot} , \bar{k}_{\bot} )
487:  \nonumber
488:  \\
489:  &  & \hspace{-25mm} -
490: \int_{ \bar{k}_{\bot}^{\prime} } \int \frac{ d \bar{q} d \bar{\epsilon}}{(2 \pi )^2 } 
491: \delta ( | \bar{q} | -1 )  [ {\bf{F}}_l ( \bar{q} , i \bar{\epsilon} , 
492:  \bar{k}_{\bot}^{\prime} ) ]_{\alpha \alpha}
493:  \nonumber
494:  \\
495:  & &  \hspace{-23mm} \times
496: \frac{ \gamma_l (  k_{\bot} + \bar{k}_{\bot }^{\prime} ,   \bar{k}_{\bot} )
497: \gamma_l (  k_{\bot} , \bar{k}_{\bot}^{\prime} )  \gamma_l (  k_{\bot} +  \bar{k}_{\bot } + \bar{k}_{\bot}^{\prime} ,   - \bar{k}_{\bot}^{\prime} )  }{
498:  [ i  \bar{\epsilon}
499:  - \alpha  \bar{q}  - \tilde{\Delta}_l ( k_{\bot} , \bar{k}_{\bot}^{\prime} )]
500:  [ i  \bar{\epsilon}
501:  - \alpha  \bar{q}  - \tilde{\Delta}_l ( k_{\bot} + \bar{k}_{\bot} , \bar{k}_{\bot}^{\prime} )]
502:  } 
503:  \; ,
504:  \nonumber
505:  \\
506:  & &
507: \label{eq:gammaflow}
508:  \end{eqnarray}
509: with initial condition $\gamma_{l=0}(  k_{\bot} , \bar{k}_{\bot} )  =1$.
510: A graphical representation of Eq.~(\ref{eq:gammaflow}) is
511: shown in Fig.~\ref{fig:threevertex}.
512: %
513: %
514: \begin{figure}[tb]
515: % \begin{center}
516:   \centering
517: %  \psfrag{t}{$\theta$}
518:   \epsfig{file=fig2.eps,width=55mm}
519: %   \end{center}
520:   \vspace{-4mm}
521:   \caption{%
522: Diagrammatic representation of the flow equation (\ref{eq:gammaflow})
523: for the three-legged vertex with one bosonic (wavy line) and
524: two fermionic (solid lines with arrows) external legs.
525: The thick wavy line with a slash denotes the bosonic single scale propagator.
526: Additional contributions involving irrelevant higher order vertices
527: are omitted, see Refs.~[\onlinecite{Schuetz05,Ledowski07}].
528: }
529: \label{fig:threevertex}
530: \end{figure}
531: %
532: %
533: 
534: {\it{Results.}}
535: Eqs.~(\ref{eq:flowr})-(\ref{eq:gammaflow}) form a closed system of flow 
536: equations for the rescaled self-energy at the FS 
537: $r_l ( k_{\bot})$, the flowing
538: anomalous dimension $\eta_l ( k_{\bot} )$, and the three-legged vertex
539: $\gamma_l ( k_{\bot} , \bar{k}_{\bot} )$.
540: Of course, these equations can only be solved numerically, but the
541: qualitative behavior of the solutions
542: can also be extracted analytically.
543: To begin with, let us establish the 
544: relation with  the perturbative Eq.~(\ref{eq:IE}).
545: We set $g_4=0$ from now on, because
546: the dominant renormalization of the FS is due to the
547: $g_2$-process.
548: In the simplest approximation, we set
549: $\gamma_l ( k_{\bot} , \bar{k}_{\bot} ) \approx 1$, 
550: $\eta_l ( k_{\bot} ) \approx 0$
551: and replace the flowing FS 
552: $\tilde{k}_{F,l} ( k_{\bot} ) = k_F ( k_{\bot}) / \Lambda - 
553: r_l ( k_{\bot} )$ by $k_F ( k_{\bot} ) / \Lambda $.
554: Then we obtain from Eqs.~(\ref{eq:r0flow}) and (\ref{eq:dotGamma})
555: to leading logarithmic order
556: \begin{equation}
557:  \frac{\delta k_F  ( k_{\bot} )}{\Lambda_0} 
558:  = [ ( 1 - g_2^2)^{-\frac{1}{2}} -1 ]  
559: %\frac{ 1 - \sqrt{ 1 - g_2^2}}{\sqrt{ 1 - g_2^2}}
560:  \int_{\bar{k}_{\bot}} 
561:  \tilde{\Delta} ( k_{\bot} , \bar{k}_{\bot} )
562:  \ln  |  \tilde{\Delta} ( k_{\bot} , \bar{k}_{\bot} ) |
563:  \; .
564:  \end{equation}
565: Expanding in harmonics we obtain as before
566: $t / t_0 = [ 1 + R ( t )]^{-1}$, but now with
567:  \begin{eqnarray}
568:  R ( t ) & \approx &    
569:  [ ( 1 - g_2^2)^{-\frac{1}{2}} -1 ]  
570: %\frac{ 1 - \sqrt{ 1 - g_2^2}}{ \sqrt{ 1 - g_2^2}}
571:   \ln ( {\Lambda_0  }/{ | t|  } )
572: \; ,
573:  \label{eq:R2}
574: \end{eqnarray}
575: which reduces to Eq.~(\ref{eq:Rdef}) to leading order in $g_2$.  
576: Obviously, $R ( t ) \rightarrow \infty$ for $g_2 \rightarrow 1$,
577: indicating a confinement transition at strong coupling, where the renormalized
578: interchain hopping $t$ vanishes.
579: However, from our previous work
580: on two coupled chains \cite{Ledowski07}
581: we know that vertex corrections and
582: wave-function renormalizations
583: can possibly change this scenario.
584: 
585: To investigate this, let us first consider the
586: RG flow of the vertex $\gamma_l ( k_{\bot} , \bar{k}_{\bot} )$ numerically.
587: Representative results are shown in Fig.~\ref{fig:flowvertex}.
588: % %
589: % \begin{minipage}{27.0mm}
590: % %  \epsfig{file=3.2.eps,width=28mm}
591: %   \epsfig{file=Ev0.4T3.2.eps,width=27.0mm}
592: %  \end{minipage}
593: %  \begin{minipage}{24.4mm}
594: % % \epsfig{file=6.4.eps,width=28mm}
595: %   \epsfig{file=Ev0.4T6.8.eps,width=24.4mm}
596: %  \end{minipage}
597: %  \begin{minipage}{24.4mm}
598: % % \epsfig{file=20.eps,width=28mm}
599: %  \epsfig{file=Ev0.4T20.eps,width=24.4mm}
600: %  \end{minipage}
601: % %   \end{center}
602: % \hspace{1mm}
603: % \begin{minipage}{4.5mm}
604: % % \epsfig{file=20.eps,width=28mm}
605: %  \epsfig{file=Legend.eps,width=4.5mm}
606: %  \end{minipage}
607: %
608: \begin{figure}[tb]
609:  \vspace{5mm}
610: % \begin{center}
611:   \centering
612: \epsfig{file=fig3.eps,width=8.4cm}
613: %\epsfig{file=fig3.eps,width=8.4cm}
614: % \begin{minipage}{27.0mm}
615: %  \epsfig{file=fig3a.eps,width=27.0mm}
616: %  % \epsfig{file=Ev0.4T3.2.eps,width=27.7mm}
617: %  \end{minipage}
618: %  \begin{minipage}{24.4mm}
619: %  \epsfig{file=fig3b.eps,width=24.4mm}
620: % %  \epsfig{file=Ev0.4T6.8.eps,width=24.4mm}
621: %  \end{minipage}
622: %  \begin{minipage}{24.4mm}
623: %  \epsfig{file=fig3c.eps,width=24.4mm}
624: % % \epsfig{file=Ev0.4T20.eps,width=24.4mm}
625: %  \end{minipage}
626: % %   \end{center}
627: % \hspace{1mm}
628: % \begin{minipage}{4.5mm}
629: %  \epsfig{file=fig3d.eps,width=4.5mm}
630: % % \epsfig{file=Legend.eps,width=3.5mm}
631: %  \end{minipage}
632: 
633: 
634:   \caption{%
635: (Color online) Typical evolution of the vertex $\gamma_l ( k_{\bot} , \bar{k}_{\bot} )$
636: for different values of the flow parameter $l$. 
637: To evaluate the flow we have expanded in Eqs.~(\ref{eq:dotGamma}) and (\ref{eq:gammaflow}) up to $g_2^2$.
638: We have assumed a
639: harmonic bare FS with amplitude $t_0 / \Lambda_0 = 10^{-3}$
640: and bare coupling $g_2 =0.4$ 
641: From left to right 
642: $l= \frac{1}{2}l_c, l_c,3l_c$ and $\gamma_{\min}= 0.999989,0.981,0.85$,
643: where  the crossover scale $l_c$ (see text) can be approximated by
644: $l_c \approx - \ln (2 t_0 / \Lambda_0 )$ for small $g_2$.
645: }
646:   \label{fig:flowvertex}
647: \end{figure}
648: %
649: %
650: Obviously, 
651: the dependence of $\gamma_l ( k_{\bot} , \bar{k}_{\bot} )$
652:  on the fermionic momentum
653: $k_{\bot}$, which develops at intermediate scales $l$, is smoothed out again at
654: the scale $l_c$ where the
655: reduced cutoff $\Lambda_c = \Lambda_0 e^{-l_c} $ becomes comparable 
656: with the warping of the renormalized FS.
657: Defining the flowing dimensionless 
658: nearest neighbor interchain hopping $\tilde{t}_l$ via
659: $\tilde{k}_{F,l} = \bar{k}_F / \Lambda + \tilde{t}_l
660:  \cos ( k_{\bot} a_{\bot} ) + \ldots$, 
661: the crossover scale $l_c$ can be defined 
662: self-consistently via
663: $2 \tilde{t}_{l=l_c} = 1$.
664: %The corresponding RG flow of the
665: %rescaled first harmonic $\tilde{t}_l$ of the
666: %FS, defined via 
667: %$\tilde{k}_{F,l} = \bar{k}_F / \Lambda + \tilde{t}_l
668: % \cos ( k_{\bot} a_{\bot} ) + \ldots$,
669: %is shown in Fig.~\ref{fig:flowA}.
670: %
671: %
672: % \begin{figure}[tb]
673: % % \begin{center}
674: %   \centering
675: % %  \psfrag{t}{$\theta$}
676: %   \epsfig{file=FlowAmplitude0.001.eps,width=50mm}
677: % %   \end{center}
678: %   \vspace{-4mm}
679: %   \caption{%
680: % (Color online) RG flow of the dimensionless interchain hopping
681: %  $\tilde{t}_l$
682: % for the same parameters as in Fig.~\ref{fig:flowvertex}. 
683: % }
684: %   \label{fig:flowA}
685: % \end{figure}
686: % %
687: %
688: %Clearly, the FS flow is dominated by the regime
689: %$0 < l < l_c$.
690: Because of the weak  dependence
691: of  $\gamma_l ( k_{\bot} , \bar{k}_{\bot} )$
692: on  the fermionic momentum ${k}_{\bot}$, we may approximate
693: $\gamma_l ( k_{\bot} , \bar{k}_{\bot} ) \approx
694: \gamma_l ( 0 , \bar{k}_{\bot} ) \equiv \gamma_l (\bar{k}_{\bot} )  $.
695: Moreover, using the fact that close to the confinement transition $|\tilde{t}_l| \ll 1$,  
696: we obtain from
697: Eqs.~(\ref{eq:flowr})-(\ref{eq:gammaflow})
698: to leading order in $\tilde{t}_l$ 
699: the following RG flow equation for the effective interaction
700:  $g_l ( \bar{k}_{\bot}  ) = g_2 \gamma_l^2 ( \bar{k}_{\bot} )$,
701: \begin{equation}
702:  \partial_l g_l ( \bar{k}_{\bot} ) = -
703:  \frac{   4 \sin^2 ( \bar{k}_{\bot} a_{\bot}/2  )
704:    g_l ( \bar{k}_{\bot} ) u_l^2  \tilde{t}_l^2   }{
705:  \sqrt{ 1 - u_l^2} [ 1 + \sqrt{ 1 - u_l^2} ]^3}
706:  \; ,
707:  \label{eq:uflowfinal}
708:  \end{equation}
709:  with $u_l = g_l ( \pi / a_{\bot} )$.
710: Note that the flow of $g_l ( \bar{k}_{\bot}  )$ is driven by the
711: component  $ g_l ( \pi / a_{\bot} )$  of the interaction involving 
712: momentum transfer $\bar{k}_{\bot} = \pi / a_{\bot}$; 
713: in a simplified two-chain model 
714: this corresponds to the pair-tunneling process~\cite{Fabrizio93}.
715: The flow of the rescaled interchain hopping
716:  $\tilde{t}_l$ is determined by
717:  $\partial_l \tilde{t}_l =  ( 1 - \bar{\eta}_l ) \tilde{t}_l + O ( \tilde{t}_l^2 )$,
718: %  \begin{equation}
719: %  \partial_l \tilde{t}_l =  
720: % %=  ( 1 - \eta^A_l ) \tilde{A}_l + O ( \tilde{A}_l^2 )
721: % \left[ 1- 2 
722: %  \int_{ \bar{k}_{\bot}}   
723: %  \sin^2 (  \bar{k}_{\bot}  a_{\bot} /2 ) 
724: %  \frac{ 1 - \sqrt{ 1 - g_l^2 ( \bar{k}_{\bot} ) }}{
725: %   \sqrt{ 1 - g_l^2 ( \bar{k}_{\bot} ) } } \right] 
726: %  \tilde{t}_l
727: %  \; ,
728: %  \label{eq:Aflow2}
729: %  \end{equation} 
730: where  $\bar{\eta}_l$ is
731: the weighted FS average 
732:  $\bar{\eta}_l =
733:     2 
734:    \int_{ {k}_{\bot}}   
735:   \sin^2 (  {k}_{\bot}  a_{\bot} /2 )
736:  \eta_l( {k}_{\bot} )$ 
737: of the flowing anomalous dimension, which for small
738: $\tilde{t}_l$ can be approximated by
739: $ \eta_l( k_{\bot} ) =
740:   [  1 - g_l^2 ( k_{\bot} )  ]^{- \frac{1}{2}}  -1$.
741: % \; .
742: % \label{eq:etadef}
743: % \end{equation} 
744: %\begin{equation}
745: % \bar{\eta}_l =
746: %    2 
747: %   \int_{ \bar{k}_{\bot}}   
748: %  \sin^2 (  \bar{k}_{\bot}  a_{\bot} /2 )
749: % \eta_l( \bar{k}_{\bot} ) \; , \; 
750: %  \eta_l( \bar{k}_{\bot} ) =
751: %  \frac{ 1 - \sqrt{ 1 - g_l^2 ( \bar{k}_{\bot} ) }}{
752: %   \sqrt{ 1 - g_l^2 ( \bar{k}_{\bot} ) } }
753: %  \; ,
754: %  \label{eq:etat}
755: %  \end{equation}
756: From the numerical solution of these equations we find that,
757: for a given  $t_{\bot , 0} \ll v_F \Lambda_0$, there exists  a critical value of the
758: bare interaction $g_{2}$ where the renormalized $\tilde{t} = \tilde{t}_{\infty}$ 
759: vanishes, corresponding to a confinement 
760: transition to a state with vanishing interchain hopping $t_{\bot}$ and flat FS. 
761: In Fig.~\ref{fig:flowAu} we show the ratio $t_{\bot} / t_{\bot ,0}$
762: together with the renormalized interaction $u = g_{\infty} ( \pi / a_{\bot} )$ 
763: as a function of
764: the bare interaction for small $\tilde{t}_0 = 2 t_{\bot ,0} / (v_F \Lambda_0)$. 
765: %
766: %
767: \begin{figure}[tb]
768: % \begin{center}
769:   \centering
770: %  \psfrag{t}{$\theta$}
771: \epsfig{file=fig4.eps,width=45mm}
772: % \epsfig{file=Numerics1.eps,width=45mm}
773: %   \end{center}
774:   \vspace{-4mm}
775:   \caption{%
776: (Color online) 
777: Renormalized nearest neighbor interchain hopping 
778: %$\tilde{t} = \tilde{t}_{\infty}$ 
779: $t_{\bot}$ and renormalized  
780: interaction $u = g_{\infty} ( \pi / a_{\bot} )$
781: as a function of the bare
782: interaction $g_2$ for  $\tilde{t}_0 = 2 t_{\bot,0} / (v_F \Lambda_0 ) = 0.1$ 
783: as obtained from the numerical solution
784: of Eqs.~(\ref{eq:dotGamma}) and (\ref{eq:gammaflow}).
785: For the perturbative curves we have expanded in these equations up 
786: to order $g_2^2$.}
787:   \label{fig:flowAu}
788: \end{figure}
789: %
790: %
791: A projection of the RG flow in the $\tilde{t}$-$u$ plane  
792: %numerical solutions of the full equations 
793: %(\ref{eq:flowr})-(\ref{eq:gammaflow})
794: is shown in Fig.~\ref{fig:flowdiagram}.
795: %
796: %
797: \begin{figure}[tb]
798: % \begin{center}
799:   \centering
800: %  \psfrag{t}{$\theta$}
801: \epsfig{file=fig5.eps,width=45mm}
802: %  \epsfig{file=FlowV6.eps,width=55mm}
803: %   \end{center}
804:   \vspace{-4mm}
805:   \caption{%
806: (Color online)
807: Projection of the RG flow  in the  $\tilde{t}$-$u$ plane.
808: The trajectories are obtained from the numerical solution of  Eq.(\ref{eq:uflowfinal}) and
809: $\tilde{t}_l = ( 1 - \bar{\eta}_l ) \tilde{t}_l$, where $\bar{\eta}_l$ is the
810: weighted FS average of the flowing anomalous dimension 
811: $\eta_l ( k_{\bot} )$  defined in the text.
812: %Eq.~(\ref{eq:etadef}).
813: }
814:   \label{fig:flowdiagram}
815: \end{figure}
816: %
817: %
818: At the confinement transition the system is certainly not a
819: Fermi liquid, because the anomalous dimension
820: $\eta_l ( k_{\bot} )$  is larger than unity and a
821: sharp FS cannot be defined~\cite{Kopietz01}.
822: The physical properties of the model in the confined phase
823: remain to be explored. 
824: 
825: 
826: 
827: 
828: {\it{Summary and conclusions.}}
829: We have
830: shown that sufficiently strong interactions 
831: involving momentum transfers parallel to the chains can lead to a 
832: confinement transition in highly anisotropic
833: quasi-one-dimensional metals.
834: At the confinement transition the curvature
835: of the two disconnected sheets of the FS vanishes, so that
836: a coherent motion of the electrons perpendicular to the chains
837: is not possible and the electronic motion is one-dimensional.
838: Our calculation thus  supports the existence of a confined state
839: in quasi-one-dimensional metals, as originally suggested  by
840: Clarke, Strong, and Anderson \cite{Clarke94}.
841: In the confined state the system is a non-Fermi liquid with large anomalous dimension.
842: From the numerical solution of our  functional RG equations we found no
843: evidence for a truncation transition~\cite{Furukawa98,Ferraz03}, where
844: only certain sectors of the FS are washed out by interactions.
845: 
846: 
847: Because in this work we have  considered only spinless fermions 
848: and interactions
849: which do not transfer momentum between the two disconnected sheets of the FS,
850: our model (\ref{eq:Sdef}) is too simple  to
851: describe the competition between confinement and the
852: tendency to develop  some kind of 
853: long-range order, such as  charge-density or  spin-density waves.
854: In principle, spin fluctuations can be
855: taken into account with the help of another Hubbard-Stratonovich field,
856: but the analysis of the resulting  RG equations for the coupled 
857: boson-fermion model requires a substantial extension of our
858: calculation.
859: 
860: 
861: 
862: 
863: 
864: %while
865: %other sectors remain well defined with finite curvature.
866: %
867: % At this point we cannot exclude the possibility
868: % that in a certain parameter regime 
869: % a more thorough numerical analysis of
870: % our non-perturbative FRG
871: % equations (\ref{eq:flowr})-(\ref{eq:gammaflow})
872: % would predict a truncated FS, where only certain 
873: % parts of the FS remain well-defined \cite{Furukawa98,Ferraz03}.
874: %Our method can also be used to calculate the renormalization
875: %of an almost square shaped FS, which has been observed
876: %in the underdoped cuprates~\cite{Abrecht03}.
877: 
878: We thank A. Ferraz and F. H. L. Essler for useful discussions.
879: 
880: \vspace{-7mm}
881: 
882: \begin{thebibliography}{99}
883: %
884: \bibitem{Pomeranchuk58}
885: I. J. Pomeranchuk, Zh. Eksp. Teor. Fiz. {\bf{35}}, 524 (1958)
886: [Sov. Phys. JETP {\bf{8}},  361 (1958)].
887: %
888: \bibitem{Lifshitz60}
889: I. M. Lifshitz, Zh. Eksp. Teor. Fiz. {\bf{38}}, 1569 (1960)
890: [Sov. Phys. JETP {\bf{11}}, 1130 (1960)].
891: % 
892: \bibitem{Quintanilla06}
893:  J. Quintanilla and A. J. Schofield, Phys. Rev. B {\bf{74}}, 115126 (2006).
894: %
895: \bibitem{Furukawa98}
896: N. Furukawa, T. M. Rice, and M. Salmhofer, Phys. Rev. Lett.
897: {\bf{81}}, 3195 (1998);
898: C. Honerkamp, M. Salmhofer, N. Furukawa, and T. M. Rice,
899: Phys. Rev. B {\bf{63}}, 035109 (2001).
900: %
901: \bibitem{Ferraz03}
902: A. Ferraz, Phys. Rev. B {\bf{68}}, 075115 (2003); H. Freire, E. Correa, and
903: A. Ferraz, {\it{ibid.}} {\bf{71}}, 165113 (2005).
904: %
905: \bibitem{Clarke94}
906: D. G.  Clarke, S. P. Strong, and P. W. Anderson,
907: Phys. Rev. Lett. {\bf{72}}, 3218 (1994); 
908: {\it{ibid.}} {\bf{74}}, 4499 (1995); D. G. Clarke and S. P.
909: Strong, Adv. Phys. {\bf{46}}, 454 (1997).
910: %
911: \bibitem{Wen90}
912: X. G. Wen, Phys. Rev. B {\bf{42}}, 6623 (1990). 
913: %
914: \bibitem{Kopietz94}
915: P. Kopietz, V. Meden, and K. Sch\"{o}nhammer, Phys.  Rev. Lett.
916:  {\bf{74}}, 2997 (1995); Phys. Rev. B {\bf{56}}, 7232 (1997).
917: %
918: \bibitem{Boies95}
919: D. Boies, C. Bourbonnais and A.-M. S. Tremblay, Phys. Rev. Lett. {\bf{74}}, 968 (1995). 
920: %
921: \bibitem{Arrigoni98}
922: E. Arrigoni, Phys. Rev. Lett. {\bf{80}}, 790 (1998);
923: {\it{ibid.}} {\bf{83}}, 128 (1999);
924: Phys. Rev. B {\bf{61}}, 7909 (2000).
925: %
926: \bibitem{Schuetz05}
927: F. Sch\"{u}tz, L. Bartosch, and P. Kopietz,
928: Phys. Rev. B {\bf{72}}, 035107 (2005).
929: %
930: \bibitem{Ledowski07}
931: S. Ledowski and P. Kopietz, Phys. Rev. B {\bf{75}}, 045134 (2007).
932: %
933: \bibitem{Neumayr03}
934: A. Neumayr and W. Metzner, Phys. Rev. B {\bf{67}}, 035112 (2003).
935: %
936: % \bibitem{Pomenanchuk58}
937: % I. J. Pomeranchuk, Zh. Eksp. Teor. Fiz. {\bf{35}}, 524 (1958)
938: % [Sov. Phys. JETP {\bf{8}},  361 (1958)].
939: %  
940: %   
941: % \bibitem{Furukawa98}
942: % N. Furukawa, T. M. Rice, and M. Salmhofer, Phys. Rev. Lett.
943: % {\bf{81}}, 3195 (1998);
944: % C. Honerkamp, M. Salmhofer, N. Furukawa, and T. M. Rice,
945: % Phys. Rev. B {\bf{63}}, 035109 (2001).
946: %  %
947: %\bibitem{Dusuel03}
948: %S. Dusuel and B. Dou\c{c}ot, Phys. Rev. B {\bf{67}}, 205111 (2003).
949: %
950: \bibitem{Fabrizio93}
951: M. Fabrizio, Phys. Rev. B {\bf{48}}, 15838 (1993);
952: S. Ledowski, P. Kopietz, and A. Ferraz, {\it{ibid.}} {\bf{71}}, 235106 (2005).
953: %
954: \bibitem{Kopietz01}
955: P. Kopietz and T. Busche, Phys. Rev. B {\bf{64}}, 155101 (2001);
956: %
957: %\bibitem{Ledowski03}
958: S. Ledowski and P. Kopietz, J. Phys.: Condens. Matter {\bf{15}}, 4779 (2003).
959: %
960: %
961: %\bibitem{Nozieres64}
962: %P. Nozi\`{e}res, {\it{Theory of Interacting Fermi Systems}}, (Benjamin, New York, 1964).
963: %.
964: %
965: %\bibitem{Kopietz01}
966: %P. Kopietz and T. Busche, Phys. Rev. B {\bf{64}}, 155101 (2001);
967: %M. Salmhofer and C. Honerkamp, Prog. Theor. Physics {\bf{105}}, 1 (2001).
968: %
969: %\bibitem{Abrecht03}
970: % M. Abrecht, D. Ariosa, D. Cloetta, 
971: %S. Mitrovic, M. Onellion, X. X. Xi, G. Margaritondo, and D. Pavuna, 
972: %Phys. Rev. Lett. {\bf{91}}, 057002 (2003). 
973: %
974: %\bibitem{Roldan06}
975: %R. Rold\'{a}n, M. P. L\'{o}pez-Sancho, F. Guinea, and
976: %S.-W. Tsai, cond-mat/0612368.
977: %
978: 
979: \end{thebibliography}
980: 
981: 
982: \end{document}