cond-mat0204585/ett.tex
1: % INJ
2: 
3: \documentclass[aps,prb,superscriptaddress,showpacs,twocolumn]{revtex4}
4: %\documentclass[aps,prb,superscriptaddress,showpacs,preprint]{revtex4}
5: \usepackage{graphicx}
6: \usepackage{amssymb}
7: \usepackage{amsmath}
8: 
9: \newcommand{\bk}{{\bf k}}
10: \newcommand{\bq}{{\bf q}}
11: \newcommand{\bp}{{\bf p}}
12: \newcommand{\bx}{{\bf x}}
13: \newcommand{\Li}{{\mathop{\rm{Li}}\nolimits}}
14: \renewcommand{\Im}{{\mathop{\rm{Im}}\nolimits}}
15: \newcommand{\Gi}{\rm{Gi}}
16: 
17: \begin{document}
18: 
19: \title{Effects of an electronic topological transition for anisotropic
20:    low-dimensional superconductors}
21:  
22: \author{G. G. N. Angilella}
23: \affiliation{Dipartimento di Fisica e Astronomia, Universit\`a di
24:    Catania,\\ and Istituto Nazionale per la Fisica della Materia,
25:    UdR di Catania,\\ Corso Italia, 57, I-95129 Catania, Italy}
26: \author{E. Piegari}
27: \affiliation{DMFCI, Universit\`a di
28:    Catania, Viale A. Doria, 6, I-95125 Catania, Italy}
29: \affiliation{Istituto Nazionale per la Fisica della Materia,
30:    UdR di Firenze,\\ Via G. Sansone, 1, I-50019 Sesto Fiorentino (FI),
31:    Italy}
32: \author{A. A. Varlamov}
33: \affiliation{Istituto Nazionale per la Fisica della Materia,
34:    UdR di Tor Vergata,\\ DSTFE, Universit\`a di Roma ``Tor Vergata'',\\
35:    Via di Tor Vergata, 110, I-00133 Roma, Italy}
36:     
37: \date{\today}
38: 
39: \begin{abstract}
40: We study the superconducting properties of a two-dimensional superconductor in
41:    the proximity to an electronic topological transition (ETT).
42: In contrast to the 3D case, we find that the superconducting gap at
43:    $T=0$, the critical temperature $T_c$, and the impurity scattering rate
44:    are characterized by a nonmonotonic behavior, with maxima occurring
45:    close to the ETT.
46: We derive analytical expressions for the value of such maxima both in
47:    the $s$-wave and in the $d$-wave case.
48: Such expressions are in good qualitative agreement with the
49:    phenomenological trend recently observed for $T_c^{\rm{max}}$ as a
50:    function of the hopping ratio $t^\prime /t$ across several cuprate
51:    compounds.
52: We further analyze the effect of an ETT on the Ginzburg-Landau stiffness
53:    $\eta$.
54: Instead of vanishing at the ETT, as could be expected, thus giving
55:    rise to an increase of the fluctuation effects, in the case
56:    of momentum-independent electron-electron interaction, we find
57:    $\eta\neq 0$, as a result of an integration over the whole Fermi
58:    surface.\\
59: \pacs{% 
60: 74.20.-z,
61: %Theories and models of superconducting state
62: 74.62.-c,
63: %Transition temperature variations
64: 74.40.+k
65: %Fluctuations
66: %(noise, chaos, nonequilibrium superconductivity, localization, etc.)
67: }
68: \end{abstract} 
69: 
70: \maketitle
71: 
72: \section{Introduction}
73: 
74: The unconventional properties of the normal and superconducting
75:    states of several low-dimensional novel electronic materials is a
76:    source of continuous interest and research.
77: Such materials include the high-$T_c$ cuprate superconductors
78:    (HTSC) \cite{Anderson:97},
79:    as well as some organic superconductors based on doped
80:    BEDT-TTF layers, and the ruthenates.
81: In these materials, the interplay between their reduced dimensionality
82:    and the strength of the effective electron-electron interaction is
83:    believed to be the key for the elusive nature of their normal
84:    state, as well as for the anisotropic gap characterizing their
85:    superconducting state.
86: 
87: A feature common to almost all the material classes listed
88:    above is a quasi-2D dispersion relation, arising from their layered 
89:    structure and stabilized by the tendency to confined coherence
90:    within layers, due to strong correlations
91:    \cite{Clarke:97}.
92: Indeed, flat bands have been observed in nearly all hole- and
93:    electron-doped superconductors \cite{Shen:95}, in the $\kappa$
94:    phase of BEDT-TTF organic superconductors \cite{McKenzie:97}, as
95:    well as in the noncuprate layered superconductor Sr$_2$RuO$_4$
96:    (Ref.~\onlinecite{Yokoya:96}).
97: Clear evidence for a 2D Fermi surface changing topology as a function
98:    of doping has been recently provided by ARPES measurements in LSCO
99:    \cite{Ino:01}.
100: In particular, the role of the proximity to an electronic
101:    topological transition in establishing the unconventional
102:    properties especially of the cuprates has been very early
103:    emphasized (see Ref.~\onlinecite{Markiewicz:97} for a
104:    review).
105: Therefore, in the following we will be mainly concerned with the case
106:    of the high-$T_c$ cuprates~\cite{note:RBCO}.
107: 
108: The term electronic topological transition (ETT) has been proposed in
109:    order to describe the phenomena related to a change of the
110:    connectivity number of the components of the Fermi surface (FS)
111:    \cite{Varlamov:89}. 
112: Such a transition can be driven by several causes such as isotropic
113:    pressure, anisotropic deformation, and the introduction of isovalent
114:    impurities. 
115: All these influences can be parametrized by their effect on the
116:    chemical potential $\mu$ passing through the critical value
117:    $\varepsilon_c$, corresponding to the transition point. 
118: Indeed, such a critical point can be well defined only in a pure
119:    metal at  $T=0$ where a true phase transition
120:    of order $2\frac{1}{2}$ occurs~\cite{Lifshitz:60},
121:    according to Ehrenfest classification.
122: Typical manifestations of an ETT consist in cusp-like anomalies of
123:    physical quantities such as the specific heat \cite{Dorbolo:98},
124:    the density of states (DOS), and the conductivity, as well as in
125:    the appearance of 
126:    asymmetric singularities of the thermal expansion coefficient and
127:    thermoelectric power in the dependence of all these quantities on
128:    $z=\mu -\varepsilon_c$. 
129: A non-zero temperature or the presence of electron scattering results in
130:    the smearing of these anomalies and, strictly speaking, in washing out
131:    the notion of a $2\frac{1}{2}$-order phase transition itself.
132: Moreover, the occurrence of an ETT can be masked by an intervening
133:    structural transition, as could be induced by external pressure.
134: The effects of an ETT on the properties of metals and alloys have been
135:    thoroughly investigated as well
136:    \cite{Varlamov:89,Blanter:94,Bruno:94}.
137: 
138: In lower dimensional metallic systems, an ETT is characterized by yet
139:    stronger anomalies.
140: In particular, the DOS of a 2D metal increases
141:    logarithmically near an ETT, instead of displaying a square-root
142:    cusp, as in the 3D case \cite{note:fluctuations}.
143: Therefore, it has been suggested that an ETT may
144:    be a clue for the understanding of the anomalous superconducting
145:    state of the high-$T_c$ cuprates \cite{Markiewicz:97}.
146: In particular, it is well known that the presence of an ETT in the
147:    spectrum of a 2D superconductor induces a nonmonotonic dependence
148:    of the critical temperature on doping or applied pressure
149:    \cite{Tsuei:90,Markiewicz:97}, in qualitative agreement with the
150:    available experimental results \cite{Zhang:93,Wijngaarden:99}.
151: This has to be contrasted with the 3D case, where an ETT only gives
152:    rise to a step-like behavior in the $z$-dependence of $T_c$
153:    \cite{Makarov:65}.
154: 
155: Moreover, it has been proposed that the proximity to an ETT may be the 
156:    origin of the unconventional normal state of the HTSC.
157: In particular, a marginal Fermi liquid
158:    \cite{Pattnaik:92,Newns:92,Newns:92a,Gopalan:92} or a non Fermi liquid
159:    \cite{Dzyaloshinskii:96} behavior can be naturally derived for a 2D 
160:    electron system near a Van~Hove singularity.
161: More generally, it has been argued that the anomalous
162:    finite-temperature phenomenology of the cuprates stems from the
163:    competition of several broken-symmetry states intervening near one
164:    and the same quantum critical point (QCP)
165:    \cite{Vojta:00,Chakravarty:01}.
166: Recently, Onufrieva \emph{et al.} \cite{Onufrieva:99a,Onufrieva:99b}
167:    showed that an ETT occurring in a 2D square lattice with hopping
168:    beyond nearest neighbors is a QCP, with
169:    two aspects of criticality: the first is related to the singular
170:    behavior of the thermodynamic properties (Van~Hove singularity),
171:    while the second is related to the existence of the critical line
172:    $T=0$, $z>0$ of static Kohn singularities
173:    \cite{Onufrieva:99a,Onufrieva:99b}.
174: The proximity to an ETT may be characterized by spin density wave (SDW),
175:    charge density wave (CDW), and $d$-wave superconducting (dSC)
176:    instabilities, depending on the appropriate interaction channels
177:    included in the analysis.
178: Onufrieva \emph{et al.} argued that SDW fluctuations dominate in the
179:    case of the high-$T_c$ cuprates \cite{Onufrieva:00}.
180: On the other hand, recent studies focussed on the competition between AFM
181:    and AFM-mediated $d$-wave pairing via a diagrammatic approach
182:    \cite{Onufrieva:96}, among dSC,
183:    AFM, and $\pi$-triplet pairing at a mean field level in the
184:    presence of backward scattering \cite{Murakami:98}, among dSC, AFM, 
185:    and CDW within the renormalization group (RG) approach
186:    \cite{Alvarez:98a,Alvarez:98,Gonzalez:00}, among dSC, AFM, and FM
187:    within the RG and the parquet approaches \cite{Irkhin:01}, or
188:    between dSC and an excitonic ordered state \cite{Kiselev:00}.
189: More recently, it has been shown \cite{Honerkamp:01} that elastic
190:    umklapp scattering near a Van~Hove singularity may give rise to an
191:    RVB-like, insulating spin liquid state \cite{Furukawa:98}, which
192:    exhibits both $d$-wave superconducting and AFM correlations,
193:    without being characterized by true symmetry breaking, as is
194:    typical of a quantum ordered state.
195: The competition between superconductivity and various kinds of density 
196:    waves in several low-dimensional electron systems in the presence of a 
197:    Van Hove singularity has been reviewed both from the experimental
198:    and the theoretical point of view in Ref.~\onlinecite{Gabovich:01}.
199: 
200: In this paper, we will concentrate on a single superconducting
201:    instability (towards either an $s$-wave or a $d$-wave superconducting
202:    state), in the weak coupling limit, thus neglecting altogether any
203:    other competing ordered phase, for a 2D electron system near an
204:    ETT.
205: This is of course justified only if all other instabilities are
206:    characterized by weaker couplings, which may not be the case for
207:    the cuprates.
208: However, such an approximation will enable us to derive an analytical
209:    expression for the maximum gap $\Delta_0$ near the ETT as a
210:    function of the band details.
211: Such results are in good qualitative agreement with recent studies of
212:    $T_c$ in the cuprates correlated with material dependent
213:    properties, such as the ratio of next-nearest to nearest neighbor
214:    hopping \cite{Pavarini:01}.
215: 
216: We will also study the effect of an ETT on the Ginzburg-Landau
217:    stiffness $\eta\propto\Gi^{-1}$, where $\Gi$ is the
218:    Ginzburg-Levanyuk parameter, which characterizes the manifestation range of
219:    fluctuations near $T_c$ (Ref.~\onlinecite{Varlamov:99}). 
220: In the case of an isotropic FS, $\eta$ is proportional to the square
221:    of the Fermi velocity.
222: In the vicinity of an ETT, due to the presence of `slow' electrons
223:    near the saddle point in the electronic spectrum, one may
224:    expect an increase of fluctuations.
225: However, we will show that, in the case of a momentum-independent
226:    electron-electron interaction, all electronic states on the FS
227:    participate in establishing the superconducting correlations.
228: Such correlations give rise to a superconducting stiffness, whose
229:    value is of the same order of magnitude of the result obtained for
230:    an isotropic electronic spectrum \cite{Perali:00}, to the lowest
231:    order in $z/E_{\mathrm{F}}$.
232: 
233: The paper is organized as follows.
234: In Sec.~\ref{sec:model}, we introduce a dispersion relation beyond
235:    nearest neighbors for an electron system in a 2D square lattice, as 
236:    is typical for the cuprates, and discuss its corresponding singular 
237:    DOS.
238: In Sec.~\ref{sec:super}, we study the superconducting gap $\Delta_0$ at
239:    $T=0$ as a function of the critical parameter $z$, in the case of
240:    $s$- and $d$-wave pairing.
241: In Sec.~\ref{sec:impurities}, we discuss the effect of impurities on
242:    the normal state DOS and show that the proximity to an ETT gives
243:    rise to a nonmonotonic $z$-dependence of the renormalized quasiparticle
244:    inverse lifetime $\tau^{-1}$.
245: In Sec.~\ref{sec:GL}, we calculate the Cooper pair propagator near an
246:    ETT, and discuss the effects of an ETT on the superconducting
247:    fluctuations.
248: We eventually summarize in Sec.~\ref{sec:conclusions}.
249: 
250: 
251: \section{The model}
252: \label{sec:model}
253: 
254: Detailed band structure calculations within the local-density approximation
255:    (LDA)~\cite{Andersen:95}, as well as ARPES
256:    experiments~\cite{Randeria:99}, show that a realistic 
257:    tight-binding approximation for the band dispersion of most
258:    quasi-2D cuprates has to be expanded at least up to
259:    next-nearest neighbors hopping.
260: We then assume the following rigid band dispersion relation for a
261:    tetragonal lattice: 
262: \begin{equation}
263: \xi_\bk =\varepsilon_\bk -\mu =-2t(\cos k_x+\cos
264:    k_y)+4t^{\prime}\cos k_x\cos k_y-\mu ,   
265: \label{eq:bdisp}
266: \end{equation}
267: where $\mu $ denotes the chemical potential, and the components of
268:    the wavevector $\bk$ are measured in units of the inverse
269:    lattice spacing.
270: A non-zero value of the hopping ratio $r=t^\prime /t$, measuring the
271:    ratio of next-to-nearest vs nearest neighbors hopping, slightly
272:    modulates the actual shape of the Fermi line $\xi_\bk =0$,
273:    and in particular destroys perfect nesting at $\mu =0$ as well as
274:    electron-hole symmetry (Fig.~\ref{fig:FS}). 
275: In order to have a flat minimum in $\xi_\bk$ around the
276:    $\Gamma$ point, as is observed experimentally for the majority of
277:    the cuprates \cite{Dessau:93,Abrikosov:93,Gofron:94}, the condition
278:    $0<r< \frac{1}{2}$ must be fulfilled.
279: The role of an extended saddle point in stabilizing superconductivity
280:    against other possible low-energy instabilities has been
281:    established within the renormalization group (RG) approach in the
282:    weak-coupling limit \cite{Alvarez:98}.
283: Moreover, it has been shown that an increase of $r$ in the mentioned range
284:    correlates with an increase of the maximum
285:    $T_c$ across different classes of cuprate superconductors
286:    \cite{Pavarini:01}. 
287: However, changes of the \emph{shape} of the FS
288:    resulting from screening effects can also correlate with changes
289:    in the superconducting properties in an indirect way.
290: Indeed, a deformation of the FS
291:    also induces a change of the phase space effectively
292:    probed by the electron-electron interaction \cite{Hodges:71}. 
293: This is particularly relevant for several models proposed for the
294:    HTSC, characterized by effective interactions peaked at
295:    $X=(0,\pi )$, namely exactly where the
296:    Fermi line is most sensible to a change in the hopping range $r$.
297: Possible realizations of such a strongly anisotropic
298:    $\bk$-dependent potential include the interaction mediated by
299:    antiferromagnetic spin fluctuation \cite{Millis:90} or by
300:    quasicritical stripe fluctuations, due to the
301:    proximity to a QCP near optimal doping at $T=0$
302:    (Refs.~\onlinecite{Emery:93,Emery:95a,Castellani:95}), as well as
303:    electron-electron interactions enhanced by interlayer
304:    pair-tunneling (ILT) \cite{Chakravarty:93,Angilella:99}. 
305: 
306: \begin{figure}[ht]
307: \centering
308: \includegraphics[height=0.9\columnwidth,angle=-90]{fig1.ps} 
309: \caption{Fermi line $\xi_\bk = 0$, Eq.~(\protect\ref{eq:bdisp}), for a
310:    value of the hopping ratio $r=0.45$.
311: As the chemical potential varies from the bottom to the top of the
312:    band, the Fermi line changes topology, evolving from a closed
313:    contour around the $\Gamma$ point, to a contour, whose
314:    continuation in the higher order Brillouin zones closes around
315:    $M=(\pi,\pi)$.
316: The change of topology (ETT) occurs when the Fermi line touches the
317:    zone border, i.e. at $X=(\pi,0)$, and symmetry related points.
318: The thicker solid lines evidence the hyperbola-like shape of the Fermi
319:    line, Eq.~(\protect\ref{eq:hyperbola}), around $X$.
320: }
321: \label{fig:FS}
322: \end{figure}
323: 
324: 
325: As the chemical potential $\mu$ in Eq.~(\ref{eq:bdisp}) varies
326:    from the bottom, $\varepsilon_\perp =-4t (1-r)$, to the
327:    top of the band, $\varepsilon_\top =4t(1+r)$, the Fermi
328:    line $\xi_\bk =0$ evolves from an electron-like contour,
329:    closed around the $\Gamma$ point, to a hole-like contour, whose
330:    continuation into higher order Brillouin zones closes around the
331:    $M=(\pi ,\pi )$ point.
332: In doing so, an ETT is passed exactly at $\mu
333:    =\varepsilon_c=-4t^\prime$, where the Fermi line touches the
334:    zone boundaries, and assumes the asteroid-like shape depicted in
335:    Fig.~\ref{fig:FS}.
336: It has been shown that such a critical value for the Fermi energy is
337:    stable against the RG flow for any repulsive electron-electron
338:    interaction \cite{Gonzalez:96,Gonzalez:01}.
339: Such a result has been recently confirmed also when self-energy
340:    effects to the quasiparticle dispersion relation are included, thus 
341:    demonstrating that the pinning of the Fermi surface to a Van~Hove
342:    singularity can actually take place for a rather wide range of hole 
343:    concentration \cite{Irkhin:01b}.
344: The condition $\mu = \varepsilon_c$ corresponds to having a saddle
345:    point at $X=(\pi ,0)$ in the 
346:    single-particle dispersion relation $\varepsilon_\bk$,
347:    which, for small wavevector displacements from $X$ and symmetry
348:    related points, can be expanded as 
349: \begin{equation}
350: \varepsilon_\bk -\varepsilon_c\sim \frac{p_{1}^{2}}{2m_{1}}-%
351: \frac{p_{2}^{2}}{2m_{2}} \equiv \epsilon_\bp,  
352: \label{eq:hyperbola}
353: \end{equation}
354: where $p_1=k_x$, $p_2=k_y-\pi $, and $m_{1,2}=[2t(1\pm 2r)]^{-1}$ are
355:    the eigenvalues of the effective mass tensor \cite{Morse}. 
356: Here and below, we choose units such that $\hbar $ and the lattice
357:    spacings are set equal to unity.
358: 
359: For $\varepsilon_\perp \leq \varepsilon \leq \varepsilon_\top$,
360:    the density of states $\nu (\varepsilon )=\oint_{\varepsilon_\bk
361:    =\varepsilon }d\Omega_\bk |\nabla_\bk
362:    \varepsilon_\bk |^{-1}$ can be computed
363:    analytically as \cite{Xing:91}
364: \begin{equation}
365: \nu (\varepsilon )=\frac{1}{\pi
366:    ^{2}}\frac{1}{\sqrt{4t^{2}-\varepsilon_c \varepsilon }}K\left[
367:    \frac{1}{2}\sqrt{\frac{16t^{2}-(\varepsilon +\varepsilon_c
368:    )^{2}}{4t^{2}-\varepsilon_c \varepsilon }}\right] , 
369: \label{eq:DOSexact}
370: \end{equation}
371: where $K(k)$ denotes the complete elliptic integral of first kind of
372:    modulus $k$ (Ref.~\onlinecite{GR}).
373: At $\varepsilon = \varepsilon_c$, $\nu (\varepsilon )$ diverges
374:    logarithmically.
375: Making use of the appropriate asymptotic expansion for $K(k)$
376:    (Ref.~\onlinecite{AS}), it is then customary to identify a regular
377:    and a singular contribution to $\nu (\varepsilon )$ as
378: \begin{equation}
379: \nu (\varepsilon ) = \nu_0 (\varepsilon ) + \delta \nu (\varepsilon ),
380: \label{eq:DOScontribs}
381: \end{equation}
382: with $\nu_0 (\varepsilon )$ being a continuous function of
383:    $\varepsilon$ over the whole bandwidth such that $\nu_0 (\varepsilon_c )
384:    = 0$, and
385: \begin{equation}
386: \delta\nu(\varepsilon) = 2\rho \ln
387:    \left| \frac{4\sqrt{2}/\pi^2 \rho}{\varepsilon - \varepsilon_c} \right|,
388: \end{equation}
389: where $\rho^{-1} = 4 \pi^2 t\sqrt{1-4r^2}$ (Fig.~\ref{fig:DOS}).
390: In the following, we shall often make use of the critical
391:    parameter $z = \mu-\varepsilon_c$, measuring the
392:    distance of the chemical potential from the ETT.
393: 
394: \begin{figure}[ht]
395: \centering
396: \includegraphics[height=0.9\columnwidth,angle=-90]{fig2.ps}
397: \caption{Total density of states ($\nu$), Eq.~(\protect\ref{eq:DOSexact}), and
398:    regular ($\nu_0$) and singular ($\delta\nu$) contributions to the DOS,
399:    Eq.~(\protect\ref{eq:DOScontribs}), as a function of energy
400:    $\varepsilon$, ranging from the bottom ($\varepsilon_\bot$) to the
401:    top ($\varepsilon_\top$) of the band.
402: The inset shows the effect of a non-zero energy broadening
403:    $\Gamma$ (here, $\Gamma\sim 0.5$~\% of the total bandwidth) on the
404:    same quantities [see Eq.~(\protect\ref{eq:dnuG})].
405: }
406: \label{fig:DOS}
407: \end{figure}
408: 
409: 
410: \section{The effect of an ETT on the superconducting gap}
411: \label{sec:super}
412: 
413: We will now discuss the effects of the proximity to an ETT on the
414:    superconducting properties of a 2D, single-layer
415:    system, both for an $s$-wave and for a $d$-wave order parameter, in 
416:    the weak coupling limit.
417: Our starting point will be the BCS equation for the gap function
418:    $\Delta_\bk$, which at $T=0$ reads as
419: \begin{equation}
420: \Delta_\bk = -\frac{1}{N} \sum_{\bk^\prime} V_{\bk\bk^\prime}
421:    \frac{\Delta_{\bk^\prime}}{2 E_{\bk^\prime}} .
422: \label{eq:BCST=0}
423: \end{equation}
424: Here, $E_\bk = \sqrt{\xi_\bk^2 + |\Delta_\bk |^2}$ is the upper branch
425:    of the
426:    superconducting excitation spectrum, $V_{\bk\bk^\prime}$ denotes
427:    the interparticle potential, and the sum runs over all the $N$
428:    $\bk$ points in the 1BZ.
429: In the case of $s$-wave symmetry, we assume $V_{\bk\bk^\prime} = -\lambda$,
430:    i.e. a constant over the whole 1BZ,
431:    whereas in the $d$-wave case we take the potential in the separable
432:    form $V_{\bk\bk^\prime} = -\lambda g_\bk
433:    g_{\bk^\prime}$, $g_\bk = \frac{1}{2} (\cos k_x - \cos k_y )$ being
434:    the lowest-order lattice harmonic corresponding to $d$-wave
435:    symmetry.
436: Accordingly, one has $\Delta_\bk = \Delta_0$ in the $s$-wave case, and
437:    $\Delta_\bk = \Delta_0 g_\bk$ in the $d$-wave case, respectively.
438: It is worth emphasizing that, in both cases, the coupling constant
439:    $\lambda>0$ has been assumed independent of doping.
440: This amounts to neglecting higher-order correlation effects among
441:    interacting particles \cite{Perali:00}.
442: Moreover, the weak coupling hypothesis allows us to neglect
443:    renormalizations of the shape of the Fermi surface, which are
444:    certainly expected in the strong coupling limit, and are known to
445:    give rise to another kind of ETT as well \cite{Chubukov:97a}.
446: 
447: Eq.~(\ref{eq:BCST=0}) implicitly neglects the possibility of any
448:    pairing instability other than singlet superconductivity in the
449:    Cooper channel (characterized by a pair relative momentum ${\bf
450:    P}=0$).
451: Possible alternative intervening pairing instabilities include, \emph{e.g.,}
452:    antiferromagnetism and the $\pi$-triplet paired state
453:    \cite{Murakami:98}.
454: Such instabilities would be characterized by large momentum transfer
455:    near the `hot spots' $(0,\pi)$ and $(\pi,0)$.
456: They have been shown to coexist and win out singlet superconductivity
457:    at a mean field level near half-filling, when backward scattering
458:    is a relevant process \cite{Murakami:98}.
459: However, within our weak-coupling approximation, it is consistent to
460:    retain only one kind of instability (namely, Cooper pairing in the
461:    singlet, ${\bf P}=0$ channel, with either $s$- or $d$-wave
462:    symmetry), under the assumption that other instabilities are
463:    characterized by weaker couplings.
464: 
465: We will first analyze the gap equation, Eq.~(\ref{eq:BCST=0}), close
466:    to an ETT ($|z|\ll 4t$).
467: In the $s$-wave case, the summation over the 1BZ in
468:    Eq.~(\ref{eq:BCST=0}) can be transformed into an integral over
469:    energy weighted by the DOS, which we approximate by its singular
470:    part $\delta\nu(\varepsilon)$ in Eq.~(\ref{eq:DOScontribs}).
471: The Fermi line corresponding to the ETT divides the 1BZ in two
472:    regions, $\varepsilon_\bk <0$ and $\varepsilon_\bk >0$, which are
473:    electron-hole conjugated of each other.
474: Separating the contributions coming from these regions, the gap
475:    equation can be compactly written as:
476: \begin{equation}
477: \frac{1}{\lambda\rho} = S_+ + S_- ,
478: \label{eq:gapeqs}
479: \end{equation}
480: where $S_\pm$ represent the pairing susceptibility integrated between
481:    the ETT and either band edges (see App.~\ref{app:conti1} for
482:    details).
483: 
484: While in 3D BCS theory $S_\pm$ are logarithmically divergent in the limit
485:    $\Delta_0 \to0$ (Ref.~\onlinecite{AGD}), the proximity to an ETT in 2D makes
486:    them divergent as $\sim\ln^2 \Delta_0$.
487: Direct inspection of Eq.~(\ref{eq:gapeqs}) as well as numerical
488:    calculations show that $\Delta_0 (z)$ is maximum near the ETT.
489: Such a nonmonotonic dependence of the superconducting gap on the
490:    critical parameter $z$ is in agreement with the phenomenology of
491:    the HTSC, where $\Delta_0$ and $T_c$ are characterized by a
492:    parabola-like dependence on doping \cite{Zhang:93}.
493: This has to be contrasted with the step-like behavior observed in the 
494:    3D case \cite{Makarov:65}, where it has to be emphasized that no
495:    divergence occurs in the DOS at the ETT.
496: However, due to the electron-hole symmetry breaking induced by a
497:    non-zero hopping ratio $r$, $\Delta_0 (z)$ is not an
498:    even function of $z$, and its maximum will actually occur not
499:    exactly at $z=0$, as will be discussed below.
500: Solving Eq.~(\ref{eq:gapeqs}) for $\Delta_0$ in the
501:    weak-coupling limit ($\lambda\rho\ll 1$, $\Delta_0 \ll 4t$),
502:    at $z =0$ one finds:
503: \begin{widetext}
504: \begin{equation}
505: \Delta_0 (z=0) \simeq \frac{8\sqrt{2}}{\pi^2 \rho} \exp\left(
506: - \sqrt{\frac{1}{\lambda\rho} + \frac{1}{4} \ln^2 \frac{1+2r}{1-2r}}
507:    \right),
508: \qquad\qquad\mbox{($s$-wave)}
509: \label{eq:deltamaxs}
510: \end{equation}
511: which can be then taken as a first approximation to the gap maximum at
512:    $T=0$, in the $s$-wave case.
513: Following the same procedure, qualitatively similar results can be
514:    derived for the critical temperature $T_c$ as a function of
515:    $z$ (see also Refs.~\onlinecite{Markiewicz:97,Tsuei:90}).
516: 
517: In the $d$-wave case, due to the anisotropic $\bk$-dependence of the
518:    integrand in Eq.~(\ref{eq:BCST=0}), it is not possible to
519:    explicitly separate the integration over energy, and a different
520:    approach must be followed (see App.~\ref{app:conti1} for details).
521: However, the proximity to an ETT does endow the pairing susceptibility
522:    with an analogous asymptotic low-$\Delta_0$ behavior, like in the $s$-wave
523:    case, which eventually results in the following weak-coupling
524:    expression for the gap amplitude at $T=0$, $z=0$
525:    ($\lambda\rho\ll1$, $\Delta_0 \ll 4t$):
526: \begin{equation}
527: \Delta_0 (z=0) \simeq 4 t f_1 (r) \exp\left( - \sqrt{\frac{1}{\lambda\rho} +
528:    f_2 (r)} \right),
529: \qquad\qquad \mbox{($d$-wave)}
530: \label{eq:deltamaxd}
531: \end{equation}
532: where $f_1 (r) = 2b^{-1} (1-4r^2) (\sqrt{1+2r}+\sqrt{1-2r})^{-1}$,
533:    $f_2 (r) = \ln^2 (b\pi\sqrt{1-4r^2}) + 2 \ln\sqrt{1-4r^2}
534:    \ln[2\pi^{-1} \sqrt{1-4r^2} (\sqrt{1+2r}+\sqrt{1-2r})^{-1}]$, and
535:    $b=e^2 /8$.
536: \end{widetext}
537: 
538: Fig.~\ref{fig:deltamax} shows $\Delta_0 (0)$ both in the $s$-
539:    and in the $d$-wave case, Eq.~(\ref{eq:deltamaxs}) and
540:    (\ref{eq:deltamaxd}), respectively, as a function of the hopping
541:    ratio $r$, for several values of $\lambda/t$.
542: In view of the fact that $T_c \propto \Delta_0$, as in any mean-field
543:    theory, Fig.~\ref{fig:deltamax} is in good qualitative agreement with
544:    Fig.~5 of Ref.~\onlinecite{Pavarini:01}, showing a direct correlation
545:    between the experimental $T_c^{\rm{max}}$ and the hopping range
546:    $r$ for several cuprate compounds.
547: Moreover, our results suggest that such an effect is a general
548:    consequence of the proximity to an ETT, and is roughly independent
549:    of the superconducting pairing symmetry.
550: 
551: \begin{figure}[ht]
552: \centering
553: \includegraphics[height=0.9\columnwidth,angle=-90]{fig3.ps}
554: \caption{Normalized gap amplitude $\Delta_0 (z=0,r)/\Delta_0 (z=0,r=0)$ at
555:    $T=0$, as a function of the hopping ratio $r=t^\prime /t$, for
556:    different couplings $\lambda/t = 0.9\div 1.1$ (bottom to top).
557: Continuous lines refer to the $s$-wave case,
558:    Eq.~(\protect\ref{eq:deltamaxs}), while dashed lines refer to the
559:    $d$-wave case, Eq.~(\protect\ref{eq:deltamaxd}).
560: One can recognize the direct correlation between $T_c^{\rm{max}}
561:    \propto \Delta_0 (z=0)$ and $r$, as observed in
562:    Ref.~\protect\onlinecite{Pavarini:01}.
563: }
564: \label{fig:deltamax}
565: \end{figure}
566: 
567: 
568: Expanding $\Delta_0 (z)$, as implicitly defined by
569:    Eq.~(\ref{eq:gapeqs}) around $z=0$, one finds that the
570:    maximum of $\Delta_0$ actually occurs at a larger value of
571:    the critical parameter, which, in the $s$-wave case, is given by
572: \begin{equation}
573: z_{\rm{max}} \simeq \frac{1}{8t} \Delta_0^2 (0) \ln
574:    \frac{1+2r}{1-2r}.
575: \label{eq:zmaxs}
576: \end{equation}
577: A qualitatively analogous result applies to the $d$-wave case.
578: Therefore, as an effect of the electron-hole asymmetry induced by a
579:    non-zero hopping ratio $r$, the maximum in $\Delta_0$ is
580:    actually located in the hole-like region ($z_{\rm{max}} >0$),
581:    in agreement with the phenomenology of some hole-doped cuprate
582:    compounds.
583: For instance, a representative high-$T_c$ cuprate such as LSCO is
584:    characterized by an optimal doping level of $x_{\rm{opt}}
585:    \simeq 0.15$, lying in the hole-doped region, while a
586:    doping-dependent crossover from a hole-like to an electron-like FS
587:    has been clearly observed at a somewhat larger doping
588:    $x_c \simeq 0.20$ (Ref.~\onlinecite{Ino:01}).
589: On the contrary, no direct evidence of a change in the FS topology has
590:    been so far reported for Bi-2212 (Ref.~\onlinecite{Loeser:96}), whose FS
591:    displays a hole-like character at all dopings, including optimal
592:    doping.
593: This implies that the ETT is located at a much larger distance from optimal
594:    doping, which is consistent with Eq.~(\ref{eq:zmaxs}) above, given
595:    the larger value of the gap amplitude of Bi-2212 than that of LSCO.
596: 
597: \section{Effect of impurities}
598: \label{sec:impurities}
599: 
600: We now turn to consider the more realistic case, in which electron
601:    scattering from non-magnetic impurities is included.
602: Here, we will be mainly concerned with the normal state properties.
603: A finite quasiparticle lifetime induces a
604:    broadening of the energy linewidth of a quasiparticle state.
605: Therefore, the use of a quasiparticle description and the definition
606:    of a Fermi surface for impure metals can, at first sight, be
607:    objected.
608: Indeed, quasimomentum is a `good' quantum number
609:    only for electrons moving in a periodic potential \cite{Landau:X}.
610: Scattering of electrons on impurities results in momentum relaxation
611:    and in the corresponding smearing of the Fermi surface in momentum
612:    space.
613: The characteristic scale of such a smearing is $\sim\tau^{-1}$, where
614:    $\tau$ is the elastic relaxation time at low temperatures.
615: The value of $\tau$ can easily exceed the quasiparticle energy $\sim
616:    T$ even for moderate impurity concentrations.
617: 
618: Nevertheless, elastic scattering does not result in energy relaxation.
619: This means that, in principle, one can solve exactly the eigenvalue
620:    problem for the Hamiltonian of the electron in a lattice with some
621:    specific realization of the impurity potential.
622: The eigenstates of such Hamiltonian can then be chosen as a basis in
623:    the Hilbert space and the `Fermi surface' in this space can be
624:    defined as the surface separating the low-energy occupied
625:    eigenstates from the high-energy empty eigenstates at zero
626:    temperature.
627: It is evident that the Fermi surface defined in this way does exist, and
628:    that the elastic scattering has no effect on the quasiparticle
629:    lifetime in the vicinity of the Fermi surface (see also
630:    Refs.~\onlinecite{Varlamov:89,Blanter:94}).
631: 
632: The effect of a nonvanishing impurity scattering rate can then
633:    be accounted for in the DOS by means of a convolution between
634:    $\nu(\varepsilon)$ and a Lorentzian of finite width $\Gamma$:
635: \begin{equation}
636: \nu_\Gamma (\varepsilon) = \int d\xi \, \frac{1}{\pi}
637:    \frac{\Gamma}{(\xi-\varepsilon)^2 + \Gamma^2} \nu(\xi).
638: \end{equation}
639: Such a procedure \cite{Blanter:94} effectively smears out the
640:    logarithmic singularity in 
641:    the DOS at the ETT into a pronounced maximum of finite width
642:    $\sim\Gamma$ (see inset in Fig.~\ref{fig:DOS}).
643: Nonetheless, it is still possible to separate a `regular' and a
644:    `singular' contribution to $\nu_\Gamma (\varepsilon)$ as 
645: \begin{equation}
646: \nu_\Gamma (\varepsilon) = \nu_\Gamma^0 (\varepsilon) +
647:    \delta\nu_\Gamma (\varepsilon),
648: \end{equation}
649: with $\delta\nu_\Gamma (\varepsilon)$ now given by \cite{Blanter:94}:
650: \begin{equation}
651: \delta\nu_\Gamma (\varepsilon) = 2\rho \ln
652:    \frac{4\sqrt{2}/\pi^2 \rho}{\sqrt{(\varepsilon - \varepsilon_c )^2
653:    + \Gamma^2}} .
654: \label{eq:dnuG}
655: \end{equation}
656: From a physical point of view, the energy linewidth broadening
657:    associated to impurity scattering has the effect of `blurring' the
658:    Fermi line.
659: As a consequence, one expects that the ETT occurs slightly farther
660:    from $\mu=\varepsilon_c$ (i.e., for $|z|>0$),
661:    as soon as such a blurred Fermi line touches the border of the 1BZ
662:    \cite{Varlamov:89}.
663: 
664: Thus far, we have assumed a constant energy linewidth $\Gamma$ over
665:    the whole band.
666: This is clearly an approximation, since the quasiparticle lifetime
667:    $\tau_\bk$ is generally an anisotropic quantity over the 1BZ
668:    \cite{Hlubina:95}.
669: The last statement holds true even in the simplest case of isotropic
670:    impurity scattering, due to the anisotropy of the single-particle
671:    band structure.
672: In particular, the proximity to an ETT in 2D endows the
673:    quasiparticle lifetime with a nonmonotonic behavior, in contrast to
674:    the 3D case, where a step-like $z$-dependence was
675:    found \cite{Varlamov:89}.
676: Following Ref.~\onlinecite{Varlamov:89}, one can write the self-consistent
677:    equation for the retarded quasiparticle self-energy
678:    $\Sigma^{\rm R}$ due to impurity scattering as
679: \begin{equation}
680: \Sigma^{\rm R} (\omega,z) = \frac{n_i |u_0 |^2}{(2\pi)^2} \int d^2 \bp \,
681: \left[ \epsilon_\bp + z + \omega - \Sigma^{\rm R} (\omega,z)
682:    \right]^{-1} ,
683: \end{equation}
684: where $\epsilon_\bp$ is the asymptotic single-particle dispersion
685:    relation near the saddle point defining the ETT,
686:    Eq.~(\ref{eq:hyperbola}), $n_i$ denotes the concentration of
687:    impurities, and $u_0$ is the impurity scattering strength, here
688:    assumed independent of $\bp$.
689: Performing the integrations as in
690:    Refs.~\onlinecite{Varlamov:85,Varlamov:89}, but now for the 2D case,
691:    one arrives at the self-consistent expression:
692: \begin{equation}
693: \Sigma^{\rm R} = -\frac{i}{2\tau_0} \ln \left(
694: \frac{\sqrt{1+z} + \sqrt{1 + \omega +
695:    \Sigma^{\rm R}}}{
696: \sqrt{-\omega-z+\Sigma^{\rm R}}}
697: \right),
698: \label{eq:Sigma}
699: \end{equation}
700: where $\tau_0^{-1} = \pi^{-1} n_i |u_0 |^2 (m_1 m_2)^{1/2}$, and all
701:    energies are in units of a cut-off energy $\sim\rho^{-1}$.
702: Fig.~\ref{fig:sigma} shows the renormalized quasiparticle inverse
703:    lifetime $\tau^{-1} = -2 \Im \Sigma^{\rm{R}}
704:    (\omega = 0,z)$ as a function of the critical
705:    parameter $z$, for several values of $\tau_0^{-1}$.
706: As anticipated, one observes a maximum in $\tau^{-1}$ at
707:    $z\gtrsim 0$, as an effect of crossing an ETT.
708: 
709: \begin{figure}[ht]
710: \centering
711: \includegraphics[height=0.9\columnwidth,angle=-90]{fig4.ps}
712: \caption{Renormalized quasiparticle inverse lifetime
713:    $\tau^{-1} = -2 \Im \Sigma^{\rm R}$,
714:    Eq.~(\protect\ref{eq:Sigma}), resulting from isotropic impurity
715:    scattering.
716: Different curves correspond to increasing values of
717:    $\tau_0^{-1}$ (bottom to top).
718: The proximity to an ETT induces a nonmonotonic dependence of
719:    $\tau^{-1}$ on the critical parameter $z$, with
720:    $\tau^{-1}$ assuming its maximum value at $z\protect\gtrsim
721:    0$.
722: }
723: \label{fig:sigma}
724: \end{figure}
725: 
726: 
727: The study of a 2D superconductor in the dirty limit goes
728:    beyond the reach of the present analysis.
729: The dependence of $\Delta_0$ as well as of the DOS
730:    on the impurity concentration has been
731:    derived in the $d$-wave case in Ref.~\onlinecite{Sun:95}, for an
732:    isotropic dispersion relation.
733: In $d$-wave superconductors, Dirac-like single-particle excitations
734:    can be created at virtually no energy cost near the gap nodes
735:    \cite{Lee:97,Lee:97a,Balents:98}.
736: Within the QCP scenario, long-range interaction between such gapless
737:    modes is mediated via the fluctuations of an intervening order
738:    parameter at $T=0$.
739: Current proposals for the HTSC include the possibility of the
740:    proximity to a quantum ordered phase characterized by either charge 
741:    or (AFM) spin fluctuations, as well as fluctuations related to the
742:    opening of another subdominant contribution to the superconducting
743:    OP, usually accompanied by time-reversal breaking
744:    \cite{Vojta:00,Chakravarty:01}.
745: Recent results for 2D $d$-wave superconductors in the presence of
746:    disorder yield corrections to the density of states coming both
747:    from the diffusion (${\bf Q}=0$) and the Cooperon mode
748:    \cite{Khveshchenko:01}, as well as from the diffusive mode with
749:    ${\bf Q}=(\pi,\pi)$ \cite{Yashenkin:01}.
750: 
751: \section{Ginzburg-Landau stiffness near an ETT}
752: \label{sec:GL}
753: 
754: As is well known, the normal state of HTSC is characterized by several
755:    anomalous properties at the transition edge.
756: Such properties include a peak in the $c$-axis resistivity, an
757:    anomalously large sign-changing $c$-axis magnetoresistance, as well
758:    as the opening of a pseudogap, which is observed both in the
759:    $c$-axis optical conductivity, in tunneling experiments, and  
760:    in the NMR relaxation rate (see Ref.~\onlinecite{Varlamov:99} for a
761:    review).
762: On the basis of the Fermi liquid theory, it has been recently
763:    demonstrated that the renormalization of the one-electron DOS
764:    in the vicinity of the Fermi level due to the
765:    electron-electron interaction in the Cooper channel is able to
766:    explain satisfactorily many of these pseudogap-like manifestations
767:    both in the overdoped and in the optimally doped compounds
768:    \cite{Varlamov:99}.
769: Moving across the phase diagram of the HTSC from the overdoped,
770:    bad metallic region, towards underdoping, the enhancement of the
771:    mentioned effects correlates with an
772:    increase of the Ginzburg-Levanyuk parameter, ${\Gi}_{\mathrm{(2D)}}
773:    \approx T_c /E_{\rm{F}}$, thus making perturbation theory less reliable
774:    \cite{Varlamov:99}. 
775: Nevertheless, such a rapid growth of the normal state anomalies with
776:    the decrease of doping
777:    strongly overcomes the theoretical predictions, thus making
778:    it difficult to attribute such an effect to the mere shrinking of
779:    the FS.
780: 
781: Indeed, ARPES studies indicate a marked increase of the
782:    FS anisotropy in the $ab$-plane with underdoping, which is
783:    accompanied by the development of an
784:    extended saddle point in the electronic spectrum
785:    \cite{Randeria:99}. 
786: One can identify two characteristic energy scales related with such a
787:    FS, namely the size of its `bulk' part, $E_{\rm{F}} \approx 0.3$~eV, and
788:    the `width' of the saddle point, $|z| = |\mu - \varepsilon_c
789:    |\approx 0.01$~eV.
790: The large difference in the magnitude of such energy scales is
791:    therefore suggestive of a crossover, related to the special role
792:    played by the electronic states lying close to the saddle point
793:    (the so-called `slow' electrons).
794: Intuitively, one may expect a replacement of $E_{\rm{F}}$ with a
795:    small $z$ in the denominator of $\Gi_{\mathrm{(2D)}}$,
796:    which would result in the breakdown of the perturbative approach
797:    developed in Ref.~\onlinecite{Varlamov:99}.
798: Nevertheless, as is demonstrated below, the sole existence of an ETT in the
799:    electronic spectrum is not able to change the character of the isotropic
800:    electron-electron interaction in the Cooper channel.
801: This implies that the breakdown of fluctuation theory in the
802:    underdoped compounds has to be related to some other properties of
803:    the HTSC.
804: In order to substantiate the above statements, in the present section
805:    we shall derive the momentum dependence of the
806:    two-particle Green function in the Cooper channel near $T_c$ and
807:    close to an ETT.
808: 
809: In the case of a 2D electron system characterized by the approximate 
810:    spectrum given by
811:    Eq.~(\ref{eq:hyperbola}), close to an ETT,
812:    the single-particle Green function can be written as
813: \begin{equation}
814: G(\bp,\omega_n; z ) = (i\omega_n - \epsilon_\bp + z)^{-1} .
815: \end{equation}
816: Here, the electron quasi-momentum $\bp$ is measured from the saddle
817:    point location, as in 
818:    Eq.~(\ref{eq:hyperbola}), and $\omega_n =2\pi T(n+\frac{1}{2})$ are the 
819:    fermionic Matsubara
820:    frequencies. As already emphasized in Sec.~\ref{sec:model},
821:    the condition $z >0$
822:    describes the case of an open Fermi surface without any disrupted
823:    neck, whereas the opposite one, $z<0$, is appropriate to a closed
824:    Fermi surface with respect to the $\Gamma$ point
825:    (Fig.~\ref{fig:FS}).
826: The two-particle Green function in the Cooper channel $L(\bq,\Omega_{\nu}
827:    )$ can be expressed within the ladder approximation by means of
828:    the 
829:    polarization operator $\Pi (\bq,\Omega_{\nu} )$ as \cite{Varlamov:99}
830: \begin{equation}
831: L^{-1} (\bq,\Omega_{\nu}; z, T ) = \lambda^{-1} - \Pi (\bq,\Omega_{\nu};
832:       z, T ).
833: \label{eq:pol1}
834: \end{equation}
835: Here, $\lambda>0$ denotes the momentum independent effective
836:    electron-electron interaction, $\bq$ is the Cooper pair
837:    momentum, and 
838: \begin{widetext}
839: \begin{equation}
840: \Pi (\bq,\Omega_\nu ;z,T) = T \sum_{\omega_n} \int \frac{d^2
841:    \bp}{(2\pi)^2} G(\bp+\bq,\omega_{n+\nu}; z ) G(-\bp,-\omega_n ;z )
842: \equiv T \sum_{\omega_n} I (\bq,\omega_{n+\nu} , -\omega_n ;z),
843: \label{eq:pol2}
844: \end{equation}
845: \end{widetext}
846: with $\Omega_\nu = 2\pi T\nu$ the bosonic Matsubara frequencies.
847: 
848: The superconducting critical temperature can be characterized 
849:    as the temperature at which $L$ presents a pole at $\bq=0$ and
850:    $\Omega_\nu=0$. 
851: The procedure to deal with the integral in Eq.~(\ref{eq:pol2}) is
852:    outlined in App.~\ref{app:conti2}.
853: One eventually arrives at the result:
854: \begin{equation}
855: \Pi(0,0;z,T) = \frac{m}{\pi} T \sum_{\omega_n \geq
856:    0}^{\omega_{\rm{D}} / (2\pi 
857:    T)} \frac{1}{\omega_n} \ln \left( \frac{\omega_{\rm{D}}^2}
858:     {\omega_n^2 + z^2}
859:    \right),
860: \label{eq:pol3}
861: \end{equation}
862: where $m= \sqrt{m_1 m_2}$ is the geometric average mass around the
863:    saddle point, and the Debye frequency $\omega_{\rm{D}}$ has
864:    been introduced as a cut-off in the summation over the Matsubara
865:    frequencies. 
866: The equation for the critical temperature then reads
867: \begin{equation}
868: \lambda^{-1} = \frac{m}{\pi} T_c \sum_{\omega_n \geq 0}^{\omega_{\rm{D}} 
869:    / (2\pi T_c )} \frac{1}{\omega_n} \ln \left(
870:      \frac{\omega_{\rm{D}}^2}
871:        {\omega_n^2 + z^2}\right).
872: \end{equation}
873: For small $z$ ($|z|\ll \omega_{\rm{D}}$), one recovers the well-known result
874:    \cite{Abrikosov:93}:
875: \begin{equation}
876: T_c \sim \frac{\omega_{\rm{D}}}{2\pi} \exp \left(
877:    -\frac{1}{\sqrt{\lambda\rho}} \right),
878: \label{eq:Tcsqrt}
879: \end{equation}
880: where $\rho = m/(2\pi^2 )$ denotes the DOS at the saddle point.
881: Eq.~(\ref{eq:Tcsqrt}) is in agreement with the $s$-wave result for
882:    $\Delta_0$, Eq.~(\ref{eq:deltamaxs}), when the assumption
883:    of a phonon-mediated pairing mechanism is made and the limit $r\to0$ is
884:    taken.
885: Vertex and cross corrections to Eq.~(\ref{eq:Tcsqrt}) in terms of the
886:    Migdal adiabaticity parameter $\omega_{\mathrm{D}} /
887:    E_{\mathrm{F}}$ have been shown to decrease the enhancement of
888:    $T_c$ due to the proximity to an ETT
889:    \cite{Cappelluti:96a,Cappelluti:96b}.
890: Analogously, for the two-particle Green function close to
891:    the superconducting transition and in the proximity of an ETT
892:    ($|z|\ll T \sim T_c$), one finds
893: \begin{equation}
894: L^{-1} (0,0;z,T) = - \rho\ln\left(
895:    \frac{\omega_{\rm{D}}}{2\pi T_c} \right) \frac{T-T_c}{T_c}.
896: \end{equation}
897: One observes that the presence of the ETT results in the appearance of
898:    the additional large factor $\sim\ln(\omega_{\rm{D}}/T_c)$ in front
899:    of the reduced temperature.
900: 
901: In order to determine the superconducting stiffness tensor $\eta_i
902:    (z)$, one is led to consider the $\bq$-dependence of the
903:    polarization operator, Eq.~(\ref{eq:pol2}).
904: Expanding $I(\bq,\omega_n,-\omega_n;z)$ in Eq.~(\ref{eq:pol2}) for
905:    small $q$ and $z$, up to quadratic order in $q$, one has
906: \begin{widetext}
907: \begin{equation}
908: I(\bq,\omega_n, -\omega_n ; z) = I_0 (0, \omega_n, -\omega_n;z) 
909: + I_1 (q^2, \omega_n, -\omega_n ;0) + I_2 (q^2, \omega_n, -\omega_n ;z),
910: \label{eq:pol4}
911: \end{equation}
912: \end{widetext}
913: where $I_j$ are defined in App.~\ref{app:conti2}, and $I_0$ has been used
914:    above for the definition of the critical temperature, Eq.~(\ref{eq:pol3}).
915: One eventually finds for the $q^2$-dependence of the two-particle
916:    Green function in the limit $|z|\ll T\sim T_c$ the relatively
917:    classical form:
918: \begin{equation}
919: L^{-1} (\bq,0;0,T) = - \rho \left[
920:    \ln \left(\frac{\omega_{\rm{D}}}{2\pi T_c} \right)
921:    \frac{T-T_c}{T_c} + \eta_1 q_1^2 + \eta_2 q_2^2
922:    \right],
923: \label{eq:L-1}
924: \end{equation}
925: where the components of the superfluid stiffness tensor are given by
926: \begin{equation}
927: \eta_i (z) = \frac{7\zeta(3)E_{\rm{F}}}{8\pi^2 T^2 m_i} ,
928: \end{equation}
929: to the lowest order in $z/E_{\rm{F}}$.
930: The latter expression is analogous to the result obtained
931:    in the standard 2D isotropic case, with $\xi_\bp = p^2 / (2m) -
932:    \mu$, where the superfluid stiffness reads $\eta = 7\zeta(3)
933:    E_{\rm{F}} / (16 \pi^2 T^2 m)$, and the DOS per spin is
934:    $\rho_{\mathrm{(2D)}} = m/(2\pi)$ (Ref.~\onlinecite{Varlamov:89}).
935: It is worth noting that in Eq.~(\ref{eq:L-1}) the effective mass of
936:    the fluctuating Cooper pair gets increased by a factor $\ln
937:    (\omega_{\rm{D}} /2\pi T_c )$, with respect to the case in which a
938:    parabolic spectrum is assumed.
939: This implies a reduced role of fluctuations near an ETT.
940: Indeed, our results demonstrate that the temperature range of the
941:    fluctuation regime is governed by essentially the same $\Gi$, while
942:    the propagator's effective mass is enhanced.
943:    
944: Summarizing, the results obtained above shows that a topological
945:    singularity in the electronic spectrum practically does not affect
946:    the Ginzburg-Landau stiffness, in contrast to what was
947:    intuitively speculated \cite{Varlamov:99}.
948: The reason thereof is that the value of $\eta$ is formed by gathering the
949:    contributions of the electronic states belonging to the whole Fermi
950:    surface, not only by the `slow' ones. 
951: Finally, we note that in the approach we followed, only the
952:    polarization loop Eq.~(\ref{eq:pol2}) is critical, since Cooper
953:    pairing of non-zero center-of-mass momentum is not taken into
954:    account \cite{AGD}.
955: 
956: \section{Conclusions}
957: \label{sec:conclusions}
958: 
959: We have reviewed the effects of an electronic topological transition
960:    on the superconducting properties of a 2D electron system, with an
961:    energy spectrum characterized by a minimum at the $\Gamma$ point
962:    and an extended, doping dependent saddle point at $(\pi,0)$, as is
963:    typical for most single-layered, hole-doped HTSC.
964: We analytically derived the expressions for the superconducting gap
965:    $\Delta_0$ at $T=0$ close to an ETT, both in the $s$-wave and in
966:    the $d$-wave case.
967: In contrast to the 3D result \cite{Makarov:65}, $\Delta_0$ turns out
968:    to be characterized by a nonmonotonic behavior as a function of the
969:    critical parameter $z$, with a maximum at $z\simeq0$, i.e. close to
970:    the ETT.
971: Due to the electron-hole symmetry-breaking induced by a non-zero value
972:    of the hopping ratio $r$, we find that the maximum of $\Delta_0$
973:    actually occurs at $z\gtrsim 0$, \emph{i.e.} it is slightly shifted
974:    towards the hole-like region, as observed in LSCO \cite{Ino:01}.
975: We point out that in previous calculations
976:    \cite{Tsuei:90,Pattnaik:92,Newns:92,Newns:92a,Onufrieva:00} the
977:    maximum of $T_c$ as a function of the critical parameter $z$ occurs 
978:    at the ETT, $z=0$.
979: This result has been often used as an objection against the relevance
980:    of the Van~Hove scenario for the cuprates, since ARPES data show
981:    that the optimal doping does not correspond to the critical point
982:    $z=0$, and that the FS preserves a hole-like character over the
983:    entire doping range for almost all hole-doped compounds (see,
984:    however, also the recent results of Ino \emph{et al.} for the LSCO
985:    compound \cite{Ino:01}).
986: On the contrary, we have shown that the observed difference between
987:    optimal doping and the doping actually corresponding to the ETT can 
988:    be justified by taking into account an electronic spectrum beyond
989:    the hyperbolic approximation.
990: Moreover, we find that the dependence of $\Delta_0$ on the hopping
991:    ratio $r$ is in good qualitative agreement with the
992:    phenomenological results collected by Pavarini \emph{et al.} for
993:    several HTSC materials \cite{Pavarini:01}, thus demonstrating the
994:    role of the band structure peculiarities and, in particular, of
995:    next-nearest neighbor hopping, in stabilizing high-temperature
996:    superconductivity in the cuprates.
997: 
998: In the presence of impurities, the Fermi line is
999:    effectively smeared, and one expects the anomalies due to the
1000:    proximity to an ETT to occur at a larger value of the critical
1001:    parameter, as soon as such a `blurred' Fermi line touches the zone
1002:    border.
1003: We also derived the energy dependence of the retarded quasiparticle
1004:    self-energy $\Sigma^{\rm{R}}$ due to impurity scattering in the
1005:    2D case, for a simplified, hyperbolic dispersion relation.
1006: In contrast to the 3D case, $\Sigma^{\rm{R}}$ is again
1007:    characterized by a nonmonotonic $z$-dependence, thus confirming
1008:    that the quasiparticle lifetime $\tau_\bk$ is generally an
1009:    anisotropic quantity over the 1BZ.
1010: 
1011: Finally, we addressed the issue of the range of fluctuations near $T_c$.
1012: By explicitly computing the two-particle propagator in the Cooper
1013:    channel near $T_c$, we derived the expression for the superfluid
1014:    stiffness $\eta$ close to an ETT.
1015: Although the Fermi velocity vanishes at the saddle point, we
1016:    find a non-zero value of $\eta$, in complete analogy with the
1017:    Ginzburg-Landau result for an isotropic electronic spectrum, thus
1018:    showing that all electronic states participate in establishing the
1019:    superconducting correlations.
1020: Moreover, our results show that the role of fluctuations even
1021:    diminishes near the ETT.
1022: Indeed, while their temperature range is determined by about the same
1023:    value of $\Gi$, the effective mass of the fluctuating Cooper pair
1024:    increases by a factor $\ln (\omega_{\mathrm{D}} / 2 \pi T_c )$ with 
1025:    respect to the case of a parabolic spectrum.
1026: Therefore, in the denominator of any fluctuation contribution there
1027:    appears a large logarithm, which implies a relative suppression of
1028:    fluctuation effects near the ETT.
1029: 
1030: \begin{acknowledgments}
1031: We thank J. V. Alvarez, G. Balestrino, B. Ginatempo, F. Onufrieva,
1032:    P. Pfeuty, P. Podio-Guidugli, R. Pucci, and G. Savona for
1033:    useful discussions.
1034: Partial support from MURST through COFIN fund 2001 and the
1035:    E.U. through the F.S.E. Program (G. G. N. A. and E. P.), and from
1036:    I.N.F.M. and I.S.I. (Torino) through the `Progetto Stage'
1037:    (G. G. N. A.) is gratefully acknowledged.
1038: G. G. N. A. also acknowledges the D.S.T.F.E. ``Tor Vergata'' and
1039:    P.~G. Nicosia for warm hospitality during the period in which the
1040:    present work was brought to completion. 
1041: \end{acknowledgments}
1042: 
1043: \appendix
1044: 
1045: \section{Evaluation of the pairing susceptibility in the
1046:    \lowercase{$d$}-wave case} 
1047: \label{app:conti1}
1048: 
1049: We briefly outline the derivation of the asymptotic dependence of the
1050:    integrated pairing susceptibility close to an ETT in 2D. 
1051: In the $s$-wave case, changing integration variables in the gap
1052:    equation, Eq.~(\ref{eq:BCST=0}), from wavevector $\bk$ to energy,
1053:    we can write the pairing susceptibility integrated between the ETT
1054:    and either band edges as
1055: \begin{equation}
1056: S_\pm  = - \int_0^1 \frac{\ln (a_\pm
1057:    \xi) \,d\xi}{\sqrt{(\xi \mp \zeta_\pm )^2 + \delta_\pm^2}} ,
1058: \label{eq:pairings}
1059: \end{equation}
1060: where we have introduced the dimensionless auxiliary quantities
1061:    $\zeta_\pm = z/(4t\pm 8t^\prime )$, $\delta_\pm = \Delta_0 /
1062:    (4t\pm 8t^\prime )$, and $a_\pm = (1\pm
1063:    2r)/(4\sqrt{2}\sqrt{1-4r^2})$.
1064: 
1065: In the $d$-wave case, the double integration over wavevector $\bk$
1066:    cannot be reduced to a simple integration over energy, and one has
1067:    to change variables to $\xi=\pm (\varepsilon_\bk
1068:    -\varepsilon_c)/[4t(1\pm 2r)]$, $g=g_\bk$.
1069: In such case, Eq.~(\ref{eq:BCST=0}) can be written as
1070: \begin{equation}
1071: \frac{\pi^2 t}{2\lambda} = D_+ + D_- ,
1072: \label{eq:gapeqd}
1073: \end{equation}
1074: where the integrated pairing susceptibility now reads:
1075: \begin{equation}
1076: D_\pm = \frac{1}{4} \int_0^1 d\xi
1077:    \int_0^{1-\xi} dg \frac{g^2}{\sqrt{(\xi\mp\zeta_\pm)^2 +
1078:    \delta_\pm^2 g^2}} \frac{1}{\sqrt{J_1 J_2 J_3}} ,
1079: \label{eq:pairingd}
1080: \end{equation}
1081: and
1082: \begin{subequations}
1083: \begin{eqnarray}
1084: J_1 &&= (1-2rg)^2 + 4r[g-r+(1\pm 2r)\xi],\\
1085: J_2 &&= (1+g)^2 - (1-\sqrt{J_1})^2/(4r^2),\\
1086: J_3 &&= (1-g)^2 - (1-\sqrt{J_1})^2/(4r^2).
1087: \end{eqnarray}
1088: \end{subequations}
1089: Eq.~(\ref{eq:pairingd}) leads to hyperelliptic integrals,
1090:    that cannot be generally expressed in terms of known special functions
1091:    \cite{Liu:97}.
1092: We next set $\Omega=-(1-\sqrt{J_1})/(2r)$, with $\Omega\to\xi$ as
1093:    $r\to0$, with which $D_+$ in Eq.~(\ref{eq:pairingd}) transforms
1094:    into
1095: \begin{widetext}
1096: \begin{equation}
1097: D_+ = \frac{1}{4} \int d\Omega \int dg \frac{g^2}{\sqrt{[r(v^2 - g^2) -
1098:    z]^2 + \Delta_0^2 g^2}}
1099:    \frac{1}{\sqrt{[(1+\Omega)^2 - g^2][(1-\Omega)^2 -g^2]}} ,
1100: \label{eq:hyper}
1101: \end{equation}
1102: \end{widetext}
1103: where all energies are in units of $4t$ and the integration is be to
1104:    performed over the curvilinear triangle 
1105:    defined by $g\geq0$, $g\leq 1-\Omega$, and $v^2 - g^2 \geq0$, with
1106:    $v^2 = 1+\Omega^2 + \Omega/r$.
1107: Such triangle represents the hole-like region of the 1BZ,
1108:    $\varepsilon_\bk \geq\varepsilon_c$, in these new coordinates.
1109: 
1110: In the limit $r\to0$, one branch of the hyperbola defined by $g=v$
1111:    reduces to the $\Omega=0$ axis, and $r (v^2 -g^2) \to\Omega$.
1112: A further change to polar coordinates as $\Omega=\rho\cos\theta$,
1113:    $g=\rho\sin\theta$ then allows to exactly decouple the two
1114:    integrations in Eq.~(\ref{eq:hyper}), the integration over $\rho$
1115:    leading to elliptic integrals.
1116: Extracting the logarithmic divergence of these latter near
1117:    $\theta=\pi/2$ (corresponding to the ETT, together with $\rho=1$)
1118:    yields the answer [see Eq.~(\ref{eq:continuation}) below, with
1119:    $r=0$]. 
1120: 
1121: In the case $r\neq0$, we could not find any such simple change of
1122:    variables, allowing to exactly decouple the integrations in
1123:    Eq.~(\ref{eq:hyper}).
1124: However, since only the behavior of the dispersion relation close to
1125:    the ETT is believed to determine the asymptotic properties of the
1126:    integrated pairing susceptibility, we may linearly expand $r (v^2
1127:    -g^2)$ near $\Omega=0$, $g=1$ and set $\Omega
1128:    +2r(g-1)=\rho\cos\theta$. 
1129: Within such approximation, Eq.~(\ref{eq:hyper}) then reads:
1130: \begin{widetext}
1131: \begin{equation}
1132: D_+ = \frac{1}{4} \int_0^{\pi/2} d\theta \frac{\sin^2
1133:    \theta}{\sqrt{\cos^2 \theta + \Delta_0^2 \sin^2 \theta}} 
1134:    \frac{|\alpha_- \beta_+ |}{1-4r^2}
1135:    \int_0^{\alpha_-} d\rho \frac{\rho^2}{\sqrt{(\alpha_+ +
1136:    \rho)(\alpha_- -\rho)(\beta_+ +\rho)(\beta_- -\rho)}} ,
1137: \label{eq:continuation}
1138: \end{equation}
1139: \end{widetext}
1140: where
1141: \begin{subequations}
1142: \begin{eqnarray}
1143: \alpha_\pm &&= \frac{1\pm 2r}{\cos\theta+(1-2r)\sin\theta} ,\\
1144: \beta_\pm &&=  \frac{1\pm 2r}{\cos\theta-(1+2r)\sin\theta} .
1145: \end{eqnarray}
1146: \end{subequations}
1147: Following standard methods (see, e.g., Ref.~\onlinecite{Hancock:10}),
1148:    the inner integral can be now expressed as a combination of
1149:    elliptic integrals, which diverge logarithmically as $\theta\to\pi/2$,
1150:    whence Eq.~(\ref{eq:deltamaxd}) follows.
1151: 
1152: \section{Evaluation of the polarization operator}
1153: \label{app:conti2}
1154: 
1155: The momentum integration in Eq.~(\ref{eq:pol2}) for the polarization
1156:    operator can be performed by reducing the integration domain to the
1157:    first quarter of the 1BZ, and dividing the latter into the two triangles
1158:    defined by $\{ \bp \,:\, p_2 \leq (m/m_1 ) p_1 ;\,\, p_1\geq0 \}$, and $\{
1159:    \bp \,:\, p_2 \geq (m/m_1 ) p_1 ;\,\, p_1 \geq0\}$, respectively.
1160: One finds:
1161: \begin{widetext}
1162: \begin{equation}
1163: \Pi(0,0;z,T) = T \sum_{\omega_n}I(0,\omega_n, -\omega_n;z) 
1164:  = \frac{2m}{\pi^2} T \sum_{\omega_n} \left[ f(\omega_n, -\omega_n ;
1165:    z) + f(\omega_n, -\omega_n ; -z) \right],
1166: \end{equation}
1167: where
1168: \begin{equation}
1169: f(\omega_n, -\omega_n; z) =  
1170: \int_0^\infty dx_1 \int_0^x dx_2 \frac{1}{x_2^2 - x_1^2
1171:    -i\omega_n + z} \cdot \frac{1}{x_2^2 - x_1^2
1172:    +i\omega_n + z} .
1173: \end{equation}
1174: Performing the inner integration, one obtains
1175: \begin{equation}
1176: f(\omega_n, -\omega_n ;z) = \frac{1}{2i\omega_n}
1177: \int_0^\infty dx \left[ \frac{\ln(x-i\sqrt{z-i\omega_n-x^2})}
1178: {\sqrt{z-i\omega_n -x^2}}-\frac{\ln(x+i\sqrt{z-i\omega_n-x^2})}
1179: {\sqrt{z-i\omega_n -x^2}}- \rm{H.c.} \right] .
1180: \label{a1}
1181: \end{equation}
1182: \end{widetext}
1183: The last integration can be carried out observing that 
1184: \begin{equation}
1185: \frac{\ln (x \pm i \sqrt{z-i\omega_n -x^2} )}{\sqrt{z-i\omega_n -x^2}}
1186:    = \pm \frac{i}{2} \frac{d}{dx} \ln^2 ( x \pm i \sqrt{z-i\omega_n
1187:    -x^2} ),
1188: \end{equation}
1189: and choosing the branches of the logarithms in order to make them
1190:    complex conjugated of each other.
1191: Finally, we obtain:
1192: \begin{equation}
1193: I(0,\omega_n, -\omega_n; z)= \frac{m}{2 \pi}\frac{1}{|\omega_n|} \ln
1194: \frac{{\omega_{\rm{D}}^2}}{z^2 + \omega_n^2} .
1195: \end{equation}
1196: 
1197: In order to calculate the GL stiffness, let us now expand the
1198:    polarization operator up to quadratic order in $q$.
1199: One has:
1200: \begin{equation}
1201: \Pi (q,0;z,T)= \Pi (0,0;z,T) + \rho \sum_{i=1,2} \eta_i q_i^2 ,
1202: \end{equation}
1203: where the components $\eta_i$ of the stiffness tensor are defined by
1204: \begin{widetext}
1205: \begin{equation}
1206: \rho \sum_{i=1,2} \eta_i q_i^2
1207: =T\sum_{\omega_n}[I_1(q^2,\omega_n,-\omega_n;0) + I_2
1208: (q^2,\omega_n,-\omega_n ;z) + \ldots].
1209: \end{equation}
1210: One explicitly finds:
1211: \begin{equation}
1212: \eta_i (z) = \frac{1}{2m_i} T \sum_{\omega_n} \int 
1213: \frac{d^2 \bx }{x_2^2 - x_1^2 -i\omega_n}
1214: \cdot \frac{x_i^2}{(x_2^2 - x_1^2 + i\omega_n )^3} 
1215: - \frac{z}{2m_i} T \sum_{\omega_n} \int d^2 \bx
1216: \sum_{k=0}^1 \frac{3^k}{(x_2^2 -x_1^2 + i\omega_n )^{3+k}} \cdot
1217:    \frac{x_i^2}{(x_2^2 -x_1^2 -i\omega_n )^{2-k}} 
1218: \end{equation}
1219: to the first order in $z/E_{\rm{F}}$.
1220: The first integral, giving the principal contribution to the
1221:    stiffness, can be evaluated using polar coordinates as:
1222: \begin{eqnarray}
1223: \eta_i (z=0) &&= \frac{T}{m_i} \sum_{\omega_n} \frac{1}{\omega_n^2}
1224: \int_0^{2\pi} d\phi \int_0^\infty dr \frac{r^3 (r^4 \cos^2 2\phi -1)
1225:    \cos^2 \phi}{(r^4 \cos^2 2\phi +1)^3} \nonumber \\
1226: &&= - \frac{\pi T}{4m_i} \sum_{\omega_n} \frac{1}{\omega_n^2}
1227: \int_0^\infty dt \left[ \frac{\partial}{\partial\alpha} +
1228:    \frac{\partial^2}{\partial \alpha^2} \right]
1229:    \left. \frac{1}{\sqrt{\alpha} \sqrt{t+\alpha}} \right|_{\alpha=1}
1230:    \nonumber \\
1231: &&\sim \frac{\pi T}{m_i} \sum_{n\geq 0} \frac{1}{\omega_n^3} E_{\rm{F}} =
1232: \frac{7\zeta(3)E_{\rm{F}}}{8\pi^2 T^2 m_i} .
1233: \end{eqnarray}
1234: \end{widetext}
1235: 
1236: \bibliographystyle{apsrev}
1237: \bibliography{a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z,zzproceedings,Angilella,notes}
1238: 
1239: \newpage
1240: 
1241: 
1242: \newpage
1243: 
1244: 
1245: \newpage
1246: 
1247: 
1248: \newpage
1249: 
1250: 
1251: 
1252: \end{document}
1253: