cond-mat0401258/Ylm.tex
1: % INJ
2: 
3: \documentclass{elsart}
4: 
5: \usepackage[dvips]{graphicx}
6: \usepackage{amsmath}
7: 
8: 
9: \newcommand{\bk}{{\bf k}}
10: \newcommand{\bq}{{\bf q}}
11: \newcommand{\bx}{{\bf x}}
12: \newcommand{\br}{{\bf r}}
13: \newcommand{\eF}{\varepsilon_{\mathrm{F}}}
14: \newcommand{\vF}{v_{\mathrm{F}}}
15: \newcommand{\kB}{k_{\mathrm{B}}}
16: \newcommand{\kF}{k_{\mathrm{F}}}
17: \let\onlinecite\cite
18: 
19: %\newcommand\modified[1]{{\bf #1}}
20: \newcommand\modified[1]{{\relax #1}}
21: \newcommand\marginmodified[1]{{\marginpar{
22: \framebox{
23: \begin{minipage}[c]{1.3truecm}
24: {\bf #1}
25: \end{minipage}
26: }}}}
27: 
28: \begin{document}
29: 
30: \runauthor{Angilella, Leys, March, and Pucci}
31: \runtitle{Correlation between characteristic energies ...}
32: \journal{Phys. Lett. A (accepted)}
33: \received{\today}
34: 
35: \begin{frontmatter}
36: \title{Correlation between characteristic energies
37:    in non--\lowercase{$s$}-wave pairing
38:    superconductors}
39: 
40: \author[CT]{G. G. N. Angilella}
41: \author[RUCA]{F. E. Leys}
42: \author[Oxford,RUCA]{N. H. March}
43: \and
44: \author[CT]{R. Pucci}
45: \address[CT]{Dipartimento di Fisica e Astronomia, Universit\`a di
46:    Catania, and Istituto Nazionale per la Fisica della Materia, UdR
47:    di Catania,\\ Via S. Sofia, 64, I-95123 Catania, Italy}
48: \address[RUCA]{Department of Physics, University of Antwerp (RUCA),\\
49:    Groenenborgerlaan 171, B-2020 Antwerp, Belgium}
50: \address[Oxford]{Oxford University, Oxford, England}
51: 
52: \date{\today}
53: 
54: \begin{abstract}
55: By solution of the Bethe-Goldstone equation for the Cooper pairing
56:    problem, an approximate analytic relation is derived between
57:    coherence length $\xi$ and the binding energy of the Cooper pair.
58: This relation is then qualitatively confirmed by numerically
59:    solving the corresponding self-consistent gap equations, following
60:    the crossover 
61:    from weak to strong coupling, in non--$s$-wave superconductors.
62: The relation applies to non-conventional superconductors, and in
63:    particular to heavy Fermions and to high-$T_c$ cuprates.
64: Utilizing in addition a phenomenological link between $\kB T_c$ and a
65:    characteristic energy $\varepsilon_c = \hbar^2 / 2m^\ast \xi^2$, with
66:    $m^\ast$ the effective mass, major differences are exposed in the
67:    functional relation between $\kB T_c$ and $\varepsilon_c$ for $s$-wave
68:    materials and for non-conventional superconductors.
69: The relation between critical temperature and $\varepsilon_c$ thereby
70:    proposed correctly reflects the qualitative properties of heavy
71:    Fermion superconductors.
72: \begin{keyword}
73: \PACS
74: 74.20.-z,
75: % Theories and models of superconducting state,
76: 74.72.-h, 
77: % High-$T_c$ compounds,
78: 74.70.Tx 
79: % Heavy-fermion superconductors.
80: \end{keyword}
81: \end{abstract}
82: %\begin{keyword}
83: %\end{keyword}
84: \end{frontmatter}
85: 
86: \newpage
87: 
88: \section{Introduction}
89: \label{sec:intro}
90: 
91: In earlier work \cite{Angilella:00b,Angilella:01a}, we have discussed
92:    the possible correlation between the critical temperature $T_c$ in
93:    anisotropic superconductors and a `natural' energy scale
94:    $\varepsilon_c \sim \hbar^2 / m^\ast l_c^2$ involving the effective
95:    mass $m^\ast$ and some characteristic length $l_c$.
96: Here, by anisotropic superconductors we mean a superconductor
97:    characterized by non-spherically symmetric pairing, giving rise to
98:    an anisotropic $\bk$-dependence of the gap energy in momentum
99:    space.
100: Well-known instances of such materials are the heavy Fermion
101:    compounds, most of them being characterized by a $p$-wave order
102:    parameter \cite{Sigrist:91,Heffner:95}, and the high-$T_c$,
103:    cuprates, whose order parameter displays $d$-wave symmetry
104:    \cite{Annett:90a}.
105: The most natural physical choice for the characteristic length $l_c$
106:    entering the definition of $\varepsilon_c$ above was the coherence
107:    length $\xi$ \cite{Angilella:00b}.
108: In the case of anisotropic superconducting materials in the presence
109:    of magnetic fluctuations, we later correlated the coherence length
110:    $\xi$ to the spin-fluctuation temperature $T_{\mathrm{sf}}$ as $\kB
111:    T_{\mathrm{sf}} \sim \hbar^2 / m^\ast \xi^2$ \cite{Angilella:01a}.
112: 
113: The possible enhancement of the critical temperature $T_c$ in an
114:    anisotropic superconductor was already studied in the early work of
115:    Markowitz and Kadanoff \cite{Markowitz:63} within a weak-coupling,
116:    BCS-like approach, and more recently revived in
117:    Refs.~\onlinecite{Whitmore:84,Beal-Monod:95,Valls:95} as possibly
118:    relevant for the high-$T_c$ cuprates.
119: A generalization of Markowitz and Kadanoff's results to the strong and
120:    intermediate coupling regime however showed that in the strong
121:    coupling limit anisotropy is effectively averaged out, and the gap
122:    tends to become isotropic \cite{Combescot:91}.
123: 
124: In the case of the high-$T_c$ cuprates, the strong coupling limit
125:    corresponds to the underdoped region of their phase diagram, which
126:    is usually interpreted in terms of a crossover between
127:    Bose-Einstein condensation of strongly coupled preformed pairs
128:    above $T_c$, and weak coupling, BCS-like superconductivity in the
129:    overdoped regime \cite{Randeria:95}.
130: On the other hand, the coherence length $\xi$ can be continuously
131:    connected to the characteristic size of the preformed bosonic pairs
132:    in the normal state of (underdoped) high-$T_c$ cuprates.
133: Therefore, the coherence length may serve to parametrize the crossover
134:    from weak to strong coupling, with $\kF \xi$ decreasing in going
135:    from weak coupling, characterized by large superconducting pair
136:    fluctuations, to strong coupling, $\kF$ being the Fermi momentum
137:    \cite{Pistolesi:94} (see also Refs.~\onlinecite{Andrenacci:00,Perali:02}, and
138:    refs. therein).
139: In the underdoped regime, a relatively short coherence
140:    length is consistent with the idea of preformed pairs localized in
141:    real space \cite{Nozieres:85,Egorov:94,March:94,Timusk:99}.
142: In such a limit, it is interesting to investigate how the internal
143:    structure of these preformed pairs is related to the overall
144:    symmetry of the many-body order parameter \cite{Andrenacci:03}.
145: 
146: Here, we derive an explicit expression for the
147:    coherence length in terms of the pair binding energy for the Cooper
148:    problem in anisotropic superconductors.
149: Such an expression [see Eq.~(\ref{eq:xiellneq0}) below] explicitly
150:    contains the quantum number $\ell$ of the pair relative angular
151:    momentum, which is usually employed to parametrize the anisotropic
152:    character of the order parameter, $\ell=0,1,2$ corresponding to
153:    $s$-, $p$-, and $d$-wave symmetry, respectively.
154: While this expression correctly reduces to the standard one for
155:    isotropic $s$-wave superconductors, in the case $\ell>0$ it agrees
156:    qualitatively with the phenomenological dependence of $\kB
157:    T_c$ on the characteristic energy $\varepsilon_c$, proposed in
158:    Ref.~\onlinecite{Angilella:00b} for the heavy Fermion compounds as
159:    well as for the high-$T_c$ cuprates.
160: Such a dependence of the characteristic energy for superconductivity
161:    on $\varepsilon_c$ is 
162:    then qualitatively confirmed by numerically solving the
163:    self-consistent gap equations for the maximum gap at $T=0$ in the
164:    case of an anisotropic superconductor in the crossover between the
165:    weak- and strong-coupling limits, as a function of the
166:    dimensionless crossover parameter $\kF \xi$, now employing a more
167:    general definition of the coherence length $\xi$ for anisotropic
168:    superconductors.
169: 
170: The paper is organized as follows.
171: In Sec.~\ref{sec:model}, after briefly reviewing the Bethe-Goldstone
172:    equation for the Cooper problem in the case of isotropic
173:    superconductors, we generalize and solve it for non--$s$-wave
174:    superconductors, characterized by a superconducting instability in
175:    the $(\ell,m)$ channel of relative angular momentum quantum
176:    numbers.
177: Our results are discussed in relation to the
178:    earlier phenomenological findings in
179:    Ref.~\onlinecite{Angilella:00b}.
180: In Sec.~\ref{sec:crossover} we analyze the gap equations arising from
181:    the corresponding many-body problem at the mean-field level.
182: Although we now have to resort to numerical integration (at least in
183:    the non--$s$-wave case), our results qualitatively confirm the
184:    approximate relation between characteristic energies derived in
185:    Sec.~\ref{sec:model}, also in the crossover from weak to
186:    strong coupling.
187: In Sec.~\ref{sec:conclusions}
188:    we eventually summarize and propose some directions for future work.
189: 
190: \section{Bethe-Goldstone equation for non--\lowercase{$s$}-wave superconductors}
191: \label{sec:model}
192: 
193: The Bethe-Goldstone equation for the Cooper
194:    problem \cite{Bethe:57} in momentum space reads:
195: \begin{equation}
196: (\varepsilon - 2\xi_\bk )\psi_\bk = \sum_{\bk^\prime}
197:    V_{\bk\bk^\prime} \psi_{\bk^\prime} ,
198: \label{eq:BG}
199: \end{equation}
200: where $\psi_\bk$ is the Fourier transform of the pair wave-function
201:    with wave-vector $\bk$ (here, we assume a spin-singlet state, with
202:    zero total momentum), $\xi_\bk = \hbar^2 k^2 / (2m^\ast ) - \mu$ is the
203:    energy of a single electron with respect to the chemical
204:    potential $\mu$,
205:    $m^\ast$ is the effective mass,
206:    $\varepsilon$ is the binding energy of the electron pair, and
207:    $V_{\bk\bk^\prime}$ is the Fourier transform of the
208:    electron-electron interaction (see also Ref.~\onlinecite{Ketterson:99}
209:    for a pedagogical review).
210: In the weak-coupling limit, we may safely assume $\mu=\eF$
211:    at $T=0$, with $\eF$ the Fermi energy.
212: However, we anticipate that this identification will be superseded in
213:    Sec.~\ref{sec:crossover}, where the crossover from weak to strong
214:    coupling will be addressed more in detail.
215: In the case of anisotropic superconductors, we adopt the spirit of the
216:    Anderson-Brinkman-Morel model for $p$-wave superfluidity in $^3$He
217:    \cite{Anderson:61}, and expand the electron-electron interaction in
218:    spherical harmonics around the Fermi surface as
219: \begin{equation}
220: V_{\bk\bk^\prime} = 
221:  -\frac{1}{\Omega} \sum_{\ell=0}^\infty \sum_{m=-\ell}^\ell V_{\ell
222:    m} Y_{\ell m} (\hat{\bk}) Y_{\ell m} (\hat{\bk}^\prime ),
223: \label{eq:Yexpansion}
224: \end{equation}
225: for $|\xi_\bk |$, $|\xi_{\bk^\prime }| < \Lambda$, and
226:    zero otherwise, 
227:    where $\hat{\bk}$ is the unit vector pointing along the direction
228:    of $\bk$, $\Omega$ denotes the volume of the system, and
229:    $\Lambda$ is an energy cut-off, characterized by the nature of
230:    the interaction. 
231: In the case of conventional superconductivity ($s$-wave pairing, or
232:    $\ell=0$), it would be natural to identify such an energy scale with
233:    the Debye energy, as in BCS theory.
234: 
235: The use of spherical harmonics to expand the electron-electron
236:    interaction in anisotropic pairing superconductors characterized by
237:    a spherical Fermi surface was earlier considered by Markowitz and
238:    Kadanoff \cite{Markowitz:63} in the weak-coupling regime (see also
239:    Refs.~\onlinecite{Beal-Monod:95,Valls:95}), and later by Combescot
240:    \cite{Combescot:91} in the strong and intermediate coupling regime.
241: A related model of exotic Cooper pairing with finite angular momentum
242:    has been discussed also in Ref.~\onlinecite{Quintanilla:01}, where
243:    rotational symmetry breaking ($\ell>0$) is due to the interplay
244:    between the finite range of the attractive potential and the
245:    interelectronic average distance.
246: Within such model, the authors of Ref.~\onlinecite{Quintanilla:01}
247:    also derive the $\ell$-dependence of the critical temperature in
248:    the Bose-Einstein limit.
249: In the more realistic case of a non-spherical Fermi surface, spherical
250:    harmonics are naturally replaced by Allen's Fermi surface harmonics
251:    \cite{Allen:76}, which have been used by Whitmore \emph{et al.}
252:    \cite{Whitmore:84} in extending the results of
253:    Ref.~\onlinecite{Markowitz:63} within the framework of Eliashberg
254:    equations for $T_c$.
255: Moreover, spherical harmonics afford a natural classification of
256:    anisotropic superconductors (such as heavy Fermion compounds
257:    \cite{Sigrist:91,Heffner:95} as well as high-$T_c$ cuprates
258:    \cite{Annett:90a}) in terms of their pairing symmetry.
259: 
260: Within the weak-coupling approximation, the largest attractive
261:    coupling constant $V_{\ell m} \equiv V > 0$ in
262:    Eq.~(\ref{eq:Yexpansion}) gives rise to a pairing
263:    instability in the $(\ell,m)$ channel.
264: In the following, we shall neglect other instabilities, which may
265:    arise well below $T_c$, corresponding to mixed symmetry pairing.
266: This amounts to retaining the $(\ell,m)$ term only in the expansion
267:    Eq.~(\ref{eq:Yexpansion}).
268: In this case, Eq.~(\ref{eq:BG}) has a solution given by
269: \begin{equation}
270: \psi_\bk = \alpha \frac{Y_{\ell m} (\hat{\bk})}{\varepsilon - 2 \xi_\bk}
271: \label{eq:BGsol}
272: \end{equation}
273: belonging to the eigenvalue
274: \begin{equation}
275: \varepsilon \simeq -2\Lambda \exp\left( - \frac{2}{VN(0)} \right),
276: \end{equation}
277: where $N(0)$ is the density of states at the Fermi level, and $\alpha$
278:    is a normalization constant.
279: Such a solution corresponds to a bound state ($\varepsilon<0$), and a
280:    further mean-field analysis of the many-electron problem (see also
281:    Sec.~\ref{sec:crossover}) shows that
282:    it indeed corresponds to a superconducting state, characterized by
283:    a gap function at $T=0$:
284: \begin{equation}
285: \Delta_{\hat{\bk}} = 2\Lambda\Gamma e^{-1/N(0)V} Y_{\ell m} (\hat{\bk}),
286: \end{equation}
287: having the symmetry of the attractive channel under consideration,
288:    where:
289: \begin{equation}
290: \ln\Gamma = -\int d\Omega_\bk |Y_{\ell m} (\hat{\bk})|^2 \ln |Y_{\ell m}
291:    (\hat{\bk})|,
292: \end{equation}
293: the integration being carried over the unit sphere \cite{March:67}.
294: The anisotropic $\bk$-dependence of the pair wave-function $\psi_\bk$
295:    in Eq.~(\ref{eq:BGsol}) in the case $\ell>0$ provides interesting
296:    information on the internal structure of a Cooper pair and its
297:    connection with the overall symmetry of the many-body gap function.
298: 
299: Given the pair wave-function $\psi_\bk$, the coherence length $\xi$ is
300:    naturally defined by
301: \begin{equation}
302: \xi^2 = \frac{\sum_\bk | \nabla_\bk \psi_\bk |^2}{\sum_\bk |\psi_\bk
303:    |^2} .
304: \label{eq:xidef}
305: \end{equation}
306: Passing to the continuous limit, in the isotropic, $s$-wave case
307:    ($\ell=0$), one obtains \cite{Ketterson:99}
308: \begin{equation}
309: \xi^2 = \frac{2\hbar^2}{3m^\ast \eF} \frac{1}{x^2},
310: \label{eq:xi0}
311: \end{equation}
312: where $x=|\varepsilon|/(2\eF)$ measures the binding energy of a pair in
313:    units of the energy of an unbound pair.
314: Apart from a numerical factor, from Eq.~(\ref{eq:xi0}) one thus recovers the
315:    correct order of magnitude relation between the critical
316:    temperature, the Fermi velocity $\vF = \hbar^{-1} d\xi_\bk /dk$,
317:    and the coherence length: 
318: \begin{equation}
319: \kB T_c \sim |\varepsilon| \sim \frac{\hbar \vF}{\xi} .
320: \label{eq:Tcxi0}
321: \end{equation}
322: 
323: In the anisotropic case, for a pairing instability in the $(\ell,m)$
324:    channel, with the pair wave-function given by Eq.~(\ref{eq:BGsol}),
325:    the denominator in Eq.~(\ref{eq:xidef}) is easily integrated in the
326:    continuous limit as
327: \begin{eqnarray}
328: \sum_\bk |\psi_\bk |^2  &\to& \Omega \int d^3 \bk |\psi_\bk |^2
329:   \nonumber \\
330: &&=
331:    \frac{\Omega \alpha^2}{4\pi} \int_0^\infty dE
332:    \frac{N(E)}{(\varepsilon -2E)^2} \int d\Omega_\bk |Y_{\ell m}
333:    (\hat{\bk})|^2 \nonumber\\
334: &&\approx - \Omega \alpha^2
335:    \frac{N(0)}{8\pi\varepsilon} ,
336: \label{eq:numerator}
337: \end{eqnarray}
338: where $N(E)/4\pi$ is the density of states per unit solid angle, and
339:    use has been made of the normalization condition of the spherical
340:    harmonics.
341: In the case $\ell\neq0$, the anisotropic $\bk$ dependence of the pair
342:    wave-function Eq.~(\ref{eq:BGsol}) gives rise to two contributions
343:    in the numerator of Eq.~(\ref{eq:xidef}), according to the chain rule
344: \begin{equation}
345: |\nabla_\bk \psi_\bk |^2 = \frac{\alpha^2}{k^2}
346:    \frac{|\nabla_{\hat{\bk}} Y_{\ell m}
347:    (\hat{\bk})|^2}{(\varepsilon-2\xi_\bk )^2} + 4 \alpha^2 \hbar^2
348:    \vF^2 \frac{| Y_{\ell m} (\hat{\bk})|^2}{(\varepsilon-2\xi_\bk )^4}.
349: \end{equation}
350: Here, we have made use of the decomposition $\nabla_\bk = k^{-1}
351:    \nabla_{\hat{\bk}} + \hat{\bk} \partial/\partial k$, where
352:    $\nabla_{\hat{\bk}}$ denotes the angular part of the gradient
353:    operator in momentum space, and of the fact that
354:    $\xi_\bk$ depends only on $k=|\bk|$.
355: Passing to the continuous limit, integrating separately over the
356:    angles as in Eq.~(\ref{eq:numerator}), and making use of the
357:    identity
358: \begin{eqnarray}
359: \int d\Omega_\bk |\nabla_{\hat{\bk}} Y_{\ell m} (\hat{\bk})|^2 &=& -
360:    \int d\Omega_\bk Y_{\ell m}^\ast (\hat{\bk}) \nabla_{\hat{\bk}}^2
361:    Y_{\ell m} (\hat{\bk}) \nonumber\\
362: &&= \ell(\ell+1),
363: \end{eqnarray}
364: which follows from a variant of the Green's formula over the unit
365:    sphere, one eventually obtains
366: \begin{equation}
367: \xi^2 = \frac{\hbar^2}{2m^\ast \eF} \left[ \frac{4}{3} \frac{1}{x^2} + \frac{\ell
368:    (\ell+1)}{1+x} \left( 1 - \frac{x\ln x}{1+x} \right)
369:    \right].
370: \label{eq:xiellneq0}
371: \end{equation}
372: Equation~(\ref{eq:xiellneq0}) implicitly relates the binding energy
373:    $|\varepsilon|$ of a Cooper pair to its characteristic size $\xi$
374:    for anisotropic pairing superconductors.
375: 
376: In the case of non--$s$-wave superconductors,
377:    Equation~(\ref{eq:xiellneq0}) manifestly includes an `orbital'
378:    contribution, proportional to $\ell(\ell+1)$, as earlier surmised
379:    in Ref.~\cite{Angilella:00b} on the basis of phenomenological
380:    considerations.
381: There, we proposed the existence of
382:    a phenomenological relation linking $\kB T_c$ to the characteristic
383:    energy 
384: \begin{equation}
385: \varepsilon_c = \frac{\hbar^2}{2 m^\ast \xi^2 },
386: \label{eq:00b}
387: \end{equation}
388: in the case of anisotropic superconductors. 
389: That the effective mass $m^\ast$ should enter inversely in determining
390:    the scale of $\kB T_c$ was earlier recognized by Uemura \emph{et
391:    al.} \cite{Uemura:91}, who did not, however, include the coherence
392:    length in their analysis \cite{Pistolesi:94}.
393: By comparing the experimental values for several heavy Fermion
394:    compounds ($p$-wave superconductors, $\ell=1$) as well as
395:    high-$T_c$ superconductors ($d$-wave superconductors, $\ell=2$), in
396:    the anisotropic case we
397:    found that $\kB T_c = f(\varepsilon_c)$ deviates from the square-root
398:    behavior, $\kB T_c \propto \sqrt{\varepsilon_c}$, that is easily
399:    derived from Eqs.~(\ref{eq:Tcxi0}) and (\ref{eq:00b}) for isotropic, $s$-wave
400:    superconductors. 
401: In particular, in the heavy Fermion case, we noticed a large initial
402:    slope in $f(\varepsilon_c )$ for $\varepsilon_c =0$, and a tendency of
403:    such function to approach saturation, as $\varepsilon_c$ increases.
404: 
405: In view of the fact that $\kB T_c \sim |\varepsilon| \sim x \eF$
406:    \cite{Ketterson:99}, 
407:    Eq.~(\ref{eq:xiellneq0}) implicitly defines the generalization to
408:    the anisotropic case of the relationship between $\kB T_c$ and
409:    $\varepsilon_c$ we were looking for.
410: Equation~(\ref{eq:xiellneq0}) correctly reduces to Eq.~(\ref{eq:xi0}) in the
411:    isotropic case ($s$-wave superconductors, $\ell=0$).
412: In the case of non--$s$-wave superconductors ($\ell>0$),
413:    Eq.~(\ref{eq:xiellneq0}) indeed increases with increasing $\varepsilon_c$,
414:    starting with a logarithmically infinite slope at $\varepsilon_c =
415:    0$, and tending to a saturation value as $\varepsilon_c \to\infty$.
416: Such a limit is \emph{e.g.} approached for $\kF \xi\ll 1$.
417: Setting $a=4/[3\ell(\ell+1)]$ and performing an asymptotic analysis of
418:    Eq.~(\ref{eq:xiellneq0}), in the limit $\kF \xi\ll 1$ ($\ell>0$) one
419:    obtains
420: \begin{equation}
421: x = \frac{1+a}{W [(1+a)/e]},
422: \label{eq:Lambert}
423: \end{equation}
424: where $W(z)$ is Lambert's function \cite{Lambert}, and in the limit of
425:    very large anisotropy ($\ell\gg1$): 
426: \begin{equation}
427: \lim_{\ell\to\infty} x = W^{-1} (e^{-1} ) = 3.591\ldots .
428: \label{eq:Lambertinf}
429: \end{equation}
430: Figure~\ref{fig:multil} shows the dependence of the dimensionless
431:    measure of the
432:    pair binding energy $x=|\varepsilon|/2\eF$ on $\varepsilon_c / \eF
433:    = (\kF \xi)^{-2}$,
434:    implicitly defined by
435:    Eq.~(\ref{eq:xiellneq0}) in the cases $\ell=0,1,2$, corresponding
436:    to $s$-, $p$-, and $d$-wave superconductivity, respectively.
437: Keeping fixed all other variables, one notices that for $\xi>\xi_0$,
438:    where
439: \begin{equation}
440: \frac{\hbar^2}{2m^\ast \xi_0^2 \eF} = \frac{3}{4W^2 (e^{-1})} = 9.672\ldots,
441: \label{eq:Lambertxi0}
442: \end{equation} 
443: $|\varepsilon|$ increases as $\ell$ increases while,
444:    for $\xi<\xi_0$, $|\varepsilon|$ decreases as $\ell$ increases, although
445:    without large deviations from the limiting value,
446:    Eq.~(\ref{eq:Lambertinf}).
447: Therefore, at least for sufficiently weakly coupled superconductors
448:    ($\xi>\xi_0$), anisotropy enhances superconductivity, the 
449:    degree of anisotropy being here parametrized by the order $\ell$ of
450:    the spherical harmonic modelling the $\bk$ dependence of the order
451:    parameter.
452: 
453: \section{Mean-field analysis of the many-body problem and the
454:    crossover from weak to strong coupling}
455: \label{sec:crossover}
456: 
457: 
458: The definition of the Cooper pair size $\xi$ we employ in
459:    Eq.~(\ref{eq:xidef}) makes use of the pair wave-function $\psi_\bk$
460:    for the Cooper problem, in which the only many-body effect is that
461:    of forbidding the occupancy of states below the Fermi level, $\eF$.
462: Within BCS theory, \emph{i.e.} at the mean field level, a
463:    self-consistent treatment of the superconducting instability
464:    affords a better definition of the pair wave-function, which in
465:    momentum space is given by \cite{Ketterson:99} $\psi_\bk =
466:    \Delta_\bk /(2 E_\bk )$, where $\Delta_\bk$ is the gap function, now
467:    depending on $\bk$ as a vector, and $E_\bk = (\xi_\bk^2 +
468:    |\Delta_\bk |^2 )^{1/2}$ the upper branch of the quasiparticle
469:    dispersion relation.
470: With this definition of $\psi_\bk$, when $E_\bk$ is allowed to vanish
471:    locally on the Fermi surface, as is the case for $p$- and $d$-wave
472:    superconductors, it has been recently shown that nodal
473:    quasiparticles give rise to a logarithmically divergent
474:    contribution to $\xi$, as defined by Eq.~(\ref{eq:xidef})
475:    \cite{Benfatto:02}.
476: While such a drawback does not arise in $s$-wave superconductors
477:    \cite{Marini:98}, in the case of anisotropic superconductors the
478:    coherence length must be defined in terms of the range
479:    in real space of
480:    the static correlation function for the modulus of the
481:    order parameter \cite{Pistolesi:96,Benfatto:02}.
482: The fact that our result Eq.~(\ref{eq:xiellneq0}) provides a finite
483:    estimate for $\xi$ also in the anisotropic case clearly depends on
484:    our approximate choice for $\psi_\bk$, which solves only the
485:    two-body Cooper problem.
486: 
487: In order to
488:    discuss the limits 
489:    of validity of the approximate results 
490:    derived in Sec.~\ref{sec:model}, we will then work out numerically
491:    the mean-field solution to the corresponding BCS Hamiltonian, both
492:    in the weak and in the strong-coupling limit.
493: We start by reviewing the model and notations set out in
494:    Refs.~\cite{Pistolesi:94,Pistolesi:96,Marini:98,Benfatto:02} (see also
495:    Ref.~\cite{Babaev:01}).
496: 
497: We shall consider the following Hamiltonian for a superconducting
498:    system in three dimensions \cite{note:static}:
499: \begin{equation}
500: H = \sum_{\bk\sigma} \xi_\bk c^\dag_{\bk\sigma} c_{\bk\sigma} +
501:    \sum_{\bk\bk^\prime \bq} V_{\bk\bk^\prime} c^\dag_{\bk\uparrow}
502:    c^\dag_{-\bk+\bq\downarrow} c_{-\bk^\prime+\bq\downarrow}
503:    c_{\bk^\prime\uparrow} , 
504: \label{eq:hamiltonian}
505: \end{equation}
506: where $c^\dag_{\bk\sigma}$ [$c_{\bk\sigma}$] is the creation
507:    [destruction] operator for an electron with wave-vector $\bk$ and
508:    spin projection $\sigma\in\{\uparrow,\downarrow\}$ along a
509:    specified direction, $\xi_\bk = \hbar^2 k^2 /(2m^\ast ) -\mu$ is
510:    the dispersion relation for free electrons with effective mass
511:    $m^\ast$, measured with respect to the chemical potential $\mu$,
512:    and
513: \begin{equation}
514: V_{\bk\bk^\prime } = -\frac{V}{\Omega}
515:  Y_{\ell m} (\hat{\bk}) Y_{\ell m} (\hat{\bk}^\prime )
516: \label{eq:interY}
517: \end{equation}
518: is the projection of the electron-electron interaction along the
519:    $(\ell,m)$ channel, as in
520:    Eq.~(\ref{eq:Yexpansion}), which we assume
521:    to be attractive ($V>0$) \cite{note:Coulomb}.
522: Equation~(\ref{eq:interY}) generalizes to the non--$s$-wave case the
523:    contact potential in real space discussed \emph{e.g.} by Marini
524:    \emph{et al.} \cite{Marini:98}.
525: Standard diagonalization techniques then lead to the mean-field
526:    coupled equations for the gap energy $\Delta_\bk = \Delta_0 Y_{\ell
527:    m} (\hat{\bk})$ and the particle density $n$ at $T=0$ (see
528:    Ref.~\cite{Babaev:01} for the case $T\neq0$):
529: \begin{subequations}
530: \begin{eqnarray}
531: \label{eq:BCSa}
532: \frac{1}{V} &=& \frac{1}{\Omega} \sum_\bk \frac{|Y_{\ell m}
533:    (\hat{\bk})|^2}{2E_\bk} ,\\ 
534: \label{eq:BCSb}
535: n &=& \frac{2}{\Omega} \sum_\bk v^2_\bk ,
536: \end{eqnarray}
537: \label{eq:BCS}
538: \end{subequations}
539: where $v_\bk^2 = \frac{1}{2} (1 - \xi_\bk /E_\bk )$, together with
540:    $u_\bk^2 = 1-v_\bk^2$, are the usual coherence factors of BCS
541:    theory.
542: 
543: Owing to our choice of a free particle dispersion relation (all band
544:    effects are embedded in a single parameter, namely the effective
545:    mass $m^\ast$), and of a contact potential in real space, the sums
546:    over $\bk$ in Eq.~(\ref{eq:BCSa}) above 
547:    give rise to an ultraviolet divergence, which requires a suitable
548:    regularization.
549: In three dimensions, it is customary to intoduce the scattering
550:    amplitude $a_s$ \cite{Pistolesi:96,Marini:98,Pieri:00}, which for our
551:    anisotropic interaction reads:
552: \begin{equation}
553: \frac{m^\ast}{4\pi a_s} = -\frac{1}{V} + \frac{1}{\Omega} \sum_\bk
554:    \frac{m^\ast}{k^2} |Y_{\ell m} (\hat{\bk})|^2 .
555: \label{eq:scatt}
556: \end{equation}
557: Subtracting Eq.~(\ref{eq:scatt}) from Eq.~(\ref{eq:BCSa}) one has:
558: \begin{equation}
559: -\frac{m^\ast}{4\pi a_s} = \frac{1}{\Omega} \sum_\bk \left(
560:    \frac{1}{2E_\bk} - \frac{m^\ast}{k^2} \right) |Y_{\ell m}
561:    (\hat{\bk})|^2 .
562: \label{eq:BCSc}
563: \end{equation}
564: Following Ref.~\cite{Marini:98}, we render Eqs.~(\ref{eq:BCS}) and
565:    Eq.~(\ref{eq:BCSc}) dimensionless, by introducing the dimensionless
566:    quantities 
567: \begin{alignat*}{2}
568: x^2 &= \frac{\hbar^2 k^2}{2m^\ast} \frac{1}{\Delta_0} ,
569: & \quad x_0 &= \frac{\mu}{\Delta_0} ,\\
570: \xi_x &= \frac{\xi_\bk}{\Delta_0} = x^2 -x_0 , & \quad E_x &=
571:    \frac{E_\bk}{\Delta_0} = \sqrt{\xi_x^2 + |Y_{\ell m} (\hat{\bx})|^2}
572:    ,
573: \end{alignat*}
574: and the Fermi energy $\eF = \hbar^2 \kF^2 /(2m^\ast ) = \hbar^2
575:    (3\pi^2 n)^{2/3} /(2m^\ast )$.
576: In particular, Marini \emph{et al.} \cite{Marini:98} observe that $x_0
577:    =\mu/\Delta_0$ can be used as a parameter for the crossover between
578:    the strong-coupling, Bose-Einstein (BE) limit ($x_0 \ll 0$) and the
579:    weak-coupling, BCS limit ($x_0 \gg 0$).
580: Moreover, on one hand, the expression resulting from the dimensionless version
581:    of Eq.~(\ref{eq:BCSc}) for $(\kF a_s )^{-1}$ can be inverted to
582:    obtain $x_0$ as a function of the dimensionless scattering length
583:    $\kF a_s$.
584: On the other hand, one may alternatively use $\kF \xi$ as the
585:    independent variable in place of $\kF a_s$ \cite{Marini:98}.
586: Indeed, it was earlier recognized by Pistolesi and Strinati
587:    \cite{Pistolesi:94} (following the seminal work of Nozi\`eres and
588:    Schmitt-Rink \cite{Nozieres:85}) that a natural variable which can
589:    be used to follow the crossover from strong-coupling to weak-coupling
590:    superconductivity is the product $\kF\xi$ of Fermi wave-vector
591:    $\kF$ times the coherence length $\xi$ for two-electron
592:    correlation.
593: In the BE limit, electrons are expected to bind in quasi-bound pairs
594:    localized in real space (Schafroth pairs \cite{Schafroth:54}), thus
595:    realizing the condition $\kF \xi \ll 1$, while the BCS limit
596:    corresponds to loosely coupled pairs, with $\kF \xi \gg 1$,
597:    localized in momentum space close to the Fermi energy.
598: 
599: A zero-temperature calculation of the coherence length for a
600:    three-dimensional, $s$-wave 
601:    superconductor along the crossover between the weak- and the
602:    strong-coupling limits has been performed both numerically
603:    \cite{Pistolesi:96} and analitically \cite{Marini:98}.
604: The case of a two-dimensional, $d$-wave superconductor has been
605:    discussed in Ref.~\cite{Benfatto:02}, where a dispersion relation
606:    typical of the cuprate superconductors has been explicitly
607:    considered.
608: As anticipated above, in the non--$s$-wave case ($\ell\neq0$), the
609:    standard definition, Eq.~(\ref{eq:xidef}), of the coherence length
610:    leads to unphysical divergences.
611: One may still conveniently define a coherence length as the range in
612:    real space of the static correlation function $X_\Delta (\br)$ for
613:    the modulus of the order parameter \cite{Pistolesi:96,Benfatto:02}
614:    as
615: \begin{equation}
616: \xi^{-1} = - \lim_{r\to\infty} \frac{\log X_\Delta (r)}{r} .
617: \end{equation}
618: The Fourier transform $X_\Delta (\bq)$ of such a function has been
619:    derived for an $s$- and a $d$-wave superconductor in
620:    Refs.~\cite{Pistolesi:96} and \cite{Benfatto:02}, respectively.
621: In the case of our anisotropic interaction, Eq.~(\ref{eq:interY}), it
622:    reads:
623: \begin{eqnarray}
624: X_\Delta (\bq)^{-1} &=& \frac{1}{V} - \frac{1}{2\Omega} \sum_\bk 
625:    |Y_{\ell m} (\hat{\bk})|^2 \frac{(u_{\bk +\bq/2} u_{\bk-\bq/2} -
626:    v_{\bk+\bq/2} v_{\bk-\bq/2} )^2}{E_{\bk+\bq/2} + E_{\bk-\bq/2}}
627:    \nonumber \\
628: &=& \frac{1}{2\Omega} \sum_\bk |Y|^2 \left[
629:    \frac{1}{E} - \frac{1}{E_+ + E_-} \left( 1 + \frac{\xi_+ \xi_- -
630:    |\Delta_+ | |\Delta_- |}{E_+ E_- } \right) \right],
631: \label{eq:LaraX}
632: \end{eqnarray}
633: where $E\equiv E_\bk$, $Y\equiv Y_{\ell m} (\hat{\bk})$, $\xi_\pm$,
634:    $\Delta_\pm$, 
635:    $E_\pm$ are calculated at momenta $\bk \pm \bq/2$, respectively,
636:    and use has been made of the gap equation, Eq.~(\ref{eq:BCSa}), in
637:    going into the second line.
638: 
639: First of all, since the summand in Eq.~(\ref{eq:LaraX}) depends only
640:    on $k=|\bk|$, $q=|\bq|$, and on the relative angle between $\bk$
641:    and $\bq$, passing to the continuum limit and transforming the sums
642:    over wave-vectors into an integral, one has $X_\Delta (\bq )\equiv
643:    X_\Delta (q)$.
644: Moreover, $X_\Delta (q)$ is an even function of its argument.
645: Back to real space, one analogously has $X_\Delta (\br )\equiv
646:    X_\Delta (r)$.
647: Then, the asymptotic behaviour of $X_\Delta (r)$ at large distance $r$
648:    will be governed by the behaviour of its Fourier transform
649:    $X_\Delta (q)$ at small wave-vector $q$.
650: In particular, assuming the expansion \cite{Pistolesi:96,Benfatto:02}:
651: \begin{equation}
652: X_\Delta^{-1} (q) = a + b q^2 + O(q^4 ),
653: \end{equation}
654: it is straightfoward to show \cite{Pistolesi:96} that
655: \begin{equation}
656: \xi^2 = \frac{b}{a} .
657: \label{eq:xidefnew}
658: \end{equation}
659: Such a definition of the coherence length is now consistent also for
660:    nodal superconductors \cite{Benfatto:02}.
661: It reduces to Eq.~(\ref{eq:xidef}) in the weak-coupling limit for the
662:    $s$-wave case, and applies also to the strong-coupling regime,
663:    provided $b>0$ (given that $a>0$, identically, provided $\Delta_0
664:    \neq0$), \emph{i.e.} 
665:    provided that $X_\Delta (q)$ has its absolute minimum at $q=0$
666:    \cite{note:Lara}.
667: 
668: We have numerically solved the gap equations, Eqs.~(\ref{eq:BCS}), and
669:    evaluated the dimensionless coherence length $\kF\xi$ according to
670:    this more general definition, Eq.~(\ref{eq:xidefnew}), for the
671:    anisotropic potential Eq.~(\ref{eq:interY}), with $\ell=0,1,2$,
672:    corresponding to $s$-, $p$-, and $d$-wave pairing, respectively.
673: Here, a convenient measure of the characteristic energy for
674:    superconductivity may be taken as the maximum gap
675:    $\Delta_{\mathrm{max}}$, where
676:    $\Delta_{\mathrm{max}} \propto \kB T_c$ holds also for anisotropic
677:    superconductors \cite{Combescot:91}.
678: Such a quantity is readily extracted
679:    from the solution of the gap equations, Eqs.~(\ref{eq:BCS}).
680: In Figure~\ref{fig:delta}, we plot the dimensionless characteristic
681:    energy $2\Delta_{\mathrm{max}}/\eF$ versus $\varepsilon_c /\eF =
682:    (\kF \xi)^{-2}$ in the cases $\ell=0,1,2$.
683: As $\kF \xi$ decreases [$(\kF \xi)^{-2}$ increases], one crosses over
684:    from the weak-coupling, BCS limit into the strong-coupling, BE
685:    limit \cite{Pistolesi:94}.
686: Despite a mean-field solution of the self-consistent gap equations has
687:    been now taken into account in the calculation, and a more general
688:    and consistent definition of the coherence length has been
689:    employed, our numerical results are in good qualitative agreement
690:    with the approximate results of Sec.~\ref{sec:model}, with
691:    $2\Delta_{\mathrm{max}}/\eF$ increasing as a function of
692:    $\varepsilon_c /\eF$ with a steeper slope, as $\ell$ increases from
693:    $\ell=0$ ($s$-wave) to $\ell=2$ ($d$-wave), all curves tending to
694:    saturation, at least within the numerically accessible range of the
695:    crossover parameter $\kF \xi$.
696: These results are in agreement with the phenomenological plots correlating
697:    characteristic energies for both the heavy Fermion and cuprate
698:    superconductors in Refs.~\cite{Angilella:00b,Angilella:01a}.
699: 
700: 
701: \section{Summary and future directions}
702: \label{sec:conclusions}
703: 
704: We have studied the Cooper problem for an anisotropic superconductor
705:    characterized by an electron-electron interaction expanded in terms
706:    of simple spherical harmonics over the Fermi sphere.
707: In the weak-coupling limit of a superconducting instability in the
708:    $(\ell,m)$ channel, we have derived an analytical expression for
709:    the relation between the pair binding energy and the correlation
710:    length, for arbitrary relative angular momentum quantum number
711:    $\ell$.
712: While such an expression correctly reduces to the standard one in the
713:    $s$-wave case ($\ell=0$), in view of the fact that $\kB T_c$ scales
714:    with such a pair binding energy, in the case of non--$s$-wave
715:    superconductors ($\ell>0$) our expression agrees qualitatively
716:    with the phenomenological correlation between $\kB T_c$ and
717:    the characteristic energy $\hbar^2 / (2 m^\ast \xi^2 )$ earlier found in
718:    Ref.~\onlinecite{Angilella:00b}.
719: These results have been confirmed by a numerical solution of the
720:    self-consistent gap equations in the crossover between the weak-
721:    and the
722:    strong-coupling limits, where a more general definition for the
723:    coherence length has been employed, applying to anisotropic
724:    superconductors.
725: 
726: It may, in the future, prove of significance to include the effect of
727:    Coulomb interaction on the formation of Cooper pairs in
728:    superconducting assemblies along lines such as those laid down in
729:    Ref.~\onlinecite{Lal:92b} (see also Ref.~\onlinecite{Mila:91}).
730: The question arises then as to whether, in the future, it will be of
731:    significance in making the present study the basis of fully
732:    quantitative calculations, to generalize beyond the result here,
733:    and of course, of many earlier workers, that the Cooper pairs
734:    formed as a consequence of an attractive interaction correspond to
735:    a single value of the binding energy at a given temperature.
736: There is some evidence (see, \emph{e.g.,} Refs.~\onlinecite{Lal:91,Lal:92a})
737:    that some physical properties, and in particular specific heat and
738:    tunneling spectra of cuprate materials, may require generalization
739:    to Cooper pairs with a finite range of binding energies.
740: We do not anticipate that the qualitative trends proposed in the
741:    present paper will be grossly affected.
742: 
743: 
744: 
745: 
746: 
747: \begin{ack}
748: G.G.N.A. gratefully acknowledges Dr. N. Andrenacci for useful and
749:    stimulating discussions.
750: F.E.L. thanks Professor K. Van Alsenoy for much motivation and
751:    encouragement.
752: F.E.L. also acknowledges financial support from the ``Program for the
753:    Stimulation of Participation in Research Programs of the European
754:    Union'' of the University of Antwerp.
755: N.H.M. and F.E.L. thank the Department of Physics and Astronomy,
756:    University of Catania, for the stimulating environment and much
757:    hospitality.
758: \end{ack}
759: 
760: \bibliographystyle{elsart-num}
761: \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}
762: 
763: \newpage
764: 
765: \begin{figure}
766: \centering
767: \includegraphics[height=\columnwidth,angle=-90]{multil.ps}
768: \caption{Dependence of the normalized Cooper pair binding energy
769:    $|\varepsilon|/2\eF$ 
770:    on the characteristic energy $\varepsilon_c /\eF = [\hbar^2 /
771:    (2m^\ast \xi^2)]/\eF = (\kF\xi)^{-2}$, 
772:    as implicitly defined by Eq.~(\protect\ref{eq:xiellneq0}), for
773:    $\ell=0,1,2$, corresponding to $s$-, $p$-, and $d$-wave
774:    superconductivity, respectively.
775: See text for discussion.
776: }
777: \label{fig:multil}
778: \end{figure}
779: 
780: \begin{figure}
781: \centering
782: \includegraphics[height=\columnwidth,angle=-90]{delta.ps}
783: \caption{Dependence of twice the normalized maximum gap energy
784:    $2\Delta_{\mathrm{max}}/\eF$ 
785:    on the characteristic energy $\varepsilon_c /\eF = (\kF \xi)^{-2}$, for
786:    $\ell=0,1,2$, corresponding to $s$-, $p$-, and $d$-wave
787:    superconductivity, respectively.
788: As $\kF\xi$ decreases [\emph{i.e.,} $(\kF\xi)^{-2}$ increases] one
789:    crosses over from the weak-coupling, BCS limit, to the
790:    strong-coupling, BE limit.
791: }
792: \label{fig:delta}
793: \end{figure}
794: 
795: 
796: \end{document}
797: 
798: