1: \documentclass[letterpaper,english,amssymb]{revtex4}
2:
3: %\usepackage[T1]{fontenc}
4: %\usepackage[latin1]{inputenc}
5: \usepackage{graphicx}
6: \usepackage{amssymb}
7: \usepackage{overpic}
8:
9: %\usepackage{feynmf}
10: \usepackage{babel}
11:
12: %\renewcommand{\baselinestretch}{2.0}
13: \newcommand\HFILL{\hspace*{\fill}}
14: \newcommand\VFILL{\vspace*{\fill}}
15:
16: %\makeatletter
17:
18: %\makeatother
19:
20: \newcommand{\beq}{\begin{equation}}
21: \newcommand{\eeq}{\end{equation}}
22: \newcommand{\bea}{\begin{eqnarray}}
23: \newcommand{\eea}{\end{eqnarray}}
24:
25: \newcommand{\rmd}{{\rm d}}
26: \newcommand{\rmi}{{\rm i}}
27:
28: \newcommand{\kb}{{\bf k}}
29: \newcommand{\qb}{{\bf q}}
30:
31: \begin{document}
32:
33: \title{Renormalization-group approach to superconductivity: from weak to
34: strong electron-phonon coupling}
35: \author{S.-W.~Tsai$^1$, A.~H.~Castro~Neto$^1$, R.~Shankar$^2$, and D.~K.~Campbell$^1$}
36:
37: \affiliation
38: {$^1$ Department of Physics, Boston University, Boston, MA 02215\\
39: $^2$ Sloane Laboratory of Physics, Yale University, New Haven, CN 06520}
40:
41:
42:
43: \date{\today}
44:
45: \begin{abstract}
46: We present the numerical solution of the renormalization group (RG) equations
47: derived in Ref.\onlinecite{Tsai}, for the problem of superconductivity in the
48: presence of both electron-electron and electron-phonon coupling at zero
49: temperature. We study the instability of a Fermi liquid to a superconductor
50: and the RG flow of the couplings in presence of retardation effects and
51: the crossover from weak to strong coupling. We show that our numerical
52: results provide an ansatz for the analytic solution of the problem in the asymptotic
53: limits of weak and strong coupling.
54: \end{abstract}
55:
56: \maketitle
57:
58: The renormalization-group approach to interacting fermions in more than one
59: spatial dimension \cite{shankar} has been extensively applied to the study of
60: instabilities of the Fermi liquid state, and has become a major tool in the
61: study of correlated electron systems. Nevertheless, even for weak
62: electron-electron interactions, the picture is far from being complete, since
63: electrons in solids also interact with bosonic modes such as
64: phonons. Therefore, the development of an RG scheme \cite{Tsai} that includes
65: both electron-electron and electron-phonon interactions on an equal footing is
66: an important advance. Experimental evidence show that in many strongly
67: correlated systems, such as high-temperature superconductors and organic
68: charge transfer salts, both electron interactions and phonons seem to play
69: an important role\cite{shen,organics,mgb2,millis,c60}.
70:
71: The renormalization-group approach to interacting fermions coupled to phonons
72: was presented in Ref. [\onlinecite{Tsai}]. This approach takes retardation
73: effects and the
74: presence of multiple energy scales fully into account. For a circular Fermi
75: surface,
76: the RG equations predict the onset of the superconducting instability in
77: agreement
78: withe Eliashberg's superconducting theory \cite{eliashberg}. A large-$N$
79: analysis,
80: where $N \approx E_F/\Lambda$ is the number of patches in the Fermi surface
81: ($E_F$
82: is the Fermi energy and $\Lambda$ is the size of the patch),
83: shows that in this case Eliashberg theory is asymptotically exact and Migdal's
84: theorem \cite{migdal} emerges as a consequence of the $1/N$ expansion.
85: Here we present a numerical solution
86: of the RG equations at $T=0$, showing how the couplings, which are functions of
87: frequencies, flow with the RG procedure. In the limits of weak and strong
88: phonon coupling, simple analytical expressions are extracted on the basis
89: of the numerical solution.
90:
91: For completeness, in Sec. \ref{sec:derivation} the RG equations for the
92: self-energy and
93: interaction couplings are derived. In Sec. \ref{sec:flows} we present the
94: numerical results, and the analytical expressions associated with the
95: asymptotic limits.
96: Sec. \ref{sec:conclusion} contains the concluding remarks.
97:
98: \section{Derivation of the RG equations}
99: \label{sec:derivation}
100:
101: The action that describes electron-electron and electron-phonon interactions
102: can be
103: written as $S(\psi,\phi) = S_e(\psi) + S_{ph}(\phi) +
104: S_{e-ph}(\psi,\phi) + S_{e-e}(\psi)$,
105: where $\phi$ are bosonic fields, $\psi$ are fermionic (Grassman) fields (we
106: use units such that $\hbar = 1 = k_B$),
107: \begin{eqnarray}
108: S_e = \int_{\omega \kb \sigma}
109: \psi^{\dagger}_{k \sigma} (i\omega - \epsilon_{\kb}) \psi_{k \sigma} \, ,
110: \end{eqnarray}
111: is the free electron action, $\epsilon_{\kb}$ is the electron dispersion as
112: a function of momentum $\kb$, and
113: \begin{eqnarray}
114: S_{ph} = \int_{\Omega \qb} \phi^{\dagger}_{q} (i\Omega - w_{\qb})
115: \phi_{q} \, ,
116: \end{eqnarray}
117: is the free phonon action where $w_{\qb}$ is the phonon dispersion
118: ($\sigma=\uparrow,\downarrow$ is the electron spin,
119: $k = \{\omega, \kb\}$ and $q = \{\Omega, \qb\}$,
120: where $\omega,\Omega$ are fermionic and bosonic Matsubara frequencies,
121: respectively, and $\kb,\qb$ are the momenta).
122: The electron-phonon interaction can be written as:
123: \begin{eqnarray}
124: S_{e-ph} =
125: \int_{\omega \kb \sigma} \int_{\Omega \qb} g(q) \psi^{\dagger}_{k+q \sigma}
126: \psi_{k \sigma}(\phi_q + \phi^{\dagger}_{-q}) \, ,
127: \end{eqnarray}
128: where $g(q)$ is the electron-phonon coupling constant. The electron-electron
129: interactions have the form:
130: \begin{eqnarray}
131: S_{e-e} = \frac{1}{2}
132: \prod_{i=1}^3 \int_{\omega_i \kb_i \sigma \sigma'} u(k_4,k_3,k_2,k_1)
133: \psi^{\dagger}_{k_4 \sigma} \psi_{k_2 \sigma} \psi^{\dagger}_{k_3
134: \sigma^{\prime}} \psi_{k_1 \sigma^{\prime}} \, ,
135: \end{eqnarray}
136: where $u(\{k_i\})$ is a general spin-independent electron-electron coupling
137: which depends on the momenta and frequencies of the electrons, and
138: $k_4=k_1+k_2-k_3$ due to momentum conservation.
139:
140: Since the bosonic action is quadratic in the boson fields, they can be
141: integrated
142: out exactly, leading to an effective electron-electron problem with retarded
143: interactions:
144: \begin{eqnarray}
145: \tilde{u}(k_4,k_3,k_2,k_1) =
146: u(k_4,k_3,k_2,k_1)- 2 g(k_1,k_3) g(k_2,k_4) D(k_1-k_3) \, ,
147: \label{eq:initial}
148: \end{eqnarray}
149: where
150: \begin{eqnarray}
151: D(q) = \omega_{{\bf q}}/(\omega^2+\omega_{{\bf q}}^2)
152: \end{eqnarray}
153: is the phonon propagator. Here we consider the case of a circular Fermi
154: surface and anisotropic Einstein
155: phonons, with $g(k,k^{\prime})=g_0$ and $\omega_{{\bf q}}=\omega_E$. In what
156: follows, it is also convenient to define the dimensionless electron-phonon
157: coupling
158: constant
159: \begin{eqnarray}
160: \lambda=2N(0)g_0^2/\omega_E \, .
161: \end{eqnarray}
162:
163: In the Kadanoff-Wilson approach to RG the flow equations are obtained by
164: computing the corrections to the couplings of the theory as the energy modes
165: in a energy shell between $\Lambda$ and $\Lambda+d\Lambda$ are integrated
166: out. The momenta, frequencies, and fields are then rescaled in such a way that
167: the quadratic terms remain unchanged. This second step presents a difficulty
168: in the electron-phonon problem. The momenta of the electrons scale
169: differently in the directions parallel and perpendicular to the Fermi surface,
170: while the momenta of the phonons scale isotropically. In our RG treatment of
171: the electron-phonon problem \cite{Tsai}, we use the quantum field theory
172: version of the RG, with no rescaling. In this approach, corrections to the
173: vertices of the model are written in term of running, {\it i. e.}
174: cut-off--dependent,
175: coupling functions. The RG flow equations for these running couplings are
176: then obtained by imposing that the vertices are cut-off independent.
177: % rescaling the
178: %momenta and frequencies the coupling constants
179: %do that to absorb the change in the cut-off energy scale $\Lambda$
180: %\cite{Tsai}. As shown in Ref.\onlinecite{Tsai} the electron two-point vertex
181: %in leading order in $1/N$, is given by:
182:
183: We start with the two-point vertex, which at one loop is given by \cite{Tsai}
184: \begin{eqnarray}
185: \Gamma^{(2)}(\Sigma(\omega, {\mathbf k})) = \Sigma_{\ell}(\omega,{\mathbf k}) -
186: \int_{-\infty}^{\infty} \frac{d\Omega}{2\pi} \int_{(\Lambda)}
187: \frac{d^2{\mathbf k}^{\prime}}{(2\pi)^2}
188: \frac{e^{i\Omega 0_+}}{i\Omega - \epsilon_{\mathbf k^{\prime}} -
189: \Sigma_{\ell}(\Omega,
190: {\mathbf k}^{\prime})} \tilde{u}(k, k^{\prime}, k, k^{\prime})
191: \label{eq:gamma2}
192: \end{eqnarray}
193: where $k=(\omega,{\mathbf k})$ and $k^{\prime}=(\Omega,{\mathbf
194: k}^{\prime})$, and
195: $\Sigma_{\ell}(\omega,{\mathbf k})$ is the electronic self-energy which now
196: flows under the RG, and the RG parameter $\ell = \log(\Lambda_0/\Lambda)$, so
197: that $d\ell=-d\Lambda/\Lambda$.
198: The RG equation for the electron self-energy is obtained by
199: imposing the condition of renormalizability of the theory, that is,
200: \begin{eqnarray}
201: \frac{d \Gamma^{(2)}}{d\ell} = 0 \, ,
202: \end{eqnarray}
203: where $\ell = \log (\Lambda_0/\Lambda)$ and $d \ell = - d\Lambda/\Lambda$.
204:
205: Notice that the interaction that appears in
206: Eq. (\ref{eq:gamma2}), $\tilde{u}(k,k^{\prime},k,k^{\prime})$, only describes the
207: forward scattering channel ($k_4=k_1=k$, $k_3=k_2=k'$) which, in the large
208: $N$ limit,
209: does not flow under RG \cite{Tsai}.
210: Therefore, it can be substituted by its unrenormalized value given by
211: Eq. (\ref{eq:initial}).
212: The electron-electron part of the interaction, $u(k,k^{\prime},k,k^{\prime})
213: = u_0$ contributes only to the real part of the self-energy and gives a shift
214: in the chemical potential that can be reabsorded into the definition of the
215: chemical potential \cite{shankar}. The electron-phonon part
216: %of the interaction can be average out over the Fermi surface and,
217: for $\omega > 0$, leads to the following RG equation for the imaginary part of
218: the self-energy, $\Sigma^{''}(\omega)$:
219: \begin{eqnarray}
220: \frac{\partial}{\partial \ell} \Sigma^{''}_{\ell}( \omega)
221: = - \int_{-\infty}^{+\infty} \frac{d \Omega}{\pi}
222: \lambda \omega_E D(\Omega-\omega)
223: \frac{\Lambda_{\ell} (\Omega-\Sigma^{''}_{\ell}(\Omega))}{
224: (\Omega-\Sigma^{''}_{\ell}(\Omega))^2 + \Lambda_{\ell}^2} \ \ ,
225: \label{eq:rgself}
226: \end{eqnarray}
227: where $\Lambda_{\ell} = \Lambda_0 e^{-\ell}$ (where $\Lambda_0 < E_F$ is the
228: cut-off in the beginning of the RG flow and it is assumed to be much larger
229: than the other energy scales in the problem). There is no dependence of
230: $\Sigma^{''}_{\ell}$ on the direction of the momentum ${\mathbf k}$ because
231: we are considering the case of isotropic Fermi surface, and the dependence on
232: the magnitude of ${\mathbf k}$ is irrelevant \cite{shankar}.
233: It is convenient to write $\Sigma^{"}(\omega) = [1-Z(\omega)]\omega$,
234: and the solution of Eq. (\ref{eq:rgself}) becomes:
235: \begin{eqnarray}
236: Z_{\ell}(\omega) = 1
237: + \frac{1}{\omega} \int_{-\infty}^{+\infty} \frac{d \Omega}{2}
238: \lambda \omega_E D(\Omega-\omega) \Omega
239: F_{\ell}(\Omega)
240: \label{eq:self}
241: \end{eqnarray}
242: where
243: \begin{eqnarray}
244: F_{\ell}(\omega) = \frac{2}{\pi} \int_{W_{\ell}(\omega)}^{\infty}
245: \frac{dW^{\prime}}{\left(Z_{W^{\prime}}(\omega)/Z_{W}(\omega)\right)^2
246: \omega^2 + W^{\prime 2}} \ \ ,
247: \label{fe}
248: \end{eqnarray}
249: where we have introduce $W_{\ell}(\omega) = \Lambda_{\ell}/Z_{\ell}(\omega)$
250: is the new renormalized cut-off running scale.
251: The dependence of $Z_W(\omega)$ on $W$ is weak (see below) and therefore we can write
252: $Z_W(\omega) \approx Z_{W^{\prime}}(\omega)$. In this case (\ref{fe})
253: can be solved at once:
254: \begin{eqnarray}
255: F_{\ell}(\omega) \approx \frac{1}{|\omega|} \left[1-\frac{2}{\pi} \tan^{-1}
256: \left(\frac{W_{\ell}(\omega)}{|\omega|}\right)\right] \, .
257: \label{bigf}
258: \end{eqnarray}
259: In what follows we will be only interested in the low frequency behavior, in which
260: case $W_{\ell}(\omega)$ can be safely replaced by $W_{\ell}(\omega=0)$, the
261: expression for $Z_{\ell}(\omega)$ can be easily evaluated from (\ref{eq:self}) to give
262: \begin{eqnarray}
263: Z_{\ell}(\omega) = 1 - \lambda \frac{\omega_E}{\omega}
264: \left[\tan^{-1}\left(\frac{W_{\ell}+\omega_E}{|\omega|}\right) -
265: \frac{\pi}{2}\right] \, .
266: \label{big_Z}
267: \end{eqnarray}
268: Notice that indeed the dependence of $Z_W$ on $W$ is weak, as assumed previously.
269: In the static limit, $\omega \to 0$, one obtains
270: \begin{eqnarray}
271: Z_{\ell}(0) = 1 + \lambda \omega_E/(W_{\ell}+\omega_E) \, .
272: \end{eqnarray}
273:
274: So far we have considered the renormalization of the self-energy.
275: The renormalization of the interaction in the Cooper channel ($k_4=-k_3$ and $k_2=-k_1$)
276: can be obtained in a completely analogous way.
277: Since we are considering the case of a circular Fermi surface, we can focus
278: on the $s$-wave component of the BCS interaction
279: $\tilde{v}(\omega_1,\omega_3) \! = \! N(0) \int d\theta_1/2\pi
280: \int d\theta_3/2\pi \tilde{u}(-k_3,k_3,-k_1,k_1)$.
281: The RG equation for this interaction can be shown to be \cite{Tsai}:
282: \begin{eqnarray}
283: \frac{\partial}{\partial \ell} \tilde{v}(\omega_1,\omega_2,\ell)
284: = - \int_{-\infty}^{+\infty} \frac{d \omega}{\pi}
285: \frac{\Lambda_{\ell}}{[\omega-\Sigma_{\ell}^{''}(\omega)]^2
286: +\Lambda_{\ell}^2} \tilde{v}(\omega_1,\omega,\ell) \,
287: \tilde{v}(\omega,\omega_2,\ell) \, ,
288: \label{main_rg}
289: \end{eqnarray}
290: where the initial condition for the flow is given by
291: $\tilde{v}(\omega_1,\omega_3,\ell=0) = u_0 - \lambda
292: \omega_E D(\omega_1-\omega_3)$. Equation (\ref{main_rg}) can be
293: written in matrix equation as:
294: \begin{eqnarray}
295: \frac{\partial {\bf U}}{\partial \ell} = - {\bf U} \cdot {\bf M} \cdot {\bf
296: U}
297: \label{main_rg_matrix}
298: \end{eqnarray}
299: where
300: \begin{eqnarray}
301: U_{ij}(\ell) &=& \tilde{v}(\omega_i,\omega_j,\ell)
302: \\
303: M_{ij}(\ell) &=& \Lambda_{\ell}\delta_{ij}/
304: [(\omega_i-\Sigma^{''}_{\ell}(\omega_i))^2 + \Lambda_{\ell}^2] \, .
305: \end{eqnarray}
306: Formally, we can rewrite (\ref{main_rg_matrix}) as
307: \begin{eqnarray}
308: \frac{\partial {\bf U}^{-1}}{\partial \ell} = {\bf M}
309: \label{inverse}
310: \end{eqnarray}
311: since ${\bf U}^{-1} \partial_{\ell} {\bf U} {\bf U}^{-1} = -
312: \partial_{\ell}{\bf U}^{-1}$. The solution of (\ref{inverse}) reads:
313: \begin{eqnarray}
314: {\bf U}^{-1}(\ell) = {\bf U}^{-1}(0) + {\bf P}(\ell)
315: \end{eqnarray}
316: where
317: \begin{eqnarray}
318: {\bf P}(\ell) = \int_0^{\ell} d\ell' {\bf M}(\ell') \, ,
319: \end{eqnarray}
320: which can inverted to give:
321: \begin{eqnarray}
322: {\bf U}(\ell) =
323: [{\bf 1} + {\bf U}(0) \cdot {\bf P}(\ell)]^{-1} \cdot {\bf U}(0) \, .
324: \label{uell}
325: \end{eqnarray}
326:
327: Eq.(\ref{uell}) allows the study of instabilities of the Fermi liquid
328: state towards superconducting instabilities. The instabilities
329: occur when one of the couplings diverges under the RG flow.
330: These instabilities happen at some finite energy (or temperature)
331: scale $\ell_c$ at which one of the eigenvalues of the
332: coupling matrix ${\bf U}(\ell_c) $ diverges at $\ell = \ell_c$.
333: Notice that the condition for the instability can be written as:
334: \begin{eqnarray}
335: \det[{\bf 1}+{\bf U}(0)\cdot{\bf P}(\ell_c)]=0 \, .
336: \end{eqnarray}
337: Hence, the problem reduces to the calculation of the zeros of a determinant
338: or, equivalently,
339: to the problem of finding the zero eigenvalue of the
340: matrix $\left[{\bf 1} + {\bf U}(0) \cdot {\bf P}\right]$:
341: \begin{eqnarray}
342: [{\bf 1} + {\bf U}(0) \cdot {\bf P}(\ell_c)] \cdot {\bf f} =0
343: \label{eigenvalue_problem}
344: \end{eqnarray}
345: where the $f_i$ is the eigenvector of the problem.
346: Eq. (\ref{eigenvalue_problem}) can be written explicitly as:
347: \begin{eqnarray}
348: f(\omega) = - \int_{-\infty}^{+\infty} \frac{d \Omega}{\pi}
349: \int_{\Lambda_c}^{\Lambda_0}
350: d\Lambda \frac{1}{[\Omega Z_{\ell}(\Omega)]^2+\Lambda^2}
351: \left[u_0 - \frac{\lambda \omega_E^2}{(\omega-\Omega)^2+\omega_E^2}\right]
352: f(\Omega) \, .
353: \label{int_eq_2}
354: \end{eqnarray}
355: Here we can, similarly to the expression for the self-energy (\ref{eq:self}),
356: write
357: \begin{eqnarray}
358: Z_{\ell_c}(\omega) \phi_{\ell_c}(\omega) =
359: - \int_{-\infty}^{+\infty} \frac{d \omega'}{2}
360: \left[u_0-\frac{\lambda \omega_E^2}{(\omega-\omega')^2+\omega_E^2}\right]
361: F_{\ell_c}(\omega') \, \,
362: \phi_{\ell_c}(\omega') \, ,
363: \label{appr_1}
364: \end{eqnarray}
365: where we have defined $\phi_{\ell_c}(\omega) = f_{\ell_c}(\omega)/Z_{\ell_c}(\omega)$.
366: Equations (\ref{big_Z}) and (\ref{appr_1}) determine the energy scale $W_c = W_0 e^{-\ell_c}$
367: at which the renormalization group equations for the scattering in the Cooper channel
368: diverge as one renormalizes the problem from high to low energies. Below this energy scale
369: the Fermi liquid breaks down and superconductivity sets in. Thus, we can associate
370: $W_c$ with the superconducting gap, $\Delta$. In fact, we will show that the equation for
371: $W_c$ gives exactly the same result obtained from Eliashberg's theory for strongly coupled
372: superconductors \cite{eliashberg}.
373:
374: \section{Solution of the RG flow equations}
375: \label{sec:flows}
376:
377: Traditionally the superconducting temperature has been calculated directly from
378: Eliashberg's equations \cite{eliashberg}. Nevertheless, the
379: formalism presented in the previous section allows the solution of the
380: problem by
381: solving equations (\ref{big_Z}) and (\ref{uell}), instead. The advantage of
382: such
383: a procedure is clear since it is not necessary to solve integral equations.
384:
385: It is interesting to investigate how the coupling matrix evolves under the RG
386: flow in different regimes since it provides an insight on how to solve the
387: problem analytically in some asymptotic limits. The simplest case occurs
388: when there are no phonons present (as discussed in Ref.\onlinecite{shankar})
389: where a full analytical solution is possible.
390: For the case of $\lambda =0$, Eq. (\ref{big_Z}) gives $Z_{\ell}=1$ and (\ref{uell})
391: becomes independent of the external frequency and therefore we must have
392: $\phi(\omega) = C$ is a non-zero constant. In this case (\ref{uell}) gives:
393: \begin{eqnarray}
394: 1 &=& - u_0 \int_{-\infty}^{+\infty} \frac{d \omega'}{2} F_{\ell_c}(\omega')
395: \nonumber
396: \\
397: &=& -u_0 \ln(W_0/W_c)
398: \end{eqnarray}
399: where we have used (\ref{bigf}). Notice that because $W_c < W_0$ the above
400: equation only has solution if $u_0<0$, that is, if the unrenormalized electron-electron
401: interactions are attractive to start with. The solution of the above equation
402: gives:
403: \begin{eqnarray}
404: \Delta = \Lambda_0 e^{1/u_0} \, ,
405: \end{eqnarray}
406: where $W_0 = \Lambda_0$ since $Z=1$.
407: Since, in our case we assume $u_0>0$, no superconducting instability
408: can exist in the absence of phonons.
409:
410: Instead of only focusing on the solution of the RG equation at the
411: instability, at which point the zero eigenvalue condition
412: (\ref{eigenvalue_problem}) holds, we have also obtained the full solution for
413: the RG flow by solving (\ref{eq:self}) and (\ref{uell}) numerically. Fig.
414: \ref{fig:weak} shows the evolution of ${\bf U}(\ell)$ with $\ell$. Each panel
415: in Fig. \ref{fig:weak} represents the $N$x$N$ matrix ${\bf U}(\ell)$ at a given
416: $\ell$. The frequency has been cut-off so that $|\omega| < \Lambda_0$ and has
417: been discretized in $N=200$ divisions in this interval. In a typical flow in
418: the weak-coupling limit (in the case of Fig. \ref{fig:weak} $\lambda = 0.3$,
419: $\omega_E=10$, and $u_0=0.1$), ${\bf U}$ flows from the initial condition where
420: $\tilde{v}(\omega,-\omega)$ are the largest couplings to the instability point
421: where the couplings that first diverge are the $\tilde{v}(\omega_i,\omega_j)$
422: where $|\omega_i|, |\omega_j| < \omega_E$.
423:
424: \begin{figure}[htb]
425: \includegraphics[scale=0.9]{g03.eps}
426: \caption{Plots of the $N$x$N$ matrix $U$ at different RG scales
427: $\ell$. Here the number of frequency divisions $N\!=\!200$, and the value
428: of the parameters used are $\lambda\!=\!0.3$,
429: $\Lambda_0\!=\! 100$, $\omega_E\!=\! 10$,
430: $u_0\!=\! 0.1$. Panels correspond to $\ell$ = 0, 2.5, 3, 5, 6.5, 6.9, 7.1,
431: and 7.19.}
432: \label{fig:weak}
433: \end{figure}
434:
435: We can clearly see from Fig.\ref{fig:weak} that, as the RG develops,
436: the matrix ${\bf U}$, which at the beginning of the RG has its largest
437: elements along
438: the diagonal, acquires a cross-like shape, at which point its largest matrix
439: elements occur near the origin where the frequencies are very small. The
440: region with largest matrix elements occurs within the dark circle of size
441: $\omega_E$ around the origin. Moreover, within this circle the matrix
442: elements are essentially constant showing that there is very little frequency
443: dependence if the frequencies are smaller than $\omega_E$.
444: %Outside the circle
445: %the matrix elements are much smaller and can be neglected.
446: Hence, $\omega_E$ is the scale that separates high from low
447: energy physics in the weak coupling regime. Thus, the full frequency dependence
448: of the phonon propagator is irrelevant and we can safely approximate:
449: \begin{eqnarray}
450: \lambda \omega_E D(\omega) \approx \left\{
451: \begin{array}{ll}
452: \lambda & \hskip 0.5cm {\rm if} \ \ \ |\omega| < \omega_E\\
453: 0 & \hskip 0.5cm {\rm if} \ \ \ |\omega| > \omega_E\\
454: \end{array}
455: \right. \, .
456: \label{phonon_prop_app}
457: \end{eqnarray}
458: Hence, from Eq. (\ref{big_Z}) we have
459: $Z_{\ell}(\omega) \approx 1+ \lambda$ which is independent of $\ell$.
460:
461: Even tough the most divergent elements of the matrix ${\bf U}$ are the ones
462: at small frequencies, it is important to include all the matrix elements
463: since the RG equation couples low and high frequency processes.
464: Again, making use of the fact that $\omega_E$ is the energy scale that
465: separates high and low energy physics in this limit,
466: %Since there are only two scales in the weak coupling regime
467: we can propose the {\it ansatz}:
468: \begin{eqnarray}
469: \phi(\omega) = \left\{
470: \begin{array}{ll}
471: \phi_0 & \hskip 0.5cm {\rm if} \ \ \ |\omega| < \omega_E\\
472: \phi_{\infty} & \hskip 0.5cm {\rm if} \ \ \ |\omega| > \omega_E\\
473: \end{array}
474: \right.
475: \label{ansatz_weak}
476: \end{eqnarray}
477: where $\phi_0$ and $\phi_{\infty}$ are unknowns that to be calculated.
478: Substituting (\ref{phonon_prop_app}), (\ref{ansatz_weak}), and the expression
479: for $Z_{\ell_c}$ into (\ref{uell}) one finds:
480: \begin{eqnarray}
481: \left[1+\lambda+(u_0 - \lambda) \, h(\omega_E/\Lambda_c)\right] \phi_0 +
482: (1+\lambda) (\ell_c - h(\omega_E/\Lambda_c)) \phi_{\infty}&=&0 \, ,
483: \nonumber
484: \\
485: u_0 h(\omega_E/\Lambda_c) \phi_0 + (1+\lambda) \left[1+(\ell_c
486: -h(\omega_E/\Lambda_c))u_0\right] \phi_{\infty} &=& 0 \, ,
487: \label{eq_weak}
488: \end{eqnarray}
489: where
490: \begin{eqnarray}
491: h(z) = \frac{2}{\pi} \int_0^{z} dx \, \frac{\tan^{-1}(x)}{x} \, ,
492: = \frac{z}{2 \pi} \Phi(-z^2)
493: \label{fc}
494: \end{eqnarray}
495: and where $\Phi(x)$ is a LerchPhi function, $\Phi(x) = \sum_{k=0}^{\infty}
496: x^k/(k+1/2)^2$. The solution of (\ref{eq_weak}) is given by the determinant
497: equation:
498: \begin{eqnarray}
499: \left|
500: \begin{array}{ll}
501: 1+\lambda+(u_0 - \lambda) \, h(\omega_E/\Lambda_c) & \hskip 0.5cm
502: (1+\lambda) (\ell_c - h(\omega_E/\Lambda_c)) \, u_0 \\
503: u_0 \, h(\omega_E/\Lambda_c) & \hskip 0.5cm (1+\lambda)(1+(\ell_c
504: -h(\omega_E/\Lambda_c))) \, u_0 \\
505: \end{array}
506: \right| = 0 \, ,
507: \end{eqnarray}
508: or equivalently,
509: \begin{eqnarray}
510: \lambda u_0 h^2(\omega_E/\Lambda_c) - \lambda(1+u_0+ u_0 \ell_c)
511: h(\omega_E/\Lambda_c) + (1+\lambda) \, (1 + u_0 \ell_c) = 0 \, ,
512: \end{eqnarray}
513: which is a transcendental equation for $\Lambda_c$. The last equation can be
514: rewritten in a more interesting form by defining $\Lambda_c = x_c \omega_E$
515: where $x_c$ is the variable of the problem (notice that $\ell_c =
516: \ln(\Lambda_0/\Lambda_c) = \ell_E + \ln(1/x_c)$ where $\ell_E =
517: \ln(\Lambda_0/\omega_E)$) in order to get:
518: \begin{eqnarray}
519: h^2(1/x_c) - \left[1/\mu^* +1+\ln(1/x_c)\right] h(1/x_c) +
520: \frac{1+\lambda}{\lambda} \left[1/\mu^* + \ln(1/x_c)\right] = 0 \, ,
521: \label{trans_weak}
522: \end{eqnarray}
523: where
524: \begin{eqnarray}
525: \mu^* = \frac{u_0}{1+ u_0 \ln(\Lambda_0/\omega_E)} \, ,
526: \end{eqnarray}
527: is the renormalized electron-electron interaction at the scale of phonon
528: energy $\omega_E$. This is the Anderson-Morel potential\cite{Morel}, which
529: emerges naturally from the solution to the RG flow.
530:
531: Note that the
532: renormalized interaction is really weakly dependent on the bare
533: electron-electron interaction since as $u_0$ increases it saturates very fast
534: at $\mu^* \approx 1/\ln(\Lambda_0/\omega_E)$ which is interaction
535: independent. In fact, in ordinary superconductors $\mu^*$ varies little from
536: material to material ($0.1<\mu^*<0.25$) and this can be understood by the
537: logarithmic dependence with the cut-off $\Lambda_0$ and the phonon frequency
538: $\omega_E$ (if we estimate $\Lambda_0 \approx 10^4$K while
539: $\omega_E \approx 10^2$K we get $\mu^* \approx 0.217$).
540:
541: When $x_c \ll 1$ ($\Lambda_c \ll \omega_E$) we can approximate $h(1/x_c)
542: \approx \ln(1/x_c)$ and Eq. (\ref{trans_weak}) becomes
543: \begin{eqnarray}
544: \left(\frac{1}{\mu^*}+1-\frac{1+\lambda}{\lambda}\right) \ln(1/x_c) \approx
545: \frac{1+\lambda}{\lambda \mu^*} \, ,
546: \end{eqnarray}
547: which leads to
548: \begin{eqnarray}
549: W_c \approx \omega_E \, \exp [- (1+\lambda)/(\lambda - \mu^*)]
550: \end{eqnarray}
551: which is the MacMillan expression for the superconducting gap at zero
552: temperature \cite{mc,allen}.
553:
554: \begin{figure}[htb]
555: \includegraphics[scale=0.9]{g4.eps}
556: \caption{Plots of the $N$x$N$ matrix $U$ at different RG scales
557: $\ell$. Here the number of frequency divisions $N\!=\!200$, and the value of
558: the other parameters are $\lambda\!=\!4$, $\Lambda_0\!=\!100$,
559: $\omega_E\!=\! 10$, and
560: $u_0\!=\!0.1$. Panels correspond to $\ell$ = 0, 1, 2, 2.5, 3, 3.13, 3.157,
561: 3.172. The scale $2W_c \approx 40$ distinguishes the high and low
562: frequencies close to $\ell_c$.}
563: \label{fig:strong}
564: \end{figure}
565:
566: In the strong-coupling regime ($\lambda > 1$) the situation is rather different.
567: Fig. \ref{fig:strong} shows the numerical results for the full solution of
568: the RG flow for parameters in strong coupling. The form of the
569: matrix ${\bf U}(\ell)$ is very different from the weak-coupling limit and the
570: important scale separating the high and low energy physics is $2W_c$.
571: We have seen that for $\lambda <1$ the characteristic energy scale
572: that separates high from low energy physics is $\omega_E$.
573: When $\lambda > 1$ we expect $W_c > \omega_E$ in which case
574: $\omega_E$ is not the characteristic energy scale of the instability.
575: Instead we expect $W_c$ to act as the characteristic energy
576: scale in separating the low and high energy physics. Thus, it is natural to
577: make the following {\it ansatz}:
578: \begin{eqnarray}
579: \phi(\omega) = \left\{
580: \begin{array}{ll}
581: \phi_0 & \hskip 0.5cm {\rm if} \ \ \ |\omega| < 2 W_c\\
582: 0 & \hskip 0.5cm {\rm if} \ \ \ |\omega| > 2 W_c\\
583: \end{array}
584: \right. \, .
585: \label{ansatz_strong}
586: \end{eqnarray}
587: where $\phi_0$ is an unknown.
588: Since, as discussed previously, $\mu^*$ is bounded from above,
589: the electron-electron interaction will play a minimal role in the problem and
590: can be disregarded.
591:
592: In this regime (\ref{appr_1}) becomes ($\phi_0 \neq 0$):
593: \begin{eqnarray}
594: Z_c(0) = 2 \lambda \omega^2_E \int_0^{2 W_c} \frac{d \omega}{\pi}
595: \int_{W_c}^{+\infty} d\Lambda \frac{1}{(\omega^2+\Lambda^2) \,
596: (\omega^2+\omega_E^2)}
597: \approx \frac{2 \lambda}{\pi} \left( \frac{\pi \omega_E}{2 W_c}
598: -\frac{\pi \omega_E^2}{4 W_c^2}\right) \, .
599: \label{eq_strong_1}
600: \end{eqnarray}
601: But from (\ref{big_Z}) we have $Z_{c}(0) \approx 1 +
602: \lambda (\omega_E/W_c) - \lambda (\omega_E/W_c)^2$,
603: which substituted in (\ref{eq_strong_1}) gives
604: \begin{eqnarray}
605: W_c \approx \sqrt{\lambda} \omega_E \, ,
606: \end{eqnarray}
607: the strong coupling result of Eliashberg
608: equation found by Allen and Dynes \cite{allen}.
609:
610:
611: \section{Discussion and conclusion}
612: \label{sec:conclusion}
613:
614: We have numerically solved the RG equations for the flow of the BCS coupling
615: matrix for interacting electrons coupled to phonons, for the case of circular
616: Fermi surface and Einstein phonons. Fig. \ref{fig:lamc}
617: shows the result for the energy scale of divergence in the Cooper channel,
618: for a range of $\lambda$, and the expected analytical result at weak and
619: strong coupling limits. One can clearly see that the approximate analytical
620: solutions, based on the evolution of the coupling matrix under the RG, give a
621: very good description of the superconducting gap in the asymptotic
622: limits. Thus, our RG scheme can used as an alternative way to solve for the
623: superconducting properties of a Fermi liquid and is in full agreement with
624: Eliashberg's theory.
625:
626:
627: \begin{figure}[htb]
628: \includegraphics[scale=0.45,angle=0]{wc.eps}
629: \caption{Plot of $W_c$ ($\sim $ superconducting gap) in units of
630: $\omega_E$ versus $\lambda$ (black circles) ($N=200$,
631: $\Lambda_0 = 100$, $\omega_E=10$, $u_0=0.1$). The dashed line is the
632: fit at small $\lambda$'s using the MacMillan formula and the
633: solid line is the fit to $\sqrt{\lambda}$ at large $\lambda$'s (Allen-Dynes). }
634: \label{fig:lamc}
635: \end{figure}
636:
637: Unlike mean-field approaches like Eliashberg theory, the RG approach, when
638: solved
639: numerically, can be easily extended to any shape of the Fermi surface or
640: anisotropic phonons. It also addresses the question of competition between
641: different types of instabilities, as was done extensively for the case
642: without phonons\cite{zanchi,halboth,marston}. Furthermore, the approach
643: presented here at zero temperature can be easily extended to finite
644: temperature.
645:
646:
647: \acknowledgments
648: We thank J.~Carbotte, A.~Chubukov, C.~Chamon, J.~B.~Marston,
649: G.~Murthy, and M.~Silva-Neto for illuminating discussions
650: and the Aspen Center for Physics for its hospitality during the
651: early stages of this work. A.~H.~C.~N. was supported
652: by NSF grant DMR-0343790. R.~S. was supported by NSF grant
653: DMR-0103639.
654:
655: \begin{thebibliography}{20}
656:
657: \bibitem{Tsai}S.-W. Tsai, A. H. Castro Neto, R. Shankar, and D. K. Campbell,
658: {\it ``Renormalization Group Approach to Strong-Coupled Superconductors''},
659: to appear in Phys.Rev.B.
660:
661: \bibitem{shankar}R.~Shankar, Rev.~Mod.~Phys. {\bf 68}, 129 (1994).
662:
663: \bibitem{shen}A. Lanzara, {\it et al.},
664: %P. V. Bogdanov, X. J. Zhou, S. A. Kellar,
665: % D. L. Feng, E. D. Lu, T. Yoshida, H. Eisaki, A. Fujimori, K. Kishio,
666: % J. I. Shimoyama, T. Noda, S. Uchida, Z. Hussain, and Z. X. Shen,
667: Nature {\bf
668: 412}, 510 (2001); X. J. Zhou, {\it et al.},
669: %J. Shi, T. Yoshida, T. Cuk, W. L. Yang,
670: % V. Brouet, J. Nakamura, N. Mannella, S. Komiya, Y. Ando, F. Zhou,
671: % W. X. Ti, J. W. Xiong, Z. X. Zhao, T. Sasagawa, T. Kakeshita, H. Eisaki,
672: % S. Uchida, A. Fujimori, Z. Zhang, E. W. Plummer, R. B. Laughlin,
673: % Z. Hussain, and Z.-X. Shen,
674: {\it ``Multiple Bosonic Mode Coupling in
675: Electron Self-Energy of (La$_{2-x}$Sr$_x$)CuO$_4$''}, cond-mat/0405130.
676:
677: \bibitem{organics}R.~H.~McKenzie, Science {\bf 278}, 820 (1997).
678:
679: \bibitem{mgb2}H.~J.~Choi,
680: M.~L.~Cohen, and S.~G.~Louie,
681: Physica {\bf C385}, 66 (2003).
682:
683: \bibitem{millis}A.~Deppeler,~and~A.~J.~Millis,~Phys.~Rev.~B {\bf 65}, 224301
684: (2002).
685:
686: \bibitem{c60}J.~E.~Han,
687: %{\it et al.},
688: O.~Gunnarsson, and V.~H.~Crespi,
689: Phys.~Rev.~Lett. {\bf 90}, 167006 (2003).
690:
691: %\bibitem{polchinski}J.~Polchinski, {\it ``Effective field theory and the
692: % Fermi surface''}, 1992 TASI Proceedings, edited by J.~Harvey and
693: % J.~Polchinski (World Scientific, Singapore, 1993).
694:
695: %\bibitem{next}S.-W.~Tsai {\it et al}., unpublished.
696:
697: \bibitem{eliashberg}G. M. Eliashberg, Zh. Eksp. Teor. Fiz. {\bf 38}, 966
698: (1960) [Sov. Phys. JETP {\bf 11}, 696 (1960)].
699:
700: \bibitem{migdal}A.~B.~Migdal, Zh.~Eksp.~Teor.~Fiz. {\bf 34}, 1438 (1958)
701: [Sov.~Phys.~JETP {\bf 7}, 996 (1958)].
702:
703: %\bibitem{hooft}G.~t`Hooft, Nucl.~Phys. {\bf B 72}, 461 (1974).
704:
705: \bibitem{carbotte}See, J.~P.~Carbotte, Rev.~Mod.~Phys. {\bf 62}, 1027 (1990),
706: and references therein.
707:
708: \bibitem{mc}W.~L.~McMillan, Phys.~Rev. {\bf 167}, 331 (1968).
709:
710: \bibitem{allen}P.~B.~Allen, and R.~C.~Dynes, Phys.~Rev.~B {\bf 12}, 905 (1975).
711:
712: %\bibitem{cdw}Our methods can be used to study the competition between charge
713: %density wave and superconductivity in the strong coupling regime
714: %\cite{next}, something that cannot easily be achieved in a
715: %mean-field approach.
716:
717: %\bibitem{det}C. S. Owen and D. J. Scalapino, Physica {\bf 55}, 691 (1971);
718: % G. Bergmann and D. Rainer, Z. Phys. {\bf 263}, 59 (1973).
719:
720: %\bibitem{rowell}
721: %W. L. McMillan and J. M. Rowell, Phys. Rev. Lett. {\bf 14},
722: % 108 (1965),
723: %W. L. McMillan and J. M. Rowell, in {\it ``Superconductivity''}, vol. 1,
724: %R.~D. Parks, ed. (Dekker, New York, 1969).
725:
726: %\bibitem{grimvall}G.~Grimvall, {\it ``The Electron-Phonon Interaction in
727: % Metals''} (North-Holland, 1981).
728:
729: %\bibitem{leavens}C.~R.~Leavens, and J.~P.~Carbotte, Can.~J.~Phys. {\bf 49},
730: % 724 (1971).
731:
732: \bibitem{Morel} P. Morel and P. W. Anderson, Phys. Rev. {\bf 125}, 1263
733: (1962).
734:
735: \bibitem{zanchi}D.~Zanchi, and H.~J.~Schulz, Phys.~Rev. B {\bf 61}, 13609
736: (2000);
737: \bibitem{halboth}C.~J. Halboth and W.~Metzner, Phys.~Rev.~B {\bf 61}, 7364
738: (2000);
739: \bibitem{marston}S.-W.~Tsai and J.~B. Marston, Can.~J.~Phys. {\bf 79} 1463
740: (2001).
741:
742: \end{thebibliography}
743:
744:
745:
746: \end{document}
747:
748: