cond-mat0203068/wf.tex
1: \documentclass[draft,prb,twocolumn,showpacs,nobibnotes]{revtex4}
2: 
3: \usepackage{bm}
4: \usepackage[dvips,final]{graphics}
5: 
6: \begin{document}
7: 
8: \title{Violation of the Wiedemann-Franz Law in a Large-N Solution of the t-J Model} 
9: 
10: \author{A. Houghton}
11: \author{S. Lee}
12: \author{J. B. Marston} 
13: \affiliation{Department of Physics, Brown University, Providence, RI
14: 02912-1843}
15: 
16: \date{\today}  
17: 
18: \begin{abstract}
19: We show that the Wiedemann-Franz law, which holds for Landau Fermi liquids, 
20: breaks down in a large-n treatment of the t-J model.  
21: The calculated ratio of the in-plane 
22: thermal and electrical conductivities agrees quantitatively with 
23: experiments on the normal state of the electron-doped Pr$_{2-x}$Ce$_x$CuO$_4$ ($x = 0.15$) 
24: cuprate superconductor. 
25: The violation of the Wiedemann-Franz law in the uniform phase contrasts with 
26: other properties of the phase that are Fermi liquid like.
27: \end{abstract}
28: 
29: \pacs{71.10.-w, 71.10.Hf, 71.27.+a, 72.10.-d}
30: 
31: \maketitle
32: 
33: A recent experiment that measured the electrical and thermal conductivities of
34: the copper-oxide superconductor Pr$_{2-x}$Ce$_x$CuO$_4$ in its normal state 
35: found striking deviations from the Wiedemann-Franz law\cite{hill}.  Simply
36: stated, the Wiedemann-Franz law says that fermion quasiparticles transport
37: both electrical and heat currents, with the ratio of the 
38: heat conductivity $\kappa$ to the electrical conductivity $\sigma$ given by:
39: \begin{equation}
40: {{\kappa}\over{\sigma T}} = {{\pi^2}\over{3}}~ \bigg{(} {{k_B}\over{e}} \bigg{)}^2
41: \end{equation}
42: Ordinary Landau Fermi liquids respect the Wiedemann-Franz law, so deviations from
43: it indicate the presence of non-Fermi liquid physics.  
44: The results of Ref.~\onlinecite{hill}
45: are therefore broadly consistent with other experimental 
46: evidence that points to non-Fermi 
47: liquid behavior in the cuprate phase diagram. 
48:  
49: One possible interpretation of the breakdown of the Wiedemann-Franz law 
50: is that the quasiparticle fractionalizes into separate carriers of charge
51: and spin.  To see what effects such fractionalization might have, 
52: consider first heat transport in a 
53: system of weakly-interacting electrons, which could be viewed as a crude approximation
54: to electrons in the highly overdoped region of the cuprate phase diagram.  
55: The electrons transport both charge and heat. The electrical conductivity is given by 
56: the Drude formula $\sigma = n e^2 \tau / m$ and the thermal conductivity $\kappa \propto T$ because
57: while each quasiparticle carries fixed charge $e$, it only carries an
58: energy of order the temperature.  Next consider a model for the undoped cuprates: the N\'eel 
59: ordered antiferromagnetic insulator with zero  electrical conductivity.
60: Phonons and spinwaves transport heat, and as both excitations are bosonic in character 
61: with linear dispersions at low energy, each contributes 
62: similarly, yielding a thermal conductivity $\kappa \propto T^3$.  
63: 
64: Leaving aside these ordinary states of matter, consider the case in which the spins in 
65: the insulator, instead of ordering, fractionalize into spinons with an extended 
66: Fermi surface and a non-zero density of states\cite{PWA,BZA}.  
67: Now spins transport heat in much the same way as charges do
68: in the non-interacting electron system, with $\kappa \propto T$.  The Lorenz ratio 
69: is infinite, and remains significantly larger than one upon doping with holes or electrons.    
70: 
71: Just this scenario is predicted in a large-n treatment of the t-J model on the 
72: square lattice.  We follow the approach of Refs.~\onlinecite{AM,MA,Ted}.    
73: Implementing the single-occupancy constraint by introducing slave-boson operators
74: $b_i$, the t-J model may be written: 
75: \begin{eqnarray}
76: H &=& -t~ \sum_{<i,j>}~ (c^{\dagger \alpha}_i b_i c_{j \alpha} b^\dagger_j + H.c.)
77: \nonumber \\
78: &+& J~ \sum_{<i,j>}~ (\vec{S}_i \cdot \vec{S}_j - \frac{1}{4} n_i n_j)
79: \nonumber \\
80: &+& {{1}\over{2}}~ \sum_{i \neq j}~ V(|\vec{r}_i - \vec{r}_j|)~ n_i n_j \ .
81: \label{tJ}
82: \end{eqnarray}
83: Number and spin operators are related to the electron creation and annihilation 
84: operators by $n_i \equiv c^{\dagger \alpha}_i c_{i \alpha}$ and
85: $\vec{S}_i \equiv (1/2) c^{\dagger \alpha}_i \vec{\sigma}_\alpha^\beta c_{i \beta}$
86: where there is an implicit sum over pairs of raised and lowered Greek indices.
87: The single-occupancy constraint is now holonomic, 
88: $b^\dagger_i b_i + c^{\dagger \alpha}_i c_{i \alpha} = 1$,
89: with the physical meaning that only a hole, or a single electron, may occupy each site
90: of the lattice.  (In the following 
91: we generally refer to hole doping with the understanding
92: that our calculations apply equally well to electron doped systems.) 
93: Included in Eq.~\ref{tJ} is the off-site Coulomb repulsion $V(r)$; however, in the uniform
94: and staggered flux phases discussed below it plays no role other than to contribute an 
95: additive constant to the total energy. 
96: 
97: Because spin-exchange involves no net flow of charge, electrical current 
98: only arises from the hopping term.  The continuity equation relates the time
99: rate of change of the occupancy on a given site to the lattice divergence of the 
100: current flowing into the site:
101: \begin{equation}
102: e * {{d n_j(t)}\over{dt}} = -i * e * [n_j,~ H] 
103: = - \sum_{\hat{e} = \hat{x}, \hat{y}} {{j^e_{j, j + \hat{e}} - j^e_{j - \hat{e}, j}}\over{a}}  
104: \end{equation}
105: where $a$ is the lattice spacing between copper atoms.  Thus
106: \begin{equation}
107: j^e_{j, j + \hat{e}} = -i e * t * a * (c^{\dagger \alpha}_{j} b_j 
108: c_{j + \hat{e} \alpha} b^\dagger_{j + \hat{e}} - H.c.)
109: \end{equation}
110: is the electric current flowing from site $j$ into site $j+\hat{e}$ along the link connecting
111: the two sites.  We emphasize that neither the spin-spin exchange interaction $J$ 
112: nor the Coulomb interaction $V(r)$ appear in the expression
113: for the electrical current.  This result, which is a direct consequence 
114: of the gauge invariance of the spin-exchange and Coulomb interactions 
115: in our microscopic calculation, contrasts with that obtained recently\cite{yang} 
116: within the more phenomenological ``d-density wave'' picture\cite{nayak}.  
117: 
118: The heat current $j^q_{j, j+\hat{e}}$ can be found by taking the time derivative of the 
119: Hamiltonian density:
120: \begin{equation}
121: {{d h_j(t)}\over{dt}} = 
122: - \sum_{\hat{e} = \hat{x}, \hat{y}} {{j^q_{j, j+\hat{e}} - j^q_{j - \hat{e}, j}}\over{a}}  
123: \end{equation}
124: where 
125: \begin{equation}
126: h_j = \sum_{i = j \pm \hat{x}, \hat{y}}~ (-t~ c^{\dagger \alpha}_i b_i c_{j \alpha} b^\dagger_j + H.c.)
127: + J~ \vec{S}_i \cdot \vec{S}_j  
128: \label{ham-density}
129: \end{equation}
130: and the sum is only over the four sites $i$ that are nearest-neighbors of site $j$. 
131: We have dropped the $n_i n_j$ interaction terms.  These do not contribute to
132: the DC thermal conductivity because $\langle n_i \rangle$ remains unchanged in 
133: the presence of currents.  However, in contrast to the electrical current, 
134: the heat current has contributions both from hopping, and from spin-exchange:
135: \begin{eqnarray}
136: j^q_{j, j+\hat{e}} &=& t * a * [ c^{\dagger \alpha}_{j} b_j 
137: \partial_t (c_{j + \hat{e} \alpha} b^\dagger_{j + \hat{e}}) + H.c.]
138: \nonumber \\
139: &+& {{J * a}\over{2}} * [ c^{\dagger \alpha}_j (\partial_t c_{j + \hat{e} \alpha}) 
140: c^{\dagger \beta}_{j + \hat{e}} c_{j \beta} 
141: \nonumber \\
142: &+& c^{\dagger \alpha}_j c_{j + \hat{e} \alpha} 
143: (\partial_t c^{\dagger \beta}_{j + \hat{e}}) c_{j \beta} ]\ . 
144: \label{jq}
145: \end{eqnarray}
146: 
147: The model generalizes from the physical case of $n = 2$ (up and down spins) 
148: to arbitrary (even integer) values of $n$ by letting the Greek indices run over $1, \cdots, n$.
149: In the large-n limit the spin-spin interaction factorizes in the particle-hole channel.
150: Formally this factorization is implemented via a Hubbard-Stratonovich transformation 
151: within the functional integral approach. Complex-valued mean-fields along the bonds
152: are introduced:
153: \begin{equation}
154: \chi_{i j} = \frac{J}{n} \langle c^{\dagger \alpha}_i c_{j \alpha} \rangle\ .
155: \end{equation}
156: The $\chi_{ij}$ fields function as the order parameter, and as they are spin-rotation 
157: invariant, there is no possibility
158: of N\'eel or other spin order, and the mean-field Hamiltonian respects spin-rotational symmetry.
159: Furthermore, at sufficiently low temperatures the (holon) boson
160: fields condense\cite{condense}, 
161: and we may make the replacement $b_i = b^\dagger_i = \sqrt{x}$ where
162: $x$ is the hole density.  The mean-field Hamiltonian, which is exact in the 
163: $n \rightarrow \infty$ limit, then takes the form: 
164: \begin{equation}
165: H_{MF} = \sum_{<i,j>}~ [ {{n}\over{J}}~ |\chi_{i j}|^2 -
166: (t~ x + \chi_{i j}) (c^{\dagger \alpha}_j c_{i \alpha} + H.c.) ]\ .
167: \label{H_mf}
168: \end{equation}
169: For parameters appropriate to the cuprates, $t = 0.44$ eV (following
170: Hybertsen {et al.}\cite{hybertsen}) and $J = 0.13$ eV (obtained by
171: Singh {\it et al.}\cite{singh}),
172: the minimum energy configuration has $\chi_{ij}$ both real and constant 
173: when the doping exceeds $x > 0.12$.  There are no broken symmetries in this 
174: ``uniform'' phase.  Upon suppressing dimerization with the addition of a
175: biquadratic spin exchange interaction\cite{MA,Vojta}, 
176: a staggered flux (SF) phase with counter-circulating
177: orbital currents\cite{AM,MA,Ted,john} occurs for $x < 0.12$, that is, in the
178: underdoped region of the phase diagram. (The biquadratic interaction which simultaneously
179: exchanges four fermions disappears in the 
180: physical $n = 2$ limit of up and down spins,
181: and thus does not alter the physics of the t-J model\cite{MA}.)
182: More realistic models include non-zero next- and third-nearest neighbor 
183: hopping amplitudes\cite{OKA} but these terms do not qualitatively affect 
184: the phase diagram or transport behavior.
185: 
186: The expressions for the two currents simplify in the large-n limit.  As the boson operators
187: may be replaced by the c-number $\sqrt{x}$, the electrical current becomes: 
188: \begin{equation}
189: j^e_{j, j+\hat{e}} = -i e * t * x * a * (c^{\dagger \alpha}_{j} c_{j + \hat{e} \alpha} - H.c.) \ .
190: \label{je-mf}
191: \end{equation}
192: Upon further replacing the fermion bilinear operator $c^{\dagger \alpha}_i c_{j \alpha}$ 
193: with ${{n}\over{J}}~ \chi$, the heat current also simplifies\cite{why}: 
194: \begin{equation}
195: j^q_{j, j+\hat{e}} = (t * x + \chi) * a * (c^{\dagger \alpha}_{j} \partial_t c_{j + \hat{e} \alpha} + H.c.)\ .
196: \label{jq-mf}
197: \end{equation}
198: As the heat current differs only by the $(t * x + \chi)$ prefactor from that of a 
199: non-interacting tight-binding system, in the low-temperature limit 
200: the Lorenz ratio for in-plane transport is simply:
201: \begin{equation}
202: {{\kappa}\over{\sigma T}} = {{\pi^2}\over{3}}~ \bigg{(} {{k_B}\over{e}} \bigg{)}^2 
203: * \bigg{(} {{t x + \chi}\over{t x}} \bigg{)}^2\ . 
204: \label{wf}
205: \end{equation}
206: Thus for any $\chi \neq 0$ the ratio differs from unity, indicating a breakdown of 
207: Fermi liquid theory.  Note that all details of the scattering mechanisms cancel out in
208: the ratio.  Direct calculation of the two conductivities in linear response shows that
209: the integrals over momentum have identical form. Only the frequency integrals differ; 
210: hence for static impurities in the weak-scattering limit 
211: the Lorenz ratio is given by Eq.~\ref{wf}.  We note that while the order
212: parameter $\chi$ is perturbed by the application of external electric fields
213: or thermal gradients, the perturbation in $\chi$ does not alter the DC response\cite{caroli}. 
214: 
215: In the SF phase the $\chi$-fields are complex numbers\cite{AM,MA,Ted}, 
216: and the prefactor $(t x + \chi)^2$ should be replaced by $|t x + \chi |^2$.  
217: Specifically, for $\chi = |\chi| \exp(i \theta)$, with phase $\theta$, 
218: the Lorenz ratio generalizes to\cite{in_prep}: 
219: \begin{equation}
220: {{L}\over{L_0}} =  
221: {{(t x)^2  + |\chi|^2 + 2 t x |\chi| \cos(\theta)}\over{(t x)^2}}  
222: \label{wf-sf}
223: \end{equation}
224: where $L \equiv \kappa / (\sigma T)$ 
225: and  $L_0 \equiv (\pi^2/3) (k_B /e)^2 \approx 2.45 \times 10^{-8} {\rm W \Omega K^{-2}}$ 
226: is the Lorenz number.
227: 
228: We turn now to a comparison of our predictions, Eqs.~\ref{wf} and \ref{wf-sf}, 
229: with experiments.  In Fig.~\ref{fig:ratio-x}
230: we plot the Lorenz ratio as a function of the doping.  As expected, the 
231: ratio diverges as the insulating limit $x \rightarrow 0$ is approached because 
232: the spinons transport only heat, not charge.  In the opposite limit of large doping 
233: $\chi \rightarrow 0$, and the ratio approaches unity. Landau Fermi liquid  
234: theory is recovered in the dilute limit of widely spaced electrons.  
235: We emphasize that the uniform phase with
236: $\chi_{ij}$ constant and real does not break any symmetries.  
237: It exhibits weak pseudogap behavior 
238: because, according to the mean-field equations, $|\chi|$ increases
239: slightly in size at low temperatures, which in turn increases 
240: the quasiparticle bandwidth (see Eq.~\ref{H_mf}) 
241: and decreases the density of states (DOS)\cite{chung}. 
242: For example, at a hole concentration of 
243: $x = 0.15$, $|\chi| = 0.024$ eV at $T = 500$ K rising
244: to $|\chi| = 0.026$ eV at zero temperature; consequently the DOS drops by 2\%.  
245: This contrasts with the strong pseudogap behavior of the SF phase which has a gap along 
246: most of the Fermi surface and which breaks time reversal invariance.  
247: In either phase, however, the fermionic quasiparticles are non-interacting in
248: the $n \rightarrow \infty$ limit and hence behave as long-lived Landau quasiparticles 
249: such as those found in ordinary Fermi liquids.  
250: 
251: We note that the Wiedemann-Franz law is strongly violated in s-wave superconducting
252: states because while Cooper pairs carry charge, the condensate has no entropy. 
253: In a d-wave superconductor, quasiparticle excitations at the nodes 
254: result in a modified Wiedemann-Franz law\cite{durst}.  
255: The violation that we find occurs in the normal state, and is a consequence of 
256: the spin-charge separation inherent in the large-n solution of the t-J model,
257: and not of any incipient superconducting tendencies. 
258: 
259: \begin{figure}[h,t]
260: \resizebox{8cm}{!}{\includegraphics{ratio-x.eps}}
261: \caption{The Lorenz ratio (Eqs.~\ref{wf} and \ref{wf-sf}) as
262: a function of the doping for $t = 0.44$ eV, $J = 0.13$ eV.
263: The ratio approaches unity in the dilute limit, $x \rightarrow 1$.}
264: \label{fig:ratio-x}
265: \end{figure}
266: 
267: \begin{figure}[h,t]
268: \resizebox{8cm}{!}{\includegraphics{ratio-T.eps}}
269: \caption{The Lorenz ratio (Eqs.~\ref{wf} and \ref{wf-sf}) as a function of temperature
270: at dopings $x = 0.06$, $0.15$, and $0.26$ 
271: for $t = 0.44$ eV, $J = 0.13$ eV.  In the underdoped $x = 0.06$ case
272: there is a transition to the staggered flux phase at a temperature of
273: approximately $300$ K.} 
274: \label{fig:ratio-T}
275: \end{figure}
276: 
277: In Fig.~\ref{fig:ratio-T} we plot the temperature dependence of the 
278: Lorenz ratio for three dopings at which transport experiments have been 
279: conducted: $x = 0.06$, $0.15$, and $0.26$. For a single crystal of the
280: La$_{2-x}$Sr$_x$CuO$_4$ material with hole doping $x = 0.06$, the resistivity
281: was measured upon suppressing the superconductivity by application of a 
282: 18 T magnetic field along the c-axis\cite{takeya}.  The thermal conductivity
283: was, however, measured in the superconducting state, so it is not possible to 
284: extract a real Lorenz ratio.  Nevertheless it is intriguing that 
285: $L / L_0 \approx 5$ at low temperatures, based on the numbers appearing in the
286: inset to Fig. 3 of Ref.~\onlinecite{takeya}.  This compares 
287: reasonably well with the theoretical value of $4.1$ seen in Fig.~\ref{fig:ratio-T}.
288:     
289: At optimal doping, experimentally available magnetic fields can only eliminate
290: superconductivity in electron-doped compounds.  In Ref.~\onlinecite{hill} a 13 T field
291: was applied to Pr$_{2-x}$Ce$_x$CuO$_4$ at $x = 0.15$ to access the normal state. 
292: The measured ratio of $L / L_0 \approx 2$ found at temperatures
293: above $0.3$ K is again in reasonable quantitative agreement with 
294: the theoretical value of $1.95$.  At temperatures below $0.18$ K, however,
295: the experimentally determined ratio drops rapidly below one.  We have no explanation for
296: the observed behavior at the lowest temperatures\cite{si}.
297: 
298: Finally, in the highly overdoped regime Proust {\it et al.}\cite{proust} 
299: have studied the Tl$_2$Ba$_2$CuO$_{6+\delta}$ material at a hole concentration
300: of $x \approx 0.26$.  Superconductivity was suppressed in a 13 T field, and
301: $L / L_0 = 0.99 \pm 0.01$ in good agreement with the
302: Wiedemann-Franz law for Fermi liquids.  The theoretical value of the ratio is $1.5$.
303: We speculate that the persistence of non-Fermi liquid behavior at large doping
304: in the mean-field theory is an artifact of the large-n approximation.  Finite-n 
305: corrections could possibly restore Fermi liquid behavior in the overdoped region.  
306: At large doping $|\chi|$ is small compared to the effective hole hopping amplitude $t x$,
307: so fluctuations in $\chi$ 
308: may be expected to be relatively more important than at low doping.
309: 
310: In summary we have shown that the Wiedemann-Franz law is 
311: violated in a mean-field treatment of the t-J model. 
312: Our analysis, which holds for weak scattering, is exact in the 
313: $n \rightarrow \infty$ limit.  The Lorenz ratio is significantly larger than one
314: both in the uniform phase ($x > 0.12$) and in the SF phase ($x < 0.12$).  
315: The theoretical prediction is in reasonably good quantitative agreement
316: with existing experimental measurements on the cuprate materials.  
317: 
318: {\it Note added: } After this work was completed a paper by Kim and Carbotte (KC)
319: appeared\cite{kim} that examined the Wiedemann-Franz law within the context of
320: the phenomenological d-density wave picture.  There are several differences between 
321: their work and ours.  The main difference is that we study both the 
322: uniform phase which has no broken symmetries, and the SF or d-density phase with
323: time-reversal breaking counter-circulating currents.
324: We find that the Lorenz ratio is significantly larger than one in both phases.
325: Furthermore, at low temperatures KC find only small deviations 
326: from the Wiedemann-Franz law.  This is due in part to the fact that their d-density 
327: order parameter (the analog of our $\chi_{ij}$) was chosen to be purely imaginary 
328: (equivalent to setting $\theta = \pi/2$ in our Eq.~\ref{wf-sf}) and also because  
329: their kinetic energy is not rescaled by the slave-boson doping factor, $x$, as it
330: is in our microscopic analysis of the t-J model.  KC also find a 
331: large temperature variation in the Lorenz ratio for the case of
332: strong scattering because the quasiparticle lifetime has a strong frequency dependence.
333: 
334: %%%%%%%%%%%%%%%%%%%%%%
335: We thank John Fj{\ae}restad and Louis Taillefer for helpful comments.
336: This work was supported in part by the NSF under grant No. DMR-9712391.
337: 
338: \begin{thebibliography}{}
339:   
340: \bibitem{hill} R. W. Hill, C. Proust, L. Taillefer, P. Fournier, and R. L. Greene,
341: Nature {\bf 414}, 711 (2001).
342: 
343: \bibitem{PWA} P. W. Anderson, Science {\bf 235}, 1196 (1987).
344: 
345: \bibitem{BZA} G. Baskaran, Z. Zou, and P. W. Anderson, Solid State Comm. {\bf 63},
346: 973 (1987).
347: 
348: \bibitem{AM} I. Affleck and J. B. Marston, Phys. Rev. B{\bf 37}, 3774 (1988). 
349: 
350: \bibitem{MA} J. B. Marston and I. Affleck, Phys. Rev. B{\bf 39}, 11538 (1989).
351: 
352: \bibitem{Ted} T. C. Hsu, J. B. Marston, and I. Affleck, 
353: Phys. Rev. B {\bf 43}, 2866 (1991).
354:   
355: \bibitem{yang} X. Yang and C. Nayak, cond-mat/0108407.   
356: 
357: \bibitem{nayak} S. Chakravarty, R. B. Laughlin, D. K. Morr, and C. Nayak, 
358: Phys. Rev. B {\bf 63}, 094503 (2001).
359: 
360: \bibitem{condense} Strictly speaking bosons cannot condense in a purely
361: two-dimensional system at any non-zero temperature.  In practice, however, weak
362: interlayer tunneling stabilizes the condensate.  As the temperature scale for
363: condensation is determined primarily by the in-plane hopping amplitude $t$, 
364: and since $t \gg k_B T$, we
365: assume that the bosons remain condensed at the temperatures relevant for the
366: experiments we consider here. 
367: 
368: \bibitem{hybertsen} M. S. Hybertsen, E. B. Stechel, M. Schluter, and 
369: D. R. Jennison, Phys. Rev. B {\bf 41}, 11068 (1990).
370: 
371: \bibitem{singh} R. R. P. Singh, P. A. Fleury, K. B. Lyons, and P. E. Sulewski,
372: Phys. Rev. Lett. {\bf 62}, 2736 (1989).
373: 
374: \bibitem{Vojta} M. Vojta, Y. Zhang, and S. Sachdev, Phys. Rev. B{\bf 62} 6721 (2000).
375: 
376: \bibitem{john} The SF phase was recently shown to occur in a half-filled
377: two-leg ladder.  See: J. O. Fj{\ae}restad and J. B. Marston, cond-mat/0107094 (to
378: appear in Phys. Rev. B) and J. B. Marston, J. O. Fj{\ae}restad, and A. Sudb{\o}, 
379: cond-mat/0202188. 
380: 
381: \bibitem{OKA} O. K. Andersen, A. I. Liechtenstein, O. Jepsen, and F. Paulsen,
382: J. Phys. Chem. Solids. {\bf 56}, 1573 (1995).
383: 
384: \bibitem{why} It is straightforward to verify that the heat 
385: current-current correlation function is the same in the large-n limit 
386: for either form of the current operator, Eq.~\ref{jq} or Eq.~\ref{jq-mf}.  
387: 
388: \bibitem{caroli} The BCS pairing order parameter $\Delta$ is likewise perturbed in a 
389: superconductor, but in the absence of an external magnetic field, the perturbation 
390: has no effect on the low-frequency conductivities. 
391: See: C. Caroli and K. Maki, Phys. Rev. {\bf 159}, 306 (1967); {\bf 159}, 316 (1967).
392: 
393: \bibitem{in_prep} A. Houghton, S. Lee, and J. B. Marston, in preparation.
394: 
395: \bibitem{chung} C. H. Chung, J. B. Marston, and R. H. McKenzie, 
396: J. Phys. C. {\bf 13}, 5159 (2001). 
397: 
398: \bibitem{durst} A. C. Durst and P. A. Lee, Phys. Rev. B {\bf 62}, 1270 (2000). 
399: 
400: \bibitem{takeya} J. Takeya, Y. Ando, S. Komiya, and X. F. Sun, Phys. Rev. Lett.
401: {\bf 88}, 077001 (2002).
402: 
403: \bibitem{si} For one possible explanation see: Q. Si, Physica C {\bf 364-365},
404: 9 (2001).
405: 
406: \bibitem{proust} C. Proust, E. Boaknin, R. W. Hill, L. Taillefer, and A. P. Mackenzie,
407: cond-mat/0202101.
408: 
409: \bibitem{kim} W. Kim and J. P. Carbotte, cond-mat/0202514.
410: 
411: \end{thebibliography}
412: \end{document}
413: