1: %
2: %This paper has 3 figures:
3: %
4: %Fig.\ \ref{fig:fisk} contained in postscript file: allen_fig1.ps
5: %Fig.\ \ref{fig:nikolic} contained in postscript file: allen_fig2.ps
6: %Fig.\ \ref{fig:various} contained in postscript file: allen_fig3.ps
7: %
8: %\documentstyle[epsf,aps]{revtex}
9: %\documentstyle[psfig,aps]{revtex}
10: %\documentstyle[epsf,psfig,twocolumn,aps,prl,floats]{revtex}
11: \documentstyle[psfig,twocolumn,aps]{revtex}
12: \begin{document}
13:
14: \draft
15: \title{Metals with Small Electron Mean-Free Path: Saturation
16: {\sl versus} Escalation of Resistivity}
17:
18: \author{ Philip B. Allen}
19: \address{Department of Physics and Astronomy, State University of New York,
20: Stony Brook, New York 11794-3800}
21:
22:
23: \date{\today}
24:
25: \maketitle
26:
27: \begin{abstract}
28: \begin{center} Abstract
29: \end{center}
30: Resistivity of metals is commonly observed either to `escalate'
31: beyond the Ioffe-Regel limit (mean free path $\ell$ equal
32: to lattice constant $a$) or to `saturate' at this point.
33: It is argued that neither behavior is well-understood, and
34: that `escalation' is not necessarily more mysterious than `saturation.'
35:
36: \end{abstract}
37: \pacs{72.15.Eb, 72.15.Lh, 72.15.Rn}
38:
39: %\narrowtext
40:
41: %\section{introduction}
42:
43: During the period 1954-1986, `high temperature' superconductivity
44: occured primarily in Nb$_3$Sn and related materials with the
45: A15 crystal structure. Although peculiar resistivity $\rho(T)$
46: had been noticed \cite{Woodward} and modeled (corrected for
47: non-zero $k_B T/\epsilon_F$) \cite{Cohen}, it was not until the
48: Fisk-Lawson paper \cite{Fisk1} that the general nature of the
49: peculiarity was known. At first the peculiarity seemed to correlate
50: with high $T_c$ and therefore large electron-phonon coupling, but
51: the Fisk-Webb paper \cite{Fisk2} showed that even a low $T_c$
52: material had `saturating' resistivity when the resistivity
53: became large enough. Data from these two classic papers
54: is shown in Fig. \ref{fig:fisk}.
55: Fisk and Webb observed that the mean free path $\ell$
56: was becoming nearly as small as the lattice constant $a$. This
57: violation of the condition $\ell \gg a$ (required
58: for validity of the quasiparticle picture and the Bloch-Boltzmann
59: resistivity theory) seemed from the available data
60: to lead quite generally to
61: `saturation' of resistivity.
62: Unfortunately, correcting theory for non-zero $a/\ell$ (with
63: values of order 1) is more important, and
64: much more difficult than correcting
65: for non-zero $k_B T/\epsilon_F$ (which is usually small.)
66:
67:
68: To strengthen the case,
69: `saturation' could be provoked not only
70: by the thermal disorder of high $T$ lattice vibrations, but also
71: by static disorder, such as radiation damage \cite{Gurvitch}.
72: It could also be modelled by a `shunt-resistor model'
73: \cite{Wiesmann} $1/\rho(T)=1/\rho_{\rm Boltz}+1/\rho_{\rm sat}$.
74: If we take the saturation value $\rho_{\rm sat}$ to be the
75: same as the Boltzmann value $\rho_{\rm Boltz}$ extrapolated
76: to the Ioffe-Regel point $\ell=a$, then the `shunt-resistor'
77: formula can be written $\rho(T)=\rho_{\rm Boltz}(T)/(1+a/\ell(T))$.
78: When $a/\ell(T)$ is large and the second term dominates the
79: denominator, then $\ell(T)$ cancels from the formula and $\rho$
80: becomes independent of $T$.
81:
82: In spite of the explicit warning \cite{Fisk2} that no theory
83: of `saturation' existed, it was often assumed that the
84: case was closed. The reasoning \cite{Gurvitch2}
85: was that $\ell$ cannot sensibly
86: become smaller than $a$, so one should just use the Boltzmann
87: result with $\ell$ replaced by $a$. The fault with this argument
88: \cite{Allen}, that it denies the possibility of Anderson localization
89: under static disorder, was easy to forget. A careful
90: theoretical discussion has to mention that if $\ell$ is not
91: longer than $a$, then the concept of a mean free path has to be
92: abandoned. One has no right to take a formula which contains
93: a factor $1/\ell$ and replace it with $1/a$, because the theoretical
94: framework which provided the formula has already been destroyed.
95:
96: \par
97: \begin{figure}[t]
98: \centerline{\psfig{figure=allen_fig1.ps,height=5.0in,width=3.4in,angle=0}}
99: \caption{Upper panel: resistivity from Ref.
100: \protect\cite{Fisk1} normalized to the value at 273 K.
101: The `bulge' in Rh$_{17}$S$_{15}$ was originally
102: correlated with the higher superconducting
103: transition. Lower panel: absolute resistivity from Ref.
104: \protect\cite{Fisk2}
105: of a good superconductor (Nb$_3$Sn, $T_c$=18 K) saturates
106: at the same final resistivity value as a poor superconductor (Nb$_3$Sb).
107: This shows that `saturation' correlates with
108: magnitude of resistivity rather than $T_c$
109: or strength of electron-phonon coupling.}
110: \label{fig:fisk}
111: \end{figure}
112: \par
113:
114: In recent years it became possible to calculate resistivity
115: for highly disordered metals (with non-interacting electrons)
116: by the Kubo formula or an equivalent
117: Landauer formulation. The results \cite{Nikolic} for the most familiar model
118: are shown in Fig. \ref{fig:nikolic}. Rather than saturating,
119: the resistivity escalates. In the center of the band formed
120: by a single $s$ orbital on a simple cubic lattice, Boltzmann
121: theory (in Born approximation, at $T=0$) gives \cite{Nikolic}
122: $\rho_{\rm Boltz}=(\hbar a/e^2)(6.94a/\ell)$, and
123: $a/\ell=0.0283(W/t)^2$, where $t$ is the hopping parameter and
124: $\pm W/2$ is the interval of the random on-site disorder potential.
125: Numerically converged results for large disorder are plotted
126: versus $a/\ell$, with $a$ set arbitrarily
127: at $3\AA$. Of course, $\ell$ is meaningless over
128: most of the range shown, and the horizontal axis is just
129: an alternate parameterization of $(W/t)^2$. There are three
130: things to notice: (1) Boltzmann theory continues to give
131: a good answer even at $a/\ell=1$ where it is invalid; (2)
132: there is a very large interval between the Ioffe-Regel point
133: $a/\ell=1$ and the point where states become all localized,
134: $(a/\ell)_c=7.71$ \cite{Slevin}; (3) rather than saturating,
135: the resistivity escalates toward the scaling region where it
136: diverges as $((a/\ell)_c -a/\ell)^{-\nu}$ with critical
137: exponent $\nu\approx 1.57$.
138:
139: \par
140: \begin{figure}[t]
141: \centerline{\psfig{figure=allen_fig2.ps,height=3.0in,width=3.4in,angle=0}}
142: \caption{Resistivity of a half-filled
143: $s$-band on a cubic lattice with random on-site
144: potential fluctuations, plotted {\sl versus}
145: $a/\ell$. Here $\ell$ is not a physical
146: mean free path, but instead the value given by Boltzmann
147: theory, extended beyond is range of validity.
148: The points were calculated in ref.
149: \protect\cite{Nikolic}.
150: The solid line is the Bloch-Boltzmann prediction in Born approximation.
151: The dashed line is the expected saturated form, $\rho_{\rm Boltz}/(1+a/\ell)$,
152: and the dot-dashed line is a naive scaling formula
153: $\rho_{\rm Boltz}/(1-\ell_c/\ell)$, with $\ell_c=a/7.71$, the observed
154: point of the Anderson transition.}
155: \label{fig:nikolic}
156: \end{figure}
157: \par
158:
159: One of the first indications that `saturation' was not ubiquitous
160: was the observation by Gurvitch and Fiory \cite{Gurvitch3} that
161: resistivity of high $T_c$ superconductors seems to fail to saturate.
162: The failure of `saturation', which I will call `escalation',
163: is probably not rare after all. Emery and Kivelson \cite{Emery}
164: call such metals `bad metals.'
165: Fig. \ref{fig:various} shows resistivities
166: of several metals. Alkali-doped C$_{60}$ saturates (if at all) \cite{Hou}
167: only for very short $\ell$) (less than $1\AA$.) Among oxide metals,
168: saturation seems rare for simpler compounds like Sr$_2$RuO$_4$
169: \cite{Pavuna,Tyler}, and may occur at very large resistivity values
170: for certain complex oxides like La$_4$Ru$_6$O$_{19}$ \cite{Khalifah}.
171: Quite likely the mean free path in this last case is also very small.
172:
173:
174: \par
175: \begin{figure}[t]
176: \centerline{\psfig{figure=allen_fig3.ps,height=3.0in,width=3.4in,angle=0}}
177: \caption{Resistivity {\it versus} temperature for various metals.
178: Data are from Ref. \protect\cite{Hou} (K$_3$C$_{60}$);
179: Ref. \protect\cite{Pavuna} (Sr$_2$RuO$_4$);
180: Ref. \protect\cite{Khalifah} (La$_4$Ru$_6$O$_{19}$). The error bars
181: represent the error estimated in the geometric normalization for
182: K$_3$C$_{60}$.}
183: \label{fig:various}
184: \end{figure}
185:
186: There are exotic theories which attribute \cite{Kiv} `bad metal' behavior
187: to quasi-1d electron conduction on fluctuating stripes. One can naively
188: model this by meandering stripes containing metallic electrons whose
189: resistivity is not high.
190: Dilution of the metallic stripes in an intervening non-metallic
191: phase causes an apparent high resistivity and short mean-free path.
192: Arnason {\it et al.} \cite{Arnason} reported
193: an interesting experiment where a static realization
194: of this geometry was intentionally created.
195:
196: Several recent theories \cite{Millis,Merino,Gunnarsson1,Gunnarsson2}
197: attempt to find or explain saturation behavior.
198: Millis {\sl et al.} \cite{Millis} find a tendency in the direction of saturation
199: by applying the `dynamical mean field theory' approximation to
200: a model with electron-phonon interactions and disorder, but
201: no Coulomb scattering. Merino and McKenzie \cite{Merino}
202: find a similar effect in a model with on-site (Hubbard)
203: Coulomb repulsion and no phonons. Both theories have only a single
204: band. Gunnarsson and Han \cite{Gunnarsson1} use a 3-fold degenerate
205: band to model doped C$_{60}$, and Calandra and Gunnarsson
206: \cite{Gunnarsson2} use a 5-fold degenerate band to model
207: Nb$_3$X. These interesting quantum-Monte-Carlo (QMC) calculations
208: find no saturation in the C$_{60}$ model, and saturation
209: in the Nb$_3$X model. The difference apparently is in the
210: form of electron-phonon coupling used (Coulomb interactions
211: were omitted). When phonons were coupled to on-site potential
212: as in Fig. \ref{fig:nikolic}, saturation was not found,
213: whereas when coupled to the hopping matrix element, vibrations
214: were found to cause saturating resistivity. In this latter
215: case, theoretical QMC results could be fit by Kubo theory
216: calculated under the assumption that vibrations could be treated
217: adiabatically, {\it i. e.}
218: modelled as static disorder. This approximate analysis
219: agrees with qualitative arguments I made with Chakraborty long ago
220: \cite{Chakraborty}, emphasizing the importance of allowing
221: a multi-orbital band structure.
222:
223: The conclusions to draw from this are: (1) there still is
224: no theory of saturation, although computational results are
225: giving a first outline; (2) it seems no harder to account
226: for escalation than for saturation. Therefore it is
227: perhaps an unproven guess that metals with escalating
228: resistivity are more exotic than metals
229: which show saturation. However, an exotic origin of
230: escalation has been nicely demonstrated in the case of
231: high $T_c$ cuprates. Resistivity versus doping and
232: temperature was re-examined by Ando {\it et al.} \cite {Ando}.
233: They make a very convincing case that the mechanism of
234: transport is intimately related to antiferromagnetic
235: correlations over a wide doping range. A candidate
236: picture by which this can happen is the model of
237: self-organizing stripe inhomogeneities \cite{Kiv}.
238:
239:
240: \acknowledgements
241: I thank Z. Fisk and B. Nikoli\'c for their many contributions
242: to this work, and S. A. Kivelson for very helpful comments on
243: the manuscript. This work was supported by NSF grant no. DMR-0089492.
244:
245:
246:
247:
248: %\begin{thebibliography}{99}
249: \begin{references}
250:
251: \bibitem{Woodward} D. W. Woodward and G. D. Cody,
252: Phys. Rev. {\bf 136} (1964) A166.
253:
254: \bibitem{Cohen} R. W. Cohen, G. D. Cody, and J. J. Halloran,
255: Phys. Rev. Lett. {\bf 19} (1967) 840.
256:
257: \bibitem{Fisk1} Z. Fisk and A. C. Lawson,
258: Solid State Commun. {\bf 13} (1973) 277.
259:
260: \bibitem{Fisk2} Z. Fisk and G. W. Webb,
261: Phys. Rev. Lett. {\bf 36} (1976) 1084.
262:
263: \bibitem{Gurvitch} M. Gurvitch, A. K. Ghosh, B. L. Gyorffy, H. Lutz,
264: O. F. Kammerer, J. S. Rosner, and M. Strongin,
265: Phys. Rev. Lett. {\bf 41} (1978) 1616.
266:
267: \bibitem{Wiesmann} H. Wiesmann, M. Gurvitch, H. Lutz, B. Schwartz,
268: M. Strongin, P.B. Allen and J.W. Halley,
269: Phys. Rev. Lett. {\bf 38} (1977) 782.
270:
271: \bibitem{Gurvitch2} M. Gurvitch,
272: Phys. Rev. B {\bf 24} (1981) 7404.
273:
274: \bibitem{Allen} P.B. Allen,
275: in {\sl Superconductivity in d- and f-Band Metals},
276: H. Suhl and M.B. Maple, eds.
277: (Academic Press, NY 1980) p. 291;
278: P.B. Allen,
279: in {\sl Physics of Transition Metals, Leeds, 1980}
280: (Inst. Phys. Conf. Ser. No. 55, 1981) p. 425.
281:
282:
283: \bibitem{Nikolic} B. Nikoli\'c and P. B. Allen,
284: Phys. Rev. B {\bf 63} (2000) 020201.
285:
286: \bibitem{Slevin} K. Slevin, T. Ohtsuki, and T. Kawarabayashi,
287: Phys. Rev. Lett. {\bf 84} (2000) 3915.
288:
289: \bibitem{Gurvitch3} M. Gurvitch and A. T. Fiory,
290: Phys. Rev. Lett. {\bf 59} (1987) 1337.
291:
292: \bibitem{Emery} V. J. Emery and S. A. Kivelson,
293: Phys. Rev. Lett. {\bf 74} (1995) 3253.
294:
295: \bibitem{Hou} J. G. Hou, L. Lu, V. H. Crespi, X.-D. Xiang,
296: A. Zettl, and M. L. Cohen,
297: Solid State Commun. {\bf 93} (1995) 973.
298:
299: \bibitem{Pavuna} D. Pavuna, H. Berger, and L. Forro,
300: J. Eur. Ceram. Soc {\bf 19} (1999) 1518.
301:
302: \bibitem{Tyler} A. W. Tyler, A. P. Mackenzie, S. NishiZaki,
303: and Y. Maeno,
304: Phys. Rev. B {\bf 58} (1998) 10107.
305:
306: \bibitem{Khalifah} P. Khalifah, K. D. Nelson, R. Jin, Z. Q. Mao, Y. Liu,
307: Q. Huang, X. P. A. Gao, A. P. Ramirez, and R. J. Cava,
308: Nature {\bf 411} (2001) 669.
309:
310: \bibitem{Kiv} S. A. Kivelson, E. Fradkin, and V. J. Emery,
311: Nature {\bf 393} (1998) 550.
312:
313: \bibitem{Arnason} S. B. Arnason, S. P. Herschfield, and A. F. Hebard,
314: Phys. Rev. Lett. {\bf 81} (1998) 3936.
315:
316: \bibitem{Millis} A. J. Millis, L. Hu, and S. Das Sarma,
317: Phys. Rev. Lett. {\bf 82} (1999) 2354.
318:
319: \bibitem{Merino} J. Merino, and R. H. McKenzie,
320: Phys. Rev. B {\bf 61} (2000) 7996.
321:
322: \bibitem{Gunnarsson1} O. Gunnarsson and J. E. Han,
323: Nature {\bf 405} (2000) 1027.
324:
325: \bibitem{Gunnarsson2} M. Calandra and O. Gunnarsson,
326: cond-mat/0106397.
327:
328: \bibitem{Chakraborty} B. Chakraborty and P.B. Allen,
329: Phys. Rev. Lett. {\bf 42} (1979) 736;
330: P.B. Allen and B. Chakraborty,
331: Phys. Rev. B {\bf 23} (1981) 4815.
332:
333: \bibitem{Ando} Y. Ando, A. N. Lavrov, S. Komiya,
334: K. Segawa, and X. F. Sun,
335: Phys. Rev. Lett. {\bf 87} (2001) 017001.
336:
337:
338: %\end{thebibliography}
339: \end{references}
340:
341:
342:
343: \end{document}
344: