1: \documentclass[epj]{svjour}
2: \usepackage{graphics}
3: \usepackage{psfig}
4: \newcommand{\bsim}{\mbox{\raisebox{-0.1cm}{$\;
5: \stackrel{\textstyle>}{\sim}\;$}}}
6: \newcommand{\lsim}{\mbox{\raisebox{-0.1cm}{$\;
7: \stackrel{\textstyle<}{\sim}\;$}}}
8:
9:
10: \begin{document}
11: %\psdraft
12: \title{
13: Spin susceptibility in small Fermi energy systems:
14: effects of nonmagnetic impurities}
15: \subtitle{}
16: \author{E. Cappelluti\inst{1}
17: \and C. Grimaldi\inst{2} \and L. Pietronero\inst{1,3}}
18: \institute{Dipartimento di Fisica, Universit\'{a} ``La Sapienza",
19: P.le A. Moro 2, 00185 Roma, and INFM Roma1, Italy
20: \and Ecole Polytechnique F\'ed\'erale de Lausanne, IPR-LPM,
21: CH-1015 Lausanne, Switzerland
22: \and Istituto di Acustica ``O.M. Corbino'',
23: CNR, Area di Ricerca Tor Vergata, Roma, Italy}
24: \date{\today/ \mbox{}}
25: \abstract{
26: In small Fermi energy metals, disorder
27: can deeply modify superconducting state properties
28: leading to a strong suppression of the critical temperature $T_c$.
29: In this paper, we show that
30: also normal state properties can be seriously
31: influenced by disorder when the Fermi energy $E_{\rm F}$ is
32: sufficiently small. We calculate the normal state spin
33: susceptibility $\chi$ for a
34: narrow band electron-phonon coupled metal as a function of
35: the non-magnetic impurity scattering rate $\gamma_{\rm imp}$.
36: We find that as soon as
37: $\gamma_{\rm imp}$ is comparable to $E_{\rm F}$, $\chi$
38: is strongly reduced with respect to its
39: value in the clean limit. The effects of the electron-phonon interaction
40: including the
41: nonadiabatic corrections are discussed. Our results strongly
42: suggest that the recent finding
43: on irradiated MgB$_2$ samples can be naturally
44: explained in terms of small $E_{\rm F}$ values
45: associated with the $\sigma$-bands of the boron plane,
46: sustaining therefore the hypothesis
47: that MgB$_2$ is a nonadiabatic metal.
48: \PACS{
49: {74.25.-q}{General properties; correlations
50: between physical properties in normal and superconducting states} \and
51: {71.28.+d}{Narrow-band systems; intermediate-valence solids} \and
52: {74.62.Dh}{Effects of crystal defects, doping and substitution}
53: }
54: }
55: \maketitle
56:
57:
58:
59: \section{Introduction}
60:
61:
62: Scattering from weak disorder or diluted non magnetic impurities
63: plays a marginal
64: role on many thermodynamics quantities of conventional metals.
65: Most peculiar is the absence of any reduction on the critical
66: temperature $T_c$ in conventional isotropic $s$-wave superconductors
67: as stated by the Anderson's theorem and as confirmed
68: by several experimental measurements \cite{anderson}.
69:
70: This insensitivity stands out in particular in comparison with
71: $d$-wave superconductors where the strong an\-iso\-trop\-y
72: of the order parameter
73: leads to a suppression of $T_c$ \cite{dwave}.
74: In that case for instance the reduction on
75: $T_c$ upon disorder can give
76: qualitative information of the microscopic characteristic on the pairing
77: ($d$- vs. $s$- wave symmetry, local vs. long-ranged interaction, etc \ldots)
78: \cite{flatte}.
79:
80: A conventional role of nonmagnetic impurities is recently questioned
81: in some high-$T_c$ superconductors, as MgB$_2$ and fullerene compounds.
82: In these materials a notable reduction of $T_c$ upon disorder
83: has been reported in spite of the $s$-wave symmetry of
84: both of them \cite{watson,karkin}.
85: Quite remarkable is also the reduction of the density of states (DOS)
86: inferred by NMR measurements of
87: the nuclear spin-lattice relaxation rate $T_1$
88: [$1/T_1 T \propto N_0^2$ where $N_0$ is the electronic density of
89: states at the Fermi level
90: and $T$ is the temperature].
91: In Ref. \cite{gerashenko}
92: a reduction of 62 \% of $1/T_1 T$ upon disorder
93: was reported by $^{11}$B NMR measurements
94: in contrast with the conventional theory of non magnetic impurity
95: scattering which would predict no effect of the magnetic susceptibility.
96: An additional puzzling feature is the discrepancy between
97: spin-lattice relaxation rate measurements performed
98: on $^{11}$B NMR and on $^{25}$Mg. No reduction of $1/T_1 T$ was
99: indeed observed on magnesium atoms. The authors of
100: Ref. \cite{gerashenko} speculate this difference could be related to the
101: different nature of the electronic states: magnesium atoms would mainly
102: probe the $\pi$ bands of MgB$_2$ through the hybridization of B($2p_z$)
103: orbitals with Mg(s) states, while the spin-lattice relaxation rate
104: on the boron is expected to be very sensitive to the twodimensional
105: $\sigma$ bands formed by B($2p_x 2p_y$).
106:
107: The evidences of anomalous effects of disorder and non magnetic
108: impurities in these systems prompt thus some intriguing open questions:
109: $i$) which is the origin of the suppression of $T_c$ in $s$-wave
110: systems as MgB$_2$ and fullerenes? $ii$) which is the origin of the
111: reduction of the density of states as probed by
112: spin-lattice relaxation rate $1/T_1 T$ measurements?
113: $iii$) which is the origin of the different behaviour of Mg and B
114: NMR measurements?
115:
116: The point ($i$) was previously addressed in Ref. \cite{sgp}
117: in the context of a nonadiabatic theory of superconductivity \cite{gps} where
118: impurity effects in the nonadiabatic channels were shown
119: to suppress $T_c$ even for purely isotropic $s$-wave
120: superconductors.
121: In this paper we extent our analysis to the spin susceptibility.
122: In particular
123: we show that a possible unifying explanation of all this
124: complex anomalous scenario could come from taking into account
125: in a coherent way the small Fermi energy nature of these materials.
126: This is clearly unavoidable in C$_{60}$ compounds where the narrow
127: bandwidth of the $t_{1u}$ bands (but also the $h_{u}$ bands for hole doped C$_{60}$)
128: results in a Fermi energy $E_{\rm F} \sim 0.25$ eV \cite{gunny}. This is also true
129: in MgB$_2$ where the low hole filling of the 2D $\sigma$ bands leads
130: to $E_{\rm F}^\sigma \sim 0.4-0.6$ eV \cite{an,kortus}. These values of $E_{\rm F}$
131: are at least one order of magnitude less than in common metals and in conventional
132: superconductors. The discrepancy between Mg and B NMR measurements can
133: be thus related to the probing of different bands ($\pi$ on Mg,
134: $\sigma$ on B), and, in the last analysis, to the different magnitude
135: of the Fermi energies ($E_{\rm F}^\pi \sim 5$ eV $\gg E_{\rm F}^\sigma$).
136:
137: On microscopic grounds, small Fermi energy effects are operative as soon
138: as $E_{\rm F}$ becomes of the same order of the other relevant energy
139: scales. For an electron-phonon system in the presence of
140: non magnetic impurities as we consider here, $E_{\rm F}$ should be thus
141: compared with the characteristic phonon energy scale $\omega_{\rm ph}$
142: and with the impurity scattering rate $\gamma_{\rm imp}$.
143: The breakdown of the adiabatic hypothesis
144: ($E_{\rm F} \gg \omega_{\rm ph}$) in a small Fermi energy system implies
145: the onset of new channels of electron-phonon interaction which
146: need to be taken into account. On the other hand the finiteness of the
147: ratio $\gamma_{\rm imp}/E_{\rm F}$ gives rise to anomalous impurity
148: effects which have to be analyzed in the presence of the same
149: electron-phonon interaction since
150: electron, phonon and impurity energy scales could be all of
151: the same magnitude: $E_{\rm F} \sim \omega_{\rm ph}
152: \sim \gamma_{\rm imp}$.
153:
154:
155:
156: \section{The model}
157:
158:
159: In this section, we derive the electron spin susceptibility
160: by employing the Baym-Kadanoff technique which permits to derive,
161: within a conserving theory,
162: higher order response functions as functional derivatives
163: of the single particle Green's function
164: in the presence of an external field \cite{baym}.
165: This approach is thus an appropriate starting point
166: to study small Fermi energy systems
167: where the violation of the Migdal's theorem valid
168: for $E_{\rm ph} \gg \omega_{\rm ph}$ requires a generalization
169: of the conventional theory in the nonadiabatic regime.
170:
171: Objects of our investigation is the non magnetic impurity effects
172: on the spin susceptibility in small Fermi energy systems
173: in the presence of a sizable electron-phonon interaction.
174: NMR techniques can probe
175: the electron density of states by means of different ways.
176: Most direct is the evaluation of the static uniform
177: limit $\chi$ of the generalized electron
178: spin susceptibility $\chi({\bf q},\omega)$:
179: \begin{equation}
180: \chi = \lim_{{\bf q} \rightarrow 0} \lim_{\omega \rightarrow 0}
181: \chi({\bf q},\omega)
182: \label{pauli}
183: \end{equation}
184: which, for a non interacting system with large Fermi energy,
185: is simply $\chi \propto N_0$.
186: Electron-electron exchange interaction gives rise however
187: to the so called Stoner enhancement: $\chi \propto N_0/(1-I)$
188: ($I$ being the Stoner factor).
189: Experimentally the static uniform limit $\chi$ of
190: electron spin susceptibility can be measured
191: by a proper analysis of the Knight shift after the orbital
192: contribution is subtracted.
193:
194:
195: Similar information are obtained by
196: spin-lattice relaxation rate $T_1$ which can be also mainly related,
197: after subtraction of orbital terms,
198: to the electron spin susceptibility through the relation:
199: \begin{equation}
200: \frac{1}{T_1T} \propto
201: \lim_{\omega \rightarrow 0}
202: \sum_{\bf q} A^2({\bf q}) \frac{\chi({\bf q},\omega)}{\omega},
203: \label{t1}
204: \end{equation}
205: where $A({\bf q})$ is the form factor relative
206: to the particular nucleus. As pointed out in the introduction,
207: $1/T_1 T \propto N_0^2$ in large Fermi energy systems.
208:
209: In the following we focus on the static uniform spin susceptibility
210: $\chi$
211: which permits a more direct comparison with the density of states
212: and which is only slightly affected by different form factors $A({\bf q})$.
213: As it will be clear in the following the anomalous effects of non magnetic
214: impurities are essentially related to the similar energy scales
215: of $\gamma_{\rm imp}$, $E_{\rm F}$ and $\omega_{\rm ph}$. In this situation
216: impurity scattering leads to an effective renormalization of the
217: generalized spin susceptibility
218: which is expected to appear in similar way
219: both in the static uniform limit $\chi$
220: and in the spin-lattice relaxation rate.
221: In this perspective the reduction of the magnetic susceptibility
222: upon disorder pointed out by NMR technique should be read more
223: as an anomalous renormalization effect appearing in small
224: Fermi energy system than as a real reduction of the density of states.
225:
226: In Quantum Field Theory
227: the electron spin susceptibility is usually related
228: to the one particle Green's function $G$ through the
229: relation:\cite{note-localeffects}
230: \begin{equation}
231: \chi(T)=
232: -2\mu_{\rm B}^2T\sum_n\sum_{{\bf k}}
233: G({\bf k},n)^2\Gamma({\bf k},n),
234: \label{chi}
235: \end{equation}
236: where $G({\bf k},n)$ is the electron propagator at finite temperature
237: expressed in
238: Matsubara frequencies
239: \begin{equation}
240: G^{-1}({\bf k},n)=i\omega_n-\epsilon({\bf k})+\mu-
241: \Sigma({\bf k},n),
242: \label{green}
243: \end{equation}
244: and $\Gamma({\bf k},n)$ is the spin vertex function.
245: The Baym-Kadanoff formalism provides a powerful technique to related
246: the spin vertex function $\Gamma({\bf k},n)$ to $G({\bf k},n)$.
247: Following the standard procedure we generalize
248: the Green's function in Eq. (\ref{green}) in the presence
249: of an external magnetic field $h$:
250: \begin{equation}
251: G^{-1}_\sigma({\bf k},n)=i\omega_n-\epsilon({\bf k})+\mu
252: +h\sigma-\Sigma_\sigma({\bf k},n).
253: \label{greenh}
254: \end{equation}
255: The spin vertex function is thus obtained
256: as functional derivative of the Green's function $G_\sigma$
257: in the presence of the external magnetic field for $h \rightarrow 0$ \cite{cgp}:
258: \begin{eqnarray}
259: \Gamma({\bf k},n)&=&\frac{1}{2}\sum_{\sigma}\sigma
260: \left[\frac{d G_{\sigma}^{-1}({\bf k},n)}{dh}\right]_{h=0}
261: \nonumber\\
262: &=&
263: 1-\frac{1}{2}\sum_{\sigma}\sigma
264: \left[\frac{d \Sigma_{\sigma}({\bf k},n)}{dh}\right]_{h=0}.
265: \label{gammaspin}
266: \end{eqnarray}
267: The set of Eqs. (\ref{chi})-(\ref{gammaspin}) defines a self-consistent
268: method to obtain the spin susceptibility from the knowledge of
269: the self-energy. The complex nature of the interactions in the systems
270: is thus hidden in the specific form of the self-energy which needs
271: to be explicitly provided.
272:
273: In order to investigate the interplay between non magnetic impurities and
274: the electron-phonon interaction
275: in small Fermi energy systems an appropriate approach is
276: the nonadiabatic theory which accounts for the
277: additional interaction channels arising
278: when $E_{\rm F} \sim \omega_{\rm ph}$.
279: The formal derivation of the nonadiabatic theory has been already
280: presented in several papers where we refer for more
281: details \cite{sgp,gps,cgp}.
282: Here we focus on the role of non magnetic impurities.
283: In the spirit of the Baym-Kadanoff theory our starting point
284: will be the self-energy which is diagrammatically depicted in
285: Fig. \ref{f-self}.
286: \begin{figure}
287: \centerline{\psfig{figure=fig1.eps,width=8cm}}
288: \vspace{1mm}
289: \caption{Diagrammatic picture of the electron self-energy.
290: Legend of the pictorial elements:
291: electrons (solid line), phonons (wavy lines), elastic scattering
292: (dotted lines) with dilute non magnetic impurities (cross),
293: electron-electron repulsion (dashed line).}
294: \label{f-self}
295: \end{figure}
296: The first two diagrams represent the electron-phonon interaction
297: in nonadiabatic regime ($E_{\rm F} \sim \omega_{\rm ph}$) including
298: the first order vertex processes; the third diagram is the
299: self-energy in Born approximation for impurities of density $n_{\rm imp}$
300: interacting with electrons via a scattering potential $v_{\rm imp}$.
301: The last diagram is the exchange electron-electron
302: interaction: this term is just a constant and does not play any role
303: in the self-energy, but it gives rise to the Stoner factor
304: in the spin susceptibility.
305:
306: Fig. \ref{f-self} defines in an unambiguous way the self-energy
307: and the one particle properties of the system. Standard procedure
308: in isotropic materials is to replace the self-en\-er\-gy $\Sigma({\bf k},n)$
309: with its Fermi surface average:
310: $\Sigma({\bf k},n) \rightarrow \Sigma(n) \equiv
311: \langle \langle \Sigma({\bf k},n) \rangle \rangle_{\rm FS}$.
312: It is convenient to take into account self-energy effects
313: in the electronic Green's function is to introduce the
314: renormalized Matsubara frequencies defined as $iW_n = i\omega_n -\Sigma(n)$.
315: In addition, for sake of simplicity we consider a half-filled band
316: with bandwidth $E$ and constant density of states
317: $N(\epsilon) = N_0$ [$-E/2 \le \epsilon \le E/2$]. The parameter
318: $E/2$ represents thus the Fermi energy $E_{\rm F}=E/2$.
319: Within these assumptions the analytic expression of the renormalized
320: Matsubara frequencies $W_n$ involving the self-energy
321: depicted in Fig. \ref{f-self} reads:
322: \begin{eqnarray}
323: W_n &=& \omega_n
324: -2T \sum_M V(n,m) \arctan\left(\frac{E_{\rm F}}{W_m}\right)
325: \nonumber\\
326: && +2\gamma_{\rm imp}
327: \arctan\left(\frac{E_{\rm F}}{W_n}\right),
328: \label{selfeq}
329: \end{eqnarray}
330: where
331: \begin{equation}
332: \label{V1}
333: V(n,m)=\lambda D(n-m)[1+\lambda P(Qc;n,m)]
334: \end{equation}
335: is the nonadiabatic electron-phonon kernel
336: appearing in the self-energy equation and where
337: we have neglected the
338: electron-electron exchange interaction which leads just to a constant
339: term. In Eq. (\ref{V1}), $D(n-m)$ is the phonon propagator which
340: for a single Einstein mode $\omega_{\rm ph} = \omega_0$ reduces
341: simply to $D(n-m)=-\omega_0^2/[(\omega_n-\omega_m)^2+\omega_0^2]$ and
342: $P(Q_c;n,m)$ is the vertex function \cite{sgp,cgp}:
343: \begin{eqnarray}
344: \label{vertex2}
345: &P&(Q_c;n,m) = -T\sum_lD(n-l)
346: \left\{\frac{}{}B(n,m,l)\right. \nonumber \\
347: & + & \frac{A(n,m,l)-B(n,m,l)(W_l-W_{l-n+m})^2}
348: {(2E_FQ_c^2)^2} \nonumber \\
349: & \times & \left.\left[R(Qc;n,m,l)-1-\log\left(\frac{1+R(Q_c;n,m,l)}{2}\right)
350: \right]\right\} , \nonumber \\
351: \end{eqnarray}
352: where
353: \begin{eqnarray}
354: \label{a}
355: A(n,m,l) & = & (W_l-W_{l-n+m})\left[\arctan\left(\frac{E_F}{W_l}\right)
356: \right. \nonumber \\
357: & - &\left.\arctan\left(\frac{E_F}{W_{l-n+m}}\right)\right] ,
358: \end{eqnarray}
359: \begin{eqnarray}
360: \label{b}
361: B(n,m,l) & = & (W_l-W_{l-n+m})\frac{E_FW_{l-n+m}}
362: {[E_F^2+W_{l-n+m}^2]^2} \nonumber \\
363: & - & \frac{E_F}{E_F^2+W_{l-n+m}^2} ,
364: \end{eqnarray}
365: \begin{equation}
366: \label{r}
367: R(Q_c;n,m,l)=\sqrt{1+\left(\frac{4E_FQ_c^2}{W_l-W_{l-n+m}}\right)^2} .
368: \end{equation}
369: The dimensionless parameter $Q_c=q_c/2k_{\rm F}$, where $k_{\rm F}$ is the Fermi momentum,
370: takes into account the upper cutoff $q_c$ for the momentum transfer in the
371: electron-phonon interaction. This cutoff has been introduced to simulate a
372: momentum dependent renormalization due to possible
373: strong electronic correlations \cite{kulic}.
374: For weak correlated metals $Q_c\simeq 1$, while $Q_c\ll 1$ when correlation is strong.
375: As we are going to see, the parameter $Q_c$ plays only a marginal
376: on the static spin susceptibility, whereas it strongly affects
377: the superconducting critical temperature \cite{gps}.
378:
379: In the formula for $W_n$, Eq. (\ref{selfeq}), $\gamma_{\rm imp}$ is the
380: impurity scattering rate which in the Born approximation reduces to
381: $\gamma_{\rm imp}=\pi n_{\rm imp}N_0 v_{\rm imp}^2$ \cite{rick}. This expression holds true for
382: low values $n_{\rm imp}$ of impurity concentrations and
383: weak scattering potential $v_{\rm imp}$. An expression of
384: $\gamma_{\rm imp}$ valid also for strong, but diluted, impurity interactions is
385: provided by the $T$-matrix approximation:
386: $\gamma_{\rm imp}=
387: \pi n_{\rm imp}N_0 v_{\rm imp}^2/[1+(\pi N_0 v_{\rm imp})^2]$.
388:
389:
390:
391:
392: Using the Baym-Kadanoff formalism we are able
393: to obtain also an analytic expression
394: for the spin vertex function $\Gamma({\bf k},n)$.
395: The diagrammatic expression of $\Gamma({\bf k},n)$ corresponding
396: to the self-energy depicted in Fig. \ref{f-self} is shown
397: in Fig. \ref{f-gamma}.
398: \begin{figure}
399: \centerline{\psfig{figure=fig2.eps,width=8cm}}
400: \vspace{1mm}
401: \caption{Diagrammatic picture of spin vertex function $\Gamma$.
402: See caption of \protect Fig. \ref{f-self} for a legend
403: of the pictorial elements.}
404: \label{f-gamma}
405: \end{figure}
406: In the isotropic case we have considered here
407: we can replace also the spin vertex function $\Gamma({\bf k},n)$
408: with its Fermi surface average $\Gamma(n)$.
409: The momentum average of the spin vertex implies that the
410: momentum correlations between spin up electrons and spin down holes
411: are taken into account only at a mean level through the parameter $Q_c$.
412: This will be a poor approximation when the dispersion
413: of collective modes is investigated, while it is
414: expected to not affect in a qualitative way
415: the static uniform spin susceptibility.
416: Disregarding the momentum dependence of
417: $\Gamma({\bf k},n)$ and we get thus:
418: \begin{eqnarray}
419: \Gamma (n)&=&1 + T\sum_{m}\left[I+V_\Gamma(n,m)\right]
420: \frac{2E_F}{W_m^2+E_F^2}\Gamma(m)
421: \nonumber\\
422: &&
423: -\gamma_{\rm imp}\frac{2E_F}{W_n^2+E_F^2}\Gamma(n),
424: \label{gammavertex}
425: \end{eqnarray}
426: where the quantity $I=N_0 U$ is the Stoner factor arising from the
427: electron-electron exchange interaction and the last term
428: comes from the impurity scattering processes.
429: Moreover
430: \begin{eqnarray}
431: \label{ave6}
432: V_\Gamma(n,m)&=&\lambda D(n-m)\left[
433: 1+2 \lambda P(Q_c;n,m)\right] \nonumber \\
434: &+&\lambda^2 C(Q_c;n,m)
435: \end{eqnarray}
436: describes
437: the electron-phonon processes in nonadiabatic regime which include
438: electron-pho\-non vertex $P(Q_c;n,m)$ given by Eqs.(\ref{vertex2}-\ref{r}) the
439: and cross diagrams:
440: \begin{eqnarray}
441: \label{crossb}
442: &C&\!(Q_c;n,m) = T\!\sum_l D(n-l)D(l-m) \nonumber \\
443: & \times &\!\left\{2B(n,m,l)+\arctan\left(\frac{4E_FQ_c^2}
444: {|W_l-W_{n+m-l}|}\right)\right. \nonumber \\
445: & \times &\left.
446: \frac{A(n,m,l)-B(n,m,l)(W_l-W_{n+m-l})^2}
447: {2E_FQ_c^2|W_l-W_{n+m-l}|}\right\} ,
448: \end{eqnarray}
449: where $A(n,m,l)$ and $B(n,m,l)$ are given by equations
450: (\ref{a}) and (\ref{b}), respectively.
451: Note that
452: $V_\Gamma(n,m)$ is deep\-ly different from $V(n,m)$
453: since the first describes electron-phonon scattering in the spin
454: electron-hole channel, and the second one the electron-phonon interaction
455: in the single particle propagator.
456:
457: \section{Results and discussion}
458:
459:
460: Eqs. (\ref{selfeq})-(\ref{crossb})
461: can be solved in a self-consistent iterative way to obtain
462: $W_n$ and $\Gamma(n)$.
463: Eq. (\ref{chi}), in its isotropic form:
464: \begin{equation}
465: \chi(T)=
466: \chi_{\rm P} T\sum_n
467: \frac{2E_{\rm F}}{W_n^2+E_{\rm F}^2}
468: \Gamma(n),
469: \label{chi-iso}
470: \end{equation}
471: provides finally the spin susceptibility as function
472: of ge\-ner\-ic impurity scattering rate $\gamma_{\rm imp}$,
473: electron-phonon coupling constant $\lambda$, adiabatic ratio
474: $\omega_0/E_{\rm F}$ and momentum cut-off $Q_c$.
475: Here $\chi_{\rm P}$ is the free electron Pauli spin susceptibility
476: $\chi_{\rm P} = 2\mu_{\rm B}^2 N_0$.
477:
478: In order to point out the role of small Fermi energy in
479: the impurity scattering effects on the spin susceptibility,
480: we consider for the moment the simple case of no electron-phonon
481: interaction ($\lambda=0$). In this case the only energy scales
482: in the system are $\gamma_{\rm imp}$ and $E_{\rm F}$.
483: From Eq. (\ref{chi-iso}), equation (\ref{gammavertex})
484: has thus the simple self-consistent
485: solution as function of the spin susceptibility itself:
486: \begin{equation}
487: \Gamma(n)=\frac{\displaystyle 1+I (\chi/\chi_{\rm P})}{\displaystyle 1+
488: \gamma_{\rm imp}\frac{\displaystyle 2E_F}{\displaystyle W_n^2+E_F^2}},
489: \end{equation}
490: and the spin susceptibility $\chi$ recovers the usual Stoner-like
491: expression:
492: \begin{equation}
493: \chi = \frac{\chi_0}{1-I (\chi_0/\chi_{\rm P})},
494: \end{equation}
495: where the bare spin susceptibility $\chi_0$ is now affected by the
496: non magnetic impurity scattering:
497: \begin{equation}
498: \chi_0 = \chi_{\rm P}
499: T\sum_{m}
500: \frac{2E_F}{W_m^2+E_F^2 + \gamma_{\rm imp}2E_F}.
501: \label{chi0}
502: \end{equation}
503: For large Fermi energy systems, $E_{\rm F} \gg \gamma_{\rm imp}$,
504: equation (\ref{chi0}) reduces to the Pauli spin susceptibility
505: $\chi_0 = \chi_{\rm P}$. It is thus clear the non magnetic
506: impurity effects can appear only if the Fermi energy is small enough
507: to be comparable with $\gamma_{\rm imp}$. Note that in the presence
508: of electron-phonon interaction an additional energy scale is provided
509: by $\omega_{\rm ph}$, so that additional anomalous impurity effects are
510: ruled by the additional parameter $\gamma_{\rm imp} / \omega_{\rm ph}$.
511:
512:
513: \begin{figure}
514: \centerline{\psfig{figure=fig3.eps,width=8cm}}
515: \vspace{1mm}
516: \caption{Spin susceptibility $\chi$ as function of the
517: impurity scattering rate $\gamma_{\rm imp}$ in the presence of
518: electron-phonon interaction($\lambda=0.4$, left panel;
519: $\lambda=1.0$, right panel, $\omega_0/E_{\rm F}=0.7$ and
520: electron-electron exchange repulsion $I=0.4$.
521: Solid lines: nonadiabatic vertex corrected theory
522: with different values of $Q_c$ (from the top to the bottom:
523: $Q_c=0.1, 0.3, 0.5, 0.7, 0.9$); dashed line: non crossing
524: approximation.}
525: \label{f-chi-vs-g}
526: \end{figure}
527:
528: In Fig. \ref{f-chi-vs-g} we plot the behaviour of the static
529: spin susceptibility $\chi$ in a small Fermi energy
530: system ($\omega_0/E_{\rm F}=0.7$)
531: as function of the
532: impurity scattering rate $\gamma_{\rm imp}$.
533: The data are normalized with respect to the
534: ``pure'' limit $\gamma_{\rm imp} \rightarrow 0$.
535: Left panel refer to a weak coupling electron-phonon case
536: ($\lambda=0.4$), right panel to strong coupling ($\lambda=1.0$).
537: In both the case a Stoner factor $I=0.4$ was considered.
538: Solid lines represent the nonadiabatic vertex corrected theory
539: with different values of $Q_c$ (from the top to the bottom:
540: $Q_c=0.1, 0.3, 0.5, 0.7, 0.9$) and the dashed line the non crossing
541: approximation where only finite bandwidth effects were retained
542: [$P(Qc;n,m)=C(Q_c;n,m)=0$ in Eqs. (\ref{V1})-(\ref{ave6})].
543: Fig. \ref{f-chi-vs-g} shows a strong reduction of $\chi$ due to the
544: impurities scattering with respect to a
545: large Fermi energy case ($E_{\rm F} \gg \omega_0$, dotted line).
546: We observe only a weak dependence on the electron-phonon
547: coupling (right panel data are slightly higher than the left
548: panel), while the introduction of the nonadiabatic vertex and cross
549: diagrams essentially leads to a spread of the different curves
550: for different values of $Q_c$.
551:
552: From this behaviour we can argue that the electron-phonon interaction
553: $\lambda$ plays a secondary role in the reduction of $\chi$ due
554: to impurity scattering. In similar way a marginal role
555: is recovered for the electron-electron interaction (Stoner factor $I$).
556: As matter of facts, the leading effects are ruled by the comparison
557: between the energy scales $\gamma_{\rm imp}$, $\omega_0$ and
558: $E_{\rm F}$. In order to highlight this point we
559: compare in Fig. \ref{f-chi-vs-g2} the dependence of
560: \begin{figure}
561: \centerline{\psfig{figure=fig4.eps,width=8cm}}
562: \vspace{1mm}
563: \caption{Spin susceptibility $\chi$ as function of the
564: impurity scattering rate $\gamma_{\rm imp}$
565: for a small Fermi energy ($\omega_0/E_{\rm F}=0.7$, left panel)
566: and for a large Fermi energy
567: system ($\omega_0/E_{\rm F}=0.1$, right panel).
568: Solid lines corresponds to different values of $\lambda$:
569: (from bottom to the top) $\lambda=0.2, 0.4, 0.6, 0.8, 1$.
570: Other parameters: $Q_c=0.4$ and $I=0.4$.
571: Dotted line: infinite Fermi energy case.}
572: \label{f-chi-vs-g2}
573: \end{figure}
574: the magnetic susceptibility $\chi$ on the impurity scattering
575: rate for a small Fermi energy ($\omega_0/E_{\rm F}=0.7$,left panel)
576: and for a large Fermi energy
577: system ($\omega_0/E_{\rm F}=0.1$, right panel).
578: Solid lines corresponds to different values of $\lambda$
579: (from bottom to the top): $\lambda=0.2, 0.4, 0.6, 0.8, 1$.
580: Here we set $Q_c=0.4$ and $I=0.4$ but, as above discussed, different values
581: would not change the physics.
582: Note the remarkable difference between left and right panel.
583: The same impurity scattering rate, which in the presence of a small
584: Fermi energy
585: ($\omega_0/E_{\rm F}=0.7$) would lead to a reduction of $\chi$
586: of about 33-47\%, gives rise only to a 12-14\% reduction when
587: Fermi energy is considerably increased
588: ($\omega_0/E_{\rm F}=0.1$), and to no reduction
589: at all for the infinite Fermi energy case (dotted line).
590:
591: Curiously the presence of the electron-phonon interaction
592: decreases the sensitivity of $\chi$ to impurity scattering rate.
593: This can be understood considering that the the electron-phonon
594: scattering reduces by itself the magnetic susceptibility $\chi$ \cite{cgp},
595: so that further reduction by non magnetic impurity scattering
596: is disfavored. For the same reasons a stronger Stoner factor
597: $I$ would enhance the reduction of $\chi$.
598:
599: \section{Disorder and nonmagnetic impurities in real materials
600: (MgB$_2$, fullerenes, \ldots)}
601:
602: We are now in the position to re-address the open questions
603: arisen in the introduction, concerning namely:
604: the origin of the
605: reduction of the density of states as probed by
606: NMR susceptibility measurements;
607: the discrepancy between the different behaviour of Mg and B
608: NMR measurements.
609: In particular we suggest that the reduction of the
610: spin-lattice relaxation rate $1/T_1 T$ upon induced disorder
611: could reflect the small Fermi energy nature of the electronic
612: structure probed by the experiments.
613: Within this context the insensitivity to disorder scattering of
614: the spin-lattice relaxation rate $1/T_1 T$ probed on $^{25}$Mg in contrast
615: to the marked reduction of $^{11}$B NMR measurements acquires a natural
616: explanation related to the different Fermi energy scales involved
617: in the two cases.
618: $^{25}$Mg NMR measurements mainly probe the $\pi$ band structures with
619: high Fermi energy $E_{\rm F} \sim 5$ eV, Using a typical phonon frequency
620: $\omega_0 \simeq 70$ meV \cite{renker} we estimate
621: $\omega_0/E_{\rm F}^{\pi} \sim 0.014$ which yields a negligible
622: dependence of the spin susceptibility on the amount of disorder.
623: On the other hand NMR on the $^{11}$B boron nucleus is strongly
624: coupled with the in-plane $\sigma$ orbitals with
625: Fermi energy $E_{\rm F} \sim 0.4-0.6$ eV. The same phonon
626: frequency scale gives thus $\omega_0/E_{\rm F}^{\sigma} \sim 0.12-0.14$,
627: where visible impurity scattering effects are expected.
628: We conclude that the small Fermi energy of the $\sigma$ bands
629: is the major responsible for the reduction of
630: the magnetic susceptibility when probed on $^{11}$B nuclei compared
631: with NMR measurements on the same quantity on $^{25}$Mg.
632:
633: Note that the usual two band model within the Migdal-Eliashberg
634: framework, with different electron-phonon coupling
635: for $\sigma$ and $\pi$ bands
636: ($\lambda_\sigma \sim 1$, $\lambda_\pi \sim 0.2$),
637: can not alone explain the discrepancy between the reduction rate
638: probed by NMR. Indeed: a) electron-phonon interaction
639: does not affect impurity scattering for high Fermi energy systems
640: ($E_{\rm F} \gg \gamma_{\rm imp}, \omega_0$);
641: b) the larger electron-phonon coupling constant in the $\sigma$ bands
642: would predict a smaller reduction rate of $\chi$ as compared
643: with the smaller $\lambda$ of the $\pi$ bands.
644:
645: The above discussion suggests that the reduction of $\chi$
646: upon disorder is a possible tool to point out nonadiabatic effects
647: in small Fermi energy materials where $\omega_0 \sim E_{\rm F}$.
648: In this perspective it is interesting to compare the simultaneous
649: reduction of $\chi$ and $T_c$ as function of the impurity scattering
650: rate $\gamma_{\rm imp}$. In this way in principle one can trace out
651: the effects of non magnetic impurity scattering in small Fermi
652: energy systems as functions of physical measurable quantities
653: as $\chi$ and $T_c$ avoiding the use of the unaccessible parameter
654: $\gamma_{\rm imp}$.
655:
656: A conserving derivation of the superconducting equations in a
657: fully consistent way with the evaluation of the spin susceptibility
658: follows once again the Baym-Kadanoff theory based on Fig. \ref{f-self}
659: written in Nambu notation.
660: A formal derivation of those equations and some technicalities
661: about the numerical calculations of $T_c$ were discussed in
662: Ref. \cite{sgp} where we refer for more details.
663: In Fig. \ref{f-tcvschi}a we plot
664: \begin{figure}
665: \centerline{\psfig{figure=fig5a.eps,width=8.5cm}}
666: \vspace{1mm}
667: \centerline{\psfig{figure=fig5b.eps,width=6cm}}
668: \vspace{1mm}
669: \caption{(a) Reduction of $T_c$ (left panel) and $\chi$
670: (right panel) as function of $\gamma_{\rm imp}$
671: for the cases ($\lambda=1.0$, $\omega_0/E_{\rm F}=0.2$, $I=0.2$)
672: and ($\lambda=0.2$, $\omega_0/E_{\rm F}=0.02$, $I=0.2$),
673: qualitatively representative respectively of the $\sigma$ and $\pi$ bands
674: in MgB$_2$. Different lines corresponds to different $Q_c$'s
675: (from the bottom to the top in left panel, from the top
676: to the bottom in right panel): $Q_c = 0.1, 0.3, 0.5, 0.7, 0.9$.
677: (b) Corresponding plot of $T_c$ vs. $\chi$ as varying $\gamma_{\rm imp}$
678: ($\gamma_{\rm imp}=0$ at the left end,
679: $\gamma_{\rm imp}=0.5 \omega_0$ at the right end)
680: for ($\lambda=1.0$, $\omega_0/E_{\rm F}=0.2$, $I=0.2$).
681: Different values of $Q_c$ are reported as in the previous captions.}
682: \label{f-tcvschi}
683: \end{figure}
684: $\chi$ and $T_c$ as functions of $\gamma_{\rm imp}$ for $I=0.2$
685: and the couples of parameters ($\lambda=1.0$, $\omega_0/E_{\rm F}=0.2$),
686: ($\lambda=0.2$, $\omega_0/E_{\rm F}=0.02$). These cases should
687: be qualitatively representative of the MgB$_2$ $\sigma$ and $\pi$ bands
688: which are respectively: strong coupled with small Fermi energy;
689: and weak coupled with large Fermi energy. Note that a significant
690: dependence on $\gamma_{\rm imp}$, for both $T_c$ and $\chi$
691: is observed only for ($\lambda=1.0$, $\omega_0/E_{\rm F}=0.2$)
692: which represents the case of $\sigma$ bands. The $T_c$ vs. $\chi$ plot
693: is shown in Fig. \ref{f-tcvschi}b for
694: ($\lambda=1.0$, $\omega_0/E_{\rm F}=0.2$, $I=0.2$). A reduction of
695: $T_c$ of the order of $30-80 \%$ is predicted for a reduction
696: of $\chi$ of $\sim 20 \%$, depending on the parameter $Q_c$.
697: The general trend is thus in agreement with the experimental
698: data reported in Ref. \cite{gerashenko}.
699:
700: In MgB$_2$, where the electronic correlation is thought to be negligible,
701: there is no reason to expect a significant
702: momentum selection and $Q_c$ is expected to be $Q_c \sim 1$.
703: In this situation our analysis would underestimate the suppression of
704: $T_c$ ($\Delta T_c / T_c \sim 30 \%$) and $\chi$
705: ($\Delta \chi / \chi \sim 20 \%$) when compared with the
706: experimental scenario, although
707: some care should be used to extrapolate from the static magnetic
708: susceptibility $\chi$ to $1/ T T_1$.
709:
710: It is clear however that
711: additional ingredients are
712: required to be taken into account for a quantitative analysis of the
713: experimental data
714: of $T_c$ vs. induced disorder.
715: In particular the discussion in terms of two
716: separated $\sigma$ and $\pi$ bands is expected to be a poor description
717: for the superconducting properties of a complex multiband system as MgB$_2$.
718: On this basis we conclude that further investigation is needed to
719: account in a fully satisfactory way for the anomalous dependence
720: of $T_c$ on the amount disorder. On the other hand
721: the reduction of the spin susceptibility
722: upon disorder and non magnetic impurities in MgB$_2$
723: could be qualitatively understood within the present analysis.
724: In paricular our results suggest that a primary role
725: could be played by the small Fermi energy effects driven in MgB$_2$ by the
726: closeness of the chemical potential to the top
727: of the $\sigma$ band. In this framework
728: the different behaviour of B and Mg NMR data receives a natural explanation.
729:
730: Interesting perspectives are also opened in regards to the fullerene
731: based materials. The analysis is indeed simplified in these compounds
732: as a single Fermi energy is present. The extreme smallness of $E_{\rm F}$
733: is fullerenes ($E_{\rm F} \simeq 0.25$ eV) suggests that disorder
734: or nonmagnetic impurity effects could lead to even more marked reduction
735: of $T_c$ and $\chi$ than in MgB$_2$.
736: A suppression of $T_c$ upon induced disorder as previously been reported
737: in Ref. \cite{watson}
738: At our knowledge no measurements of magnetic susceptibility
739: as function of disorder or impurity amount as been at the present
740: performed. Experimental work along this line is thus encouraged.
741:
742: \acknowledgement
743:
744: This work was partially supported by INFM Research Project
745: PRA-UMBRA
746:
747: \endacknowledgement
748:
749: \begin{thebibliography}{99}
750:
751: \bibitem{anderson}
752: P. W. Anderson, J. Phys. Chem. Solid, {\bf 11}, 26 (1959).
753:
754: \bibitem{dwave}
755: A. B. Aleksashin {\it et al.}, Physica C {\bf 153-155}, 339 (1988);
756: Gang Xiao {\it et al.}, Phys. Rev. B {\bf 42}, 8752 (1990);
757: W. N. Hardy {\it et al.}, Phys. Rev. Lett {\bf 70}, 3999 (1993);
758: E. R. Ulm {\it et al.}, Phys. Rev. B {\bf 51}, 9193 (1995).
759:
760: \bibitem{flatte}
761: M. E. Flatt\'e and J. M. Byers, Solid State Phys. {\bf 52}, 137 (1996).
762:
763: \bibitem{watson}
764: S. K. Watson {\it et al.}, Phys. Rev. B {\bf 55}, 3866 (1997).
765:
766: \bibitem{karkin}
767: A. E. Kar'kin {\it et al.}, JETP Letters {\bf 73}, 570 (2001).
768:
769: \bibitem{gerashenko}
770: A.P. Gerashenko {\em et al.}, Phys. Rev. B {\bf 65}, 132506 (2002).
771:
772: \bibitem{sgp} M. Scattoni, C. Grimaldi and L. Pietronero,
773: Europhys. Lett. {\bf 47}, 588 (1999).
774:
775: \bibitem{gps}
776: L. Pietronero, S. Str\"assler, and C. Grimaldi,
777: Phys. Rev. B {\bf 52}, 10516 (1995);
778: C. Grimaldi, L. Pietronero, and S. Str\"assler,
779: Phys. Rev. B {\bf 52}, 10530 (1995).
780:
781: \bibitem{gunny}
782: O. Gunnarsson, Rev. Mod. Phys. {\bf 69}, 575 (1997).
783:
784: \bibitem{an}
785: J. M. An and W. E. Pickett, Phys. Rev. Lett. {\bf 86}, 4366 (2001).
786:
787: \bibitem{kortus}
788: J. Kortus {\it et al.}, Phys. Rev. Lett. {\bf 86}, 4656 (2001).
789:
790: \bibitem{baym}
791: G. Baym and L. P. Kadanoff, Phys. Rev. {\bf 124}, 287 (1961).
792:
793: \bibitem{note-localeffects}
794: We neglect here non local effects appearing in the
795: neighbouring of a single impurity. In this perspective only
796: physical quantities averaged on the impurity configuration,
797: which do not break translational invariance,
798: will be in the following involved.
799:
800: \bibitem{cgp} E. Cappelluti, C. Grimaldi and L. Pietronero,
801: Phys. Rev. B {\bf 64}, 125104 (2001).
802:
803: \bibitem{kulic}
804: M. L. Kuli\'c and R. Zeyher, Phys. Rev. B {\bf 49}, 4395 (1994);
805: R. Zeyher and M. Kuli\'c, Phys. Rev. B {\bf 53}, 2850 (1996).
806:
807: \bibitem{rick}
808: G. Rickayzen, {\it Green's Functions and Condensed Matter}
809: (Academic Press, New York, 1980).
810:
811: \bibitem{renker} B. Renker, K.B. Bohnen, R. Heid, D. Ernst,
812: H. Schober, M. Koza, P. Adelmann, P. Schweiss and T. Wolf,
813: Phys. Rev. Lett. {\bf 88}, 067001 (2002).
814:
815: \end{thebibliography}
816:
817:
818: \end{document}
819: