VOLUME 88, NUMBER 5 P HYS ICAL REVIEW LETTER S4 FEBRUARY 2002 Download full-text
Comment on “Spinless Impurities In High-Tc
Cuprates: Kondo-Like Behavior”
Bobroff et al., in an astute study , compare the
effects of Zn and Li doping in YBa2Cu3O72d (Y123)
and Y0.8Ca0.2Ba2Cu3O72dusing89Y and7Li NMR.
NMR allows precise determination of the T dependence
of the local spin susceptibility, xs, especially in overdoped
samples where the89Y satellites (which also probe the
local xs) could no longer be distinguished from the main
resonance (probing the bulk xs). From the T dependence
of the local xs, which progressed from Curie-like to a
1??T 1 u? Kondo-like variation, these spinless substi-
tutions were inferred to induce a local moment on the
nearest-neighbor (nn) Cu sites. The new energy scale, u,
rose with doping to $100 K. The authors then conclude
that even in overdoped materials there is still a substantial
induced moment as evidenced by the strong 1??T 1 u?
dependence of xs. We question this.
There are two alternative sources of the apparent in-
duced moment. The short range antiferromagnetic (AF)
background may induce an unbalanced quasistatic moment
on the four nn Cu sites adjoining the spin vacancy .
Alternatively, a Curie-like xscould arise merely from a
sharp low-energy peak in the density of states (DOS), as
recently observed by scanning tunneling microscopy at Zn
impurity sites . Such a resonance peak can originate
from the presence of a gap in the DOS, whether in the
superconducting (SC) or normal state (i.e., the pseudo-
gap) , and does not implicate (indeed may rule out) a
local moment. In fact, the Zn resonance lies at a small
negative bias (V ? 1.5 meV), thus cutting off the diver-
gence of the Curie dependence at low T with the func-
tional form tanh?V?kBT?. The Kondo-like T dependence
also cuts off the Curie divergence, but the two functions
differ substantially at low T. Elsewhere , we showed
that the short-range low-energy AF correlations seem to
die out abruptly at the critical doping point, pc? 0.19,
where the pseudogap disappears. Thus, either way, the ap-
parent moment should disappear in the lightly overdoped
region, p . pc, either through loss of AF correlations or
loss of the pseudogap. A Li-induced moment in the heav-
ily overdoped region seems to contradict this scenario.
First, we show that a Kondo-like overdoped T depen-
dence is intrinsic. Figure 1 shows89Ks for a series of
overdoped samples of pure Y0.8Ca0.2Ba2Cu3O71d. The
suppression of Ksin the underdoped and optimally doped
samples is associated with the pseudogap [4,5] and mod-
eled here by a triangular gap . For p . pc,89Ks ex-
hibits an upturn that closely fits a 1??T 1 u? dependence
(dashed curves), in all cases with u ? 110 K.
Now Bobroff et al. point out that the T dependence of
the observed local Kondo-like xsis much stronger than
that of the intrinsic xsshown in Fig. 1a. But7Ksmay be
used to calculate the local89Ks that would be probed by
the89Y nucleus in the satellite resonance arising from a Y
89Y Knight Shift (ppm)
FIG. 1 (color).
Cu3O72dfitted to 1??T 1 110?. Bold curve denotes critical dop-
ing, pc? 0.19. Calculated satellite shifts are shown by crosses.
89Y Knight shift for overdoped Y0.8Ca0.2Ba2-
atom adjoining the Zn or Li impurity. This is shown for
pure Y123 (blue crosses) and for Y0.8Ca0.2Ba2Cu3O72d
(red plusses), both fully oxygenated. Here the T depen-
dence is very strong, possibly sufficient, even with typical
finite linewidths, to observe the satellites. They are not
observed. A resolution of the whole problem, we suggest,
is that Li causes local excess doping due to incomplete
screening. The overdoped “moment” may then indeed be
absent and the strong Kondo-like T dependence of7Ksis
just the intrinsic xsof a more overdoped sample. In this
way there is no new energy scale, just a crossover from the
low cutoff energy of the underdoped extrinsic resonance
(15 K) to the higher energy cutoff scale (110 K) of the
intrinsic overdoped xs. These are distinguished not just
by their energy but their functional form. It would seem,
then, that an induced moment in the overdoped region re-
mains uncertain. Tunneling studies could provide a clear
answer to this question, where the resonance in the normal
state (or in vortex cores in the SC state) may be expected
to abruptly disappear for p . pc.
J.L. Tallon,1,2J.W. Loram,2and G.V.M. Williams1
P.O. Box 31310
Lower Hutt, New Zealand
Cambridge CB3 0HE, United Kingdom
Received 7 July 2000; published 16 January 2002
PACS numbers: 74.25.Nf, 74.62.Dh, 74.72.Bk
 J. Bobroff et al., Phys. Rev. Lett. 83, 4381 (1999).
 S.H. Pan et al., Nature (London) 403, 746 (2000).
 H. Kruis, I. Martin, and A. Balatsky, Phys. Rev. B 64,
 J.L. Tallon and J.W. Loram, Physica (Amsterdam) 349C,
 H. Alloul et al., Phys. Rev. Lett. 63, 1700 (1989).
059701-10031-9007?02?88(5)?059701(1)$20.00© 2002 The American Physical Society059701-1