PHYSICAL REVIE%'
8
VOLUME 46, NUMBER 1 1JUI.
Y 1992-IThermal fluctuations
in
the
microwave
conductivity
of
Bi2Sr2CaCu2os
M.
L.
Horbach and W.van SaarloosInstitute Lor-entzforTheoretical Physics, University
of
Leiden,P
0
B.ox. 9506,2300RALeiden, TheNetherlands (Received 3February 1992)Recently, a peak close to T,has been observed in the microwave conductivity ofsingle crystals ofthe high-temperature superconductor Bi2Sr2CaCu208. This peak was interpreted as a coherence peak. In this paper we investigate an interpretation in terms ofthermal-fluctuation effects. The fluctuation con-tribution to the conductivity calculated by Aslamasov and Larkin (generalized to 6nite frequencies by Schmidt) isofthe magnitude ofthe observed effect, but leads to a narrow peak at T,
.
In microwave ex-periments in the gigahertz range, however, which probe a surface layer with a distribution of T,s, thermal fluctuations lead to a broader peak slightly below the dc critical temperature, as observed. Strong pair breaking tends to shift this peak somewhat further below the dc critical temperature, and also suppresses the importance ofother fluctuation contributions (Maki-Thompson), which in principle could lead to a fluctuation peak in the nuclear-spin-relaxation rate. Our results are consistent with the conclusion that there are no true coherence peaks in the conductivity ornuclear-spin relaxation as a re-sult ofstrong pair breaking.I.
ABSENCEOFCOHERENCE PEAKSA well-known characteristic feature
of
weak-couplingBCS
superconductors is the existenceof
a so-calledcoherence
peak'
in both the nuclear-spin-relaxation rate1/T, T
and the microwave conductivityo,
( )co/o,„(co)atfrequencies much smaller than b,
(0),
the gap at zerotem-perature. The peak in these quantities as a function
of
temperature appears typically at about0.
8T,
and has a widthof
about0.
4T,
.According to
BCS
theory, a coherence peak reflects propertiesof
the quasi-particle spectrum andof
the singular natureof
the weak-coupling quasiparticle densi-tyof
states in the superconducting state just above the gap,X,
(E
)=
EN„(E
)/+E
b,
whereX,
~„~(E—) is thequasiparticle density
of
states in the superconducting (normal) state.It is an established experimental fact that the high-T,
cuprate superconductors show no peak below
T,
for1/T,
T. Several explanations for this have been put for-ward. Strong inelastic scattering (with a bosonic mode which presumably isof
electronic origin ) leads to a smearingof
the density-of-states singularity and fillingof
the gap and, depending on the strengthof
the coupling,to the suppression
of
the coherence peak. An implicationof
such an explanation is that the coherence peak in the microwave conductivity should be absent as well, whenmeasured at sufficiently low frequencies.
Several groups, however, have observed peaks in
the conductivity at frequencies up to 60
GHz.
These peaks look rather different from the weak-couplingBCS
coherence peaks, in the sense that they occur very close
to
T,
and that they are very narrow, the width typically 1—
3K
(see insetof Fig. 3).
Also this could be due to strong coupling, as was suggested by Holczer etal.
Alarge value
of
2b,(0)/kT,
and a constant b,(T)
from zero temperature almost up toT,
results in a narrowcoher-0.85 0.90 0.95 1.00
FIG. 1. A narrow coherence peak obtained for 24(0)/kT,
=9
and a gap which istemperature independent up to about 0.85T,. Both anomalous features are necessary to pro-duce such a narrow peak. The dashed line is a weak-couplingBCScoherence peak. Horizontal axis inunits ofT,
.
ence peak (Fig. 1). Large pairbreaking, however, as it
usually occurs in the case
of
strong coupling, would des-troy this peak. In any case, it seems that strong coupling cannot explain both the absenceof
a peak in the1/T, T
data and the presenceof
anarrow peak in theconductivi-ty.
A fit
of
the narrow conductivity peak observed on granular YBa2Cu307 films by Kobrin etal.
within weak-couplingBCS
theory was presented inRef.
10,withas ingredients a temperature-dependent mean free path and effective carrier mass and a temperature-dependent mixture
of
normal and superconducting regionsof
the sample close toT,
~ A large mean free path nearT,
suppresses the coherence peak, while the mixture
of
nor-mal and superconducting regions leads to the appearanceof
an additional peak.A mechanism which has different effects on the nuclear-spin relaxation and the conductivity is provided by the spin-bag approach. In this approach a collective mode exists which renormalizes the interaction which is
relevant for the nuclear-spin relaxation, but which does
not affect the conductivity. Consequently this approach
predicts"
a coherence peak in the conductivity but not inthe nuclear-spin relaxation.
If
the narrow conductivity peak is indeed a coherence peak itmight be viewed as ex-perimental support forthis picture.It
must be mentioned that at relatively high frequen-cies (THz) a broad conductivity peak has been ob-served. ' An explanation forthis peak has been given''
in termsof
a competition between an increasedquasipar-ticle lifetime
r
(and consequently an increased diffusion constantD=v~r/d)
when the temperature is lowered throughT,
and a decreaseof
the densityof
states N, inthe gap region. The details
of
this mechanism appear todepend sensitively on the opening up
of
the gap asa func-tionof
temperature. In particular, pair-breaking effectscan lead togapless superconductivity. The precise way in which a real gap is filled in, related to the pair-breaking
rate, then determines whether the decrease
of
thequasi-particle density
of
states close to the Fermi energy is lowered sufficiently in order to overrule the increaseof
the diffusion constant. This delicate mechanism would not lead to a peak in the nuclear relaxation rate,can-sistent with experiment.
Recently, Marsiglio' found that, within the frame-work
of
Eliashberg theory, the conductivity coherencepeak (not the
I/T~
T
peak) disappears in the clean limit.In view
of
Holczer's observationof
a peak in thecon-ductivity, he concludes that the clean limit can be ruled
out. Also, based on the same argument, he rules out very strong coupling, since then the peak disappears. These statements
of
course depend heavily on the interpretationof
the observed conductivity peaks ascoherence peaks. We shall concentrate our attention on these narrow conductivity peaks. Our aim is toshow that the peak in Bi2Sr2CaCuz08 observed at 60 GHz by Holczer etal.
, rather than being coherence peaks, might well be due tothermal fluctuations. The picture then is that although a real coherence peak is absent, presumably due to strong-coupling effects, a fluctuation-induced peak may arise. In
the experiments on YBa2Cu307 at lower frequencies, ' however, peaks near
T,
are not predominantly due tofluctuation effects. As we will discuss, the large peaks near
T,
at low frequencies found in some experiments inthese materials are presumably due to an experimental
artifact identified by Olsson and
Koch.
' This mechanism may also play a role in Bi2SrzCaCu208, however.The importance
of
thermal fluctuations for the high-temperature superconductors in general is suggested bythe effective two-dimensionality
of
theCu-0
layers, the high transition temperature, and the short coherencelength. Fluctuations affect the conductivity and the nuclear-spin relaxation rate. Strong pair breaking has a
small effect on certain fluctuation contributions (Aslamasov-Larkin diagram' ) while it suppresses others
(Maki-Thompson diagram'
'
). Since the nuclear-spin relaxation, unlike the conductivity, is not affected by the former typeof
diagram, ' we do not expect a large effectof
thermal fluctuations on the nuclear-spin relaxationrate.
We shall see that, although fluctuations lead
to
a peakin the conductivity right at
T„a
distributionof T,
sleads to a peak below the dc critical temperature, in agreement with experiment.
II.
FLUCTUATION-INDUCED CONDUCTIVITY PEAK Above the critical temperature, loweringof
the tem-perature leads to the anticipationof
the superconductingstate due to thermal fluctuations and thus the dc
resis-tance decreases. The contribution to the static conduc-tivity from fluctuations
of
the order parameter was calcu-lated by Aslamasov and Larkin' and generalized for the frequency-dependent conductivity by Schmidt. We willfocus on this contribution here, and will discuss other
terms (such as the Maki-Thompson term) later. The re-sult
of
Schmidt, forthe real part0'
of
the conductivityof
films with thickness d&g(
T),
where g(T) is thecorrela-tion length
of
the fluctuations, isgiven by e16k'de co'
2
,arctan
1
i
In/I+co'
/(T&
T,
),
(1)where
e=(
~T
T,
~)/T,
—
and co'=fiasco/16k&T,
eBelow
T,
the effectof
variationsof
the orderparame-ter around its nonzero mean-field value on the real part
of
the frequency-dependent conductivity were also studied by Schmidt, e 1 4%de ((+co"2)
2 CO 11+m"
ln2(1+co"
)4
where co"=2''.
The expressions (1) and (2) have the scaling form
o'(co)
~
(1/co)F(co/e).
The functionF(x
) goes to a nonzero constant for large valuesof
x,
while it ispropor-tional to
x
forsmallx.
This results (1) and (2)join at
T„
leading to a max-imum fluctuation contribution o2D(T,
)=
(e kT,
s)/
(A'co%'d) at
T,
.
Note, however, that due to thelogarith-mic terms
do'/dT
is infinite atT,
.The width
hT
of
the peak at halfof
its height follows from the criterion (Rirco)/(Skosh,T)
=1.
The valueof
the functionF(x)
then is approximately halfof
its limiting value for largex, i.
e., its value atT,
.
For
a frequencyof
60 GHz this yields
hT
=
1K.
Below the GHz regime the width is unmeasurably small.Although the analysis
of
Schmidt is based on aCxauss-ian fluctuation theory, asimple scaling analysis ' shows that the 1/co behavior at
T,
actually holds moregeneral-ly. In three dimensions,
o'
behaves asI/&co
atT,
in the Gaussian theory, and as co"'
'
according to the scaling434 M. L.HORBACH AND W.van SAARLOOS 46
Equations (1) and (2) have been derived for a homo-geneous film
of
thickness d. In the cuprates the conduc-tion takes place in theCu-0
layers, which are onlyweak-lycoupled. This isespecially true forBizSrzCaCuz08. As
will be discussed further below, except for a region ex-tremely close to
T„
this material behaves essentially two dimensionally. The distance d should then be taken as the distance between theCu-0
layers.In
Fig.
2 we show the resultsof
the computationof
o,
(co)/cr&„(co) with the 2D fluctuation effects taken intoaccount. The normal-state conductivity is taken to be temperature independent, and the fluctuation contribu-tion is added to a behavior without a coherence peak, as could be the result
of
strong-coupling effects. The latteris indicated by the dashed line. The drop
of
themean-field behavior below
T,
is rather drastic in Fig. 2.Whether this is really the casein the cuprates depends on the details
of
the strong-coupling effects. A more smooth behavior is possible. The solid line is the resultinclud-ing the Aslamasov-Larkin-Schmidt fluctuation
contribu-tion. The frequency is taken tobe the frequency at which the experiment
of Ref.
6 (on Bi2Sr2CaCuzOs) was per-formed, 60 GHz, and the normal-state resistance per square was taken to be 3000,
a value reported forBizSrzCaCuz08 in Ref. 26. Thus, without adjustable pa-rameters for the fluctuation contribution, the conductivi-ty enhancement we find is close to the height
of
the peak, which is observed experimentally. In fact, the predicted peak value2.
9 is higher than the experimental valueof
1.
9.
Also notice that at frequenciesof
the orderof
60 GHz the fluctuation peak at halfof
its height isof
theor-der
of
1K,
while at lower frequencies the peak becomes very narrow, as mentioned before. As discussed above, at frequencies much smaller than 60 GHz, fluctuation peaks become extremely narrow (in the absenceof
anybroaden-ing effects). At frequencies much higher than 60 GHz,
on the other hand, the height
of
the peak is too small tolead to observable effects.
) ai
GIN
Of
course, the Auctuation-induced conductivity peaks right atT„while
the experimentally observed peak isslightly below
T, .
We will come back to this inSec.
IV,where we will argue that in the presence
of
adistributionof T,
's the fluctuation peak occurs slightly below the dc critical temperature.Since the
Cu-0
planes are weakly coupled, acrossover to three-dimensional behavior is expected close toT, .
Within the Lawrence-Doniach model, in which the cou-pling is the Josephson type, the crossover occurs whenthe correlation length in the direction perpendicular to
the
Cu-0
layers, g,(T),
becomes comparable to thedis-tance between the
Cu-0
layers. In the caseof
a high an-isotropy g,(T)/g
(T)
is very small, smaller than0.
02 inBizSrzCaCuz08. The crossover to three dimensions then occurs immeasurably close to
T„
typically at0.
9995T,
. Right atT„where
g, diverges, the fluctuation contribution to 0'(co) for the anisotropic three-dimensional case is the three-dimensional Aslamasov-Larkin result enhanced by the anistropy factor g„ /g, . This crossover from the 2D to3D
fluctuation conductivi-ty has been observed experimentally in the dc resistivityvs temperature
of
YBazCu307 at about 1K
away fromT„but,
as mentioned above, for BizSrzCaCuz08 thiscrossover is unimportant. Note, however, that due tothe smaller normal-state conductivity
of
YBazCu307 the effectof
the smaller anisotropy, which leads to a smaller peak in cr&/0.&„due to a crossover to three-dimensional fluctuations, is partially undone. In particular, at 60 GHz the fluctuation effect might still be measurable in YBazCu307. Indeed, inRef.
9 a peakof
height1.
9 in o',/O.,
„
for YBazCu307 at 58.9 GHz was reported.Recently, experiments '
have been performed on
YBazCu307 thin films for frequencies between 50 kHz
and 500 MHz, which yield very sharp enhancements in cr' slightly below
T,
. The observed 1/co frequency depen-denceof
the peak height does not agree with the three-dimensional Aslarnasov-Larkin-Schmidt formula, which yields a 1/&co behavior. Furthermore, as men-tioned before, at frequencies lower than the GHz regime the fluctuation-induced peak is, without broadening dueto a distribution
of T,
's extremely narrow. These effects seem therefore not due to fluctuations. As discussed fur-ther in Sec. IV, these peaks are likely to bean experimen-tal artifact.III.
EFFECTS OFPAIR BREAKINGI I I I I
.94 .96
j
.98 1.00 1.02 T
FIG.
2. Fluctuation-induced peak, at afrequency of60GHz.The normal-state resistance per square ofaCu-0 layer istaken
to be 300 Q. The fluctuation contribution issuperimposed on a mean-field behavior with asuppressed coherence peak, indicat-ed by the dashed line, presumably due to the strong-coupling effects. The dotted curve isthe result with the pair-breaking pa-rameter p
=0.
2.Inline with the fact that pair breaking becomes impor-tant for strong coupling, pair-breaking effects play a role
in the high-temperature superconductors: it has been es-timated that the actual
T,
is a factorof
2 1ower than what it would have been without pair breaking. Pairbreaking can suppress the coherence peak, but it does not affect the Aslamasov-Larkin-Schmidt fluctuation
contri-bution (the Cooper-pair conductivity) above the critical
temperature. The Maki-Thompson contribution'
'
(the contributionof
electron-hole pair scattered into another electron-hole pair by exchangeof
a Cooper-pairn.
[1
pf'(p+
,
')—]-2/'(p+
—,')(3)
Here
f'(x
) is the trigamma function ' and p is a pair-breaking strength parameter, which is related to the quasiparticle scattering time ro by p=
fi/(4771oksT
). With roof
the orderof
10 ' (the value determined ex-perimentally inRef.
32), p isclosetoT, of
the order0.
1.
The effect
of
pair breaking on the fluctuation-induced conductivity peak is indicated inFig. 2.
Here we have taken a temperature-dependent pair breaking proportion-al to (T/T,
),
as was considered inRef.
33in an analysisof
the suppressionof
the NMR coherence peak dueto
pair breaking. The fluctuation contribution is enhanced belowT, .
IV. DISTRIBUTION OF T,'s
In a homogeneous sample with one critical tempera-ture, the fluctuation contribution to o.' peaks at
T„
whereas in the experiment by Holczer etal.
a peak is observed slightly belowT, .
We shall argue that inhomo-geneities that leadto
a distributionof T,
s in a mi-crowave experiment naturally shift the fluctuation peakto
below the critica1 temperature as obtained in adc mea-surement.The
T,
reported inRef.
6was obtained from a dc resis-tivity measurement, which yields the highest temperatureforwhich there exists apercolating superconducting path in the bulk
of
the sample. Ina microwave experiment, onthe other hand, asurface layer is probed which is
expect-breaking.
For
instance, experiments on aluminum films' show asharpeningof
the resistive transition with the ad-ditionof
magnetic impuritiesor
with the applicationof
aparallel magnetic field, both
of
which are pair breaking effects forBCS
superconductors.For
small pair breaking the Maki-Thompson contribution is, well away fromT„
typically one orderof
magnitude larger than the Aslamasov-Larkin contribution. Consistent with our as-sumption that the mean-field coherence peak is absent dueto
strong coupling, it is consistent to neglect Maki-Thompson-type fluctuation contributions.In
any case,these contributions would enhance the conductivity even
more.
Maniv and Alexander, and more recently Kuboki and Fukuyama, have predicted that fluctuations can also enhance the nuclear-spin relaxation rate. However, the enhancement is typically expected tobe weak. Moreover, Aslamasov-Larkin-type diagrams do not occur in the
lo-cal spin susceptibility, which is measured in the nuclear-spin relaxation. Only Maki-Thompson type diagrams determine the fluctuation correction in this case. The
contribution
of
the latter is suppressed in the presenceof
strong pair breaking. Possibly this is the reason that no fluctuation-induced peak is observed in the NMR experi-ments.Below
T,
pair breaking does affect the result (2); it nat-urally enhances the fluctuation effect. This results'
in a changeof
the characteristic time scaleof
the fluctuations,'roL=(~)[16k'(
T
—
T,
)],
toroL/f
(p),
where&IN
.94 .96 .98 1.00 1.02
FIG.
3. Fluctuation-induced peak in case ofa distribution ofT,'s. The horizontal axis is given in units ofT,
'.
The dashed curve is obtained with the pair-breaking parameter p=0.
2. The inset shows some ofthe data points of Ref.6 from T=74
to 100K,with peak height of 1.9and T,
'=91
K.
ed to be
of
a poorer quality than the bulk. Therefore, the dc-transition temperature, which we denote byT,
',
lies on the high-temperature sideof
the distributionof T,
sinthe surface layer. The thickness
of
the surface layer isgiven by the penetration depth, which isthicker forlower frequencies.
In
Fig.
3 weshow results forthe caseof
a Gaussian dis-tributionof T,
's from0.99T,
'
toT,
'.
Except for the widthof
the distributionof
critical temperatures, also the precise formof
the mean-field strong-coupling curve influences the widthof
the peak.Recently, Olsson and
Koch'
have pointed out that a distributionof
critical temperatures can also give rise toa peak in0'
when calculated from the measured complex impedance, which involves both the real part cr' and the imaginary partcr".
BelowT„cr"
has a contributionof
the formp,
/ice from the superfiuid condensate with den-sityp,
.
Since only the total impedanceof
the sample is determined,o'
and 0."
from regions below and aboveT,
get strongly mixed. The widthof
the resulting apparent peak in0'
appears to be roughly the same as the widthof
the distributionof T,
s. Unfortunately, this additional complication will make it quitediScult
to disentangle a fluctuation peak from such nonintrinsic behavior without independent information onthe sample quality, especially since the above effect depends both on the distributionof
T,
's and the behaviorof p,
.
V. CONCLUSIONS
In this paper, we have focused mainly on the experi-ments by Holczer et
al.
on single crystalsof
the highlyanisotropic material BizSr2CaCu208.
For
this material, and for the frequency they used (60 GHz), we find thatthe 2D Aslamasov-Larkin fluctuation contribution is
of
the same order as the peak which is seen experimentally.The location
of
the peak, slightly below the critical tem-perature as obtained from a dc-resistivity measurement,436 M.L.HORBACH AND W.van SAARLOOS
the bulk.
Low-frequency experiments ' (50 kHz
—
500 MHz) on thin filmsof
the less anisotropic YBa2Cu307 also yieldpeaks just below
T,
'.
The frequency dependenceof
their magnitude is inconsistent with the 3D fluctuationcon-ductivity. Also, the shift
of
the peak tobelowT,
'
as wedescribed does not apply for these low-frequency experi-ments on this films, since the penetration depth exceeds the film thickness. Olsson and Koch have observed that sample inhomogeneities can give an apparent peak in
0'
as a function
of
temperature. This effect possibly plays also a role in Holczer's experiment. Therefore, the pre-cise originof
the enhancement seen in BizSr2CaCu208 can only be determined by a more precise analysisof
thedata (for instance, the frequency dependence) and the sample quality. But in any case, we have shown that the fluctuation enhancement is a large effect in
Bi2Sr2CaCu208. Nevertheless, whatever the relative im-portance
of
the two effects is, it appears justified tocon-clude that, contrary
to
the authors' interpretationof
their data, these provide no evidence for the existenceof
acoherence peak.The Gaussian theory,
i.
e.,the theoryof
noninteractingCooper propagators, the modes which signal the instabili-ty
of
the normal state and which drive the phasetransi-tion, is valid only outside the critical region around
T,
.
In the critical regime the Cooper-pair propagator is
re-normalized due to its self-interaction. As discussed in
more detail by Fisher, Fisher, and Huse, critical fluc-tuations may be observable in BizSr2CaCu208. Our analysis shows that in suSciently high-quality Bi2Sr2CaCu208 single crystals, in which the effect dis-cussed by Olsson and Koch disappears, the fluctuation peak is large and measurable.
For
such samples it maythen be possible to see the effects
of
critical fluctuationsin the temperature dependence
of
0.'.
ACKNOWLEDGMENTS
We gratefully acknowledge collaboration with
D.
A.
Huse in the initial stageof
this research. We also have enjoyed stimulating discussions with H.B.
Brom,R.
H.
Koch,
J.
Moonen, and H.K.
Olsson. This work was supported financially by the Dutch Foundation forFun-damental Research on Matter (FOM).
'J.
R.Schrieffer, Theoryof
Superconductivt'ty (Benjamin, NewYork, 1964).
M.Tinkham, Introduction to Superconductivity (McGraw Hill, New York, 1975).
W.W.Warren,
Jr.
,R.
E.
Walstedt, G.F.
Brennert, G. P.Espi-nosa, and
J.
P.Remeika, Phys. Rev. Lett. 59, 1860 (1987).4P.
B.
Allen and D.Rainer, Nature 349, 396 (1991).5C. M. Varma, P.
B.
Littlewood, S. Schmitt-Rink,E.
Abra-hams, and A.E.
Ruckenstein, Phys. Rev. Lett. 63, 1996 (1989);64,497(E) (1990).K.
Holczer, L.Forro, L.Mihaly, and G. Griiner, Phys. Rev. Lett. 67,152(1991).7H.
K.
Olsson andR.
H. Koch,Physics C 185-189,1849(1991). H.B.
Brom (private communication).P.H. Kobrin,
J.
T.
Cheung, W. W.Ho, N.Glass,J.
Lopez,I.
S.Gergis,
R.
E.
DeWames, and W.F.
Hall, Physica C 176,121(1991).
N.
E.
Glass and W.F.
Hall, Phys. Rev. B44, 4495 (1991).'J.
R.
Schrieffer (unpublished).M. C. Nuss, P. M. Mankiewich, M. L. O' Malley, E. H.
Westerwick, and P.
B.
Littlewood, Phys. Rev. Lett. 66,3305(1991).
'
E.
J.
Nicol andJ.
P.Carbotte, Phys. Rev. B44, 7741(1991). '4F.Marsiglio, Phys. Rev.B44, 5373(1991).' H.K.Olsson and
R.
H. Koch (unpublished).L. G. Aslamasov and A.
I.
Larkin, Phys. Lett. 26A, 238(1968).
'
K.
Maki, Prog. Theor. Phys. 40, 193(1968). 'R.
S.Thompson, Phys. Rev. B 1,327(1970). 'R.
D.Parks, Proc. LT12-Kyoto XX, 217 (1970).H.Schmidt, Z.Phys. 216, 336 (1968). H.Schmidt, Z.Phys. 232, 443(1970).
D.S.Fisher, M. P.A.Fisher, and D.A. Huse, Phys. Rev. B
43, 130 (1991).
A.Dorsey, Phys. Rev.B43, 7575(1991).
24W. E.Lawrence and S.Doniach, Proc. LT12-Koyto XX,361
(1970)~
M.L.Horbach, W.van Saarloos, and D.A.Huse, Phys. Rev. Lett. 67,3464(1991).
S.Martin, A.T.Fiory,
R.
M. Fleming, G. P. Espinosa, andA.S.Cooper, Phys. Rev.Lett.62, 677(1989). V.A.Gasparov, Physica C178,449(1991).
K.
Semba, T.Ishii, and A.Matsuda, Phys. Rev. Lett. 67,769 (1991).T.Maniv and S.Alexander,
J.
Phys. C 9,1699(1976).K.
Kuboki and H. Fukuyama,J.
Phys. Soc. Jpn. 58, 376(1989).
3'Handbook of Mathematical Functions, edited by M. Abramowitz and
I.
Stegun (Dover, New York, 1964).D.A. Bonn, P.Dosanjh,
R.
Liang, and W.N. Hardy, Phys.Rev. Lett. 68, 2390(1992)~