You are on page 1of 12

Microwave Spectroscopy of Thermally Excited Quasiparticles in YBa2Cu3O6 99 :

A. Hosseini, R. Harris, Saeid Kamal, P. Dosanjh, J. Preston, Ruixing Liang, W.N. Hardy and D.A. Bonn
Dept. of Physics and Astronomy, University of British Columbia, 6224 Agricultural Rd., Vancouver, BC, V6T 1Z1, Canada
(February 9, 1999)
We present here the microwave surface impedance of a high purity crystal of Y Ba2 Cu3 O6 99 mea-
:

sured at 5 frequencies between 1 and 75 GHz. This data set reveals the main features of the
conductivity spectrum of the thermally excited quasiparticles in the superconducting state. Below
20 K there is a regime of extremely long quasiparticle scattering times, due to both the collapse
of inelastic scattering below T and the very weak impurity scattering in the high purity BaZrO3 -
c

grown crystal used in this study. Above 20 K, the scattering increases dramatically, initially at least
as fast as T 4 .
Valid PACS appear here.

I. INTRODUCTION ments [9] as well as in other electromagnetic absorption


measurements at microwave [2,10{16], far infrared [17]
Over the past few years measurements of electrody- and THz frequencies [8,18,19]. The rapid increase in
namic properties at microwave frequencies have proven quasiparticle scattering time is well established by these
to be a fruitful technique for studying the superconduct- measurements and the very long quasiparticle mean free
ing state of the high temperature superconductors. A paths resulting from this have been corroborated further
key strength of the technique is that measurements of by the thermal Hall e ect measurements of Krishana et
the real and imaginary part of the surface impedance al. [20]. However, obtaining a quantitative determination
Zs (!; T ) provide complementary information on two as- of the scattering time and the details of its temperature
pects of the superconducting state; the super uid density dependence has been hampered by the need to use models
and the low energy excitations out of the condensate. to interpret the existing microwave data. The problem
Measurements of the imaginary part of the surface has been that although the step from Rs (T ) and (T )
impedance Xs (T ) provide a direct measure of the pen- to the conductivity 1 (T ) is a matter of straightforward
etration depth (T ), which is determined by the tem- superconductor electrodynamics when in the local limit,
perature dependence of the super uid density. The the extraction of a scattering time from 1 (T ) su ered
widespread observation of a linear temperature depen- from a dependence on an assumed model for the shape
dence of (T ) at low T in many of the superconducting of the quasiparticle conductivity spectrum 1 (!).
cuprates [1{5] has been a key piece of evidence suggesting In this paper we present measurements at 5 microwave
nodes in the energy gap in these materials. Near Tc the frequencies, giving enough spectroscopic detail over a
temperature dependence of (T ) has provided evidence wide enough frequency range to produce a rather com-
of 3DXY critical uctuations over a wide temperature plete picture of the evolution of 1 (!; T ) in the super-
range in Y Ba2 Cu3 O7  [6]. conducting state. These results support the early model
Measurements of the real part of the surface impedance based on an ansatz that the thermally excited quasiparti-
Rs (T ), when combined with the measurements of (T ), cles have a Drude-shaped conductivity spectrum and now
can be used to determine the real part of the microwave provide a much clearer measurement of the temperature
conductivity 1 (T ), which is essentially electromagnetic dependence of the quasiparticle scattering rate below Tc.
absorption by quasiparticles excited out of the conden- We nd that this scattering rate becomes extremely small
sate (either thermally excited quasiparticles or excita- and appears to become essentially temperature indepen-
tions created by the absorption of photons). Early mea- dent below about 20 K. At higher temperatures the scat-
surements of Rs (T ) at a few GHz exhibited a broad peak tering rate increases rapidly, initially at least as fast as
below Tc , caused by a very large peak in 1 (T ) [7] at T 4. In the following section we will describe the tech-
these low frequencies. This peak was also observed at niques used to produce this information and in Section
THz frequencies by Nuss et al. [8], and was attributed III we will present the results along with details of the
to a competition between two temperature dependences; extraction of the microwave conductivity from surface
the overall decrease with temperature of the number of impedance measurements. In Section IV the microwave
thermally excited quasiparticles competing with a rapid conductivity spectra will be presented along with the re-
increase below Tc of the transport scattering time of these sults of ts that generate the temperature dependence of
quasiparticles. This rapid increase in scattering time has the quasiparticle scattering rate. In Section V we will
been interpreted as a collapse of the inelastic scattering point out some of the implications of these results and
processes responsible for the large normal state resistiv- in particular compare them to the present literature on
ity of the high temperature superconductors and is an conductivity in a d-wave superconductor.
e ect that is observed in thermal conductivity measure-

1
II. EXPERIMENTAL TECHNIQUES weak in the c^-direction, except near Tc [21,22]. The ab-
solute value of Rs (T ) for the c^-direction is also quite low
The property being directly probed in a microwave [22].
measurement on a superconductor is the complex sur- The measurements at 1 GHz were performed in a loop-
face impedance Zs (T ) = Rs (T ) + iXs (T ). One simpli- gap resonator initially designed for measurement of (T )
fying feature of the high temperature superconductors [1]. Like most of the resonators used in this study, the
is that the super uid response falls in the limit of local loop-gap is plated with a Pb:Sn alloy which is super-
electrodynamics, so that Rs (T ) and Xs (T ) are related to conducting below 7 Kelvin and has very low microwave
the complex conductivity (!; T ) = 1 (!; T ) i2(!; T ) loss at 1.2 Kelvin. In the case of the 1GHz loop-gap,
in a straightforward way. This simple limit arises from the cavity Q can be as high as 4  106 at low tem-
the very small coherence length in these materials, which perature. Another feature common to all of the mea-
guarantees that   . A particularly simple regime of surements is that the sample is mounted on a thin sap-
the local limit occurs when 2  1 below Tc and at low phire plate with a tiny amount of silicone grease, with
frequency, in which case one obtains the relations the thermometry and sample heater located at the other
end of the sapphire plate, outside of, and thermally iso-
2
Rs (T ) = 20 !23 (T )1 (!; T ) (2.1) lated from the resonator. In this way the resonator can
be held xed at the regulated 4 He bath temperature
Xs (T ) = 0 !(T ) : while the sample temperature is varied. A unique fea-
A more detailed discussion of the electrodynamics and ture of the 1 GHz loop-gap system is that the sample is
the extraction of 1 (!; T ) from Zs (T ) will be presented held xed in the resonator, with the sapphire plate and
in Section III, but the equations above are useful for quick thermometry stage supported by a thin walled quartz
estimates and for understanding the main features of the tube which sustains the temperature gradient between
microwave properties of these superconductors. the 4 He bath and the sample. This means that the sam-
The expression for Rs (T ) in Eq. 2.1 embodies what ple cannot be removed from the resonator during the
is most dicult about the microwave measurements. It measurements, but we nd this restriction is necessary
contains both 1 and a term of the form !2 3 , and physi- for the high precision measurements of (T ) which rely
cally can be interpreted as the microwave absorption pro- on the sample being held rigidly in a position that does
cesses (1 ) occuring within the rather shallow depth into not vary when the temperature is changed. Without
which the microwaves penetrate (hence the term !2 3 ). this type of construction, motion of the sample and sap-
This screening length set by the super uid makes Rs ex- phire plate in the elds of the resonator can give rise
tremely small below Tc , such that for small single crys- to experimental artifacts. This is because (T ) is de-
tals, one is forced to employ cavity perturbation in very termined from very small changes in the resonator fre-
high Q microwave resonators. This is most often achieved quency as the sample temperature is changed. This tech-
by the use of microwave cavities made from conventional nique gives highly precise and reliable temperature de-
superconductors cooled to low temperature. This xed pendences relative to the base temperature, but is limited
frequency type of measurement must be performed with to measuring di erences; (T ) = (T ) (1:2 K ) and
several resonators if one is to build up a complete picture Rs (T ) = Rs (T ) Rs (1:2 K ). The value of Rs (1:2 K )
of the microwave conductivity spectrum 1 (!). at 1 GHz is estimated by comparing to separate experi-
The data presented here involve measurements with 5 mental runs where the resonator is loaded with only the
resonators spanning a 1 to 75 GHz range and utilizing a sapphire sample holder. The need to do this in two sep-
number of variations on the basic method of cavity per- arate experimental runs gives the 1 GHz Rs (T ) mea-
turbation. The common feature is that all of the mea- surements a relatively large uncertainty in the form of
surements have been performed on the same sample, a a temperature-independent background value (0:7
),
thin plate of Y Ba2 Cu3 O6:993 oriented such that the mi- although the resolution of the temperature dependence
crowave magnetic eld of each cavity lay in the plane of is much better than this. We do not attempt to ex-
the plate (H~ rf k ^b). This geometry has the advantage tract (1:2 K ) from the microwave measurements, but
that demagnetization factors are small, so the surface instead use values inferred from muon spin rotation mea-
current distributions are fairly uniform and are similar surements [23]. None of the analysis discussed in this
in all of the measurements, making comparison from fre- paper depends sensitively on this choice of (1:2 K ).
quency to frequency quite reliable. The disadvantage is The 2 GHz measurements have been performed in a Nb
that although this geometry mainly measures the surface split-ring resonator that will be described in more detail
impedance for currents running across the crystal face in elsewhere [24]. Measurements at 13.3, 22.7, and 75.1 GHz
the a^ direction, there is some admixture of c^-axis surface have been performed in the axial microwave magnetic
impedance coming from currents running down the thin elds of the TE011 modes of three right-circular cylindri-
edge of the crystal. However, we have previously shown cal cavities. In all of the measurements other than those
that these e ects are small for a thin crystal, because the at 1 GHz, the sample can be easily moved in and out of
temperature dependences of Rs (T ) and Xs (T ) are quite the resonator. This freedom to move the sample makes it
dicult to measure (T ) at the higher frequencies, due to

2
problems of controlling motion of the sample as the tem-
perature is changed. However, the sample motion is much
too slight to have any in uence on the measurements of
Rs (T ). The ability to measure Q0 by pulling the sam- 0

ple out rather than doing a separate experiment makes 10


the background uncertainty in these measurements much 1.14 GHz
2.25 GHz
less signi cant than it is in the 1 GHz data. The sur- 13.4 GHz
face resistance is determined from Rs (T ) / 1=Qs 1=Q0 22.7 GHz
where Q0 is the Q of the empty cavity and Qs is the Q 10
−2 75.3 GHz

with the sample inserted. A small correction for other


sources of loss must be made by measuring the sapphire
holder and grease without the sample. Then the only re-

Rs(Ω)
maining uncertainty in the background is the in uence of −4

non-perturbative e ects, which we check by measuring a 10

superconducting Pb:Sn sample at 1.2 K. With these cor-


rections, the estimated uncertainty in the background is
0.7, 0.2, 10, 20, and 360 
for the 1.1, 2.2, 13.4, 22.7,
and 75.3 GHz data respectively. The calibration of the 10
−6

absolute value of the surface resistance is obtained by


measuring a Pb:Sn reference sample whose normal state 0 20 40 60 80
surface resistance is governed by the classical skin e ect. T(K)
These calibrations, together with the statistical scatter
in the data amount to a total uncertainty of  10% or FIG. 1. The surface resistance R (T ) for currents running
less at each frequency. s

in the a^ direction of a high purity crystal of Y Ba2 Cu3 O6 993 .


The sample used for these measurements was a single :

crystal of Y Ba2 Cu3 O6:993 grown by a ux growth tech-


nique using BaZrO3 crucibles [25]. The BaZrO3 , which
does not corrode during crystal growth, yields crystals
of much higher purity [25,26] (better than 99.99%) than
those grown in more commonly used crucibles such as
yttria-stabilized zirconia and alumina. The sample was
detwinned at 250 C under uniaxial stress and then an-
nealed at 350 C for 50 days to produce a sample with 1.14 GHz
nearly lled chain oxygen sites. This gives a slightly 2 2.25 GHz
lower Tc (88.7 K) than the maximum obtained near an 13.4 GHz
oxygen concentration of 6.91 [25], but provides partic- 22.7 GHz
ularly defect-free samples without the oxygen vacancy 75.3 GHz

clustering discovered by Erb et al. [27]. The dimensions


Ωs )

of the sample were initially 2x1x0.02 mm3 for the 1 GHz


2

measurements, which have been reported elsewhere [28].


−25

Subsequent measurements in the other resonators were


Rs/ω (10

1
performed on smaller pieces cleaved from the original
2

crystal, but measurements at 22.7 and 75.3 GHz were


repeated on di erent pieces to ensure that the sample
was uniform.

III. EXPERIMENTAL RESULTS AND ANALYSIS 0


0 20 40 60 80
Fig. 1 shows the surface resistance of Y Ba2 Cu3 O6:993 T(K)
for currents running in the a^ direction. At all 5 of the
frequencies shown in the gure, the rapid drop in Rs (T ) FIG. 2. The same measurements of surface resistance as
below Tc is due to the onset of screening by the super- shown in Fig. 1, but with the behaviour below T emphasized c

uid, with the overall magnitude of the drop depending by dividing out a frequency-squared dependence associated
strongly on frequency as expected from the term in Rs (T ) with super uid screening.
that varies as !2 3 (T ).
For easier comparison of the di erent frequencies, it is

3
convenient to plot the low loss data in the superconduct-
ing state as Rs =!2 versus temperature, as shown in Fig. 1.0
2. One feature of these gures is that above 65 K the loss
scales as !2 to within 8%, a scaling that is expected if
the microwave 1 (!) is frequency independent above 65
K. This agreement lies within the estimated uncertainty
in background and calibration constants, and provides a
check of the degree to which one can compare the mea-

λ (0)/λ (T)
surements at di erent frequencies.

2
The broad peak in Rs (T ) was originally attributed to 0.5

a quasiparticle scattering time  that increases rapidly

2
with decreasing temperature below Tc and competes with
a density of these thermally excited qusiparticles that de-
creases with temperature [7]. For the early measurements
near 2 GHz it was suggested that the quasiparticle scat-
tering time reaches a limiting value near 30 K, possibly
due to impurity scattering, at which point the quasipar- 0.0
ticle density takes over and causes Rs (T ) to fall again 0 20 40 60 80
with further decreases in temperature [29]. The specula- T(K)
[t]
tion that impurities are involved in the turnover at 30 K
was partly checked by studying samples doped with Ni FIG. 3. The temperature dependence of the a^ axis pene-
and Zn [30]. These doping studies showed either a smaller tration depth of Y Ba2 Cu3 O6 95 plotted as 1=2 (T ), which is
:

peak shifted to higher temperature or no peak at all, con- a measure of the super uid density via n e2 =m = 1=(0 2 ).
s

sistent with the quasiparticle scattering rate running into


an impurity limit at higher temperature, even when only
0.15% Zn was added to the sample. This sensitivity to tract 1 (!; T ) from these measurements of Rs (!; T ). At
such low levels of impurities raises a serious concern over the lowest frequency of 1 GHz, where !  1 and we
these earlier crystals grown in yttria-stabilized zirconia have simultaneous measurements of Rs (T ) and (T ), it
crucibles, because during crystal growth the residual im- is straightforward to extract 1 (T ) with only an assump-
purity level due to uptake of material from the corroding tion that the electrodynamics are local. The 1 GHz mea-
crucible reaches the 0.1% level. The results shown here surements of (T ) used in this analysis are shown in Fig.
on a new higher purity sample con rm the original spec- 3, plotted as 1=2 (T ).
ulation that it is this residual impurity scattering that The screening by the super uid follows the local Lon-
limits the increase in quasiparticle lifetime, even in quite don model if a superconductor is in the limit   .
pure crystals. The low frequency surface resistance of Strictly speaking, this local limit is more complicated for
the new crystal rises a factor of four above the minimum the case of a superconductor with nodes in the energy
near 70 K, higher than the factor of two or less observed gap because the coherence length is then k-dependent
in earlier measurements of samples at 2 and 4 GHZ [29]. and becomes large in the node directions. However, the
The changes are consistent with the new samples having consequences of this type of non-locality for the electro-
lower impurity scattering and so reaching a higher quasi- dynamics would only be observable at low temperatures
particle scattering time limit at a lower temperature than and the e ect would be small in the measurement geom-
was seen in the yttria-stabilized zirconia grown crystals. etry used here [31]. A much more serious problem for
The frequency dependence of the peak in Rs (T ) pro- the data analysis is that at the higher frequencies where
vides further evidence of the rapid increase in the quasi- !  1, the thermally excited quasiparticles also con-
particle lifetime. In much higher frequency measure- tribute to the screening of microwave elds and it is then
ments on thin lms, Nuss et al. [8] were the rst to ob- incorrect to use the 1 GHz measurements of (T ) in order
serve a similar broad peak in the THz range that shifted to extract 1 (T ) from Rs (T ). E ectively, the penetration
up in temperature and decreased in size at higher fre- depth becomes frequency dependent, a phenomenon that
quencies. They pointed out that this could be accounted has been observed directly in mm-wave measurements on
for by relaxation e ects. If the quasiparticle lifetime in- particularly high quality thin lms [16]. Ideally this prob-
creases suciently that !  1, then the sample enters a lem can be solved by measuring both Rs (T ) and Xs (T ) at
regime where 1 (T ) falls as  continues to increase. So for each measurement frequency, but we have done this only
high frequency measurements the temperature at which at 1 GHz, for reasons discussed in the previous section.
Rs (T ) reaches its peak roughly indicates where ! = 1 It is possible to work around the fact that we have
at each measurement frequency. The fact that we ob- measurements of both the real and imaginary part at
serve these relaxation e ects at 13 GHz indicates that only one frequency, as long as there is adequate infor-
the scattering time  exceeds 10 ps in the new samples. mation on the frequency dependence of the real part of
To better understand the data it is desireable to ex- the surface impedance. The problem is analogous to the

4
one faced in far infrared spectroscopy of opaque samples, delta-function can be neglected, leaving
where the re ectance, but not the phase of the re ected  
light, is measured over a wide frequency range. Kramers-
Kronig relations are the usual solution if only one of the
(!; T ) = 1N (!; T ) i 2N (!; T ) +  !12 (T ) :
0
optical constants is known, but is known over a wide fre- (3.2)
quency range. In principle, data at the ve microwave
frequencies presented here could be connected to far in- Thus, in general the real part of the conductivity comes
frared measurements of re ectance in order to perform from the normal uid. The imaginary part has contribu-
this analysis and extract 1 (!) and 2 (!). However, such tions from both the normal and super uid, although the
an analysis is dicult because there still exists a substan- super uid dominates at low frequency.
tial gap in the mm-wave frequency range between our In the local limit the connection between 1 (!; T ) and
highest frequency measurement and the lowest frequency the surface impedance is made via
far infrared measurements on crystals. Some data have
been obtained in the mm-wave region using techniques  i !  12
such as time domain THz measurements [8,18,19] and Zs = Rs + iXs =  0 i : (3.3)
direct infrared absorption [32], but these have all been 1 2
performed on lms rather than untwinned single crys- Thus, at 1 GHz where we have measurements of both Rs
tals. The quasiparticle scattering rate is typically much and Xs , 1 and 2 can be extracted using
higher in lms than it is in single crystals, so data taken
on such di erent samples cannot be analyzed together in
this way. An alternative to a Kramers-Kronig analysis 1 = 20 ! (R2R+s XXs 2 )2 (3.4)
is to t the frequency dependent surface resistance to a X
s
2 R
s
2
model, but this is not a very satisfactory procedure if 2 = 0 ! (R2s+ X 2s)2 :
one only has data at ve microwave frequencies and one s s
doesn't know the shape of the conductivity spectrum a At higher frequencies, where we only have measure-
priori. A further problem with both of these techniques
is that the far infrared data is only available at a couple ments of Rs (T ), it is useful to write 1 in terms of Rs
of temperatures below Tc, so they can not be used to do and 2 in the following way [29]:
a complete analysis of the data presented here. 2"  2  1 =2 # 2 31=2
Our main approach to analyzing the surface resistance 
data is to use the Rs (T ) measurements at 1 GHz to arrive 1 = 4 2  4s 2 s
s
22 5 (3.5)
at an estimate of how much screening by the thermally p
excited quasiparticles must be included when extracting with the sign choice + ( ) for 1 > (<) 32
1 (T ) from Rs (T ) at higher frequencies. Although the
procedure involves some assumptions about the model and where s = 2R0 !2 :
for the screening, we will show that the corrections are s

small enough that the e ect of uncertainty in the choice of In the normal state, where 1  2 at low frequen-
model does not signi cantly a ect the conductivities that cies, this expression reduces to the classical skin e ect
we extract in the analysis. We begin by writing down a result 1 = 0 !=2Rs2. Eq. 3.5 continues to be valid right
general 2- uid expression for the microwave conductiv- through the superconducting transition and can be used
ity that includes contributions to the real and imaginary to extract 1 from measurements of Rs provided that one
part from both the super uid and normal uid (the con- also has some information on 2 . The extent to which
ductivity due mainly to thermally excited quasiparticles), one can use this expression in the superconducting state
depends upon the relative importance of the normal uid
(!; T ) = 1S i2S + 1N i2N (3.1) contribution to 2 . In regimes where this contribution is
=  nms e (!) ns e2 +  small, 2 is simply given by 2S = (0 !2 (T )) 1 which
2
im ! 1N i2N can be calculated from the penetration depth measure-
ments. To clearly illustrate the problem that occurs when
where 1S and 2S are the real and imaginary parts of the the normal uid contribution to 2 is not small, we ex-
super uid conductivity and 1N and 2N are the real and amine a particular version of the 2- uid model where the
imaginary parts of the normal uid conductivity. The su- normal uid conductivity is assumed to have a Drude
per uid contribution consists of a delta-function at ! = 0 frequency dependence:
with an oscillator strength given by the super uid den-
sity divided by the e ective mass (ns =m), and an imag- 2  
1N i2N = nmn e 1 + (! )2 i 1 +!(! )2 (3.6)
2
inary part that is the inductive response of the super-
uid at nonzero frequencies. This super uid response
can be expressed in terms of the penetration depth us- where  is the scattering time of the normal uid and
ing ns e2=m = (0 2 (T )) 1 , and away from ! = 0 the nn =m is the normal uid density over the e ective mass.

5
The ratio of the normal uid and super uid contributions
to the e ective screening is then 6

2N = nn (! )2 :
1.14 GHz

(3.7)
2.25 GHz

2S ns 1 + (! )2 13.4 GHz


22.7 GHz

The normal uid contribution to screening is thus unim-


75.3 GHz

portant at low frequency (!  1) or when the normal 4


uid density is small (nn =ns  1, which occurs at low

σ1(10 Ω m )
−1
temperatures). Conversely, diculties arise when ! > 1

−1
and nn =ns is not small, which is likely the case for our

7
data at 13 and 22 GHz in the temperature range from
20{40 K, and for the 75 GHz data from 20{60 K. 2
For the conductivities that we will show below we
take the following approach to estimating the normal
uid contribution to 2 . At 1 GHz 1 is calculated di-
rectly from the simultaneous measurements of Rs (T ) and
Xs (T ). Then we extract an estimate of  (T ) using the 0
Drude model above (Eq. 3.6). The normal uid density 0 20 40 60 80
is calculated from the penetration depth measurements T(K)
by assuming that
FIG. 4. The temperature dependence of the a^-axis mi-
nn e2 (T ) + ns e2 (T ) = ns e2 (T = 0) : (3.8) crowave conductivity of Y Ba2 Cu3 O6 993 extracted from the
m m m
:

surface resistance measurements of Fig. 1. The sharp spike


which amounts to assuming that all of the low frequency near T is a result of superconducting uctuations and the
c

oscillator strength is being shifted from the normal uid broad peak at lower temperatures is caused by the increase
to the super uid as the temperature is decreased. The in the scattering time of thermally excited quasiparticles.
 (T ) and nn (T ) derived from the 1 GHz data can then be
used to make an estimate of the normal uid contribution of Y Ba2 Cu3 O7  crystals have shown the presence of an
to 2 needed to analyze the data at higher frequencies. extra sample-dependent peak in 1 (T ) just below Tc, a
When this analysis is performed, we nd that the normal feature that was discussed by Olson and Koch [34] and
uid screening in uences the extraction of 1 from Rs (T ) by Glass and Hall [35] who attributed it to having a sam-
only at the level of 20% or less. This can be understood ple with a broadened transition. A similar feature was
if one notices that !  1 at temperatures of 50 K or also reported recently by Srikanth et al. for BaZrO3 -
less at which point the normal uid density has already grown crystals [36], but since we see no sign of this in
fallen below 20% of the total oscillator strength. There our new high purity samples, we conclude that such fea-
is some uncertainty associated with assuming a Drude tures are associated with a spread in Tc0 s in the surface
form for 1 (!) (Eq. 3.6) and assuming the redistribution of the sample. The main feature that we do observe in
of oscillator strength implied by the two uid model (Eq. the conductivity is that 1 (T ) has a large broad peak,
3.8). However, we will show that for temperatures up rising to nearly 25 times the normal state conductivity.
to 45 K, deviations from a Drude spectrum are barely The peak rises a factor of two higher than was seen in
discernible within our experimental uncertainties and the measurements at 2 and 4 GHz on an earlier generation
two uid model is obeyed at microwave frequencies to of crystals grown in yttria-stabilized zirconia crucibles
within 10%. Thus, the maximum systematic error that [30] and it crests at a lower temperature, 25 K instead
might be introduced by this analysis amounts to 3% at of the 35 K turnover observed in the lower purity crys-
worst at 75.2 GHz, only about 1% at 13 and 27 GHz, and tals. This e ect of very low levels of impurities is not
is completely negligible at 1 and 2 GHz. Furthermore, the consistent with the suggestion of Klein et al. [37] that
values of  (T ) that we use in the analysis above turn out this feature is the result of BCS-type coherence e ects.
to be consistent with the conductivity spectra 1 (!) that A coherence peak, essentially a density of states e ect,
will be discussed below, so the correction for screening by is observed near Tc in 1 (T ) of s-wave superconductors
the normal uid is self-consistent. such as Pb [38]. However, the peak observed here at 1
Fig. 4 shows the conductivities extracted from the GHz is much too large and too low in temperature to be
Rs (T ) data of Fig. 1, using the methods described above. attributed to such an e ect. Furthermore, a strong co-
The sharp upturn at Tc marks the presence of super- herence peak has not been seen in NMR measurements
conducting uctuations, which have been discussed in of T1 in this material [39] nor is it expected in a d-wave
greater detail by Kamal et al. [6] and Anlage et al. [33] superconductor, the now widely accepted pairing state of
and are not the main focus of the work presented here. A Y Ba2 Cu3 O7  . [40,41]
number of previous measurements on earlier generations

6
4K 35 K
10 K 50 K
20 K

4 4
σ1(10 Ω m )

σ1(10 Ω m )
−1

−1
−1

−1
7

7
2 2

0 0
0 10 20 30 40 50 0 10 20 30 40 50
10 −1 10 −1
ω(10 s ) ω(10 s )
[t] [t]
FIG. 5. The conductivity spectrum at 3 selected tempera- FIG. 6. The conductivity spectrum at 2 temperatures
tures between 4 and 20 K, extracted from the 1 (T ) curves above 20 K, extracted from the 1 (T ) curves of Fig. 4. Above
of Fig. 4. In this temperature regime, the conductivity due 20 K, the width of the conductivity peak broadens rapidly,
to thermally excited quasiparticles has a nearly temperature stretching out of the microwave frequency range above 55 K.
independent width of 91 GHz and a nearly temperature in- These spectra continue to be reasonably well t by Drude
dependent shape that is close to a Drude lineshape (the lines lineshapes, as shown by the lines in the gure.
are Drude ts).
IV. CONDUCTIVITY SPECTRA AND
Thus, in the absence of strong coherence e ects, we QUASIPARTICLE LIFETIME
have attributed the rise in 1 (T ) below Tc to a rapid in-
crease in the scattering time  of thermally excited quasi- The evolution of the conductivity with temperature is
particles, as discussed in the Introduction. The fact that better illustrated in Figs. 5 and 6 where we show the con-
this peak rises higher and turns over at a lower tempera- ductivity spectrum 1 (!) at several representative tem-
ture in the new, higher purity crystals is consistent with peratures. In fact, the central technical achievement of
this interpretation. That is, in the higher purity sample, this work is that we now have measurements at enough
 runs into its impurity limit at a somewhat lower tem- frequencies that both the shape of 1 (!) and its tem-
perature than it did in the earlier generations of crystals perature dependence are quite clear. Fig. 5 shows the
and this impurity-limited scattering time is very large in conductivity spectrum at three temperatures below 25
the new BaZrO3 -grown crystals. An estimate of the in- K, the temperature range where we have argued above
crease can be made as follows. The penetration depth that the conductivity is dominated by thermally excited
measurements indicate that more than 90% of the nor- quasiparticles scattered by a low level of impurities. We
mal uid density is gone at 25 K (see Fig. 3), so the will not concentrate here on the lowest temperature data,
25-fold increase in 1 between Tc and 30 K implies an where the loss becomes very small and dicult to mea-
increase in  by at least a factor of 250 over the scatter- sure accurately by most cavity perturbation techniques.
ing time just above Tc. This is actually an underestimate Fig. 5 shows that from 4 to 20 K, the conductivity con-
because not all of the far infrared oscillator strength in sists of a very narrow peak whose width is largely temper-
the normal state ends up condensed into the super uid ature independent. The lines in the gure are weighted,
at low temperatures. Such a huge increase in lifetime least squares ts to a Lorentzian lineshape (the real part
is consistent with the relaxation e ects observed in the of Eq. 3.6), demonstrating that the conductivity spec-
data at 13 GHz and higher. As a rough illustration of trum of the thermally excited quasiparticles is Drude-
this agreement, far infrared measurements indicate that like. The main contributions to the error bars in these
!  1 at about 3000 GHz just above Tc, so a 300-fold spectra are a combination of uncertainty in the individual
increase in  below Tc would mean !  1 at about 10 measurements' calibration constants, plus a substantial
GHz, leading to the considerable frequency dependence uncertainty in the background loss in the 1.1 GHz mea-
in 1 (!) that we observe in the microwave range. surements, and scatter in the 2.2 GHz measurements.
Within these error estimates (mostly sytematic error)
there is very little clear deviation from a Drude lineshape

7
when examining an individual spectrum at a particular
temperature. The most noticeable deviation is in the
spectrum at 4 K, where the data may be taking on a 60 From fits
From 1/λ
2

slightly more cusp-like shape than the Drude curve. The


nearly Lorentzian lineshape largely con rms the ansatz
originally used by Bonn et al. [29] to analyze the early
Rs (T ) measurements.

nne /m (10 Ω m s )
−1 −1
An interesting cross-check of these ts to 1 (!) is
40

to compare the oscillator strength in the normal uid

−1
conductivity peak, which is given by the t parameter

17
nn e2 =m, to the super uid density ns e2 =m = (o 2 ) 1
extracted from the penetration depth measurements. If 20
one assumes that all of the oscillator strength ends up

2
in the super uid -funtion as T!0, then the super uid
density is related to the normal uid density via Equa-
tion 3.8. Fig. 7 shows a comparison between the normal
uid density inferred from the ts to the 1 (!) peaks 0
and the normal uid density inferred from the penetra- 0 10 20 30 40
tion depth via Eq. 3.8. This gure indicates that the T(K)
normal uid density does nearly track the decline of the
super uid density with increasing temperature, which in- FIG. 7. A comparison of the normal uid oscillator
dicates that our use of a two- uid model to describe the strength determined in two ways; from Drude ts to spec-
screening is a reasonable procedure. The increasingly se- tra like those of Figs. 5 and 6 and from the disappearance of
rious deviation betweeen the curves above 30 K is an oscillator strength in the super uid response, as measured by
indication that the Drude lineshapes do not completely 1=2 (Fig. 3).
keep track of where all of the oscillator strength is going
as temperature increases. There is also some deviation been obtained in measurements on thin lms by high fre-
at low temperatures, taking the form of a normal uid quency microwave and THz techniques [8,19]. However,
oscillator strength that is extrapolating linearly towards the much higher level of defects in such samples limits
a non-zero value as T!0. This is an indication of the the temperature range over which the evolution of the
presence of residual conductivity in the low temperature, inelatic scattering rate can be observed. In the high pu-
low frequency limit, which is expected for a d-wave su- rity crystals, the temperature dependent scattering can
perconductor [42]. be tracked all the way down to 20 K. A t to the form
The quality of the ts to 1 (!) and the agreement in 1= (T ) = A + B (T=Tc)y from 4 to 40 K yields an ex-
the oscillator strengths shown in Fig. 7 indicate that ponent y = 4:2  0:1, with coecients A = 5.2(0.4)
the Drude ts do provide a reasonable measure of the 1010 s 1 and B = 4.6(0.9) 1012 s 1 , and is shown
width of the peaks from 4 to 45 K. The temperature as the solid curve in Fig. 8. The uncertainty in the ex-
dependent width coming from these ts, which we in- ponent indicates the maximum range that is consistent
terpret as the scattering rate of the thermally excited with the data to within the estimated errors shown in
quasiparticles, is shown in Fig. 8. One of the key re- Fig. 8.
sults of these measurements is that the width of the nor-
mal uid peak is very small, 91 GHz, and it is nearly
temperature-independent up to 20 K. The main change in V. DISCUSSION
the spectra in this temperature range is an increase in os-
cillator strength, due to the shift of spectral weight from For the purposes of discussing the data, we have di-
the super uid response (a -function at !=0) to the mi- vided the conductivity spectra up into the two regimes
crowave conductivity. The narrow width suggests a very discussed in the previous section. In the range below
long quasiparticle scattering time of 1.8(0:2)  10 11 s. about 20 K where the width of the peak in 1 (!) is
One sees in Fig. 6 that above 25 K the conductivity narrow and nearly temperature independent, studies of
peak broadens rapidly and by 60 K the width becomes samples over a wide range of purities indicate that this
much greater than the frequency range of the microwave regime is governed by thermally excited quasiparticles
measurements, so we have no direct measure of the width being scattered by impurities or other defects [30,50].
and shape of 1 (!) above 60 K. The exceptionally low Waldram et al. have pointed out the possibility that
level of defects in the new BaZrO3 grown crystals, cou- non-local e ects might come into play in the conductiv-
pled with the measurements at 5 microwave frequencies, ity in this regime, leading to an e ective scattering rate
allows us to determine the temperature dependence of the that is not in uenced by the density of residual impurities
inelastic scattering rate over a fairly wide temperature [43]. However, the considerable narrowing of 1 (!) that
range from 20 to 40 K. Qualitatively similar results have

8
Fermi surface. In particular, it has been pointed out by
2 Hirschfeld et al. that elastic impurity scattering in this
situation should lead to a frequency and temperature de-
pendent scattering rate because of the restricted phase
space into which the quasiparticles at the nodes can scat-
ter [44,46]. So, with this possible con ict between theory
and the phenomenological model in mind, we will make
some more detailed comparison between the microwave
1/τ(10 s )
11 −1

conductivity data and the relevant theoretical calcula-


1 tions.
The transport properties (both microwave conductiv-
ity and thermal conductivity) of a dx2 y2 superconductor
have been the subject of considerable theoretical e ort
recently [42,44{48]. This work builds on earlier calcula-
tions of the transport properties of anisotropic supercon-
ductors, aimed primarily at explaining and predicting the
0 properties of heavy fermion superconductors [51,52]. The
0 10 20 30 40 high temperature superconductors now o er an opportu-
T(K) nity to test these calculations in a situation where we
have a relatively simple anisotropic pairing state. Such
FIG. 8. The scattering rate of the thermally excited quasi- comparisons are somewhat complex because the question
particles, as inferred from the width of Drude ts to the con- of transport properties at low temperatures is inherently
ductivity spectra of Figs. 5 and 6. The solid curve is a t to a question of understanding impurity e ects. This is es-
a scattering rate that increases as T 4 2 .
:
pecially the case for anisotropic superconductors because
the presence of impurities has a strong impact on the
excitation spectrum near gap nodes, particularly in the
we have observed upon going from YSZ-grown crystals limit of unitary scattering.
to the higher purity BaZrO3 -grown crystals indicates A key e ect of impurities in an anisotropic supercon-
that the samples are still in a regime where impurities ductor is to produce a band of impurity states with a
play a role in the low frequency scattering. We have width , thus giving the superconductor a non-zero den-
previously suggested an intuitively appealing way to in- sity of states at the Fermi level. One surprising conse-
terpret the spectra in this regime, based on a speci c quence of these states is a universal conductivity limit at
version of the two uid model [29]. In this phenomeno- low frequency as T ! 0, rst pointed out by P.A. Lee
logical picture, the quasiparticles excited near the nodes [42]. This residual conductivity is independent of impu-
in the energy gap have a temperature-independent scat- rity concentration, the result of a cancellation between
tering rate due to elastic scattering by impurities and the density of states induced by the presence of the im-
a conductivity spectrum with a Drude lineshape whose purities and the lifetime associated with those states. A
width is set by this scattering rate 1=i just like impu- version of this universal limit has been observed by Taille-
rity scattering in a normal metal. In the high purity fer et al. in thermal conductivity measurements of pure
samples this 1=i would correspond to a strikingly long and Zn-doped Y Ba2 Cu3 O6:9 below 1 Kelvin [53]. To the
mean free path of 4 m if one takes the Fermi velocity best of our knowledge, this limit has not yet been de ni-
to be vF = 2  107cm=s. In this particular two uid tively observed in microwave conductivity measurements,
model, the only source of temperature dependence in the in part due to sensitivity problems in the type of cavity
low temperature microwave conductivity is the density perturbation measurement being discussed in this arti-
of the thermally excited quasiparticles, which increases cle. Instead, our main concern here will be the behaviour
linearly with temperature. This is a straightforward con- of this conductivity as temperature and frequency are
sequence of the linear dispersion of the gap function near increased, which involves conductivities that are sub-
the nodes and is also intimately connected to the linear stantially larger than the T!0 limit and are thus eas-
temperature dependence of the penetration depth (they ily measurable with the methods discussed above. This
are Kramers-Kronig related). microwave conductivity has been the subject of consid-
Parameterizing the normal uid response in this way, erable theoretical e ort, including both numerical work
with a temperature dependent density and a scattering and analytical results in certain limits [44{48].
rate, has been partly justi ed by calculations of the mi- One well studied case involves the electrodynamic
crowave conductivity of a dx2 y2 superconductor [44]. properties at temperatures and frequencies below the im-
However, it is not so obvious that a temperature indepen- purity bandwidth , in the unitary scattering limit (scat-
dent scattering rate is expected for impurity scattering tering phase shift ! =2). In this limit the impurity
of these thermally excited quasiparticles near the nodes, bandwidth is given roughly by  ( )1=2 where 
since they di er greatly from free carriers at an ungapped is the magnitude of the gap and is the elastic scat-

9
1.14 GHz
tering rate that the impurities would contribute to the 2.25 GHz
normal state resistivity. For !; T < , where the trans- 13.4 GHz
port properties are dominated by this impurity band, it 4
22.7 GHz
has been shown that both 1 and  vary as T 2 [44,45]. 75.3 GHz

σ1(10 Ω m )
−1
This quadratic behaviour has been seen in Zn-doped sam-
ples of Y Ba2 Cu3 O6:95 , where it was found that at a Zn

−1
impurity concentration as low as 0.15 % the crossover

7
energy scale is already > 4 K [49,30]. However, Zn
substitution for planar Cu's is the only impurity that we 2
have found that clearly gives this unitary scattering be-
haviour. Ni substitution for Cu, Ca substitution for Y,
and the chain oxygen vacancies all have much weaker ef-
fects, even at defect levels of 1% or more [21]. The previ-
ous generation of YSZ-grown crystals showed only slight
curvature in (T ) and 1 (T ) below 4 Kelvin and the new 0
BaZrO3 -grown crystals show no hint of T 2 temperature
dependence down to 1.2 Kelvin. The relative rarity of T 2
behaviour (though it is common in lms for reasons that
remain unclear [54]) leads us to consider the opposite
limit for the strength of the scattering, the Born limit.
For impurity scattering in the Born limit, the crossover 0.4
energy scale is exponentially small, so one does not ex-
pect to see the universal conductivity limit until measure-
ments are performed well below 1 Kelvin. In fact, in the
microwave measurements presented here we are also not
necessarily at low enough or frequency to observe any 0.2
simple limiting behaviour. So, we compare our results
qualitatively to numerical calculations performed in the
Born limit by Hirschfeld et al. [46]. They found that at
very low frequency 1 (T ) rises rapidly from the universal
zero temperature limit to a much larger conductivity that
depends upon the impurity scattering rate. It remains 0.0
0 5 10 15 20
fairly temperature independent until inelastic scattering T(K)
processes become important. At higher frequencies 1 (T ) FIG. 9. This detailed view of the temperature dependence
becomes smaller and moves through a whole range of be- of the microwave conductivity below 20 K shows a gradual
haviours, varying from mostly sub-linear in T at low fre- evolution of the shape of of 1 (T ), from concave up at high fre-
quency, through a quasi-linear temperature dependence quencies to concave down at the lowest frequencies. The quite
at intermediate frequencies, to a faster than linear tem- linear temperature dependence seen here at 13 GHz seems to
perature dependence at high frequencies. Figure 9 shows be an intermediate behaviour.
behaviour in the measured microwave conductivity that
is similar to this Born limit result in some of its qualita- sub-linear as expected in the Born limit. However, one
tive features. The overall conductivity has a magnitude perhaps important di erence from the theoretical calcu-
that varies with purity, from the quite high values seen lations is that the trend does not to continue below 2
here, through to very low, at conductivities observed GHz. The data stops evolving towards the expected low
in Ni-doped samples [30]. More importantly, here we do frequency behaviour, which is a rapid leap upwards to a
clearly see for the rst time that the linear behaviour constant value. The reason for this is not yet clear, so
of 1 (T ) is an intermediate behaviour, albeit one that we are left for the moment with qualitative features that
survives over a substantial range of frequency and tem- seem only weakly in accord with Born limit scattering.
perature. The evolution in the shape of 1 (T ) at low Other features of the data echo the expectations of a con-
T clearly falls outside of our phenomenological model, ductivity spectrum with a Drude-like lineshape that has
which would have predicted a linear temperature depen- a temperature independent width; namely, the similarity
dence at all of the frequencies shown in Fig. 9. That in the 1 and 2 GHz curves and the relatively large fre-
is, if 1 (!) were really Drude-shaped with a temperature quency and temperature range over which 1 (T ) is nearly
independent width, then 1 (T ) would exhibit the same linear in T. It is the latter aspects of the data that lead
temperature dependence at all frequencies; namely, the to the Drude model being a fairly good description of the
linear temperature dependence of the normal uid den- conductivity spectra.
sity. Although the foregoing discussion indicates that the
At our lowest frequencies the data does move towards

10
1 (!) spectra cannot be perfectly Drude shaped, there is collapses rapidly and it becomes easily discernible in the
still a characteristic width to the peaks that is well pa- microwave spectral range at temperatures below 55 K.
rameterized by the Drude ts. Thus, the plot of scatter- In the range between 45 K and 20 K we nd that the
ing rates shown in Fig. 8 provides a reasonable measure collapse of the scattering rate varies at least as fast as
of the narrowness of the peak at low temperature and T 4. By 20 K the width becomes extremely narrow and
its rapid broadening above 20 K. Just above 20 K, the nearly temperature independent. This narrow width of
initial onset of this increase in scattering appears to be only 9 GHz corresponds to a mean free path as high as
at least as rapid as T 4 and must rise even more quickly 4 m if we interpret the width as being a direct mea-
at higher temperatures in order to meet the width ob- sure of the elastic scattering rate due to impurities. We
served in the normal state far infrared measurements. A nd that some features of 1 (!; T ) below 20 K are in ac-
rapid temperature dependence of the quasiparticle scat- cord with quasiparticle scattering in the Born limit for a
tering time would be expected in any situation where the dx2 y2 superconductor, in particular, a gradual evolution
inelastic scattering comes from interactions that become of the shape of 1 (T ) from sublinear T dependence at low
gapped below Tc . A number of early calculations tackled !, to quasilinear and then faster at higher !. However,
the problem of the collapse of the scattering rate below there remain discrepancies that might best be settled by
Tc in this way. Early on, Nuss et al. explained the peak a detailed numerical calculation aimed at tting the ob-
in their THz conductivity measurements in this manner servations presented here. Such ts must come to grips
[8]. Littlewood and Varma studied this type of e ect with the observation that the behaviour of 1 (!; T ) below
in a marginal Fermi liquid [55], the idea being that be- 20 K is reasonably well described by a model involving
low Tc a gap opens up in the scattering spectrum. Statt quasiparticle scattering with a temperature-independent
and Grin similarly studied the e ect of the opening of scattering rate.
a gap in the spectrum of spin uctuations [56]. All of
this work predated the solid identi cation of the dx2 y2
pairing state, so isotropic s-wave gaps were assumed in ACKNOWLEDGMENTS
the calculations. Quinlan et al. studied a model in
which quasiparticle lifetimes were associated with spin We are greatly indebted to A.J. Berlinsky and C.
uctuation scattering. They studied the e ects of both Kallin for many helpful discussions regarding transport
s wave and d wave gaps opening up in the spin uc- properties in d-wave superconductors. We wish also to
tuation spectrum [57]. Since we now know that the gap acknowledge helpful conversations with D.J. Scalapino,
in this material has dx2 y2 symmetry, this latter calcula- P.J. Hirschfeld, P. A. Lee and G. Sawatsky, and are grate-
tion is most directly relevant to our measurements here. ful for the opportunity to carry out some of this work at
In particular they found that at temperatures well below the Aspen Center for Physics. We also thank J. Trodahl
Tc, the quasiparticle lifetime increases as T 3 and even for his contribution to the design of the 75 GHz cavity
faster than this as Tc is approached. In a comparable setup. This research was supported by the Natural Sci-
temperature range, the transport scattering rate that we ence and Engineering Research Council of Canada and
extract from the width of 1 (!) is closer to T 4. Thus, the the Canadian Institute for Advanced Research. DAB ac-
temperature dependence of the inelastic scattering rate knowledges support from the Sloan Foundation.
seems to be about one power of T faster than the lifetime
calculations based on a gapping of the spin uctuation
spectrum. One must note, however, that the quasipar-
ticle lifetime can in principle di er from the electronic
transport scattering time, since charge transport is most
strongly a ected by backscattering.
 Permanent address Dept. of Physics and Astronomy, Mc-
Master University, 1080 Main St. W, Hamilton, ON, L8S
VI. CONCLUSIONS 4M1, Canada.
[1] W.N. Hardy, D.A. Bonn, D.C. Morgan, Ruixing Liang,
We have presented here the most complete microwave and Kuan Zhang, Phys. Rev. Lett. 70, 3999 (1993).
measurements yet obtained on a crystal of a high tem- [2] D.M. Broun, D.C. Morgan, R.J. Ormeno, S.F. Lee, A.W.
perature superconductor. These spectra reveal a wealth Tyler, A.P. MacKenzie, and J.R. Waldram, Phys. Rev.
of detail regarding the conductivity spectrum 1 (!) of B 56, 11443 (1997).
the thermally excited quaiparticles in the superconduct- [3] L.A. de Vaulchier, J.P. Vieren, Y. Guldner, N. Bontemps,
ing state. We nd that 1 (!) has a Drude-like shape. R. Combescot, Y. Lemaitre, and J.C. Mage, Europhys.
Lett. 33, 153 (1996).
Although in detail there may be deviations from this [4] Shih-Fu Lee et al., Phys. Rev. Lett. 77, 735 (1996).
shape, there is nevertheless a well de ned characteristic [5] T. Jacobs, S. Sridhar, Qiang Li, G.D. Gu, and N.
width which can be associated with the transport scat- Koshizuka, Phys. Rev. Lett. 75, 4516 (1995).
tering time of the quasiparticles. Below Tc the width

11
[6] S. Kamal, D.A. Bonn, Nigel Goldenfeld, and P.J. [32] D. Miller, P.L. Richards, S. Etemad, A. Inam, T.
Hirschfeld, Phys. Rev. Lett. 73, 1845, (1994). Venkatesan, B. Dutta, and X.D. Wu, Phys. Rev. B 47,
[7] D.A. Bonn, P. Dosanjh, R. Liang, and W.N. Hardy, Phys. 8076 (1993).
Rev. Lett. 68, 2390 (1992). [33] M.S. Pambianchi, S.N. Mao, and S.M. Anlage, Phys.
[8] Martin C. Nuss, P.M. Mankiewich, M.L. O'Malley, E.H. Rev. B 53, 4477 (1995).
Westerwick, and Peter B. Littlewood, Phys. Rev. Lett. [34] H.K. Olson and R.H. Koch, Phys. Rev. Lett. bf 68, 2406
66, 3305 (1991). (1991).
[9] R.C. Yu, M.B. Salamon, J.P. Lu, and W.C. Lee, Phys. [35] N.E. Glass and W.F. Hall, Phys. Rev B 44, 4495 (1991).
Rev. Lett. 69, 1431 (1992). [36] H. Srikanth, B.A. Willemsen, T. Jacobs, S. Sridhar, A.
[10] T. Shibauchi, A. Maeda, H. Kitano, T. Honda, and K. Erb, E. Walker, and R. Flukiger, Phys. Rev. B 55, 14733
Uchimokura, Physica C 203, 315 (1992). (1997).
[11] Jian Mao, D.H. Wu, J.L. Peng, R.L. Greene, and Steven [37] O. Klein, Phys. Rev. Lett. 72, 1390 (1994).
M. Anlage, Phys. Rev. B 51, 3316 (1995). [38] K. Holczer, O. Klein, and G. Gruner, Sol. St. Comm. 78,
[12] Kuan Zhang, D.A. Bonn, Ruixing Liang, D.J. Baar, 875 (1991).
W.N. Hardy, D. Basov and T. Timusk, Phys. Rev. Lett. [39] P.C. Hammell, M. Takigawa, R.H. He ner, Z. Fisk, and
73, 2484 (1994). K.C. Ott, Phys. Rev. Lett. 63, 1992 (1989).
[13] Steven M. Anlage, Dong-Ho Wu, J. Mao, X.X. Xi, T. [40] D.J. Scalapino, in High Temperature Superconductiv-
Venkatesan, J.L. Peng, and R.L. Greene, Phys. Rev. B ity, edited by K.S. Bedell, D. Co ey, D. Meltzer, D.
50, 523 (1994). Pines, and J.R. Schrie er (Addison-Wesley, Reading,
[14] T. Shibauchi, H. Kitano, K. Uchinokura, A. Maeda, T. MA, 1990).
Kimura, and K. Kishio, Phys. Rev. Lett. 72, 2263 (1994). [41] N. Bulut and D.J. Scalapino, Phys. Rev. Lett. 67, 706
[15] H. Kitano, T. Shibauchi, K. Uchinokura, A. Maeda, H. (1992).
Asaoka, and H. Takei, Phys. Rev. B 51, 1401 (1995). [42] P.A. Lee, Phys. Rev. Lett. 71, 1887 (1993).
[16] U. Dahne, Y. Goncharov, N. Klein, N. Tellmann, G. Ko- [43] J.R. Waldram, P. Theopistou, A. Porch, and H.-M.
zlov, and K. Urban, J. Supercond. 8, 129 (1995). Cheah, Phys. Rev. B 55, 3222 (1997).
[17] D.B. Romero, C.D. Porter, D.B. Tanner, L. Forro, D. [44] P.J. Hirschfeld, W.O. Puttika, and D.J. Scalapino, Phys.
Mandrus, L. Mihaly, G.L. Carr, and G.P. Williams, Phys. Rev. Lett. 71, 3705 (1993).
Rev. Lett. 68, 1590 (1992). [45] Peter J. Hirschfeld and Nigel Goldenfeld, Phys. Rev. B
[18] S. Spielman, Beth Parks, J. Orenstein, D.T. Nemeth, 48, 4219 (1993).
John Clarke, Paul Merchant, and D.J. Lew, Phys. Rev. [46] P.J. Hirschfeld, W.O. Puttika, and D.J. Scalapino, Phys.
Lett. 73 1538 (1994). Rev. B 50, 10250 (1994).
[19] A. Pimenov, A. Loidl, G. Jakob, and H. Adrian, preprint. [47] D. Xu, S.K. Yip, and J.A. Sauls, Phys. Rev. B 51, 16233
[20] K. Krishana, J.M. Harris, and N.P. Ong, Phys. Rev. Lett. (1995).
75, 3529 (1995). [48] S. Hensen, G. Muller, C.T. Rieck, and K. Scharnberg,
[21] W.N. Hardy et al., in Proc. of the 10 Anniversary HTS
th
Phys. Rev. B 56, 6237 (1997).
Workshop (World Sci., Singapore, 1996); D.A. Bonn et [49] D. Achkir, M. Poirier, D.A. Bonn, Ruixing Liang, and
al., in Proc. of the 21st Low Temp. Phys. Conf., Czech. W.N. Hardy, Phys. Rev. B 48, 13184 (1993).
J. Phys. 46, 3195 (1996). [50] Kuan Zhang, D.A. Bonn, Ruixing Liang, D.J. Baar, and
[22] A. Hosseini, Saeid Kamal, D.A. Bonn, Ruixing Liang, W.N. Hardy, Appl. Phys. Lett. 62, 3019 (1993).
and W.N. Hardy, Phys. Rev. Lett. 81, 1298 (1998). [51] P.J. Hirschfeld, P. Wol e, J.A. Sauls, D. Einzel, and
[23] J.L. Tallon, C. Bernhard, U. Binninger, A. Hofer, G.V.M. W.O. Puttika, Phys. Rev. B 40, 6695 (1989).
Williams, E.J. Ansaldo, J.I. Budnick, and Ch. Nieder- [52] R.A. Klemm, K. Scharnberg, D. Walker, and C.T. Rieck,
mayer, Phys. Rev. Lett. 74, 1008 (1995). Z. Phys. 72, 139 (1988).
[24] B. Gowe, P. Dosanjh, and D.A. Bonn, to be published. [53] Louis Taillefer et al., Phys. Rev. Lett. 79, 483 (1997).
[25] Ruixing Liang, D.A. Bonn, and W.N. Hardy, Physica C [54] D.A. Bonn and W.N. Hardy, in Physical Properties of
304, 105 (1998). High Temperature Superconductors, Vol. V, ed. by D.M.
[26] A. Erb, E. Walker, and R. Flukiger, Physica C 245, 9 Ginsberg (World Scienti c,Singapore, 1996).
(1996). [55] P.B. Littlewood and C.M. Varma, Phys. Rev. B 46, 405
[27] A. Erb, J.Y. Genoud, F. Marti, M. Daumling, E. Walker, (1992).
and R. Flukiger, J. Low Temp. Phys. 105, 1023 (1996). [56] B.W. Statt and A. Grin, Phys. Rev. B 46, 3199 (1992).
[28] S.Kamal, Ruixing Liang, A. Hosseini, D.A. Bonn, and [57] S.M. Quinlan, D.J. Scalapino, and N. Bulut, Phys. Rev.
W.N. Hardy, Phys. Rev. B 58, 8933 (1998). B 49, 1470 (1994).
[29] D.A. Bonn, Kuan Zhang, Ruixing Liang, D.J. Baar, D.C.
Morgan, P. Dosanjh, T.L. Duty, A. MacFarlane, G.D.
Morris, J.H. Brewer, W.N. Hardy, C. Kallin, and A.J.
Berlinsky, Phys. Rev. B 47, 11314 (1993).
[30] D.A. Bonn, S. Kamal, Kuan Zhang, Ruixing Liang, D.J.
Baar, and W.N. Hardy, Phys. Rev. B 50, 4051 (1994).
[31] Ioan Kosztin and Anthony J. Leggett, Phys. Rev. Lett.
79, 135 (1997).

12