1lI Y S I C A L R F. V I F. M' %'Of l.

ME $88, NUMBER ho r~ EC FM SI:a

Electrical Conductivity of a Superconductor*

W. E. MAsKERq S. MARWEI JA~t AND R. D. PARKs
Department of Physics and Astronomy, University of Rochester, Rochester, %em F'ork 746Z7
(Received 25 July 1969; revised manuscript received 15 September 1969)

A model is proposed which describes the dc electrical conductivity of a superconductor both above and
below the classical critical point. The approach is to include in the calculation of superfluid density and the
electrical conductivity, the effect of the fourth-order term of the Ginzburg-Landau theory, which represents
the interaction between superQuid excitations. The resulting expression for the conductivity simplifies to the
Aslamazov-Larkin result above T, and yields, for "two-dimensional" samples, an exponential dependence
of the superAuid conductivity on d T (= T.—T} below T,. Calculated results for the one- and three-dimen-
sional cases are also given. Experimental studies of the electrical conductivity of "two-dimensional" Al
films were made to test the model. Good agreement was obtained in the region below T„even when the
sample resistance was followed over five decades and sample mean free paths were varied over two decades.

I. INTRODUCTION reported the sizable broadening for the resistive transi-

1' "NYIL rather recently it was generally accepted tion in thin amorphous Bi films, and the interpreting
of these results by Ferrell and Schmidt" as arising from
that it would be virtually impossible to observe
thermodynamic fluctuations. The theory was presented
thermodynamic fluctuations or critical-point eRects
by Aslamazov and Larkin, '3 by use of a microscopic
in superconductors. This pessimism was generated in
approach. Later, Abrahams and %oo" and Schmid"
part by (correct) estimates of the extremely narrow tem-
obtained essentially the same result from the linearized
perature interval over which anomalous eRects in
time-dependent Ginzburg-Landau (GL) theory. Further
equilibrium properties, such as specific heat, should be
observable in bulk superconductors. ' The outlook was
discussion has been given by KadanoR and Laramore. "
The main result of these studies is the prediction that
brightened by Little's' discussion of the feasibility of
thermodynamic fluctuations of the order parameter
observing the eRect of thermodynamic fluctuations on
above T, lead to an excess conductivity 0.' in the normal
the behavior of superconducting microgeometries. The
state, given by a'=o„ra/e, where Tp=e'/16dh 0 d is
first experiments'4 which provided at least qualitative
the film thickness, O. „ is the normal conductivity, and
evidence for fluctuation eRects in superconductors were
c= (T T.)/T. . Thi— s result is valid only in the two-
reported shortly thereafter. These experiments and sub-
sequent related work ' focused on the critical current
dimensional regime Ld& $(T)), in the dirty limit
behavior of "one-dimensional" samples in the classical' (l«$0, where $0 is the BCS coherence length and l the
mean free path), and in the classical temperature region
region below T,. These experiments were particularly
where the mean-field theory is expected to be a valid
dificult because of sample preparation problems and
the smallness of the measured eRects. ' approximation. The last constraint requires that e(&T
and ~)&2ro, where 2~0 approximately defines the width
More recently, attention has shifted to the problem
above T, as a result of the paper by Glover" which
of the critical region. '3 The one- and three-dimensional
regimes, which are less accessible experimentally
*Work supported in part by the U. S. Air Force Ofhce of because of sample preparation difhculties in the first
Scientific Research and the U. S. Army Research OfIjLce (Durham). case and the smallness of the eRect in the second, have
t On leave from the Institute of Physics, University of Zagreb, also been treated. '~i5
Zagreb, Yugoslavia. Glover's results on amorphous Bi films" and subse-
' D. J. Thouless, Ann. Phys. (X. Y.) 10, 553 (1960}.
"
quent results by S trongin et al. on Pb and Al films and
~ W. A.
Little, Phys. Rev. 156, 396 (1967).
'R. D. Parks and R. P. GroB, Phys. Rev. Letters 18, 342
{1967).
"
by Smith et al. on Pb films are alI. in reasonable accord
'T. K. Hunt and J. E, Mercereau, Phys. Rev. Letters 18, 551 with the predictions of the Aslamazov-Larkin theory.
(1967). The purpose of the present work was to extend both
' R. P. Gro6, S. Marcelja, W. E. Masker, and R. D. Parks, the theory and experiments to include the entire transi-
Phys. Rev. Letters 19, 1328 (1967).
'W. W. Webb and R. J. Warburton, Phys. Rev. Letters 20,
461 (1968}. ~ R. A. Ferrell and H. Schmidt, Phys. Letters 25A, S44 (1967).
' J. S. Langer and V. Ambegaokar, Phys. Rev. 164, 498 (1967}. "L. G. Aslamazov and A, I. Larkin, Fiz. Tverd. Tela 10,
Samples for which two dimensions are smaller than the tem-
perature-dependent
—
1104 (1968) I English transl. : Soviet Phys. Solid State 10, 875
coherence length g(T}. (1968)g.
OFor a discussion of the critical and classical regions, see '4E. Abrahams and J. W. F. Woo, Phys. Letters 27A, 117
K. Maki, Progr. Theoret. Phys. (Kyoto) 40, 193 (1968); B. I. (1968).
Halperin and P. C. Hohenberg, Phys. Rev. 177, 952 (1969).
" "A. Schmid, Z. Physik 215, 210 (1968).
R. D. Parks, in Proceedings of the Conference on Ftuctuations "L. P. Kadano6 and G. Laramore, Phys. Rev. 175, 579
in Semiconductors, Asilomar, California, 1968', edited by W. S. (1968).
Goree and F. Chilton (Stanford Research Institute, Menlo Park, "M. Strongin, O. F. Kammerer, J. Crow, R. S. Thompson, and
Calif. , 1968), p. 141; W. W. Webb, ibid. , p. 159; T. K. Hunt, H. L. Fine, Phys. Rev. Letters 20, 922 (1968).
Phys. Rev. 177, 749 (1969). ' R. O. Smith, B. Serin, and E. Abrahams, Phys. Letters 28A„
" R. E. Glover, Phys. Letters 25A, 542 (1967). 224 (1968)~

7'4$
 3'IASKFR, MARPFI. JA, 3'XD PARKS
"
tion region. Extending the theory from the classical parameter field, ' nI, is given by
region above T, towards T, requires the inclusion of the
nq = (expI (a+bk'+b &1/1'&)/k~T] —1)—', (5)
interaction between superfluid excitations, which be-
comes important as the density of excitations increases. and the average value of the superfluid density follows
Such interactions are represented by the fourth-order in a self-consistent way,
term of the GL free-energy functional. The model pre-
sented in the following section consists of including this
term, in Hartree approximation, in the calculation of
electrical conductivity. Detailed calculations for the Using the same approximation in the time-dependent
two-dimensional case are presented in Sec. II, and the GL equation, one obtains
results for the one- and three-dimensional cases are
given in Appendix A. Preliminary results on the calcula- 4(~/~f)4=a4+4&141'&+b~V (7)
tion of specific heat within the spirit of the model are where y I =nb/8k~TP(0)] is calculated in Ref. 21.
presented in Appendix B. The remainder of the paper Assuming the usual relaxation form for the solution
reports new experimental results on Al 6lms and com- —)e '~ I, we obtain for the lifetime
fI, (x—
pares these results with the predictions of the model.

II. THEORY 8ksTP(0) a+bk'+b(1/1'&

For the reason of simplicity, with the we work o'= (e'/m)Pk
The superfluid conductivity n~r~ is then
Ginzburg-Landau formulation of the theory of super-
conductivity. In the transition region, this approach is me'hb
equivalent to the BCS formulation and leads to the
same results. The free-energy density for a given con- 8m@(0) ~ (a+bk'+b(I/I'&)'
figuration of the order paranieter is
together with Eq. (6) for (1/1'&. In the equation for
f(AT) = a141'+l b + I'+b I I ~+
I' a', we have used the linearized form of Eq. (5) in order
where the phenomenological
to simplify the algebra. For any real situation, this is
constants a, b, and 8 are
an extremely good approximation. Equation (9) reduces
defined in Sec. V A. If we introduce tb=Pqnqe'~' and
integrate over r, the free energy becomes (we consider
to the Aslamazov-Larkin result" in the classical region
above T,
unit volume)
The above results are easily reproduced within the
F(P, T) =Q (a+bk')n~ ng 2b Q rlk~o rig o nkng . (2) microscopic theory. However, a number of theoretical
k, k', Q objections can be raised against the proposed approxi-
mation scheme. To begin, the Hartree approximation
In the quantized version of the theory gkt, qk may be which we use in Eqs. (5) and (7) is a self-consistent
regarded as creation and annihilation operators for treatment of the first order in the perturbation theory
superfluid excitations. The fourth-order term in its and will no longer be sufhcient at high superfluid
original form makes impossible the exact calculation of densities. In three dimensions the approximation is not
the partition function and correspondingly all other good close to the new (renormalized) critical tempera-
thermodynamic quantities. For any general state ture. In one and two dimensions, the present approxi-
described by occupation numbers n& the expectation mation gives no exact phase transition, but nevertheless,
value of the fourth-order term is one expects the first-order theory to be a poor approxi-
. mation when the Cooper pair density becomes quite
gn, gn, , I
b Q
k, k', Q
q-g~o'q); o'rl), rlj, . ng, ng,
I ) large. This raises the more general question of the exis-
tence of an exact superconducting phase transition in
=b Pk ng Pnp
k'
. (3) one and two dimensions. At the moment, the theoretical
situation on this equation is open" and only a number
At this point, we invoke a Hartree-like approximation of plausibility arguments can be invoked for either of the
by replacing the sum over v~ by its average value, two possible answers. The last objection which we men-
tion is the use of Bose statistics to describe a non-
Qk' ng' Q ng n Q ng = (I& I
'& Q ng. (4) equilibrium system of Cooper pairs. As well as the
k k k
previous assumptions, this approximation will become
If we further assume the Bose character of the order- worse at higher pair densities.
~ For a discussion, see M. Revzen, Phys. Rev. 185, 337 (1969).
"Preliminary
reports of this work are given in {a) S. Marcelja, "A. Schmid, Physik Kondensierten Materie 5, 302 (1966).
Phys. Letters 2SA, 180 {1968);(b) S. Marcelja, K. E. Masker, and "See, e.g. , K. D. Grobman, Phys. Rev. 182, 297 (1969),
R. D. Parks, Phys. Rev. Letters 22, 124 (1969). and references therein.
 I. . I. I'C'I RICAL CON I) t.'C'I IV I I'Y Ol A SU VEI~CO th IDUCTOR

In the reniaining part. of this paper we shall conipare e.g. , Ref. 19(b)],
the presented model with the experimental results on
thin films. The good agreement which is obtained
2(= [b/(a+b(1(4 I(')) j"' (14)
supports our belief that the model has significantly which describes the range of the order in the system.
extended the calculations of Aslamazov and Larkin, " %ith decreasing temperature R becomes larger. To find
describing the electrical conductivity throughout the the temperature dependence of R below T„we express
superconducting phase transition. Eq. (10) as
In the present work, we are interested in the solutions e'«6 (""22
(&—
of the preceding equations applicable to thin films.
a+b((bead)')+bQ'=kd)T[ln(ebo""er )
Equations (6) and (9) become ln(1 e
—4«6((p(')/)62)F)
j (13)
Below T„(~ jb~') is large and approximately equal to
a/b —
In di. rty or moderately dirty metals bQ'/ksT»1
and one obtains

exp
a+5(k2+k, r)+b(~ij ~')
kgT
— 1, (10)
a+b(~ p 2) ~
k Te+4

The correction for film thickness is now similar to the

6o()b)dT6
(16)

Ze2A8
one introduced earlier"; the only difference is that the
range of the order comes in place of the coherence
8m@ (0)« length $(T). If we introduce the notation
(e(
(12k
(11) G(x) = P — =-1 1+—coth- (17)
[a+b( ~
lk ~
')+b(k'+k, ') j' n=o j.+n2x2 2 x x

where the film is assumed to be in the xy plane. Contrary Eq. (12) may be written
to previous work, ""
we have also introduced the cutoff
momentum Q in the integrations. The justification for I
e2 7' 1

this is that the GL theory contains only first-order

derivatives in the expression for free energy and is not
16bd r(0)( +b((0~')) ( ) d 1+20/2'(0)
valid for very high q values. For superconductors the 7r
XG— (18)
highest value of q follows from the microscopic deriva-
tion of the GL theory, and is of order 1/$(0).00 The («[(a+b(14 I'))/b+I/5'(0) j'"/—
introduction of the momentum cutoff Q= 1/$(0) in For convenience we did not stop the summation over
Eq. (11) does not affect the result close to T, (6&0.1), k, at the maximum value Q. This introduces negligible
but acts to reduce the superHuid conductivity far above error, as long as «((T),
since the first few terms in
the transition. The approach, however, is superQuous the sum are by far the most important; for d&(f. (T) the
for sufficiently large e (i.e. , e 1), where the GL expan- term in brackets may be replaced by unity. It can be
sion itself is no longer a good approximation. After readily seen that the correction for finite film thickness
integration, (11) becomes will have negligible e6ect on the result below T,. The
e2 reason for this is apparent: The contribution to o-'
comes only from low k values and only the first term in
0
.
16k« P(0) -. a+b(~ik~r)+6k, r the k. sum is important as the range of order becomes
much larger than the film thickness. In computing the
value of o' from Eqs. (10) and (16), we can also neglect
(12) the summation over k, in Eq. (10). The correction to
a+b ( 2)+b/p(0) +bk, r
t
bead(

( j~ i)2becomes important only for

~
T„when T)
The boundary condition is that the derivative of P a»b(~b(d ~'), and then it does not change the result for
dr' in Eq. (11) or Eq. (18).
should vanish at the surface. Both the real and imaginary
f
parts of can be expressed as Fourier series in cos(k r). Finally, we give the asymptotic form of the conduc-
tivity for T«'1„
This leads to the possible k, values
k, = rrrr/d, (13) e2 h' 47rbdc
0' exp (19)
where n is a positive integer (including zero). 16bd 2mb Tb'(0) bb T)
The correction for finite film thickness takes a par-
ticularly simple form if we introduce the length [see, This form is correct in the limit bQ'»1; for bQ'&(1 the
factor Ar/2mksTP(0) would be absent,
~ P. G. de Gennes, Superconductivity of Meats and A/loys
(W. 2(t. Benjamin, Inc. , New York, 1966), Chap. 7. 2~ H. &chml. dt, Z. Phys)Ik 216, 336 (1968),
 MASKFR, ih'IAR(. LLJA, AiV I) I ARK'
TAsz. E I. Sample parameters. The method used to determine effect on superconducting properties is the same whether
~, and T, is discussed in Sec. V A.
its origin lies in impurity scattering or tunneling resist-
Thickness Oa Tc
ance. We admit that this method is open to challenge
Sample (L)
fg
(0/sq) ('K) 10'ee when p F/, ff (1; yet, as will be shown, the results appear
to be consistent with this assumption even for
I 200 107 2. 155 0.381
II 160 203 2. 188 0.817 Ppl. ii 0.1.
III 120 350 2.325 1.32 In the present study, thin films were evaporated
IV 190 1070 2.174 4.04
V 190 1444 2. 123 4.86 from aluminum-wetted tungsten wire sources onto room-
VI 170 4454 1.817 22.9 temperature glass substrates at oxygen pressures of
VII
VIII
170
130
4741
5106
1.900
1.775
21.7
15.4 10 4 to 10 ' Torr. Evaporation rates were of the order
IX 170 6010 1.980 25.2 of 100 A/sec. The resistivity of the sample could be
controlled by the amount of Al evaporated, the oxygen
pressure, and the speed of evaporation. It was found
The analogs of Eqs. (6), (9), and (19) for the one- that for thin films (150 A or less) the resistance per
and three-dimensional cases are given in Appendix A. square could be increased appreciably by merely expos-
ing the sample to air for a period of days (or weeks). 's
III. SAMPLE PREPARATION Films prepared in this manner for the present study
ranged in resistance from 100 to 6000 0/sq.
The system chosen to experimentally challenge the
After deposition the film was cut into a rectangular
preceding model was that of thin, short mean-free-path
zigzag pattern with a sharp tungsten needle mounted
Al films. This system is particularly attractive because
on a precision micromanipulator. This provided a
of the ease and reproducibility with which films of
sample with a large length to width ratio and corre-
widely varying resistivities can be made. The films
spondingly a large, easily measurable resistance. In
were prepared according to the procedure of Abeles
general, the samples were constrained to areas smaller
et c/. ,2' wherein Al is evaporated in the presence of
than 1 cm square in order to reduce the possibility of
oxygen. Such films have been shown to be composed of
very small crystallites with dimensions & 100 A. " thickness variations due to evaporation source asym-
metry or other causes. (For some of the "cleaner"
It has been shown theoretically by Parmenter" and
samples, larger areas were required in order to provide
by Abeles, Cohen, and StowelP' that hypothetical
enough sample resistance for accurate measurement at
granular systems can be described by an eGective mean
low currents. ) An auxiliary benefit of the trimming
free path, even though the resistance is dominated by
the tunneling barrier resistance between grains (or operation was the presumed elimination of tapered
crystallites). The effective transport mean free path is edges, which always result from using evaporation
— masks to define geometries.
given by t, « dt/(1 t), w— here d i—
s the grain size and t
Film thicknesses which ranged 100—200 A were
the transmission coeScient between grains, and is
measured with a Tolansky interferometer to an accuracy
related to the resistivity, as in ordinary metallic con-
of &20 A. More accurate film thickness measurements
duction, by (in the free-electron theory)
were unnecessary since the film thickness does not enter
(pl, ii) '=-', E(0)ape', (2o) the theory except through a small corrective term Lthe
where p is the resistivity, 1V(0) is the density of states j,
quantity G in Eq. (16) this term being completely
unimportant in the region of interest T& T, .
at the Fermi surface, and ep is the Fermi velocity.
While this result is strictly valid, within the approxi-
IV. EXPERIMENTAL METHOD
mations of the model, only for pp/, «)&1, where p p is the
momentum at the Fermi surface, critical magnetic field The resistance of the Al samples was measured in the
measurements made on granular Al films" imply that temperature range 4.2—1.07'K by measuring the voltage
the effective mean-free-path representation is a good across the sample in the presence of a constant measur-
approximation even when pp/, ff 1. In particular, it ing current. In the range 10 pV —2 V, the voltage was
was shown empirically that $=()ol, «)"2 even when measured to within 0.1 pV with a Leeds and Northrup
p p/ f f~ 1, a relation that is known to be valid when the K-5 potentiometer. Lower voltages (0.05—10 yV) were
mean free path is determined by impurity scattering. measured with a Keithley 1508 microvoltmeter. A
Therefore, when discussing the experimental results, heavy-duty 12-V automobile battery provided a mea-
we shall assume that all of the samples studied can be suring current which was constant to within one part in
described by an eA'ective mean free path given by 105. The measuring current densities ranged2 —20A/cm2.
p /, f f —const, and that the role of /, ff in terms of its It was found that within this range (typically 0.5—5 tiA,
~' B. Abeles, R. W. Cohen, and G. W. Cullin, Phys. Rev. Letters
depending on the sample and temperature) the resist-
17, 632 (1966). ~ This method of achieving higher resistances has also been
' R. H. Parmenter, Phys. Rev. 154, 353 (1967). used by Strongin and co-workers, e.g., Ref. 17, and M. Strongin,
~'B. Abeles, R. %. Cohen, and R. W. Stowell, Phys. Rev. O. F. Kammerer, and A. Paskin, Phys. Rev. Letters 14, 949
Letters 18, 902 {1967). (1965).
188 EI. ECTRlCAI. CON DUC 1'IVITY OF A SUPF RCONDUC'1 OR

ance of the sample was ohmic and the result inde-

pendent of measuring current; the measuring currents
were, nonetheless, large enough to avoid the possible
influence from noise currents generated by room-
temperature circuitry. All measurements were made
inside a copper-screened room to eliminate possible
noise problems from external sources.
The temperature of the samples, which were im-
mersed in liquid helium, was determined from a carbon K

resistance thermometer placed in close proximity to the Q -8t-

R,"= 4454Q/o
sample. The carbon thermometer was calibrated in
each run against the vapor pressure of the helium bath. -IO—
The temperature was controlled with a planar dia-
phragm manostat which was inserted in the pumping -l2—
line to the Dewar. Regulated in this manner, the tem-
perature was stable to approximately five parts in 10' -l4-
in the normal helium range and one part in 10' in the
L4 6
1. l. 8 2.0 2.2 2.4 2.6 2.8 30
superfluid range. T ('K)

Since, as will be shown below, small magnetic fields FIG. 1. Natural logarithm of resistance versus temperature for
were found to influence the sample behavior, it was sample VI. Solid line: plot of Eq. (18) (which simplihes to Eq.
necessary to provide magnetic shielding. The samples (19) for T«T, j, with Rp" = 4454 0/sq and T, = 1.817'K.
were mounted on a Formica insert and kept at least
10 in. away from the nearest metal support. The glass A check on this procedure was obtained by comparing
Dewar itself was placed inside a 4)&10-in. cylindrical the resistor with a germanium thermometer in the
magnetic shield which reduced the earth's field to less range 4—10'K.
than 10 ' Oe. A copper solenoid magnet shrouded the
Dewar inside the shield and could be used to provide
7. EXPERIMENTAL RESULTS
AND DISCUSSION
small fields longitudinal to the Dewar's axis. The
magnetic field inside the Dewar was measured before Nine Al samples, prepared in the manner discussed
each run with a Hewlett Packard model 4283 milliam- in Sec. III and having the parameters listed in
meter equipped with a Model 3529A magnetometer Table I, were used in the study. As seen from the Table
probe. The field longitudinal to the Dewar axis could the T,'s of the films varied and were always larger than
be reduced to less than 2X10 4 Oe by applying small the T, of bulk Al. This is not a new result; the enhance-
currents through the magnet. ment of T, in such granular (or amorphous) films has
The samples were mounted with the sample surface been discussed extensively. "
This provides no problem
normal to the Dewar axis. Perpendicular magnetic fields in the present study, if we assume, as we shall below,
could be applied with the solenoid in order to study the that the BCS law of corresponding states holds for the
effect of field on the measured resistivity. Because of the hlms, which allows one to scale the relevant parameters
small sample area, field uniformity was not a problem. according to T..
The applied fields were kept deliberately small (less
than 10 Oe) to avoid inducing a magnetic moment in A. T & T,
the shield. In order to test the model presented in Sec. II, in
In order to fix the value of normal resistance and to particular, Eqs. (10) and (11), the full resistive transi-
examine the "semiconducting" behavior exhibited by tions of the samples were measured. The results for two
the samples above T, (see Sec. V C), resistance measure- extremely short mean-free-path samples are shown in
ments were taken in the range 4.2 —20 K for some of the Fig. 1 and Fig. 2, and those for a relatively clean sample
samples. In these instances the samples were run inside in Fig. 3. The log plots tend to deemphasize the
a copper vacuum can immersed in liquid helium. An Aslamazov-Larkin (AL) region (T)&T,) in favor of the
Advance wire heater was used to raise the sample tem- region T& T„which is the region. of particular interest
perature to 20'K. The temperature measurement in in the present study.
this range was provided by a carbon resistance thermom- The solid curves are plots of Eq. (11) Lsolved to-
eter which was calibrated in the liquid-helium range gether with Eq. (10)$. The thickness correction is
against the vapor pressure of helium and at approxi-
~ See, e.g. , Refs. 25, 26, 28; R. Hilsch, Eon-Crystalline Solids,
mately 20'K against the boiling point of liquid hydro- edited by V. D. Freschette (John %'iley & Sons, Inc. , New York,
gen. Temperatures between 4 and 20'K were deter- 1958); M. Strongin, O. F. Kammerer, J. E. Crow, R. D. Parks,
mined using a standard-resistance — versus —temperature D. H. Douglass, Jr. , and M. A. Jensen, Phys. Rev. Letters 21,
1320 (1968); J. W. Garland, K. H. Bennemann, and F. M,
relation known to hold for the type of resistor used. Mueller, Phys. Rev. Letters 21, 1315 (1968),
 IX'I ASKLI~, AIARCLl. Jr%, AIM I) I'Ak V, S 1SS

below T, . The values of e, for the nine samples studied,

which were extracted from theoretical fits to the data
such as those shown in Figs. 1—3, are listed in Table I
-2— and plotted as a function of R~" in Fig. 4. The fact
that the straight-line fit to the data has a slope of
-4—
unity confirms the predicted linear dependence of e,
on R~". The equation of the line corresponds to
-6—
=0.40/ 10 ' R&". The parameter &, can be expressed
-8— using the microscopic values for the constants in the
D
M= IQ2 GL free energy expansion. 's Introducing b = ttt'/2m,
-IO
I a =b(T —T.)/T. e(0),
$(0) =0.8&(tel tt)'
0 Oe
and b=1 02A.: /lkkrl, tt'm,
I
12
where llr is
the electron density, gives
-14- CI
k

OIO
I
I4

0
mvk (1.04X10 ')
(23)
2
1. 1. 4 6
1. 8
1. 2.0 22 2.4 26 2.8 3.0 32
Xe'l, fg($
T ('K)

I'I(:. 2. Resistive transition for sample VIII measured in Finally, using LEq. (20)],
"zero" magnetic field & 2 /10 4 Oe] and in applied perpendicular
I
fields. Solid line: plot of Eq. (18) with Rz"=5106 f)/sq and
T, = 0. 1775'K. (24)

negligibly small for T& T, and Eq. (17) is adequate to

one obtains ~, =1.04X10 'R~", which is about 2.6
treat the classical region, corresPonding to In(Rtj/Rck") times the value determined from experiment. In view
&— 3 (Rck is the resistance per square and n refers to of this discrepancy, it should be noted that the ex-
the normal state), which spans a sizable fraction of the
pression for e. below T, Eq. (23)] contains a number of
measured curves. Equation (19) can be expressed in the l

material constants, in contrast to the expression above

more useful form
T. (o.'=e'/16hd). We have eliminated the constants by
use of Eq. (24), which might not be a good approxima-
(T' == e'-'
—— exp
cTc
(21) tion for the amorphous Al films. A similar difficulty was
lkkk2 k, rk'(tl), T) encountered in the study by Abeles et al. ,27 in which the
value of pl, ff determined from upper critical magnetic
where e„= ebkkkT, , /4trbkla
field measurements on granular Al fi]ms was at least a
Two adjustable parameters were used in fitting the
theoretical curves in Figs. 1—3 (and the corresponding
curves for the other samples). These were T, and e., .
0-
which were chosen to give the best agreement in the
exponential region. Changing T. has the effect of trans-
lating the entire curve horizontally, while changing
e, has the effect of changing the slope of the curve in -4—
the exponential region. The T, s determined in this
manner were in reasonable agreement with those -6— H=2. 0 Oe AMPLE II
CD j I
determined by fitting the data to the theory in the
~
~
If ~

D
v ~~
~
~
~~
Tc = 2. 188 'K
Aslamazov-Larkin region above T, (see Sec. V C).
tt:
-8 t--. I ~I Ro" = 202. 6' /o

(
L~
~
~
~
~ ~

The crucial test of the theory in the region T T,

below T, reduces to that of demonstrating that the 10 I
.'037 Oe
absolute magnitude of ~, and the functional dependence
-12-
of e, on R~" are consistent with the theoretical pre- &2 x 10 ape

dictions LEq. (21)]. For a'))o we have o'=1/p, so

-14
that we may write k k

213 215 217 219 221 223 225 227

T( K)
d on (Ro/Rck") ] T.
(22)
FIG. 3. Resistive transition for sample II measured in zero tkeld
and in applied perpendicular fields. Solid line: plot of Eq. (18)
dT ecT with Rg"= 202. 6 0/sq and T, = 2. 188'K.

which gives the slope of the resistance tail in the region

well below T.. The measurement of this sloPe Provides
"See, e.g. ,A. I.. Fetter and P. C. Hohenberg, in Superconduc-
edited by
/ivi~y, R. D. parks (Marcel Degker, Inc. , New borg,
@ t.-onvenient test of the theory in the classical region 1969),~p. 872,
188 F I F CTRICAI. COX DUCTIX" I TY OV A SU PF. RCO~'DUCTOR

factor of 2 greater than the value determined from 10 000

Eq. (24)" 0 /gI

In summarizing the above results, we note that the
experimental results below T, are in excellent agree- 3000—
rnent with the model in terms of confirming the pre-
dicted temperature dependence of 0-' and the functional
dependence of 0' on sample parameters. There remains l000—

the discrepancy in the absolute magnitude of e, (see CO

IL
Xote added in proof)
As seen in Figs. 1 and 3, the presence of a sma, ll
perpendicular magnetic field leads to a severe distor-
tion of the resistive tail. It was determined that the
IOO-
magnetoresistance is ohmic and, at temperatures for 0003
I

.Ol 03 .I
, OOOI OOI
which R~)R (i.e. , where the magnetoresistive tail is
well separated from the zero-field tail in Fig. 1 or FH:. 4. Circles: Rg" versus the parameter e, for the nine samples
3), linear with field. These properties of the magneto- whose parameters are listed in Table I; the values of ~, are ex-
resistance suggest that it results from current-induced tracted from the best theoretical 6ts to the R-versus-T data
"
vortex low. In addition, it was found that for currents
—
t e.g. , see Figs. 1 3]. Solid line: best straight-line fit to the data.
much larger than those used in the experiment a
V~12 component appeared in the voltage (in the problems, in the study of the sampl. es reported in
absence of field), which is expected in the flux flow state, Table I, they represented serious obstacles in the study
when it is the field associated with the current, which
of cleaner samples (e.g. , Ro" &10 fi/sq), because of the
extremely rapid drop oR of resistance in such samples.
produces the vortex state. The fact that flux flow is
observed for perpendicular magnetic fields as small as The resistive transitions in such samples were com-
1 Oe and current densities as small as 10 At'cm' in such pletely dominated by these deleterious eRects.
samples is not surprising. The field at which vortices
enter (H, i) is expected to be vanishingly small in such
B. T T,
thin, short mean-free-path samples"; the small depin- %hile the agreement between the model and experi-
ning currents result presumably from the fact that rnent is remarkably good in the region below T„ it is
the scale of the inhomogeneities which act as pinning clear from Figs. 1—3 that the agreement is less good in
sites is small compared to the size of the vortex core" the immediate vicinity of T,. The mismatch between
&(T), the size of the region over which the order experiment and theory in the critical region seems to
parameter is depressedj. This is distinctly different diRer somewhat from sample to sample even though
than the situation in clean films or bulk materials, where the sample parameters are quite similar (e.g. , Figs. 1
the crystallite size is usually much larger than &(T). and 2), thus suggesting that at least. part of the problem
A more complete report on Aux-flow phenomena in. such might lie in the experiment.
granular systems will be published elsewhere. '4 These
observations are relevant to the present study only in C. T & T,
that they document the need for canceling the earth' s (Aslamazov-Larkin Regime)
magnetic field in the experiments.
The experimental investigation of the Al. regime
Sample inhomogeneity can also lead to anomalous
tail broadening, which appears, in plots like those of poses special problems not encountered in. the T&T,
studies. The most serious of these is the extreme sensi-
Figs. 1 or 3, very much like the field-induced anomalies
tivity of the excess conductivity on the normal resist-
(e.g. , Fig. 1 of Ref. 19b), except that it is relatively
ance. A complicating aspect encountered in granular
insensitive to magnetic field and does not vanish in the
Al films (or amorphous films prepared at cryogenic
absence of field. While magnetoresistance (for
H & 2&(10 4 Oe) and sample inhomogeneity posed no temperatures) is the semiconducting behavior exhibited
by such samples. For example, all of the samples used
"Note that there is an error in the value of /, ff used in Ref. 27, in the present study (Table I) exhibited small negative
which reflects the same error in earlier sources (e.g. , Ref. 13 of coeScients of resistance above T, . This, which is not a
Ref. 27). The value of pl, ff for Al, based upon the measured values
of the electronic specific-heat constant t N. E. Phillips, Phys. Rev. new result, "
arises presumably from the temperature-
114, 676 (1959)j, and the Fermi velocity from anomalous skin
effect measurements |
E. Fawcett, in The Eer~ni Surface, edited by
W. A. Harrison and M. B. Webb (John Wiley 8z Sons. , Inc. ,
"
dependent electron tunneling between Al grains sepa-
rated by oxide. This behavior is shown in Fig. 5, where
New York, 1961), p. 197j, is pl, @=0.42X10 "
0 cm2, approxi- the temperature dependence up to 20'K is shown for
mately four times smaller than the value used in Ref. 27.
"For a review of flux flow in superconductors, see Y. B. Kim "See, e.g. , O. F. Kammerer, D. H. Douglass, and M. Strongin,
and M. J. Stephen, in Superconductivity, edited by R. D. Parks Bull. Am. Phys. Soc. 12, 417 (1967).
(Marcel Dekker, Inc. , New York, 1969), p. 1107. g6
See, e.g. , Neubauer and R. H. Wilson, in Basic Problefns in
~ See, e.g. , Ref. 30, p. 845. Thin Flm Physics, edited by R. Neidermayer and H. Mayer
"P. Horn, K. E. Masker, and R. D. Parks {to be published). (Vandenhoeck R Ruprecht, Gottingen, 1966), p. 579.
'752 M ASKER, M ARCEI. J A, AN D PARKS 188
l200 TABLE II. Illustration of semiconducting behavior above T, for
three representative samples.

o"(10') —Eo"(20')
IOOO—

E.o"(T,)
Sample (0//sq) &o"(10 )
800—
II 203 0.000452
CD IV 1070 0.0109
VIII 5106 0.0355
Tc = 2. 174 K

R~ - l07'OQ&a

VI. SUMMARY
The theory of conductivity in superconductors has
been extended from the classical region above T,
6 8
(AL region) to include the complete transition region.
iO

T~)
FIG. 5. Plot of resistance versus temperature for a representative
The approach was to calculate the superfluid density
sample, which reveals semiconducting behavior above T, . Solid and excitation lifetime within the GL formalism, in-
line: best 6t to the data. cluding the fourth-order term, which represents an
interaction between superfluid excitations. The result
sample IV, which has an intermediate value of Eo". for conductivity simplifies to the AL result for T)&T,
The dependence of this semiconductivity on the mean and gives strikingly diferent results for diferent dimen-
free path is illustrated in Table II, where the diff'erence sionalities in the classical region below T, : for one-
between Ro"(20') and Ro" (10') is shown for three dimension, 0.'~ — e', for two-dimensions, 0-'~e '~'",
samples which span a large range of Eo". Despite the where e=(T T, )/T, — and e, is a sample-dependent
semiconducting behavior, it is possible to choose a parameter; and for infinite three-dimensional samples,
value of Ro" which is correct to within 1 or 2%%uo. a = ~ for T & T,* (thermodynamic
'
fluctuations are
This uncertainty is inconsequential in the theory for manifested only as a small shift in T, to a lower value
T&T. or in plots such as those in Figs. 1—3 which T,*).
deemphasize the AL region, but is of crucial importance Experimental results on the electrical conductivity
when determining a' for T»T, .
Taking the lead of other investigators we "' ' of short, mean-free-path two-dimensional Al films which
span a large range of resistivities are in reasonable agree-
could force the data into the Curie-Weiss fit predicted ment with the model in the region below T,. Experi-
by AL, i.e., mental results on extremely short mean-free-path Al
a'=a a„= r—oT, /(T T, ), —(25) films in the region above T, are in reasonable agreement
with the AL (or present) theory; however, the excess
by treating both T. and a„as arbitrary (constant) conductivity exhibited by moderately clean Al films
parameters, and then determine if Tp 1' =Tp .' Doing
this we obtain, as did Strongin and co-workers, reason- " is larger than expected. Further work on long mean-
free-path films is in progress.
able agreement with the AL theory for Al films with iVote added in proof We have .recently become aware
R&"&3000 0/sq. For cleaner Al samples the excess of the size-effect work of I. Holwech and J. Jeppesen
conductivity, determined in this way, is larger than that /Phil. Mag. 15, 217 (1967)$ which leads to the value
predicted by AL, the discrepancy growing with increas- "
pl=0. 82X10 0 cm' for Al at 4.2'K, which is approxi-
ing mean free path (e.g. , for Ro" —100 0/sq, ro'*~' mately twice the value determined from a free-electron
4ro"L). These results would seem to indicate (1) calculation. If the parameters a and b in Eq. (21) are
the breakdown of the AL theory for clean Al, or (2) evaluated in terms of the thermodynamic critical field
sample inhomogeneity. The fact that our results on and the (l-dependent) coherence length of bulk Al,
clean Al films have been corroborated in a second instead of free-electron parameters, and if l is elimi-
laboratory, ' and that the behavior of the cleaner nated in favor of p (or Ro") by using the Holwech and
samples used in the above study agrees with theory in Jeppesen result above, this leads to e, =0.43&(10 ' Ro"
the classical region below T„but not in the AL region, instead of ~, =1.04X10 ' Eo" and the factor of 2.6
tends to support the former hypothesis. Since the discrepancy between experiment and theory discussed
T&)T, study requires further experimental work be- in Sec. V A is reduced to 1.1. This, rather than closing
cause of the special problems associated with the AL the question of a possible discrepancy, elucidates the
region, as well as being not directly relevant to the main fact that the value of e, is sensitive to ma, terial
theme of this paper, a full report on it will be published constants.
separately.
"
It is possible to test the temperature dependence predicted ACKNOWLEDGMENTS
by Aslamozov-Larkin only if T, or Eo", or both, can be accurately
measured. It is a pleasure to thank Dr. R. Thompson for crit-
'8 J. E. Crow and M. Strongin (private communication). ically reducing the manuscript. We are grateful also
 I'I. EC'l'R I CAL CONDUCTOR% IT Y OF A SU PERCO X Dtr CTOR 733

to Professor V. Ambegaokar, Professor R. Ferrell, represent improvement compared to linearized theory:

Professor J. Langer, Professor K. Maki, and Professor
A. Schmid for helpful discussions. k r a+5(IWI'/)"'
Q-
2x'5
APPENDIX A: ELECTRICAL CONDUCTIVITY IN
bQ2 1/2-
ONE AND THREE DIMENSIONS
stan ' (A4)
In this Appendix we list the results for superfIuid o+b &141')
density and superQuid conductivity in one and three
dimensions. "
In one dimension the solutions corre- This equation has a solution with a+b& I&I'& different
sponding to Eqs. (6), (9), and (17) are straightforward from zero only if T is larger than the new critical tern-
and read as follows: perature T,*. When T= T,*, a+b&I PI'& =0; this is
the point of Bose-Einstein condensation and below this
kpT point the superfluid density has the same form as in the
mean-field theory without fluctuations. The shift in
~1:b(&+b &141')))'" the critical point is given by
I /2
bQ2 k/r Tb& (0)
)&tan ' (Ai) AT, = T, —T,*=T,
+b&l~l'& 47r'8

me'PI' The conductivity is infinite below the new critical

(A2) point and above the critical point is given by
16AA p(0) (/s+b(lip I'))"'
I
or, below T„ (A6)
328(0) L2~(~+b&141'))3"'
tre'bsl al' 1.71X10' eL '
Experimentally, this situation is indistinguishable from
0
2AAP(0)b'k Tess AP(0)
(A3)
the result of the linearized theory. ""
In the case when one, two, or three dimensions are
where A is the cross-sectional area of the sample, L is finite, we must retain the summation over k. In this
the length, and R" is the normal resistance in ohms. case the superQuid conductivity below the transition
The results are insensitive to the value of the momentum is no longer perfect. For example,i'll if the sample is a
cutofI' and the form of the theory for high q values, and cube of side L, we have
are expected to be a very good approximation. The
resistanc e-producing mechanism described here is
entirely different from and independent of the one (A7)
proposed for one-dimensional systems by I.anger and
Ambegaokar, ' which leads to an exponential rather ~e'h, l
than cubic dependence of fT' on ~.~ (AS)
The situation is more complicated in three dimensions. S~ts(0)Ls a L~+bk'+b&I41'&jr
First we have to distinguish between two qualitatively —
different situations: an infinite 3-D system and a and for e((e, (retaining only the first term in the
3-D system of finite size. In the latter case, one may not summation),
replace the summation over k by integration. For the kgT
(A9)
infinite 3-D case the first-order correction is insufFicient Ls(a+b s&)
close to T, . Nevertheless, we give the results since they
& I

ol
(I+I') = a/b k//T/L'r/, — —
(A10)
'9 Throughout this paper the dimensionality of the quantities alld
( ~f ls) and b changes according to the number oi dimensions of the
system. The 3-D numerical value for b must be divided by the
sample thickness for the two-dimensional case or the sample cross (A11)
section in the case of one dimension. S e(0)L'(+b(lal'))-'
We believe that for temperatures close to T, where the long- 01
range order is incomplete, the present model is applicable. It is ~e'h5 a'L'
possible that at sufliciently low temperatures long-range order (A12)
prevails and the mechanism proposed by I.anger and Ambegaokar
is the dominant loss mechanism in one dimension. However, SmP(0) b'k 'Ts
whether exact long-range order appears at any temperature in
one- or two-dimensional systems is an open question, and one that In the dirty limit the last equation reduces to
is intimately related to the discussion of the validity of the
approximations used in the present model, given in Sec. II. a' = 0.7 2 (m'Lsk//Tt///ks) o „. (A13)
As expected , the superQuid conductivit i
1 (1 h h pp
en current experiments4').

APPENDIX B: SPECIFIC HEAT AT THE

SUPERCONDUCTING TRANSITION
The calculation of specific heat at the su

b i
h as b een the sub ject

h eoretic expansions
p
oin . C orrespondingl
critical point
ect of
tions. '4'4' All of these calculations a

the r
',
o several investiga-

divergent at the
~
C3
U

UJ
CL
EA

th w h ere t h e correction to the s

llc p d t o th e d'iscontinuit y which w is the result
of mean-field theories
a roxim
The self-consistent approximation from Sec. II can
ie to t is problem. Thee free ree energy
e of a super-
Quid excitation -2 -I 0
b(k) = a+5k'+bn (Bfh l05 (T-Tc~/ Tc

where —(~tP~')
n= describes the free ener r Frc. 6. xpected specific-heat an

od oh o 1 fi
ergy o condensation for D hdl . 1 '
1 h
e, i is necessary to inte g ratee fro
rom zero super-
Quid density t o i't s fi nal1 value for the corn
possi le approximations 1
h fll
(a+bn)(1n+5 Q ¹ng. (B2)
F =+ (La+-,'bn(T)+8k')nk(T) }. (B4)
0 k

en tth e partition function

The central result is then
ince the interaction of the excitations
1
.+~~+-z., ).. . (s3)
ree energy per excitation and the
e
n 'ons. nce t e approximation for the
en unable to calculate ~83'
%e have been a ion of the
order to obtain thee comp/etc free ener
er o approximations ma y b
p or ar, b
specific heat is straightforwar bu tso e ateg
imensions, the fi h t h
rea s own close to T,. a a t e new (shifted tran
' '
cases are more amenable e t o approximation H ever, as notedd above th'
schemes, since in these i era y. In the case of two
*

ou t e transition region. One of the

1 11o o ( )
o ence, the specific-heat eak
41 See ab h i I va 1ue at T,) is round
1 sica

"'Tea""'" ""'""'
1
2A. P. Levanyik, I"u. d.
d (963)
1 gls
5, 1776 th ob serveddiin near-monolayer
films. "
thi L ff b een o helium
AIL k dsAPk Fi
rim i Teor ~~ 3
'
y o M y and Technical Repo t Xo. 44 The details of t e calculation will bee descnbed elsewhere.
se, p sica 15, 860 {1949).