VOLUME 10' NUMBER 3 PHYSICAL REVIEW LETTERS 1 PEsRUwmv 1963

a point-by-point measurement would be. In conclusion, a caution should be inserted.

The method proposed here can also be used to The derivation given here has assumed an inde-
determine the shape of the Fermi surface in semi- pendent-particle model for the electrons in the
metals and semiconductors. In fact, measure- metal. If many-body effects are important they
ments of this effect in bismuth have already been may invalidate the results given here.
made. ' In these materials it is not always true The author is happy to acknowledge very in-
that w «&~, and the equations given here have to formative correspondence with Professor T. Kjel
be modified accordingly. In particular, 0& in daas, Professor A. B. Pippard, and Professor
Eqs. (1'), (5'), and (7) has to be replaced by R. Chambers.
k~(I - &u/~c) '. In these materials, because w
=~&, the same equipment that measures the R. Bowers, C. Legendy, and F. Rose, Phys. Rev.
properties of the helicons (&u and k&) can also Letters 7, 339 (1961).
measure w and thus determine the same geo- 2F. Rose, N. Taylor, and R. Bowers, Phys. Rev.
metric properties as for metals. 127, 1122 (1962).
Jones and R. Chambers (to be published). The
It has been pointed out to the author after this
author is indebted to Dr. Chambers for a copy of their
paper was written that the analysis presented paper before publication.
here for helicons has already been done for the ~P. Cotti, P. %yden, A. Quattropani, Phys. Letters
case of circularly polarized transverse sound 1, 50 (1962).
waves. "~" The general ideas in the analysis 5P. Aigrain, Proceedings of the International Con-
for both helicons and circularly polarized sound ference on Semiconductor Physics, Prague, 1960
waves propagating along a magnetic field are the (Czechoslovakian Academy of Sciences, Prague, 1961),
same, but the details are somewhat different. p. 225.
6R. G. Chambers, Phil. Nag. 1, 459 (1956).
However, helicons appear to be a much more P. B. Miller and K. K. Haering, Phys. Rev. 128,
useful means than circularly polarized sound 126 (1962).
waves to study this onset of the Doppler-shifted J. E. Drummond, Phys. Rev. 112, 1460 (1958).
absorption. It is not possible to produce circu- ~A. B. Pippard, Reports on Progress in Physics
larly polarized transverse sound waves except (The Physical Society, London, 1960), Vol. 23, p. 176.
along a few high-symmetry directions. "~' Hel- The techniques used in deriving the equations of this
icons do not have this limitation and can propa- paper follow very closely the methods described in
this reference.
gate in all directions. It also appears that the OJ. Kirsch and A. G. Redfield, Bull. Am. Phys. Soc.
distinction between the onset of absorption at a 7, 477 (1962).
point or at a finite-sized orbit is more striking "T. Kjeldaas, Phys. Rev. 113, 1473 (1959).
for helicons. '2Reference 9, pp. 252-255.

OBSERVATION OF PERSISTENT CURRENT IN A SUPERCONDUCTING SOLENOID

J. File and R. G. Mills
Plasma Physics Laboratory, Princeton University, Princeton, New Jersey
(Received 21 September 1962; revised manuscript received 26 December 1962)

The classical experiment of Onnes, ' in which a netic resonance (NMR) techniques to the measure-
persistent current is induced in a closed super- ment of the field. A double layered solenoid of
conducting circuit and the trapped magnetic field 984 turns of 0. 020-in. diameter Nb-25% Zr alloy
observed over a period of time, has been repeated approximately 4 in. in diameter and 10 in. long
several times ' ' From the length of the period
~ as shown in Fig. 1 was constructed to provide a
of observation and the accuracy of the measure- homogeneous field. The measured axial field pro-
ment, one can set a lower limit to the time con- file of the central 0. 5 in. is shown in Fig. 2. The
stant of the circuit. Apparently the highest value terminals of the coil are permanently connected
to this limit was set in the experiment of Collins' by spot welding.
at a value of approximately 250 years. After inducing a persistent current in the coil,
To extend this limit by several orders of rnagni- its magnetic field was measured by NMR tech-
tude, we undertook to apply modern nuclear mag- niques and recorded with time. The first run ex-
VOLUME 10, NUMBER 3 PHYSICAL RE VIE%' LETTERS 1 FEBRUARY 1963

8—
2I04.

I I

PROBE SIZE

2I04.4

2104.2
—
0.3 —
0.2 -01 0 i O. I t0, 3
QISTANCE FROM COIL CENTERLINE - INCHES

FIG. 2. Measured axial field profile.

the data but eliminates the possibility of a position-

FIG. 1. Construction of solenoid. al shift indicating a false decrease.
Run two extended over. a period of 37 days, be-
ginning 9 days after the end of run one. The data
tended over a period of 21 days, and was termi- are shown in Fig. 3. The solid lines are least-
nated at that time because of two coincidental ac- squares fits of the data to the leading terms of an
cidents —one a failure in the electronic system, exponential decay,
and the other a mechanical shock to the system
which probably disturbed the position of some of a =B,(l- t/~), (i)
the turns and the position of the probe. The super- where B = observed field strength, Bo = field
current, however, was maintained, and the experi- strength at time zero, and r = time constant of the
ment continued. The observed data of run one circuit, giving the results shown in Table I. The
seemed to indicate a real field decay. Since the result of run two is consistent with that of run one.
probe had been displaced from the field maximum, An attempt has been made to correlate the data
it was not possible to determine the contribution with the work of Kim, Hempstead, and Strnad
to apparent decrease in field by this mechanical based on the flux-creep theory of Anderson. In '
shift. their work, field decays were observed of the form
A slightly inferior set of electronics was sub-
stituted and the measuring technique changed to B =8 - cln(t/t ),
C C
reposition the probe to field maximum before each
measurement. This introduces more scatter in where B is the observed field as a function of time,

Flf LD RUN I
GAuSS
8=2I04.88B2{I-800xl0 t) IIIII7fRRUp7IQQ
RUN 2
264884 8=2I04.88I I 8{I-Q,6lxl0 t)

0 0
a dl
2I049$
0
o + 0op 0
O 0
0
e
e~
0
2I04$%

240 480 ?20 950 I200 I440 l580

7l IIII E —HOURS

FIG. 3. Experimental data for runs one and two.

VOLUME 10, NUMBER 3 PHYSICAL REVIEW LETTERS 1 FEBRUARY 1963

Table I. Results of least-squares fit.

Run one Run two

80 (gauss) 2104. 88232 0. 000049 + 2104. 88118 + 0. 000085

Observed slope (gauss jh) (1.684 + 0. 19) x 10~ (2. 022 + 0. 15) x 10
Time constant (yr) 144 500+ 16300 119 450 + 8700

Upper limit to the equivalent 4. 3 x 10 2~

4. 9 x 10
resistivity Nb-25% Zr wire
(ohm-cm)
Maximum resistance of spot 6. 5 x 10-~~ 7. 4 x 10 '~

weld, assuming it to be en-

tire resistance of circuit (ohm)

t, and the terms B, c, and t are constants. in runs one or two. However, if the effect de-
The decay rates given, respectively, by the two scribed by relationship (2) were the only decay
expressions (1) and (2) are mechanism operating, the slope would be so much
greater at the higher field that the decay would be
dB/dt = -Bo/T
very apparent.
This follows from (4) by taking the ratio of
slopes at two different field levels, B, and B,:
dB/dt = (c/t ) e-xp[(B — B)/c]. (4)

The runs one and two were made at approximately

600 gauss below critical field, which is too low to
d 8/dt
dBJdt
, (B—B,
( c ),
allow a comparison between relationships (l) and Expression (2) may be rewritten as
(2) in a, relatively short period of time. Conse-
B =B'- clnt,
quently, run three was undertaken at approximately
36 gauss below the critical field. The data for run where
three are shown in Fig. 4. Difficulty in attaining B' =8 +clnt
a field close to the critical field required the coil C C
to be driven normal and re-energized some 50
and
times. The cycling caused the homogeneity of
the field over the length of the sample to deterio- dB/dt = c/t. -
rate by about an order of magnitude. This re-
sulted in a wider scatter of the data and consider- In this theory, B' is not uniquely defined, but
ably less accurate measurement of the slope than it is reasonable to assume that B' is approximated

FIELD I

GAUSS
———— RUN
— 8 = 2684.24584 ( +
5
I 3.68 x IO tj

0
o o 0
0 0
2684.25—
0
0 —oo O—
0
\l
0

0
0 o ~ o op
0
00 0 0
0
0
'0 0
0

l00 150 200

T I ME - HOURS

FIG. 4. Experimental data for run three.

Vox, UMz 10, NUMszR 3 PHYSICAL REVIEW LETTERS 1 FEBRUARY 1963

by the highest field experimentally attainable, that change of runs one and two.
is, a quantity very close to critical field. Having All of the runs of this experiment were made in
made this assumption, the constant c then can be the superconducting portion of the critical state
evaluated from (5), (7), and the experimental re- diagram, whereas the experiments of Kim, Hemp-
sults of run two. For an observed 8' of 2720 gauss, stead, and Strnad were done just outside this re-
c is calculated to be about 37 gauss per e-fold. gion. The decay laws may be different on the two
Applying (5) to runs two and three, it is computed sides of the critical state curve.
that the ratio of the decay rate of run three to that
of run two should be 6. 6x10', in which case the
'H. Kamerlingh Onnes, Konikl. Ned. Akad. Weten-
field as a function of time would be represented by
schap. , Proc. 23, 12, 278 (1914) {Commun. Nos. 104b
a vertical line on Fig. 4. and 140c from Phys. Lab. Leiden).
It is quite clear that if flux creep is present in 2R. F. Broom, Nature 190, 113 (1961).
this experiment, it is masked by a stronger effect. 3S. C. Collins (unpublished); quoted by J. %. Croute,
Two possible explanations other than energy dis- IBM J. Res. Develop. 1, 295 (1957).
sipation in the superconductor are these.' 4D. J. Quinn, III, and%'. B. Ittner, III, J. Appl.
(a) The spot weld may have a finite resistance. Phys. 33, 748 (1962).
A. F. Hildebrandt and D. D. Elleman, Bull. Am.
(b) The radial magnetic pressure of about 2. 7
Phys. Soc. 5, 111 {1960).
psi causes a tensile stress of 175 psi in the wire. ~P. Grassman, Physik. Z. 37, 569 (1936).
If the wire is experiencing a mechanical creep, YY. B. Kim, C. F. Hempstead, and A. R. Strnad,
then an average increase of coil radius of only Phys. Rev. Letters 9, 306 (1962).
two microinches would explain the total field P. %. Anderson, Phys. Hev. Letters 9, 309 (1962).

GAPLESS SUPERCONDUCTIVITY
J. C. Phillips*
Cavendish Laboratory, University of Cambridge, Cambridge, England
(Received 10 December 1962)

The possibility that a metal containing paramag- tation.

netic impurities could become superconducting According to AG, at T = 0 the Fourier compo-
without possessing a well-defined energy gap has nents of the quasi-particle correlation function
been discussed' by Abrikosov and Gor'kov (AG). G(p, & ) are given by
Recently two effects of paramagnetic impurities ('- &').
on the superconducting characteristics of thin In G(p, ~) = (~ + &)/(~'-
films have been studied. ' These are the dc resist- In the absence of impurity scattering the poles of
ance, which gives the transition temperature T~ (1) give the usual quasi-particle energies
as a function of c, the atomic percent of magnetic
~ (( 2 + g 2) 1/2 (2)
impurities, and the I - 1/' tunneling characteristic.
The result was found that b, the energy gap de- where ( =v, (p- po). In the presence of impurity
rived from the I- t/' characteristic, appeared to scattering, ~ and b, are altered to
decrease with c almost twice as fast as T~. It = (u+ u')"',
disappeared (I- V Ohmic) at c = 1%, although Tz
(u (1/2T, )u/(1- (3)
was still nearly half the value in the pure metal. E, = ~+ (I/2T, ) 1/(1- u')"', (4)
Both the theoretical and experimental results
where &u is the (real) energy variable. Here u is
just quoted appear to contradict the sum rule'
an implicit function of ~defined by u =~/E. The
which relates the supercurrent strength (i. e. ,
total scattering rate is 1/T„and the spin-flip
the penetration depth in the London limit) to the
difference in area [o„(e)- gz(cu)j/&u. The purpose
scattering rate is proportional to
of this note is to clarify the meaning of "energy 1/T = 1/2T, —1/2T, . (5)
gap" in a superconductor containing paramagnetic
impurities. To do so we need not change any of From (3)-(5) one finds a consistency equation
AG's equations, but only their physical interpre- which determines the contribution of each state