Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Determination of the temperature dependence of the superconducting energy gap of

lead by means of single-electron tunneling

Jane Doe
Brown University
Department of Physics
(lab conducted in collaboration with John Smith)
(Dated: August 19, 2014)
The superconducting gap energy of lead was measured at a range of temperatures for com-
parison with
q the predicted dependence for weakly-coupled quasiparticles given by BCS theory:
3.10kb Tc 1 − TTc , which is accurate near the critical temperature, Tc = 7.2 K, as well as
2∆0 = 3.50kb Tc , which is relevant near T = 0. This was accomplished by observing single-electron
tunneling through an oxide barrier between lead and aluminum at temperatures for which lead
is superconducting and aluminum is not (1.1 K < T < 7.2 K). In order to improve on previous
results, which have taken the inection point in the I(V ) curve to indicate the value of ∆(T ) [3],
the displacement of the inection point is approximately corrected for using values obtained by
numerical determination of the tunneling current based on the density of state given by BCS the-
ory. The dependance of current on a voltage bias was compared with numerical integration of a
semiconductor-like model of the system built on BCS theory. As expected, the resulting gap ener-
gies were overpredicted by the asymptotic approximation, with agreement getting worse at lower
temperatures. Because the experimental apparatus used was limited to minimum temperatures of
about 1.5K , the expected constant ∆ → ∆0 as T → 0 could not be conrmed absolutely; however, if
the energy gap at the three lowest temperatures are clustered in a manner which suggest a likelihood
they had leveled o at a constant ∆0 , in which case, an error-weighted average of these data points
predicts ∆0 = 0.001360±0.000030 eV, or approximately 4.4kb Tc , which is within the range of typical
energy gaps reported by Tinkham [8] of 3.0kb Tc . 2∆ . 4.5kb Tc . This value would also be in agree-
ment with the gap energy reported by Giaever (1960) at T = 1.6 K of ∆0 = 0.001303 ± 0.000031
[3]. Uncertainty in the experiment was found to be dominated by noise in the current signal and
the postprocessing required to extract the energy gap.

I. INTRODUCTION coulomb repulsion, even slightly, electrons will pair to-


gether, becoming bosonic quasiparticles [6]. Once paired,
only thermal excitation or external forces of a sucient
While phenomenological models were eective at
magnitude will break the so-call Cooper pairs. This ex-
describing many pure superconducting states for the
citation energy, ∆, is shown in BCS theory to be related
rst half of the 20th century, the details of the process
to the interaction potential between the electrons as [8]:
were a black box until BCS theory of Bardeen, Cooper,
and Schrieer and, to a lesser extent, Ginzburg-Landau
theory were developed to describe the phenomenon h i
Tanh 1
p
2β ζ 2 + ∆2
on a microscopic scale [8]. BCS theory motivated
Z ~ωc
1
= dζ , (1)
the London equations and explained the Meissner
p
N (0)V 0 ζ 2 + ∆2
eect. It also predicted a non-conducting energy gap
and a veriable form of the excitation energy as a where ~ωc is the cuto energy of the electrons at which
function of temperature, as well as other constants of the lattice-induced attraction of the electrons is canceled
the phenomenological equations [2, 68]. Despite the by the coulomb potential, N (0) is the density of states of
advantage of this understanding, superconductivity is the material at T = 0 K, V is the interaction potential, β
still a subject driven by empericism more than theory; is the Boltzmann factor, and ∆ is the excitation energy.
new superconductors and their specic properties have From this relation, the approximate forms of ∆(T ) can
generally been found by probing new materials without be shown to be [7, 8]:
precise expectations. Nevertheless, if emperical luck
is to be augmented to realize the potential of high-
temperature superconductivity, the microscopic details ∆(T ) = 3.064kb (Tc − T ) T → Tc (2)
ought to be explored and expanded. It is with this in ∆(0) = 1.76kb Tc T →0 , (3)
mind that the temperature dependence of the energy
gap is explored, reproducing a key success of BCS theory. where Tc is the critical temperature marking the onset
of superconductivity; these can be seen in gure 1. The
BCS theory recognized the important role that small excitation energy, ∆, can be thought of as similar to the
ion-lattice-based attractive forces between conducting band gap in semiconductors; free fermions may only oc-
electrons; when such a force is strong enough to overcome cupy states at an energy of ∆ or more above the ground
2

rent, as shown in gure 2. This feature can be used


to discern the energy gap from smoothed I-V curves at
nonzero temperatures.

FIG. 1: The predicted asymptotic behavior of the


energy gap, ∆, of lead as a function of temperature
according to BCS theory. The intermediate energies can
be evaluated numerically from the theory, but are not
done so here.
FIG. 2: Numerically integrated dV
dI
vs bias voltage for
lead at the tempearatures used in this experiment.
superconducting state. Cooper pairs, being bosonic in
nature, will fail to tunnel as would a typical electron (see
Josephson tunneling), thus, the resistance across a bar-
rier separating a superconductor and another conductor,
II. EXPERIMENT
normal or super, will be very large at low bias voltages,
and ohmic with large bias voltages across the barrier.
The density of states of such a superconductor is de-
scribed by [8]:

E
N (E) = √ forE > ∆ (4)
E2 − ∆2
N (E) = 0 forE < ∆ , (5)

which blows up at the gap energy. This can be used to


model the current due to the tunneling of free electrons
from a superconducting material to a normal material at
nite temperatures:

Z ∞
I∝ N () (f () − f ( + E)) d , (6)

where E is the bias energy of a single electron driving FIG. 3: A schematic of the slide on which the
the tunneling and f () is the fermi function of a given lead-oxide-aluminum junction is axed.
electron energy,  and has the form:
In order to determine the temperature dependenceof
1 the superconducting gap energy of lead, a lead-barrier-
f () = (7) aluminum junction was chosen for study. The low
e/kb T + 1
critical temperature of aluminum of 1.1 K ensured
At T = 0, this current drops o abruptly at E = that the junction tunneling would be in a consistent
∆; however, at nonzero temperatures, the function is state throughout the range of temperatures the dewar
smoothed out and passes through zero. The inection apparatus (see gure 4) was capable of reaching. The
point of this function approaches ∆ as T → 0 and drifts junction was formed as the overlap between thin lms
right slightly at higher temperatures, which can be ob- deposited by a thermal evaporator. The lead lm was
served with greater ease in the rst derivative of the cur- measured to be 130 nm thick, while the aluminum lm
3

The superconductive properties of each junction were


recorded using a 4-wire measurement setup as depicted
in gure 3. A separate circuit outtted with a 36 Ω resis-
tor ensured that an accurate current reading was being
taken. The voltage bias, which was varied dependently
on the current produced by a wave function generator,
was amplied by a dierential amplier, before being
routed to both an oscilliscope and a DAQ for process-
ing. Data sets were taken at six dierent pressures for
all three junctions.

III. RESULTS AND ANALYSIS

Despite initial concernes over alignment of the de-


posited layers of metallic lm on the junction slide, the
experiment produced reasonable data on all three junc-
tions. The only concern was that there was a slight oset
of the bias voltage from zero compared with ground. The
reason for this remains unknown; however, the eect was
observed to be small. One of the junctions (the lower
one, which was the focus of the forthcoming analysis) ex-
hibited slightly nonlinear current as a function of time.
This eect was found to dissappear at cryogenic temper-
atures. The resistances of each of the junctions on the
slide at room temperature were determined using a 4-
point measurement and are given in table I. These are
on the high end of the expected range of values (deter-
mined via personal communication through Dean Hudek
FIG. 4: A schematic of the nested dewars containing as opposed that listed in the lab manual [1]).
the slide.
Junction Upper Middle Lower
Resistance 17.63Ω 10.08Ω 14.40Ω

beneath it was measured to be 140nm thick. A thin TABLE I: Resistances of each of the junctions used in
oxide layer was allowed to form in between the two the experiment. Each was found to respond in phase
layers by exposing the aluminum to the atmosphere for with a slow driving frequency, indicating suitability of
15 minutes. The ends of the lm strips were supported the junctions for the experiment [1].
with hard chromium 150nm thick to aid in contact
with the cryogenic apparatus. The slide featured three Processing the resulting data was a time intensive task
potential junctions, only one of which was analyzed in due to the careful smoothing process required, especially
this article for reasons to be discussed in the next section. at higher temperatures, to achieve reasonable dVdI
curves.
As a result, only a single junction was analyzed. The
The slide was lowered to the appropriate supercon- lower junction was selected because it was characterized
ducting temperatures in the range of 1.49 K < T < 4.20 by an intermediate amplitude response to the driving
K using the dewar apparatus shown in gure 4. The frequency and intermediate resistance, making it most
setup featured two nested dewars, each with an outer representative of the set of junctions on the slide. Data
buer which can be evacuated during the experiment. was recorded at 100khz and, particularly with the cur-
The outer dewar was kept full of liquid nitrogen, while rent channel, featured considerable noise. A two-stage
the inner dewar was, after being precooled, lled with smoothing process was necessary to generate the desired
liquid helium 4. The slide was suspended in the center functional dependence in which the value at a given time
of the bottom of the inner dewar within the liquid is the result of a weighted average over a substantial
helium. Helium vapor above the liquid helium was then period of data. This would be expected to aect the
pumped out of the central chamber to further reduce the form of the voltage bias data, so care was taken to keep
temperature of the junction through evaporative cooling. the smoothing much smaller than features exhibited in
The minimum pressure achievable by this system was 3.5 that data. The current data, on the other had, was
Torr, which results in a temperature of 1.49K in helium 4. approximately linear and could be smoothed over larger
4

(a) The current vs. time: 1.49 K. (b) The current vs. time: 4.2 K.

(c) dI
dt
vs. time: 1.49 K. (d) dI
dt
vs. time: 4.2 K.

(e) The bias voltage vs. time: 1.49 K. (f) The bias voltage vs. time: 4.2 K.

(g) dv
dt
vs. time: 1.49 K. (h) dv
dt
vs. time: 4.2 K.

FIG. 5: A comparison between the behavior of various properties at the lowest and highest temperatures
investigated, 1.49 K and 4.20 K. Only data from the lowest of three barriers is displayed.
5

(a) dI
dV
vs. bias voltage: 1.49 K. (b) dI
dV
vs. bias voltage: 2.14 K.

(c) dI
dV
vs. bias voltage: 2.29 K. (d) dI
dV
vs. bias voltage: 3.48 K.

(e) dI
dV
vs. bias voltage: 4.20 K. (f) Numerically integrated curves for dV
dI
vs. bias
voltage are provided again for comparison with BCS
predictions.

FIG. 6: dV
dI
vs bias voltage at dierent temperatures along with theoretical curves obtained from numerical
integration of the tunneling current (equation 6). The distance from the line of symmetry to the peaks were
corrected by the numerically integrated peaks to yield values for the energy gap at each temperature. Note to
Students: uncertainty documentation is lax here. ±σ should have been included on temperatures and pressures in
the subcaptions.
6

FIG. 7: ∆ as a function of temperature, the primary result of this investigation. The predictions of BCS theory are
shown in the T → Tc limit (equation 3) as the blue line as well as ∆0 with the red × at T = 0 K. The measured
value of ∆(1.6 K= 4.2kb Tc = 0.0013 eV reported by Giaever in 1960 is indicated by the black × [3], while the range
of typical values for superconductors (in terms of Tc ) given by Tinkham is indicated by the vertical red line on the
y-axis at T = 0 K [8]. If constant prole is assumed to have been achieved below T = 2.5 K, the present data yields
a gap energy of ∆0 = 0.001360 ± 0.000030 eV, indicated by the dotted horizontal red line.

integration times without aecting the local prole. Temp I dI V dV


There were limitations to this as a slight response to the 1.49 K 10−2 s 10−3 s 2 × 10−3 s 3 × 10−4 s
superconducting transitions is observable in the current 2.14 K 10−2 s 2 × 10−3 s 10−3 s 5 × 10−4 s
(see gures 5a and 5b), showing limitations of the wave 2.29 K 10−2 s 2 × 10−3 s 10−3 s 5 × 10−4 s
function generator used. The smoothed current and bias 3.48 K 2 × 10−2 s 2 × 10−3 s 10−3 s 5 × 10−4 s
voltage could then be converted to dierential data via 4.20 K 1.6 × 10−2 s 2 × 10−3 s 10−3 s 5 × 10−4 s
dφi = φi+1 − φi−1 for the general property φ. A second,
limited smoothed needed to be performed on the two TABLE II: Integration times for smoothing. Smoothing
sets of dierential data. All smoothing integration times was performed on Kaleidagraph using the built-in
are listed in table II. smooth macro. Only the lower junction data was
processed due to the time-intensive nature of this
The rst point of interest encountered in the processed process, both in terms of computational and wall-time.
data is in the aect of the superconducting transition
on the current measured passing through the junction,
seen in gures 5a and 5b, but much more apparent in higher temperatures; in particular, the dependence at
dt in gures 5c and 5d. While it is understandable that positive bias voltages near peak dV features a supressed
dI dI

this sudden transition in conductivity occuring at the peak. The resulting curves are asymmetrical at higher
junction might aect the current in the circuit, it is not temperatures. While the smoothing does have some
clear why the aect changes parity from an odd to an eect on the prole of dV dI
, changing the integration
even eect as the temperature lessens. time did not noticably mitigate the suppression of the
positive voltage maximum.
The dV
dI
curves depicted in gure 6 were the purpose
of smoothing the data to take derivatives, yet it is clear The aected peak at positive bias voltages at higher
that the resulting curves do not appear as expected at temperatures is dealt with by utilizing the symmetry
7

try) and the uncertainty in the peak value, which was


chosen to encompass any uncharacteristic uctuations
in the viscinity, which tended to persist across cycles,
as well as the variance of peak values from cycle to
cycle. The uncertainty in temperature was determined
from drift in pressure propogating through emperically
determined temperature tables for liquid helium. These
were found at all pressures to result in < 4% uncertainty
in temperature, which was subsequently used for all
temperatures. The uncertaintly in the resistance of the
resitor was unknown, so a value of 1% was assumed. All
of the sources of uncertainty contribute to the error bars
displayed in gure 7.

The primary result of this experiment is the tempera-


ture dependence of the gap energy, which is depicted in
gure 7. The values are seen to fall below the T → Tc be-
havior, as expected; however, they are signicantly above
the predictions of BCS theory and at the upper end of
the approximate range of typical values of ∆0 described
by Tinkham (1996) [8]. This range of values is conrmed
FIG. 8: The bias energy shift in the predicted inection by the experimental result of Giaever (1960), who re-
point of I(V ) provided by numerical integration of the ported a gap energy of ∆0 = 0.001303 ± 0.000031 at 1.6
tunneling current (equation 6. A quadratic t is K [3]. Given that the three lowest temperature gap en-
provided for future corrections. ergies from the present study, as well as the gap energy
of Glaever are all clustered around a similar value, it
may be assumed that the gap energy has approached a
at low voltages. The prole at dV
dI
= 0.05 Amps/Volt constant at these temperatures. These three data points
was used to determine the line of symmetry of the may be combined using an uncertainty-weighted average
dependence, which could then be used along with the such that (provided for student reference) [5]:
negative bias voltage maximum to compute the gap
energy. X ∆i

As was stated, the resulting energy at the inection i


σi2
¯
∆= X (8)
point (peak dV dI
) is corrected using the peak in the 1
numerically determined form of the rst derivative of σ 2
i
i
equation 6 initially assuming values of ∆ given by BCS v
theory in equation 3. While the inection point of I(V )
u 1
σ̄ = u (9)
is not observed to drift much in terms of energy, its
uX 1
t
relative position with respect to the gap energy gets σi2
i
considerably larger as the temperature is increased.
This is depicted in gure 8. At T = 1.6 K, at which This results in a predicted gap energy at T = 0 of
Giaever (1960) conducted this experiment, the error is ∆0 = ∆ ¯ T <2.5K = 0.001360 ± 0.000030, which is in
only 11%; however, in this investigation, the error would agreement with Giaever up to uncertainty and still a
have reached as high as 48% at T = 4.20 K. Therefore, reasonable value according to Tinkham [3, 8].
this correction is both necessary due to the present
measurements above 3 K and an improvement over some The deviation from the expected prole of as well as
of the original measurements in the literature [3]. the assymetry present in dV
dI
merits serious consideration
of potential sources of error. Of note is the smoothing
Due to the measured gap energies dependence on process, which is an indirect manifestation of the error
specic features in I(V ), propogation of the current associated with the noise in the current signal and, to
error through the smoothing process and the inection a lesser extent, the voltage bias signal. Smoothing the
detection was not done explicitly; rather, the error of bias voltage data would have preferrentially increased the
this was wrapped up in a conservative estimate of the gap size; however, the integration time of the bias volt-
inection point uncertainty. The uncertainty in the bias age smoothing was very small. Error was more likely to
energy of the peak of dVdI
was the root mean square of be introduced by the very large integration time of the
the uncertainties of the two values at which dV
dI
= 0.05 current, which would have acted to spread out and dis-
Amps per Volt (used in determining the line of symme- tort the slight uctuations in the current at the onset
8

of superconductivity. Indeed, this is where the great- approximated due to the insensitivity of gap energy to
est deviations from the expected behavior are observed. temperature as T → 0, which resulted in a value of
While the smoothing process is symmetrical and could ∆0 = 0.001360 ± 0.000030 eV. The results are consis-
not account for the assymetry observed at high temper- tant with a published gap energy in a similar experiment
atures alone, the current can be observed to transition [3] and expected values for superconcductors [8], but also
from the appropriate odd parity at lower temperatures deviated somewhat from the purely theoretical BCS pre-
to an unexpected even parity at higher temperatures in dictions, which must only approximately describe the mi-
gures 5a and 5b. This is likely related to the consis- croscopic processes involved in superconductivity. BCS
tant bias voltage oset of 0.125 ± 0.02 mV observed at all theory is based on a number of assumptions, including
temperatures, the other notably breaking of symmetry weak electron coupling (valid for metallic superconduc-
across current direction. It is estimated that the cause tors) and the simplication of the attractive interaction
of this error is due to the quality of the fabricated junc- potential between paired electrons to a constant [2, 8].
tions and can be mitigated by experimenting with slower The latter approximation may be the source of the vari-
deposition rates of the lead, as well as using a thinner ation between theory and observations of energy gaps at
aluminum oxide layer. very low temperatures. Future experiments may benet
from conrming zero oset in the bias voltage signal of
each junction prior to performing the experiment as well
as taking steps to reduce noice in the current signal to
IV. CONCLUSIONS
more accurately the slight uctuations around the onset
of superconductivity, perhaps by introducing a second
The temperature dependence of the superconducting dierential amplier. As a whole, the experiment takes
gap energy of lead was measured at discrete points the as step towards validating the functional dependence of
range of 1.49 K < T < 4.20 K and is given in gure the gap energy of lead on temperature given by BCS the-
7. The gap energy at T = 0 was able to be roughly ory.

[1] Stephen Albright and Dean Hudek. Electron para- [5] Ifan G. Huges and Thomas P. A. Hase. Measurements and
magnetic resonance. Online; accessed 24-October- their Uncertainties: A Practical Guide to Modern Error
2013.<https://wiki.brown.edu/conuence/download/ Analysis. Oxford University Press, 2010.
attachments/5896/TF+Tunneling.pdf?version=13& [6] Charles. Kittel. Introduction to Solid State Physics. John
modicationDate=1375924481000>. Wiley & Sons, Inc., 1967.
[2] J. Bardeen, L. N. Cooper, and J. R. Schrieer. Theory of [7] E. A. lynton. Superconductivity. Methuen & Co, ltd.,
superconductivity. Physical Review, 108:11751204, 1957. 1962.
[3] Ivar Giaever. Energy gap in superconductors measured by [8] Michael Tinkham. Introduction to Superconductivity.
electron tunneling. Physical Review Letters, 5:147148, McGraw-Hill, Inc., 1996.
1960. [9] various manufacturers. Phys 1560 & 2010 equip-
[4] D. M. Ginsberg. Experimental foundations of the bcs the- ment manuals. Online; accessed 7-October-
ory of superconductivity. American Journal of Physics, 2013.<https://wiki.brown.edu/conuence/pages/
30:433438, 1962. viewpage.action?pageId=29406>.

You might also like