...

Shot Noise Suppression in an Atomic Point Contact

by user

on
Category: Documents
1

views

Report

Comments

Transcript

Shot Noise Suppression in an Atomic Point Contact
A New Resistance Standard
Microwave Measurements of Shot Noise Suppression in an
Atomic Point Contact
Ben Zimmerman1
Advisor: Konrad Lehnert2
August 10, 2006
1 University
2 JILA,
of Chicago, [email protected]
National Institute of Standards and Technology, University of Colorado at Boulder
Abstract
In this paper we investigate electron transport probabilities by simultaneously measuring conductance
and shot noise in an atomic point contact. The results show a characteristic noise spectrum for shot
noise at many resistances, with observed suppression of shot noise at integer multiples of the quantum of
conductance in accordance with theoretical expectations. We are able to use our shot noise suppression
measurements as a standard resistor, from which other unknown resistances can be calibrated.
Introduction and Motivation
Possible Applications
Instruments that can combine high precision with
Shot Noise as a New Resistance Standard low maintenance are rare and in high demand. Indeed, a great deal of the cost of micro electro-mechanical
The noise in the electrical current through an atom- devices (MEMS) comes from frequent calibration.
ically thin gap in a wire is related to the probabil- Accelerometer MEMS, for example, are used to deity that discrete electrons will be able to jump (or ploy airbags in cars, inform the inertial guidance
rather tunnel through) that gap [17]. This mani- systems of missiles, and direct satellites. Such defestation of the discreteness of charge is known as vicies usually consist of little more than a cantilever
shot noise, and has many interesting qualities. One beam with some type of deflection sensing circuitry.
demonstrated behavior of shot noise is that it de- As an accelerometer experiences a jerk, the cancreases theoretically to zero when the probability of tilever beam will deflect a piezo-electric crystal, proelectron transport is 100 percent [17].
viding a measurable voltage to the device’s circuit.
However, over the course of the device’s life as it
This is reasonable–if every electron is tunneling through
bends again and again, resistances will change as the
the gap, then there should be no fluctuations in the cantilever and piezo change. Instead of calibrating
current and hence no shot noise. What is particu- it over and over (a process that is costly for airbags
larly interesting is that in an atomic point contact and early impossible to do for missiles or satellites),
(essentially a wire with a gap) the transmission prob- a shot noise resistor within the device’s digital cirability becomes 100 percent at integer values of 2e2 /h,cuitry could calibrate all the resistances at once.
which equals 1/12.9 kΩ. Thus, shot noise is suppressed at resistances determined entirely by fundamental physical constants. Measurements of shot
noise at a variety of resistances, therefore, should Noise Due to Electron Tunneling
show minima at certain well-defined resistances. If
the resistance scale was previously undetermined,
Electrical current across a conducting wire will flucthen the shot noise minima would be enough to calituate over time. These fluctuations occur about an
brate it. This is how the physical phenomena of shot
average value–at one time it will be at one current,
noise can be used as a resistance standard.
another time at some other value, but over time the
The quantum Hall effect is used worldwide to main- average value of the current, i(t)i(t − τ )τ (where
tain and compare the unit of resistance, and for good tau is some interval between the present measurereason: ”the reproducibility reached today is almost ment time and another) can be found.
two orders of magnitude better than the uncertainty
The similarity of the current at t and at t − τ can be
of the determination of the ohm” using the best known
measured via the cross-correlation, a function of the
values of Planck’s constant h and the unit of charge
time between measurements used to find features in
e [1], about one part per billion. Such remarkable
an unknown signal by comparing it to a known one.
precision cannot be matched by a shot noise resisFor our experiment, we use one measured value of
tance standard yet. However, the intense magnetic
current in order to determine the value at some later
fields required to measure quanta of resistance via
time: we cross-correlate the current signal with itthe hall effect are difficult to produce on the small
self, which is known as an auto-correlation function
scale of a device. Because of this, using an atomic
[2] of the current, RI (τ ).
point contact to measure shot noise suppression has
many practical applications for calibrating resistance. The Fourier transform of an autocorrelation function gives the spectral density, which can be written
in this context in terms of the current as
screen out the current fluctuations from any indiI ω =
SI (ω),
(1) vidual electron. However, when charges are tunnelω
ing through a potential barrier or an atomically-thin
wire (where the conductor’s length is less than the
over a range of frequencies ω. Electrical current mean free path of an electron), the discreteness of
fluctuations resulting in a measurable noise power charge manifests itself as shot noise.
SI (ω) can be caused by a number of physical phenomena [5]. In this paper, the motion of electrons Electrons tunnel across a potential barrier with no
leads to two dominant types of noise : thermal noise time or frequency dependence. So for a conducting
wire with a small gap in it, current measurements
and shot noise.
should have detectable fluctuations–noise–due to individual electrons randomly crossing the gap. This
can be seen in the classical result for the noise power
Thermal Noise
of shot noise,
2
A conducting material at a nonzero temperature will
spontaneously have an electrical current due to phoS(ν) = 2eI¯
(3)
non interactions with free electrons. At equilibrium
with no applied voltage but a finite temperature, phonons
where I¯ is the average current. Although up to this
in the metal wire will agitate charge carrying elec- point we have discussed noise in terms of fluctuatrons. Movement of electric charge defines a cur- tions in the current, ohm’s law provides an insightful
rent, so this spontaneous electron movement leads version of the classical shot noise equation, namely,
to a measurable noise in the current across the wire
without any applied voltage.
S(ν)SN = 2eGV
(4)
This kind of thermal noise is known as JohnsonNyquist noise after the Bell Laboratory researchers
In this case, G is conductance and V the applied
who first discovered and explained it in 1928 [3][4].
voltage. So although shot noise occurs randomly,
They determined that in terms of the noise spectral
it (and thereby the likelihood of electron tunneling
density, thermal noise is
events) can be increased by applying a voltage across
the barrier. This is discussed further in the next secS(ν)JN = 4kbT G.
(2) tion, where it is shown mathematically that electron
tunneling is the main source of noise in an atomic
point contact.
Note that thermal noise is linearly dependent with
temperature; the higher the temperature, the more
thermal noise in a conductor. In order to measure
shot noise we must perform experiments at a low Electron Tunneling
enough temperature to minimize this noise.
The voltage dependence of noise in an atomic point
contact is what makes shot noise so useful. Shot
noise has previously been used as a thermometer
Shot Noise
that doesn’t require calibration [16], to determine
the gain and noise temperature of an amplifier [10],
The fundamental source of shot noise is the discreteand to study the fundamental physics involved in
ness of charge. Charge being carried by individual
electron transport. As this paper tries to shine light
electrons is usually not apparent in conductors–the
on fundamental transport physics allowing for the
mass of other electrons in the electron sea strongly
possibility of a new resistance standard, it follows
that electron tunneling must be discussed. Indeed,
we will show that this view of noise power is much
more enlightening than the classical equations for
thermal noise and shot noise.
Figure 2: A diagram of two groups of electrons separated by
Figure 1: A diagram of two groups of electrons separated
by a barrier in equilibrium. I R , the current from electrons
tunneling from the left to the right, and I L , the current for
electrons going left, are equal. Graphic from [16].
An atomic point contact can be thought of as a tunnel junction with electrons passing from one conductor to another across either an empty barrier or
an atomically-thin wire. This arrangement can be
represented by two groups of electrons spread across
a continuous set of energies separated by an empty
space as in Figure (1). Thermal energy, kB T , smears
out the top of both energy distributions. The rate at
which electrons can move from one group to another
is simply the current, with the average current equal
to the difference between the two rates. As shown in
Figure (2), when a voltage is applied the energy levels adjust to the potential difference, making electrons more likely to move in one direction than the
other.
a potential barrier with an applied bias eV . This applied bias
changes the difference in energy levels between the two groups,
making electron transport easier and I R larger than IL . Each
group is represented by a Fermi function at a finite temperature
T that causes the spread in values at high energies. Graphic
from [16].
2πe
|l|M(EF )|r|2D(EF )2
[fr (E) − fl (E)]dE
(5)
When evaluated, this integral yields I = V /R, showing that a tunnel junction such as an atomic point
contact is simply a ohmic resistor.
Furthermore, the noise spectral density of the current can be written as SI (V ) =
2
R
{fr (E)[1−fl (E)]+fl (E)[1−fr (E)]}dE (6)
Because the two groups of electrons distributed evenly which when evaluated [16] gives a second important
across energies are easily represented by Fermi func- result: the noise from electrons tunneling is
tions, two interesting and important results can be
obtained by applying the concepts of solid state physics
eV
(7)
S(ν) = 2eV G coth
to this problem. The total current can be written
2kB T
simply as I = IR − IL , but if we model the electron seas as Fermi functions, the current can also be
This is the general expression for noise–both therwritten as I =
mal and shot noise–due to electrons tunneling through
a potential barrier. It takes into account the experimental energy limits hinted at in earlier descriptions
of noise above, as well as in Figures (1) and (2).
In the high-temperature limit where the thermal energy, kB T , is much larger than the electrical energy
applied, eV , the noise due to electrons tunneling becomes
P(V)
T>0
T=0
S(ν) = 4kbT G.
(8)
V
the classical equation for Johnson noise. If, how4kBT/e
ever, kB T << eV , then equation (7) is dominated
by shot noise,
Figure 3: A theoretical plot of noise power P (V ) from elecS(ν) = 2eGV,
(9)
Rolf Landauer was the first to clearly point out that
in some systems, ”shot noise and thermal equilibrium noise are special limits of a more general noise
formula,” [17], as we have shown.
Neither thermal noise nor shot noise depend on the
frequency of the measurement–they are both white
noises. However, a fundamental difference between
the two, as can be seen from equations (9) and (8)
is that shot noise depends on the potential across the
barrier while thermal noise does not. This means
that, as shown in Figure (3), a plot of noise versus
voltage will show a flat level of thermal noise with
shot noise linearly increasing with voltage.
tron tunneling in an atomic point contact as a function of voltage V . At the zero temperature limit, theoretical shot noise
is plotted as a blue dotted line. The green dotted line indicates the level of thermal noise. Interestingly, the temperature
can be determined from the length of the dotted green line.
Graphic inspired by [16].
Methods and Materials
Mechanical Break Junction
Deep within a helium dewar, at the very end of a
long stainless steel evacuated probe, is a vibrationisolated apparatus that holds a mechanical break junction, the heart of our experimental setup. This technique has been used widely to create an atomic point
contact [10][11], but the purpose of a mechanical
break junction (or MBJ) must be made clear and our
Note again that because thermal noise increases lin- design is innovative in ways worth noting.
early with temperature, at higher temperatures the
level of thermal noise will overwhelm the shot noise. Atomic point contacts can be made by bringing an
In order to reduce the amount of thermal noise and atomically sharp tip of an atomic force microscope
successfully measure shot noise we perform our ex- [12] or an STM tip [13] near a substrate covered in
nano-wires or GaAs-AlGaAs heterostructures; anyperiments at the low temperature of 4.2 K.
thing sharp, conducting, and dense enough so that
In addition to working at low temperatures, how- the sharp tip is likely to be atomically-close to one
ever, we must also make the length of the conductor of them, like a javelin over a field of grass. A meas short as possible, if not break it outright. This chanical break junction, however, creates an APC
was done by using a mechanical break junction to by by pulling apart a piece of conductor to stretch it
stretch a gold wire and then form an atomic point until it breaks. The MBJ leaves the conducting wire
contact, as described below.
either in two parts with atomically-sharp points, or
with an atomically thin wire between them [14].
driving rod
tensioning
spring
piezoelectric
crystal stack
flexible
substrate
clamp
rod at the top of Figure (4). The forked connection allows us to bring the atomically-sharp points
of the stretched and broken wire together and then
disengage, switching to the finer control of the piezo
stack actuator, which is rated to move 11.6 ± 2.0μm
when 100 V is applied at room temperature. At
4.2 K we expect perhaps 10 percent of that range
of motion, yet even that is enough for our measurements.
countersupport
notched gold wire
circuit board
Figure 4: A diagram of our mechanical break junction setup.
During an experiment, the forked driving rod is screwed down,
bending the flexible beryllium-copper substrate, which then
stretches the clamped-down and notched gold wire until it
breaks. Once broken, the piezoelectric crystal is used to finely
tune the bend of the substrate and thereby the separation of the
gap in the gold wire, which is connected to a circuit board to
make measurements. The entire apparatus operates in vacuum
at 4.2 K
Piezoelectric crystals transform electrical energy into
precisely controlled mechanical displacements[15].
While making measurements of shot noise suppression we need to take data across a wide range of
resistances. Since the separation of the atomic point
contact determines the transmission probability of
electrons and therefore the resistance of the wire
(see equation (5)), we can use a piezo actuator to
finely adjust the resistance of the APC.
From there, however, we need to be able to measure the resistance as well as the fluctuations in the
APC current. Thus, at the bottom of Figure (??), the
notched gold wire is attached to a circuit board with
In our experiments we use a gold conducting wire
the measurement circuit described below.
notched with a razor blade. The internal stresses,
reduced connection, and brittleness caused by the
liquid helium temperature (4.2 K) makes breaking
the wire easier. However, in order to make electron Measurement Circuit
transport measurements, we also need to be able to
change the separation of the wires–our quantum tun- After using a mechanical break junction held at 4 K
neling barrier–from open to close and many places to stretch a gold wire and create atomic point coninbetween. Controlled bending is accomplished by tacts, we then made simultaneous measurements of
attaching the gold wire to a appropriately flexible current and noise power as a function of voltage.
piece of 0.25 mm-thick beryllium-copper, which is The current measurements allowed us to find the reelectrically isolated from the conducting wire by a sistance of the APC, which was then used to inveslayer of insulating kapton tape about 0.08 mm thick. tigate the suppression of shot noise.
The gold wire is held in place on the berylliumcopper by two clamps, which is pressed against three This is trickier than it sounds, however. Because
countersupports by the driving rod and piezoelectric shot noise occurs due to individual electrons movcrystal stack.
ing across a potential barrier, it occurs very quickly.
Moreover, the fluctuations in the current that we are
The separation between the two parts of the broken measuring are small. In our experiments we used
wire comes is roughly controlled by the extention a microwave amplifier to account for the problems
of the driving rod. At the top of the driving rod, with measuring shot noise and allow for a faster
a hand screw extends out of our cryostat probe. It readout.
then passes into the probe via an ultra torr fitting,
down the 1.5 m long interior, and meets the driving The circuit inside the helium cryostat along with the
amplifier
S/A
+
RF
DC
RF/DC
will be inaccurate [21]. Therefore, in addition to
our measurements of conductance and shot noise,
we use a slightly different circuit to measure Γ for
each APC (Figure (6)). These results for a typical
sample are shown in Figure (7).
amplifier
300 K
S/A
4.1 K
directional
coupler
L
R
RF
signal
APC
+
RF
C
DC
RF/DC
300 K
Figure 5: The circuit diagram for the circuit used to make
simultaneous conductance and noise measurements. A DC
power supply at room temperature applies a voltage to the
atomic point contact and accompanying circuit in the helium
dewar. This circuit matches the impedance of the shot noise
measurement circuit, which amplifies and then measures the
noise power of the APC.
APC is used to effectively couple the noise power
from the high impedance APC to the microwave
amplifier at the top of the circuit (see Figure(5)).
This is done by carefully matching impedances via
selection of the inductor and resistor to account for
the capacitance inherent in the system. Because the
selection of appropriate resistors and inductor values for impedance matching is a fundamental exercise and is explored more thoroughly elsewhere [10]
[21], we will now focus on a direct measurement of
a mismatch, the reflectance coefficient Γ.
4.1 K
L
C
Figure 6: This circuit diagram used for measuring the reflectance coefficient Γ has a slight difference between the conductance noise measurement circuit: a directional coupler allows us to send a signal down at the APC circuit and then
measure the reflected signal. The signal is at the resonant frequency of the APC, as measured with a network analyzer.
Just as P = IR, the total noise power can be expressed as
SP =
Reflectance Calibration
The reflectance coefficient, Γ, indicates the amplitude of the reflected voltage and current standing
wave on a conducting wire, where the other part of
the superposition is the incident signal. To obtain
no reflected power, the load impedance must match
the characteristic impedance of the tramsission line,
which in this experiment are BNC cables. If the load
is mismatched, however, the power being measured
APC
SI
= SI R
G
(10)
If there is a power mismatch in our measurement circuit, then the available power Pav is reduced to only
the fraction that is coupled into the sample. Since
the available power is equal to the total noise power,
we can write
Pav (1 − Γ2 ) =
SI
G
(11)
Gamma (dimensionless)
1
0.8
0.6
0.4
0.2
0
0
1
2
Resistance (ohms)
3
5
x 10
Figure 7: A plot of the reflectance coefficient Γ as a function
of resistance. Interestingly, this APC is best coupled near 30
kΩ, as the minimum indicates.
which is equivalent to the result that the noise power,
instead of being equal to 2eV G as before, is really
Figure 8: The noise power from an atomic point contact plotted as a function of voltage for three different resistance values. As the resistance approaches 12.9 kΩ, the first quantum
of resistance where the tunneling probability T = 1, the slope
the noise is reduced.
separations of the APC and therefore three differ(12) ent resistance values. As the resistance increases,
SI = 2eV G(1 − Γ )
the slope of noise ”V” arms becomes smaller and
Thus, is is clear that to return the expected value smaller until, at 12.9 kΩ, the noise curve looks alof noise power, we must have to normalize SI by most flat in comparison with the other two. This is
due to shot noise suppression, which will be better
(1 − Γ2 ).
explained in the next section.
The measurements for Γ are taken for a wide range
of resistances, allowing us to normalize each measurement of shot noise. One aspect of our shot noise
Results
measurements is particularly well-suited for measuring shot noise suppression, and must be briefly
discussed.
Suppression of Shot Noise
2
Shot Noise Measurements
Shot noise increases linearly with voltage. However, even at zero volts the noise can be offset significantly from zero depending on the microwave amplifier being used. This behavior is the reason why,
in these experiments, we measured the supression of
the slope of shot noise, not exactly shot noise itself.
Measurements of shot noise in two-dimentional electron gases in quantum hall experiments [18] show
that the conductance G increases in quantized steps
of G0 = 2e2 /h [18] known as the quantum of conductance. In some of these experiments, shot noise
was observed to be suppressed around integer multiples of G0 [6]. Indeed, previous experiments have
shown similar suppression of shot noise in atomic
point contacts, most notably from the research by
van Ruiteenbeek et al. [20].
Shot noise, when plotted as a function of voltage,
looks like a ”V”. Three such ”V”s are shown in The suppression of shot noise at multiples of the
Figure (8), where each was measured at different
quantum of conductance can be explained with theory about the nature of electron tunneling through
atomic point contacts. Rolf Landauer and Markus
Büttiker give the conductance of a mesoscopic1 conductor as [7]
N
2e2 G=
Ti
h n=1
3
2.5
2
1.5
(13)
1
0.5
where Tn are the probabilities of transmission for
conducting channels. Note that for one channel the
conductance is simply G = 2e2 /hT , and when that
channel is perfectly correlated i.e. the probability of
an electron tunneling through that channel is 1, then
the conductance is G0 .
Landauer-Büttiker theory changes the equation for
the shot noise spectral density, previously S(ν) SN =
2eGV in terms of conductance, to
N
2e2 V
S(ν) = 2e
Ti (1 − Ti ).
h
i=1
(14)
The extra term (1 −Ti ) in the sum is due to the Pauli
exclusion principle, which states generally that no
two fermions can occupy the same quantum state.
As electrons are fermions, the Pauli exclusion principle prevents two electrons to tunnel through the
same conducting channel at once. It is this (1 − T )
term–and more broadly the behavior of fermions in
conducting channels–that causes the suppression of
shot noise shown above in Figure (8) as well as more
fully in Figure (9) below.
Our measurements of noise power spectral density
across a larger range of conductances are plotted in
Figure (10). At larger conductances, although the
shot noise still comes to a point at the quanta of conductance (implying that it is still being suppressed),
the overall level of shot noise is increasing. This is
due to more and more conducting channels opening
1
Mesoscopic is an intermediate scale between the macroscopic world and the point where the behavior of each individual atom becomes significant, roughly ten nanometers.
0
0
0.2
0.4
0.6
0.8
1
1.2
6
Transmission Tn
Noise Power (arbs)
5
4
3
2
1
0
0
1
2
3
4
2
Conductance (2e /h)
Figure 10: Shot noise measurements (in green) with minima
corresponding to the theoretical prediction from equation (14)
in black. Instead of being completely suppressed, however,
the noise increases with increasing conductance. This is due
to more and more conducting channels opening up at higher
conductances, which is unaccounted for in the simple model of
equation (14).
channels opening at once would allow for the overall rise in shot noise seen in Figure (10). Equation
(14), when the Ti are as shown in Figure (11), are
plotted below in Figure (12).
Determining Unknown Resistance Values
In the previous section we showed that shot noise
is suppressed at integer multiples of G 0 . Indeed, it
seems that our measured data not only confirmed
the Landauer-Büttiker theory but showed evidence
of multiple conducting channels open at once. Now
that the suppression of shot noise in our system has
been established, we can finally perform what this
paper set out to do in the beginning: use shot noise
suppression as a resistance standard.
To provide an adequate test of this standard, we took
four resistors between 1.2 kΩ and 120 kΩ, concealed
their markings, and labeled each with a randomlygenerated number. Keeping their true resistances
unknown, we then determined their resistance values by analyzing the shot noise suppression of an
connected atomic point contact. This analysis was
1.0
0.8
0.6
0.4
0.2
0
0
1.0
0.8
0.6
0.4
0.2
0
0
1.0
0.8
0.6
0.4
0.2
0
0
1
2
3
4
1
2
3
4
1
2
3
4
Conductance G0
Figure 11: Three diagrams of the probability of electrons
tunneling, T , as a function of conductance, with each line representing a conducting channel in an APC. In the top diagram,
onlychannel9.32rtimnce,43453(makcting)-32rthet(gr)12ransmissim,
surement of conductance is necessarily arbitrary2 ,
we can then find two minima of the shot noise and
know, due to theory described above, the resistance
of those two minima. This allows us to divide out
the applied voltage and write, for two APC-derived
resistances R1 and R2 ,
6
Noise Power (arbs)
5
4
3
2
R? =
R1 R2 (V2 − V1 )
R2 V1 − R1 V2
(17)
1
0
0
1
2
3
4
For the first and second minima of shot noise, R 1
should be equal to h/2e2 and R2 equal to h/4e2 .
This then makes R? equal to
5
2
Conductance (2e /h)
Figure 12: Shot noise as a function of conductance with
three plots of equation (14) in black, blue, and red for the upper, middle, and lower diagrams in Figure (11), respectively.
As more and more channels open, the theory better fits the
measured data.
_
V
R
h (V1 − V2 )
2e2 (2V2 − V1 )
(18)
Switching to an unknown resistor, measuring shot
noise across a wide range of arbitrary conductances,
fitting curves to find the minimums at the first and
second quanta of resistance, finding the corresponding voltage across the APC, and plugging them into
equation (17), we are able to make bridge measurements to determine the resistance of an unknown
bias resistor. An example of this curve fitting is
shown in Figure (14)
R?
+
R? =
VR
Figure 13: In this diagram of a voltage divider, R ? represents our bias resistor, R is the atomic point contact, V the
applied voltage and V R the voltage across the APC.
voltage divider. In this setup, a voltage is applied
to two resistors in series, R and R? , where R is the
resistance of an APC and R? that of a resistor with
unknown value. The voltage across R can be meaFigure 14: Noise data near the second quantum of conducsured, with the relation between it and the applied tance for resistor 225 is plotted as a function of conductance
voltage and two resistors being
in arbitrary units. A model is fit to the data in order to find the
minimum.
R
VR
=
V
R? + R
2
Our bias resistor allows us to determine the resistance of a
(16) sample, and therefore if the bias resistor is replaced with a re-
where VR is the measured voltage across the APC.
Using our experimental setup with an unknown bias
resistor, we can take measurements of shot noise
as a function of conductance. Although this mea-
sistor of unknown value, the measured conductance would be
uncalibrated. From one conductance to another it would still
be relatively correct (presuming that the voltage measurement
is linear), just in arbitrary units of conductance.
Method 2: Recalibration
Because our bias resistor allows us to determine the
resistance of a sample, when the bias resistor is replaced with a resistor of unknown value, the measured APC conductance–such as the x-axis of Figure (10)–would be uncalibrated. Recall however how
simple the calibration is, and moreover, that the calibration of the APC’s conductance with a known bias
resistor can be just as easily made of the bias resistor
if the APC’s conductance is known.
Fitting a curve to noise data near a minimum of shot
noise (as in Figure (14)) determines the exact arbitrary conductance where Landauer-Büttiker theory
predicts the APC’s conductance to be G0 . Multiplying the arbitrary conductance by the reciprocal
of G0 gives the bias resistor’s value. This is because equation (15) for an unknown resistor makes
Rbias = 1 so that
Rarb =
VAP C
Vbias
(19)
Sources of Error and Results
In the bridge measurement, we used two minima of
shot noise and assigned them the resistances R 1 or
R2 , claiming that because of equation (14) the minima would have to correspond to R1 = h/2e2 and
R2 = h/4e2 . But what if R1 or R2 are not integer mulitples of G0 ? That is, what if noise power is
at a minimum and the APC has a conductance other
than G0 ? In that case, R1 and R2 would no longer be
standard resistances. The noise minima are unlikely
to shift if, as is usually assumed for mesoscopic conductors such as an APC, there is only one contributing channel [17], and this is almost certainly the
case for the first quantum of conductace. However,
we have already shown evidence for multiple channels being open simultaneously at the second noise
minimum. Not only could these extra channels lead
to shot noise not being entirely suppressed, but extra open channels could also make the noise minimum differ from where it would be when the sum
in equation (14) is equal to one. The uncertainty in
the location of the second minimum is one possible
source of error in the bridge measurement, which
depends on that second measurement to determine
an unknown resistor. Moreover, due to the geometry of most APC samples used, measurements near
the second quantum of conductance were difficult
to obtain–often the conductance would bounce back
and forth from one conductance to another, giving
inaccurate noise data.
The recalibration method of determining a resistor’s
resistance also has inherent sources of error, but most
probably in the measurement of voltage–the recalibration method only uses the first noise minimum
and is thus free of any uncertainty in the conductance of the second. One source of error in measuring V comes from the voltage source itself. The
voltage applied by it is offset a small fraction, perhaps 0.3 microvolts. Of course, since we are applying a voltage of 0.3 milivolts, this error would be
near one part in a thousand. Larger sources of error,
perhaps as high as one part in ten, exist due to the
fact that we are taking a measurement across lines
that could act as a thermocouple and add voltage
due to the extremely large thermal gradient. Thermocouples are usually made of two different metals
that, when across a thermal gradient, have a potential difference between them. In our experiment, the
stainless steel cryostat is the ground, while our RF
and DC measurements are carried through a copper
BNC wire, both of which go from (4 K) to (300 K).
This could be solved by measuring the voltage across
the APC with two wires of the same material.
Despite these sources of error, we found the following results. Using the bridge measurement technique, we obtained the values 6.25 ± 0.24 kΩ, 18.98
± 0.23 kΩ, 13.59 ±0.20 kΩ, and 55.67 ± 0.48 kΩ
for the four unknown resistors labeled 225, 640, 423,
and 582 respectively. With the recalibration method
the derived resistances for 225, 640, 423, and 582
were 7.11 ± 0.37 kΩ, 21.98 ± 0.75 kΩ, 12.78 ±
0.51 kΩ, and 55.84 ± 2.84 kΩ. After these resistances were found from the noise, we measured the
unknown resistors with a Agilent 34401A 6 1/2
digit multimeter. They were 7.06902 kΩ for 225,
21.6989 kΩ for 640, 12.8957 kΩ for 423, and 55.495 data analysis, equipment setup, problem-solving, and
kΩ for 582. The three data sets, with error bars, are general camaraderie.
plotted in Figure (15).
This REU project was funded by JILA, the National
Institute of Standards and Technology , and the National Science Foundation under NSF award number
0353326.
Figure 15: This figure shows all four test resistors and their
measured values as a black ring. The bridge measurements
(plotted as blue squares) are off by as much as 13 percent,
while the red resistance measurements found via the recalibration method (in red) are off by at most 1.3 percent.
Conclusion
Although there are important sources of error that
could be removed, the general conclusion that can
be drawn from the results of this work is that using
the shot noise of an atomic point contact as a resistance standard is very promising.
Acknowledgements
This paper is the result of a joint REU project with
Brandon M. Smith–almost all of its success must be
credited to his previous year of effort and his dedicated partnership during the summer.
I would like to thank my advisor Konrad Lehnert for
such an interesting project to spend a summer on
as well as for many insightful conversations along
the way. I also must thank Nathan Flowers-Jacobs,
Cindy Regal, and Manuel Castellanos-Beltran–the
other members of the Lehnert lab–for great help with
References
[1] B. Jeckelmann, B. Jeanneret. The Quantum [12] E. S. Snow, D. Park, and P. M. Campbell.
Single-atom Point Contact Devices Fabricated
Hall Effect as an Electrical Resistance Stanwith an Atomic Force Microscope. Applied
dard. Measurement Science and Technology
Physics Letters 69 2, 269-271 (July 1996).
14, 1229-1236.
[2] M.B. Priestley. Spectral Analysis and Time [13] L. Olesen, E. Laegsgaard, et alii. Quantized
conductance in an atom-sized point contact.
Series. London, New York: Academic Press,
Physical Review Letters 72, 2251 (December
1982
1993).
[3] J. Johnson. Thermal Agitation of Electricity in
[14] A. I. Yanson. Atomic Chains and Electronic
Conductors. Physical Review 32, 97 (1928).
Shells: Quantum Mechanisms for the Forma[4] H. Nyquist. Thermal Agitation of Electric
tion of Nanowires. PhD thesis, Universiteit
Charge in Conductors. Physical Review 32,
Leiden, February 2001.
110 (1928).
[15] W.P. Mason. Piezoelectric Crystals and their
[5] C.W.J. Beenakker, C. Schonenberger. QuanApplication to Ultrasonics. Van Nostrand,
tum Shot Noise. Physics Today, 37 (May
1950.
2003).
[16] L. Spietz. The Shot Noise Thermometer. PhD
[6] M. Reznikov, M. Heiblum, Hadas Shtrikman,
thesis, Yale University, May 2006.
and D. Mahalu. Temporal Correlation of Electrons: Suppression of Shot Noise in a Ballistic [17] R. Landauer. Solid-State Shot Noise. Physical
Review B 47, 16427 (1993).
Quantum Point Contact. Physical Review Letters 75, 3340 (1995).
[18] B.J. van Wees et al.. Physical Review Letters
60, 848 (1988)
[7] M. Buttiker, Phys. Rev. Lett. 65, 2901 (1990).
[19] L. Spietz, K. Lehnert, I. Siddiqi and R.J.
Schoelkopf. Primary Electronic Thermometry
[9] V. F. Gantmakher, Rep. Prog. Phys. 37, 317
Using the Shot Noise of a Tunnel Junction.
(1974)
Science 300, 1929 (June 2003).
[8] W. A. Little, Can. J. Phys. 37, 334 (1959)
[10] J. Bert. ”The Noise is the Signal” Final Pa- [20] H.E. van den Brom, J.M. van Ruitenbeek.
per: REU Summer 2004. University of ColQuantum Suppression of Shot Noise in Atomorado JILA, August 2004.
Size Metallic Contacts. Physical Review Letters bf 82, 1526, (February 1999).
[11] J. Moreland and J. W. Ekin. Electron Tunneling Experiments Using Nb-Sn ”Break” Junc- [21] D.M. Pozar. Microwave Engineering. John
tions. Journal of Applied Physics 58 10, 3888
Wiley and Sons Inc., second edition, 1998.
(November 1985).
13
Fly UP