Sei sulla pagina 1di 5

Noise Thermometers

From a theoretical point of view, this type of thermometer is particularly


attractive, since only fundamental thermodynamic relations are necessary
to calculate the magnitude of the thermodynamic fluctuations. Noise
thermometers are based on the statistical thermal motion (Brownian
motion) of the conduction electrons in resistors. Independently, in 1927
Johnson and Nyquist investigated experimentally and theoretically the
. [voltage fluctuations across a resistor [4], [5

According to the Nyquist theorem, the mean square of the voltage


:fluctuations is given by

VN2= 4kBTR Δf

Here, R denotes the resistance and Δf the bandwidth of the measurement.


This expression is an approximation and is only valid if the upper limit of
the bandwidth in the measurement fulfills the condition Δfmax << kBT/h. In
practice, this is not a serious limitation, since in most cases the
bandwidths that can be realized are much smaller. For example, at 1K one
.finds kBT/h ≈ 20 GHz

Figure 12.3 shows the mean square voltage for different conductors as a
function of the resistance. As expected from (12.4) a linear relation
between the resistance and the mean square voltage is found,
.independent of the type of the conductor

To determine the temperature, the quantities VN2, Δf, and R must be


known [2],[3]. The measurement of each of these quantities has its specific
demands and attendant problems. For example, the voltage fluctuations
VNare extremely small and must be amplified. In doing this, the amplifier
noise is superimposed on the thermodynamic fluctuations. In addition, the
amplification factor must be known precisely. A further problem is the
accurate determination of the bandwidth, because it strongly depends on
the specific filter characteristic of the setup. Therefore, absolute
1

measurements have a typical error margin of a few per cent. Because of


this, a noise thermometer is usually calibrated against a second
thermometer, at least at one temperature, to obtain the amplification
.factor and the bandwidth
With the use of modern analogue-to-digital converters and FFT this problem has become 1
much less significant

To get a feeling for the difficulty of measuring voltage fluctuations we


consider a specific example: Assuming a resistance of 10 kΩ, a bandwidth
of Δf = 105 Hz, and a temperature of 1 K, the voltage fluctuations are of
the order of VN21/2≈ 2 × 10−7 V. The corresponding power is only P ≈ 10−18
W. Precise measurements of these very small signals require extremely
good amplifiers. In many cases, SQUIDs are used for noise thermometry,
since conventional semiconductor amplifiers do not fulfill the required
.criteria

There are several types of noise thermometers based on SQUID amplifiers.


One method is to connect the resistor in parallel with a Josephson contact,
that serves as a precise voltage–frequency converter making use of the ac
Josephson effect (see, for example, [6]). In this case, the transfer function
contains only fundamental constants. A voltage V across the Josephson
contact results in a frequency of fJ = (2e/h)V . Therefore, for the
determination
of the voltage fluctuations, only a frequency counter is necessary.
However, the amplitude of the oscillations is extremely small and special
techniques are required for the measurement, which we will not discuss in
detail here. An advantage of this method is that the result is independent
.of the amplification factor

In another technique, the voltage fluctuations across the resistor are


inductively coupled into the SQUID with the help of a coil, i.e., the SQUID
acts as a low-noise preamplifier of the current caused by the voltage
fluctuations (see, for example, [7]). Analogous to (12.4), the current
fluctuations can be expressed by

with τ = Ltot/R being the time constant. Here, Ltot denotes the total
inductance of the input circuit. Modern low-noise dc SQUIDs thermally
anchored at a fixed temperature of 4.2K or 1.2K are well suited for this
application. A sketch of a recent realization of such a current-sensing
.noise thermometer is shown in Fig. 12.4a
Fig. 12.4. (a)Schematic diagram of a current-sensing noise thermometer
(after [7]). A superconducting aluminum fixed-point thermometer is
incorporated for a one-point calibration. (b) Temperature derived from the
noise of a 29-mΩ resistor plotted versus the temperature measured with a
3He melting-curve thermometer (open circles) and with a resistive
[
thermometer (closed circles) [7
At very low temperatures, a potential problem arises from the occurrence
of hot electrons because of the weak electron–phonon coupling. To avoid
this phenomenon, the noise resistor is grounded at the base of the copper
holder. In this way, the electrons in the noise resistor are in direct thermal
contact with the free electrons of the copper support. The temperatures
derived from the noise of a 29-mΩ resistor are plotted in Fig. 12.4b versus
the temperature measured by other thermometers. Obviously, very good
agreement is found in the whole temperature range between 24mK and
.4.2 K

Very recently, a novel noncontact noise thermometer has been developed


[8]
. In this thermometer, the magnetic Johnson noise from a solid gold post
was measured inductively using a low-noise dc SQUID operated at 1.2K.
The main advantage of this technique is that no leads are in contact with
the conductor that is used for noise thermometry. A schematic of the
setup is shown in Fig. 12.5a. The pickup loop was arranged as a first-order
gradiometer to suppress possible fluctuations of external magnetic fields
.that couple simultaneously into both loops
Fig. 12.5. (a) Sketch of an inductively coupled noise thermometer. (b)
Spectral density of the flux noise caused by the magnetic Johnson noise of
a gold post plotted against the temperatures of a superconducting fixed-
[
point device SDR1000 [8

At sufficiently low frequencies ω << ωro, the power density of the flux
noise through the cross section of a cylindrical sample with radius and
[
conductivity σ is given by [8

S∅=4kBTμ02 σGr3

where G is a geometry-dependent numerical constant, which is of the


order of 0.1 for a single turn pickup loop tightly wound around the
conductor. Depending on the geometry of the setup, the so-called roll-off
frequency ωro can be determined either by the radius of the sample and its
conductivity via the skin depth, or by the total inductance of the
superconducting flux transformer, which forms an RL low-pass filter with
the real part of the complex impedance of the pickup coil. As shown in Fig.
12.5b, successful operation of this device was demonstrated by
investigating the flux noise at temperatures predestined by the calibrated
superconducting fixed-point thermometer SDR1000 . As expected, the
.power density of the flux noise varies proportional to temperature

: References
C.Enns, S.Hunklinger, Low-Temperature Physics, 2005, Springer [1]
Berlin Heidelberg New York, ISBN-10 3-540-23164-1

R.P. Hudson, H. Marshak, R.J. Soulen Jr., D.B. Utton, J. Low Temp. [2]
(Phys. 20, 1 (1975

G. Schuster, D. Hechtfischer, B. Fellmuth, Rep. Prog. Phys. 57, [3]


(187 (1994
J.B. Johnson, Nature 119, 50 (1927); Phys. Rev. 29, 367 (1927) [4]
(and 32, 97 (1928

(H. Nyquist, Phys. Rev. 29 614 (1927) and 32, 110 (1928 [5]

S. Menkel, D. Drung, Y.S. Greensberg, T. Schurig, J. Low Temp. [6]


(Phys. 120, 382 (2000

C.P. Lusher, J. Li, V.A. Maidanov, M.E. Digby, H. Dyball, A. Casey, [7]
J. Ny’e ki, V.V. Dmitriev, B.P. Cowan, J. Saunders, Meas. Sci. Technol.
(12, 1 (2001

A. Netsch, E. Hassinger, A. Fleischmann, C. Enss, to be published [8]

J.B. Johnson, Nature 119, 50 (1927); Phys. Rev. 29, 367 [623]
((1927) and 32, 97 (1928

Potrebbero piacerti anche