Noise Thermometry Using Josephson Junctions

Abstract

As can be seen from the defining equation,

$$ \left. {\left\langle {V_n^2 (t)} \right.} \right\rangle = \int {4RkTd\nu } $$

(1)

thermal noise voltages, V_n(t), are so small that amplification must precede further electronic manipulation of V_n(t) (i.e., squaring and averaging) before a noise temperature can be measured. Unfortunately, room-temperature amplifiers contribute their own noise to the measured output of the complete circuit, thus rendering noise temperature measurements meaningless for temperatures where $\left. {\left\langle {V_n^2 (t)} \right.} \right\rangle$ is less than the amplifier noise. Much of the amplifier noise can be eliminated, however, by using a more sophisticated technique consisting of cross-correlating the outputs of two identical amplifiers connected to the same resistor [1]. In this way temperatures as low as 2 K have recently been measured by KLEIN, KLEMPT and STORM even though the noise of each amplifier was equivalent to a much higher noise temperature [2]. This cross-correlation technique does not completely eliminate amplifier noise, however, so that even it ceases to be useful at temperatures equivalent to this residual noise. It has been very fortunate, therefore, that low-noise cryogenic amplifiers based on the Josephson effect have been available to extend noise thermometry far into the extreme cryogenic limit beyond the reach of the conventional techniques mentioned above. In fact, two distinct types of Josephson junction devices (often referred to generically as SQUIDs for Superconducting Quantum Interference Devices) have proven successful in measuring temperatures as low as a few millikelvins with amplifier noise equivalent to a noise temperature of 0.1 mK. This article will discuss both types, compare their performance and their most appropriate regimes of application, and the temperature scales developed with them.