Macroscopic, Sub-Kelvin Refrigeration from Electron Tunneling
Authors: Peter Lowell, Galen O’Neil, Jason Underwood, Joel Ullom
National Institute of Standards and Technology, Boulder, Colorado, USA
Modern science often requires sub-Kelvin temperatures in order to observe quantum mechanical effects. Examples include quantum information studies, quantum computing and quantum sensors that can be used for measurements of radiation from microwave to gamma-ray wavelengths with unprecedented sensitivity. Cryogenic temperatures of 4K are easily reached using commercial mechanical refrigerators. The simplest method of reaching lower temperatures is by vacuum pumping on cryogenic liquids such as liquid 3He, which can reach as low as about 300mK. Achieving yet lower temperature is challenging and requires large, complex and expensive refrigerators. Our goal is to improve the availability of sub-300mK temperatures by creating simpler refrigeration techniques.
Normal metal–insulator–superconductor (NIS) tunnel junctions can be utilized for a new type of refrigeration for cooling below 300mK using quantum mechanical effects. When a voltage bias is applied to a NIS junction, it cools the electrons in a normal metal because only the hottest electrons from the normal metal can tunnel into the superconductor, which causes the electron gas remaining in the normal metal to cool down. The selective tunneling is made possible by the energy gap in the superconductor. The effect is analogous to blowing on a cup of coffee; the stream of air removes the hottest particles, which causes the coffee to cool down. The cooling power of a NIS junction is controlled by the voltage bias, so any temperature between the launch and base temperature is accessible by changing the voltage bias. Also, since individual NIS devices provide the cooling power, simply adding more junctions can increase the cooling power.
As a result of electron-phonon decoupling at sub-Kelvin temperatures, NIS refrigerators ordinarily cool the electrons, while the atomic lattice remains at roughly the same temperature. To cool the atomic lattice, we extended the normal metal containing the cold electrons onto a thin, thermally isolating membrane. Such a structure allows the NIS refrigerator to cool down the atomic lattice along with the electrons. The NIS junctions will then cool anything connected to the membrane. The downside to this technique is that the membrane is small and fragile, which makes it difficult to connect to other objects.
The ability of NIS junctions to behave as refrigerator is well known  but they have never been able to be used as a general refrigerator since scientists could not attach arbitrary payloads. The novel process, as described in Applied Physics Letters , was to make NIS refrigerators behave more like a general-purpose refrigerator, where the user can cool any arbitrary object, much like one can cool an arbitrary object inside a kitchen refrigerator. This was difficult because of the fragility of the membranes and the low cooling power of a single NIS device. To build a general-purpose refrigerator, we suspended a copper block with thin Kevlar cords to minimize stray power loads. This copper block was connected to the cooled membrane of the NIS refrigerator by tiny gold wires, which required extremely delicate microassembly to avoid breaking the fragile membrane.
Figure 1: Photograph of NIST's prototype solid-state refrigerator uses quantum physics in the square chip mounted on the green circuit board to cool the much larger copper platform (in the middle of the photo) below standard cryogenic temperatures. Other objects can also be attached to the platform for cooling.
Measurements of our refrigerator (see Figure 1) showed that over the course of about 18 hours, the copper block was cooled from 290mK to 256mK with about 700pW of cooling power at 290mK. This is the same fractional temperature reduction that is achieved by a kitchen refrigerator. One of the current refrigeration technologies used to reach sub-300mK temperatures, the dilution refrigerator, resembles a kitchen refrigerator in that it relies on compressors and pumps, which makes it complicated to use and prone to mechanical failure. In comparison, our NIS refrigerator is powered by quantum mechanics and has no moving parts. It only requires a small voltage bias to operate and can be powered off of a small battery. The simplicity of operation and relatively small size, several inches on a side, provide advantages over dilution and adiabatic demagnetization refrigerators.
Although we have demonstrated cooling a block of copper about one million times larger than the refrigerators themselves, these refrigerators aren’t quite ready for commercialization. We will improve the NIS refrigerators so they can cool from 300mK to 100mK by using more devices, and making more aggressive design choices. We can further expand the temperature range of our refrigerator using similar devices made of different materials. For example, electron coolers have demonstrated cooling from 1K to 400mK  and below 100mK  based on the same quantum mechanical principles. We can apply the same techniques to build general-purpose refrigerators based on these electron coolers. Eventually, our goal is to develop a composite, multi-stage cooler that can cool from 1K to below 100mK.
This work is supported by NASA.
 Juha T Muhonen, Matthias Meschke and Jukka P Pekola, "Micrometre-scale refrigerators". Reports on Progress in Physics, 75, 046501 (2012). Abstract.
 Peter J. Lowell, Galen C. O'Neil, Jason M. Underwood, and Joel N. Ullom, "Macroscale refrigeration by nanoscale electron transport". Applied Physics Letters, 102, 082601 (2013). Abstract.
 O. Quaranta, P. Spathis, F. Beltram, and F. Giazotto, "Cooling electrons from 1 to 0.4 K with V-based nanorefrigerators", Applied Physics Letters, 98, 032501 (2011). Abstract.
 Galen C. O'Neil, Peter J. Lowell, Jason M. Underwood, and Joel N. Ullom, "Measurement and modeling of a large-area normal-metal/insulator/superconductor refrigerator with improved cooling", Physical Review B, 85, 134504 (2012). Abstract.