Siegel der Universität

Universität zu Köln
Mathematisch-Naturwissenschaftliche Fakultät
Fachgruppe Physik

I. Physikalisches Institut

Hot Electron Bolometer (HEB) Mixers

Printer-friendly versionPDF version
Fig. (1) Electron micrograph of a 1.9 THz waveguide HEB device. From left to right: RF waveguide probe antenna on 2 µm Si membrane, opening is for waveguide, cpw transmission line, HEB microbridge in narrow gap near the image center, tuning capacitor, outside shows beamleads that suspend the device in the waveguide housing and also serve as dc and rf ground contacts.


In the early 90s a new heterodyne mixer concept was introduced based on superconductive Hot Electron Bolometer (HEB) mixers (e.g. for a comprehensive overview see this review paper by Shurakov et al., Supercond. Sci. Technol. 29 (2016) 023001). The HEB mixers rely on the non-linear temperature dependence of the resistance of a superconducting microbridge near its superconducting transition temperature Tc. Bolometer based mixers do not follow the signal and local oscillator (LO) electrical field instantaneously (as is the case for SIS mixers), but rather react to the square of the sum of the fields. An intermediate frequency (IF) signal (3 magnitudes lower in frequency for THz receivers) is thus generated by the beat frequency between signal and LO.
The device essentially is a superconducting microbridge with typical dimensions of 3 µm width, 200 nm length and only 4 nm thickness typically made of the superconductor material niobium nitride (NbN). To make the bolometric response fast (for high enough receiver IF of a few GHz), only the electron bath must be heated above the lattice temperature creating the "Hot Electrons". This occurs if the strip is very thin (3-4 nm). The bolometer material therefore must have a fast electron-phonon interaction to enable cooling of the electron via the lattice phonons. For the most common phonon-cooled HEB device a fast phonon escape into the substrate, made possible by a good (lattice) match between microbridge and substrate, must be ensured in order to cool the microbridge on a short enough time constant and thus high enough IF for the receiver application.
HEB mixers are presently the lowest noise (most sensitive) mixers at frequencies > 1 THz where SIS mixers cease to work due to their operation limit determined by the superconducting gap. Semiconductor Schottky diode mixers are still significantly more noisy. For our THz astronomy heterodyne receiver applications within the research framwork of our institute we therefore choose HEB mixers. 

Hot Electron Bolometer Development 

We have researched HEB mixer development for application in radioastronomical Terahertz receivers since the mid 90s, starting with the diffusion-cooled sister (cooling of the phonons via the normal conducting contacts) to aforementioned principle (Fiegle et. al, IEEE Trans. Appl. Supercon. vol. 7, no. 2, 1997). From the beginning we choose to develop mixer hardware based on waveguide feedhorn radiation coupling rather than the more common lens based systems. In our reasoning this ensures a better defined beam pattern, which is crucial for efficient radiation coupling of the mixer through a complicated telescope system at an observatory. A well-definded input beam into the mixer translates to well-defined beam on the sky.
Our main observatory platform for development is the modular GREAT (German REceiver for Astronomy at Terahertz frequencies) heterodyne instrument for SOFIA (Stratospheric Observatory for Infrared Astronomy) observatory, with the Cologne groups (instrumentation, software and astrophysics) having a very strong contribution to the German consortium that develops and operates the instrument (Figs. (2) and (3)). We have developed and sucessfully operated all mixers in the various channels of the receivers, i.e. 1.4 THz, 1.9 THz, 2.5 THz and 4.7 THz, with first light on SOFIA being in 2011.
We have achieved equivalent receiver noise temperatures of 500 K at 1.5 THz, 900 K at 1.9 THz (GREAT single-pixel receivers L1 and L2 on SOFIA), 900 K at 2.7 THz, and in 2014 an excellent 800 K at 4.7 THz with the GREAT H receiver, see below for more details.

HEB mixers for THz heterodyne focal plane arrays

Focus of our HEB mixer development in the last few years lies on the two "up"GREAT focal plane array extensions to the GREAT platform on SOFIA (Figs. 2 and 3), which will be described seperately below. In order to improve the overall efficiency of an astronomical heterodyne receiver besides, naturally, using the most sensitive mixers available, the way to go is to incorporate more pixels into the receiver front end. Unlike everyday used commercially developed technology such as digital cameras with millions of CMOS or CCD type detectors, the heterodyne receivers in radioastronomy use very few pixels due to that fact the superconducting detector technology needed to mature.
At Terahertz frequencies only in the last 10 years there has beein significant momentum to implement focal plane array type receivers, i.e. receivers with several spatial pixel. It is important to understand even with only one or a few spatial pixels a lot of science can be done, as the heterodyne system multiplexes each spatial pixel into many thousand spectroscopic frequency "pixels" (i.e. channels) covering the receiver IF range. After the pioneering observational efforts with the Herschel Space Telecsope and also SOFIA astronomers worldwide are seeking to gain a better larger scale overview of the astronomical objects of interest, e.g. large star formation regions. For this instruments with efficient mapping capabilties are needed, which motivates our "up"GREAT receiver extension.
Fig. (2)  The GREAT modular receiver platform installed on the instrument flange of the telescope inside the SOFIA aircraft. Fig. (3)  SOFIA observatory in flight with door to telescope opened in the rear of the aircraft (© NASA).

1.9 THz waveguide HEB mixers for the upGREAT Low Frequency Array

Fig. (4)  Seven fully assembled and qualified mixer units prior to delivery for receiver integration at the MPIfR. These 1.9 - 2.5 THz mixers populate one of two polarisation diplexed sub-array in the upGREAT Low Frequeny Array receiver. Fig. (5)  Trec(IF) lab test data of the seven mixers delivered for one of the sub arrays of the upGREAT LFA receiver. The Trec data is corrected for beamsplitter transmission because various coupling factors were used.

Development of upGREAT LFA mixers

For the 1.8 - 2.5 THz bandwidth specification of the 14-pixel low frequency array (LFA) we have developed a HEB mixer that ensures compatibility to the array application (Fig. 1 shows an close-up of the front RF structures of such a device), mainly a low and reproducible local oscillator pump power requirement. Uniformity of device characteristics is the most demanding criteria for array mixer fabrication. Luckily our device fabrication is able to deliver very uniform devices (Figs. 5-7). We delivered all 14 mixers plus replacements in 2015 when the LFA receiver went into commissioning. The commissioning of the receiver was highly successful and since early 2016 the LFA is fully commissioned and in regular science operation, e.g. during the SOFIA flights from New Zealand for observations of southern hemisphere objects. The upGREAT LFA is the first functional array receiver at 1.9 THz and the gain in mapping speed experienced during commissioning fully confirms the necessity for focal plane arrays.
Fig. (6)  DC IV curves of one batch of HEB devices fabricated for the upGREAT V array. Fig. (7)  DC RT curves of one batch of HEB devices fabricated for the upGREAT V array.

4.7 THz waveguide HEB mixer for the upGREAT High Frequency Array

Fig. (8)  Optical microscope image of the central part of the 4.7 THz device. The waveguide opening (dark area) measures 24 µm x 48 µm. The Silicon substrate of the device (greyish vertical strip) is only 2 µm thick. Fig. (9)  Measured uncorrected double sideband receiver noise temperatures (Trec,DSB) vs. intermediate frequency of 4.7 THz HEB mixer.

HEB mixer for observation of the [OI] fine structure transition at 4745 GHz

Besides the [CII] fine structure transition of the ionized carbon atom the [OI] fine structure transition, the (3P1-3P2) transition of atomic oxygen at the frequency of 4.74477749(16) THz (63 µm, ref. CDMS), of  atomic oxygen is another of the main "cooling lines" of the ISM. The designation cooling line describes that this one transition by itself contributes significantly to the overall radiation budget of e.g. star formation regions in the ISM. Thus a lot of information is contained in observations of this line. But the [OI] transition has hardly been observed with sufficient spectroscopic resolution due to the fact that this atomic line can only be detected from an airborne or space observatory due to the water vapor absorption in the atmosphere. The only high resolution spectra obtained were with the Kuiper Airborne Observatory (KAO), the predecessor of SOFIA, with a Schottky diode mixer that was very insensitive at these frequencies (Boreiko and Betz, ApJ, 464, pp. L83–L86, 1996). With the development of much more sensitive superconducting HEB mixers, shown in the lab to work with much higher sensitivites, it was obvious to add such a receiver to the GREAT system. 

Prototype mixer for GREAT H channel

In December 2013 the development of our first 4.7 THz waveguide superconducting hot-electron bolometer (HEB) mixer was completed and went into lab characterisation. The 5.5 nm thickness x 300 nm length x 3600 nm width NbN nanobridge is integrated into a Au normal-metal planar circuit on a 2 µm thin silicon substrate. As with all other mixers the NbN thin film, the device circuitry and the mixer block were designed, developed and fabricated in house in our microfabrication lab and in our fine mechanics workshop. We use a CuTe alloy as a material for blocks because it has a higher hardness than pure Cu to avoid burrs which are a critical issue given the small dimensions of the structures. The 24 µm x 48 µm (22 µm depth) waveguide and the 16 µm x 105 µm (4 µm depth) substrate channel are stamped into the block with the required precision of +/- 1 µm. This was made only possible with new Kern Nano™ submicrometer precise CNC that become available in our shop.
The device is adjusted on the waveguide with a hexapod-nanomanipulator and is ultrasonically bonded to the block (see Fig. 8). The measured mixer direct detection response is in a good agreement with the 3D EM circuit simulation over the design RF bandwidth. In cooperation with the Max-Planck-Institut für Radioastronomie (MPIfR) and the Deutsches Zentrum für Luft- und Raumfahrt (DLR) in Berlin, we measured our mixer in heterodyne detection in the GREAT test setup at the MPIfR with a quantum cascade laser (as local oscillator) from DLR Berlin. The measured noise temperature vs. the intermediate frequency is shown in Fig. 9, which is one of the best results worldwide. The 3 dB noise roll-off of the IF is about 3.3 GHz.
In May 2014, this mixer was flown on the "first light" mission for the GREAT "H" High-frequency-channel on SOFIA, the Stratospheric Observatory for Infrared Astronomy. The flight campaign was a huge success, yielding more high-resolution [OI] spectra than were ever taken before. The sensitivity of the GREAT H receiver is magnitudes higher than during the pioneering efforts of Boreika&Betz. Below is an example of the data taken during this early campaign (Figs. 10 and 11).
Fig. (10) Colour coded intensity map of the planetary nebula NGC 7027, a final stage of stellar evolution, in the [OI] line at 63 µm, taken on SOFIA in May 2014 with our 4.7 THz mixer. The grey circle represents the SOFIA telescope beam diameter at 63 µm wavelength. Fig. (11) A single spectrum of the map of NGC 7027 showing the complex velocity structure of the planetary nebula that is only visible in high-resolution heterodyne spectroscopy. The integration time was only 2 minutes! The excellent signal-to-noise ratio confirms the very high sensitivity of the HEB mixer. The narrow dip is the [OI] line of the earth's upper atmosphere seen in absorption from the flight level at about 43000 feet. The expansion velocity of +- 25 km/s corresponds to 90000 km/h. TNT detonates with 9.6 km/s.

Development of upGREAT HFA mixers

Moving on from the hugely successful GREAT H receiver we then were tasked with the development of mixer hardware for the 7 pixel upGREAT high frequency array (HFA), which we delivered over the summer 2016 including some replacements. The HFA started its commissioning flights in October 2016 with excellent science return.

Current HEB mixer research

RF circuits and integration

We are currently developing a 1.9 THz balanced HEB mixer, which utilizes a one chip planar circuit for the whole RF part including the microbridges. The balanced circuit provides a second input port to the mixer, supplying separate ports for LO and signal, which we expect to simplify the optics of  focal plane array receivers. The possible suppression  the LO-noise in the mixer will be investigated.

Development of 11 THz HEB mixer

The para-H2 molecule has a lowest energy transition J=2→0 at 10.623 THz (28.22 µm) that can be directly observed. Due to the physical conditions being very cold narrow line absorption is expected, for which a heterodyne system is most sensitive. We plan to develop a HEB mixer in this frequency range. Due to the approx. factor 2 higher frequency as compared to our current achievements at 4.7 THz, which was on the limit that current technology and tolerances would allow, we will need to implement new technologies for the mixer.


There is considerably room for improvement in the performance characteristics of our HEB mixers. The mixer noise performance is still 5 to 10 times the quantum limit, and the IF bandwidth is only 4 GHz.
Improvements are thought to come through the development of new materials both for the HEB layer (on a new substrate as well), or as buffer layer between HEB and substrate.
Worldwide there is currently research in:
1. Newly available superconducting materials, in particular MgB2, with signifincantly higher Tc than NbN, >30 K vs. 10 K, have been developed by other groups for processing of thin HEB films. The higher Tc helps twofold: Firstly, the electron-phonon interaction time taue-ph scales ~ Tc-3. Secondly, the phonon escape time inversely scales with the thickness of the microbridge layer tauph-esc ~ d, and utilizing a microbridge material that stays superconducting for for very thin films (nanometers) is critical.
2. Improving the phonon acoustic mismatch between the microbridge and substrate, which arises from the the materials dependent phonon velocities, by either selection a more suitable materials combination and/or introducing a buffer layer at the interface with certain matching characteristics. Currently GaN is being explored by other groups and shows promising results.

Device Physics

A clear physical understanding of the origin(s) of the mixer noise in the driven HEB microbridge is still not on hand. We need to base our current development on various empirical or semi-empirical approaches. Therefore there is a strong research opportunity is on hand, which, ultimately, could lead to further improvement of device performance.