The upGREAT Dual Frequency Heterodyne Arrays for SOFIA

1. Introduction — Scientific Motivation

Since the turn-off of the HIFI/Herschel instrument (de Graauw et al., 2010; Pilbratt et al., 2010), there has been very limited number of high-resolution spectroscopy instruments capable of performing far-infrared wavelengths observations. For example, the balloon-borne Stratospheric Terahertz Observatory (STO-2) performed high-resolution observations in the THz range in December 2016 for about $\sim$500h. On the other hand, the spectrometer GREAT (Heyminck et al., 2012) aboard the SOFIA airborne observatory (Young et al., 2012) is the only facility that provides regular high-resolution spectroscopy at far-infrared wavelengths.

GREAT saw first light on April 1st 2011 during SOFIA’s early science phase, and since then has been operated on more than 135 science flights, collecting unique science data for more than 1000h. Starting as a dual-color single-pixel receiver operating in parallel in the 1.5 and 1.9THz frequency bands, a major upgrade occurred in 2015 when the low-frequency array receiver upGREAT/LFA was installed, spatially multiplexing now with $2\times 7$ pixels in the 1.9THz band (Risacher et al., 2016a). A year later, in May 2016, the second frequency band, the high-frequency array upGREAT/HFA operating at 4.7THz was successfully commissioned. As the SOFIA project had only one flight compressor in 2015/2016, only one of the arrays could be operated at a given time. In 2017, the phase 2 SOFIA infrastructure was completed, for the first time allowing two compressors, and therefore two closed cycle systems to operate independently and simultaneously. upGREAT in its full dual-array configuration was then commissioned in May/June 2017 and has been operated since during 25 science flights. This paper describes the development of the instrument and its in-flight performance.

2. SOFIA Observations — Atmospheric Conditions

Observations in the THz range are almost impossible to achieve from ground, therefore only airborne, balloon borne or space telescopes can perform such observations. For SOFIA, flying regularly at altitudes above 41,000 feet, a large part of this spectrum becomes accessible with high atmospheric transmission. Even though it flies above 99% of the Earth’s atmosphere, it still suffers from absorption from several atmospheric molecular species, e.g. H₂O, O₃. Figures 1 and 2 show the typical atmospheric transmission for the two upGREAT bands, for 1.8–2.07THz and around 4.745THz. Some of the main lines of interest are shown overlaid. Figure 1 shows that a large part of the band is not accessible, and even in the regions having good atmospheric transmission, not all frequencies are tunable with the current local oscillators (LOs) (Sec. 4.2). For the higher frequencies, as can be seen in Fig. 2, the [OI] observations are very challenging to perform, even from SOFIA, and require precipitable water vapors (PWV) well below 10$\mu$m, in order not to be too degraded by the atmospheric losses. Such PWV values are difficult to achieve when SOFIA is stationed in Palmdale, CA, where it normally operates. However, SOFIA is deployed yearly to New Zealand and is stationed in Christchurch during the Southern winter. There, the atmospheric conditions are usually of the order of a few microns of PWV, hence allowing more efficient [OI] observations.

3. Instrument General Description

A full detailed description of the receivers hardware can be found in Risacher et al. (2016a) and the first commissioning results of the low frequency upGREAT receiver are presented in Risacher et al. (2016b). We present here the instrument performance as of 2018, highlighting the performance of the high frequency array for 4.7THz operation, and describing the improvements and experience gained since the installation of those systems. Since May 2017, the two upGREAT arrays are operating in parallel, with 21 pixels. The main instrument characteristics are summarized in Table 1.

The main components of the upGREAT LFA/HFA configuration, are shown in Fig. 3. The astronomical signal is directed through the SOFIA telescope main optics and the first optical component from the GREAT system is the derotator optics, a “K-mirror” rotating set of mirrors, which allows to compensate for the sky rotation in the instrument focal plane (described in detail in Risacher et al. (2016b)). The frequency selection is done via a dichroic filter from QMC Instruments Ltd., which separates in transmission the low frequencies below 3THz (to the upGREAT LFA system), and reflects the 4.7THz signal (to the HFA receiver) with less than 7% and 2% losses, respectively. This component allows the simultaneous operation of both upGREAT arrays. Then the LO references are combined with the astronomical signal, via dedicated pre-adjusted optics, which also reimage as needed the optical parameters via pairs of Gaussian telescope optics (details in Risacher et al. (2016a) and in following sections). The LO references are located inside elongated housings mounted under the Science Instrument (SI). For the upGREAT/LFA system, the LO systems are based on multiplier chains from VDI Inc. and for the HFA system, the LO system is using quantum-cascade devices (detailed in Secs. 5.2 and 5.3). Details on the cryostat, mixer design, IF system and backends can be found in Risacher et al. (2016a). The superconducting Hot Electron Bolometers (HEB) mixers are developed and fabricated by the I. Physikalisches Institut der Universität zu Köln (Pütz et al., 2012; Büchel et al., 2015).

4. LFA Array

4.1. Initial design and optics modifications

Compared to the original design for the upGREAT/LFA array, several things were modified since the first installation onboard SOFIA (Risacher et al., 2016a). The optics plate and the LO optics were redesigned. The first implementation of the LO optics used two separate LO chains, having output signals whose polarization is perpendicular to each other and combined through a polarizer grid, in order to propagate both LO signals on a nominal common axis. Then, on the main optics plate, another polarizer grid would divide them again while being combined with the corresponding RF signals for each polarization. When using one LO chain at a time, illuminating its corresponding subarray worked fine, but when operated in parallel, severe interferences between the two LO systems occurred. The solution was to decouple completely the two LO paths, keeping one on-axis (for the so-called $H$-polarization subarray), while the other LO had a tilted axis. Therefore, the whole LO and LFA optics plate was redesigned to accommodate this change, which proved to be successful as the strong interferences disappeared. Figure 4 shows the updated diagram of the LO chains.

4.2. LO limitations and LO improvements

The upGREAT/LFA receiver is designed to cover the 1.8–2.5THz range. As such, the HEB mixers and the optics are all designed to cover that frequency range. Currently, the upGREAT/LFA system is offered to the community in the 1.83–2.006THz range. This frequency limitation comes from the LO technology. The VDI LO multiplier sources allow to cover a moderate region around 1.9THz with very high output power. This development proved to be very challenging, and only the range 1.83–2.006THz could be covered with several LO AMC chains, and only one VDI system was capable of extending as far as 2.07THz. This is especially important, as there are several astronomical lines of interest in that upper frequency range. Several attempts to replicate this performance with other AMC chains have not been successful until now. Therefore, observations in the 2.006–2.07THz range can only be offered through consortium flights, shared-risks, until more funding is allocated for more robust, additional LO sources.

Another major change was the redesign and rearrangement of the LO components inside the LO main enclosure. Indeed, it was found out that the thermal dissipation and air flow was not sufficient to adequately cool the main LO components, especially the power amplifiers from Spacek Labs. Each LO system generates a lot of power, 50W at 26GHz, 250mW at 220GHz, and 20–30$\mu$W at 1.9THz (or 40–60$\mu$W when the last two triplers are cooled down to 100K). They dissipate each 50W, then the Stirling cooler for the triplers also dissipates around 100W, therefore all in all, about 300W are dissipated inside this small LO enclosure. Under normal laboratory conditions, at 20^∘C, the power amplifiers case temperature would be around 35–40^∘C. But once the instrument is installed onboard SOFIA, particularly during summer months in California’s Mojave Desert, it has proven difficult to manage the cabin temperature in the hours leading up to take-off. Cabin temperatures approaching 30^∘C in the aft cabin area can result in component instability and even failure. To improve this situation, components were rearranged, the air flow was improved, power amplifiers are now cooled conductively using heat sinks with cooling pipes together with more powerful fans. These changes resulted in $\sim 10^{\circ }$C lower temperatures in the LO enclosure. Though these design changes certainly improved the thermal overtemperature situation within the LO enclosure, the stability of these components is still susceptible to large cabin air temperature fluctuations in the first few hours of a flight.

There were also relatively strong interferences, spurs, present in many places in the RF band, originating from the LO chains. VDI has been optimizing and improving on the situation since 2015. Several iterations of multiplier doublers have helped to clean up most of the band. A major breakthrough in this regard occurred in 2017, when VDI identified a source of interference causing a forest of narrow spurs evenly spaced around 1GHz IF (at the IF output of the receiver). These narrow spurs had been present from the beginning for all pixels, and were flagged during data reduction. They originated from pickup from the multipliers bias lines of a small part of the input RF signal (via radiation). A redesign of this VDI electronic controller is ongoing, but the short term solution was to include 10MHz low pass filters for every multiplier bias line. This effectively solved the LO-based issues (while observations may still suffer from SOFIA-internal but also externally rooted RFI).

The most recent improvement performed in 2018 was the addition of LO attenuation wire grids, placed at the LO beam outputs (see Fig. 5), in order to independently adjust the LO levels for both sub-arrays. Before that, LO adjustment was performed electronically, or not done at all, therefore occasionally resulting in overpumped mixers, with some degradation in performance.

5. HFA Array

The second upGREAT array for high frequencies (HFA) was completed in mid-2016, and its commissioning started in November 2016. A view of the two closed-cycle receivers (upGREAT LFA+HFA) is shown in Fig. 6.

5.1. Initial design, optics

As described in Risacher et al. (2016a), the cryostat was fabricated at CryoVac GmbH, integrated with a transMIT pulse tube 2-stage cooler (PTD-406C). The cryostat design is identical to the lower frequency array system (LFA), with some differences:

5.2. Mixers fabrication

The HEB device fabrication, the feedhorns manufacturing and the subsequent integration of the mixer block proved to be challenging tasks. The feedhorns used are based on the smooth-wall spline horns (Granet et al., 2004) and are a scaled version of the LFA horns. They were fabricated by electro-forming through Radiometer Physics, and they proved to be very difficult to fabricate with the required accuracy (low yield), mainly due to the difficult removal of the mandrel material inside the small waveguide section. Also in the block manufacturing in the Cologne workshop and especially in the assembly tolerances the approximate factor of 2 reduction in all dimensions compared to the LFA made a notable difference. More devices became damaged upon assembly and all mixers had to be pre-tested as single pixel in Cologne to ascertain their performance as a check on the assembly tolerances. The device fabrication was also not straightforward. As was done for the upGREAT/LFA mixer junctions, the HEB film thickness was decreased in order to reduce the LO power requirement for the array. However, those reduced-thickness devices did not perform as well as the original 4.7THz single pixel HEB mixer that flew on SOFIA (Büchel et al., 2015). Also, it appeared that the extremely thin NbN films suffered more from environmental effects resulting in performance deterioration over several cooling cycles. Finally, as there is sufficient LO power with the QCL technology (see Sec. 5.3), the original HEB device was kept and the upGREAT/HFA devices are based on it, reproducing the previous single pixel performance. A cross-section view of the HFA mixer block and chip assembly can be seen in Fig. 7.

5.3. QCL local oscillators

The LO technology uses Quantum-Cascade Lasers (QCL). Two groups, the Institute of Optical Sensor Systems, German Aerospace Center (DLR) and the Physikalisches Institut der Universität zu Köln, (KOSMA hereafter) have carried out parallel developments, and both have succeeded in providing flightworthy hardware. The DLR QCL has been flying since 2016 with the HFA array, and the KOSMA QCL started flying since May 2018.

5.3.1. DLR QCL LO evolution and performance

The DLR QCL system (see Fig. 8) is a further development of the GREAT single pixel 4.7THz receiver, which had been successfully operated on board of SOFIA from May 2014 to November 2016 (Richter et al., 2015). The QCL LO system consists of a 4.7-THz QCL with a compact, low-input-power Stirling cooler (Ricor K535). The QCL is mounted inside a vacuum housing on the cold finger of the cryocooler, which provides a temperature between 30 and 80 K, necessary for the operation of the QCL. No liquid cryogens are required. The active medium of the QCL is based on a hybrid design and has been developed for continuous-wave operation, high output power, and low electrical pump powers (Schrottke et al., 2013). Efficient carrier injection is achieved by resonant longitudinal optical phonon scattering. This design allows for an operating voltage below 6V. The amount of generated heat complies with the cooling capacity of the Stirling cooler of 7W at 65K with 240W of electrical input power (Richter et al., 2010). The QCL has a lateral distributed feedback grating, which is optimized for the transition frequency of OI at 4.745THz. Due to the DFB grating, the QCL has a single mode emission. The radiation is coupled out through one of the end facets of the single-plasmon waveguide and focused with a parabolic mirror. The beam profile is almost Gaussian with an M² value of approximately 1.2. A flat mirror which is mounted on a goniometer allows aligning of the QCL-beam in order to direct it to a Fourier grating to multiplex the single beam into seven equal beams (see Sec. 5.4). The emission frequency of the LO is determined by measuring the absorption spectrum of CH₃OH and comparing this with data from the literature. The LO covers a frequency range from $-$1.5 to $+$6.5GHz around the rest frequency of [OI]. The frequency is tunable by changing the laser driving current as well as by changing the heat sink temperature. At a constant heat sink temperature the frequency coverage of the LO is up to 4GHz, for example at 52K it ranges from $-$1.1 to $+$2.6GHz. In general, tuning of the frequency by current is preferred, because it is faster than tuning by changing the heat sink temperature, which requires more time, because thermal equilibrium has to be established. The heat sink temperature is kept constant within $\pm 1$mK by a dedicated control loop. The frequency stability is about a few MHz (Richter et al., 2015). The output power varies with the laser driving current and with the heat sink temperature. For all LO frequencies it is above 1.2mW with a maximum power reaching 2.2mW. This includes absorption loss by the window in the vacuum housing of the Stirling cooler, as well as atmospheric absorption loss. Corrected for these losses the output power of the QCL is up to 6mW. The LO system is fully controlled by a dedicated software which controls the Stirling cooler itself, the electrical input power of the QCL, and the heatsink temperature.

5.3.2. KOSMA QCL LO performance

The KOSMA QCL LO (see Figs. 9 and 10) relies on a single mode, double metal QCL with integrated patch antenna array (Bosco et al., 2016; Justen et al., 2016) delivering about 2.4mW optical power in the fundamental Gaussian mode. About 25 to 30 percent of the power is needed to pump the 7-pixel HEB mixer array. The power can be varied by a polarizer grid in the HFA optics compartment. The QCL can be tuned from $-$2.5 to $+$2.9GHz around the OI line at 4744.8GHz by controling the bath temperature to a few mK between 40 and 70K. The QCL is cooled by a Stirling cryocooler. No liquid cryogens are required. The beam position and beam angle are adjusted with two tip-tilt mirrors which couple the beam into the HFA optics compartment. In order to avoid absorption of the LO signal by water in the surrounding air, the beam path is contained in a hermetic housing, which is open to the inside of the instrument optics compartment and therefore experience the same external air pressure (typically about 160mBar) during flight. The LO system is controlled remotely over an Ethernet connection by a dedicated computer which controls the Stirling cooler, vibration control, QCL current source, QCL temperature control and mirror positions. A prototype of this QCL was used in Cologne to pre-test the 4.7THz mixers in a single pixel set-up.

5.3.3. QCL LO frequency stability

Currently, the QCL LOs are free-running and their absolute frequency stability is set by the temperature and current stability of the lasers. For the upGREAT/HFA observations, the [OI] frequency scale is established by referencing to the telluric [OI] that is prominently present in all data, to a precision of better than 0.1km/s, however this requires additional post processing steps in the data reduction. For the DLR QCL, the estimated linewidth is of a few MHz and drifts can be of the same order. When using the KOSMA QCL, measurements of the beat signal between the QCL and a microwave source showed that the QCL frequency jitter is smaller than 1MHz (0.05km/sec). Similarly, the laser frequency drift during observations, as measured by comparison with the telluric OI absorption line is also better than 1MHz. Several options are under study to achieve a frequency/phase stabilization of the QCL, the main obstacle for this being the limited physical space available in the instrument to place additional components.

5.4. LO distribution

For the LO distribution, as in the LFA array, one single LO source is equally multiplexed into seven beams via a phase grating. This collimating Fourier grating (CFG), distributes the LO signal to the array elements (Heyminck & Graf, 2001). The grating structure (shown in Fig. 11) is the same as in the LFA band (Risacher et al., 2016a) or in the CHAMP+ receiver (Kasemann et al., 2008). The nominal efficiency^(a) of the grating is approximately 90%. The RMS power imbalance between the seven beams is ideally less than 0.5% at the designed frequency. The CFG has a diameter of 20mm, a focal length of 135.3mm and operates at a reflection angle of 24^∘. In order to comply with opto-mechanical boundary conditions, the grating structure is designed to produce a diffraction pattern rotated by 9^∘ with respect to the reflection plane.

6. Instrument Installation Onboard SOFIA

This section describes the setup and typical installation procedure of the SI into the aircraft. The various SOFIA SIs are connected to a main reference, the SI flange, which is at the end of the Nasmyth Tube. This is inherited from the Kuiper Airborne Observatory (KAO), SOFIA’s predecessor, with a 36-inch diameter primary mirror on a Lockheed C-141A platform which was operational from 1975–1995. Typically, an instrument is installed aboard SOFIA and stays there for a flight series lasting from one to several weeks. Figure 12 shows the GREAT instrument with the upGREAT arrays on the aircraft just before the final installation and connection to the SI flange.

After the SI is connected mechanically, all of the electrical connections are done and tested. Typically, as the instrument has been undergoing testing in the laboratory facilities before the installation, it is already cold. The helium lines are connected from the instrument to the patch panel, and the compressors are started as soon as possible to resume its cooling. As long as the SI installation can be done in under 2h, this procedure allows adequate time for thermal recovery of the closed-cycle systems, resulting in an operational system within half a day. Figure 13 shows the typical work configuration during ground testing or flights.

6.1. Cryocoolers onboard SOFIA

The cryocooler system is permanently installed as part of the observatory on SOFIA and includes components located primarily on the aft upper deck but with a remote control panel located in the main cabin. The cryocooler system provides closed-loop cryogenic cooling for SIs with the following benefits:

The cryocooler system capable of driving two cold heads, completed Integration and Test (I$\&$T) and Verification and Validation (V$\&$V) in 2017 (overview shown in Fig. 14, photograph in Fig. 15). It consists of two ruggedized Commercial Off The Shelf (COTS) liquid-cooled scroll Cryomech compressors that use helium as an operating fluid to cryogenically cool the SI. Operation of the cryocooler is largely automatic, with two modes, a ground mode and flight mode that determine where and how the waste heat from the compressor(s) is rejected. The cryocooler has automatic monitoring and is designed to shut itself off in the event of critical faults. The cryocooler draws about 23kVA from the observatory bus 4, 115 VAC, 400Hz, three-phase power source, when both compressors are operating at maximum load. Eight frequency converters in the aft upper deck convert the input power from 400Hz to 60Hz for the compressors, coolant pump and controller operation. The two heat exchanger axial fans operate from unconverted 400Hz, three-phase power from observatory bus 4.

The cryocooler power source is downstream of the cabin depressurization relay which will remove power from all components in the event of depressurization.

To maintain ideal SI detector temperature, the cryogenic cooling system should operate continuously, although short (30–45min) interruptions are tolerable and occur regularly when the plane is towed between the hangar and the airfield, or during maintenance.

7. Performance on Sky

The performance of the upGREAT/LFA system was already shown in Risacher et al. (2016b). We present here some results for the upGREAT/HFA system, taken during flight campaigns in 2017 and 2018.

7.1. In-flight instrument performance verification

7.1.1. upGREAT/HFA receiver sensitivities and stability

The upGREAT/HFA best achievable sensitivity can be seen in Fig. 16. The left plot shows the HEB mixers IV curves for the best LO distribution between pixels. The mixers achieve their optimum performance only over a narrow range in LO power, which is not identical for all pixels. The right plot shows the measured double sideband receiver noise temperature over its IF range with bias settings of 1.5mV for all pixels. For projects requiring best stability, different bias settings are recommended, with slightly degraded sensitivities. During normal observations, the time allocated for tuning optimizations can be less than a few minutes, therefore there is generally insufficient time to ideally set the parameters (the most critical one is the LO power distribution between the various pixels) and achieving the performance shown in Fig. 16. As a consequence, some of the pixels can have degraded performance. Thus, for each flight leg, the project scientist has to trade-off between observing time and best possible performance.

To assess the instrument stability performance, a typical figure-of-merit is the Allan variance. The results for the upGREAT/LFA receiver were detailed in Risacher et al. (2016a), in Sec. IV.B.4. Under best stable ambient temperature conditions, the spectroscopic stability between 1.4MHz width channels is typically above 80–100s (Allan time). But if ambient temperature is not stable and drifts for example more than 1 degree per hour, the stability can easily degrade to 30–40s. For the upGREAT/HFA receiver, performance is similar, only slightly worse, with spectroscopic Allan times typically better than 60s. It also appears that the upGREAT/HFA system is more sensitive to the pulse-tube induced mechanical vibrations, which modulate the LO coupling causing bandpass changes and ultimately the baseline can be dominated by the standing wave originating from the mismatch between the HEB mixers and the cryogenic amplifiers. There are several ways to mitigate this known effect. For chopped observations, data can be read-out synchronously with the wobbler phase, effectively suppressing this effect. With this solution in chopped mode, no degradation or limitation due to this effect is seen in the baseline quality up to integrations of $\sim$2h (more is difficult to accommodate with an airborne observatory). An example is given in Fig. 17 showing the typical achieved baseline quality.

If total power observations are requested (e.g. on-the-fly modes, up to 20s on phase), procedures have been developed to minimize this baseline modulation during post-processing (Higgins, in prep.). This takes advantage of the fact that the spectral pattern is constant for a given tuning (pixel-based) and hence can be fit to observations of “blank” sky (an approach similar to what has been done with Herschel/HIFI in Kester et al. (2014)).

7.1.2. Array parameters determination

The necessary steps to be performed in the very first flight after (re-)installation of the instrument (or after a change of its configuration that might affect the optical alignment) are described for the LFA in Risacher et al. (2016b). For the upGREAT LFA$+$HFA characterization, we basically follow the same procedures, with parameters adjusted to the smaller (HFA) beam and the more compact array configuration:

The first step is to determine the arrays geometry in the focal plane and hence their optical alignments to the optics of the telescope’s imager, which later on will control the tracking of the telescope (and hence the pointing of the GREAT instrument). Once the array geometry has been derived, the actual sky coordinate of each pixel will be introduced to the raw data header of each observation dump (typically every 0.3–1.0s). The alignment of, in particular, the HFA array (means, determining the pixel positions precisely) is challenging even on the brightest calibrators, requiring excellent residual atmospheric transmission. As an example, during our first and only flight out of Palmdale after re-installation for the OC6-G series on May 22nd, 2018, flight planning enforced observations of Jupiter starting at low flight altitudes (36kft) and mediocre atmospheric conditions (15$\mu$m PWV). Short scale atmospheric instabilities (see Fig. 2 for a display of the limited atmospheric transmission near the [OI] line) made it impossible to derive the HFA parameters, and only with our first southern deployment flight (cruising now at 41–43kft in cold southern skies), the array parameters could be derived. From two observations of Mars in June 2018 (on June 16th, the diameter of the planet was 18.9${}^{\prime \prime }$), we determined the offset and angle of the rotator axis and the array geometries. We derive the individual pixel positions with an rms deviation of typically 1${}^{\prime \prime }$. The arrays’ central positions were co-aligned with respect to the central HFA pixel to 0.3${}^{\prime \prime }$ (LFAV) and 1.0${}^{\prime \prime }$ (LFAH), so well tolerable within their 14.1${}^{\prime \prime }$ main beam. In practice, for almost all projects we “tracked” on the central pixel of the HFA because of the higher tracking requirement of its smaller beam.

With extended Mars, the HFA beam sizes could not be well constrained during this year’s observations but as the internal array optics has not changed since the commissioning of the individual arrays in 2016 (when Mars was more compact, 7.6${}^{\prime \prime }$), it is safe to adopt the beam sizes determined from those observations. On June 26th, 2018, we derived the coupling efficiencies of the LFA and HFA arrays with dedicated cross-scans of the individual array pixels across Mars. In Table 2, we quote the average main beam efficiency per sub-array, with the standard deviation.

There are several science applications for which the absolute pointing accuracy of the system (telescope plus instrument) is critical. In some cases of observations towards dark cloud complexes (a prominent example is IRAS16293) where there are no tracking stars available for the telescope’s focal plane imager and positioning based on gyroscopes and the wide field imagers may be unstable, drifting on short timescales to as much as 3–5${}^{\prime \prime }$, which is unsuitable for observations with a 6${}^{\prime \prime }$ beam. Under nominal circumstances, with a bright star in the FPI field, optical tracking is good to 0.5arcsec. Tracking the instrument on the central HFA pixel, additional uncertainty comes from the precision of the determinations of the rotator offset and axis and of the pixel’s offset from the rotator axis. The combination of these errors is obviously difficult to predict, but experience has shown that achieving an absolute positioning of HFAV00 to better than 1.5${}^{\prime \prime }$ is challenging.

7.2. Science use/operation

Since mid-2017 GREAT is operated in two configurations only: upGREAT/LFA with HFA and upGREAT/HFA with 4GREAT (Duran et al., in prep), thereby simplifying the operation and minimizing the requests for configuration change, without compromising on the frequency coverage. Since first light on April 01, 2011, GREAT has logged 137 science flights with more than 1000 successful science hours — including 63 flights with upGREAT/LFA, 39 with upGREAT/HFA (25 flights in their joint configuration, since June 2017). More than 200 science projects have been successfully supported. The spatial multiplexing of the arrays now makes it possible to efficiently address large-scale mapping projects within a reasonable investment of observing time. In Fig. 18, we present an example of a large-scale [OI] 63$\mu$m fully-sampled map towards the prominent NGC3603 star forming region, covering an area of 160${}^{\prime \prime }\times 150^{\prime \prime }$ ($4.7\times 4.4$pc²). The map is composed of 2300 spectra, acquired in chopped on-the-fly mode (see Risacher et al. (2016b) for the description of the GREAT observing modes) during a short 45min observing leg in summer 2017 (with excellent atmospheric conditions at 43kft flight altitude). The sensitivity across the map is very uniform, the noise rms of the data cube (in the velocity channels free of emission) is 2K (with a spectral resolution of 0.5km/s)

Several legacy-type projects have been completed/launched with upGREAT during the last two years: e.g. a 1-square degree region of the Orion Molecular Cloud (Pabst et al. submitted), a fully-sampled map of the M51 galaxy (PIs: J. Pineda, J. Stutzki) is nearing completion. This year the distribution of [CII] and [OI] across the Central Molecular Zone ($1.5\times 0.33$deg², composed of 2.3 million spectra) (PIs: A. Harris, R. Güsten) has been measured during the southern deployment. None of these projects would have been possible without the enhanced mapping speed of upGREAT. All of these examples, together with future legacy projects that will be selected from the explicit calls for legacy proposals starting with observing cycle 7, will contribute to the GREAT legacy of the SOFIA project.

8. Future Work

The main current limitation for the upGREAT/LFA is the LO system. The solid state multiplier chains provide just enough output power, but are narrow in frequency range and quite fragile. A QCL option would provide much more output power but they are even more narrow band, and having a frequency locked system is not trivial. When the technology allows covering much more bandwidth, with enough output power, the entire range of the 1.8–2.5THz could be covered at once.

The next natural step for the upGREAT/HFA array would be to populate its second polarization with seven additional pixels (currently on hold due to lack of funding). The cryostat is already prepared for it and has all the wiring in place. What is needed therefore is to fabricate seven new mixer blocks with the feedhorns. The LOs have enough power to pump 14 pixels. Currently, less than 5% of the LO power is coupled to the mixers, therefore 95% of it is being terminated into an absorber. It will make sense to use this 95% of LO power for the second polarization. Another future improvement will be to frequency-lock, or better yet, phase-lock the LO, in order to achieve a better frequency accuracy.

Acknowledgments

We thank the SOFIA engineering and operations teams whose support has been essential for the successful installation and operation of the upGREAT instrument. We thank the SOFIA project office at DLR, and in particular A. Himmes and D. Lilienthal for their outstanding continuous support during the development years of GREAT. We are grateful to J.-P. Perez-Beaupuits for kindly providing upGREAT results (Fig. 15) prior to publication. The development of upGREAT was financed by the participating institutes, by the Federal Ministry of Economics and Technology via the German Space Agency (DLR) under Grants 50 OK 1102, 50 OK 1103 and 50 OK 1104 and within the Collaborative Research Centre 956, sub-projects D2 and D3, funded by the Deutsche Forschungsgemeinschaft (DFG).

Notes