2.1.

FIFI-LS Overview

In-depth information on FIFI-LS was published by Fischer et al.¹ and Looney et al.³ Here we give a coarse overview of the instrument design, concentrating on subsystems and concepts that are relevant to the instrument upgrade.

FIFI-LS is a dual-channel imaging spectrometer. Both spectral channels operate simultaneously, allowing observations of two spectral lines at the same time. The long-wavelength (red) channel operates between 115 and $203\mu \mathrm{m}$ and the short-wavelength (blue) channel between 51 and $125\mu \mathrm{m}$. The instantaneous spectral coverage is 750 to $2350\mathrm{km}/\mathrm{s}$ (0.51 to $0.90\mu \mathrm{m}$) in the red and 800 to $3000\mathrm{km}/\mathrm{s}$ (0.19 to $0.71\mu \mathrm{m}$) in the blue channel, depending on the observed wavelength. The complete spectral range of the instrument could be covered with about 125 settings in the red and 170 settings in the blue channel, without overlap. The spectral resolution is also wavelength dependent and ranges from 150 to $550\mathrm{km}/\mathrm{s}$ (0.04 to $0.16\mu \mathrm{m}$). Spatially, each channel has a $5\times 5$ spatial pixel (spaxel) field of view (FOV). In the red channel, the spaxel size on the sky corresponds to $12.2\times 12.5$ square arc sec, in the blue channel to $6.14\times 6.25$ square arc sec.

Both channels are equipped with a reflective integral field unit (IFU), which reorganizes the square $5\times 5\text{spaxel}$ FOV into a $(25+4)\times 1\text{spaxel}$ slit, which is coupled into a quasi-Littrow mounted spectrometer. Figure 1 shows the IFU and a schematic of how the two-dimensional FOV is reorganized into a one-dimensional slit. The IFU comprises three sets of five mirrors each. The first set is the slicer mirrors. These mirrors are in a focal plane of the instrument. They slice the $5\times 5\text{spaxel}$ FOV into five image slices and reimage the entrance pupil for each slice onto one of the five capture mirrors, which form the second set of mirrors. The capture mirrors are positioned in the plane of a pupil image. Each of the five mirrors refocuses its slice onto a slit mirror in the third set of mirrors of the IFU. Therefore, the slit mirrors are again in a focal plane. The focused image of the slit has a length of $25+4\text{spaxels}$ at this position. The space for the four additional spaxels is gaps between the five individual slices. These gaps are introduced to avoid crosstalk between the edges of the slices. Finally, the slit mirrors ensure that the pupil image of each slice is correctly centered on the grating once it passes the collimating optics. The dispersing element in each spectrometer channel is a reflective grating. Each grating may be rotated within a total range of about 40 deg at sub-arc second precision, to select the observed wavelength. The gratings have been optimized for their specific wavelength ranges. The simulated grating efficiencies are shown in Fig. 2. Especially, the blue channel grating is highly polarizing toward the ends of the covered wavelengths. The polarization has to be taken into account when optimizing the new detector arrays, when pixels have polarization dependent quantum efficiency (QE).

Fig. 1

IFU concept of FIFI-LS. (a) Close-up view of the IFU mirror components for the red channel with the rays for the central slice shown.³ The blue channel is similar. (b) Illustration of the field-imaging concept in FIFI-LS. The $5\times 5\text{spaxel}$ FOV imaged in the plane of the slicer mirrors is sliced into five slices and reorganized on a $(25+4)\times 1\text{-spaxel}$ pseudoslit, imaged in the plane of the slit mirrors. The pseudoslit is dispersed along the spectral direction in the spectrometer optics and imaged onto the 400-pixel detector with 25 pixels in the spatial direction along the slit and 16 pixels in spectral direction perpendicular to the slit. The rainbow colored spaxels and highlighted columns of spectral pixels (spexels) give an idea on how the spatial information of the spaxels is reorganized and finally spectrally dispersed. The right side of the schematic shows how the FOV and detector size could be increased for FIFI+LS; this is explained in Sec. 3.

Fig. 2

Simulated grating efficiencies of the FIFI-LS gratings. The red channel operates in first order only, for the blue channel first and second orders are used—in operation, a filter changer switches between filter sets optimized for either order. The gratings have polarizing effects. The TE polarization has the $E$-field parallel to the grating grooves, whereas the TM polarization has the $E$-field perpendicular to the groves. Simulation by Norbert Geis, Max Planck Institute for Extraterrestrial Physics.

The core of the instrument is its two 400-pixel gallium-doped germanium photoconductor detector arrays.⁴ Each array is assembled from 25 modules—one for each spaxel—with 16 pixels in the spectral direction. For the red channel, the modules have been specially designed to apply strong uniform stress to the 16 pixels in order to shift their spectral response toward longer wavelengths. The modules in the blue channel apply only enough stress to ensure the mechanical stability of the pixel stack. The detectors have been assembled pixel by pixel, manually. The pixel yield is $>95\%$. The photon dominated NEP of the detectors is about $5\times {10}^{-17}\mathrm{W}{\mathrm{Hz}}^{-1/2}$. The dark NEP is about $5\times {10}^{-18}\mathrm{W}{\mathrm{Hz}}^{-1/2}$.⁴ The QE as inferred from measurements assuming background limited performance ranges from 25% to 35%.³ Light is coupled to the individual pixels via funnel shaped light cones. The area of a pixel in the focal plane of the detector is about $3.9\times 3.9\mathrm{m}{\mathrm{m}}^2$. Each detector module is read out by a cold read-out electronic (CRE) circuit. The CREs are specially designed complementary metal–oxide–semiconductor (CMOS) circuits, originally developed for the Herschel-photodetector array camera & spectrometer (PACS).⁵ The detector current is read out by a capacitive feedback, integrating amplifier. A sample-and-hold circuit within the CRE acts as analog memory between integrator and multiplexing circuit. The capacitor voltages are sampled with a rate of 250 Hz. The measurement is done in a sampling up the ramp technique, typically resetting the capacitor voltage every 32 samples. Therefore, the detector electronics supply a rate of about 7.8 ramps per second with 32 samples on each ramp.

Both detectors are cooled to temperatures of 1.6 K to attain the NEP mentioned above. The spectrometer optics are cooled to 6 K and the entrance optics to 77 K, reducing the thermal background within the instrument. Cooling of the instrument is accomplished by the four-layer bath cryostat design shown in Fig. 3. The first layer is the evacuated cryostat shell at room temperature. The second layer is cooled to 70 K by a 25-L liquid nitrogen vessel attached to a radiation shield and the optical bench for the entrance optics. On the third layer, the cooling is provided by a 35-L liquid helium vessel. The optical bench with the IFU and spectrometer optics belongs to this level. Due to the heat influx and the heat produced on the optical bench by electronic components, the temperature is typically 6 K, which is slightly higher than the boiling temperature of liquid helium at 4 K. The innermost layer is occupied by the two 400 pixel detector arrays and the CREs. This layer is cooled by a small 2.8-L superfluid liquid helium vessel. This vessel is pumped in order to attain the operating temperature of 1.6 K.

Fig. 3

CAD drawing of the FIFI-LS cryostat attached to the SOFIA telescope flange.¹

2.2.

Science with FIFI-LS

With its sensitivity (see Fig. 4), wavelength coverage, spectral and spatial resolution, FIFI-LS plays a major role in investigations of the ISM both extragalactic and within our galaxy, as well as for investigations of (primarily massive) star formation since 2014. The wide wavelength coverage of FIFI-LS allows the observation of a multitude of diagnostic lines in the far-infrared (FIR). Table 1 lists all lines observed to date. The key features observed most often with FIFI-LS are the bright FIR fine structure lines of [CII] at $158\mu \mathrm{m}$, [OI] at 63 and $146\mu \mathrm{m}$, [OIII] at 52 and $88\mu \mathrm{m}$, [NII] at $122\mu \mathrm{m}$, [NIII] at $57\mu \mathrm{m}$, and molecular lines like the CO rotational lines from $J=13$ to 12 up to $J=38$ to 37. FIFI-LS observations also provide the underlying continuum flux for sufficiently bright objects. Below, we summarize a few of the published FIFI-LS results.

Fig. 4

Sensitivities of FIFI-LS⁶ (thin solid lines) and FIFI+LS modeled upgrade sensitivities (thick solid, dashed, and dotted lines). The values are calculated for an $S/N$ of 4 in 900 s on-source integration time, not including the larger FOV of FIFI+LS. A penalty of √2 is included in all values to account for the differential observation schemes typical for FIR observations. (a) The minimum detectable continuum flux assuming that all spectral pixels are coadded for a single spatial pixel. (b) The minimum detectable line flux when using the optimal number of spectral pixels to retrieve a line for pure detection. The thick solid lines are the baseline for the FIFI+LS upgrade assuming similar quantum efficiencies for the new KID-arrays and the current detectors. The dashed/dotted lines would result from the KID-arrays realizing double $(\sim 60\%)/\text{triple}$ ($\sim 90\%$) the QE of the current setup. The spectral coverage in the blue channel is achieved using first or second diffraction order of the reflective gratings (green and blue curves). The spectral coverage in the red channel uses first diffraction order (red curves). Toward $200\mu \mathrm{m}$, the KID-arrays are expected to perform much better than the current detectors, which have a cutoff wavelength close to $200\mu \mathrm{m}$.

Table 1

Spectral lines observed with FIFI-LS1—there are no gaps in FIFI-LS’s wide spectral coverage, allowing the observation of a multitude of diagnostic lines in the FIR.

FIFI-LS has efficiently mapped complete galaxies in [CII], to study how star formation rates and [CII] emission correlate in different environmental conditions. For example, Pineda et al.⁷ probed the relationship between SFR and [CII] over a wide range of conditions by mapping the whole disk of the M51 galaxy including its companion M51b. The preliminary results are that SFR and [CII] emission are well correlated over the disk of M51 including the center, spiral arms, and interarm regions, in good agreement with the KINGFISH galaxy relationship.⁸ However, the M51b galaxy shows a significant deficit of [CII] emission with respect to the SFR rate estimated from FIR-continuum. M51b is a barred lenticular galaxy in a poststarburst phase, in which massive star-formation is suppressed.⁹ The lack of massive star formation in M51b is consistent with the faint [CII] emission detected, but it is inconsistent with the large SFR estimated in this galaxy. The bright $24\text{-}\mu \mathrm{m}$ dust continuum emission, which dominates the SFR estimate, is likely heated by an active galactic nucleus, therefore not tracing star formation activity. The observed [CII] deficit in M51b suggests that this galaxy is a nearby analog of ultra-luminous infrared galaxies and represents an important laboratory, in which to study the origin of the [CII] deficit observed in these galaxies.

Bigiel et al.¹⁰ presented the results for another full-galaxy [CII] map with FIFI-LS of the nearby spiral galaxy NGC6946. The data allowed disentangling the [CII] luminosities of the arms (73%), center (19%), and interarm regions (8%) of NGC6946. A comparison of the radial profiles of [CII] emission to various gas and star formation rate tracers ([CII]/TIR, [CII]/CO, and [CII]/PAH) reveals a pronounced “[CII]-deficit” in the center. In combination with models, a radially declining trend of alpha_CO is found, reflecting the radially declining metallicity gradient in this galaxy. Finally, the overall low alpha_CO values argue for a surprisingly low-dark molecular gas content in this galaxy, which is in contrast to estimates in the Milky Way.

FIFI-LS has also studied the [CII] emission in relation to other SFR tracers in multiple other galaxies. One example are the observations of the central region of the active galaxy NGC 4258 by Appleton et al.¹¹ The results suggest that a significant share of the [CII] emission from NGC 4258 is not associated with star formation but is excited by shocks and turbulence induced by the highly inclined jet of the galaxy. Observing with FIFI-LS, Smirnova-Pinchukova et al.¹² found excessive [CII] emission in an active galactic nucleus (AGN) galaxy as a likely signature of an AGN-driven outflow.

Observations of the galactic center presented by Iserlohe et al.¹³ showed how the ability of FIFI-LS to quickly map large areas in multiple fine diagnostic lines was used to characterize photodissociation regions (PDR). Line emission maps of four lines: [CII], [OI] 63 and $145\mu \mathrm{m}$, and CO $J=14$ to 13 as well as the FIR flux derived using FIFI-LS, FORCAST,¹⁴ and PACS continuum data were used to infer the density and FUV-field using PDR models.¹⁵ The analysis by Iserlohe et al. shows that the [OI] line fluxes at $\sim 63\mu \mathrm{m}$ have to be affected by self-absorption, whereas there is significant contribution to the [CII] line flux not originating from the probed PDR. The ability to measure the [OI] line at $145\mu \mathrm{m}$ and the CO line in addition to [CII] and [OI] at $63\mu \mathrm{m}$ were the key to derive the correct PDR parameters.

3.1.

Concept Overview and Estimated Performance Improvements

The sensitivity, FOV, and instantaneous spectral coverage of FIFI-LS are currently limited by its 90s-era photoconductor arrays. All of these performance parameters can be improved significantly by upgrading the instrument with new KID arrays. These arrays may be produced with an order of magnitude more pixels, enabling the increase of the FOV and the instantaneous spectral coverage of the instrument. KIDs have less generation–recombination ($g-r$) noise than photoconductors, which translates into a $\sqrt{2}$ sensitivity gain for background-limited operation. Sensitivity data for the current FIFI-LS instrument⁶ and the modeled sensitivities of the upgrade are shown in Fig. 4. The baseline improvement model (thick solid line) relies on the negligible $g-r$-noise; further improvements are possible by increasing the QE of the KIDs beyond the $\sim 30\%$ of the current FIFI-LS detectors.

The presented instrument upgrade concept, called FIFI+LS, will maximize the science output while keeping cost and development time low by utilizing most of the original FIFI-LS design and components. This is possible because the original instrument design anticipated future detector upgrades and the entrance optics were designed to accommodate a larger FOV. Table 2 shows the baseline parameters for the FIFI+LS upgrade in comparison to the current FIFI-LS instrument.

Table 2

Upgraded FIFI+LS instrument parameters compared to FIFI-LS.

On the detector level, the instrument’s sensitivity will improve by a factor of 1.4 to 2.5. The lower end of the estimate is the gain from the reduced $g-r$-noise. Additional gains are expected from improvements of the QE. The current FIFI-LS detectors have a QE of 25% to 35%.⁴ Depending on the implementation, the new KIDs arrays could yield QEs of 40% to 80%. The FOV of the instrument will increase by a factor of 1.75. This increase will enable faster mapping and more efficient observation modes (e.g., Lissajous scanning).

The increased number of spectral pixels will have three distinct advantages. (1) It will facilitate the observation of wide lines, especially in extragalactic sources, providing overall better baseline coverage. (2) It will enable the observation of atmospheric water features concurrent with the observations for most of the important FIR-diagnostic lines. This will improve the accuracy of the atmospheric calibration and strongly reduce the overhead necessary for atmospheric calibration observations. (3) The detector may be split into two (or more) sections in spectral direction, each optimized for a subband of the spectral coverage in the respective channel. This will allow increasing the sensitivity at key lines (e.g., [CII] and [OI]). Figure 5 shows the increased spectral coverage close to the [CII] line along with ATRAN¹⁶ models of the atmospheric transmission. The two visible water absorption features are routinely used to measure precipitable water vapor with FIFI-LS¹⁷ at specific times in a flight. FIFI+LS could continually monitor water vapor for [CII] observations in parallel with the astronomical observation with no need for additional observing time.

Fig. 5

Atmospheric transmission for three different values of precipitable zenith water vapor close to the [CII] fine structure line. The current instantaneous spectral coverage is within the dashed lines, and the upgrade covers the full spectral width shown. The telluric water features at 155.6 and 156.2 are used for atmospheric calibration.

Through a careful redesign of the spectrometer optics, it will be possible to improve the spectral resolution in the blue channel and to increase the spectral range of the red channel to include the [NII] $205\text{-}\mu \mathrm{m}$ line. In all, we expect the improvements to increase the observing speed by a factor of 3.5 to 12 for compact sources and for mapping. This gain in efficiency will increase the observable universe for SOFIA/FIFI-LS and provide more efficient use of SOFIA’s precious observing time.

The three major areas of the instrument that are affected by the upgrade are the detectors and detector readout, the IFU and spectrometer optics, and the cooling concept. The necessary changes are addressed in the following sections.

3.2.

KIDs Arrays

There have been important improvements in detector technology since FIFI-LS and PACS fielded the largest gallium-doped germanium photoconductor detector arrays (400 pixels), especially in respect to KIDs. The Caltech/JPL group is developing sensitive feedhorn-coupled aluminum KIDs for use at 240 to $420\mu \mathrm{m}$ in the balloon borne spectrometer TIM.¹⁸ TIM will deploy $\sim 3600$ KIDs using $\sim 1000\text{pixel}$ subarrays and has an NEP requirement of $1\times {10}^{-17}\mathrm{W}{\mathrm{Hz}}^{-1/2}$, similar to the photon-limited NEP needed for FIFI+LS. TIM prototype detectors have demonstrated NEPs of $4\times {10}^{-18}\mathrm{W}{\mathrm{Hz}}^{-1/2}$,¹⁹ and simulations indicate an 80% dual-polarization optical efficiency across the 317- to $420\text{-}\mu \mathrm{m}$ band.²⁰ This design will be scaled to the shorter wavelengths of the FIFI+LS red channel. As the responsivity of these KIDs scales inversely with the inductor volume,¹⁹ shrinking the absorber to scale the design to shorter wavelengths will result in a moderate reduction in the NEP. With a factor of 2 reduction in the linear scale and assuming an 80% QE, the detector NEP referenced to the front of the feedhorn for the red channel detectors will be $1.9\times {10}^{-18}\mathrm{W}{\mathrm{Hz}}^{-1/2}$.

The KIDs developed for TIM will also form the basis for the FIFI+LS blue channel detectors but will use a modified absorber and optical coupling scheme appropriate for the shorter wavelengths. Recent simulation work and initial testing was presented by Perido et al.²¹ in a study for detectors for the galaxy evolution probe (GEP).²² Their work addresses KIDs designed for wavelengths between 10 and $100\mu \mathrm{m}$. GEP will need sensitivities as low as $1\times {10}^{-19}\mathrm{W}{\mathrm{Hz}}^{-1/2}$ in this wavelength range, which is much better than what is necessary for FIFI+LS. As with the TIM detectors, Perido et al. were pursuing lumped element aluminum. Their simulation work has shown that thin aluminum film may be patterned on a silicon substrate to produce good optical absorption at wavelengths as short as $10\mu \mathrm{m}$. The short meander design studied by Perido et al.²¹ used a 200-nm wide absorber line that is interrupted with a meander to increase the resistance per unit length while simultaneously introducing a distributed capacitance to compensate for the accompanying distributed inductance. For backside illumination, this design achieves a dual-polarization efficiency of 58% over the 10- to $30\text{-}\mu \mathrm{m}$ band, a fractional bandwidth larger than that of the 51- to $125\text{-}\mu \mathrm{m}$ FIFI+LS blue channel spectral range. This absorber geometry will be the starting point of a future optimization targeting the FIFI+LS blue band. The absorber will be illuminated with a microlens equipped with a broadband antireflective layer.

Perido et al.²¹ presented dark-noise measurements of KIDs patterned using their short meander design. The noise is dominated by the two-level system (TLS) noise common to KIDs, and when operating at 200 mK the electrical noise is $S_{xx}=6\times {10}^{-17}{\mathrm{Hz}}^{-1}$. The detector responsivity was not measured, but following the approach described by Hailey-Dunsheath et al.,²³ conservative assumptions about the aluminum film quality lead to an estimated responsivity of $R_x=1.8\times {10}^9{\mathrm{W}}^{-1}$ and a TLS-limited NEP of $4.3\times {10}^{-18}\mathrm{W}{\mathrm{Hz}}^{-1/2}$. To operate at the somewhat longer wavelengths of the FIFI+LS blue band, the absorber studied by Perido et al. will be increased to $120\mu \mathrm{m}$ in diameter, which represents a factor of 4 increase in the absorber area and a corresponding increase to the NEP. The TLS noise will be reduced, however, by increasing the size of the capacitor. The small size of the capacitor employed by Perido was determined by their 0.4 mm detector pitch, which is significantly smaller than the $\sim 0.8\text{-}\mathrm{mm}$ pitch of the FIFI+LS focal plane. A detailed capacitor geometry will be the subject of a future optimization, but it is well established that TLS noise is reduced by increasing the capacitor size,²⁴ and we expect a factor of 2 reduction in the TLS noise will be readily achieved. Assuming a 50% QE, this leads to an NEP referenced to the front of the microlens of $1.6\times {10}^{-17}\mathrm{W}{\mathrm{Hz}}^{-1/2}$.

The performance of the FIFI+LS KIDs will be similar to those developed for TIM. The diffraction-limited spot size produced by the microlens on the blue channel detectors will be smaller than the wavelength. Assuming a $125\text{-}\mu \mathrm{m}$ diameter absorber the inductor volume would be a factor of $\sim 3$ larger than in the TIM prototypes. The capacitor area would be reduced to maintain the LC resonance frequency at $\sim 300\mathrm{MHz}$, and the expected detector NEP would be $\sim 1.5\times {10}^{-17}\mathrm{W}{\mathrm{Hz}}^{-1/2}$. The absorber geometry, pixel size, and layout of the microlens and feedhorn arrays for the blue and red channels has to be refined during a further study. In addition, the detector response within the FIFI+LS spectral bands (especially the large bandwidth of the blue channel) has to be optimized and strategies for maximizing array efficiency at the most used wavelengths have to be investigated, all while retaining background-limited performance. One possible strategy is to split the blue channel detector into two sections along the spectral direction. Those sections would be optimized individually (e.g., considering the polarizing effects of the grating shown in Fig. 2).

The switch from background-limited photoconductors to KIDs will improve the point source sensitivity. A photon absorbed in a photoconductor excites a single-charge carrier, and the stochastic recombination of these charge carriers increases the fundamental photon noise by a factor of $\sqrt{2}$. In contrast, a photon absorbed in a KID breaks a large number of superconducting electron pairs ($\sim 17$ for $\lambda =100\mu \mathrm{m}$ in an aluminum film), and the associated recombination noise is negligible. Additionally, the absorption efficiency in the FIFI+LS KIDs is anticipated to be $\sim 40\%$ to 80%, an improvement on the 25% to 35% achieved with the FIFI-LS photoconductors. As a result of these two terms, we expect the KID arrays will provide at least a factor of $\sqrt{2}$ sensitivity increase based on the $g-r$ noise reduction and an increase up to a factor of 2.5, if we can realize the 80% QE.

The KIDs arrays will also require a redesign of the readout electronics for the FIFI+LS upgrade. Each KID forms a microwave resonant circuit with changes to the absorbed optical power driving shifts in the resonant frequency. The readout electronics drive each KID at its resonant frequency and determine the frequency shifts by monitoring the complex transmission of the drive tone. We will use the same FPGA-based readout originally developed for the BLAST-TNG instrument²⁵ and further developed for TIM and other instruments fielding large-format KID arrays.²⁶

3.3.

Optics

The general optical layout of FIFI+LS will remain the same as FIFI-LS. The upgrade will affect the IFU, the collimator optics, and the final detector camera optics. The entrance optics can remain unchanged as they were already designed to support an FOV up to $90\mathrm{arc}\sec \times 90\mathrm{arc}\sec$. The diffraction gratings will also be reused for the upgrade. The gratings are already the optimal size for the available space and the groove distance and profile are optimal for the FIFI+LS wavelength range. Not including the detectors, the gratings are the most expensive parts of the instrument optics. Retaining the original gratings saves upgrade costs on the order of $450k.

Each IFU is composed of three sets of mirrors as explained in Sec. 2.1. Each mirror-set contains as many mirrors as there are slices of the FOV. The increase of the FOV from $30\mathrm{arc}\sec \times 30\mathrm{arc}\sec$ to $45\mathrm{arc}\sec \times 35\mathrm{arc}\sec$ in the blue channel and from $60\mathrm{arc}\sec \times 60\mathrm{arc}\sec$ to $90\mathrm{arc}\sec \times 70\mathrm{arc}\sec$ in the red channel necessitates an increase in spaxel number to sample the FOV sufficiently. The baseline design uses $9\times 7\text{spaxels}$, with 7 slices of 9 spaxels in the IFU. Therefore, the upgraded IFU will need seven instead of five mirrors for each of its three mirror sets. This baseline design of the IFU has already been modeled in the optical software ZEMAX. Figure 6(a) shows this seven-slice model in a view similar to the original setup shown in Fig. 1. The width of the original five-slice IFU setup in Fig. 1 is driven by the capture mirror set, which needs 86 mm. The new seven-slice setup will need 105 mm, driven by the wider fold mirror R5. The capture mirror width for the new setup would be even wider than 105 mm but was reduced by reducing the individual mirror widths. This reduced width still contains the full geometric beam with additional margin for diffraction. The $\sim 20\text{-}\mathrm{mm}$ overall increase in width can be accommodated within the cryostat. The structures closest to the IFU holder are more than 10 mm away on either side. This is visible in the CAD-image in Fig. 6(b). The ZEMAX model shows that the optics will remain diffraction limited even in the corners of the larger FOV. The final optimization of the complete spectrometer optics in both channels has to be done in a future study. The currently baselined spaxel pitch of 5 arc sec in the blue and 10 arc sec in the red channel is a compromise between instantaneous spatial resolution and light losses due to diffraction within the IFU. Within the future study, it has to be determined whether this baseline is the best solution in terms of mapping speed and sensitivity with a full model of the optics.

Fig. 6

(a) Seven-slice IFU concept of FIFI+LS for the red channel. The blue channel is similar. (b) Front-diagonal- (lower image) and top- (upper image) view of the current FIFI-LS five-slice IFU of the red channel. The IFU holder structure is the translucent structure shown. The closest structure to the right of the IFU holder is the center baffle blocking stray light from the blue channel. The closest structures to the left are a mirror holder and the holding frame for the diffraction grating. All structures are more than 10 mm away from the IFU holder.

3.4.

Cooling

The new KID arrays operate at temperatures below 250 mK, which is considerably lower than the current FIFI-LS detectors’ operating temperature of 1.6 K. It will, therefore, be necessary to redesign the instrument cooling. The original cryostat provides four temperature levels: room temperature, liquid nitrogen cooled at 77 K, liquid helium cooled at 6 K, and pumped liquid helium at 1.6 K. The entrance optics are on the 77-K level, the spectrometer optics on the 6-K level, and the detectors on the 1.6-K level.

There are multiple options to realize the necessary detector cooling within the current layout of the FIFI-LS cryostat with different impacts to the overall design. An option with limited impact would only remove the 2.8-L superfluid helium vessel and use the space freed by the vessel and the KIDS detectors, which are smaller in volume compared to the current photoconductor arrays, to accommodate the new detector cooling systems. A more invasive option would require the removal of the 2.8-L superfluid helium vessel and the 35-L helium vessel. This option uses a pulse tube cryocooler to cool the spectrometer optics and provide the intercept for the detector cooling system. This concept takes advantage of the cryocooler compressors installed on SOFIA, which are used for the cryocooling of upGREAT.²⁷

The detector cooling will be realized by an adiabatic demagnetization refrigerator (ADR) to reach temperatures $<250\mathrm{mK}$. This approach is similar to the approach taken by the SOFIA instrument HAWC+.^28,29 HAWC+ has an estimated detector heat load on the ADR of $5\mu \mathrm{W}$²⁸ and an additional parasitic load of $18.5\mu \mathrm{W}$²⁸ excluding vibrational heating. The vibrational heating initially tripled the total heat load on the ADR to $\sim 70\mu \mathrm{W}$.³⁰ This was greatly reduced by tuning the mechanical resonances of the system. The HAWC+ADR design supplies a 170-mK operating temperature for the duration of complete SOFIA science missions.²⁹ For our KID arrays, we will aim for higher operating temperatures between 200 and 250 mK. The FIFI+LS baseline detectors have a total of 9856 pixels with an estimated saturation power of 2 pW, which also corresponds to the maximum optical loading expected, resulting in a 20-nW optical load on the detectors. In comparison, the slightly smaller HAWC+ detectors have an estimated background load of $60\mathrm{pW}/\text{pixel}$ giving 230 nW on its three $32\times 40\text{pixel}$ detector arrays.²⁹ HAWC+ routes 822 wires from its 1-K stage to its 170-mK ADR stage for detector readout and housekeeping. The FIFI+LS readout will need a pair of wires for every 750 pixels resulting in only 28 readout wires plus a few additional wires for temperature monitoring. Given these comparative figures between the FIFI+LS baseline design and HAWC+, we are confident that an ADR design similar to the one used in HAWC+ will provide sufficient cooling for the detector stage of FIFI+LS.

The optimal layout for the cryogenic cooling has to be defined in a future study. For the pulse tube option, it needs to be decided whether a standard commercial two-stage option or a custom single-stage option directly coupled to the liquid nitrogen reservoir would be the best. The current heat load on the liquid helium cooled level at 6 K is 1.3 W. TransMIT, the supplier of the upGreat cryocoolers, offers a model that can accommodate 1.17 W at 4.2 K³¹ when paired with one of the two compressor systems on SOFIA. We are, therefore, confident that an optimized design will allow accommodating the heat load from the FIFI+LS system at a temperature below the 6-K of the current spectrometer optics. This would reduce the instrument internal background in the red channel. For the detector cooling, the two considered options include a dual-stage ADR system and a He4 or He3 sorption cooler backed single-stage ADR.