1.1.

Scientific Motivation for Far-IR Spectroscopy

A number of high-priority Origins scientific investigations require sensitive, wide-bandwidth spectroscopy covering the full the far-IR. The extragalactic science cases, in particular, center around charting the cosmic history of the contents and processes within galaxies from the first billion years to the present day using rest-frame mid- and far-IR spectral tracers. The various mid-IR and far-IR features originate from a wide range of interstellar medium phases: ionized, neutral atomic, and molecular material, and in aggregate they constrain heavy-element contents, local UV radiation field conditions, which reveal stellar properties and physical conditions (temperature and density).^1–3 Crucially, the features in this waveband are largely immune to dust obscuration, providing an unbiased census of the interplay between stars, black holes, and the interstellar material in any given observation.

Of particular interest for Origins, and driving the sensitivity, is the epoch of reionization and the dawn of galaxies’ first billion years. As the Universe was enriched from primordial hydrogen and helium to a medium that contains heavy elements and dust grains, the key cooling pathways shifted from the quadrupole pure rotational ${\mathrm{H}}_2$ lines (28, 17, 12, 9.7, 8.0, 6.9,…, $\mu \mathrm{m}$) (e.g., Refs. 4 and 5) to a combination of atomic fine-structure transitions and dust thermal emission. Origins will probe both sides of this transition out to $z\sim 8$ to 10 using ${\mathrm{H}}_2$ lines, fine-structure lines, and polycyclic aromatic hydrocarbon (PAH) features at 6.2 to $17\mu \mathrm{m}$, which have $15\times$ more power than the brightest atomic cooling lines.⁶ Both the ${\mathrm{H}}_2$ lines and most of the PAH features are redshifted out of the James Webb Space Telescope (JWST) band, but not into the atacama large millimeter array (ALMA) windows in the $z\sim 5$ to 10 era. Origins with the Origins Survey Spectrometer (OSS) spectrograph will detect these powerful bands at early epochs (Fig. 7).

Measuring these features does not require high spectral resolution, as integrated line fluxes are all that is required. However, accessing the full-time history from the first billion years to the present day does require both exquisite sensitivity (because of $1/d_{\text{Luminosity}}^2$ dimming) and coverage across the full far-IR to capture the wide intrinsic wavelength distribution of spectral features—approximately uniformly in $\log \lambda$ and considering the large range of redshifts.

In addition to pointed observations of individual sources and fields identified a priori, similar to what is done with ALMA and will be done with JWST, the OSS on Origins is designed to be a powerful wide-field spectral mapper. In its blind survey mode, millions of galaxies will be identified with redshifts and physical parameters immediately encoded.¹⁰ These requirements set the basic architecture for the OSS: a suite of $\mathcal{R}=300$ wideband long-slit grating spectrometers with detectors operating at or near the fundamental limits.

The resulting excellent sensitivity to spectral line surface brightness (velocity-integrated, spatially unresolved line emission with units of $\mathrm{W}{\mathrm{m}}^{-2}{\mathrm{sr}}^{-1}$, e.g., from diffuse gas or unresolved faint galaxies) also enables three-dimensional (3D) intensity mapping of residual line emission fluctuations after individual sources are extracted. The resulting intensity maps can yield the total cosmic luminosity density—including all sources even those too faint to detect individually—in the key far-IR lines, charting aggregate properties of galaxies through time.^11–17

For pointed observations, the high-resolution mode is also available (see below). This offers a direct measure of stellar and black-hole feedback through outflow signatures in OH and other transitions, potentially a key ingredient in the decline of the cosmic star formation rate density over the last 8 billion years.¹⁸

Another driving case for the OSS is understanding the formation of planetary systems. This requires tracking the evolution of gas mass and composition in protoplanetary disks from the early (protostellar) phase to the late (debris-disk) stages where most of the gas has been dispersed or absorbed into giant planets. For early stage, gas-rich disks, the fundamental of hydrogen deuteride (HD) at $112\mu \mathrm{m}$ is an excellent probe of total mass. It is an optically thin chemical analog of ${\mathrm{H}}_2$ and has only a modest temperature dependence. The challenge for studying protoplanetary disks is not so much the raw spectral line sensitivity—even the relatively weak HD integrated line flux is typically high by comparison with lines in distant galaxies—but extracting line fluxes accurately in the face of a bright dust continuum. HD, in particular, can be easily washed out by the bright disk continuum so it requires a high resolving power ($R$ ∼ several thousands) for reliable measurement. Gas mass evolution is the first step, but with Origins we will also study the conditions for Earth-like habitability by tracking water in disks with low-lying far-IR transitions that probe the bulk of the gas including that in the snow line. Distinguishing material interior to the snow line from material further out requires line-width measurements with a resolving power of 25,000 at wavelengths as long as $179\mu \mathrm{m}$, a key ground-state water transition. These higher-resolving-power requirements are satisfied with the addition of a Fourier-transform spectrometer (FTS) which operates in tandem with the grating spectrometer suite. Finally, in order to study the structure of disks through Doppler tomography, the OSS includes a very-high-resolution mode using an etalon in concert with the FTS to provide $\delta v\sim 1$ to $2\mathrm{km}/\mathrm{s}$ for HD $112\mu \mathrm{m}$ and ${\mathrm{H}}_2\mathrm{O}$ $179\mu \mathrm{m}$.

3.1.

Grating Module Designs

Each grating module forms a light–tight enclosure that includes the detector array and for which the only opening is a slit. Requirements that met by the optical design include as follows.

The optical surfaces in the spectrometers are numerically optimized with many ($\sim 20$) degrees of freedom, essentially approaching free-form surfaces, which is necessary to accommodate the large range of spatial and spectral fields. This approach is fully consistent with the machined aluminum construction. No constraints are placed on distortion in the focal plane, or “smile,” which is the departure from a straight line of the spectrum from a given spatial position. This is because imaging a slit to a slight arc on a top-dimensional (2D) rectangular pixel array simply results in a slight wavelength shift of the pixels as one moves along the slit direction in the focal plane, an issue easily handled in software and potentially providing a spectral sampling benefit. Alternatively, if desired, a focal plane could be built with a custom arrangement of pixels in curved rows to match a curved slit image, since each focal plane is already fully custom. In any case, the slit shape at the entrance to the spectrometer (and thus on the sky) need not be curved. The size and mass of the short-wavelength gratings are driven primarily by the pixel pitch and thus the array size, whereas the long-wavelength modules are driven by the size of the grating and associated pupil. Figure 2 shows the shortest and longest of the grating module designs. The strawman detector array formats are tabulated for each band in Table 1, under the assumption of $1.13f\lambda$ sampling spatially by $R=300$ sampling spectrally. As is typical of wideband grating systems that use a large incidence angle, the systems are highly anamorphic, with spatial focal ratio ($f/\#$) much faster ($\sim 2\times$) than the spectral focal ratio ($f/\#$). Pixels are, therefore, rectangular, elongated in the spatial direction. Additionally, the physical size of an $R=300$ bin changes across the band, so each focal plane will have at least two spectral pitches across the full band to preserve a given fractional bandwidth per pixel.

Fig. 2

Grating module designs for (a) band 1 and (b) band 6, the shortest- and longest-wavelength bands in OSS. In both cases, the $f/14.4$ telescope focus is shown, positioned at the input slit for band 6, and the conjugate to it for band 1. The grating optical designs and focal-plane dimensions shown here represent larger focal planes than are currently baselined (pixel count is limited by readout), and the optics accommodate these, thereby allowing for an upscope. Band 1 in particular can accommodate a large slit up to 250 beams. The band 6 focal plane uses the four spatial fields above and below the center, which is obscured by the grating.

3.2.

Interferometer Design

The design of the interferometer (Fig. 3) is driven by the need to incorporate high spectral resolution at the long OSS wavelengths while maintaining the excellent sensitivity and multi-band capability of the base grating suite. The approach is to intercept the full-band light from a small field and process it with a FTS before sending it to the grating modules for detection. This ensures that the basic sensitivity to spectral lines provided by the $R=\nu /\delta \nu =\lambda /\delta \lambda =300$ detector bandwidth is preserved. The system can be considered as an extreme implementation of a band-limited FTS with many output bands.

Fig. 3

Optical layout for the OSS interferometer. Light is intercepted in the converging beam from the telescope and is collimated with a diameter of 8 cm. After processing by the interferometer, the light is reinserted into the original light path from the telescope to the grating modules. The interferometer is engaged with a sliding stage containing POM1 and POM4. It only accesses a small portion of the grating slits as it is designed primarily for single-source or small-field measurements. On the right, the mirrors are numbered. The single moving stage carries mirrors 11, 12, 15, 16, 23, 24, 27, and 28. A detail not depicted in this figure is the retroreflectors 17 and 29; they are roof-top mirrors oriented at 45 deg with respect to the interferometer plane to provide the polarization rotation (see Ref. 28).

The baseline design has a 30-cm physical throw providing 2.4 m of optical path difference (OPD) between the arms of the interferometer. This provides a wavelength-dependent maximum resolving power of $43,000\times [112\mu \mathrm{m}/\lambda ]$, a good match to expected linewidths of the HD $112\mu \mathrm{m}$ transition in protoplanetary disks (see Origins study report, Sec. 1.2.2.5). Shorter FTS scans can be employed; this reduces $R$, but the $1/\lambda$ dependence is always present since the single FTS scan is common to all six bands. The collimated beam diameter is 8 cm, which ensures divergence (diffraction) is not a concern for the FTS operation, even at long wavelengths. The long-extra optical path between the pickoff/reinsertion position and the interferometer retroreflectors necessitates pupil reimaging to keep the system compact. This is provided by two sets of confocal off-axis Cassegrain telescopes that image the telescope pupil to the back of the interferometer. The FTS architecture is a Martin–Puplett polarizing FTS (Refs. 27; 28), which is optimal for this application. Relative to other approaches to Fourier-transform spectroscopy, this system uses a single-input port and single-output port. It requires a static polarizing grid at the input prior to the beam splitter (e.g., at or near POM2), and a linearly polarized detector, but both output polarizations can be used with independent detectors, exactly as is baselined for the OSS gratings. In each of these output polarizations, the system produces an interferogram with no lost light, unlike a Michelson architecture, which shares power between two ports.

In addition to the FTS, a second, higher-resolution capability enables line-profile Doppler tomography measurements of HD and ${\mathrm{H}}_2\mathrm{O}$ lines around the $112\text{-}\mu \mathrm{m}$ band. (Operation at other wavebands may be possible, a subject for phase-A study.) This ultra-high-resolution mode is provided with an etalon (Fabry–Perot interferometer) consisting of a 26-cm long cavity formed by two partially reflecting mirrors that can be inserted into the beam before the FTS beam splitter. With a design finesse of 70, the etalon provides a resolving power of 325,000 ($0.9\mathrm{km}/\mathrm{s}$) at $112\mu \mathrm{m}$, or 200,000 ($1.5\mathrm{km}/\mathrm{s}$) at $179\mu \mathrm{m}$. The etalon uses the same 8-cm collimated beam provided for the FTS; however, to allow for beam walk in the etalon without substantial signal loss, the mirrors must be 14 cm in diameter. The mirrors in the FTS are similarly oversized. Once in the optical train, the cavity spacing can be changed by $\sim 1\mathrm{mm}$, providing a wavelength scan of a few orders, which is ample to provide complete coverage. The etalon mirrors will either be stretched metal mesh with high heritage (Sec. 10), and recent demonstrations of $\mathcal{R}=\mathrm{100,000}$,²⁹ or more modern patterned silicon,³⁰ which could improve performance for wider band coverage.

3.3.

Full Optical Configuration

Since all six grating modules must be fed through their slits at the telescope focus, and these slits overlap, packaging is an important consideration, and the solution has been obtained through careful iteration with the observatory mechanical and optical designs. Figure 4 shows views of the full configuration with all grating bands, the interferometer, and etalon, as well as the field of view of the OSS in the Origins focal plane.

Fig. 4

The OSS optical and mechanical design has at its core, the six grating models (right). Light from the telescope comes in from the top (blue rays) and is split into the six bands via polarizing grids and dichroics. The FTS spectrometer (back left) picks off light near the entrance of the instrument, passing it through the interferometer and then back through the grating modules. For the highest spectral resolution, an etalon (turquoise tube) is inserted in the light departing from the FTS.

4.1.

Pointed Low-Resolution Spectroscopy

OSS is very efficient for pointed observations since the slits from all six bands overlap so that a complete spectrum is obtained over the full wavelength range simultaneously. For this mode, the Origins field steering mirror (FSM) will be used to chop a source back and forth with a selectable throw. For point sources, a distance of three long-wavelength beams or 1 arc min ensures that the flux in from an unresolved target source is negligible in the off-source position. The chop throw could extend up to the length of the short-wavelength (band 1) slit—2.7 arc min to ensure full efficiency with the source always coupled through the slit. Chop frequencies in this mode are expected to be as low as 0.5 Hz, as the detectors will have good stability to this level. The Origins FSM is capable of faster chopping (up to 10 Hz) if needed to overcome other sources of variability.

4.2.

Mapping Modes

The OSS slits can also be rastered around the sky to generate maps. In this mode, the scan rate must not be so fast that the short-wavelength beams smear when sampled with the finite bandwidth of a detector. These modes together with the observatory scan speed, therefore, set a requirement on the detector time constant $\tau$. As a compromise between agility and the need to have a reasonable attitude control system, Origins selected $60\mathrm{arc}\sec /\mathrm{s}$ as the maximum observatory scan speed. Operating at this maximum speed then requires $\tau$ at the short wavelength ($25\mu \mathrm{m}$) of at most 5 ms, which is comfortably within the Origins 3 ms requirement. This maximum speed allows the low-frequency $f_{\text{knee}}$ to be as high as 17 mHz for recovering the large-scale structure in the intensity mapping experiments. The 3-ms requirement has been comfortably met with all of the detector options under consideration.

4.3.

High-Resolution Mode

With using the FTS, the modulation is naturally provided by the OPD change, so chopping is not required. A given rate of change of the OPD scan $v_{\mathrm{OPD}}$ connects the narrow spectral bandwidth of each grating channel to a correspondingly narrow audio-band signal via

4.4.

Very-High-Resolution Mode

When using the etalon in the nominal design with cavity finesse = 70 and cavity spacing of 26 cm, the order number $m_{\mathrm{FP}}$ will be several thousand (e.g., $m_{\mathrm{FP}}=4600$ at HD $112\mu \mathrm{m}$). Multiple orders will thus be coupled simultaneously to a single $R=300$ grating channel (e.g., $4600/300\sim 15$ at $112\mu \mathrm{m}$), and an FTS scan will be required to resolve these orders from one another. To fully ensure spectral purity, each step in a high-resolution etalon spectrum will, therefore, require an order-sorting FTS scan. This scan, however, need only be long enough to clearly resolve the orders from one another, that is it requires $R=2\times m_{\mathrm{FP}}$ or 10,000 at HD $112\mu \mathrm{m}$. This is about 1/4 of the total FTS range. Additionally, the etalon is not required to operate at wavelengths shorter than $100\mu \mathrm{m}$, so the fringe rate constraint described above only applies to $100\mu \mathrm{m}$. Thus, $v_{\mathrm{OPD},\max }$ for the order sorting operation is $5.3\mathrm{mm}/\mathrm{s}$, and the order-sorting scan requires $\sim 196\mathrm{s}$. A 20-position etalon scan will thus require 1.1 h, and a 140-position etalon scan covering a full free spectral range (the space between orders) to produce a Nyquist-sampled complete spectrum requires 140 positions, or nearly 8 h. A subject for phase-A study is the potential to reduce this time with a custom FTS scan, which uses many fewer (e.g., $\sim 60$) discrete positions sampled sparsely across the scan instead of a continuous or fully-sampled scan. This is expected work because the FTS scan only needs to discriminate among the few orders within a single $\mathcal{R}=300$ grating channel, so aliasing signals from much higher or lower frequencies, as is produced when large steps are taken in the FTS scan, would not couple to the detector.

5.1.

Readout Electronics

All three of the detector approaches for OSS (as well as for the Origins far-IR imager/polarimeter (FIP)) will employ an rf/microwave frequency-domain multiplexing readout in which each pixel is coupled to a narrow-band ($Q\sim {10}^5$) superconducting resonator. The warm electronics generates a comb of frequencies, typically one per pixel, and this comb interacts with the array and returns to the warm electronics. Each tone incurs a frequency or amplitude shift as it interacts with its corresponding detector pixel. The returning comb signal is amplified in the cold with two stages of HEMT (high-electron mobility transistor) amplifiers, one at 4 K and one at $\sim 35\mathrm{K}$. In the warm electronics, it is then digitized and demodulated to extract each frequency’s information. No fundamentally new algorithms or approaches are required; it is just a question of matching capability with the Origins system resources.

The readout system for OSS and FIP places demands on the Origins system in two ways: the total electrical power budget (for the processing of the readout datasteam in the warm spacecraft) and on the 4-K cooling budget (for both amplifier dissipation and conducted loads through the coaxial lines). At present, the 4-K dissipation of the readout system is the primary limitation on total pixel count. Up to $2\times$ more pixels could be accommodated if readout dissipation is reduced, and/or as the current 100% margin on cooling capability is relaxed as the design matures.

For the OSS point design, we adopted the case with the greatest system-level impact: TES bolometers with the microwave MUX.^59,60 Figure 6 shows the approach. The baseline resonator frequencies are 4 to 8 GHz, and the resonator spacing is 2 MHz. Each readout circuit, therefore, can process 2000 pixels, and when partitioned into integer readout circuits for the six arrays, the total number of circuits required for all six is 32. The KIDs and possibly QCDs will also likely operate at $\sim 2000\text{pixels}$ per circuit, but will likely use lower readout frequencies (e.g., 0.5 to 2 GHz). This could translate to lower required power dissipation in the cryogenic amplifiers, allowing more circuits and larger arrays.

Fig. 6

The readout system consists of 32 circuits, each carrying 4 GHz of information bandwidth. Detectors are coupled to microresonators, 2000 of which can be read out in each line. On the warm spacecraft side, electronics create the waveform that interacts with the array, and then digitizes and analyzes the return waveform to extract phase shifts that encode the optical power on the detector. This basic approach accommodates all viable detector technologies under the development for Origins.

The warm side approach is described more fully in the Origins Technology Development Plan. The baseline uses the new Xilinx rf-system on chip (SoC),⁶¹ which integrates the analog to digital converters (ADC) with programmable digital signal processing logic [floating point gate array (FPGA)-like] in a single chip based on 16-nm-gate complementary metal-oxide semiconductor transistors. A single chip provides eight channels, each capable of processing 2 GHz at 12 bits depth, and the associated waveform generation capability. When compared with our existing 500 MHz (0.5 GHz) single-channel FPGA-based systems now in use, the new rfSoC represents $\sim 8\times (2/0.5\mathrm{GHz})=32$ times the processing capability. Abaco released a board with this chip in late 2018. Initial power dissipation estimates for this board are 50 W at full utilization.

For the system design, OSS uses this power dissipation fiducial of 50 W per 16 GHz of information bandwidth for the ADC and digital spectrometer logic. This is a conservative estimate—while neither of these existing systems are flight-qualified, mixed-signal application-specific integrated circuits, which integrate ADCs and signal processing, have been developed for flight implementation of similar applications,⁶² and these systems typically have lower power dissipation than their programmable-logic counterparts.

5.2.

Detector Technology Development Plan

The heritage of the OSS detector system and its path to flight readiness is detailed in the Origins Technology Development Plan,⁶³ a supplement to the full report. In this, we propose a 7-year program beginning in 2021 designed to dovetail with the flight implementation program beginning around 2025. The program serves both OSS and the imager-polarimeter FIP. It will achieve technology readiness level (TRL)-5 (performance at representative array/multiplexing scale) in 2025 and TRL-6 (with environmental qualification) in 2027. The strategy is to pursue two technology paths, TES bolometers, and KIDs, and to also fund promising lower TRL technologies such as the QCDs with $\sim 20\%$ of the program budget. This approach is chosen to mitigate mission risk of inadequate detector development in time for mission implementation. Unlike a traditional grant-based program, we will implement a review and progress assessment program, so that a technology that meets Origins requirements can be chosen as early as possible. For example, if during the initial phases of the program, one of the lower-TRL technologies takes off and appears more likely to achieve the requirements than KIDs or TES bolometers, then resources can be redirected to maturing that technology.

5.3.

Data Rate

Far-IR direct detectors must be sampled continuously, and the OSS data rate is determined by the pixel count (60,000), sampling rate, and required bit depth. The detectors will be sampled at 200 Hz to support an electrical bandwidth of 100 Hz, a comfortable factor of 2 above the 50-Hz 3-dB frequency of the detector with its 3-ms time constant. 12 bits is sufficient to capture the detector’s dynamic range under which it is photon noise limited if a logarithmic scaling is used. The data rate to ground without any compression is thus $144\mathrm{Mbits}/\mathrm{s}$ (where $1\mathrm{Mbit}=\mathrm{1,000,000}\mathrm{bits}$), a figure which can be accommodated by the emerging optical communication systems. In the high-resolution modes, since only a small fraction of the array is used, the rate could be smaller, but we expect that the grating data may be valuable for other purposes during these observations (e.g., for monitoring background and taking potentially serendipitous measurements along the slit).

6.1.

Sensitivity Model

The relationship between the detector performance and the estimated on-sky sensitivity are similar to what has been obtained in ground-based grating spectrometers Z-Spec⁶⁴ and ZEUS,^35,65 providing confidence in the estimation. The model assumes a 5.9-m-diameter telescope with 72% aperture efficiency to a point source, and 25% total instrument transmission (in a single polarization) for the base low-resolution mode. Detectors are assumed to operate with $\mathrm{NEP}=3\times {10}^{-20}\mathrm{W}{\mathrm{Hz}}^{-1/2}$, a figure which has been demonstrated in the lab with the QCD. The line-detection sensitivity includes a factor of $\sqrt{N_{\mathrm{eff}}}$, the effective number of pixels used to optimally extract the line flux. For the OSS grating modules, which have detectors sampling and the spectrometer intrinsic resolving power comparable at $\mathcal{R}=300$, $N_{\mathrm{eff}}\sim 1.6$, this adds a factor of 1.26.

6.2.

Bright Sources and High-Resolution Modes

Further applications of the sensitivity model are shown in Fig. 8. Bright sources add shot noise and degrade the sensitivity in a straightforward manner. For spectroscopy of solar-system objects, OSS must permit observations of sources up to 5 Jy at $128\mu \mathrm{m}$ without saturation. This translates to a dynamic range from the Zodi background loading of 1000, a requirement on the detector system.

Fig. 8

(a) Sensitivity degradation of the base grating mode in the face of photon noise from bright sources. Also plotted for comparison are sensitivities possible with possible optimized heterodyne receivers operating near the quantum limit (dark green) and with $T_{\mathrm{sys}}=1000\mathrm{K}$ (brown), assuming a linewidth of $3\mathrm{km}/\mathrm{s}$. (b) The high-resolution mode sensitivities. The FTS has a sensitivity penalty due to on-source time and losses. The etalon compounds this with additional losses and the etalon must scan, a factor not included in these plots. The etalon is baselined to cover the 100- to $200\text{-}\mu \mathrm{m}$ range, with coverage over the full band as a goal, but achieving the etalon sensitivity beyond $200\mu \mathrm{m}$ will require careful attention to stray light from the instrument itself.

For the high-resolution modes, the estimated sensitivities include the additional signal loss (52%) and loading from the 4.5-K FTS optics, and another factor of 0.25 transmission through the etalon (49% due to Ohmic loss in the mirrors × an estimated 75% due to beam walk-off in the large cavity). The very-high-resolution mode requires stepping the etalon to create a full spectrum, an additional sensitivity factor that is not included in the curves in Fig. 8.

7.1.

OSS Thermal Design

The interferometer optics operate at the observatory-provided base temperature of 4.5 K, but the grating modules as well as the band separation elements (polarizing grid and dichroic filters) are all cooled further to ensure that their thermal emission does not impact sensitivity. In particular, the potentially lossy grid and dichroic filters before the grating slits (termed the band separation optics) must be cooled to below the microwave background temperature to ensure that they do not add background. The gratings modules themselves need to be below 0.9 K—at least at the long wavelengths—to ensure that the integrated power through the bandpass filter with the full native detector étendu ($\mathrm{A}\mathrm{\Omega }$) does not saturate the detector or degrade the noise equivalent power. These requirements are met with a continuous adiabatic demagnetization refrigerator (CADR), which stages from the observatory-provided 4.5-K system. A schematic is provided in Fig. 9. The CADR is adapted from the high-heritage Goddard Space Flight Center (GSFC) designs of the Hitomi ADR and subsequent technology advances.^66–68 It uses 7 salt pills to provide 3 continuously-cooled stages: at 1.5 K (for the polarizing grid and dichroic filters), 0.7 K (for the grating spectrometer enclosures), and 0.05 K (for the focal plane arrays themselves). A key advantage of the ADR approach is the high Carnot efficiency, and the CADR heat rejection is only about half of the 4.5-K budget even though it is designed to provide 100% lift margin (that is $2\times$ what is needed) at all of its actively cooled stages. Further information is provided in the Origins Technology Development Plan. A suspension system has been designed together with the cooling system; its schematic is included in Fig. 9. The band separation optics and the grating optics modules are suspended kinematically from 4.5 K with bipods made of titanium 15-3-3-3, a low-thermal-conductivity alloy. Although composites offer lower thermal conductance at these temperatures, the use of metal eliminates a potential source of water contamination curing cool down. For the grating optics modules cooled by the 0.7 K stage, the titanium supports couple directly from 4.5 K mechanically, but heat is intercepted at the strut midpoints by the 1.6-K stage. Strut dimensions are tuned for each band’s mass, ensuring a resonant frequency that satisfies a standard mass acceleration curve. Inside each grating module, the focal plane assemblies are also mounted kinematically with Ti-15-3-3-3 bipods, in this case designed to provide resonant frequencies of at least 100 Hz.

Fig. 9

OSS thermal architecture. (a) The seven-stage continuous ADR is actually a five-stage ADR, similar to that used in the FIP instrument, and a two stage ADR. The ADRs are controlled from a single ADR controller on the warm side of the spacecraft. (b) Low-thermal-conductivity titanium bipods mount the spectrometer boxes from the optical bench, and the focal plane assemblies inside the spectrometers. These comfortably support launch loads, yet enable the 50-mK cryogenic system; no special provisions are required to survive launch. Though numerous, the cooling straps from the ADR are the size of large gauge copper wires and weigh a combined total of 1 kg.

Table 4 summarizes the calculated heat load and available lift with the cooler for each of the three OSS cooling stages. The dominant load on the sub-4K OSS-cooled stages is the conducted loads through the mechanical suspension. Wiring parasitics are negligible because superconducting cables are used. The 4.5-K situation is more challenging relative to the adopted allocation from the observatory, as this stage must accept the loads from the nonsuperconducting harnesses, and 4.5 K houses the first stage low-noise amplifier. The 0.38-mW per amplifier × 31 amplifiers correspond to commercially available devices operating at up to 8 GHz, but the team expects improvements in amplifier power dissipation will be possible by the time Origins is built. A straightforward descope to ease this aspect is to not operate all six bands simultaneously. (Parasitic conduction in harnesses is book-kept at observatory level, but is much smaller than the amplifier dissipation.) The design includes thermal straps made of pure annealed copper for all stages, sized for the load, and distance required. These copper straps along with the indium used for thermal connections total $\sim 8\mathrm{kg}$ for OSS.

Table 4

OSS cryogenic stages: thermal loads and available lift

7.2.

Mechanical Approach

Each of the OSS grating modules and the interferometer bench are discrete optomechanical elements that will be fabricated and tested independently and then mounted kinematically in the full instrument. They can be internally aligned and are subjected to environmental testing as units before being integrated into the full instrument. Aluminum spectrometer optical elements (mirrors and gratings) are the baseline, but the structural benches are baselined as beryllium to increase stiffness and reduce mass.

7.3.

Mechanisms

OSS has four mechanisms tabulated in Table 5 with their travel and accuracy requirements Me1 is the FTS interferometer scan mechanism that moves the rooftop mirror carriage. The required accuracy is driven by ensuring that in a given sample in the interferogram, the power uncertainty due to OPD error is less than the uncertainty due to the other noise sources (only photon noise in the optimal case). Quantitative assessment requires simulations, as was done for Herschel SPIRE,^69,70 a subject for phase-A study. But a key point is that for random errors in OPD, the required accuracy scales as $\lambda$ and the ratio $(\mathrm{NEP}/\sqrt{\tau })/P$. With its lower background, the latter ratio is more than $4000\times$ greater for OSS than for SPIRE for a given sample time. We, therefore, expect random errors will be much less stringent for OSS than for SPIRE, even with the shorter wavelengths. Pending the simulation work, we adopt for the purposes of the study a value of $\lambda /4$ in OPD, which is more stringent than the scaling would suggest. This translates to $0.8\mu \mathrm{m}$ in physical displacement with the $8\times$ path folding. In addition to verifying this, a key subject for phase-A simulation work is to assess the requirements on nonrandom (systematic) errors. These will have to be controlled to a greater level to avoid impacting long observations. The angular runout requirement is given by the need to maintain good pupil overlap in the interferometer over its full scan length.

Table 5

OSS mechanism requirements

Me2 is for the interferometer insertion; it simply moves the carriage containing the pickoff and reinsertion mirrors into the beam. Position accuracy is not a strong constraint, but this carriage must have an orientation accurate to 20 arc sec to ensure the interferometer pupil matches the telescope pupil. Me3 scans the etalon, it needs only cover a couple × the free spectral range (FSR) at the longest wavelength ($\mathrm{FSR}={\lambda }_{\max }/2$), so a total of $600\mu \mathrm{m}$. However, it must provide translation knowledge to $0.2\mu \mathrm{m}$, $\sim 1/4$ of a resolution element at $112\mu \mathrm{m}$. Furthermore, it must maintain parallelism to 0.3 arc sec to ensure that the cavity finesse is not impacted. Finally, Me4 allows the etalon assembly to be inserted and removed from the interferometer beam. Here, the positional accuracy is not critical, but the etalon must be oriented to within 10 arc sec once inserted to avoid beam walk-off in the etalon. The FTS and etalon scan mechanisms and the associated lubricants are specified for a minimum of 10 years and one million cycles, ample to accommodate any conceivable combination of observations over Origins’ mission life. The insertion mechanisms for the FTS pickoff and etalon are specified for 2000 cycles, a figure which is likely ample, but which should be revisited in phase-A as observing scenarios become clearer.

Detailed mechanism designs will be undertaken in phase A, but Fig. 10 shows an approach that meets the FTS translation requirements and was used in the study. This dual-stage parallelogram flex-band type system has essentially no friction. The pictured prototype was developed at NASA Goddard⁷¹ for use in Herschel SPIRE;⁶⁹ it will be scaled up by a factor of 3 to achieves the required OSS linear travel of 300 mm. A superconducting linear servo motor has the required precision and low losses allowing for a dissipation of $<1\mathrm{mW}$. For the etalon, a piezoelectric inchworm actuator will be used. Position, to the accuracy required, will be read out with either a modified Kaman DIT⁷² or a modified Mad City Labs NanoAlign-3;⁷³ both product families have some flight heritage. The launch locks will be actuated using typical nonexplosive actuators such as the NEA 9100. For the etalon and FTS insertion mechanisms, which are used infrequently, a four-bar linkage connected to a brushless dc motor is used.

Fig. 10

Flex type mechanism for use with the OSS FTS. This mechanism is free of friction, can be operated in any orientation, and has a stroke of 10 mm. For OSS, this would be scaled up by a factor of 3.

9.1.

Optical Alignment

The grating modules can be aligned readily either by the vendor or by instrument team members—the elements within each grating module should be aligned sufficiently just from machining tolerances. A good example of this approach is the aluminum grating spectrometer modules of the Spitzer Infrared Spectrograph (IRS), which had even tighter tolerances due to the shorter wavelengths. An aspect not used on Spitzer, but potentially useful for OSS is the fact that the grating module optics, along with the large, field-filling arrays, allows adjustment of the slit position by several beams with respect to the rest of the optics to obtain a desired slit position relative to the grating modules box. Each slit opening is an independent piece that can be adjusted before being clamped into position. The OSS requires that the six grating modules be positioned so that their slits overlap (field alignment), and that they each have their pupil aligned with the telescope pupil (pupil alignment). Meeting the first requirement consists of simply moving the grating module (or, for small adjustments, moving the slit openings) so that the slits align in position, using a telescope simulator which produces a point-source image. The second requires tipping/tilting the grating modules about the slit positions so that they point to a common pupil in the simulator. These tests will have to be performed with a cryogenic mounting of the grating modules.

Alignment of the interferometer begins with a self-alignment ensuring that the collimated beam remains centered as it traverses the mirrors and retroreflectors, and that the two sides of the interferometer recombine with good overlap. This should be possible to do warm if good practices are used in the machining and assembly of the FTS bench (uniform materials and thermal stress relieving), but it must be verified cold. Aligning the interferometer with the grating system requires positioning the interferometer insertion mirrors to provide an extremely well-collimated beam to permit the etalon operation. Here, the distance along the optical axis from the telescope is important, and this may require a cryogenic test. The interferometer must then have its pupil aligned with the telescope pupil; this consists of tilting the interferometer about the insertion mirror. Finally, the reinjection mirror must then be adjusted to ensure the interferometer’s pupil aligns with those of the gratings. The team expects precision machining of the interferometer insertion/reinjection mirror pair as a single carriage will result in having no need to make adjustments. Ultimately, a cold test must be done to validate the results of these alignment activities. The two key success criteria are: (a) transmission efficiency from the OSS front focus (corresponding to the telescope focus) through the interferometer to the detectors, likely with a chopped thermal load viewed through a small aperture and (b) spectral purity (especially for the etalon) verified with a narrowband source such as a far-IR laser or THz multipler source.

With this full system aligned in the flight truss, which is brought as a unit to the telescope, a full alignment check of all elements should not be required on the observatory. Simply verifying pupil overlap (also known as pupil alignment) of the grating modules with the observatory should be sufficient since the Origins telescope has sufficient actuation authority to adjust the telescope boresight and focus depth with respect to the instrument using actuation at both the secondary (6 degrees of freedom) and FSM (2 degrees of freedom).

9.2.

Test Facilities

OSS will require cryogenic test facilities, but they will be straightforward and pose no particular technical challenges. The detector arrays and individual grating modules will require a refrigerator with 2-K cooling of volumes on order $40\times 40\times 50{\mathrm{cm}}^3$, with cold fingers allowing the arrays to cool to 50 mK. This will verify basic sensitivity and stray-light rejection at the individual module level. Such systems are available commercially. Examples include the closed-cycle dilution refrigerators made by BlueFors, Oxford, and Leiden Cryogenics. For the multi-spectrometer alignment, a larger cryogenic test is required, simultaneously encompassing 1.5 m to accommodate the optical bench and at least three spectrometers. However, for this test, the full structure only needs to be cooled to $\sim 30\mathrm{K}$, as this is sufficient to incur all coefficient of thermal expansion related alignment effects. Cold fingers are required at 250 mK to allow the detectors to operate in a high-background mode and verify alignment. Full sensitivity is not required. These tests can be carried out in one of the existing thermal-vacuum chambers at JPL or GSFC.

9.3.

Calibration

In considering calibration uncertainties for measurements where the radiometric signal to noise is sufficient, we expect to achieve $\le 1\%$ relative measurement accuracy and $\le 5\%$ absolute flux calibration. Every detector pixel in the OSS spectrometer modules will have four key measurements generated in integration and test in a dedicated low-background spectrometer testbed that can house any of the grating modules. (1) Sensitivity will be measured with a temperature-controlled slit-filling load (blackbody source). (2) Spectral response described below. (3) Optical efficiency is obtained from the response as the slit-filling load is elevated in temperature, by comparing blackbody curves integrated over the measured bandpasses. (4) Beam shape is measured via an external (to the testbed) chopped warm source rastered in the far field of the grating modules. This is essentially verification to reveal any problems with grating module optics so they can be fixed ahead of full integration. Tests 2 and 4 will couple the grating module to outside the testbed, so will require substantial blocking filters to stay within the dynamic range of the OSS detector arrays—this can be accomplished with a combination of dielectric blockers and resonant mesh filters.

As for Herschel and Spitzer, radiometric scientific calibration is ultimately tied to astronomical sources. Given our experience with Herschel, we do not believe an on-board calibration source is needed, and therefore, the baseline OSS design does not include one. However, we will study this question further in phase A.

A key aspect unique to the grating spectrometer is the need for accurate measurement of the spectral response of every detector. This may reveal low-level grating ghosts and other peculiarities that do not impact sensitivity, but could be important in deep observations of sources with a large spectral dynamic range (very bright lines and faint lines or faint lines on a bright continuum). We will measure spectral profiles with a dynamic range of at least 1000 and a resolving power of at least $3000\times (100\mu \mathrm{m}/\lambda )$ using a long-path FTS, with the spectral scale anchored by spectral standards such as absorption-cell measurements and a far-IR laser. Once the OSS interferometer is integrated with the instrument, these profiles can be verified with the OSS FTS.

We do not expect the wavelength calibration and spectral profiles to change from the lab to flight, since there are not moving parts, but the on-board FTS will continue to be a benefit for OSS through the mission, as it will provide long-term tracking of the grating system spectral response in the unlikely event that something changes in a grating module. The etalon wavelength scale will be established during integration with a far-IR laser and/or a narrow-band THz source and confirmed and periodically monitored on orbit with astronomical line-emitting spectral standards observed on Herschel and Stratospheric Observatory for Far-Infrared Astronomy (SOFIA) such as IRC+10216.