FOSSIL's preliminary thermal architecture

FOSSIL (FTS fOr CMB Spectral diStortIon expLoration) is a proposed ESA M8 mission designed to measure the spectral distortions monopole of the Cosmic Microwave Background (CMB) with a sensitivity three orders of magnitude beyond the COBE/FIRAS legacy measurement. Achieving this sensitivity demands an extraordinarily challenging cryogenic architecture: the scientific instrument must be maintained at 4.5 K, while the detector focal plane assembly operates at 50 mK. This paper presents an overview of the preliminary thermal architecture of the FOSSIL payload, from the spacecraft service module at 293 K down to the sub-kelvin detector stage. The design draws on heritage from the Planck and ARIEL missions and relies on a staged passive cooling chain comprising a multi-layer insulation blanket, three V-groove radiators (operating at approximately 130 K, 90 K, and 50 K), and a 25 K actively cooled shield fed by an ESA-provided 4 K mechanical cryocooler. Sub-kelvin temperatures are achieved via a multi-stage adiabatic demagnetisation refrigerator (ADR) developed for NewAthena/X-IFU, providing continuous cooling at 1.8 K and 350 mK, and 50 mK with an 80% duty cycle. The Focal Plane Assembly (FPA), housing four Kinetic Inductance Detector (KID) arrays at 50 mK, is thermally isolated from the 4.5 K bench via a carbon-fibre reinforced polymer (CFRP) hexapod with staged heat interception. The paper outlines the staged cooling concept and shows that the architecture closes with positive thermal margins at every stage; the detailed steady-state thermal budget will be presented in a forthcoming dedicated paper.

The thermal requirements flow directly from the science objectives. The fundamental measurement principle is a differential comparison between sky emission and an actively cooled blackbody internal reference (BBIR) using the FTS. To suppress instrumental self-emission, the optimal instrument temperature is as close as possible to the CMB monopole temperature; a practical upper limit of 4.5 K is adopted for the instrument stage, enabled by a 4 K mechanical cryocooler currently under development at ESA. The detector units must operate at 50 mK for two reasons: the KID materials must operate well below their superconducting critical temperature Tc (for the LF band, the absorbing resonators require Tc ∼ 0.7 K, achieved through the proximity effect in a tri-layer Al-Ti-Au thin film; operating at T ≲ Tc/10 ensures a stable superconducting regime), and achieving the photon-noise-limited regime requires that the detector NEPdet ≲ 3 × 10−17 W Hz−1/2 remain at least one order of magnitude below the mean photon noise. State-of-the-art KIDs have demonstrated NEPdet ∼ 3 × 10−20 W Hz−1/2, comfortably exceeding this requirement.

The thermal architecture follows a layered approach, with each cooling stage progressively reducing the load intercepted by the following one, mirroring the approach employed on Planck. The architecture maximises the benefit of the L2 orbit, which provides an exceptionally stable thermal environment with a tightly constrained solar aspect angle of approximately ±10°. The heritage of Planck demonstrates the thermal stability achievable at L2, where the HFI bolometer plate was actively stabilised below 20 nK Hz−1/2 in the science band and the 4 K optical stage met a 10 µK Hz−1/2 requirement.

The passive cooling chain draws directly on the Planck heritage, whose three V-groove radiators and MLI architecture demonstrated purely passive cooling from the SVM temperature down to ∼50 K at L2. A 20-layer MLI blanket covering the SVM top surface forms the first thermal barrier between the service module at ≈293 K and the payload. Three V-groove radiators, inclined at increasing angles with respect to the spin axis, then progressively intercept and re-radiate heat toward deep space, reaching equilibrium temperatures of approximately 130 K, 90 K and 50 K. The panels are modelled as aluminium honeycomb sandwich structures following the ARIEL heritage, with a surface treatment strategy (VDA hot faces, low-emittance cold faces, high-emissivity space-facing surfaces) validated on Planck. The structural support struts between the SVM and each V-groove stage are made of GFRP G-10, with heat intercepted at each stage via collars and braided straps. On top of the passive stack sits the 25 K shield, actively cooled by the ESA-provided 4 K cooler chain. It provides a further interception stage before the 4.5 K instrument and thermally pre-conditions the Low Noise Amplifiers (LNAs) of the KID readout, which operate between 4 and 30 K. The structural support between the SVM, the 25 K shield and the 4.5 K bench is provided by a hexapod of six CFRP struts, sharing the ARIEL payload support heritage.

The FOSSIL instrument is mounted on a 4.5 K aluminium bench surrounded by a light-tight Aluminium baffle. The bench houses all optical and quasi-optical elements: the two off-axis dual-mirror telescopes (primary 420 mm, secondary 200 mm), the Martin-Puplett FTS with its scanning mirror mechanism, the BBIR, the two beam-switching mechanisms (BSMs), the dichroic beam splitters, and the FPA. Cooling the full enclosure to 4.5 K minimises the instrumental self-emission, reducing the differential signal between the instrument and the sky to a level manageable by the BBIR-based calibration scheme. The dominant heat sources at the 4.5 K stage are the FTS scanning mirror mechanism, achieved through a stiffness-compensated reaction-less design; the two BSMs, building on cryogenic wheel-mechanism heritage from ISO; the ADR heat sink during re-magnetisation; the SVM harnesses; and parasitic radiative and conductive loads from the 25 K shield. The total stage load remains comfortably within the 4 K cooler capacity.

Cooling below 4.5 K is provided by a multi-stage ADR derived from the design developed for the NewAthena/X-IFU instrument, also benefiting from SPICA/SAFARI and LiteBIRD developments. It provides three successive stages: a T2 continuous stage at 1.8 K, which intercepts the dominant parasitic loads from the FPA support and maintains the BBIR within its 2.5–2.9 K range; a T1 continuous stage at 350 mK for intermediate interception; and a T0 single-shot stage reaching 50 mK, reaching 50 mK of continuous science operations during 28 with a duty cycle exceeding 80%. This stage directly cools the detector units via a copper thermal strap. The demonstrated thermal stability of the X-IFU ADR provides ample margin for stable KID operation at 50 mK.

The FPA, whose multi-stage thermal isolation draws on the philosophy demonstrated on Planck HFI and recent structural and thermal model developments, houses four detector units (two Low-Frequency and two High-Frequency Detector Units), each consisting of a smoothwall multimoded feedhorn coupled to a segmented KID array. It is built around an isostatic hexapod surrounded by a 1.8 K radiation shield. Maintaining all four detector units at 50 mK from the 4.5 K bench requires very effective thermal isolation through a multi-stage interception scheme. The primary structural supports between the 4.5 K bench and the 1.8 K interface are CFRP T700 struts, whose low thermal conductivity minimises the conductive load reaching the cold stages. The struts are split at the 1.8 K and 320 mK stages, with thermal collars and braided straps at each intercept point, preventing continuous conductive heat propagation from 4.5 K to 50 mK. Below 320 mK, two complementary phenomena provide additional passive isolation: all structural interfaces use Al-6061-T6, which becomes superconducting below 1.1 K and whose thermal conductivity then drops dramatically; and at the 320 mK–50 mK boundary, the Kapitza boundary resistance further reduces parasitic loads on the coldest stage. Copper thermal straps redirect heat from the structural supports to the ADR cold stages.

In science observation modes, one FTS input arm is directed at the sky while the other is directed at the BBIR, which must represent a known, stable, nearly perfect blackbody emission at temperatures bracketing T0 ≈ 2.7255 K, with setpoints spanning 2.5–2.9 K. The assembly comprises a primary absorbing region actively controlled over 2.5–2.9 K and a secondary high-frequency calibration cavity, independently heated to ∼20 K during dedicated sequences. The driving thermal requirements are a temperature stability of a few µK over the FTS scan duration (∼2 s), and a limited parasitic load to the 1.8 K ADR stage. The BBIR is thermally isolated from the 4.5 K bench via insulating struts, with an Al-6061-T6 substrate to minimise gradients. The absorbing surface design draws on heritage from ARCADE and the MetOp-SG MWI calibration target. Temperature is monitored by Cernox thermometers, with an absolute thermometry uncertainty of ∆T0 ≈ 100 µK from ground calibration, sufficient for FOSSIL’s measurement goals.

A preliminary steady-state thermal model has been developed for the full chain, from the SVM panel at 293 K down to the 50 mK detector stage, accounting for the dominant radiative and conductive loads across each interception stage. At this early design stage the model retains only first-order contributions and conservative assumptions; it is intended to verify overall feasibility rather than to provide consolidated figures. The passive chain provides comfortable margins at the 130 K, 90 K and 50 K V-grooves, which can be further tuned through geometric adjustments of the V-groove surface area and inclination angles. On the active side, the 25 K shield, 4.5 K bench, and the 1.8 K, 320 mK and 50 mK ADR stages all close with positive margins against their respective cooling capacities. The key outcome at this preliminary stage of the design is that the FOSSIL thermal architecture closes consistently, with positive margins at every temperature level, confirming its compatibility with the mission requirements.

The preliminary thermal architecture of FOSSIL addresses an extraordinary engineering challenge: maintaining a complex optical instrument at 4.5 K and its detectors at 50 mK, while preserving a blackbody calibration reference stable to a few µK at ∼2.7 K, all within the resource constraints of an ESA medium-class mission. The baseline design demonstrates that this challenge is addressable through a staged passive cooling chain (MLI + 3 V-grooves + 25 K active shield) drawing on validated Planck and ARIEL heritage; an ESA-provided 4 K cryocooler bridging the SVM temperature to 4.5 K; a multi-stage ADR developed for NewAthena/X-IFU providing 1.8 K, 350 mK and 50 mK cooling; and a CFRP hexapod FPA support exploiting the superconducting transition of Al-6061-T6 and the low conductivity of CFRP to minimise thermal conductance at the coldest stages. While the present architecture remains a baseline concept to be consolidated and optimised during Phase A, it benefits from a clear TRL development path: the multi-stage ADR will reach TRL-6 through NewAthena/X-IFU, and future instruments (BISOU, TMS, COSMO) will further validate key sub-systems ahead of FOSSIL Phase A, targeting a launch in the early 2040s.

Improvements to AI Systems Based on This Paper:

1. Cryogenic Thermal Architecture Optimization AI

• Improvement: Train an AI model on the staged cooling chain (MLI, V-grooves, active shields, ADR stages) to automatically optimize thermal budgets for future space missions.

• Capability: The AI can generate and validate multi-stage thermal designs (e.g., for L2 orbit payloads) by simulating radiative/conductive loads, adjusting V-groove geometries, and predicting margins—reducing manual iteration time from months to hours.

2. Superconducting Material Selection and Thermal Isolation AI

• Improvement: Incorporate the physics of Al-6061-T6 superconducting transition (Tc 1.1 K) and Kapitza resistance into a machine learning model for material selection at sub-kelvin stages.

• Capability: The AI can recommend optimal structural materials, strut geometries, and thermal intercept placements for any detector stage (e.g., 50 mK) by predicting thermal conductivity drops and parasitic load reductions, enabling faster design of ultra-low-temperature instruments.

3. Blackbody Calibration Reference (BBIR) Stability Controller

• Improvement: Use the BBIR’s requirement (temperature stability of a few µK over 2 s, absolute uncertainty 100 µK) to train a reinforcement learning agent for active temperature control.

• Capability: The AI can autonomously maintain sub-µK stability in real-time, compensating for thermal drifts from ADR cycling or FTS scanning, and self-calibrate using Cernox thermometer feedback—applicable to any precision radiometry or spectroscopy system.

4. Multi-Stage Cryocooler and ADR Duty-Cycle Optimizer

• Improvement: Model the 4 K mechanical cryocooler and multi-stage ADR (1.8 K, 350 mK, 50 mK with 80% duty cycle) as a constrained scheduling problem.

• Capability: An AI scheduler can optimize ADR re-magnetization timing, heat sink loads, and cooler power allocation to maximize science observation time, while predicting thermal transients and preventing duty-cycle violations—useful for future X-IFU or LiteBIRD-like missions.

5. Passive Cooling Chain Design Generator (V-Groove + MLI)

• Improvement: Train a generative AI on Planck/ARIEL heritage data (V-groove temperatures 130 K, 90 K, 50 K; MLI layer counts) to propose novel passive cooling architectures.

• Capability: The AI can generate and evaluate thousands of V-groove configurations (angles, surface areas, emissivities) for any orbit or spacecraft bus, outputting Pareto-optimal designs that minimize mass while ensuring positive thermal margins—accelerating Phase A concept studies.

6. Thermal Margin Prediction and Risk Assessment AI

• Improvement: Use the paper’s preliminary steady-state thermal model results (positive margins at every stage) to train a predictive AI for early-stage mission feasibility.

• Capability: The AI can ingest incomplete design parameters (e.g., harness loads, mechanism heat dissipation) and output probabilistic margin forecasts, flagging high-risk stages and suggesting design tweaks—enabling faster, more reliable thermal feasibility studies for ESA M-class proposals.

7. Focal Plane Assembly (FPA) Structural-Thermal Co-Design AI

• Improvement: Combine the CFRP hexapod design (with staged heat interception) and superconducting thermal breaks into a multi-physics AI co-optimizer.

• Capability: The AI can simultaneously optimize structural stiffness (to survive launch) and thermal isolation (to minimize 50 mK loads), using topology optimization and surrogate models—yielding lighter, more thermally efficient detector mounts for KID arrays or similar cryogenic sensors.

8. Mission-Level Thermal Stability Forecasting AI

• Improvement: Leverage L2 orbit’s stable environment (±10° solar aspect) and Planck’s demonstrated stability (20 nK Hz−1/2) to train a time-series AI for predicting long-term thermal drift.

• Capability: The AI can forecast thermal noise and stability over multi-year missions, enabling autonomous adjustment of calibration sequences (e.g., BBIR setpoints) and improving the sensitivity of CMB spectral distortion measurements—directly enhancing scientific yield.

What the improved AI system can do overall:

It can serve as an end-to-end design and operations assistant for ultra-cryogenic space instruments, automatically generating, validating, and optimizing thermal architectures from 293 K to 50 mK, while ensuring stability, margins, and duty cycles—reducing human design effort by orders of magnitude and enabling faster, more ambitious missions like FOSSIL.

FOSSIL (FTS fOr CMB Spectral diStortIon expLoration) is a proposed ESA M8 mission tailored to measure the spectral distortions monopole of the Cosmic Microwave Background (CMB) with a sensitivity three orders of magnitude beyond the COBE/FIRAS legacy measurement. Achieving this sensitivity demands an extraordinarily challenging cryogenic architecture: the scientific instrument must be maintained at 4.5 K, while the detector focal plane assembly operates at 50 mK. This paper presents an overview of the preliminary thermal architecture of the FOSSIL payload, from the spacecraft service module at 293 K down to the sub-kelvin detector stage. The design draws on heritage from the Planck and ARIEL missions and relies on a staged passive cooling chain comprising a multi-layer insulation blanket, three V-groove radiators (operating at approximately 130 K, 90 K, and 50 K), and a 25 K actively cooled shield fed by an ESA-provided 4 K mechanical cryocooler. Sub-kelvin temperatures are achieved via a multi-stage adiabatic demagnetisation refrigerator (ADR) developed for NewAthena/X-IFU, providing continuous cooling at 1.8 K and 350 mK, and 50 mK with an 80% duty cycle. The Focal Plane Assembly (FPA), housing four Kinetic Inductance Detector (KID) arrays at 50 mK, is thermally isolated from the 4.5 K bench via a carbon-fibre reinforced polymer (CFRP) hexapod with staged heat interception. We outline the staged cooling concept and show that the architecture closes with positive thermal margins at every stage; the detailed steady-state thermal budget will be presented in a forthcoming dedicated paper.