Optical development of the BISOU breadboard

BISOU (Balloon Interferometer for Spectral Observations of the primordial Universe) is an astronomical balloon-borne pathfinder developed as part of a preparatory study for a future space mission aimed at measuring spectral distortions of the cosmic microwave background (CMB). A laboratory breadboard of the instrument is being developed at the Institut d’Astrophysique Spatiale (IAS), enabling the characterization of subsystems and instrument systematic effects, particularly in the optical system. The optical system is based on a differential polarizing Fourier Transform Spectrometer (FTS) that receives inputs from both a sky-facing telescope and an internal calibration source. The FTS focal planes include sub-K detectors coupled to multimode feed horns. The full spectral band, spanning between 90 and 1500 GHz, is sub-divided into two frequency sub-bands, thanks to the use of a dichroic. The optical analysis first relies on ray-tracing simulations to establish the overall configuration of the system, before proceeding to more advanced Gaussian beam and physical optics analyses.

The measurement of the CMB spectral distortions requires an absolute comparison between the sky and a well-known reference. For this purpose, the instrument concept is based on a polarised Martin-Puplett Fourier Transform Spectrometer (FTS) with two inputs. One arm of the FTS views the sky while the other looks at an internal blackbody reference maintained at the temperature of the CMB (2.7 K). The beams are split and recombined using wire-grid polarisers, and a scanning mirror introduces an optical path difference (OPD) δ between the two inputs to produce an interferogram. The outputs of the FTS are coupled to detectors through multimode feedhorns. The measured signal consists of a constant part and a scan-modulated part that is proportional to the difference between the two inputs. Applying an Inverse Fourier Transform and adding or subtracting the known spectrum of the reference allows for sky spectrum retrieval. Finally, a dichroic is added before the focal plane to split the wide frequency range into two sub-bands, improving the sensitivity of the instrument and resulting in low frequency detection units (LFDU) and high frequency detection units (HFDU).

With a frequency range from 90 GHz to 1.5 THz and a spectral resolution of 15 GHz, the BISOU instrument design follows the measurement principle but is adapted to meet the constraints of a balloon-borne experiment. The presence of residual atmosphere at its float altitude (40 km) requires the instrument to be enclosed in a vacuum-tight cryostat in order to maintain it at a very low temperature T∼2.7 K. The optical entrance aperture will be closed by a window, transparent to millimeter and sub-millimeter waves, which, at this altitude, will be at a temperature of ∼ 270 K. Therefore, the thermal emission of this warm optical component will dominate the photon noise. Thermal filters located on each thermal shield will limit the thermal load on the cryogenic stages of the instrument and on the detectors. While one arm of the FTS is looking at the sky through a telescope, the other arm is facing the internal reference directly. This asymmetry in the optical path needs to be studied since the additional elements and reflections introduced by the telescope can add cross-polarisation effects and aberrations that will not be present in the internal reference arm. To improve the sensitivity of the instrument, the wide frequency band is split into two sub-bands using a dichroic filter with a cut-off frequency of about 300 GHz. Frequencies below this threshold are transmitted through the dichroic, while higher frequencies are reflected. This flux division significantly reduces the optical power incident on the low frequency band detectors, lowering the photon noise. This is an important consideration for a balloon-borne FTS instrument, where high-frequency emissions are dominated by the thermal emission of warm instrument components, together with the residual atmosphere. The resulting reduction in photon noise directly translates into an improved sensitivity in the low-frequency channel, where the specific spectral distortion signature needs to be measured.

The breadboard model aims to study systematic effects and validate technologies prior to their integration into the BISOU and FOSSIL designs. The BBM is therefore designed to be highly reconfigurable, allowing each component of the instrument to be characterized. The FTS will be placed on a 2 K cold plate inside a cryogenic test facility in order to test the components at the temperature at which they will operate in flight. The cryogenic chamber environment is composed of different shields at 300 K, 50 K and 4 K. In order to reach the operating temperature of the detectors, the focal plane will be cooled to 100 mK or 50 mK (depending on the chosen technology) by a dilution refrigerator from Bluefors (model SD250). Integrated within the breadboard cryostat facility, it provides 250 µW of cooling power at 100 mK and can reach temperatures as low as 30 mK, which is more than sufficient. The 2 K cold plate is cooled by the 30 mW of cooling power available on the still stage of the SD250. The 2 K baffle is thermally connected to the cold plate. The 4 K shield and the 50 K shield are cooled by a Cryomech PT415-RM (1.35 W at 4.2 K and 40 W at 45 K). The FTS motor will be located on a 4 K plate attached beneath the 2 K cold plate via insulating supports (G-10) and cooled by the 4 K stage. A key feature of the cryostat will be its 200 mm diameter window on the 300 K vacuum can. Filters will be placed on the 50 K and 4 K thermal shields, allowing us to study their systematic effects on the measurements. This aperture will allow us to couple the cryostat to an atmospheric chamber as an input in order to understand how the atmospheric variations affect the measurements. It will also allow us to use the cryogenic facility for the calibration of the final BISOU instrument. In addition, the controlled thermal environment will enable the study of systematic effects induced by thermal gradients within the optical components.

The optical design of BISOU relies on three complementary approaches. Ray tracing, performed with Zemax software, provides fast geometrical modelling to select mirror configurations and satisfy all design constraints. Gaussian beam optics then enables analytical estimation of the beam size at each optical element, which is used to optimise mirror dimensions, their illumination, and define edge taper targets. Full physical optics simulations are carried out with GRASP by propagating Maxwell’s equations through the design, capturing diffraction effects and beam deformation, which become critical at the lowest frequencies where the beam is widest. The FTS consists of five pairs of mirrors, numbered from M1 to M5, guiding the beam through polarisers. On the telescope side, mirrors M1 to M4 are (M3 to M6), while on the reference side, only M2 to M4 are ellipsoidal. M1 remains a parabolic mirror as it is facing the internal calibrator reference. M3 mirrors are the moving element, translated by the motor. The dichroic is one of the most critical components of the design. It is placed in a collimated beam in order to avoid any spurious effect which might arise from incident rays reaching the dichroic surface with different incidence angles. For this purpose, the last FTS mirror M5 has been changed from an elliptical mirror, to a parabolic mirror. The detectors are required to operate at sub-K temperature, so in order to optimise the thermo-mechanical design the detection units (detectors and their associated feedhorns) must be co-located within the same area to form a single Focal Plane. This requirement is achieved through the use of the M6 mirrors, re-directing and focusing the beams onto the feedhorns. Wire grid polarisers are located in between the path of the FTS mirrors. The wires reflect the parallel component of the incident beam while transmitting the perpendicular one. The polarisers are used to combine the beams from the two sources: the sky and the internal reference. Polarisers A and D are oriented with their wires vertical (0°), while polarisers B and C are rotated by 45°. The telescope follows the Mizuguchi-Dragone condition in its off-axis Cassegrain configuration, minimizing cross-polarization and astigmatism induced by reflections on the mirrors. The design was first established through ray tracing, then refined using Gaussian beam propagation to meet the-20 dB edge taper requirement on the 150 mm diameter primary reflector. The design presented corresponds to the current warm model configuration. Compared to the BISOU instrument and the cold breadboard, both including four Detector Units (DUs), the warm model is reduced to two DUs. This simplified configuration allows us to first characterize the optics and perform the optical alignment more easily, as the feedhorns lie in the FTS plane. In the 4-DUs configuration, the feedhorns and their associated detectors are arranged in a two-level layout, placing them outside the FTS plane. The focal plane is designed to be modular, enabling different configurations to be tested.

Additional work is planned in the near future. Gaussian beam simulations of the full instrument will be carried out to further optimize the optical design. For instance, specific studies on the de-pointing, the aberrations and the spillover induced by the movement of the M3 moving mirror will be investigated, even though, the analysis carried out on the BISOU instrument showed that moving M3 with a translation of four times the real mirror stroke does not introduce major impacts on the beam shape. The Gaussian beam simulation will then be improved by taking into account the multimode behaviour of the feedhorn to model the effective beam shape of the instrument. Moreover, special attention will be paid to the accurate calculation of the optical path difference (OPD), a critical parameter in the analysis and reconstruction of the data from the interferogram. Indeed, any systematic error in the OPD determination will introduce artificial distortions in the reconstructed spectrum. The construction of the warm breadboard model is expected to be completed by the end of 2026, marking a key milestone toward the validation of the instrument concept.

In conclusion, the paper presents the optical design of the BISOU breadboard model, developed at IAS as a critical step toward the realisation of the BISOU stratospheric balloon project and the future FOSSIL space mission. The measurement concept, based on a differential polarising Martin-Puplett FTS with two inputs, has been described alongside the cryogenic test facility that will enable the characterisation of the instrument at its operating temperature. The warm breadboard model, currently under construction, will allow for the validation the optical design and alignment procedures before the cold model is assembled.

Improvements to AI Systems Based on This Paper:

1. AI for Optical System Design Optimization

• Train a generative AI model on ray-tracing, Gaussian beam, and physical optics simulation data (from Zemax, GRASP) to automatically propose mirror configurations (e.g., M1–M6) that minimize aberrations, cross-polarization, and spillover.

• The improved AI can rapidly iterate over thousands of design variants, optimizing edge taper (e.g., -20 dB on primary reflector) and beam size at each element, reducing manual trial-and-error.

2. AI for Systematic Error Prediction and Mitigation in FTS

• Develop a machine learning model that predicts systematic effects (e.g., optical path difference errors, thermal gradients, dichroic phase shifts) from simulated interferograms and known component tolerances.

• The AI can then suggest calibration corrections or design adjustments (e.g., mirror stroke compensation, polarizer alignment) to minimize artificial spectral distortions, improving the fidelity of CMB spectral distortion measurements.

3. AI for Multimode Beam Propagation Modeling

• Create a neural network surrogate that learns the multimode behavior of feedhorns and their coupling to the FTS optics, replacing computationally expensive physical optics simulations.

• This AI can quickly predict effective beam shapes, spillover, and mode conversion for different feedhorn geometries, enabling faster optimization of focal plane layouts (e.g., 2-DU vs. 4-DU configurations).

4. AI for Cryogenic System Thermal Management

• Use reinforcement learning to optimize the thermal control of the cryostat (e.g., cooling power allocation between 2 K, 4 K, 50 K stages, and detector temperatures) under varying atmospheric loads and instrument operating conditions.

• The AI can autonomously adjust filter placements and shield temperatures to minimize photon noise from warm components (e.g., 270 K window) while maintaining detector stability at sub-K temperatures.

5. AI for Interferogram Data Reconstruction

• Implement a deep learning-based inverse Fourier transform that accounts for non-ideal OPD sampling, detector nonlinearities, and dichroic frequency-dependent losses.

• The improved AI can reconstruct sky spectra with higher accuracy from noisy interferograms, automatically separating sky signal from reference and residual atmospheric emission, even with incomplete or irregular OPD sampling.

6. AI for Automated Alignment and Calibration

• Train a computer vision system on optical alignment data (e.g., from the warm breadboard model) to detect misalignments in mirror positions, polarizer angles, or focal plane placement.

• The AI can provide real-time feedback during assembly, reducing alignment time and improving repeatability, as well as predict the impact of misalignments on final spectral quality.

7. AI for Atmospheric Emission Subtraction

• Develop a generative model that simulates residual atmospheric emission at 40 km altitude (temperature, pressure, water vapor) and its coupling to the instrument.

• The AI can then be used to subtract atmospheric contributions from measured interferograms in real time, improving the sensitivity of low-frequency channels where the CMB distortion signal resides.

8. AI for Design Reconfiguration and Modular Testing

• Build a meta-learning system that, given the breadboard’s reconfigurable design, automatically suggests optimal test configurations (e.g., which components to characterize, which filters to place) to maximize information gain about systematic effects.

• This AI can plan experiments to validate technologies for BISOU and FOSSIL, reducing the number of required test cycles.

BISOU (Balloon Interferometer for Spectral Observations of the primordial Universe) is an astronomical balloon-borne pathfinder developed as part of a preparatory study for a future space mission aimed at measuring spectral distortions of the cosmic microwave background (CMB). A laboratory breadboard of the instrument is being developed at the Institut d'Astrophysique Spatiale (IAS), enabling the characterization of subsystems and instrument systematic effects, particularly in the optical system. The optical system is based on a differential polarizing Fourier Transform Spectrometer (FTS) that receives inputs from both a sky-facing telescope and an internal calibration source. The FTS focal planes include sub-K detectors coupled to multimode feed horns. The full spectral band, spanning between 90 and 1500 GHz, is sub-divided into two frequency sub-bands, thanks to the use of a dichroic. The optical analysis first relies on ray-tracing simulations to establish the overall configuration of the system, before proceeding to more advanced Gaussian beam and physical optics analyses.