Stability comes through first
The clearest result from an early engineering test of the RISTRETTO spectrograph was its short-term stability: 2.8 m/s over four hours, within a stated 10 m/s requirement. Calibration measurements also showed a trend of about 5 m/s over a month. Fluctuations over a few minutes were highly correlated across the seven fibres, and their source was not yet resolved.
The result matters because the project targets visible reflected-light detection of exoplanets, especially Proxima b. RISTRETTO is a diffraction-limited, high-resolution spectrograph built around seven spatial channels, or spaxels, designed for a resolution of 135,000 and operated under vacuum with thermal control.
Built for repeatable alignment
The team tested the assembled subsystem in the laboratory and during initial on-sky observations. For the OHP run, the PAPYRUS fibre link fed the instrument, and first spectra were obtained one day after arrival.
The optomechanics used aluminium parts with canoe-sphere interfaces for reproducible alignment. Positioning repeatability was reported at 10 to 15 microns, while angular reproducibility was in the arcminute range.
What the optical tests found
Laboratory measurements covered wavelengths from 617 to 860 nanometres. At 635 nm, the best-performing fibre had efficiency greater than 37 per cent, with a reported uncertainty of plus or minus 7 per cent. The paper treats that figure as a lower-bound measurement rather than a final efficiency result.
A similar prototype IFU, a unit used to feed light into the spectrograph, was estimated at about 62 per cent efficiency on its central fibre and 54 to 60 per cent on the others. Those values were not confirmed for the IFU used at OHP. During a one-hour exposure, a first on-sky estimate indicated that at least 10 per cent of the light was injected into the spectrograph across all seven spaxels, but the injection efficiency was not yet fully established.
Detector maps gave a median full width at half maximum, a measure of the recorded light profile’s width, of 2.178 pixels and a median spectral resolution of 164,269.94 across 218 measured points. That median resolution was above the design value of 135,000, but the recorded profile varied across the detector and was generally too small and under-sampled. Detector focus and tilt had not been fully optimised.
Calibration and transport still matter
The preliminary wavelength solution, the calibration used to map detector positions to wavelengths, was in line with the design expectation at the centre of an order, with a 5.6 per cent difference between the experimental and design curves. The paper points to tolerances in camera focal length and grating blaze angle as possible explanations, while noting that the full solution was not complete.
Transport exposed another piece of unfinished work. The first spectra came one day after the instrument reached OHP, but after the return trip the team found a roughly 20-pixel misalignment in the dispersion direction and a small defocus. The report says these issues were to be studied and corrected in the next phase.
The thermal and vacuum tests were broadly in line with the operating expectations reported for the instrument. Optics temperature stability was better than 1.4 millikelvin in March 2026. Detector stability was about 1 mK rms when idle but roughly 50 mK rms with the system online; noise in the detector thermal sensor affected that online measurement. Pressure stayed below 10^-4 mbar and reached 6 times 10^-5 mbar over a month.
White-light tests found two series of in-focus ghost orders from the field lens and detector, with intensities below 0.5 per cent. The data-reduction pipeline was still under construction, so the reported performance remained preliminary.
A subsystem, not an exoplanet result
Taken together, these measurements describe the readiness of one instrument subsystem, not a detection of an exoplanet or its reflected light. The project targets visible reflected-light detection, but the evidence here is limited to laboratory characterization and initial on-sky tests of the assembled spectrograph.
At the time of the preprint, the subsystem was described as mostly complete, with further tests and improvements planned through the end of 2026. Those plans include improving detector alignment, tracking the few-minute stability fluctuations, confirming the actual OHP IFU efficiency, completing the wavelength solution and finishing the data-reduction software.
Paper data and sources
Original title: RISTRETTO: Assembly and Testing of the Seven-spaxel, High-resolution, Diffraction-limited Spectrograph
Authors: Leonie Hoerner, Bruno Chazelas, Nathanaël Restori et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text