COLIBRI (SVOM/FM-GFT): Instrumentation and Performances on the SVOM Alerts

Abstract: COLIBRI, the French Mexican Ground Followup Telescope (FM GFT) for SVOM, is a 1.3 meter rapid response optical facility specifically developed for prompt, multiband observations of GRB afterglows and for delivering subarcsecond localisations of optical counterparts for detailed followup studies. The telescope operates through a fully automated system that manages the entire workflow, from alert reception to counterpart identification. Commissioning results confirm that the telescope meets design specifications, and this paper presents a comprehensive performance assessment of the capabilities.

• The paper presents COLIBRI, a 1.3-m robotic optical/NIR telescope designed to localize SVOM transients within minutes and support rapid photometric-redshift measurements.
• Ten months of operations show approximately 60-second mean response times, follow-up of 71% of SVOM alerts, and counterpart detections for 65–71% of valid GRB triggers.
• The facility’s main capabilities remain incomplete: current image quality is 0.87″, above the 0.75″ requirement, while CAGIRE’s NIR channel and fast dome rotation await commissioning, leaving high-redshift results simulation-based.

COLIBRÍ, the French–Mexican Ground Follow-up Telescope (FM-GFT) for the SVOM mission, is a 1.3-m robotic optical/NIR facility at the Observatorio Astronómico Nacional (OAN) in the Sierra de San Pedro Mártir, Baja California, Mexico. This paper presents a comprehensive description of its instrumentation and a first assessment of its performance on SVOM alerts during roughly ten months of scientific operations beginning in January 2025 (2604.24259).

Context within SVOM

SVOM, launched in June 2024 from Xichang, detects GRBs with onboard localization precision of about 26 arcmin (ECLAIRs) and distributes alerts through a ground segment that refines positions to sub-arcsecond levels. The two robotic Ground Follow-up Telescopes—one managed by France and Mexico (COLIBRÍ), one by China—are tasked with providing sub-arcsecond optical localizations within 5 minutes of an alert, identifying low signal-to-noise triggers that do not prompt satellite slews (including very high-redshift candidates, $z>6$), detecting optically dark GRBs, and supplying rapid photometric redshift estimates to bridge to major facilities such as GTC, VLT, NOT, NTT, and NOEMMA.

The driving requirements are demanding: pointing delay under 20 seconds, localization better than 0.5 arcsec, three simultaneous photometric channels spanning 400–1800 nm ($Bgri$ + $zy$ in the visible; $JH$ in the NIR), a field of view covering the 26-arcmin ECLAIRs error box, and real-time data processing in under 5 minutes. Beyond SVOM, COLIBRÍ is positioned to respond to alerts from the broader multi-messenger ecosystem (VRO, CTA, SKA, KM3NeT/IceCube, Advanced LIGO/Virgo/KAGRA).

Site and infrastructure

The telescope sits at the highest point of OAN (~2940 m). The site's credentials are well established: median seeing around 0.8″ historically, ~80% usable nights, ~60% photometric nights—comparable to Mauna Kea or La Silla. The facility's own seeing monitor has since measured a median seeing of 0.72″, which the authors cite as evidence that OAN ranks among the best observing sites worldwide.

Thermal control received particular attention to preserve image quality: a low-emissivity bare-aluminum dome lined with 15 cm of polystyrene, insulated wall panels, twelve louvers for nocturnal ventilation, and strict isolation of the heat-producing control room, whose exhaust is vented ~60 m away perpendicular to prevailing winds. The design appears validated: delivered images at the start of night are typically 0.8–1.0″ FWHM in DDRAGO's channels, limited partly by the 0.38″/pixel sampling.

The compact 7.5 m dome required a custom rotation mechanism capable of 180° in 13 seconds at fast speed; commissioning of that fast mode was expected in winter 2025–2026. Electrical supply required frequency conversion (60 Hz site power to 50 Hz for the ASTELCO telescope) via a dedicated UPS chain. Because the compact dome leaves only ~2.1 m between column and building structure, the team developed bespoke handling tools—a mirror transport cart system, a dummy derotator for instrument balancing (derotators tolerate only 100 N·m torque), and a five-degree-of-freedom insertion cart for the 330 kg instrument—all manufactured at UNAM Ensenada.

Telescope and instruments

The alt-azimuth telescope, built by ASTELCO with mirrors figured by AstroOptique Cardoen (M1) and Winlight-Bertin (M2), uses unprotected pure-aluminum coatings recoated every two years with on-site tooling. Pointing delay is specified below 20 seconds with settle time under 1 second and absolute pointing accuracy below 2.5″ RMS.

DDRAGO is the primary wide-field optical imager, a "one-and-a-half" instrument whose optics also feed the NIR channel. Eleven lenses, two dichroics, a corrector plate, and three fold mirrors split the beam: $gri$ to a blue e2v 231-84 deep-depleted CCD, $zy$ to a red CCD (both 4k×4k, 0.38″/pixel, 25.9′ field), and $JH$ through the Warm Optical Bench (WOB) to CAGIRE. The WOB reimages from f/6.3 to f/3.7 and places the cold stop after all powered optics—an approach pioneered on RATIR but here with a faster beam and larger field, reducing cryostat complexity and cost. A notable supply-chain failure is documented candidly: the original WOB lenses from Trioptics had inadequate coatings and were never remedied; replacement lenses from Optimax were expected in early 2026, with CAGIRE integration deferred to end of 2026.

Measured on-axis image quality falls short of requirement at present: $d_{80}$ is 0.58″ for the instrument alone but 0.87″ combined with the telescope, against a requirement of 0.75″. The authors attribute the excess to M1 astigmatism and M1–M2 coma, with correction planned for spring 2026 and an expected final $d_{80}$ near 0.60″. Until that correction is demonstrated, the image-quality specification remains unmet.

CAGIRE is the NIR imager ($J$, $Bgri$0 bands) built around the Lynred ALFA 2k×2k HgCdTe sensor (loaned by ESA via LabEx FOCUS), operating at 100 K in Up-The-Ramp mode with 1.3 s full-frame reads, QE >70% over 0.8–2.1 μm, 22′×22′ field at 0.64″/pixel—sky-background-limited by design. Integration at IRAP was completed in summer 2025 with ESO's NGC controller; dark noise matches laboratory characterization, confirming thermal and electrical design quality. Its pre-processing software fits differential ramps per pixel, handling nonlinearity, cosmic rays, and saturation to extend effective dynamic range. Simulation studies combining an end-to-end camera model with sensor characterization indicate that four of the nine known $Bgri$1 GRBs from the past 20 years would have been detected, and artificially redshifting those events suggests afterglow detectability out to $Bgri$2, where the Lyman break exits the H band. These are simulation-based projections, not yet on-sky results.

Control system and pipeline

The TCS descends from RATIR/COATLI/DDOTI heritage code, migrated to language-neutral JSONH/JSON-RPC interfaces so new development proceeds in Python; interfacing the ASTELCO OpenTSI controller and Spectral Instruments CCD controllers each required ~1000 lines of new code. A web-based alert queue lets on-shift scientists update coordinates and filters in response to GCN Circulars.

The image-analysis pipeline—reduction, SExtractor-based alignment, SWarp stacking with iterative artifact masking, astrometry.net calibration, aperture photometry (2×FWHM), zero points from Pan-STARRS DR2 or SkyMapper DR4—runs end-to-end with an online mode delivering stacked, calibrated products and transient candidates within minutes. Transient identification flags unmatched sources and catalog sources with significant magnitude deviations; light curves use half-stack slopes. Photometric redshifts come from MCMC SED fitting of a single power-law afterglow model with four parameters ($Bgri$3, $Bgri$4, $Bgri$5, normalization), validated on simulations and archival data.

Performance on SVOM alerts

Over ten months starting January 2025, COLIBRÍ responded to approximately 140 GCNs (36% SVOM, 36% Swift, 22% Einstein Probe, remainder IceCube, CHIME, etc.). Key operational figures:

Metric	Value
SVOM alerts followed	~71% (77% excluding unobservable southern targets)
Expected follow-up fraction	~65%
GRB counterpart detection rate	65–71%; upper limits otherwise
Mean trigger-to-observation delay	~60 s

The 60-second response covers the entire chain—onboard trigger, VHF transmission, interruption of ongoing observations, and start of the alert sequence—which satisfies the sub-minute responsiveness the mission requires. Detection efficiency matches pre-launch expectations, supporting the claim that the automation delivers routine high-energy transient follow-up capability. Example science output includes deep simultaneous two-channel imaging of GRB 251013C reaching $Bgri$6 and $Bgri$7 (3σ) in 90×60 s coadded exposures.

Limitations and open items

Several caveats bear directly on the reported performance. First, the combined telescope-plus-DDRAGO image quality (0.87″ $Bgri$8) currently violates the 0.75″ requirement pending the spring 2026 wavefront correction; the projected 0.60″ is an expectation, not a measurement. Second, the NIR capability central to the high-redshift and dark-GRB science case is not yet operational: WOB lens delivery slipped due to the coating failure, and CAGIRE first light is scheduled only for end of 2026, so all high-$Bgri$9 detection figures remain simulation-derived. Third, the fast dome rotation mode had not yet been commissioned. Fourth, the detection-efficiency statistics rest on ten months of data and on verification of ECLAIRs trigger validity; seasonal weather losses account for most of the 23% missed fraction, and longer-baseline statistics will be needed to confirm stability. Finally, the second Nasmyth instrument, the TEQUILA polarimeter, remains in development.

Conclusion

This paper documents a fully automated 1.3-m rapid-response observatory that meets its core operational requirements on SVOM alerts: ~71% alert follow-up efficiency, 65–71% counterpart detection among valid triggers, and ~60 s mean response time, achieved with a modular Python/C/Bash pipeline producing calibrated photometry and transient candidates within minutes. The principal outstanding items are the telescope image-quality correction, delivery and integration of the WOB optics enabling CAGIRE's NIR channel, and commissioning of fast dome rotation—collectively targeted for completion by end of 2026, at which point the facility's full multiband, high-redshift capability can be assessed on sky rather than by simulation.