Searching for electromagnetic counterparts of gravitational wave transients

A pioneering electromagnetic (EM) observation follow-up program of candidate gravitational wave (GW) triggers has been performed, Dec 17 2009 to Jan 8 2010 and Sep 4 to Oct 20 2010, during the recent LIGO/Virgo run. The follow-up program involved ground-based and space EM facilities observing the sky at optical, X-ray and radio wavelengths. The joint GW/EM observation study requires the development of specific image analysis procedures able to discriminate the possible EM counterpart of GW trigger from background events. The paper shows an overview of the EM follow-up program and the developing image analysis procedures as they are applied to data collected with TAROT and Zadko.

💡 Research Summary

This paper reports on the first systematic electromagnetic (EM) follow‑up program to candidate gravitational‑wave (GW) events during the LIGO/Virgo science runs of late 2009 and 2010. Over two observing windows (Dec 17 2009–Jan 8 2010 and Sep 4–Oct 20 2010) a total of fourteen GW triggers were generated in near‑real time by the three‑detector network (the two LIGO interferometers and Virgo). Using a low‑latency pipeline (Omega, coherent WaveBurst, Multi‑Band Template Analysis) the triggers were uploaded to the Gravitational‑Wave Candidate Event Database (GraCEDb). The LUMIN and GEM software packages then selected statistically significant events, performed a rapid sky‑localisation based on arrival‑time differences, and produced a list of sky‑pointing positions within roughly ten minutes. Human experts validated the alerts, and if no detector‑issue was found the alerts were disseminated to partner observatories within about thirty minutes.

Because the typical sky‑error regions for low‑SNR triggers span tens of square degrees and often consist of several disconnected patches, the follow‑up strategy relied on weighting the probability map with an external galaxy catalogue (the Gravitational‑Wave Galaxy Catalog). Galaxies within 50 Mpc—the approximate horizon for binary neutron‑star mergers—were assigned weights proportional to their blue‑luminosity (a proxy for stellar mass) and inversely proportional to distance. This “galaxy‑targeted” approach reduced the required field coverage for telescopes with modest fields of view.

The EM facilities involved covered optical, X‑ray and radio wavelengths. In the optical band the robotic telescopes TAROT (two 25 cm instruments, 3.5 deg² field) and the 1 m Zadko telescope (0.17 deg² field) were used. For each GW alert the observing plan consisted of six consecutive 180 s exposures (TAROT) or 120 s exposures (Zadko) on the first night, repeated on the following three nights. TAROT reaches a limiting magnitude of ≈17.5 mag under ideal conditions, while Zadko can reach ≈20.5 mag.

A fully automated image‑analysis pipeline was developed to extract transient candidates from the large data volume. The steps are: (1) source extraction with SExtractor; (2) cross‑matching with USNO‑A2.0/B to discard known stars; (3) cross‑identification of objects present in multiple images to build light curves; (4) rejection of spurious detections (cosmic rays, asteroids) by requiring presence in at least four consecutive frames; (5) spatial filtering to retain only objects located within a radius four times the host galaxy’s major axis, thereby accounting for possible offsets (up to tens of kpc for GRB afterglows); (6) light‑curve analysis assuming a power‑law decay L ∝ t⁻ᵝ, which translates into a linear magnitude evolution m = 2.5 β log₁₀(t) + C. The “slope index” 2.5 β is computed for each candidate; a conservative threshold of 0.5 was adopted (based on Monte‑Carlo simulations) to select plausible EM counterparts. This filter efficiently removes most variable stars and active galactic nuclei while preserving the expected behaviour of GRB afterglows (slope index ≈2.5–3) and kilonovae (similar or slightly shallower).

Simulations injecting synthetic on‑axis GRB afterglows and kilonova light curves into real TAROT and Zadko images showed that, for a survey limiting magnitude of 15.5 mag, the pipeline can detect the majority of GRB afterglows out to the 50 Mpc GW horizon, and kilonovae out to ≈15 Mpc. The estimated false‑positive rate is low; the remaining contaminants after all cuts are mainly rapid Cepheid variables and a few active galactic nuclei whose light curves mimic a shallow decay.

During the two observing periods, nine of the fourteen GW alerts resulted in usable optical images. No definitive EM counterpart was identified, but the exercise validated the end‑to‑end workflow—from low‑latency GW trigger generation, through galaxy‑targeted sky‑tiling, to automated transient identification. The authors argue that with the ten‑fold sensitivity improvement expected from Advanced LIGO/Virgo, the volume probed will increase by a factor of ~10³, dramatically raising the probability of detecting both GW signals and their EM counterparts. In that regime, rapid EM follow‑up will be essential for precise localisation, host‑galaxy identification, redshift measurement, and for exploiting multimessenger observations to constrain the physics of compact‑object mergers and to provide an independent probe of cosmology.