Researchers looked for brief gravitational‑wave signals coming from the directions of high‑energy neutrinos recorded by the IceCube Neutrino Observatory while Advanced LIGO, Advanced Virgo and KAGRA were taking data in the third observing run (O3). They used a targeted, unmodeled search and found no statistically significant gravitational‑wave counterpart. Instead, they set lower limits on how far a possible gravitational‑wave source would have to be for each emission model they tested.

The team used the X‑Pipeline software to search data in a small time window of ±500 seconds around each IceCube neutrino trigger. The search covered frequencies between 20 and 500 hertz, a band that includes many expected signals such as those from merging neutron stars. By restricting the search to the known sky direction of each neutrino and a short time window, the analysis is more sensitive than a blind all‑sky search. Background noise was estimated by analyzing many off‑source time segments and by time‑shifting detector data to build a distribution of loud noise events.

Because the team did not assume a single specific waveform, they tuned the search to several representative source types. These included compact binary inspirals such as binary neutron star (BNS) and neutron star‑black hole (NSBH) mergers, with neutron star masses sampled around 1.4 solar masses and black hole masses sampled around 10 solar masses, and generic burst signals. One generic family used “sine‑Gaussian chirplets,” which are short, oscillating waveforms that model a burst of gravitational waves from a rapidly changing mass distribution. The search sensitivity was measured by injecting simulated signals and finding the amplitude recovered in 90% of trials; these numbers were then converted into distance limits for each model.