Twenty-four thousand hours of GREENBURST observations with the GBT

In addition to fast radio burst (FRB) searches carried out using dedicated surveys, a number of radio observatories take advantage of commensal opportunities with large facilities in which observations for other projects can be searched for FRBs and other transient sources. We present the results from one such effort, the first 24,186 hours of the GREENBURST search for dispersed radio pulses with the Green Bank Telescope (GBT). To date, GREENBURST has detected a total of 50 pulsars and three FRBs. One of the pulsars, PSR J0039+5407, has a period of 2.2 s and was previously unknown. Using follow-up observations with the Canadian Hydrogen Intensity Mapping Experiment, we found a timing solution for this pulsar which shows it to have a characteristic age of 2 Myr. Additional GBT observations show the pulsar has a very high nulling fraction ($\sim70-80%$). All three of the FRBs are repeating sources that were previously known and were being monitored by the GBT as part of other projects. A major challenge for GREENBURST in the discovery of new FRBs is its single beam. This makes it hard to distinguish some of the pulses from sources of radio frequency interference. We highlight this problem with a case study of an FRB-like pulse that initially passed our interference filters. Upon closer inspection, the event appears to be part of a longer-duration narrow-band source of unknown origin. Further observations and monitoring are required to determine whether it is terrestrial or celestial.

💡 Research Summary

The GREENBURST project leverages commensal observing on the Green Bank Telescope (GBT) to search for dispersed radio transients. Between March 2019 and July 2024 the system recorded 24 186 hours of L‑band data (1.4 GHz) with 4096 frequency channels, 960 MHz bandwidth, 256 µs sampling, and 8‑bit precision. A real‑time pipeline searches dispersion measures from 10 to 10 000 pc cm⁻³ and pulse widths from 256 µs to 32.8 ms. Recent upgrades focus on robust radio‑frequency interference (RFI) mitigation: higher‑order statistical tests (kurtosis, skewness), Jarque‑Bera and D’Agostino K² normality metrics, and an inter‑quartile‑range (IQR) outlier detector applied on 16 ms and 1 s blocks. A first‑moment test flags non‑linear saturation events on 4.2 s scales. After statistical cleaning, a weighted zero‑DM subtraction (Lazarus et al. 2015) removes broadband RFI while preserving astrophysical dispersion signatures.

During the campaign, GREENBURST detected 16 556 individual pulses from 49 known pulsars, including bright single pulses from the eclipsing binary B1744‑24A, giant pulses from the millisecond pulsar B1937+21, and pulses from the double‑pulsar J0737‑3039A. The system also serendipitously recorded bursts from three known repeating FRBs (20190520B, 20200120E, 20121102A) while those sources were being observed for unrelated science, demonstrating the pipeline’s ability to capture FRB events in commensal data.

A major discovery was a previously unknown pulsar, PSR J0039+5407, first seen on 9 June 2022 with S/N 14 and DM 74.2 pc cm⁻³. Subsequent analysis identified eight additional low‑significance pulses, allowing a kernel‑density‑estimate (KDE) localization within the 9.2′ GBT beam and a period of 2.2 s via a fast‑folding algorithm. Follow‑up timing with CHIME/Pulsar (134 observations over 446 days) yielded a phase‑connected solution: characteristic age ≈2 Myr, and a remarkably high nulling fraction of 70–80 %. Nulling analysis on a dedicated GBT session confirmed intermittent emission consistent with known period‑age correlations.

The paper also presents a puzzling event, GBP 220718, detected on 18 July 2022 (DM 145.5 pc cm⁻³, 16 ms width, S/N 10). The signal was confined to the lower quarter of the band and, upon deeper inspection, appeared as part of a longer‑duration, narrow‑band emission of unknown origin. Initial classification as an FRB candidate was later retracted, highlighting the difficulty of distinguishing terrestrial interference from genuine celestial bursts when operating with a single‑pixel beam.

Overall, GREENBURST demonstrates that commensal observing can substantially increase the scientific return of large radio facilities, delivering both new pulsar discoveries and validation of FRB detection pipelines. However, the single‑beam architecture and limited frequency coverage constrain the ability to reject RFI and to localize transient events. The authors argue for future upgrades such as multi‑beam receivers, broader bandwidths, and more sophisticated machine‑learning classifiers to improve FRB discovery rates. Planned extensions aim to add another ~10 000 hours of observations through 2025, with continued refinement of real‑time RFI excision and deep‑learning candidate vetting, positioning GREENBURST as a key contributor to the growing inventory of fast radio transients.