Conquering the Winner's Curse for fast radio bursts
6 November 2017
CAASTRO astronomers have done the sums again on mysterious cosmic radio bursts, finding that they may have been more common earlier in the Universe’s history.
Fast radio bursts – millisecond blips of radio waves – were discovered in 2007. We still don’t know what they are. Suggestions have ranged from neutron stars imploding to a propulsion system for alien spacecraft!
Fortunately, we can examine the origins of FRBs just by measuring how the number of bursts (N) varies with their apparent brightness (S).
If the bursts came from relatively nearby galaxies, N would have a specific relationship to S: it would follow a power law with an index of –3/2 (because the volume of space increases as distance to the power 3 and the brightness decreases to the power 2). An index of –3/2 suggests that the bursts are distributed in a manner that’s called ‘Euclidean’.
However, the power-law index could be much larger than –3/2. If so, that would imply that the bursts probably originated further away, and that the rate at which they occur has changed markedly over the lifetime of the Universe. Such a distribution would be ‘non-Euclidean’.
The relationship between N and S has been a hot topic among astronomers studying FRBs, and so CAASTRO Advisory Board member Ron Ekers (CSIRO) and CAASTRO Associate Investigator Jean-Pierre Macquart (ICRAR/Curtin University) decided to take a fresh look at it.
They found that existing estimates of the relationship have been strongly influenced by the extreme brightness of the very first FRB discovered, the so-called Lorimer Burst. The Lorimer Burst is an example of discovery bias, also known as the Winner’s Curse, in which the first detected instance of a new phenomenon is often highly unrepresentative of its underlying population. Removing the Lorimer Burst from the population statistics makes a large difference to estimates of the N-S relationship.
Macquart and Ekers re-analysed the FRB population using data from CSIRO's Parkes telescope, which has found more than half of the known FRBs. All these detections were made using the telescope’s 13-beam receiver, which looks in 13 different directions simultaneously.
The brighter an FRB is, the more likely it is to show up in more than one of the telescope’s ‘beams’, and so the fraction of multiple-beam detections to single-beam detections directly measures the ratio of extremely bright bursts to fainter bursts. A previous analysis based on this ratio had suggested that the distribution is much shallower than the Euclidean value. However, Macquart and Ekers found that it is much easier to detect fainter FRBs in multiple beams than had been supposed.
To make a new estimate, Macquart and Ekers applied techniques developed in the 1970s to measure the distribution of quasars in space. They found that the most likely N-S relationship for FRBs was steeper than the Euclidean value. This suggests that the FRBs come from far off in the Universe and that, like quasars, they were more common earlier in the life of the Universe.
The Parkes dataset is the best available to date but Macquart and Ekers’ work has highlighted some of the difficulties of interpreting it. Future telescopes such as the Australian SKA Pathfinder (ASKAP) will be free of such problems and, with their very large fields of view, will detect more of the very bright FRBs, which we need to really pin down the distribution.
Jean-Pierre Macquart & Ron Ekers, "FRB event rate counts I – Interpreting the observations". Accepted for publication in Monthly Notices of the Royal Astronomical Society. Preprint on arXiv.