Gamma-ray burst progenitors revisited

Recently, several long-duration gamma-ray bursts (GRBs) associated with kilonovae have cast doubt on the traditional, dichotomous mapping between gamma-ray duration and progenitor system. Here, we investigate the rates and properties of bursts which appear to cross this dichotomy using a sample of GRBs for which progenitor constraints are possible. We first build a sample of known Swift-detected GRBs at z<0.3, finding 8 short- and 21 long-duration GRBs. Of these long GRBs, we find 9 bursts with deep limits on supernova emission, evidence for kilonova emission, or association with a quiescent galaxy (31% $\pm$ 9% of all GRBs at z<0.3), implying that a significant fraction of nearby long GRBs likely do not come from massive stars. At z<0.3, no short GRB has an observed supernova counterpart. We find comparable numbers when expanding to z<0.5 and other gamma-ray telescopes, though we obtain a decreased fraction of bursts with robust constraints on a progenitor. We further find that the long GRBs with no associated supernovae possess on-average fainter afterglows and lie in less star-forming host galaxies than those with supernovae, supporting that these events may originate in compact object mergers. We estimate approximate volumetric rates, finding similar (on-axis) rates for short GRBs and supernova-less long GRBs of $\sim 0.5-2.5$ Gpc$^{-3}$ yr$^{-1}$, although a search for possible low-redshift hosts of the complete Swift catalog suggests that our sample may be $\sim$50% complete. If supernova-less long GRBs arise from compact object mergers, this implies that $\sim$ 30-70% of all z<0.3 Swift long GRBs may arise from mergers and that the $z<0.3$ rates of mergers from long and short GRBs are comparable. These findings hold substantial implications for gravitational-wave coincidence and heavy element enrichment.