Estimation of intrinsic fast radio burst width and scattering distributions from CRAFT data

The intrinsic width and scattering distributions of fast radio bursts (FRBs) inform on their emission mechanism and local environment, and act as a source of detection bias and, hence, an obfuscating factor when performing FRB population and cosmological studies. Here, we utilise a sample of 29 FRBs with measured high-time-resolution properties and known redshift, which were detected using the Australian Square Kilometre Array Pathfinder (ASKAP) by the Commensal Real-time ASKAP Fast Transients Survey (CRAFT), to model these distributions. Using this sample, we estimate the completeness bias of intrinsic width and scattering measurements and fit the underlying, de-biased distributions in the host rest-frame. In no case do our model fits prefer a down-turn at high values of the intrinsic distributions of either parameter in the 0.01–40 ms range probed by the data. Rather, when assuming a spectral scattering index of $\alpha = -4$ , we find that the intrinsic scattering distribution at 1 GHz is consistent with a log-uniform distribution above 0.04 ms and that this functional form is strongly favoured over the lognormal descriptions used by previous works. We also find that the intrinsic width distribution rises as a Gaussian in log-space in the 0.03 – 0.3 ms range, with a log-uniform distribution above that slightly preferred to a lognormal distribution. This confirms previous works suggesting that FRB observations are currently strongly width- and scattering-limited, and we encourage FRB searches to be extended to higher values of time-width. It also implies a bias in FRB host galaxy studies, although the form of that bias is uncertain. Finally, we find that our updated width and scattering models – when implemented in the zDM code – produces $\sim$ 10% more FRBs at redshift $z=1$ than at $z=0$ when compared to alternative width/scattering models, highlighting that these factors are important to understand when performing FRB population modelling.