Date of Graduation


Document Type


Degree Type



Eberly College of Arts and Sciences


Physics and Astronomy

Committee Chair

Sarah Burke-Spolaor

Committee Co-Chair

Maura McLaughlin

Committee Member

Casey J. Law

Committee Member

Duncan R. Lorimer


Fast Radio Bursts (or FRBs) are millisecond-duration transients of extragalactic origin. They exhibit dispersion caused by propagation through an ionized medium, and quantified by Dispersion Measure (DM). Around 800 FRBs (24 repeaters) have been discovered; so far, 24 FRBs have been confidently associated with a host galaxy. In this thesis, we discuss multiple new FRB search and analysis techniques and the corresponding tools that enable us to search for FRBs harder, localize them better, and classify candidates faster.

We discuss five open-source software suites that can be used in FRB analysis. These suites are used to distinguish between FRBs and radio frequency interference (RFI), model FRB properties, search for periodic activity, calculate the probability of an association between an FRB and the host galaxy, and unify data processing across multiple data formats.

We then present a robust comparative analysis of clustering algorithms to group candidates from REALFAST transient search system at the Karl G. Jansky Very Large Array. We design a performance metric that optimizes for a few pure clusters, i.e., clusters with either astrophysical or noise candidates. We show that using sky location along with DM/time improves clustering performance, and propose a strategy that can be used to decide which clustering algorithm is most fit for a particular application.

We present a dense sample of bursts from the repeating FRB\,121102, discovered using our software. Using the Arecibo Telescope, we detected 133 bursts in 3~hours of data observed at 1.4\,GHz. We determine the properties of the bursts using robust spectro-temporal modeling. We find that the bursts are band-limited, with a lack of emission below 1.3\,GHz. We find the wait time distribution to be log-normal in form with a peak at 75\,s. Poissonian and Weibull distributions do not describe the burst rate distribution well. The cumulative energy distribution can be described using a broken power-law model, with the break at $(2.3\pm0.2)\times 10^{37}$~ergs and a high-energy slope of $-1.8\pm0.2$.

Motivated by the banded nature of FRB\,121102 bursts, we perform a simulation study to show that commonly used analyses of band-limited FRBs lead to observational biases. We show that all the observed shapes in the energy distributions of repeaters can be explained using these biases. We then recommend techniques to correct these biases: modeling burst spectra to robustly estimate the intrinsic properties, and using bursts that are within the observing band for energy distribution analyses.

Finally, we discuss the REALFAST search and analysis pipeline, compare it to the search pipelines on single-dish telescopes, and highlight the advantages of using an interferometer. Primarily, every detection with REALFAST comes with a precise localization that can be used to associate the FRB to a host galaxy. We then discuss five repeating FRBs that were localized using REALFAST.