PhD defence by Tomek Baka 'Advancing data analysis with gravitational waves'

On July 6 2026, Tomek Baka has successfully defended his thesis. He accomplished his PhD at the Institute for Gravitational and Subatomic Physics (GRASP), Gravitational Waves Group. The defense has taken place in the Academiegebouw in Utrecht.

Summary PhD thesis

Advancing data analysis with gravitational waves: The effects of calibration, overlapping signals and dispersion

When compact objects, like black holes, collide, they emit gravitational waves. They can be detected on Earth by interferometers such as LIGO and Virgo. The raw readout of the detector (electronic signals) is converted to the corresponding contraction in spacetime, in a process known as calibration. It is subject to experimental uncertainties, which we record and take into account when analysing the observed signals. 

During the fourth observing run of the detectors, the LVK (LIGO-Virgo-Kagra) collaboration discovered that there was an error in the process, with the uncertainty applied with an opposite sign. Here, we developed a method to reweight the old result to the correct calibration, bypassing the need for time-consuming reruns. We find only negligible effects on the final results. General relativity predicts that gravitational waves always travel at the speed of light. As a consequence, there is no dispersion---the signal emitted at the source has the same shape as the signal detected at the observer. By looking for dispersion in the signal, we can therefore test general relativity. We improve upon the previous tests of the dispersion and apply it to the new data from the fourth observing run. We find no evidence of any general relativity violations and, in particular, no evidence of non-zero graviton mass. 

The next generation of gravitational wave detectors, like Einstein Telescope and Cosmic Explorer, is planned to produce a much higher volume of data than can currently be analysed. We will be detecting new signals constantly, frequently with multiple sources present at the same time, overlapping in the detectors. Analysing these kinds of signals with current techniques takes an unreasonable amount of time. We adapt an algorithm known as relative binning to make the analysis feasible. We also develop a process to predict if the presence of other signals affects the analysis, enabling us to know beforehand if we can use the current single-signal analysis techniques.

Supervisors: prof. dr. C.F.F. Van Den Broeck and dr. S.E. Caudill