Titolo della tesi: Innovative strategies to streamline continuous gravitational-wave candidates in the advanced LIGO-Virgo detector era
The goal of the LIGO-Virgo community is the gravitational-wave (GW) search and the consequent achievement of a deep knowledge of the universe. Although great results have been already obtained by detecting coalescing compact objects (such as binary Black Holes and even a Binary Neutron star), a lot of work remains still to do, especially for what concerns Continuous GWs (CWs). Those signals, due to their long duration, have a central role in the study of the Universe, but are fainter than the signals we have detected so far. This Ph.D. thesis is focused on finding innovative strategies to improve CW searches in the advanced LIGO-Virgo detector era.
The most accredited CW sources are non-axisymmetric rotating neutron stars. In general, there are four different approaches to study these signals: targeted, narrow band, directed and all-sky searches. The targeted search is a search where we assume to know all information on the source (sky position, spin frequency and its derivatives). The narrow band search widens the portion of frequency and its derivatives, working on an interval centered at the known information. The directed search considers stars for which the sky position is known, but no information on the star spin frequency and derivatives is provided. In these cases, we can use a matched-filter technique: in fact, having at most two parameters to span, it is possible perform a matched-filter search changing the parameters at a reasonable computational cost. In all-sky searches we are interested to look for CWs coming from unknown sources. In this case we cannot use the matched-filter technique due to the large size of the parameter space that must be spanned. Hence, we need to use hierarchical searches, such as the FrequencyHough algorithm, which is the core of the Ph.D. work. In the FrequencyHough search, the number of candidates obtained from the procedure is too high to be followed up, so it is necessary to rank them and deeply follow up only the most significant ones. However, it is not trivial to apply the procedure over a big portion of the frequencies we need to study as the computational cost grows linearly with frequency.
Our goal is to identify CWs in the LIGO-Virgo data and estimate their parameters (sky position, frequency and first time derivative). To this purpose, we have analyzed candidates returned by the FrequencyHough pipeline to find a method to filter out the bulk of them. We have found out that any (injected) signal creates, besides a candidate with the parameters closest to those of the injected CW signal, a set of byproducts for which a small error in one of the parameters is compensated by small errors on the others according to some rules. Those byproducts are recognizable thanks to the detection statistic that tends to be higher than the background-noise candidates. Hence, the rules of error compensation create a correlation in the parameter space that links the sky position to the frequency and its time derivative. Once we have identified those rules, and the patterns that they create, we have developed a veto procedure that uses those patterns to identify the presence of a CW signal. If we don’t have a pattern of candidates in the parameter space, it is unlikely that we have a signal strong enough to be successfully detected in the follow up step. The innovative procedure we developed consists of computing two additional Hough transforms on the set of initial candidates, check whether any correlation is present in the search parameter space and, alternatively, remove the candidates. The overall procedure removes about the 50% of false candidates (without excluding those related to the injected signals).