Hierarchical data-driven modelling of binary black hole mergers

[eng]Throughout history, human beings have received and interpreted information from distant stars and galaxies through electromagnetic waves (light). Until 2015 this was the dominant way for observing astrophysical events happening in our cosmos. However, on September 14'th 2015 a new window to the universe was opened thanks to the rst direct gravitational wave detection, a goal pursued for several decades by the LIGO/Virgo scienti c collaboration. Gravitational waves are tiny space-time oscillations propagating at the speed of light. They are a prediction of the Einstein theory of gravity and we need the most catastrophic astrophysical events to detect them. The rst observation of gravitational waves described the inspiral, merger and ringdown of two black holes with 36 and 29 solar masses located at 1300 billion light-years, where about the 5% of the total mass was radiated as gravitational waves and becoming the most powerful astrophysical event ever observed. The event was called GW150914, consistently with its the arrival date and was publicly announced on February 11'th 2016 by the LIGO Virgo collaboration. This has not been the only event observed during this thesis project. Relying on statistical criteria arguments, we can certify the observation of one additional event also compatible with the coalescense of a pair of black holes tagged as GW151226 plus a third one called LVT151012 likely from astrophysical origin but that did not reach the statistical signi cance required to be con rmed. The coalescense of binary black hole systems are an optimal candidate for the observation and study of gravitational waves. The current observations suggest that these kind of events could dominate the future ground based detections. Then, we need to optimise the theoretical waveform models to characterise the future observations. In this thesis we have given the rst steps towards a new upgrading of the nonprecessing gravitational waves models. These models result from the matching of the well known post-Newtonian (PN) and e ective-one-body (EOB) analytic formulations to the computationally expensive numerical solutions of the Einstein equations. They are de ned in the frequency domain and depend on the ratio of the two black hole masses (mass-ratio) and some e ective spin e that results from the combination of the components of the spins orthogonal to the orbital plane thus reducing the physical parameter space to only two dimensions. Then, although this current prescription have been demonstrated to be su cient for the searches of the gravitational waves in the data, they are not so optimal for the statistical inference of the spins of each BH, which is partially caused by the inherent degeneracy introduced by the e ective spin. The focus of this work has been the extension of the one-spin phenomenological models to its two-spin version by adding the subdominant e ects carried by the spin di erence terms = 1 � 2. To that end, we have employed the data of more than 400 simulations of binary black hole systems generated by four di erent codes (BAM, SpEC, LAZEV, MAYA), 23 of them generated throughout this thesis by means of the BAM code. This involved the di cult task of evolving, extracting the waves and the data postprocessing of each case. Then, we have rede ned the strategy for building higher than two dimensional ansaetze to add subdominant e ects and where we have also included the results of the extreme mass ratio limit. All this analysis has resulted in the prescription of new phenomenological models for the nal mass, nal spin and peak luminosity. The new models have been shown to improve the old descriptions of these quantities while they have clearly revealed the possible impact of the subdominant e ects in the near future phenomenological models