Experimental validation of LISA interferometric performances

A century after the formulation of General Relativity by Albert Einstein, gravitational wave astronomy has opened a new window on the Universe. It allows the study of systems such as black hole binaries, until now imperceptible with electromagnetic and neutrino astronomy. Current terrestrial detectors observe on average one source per week by measuring tiny distances fluctuations using laser interferometry. Their sensitivity is mainly limited at low frequencies by seismic noise, which restricts detection to sources emitting above 10 Hz. By constituting the first space based detector of gravitational waves, the Laser Interferometer Space Antenna (LISA) mission will supplement these observations by reaching a new frequency band going from 10^-4 to 10^-1 Hz. The instrument, consisting of 3 satellites, will form a triangular constellation in heliocentric orbit. It will detect distance fluctuations of the order of a picometer between masses in free fall distant from 2:5 million kilometers. Various noises will impact these interferometric measurements and more particularly the laser frequency noise which will be dominant by several orders of magnitude. In order to achieve the required sensitivity for gravitational wave detection, noise reduction methods are being tested using computer and instrumental simulations. Although an onboard stabilization of the laser sources, a numerical method is necessary for the laser frequency noise reduction. An algorithm called Time Delay Interferometry (TDI) was formulated in the late 90's. It allows to obtain in post-processing a data set free of this dominant noise by synthesizing an interferometer with equal arms. For several years and with the support of CNES (Centre National d'Etudes Spatiales), the electro-optical simulator LISA On Table (LOT) has been used to generate realistic LISA signals in order to test TDI at APC laboratory (Astroparticule et Cosmologie). With respect to LISA architecture, it makes interfere signals delayed one from the other. The LOT having its own sources of noise, a first part of this thesis work consisted in qualifying then improving the electronic part of the instrument. The simulations carried out were able to demonstrate the TDI efficiency in reducing laser frequency noise to the required level by simulating constellations with fixed and flexible arms. In addition, the LOT also tests a phasemeter prototype for LISA. Our studies have shown a coupling between the TDI algorithm and the phasemeter anti-aliasing filters, leading to the appearance of additional non-stationary noise. A complementary computer simulation based on an analytical model allows us to estimate the impact of this new coupling for the LISA case. (author)