Precision cosmology with LSST : Developement of an unbiased cosmic shear estimator measured on galaxy images

Overview & Synopsis

The discovery of the accelerating expansion of the Universe in 1998 turned our vision of the physical phenomena governing its evolution on its head. The cause of this acceleration, which opposes the expected action of gravity, is called "dark energy" by cosmologists, and its origin remains undetermined to this day. Several cosmological probes can be used to constrain dark energy, including cosmic shear, the result of a weak gravitational lensing effect on background galaxies caused by large-scale structures. Sensitive to both the matter distribution and the geometry of the Universe, it can be used as a powerful tool to constrain the equation of state for dark energy. The next decade marks the dawn of the era of precision cosmology, with the development of next-generation surveys such as LSST. This telescope at the Vera C. Rubin Observatory is the first ground-based survey designed for weak gravitational lensing analyses, thanks in particular to its 3200 megapixel camera and 8-meter primary mirror, which will image the sky with both unrivalled depth and spatial coverage. After 10 years, LSST will have observed several billion galaxies, enabling unprecedented cosmological measurements to be made, particularly with cosmic shear, which requires statistical detection on a large number of sources. However, shear measurement is complex and associated with many sources of systematics, including calibration of the galaxy shape estimator, and shot noise in the images. Considering LSST's expected statistics, the various sources of systematic error need to be controlled to the per mil level in order to achieve the desired accuracy on the dark energy equation of state. This is the background to the results presented in this thesis. The first part of this work was to develop an unbiased estimator of gravitational shear directly from galaxy images. The advantages of this new method compared with state-of-the-art methods such as Metacalibration are manifold: no assumptions are made about the source profile (the shape measurement is based on the second moments of the light distribution) and no distortion is applied to the galaxy image, making it possible both to avoid introducing correlated noise into the estimations and to perform shear measurements on undersampled images. The second part of this work focuses on the correction of noise bias, caused by shot noise in the images. Noise introduces a bias in the shape measurement, and therefore also biases the shear estimations. As the usual methods for calibrating this bias are based on simulations, they are subject to certain limitations, notably the assumptions chosen for image modelling or the consequent computation time required to carry out Monte Carlo loops. To overcome these limitations, we have turned our attention to developing an analytical correction for noise bias by calculating the second-order derivatives of the shape estimator. This method gives very satisfactory results for galaxy ellipticity correction, as well as promising results for shear calibration, but requires further investigation.