During the last decade, cosmology has entered a precision era, leading to the prevalence of the standard cosmological model, ΛCDM. Nevertheless the main ingredient of this model, dark energy, while dominating the energy budget of our Universe, remains mysterious and its comprehension is the current Graal of the domain. The next generation of cosmological surveys among which Large Synoptic Survey Telescope (LSST) & Euclid starting now in only a couple years will provide us with an unprecedented volume of data allowing in particular to further down the constrains on the dark energy parameters. We would like to offer a PhD student to develop several analysis in the framework of the LSST Dark Energy Science Collaboration (DESC) with the APC LSST group.
The scope of this work is the weak lensing study. Weak gravitational lensing corresponds the deflection of light from distant sources due to the bending of space-time by matter along the line of sight. Its detection and use in a tomographic manner allows a measurement of the growth rate of cosmic structures, and therefore is sensitive to dark energy. Weak lensing will be one of the main probes to tackle the dark energy problem in the next decade (with strong lensing, SNIa, galaxy clusters, and large scale structures studies).
Detecting weak lensing relies on the characterisation of the shape of galaxies which will be used in the statistical analysis of their shear to derive the matter distribution, and further down the road constrain dark energy parameters. For deep surveys like LSST and Euclid, shape measurement is not as easy as it may seem as most galaxies overlap : their images are blended. Separating them is then an essential step before measuring shapes of the individual objects, and it strongly depends on the seeing of the instrument and colors observed. This very problem would be the first point we would like to investigate with a student. The problem can be addressed in a number of ways, but one promising direction we would like to investigate is the use of deep learning to tackle this issue. Neural network are good at learning and recognizing shapes, making the most of the complex information brought by multi-wavelength images. First studies on this approach have been successful in our group and we would like to continue in this line of investigations with a student, first of all, for an internship during the 2nd semester 2017-2018. Further than that, these machine learning techniques developed at the image level could be a foundation work integrated in a study of much larger scope leading to the improvement of current shear measurements. This would then be the object of the future work of the PhD student.
When surveys like LSST and Euclid are available, all the information could be used in a « multi-instrument, multi-wavelength » survey: shapes, colors could be used in order to greatly improve our precision on shear measurements. Introducing machine learning methods starting form the images on that aspect is definitely more complex than the first study, but is based on the same technical ground. We will investigate those questions using both simulations of LSST and Euclid data, and from existing overlapping data like Hubble (HST), DES, Subaru HSC and CFHT fields that are already available. The implications of this work on a joint, multi-survey, multi-wavelength analysis would then allow to reduce drastically the bias on shear measurements, bringing an essential tool for the use of this cosmological probe in the near future and for its impact on dark energy parameters.