Probing Dark Energy using photometric measurements of supernovae

Stage numéro : M1-1819-RE-01
Laboratoire :Centre de Physique des Particules de Marseille Case 902
 163 avenue de Luminy - 13288 Marseille Cedex 9
Directeur :Cristinel Diaconu - -
Correspondant :William Gillard - -
Groupe d'accueil :Renoir
Chef de groupe :Dominique Fouchez - -
Responsable de stage :Dominique Fouchez - 04 91 82 76 49 -

Thématique : Cosmologie observationnelle

The discovery of the acceleration of the expansion of the Universe, rewarded by the Nobel prize 2011, was made thanks to type Ia Supernovae observations. Since that discovery, understanding the origin of the acceleration, from the so-called dark energy, is one of the major quest of cosmology [1]. Today and in the next generation of surveys, supernova observations will be a key point to constraint the dark energy equation of state.

The Cosmology group at CPPM is strongly engaged in this quest with his participation in large supernova programs (SNLS and SNFactory), large scale structure analysis (Boss/eBoss) and the preparation of the future surveys Euclid and LSST.

Hubble diagram which is used to constraint the cosmological models needs two ingredients : a precise measurement of the luminosity distance of the supernova and its redshift. In the past, the measurement of the redshift was made by observing the spectral lines of the supernovae and/or its host galaxy. With the next generation of surveys, one major point will be the impossibility to obtain a spectrum for all the astrophysical objects including the supernovae. In particular for the Large Synoptic Survey Telescope (LSST) which will measure the luminosity of around 27 billions of stars and galaxies during 10 years.

The goal of this internship is to develop a machine learning method to estimate photometric redshifts of the supernova host galaxies and to estimate the impact of using his/her results on the cosmological parameter measurements using real data.

The student will develop an automatic tool using the photometric images. For that he will use Machine Learning methods. One of the different methods very promising for few years is Deep Learning[2] since it shows extraordinary performances in several science fields and won several competitions. Deep Learning to estimate photometric redshifts of galaxies or quasars are already showing results better than the state of the art. The student will benefit of the work already done in this field. The student will then use his results to build an Hubble diagram and measure the cosmological parameters. The student will use SDSS and CFHTLS photometric data together with associated spectroscopy data for training and control.

[1] “ Improved cosmological constraints from a joint analysis of the SDSS-II and SNLS supernova samples”, Betoule et al, Astronomy Astrophysics, Volume 568, id.A22

[2] Photometric redshifts from SDSS images using a convolutional neural , network Pasquet et al, Astronomy and Astrophysics, Volume 621, id.A26