Sorbonne Universite
Universite de Paris
Initiative Physique des Infinis


Sur ce site

Accueil > Thèses, Stages, Formation et Enseignement > Propositions de thèses 2024 > Beyond standard galaxy clustering analysis with the DESI BGS

Beyond standard galaxy clustering analysis with the DESI BGS

par Tristan Beau - 9 novembre 2023

Titre : Beyond standard galaxy clustering analysis with the DESI BGS

Directrice/directeur de thèse : Pauline Zarrouk

Groupe d’accueil :Cosmologie et énergie noire

Collaboration : DESI

Description :

Spectroscopic galaxy surveys yield detailed 3D maps of the cosmic large-scale structures by mapping the distribution of billions of galaxies in the sky. Those maps are now a well-established cosmological probe to understand our universe’s content and its mysterious late-time expansion. Is the universe’s expansion accelerating because of dark energy, an unknown component, or is the underlying theory of gravity based on General Relativity incomplete at cosmological scales ? The latest state-of-the art galaxy clustering analyses using 2-point statistics (two-point correlation function or power spectrum) have reached 3% precision on the equation of state of dark energy and a 10% precision on the growth rate of structures at 3 effective redshifts in 0.4 < z < 0.7 (eBOSS Collaboration 2020). The sensitivity of galaxy clustering to structure growth allows us to perform direct tests of gravity, and thus to test the validity of General Relativity at cosmological scales. Forthcoming spectroscopic surveys with exquisite statistical power promise advance on this fundamental question and the Dark Energy Spectroscopic Instrument (DESI, DESI Collaboration 2016) is the first survey of the new generation that is actually taking data. DESI has started its main survey on May 2021 and it will collect the spectra of 40 million galaxies and quasars up to z=3.5 during 5 years.

Traditional cosmological approaches rely on summarising the rich but noisy 3D information with 2-point correlation function (2PCF). These analyses are limited to scales of separation above 20 Mpc/h where we can use an analytic model based on perturbation theory to describe the evolution of the density and velocity fields. Going at smaller scales brings significant constraining power but those scales cannot be modelled with perturbation theory anymore. Instead, we develop simulation-based models called emulators. Going to higher-order statistics adds non-Gaussian information encoded in the gravitational evolution of the density field and can break degeneracies between cosmological parameters. One example of beyond 2-point statistics that looks very promising is the density-split clustering that was proposed in Paillas et al. (2021). First, the galaxy density field is split into different quantiles according to the local galaxy density, then auto and cross-correlations of those densities with the entire galaxy field are used to extract the cosmological information. Using DS in Paillas, Cuesta-Lazaro, Zarrouk et al. (2022), we showed that the constraints on the matter density content, the amplitude of the matter fluctuations and the sum of the neutrino masses can be improved by a factor of 5, 6 and 8, respectively when compared to the 2PCF alone. More recently, we developed an emulator for density-split in Cuesta-Lazaro et al. (2023) that has been applied to previous galaxy survey BOSS in Paillas et al. (2023). The goal is now to develop and apply such a kind of technique to the DESI Bright Galaxy Survey (BGS), which is a very dense and complete sample of bright galaxies at z < 0.5 (Ruiz-Macias, Zarrouk et al. 2021, Zarrouk et al. 2022, Hahn et al. 2023).

During the internship, the candidate will continue developing the emulator for the DESI BGS, testing it against N-body simulations without (1st step) and with (2nd step) systematics. Then, the technique will be applied to the DESI Y1 sample to obtain the cosmological constraints and compare with 2PCF alone. The technique will be further improved for DESI Y3 during the rest of the PhD. In particular, the candidate will also develop a simulation-based inference pipeline to compare with the standard way of extracting the cosmological constraints when the likelihood is Gaussian.

Lieu(x) de travail : LPNHE

Déplacements éventuels : USA, UK

Stage proposé avant la thèse : Oui

Enregistrer au format PDF