PhD - Observational Cosmology

Supervised by Dr. Cyrille Doux at LPSC (Grenoble, France).

Title

Weak Lensing Cosmology using LSST, Euclid & Roman combined datasets.

Abstract

The accelerating expansion of the Universe, driven by dark energy, remains one of the greatest unsolved challenges in cosmology. Upcoming surveys such as the Vera C. Rubin Observatory’s LSST, ESA’s Euclid, and NASA’s Roman Space Telescope aim to probe dark energy through measurements of weak gravitational lensing (WL) and large-scale structure. However, blending—where faint overlapping galaxy sources are misidentified—introduces substantial errors in galaxy shape and flux measurements, particularly for deep, ground-based surveys like LSST. This effect presents a major obstacle to accurate WL measurements.

This project aims to develop advanced techniques to mitigate blending’s impact on cosmological analyses, maximizing the scientific return of these surveys.

Project Objectives

1. Characterize blending in LSST data
   By comparing LSST commissioning data with high-resolution space-based observations from HST and JWST, the project will quantify blending levels and assess the performance of LSST’s detection and deblending pipelines.

2. Investigate the impact of blending on multi-probe cosmology
   Using simulations, the project will analyze how blending affects galaxy lensing, clustering, and cluster lensing, evaluating the trade-off between removing blended objects and losing statistical power.

3. Develop machine-learning-based mitigation strategies
   Deep learning techniques, including graph and convolutional neural networks, will be used to create models that identify and correct for blending, minimizing biases in cosmological measurements.

4. Implement a multi-survey framework
   By leveraging overlapping data from LSST, Euclid, and Roman, the project will refine blending correction algorithms, improving catalog accuracy and tightening constraints on dark energy.

Methodology

• Use the blending entropy, a novel probabilistic technique, to quantify blending.
• Employ machine learning to predict blending scores and mitigate its effects.
• Develop joint analysis techniques for LSST, Euclid, and Roman data.

Expected Results

• Improved blending characterization tools for LSST data.
• Value-added catalogs enabling advanced blending mitigation.
• Enhanced cosmological constraints from combined LSST, Euclid and Roman data.

Expertise

This project will create cutting-edge tools to correct for blending, improving the accuracy of WL measurements and advancing our understanding of dark energy. It will also contribute to the early analysis of LSST data, optimize multi-survey data for enhanced constraints on cosmological parameters from LSST, Euclid and Roman as well as contribute to a deeper understanding of dark energy and the evolution of the Universe.