Novel Clustering Methods for Cosmology with DESI

Abstract

With its ‘5000 eyes’, the Dark Energy Spectroscopic Instrument (DESI) is mapping the universe with unprecedented accuracy and volume. Since first light in late 2019, DESI has been rapidly collecting spectral measurements of galaxies. Following the completion of main survey operations, this number is now in excess of 47 million. The three-dimensional clustering of these galaxies, inferred from their measured redshift, tells a tale of competition between cosmic expansion and gravity over the last 10 billion years. By utilising a variety of data compression techniques, we can probe the cosmological model, shedding light on the nature of the elusive dark components of the cosmic energy density—dark energy and dark matter.
One of the most prominent clustering signals is the Baryon Acoustic Oscillation (BAO) feature, a powerful geometric probe for inferring the expansion rate and matter content of the Universe. However, further information is contained within the cosmic web at scales beyond than that of the BAO imprint. In the galaxy power spectrum, this additional information manifests primarily through the characteristic turnover associated with matter-radiation equality, as well as anisotropic distortions in the observed galaxy field arising from observational effects. Modelling this full range of scales—commonly referred to as a ‘full shape’ analysis—is highly challenging, requiring an accurate and robust treatment of small-scale non-linear processes.
Additionally, one can look beyond N-point statistics entirely, capturing
fundamental processes on a variety of scales using other, more novel forms of statistical compression. Unlike standard clustering analyses, these methods are generally less mature and often lack complete analytical descriptions. One such probe is that of cos-mic voids: the most underdense regions of the Universe, occupying the space between the nodes and filaments of the cosmic web. The distribution of void sizes, or void size function (VSF), is a unique cosmological probe with an accompanying theoretical framework. However, establishing agreement between the observed VSF and theoretical predictions remains challenging due to the effects of galaxy bias and the algorithm-dependent nature of void identification.
This thesis presents three complementary investigations aimed at advancing
robust cosmological analyses beyond the BAO scale: (i) robust extraction of full-shape information from the power spectrum using DESI Data Release 1 (DR1); (ii) development of a void-finding algorithm designed to be consistent with theoretical predictions for the VSF; and (iii) forward-modelling void statistics within a simulation-based framework. Together, these studies provide novel analysis tools for robust multi-probe galaxy clustering analyses capable of meeting the precision demands of DESI and future surveys.

Date of Award	30 Jun 2026
Original language	English
Awarding Institution	University of Portsmouth
Supervisor	Seshadri Nadathur (Supervisor), Robert Crittenden (Supervisor) & Eva Mueller (Supervisor)