I develop Bayesian forward modelling frameworks that employ the cosmic large-scale structure as a probe of the physics of the Dark Universe.

Impact of modified gravity on the time delay produced by the lensing of a GW event by a galaxy cluster.

Strongly lensed gravitational waves provide exceptionally precise measurements of the associated time delays, opening the question whether robust and precise inferences of modified gravity are possible given the existing observational and theoretical uncertainties. Galaxy clusters are particularly interesting lenses since they provide a large cross section and can probe the transition from screened to unscreened gravity. In this work, we implement a phenomenological modification to gravity into the standard strong lensing formalism and inject different classes of systematic effects pertaining to the modelling of galaxy cluster lenses, binary neutron stars as sources and the surrounding structure to understand their impact on the recovered modified gravity signal. We find that cluster ellipticity, secondary structure, structures along the line of sight and mis-centring produce lens-specific biases on the gravity signature. We show that typical uncertainties on the lens mass and concentration produce coherent biases, while mis-modelling of the brightest central galaxy velocity dispersion, and lens and source redshifts yield subdominant biases.

Posterior of galaxy cluster bias, velocity dispersion and growth rate multiplied by the matter power spectrum normalisation parameter for a Euclid DR1-like catalogue.

Cluster clustering can provide constraints on the matter content of the universe, the growth of structure and modified gravity, benefitting from tight priors in cluster masses and complementarity with traditional cluster probes. In this work, we study the performance of different clustering models under a Euclid DR1-like setting and lay out criteria for the statistical performance of these models against the data to be used in the real data analysis.

Modified gravity constraints from galaxy cluster gravitational redshifts
Modified gravity parameter posteriors forecasted for a spectroscopic and a photometric galaxy survey from simulated data.

Persisting tensions in cosmological data and the unknown origin of cosmic acceleration make testing possible departures from general relativity one of the central questions in modern cosmology. In this work, we build an end-to-end pipeline to forecast the constraining power of gravitational redshifts of galaxies in clusters. We show that competitive measurements demand wide-field spectroscopic cluster surveys explicitly designed to maximise the number of spectroscopically confirmed members per cluster and to control systematic effects. In particular, efforts should prioritise mitigating offsets between the inferred and true cluster centres, which we identify as the dominant systematic for this probe.

Simulated supernovae in the reconstructed large-scale structure around the Coma cluster
Simulated supernovae (orange) in a reconstruction of the cosmic large-scale structure around the Coma cluster (Abell 1656). The purple points are galaxies in the real Universe, whereas the white ones are simulated galaxies that follow the same clustering pattern and from which we simulated supernovae.

One of the biggest puzzles in cosmology today is the Hubble tension — a mismatch in the measured expansion rate of the universe, called the Hubble constant (H₀), depending on how and where it is measured. It has been suggested that the motions of galaxies induced by gravity (peculiar velocities) in our cosmic neighbourhood may be biasing these measurements. In this study, we take a closer look at that idea. We simulate the positions of type Ia supernovae — cosmic mileposts used to measure distances — using simulated galaxy data following the clustering of real galaxies in the local Universe. We develop and validate a Bayesian Hierarchical Model which allows us to infer H₀ from SNeIa accounting for peculiar velocities self-consistently, while constraining the gravitational impact of the local Universe on the Hubble constant. We found that the effect of peculiar velocities is not sufficient to reconcile the Hubble tension alone. Our study paved the way for the use of SNeIa for inferences of the Hubble constant in the very nearby Universe.

Supernova explosions in the reconstructed large-scale structure
Illustration of supernova explosions in the large-scale structure. The colormap indicates density at 4 Mpc, constrained with data from the 2M++ galaxy catalog. Credits: Eleni Tsaprazi.

The Vera C. Rubin Observatory’s Legacy Survey of Space and Time will detect tens of thousands of supernovae per year, a detection rate which would in principle allow us to exploit supernova explosions to probe the environments of these transient phenomena to mitigate biases in cosmological inferences. In Tsaprazi et al. 2021a we studied the large-scale environments of supernovae and found that they follow the clustering of typical galaxies, despite their prevalence in star-forming environments. This study inspired the above investigation, where we investigated the clustering of supernovae in the peculiar velocity field. An exciting prospect lies in understanding its impact not only on measurements of the Hubble constant, but also on the evidence for dynamical dark energy.

Cross-correlation of reconstructed photometric galaxy coordinates with ground truth
Cross-correlation of the gridded photometric galaxy coordinates before (black) and after (orange) the application of our method with the ground truth (mock data).

Next-generation photometric surveys will deliver redshifts with an uncertainty corresponding to as much as 300 Mpc or more. Reconstructing the large-scale structure from highly uncertain galaxy locations results in structures that are distorted. Here, we developed a method that mitigates such distortions, while also constraining individual galaxy redshift probability density functions. Our framework provides Markov Chain Monte Carlo realizations of the primordial and present-day large-scale structure as constrained by galaxy clustering, accounting for survey- and data-related uncertainties. Owing to the structure formation model we employ, we reconstruct the structure formation history of the large-scale structure, as well as filaments. We provide constraints on the large-scale structure on scales much smaller than the original redshift uncertainty.

Galaxy shapes and orientations in JWST observations of Abell 2744
Galaxy shapes and orientations traced over a portion of JWST’s NIRCam image of Abell 2744. Credit: Lamman et al. 2024.

Galaxy shapes are not random, but align with the large-scale structure. This effect, known as galaxy intrinsic alignment, is linked to a multitude of cosmic phenomena. It provides constraints on galaxy formation and evolution, the initial conditions of the Universe, redshift-space distortions, as well as the gravitational wave background, among others. Constraints on intrinsic alignment further allow the decontamination of the weak lensing signal from intrinsic galaxy shape correlations which bias cosmological estimates. In Lamman et al. 2024, we created a guide to intrinsic alignment formalisms, models, estimators and useful references. My colleagues also prepared a presentation of the guide in Shaun Hotchkiss’ Cosmology Talks on YouTube.

Predicted and observed shape of the FIRST J092319.1+315950 radio galaxy
Predicted shape of the FIRST J092319.1+315950 radiogalaxy (cyan) compared to its observed shape as reported by Reyes et al. 2012 (orange). The difference between the two ellipses is due to linear alignment. The background image is taken from the SIMBAD database. Credit: Eleni Tsaprazi.

In Tsaprazi et al. 2021b, we find 4σ evidence of intrinsic alignment within a luminous red galaxy sample, by cross-correlating with three-dimensional tidal fields constrained with spectroscopic galaxy observations. Field-level approaches, like the one we present in our study, facilitate the modelling of nonlinear corrections to intrinsic alignment models and gravitational evolution, as well as joint inferences of intrinsic alignment and weak lensing.