Location
DC 2314
Candidate
Dorsa Sadat Hosseini KHajouei| Applied Mathematics, University of Waterloo
Title
The Cosmic Web Through Galaxies and Gravitational-Wave Sources: Linking Astrophysical Properties to Large-Scale Clustering
Abstract
The spatial distribution of galaxies and other tracers encodes the growth of cosmic structure, but every tracer samples the matter field in a biased way, and that bias is set by the astrophysical properties of the objects and the environments they occupy. For galaxy surveys, how bias depends on galaxy properties has become increasingly important: multi-tracer analyses benefit from populations with distinct biases, and full-shape analyses from physically informed priors on the bias parameters. Gravitational waves (GWs) from compact-binary mergers offer a complementary cosmological probe, yet most sources have no electromagnetic counterpart and hence no directly identified host galaxy, motivating statistical methods that correlate GW sources with galaxy surveys. Both reduce to the same question: how the bias of a tracer population is determined by the physical properties of the galaxies that either form it or host it. This thesis addresses that question on both fronts.
First, we propose a methodology to link binary black hole (BBH) mergers to their host galaxies through phenomenological host-galaxy probability functions that depend on stellar mass and, where spectroscopy makes them available, on star formation rate (SFR) and metallicity as well. By populating observed galaxies with mock GW sources, we compute the GW bias from the angular power spectrum of the mock catalogs for the photometric
surveys and from their three-dimensional power spectrum for the spectroscopic ones. We find that the stellar-mass dependence of the host-galaxy probability is the dominant driver of the difference between the GW and galaxy biases, with the apparent SFR dependence largely reflecting its correlation with stellar mass, and no independent metallicity dependence is detected possibly due to observational and modeling limitations. Because these observational analyses are limited by survey incompleteness, selection effects, and uncertainties in the inferred galaxy properties, we extend the study to the TNG300 hydrodynamical simulation of the IllustrisTNG project, which removes these observational limitations at the cost of a finite simulated volume and uncertain subgrid physics. There, we apply the same phenomenological prescription and, independently, a machine-learning emulator trained on population-synthesis-calibrated simulations to predict the merger rate per galaxy, in place of the phenomenological host-galaxy probability. We find GW sources are more strongly biased tracers than the galaxy population as a whole, stellar mass again dominates, and the emulator predicts additional scale-dependent differences relative to the full galaxy samples that point to contributions from correlated galaxy properties and environment.
Finally, using Dark Energy Spectroscopic Instrument (DESI) Data Release 1 (DR1) luminous red galaxies over 0.4 < z < 0.8, we define physically motivated sub-samples using pop-cosmos-inferred stellar masses and specific SFRs and fit their redshift-space power spectrum multipoles with a full-shape effective field theory model. At fixed stellar mass, quenched galaxies cluster more strongly than star-forming galaxies; the contrast persists across stellar mass and redshift and is retained by confidently classified galaxies. Together these studies show how the clustering of different tracers can depend on astrophysical properties, for galaxies and GW sources alike, providing a basis for physically informed tracer-bias modeling in future GW cosmology and multi-tracer large-scale structure analyses.