PhD Defense Public Talk

The Epoch of Reionization marks the last major phase transition of the Universe and remains one of the most active areas of research in modern astrophysics. Understanding this critical period requires robust measurements of the neutral hydrogen fraction in the intergalactic medium (IGM). The Lyman-alpha (Lyα) emission line is among the most powerful probes of the IGM neutral fraction. Its diagnostic power relies on the assumption that the intrinsic Lyα emission emerging from galaxies can be inferred. However, determining the emergent Lyα strength is complicated by the same resonant scattering physics that makes the line so sensitive to its environment. Before Lyα photons interact with the IGM, they must first escape the galaxy's interstellar medium (ISM), where dust, neutral gas, and complex gas kinematics can scatter, suppress, or destroy them. As a result, accurate inference of Lyα emission requires models capable of capturing these complexities beyond the reach of traditional single-parameter estimators.

This dissertation addresses this challenge through three interconnected themes.

(i) Understanding how galaxy properties and ISM conditions regulate Lyα escape. Using a sample of 155 star-forming galaxies at z = 1.9–3.5 from the MOSDEF survey with overlapping HETDEX Lyα detections, we combined Bayesian spectral energy distribution (SED) fitting with a newly developed nebular emission-line modeling framework to characterize the physical conditions governing Lyα escape. We find that Lyα escape is most efficient in galaxies with low stellar masses, low dust attenuation, and high ionization parameters.

(ii) Developing improved inference models that leverage correlations between galaxy properties and Lyα emission through the TONE project. We first present a pilot study of 43 confirmed Lyα emitters (LAEs), establishing the analysis framework and demonstrating that lower-mass, less dusty galaxies exhibit stronger Lyα emission. We then extend the analysis to a homogeneous sample of 11,114 LAEs across six survey fields and train a normalizing-flow neural posterior estimator to model the conditional distribution of emergent Lyα equivalent width given galaxy properties. The resulting posterior estimators are well calibrated and achieve a typical mean squared error of 0.4 dex in log(Lyα EW) on an unseen test set. These models enable improved predictions of Lyα strength and provide a pathway toward constraining the neutral hydrogen fraction during the Epoch of Reionization using photometrically characterized galaxy samples.

(iii) Investigating the role of active galactic nuclei (AGN) in high-redshift galaxies. I present evidence for AGN activity in GHZ2 (z = 12.34), one of the most distant spectroscopically confirmed galaxies known. Using the BEAGLE-AGN spectrophotometric fitting framework applied to JWST observations, we find that AGN emission contributes significantly to the high-ionization emission lines observed in GHZ2. From the inferred accretion luminosity, we estimate a black hole mass of approximately 10^7.2 M☉, assuming an Eddington ratio of 0.5. These results place GHZ2 among the most distant AGN candidates identified to date.

Together, these studies establish new observational and statistical frameworks for understanding the physical mechanisms that regulate Lyα escape, improving our ability to infer Lyα emission from galaxy properties, and interpreting the earliest generations of galaxies and black holes in the Universe.