Beecroft Institute for Particle Astrophysics and Cosmology

Abstract:

Constraining the escape fraction of Lyman Continuum (LyC) photons from high-redshift galaxies is crucial to understanding reionization. Recent observations have demonstrated that various characteristics of the Ly α emission line correlate with the inferred LyC escape fraction (f LyC esc ) of low-redshift galaxies. Using a data set of 9600 mock Ly α spectra of star-forming galaxies at 4.64 ≤ z ≤ 6 from the SPHINX20 cosmological radiation hydrodynamical simulation, we study the physics controlling the escape of Ly α and LyC photons. We find that our mock Ly α observations are representative of high-redshift observations and that typical observational methods tend to overpredict the Ly α escape fraction (f Ly α esc ) by as much as 2 dex. We investigate the correlations between f LyC esc and f Ly α esc , Ly α equivalent width (Wλ(Ly α)), peak separation (vsep), central escape fraction (fcen), and red peak asymmetry (Ared f ). We find that f Ly α esc and fcen are good diagnostics for LyC leakage, selecting for galaxies with lower neutral gas densities and less UV attenuation that have recently experienced supernova feedback. In contrast, Wλ(Ly α) and vsep are found to be necessary but insufficient diagnostics, while Ared f carries little information. Finally, we use stacks of Ly α, H α, and F150W mock surface brightness profiles to find that galaxies with high f LyC esc tend to have less extended Ly α and F150W haloes but larger H α haloes than their non-leaking counterparts. This confirms that Ly α spectral profiles and surface brightness morphology can be used to better understand the escape of LyC photons from galaxies during the epoch of reionization.

Abstract:

<jats:p>This paper presents the Learning the Universe Implicit Likelihood Inference (LtU-ILI) pipeline, a codebase for rapid, user-friendly, and cutting-edge machine learning (ML) inference in astrophysics and cosmology. The pipeline includes software for implementing various neural architectures, training schema, priors, and density estimators in a manner easily adaptable to any research workflow. It includes comprehensive validation metrics to assess posterior estimate coverage, enhancing the reliability of inferred results. Additionally, the pipeline is easily parallelizable, designed for efficient exploration of modeling hyperparameters. To demonstrate its capabilities, we present real applications across a range of astrophysics and cosmology problems, such as: estimating galaxy cluster masses from X-ray photometry; inferring cosmology from matter power spectra and halo point clouds; characterising progenitors in gravitational wave signals; capturing physical dust parameters from galaxy colors and luminosities; and establishing properties of semi-analytic models of galaxy formation. We also include exhaustive benchmarking and comparisons of all implemented methods as well as discussions about the challenges and pitfalls of ML inference in astronomical sciences. All code and examples are made publicly available at https://github.com/maho3/ltu-ili.</jats:p>