Title: Bayesian Modeling of Damped Lyman Alpha Absorber Detection Using Gaussian Processes
Abstract — Damped Lyman alpha absorbers (DLAs) are strong HI absorption features found in quasar spectra. From a galaxy formation perspective, these systems contain the majority of neutral hydrogen at z = 2 - 4 and are associated with the gas surrounding early-stage galaxies. From a cosmological perspective, DLAs are one of the major sources of systematic bias in large scale structure measurements from the Lyman alpha forest and must be masked before cosmological analysis. Traditionally, DLAs have been identified by trained astronomers through visual inspection. However, with millions of quasar spectra available from DESI, this task has become challenging and labor-intensive for humans. We have developed a Bayesian machine learning method that uses Gaussian processes to find DLAs. Our method reproduces the DLA population from the previous catalog, provides improved constraints on neutral hydrogen, and is consistent with the DLA population in cosmological simulations. Beyond DLA detection, this Bayesian approach is broadly applicable to quasar spectral analysis. In this talk, I will highlight two additional applications of Gaussian processes: (1) searching for metal absorption lines and (2) probabilistic inference of quasar redshifts.
For Zoom information, please contact the seminar coordinators.
