Oral Presentation

Detecting Multiple Damped Lyman Alpha Systems with Bayesian machine learning

Author(s): Ming-Feng Ho (UCR), Simeon Bird (UCR), Garnett Roman (WUSTL)

Presenter: Ming-Feng Ho (University of California, Riverside)

Damped Lyα absorbers (DLAs) are the nucleations of galaxies from diffuse gas which reveal themselves in the spectroscopic sightlines of quasars.
The detection of DLAs sheds light on galaxy formation at high redshift, but generating large catalogues of DLAs is usually very time-consuming and highly relies on the eyes of trained astronomers.
We develop a fully automated pipeline for detecting multiple DLAs from quasar spectroscopic sightlines.
We use SDSS DR9 to train a Gaussian process model for quasar emission spectra and use parameterized DLA absorption profiles to identify an arbitrary number of DLAs.
We improve our previous DLA detecting pipeline via considering an alternative model in Bayesian model selection for absorbers with lower column densities than DLAs, which we reduce the false positive detections in DLAs.
We also improve the sampling method for models with more than two DLAs via importance sampling.
We present the column density distribution function, the line densities, and the neural hydrogen densities of the results of our multi-DLA pipeline.
We also provide the DLA catalogues of SDSS DR12 and SDSS DR15 based on the classification results of our multi-DLA pipeline.