Using Machine Learning to Find Ghostly Damped Ly$\alpha$ Systems in SDSS DR14

TL;DR

This paper reports the discovery of 59 new ghostly damped Lyα systems in SDSS DR14, identified through machine learning, with detailed spectral analysis indicating most are likely DLAs, expanding the known sample significantly.

Contribution

The study introduces a machine learning approach to identify ghostly damped Lyα systems in SDSS data, increasing the known sample size and providing detailed spectral analysis.

Findings

Discovered 59 new ghostly absorbers in SDSS DR14.

Measured HI column density of logN(HI)=21.50 in stacked spectra.

Most ghostly absorbers are likely damped Lyα systems.

Abstract

We report the discovery of 59 new ghostly absorbers from the Sloan Digital Sky Survey (SDSS) Data Release 14 (DR14). These absorbers, with $z_{\rm abs}$$\sim$$z_{\rm QSO}$, reveal no Ly$\alpha$ absorption, and they are mainly identified through the detection of strong metal absorption lines in the spectra. The number of previously known such systems is 30. The new systems are found with the aid of machine learning algorithms. The spectra of 41 (out of total of 89) absorbers also cover the Ly$\beta$ spectral region. By fitting the damping wings of the Ly$\beta$ absorption in the stacked spectrum of 21 (out of 41) absorbers with relatively stronger Ly$\beta$ absorption, we measured an HI column density of log$N$(HI)=21.50. This column density is 0.5dex higher than that of the previous work. We also found that the metal absorption lines in the stacked spectrum of the 21 ghostly absorbers with stronger Ly$\beta$ absorption have similar properties as those in the stacked spectrum of the remaining systems. These circumstantial evidence strongly suggest that the majority of our ghostly absorbers are indeed DLAs.