Damped Lyα (DLA) absorbers are a class of quasar absorption line systems that can be used to trace hydrogen evolution and study gas, composition, kinematics, and stellar formation in galaxies or the intergalactic medium in the early universe. In this study, the number of DLA absorbers found with a deep learning algorithm is augmented using metal Mg ii absorbers as signposts. The artificial training data was generated by inserting synthetic absorption lines into spectra. Using previously identified DLAs as a test set, the model returned a 94.6% accuracy score and 1197 new DLA absorbers, which nearly doubles the previously known number of samples within the search catalog. Additionally, the column density and Doppler broadening are measured on the newly found DLA absorbers through Voigt profile fitting.
Read full abstract