Searching for Rare Quasar C I and Ca II Absorbers in the Early Universe Using Deep Neural Networks

 

The universe is still largely a mystery to scientists. Quasars are one such mystery, whose emission spectra produce absorption lines such as Ca II and C I when passing through gas of distant galaxies. This data helps astronomers understand more about interstellar gas, dust, and galaxy and star formation and evolution (including our Milky Way). However, these current absorber databases are extremely limited, and traditional methods make them hard to detect. Thus, we seek to discover more C I and Ca II absorbers by developing deep neural networks, which are more accurate and faster. We first found absorbers traditionally to produce a test set. We cropped, normalized, and handpicked through thousands of spectra and discovered 108 C I and 256 Ca II original absorbers. To obtain large training sets, we generated tens of thousands of artificial samples by inserting C I or Ca II lines at corresponding wavelengths in real spectra. We preprocessed the data and created neural network models after testing different hyperparameter configurations. Overall, our accuracies for absorber detection are 93% (C I) and 95% (Ca II), 15 times higher than traditional methods, and we added significant amounts of new absorbers to current datasets for C I and Ca II, completing our goal. As for challenges, we concluded that most false negatives were due to noise and weak lines. Furthermore, our discovered absorbers agree with statistical tests of previous studies. In the future, we plan to discover more absorbers using our models and run statistical studies on them.