Li-rich Giants Identified from LAMOST DR8 Low-Resolution Survey

TL;DR

This study employs a deep learning model to identify and analyze a large sample of Li-rich giant stars from LAMOST low-resolution spectra, significantly expanding the known population for further stellar evolution research.

Contribution

Introduces a modified convolutional neural network, Coord-DenseNet, for accurate lithium abundance estimation in giant stars from low-resolution spectra, enabling large-scale identification of Li-rich giants.

Findings

Identified 7,768 Li-rich giants from over 900,000 spectra.

Achieved good prediction accuracy with MAE=0.15 dex.

Expanded the sample size of Li-rich giants for future studies.

Abstract

A small fraction of giants possess photospheric lithium(Li) abundance higher than the value predicted by the standard stellar evolution models, and the detailed mechanisms of Li enhancement are complicated and lack a definite conclusion. In order to better understand the Li enhancement behaviors, a large and homogeneous Li-rich giants sample is needed. In this study, we designed a modified convolutional neural network model called Coord-DenseNet to determine the A(Li) of Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) low-resolution survey (LRS) giant spectra. The precision is good on the test set: MAE=0.15 dex, and {\sigma}=0.21 dex. We used this model to predict the Li abundance of more than 900,000 LAMOST DR8 LRS giant spectra and identified 7,768 Li-rich giants with Li abundances ranging from 2.0 to 5.4 dex, accounting for about 1.02% of all giants. We compared the Li abundance estimated by our work with those derived from high-resolution spectra. We found that the consistency was good if the overall deviation of 0.27 dex between them was not considered. The analysis shows that the difference is mainly due to the high A(Li) from the medium-resolution spectra in the training set. This sample of Li-rich giants dramatically expands the existing sample size of Li-rich giants and provides us with more samples to further study the formation and evolution of Li-rich giants.