Estimating stellar parameters and identifying very metal-poor stars for low-resolution spectra (R∼ 200)

• DOI: 10.1017/pasa.2023.59
• Publication Date: 2023-11-28T05:00:47Z
• Authors: Tianmin Wu; Yude Bu; Jianhang Xie; Junchao Liang; Wei Liu; Zhenping Yi; Xiaoming Kong; Meng Liu
• Keywords: LAMOST

Abstract Very metal-poor (VMP, [Fe/H]<-2.0) stars serve as invaluable repositories of insights into the nature and evolution first-generation formed in early galaxy. The upcoming China Space Station Telescope (CSST) will provide us with a large amount spectral data that may contain plenty VMP stars, thus it is crucial to determine stellar atmospheric parameters ( $T_{\textrm{eff}}$ , $\log$ g, [Fe/H]) for low-resolution spectra similar CSST $R\sim 200$ ). This study introduces novel two-dimensional Convolutional Neural Network (CNN) model, comprised three convolutional layers two fully connected layers. model’s proficiency assessed estimating parameters, particularly metallicity, from $R \sim ), specific focus on enhancing search within data. We mainly use 10 008 LAMOST DR3, 16 638 non-VMP ([Fe/H]>-2.0) DR8 experiments apply random forest support vector machine methods make comparisons. resolution all reduced $R\sim200$ match CSST, followed by pre-processing transformation input CNN model. validation practicality this model are also tested MARCS synthetic spectra. results show using constructed paper, we obtain Mean Absolute Error (MAE) values 99.40 K 0.22 dex 0.14 [Fe/H], 0.26 [C/Fe] test set. Besides, can efficiently identify precision rate 94.77%, recall 93.73%, an accuracy 95.70%. paper powerfully demonstrates effectiveness proposed ) recognizing interest population galactic work.