Refined M-type Star Catalog from LAMOST DR10: Measurements of Radial Velocities, $T_\text{eff}$, log $g$, [M/H] and [$α$/M]

TL;DR

The paper addresses the challenge of obtaining precise atmospheric parameters for the Galaxy's most common stars, M-types, from vast low-resolution spectra. It introduces a two-stage pipeline that first purifies the LAMOST DR10 M-type catalog to a clean set and then employs a label-transfer plus parameter-prediction strategy, training a CNN ensemble on APOGEE DR16 labels to predict $T_eff$, log g, [M/H], and [$alpha/M$] separately for M dwarfs and giants. The resulting Recommended Catalog achieves substantial precision improvements (external and internal) compared with prior work and provides a robust resource for Galactic archaeology, exoplanet host screening, and stellar population studies. By mitigating the synthetic gap through label transfer and leveraging ten-fold cross-validated CNNs, the method demonstrates scalable, accurate parameter estimation for large spectroscopic surveys and sets a framework for extending to higher-resolution datasets. Overall, the study delivers a purified M-type catalog, reliable RVs, and high-precision stellar parameters, advancing the utility of LAMOST data for M-dwarf and M-giant science while highlighting areas for future improvement and extension to LAMOST-MRS data.

Abstract

Precise stellar parameters for M-type stars, the Galaxy's most common stellar type, are crucial for numerous studies. In this work, we refined the LAMOST DR10 M-type star catalog through a two-stage process. First, we purified the catalog using techniques including deep learning and color-magnitude diagrams to remove 22,496 non-M spectra, correct 2,078 dwarf/giant classifications, and update 12,900 radial velocities. This resulted in a cleaner catalog containing 870,518 M-type spectra (820,493 dwarfs, 50,025 giants). Second, applying a label transfer strategy using values from APOGEE DR16 for parameter prediction with a ten-fold cross-validated CNN ensemble architecture, we predicted $T_\text{eff}$, $\log g$, [M/H], and [$α$/M] separately for M dwarfs and giants. The average internal errors for M dwarfs/giants are respectively: $T_\text{eff}$ 30/17 K, log $g$ 0.07/0.07 dex, [M/H] 0.07/0.05 dex, and [$α$/M] 0.02/0.02 dex. Comparison with APOGEE demonstrates external precisions of 34/14 K, 0.12/0.07 dex, 0.09/0.04 dex, and 0.03/0.02 dex for M dwarfs/giants, which represents precision improvements of over 20\% for M dwarfs and over 50\% for M giants compared to previous literature results. The catalog is available at https://nadc.china-vo.org/res/r101668/.

Figures (25)

• Figure 1: Workflow diagram. Panel 0 illustrates the workflow for constructing a purer M-type star catalog, which consists of four modules: the Spectral Classification Module (SCM), the Dwarf/Giant Classification Module (DGCM), the Radial Velocity Measurement Module (RVMM), and the Stellar Parameter Measurement Module (SPMM). These modules are designed for spectral classification, M dwarf (dM) and M giant (gM) classification, radial velocity (RV) measurement, and stellar parameter estimation, respectively. Panels 1 to 4 present the flowcharts of these four modules. It is important to note that in Panels 1 and 3, purple and green dots are used to indicate breakpoints to avoid overlap between connecting lines. In reality, these breakpoints lie along the same connecting line.
• Figure 2: The CNN network architecture used in this work and the outputs of the different CNN models. The CNN-SPT model is used for spectral type classification, the CNN-WDM model is used for white dwarf (WD) and M-type star classification, the CNN-dMgM model is used for dM and gM classification, and the CNN-TLMA model is used for measuring the stellar parameters. It should be noted that the outputs of CNN-SPT, CNN-WDM, and CNN-dMgM are the probabilities of each type, while the output of CNN-TLMA consists of four stellar parameters.
• Figure 3: Confusion matrix of the CNN-SPT model on the validation set.
• Figure 4: Confusion matrix of the CNN-WDM model on the validation set.
• Figure 5: Confusion matrix of the CNN-SPT model with WD type included on the validation set.