The accurate determination of stellar parameters in M-type stars holds significant importance in various astronomical studies. In our study, we employed a transfer learning method coupled with PHOENIX synthetic spectra to estimate fundamental atmospheric parameters (Teff, log g, [M/H], [α/M]) and radial velocities (RV) for an extensive dataset encompassing 581,778 M dwarfs obtained from the Low-Resolution Spectroscopic Survey (LRS) within LAMOST DR10. By utilizing the domain-adaptation approach Cycle-StarNet, we successfully bridged the gap between observed and synthetic spectra. The L-BFGS algorithm was subsequently employed to identify the best-fit synthetic spectra. Notably, this study introduced theoretical PARSEC isochrones to construct training samples within the synthetic domain. As a result, the accuracy of Teff, log g, [M/H], [α/M], and RV can reach 60 K, 0.05 dex, 0.19 dex, 0.07 dex, and 8 km/s, respectively. We further assessed the precision of these parameters through repeated observations, revealing that for spectra with signal-to-noise ratios (S/Ns) exceeding 20, the precision for these parameters and radial velocities can reach as high as 31 K, 0.034 dex, 0.083 dex, 0.029 dex, and 6 km/s.
BIO
My research primarily focuses on very low-mass stars (VLMS), specifically the spectral classification, atmospheric parameter measurement, and kinematic analysis of M dwarfs and M subdwarfs.
2020.6-present: postdoc, PKU
2014.9-2020.6: PhD, NAOC. Supervisor: A-Li Luo
