Asteroseismic based estimation of the surface gravity for the LAMOST giant stars

TL;DR

This paper presents a support vector regression method to improve surface gravity estimates for LAMOST giant stars using asteroseismic data, significantly enhancing accuracy and aiding in better stellar characterization.

Contribution

The study introduces a novel SVR-based approach trained on seismic data to refine spectroscopic surface gravity estimates for large stellar surveys.

Findings

Reduced uncertainty of gravity estimates to about 0.1 dex

Improved match to stellar isochrones and better star classification

Distance uncertainties decreased to approximately 12%

Abstract

Asteroseismology is one of the most accurate approaches to estimate the surface gravity of a star. However, most of the data from the current spectroscopic surveys do not have asteroseismic measurements, which is very expensive and time consuming. In order to improve the spectroscopic surface gravity estimates for a large amount of survey data with the help of the small subset of the data with seismic measurements, we set up a support vector regression model for the estimation of the surface gravity supervised by 1,374 LAMOST giant stars with Kepler seismic surface gravity. The new approach can reduce the uncertainty of the estimates down to about 0.1 dex, which is better than the LAMOST pipeline by at least a factor of 2, for the spectra with signal-to-noise ratio higher than 20. Compared with the logg estimated from the LAMOST pipeline, the revised logg values provide a significantly improved match to the expected distribution of red clump and RGB stars from stellar isochrones. Moreover, even the red bump stars, which extend to only about 0.1 dex in logg, can be discriminated from the new estimated surface gravity. The method is then applied to about 350,000 LAMOST metal-rich giant stars to provide improved surface gravity estimates. In general, the uncertainty of the distance estimate based on the SVR surface gravity can be reduced to about 12% for the LAMOST data.