A hybrid SLAM-Payne framework for atmospheric parameter and abundance determination of early-type Stars from LAMOST DR9 low-resolution Spectra

Abstract

Early-type stars are key drivers of Galactic chemical evolution, enriching the interstellar medium with alpha elements through powerful stellar winds and core-collapse supernovae, fueled by their short lifetimes and high masses. However, their spectra remain challenging to analyse due to rapid rotation, weak metal lines, and non-LTE effects. While large spectroscopic surveys provide extensive low-resolution data, extracting reliable parameters remains difficult due to methodological limitations for hot stars. Our goal is to develop a unified framework combining data-driven and synthetic spectral approaches to determine atmospheric parameters and abundances for hot stars using low-resolution spectra, addressing limitations in current methodologies while retaining critical spectral information. We present a hybrid approach integrating the Stellar LAbel Machine (SLAM) and the Payne frameworks, for low-resolution ($R\sim 1800$) spectra from LAMOST DR9. Our method preserves full spectral information including Balmer series and metal-line blends, employing neural-network interpolation for efficient parameter estimation ($T_\mathrm{eff}$, $\log g$, and $v\sin i$) and abundance determination for O, Mg, Si, and Fe, across 8000 - 20000 K. We derive stellar parameters and abundances for 315,822 stars with $\mathrm{SNR} \geqslant 10$ in the $r$-band. Additionally, we detect abundance trends ([$α$/Fe]-[Fe/H]) that exhibit temperature-dependent systematics and a distinct $α$-poor stellar population within $0.0 \leqslant \mathrm{[Fe/H]} \leqslant 0.5$ dex. The radial abundance gradients are negative and consistent with that derived from Cepheids, with a slope of $-0.070\pm0.007$ in \feh in the region $6 \leqslant R_\mathrm{GC}\leqslant 15$ kpc.