The metal-poor tail of the APOGEE survey II. Spectral analysis of Mg and Si in very metal-poor APOGEE spectra

TL;DR

This work develops a focused, Mg and Si–driven method to recover metallicities for APOGEE stars at the very metal-poor tail where Fe lines are weak in the H band. Using a SME-based pipeline with NLTE corrections and two narrow spectral segments around Mg I and Si I lines, the authors obtain [M/H], [Mg/H], and [Si/H] for 327 stars (∼304 with uncertainties <0.3 dex), extending APOGEE’s metal-poor tail to [M/H] ≈ -3.1. The study finds notable variations in the [Si/Mg] ratio across environments, including GC stars and dwarf galaxies, and discusses multiple enrichment channels (GC second populations, SNe II/Ia, PISN, and inhomogeneous mixing) as drivers of these patterns. The results offer new constraints on early chemical evolution and highlight the potential of upcoming infrared surveys to probe alpha-element abundances at extreme low metallicities.

Abstract

H-band spectra contain very limited spectral information for stars at the most metal-poor tail ( Fe/H < -2.5) because the available Fe lines in FGK stars in this wavelength range are weak. The first paper in this series successfully identified a sample of 327 very metal-poor stars (with [Fe/H] < -2) from the APOGEE database, 289 of which are on the red giant branch. The spectra of these stars were not properly analysed by the APOGEE main pipeline because they are very metal poor. In this work, we measure metallicities for these stars using the abundances of the elements Mg and Si. We demonstrate that the absorption lines of the elements Mg and Si are of good quality despite the challenging combination of (low) metallicity, wavelength regime, spectral resolution, and signal-to-noise ratios available for these spectra. A specialised pipeline was designed to measure the abundance of Mg and Si in APOGEE spectra and yielded a robust estimate of the overall metallicity. In order to provide reliable measurements, we tested three different sets of assumptions for Mg and Si enhancement. We present Mg and Si abundances as well as overall metallicities for 327 stars, all of which had previously gotten null values from the main APOGEE pipeline for either the calibrated M/H or [Fe/H] . The typical uncertainties for our measurements are 0.2 dex. We found five stars in our sample with unusual [Si/Mg] abundances higher than 0.5, and we connect this signature to globular cluster stars, and this might be related to specific supernova events. Our data suggest a concentration of high [Si/Mg] stars in the Sextans dwarf galaxy. Other dwarf galaxies are found to agree well with results in the literature. Our derived metallicities range between -3.1 $\leq$ [M/H] $\leq$ -2.25, thereby pushing the metal-poor tail of APOGEE results down by 0.6 dex.

Figures (12)

• Figure 1: Flowchart of the pipeline, see Sect. \ref{['sec:flowchart']} for details.
• Figure 2: Example synthetic fits to some of the program spectra, the spectra are artificially offset to make all three visible. The spectra are chosen to represent the median S/N, and the upper and lower quantiles.
• Figure 3: Residual plot for estimated metallicities against photometric and spectroscopic metallicities, shown as a histogram. The means are shown as dashed lines, and given along with the standard deviations.
• Figure 4: The left panel is a histogram of the number of stars per bin (on a log scale) with stacked bars of the original truncated metal-poor tail of APOGEE with weighted metallicities from the high-confidence sample. The reported metallicities already include the quality cuts. On the right, the same clean sample with weighted metallicities is mapped onto the celestial sphere and superimposed on the APOGEE footprint. The location of the LMC and SMC, where there is a concentration of stars, has been marked.
• Figure 5: The three plots include the clean high-confidence sample, and all APOGEE, GALAH, and SAGA stars with [Mg/H] lower than -1. The APOGEE dataset has been filtered to exclude all stars from dwarf galaxies and GCs. The running mean shown for each dataset goes from the minimum to the maximum [Mg/H] with bins every 0.5 dex for the three surveys, and 0.2 dex for the high-confidence stars due to smaller sample size and spanning a smaller metallicity range.