APOKASC-3: The Third Joint Spectroscopic and Asteroseismic catalog for Evolved Stars in the Kepler Fields

TL;DR

APOKASC-3 delivers a definitive, Gaia-calibrated census of evolved stars in the Kepler fields by combining ten independent asteroseismic pipelines with APOGEE spectroscopy and Gaia-based fundamental radii. The work systematically maps the asteroseismic observables Δν and ν_max to physical parameters (mass, radius, age) through calibrated factors f_{Δ} and f_{max}, anchored to Gaia radii and supported by multiple stellar interior models. It provides a robust treatment of uncertainties, outliers, and population effects, and demonstrates that scaling relations are most reliable on the lower RGB and RC while becoming increasingly model-dependent for luminous giants. The catalog enables precise tests of stellar physics and Galactic population studies, including a sharp thin-disk age boundary and a well-constrained thick-disk age (~9.14 Gyr), while highlighting the need for modeling individual frequencies for the most luminous giants. Overall, APOKASC-3 sets a new standard for population asteroseismology with a transparent, calibration-driven framework and a rich dataset for future surveys.

Abstract

In the third APOKASC catalog, we present data for the complete sample of 15,808 evolved stars with APOGEE spectroscopic parameters and Kepler asteroseismology. We used ten independent asteroseismic analysis techniques and anchor our system on fundamental radii derived from Gaia $L$ and spectroscopic $T_{\rm eff}$. We provide evolutionary state, asteroseismic surface gravity, mass, radius, age, and the spectroscopic and asteroseismic measurements used to derive them for 12,418 stars. This includes 10,036 exceptionally precise measurements, with median fractional uncertainties in \nmax, \dnu, mass, radius and age of 0.6\%, 0.6\%, 3.8\%, 1.8\%, and 11.1\% respectively. We provide more limited data for 1,624 additional stars which either have lower quality data or are outside of our primary calibration domain. Using lower red giant branch (RGB) stars, we find a median age for the chemical thick disk of $9.14 \pm 0.05 ({\rm ran}) \pm 0.9 ({\rm sys})$ Gyr with an age dispersion of 1.1 Gyr, consistent with our error model. We calibrate our red clump (RC) mass loss to derive an age consistent with the lower RGB and provide asymptotic GB and RGB ages for luminous stars. We also find a sharp upper age boundary in the chemical thin disk. We find that scaling relations are precise and accurate on the lower RGB and RC, but they become more model dependent for more luminous giants and break down at the tip of the RGB. We recommend the usage of multiple methods, calibration to a fundamental scale, and the usage of stellar models to interpret frequency spacings.

Figures (25)

• Figure 1: The full APOKASC-3 evolved star sample. The histogram illustrates the number of targets, as a function of metallicity, in 0.05 dex $\log{g}$ bins (right).
• Figure 2: The abundance distribution of the APOKASC-3 sample in the [$$/Fe] -- [Fe/H] plane. The red line is the criterion that we use to distinguish high [$$/Fe] from low [$$/Fe] ones. For plotting purposes later in the catalog, we use a slight modification of the Roberts2024 criterion for distinguishing the two populations. A star was classified as $$-rich if [$$/Fe] was above $(0.08 - 0.15 {\rm [Fe/H]})$ in the range $-0.4< {\rm [Fe/H]} <\ +0.2$; a threshold of $+0.14$ was assigned for [Fe/H] below $-0.4$; and a threshold of $+0.05$ was used for [Fe/H] above $+0.2$. Our criterion differs from the Roberts2024 one in that we use [$$/Fe] and they used [Mg/Fe], including a slight zero-point shift.
• Figure 3: Our four cohorts -- Gold (top left), Silver (top right), Detections (bottom left) and Non-Detections (bottom right) -- illustrated in three-D mesh plots as a function of $\log{g}$ and spectroscopic $T_{\rm eff}$. We use seismic $\log{g}$ for all categories except Non-Detections, for which we use spectroscopic $\log{g}$. The horizontal lines denote the approximate boundaries where we report masses and radii (between $\log{g}$ of 1 and 3.3). Bin sizes of 40 K in $T_{\rm eff}$ and 0.02 dex in $\log{g}$ reflect typical measurement uncertainties. The vertical scale for the Gold bin is higher than for the other panels. The large majority of lower RGB and RC stars are robustly detected. In the detected groups we also note the number of stars with detections that are outside the domain where we provide masses, ages, and radii. In total, we have 12,448 targets with full data, 1,634 with partial data, and 1,423 without seismic data. Background sources (129) and stars without valid spectroscopic data (174) are not shown.
• Figure 4: Our four cohorts -- Gold (dark yellow, circles), Silver (light gray, triangles), Detections (dark gray, squares) and Non-Detections (orange, diamonds) -- illustrated as a function of spectroscopic $_{\rm \max}$ for stars with less than 3 quarters of data (left) and for those with 3-18 quarters (right). Absolute numbers are in the top panels, and fractions in the bottom panels. The large majority of stars with good data are detected in our core domain of 1-220 $\rm{ Hz}$, while recovery is much more difficult for shorter time series.
• Figure 5: Temperature offset, as defined in vrard2024, relative to the mean RGB locus as a function of asteroseismic surface gravity for our upper RGB sample. The RC stars are shown in blue, RGB stars are red, and AGB stars are yellow. A one solar mass, solar metallicity MIST AGB track is shown for reference as a yellow line.