Asteroseismic ages for 17,000 stars in Kepler, K2 and TESS

The availability of asteroseismic constraints for tens of thousands of red giant (RG) stars has opened the door to robust age estimates, enabling time-resolved studies of different populations of stars in the Milky Way. This study leverages data from Kepler, K2, and TESS, in conjunction with astrometric data from Gaia DR3 and spectroscopic constraints from APOGEE DR17 and GALAH DR3, to infer parameters for over 17,000 RGs. We use the code PARAM to homogeneously infer stellar properties considering in detail the sensitivity of our results to different choices of observational constraints. We focus on age estimation, identifying potentially unreliable age determinations, and highlight stars with unreliable $Δν$ measurements based on comparisons using Gaia luminosities. These are particularly relevant in K2 data due to the short duration of the observations of each campaign, and therefore important to characterise for Galactic archaeology studies where the spatial range of K2 is a benefit. Thanks to the combination of data from different missions we explore trends in age, mass, and orbital parameters such as $R_\mathrm{g}$ and $Z_\mathrm{max}$, and examine time-resolved [$α$/M]-[Fe/H] planes across different Galactic regions. Additionally, we compare age distributions in low- and high-$α$ populations and chemically selected ex situ stars. The study also extends known mass-[C/N] ratio relationships to lower masses. The catalogues resulting from this work will be instrumental in addressing key questions in Galactic archaeology and stellar evolution, and to improve training sets for machine-learning-based age estimations.

💡 Research Summary

This paper presents a homogeneous analysis of more than 17,000 red‑giant (RG) stars observed by the three major space‑photometry missions—Kepler, K2, and TESS—combined with Gaia DR3 astrometry and high‑resolution spectroscopy from APOGEE DR17 and GALAH DR3. The authors use the global asteroseismic observables ν_max (frequency of maximum oscillation power) and Δν (average large frequency separation) as the seismic backbone, because ν_max scales with surface gravity and Δν with mean stellar density, allowing robust estimates of mass, radius, and, through stellar‑evolution models, age.

Data preparation includes careful quality cuts: (1) removal of stars with ν_max more than three sigma below 20 µHz to avoid contamination by early‑AGB objects whose ages are poorly constrained, and (2) exclusion of stars with Δν ≥ 21 µHz where the ν_max–Δν scaling relation breaks down, often due to Nyquist‑limit effects. These cuts are especially important for K2, where the 80‑day campaigns lead to higher noise and a larger fraction of ambiguous Δν measurements.

Gaia parallaxes are filtered using RUWE > 1.4 and binary flags, and zero‑point corrections are applied. The authors adopt the Lindegren et al. (2021) correction for Kepler and TESS, while a constant +17 µas offset is used for K2, reflecting the different systematic behaviours identified in previous studies. Sensitivity to the zero‑point choice is explicitly tested in the Bayesian inference stage.

Spectroscopic inputs are drawn from APOGEE (infrared, R ≈ 22 500) and GALAH (optical/IR, R ≈ 28 000). Quality flags (e.g., STAR_BAD, RV_FLAG for APOGEE; snr_c3_iraf, flap_sp for GALAH) are enforced, and minimum uncertainties of 0.05 dex in