[Paper Review] A 'Quick Look' at All-Sky Galactic Archeology with TESS: 158,000 Oscillating Red Giants from the MIT Quick-Look Pipeline

Marc Hon, Daniel Huber+7 more|arXiv (Cornell University)|Aug 2, 2021

Stellar, planetary, and galactic studies116 references77 citations

TL;DR

This paper presents the first near all-sky catalog of 158,505 oscillating red giants detected from TESS's prime mission data using a machine learning pipeline applied to long-cadence photometry. By leveraging Gaia parallaxes and colors, the authors derive effective temperatures, radii, and νmax values, enabling the creation of the first full-sky asteroseismic mass map that reveals large-scale Galactic structures and kinematic substructures, demonstrating TESS's transformative potential for Galactic archeology with only one month of data per target.

ABSTRACT

We present the first near all-sky yield of oscillating red giants from the prime mission data of NASA's Transiting Exoplanet Survey Satellite (TESS). We apply machine learning towards long-cadence TESS photometry from the first data release by the MIT Quick-Look Pipeline to automatically detect the presence of red giant oscillations in frequency power spectra. The detected targets are conservatively vetted to produce a total of 158,505 oscillating red giants, which is an order of magnitude increase over the yield from Kepler and K2 and a lower limit to the possible yield of oscillating giants across TESS's nominal mission. For each detected target, we report effective temperatures and radii derived from colors and Gaia parallaxes, as well as estimates of their frequency at maximum oscillation power. Using our measurements, we present the first near all-sky Gaia-asteroseismology mass map, which shows global structures consistent with the expected stellar populations of our Galaxy. To demonstrate the strong potential of TESS asteroseismology for Galactic archeology even with only one month of observations, we identify 354 new candidates for oscillating giants in the Galactic halo, display the vertical mass gradient of the Milky Way disk, and visualize correlations of stellar masses with kinematic phase space substructures, velocity dispersions, and $\alpha$-abundances.

Motivation & Objective

• To systematically detect oscillating red giants across the entire sky using TESS data, overcoming limitations of prior missions' restricted sky coverage.
• To develop and apply a machine learning pipeline trained on Kepler-like data to identify red giant oscillations in TESS's unique photometric data, despite lower frequency resolution.
• To produce a high-fidelity, near all-sky asteroseismic catalog with fundamental stellar parameters (Teff, R, νmax) for Galactic archeology.
• To demonstrate the feasibility of using short-duration TESS observations (27 days) for precise asteroseismic inference and large-scale Galactic structure mapping.

Proposed method

• A convolutional neural network (CNN) classifier is trained on power spectra images to detect the presence of oscillation power excess in red giants from TESS long-cadence light curves.
• A separate regression network estimates νmax using a negative log-likelihood loss function with explicit uncertainty estimation via an auxiliary output layer.
• The pipeline uses raw, instrument-corrected light curves from the MIT Quick-Look Pipeline (QLP) for all TESS sectors 1–26, up to TESS magnitude 13.5.
• Stellar parameters (Teff, R) are derived from Gaia parallaxes and TESS-Gaia color-color relations, enabling asteroseismic modeling.
• The method applies dropout regularization (p=0.1) to prevent overfitting, with early stopping and Adam optimization.
• Simulated power spectra generated with celerite are used to benchmark the network, though real data show higher noise and different low-frequency profiles.

Experimental results

Research questions

• RQ1 Can machine learning reliably detect oscillating red giants in TESS's all-sky, short-duration (27-day) observations despite lower frequency resolution than Kepler?
• RQ2 What is the yield of oscillating red giants across the full sky using TESS data, and how does it compare to previous missions like Kepler and K2?
• RQ3 Can νmax estimates from TESS data enable the construction of a large-scale, all-sky asteroseismic mass map consistent with known Galactic structure?
• RQ4 To what extent can TESS asteroseismology resolve kinematic and chemical substructures in the Milky Way disk with only one month of data per target?

Key findings

• The study delivers a catalog of 158,505 oscillating red giants—representing an order of magnitude increase over Kepler and K2 yields—making it the largest such sample to date.
• The authors produce the first near all-sky Gaia-asteroseismology mass map, revealing large-scale Galactic structures consistent with known stellar populations.
• The catalog reveals a clear vertical mass gradient in the Milky Way disk, with higher-mass stars concentrated in the mid-plane and lower-mass stars in the halo.
• The study identifies 354 new candidates for oscillating giants in the Galactic halo, demonstrating TESS’s sensitivity to faint, distant populations.
• Stellar masses derived from TESS νmax and Gaia data correlate strongly with kinematic substructures, velocity dispersions, and α-abundance patterns, enabling detailed Galactic archeology.
• Despite 27-day integration times, the method achieves median mass uncertainties of ~8% and age uncertainties of ~26% for bright red giants (G < 11), sufficient to resolve chemically and kinematically distinct Galactic components.