Kepler stared at one patch of sky for four years and built the foundation of modern stellar seismology. That patch, though, pointed away from the crowded plane of the Milky Way, and it caught mostly the small, common red giants that outnumber everything else. Stars above three times the Sun's mass barely showed up. They are rarer, they burn through their lives faster, and they mostly live in the galactic plane Kepler was not looking at.
Noah J. Downing and six colleagues went looking for them with TESS, the survey satellite that scans nearly the whole sky instead of one field. Starting from large existing catalogs, the team used photometric and spectroscopic data to pre-select 227 candidate intermediate-mass giants. Then they did something unglamorous and important: they reprocessed the raw TESS images by hand, star by star, drawing custom apertures around each target and removing instrumental trends with what they call a boutique detrending method.
That hand work paid off. Against the MIT Quick Look Pipeline, the standard automated processing for TESS, their light curves showed a 12 percent average gain in the power-to-background ratio inside the oscillation envelope, meaning the stellar signal stood that much taller above the noise floor. The gain held even in the galactic plane, where stars sit so close together on the sky that light from neighbors bleeds into the measurement. Crowding is exactly the reason this region has been hard to work in, and it is exactly where the massive giants live.
What a ringing star tells you
Sun-like stars and red giants are not quiet objects. Convection churns their outer layers, and the turbulence rings the whole star like a struck bell, driving sound waves that travel through the interior and set the surface pulsing in and out by tiny amounts. Those pulsations show up as regular brightness variations. Because the frequencies depend on the star's internal structure, reading them backward gives the star's mass, radius and surface gravity with a precision no other method matches from this distance.
Of the 227 candidates, the team detected these oscillations in 98. Forty-three of the detections are stars above three solar masses. Ten exceed five solar masses, which the authors describe as among the highest-mass solar-like oscillators found so far.
The new measurements also caught a problem in someone else's numbers. Surface gravity, written as log g, is one of the basic quantities spectroscopists extract when they analyze starlight split into its colors. Comparing their seismic values against the APOGEE DR19 spectroscopic survey, Downing and colleagues found that APOGEE's log g runs systematically high by 0.23 dex on average. The authors offer a likely explanation rather than a certain one: APOGEE's pipeline was calibrated against Kepler asteroseismology, and Kepler saw almost no intermediate-mass giants. A calibration built on light stars appears not to carry over cleanly to heavy ones.
The last piece of the paper is a map for other people to follow. Applying the same pre-selection to TESS targets that have stellar parameters from Gaia XP spectra, the team identified up to 37,000 candidate intermediate-mass oscillators. None of those are confirmed detections. They are a target list, sized against a confirmed sample of 98.
Why it matters
Stars in this mass range do the heavy lifting in galactic chemical evolution. They live fast, fuse elements their lighter cousins never reach, and return that enriched material when they die. Yet the models describing how they mix material and how their cores behave have been tested mostly against stars two or three times lighter, because those are the ones that had precise measurements. A sample of 43 giants above three solar masses, with masses and radii measured directly from their own oscillations, gives theorists something to check the models against.
The log g offset has a wider reach. Spectroscopic surveys measure hundreds of thousands of stars, far more than asteroseismology ever will, and their calibrations are only as good as the reference sample underneath. A 0.23 dex systematic in a whole mass range is the kind of quiet error that propagates into stellar ages, distances and any population study built on top of them. Finding it required a set of stars the calibrators did not have.
There is also a practical lesson in the detrending result. A 12 percent improvement in signal strength sounds modest, and for any single star it may be. Spread across a crowded field where automated pipelines struggle, it is the difference between a detection and a non-detection for stars near the threshold. Careful reprocessing recovered signals that were already sitting in data the community has held for years.
This is a preprint, posted to arXiv on 30 July 2026 and not yet through peer review. The detections and the offset should be read as the authors' current findings, subject to that check.