The Repository Absolute Magnitude Space

12,500 Red Giants Caught Gaia Overcorrecting Itself

September 3, 2026 · Runtime 12:08 · Watch on YouTube

Every star with a convective envelope is ringing right now. This episode measures distance by listening to that ring, never once touching a parallax -- and uses it to catch a spacecraft's own correction crossing zero.

Chapters

Transcript

Every star is ringing 0:00

Every star with a convective envelope is ringing right now. Not metaphorically. Sound waves, trapped inside, making the whole star swell and shrink by a fraction too small to see and impossible to hide from an instrument sensitive enough. Including the star you are sitting under, and the one you are about to hear.

A number that never touches a parallax 0:22

This episode ends on a number that never touches a parallax. Not once. It comes from listening to twelve thousand five hundred stars ring, and using that sound to catch a spacecraft's own correction overshooting the number it was trying to fix.

Last episode's zero point 0:38

Last episode ended on Gaia's own zero point: minus seventeen millionths of a second of arc, subtracted from every parallax in the catalogue. Tonight, a completely different method checks that number from the outside, without ever borrowing it.

What a red giant actually is 0:55

Start with what a red giant actually is. A star like the Sun spends most of its life burning hydrogen in a small, dense core. When that core runs out, the outer layers lose their support and swell, cool, and turn red. The star does not die. It inflates.

UY Scuti's gentler cousin 1:16

You already met a much more extreme cousin of this. UY Scuti, from episode one, is a red supergiant -- far more massive, and it will end in a supernova. A red giant is smaller and gentler. It is what the Sun itself will become, in about five billion years, and it is already breathing.

Sound trapped inside a star 1:38

Inside a red giant, turbulent convection kicks up sound waves, and the star's own structure traps them. They bounce back and forth through the whole interior and settle into resonance, the exact same physics as a struck bell finding its ringing tone. A resonating bell makes its surface visibly tremble.

The tremor, as brightness 1:59

So does a resonating star. The tremor shows up as a flicker in brightness, a few parts per million, rising and falling on a scale of hours. An instrument watching that flicker over weeks needs no assumption about the star at all to measure it. It just counts light.

Not noise, not rotation 2:18

That flicker is not noise, and it is not rotation, and it is not a transiting planet. It has a specific signature: an evenly spaced comb of frequencies, riding on top of the boiling surface's own random flickering. Nothing else in the data looks quite like a comb.

Two numbers, no assumption 2:37

From that flicker, two numbers come out directly. Nu max: the frequency where the ringing is loudest. Delta nu: the even spacing between its overtones. Both measured, neither assumed. The technique that finds them was not born out here. It was born much closer to home.

It started on the Sun 2:59

It was born on the Sun. Helioseismology measured the Sun's own ringing decades ago: nu max at three thousand ninety millionths of a hertz, a period of about five and a half minutes. Small, dense, fast. Now go find a much bigger bell.

Big bells ring low 3:17

A typical red giant rings at nu max of thirty two millionths of a hertz. That is a period of nearly nine hours, not five minutes. Big bells ring low, the same reason a full wine glass hums lower than an empty one, and here it is a star swollen to dozens of times the Sun's radius doing the humming.

Speeding up the ring ten million times 3:39

Speed that up by a factor of ten million and thirty two millionths of a hertz becomes three hundred twenty hertz, a low held note, with its overtones spaced forty hertz apart. Run the Sun through that exact same factor and its tone lands at thirty thousand nine hundred hertz -- above what a human ear can even register. Not a metaphor. The same resonance equation, run on a star instead of a bell.

From frequency to radius 4:06

Those two frequencies, plus the star's temperature from ordinary spectroscopy, plug into a scaling relation and return the star's radius and mass. No distance anywhere in that calculation. And the paper behind tonight's number did it more carefully than one formula.

What the instrument returned 4:24

Notice what just happened, because it is the whole show in miniature. The instrument returned a light curve. Two frequencies came out with no assumption attached. Everything after that -- radius, mass, brightness, distance -- is conversion, stacked layer by layer.

A grid of stellar models 4:44

Instead of a single equation, it fits a whole grid of stellar models to five numbers at once: nu max, delta nu, temperature, surface gravity, and chemical composition. The output is the same kind of answer, radius and mass, just checked against physics at every step instead of one shortcut formula.

No ladder underneath it 5:06

Radius and temperature give the star's true luminosity, from nothing more exotic than the law that relates heat and glow. Compare true luminosity to how bright the star looks, and the difference is distance. No angle. No parallax. No ladder underneath it at all.

Compare that to the ladder 5:26

Compare that to last episode's ladder. Parallax needs a measured angle. A Cepheid needs a parallax to calibrate its brightness. A supernova needs a Cepheid. This method needs none of the three. It only needs a star willing to ring.

158,505 oscillating giants 5:46

And this kind of ringing is not rare. TESS, scanning nearly the whole sky, found over one hundred fifty eight thousand oscillating red giants, flagged automatically by a machine-learning pipeline trained on Kepler stars. That is the raw supply. One study picked a very specific slice of it.

Why a machine had to listen 6:07

No team of people could have inspected that many power spectra by eye. The pipeline learned the comb-shaped signature from tens of thousands of confirmed Kepler examples, then searched the entire TESS sky for the same shape, automatically.

Khan et al., 2023 6:23

Khan and collaborators, in twenty twenty three, selected nearly twelve thousand five hundred red giants that had BOTH things measured: a sound-based distance, and a Gaia parallax. Same stars, two independent methods, side by side.

Three fields, three samples 6:40

The sample came from three different survey fields. Four thousand six hundred eighty seven stars from Kepler. Seven thousand and twenty four from K2, spread across seventeen separate observing campaigns. One thousand two hundred fifty three from TESS.

Checking the arithmetic 7:00

Add those three numbers and you get twelve thousand nine hundred sixty four stars. The paper's own summary rounds that down to nearly twelve thousand five hundred -- close enough for a headline, not close enough for this channel to repeat without checking.

Every star, measured twice 7:16

And every star's nu max and delta nu were extracted twice, by two independent measurement pipelines, just to make sure the numbers were not an artifact of one method's assumptions. The three fields did not agree with the Gaia correction the same way.

Two spectroscopic surveys 7:32

The temperature and chemistry inputs got the same treatment. Most of the K2 sample was cross-checked against two separate spectroscopic surveys, on two different continents. The resulting offsets agree to within a few millionths of a second of arc either way.

What offset means 7:50

Before the three results: here is what 'offset' means. Take the Gaia parallax for a star. Take its sound-based parallax. Subtract one from the other. If Gaia were perfect, and the correction were perfect, that difference would be zero, every time.

Kepler: close to zero 8:10

In Kepler, the correction works about as well as it could. The raw offset was twenty millionths of a second of arc. After applying the standard fix, it dropped to four tenths of one millionth. Almost exactly zero.

K2: past zero 8:27

K2 tells a different story. The raw offset there was negative eighteen. After the same correction, it became positive fifteen. It did not just fail to reach zero. It crossed zero and kept going -- an overcorrection, in the paper's own words.

It does not stay contained 8:46

An overcorrected parallax does not stay contained in one table. It ripples into anything built on top of it -- stellar ages, the shape of the galaxy's disk, and yes, the expansion rate this whole channel spent last episode arguing about.

TESS: barely moved 9:03

TESS is the third story. The raw offset was forty one millionths of a second of arc. After correction, thirty three. Barely moved. The same fix, three fields, three completely different outcomes.

Seventeen campaigns, one spread 9:21

K2 is not even one number. Across its seventeen separate campaigns, the raw offset ranges from minus forty three to plus two millionths of a second of arc. After correction, the range becomes minus six to plus thirty three. The spread did not shrink. It shifted.

Not an accusation 9:42

None of this is an accusation against Gaia. This is Gaia's own data being tested with an independent method, published specifically to improve the next correction. The teams involved say so directly: their model is not suitable for every field it was applied to.

The bridge to last episode 10:00

And it matters past this one spacecraft. Last episode's five sigma disagreement over the expansion of the universe rests, at its very base, on exactly this kind of correction. A method that never touches a parallax is one of the few ways to check that ladder from completely outside it.

Where each method wins 10:20

The paper even maps out where each method wins. For bright stars, Gaia is still more precise than sound ever will be. For faint stars, asteroseismology gets tight enough to expose exactly where Gaia's precision runs out.

The next Gaia release 10:37

Future Gaia releases will lean on exactly this kind of comparison to refine the correction further. A new data release is not a reset button. It is the same spacecraft, checked against sound one more time.

The measured number 10:52

So here is the number this episode ends on. Plus fifteen millionths of a second of arc. Not the size of an error. The size of an error that was supposed to go to zero, and instead walked straight through it, into the correction's own blind spot.

Ask if it overshoots 11:10

Here is the practical version. Do not just ask how much a correction changes a number. Ask whether it overshoots. A fix that crosses zero is not a smaller version of the original problem. It is a new one, wearing the old one's clothes. And ask which independent method could have caught it -- because for these stars, one already did.

Next: a shadow, 42 microarcseconds wide 11:35

Every number in this episode lived in the millionths of a second of arc. There is a smaller angle still being measured: forty two millionths of a second of arc, the width of a shadow cast by a black hole fifty five million light years away. It took a virtual telescope the size of the Earth to see it. Next time: how you build a telescope that big, without building a telescope that big. Absolute Magnitude Space. Every episode ends in a measured number.

Description and sources

A red giant has sound waves trapped inside it, resonating like a struck bell, and the resonance makes the star's brightness flicker by a few parts per million. From that flicker come two numbers with no assumption attached: nu max, the loudest frequency, and delta nu, the spacing between its overtones. The Sun does this too -- nu max of 3,090 millionths of a hertz, a five-and-a-half-minute breath. A typical red giant rings at 32 millionths of a hertz, a period of nearly nine hours: big bells ring low. Those two numbers, plus temperature, give radius and mass through a scaling relation, and radius plus temperature give true luminosity -- which compared to apparent brightness gives distance, with no parallax anywhere in the chain. Khan et al. (2023) built exactly this kind of distance for nearly 12,500 red giants from Kepler, K2 and TESS that also had Gaia parallaxes, and compared the two. In Kepler, the standard Gaia correction lands almost exactly on zero. In K2, it overshoots -- crossing from negative eighteen to positive fifteen millionths of a second of arc, an overcorrection in the paper's own words. In TESS, it barely moves a offset of minus forty-one at all. Same spacecraft, same fix, three different outcomes -- which is exactly why an independent, sound-based distance is worth having.

PRINT-READY, FROM THIS CHANNEL

The Conversion Chain, Vol. 01 — the method, and four numbers taken apart with it

https://therepository.gumroad.com/l/the-conversion-chain

PRIMARY SOURCES

Khan, S., Miglio, A., et al. (2023). Investigating Gaia EDR3 parallax systematics using asteroseismology of cool giant stars observed by Kepler, K2, and TESS. I. A&A 677, A150. arXiv:2304.07158 -- 12,500 red-giant distances; K2 offset -18 to +15 uas (overcorrection), TESS-SCVZ -41 to -33 uas, Kepler -20 to -0.4 uas.
Huber, D., et al. (2011). ApJ 743, 143 -- solar reference values nu_max = 3090 +/- 30 uHz, delta_nu = 135.1 +/- 0.1 uHz, from SOHO/VIRGO.
Bedding, T. R., et al. (2011). Nature 471, 608 -- red-clump stars accumulate near nu_max ~ 32 uHz, delta_nu ~ 4 uHz.
Hon, M., Huber, D., et al. (2021). ApJ 919, 131. arXiv:2108.01241 -- 158,505 oscillating red giants found across the TESS sky by a machine-learning pipeline.
Rodrigues, T. S., et al. (2017). MNRAS 467, 1433 -- the PARAM grid-based modelling code used to derive stellar parameters from asteroseismic and spectroscopic inputs.
Lindegren, L., et al. (2021). A&A 649, A2 and A4. arXiv:2012.01742 -- the Gaia EDR3 parallax zero-point correction model (referenced from episode 2).
Kjeldsen, H. & Bedding, T. R. (1995). A&A 293, 87 -- the scaling relations connecting nu_max, delta_nu and temperature to stellar radius and mass.

ARCHIVE PLATES: all 9 are credited on screen in the video and listed in this project's manifest.

Absolute Magnitude takes one number the internet repeats and asks what an instrument actually returned. Every episode ends on a measured figure with its source on screen, and says plainly which part was measured and which part was assumed.

Narration is synthesised speech. All footage is NASA public-domain material, credited on screen and listed above; every diagram is drawn in code from the figures in the cited papers, not generated by an image model. The sonification heard in this episode is generated from the published frequencies by this channel's own code, not from any third-party audio.

#Asteroseismology #RedGiant #Gaia #Astronomy #Space