Scrutable

A popular science article has just been published in Popular Mechanics suggesting that a study very strongly demonstrates that Newton-Einstein gravity breaks down at very low accelerations, of the order of 10E(-10) m/s². Such very low accelerations are hard to investigate in our solar system, but can be investigated in the case of very wide binary stars, which the Gaia instrument is giving us a lot of data on.

We are used to the idea that modifications to gravity might be of the order of ppm. But the paper seemingly demonstrates a modification required of the order of 30% to 40% for these very low accelerations. It's huge. It ought to be bleeding obvious whether it is there or not, assuming you can actually measure such very low accelerations with any credibility.

Much as we would like to believe someone has finally found What is Wrong With Physics, it is the kind of thing that very often turns out to be wrong. So I tried to see what might be going on here. The popular article refers to this 2023 scholarly article in the Astrophysical Journal, by Chae Kyu-Hyun of Sejong University.

Inevitably there was some debate around it, and Gaia continues to spurt out data. And so Chae went back and did further analysis, more carefully, on a bigger dataset, with collaborators. This resulted in a new paper continuing to assert these claims, in this 2024 article in Monthly Notices of the Astrophysical Society.

Not unnaturally, I remain sceptical. I have tried to find some commentary on it, but there doesn't seem to be a lot. Maybe when someone finds something so seemingly outrageous, sensible people steer clear, in the same way reputable historians don't comment on articles about the real King Arthur. It took a little searching to find a thin scatter of blog articles on it. This Big Think article seems to most clearly set out a potential origin of reasons why the author might have convinced himself of something very big that turns out not to be true. What this blog suggests is that for these very wide binaries, where Chae is finding a deviation, they orbit so very slowly, that in a mere decade of data you can hardly claim to have characterised their orbits. Measuring very low accelerations is indeed very difficult, even in systems seemingly designed to demonstrate them.

Any thoughts?

No.

Because there's no such thing as Newton-Einstein gravity. Newton's version of gravity is completely incompatible with Einstein's version.

And systematic errors are likely proportionally very large in these measurements of acceleration. Surrounding gas clouds, the ISM, outgassing of stars, stellar winds and magnetic fields* are likely affecting the results. Stars do not exist in a vacuum.

*Questions about the role of magnetic fields are obligatory following any presentation of astrophysical results. It's probably an old charter or something.

Some other things you'd have to rule out:

Undetected dim companions in the system, such as brown dwarfs, massive Jupiters or Neptune's or black holes.

Even more distant stars interacting with the binary.

Coronal mass ejections and other stellar instabilities that can produce net forces on the stars.

And all the above assumes that you have an accurate mass for both stars in the system. Typically the way you measure the mass of stars is from the acceleration they produce on each other or on their planets. So you can't then use those masses to say the acceleration is wrong.

dyqik wrote: Fri Jan 16, 2026 5:26 am No.

Because there's no such thing as Newton-Einstein gravity. Newton's version of gravity is completely incompatible with Einstein's version.

I understand your extreme distaste for the title of the author's paper, which he somehow got accepted under that title to a respectable - I guess - astrophysics journal. But I think that's a semantic issue, not a physics problem.

My guess that this horrible wording has arisen because the author used Newtonian theory to calculate his counterfactual, because for this range of values Newton produces accurate enough predictions to any level we could measure. But when he finds a measurement that is very different from his Newtonian counterfactual, it is different enough that he can realistically assert that also falsifies Einsteinian theory.

But thank you for the physics issues of why this is hard and potentially a mistake. I think I might have gleaned from that a better understanding of what is going on here.

You talk about the very serious issue of the circularity of the argument. The orbital characteristics depend upon the masses, so you can't know if the orbital characteristics are non-Newtonian, because normally you'd deduce the masses from the orbits. But I think they are - in a statistical sense - asserting that they know what the distribution of the masses ought to be, from doing that calculation to a large database of binary star, excluding the wide binaries. So I think what they then do is say that the measured distribution of orbital characteristics of wide binaries does not match the asserted distribution of object masses. But the potential problem of that is that the wide binaries might have a different mass distribution from the broader distribution of binary star masses. There is mention of calibration. And now with you saying this, I think this is the calibration issue that was being mentioned.

And you draw attention to a spectrum of other physics problems.

IvanV wrote: Fri Jan 16, 2026 11:28 am You talk about the very serious issue of the circularity of the argument. The orbital characteristics depend upon the masses, so you can't know if the orbital characteristics are non-Newtonian, because normally you'd deduce the masses from the orbits. But I think they are - in a statistical sense - asserting that they know what the distribution of the masses ought to be, from doing that calculation to a large database of binary star, excluding the wide binaries.

The astrophysics of star formation and determination of masses from star type and stellar age are absolutely not precise enough to determine mass to high enough precision to make any statement about the orbital periods expected. There's too many unobservable variables to guess the mass more accurately than within a factor of two. And more likely a factor of ten.

Wide binaries very likely have specific variations of star formation history and stellar evolution, which will place systematic errors on estimates of mass from general stellar evolution models.

The systematic errors I was mentioning are more to do with the accurate determination of the orbit from Gaia data. Gaia has only been operating for a short time, while the period of wide binaries is presumably much longer than the time it's been operating. So we've only seen a small portion of the wide binary orbits, which means we can't necessarily solve accurately for the shape of the orbit and projection effects.