Gaia found dormant black holes closer to their stars than physics allows

Three. That is how many dormant black holes ESA's Gaia has found in the Milky Way with a star still orbiting them, and two of those three should not be there at all.

Gaia does not see the black holes. It measures the tiny wobble a hidden mass imposes on a visible star, across billions of precisely tracked stellar positions, and works backwards to the thing doing the pulling.

What is wrong with two of them

Gaia BH3 behaves itself. It sits in a wide binary, roughly what you would expect from a star that collapsed without disturbing its neighbour.

Gaia BH1 and Gaia BH2 do not. Each holds a black hole of around nine solar masses paired with a much lighter companion, in an orbit far tighter than standard stellar evolution permits. The progenitor that eventually collapsed must have started out near twenty solar masses, and a star that size swells enormously on its way out. A companion orbiting that close should have been engulfed, and the pair should have merged long before there was a black hole in the system at all.

The companion is still there. That is the part that matters.

Roche lobe overflow, probably

The leading explanation is Roche lobe overflow. A star's Roche lobe is the region where its own gravity wins; beyond that boundary, the companion's pull dominates. When an expanding star's outer layers push past that line, the material transfers to the companion instead of swallowing it whole. Both BH1 and BH2 are expected to have gone through that phase, while BH3's orbit is wide enough to have skipped it.

It is a clean mechanism and it fits. It is also being tested against a sample of two systems, which is not a sample so much as an anecdote with error bars. Nobody involved is claiming otherwise.

What to watch

Gaia's later data releases are the source of more of these. The detection method scales with the number of precisely measured stellar positions, so the population of known dormant black holes should grow rather than plateau, and a handful more systems would turn the Roche lobe argument into something testable.

This is the useful kind of result. A textbook answer about how massive stars die has a hole in it, and the hole is a star that is still visible in a place where the model says nothing should have survived.

For more on what happens when the models and the observations disagree, see our pieces on three black holes in an early galaxy merger, what a black hole star actually is and the Saturn decagon Hubble spotted at the planet's south pole.