Lensed Mergers Can Be Pinned to a Galaxy's Nucleus

A strongly lensed binary black hole can be placed on the nucleus of a specific galaxy, and a paper on arXiv (2609.10285) reports that a single such measurement would be enough to reject the idea that black hole mergers simply follow the starlight. The rejection is quoted at p < 5%, and in most cases at p < 1%.

The p-values are a statement about statistics, not an observation. The paper shows what a lensed event plus follow-up imaging would let you conclude, and announces no particular merger.

Black hole binaries are theorised to form through various channels. The contrast this paper's test rests on is between two classes of them. One puts mergers wherever the stars are, spreading them through disk and bulge in proportion to the starlight; the paper calls that light-tracing. The other needs density, and the densest stellar environment in a galaxy is its nucleus, which is where a binary assembled that way stays. Whether the merger happened at the centre or not is what the test is built to answer.

Detectors cannot answer it. The paper states that current LIGO-Virgo-KAGRA sky locations make the direct study of formation channels a challenge. The rest of the argument is a workaround for that.

Lensing buys the position from geometry instead. When a galaxy or a cluster of galaxies sits between the merger and Earth, its mass splits the wavefront, and the same event arrives along more than one path: two or four copies, separated in time, with different amplitudes, and masses and spins that match. Locating a speaker in a dark room from its echoes works the same way. You never see the source. You know the shape of the walls, and the delays do the rest. Fit a mass model to the lens, and the pattern of image positions and arrival times inverts to a small patch in the source plane.

The chain runs like this. A candidate appears as two or four signals whose inferred masses and spins agree and whose arrival times fall inside a window the lens sets. The delays between those arrivals, and the ratios of their amplitudes, depend on where the line of sight cuts through the lens. The reconstruction is only as good as the delays it is built from: two images give one delay, four give three, and the extra constraints tighten the mass model. The mass doing the lensing is mostly dark matter, which cannot be seen, so it has to be inferred from the light of the lens galaxy and whatever else the image supplies. Fitting that mass distribution converts the measured delays into a probability distribution over where the merger sat in the source plane. That distribution is then laid over an image of the host. Identifying what sits at the reconstructed position takes high-resolution imaging, which puts imaging time into the follow-up chain for lensed candidates, alongside the detector network and the survey telescopes that find the hosts. If the distribution lands on the nucleus, the merger was nuclear. If it lands in a spiral arm, it was not.

How small that patch is decides whether the word nucleus means anything. Earlier work established that with current detectors at upgraded sensitivity, plus survey telescopes like Euclid, a lensed binary black hole can be tied to its host galaxy, with source-plane localisations reaching sub-galactic scales. The step reported here is the next one down: with imaging at Hubble resolution, the patch is small enough to sit on a structure inside the galaxy, demonstrated for galactic centres. What changes across the rows below is the size of the patch: a sky region, a galaxy, a structure inside it.

The logic behind the test is that a nucleus is a tiny fraction of a galaxy's light, so a merger that traced starlight would rarely reconstruct onto one. That reasoning is mine, read off the geometry; what the paper reports is the number it produces. Under a light-tracing hypothesis the reconstruction lands on the nucleus less than 5% of the time, and in most of its configurations less than 1%. The result is offered as a conservative basis for excluding light-tracing spatial distributions. Conservative reads as a claim about how the test was built, and the implication, mine rather than the paper's, is that non-rejection is the easier outcome and the test rejects anyway. What survives is a family of channels that need the centre: active galactic nuclei and nuclear stellar clusters.

Three things stay open. A nucleus picks out a class of channel rather than one mechanism, since active galactic nuclei and nuclear stellar clusters both sit at galactic centres and the paper supports them jointly. The source-plane position is inferred through a lens model and inherits that model's errors, so if a plausible error in the mass distribution slides the reconstruction off the nucleus, the rejection weakens. And the inference rests on merger location standing in for formation site, which is a reasonable proxy and still a proxy. A confirmed lensed merger reconstructed away from a nucleus would, on my reading rather than the paper's, be the cleanest challenge to this picture; the paper argues only the positive direction, that a centre-localised event supports centre-confined channels, and says nothing about what an off-centre reconstruction would establish.