A study of 104,688 quasars found a pattern consistent with supermassive black holes recoiling after mergers.Larger estimated velocity offsets were associated with greater dust reddening, matching a prediction from earlier simulations.The evidence is statistical, and alternative explanations prevent researchers from treating it as proof of black-hole ejection.
Some supermassive black holes may carry the aftermath of a violent merger through their galaxies. An analysis of 104,688 quasars found that objects with larger estimated black-hole velocity offsets tend to appear redder from obscuring dust. That relationship matches a prediction for black holes displaced from galactic centers by gravitational waves.
The distinction is essential because the measurements do not directly track black holes moving through space. They compare light from different regions around quasars and use color as an indicator of dust. Recoiling black holes offer a natural explanation, but other physical processes could contribute to the pattern.
Distribution of Δv and Δ(g − i) values for ~100,000 QSOs with available Hβ_br after quality and consistency cuts. The blue point and lines mark zero values of both parameters. Dashed lines show the 2nd, 50th, and 98th percentiles (corresponding to a 2-σ range). (CREDIT: Bence Bécsy et al, Monthly Notices of the Royal Astronomical Society 2026) How a merger can kick a black hole
Nearly all massive galaxies harbor a supermassive black hole at their center. When galaxies merge, their central black holes can form a binary system and eventually combine. During the final stages, gravitational waves carry energy and momentum away from the pair.
If those waves carry momentum unevenly in different directions, the surviving black hole recoils in response. The strength of that kick depends on properties including the original black holes’ masses and spins. Such motion is a prediction of general relativity, although establishing it observationally remains difficult.
A displaced black hole need not escape its galaxy completely. If gravity keeps it bound, it can move back and forth around the galactic center while gradually losing energy. Along that path, its speed is greatest near the center, where surrounding dust can obscure its light.
That motion provided the basis for simulations by Peter Raffai, Zoltán Haiman and Zsolt Frei. Their theoretical paper appeared online in 2015, ahead of its 2016 journal issue. It predicted a population-wide link between a recoiling black hole’s speed along the observer’s sightline and the intervening dust.
Reading motion and dust in quasar light
Quasars are exceptionally bright galactic nuclei powered by gas accreting onto supermassive black holes. A recoiling black hole can retain some nearby gas, allowing it to remain luminous after leaving the center. That gas may also provide a way to estimate the black hole’s motion relative to its surroundings.
Δ(g − i) relative colour (proxy for dust reddening) as a function of the line-of-sight peculiar velocity of the BLR relative to the NLR, Δv, obtained from the broad Hβ emission line. (CREDIT: Bence Bécsy et al, Monthly Notices of the Royal Astronomical Society 2026)
Gas close to a black hole produces broad features in its spectrum, the pattern of light spread across wavelengths. More distant gas produces narrower features that can help establish a reference velocity. Comparing their wavelength shifts gives an estimate of relative motion toward or away from Earth.
For this analysis, the team used the broad component of a hydrogen feature called H-beta. Its reference combined available oxygen emission features and a calcium absorption feature associated with stars in the host galaxy. These measurements estimate one component of motion, rather than a black hole’s full speed through space.
The researchers also compared each quasar’s color with others at the same redshift, which measures the stretching of light. That comparison accounts for systematic color changes as spectral features move through observing filters. Unusually red objects can indicate intervening dust, although quasars also differ naturally in their emitted colors.
A small relationship in a large sample
The starting catalog, from the Sloan Digital Sky Survey’s sixteenth data release, contained 750,414 quasars. Spectral coverage, measurement quality and additional selection requirements reduced the main sample to 104,688. The team excluded very small velocity offsets and extreme values vulnerable to fitting artifacts.
Individual measurements showed considerable scatter, but averages revealed a consistent trend. Quasars with larger velocity offsets in either direction tended to show more dust reddening. A statistical rank test produced a correlation coefficient of 0.118, indicating a modest relationship despite its very high statistical significance.
Fraction of dust-reddened QSOs with Δ(g − i) above various κ thresholds as a function of Hβ-derived Δv. These show a similar trend as the mean Δ(g − i) (cf. Fig. 2), with increasing fraction of dust-reddened QSOs at large |Δv| values.(CREDIT: Bence Bécsy et al, Monthly Notices of the Royal Astronomical Society 2026)
A large sample makes it possible to identify a weak relationship that would be difficult to recognize in individual objects. High statistical significance describes the evidence against a chance correlation under the statistical test, rather than the strength of the relationship itself. It also does not establish which physical mechanism produced the pattern, a separate question requiring further investigation.
The difference became clearer when the team examined particularly dust-reddened objects. With a strict color threshold, their estimated fraction rose from about 6% near zero velocity to around 20% near 2,700 kilometers per second. Those approximate values describe the trend across velocity bins, rather than directly measured black-hole speeds.
“It is very difficult to prove that the black hole in a single quasar has been ejected, because other phenomena can produce similar signatures,” Bécsy said. “What is new about our research is that, rather than looking for individual candidates, we searched statistically for traces of an entire population.” The result therefore supports an interpretation that still needs independent testing.
Tests strengthen the signal but leave questions
The positive relationship persisted when the researchers changed how they estimated wavelength shifts. It also survived changes to the minimum velocity offset used for inclusion. Those checks suggest the finding does not depend on one particular measurement procedure or selection threshold.
Comparisons using only the reference spectral features did not reproduce the same positive trend. Some instead showed weak negative relationships, and the reference velocities sometimes disagreed substantially. That result supports the distinctiveness of the main signal while also exposing uncertainties in the reference measurements.
Δ(g − i) relative colour (proxy for dust reddening) as a function of the absolute value of the line-of-sight peculiar velocity of the BLR relative to the NLR, |Δv|, obtained from the broad Hβ emission line. (CREDIT: Bence Bécsy et al, Monthly Notices of the Royal Astronomical Society 2026)
Another complication involved objects with shifts toward Earth compared with those showing shifts away. Their average dust reddening differed in a way the simple recoil picture did not readily explain. Changes in measurement methods reduced some of that difference but did not eliminate it.
Gas winds, inflows, rotating structures and orbital motion in black-hole binaries can also shift spectral features. Uneven gas distributions or viewing angles may further complicate the interpretation. The observed correlation cannot by itself establish that recoil caused those shifts, much less prove that black holes escaped their galaxies.
What the pattern could reveal about mergers
If recoil produces the signal, it could help researchers investigate how frequently supermassive black holes merge. A preliminary comparison with earlier simulations suggests roughly half the quasars might have experienced a relatively recent merger. That is a conditional estimate, not a count of confirmed displaced black holes.
The estimate depends strongly on the simulated population and how long a recoiling system remains observable as a quasar. Uncertainties in those assumptions translate into large uncertainties in inferred merger rates. Better population models are needed before the relationship can provide a reliable census of past mergers.
Such measurements could eventually complement the planned Laser Interferometer Space Antenna, or LISA, which will study gravitational waves from space. The proposed follow-up work is more immediate: repeat the analysis with other broad spectral features and improve dust estimates using spectra. Establishing whether the same pattern survives those tests would help distinguish genuine black-hole recoil from moving gas and measurement biases.
Dig deeper into recoiling black holes
These resources explain the observations, predictions and instruments behind the search for displaced black holes.
Related Stories