A galaxy collision is like a slow dance in which two partners gradually draw closer. Hidden at the center of each is a supermassive black hole — an invisible flywheel with a mass of millions or billions of suns. After the galaxies merge, these black holes begin moving toward each other, losing energy through friction with stars and gas. When the distance between them becomes tiny, the dance ends in a burst of gravitational waves — ripples in the fabric of spacetime. But how long this journey takes, from the first steps to the final chord, has remained a mystery.
The catch is that galaxies are not smooth spheres. They often contain stellar bars — elongated structures of stars that rotate as a single unit. Such a bar can interfere with the dance of the black holes: it can nudge them together or, conversely, fling them into a wider orbit. Previously, calculating this effect required long computer simulations for each pair individually. Running through thousands of cases was impossible.
The authors of the new study found a workaround. They trained a neural network on a large set of simplified calculations — a kind of training ground where a satellite black hole orbits inside the remnant of a merged galaxy. The network learned how the bar and the rotating stellar disk affect the orbit, and now it can predict the merger time almost instantly. The error is small, and the speed has increased by hundreds of times.
This emulator was applied for the first time to a whole population of galaxies taken from a cosmological simulation. The team accounted for whether a galaxy has a bar, how massive it is, and how fast it rotates. The result showed that stellar bars noticeably change the distribution of merger times. On average, the picture remains the same, but there are more pairs that come together either very quickly or very slowly. This matters for planning future observations — for example, for the space antenna LISA, which will detect gravitational waves from such mergers.
This approach, training fast neural network surrogates for complex simulations, opens the door to solving other astrophysical problems. The physics is understood for each individual case, but running through millions of variants used to hit a computational wall. Now the dance of black holes can be calculated for thousands of galaxies without losing years to waiting.
Supermassive black holes are so dense that a teaspoon of their material would weigh as much as Mount Everest — if such material could be scooped up.
In Carl Sagan's novel "Contact," aliens use gravitational waves from black hole mergers to transmit a message.