In the fields of economics, biology and even space, we can all cite situations where somebody grabs too many resources from somebody else.
Now scientists have their eye on another culprit of stunting cosmic growth, which are supermassive black holes. These matter-sucking entities tend to form in the core of huge galaxies and are enormously huge — typically many millions of times the mass of our own sun and often stretching into billions of times more massive.
Supermassive black holes have been accused of robbing galaxies of the gas these star factories require to keep producing new stars. But just like a courtroom jury, scientists wanted to gather more evidence before jumping to conclusions. Now machine learning (a kind of artificial intelligence) is racking up more evidence that indeed, black holes are selfish on a cosmic scale.
Before diving into the new study, let’s talk a bit about star formation. That’s a metric of cosmic growth, because think about it — if stars aren’t being made, the universe isn’t generating new material through the cycle of star birth and star death (the most spectacular example being supernova explosions that sprinkle matter through the local neighborhood.)
A comprehensive survey of our neighborhood called the Sloan Digital Sky Survey showed that not all galaxies are “star-forming”, meaning that new stars are popping up at very slow rates at best. And just like any neighborhood on Earth without renewal of buildings and infrastructure, over time such galaxies will decay. That has huge implications for how the universe changes over time.
Enter machine learning to dive more into what’s happening in these quiet galaxies. Astronomers used three different tools (EAGLE, Illustris and IllustrisTNG) to probe the SDSS data. Next, a machine learning tool classified the galaxies into two categories, “star-forming” and “quiescent” (meaning, those galaxies where stars are dying off.)
The algorithm further grouped the galaxies by three metrics, to help astronomers better understand why the galaxies were star-forming or not. The categories included the mass of the supermassive black holes, the total mass of all known stars in each galaxy, and the mass of each galactic “dark matter halo” — referring to a huge structure in which more mass lurks. (Dark matter is invisible to conventional telescopes and is best found through its gravitational influence on other objects.)
The study picked these three types of mass because so far, these are some of our best guesses as to why galaxies enter “semi-retirement.” The new simulations showed that supermassive black hole mass is the most crucial factor among the three types, and what was seen in theory matches the observations we have to date.
While naturally, no single study can definitively prove what’s going on, the broader implication is the role of automation in helping us learn more about the universe. In the coming decades, we’ll surely rely more on machine learning to help us classify data and to help organize our thinking for other questions, like how quickly the universe is expanding or where to find potentially habitable planets.
For now, though, supermassive black holes will have to mount a good defense to refute the findings to date, so keep your eye out for more studies in this fascinating field. The research was presented at the Royal Astronomical Society’s National Astronomy Meeting by Joanna Piotrowska, a PhD student at the University of Cambridge.