It was solely a matter of time earlier than synthetic intelligence took on dark matter. A brand new deep-learning algorithm is about to be unleashed upon pictures of galaxy clusters in quest of the telltale indicators of this invisible substance that surprisingly makes up 85% of all matter in the universe.
In accordance with the standard model of cosmology, each galaxy is surrounded by a halo of darkish matter. Equally, galaxy clusters are suffused inside huge haloes of darkish matter, which we are able to detect not directly. Scientists are additionally in a position to decide darkish matter’s distribution in a cluster by expecting the best way its gravitational affect bends space, due to this fact creating weak, and typically robust, gravitational lenses. But, regardless of the massive volumes of darkish matter within the universe, no person is aware of what it’s constituted of.
Sometimes, two galaxy clusters — containing galaxies, scorching gasoline and darkish matter — can collide. When this occurs, how the collision proceeds relies on the character of darkish matter.
All of it comes all the way down to a property of darkish matter referred to as its interplay cross part, which refers back to the foundation by which darkish matter is an unidentified kind of particle. One of many causes astronomers have had a lot problem monitoring down the id of darkish matter is that it would not appear to work together with regular matter, aside from via gravity. Nonetheless, some fashions predict that particles of darkish matter can work together with one another, and to what extent this interplay takes place relies upon upon the interplay cross part.
Associated: Dark energy remains a mystery. Maybe AI can help crack the code
So, when two galaxy clusters collide, the destiny of their darkish matter haloes relies upon upon this cross part. If the worth of the cross part is excessive, the particles within the two darkish matter haloes which might be colliding will work together, slowing the darkish matter down. Galaxies, then again, will sail on via, hardly ever really “colliding” in the best way you might suppose due to the big areas inside stars and different objects inside them. In the meantime, large clouds of hydrogen within the cluster do collide, rising scorching and radiating X-rays.
If the worth of the interplay cross part is excessive, the darkish matter will separate from the galaxies and get distributed nearer to the recent gasoline clouds.
Alternatively, if darkish matter has a small cross part, then the darkish matter and galaxies can be separated, however not by as a lot, with the darkish matter discovered between the galaxies and the recent gasoline. If the cross part is zero, that means that darkish matter is collisionless, then we should always count on the darkish matter haloes to stick with the galaxies as they might cross proper via one another with out interacting in any respect.
Nonetheless, there are a number of problems. One is that we are able to see solely snapshots of galaxy cluster collisions as a result of they happen over time and distance scales which might be far too massive to disclose progress on human timescales. Moreover, we’re seeing these snapshots all at totally different levels of collisions and from totally different angles, so no two galaxy cluster mergers look precisely the identical, and it requires a educated eye to pluck what is going on from every instance.
A second complication is the impact that winds of radiation from galaxies with energetic black holes can have. These options are generally discovered within the largest galaxies inside a cluster, comparable to M87 within the Virgo galaxy cluster. These radiation winds, described as “suggestions” as a result of they immediately have an effect on what in the end instigates them, particularly matter falling in the direction of the central black gap. This suggestions can push matter out of a galaxy and into the extragalactic medium inside a galaxy cluster, in order that peculiar matter finally ends up the place the darkish matter could be anticipated to reside.
To assist distinguish among the many prospects, David Harvey of the Ecole Polytechnique Fédérale de Lausanne in Switzerland has written a deep-learning algorithm educated on simulated pictures of galaxy cluster collisions from the BAHAMAS (Baryons and Haloes of Huge Programs) undertaking performed by researchers from Liverpool John Moores College, Leiden College, Johns Hopkins College and CNRS in France.
The simulations mannequin galaxy cluster collisions with totally different cross-sectional values, and even these with no darkish matter in any respect.
Harvey examined totally different variations of his algorithm, which is a Convolutional Neural Community (CNN) in a position to acknowledge patterns in pictures very nicely. Harvey discovered that essentially the most complicated model of his algorithm, nicknamed “Inception,” was essentially the most correct, scoring an 80% success fee when challenged to characterize the simulated cluster collisions.
A number of tasks are already imaging galaxy cluster collisions in an try to unravel the thriller of darkish matter. The Hubble Space Telescope, with help from the Chandra X-ray Observatory, has been imaging galaxy cluster collisions for a while now, most famously the Bullet Cluster in 2006. Extra just lately, the European Space Agency launched the Euclid mission, which is designed to review the so-called “darkish universe” together with the presence of darkish matter in clusters. And on a smaller scale, the high-altitude balloon mission referred to as SuperBIT flew all over the world for 2 months in 2023 imaging galaxy cluster collisions, earlier than crash-landing in Argentina. With all this observational information, and extra to come back, Harvey’s “Inception” algorithm will assist us discover a sooner reply to the puzzle that’s darkish matter.
Harvey’s algorithm and its outcomes had been described on Sept. 6 in Nature Astronomy.
