A machine learning algorithm made it easier to understand the nature of dark matter • DigiCala Mag

Scientists hope to better understand the nature of dark matter by validating a machine learning algorithm developed to detect gravitational lenses.

Earlier this year, a machine learning algorithm identified about 5,000 possible gravitational lenses that could help map the evolution of galaxies since the Big Bang, and in turn identify a host of ancient galaxies.

Now, using the Keck Observatory in Hawaii and the European Very Large Telescope in Chile, the team of Kim-Vy Tran from ASTRO 3D and the University of New South Wales has identified 77 of these lenses. have evaluated gravity. They found that 68 of these are strong gravitational lenses that cause a significant magnification of cosmic distances.

This 88% success rate shows that the developed algorithm is reliable and can identify thousands of gravitational lenses. This is despite the fact that gravitational lenses are normally difficult to discover and so far only about a hundred of them have been routinely used.

The results of the study by Kim-Wee Tran and his team, recently published in the Astronomical Journal, provide spectroscopic data from the strongest gravitational lenses ever developed using convolutional neural networks. They were identified by data scientist Dr. Colin Jacobs at Astro-TREE and Swinburne University.

This research is part of the scientific campaign “Galaxy Evolution with Gravitational Lensing by Astro 3D” (AGEL). “Spectroscopy allowed us to make a 3D map of the gravitational lenses, so we know that these objects are indeed gravitational lenses and not just detected by chance,” Tran said of the findings.

“Our goal at AGEL is to confirm with spectroscopic data about 100 strong gravitational lenses seen throughout the year in the Northern and Southern Hemispheres,” he added.

Validation of this algorithm to search for specific digital signatures in the sky was made possible by an international collaboration of researchers from Australia, the United States, the United Kingdom, and Chile. “With this algorithm, we can verify thousands of gravitational lenses versus the few we have,” Tran noted.

The phenomenon of “gravitational lensing” was first predicted by Einstein, who stated that light around massive objects in space bends, similar to light passing through a lens. In this way, the image of very distant galaxies that cannot be seen under normal conditions will be visible.

While astronomers have been using this method for years to observe distant galaxies, the new algorithm is helping to discover more lenses. “These lenses are very small, so if you have a blurry image, you won’t be able to detect them,” Tran continued.

While providing a better view of objects millions of light-years away, these gravitational lenses help to measure how much light is bent, thereby determining the mass in that particular region of space.

The researcher from the University of New South Wales explained in this regard: “We know that most matter is dark and we know that the existence of matter bends light. “So if we can measure how much light bends, we know how much matter there must be.”

Having more gravitational lenses at different distances also gives us a more complete picture of the backward course of cosmic history, almost to the time of the Big Bang. “Having more magnifying lenses gives us a better chance of seeing distant objects and gives us a better census of very young distant galaxies,” Tran said.

“And between these very early galaxies and ours, many evolutionary events are happening,” he added. From the small star formation regions that turn the primordial gas into the first stars to the Sun in the Milky Way. And thus, with these lenses at different distances, we can look at different points in the history of the universe to understand the change process of various objects from the beginning of the universe until now.

Professor Stuart Wyithe from the University of Melbourne and director of AstroTreedy also pointed out that each gravitational lens tells us something new. “Gravitational lensing, apart from being objects of beauty, provides a window to study the mass distribution in very distant galaxies that cannot be seen by other methods,” he added. “By introducing a way to use this massive data set to search for new gravitational lenses, we have an opportunity to see how galaxies grow.”

Professor Karl Glazebrook, one of Swinburne’s scientists, said about this new finding: “This algorithm was developed by Dr. Colin Jacobs. He sifted through tens of millions of images of galaxies to find 5,000 samples. “We never thought the success rate would be so high.”

He added: “We are now imaging these gravitational lenses with the Hubble Space Telescope. Images that will include from attractive and beautiful galaxies to amazing and unimaginable images.

Dr. “Tucker Jones” (Tucker Jones) from the University of Davis, another collaborator of this study, also noted: “A big step has been taken in the direction of learning how galaxies are formed and the history of the universe.”

“Normally these primordial galaxies look like faint bubbles, but magnified by gravitational lensing, we can see their structure with greater clarity,” he added. “They are ideal targets for our most powerful telescopes and give us a glimpse into the early universe.”

And that’s how “thanks to the gravitational lensing effect, we can understand what these early galaxies look like, what they’re made of, and how they interact with their surroundings.”

Cover photo: a graphic design of dark matter
Credit: ASTRO 3D.SciTechDaily

Source: SciTechDaily