Cosmic leap: NASA’s Swift satellite and AI determine the distance of the most distant gamma-ray bursts

The advent of AI has been hailed by many as a societal turning point, as it opens up a universe of possibilities to improve almost every aspect of our lives.

Astronomers are now literally using AI to measure the expansion of our universe.

Two recent studies led by Maria Dainotti, visiting professor at UNLV’s Nevada Astrophysics Center and assistant professor at the National Astronomical Observatory of Japan (NAOJ), incorporated several machine learning models to add a new level of accuracy of distance measurements for gamma rays. Bursts (GRB) – the brightest and most violent explosions in the universe.

In just a few seconds, GRBs release the same amount of energy that our sun releases over its entire life. Thanks to their brightness, GRBs can be observed at several distances, including at the edges of the visible universe, and help astronomers in their quest for the oldest and most distant stars. But, due to the limitations of current technology, only a small percentage of known GRBs have all the observational features necessary to help astronomers calculate how far away they occurred.

Dainotti and his teams combined GRB data from NASA’s Neil Gehrels Swift Observatory with several machine learning models to overcome the limitations of current observing technology and, more precisely, estimate the proximity of GRBs whose distance is unknown. Since GRBs can be observed at both distant and relatively close distances, knowing where they occur can help scientists understand how stars change over time and how many GRBs can occur in a given space and time.

“This research pushes the boundaries of gamma astronomy and machine learning,” Dainotti said. “Follow-up research and innovation will help us achieve even more reliable results and allow us to answer some of the most pressing cosmological questions, including the earliest processes in our universe and its evolution over time.”

AI pushes the boundaries of deep space observation In a study, Dainotti and Aditya Narendra, a final-year doctoral student at Poland’s Jagiellonian University, used several machine learning methods to accurately measure distance GRBs observed by the Swift UltraViolet/Optical (UVOT) space telescope and ground-based telescopes, including the Subaru Telescope. Measurements were based only on other GRB properties not related to distance. The research was published May 23 in the Letters from astrophysical journals.

“The results of this study are so precise that we can determine, using the predicted distance, the number of GRBs in a given volume and time (called rate), which is very close to the actual observed estimates,” said Narendra .

Another study by Dainotti and international collaborators successfully measured the GRB distance with machine learning using data from NASA’s Swift X-ray Telescope (XRT), the afterglows of so-called long GRB. GRBs are thought to occur in different ways. Long GRBs occur when a massive star reaches the end of its life and explodes in a spectacular supernova. Another type, known as short GRBs, occurs when the remnants of dead stars, such as neutron stars, gravitationally merge and collide with each other.

Dainotti says the novelty of this approach comes from using multiple machine learning methods together to improve their collective predictive power. This method, called Superlearner, assigns to each algorithm a weight whose values ​​range from 0 to 1, each weight corresponding to the predictive power of this singular method.

“The advantage of Superlearner is that the final prediction always outperforms the singular models,” Dainotti said. “Superlearner is also used to rule out less predictive algorithms.”

This study, published on February 26 in The Astrophysical Journal, Supplement Seriesreliably estimates the distance of 154 long GRBs for which the distance is unknown and significantly increases the population of known distances among this burst type.

Answering confusing questions about GRB training

A third study, published on February 21 in the Letters from astrophysical journals and led by Stanford University astrophysicist Vahé Petrosian and Dainotti, used Swift X-ray data to answer puzzling questions by showing that the GRB rate – at least at small relative distances – does not follow the rate of star formation.

“This opens the possibility that long GRBs at small distances could be generated not by the collapse of massive stars but rather by the merger of very dense objects like neutron stars,” Petrosian said.

With support from NASA’s Swift Observatory Guest Investigator program (cycle 19), Dainotti and colleagues are currently working to make machine learning tools publicly available through an interactive web application.