A fully automated process, including a brand new artificial intelligence (AI) tool, successfully detected, identified and classified the first supernova. The new system, developed by an international collaboration led by Northwestern University, automates the entire search for new supernovae in the night sky, effectively excluding humans from the process. This not only rapidly speeds up the process of analyzing and classifying new supernova candidates, but also prevents human error.

This week, the team updated the astronomy community on the launch and success of a new tool called the Bright Transient Survey Bot (BTSbot). Over the past six years, people have spent nearly 2,200 hours visually examining and classifying supernova candidates. Thanks to a new tool now officially available online, researchers can redirect that valuable time to other responsibilities to accelerate the pace of discovery.

“For the first time, a series of robots and AI algorithms observed, then identified, and eventually communicated with another telescope to confirm the discovery of a supernova,” said Northwestern’s Adam Miller, who led the study. “This represents a significant step forward, as further refinement of the models will allow robots to distinguish specific subtypes of starbursts. Ultimately, taking humans out of the loop will give the research team time to analyze their observations and develop new hypotheses to explain the origin of the cosmic explosions we observe.” It will give more time.”

“We achieved the world’s first fully automated supernova detection, identification and classification,” added Northwestern’s Nabil Rehemtullah, who led the technology development with Miller. “This greatly simplifies large-scale studies of supernovae, helping us better understand the life cycles of stars and the origins of elements such as carbon, iron and gold that make up supernovae.”

Miller is an associate professor of physics and astronomy at Northwestern’s Weinberg College of Arts and Sciences and a member of the Center for Interdisciplinary Studies and Research in Astrophysics (CIERA). Rehemtullah is a graduate student in astronomy in Miller’s research group.

cutting out the middleman

Humans are now working hand in hand with robotic systems to detect and analyze supernovae. First, robotic telescopes repeatedly image the same areas of the night sky, looking for new sources not seen in previous images. Then, when these telescopes discover something new, people get involved.

“Automated software provides people with a list of potential explosions, which takes time to screen for candidates and make spectroscopic observations,” Miller said. “We can know for sure that a candidate is truly a supernova by collecting its spectrum – the scattered light of the source that reveals the elements present in the explosion. There are robotic telescopes that can collect spectra, but this is also often done by humans – telescopes with spectrographs.”

Researchers developed BTSbot to eliminate this human middleman. To develop the AI ​​tool, Rehemtulla developed a machine learning algorithm with more than 1.4 million historical images from approximately 16,000 sources, including confirmed supernovae, transient starbursts, periodic variable stars, and galaxy explosions.

“The Zwicky Transit Facility (ZTF) has been operating for the last six years, during which time I and others have spent more than 2,000 hours visually examining the candidates and determining which ones to observe with spectroscopy,” said astronomer Christopher Fremling. at the California Institute of Technology (Caltech), which has developed another AI tool called SNIascore and contributed to BTSbot. “Adding BTSbot to our workflow will eliminate our need to spend time vetting these candidates.”

Early wave of success and relief

To test BTSbot’s performance, the researchers turned to the recently discovered supernova candidate SN2023tyk. ZTF, a robotic observatory that scans the night sky for supernovae, first detected the source on October 3. Examining real-time ZTF data, BTSbot found SN2023tyk on October 5.

From there, BTSbot automatically requested the spectrum of the potential supernova at Palomar Observatory, where another robotic telescope (the SED Machine) was making in-depth observations to obtain the spectrum of the source. The SED machine then sent this spectrum to Caltech’s SNIascore to determine the type of supernova: white dwarf fusion or the collapse of the core of a massive star.

After determining that the candidate was a Type Ia supernova (a star explosion in which a white dwarf in a binary star system completely explodes), the automated system publicly shared the discovery with the astronomy community on October 7.

In the early days of BTSbot’s launch, Rehemtullah felt a mixture of nerves and excitement.

“The simulated performance was great, but you never know how it will turn out in the real world until you actually try it,” he said. “We felt great relief when the observations from SEDM and the automatic classification from SNIascore came. The beauty of this is that when everything works as it should, we do not actually do anything. We sleep at night, and in the morning we see BTSbot and other AIs determinedly doing their job. ”

The Northwestern-led collaboration included astronomers from the California Institute of Technology, the University of Minnesota, Liverpool John Moores University in England, and Stockholm University in Sweden.