Caltech astronomers used a machine learning algorithm to classify 1,000 supernovae completely independently. The algorithm was applied to data captured by the Zwicky Transient Facility, or ZTF, a sky survey instrument based at Caltech’s Palomar Observatory.
says Christopher Fremling, an astronomer at Caltech and the brains behind the new algorithm, dubbed SNIascore. “SNIascore ranked the first supernova in April 2021, and a year and a half later, we hit the impressive milestone of 1,000 supernovae.”
The ZTF scans the night sky each night to look for changes called transient events. This includes everything from moving asteroids to star-eating black holes to exploding stars known as supernovae. The ZTF sends hundreds of thousands of alerts each night to astronomers around the world, notifying them of these transient events. Then astronomers use other telescopes to follow and examine the nature of the changing objects. To date, ZTF data has led to the discovery of thousands of supernovae.
But with relentless amounts of data pouring in every night, ZTF team members can’t sort through all the data themselves.
“The traditional idea of an astronomer sitting in an observatory and sifting through telescope images carries a lot of romance but drifts far from reality,” says Matthew Graham, project scientist at ZTF and professor of astronomy at Caltech.
Instead, the team developed machine learning algorithms to aid in the searches. They developed a SNIascore for the task of classifying candidate supernovae. Supernovas come in two broad categories: type I and type II. Type I supernovae are devoid of hydrogen, while type II supernovae are rich in hydrogen. The most common type 1 supernova occurs when a massive star steals matter from a neighboring star, triggering a thermonuclear explosion. A type II supernova occurs when a massive star collapses under its own gravity.
Currently, SNIascore can rank what are known as Type Ia supernovae, or “standard candles” in the sky. These are dying stars that explode in a thermonuclear explosion of constant force. Type Ia supernovae allow astronomers to measure the expansion rate of the universe. Fremling and his colleagues are currently working on expanding the algorithm’s capabilities to classify other types of supernovae in the near future.
Each night, after the ZTF picks up flashes in the sky that could be supernovae, it sends the data to a spectrometer at Palomar located in a dome a few hundred meters away, called a SEDM (Spectral Energy Distribution Machine). SNIascore works with SEDM to classify supernovae as likely Type Ia. The upshot is that the ZTF team is quickly building a more reliable data set of supernovae for astronomers to investigate further and eventually learn about the physics of powerful stellar explosions.
“SNIascore is remarkably accurate. After 1,000 supernovae, we’ve seen how the algorithm works in the real world,” Fremling says. “We have not found any clearly misclassified events since their launch in April 2021, and we now plan to apply the same algorithm with other monitoring facilities.”
Ashish Mahapal, who leads ZTF’s machine learning activities and serves as principal computational and data scientist at Caltech’s Center for Data Driven Discoveries, adds, “This work nicely illustrates how machine learning applications are developing in astronomy in near real time.”
To learn more about the new algorithm, read the full story on the ZTF website.
ZTF is funded by NSF and Partners International. Additional support comes from the Heising-Simons Foundation and the California Institute of Technology. ZTF data is processed and archived by IPAC.