AI Detects Rare Supernova SN 2023zkd in Real Time

A rare supernova named SN 2023zkd has been discovered by astronomers with the help of artificial intelligence (AI).

AI is now extensively used across industries such as finance, healthcare, manufacturing, cybersecurity, disaster management, and customer service. It is transforming the world by automating repetitive tasks, operating continuously without fatigue, processing and analyzing vast amounts of complex data, minimizing errors, and improving decision-making.

The technology, which refers to the ability of machines to carry out tasks usually related to human intelligence, is also helping us detect new objects in the sky and deepen our understanding of the universe.

For instance, a couple of months ago, AI helped an international team of astronomers unravel the mysteries of the universe by assessing data about black holes. Using more than 12 million simulations, researchers found that the central black hole of the Milky Way is revolving at nearly its maximum speed.

To achieve this, the team utilized synthetic simulations to train the network, enabling the discovery of new cosmic insights about black holes.

The network was trained to uncover information from the data behind the images of the black hole Sagittarius A* that was presented by the Event Horizon Telescope (EHT) Collaboration in 2022.

While previous studies used only a small amount of realistic synthetic data, the latest study fed millions of files into a Bayesian neural network, enabled by the high-throughput computing capabilities of CHTC, to make a more accurate comparison between the EHT data and the models.

The neural network suggested that the black hole is spinning at almost top speed, and the emission nearby is not caused by a jet but by extremely hot electrons in the surrounding accretion disk. Moreover, the magnetic fields in the disk appear to behave differently from what has been previously suspected.

According to lead researcher Michael Janssen of Radboud University Nijmegen:

“That we are defying the prevailing theory is, of course, exciting. However, I see our AI and machine learning approach primarily as a first step. Next, we will improve and extend the associated models and simulations.”

Another study from last year used AI to determine five cosmological parameters or the original “settings of the universe¹” with precision. These parameters, extracted from a dataset (SDSS) containing information about more than 100,000 galaxies, determine how the universe operates on the largest scales.

To extract valuable information out of the dataset, the team has to first train the AI on what to look for, and for that, they created 2,000 simulated universes, each one with different cosmological settings and real-life challenges experienced in galaxy surveys. 

Then, it was fed real data from the SDSS Baryon Oscillation Spectroscopic Survey, providing impressive results that make it possible for astronomers to do more with less data and push the boundaries of astronomy.

Now, in the latest discovery, scientists used a new AI system called Lightcurve Anomaly Identification and Similarity Search (LAISS), which is based on the Spotify algorithm.

Created by the global research university, UC Santa Cruz, the system compares the features of supernova SN 2023zkd against a dataset of known objects to detect any anomaly, which may indicate a rare phenomenon. When the AI finds something worth checking out, it sends candidates to researchers for analysis in real time.

Using that AI system, the astronomers were able to identify the supernova within hours of its explosion. A supernova is an extremely powerful and bright explosion of a star, which is among the most luminous events in the universe that can temporarily outshine entire galaxies. 

Supernovas make for an essential part of the cosmos, as they create and spread heavy elements, which are the fundamental building blocks for future stars, planets, and life.

These are brief events, but before the SN 2023zkd faded, the researchers were able to make rapid follow-up observations. Evidence suggests this particular explosion was the result of a massive star having a catastrophic encounter with a black hole companion.

The companion either swallowed the star partially or tore it apart before it could explode on its own.

Notably, the AI used by the astronomers to find fleeting anomalies in real-time, on the regular, and ‘without relying on luck,’ could also be used for medical diagnostics, financial-fraud prevention, and national security, showing the system’s versatility and wide-ranging capabilities.

Catching Cosmic Explosions in Their Earliest Moments

The latest discovery was reported this month by an astronomy collaboration, which noted a blast of a huge star confined in an orbit with a black hole. This discovery was made with the help of an AI system, which actively looks for stars immediately after they blow up.

The explosion’s name is SN 2023zkd, and it was first detected two years ago by the Zwicky Transient Facility. It was identified through a brand new AI model, which was designed to flag any unusual blasts or cosmic events in real time.

Having an early alert enabled the team of researchers to start follow-up observations right away, which is an important step to capture the explosion from its earliest stages, cover its full story, and find its origins.

Once the explosion was over, it was observed by telescopes from space as well as on the ground. In this case, two telescopes at Hawaii’s astronomical research observatory (Haleakalāa) were used by the UC Santa Cruz-based Young Supernova Experiment (YSE).

“Something exactly like this supernova has not been seen before, so it might be very rare.”

– Ryan Foley, Associate professor of astronomy and astrophysics at UC Santa Cruz

While humans are also good at spotting things that ‘aren’t like the others,’ the AI algorithm, he noted, can flag them much earlier than we might notice, and that is critical for time-sensitive observations.

Foley’s team actually runs YSE, a time-domain survey designed to discover new supernovae (SNe) and other cosmic transients within hours or days of their explosion. It primarily operates with Pan-STARRS telescopes that will soon be surveying 1500 square degrees of sky every three days. 

According to the official website, YSE’s goal is to find statistical samples of young, red, and rare transients. It also aims to better understand black hole variability.

Surveying about 4% of the night sky every three days has allowed the team to discover thousands of new cosmic explosions, with dozens of them just days or hours after the explosion. 

Now, they have discovered something interesting about SN 2023zkd, which was detailed in the study titled ‘Evidence for an Instability-induced Binary Merger in the Double-peaked, Helium-rich Type IIn Supernova 2023zkd.’² Published in the Astrophysical Journal, the study is funded by NASA, the National Science Foundation, the Moore Foundation, and the Packard Foundation.

According to the astronomers behind the latest discovery, a collision between the massive star and the black hole was bound to happen.

The star was locked in an orbit with the black hole, and as the energy was lost from the orbit, they got closer to each other. Their separation continued to decrease while the black hole’s intense gravity pulled dust and gas from the star into a disk.

This continued on, and before the star could devour the dense black hole, the latter’s gravitational stress caused the star to explode.

When the explosion collided with the shells of material from earlier interactions, located above and below the disk, it powered a dramatic re-brightening event.

According to Alexander Gagliano, the study’s lead author and a fellow at the NSF Institute for Artificial Intelligence and Fundamental Interactions:

“Our analysis shows that the blast was sparked by a catastrophic encounter with a black hole companion, and is the strongest evidence to date that such close interactions can actually detonate a star.” 

While it has been known that most massive stars are in binaries, Gagliano noted that “catching one in the act of exchanging mass shortly before it explodes is incredibly rare.”

But this isn’t the only interpretation. The team actually contemplated multiple originating scenarios for the supernova.

The other one, as per the team of scientists, is that the black hole ripped the massive star completely before it could explode on its own, through a process called “spaghettification.” The black hole then pulled in the debris of the star. When that debris crashed into the gas surrounding it, it produced the bright light. Data does not suggest this to be the case as strongly.

In both scenarios, the heavier black hole is the only thing that’s left behind. As per the study, the brightening optical precursor and characteristics of the explosion are found to be most consistent with an enormous and halfway stripped He star going through an instability-induced merger with a black hole companion.

Click here to learn all about the James Webb Space Telescope.

The Strange Life of Supernova SN 2023zkd

SN 2023zkd is located roughly 730 million light-years from our planet Earth. At first, it looked just like any supernova, a single burst of light that slowly fades over time. But that wasn’t the case. 

As astronomers continued to monitor the decline of SN 2023zkd over many months, they found that the supernova brightened yet again. So, the team turned to study the archival data to get insight into this unusual behavior, which revealed another unique characteristic.

Ultraviolet to infrared observations of the extraordinary supernova showed persistent and luminous precursor emission spanning several years prior to the discovery, followed by a second phase of slow brightening in its concluding year.

After the discovery, the supernova showed two comparable brightness peaks separated by 240 days, or about eight months.

Spectroscopically, it displayed strongly asymmetric, multicomponent Balmer and He I profiles. These are specific spectral lines of hydrogen (H) and helium (He) observed in stellar spectra, used to determine the age and composition of stellar populations in galaxies and other cosmic objects.

So, before the explosion even happened, the system was already slowly rising in brightness for over four years, or about 1,500 days. And this kind of prolonged activity before the explosion is seldom seen in supernovae. 

The star basically experienced two major eruptions before it died. A detailed analysis also indicated that the explosion light was the result of the material the star had shed in the years before it died. 

“2023zkd shows some of the clearest signs we’ve seen of a massive star interacting with a companion in the years before explosion,” said the study’s co-author, Ashley Villar, who’s an assistant professor of astronomy at Harvard University in Massachusetts. “We think this might be part of a whole class of hidden explosions that AI will help us discover.”

So, the brightening that occurred before the explosion was the result of the supernova’s shockwave slamming into low-density gas. Another peak, months later, was the result of a slow and sustained impact with the dense, disk-like cloud.

This particular structure, along with pre-explosion behavior, points to the dying star being under extreme gravitational stress, possibly from a nearby, compact companion like a black hole. After much discussion, the team came up with the explanation that it was a binary system with a black hole.

In order to make sure that the observations are aligned with their explanation, the team built the system and methodically demonstrated it as such.

The team “built the software platform that we use to consolidate data and manage observations. The AI tools used for this study are integrated into this software ecosystem,” Foley said.

While the latest study shows the importance of AI in spotting rare cosmic events in real-time, the astronomers also point to facilities like the Vera C. Rubin Observatory and the significant role they can play over the next decade. 

Formerly known as the Large Synoptic Survey Telescope (LSST), the observatory is located in the Chilean Andes mountains and is equipped with an 8.4-meter telescope and the largest digital camera to document the entire southern sky every few nights. Its goal is to understand the nature of dark matter, create an inventory of objects like asteroids and comets in the Solar System, explore black holes and exploding stars, and map the Milky Way.

With the upcoming Legacy Survey for Space and Time from Rubin Observatory expected to discover ∼105 yr−1, the study noted that the photometric (concerned with measuring the intensity or flux of light radiated by astronomical objects) samples of SNe IIn will dramatically increase this year. The study stated:

“Algorithms designed to flag these long-lived and re-brightening transients will play a critical role in characterizing the full breadth of strongly interacting events.”

Observations from the Rubin Observatory, combined with real-time AI detection, will enable astronomers to discover and study rarer, complex events, helping us better understand how massive stars live and die in binary systems.

“We’re now entering an era where we can automatically catch these rare events as they happen, not just after the fact. That means we can finally start connecting the dots between how a star lives and how it dies, and that’s incredibly exciting.”

– Gagliano

Meanwhile, Foley noted that while predicting the AI’s path is difficult, it is still advanced and has many uses that go beyond astronomy. He said:

“You can easily imagine similar techniques being used to screen for diseases, focus attention on terrorist attacks, treat mental health issues early, and detect financial fraud. Anywhere real-time detection of anomalies could be useful, these techniques will likely eventually play a role.”

Swipe to scroll →

Investing in Space Tech

While several public companies are strategically positioned in the field of AI space exploration technology, Northrop Grumman (NOC ) stands out as a key contractor for space missions.

This includes the largest and most complex space telescope ever built. NASA’s James Webb Space Telescope was built in partnership with Northrop Grumman, which led the design, development, and system integration of the observatory. In 2022, the telescope revealed its first image.

Northrop Grumman (NOC )

The company has also been expanding its use of AI in space to improve spacecraft operations. By developing AI robotics technology, it aims to enable operations in highly complex environments, including future space missions. Agentic AI is further planned for application across all phases of spacecraft operations.

In general, the global aerospace and defense technology company operates through a few key segments: Space Systems, Mission Systems, Defense Systems, and Aeronautics Systems.

It has a market cap of $84.8 bln with NOC shares, as of writing, trading at $592.44, up 26.24% year-to-date (YTD). It has an EPS (TTM) of 25.36 and a P/E (TTM) of 23.36. Northrop Grumman shareholders enjoy a dividend yield of 1.56%.

As for its financials, for Q2 of 2025, the company reported sales of $10.4 billion. Net earnings for the period came in at $1.2 billion, or $8.15 per diluted share. 

Its operating income was $335 million, net cash from operating activities was $557 million, and free cash flow was $468 million. Net awards in the quarter totaled $7.4 billion, while total backlog was $89.7 billion. 

“We are working with our customers to accelerate capability delivery to enable their vision of peace through strength. We continue to see growing demand globally for our broad range of product offerings.”

– CEO Kathy Warden

The company also returned over $700 million to shareholders through share repurchases and dividends.

AI’s Role in Astronomy and Beyond

AI is transforming industries, including astronomy, where it has become a critical tool for enabling scientists to capture rare and fleeting cosmic events like SN 2023zkd in real time, something nearly impossible just a few years ago.

As AI tools combine with massive sky surveys and observatories like Rubin to open the door to many more discoveries, the fact that these same techniques can also be applied to medicine, finance, national security, and beyond highlights their enormous crossover potential, signaling a new era of innovation.

Click here to learn all about investing in artificial intelligence.

References:

1. Hahn, C., Lemos, P., Parker, L., et al. Cosmological constraints from non-Gaussian and nonlinear galaxy clustering using the SimBIG inference framework. Nature Astronomy, 8, 1457–1467, published 21 August 2024. https://doi.org/10.1038/s41550-024-02344-2
2. Gagliano, A., Villar, V. A., Matsumoto, T., Jones, D. O., Ransome, C. L., Nugent, A. E., Hiramatsu, D., Auchettl, K., Tsuna, D., Dong, Y., et al. Evidence for an instability-induced binary merger in the double-peaked, helium-rich Type IIn supernova 2023zkd. The Astrophysical Journal, 989, 182, published 13 August 2025. https://doi.org/10.3847/1538-4357/adea38