A new study looks at how machine learning can be used to decipher the early stages of supernova explosions by reconstructing the light emitted during the outburst. Eleonora Parrag, a Ph.D. student at the University of Cardiff, presented the research today at the 2022 National Astronomy Meeting.
Supernova explosions are some of nature’s brightest fireworks, produced by the most massive dying stars. These can be used to probe distances in space and answer questions about our universe, as well as to produce much of the material that makes up our world.
The physics governing a supernova change hundreds of days after its explosion; snapshots of this physics can be captured in terms of a supernova’s spectrum—where light is dispersed by wavelength, similar to how we see the colours in a rainbow. The signatures of the elements in the explosion are contained in spectra, which can reveal the conditions involved.
This, however, is a limited resource. More spectra would provide important information on the ever-changing physics surrounding supernovae, as well as a greater ability to compare and study their populations across cosmic time until the universe’s birth.
Parrag’s research focuses on filling in the gaps with machine learning, or algorithms that learn by being “trained” on existing observations of hundreds of supernovae. They can generate entire artificial spectra using only a few data points obtained from previously observed supernovae. Filling in the gaps for these existing data points enables the construction of a spectrum for any previous explosion up to around 200 days after the explosion.
The researchers discover that their synthetic spectra replicate many of the features seen in real supernova explosions.
“Machine learning can help us find patterns and potentially even new ideas in physics in the huge amounts of data from supernovae we can observe now and in the foreseeable future,” says project leader Eleonora Parrag. “It’s a really promising avenue to explore in astrophysics right now,” she adds, “and I’m very excited about what we might discover about supernovae in the future.”
Further research in this area will focus on applying this algorithm to all types of supernovae, improving the algorithm, and increasing the number and variety of supernovae used in training.