New AI Method Promises to Revolutionize Cosmology by Harnessing Vast Supernova Data
Trieste, Italy & London, UK — A groundbreaking artificial intelligence method is poised to transform how scientists measure the universe, enabling the analysis of hundreds of thousands of supernovae without the need for prohibitively time-consuming spectroscopy.
Developed by researchers at SISSA, Imperial College London, and the University of Barcelona, the new AI technique—named CIGaRS (Combined Inference and Galaxy-Related Standardisation) —leverages neural networks to extract precise cosmic distances using only brightness data.
"Collecting detailed, homogeneous spectra at multiple epochs for very large samples will be impossible, given the sheer volume of data expected in the coming years."
— Professor Roberto Trotta, Imperial College London and SISSA
The Core Innovation: Simultaneous Analysis
Traditionally, scientists have used Type Ia supernovae as "standardizable candles" to measure cosmic distances. However, their light can be warped by a host of factors, including the properties of the progenitor star, interstellar dust, and the characteristics of the host galaxy.
Classic methods require detailed spectroscopic observations to correct for these effects—a process that is both expensive and slow. Next-generation surveys, such as the Vera C. Rubin Observatory, are expected to discover hundreds of thousands of supernovae per year, making full spectroscopic follow-up impossible.
CIGaRS overcomes this bottleneck by using neural networks to model supernova properties, galactic environments, and cosmological parameters all within a single, unified framework.
Proven Performance
The team tested CIGaRS on two major catalogues: one containing 1,578 supernovae and another with nearly 16,000. The results were striking:
- The method achieved precision on cosmological parameters comparable to traditional spectroscopic methods.
- CIGaRS estimates distances using brightness data alone with precision four times better than traditional photometric techniques.
The researchers claim this approach could unlock the full potential of photometric surveys. Currently, as little as 1% of all discovered supernovae are analyzed with spectra. By sidestepping this limitation, CIGaRS allows astronomers to work with far larger, more statistically powerful datasets.
This development marks a significant step toward leveraging the deluge of data expected from future sky surveys, potentially offering an unprecedented view of the universe's expansion history.