We recently reported on how the mountains of data produced by astronomical instruments are “perfect for AI.” We’ve also started reporting on several use cases for different AI algorithms. Now, a team of researchers from the University of Texas has developed a new use case that focuses on discovering the interior makeup of exoplanets by looking at a specific type of star.
That particular kind of star is known as a “polluted” white dwarf. White dwarves are the end stage of stars that are too small to go supernova. After going through a red giant phase, our sun will turn into one in a few billion years. Typically, they only have hydrogen and helium in their upper atmosphere, making them mundane by the standards of stars – unless they happen to be tearing apart one of their planets.
Every once in a while, a white dwarf draws in one of the planets in its solar system, ripping the planet apart in the process. The planet’s interior materials are then absorbed into the star’s outer shell, making them “polluted” with the heavy metals that typically comprise a planet’s interior.
Credit – Kavli Institute for Theoretical Physics
Analyzing those heavy metals in a star’s atmosphere would allow astronomers to understand the makeup of the former exoplanet. As such, finding polluted white dwarves to analyze has been a focal point of exoplanet hunters for some time. However, saying the process is time-intensive is an understatement. Astronomers have to manually check astronomical surveys to find evidence of heavy metals in white dwarves’ atmospheres, and some of those surveys, needless to say, are big.
However, searching for needles in a haystack sounds like the perfect use case for AI. So, researchers at the University of Texas did just that. They developed an algorithm using an AI technique called manifold learning and let the algorithm loose on data from Gaia, ESA’s astrometry mission. They filtered data from around 100,000 white dwarves, which resulted in 375 potentially polluted candidates.
Follow-up observations on those 375 candidates by the Hobby-Eberly Telescope and the McDonald Observatory, both of which are at least partially controlled by UT, showed that the algorithm was 99% correct in detecting the existence of heavy metals in a star’s atmosphere, thereby classifying it as “polluted.” Given the sheer volume of white dwarves in our galaxy, tens of thousands more candidates can likely be found by allowing the algorithm to trawl through other data collected on them.
What that means for astronomers is the ability to understand the interior makeup of exoplanets as their host star is ripping them apart. Understanding their interior makeup would allow astronomers to develop models about their chances for harboring life. So, this paper is a step towards developing that astrobiological model and an excellent use case for AI in astronomy. It just so happens to be built on the back of dying planets that might be taking their own form of nascent biospheres with them.
Learn More:
UT Austin – Astronomers Use AI To Find Elusive Stars ‘Gobbling Up’ Planets
Kao et al. – Hunting for Polluted White Dwarfs and Other Treasures with Gaia XP Spectra and Unsupervised Machine Learning
UT – What Can AI Learn About the Universe?
UT – Astronomy Generates Mountains of Data. That’s Perfect for AI
Lead Image:
Artist’s depiction of a star ripping apart a planet.
Credit – NASA, ESSA, Joseph Olmsted (STScI)