What do two guys from Ohio, the GAIA mission, a worldwide network of ground-based telescopes, machine learning, and citizen scientists all have to do with each other? Thanks to this interesting combo of people and computers, astronomers now have more than 116,000 new variable stars to study. Until now, they knew of about 46,000 of these stars in the Milky Way Galaxy. They had observed maybe 10,000 or so in other galaxies. The discovery gives astronomers even more chances to study variables and understand why they behave the way they do.
A time-lapse of the star Polaris, which is a Cepheid variable star, showing the cycle of its brightness changes. Courtesy TimWether CC BY-SA 4.0
Ohio State University astronomer Collin Christy and Ph.D. student Tharindu Jayasinghe recently published a paper discussing their discovery of this new trove of variable stars. Christy described the importance of these objects to astronomers. “Variable stars are sort of like a stellar laboratory,” he said in a press release statement. “They’re really neat places in the universe where we can study and learn more about how stars actually work and the little intricacies that they all have.”
Combing through Data to Find Variable Stars
It turns out variables can be somewhat elusive. That’s partly because lots of things flicker in the Universe. Supernovae flare up quickly and fade. Novae do the same thing. These don’t happen on a predictable basis, though. Variables, however, brighten and dim quite regularly. Some are bright and quite obvious, like the star Polaris (our North Star) or the variable Algol in the constellation Perseus. Others, like the Sun, change in brightness so slightly that their activity takes special techniques to measure. So, what’s a good way to sort variables out from the other things that go “blink” in the night?
It helps to start out with data about a lot of stars. Christy and collaborator Jayasinghe accessed a catalog of stellar information from the space-based GAIA mission. They also used data from the 2 Micron All Sky Survey (2MASS) and ALLWISE (a wide-field infrared data repository from the WISE mission). That gave them a huge database of stars to sift through looking for targets. It also presented a big challenge. “If you want to look at millions of stars, it’s impossible for a few humans to do it by themselves. It’ll take forever,” said Jayasinghe. “So we had to bring something creative into the mix, like machine learning techniques.”
Handling that much survey data is a process tailor-made for machine learning and artificial intelligence. Computers can run through data fairly quickly, but they need good data. The human touch was still necessary because some of the data were bad, which confused the machine learning algorithms.
Separating JUNK from Variables
Citizen scientists stepped in to help identify information about objects that weren’t variable stars. This data became known as the “JUNK data.” Christy noted this phase of the project was absolutely essential. “Having people tell us what our bad data looks like is super useful because initially, the algorithm would look at the bad data and try to make sense of it,” Christy said.
Eventually, between sorting out the JUNK data and running verified information through machine learning, the astronomers had about 400,000 variables to observe. The team turned to the All-Sky Automated Survey for Supernovae (ASAS-SN) telescope network to observe the variable star candidates. The survey’s ground-based telescopes were fitted with blue-sensitive g-band filters to look for the variables. More than half were already known to astronomers, but an amazing 116,027 of them turned out to be new discoveries.
The ASAS-SN deployed telescopes like this in the search for new variable stars. Courtesy ASAS-SN Survey.
The JUNK datasets now modify and improve the
Did you miss our previous article…