Peering in a Stellar Nursery

Credit: Stella Offner

UT astronomer Stella Offner wants to better understand how our sun formed, but its birth was almost 5 billion years ago. The stellar nursery (a.k.a., “molecular cloud”) it was born into is long gone; its many siblings have drifted off all over the galaxy. 

The complex interactions between nascent stars and the gas and dust that surround them play a fundamental role in how stars form and how quickly. Studying stars similar to the sun at earlier stages of development gives clues – provided astronomers can accurately measure properties, such as gas temperature and density. Offner wants to measure how these young stars influence and heat their birth places. That process requires a lot of observational data and is time-intensive, plus prone to significant uncertainties.

So, could AI help? A type of artificial intelligence tool better known for generating an image of Pope Francis wearing a giant puffer jacket could play an outsized new role in assessing the physical properties of molecular clouds. Offner and her team are exploring a solution that applies a diffusion model – like those behind image-generating applications DALL-E and Midjourney.

They taught their diffusion model how heat in a molecular cloud correlates to light emission by training it on physical simulations of clouds similar to one Offner studied earlier. These simulations run on supercomputers at the Texas Advanced Computing Center.

Finally, they asked the trained diffusion model to interpret light from telescope observations of the previously observed cloud, which has hundreds of forming stars. The predicted values for the heat agree well with estimates based on the young star locations. Now Offner and her team plan to apply this approach to other nearby molecular clouds to study the impact of heating on cloud conditions in a broader context.

“One of the most exciting things is that machine learning allows us to do things that we could previously only do by eye,” Offner said, “and do it much more quickly, efficiently and rigorously.”