UVic scientist reveals star formation stops in recently merged galaxies

Alf Wilson, Science

UVic Physics and Astronomy professor Sara Ellison was recently featured in New Scientist for her galaxy merger research. Ellison has been testing the predictions of galaxy-merger simulations by scrutinizing astronomical observations. Her group revealed that, contrary to modern simulations, mergers between two galaxies are up to 60 times more likely to abruptly stop forming stars.

When galaxies collide, vast amounts of gas (potential star-stuff) piles up in their centres making for a "stellar nursery." Observations have confirmed the predictions that this would trigger new star formation. Modern simulations also predicted that most of these galaxies would continue to form stars after the merger. But Ellison’s team found that this is not the norm.

We've been checking through the long list of predictions from the simulations, this was the last one – and it gave the biggest surprise. It disagrees with state-of-the-art simulations which found that rapid shut down of star formation after a merger would be rare.

–Sara Ellison, Physics and Astronomy

Also surprising, the observational evidence needed was already available in public archives. Ellison just needed to ask the right question – and then find hundreds of needles in a universe-sized haystack.

As galaxies merge, stars (and other matter) from each pull on the other, rearranging themselves and tugging on the vast regions gas between them. Robbed of angular momentum, the gas is effectively knocked out of orbit. Spiralling toward the merged galactic centre, the gas piles up eventually reaching star-forming densities. 

"More gas together would mean more star formation," explained Ellison. "But the difficulty in early simulations was how to treat the way that gas approached the black hole at the centre, as well as what would happen when that gas got swallowed."

Some of the gas spirals all the way into the black holes. But as a black hole accretes matter it gives off energy; a process called feedback. Early simulations used a very aggressive feedback model that would stop (quench) star formation. Think putting out a campfire with a waterfall.

"The details of how black holes respond to this influx of gas really was unknown," Ellison recalled. "Will the feedback be gentle or extreme? If the resulting feedback is very energetic, then star formation will be quenched. If gentle, then star formation could continue."

Over the past 5 years simulations have improved and produced outputs that are in good agreement with most observations. But to achieve this, they switched to a much more gentle mode of feedback, which rarely leads to rapid star formation quenching.

"Our findings are a call back to the drawing board," Ellison excitedly spoke to the repercussions. "There's now a tension between the observations and simulations. This is good thing – it will ultimately lead to more robust theory.

Ellison was quick to point out that the data were all sitting in public archives. And what she thinks this means for scientists today.

We live in an age with so much data. A lot of science is just trying to analyze this data. As Scientists we can lose sight of what to do with it all. We need to take a step back and ask the right questions.

—Sara Ellison, Physics and Astronomy

Clearly, Ellison asked the right question!

Finding galactic needles in a universe-sized haystack

Ellison's group needed to find a large set of recently merged galaxies and compare their history of star-formation to non-merged galaxies. But mergers are quite rare and time-consuming to find in the vast observational data archives. Bobby Bickley, a PhD student in Ellison's group, used a machine learning algorithm to detect recently merged galaxies. Here, recent means within 1 billion years!

"Single galaxies usually have well-defined familiar shapes: spirals and ellipses," Ellison described. "But recently merged galaxies look like a train wreck – there's debris everywhere."

Bickley trained a neural network to find these galactic train wrecks – identifying a mere 700 out of 170,000 searched images.

Finding out if star formation recently stopped

The next step was to check if star formation had been quenched. Scott Wilkinson, an MSc student in Ellison’s group, did this by assembling the spectra of those merged galaxies. Light from any source can be spread out into various wavelengths – the spectrum – like we see in a rainbow. The relative intensity of the wavelengths (colours) in the spectrum provides information about the galaxy’s past and present.

During star formation, light from birthing stars ionizes nearby gas resulting in tell-tale ionization lines (specific colours). If those ionization lines are not in the spectrum, then star formation is not currently happening.

To determine if there was recent star formation, Ellison’s group looked at the current mix of stars in each galaxy. Hot, super-massive stars burn out very quickly, so they’re not around for long. Lower mass stars live much longer. The evidence for each type of star can also be seen by looking for different lines in the spectra. If the spectrum shows only lower mass stars, then star formation has not happened for a long time. But if the spectrum also includes hot, super-massive stars, then star formation was happening recently.

Ellison explained the final step. "I wrote code that counted the number of mergers with this one-two punch: recent star formation but nothing current. Then compared this to a set of non-merging galaxies."

Ellison's group found that star formation shutdown was 30-60 times more common in the merged than non-merged galaxies. This is an exceptionally clear result.

It's been wonderful to have the opportunity to tackle such a long-standing question. And to have the signal be so clear-cut. The result is so obvious.

–Sara Ellison, Physics and Astronomy

Check out the original research papers:

Mergers can rapidly shut-down star formation:
Galaxy mergers can rapidly shut down star formation, Ellison et al. (2022)

Using machine learning to find galaxy mergers:
Convolutional neural network identification of galaxy post-mergers in UNIONS, Bickley et al. (2021)

Finding the signatures of rapidly quenched star formation:
The merger fraction of post-starburst galaxies in UNIONS, Wilkinson et al. (2022)