Shelest already had the map of the ISM’s dust. The plan was to combine that map with a map of the ultraviolet radiation that heats the ISM up.
By chance, Bialy located just such a map by striking up a conversation at a conference with the astrophysicist Troy Porter of Stanford University, who had made a map but had yet to publish it.
So Shelest, Bialy, and their colleagues set out to combine the 200 gigabytes of data in the two maps. At every point in space, they compared the density of the dust with the intensity of the radiation to understand where the ISM was hot and where it was cool. Eventually, they were able to combine the dust and radiation information on a three-dimensional grid.
Shelest then set out to see what the map could tell him. The biggest surprise came when he investigated the ISM’s overall temperature: Shelest found that 27% of the ISM in our galactic neighborhood was cold, 32% was warm, and the remaining 41% sat squarely in the forbidden zone as unstable medium. “I thought I was wrong,” Shelest said. “How do you go to your science adviser and tell him that you get so much unstable media?”
Jonathan Shelest, a student at the Technion, found the interstellar medium to be more unstable than expected.
But the map didn’t appear to be broken. Shelest saw no evidence of artifacts or random statistical noise in the 3D grid. Instead, the unstable ISM formed massive, coherent blankets that wrapped around the cold clouds completely.
All that midrange medium, neither hot nor cold, implies that the older theory is incomplete, said Adam Ginsburg, an astronomer at the University of Florida. “Something is taking material that should, by the static theory, be sitting in one of these other two states, and it’s stirring it up so that it’s being forced to change states.”
Desika Narayanan, an astrophysicist at the University of Florida, uses an organic metaphor. “Just like your blood in your body is going to have metabolic waste or oxygen in it, and it’s moving different things to different places,” he said, “the ISM is doing the same thing, but with energy, and it’s constantly in a state of flux.”
A Measured Approach
On the July day in 2026 that the group posted the paper, Shelest’s inbox filled with messages from excited astronomers and requests to collaborate. Still, researchers urge caution before we rewrite the textbooks.
The model could have overestimated the amount of unstable ISM if cold and warm sections were too close together for the model to tell them apart, McKee said. Alternatively, the method the team used to calculate the overall distribution of gas and dust could have led them astray, said Philipp Frank, an astrophysicist at the Kavli Institute for Particle Astrophysics and Cosmology at Stanford. The researchers used an “ensemble-averaged mean,” a calculation that naturally tends to smooth over sharp contrasts. If a cold cloud transitioned abruptly to a warm one, this method would artificially soften its steep drop in density into a more gradual slope, which would inflate the amount of unstable ISM.
Shelest and Bialy — whose analysis is currently undergoing peer review — acknowledge that limited resolution and map smoothing are both issues, but they argue that mathematical blurring alone cannot account for the majority of the unstable ISM. When the authors tested their modeling on an entirely different dust map with lower resolution, the unstable fraction shifted by only about 5%, they said.
The Churn of Creation
Galaxies are remarkably inefficient at taking raw material — gas and dust — and turning it into stars. For decades, simulations have struggled to explain why. The new map suggests an answer: Star formation is suppressed by the constant cycle of heating and cooling that churns the ISM.
“The reason why star formation is very inefficient is that feedback heats up and pushes on the surrounding gas,” said Anna Rosen, an astrophysicist at San Diego State University. That churning could explain why classic models, which assume that the ISM is in a tranquil equilibrium, don’t quite seem to match reality. Rosen said the models give us “physically grounded snapshots” — unchanging images of patches of hot and cold ISM — “but in between those snapshots, there’s still a lot going on.”
