The Recipe For Star Cluster Soup

Title: The dynamics and outcome of star formation with jets, radiation, winds, and supernovae in concert

Authors: Michael Y. Grudić, Dávid Guszejnov, Stella S. R. Offner, Anna L. Rosen, Aman N. Raju, Claude-André Faucher-Giguère, and Philip F. Hopkins

First Author’s Institution: Carnegie Observatories, Pasadena, CA

Status: Submitted to MNRAS [open access]

Simulating the formation of a star cluster is similar to following a very complicated recipe for soup. Sure, it starts with the basic ingredients – a cloud of gas and some gravity to get it to collapse, but you can’t just leave it at that. You need to mix it up a bit, stir in some turbulence with a healthy spoonful of magnetic fields. Then you’ll need some freshly harvested gas chemistry models to help with the heating and cooling, and don’t forget to gradually add stellar winds, outflows, and radiation along the way. Maybe even add in a supernova here or there. And so it continues, adding more and more hard-to-model ingredients pursuing a final product that fits the bill. Until recently, many of these ingredients were anything but household items and would be extremely challenging if not impossible to include. However, modern simulations have incredibly well-stocked pantries. State-of-the-art models are getting to the point where, while they are not perfect replicas of cluster formation, they make a good soup.

Today’s paper presents the newest generation of the STARFORGE simulations, dubbed “The Anvil of Creation”. This simulation replicates the collapse of a giant molecular cloud into a cluster of stars (the progression from molecular cloud to star cluster is shown in the movie below) while including an extremely thorough treatment of “stellar feedback” (which describes the ways in which young stars eject mass and energy back into the surrounding interstellar medium). This results in one of the most realistic star cluster formation models ever computed. But what exactly do we mean by “realistic”? The complex interweaving of different physics (magnetic fields, gravity, hydrodynamic turbulence, stellar feedback, etc.) is impossible to reproduce perfectly, so star formation simulations need to find the right corners to cut. The goal is to simplify the system while reproducing the measurable features of real star-forming regions, like the star formation efficiency (the fraction of the gas that is turned into stars) and the distribution of stellar masses produced (the much-discussed Initial Mass Function). Often, simulations succeed at reproducing some of these features but remain inconsistent with others. Today’s paper reports that the star cluster in the “Anvil of Creation” has no major, glaring inconsistencies with the observed trends of star formation- a huge victory for this type of simulation.

An animation provided by the authors showing the evolution of the simulated giant molecular cloud, “The Anvil of Creation”, as it collapses, forms stars, and is destroyed by stellar feedback from those young stars. More information, see the project’s website:

A Sinking Feeling

Stars are small. Maybe not SMALL small, but compared to the clouds of gas from which they are born, it’s often best to treat them as infinitely small point particles in the simulation. These are often called sink particles. The way these sink particles form (as protostars), affect their surroundings in the simulation, evolve as they age, and potentially die as supernovae is handled by carefully-designed “feedback” models that summarize the net effect of all the physics happening on scales smaller than the finest resolution of the simulation. The authors of today’s paper implement a very thorough set of these models governing the behavior of these sink particles. They include a wide range of these feedback mechanisms, from the bipolar outflows (columns of high-velocity gas launched by protostars as they accumulate mass) to the ionizing radiation and powerful winds generated by massive stars. The authors find that the bipolar outflows play the biggest role in putting momentum back into the cloud (at least early on in the simulation). These outflows seem to be critically important for determining the final distribution of stellar masses, which ends up being very close to the observed distribution of real star clusters (see Figure 1). Radiation, winds, and supernovae from the most massive stars are responsible for displacing and destroying the remaining gas towards the end of the simulation. It is the combined effect of all these mechanisms during the cloud’s complicated evolution that reproduces a “realistic” star cluster formation event. 

Figure 1. A comparison of the distribution of stars with different masses (the initial mass function, or IMF) from the “Anvil of Creation” simulation (the black line) compared with a standard distribution found from observations. Figure 8 from the paper.

The formation of the star cluster takes about 8 million years in the simulation. Over this time, the authors observe an accelerating collapse, the onset of star formation, and then evacuation of gas by stellar feedback (see Figure 1). One somewhat surprising result is how slowly the most massive stars form. Rather than quickly collapsing from the densest pockets of gas in the cloud, massive stars seem to slowly accumulate their mass over millions of years. Since we can’t sit around and watch the formation of real stars over millions of years, this result might inform how we interpret the effectively static freeze-frames we observe of real nearby star forming regions. The assumptions we make about the distributions of ages of stars have far-reaching impacts on how we interpret the light from unresolved young stellar populations, so a thorough theoretical understanding of how and when different types of stars emerge is crucial for studying far-away galaxies. 

So, where to go from here? There are a wealth of nuances to explore using detailed and self-consistent simulations like “The Anvil of Creation”, including the formation of protoplanetary disks and the impact of different molecular cloud environments and initial conditions. The future of star formation simulations looks bright, emerging through the rapidly receding veil of theoretical and computational challenges. 

Astrobite edited by Laila Linke

Featured image credit: A frame from the simulation fly-around movie released with today’s paper, available at 

About H Perry Hatchfield

I'm a PhD candidate in Physics at the University of Connecticut, where I study star formation and gas structure in the Milky Way's Galactic Center. I do this using radio observations of molecular clouds as well as hydrodynamic simulations, and I'm all about trying to find ways to compare these two exciting means of exploring the universe.

Leave a Reply