Early Evolution and 3D Structure of Embedded Star Clusters

TL;DR

This study investigates the very early evolution of embedded star clusters by simulating a $10^4$ M$_\odot$ cloud at 0.0683 pc resolution with Torch-AMUSE-Flash to couple gas dynamics, star formation, feedback, and direct N-body dynamics. By varying the virial parameter $\alpha$ and primordial binary prescriptions across 12 runs, the authors show that cluster morphologies, growth, and mass assembly are mainly governed by the ambient gas dynamics and recent accretion rather than internal two-body relaxation, with clusters losing up to ~50% of their mass during assembly and radii/ellipticities fluctuating on timescales as short as $\sim 0.01$ Myr. The work demonstrates that early cluster structure is a reflection of recent gas-driven history, not a predictor of future evolution, and highlights that projection effects and accretion signatures (high ellipticity and large radii) can inform observational inferences about cluster formation. These results have implications for interpreting embedded clusters in the Milky Way and for improving sub-grid treatments in larger-scale simulations of star formation and cluster assembly.

Abstract

We perform simulations of star cluster formation to investigate the morphological evolution of embedded star clusters in the earliest stages of their evolution. We conduct our simulations with Torch, which uses the AMUSE framework to couple state-of-the-art stellar dynamics to star formation, radiation, stellar winds, and hydrodynamics in FLASH. We simulate a suite of $10^4$ M$_{\odot}$ clouds at 0.0683 pc resolution for $\sim$ 2 Myr after the onset of star formation, with virial parameters $α$ = 0.8, 2.0, 4.0 and different random samplings of the stellar initial mass function and prescriptions for primordial binaries. Our simulations result in a population of embedded clusters with realistic morphologies (sizes, densities, and ellipticities) that reproduce the known trend of clouds with higher initial $α$ having lower star formation efficiencies. Our key results are as follows: (1) Cluster mass growth is not monotonic, and clusters can lose up to half of their mass while they are embedded. (2) Cluster morphology is not correlated with cluster mass and changes over $\sim$ 0.01 Myr timescales. (3) The morphology of an embedded cluster is not indicative of its long-term evolution but only of its recent history: radius and ellipticity increase sharply when a cluster accretes stars. (4) The dynamical evolution of very young embedded clusters with masses $\lesssim$ 1000 M$_{\odot}$ is dominated by the overall gravitational potential of the star-forming region rather than by internal dynamical processes such as two- or few-body relaxation.

Figures (11)

• Figure 1: Left: Example of 3D spatial clustering of the stars in S-R0 at the last snapshot, 2 Myr after the onset of star formation. Member stars for each cluster with at least 100 bound members are shown in a given colour -- blue, green, yellow, orange, red, or pink -- while unclustered stars are shown in grey. Right: Example of ellipsoids enclosing 90% of the cluster mass for the bound clusters identified in the last snapshot of S-R0, 2 Myr after the onset of star formation. The colours of the ellipsoids match those of the members stars identified on the left.
• Figure 2: Gas surface density along the $z$ axis for simulations initialized with the different virial parameters $\alpha$ (from top to bottom, $\alpha$=0.8, 2.0, 4.0), 3.0 Myr after the start of the simulation. Star formation begins at a time $t_{SF}$ (labelled for each frame) after the start of the simulation. All three simulations form single stars only and use the same random seed for star formation. Stars are shown in white, with a marker size proportional to the star's mass.
• Figure 3: SFR (top) and integrated SFE (bottom) plotted against the time since the onset of star formation for the different simulations, smoothed over 0.1 Myr using a Gaussian filter. Simulations with primordial binaries are shown in red and simulations with single stars only are shown in black. Transparent red and grey are used for the runs that do not use the default random seed (respectively B-P1, B-P2, and B-P3, and S-R1, S-R2, and S-R3). Solid lines are used for simulations with $\alpha=0.8$, dashed-dotted lines for simulations with $\alpha=2.0$, and dotted lines for simulations with $\alpha=4.0$. Simulations with different $\alpha$'s display different general trends but simulations with the same $\alpha$ and different stellar populations do not.
• Figure 4: Distribution of characteristic radius $r_{50}$ against cluster mass, for all clusters identified in each snapshot of our simulations. Brightness decreases linearly with increasing density in parameter space. The dotted lines denote constant densities and the dashed-dotted lines denote constant surface densities. A few high density clusters with masses $\sim$ 100 M$_{\odot}$ and radii < 0.01 pc (discussed in-text) lie beyond the limits of the plot. Six deeply embedded clusters with at least 100 members from the MYStIX survey Kuhn2014 are shown in red for comparison.
• Figure 5: Ellipticity (top) and ellipticity of the projected ellipses along a random direction (bottom), against effective cluster radius $r_{50}$. Density in parameter space increases linearly with decreasing brightness. Six deeply embedded clusters with at least 100 members from the MYStIX survey Kuhn2014 are shown in red for comparison.