FEAST: Probing Hierarchical Star Formation with the Spatial Distributions of Young Star Clusters

TL;DR

FEAST applies the angular two-point correlation function to the spatial distributions of young star clusters in four nearby galaxies, using JWST-emerging infrared (eYSC) and archival optical catalogs to trace hierarchical structure from embedded to older populations. The authors fit three TPCF models via MCMC, deriving the two-dimensional fractal index $D_2$ and the maximum scale of hierarchy $l_{\rm corr}$, and find strong hierarchical clustering for the youngest clusters with $D_2$ values near $1.3$, while older populations become increasingly random on timescales of $\sim 10$–$100$ Myr. Across spirals NGC 628, M51, and M83 the oldest cohorts approach Poisson distributions, whereas NGC 4449 preserves hierarchical ordering longer, likely due to weaker shear. These results support a largely universal picture of hierarchical star formation driven by ISM turbulence, while galaxy-specific dynamics set the scale and persistence of clustering; JWST FEAST data prove especially effective at probing the embedded phase of YSCs.

Abstract

We apply the angular two-point correlation function (TPCF) to the spatial distribution of young star clusters (YSCs) in four nearby star forming galaxies (NGC 628, NGC 4449, M51, and M83) in order to investigate their underlying hierarchical structuring. Using newly constructed catalogs of YSCs in the emerging phase (eYSCs), identified in the infrared with JWST, and optical YSCs detected in archival HST data, we compute TPCFs for various cluster samples and age bins across the four galaxies as part of the FEAST (Feedback in Emerging extrAgalactic Star ClusTers) program. We find clear evidence of hierarchical structuring, especially in eYSCs and YSCs with ages < 10 Myr (referred to as oYSCs), which show similar TPCFs within each galaxy. NGC 628 exhibits a clear distinction between the TPCFs of eYSCs and oYSCs, implying a shorter randomization timescale. In contrast, clusters aged 10 to 300 Myr exhibit progressively more random spatial distributions, becoming effectively random after $\sim$ 100 Myr, consistent with earlier studies. The two-dimensional fractal index $D_2$ of the YSCs underlying distribution is calculated from model fits to TPCFs. Our values of $D_2$ derived from the youngest YSC populations align better with the expected value of $D_2 \sim $1.3 for a universal star formation process compared to previous findings.

Figures (12)

• Figure 1: The positions of eYSCs (pink) and YSCs (blue) in each of this sample galaxies, with scale bars indicating length scales corresponding to $2^{'}$. Visually, eYSCs tend to trace spiral arm structure compared to YSCs, which are more evenly distributed throughout the galaxy. Background images of galaxies are taken from the Digitized Sky Survey dssref.
• Figure 2: The relative frequencies of eYSC Is (orange), eYSC IIs (pink), and YSCs (blue) for each of our galaxies, designated on the $x$-axis. The total number of clusters is shown under the galaxy name. eYSCs tend to comprise $\sim 30-50$ % of the post-cut catalog
• Figure 3: Mass plotted against age for the YSCs in each of the galaxies in this sample. Blue stars, orange circles, and pink diamonds represent YSCs, eYSCs I, and eYSCs II, respectively.
• Figure 4: A brief overview of each model used in our analysis, along with quantities derived from each. The single, piecewise, and exponential truncation power models are shown in pink, blue, and orange, respectively. The fractal index $D_2$ is calculated from the inner slope of each of the models, while each model provides different constraints on the maximum scale of hierarchy $l_{\rm corr}$. Figure adapted from menon21.
• Figure 5: The TPCFs of five cluster samples in each of our four galaxies: eYSCs I, II, I+II (pink circles, blue squares, and orange stars, respectively), oYSCs with age $\leq$ 10 Myr (purple diamonds), and YSCs with age $>$ 10 Myr (magenta crosses). Points represent the TPCF calculated in that angular separation bin, and error bars are calculated via bootstrapping. The bottom axes provide angular separation scales, while those same scales are converted in physical units of pc for ease of interpretation. Shaded regions correspond to separation bins $> \theta_{\rm max}/5$ where edge effects may influence results (see discussion in menon21).