Tracing the effects of stellar feedback on molecular gas from simulations to observations

Abstract

The interstellar medium (ISM) is a turbulent, multi-phase medium with a hierarchical structure consisting of molecular clouds (MCs), clumps, and cores. Stars form in dense cores and, over their evolution, inject mass, momentum, and energy back into the ISM through stellar feedback processes. This thesis investigates the effects of various stellar feedback mechanisms on molecular gas structures across different spatial scales and evolutionary stages. Our analysis is based on the STARFORGE simulations, which model the evolution of a giant molecular cloud (GMC) at high resolution while self-consistently incorporating all relevant star-formation processes. The simulations explicitly track individual stellar feedback processes within a GMC across multiple timesteps (snapshots), enabling the isolation of individual feedback signatures. Using hierarchical clustering, we identify molecular gas structures within a 20000 M_⊙ GMC and analyse their properties across multiple projects. <br/> We first study the effects of protostellar outflows, stellar winds, and supernovae on molecular gas cores identified in the H₂ density maps of a GMC. Stellar feedback increases the velocity dispersion and virial parameter of the cores by injecting momentum into them. In addition, cores affected by different feedback mechanisms show structural differences, e.g. cores affected by outflows and winds appear smaller and cores affected by supernovae appear larger compared to pristine cores. These results highlight the role of individual feedback mechanisms in shaping the cores. <br/> To enable a direct comparison between the MCs in simulations and observations, we use the RADMC-3D radiative transfer algorithm to post-process the STARFORGE simulation in the likeliness of ¹³CO(2-1) observations from the SEDIGISM survey. From these synthetic datacubes, we identify MCs using hierarchical clustering and analyse their observable properties in multiple snapshots. The distributions of the synthetic MCs occupy different average positions in the Larson's and Heyer's relations plots. This suggests an evolutionary sequence in which MCs form as small diffuse cloudlets, grow into massive, filamentary, star-forming complexes, and are subsequently transformed into bubbles by stellar feedback before being dispersed. <br/> To study the evolution of clumps embedded in synthetic MCs, we have created an algorithm to follow their evolution by iteratively matching clumps with their corresponding counterparts in subsequent snapshots. Analysing the clump properties as a function of their lifetime, we find that long-lived clumps are typically larger and more massive. A fraction of these survive long enough to form protostars and remain relatively unaffected by strong radiative feedback. In contrast, short-lived clumps are small, less massive structures that are rapidly dispersed by turbulence and radiation. <br/> We further use these synthetic observations to train a convolutional neural network CASI-3D to identify stellar feedback signatures in an observational dataset. The neural network successfully recovers the entire shell of a large bubble, but also identifies wind-affected gas in surrounding noisy regions. This preliminary analysis shows the ability of the neural network to detect stellar feedback signatures in previously unseen observational datasets. <br/> Throughout this thesis, we find that stellar feedback operates across a wide range of spatial scales in molecular gas. <br/> On cloud scales, feedback mainly drives fragmentation, creates cavities, and governs the morphological evolution of molecular clouds. On clump scales, feedback reduces the sizes and masses of structures by eroding them and injects momentum to enhance their internal gas motions. On core scales, feedback typically increases the velocity dispersion and the virial parameter of structures. Overall, stellar feedback is a key driver of molecular cloud evolution, shaping gas structures throughout galaxies.