The effects of subgrid models on the properties of giant molecular clouds in galaxy formation simulations

Abstract: Recent cosmological hydrodynamical simulations are able to reproduce numerous statistical properties of galaxies that are consistent with observational data. Yet, the adopted subgrid models strongly affect the simulation outcomes, limiting the predictive power of these simulations. In this work, we perform a suite of isolated galactic disk simulations under the {\it SMUGGLE} framework and investigate how different subgrid models affect the properties of giant molecular clouds (GMCs). We employ {\sc astrodendro}, a hierarchical clump-finding algorithm, to identify GMCs in the simulations. We find that different choices of subgrid star formation efficiency, $\epsilon_{\rm ff}$, and stellar feedback channels, yield dramatically different mass and spatial distributions for the GMC populations. Without feedback, the mass function of GMCs has a shallower power-law slope and extends to higher mass ranges compared to runs with feedback. Moreover, higher $\epsilon_{\rm ff}$ results in faster molecular gas consumption and steeper mass function slopes. Feedback also suppresses power in the two-point correlation function (TPCF) of the spatial distribution of GMCs. Specifically, radiative feedback strongly reduces the TPCF on scales below 0.2~kpc, while supernova feedback reduces power on scales above 0.2~kpc. Finally, runs with higher $\epsilon_{\rm ff}$ exhibit a higher TPCF than runs with lower $\epsilon_{\rm ff}$, because the dense gas is depleted more efficiently thereby facilitating the formation of well-structured supernova bubbles. We argue that comparing simulated and observed GMC populations can help better constrain subgrid models in the next-generation of galaxy formation simulations.