Cloud Properties and Correlations with Star Formation in Self-consistent Simulations of the Multiphase ISM

We apply gravity- and density-based methods to identify clouds in self-consistent numerical simulations of the star-forming, multiphase interstellar medium (ISM) and compare their properties and global correlation with the star formation rate (SFR) over time. The gravity-based method identifies bound objects, which have masses $M\sim {10}^{3}\mbox{--}{10}^{4}\,{M}_{\odot }$ at densities ${n}_{{\rm{H}}}\sim 100\,{\mathrm{cm}}^{-3}$ , and virial parameters α_v ∼ 0.5-5. For clouds defined by a density threshold ${n}_{{\rm{H}},\min }$ , the average virial parameter decreases, and the fraction of material that is genuinely bound increases, with increasing ${n}_{{\rm{H}},\min }$ . Surprisingly, clouds defined by density thresholds can be unbound even when α_v < 2, and high-mass clouds ( ${10}^{4}\mbox{--}{10}^{6}\,{M}_{\odot }$ ) are generally unbound. This suggests that the traditional α_v is at best an approximate measure of boundedness in the ISM. All clouds have internal turbulent motions increasing with size as $\sigma \sim 1\,\mathrm{km}\,{{\rm{s}}}^{-1}{(R/\mathrm{pc})}^{1/2}$ , similar to observed relations. Bound structures comprise a small fraction of the total simulation mass and have a star formation efficiency per freefall time ${\varepsilon }_{\mathrm{ff}}$ ∼ 0.4. For ${n}_{{\rm{H}},\min }=10\mbox{--}100\,{\mathrm{cm}}^{-3}$ , ${\varepsilon }_{\mathrm{ff}}$ ∼ 0.03-0.3, increasing with density threshold. A temporal correlation analysis between $\mathrm{SFR}(t)$ and aggregate mass $M({n}_{{\rm{H}},\min };t)$ at varying ${n}_{{\rm{H}},\min }$ shows that time delays to star formation are ${t}_{\mathrm{delay}}\sim {t}_{\mathrm{ff}}({n}_{{\rm{H}},\min })$ . The correlation between $\mathrm{SFR}(t)$ and $M({n}_{{\rm{H}},\min };t)$ systematically tightens at higher ${n}_{{\rm{H}},\min }$ . Considering moderate-density gas, selecting against high virial parameter clouds improves correlation with the SFR, consistent with previous work. Even at high ${n}_{{\rm{H}},\min }$ , the temporal dispersion in $(\mathrm{SFR}-{\varepsilon }_{\mathrm{ff}}M/{t}_{\mathrm{ff}})/\langle \mathrm{SFR}\rangle $ is ∼50%, due to the large-amplitude variations and inherent stochasticity of the system.