Abstract
The effects of wind-driven star formation feedback on the spatio-temporal organization of stars and gas in galaxies is studied using two-dimensional intermediate-representational quasi-hydrodynamical simulations. The model retains only a reduced subset of the physics, including mass and momentum conservation, fully non-linear fluid advection, inelastic macroscopic interactions, threshold star formation, and momentum forcing by winds from young star clusters on the surrounding gas. Expanding shells of swept-up gas evolve through the action of fluid advection to form a ‘turbulent’ network of interacting shell fragments which have the overall appearance of a web of filaments (in two dimensions). A new star cluster is formed whenever the column density through a filament exceeds a critical threshold based on the gravitational instability criterion for an expanding shell, which then generates a new expanding shell after some time delay. A filament-finding algorithm is developed to locate the potential sites of new star formation. The major result is the dominance of multiple interactions between advectively distorted shells in controlling the gas and star morphology, gas velocity distribution and mass spectrum of high mass density peaks, and the global star formation history. The gas morphology strongly resembles the model envisioned by Norman & Silk, and observations of gas in the Large Magellanic Cloud (LMC) and local molecular clouds. The dependence of the frequency distribution of present-to-past average global star formation rate on a number of parameters is investigated. Bursts of star formation only occur when the time-averaged star formation rate per unit area is low, or the system is small. Percolation does not play a role. The broad distribution observed in late-type galaxies can be understood as a result of either small size or small metallicity, resulting in larger shell column densities required for gravitational instability. The star formation rate increases with density, but dependences on gas velocity dispersion and average shell column density suggest that the dependence is multivariate. The distribution of gas velocities exhibits exponential tails over a broad range of parameter values and the velocity distribution for gas in filaments is nearly exponential. Decay simulations with no star formation suggest that the exponential tails are caused by multiple shell interactions, not individual stellar winds. The cloud mass spectra, estimated using a simplified version of the structure tree method, tend to be power laws at the higher-mass end, with an index that is nearly independent of the star formation activity or model parameters. Kinetic energy decay in simulations without star formation yields a t−1 dependence. We discuss how the simulations can be viewed in the context of various incomplete conceptual models, including collisional cloud coalescence, wind-driven turbulence, propagating star formation, forced mass-conserving Burgers turbulence, and granular fluids.