Triggering the formation of halo globular clusters with galaxy outflows

We investigate the interactions of high-redshift galaxy outflows with low-mass virialized clouds of primordial composition. While atomic cooling allows star formation in objects with virial temperatures above 10⁴ K, "minihalos" with virial temperatures below this threshold are generally unable to form stars by themselves. However, the large population of high-redshift starburst galaxies may have induced widespread star formation in neighboring minihalos, via shocks that caused intense cooling through both nonequilibrium H₂ formation and metal-line emission. Using a simple analytic model, we show that the resulting star clusters naturally reproduce three key features of the observed population of halo globular clusters (GCs). First, the 10⁴ K maximum virial temperature directly corresponds to the ∼10⁶ M⊙ upper limit on the stellar mass of such clusters, a feature that cannot be explained by any GC destruction mechanism. Secondly, the momentum imparted in such interactions is sufficient to strip the gas from its associated dark matter halo, explaining why GCs do not reside in the dark matter potential wells that are ubiquitous in galaxies. Finally, the mixing of ejected metals into the primordial gas provides a straightforward mechanism to explain the approximately 0.1 dex homogeneity of stellar metallicities within a given GC, while at the same time allowing for a large spread in metallicity between different clusters. To study the possibility of such "fine grained" mixing in detail, we use a simple one-dimensional numerical model of turbulence transport to simulate mixing in cloud-outflow interactions. We find that as the shock shears across the side of the cloud, Kelvin-Helmholtz instabilities arise, which cause turbulent mixing of enriched material into ≥ 20% of the cloud. Such estimates ignore the likely presence of large-scale vortices, however, which would further enhance turbulence generation. Thus, the global nature of mixing in these interactions is multidimensional, and quantitative predictions must await more detailed numerical studies.