The Aquarius project is the first simulation that can resolve the full mass range of potential globular cluster formation sites .
With a particle mass m _ { \mathrm { p } } = 1.4 \times 10 ^ { 4 } M _ { \odot } , Aquarius yields more than 100 million particles within the virial radius of the central halo which has a mass of 1.8 \times 10 ^ { 12 } M _ { \odot } , similar to that of the Milky Way .
With this particle mass , dark matter concentrations ( haloes ) as small as 10 ^ { 6 } M _ { \odot } will contain a minimum of 100 particles .

Here , we use this simulation to test a model of metal-poor globular cluster formation based on collapse physics .
In our model , globular clusters form when the virial temperatures of haloes first exceed 10 ^ { 4 } K as this is when electronic transitions allow the gas to cool efficiently .
We calculate the ionising flux from the stars in these first clusters and stop the formation of new clusters when all the baryonic gas of the galaxy is ionised .
This is achieved by adopting reasonable values for the star formation efficiencies and escape fraction of ionising photons which result in similar numbers and masses of clusters to those found in the Milky Way .
The model is successful in that it predicts ages ( peak age \sim 13.3 Gyrs ) and a spatial distribution of metal-poor globular clusters which are consistent with the observed populations in the Milky Way .
The model also predicts that less than 5 \% of globular clusters within a radius of 100 kpc have a surviving dark matter halo , but the more distant clusters are all found in dark matter concentrations .

We then test a scenario of metal-rich cluster formation by examining mergers that trigger star formation within central gas disks .
This results in younger ( \sim 7–13.3 Gyrs ) , more centrally-located clusters ( 40 metal-rich GCs within 18 kpc from the centre of the host ) which are consistent with the Galactic metal-rich population .
We test an alternate model in which metal-rich globular clusters form in dwarf galaxies that become stripped as they merge with the main halo .
This process is inconsistent with observed metal-rich globulars in the Milky Way because it predicts spatial distributions that are far too extended .