The interstellar medium (ISM) is a complex mixture of cold dense molecular clouds in which stars form, embedded into a hot tenuous plasma. Whilst the hot component can reach at the order of millions of degrees Kelvin, the gas within molecular clouds can cool down to some of the lowest temperatures in the Universe, reaching extremely high densities. As the range covered by physical processes in galaxy formation is so immense, current supercomputers lack the power to resolve the multi-phase nature of the ISM whilst self-consistently modelling entire galaxies.

As a result the ISM is often modelled as a single phase medium, with an average density, pressure and energy. As this single phase will have a density greater than that of the real hot phase, the resistance to outflowing supernova driven winds increases. This leads to inefficient feedback which cannot effectively regulate star formation. As a result, galaxies often produce too many stars, too early in their lifetime, leading to overly red stellar populations at present day. In addition, as metals are directly injected into the hot phase of the ISM during supernova explosions, reducing the efficiency of galactic winds inevitably leads to poor metal mixing throughout the galaxy with overly concentrated central regions.

Here in Vienna, we are implementing a new multi-phase model into the massively parallel adaptive mesh refinement code, FLASH. By seperating the hot gas from the cold

molecular clouds, the multi-phase model allows each phase to have the correct density, temperature and pressure. This ensures that when supernovae detonate, their winds can stream past the molecular clouds which are too compact to offer significant resistance. This greatly enhances the realism with which stellar feedback is modelled and leads to more dynamic star formation with molecular star forming material continuing to accrete independent of the motion of the hot gas.

Our model incorporates the new stellar hydrodynamics module developed by Nigel Mitchell, which allows the collisionless dark matter, stellar material and molecular clouds to be modelled as a collisionless fluid on the mesh. This has significant advantages over conventional particle-mesh algorithms and scales inline with the hydrodynamics routines on Peta-flop machines. The hot phase of the ISM is then modelled as an ideal gas using the conventional hydrodynamics routines.

As the new multi-phase model explicitly separates the two phases, we can begin to explore the multitude of physical processes within the ISM, unlike single phase models which heavily approximate all of these processes. Processes of particular interest include metal dependent cooling, heat conduction and inter-cloud collisions, all of which can have a profound affect upon star formation and galaxy evolution.