We are the UK part of an international research group called Horizon. Our "raison d'être" is to make breakthroughs in understanding the physics driving galaxy formation and evolution with high performance computing.

More specifically, our research programme focusses on improving the modelling of stellar and super massive black hole feedback in (cosmological) hydrodynamics simulations. The main code we use to carry out this research, called RAMSES, is grid based and takes advantage of the Adaptive Mesh Refinement (AMR) technique to achieve high spatial resolution in regions of interest within the simulated volume. It is publicly available here.

The "Nut" Simulations

Nut Simulations: Formation of first generation star clusters (bottom right circular image) in a proto-galaxy (bottom left circular image) embedded within a cosmic web of filaments, at the epicentre of a far-reaching galactic wind (seen in red in the background picture). Feedback has been a bleak problem in galaxy formation for decades. Because of the extreme range of spatial scales involved, most galaxy formation simulations use largely untested semi-analytic rules for the gas physics they do not resolve. Our strategy is to increase the resolution so that we put as much of this physics on the grid as possible. The classic example of a piece of galaxy physics that has historically been modelled with subgrid recipes is winds from galaxies. Low resolution simulations set off galactic winds by hand, literally forcing gas to move at high speeds away from galaxies instead of letting the collective action of supernovae and/or Active Galactic Nuclei (AGN) power them. Taking advantage of the DiRAC national HPC facility, Horizon-UK launched a suite of high to ultra-high resolution galaxy formation re-simulations, called the "Nut". In one of these simulations, we achieved sufficient resolution to capture, for the first time, individual supernova remnants in the Taylor-Sedov phase. In Powell et al (2011), we report our breakthrough results: a high velocity, far-reaching galactic wind is naturally generated by the combined action of supernovae in the main galaxy and its satellites. By minimizing the use of subgrid recipes, simulations like the “Nut” will give more robust predictions for forthcoming observations (Alma, E-ELT, JWST) of the high redshift Universe.

The "Horizon" Simulations

Cosmological AGN Simulations: Projected gas density (background image) of a cubic chunk of simulated Universe 100 Mpc across. It includes a still simple-minded but already state-of-the-art AGN feedback model (radio emission from the AGN outflow is marked as a pink cloud in the inset) At the centre of virtually all galaxies in the Universe sit supermassive black holes. These power tremendously energetic matter outflows from what is commonly referred to as AGN. Such outflows are believed to play a crucial role in the formation of massive galaxies. Indeed, they appear to be the leading mechanism to prevent the accumulation of cold gas in their midst, which would otherwise lead to excessive star formation. The incorporation of a robust physical model for AGN feedback into numerical cosmological simulations is therefore a major goal for many research groups across the world. Thanks to the award of a dedicated parallel supercomputer from DiRAC-1 and the award of a significant amount of computing time on the DiRAC-2 complexity cluster we have already been able to make a substantial contribution to the field (e.g. Dubois et al, 2012). However this effort is best described as the first step of an ambitious, long term programme geared towards identifying physically motivated and numerically reliable models of the feedback mechanisms relevant to galaxy formation. Our pilot study (Kimm et al, 2012) is a clear proof of concept that, coupled to a thorough comparison to the latest multi-wavelength observational data sets, the proposed research has the potential to yield a step-change in our understanding of the role played by AGNs in shaping galaxy properties throughout cosmic time (e.g. stellar masses, morphologies, colours).