ASTROPHYSICS

In project Magneticum, scientists perform simulations of which the most computational intensive one covers a cosmic volume of 1 Gpc^3. This allows the researchers, for the very first time, to self consistently study galaxy clusters and groups, galaxies, and active galaxy nuclei (AGNs) within an enormously large volume of the Universe.

Large-scale cosmological simulations utilizing the modern supercomputers play a significant role in the theoretical studies of the structure formation in the Universe. They are essential tools to accurately calculate theoretical predictions of the distribution and state of the baryonic and dark matter in the Universe. Especially in the non-linear regime of the gravitational dynamics and hydrodynamics, where galaxies and clusters of galaxies form out of the large scale structure, they are of utmost importance. But even if simulations comprise several billion particles, the resulting resolution is still poor (~30kpc) – and such simulations do not even properly resolve individual galaxies.

To overcome this shortcoming, simulations must be able to calculate 10¹¹ particles and beyond, following various physical processes. If one wants to fully exploit the potential of the upcoming large sky surveys for cosmology and study of the effects of Dark Energy, it is necessary to improve the predictions of how the large-scale structures are traced by the luminous matter. This demands a detailed description of various, complex, non-gravitational, physical processes in the simulation codes, which determine the evolution of the cosmic baryons and impact their observational properties. Amongst them are the star formation and related feedback; chemical pollution by SN Ia (Supernova Type Ia), SN II (Supernova Type II) and asymptotic giant branch (AGB) winds; transport processes like the thermal conduction; the evolution of black holes and their related active galactic nucleus (AGN) feedback as well as magnetic fields. All these must be self-consistently coupled with the underlying hydrodynamics. Being able to perform such a large volume, high-resolution cosmological simulations which follow all those details will allow to produce a theoretical counterpart to interpret the data coming from the current and the forthcoming astronomical surveys or instruments, e.g. PLANCK, SPT, DES and eROSITA.

Over the last years, a lot of efforts have been put together to reach this goal. For example, the development of the Gadget code has been pushed forward to improve the performance and scalability within the KONWIHR-III project Tuning a cosmo Gadget for SuperMUC. It is now possible to run Gadget3 exploiting 16 islands (i.e. 131,072 cores) of LRZ HPC system SuperMUC. Furthermore, first pioneering simulations handling almost 10¹⁰ particles (e.g. Magneticum Pathfinder, Fig. 1) have been performed. These simulations are used among various other projects for interpreting the observations by PLANCK and SPT to study the cluster properties in the X-ray band for future high energy resolution instruments like the Athena/Astro-H and also to produce first mock observations for the forthcoming eROSITA.

In this project, the scientists are going to perform a set of simulations and the most computational intensive of them covers a cosmic volume of 1 Gpc³. This covers a large enough volume with high enough spatial/mass resolution to follow the evolution of galaxies in detail and, especially, the AGN population. This will be supplemented by one low resolution but larger volume (40 Gpc³) simulation needed to reach a simulated volume comparable to the ongoing/future survey sizes. A final element of the simulation campaign will be a set of intermediate size simulations to explore the impact induced by the current uncertainty of the most important cosmological parameters. All simulations will follow the physical processes described above. This will allow the researchers, for the very first time, to self consistently study galaxy clusters and groups, galaxies, and AGNs within an enormously large volume of the Universe. And it will produce the above mentioned theoretical counterpart to efficiently interpret the forthcoming data.