Although smoothed-particle hydrodynamics (SPH) codes, such as Gadget-3, are widely used in the theoretical galaxy formation community, the traditional formulation of SPH has some inherent flaws that may cause inaccurate results; for example, fluid instabilities can be artificially suppressed (e.g., Agertz et al. 2007). However, Eulerian grid-based codes, the most commonly used alternative, also have drawbacks; for example, fixed-grid codes are intrinsically not Galilean invariant (Springel 2010). Springel (2010) presented a novel moving-mesh hydrodynamics code, Arepo. Because the mesh is advected with the fluid, Arepo has the advantages of SPH (e.g., Galilean invariance), and because the Euler equations are solved on a grid, it avoids the suppression of fluid instabilities and artificial lack of mixing that occurs in SPH.
Because SPH has been used for decades by many authors to study galaxy mergers, it is crucial to determine whether the numerical accuracies inherent in traditional SPH jeopardize any of the previous results. In Hayward et al. (2014a), we compared idealized simulations of isolated disk galaxies and galaxy mergers performed with the state-of-the-art Arepo moving-mesh hydrodynamics code with otherwise-identical simulations performed with the Gadget-3 SPH code. This work was the first detailed study of galaxy merger simulations performed with a moving-mesh hydrodynamics code. The comparison demonstrates that, unlike in the case of cosmological simulations (e.g., Vogelsberger et al. 2012, Keres et al. 2012), the global properties of the simulated galaxies (e.g., the star formation histories; SFHs) are very insensitive to the numerical method when black hole (BH) accretion and active galactic nucleus (AGN) feedback are not included. When BH accretion and AGN feedback are included, the SFHs differ more significantly, but the agreement is still very good. These results are reassuring because of the significant amount of conclusions that have been drawn from SPH simulations of galaxy mergers and isolated disk galaxies. The reason for the robustness of the results of these simulations is that the gas phase structure in these idealized simulations is relatively simple, unlike in cosmological simulations: because different phases of gas are not in close proximity, the inaccurate treatment of fluid instabilities and mixing in traditional SPH does not significantly affect the results.
Despite the good agreement for integrated the SFHs and BH growth histories, there are some detailed differences between the results of the two different techniques. In particular, the gas morphologies can differ significantly, as shown in the above movie. Typically, the SPH results feature more prominent hot halos (formed from shock heating during the starburst and AGN-driven outflows) and spurious clumps, whereas in the moving-mesh simulations, the clumps are effectively disrupted and gas cools more effectively out of the hot halos.
Animations that compare the evolution of the star formation rates (SFRs), BH masses, and gas morphologies and phase structures for all simulations presented in this work are available at this URL.