Cosmological simulations involving the fully covariant gravitational dynamics may prove relevant in understanding relativistic/non-linear features and, therefore, in taking better advantage of the upcoming large scale structure survey data. We propose a new 3+1 integration scheme for General Relativity in the case where the matter sector contains a minimally-coupled perfect fluid field. The original feature is that we completely eliminate the fluid components through the constraint equations, thus remaining with a set of unconstrained evolution equations for the rest of the fields. This procedure does not constrain the lapse function and shift vector, so it holds in arbitrary gauge and also works for arbitrary equation of state. An important advantage of this scheme is that it allows one to define and pass an adaptation of the robustness test to the cosmological context, at least in the case of pressureless perfect fluid matter, which is the relevant one for late-time cosmology.
IntroductionThe upcoming large scale structure observations will provide a powerful probe of relativistic/nonlinear effects in cosmology, such as those encountered in alternative descriptions of the dark sector, backreaction, matter-radiation interactions, neutrinos, cosmic defects, the inflationary phase, etc. This motivated some very recent developments in going beyond the standard simulation techniques that are the linear-Boltzmann and Newtonian N -body approaches. These include the first relativistic N -body code [1][2][3][4], where the gravitational field is treated through an appropriately truncated secondorder perturbation theory [1,5,6], Newtonian N -body simulations that include linearized radiation [7,8] or a more sophisticated estimate of the scale factor [9], and a Boltzmann method for describing non-linear effects of massive neutrinos [10]. At the same time, fully non-linear simulations of General Relativity (GR) have been performed to study aspects of both the early and late universe, using wellestablished schemes of numerical relativity (NR, [11-16] for reviews). There are simulations of the inflationary epoch in 1 + 1 [17-20] and 3 + 1 [21-23] dimensions, while for late-time cosmology there are 1 + 1 spherical collapse simulations [24, 25] and 3 + 1 simulations involving a pressureless perfect fluid field [26-31] (see also [32]). The latter are mainly motivated by the issue of cosmic backreaction which, until recently, was addressed numerically only through lattice configurations [33][34][35][36][37][38][39]. Finally, there is also a renewal of interest in 3 + 1 NR N -body simulations of the collapse dynamics [40], after the earlier work of Shibata [41,42] and the even earlier, but lower-dimensional, works of Shapiro and Teukolsky using both N -body and Boltzmann approaches [43][44][45][46][47][48] (see also [49]).Our interest here lies in 3 + 1 NR applied to cosmology. A first observation is that NR has been mostly developed to deal with strong gravity phenomena at astrophysical scales, in which case the scenarios of inte...