Tidal adaptive softening and artificial fragmentation in cosmological simulations

Traditional N-body methods introduce localized perturbations in the gravitational forces governing their evolution. These perturbations lead to an artificial fragmentation in the filamentary network of the large-scale structure, often referred to as 'beads-on-a-string'. This issue is particularly apparent in cosmologies with a suppression of the matter power spectrum at small spatial scales, such as warm dark matter models, where the perturbations induced by the N-body discretization dominate the cosmological power at the suppressed scales. Initial conditions based on third-order Lagrangian perturbation theory, which allow for a late-starting redshift, have been shown to minimize numerical errors contributing to such artefacts. In this work, we investigate whether the additional use of a spatially adaptive softening for dark matter particles, based on the gravitational tidal field, can reduce the severity of artificial fragmentation. Tidal adaptive softening significantly improves force accuracy in idealized filamentary collapse simulations over a fixed softening approach. However, it does not substantially reduce spurious haloes in cosmological simulations when paired with such optimized initial conditions. Nevertheless, tidal adaptive softening induces a shift in halo formation times in warm dark matter simulations compared to a fixed softening counterpart, an effect not seen in cold dark matter simulations. Furthermore, initializing the initial conditions at an earlier redshift generally results in haloes forming from Lagrangian volumes with lower average sphericity. This sphericity difference could impact post-processing algorithms identifying spurious objects based on Lagrangian volume morphology. We propose potential strategies for reducing spurious haloes without abandoning current N-body methods.