Halo evolution in the presence of a disc bar

Angular momentum transfer from a rotating stellar bar has been proposed by Weinberg & Katz (2002) as a mechanism to destroy dark-matter cusps in a few rotation periods. The N-body simulations performed by these authors in support of their claim employed spherical harmonics for the force computation and were, as shown by Sellwood (2003), very sensitive to inclusion of asymmetric terms (odd l,m). In order to disentangle possible numerical artifacts due to the usage of spherical harmonics from genuine stellar dynamical effects, we performed similar experiments using a tree code and find that significant cusp destruction requires substantially more angular momentum than is realistically available. However, we find that the simplified model (a N-body halo torqued by a rotating bar pinned to the origin) undergoes an instability in which the cusp moves away from the origin. In presence of this off-centring, spherical density profiles centred on the origin display an apparent cusp-removal. We strongly suspect that it is this effect which Weinberg & Katz observed. When suppressing the artificial instability, cusp removal is very slow and requires much more angular momentum to be transferred to the halo than a realistic stellar bar possibly possesses.