Paper detail

The final-parsec problem in non-spherical galaxies revisited

We consider the evolution of supermassive black hole binaries at the center of spherical, axisymmetric, and triaxial galaxies, using direct N-body integrations as well as analytic estimates. We find that the rates of binary hardening exhibit a significant N-dependence in all the models, at least for N in the investigated range of 10^5<=N<=10^6. Binary hardening rates are also substantially lower than would be expected if the binary loss cone remained full, as it would be if the orbits supplying stars to the binary were being efficiently replenished. The difference in binary hardening rates between the spherical and nonspherical models is less than a factor of two even in the simulations with the largest N. By studying the orbital populations of our models, we conclude that the rate of supply of stars to the binary via draining of centrophilic orbits is indeed expected to be much lower than the full-loss-cone rate, consistent with our simulations. We argue that the binary's evolution in the simulations is driven in roughly equal amounts by collisional and collisionless effects, even at the highest N-values currently accessible. While binary hardening rates would probably reach a limiting value for large N, our results suggest that we cannot approach that rate with currently available algorithms and computing hardware. The extrapolation of results from N-body simulations to real galaxies is therefore not straightforward, casting doubt on recent claims that triaxiality or axisymmetry alone are capable of solving the final-parsec problem in gas-free galaxies.