Radiation-driven outflow in active galactic nuclei: the feedback effects of scattered and reprocessed photons

Abstract

We perform time-dependent, two-dimensional, hydrodynamical, numerical simulations to study the dynamics of a slowly rotating accretion flow from sub-pc to pc scales under the irradiation from the central active galactic nucleus (AGN). Compared to previous work, we improve the calculation of the radiative force due to X-rays. More importantly, in addition to radiative pressure and radiative heating/cooling directly from the central AGN, in the momentum equation we also include the force due to the scattered and reprocessed photons. We find that the accretion flow properties change significantly due to this 're-radiation' effect. The inflow rate at the inner boundary is reduced, while the outflow rate at the outer boundary is enhanced by about one order of magnitude. This effect is more significant when the density at the outer boundary is higher. The properties of outflows such as velocity, momentum and energy fluxes, and the ratio of outflow rate and the accretion rate are calculated. We find that the efficiency of transferring the radiation power into the kinetic power of outflow is typically 10 -3 , far below the value of 0.05 which is assumed in some cosmological simulations. The effect of the temperature of the gas at the outer boundary (T 0 ) is investigated. When T 0 is high, the emitted luminosity of the accretion flow oscillates. This is because in this case the gas around the Bondi radius can be more easily heated to be above the virial temperature due to its high internal energy. Another question we hope to address by this work is the so-called sub-Eddington puzzle. That is, observations show that the luminosity of almost all AGNs are sub-Eddington, while theoretically the luminosity of an accretion flow can easily be super-Eddington. We find that even when the re-radiation effect is included and outflow does become much stronger, the luminosity, while reduced, can still be super-Eddington. Other observational implications and some caveats of our calculations are discussed.