1999MNRAS.304..254V

We have carried out a set of self-consistent N-body simulations to study the interaction between disc galaxies and merging satellites with the aim of determining the disc kinematical changes induced by such events. We explore a region of the parameter space embracing satellites with different masses and internal structure and orbits of various eccentricities. We find that the analytic estimates of Tóth & Ostriker are high; overestimating the disc heating and thickening resulting from the accretion process by a factor of about 2–3. We find that the heating and thickening of the disc differ for satellites on prograde and retrograde orbits. The former tend to heat the stellar disc while the latter primarily produce a coherent tilt. For instance, a satellite of a Milky Way type galaxy with an initial mass of 20 per cent of that of the disc and on a retrograde orbit increases the velocity ellipsoid at the solar neighbourhood by (Δ σ_R, Δ σ_π, Δ σ_z)_☉ ~ (11,9,6) km.s^–1 and produces a maximum increment of the vertical scalelength and the stability parameter Q, inside the solar radius, of 300 pc and 0.8, respectively, increases of about 43 and 53 per cent. The same satellite, but on a prograde orbit, leads to changes of (Δ σ_R, Δ σ_π, Δ σ_z)_☉ ~ (22,15,12) km.s^–1, Δ z_o,☉ ~ 550 pc and Δ Q_☉ ~ 1.2. Thus, disc galaxies may accrete quite massive satellites without destroying the disc, particularly, if the orbits are retrograde. We also find that a massive bulge may play a role in reducing these effects. We have quantified the importance of the responsiveness of the halo by replacing it by a rigid potential in several simulations. In these cases, the increase of the vertical scalelength is larger by a factor of 1.5–2, indicating that a self-consistent treatment is essential for obtaining realistic results. A frequent by-product of the accretion process is the formation of weak stellar warps and asymmetric discs. Finally, we have checked how well Chandrasekhar's dynamical friction formula reproduces the sinking rates in several of our experiments. We find that it works well provided a suitable value is chosen for the Coulomb logarithm and the satellite mass is taken to be the mass still bound to the satellite at each moment.