Same as Fig. 3 for the simulation with 5M_⊕ embryos. We model an outer disc here, because the convergence radius turned out to be far (21 au). After numerous exchanges, temporary co-orbitals, and one merger, a relatively stable co-orbital is formed at t ≐ 4000 P_orb. Embryo 1 isthen pushed outwards by a dynamical torque created by an underdense tadpole-like region (Pierens 2015). At t ≐ 4900 P_orb the tadpole region is refilled by a material originating from the inner spiral arm, and embryo 1 is pushed inwards, until it encounters the co-orbital. This behaviour is systematically repeated, and results in a disruption of the co-orbital, and a merger.