Slow neutron captures are responsible for the production of about $50\%$ of elements heavier than iron, mainly, occurring during the asymptotic giant branch phase of low-mass stars ($1$ $\lesssim M$/M$_{\odot}$ $\lesssim$ $3$), where the main neutron source is the $^{13}$C($\alpha$,n)$^{16}$O reaction. This last is activated from locally-produced $^{13}$C, formed by partial mixing of hydrogen into the He-rich layers. We present here the first attempt at describing a physical mechanism for the formation of the $^{13}$C reservoir, studying the mass circulation induced by magnetic buoyancy and without adding new free parameters to those already involved in stellar modelling. Our approach represents the application, to the stellar layers relevant for $s$-processing, of recent exact, analytical 2D and 3D models for magneto-hydrodynamic processes at the base of convective envelopes in evolved stars in order to promote downflows of envelope material for mass conservation, during the occurrence of a dredge-up phenomenon. We find that the proton penetration is characterized by small concentrations, but extended over a large fractional mass of the He-layers, thus producing $^{13}$C reservoirs of several $10^{-3}$ M$_{\odot}$. The ensuing $^{13}$C-enriched zone has an almost flat profile, while only a limited production of $^{14}$N occurs. In order to verify the effects of our new findings we show how the abundances of the main $s$-component nuclei can be accounted for in solar proportions and how our large $^{13}$C-reservoir allows us to solve a few so far unexplained features in the abundance distribution of post-AGB objects.

The advection of thermonuclear ashes by magnetized domains emerging from near the H-shell was suggested to explain AGB star abundances. Here we verify this idea quantitatively through exact MHD models. Starting with a simple 2D geometry and in an inertia frame, we study plasma equilibria avoiding the complications of numerical simulations. We show that, below the convective envelope of an AGB star, variable magnetic fields induce a natural expansion, permitted by the almost ideal MHD conditions, in which the radial velocity grows as the second power of the radius. We then study the convective envelope, where the complexity of macro-turbulence allows only for a schematic analytical treatment. Here the radial velocity depends on the square root of the radius. We then verify the robustness of our results with 3D calculations for the velocity, showing that, for both the studied regions, the solution previously found can be seen as a planar section of a more complex behavior, in which anyway the average radial velocity retains the same dependency on radius found in 2D. As a final check, we compare our results to approximate descriptions of buoyant magnetic structures. For realistic boundary conditions the envelope crossing times are sufficient to disperse in the huge convective zone any material transported, suggesting magnetic advection as a promising mechanism for deep mixing. The mixing velocities are smaller than for convection, but larger than for diffusion and adequate to extra-mixing in red giants.