Neutrino-driven convection versus advection in core collapse supernovae

A toy model is analyzed in order to evaluate the linear stability of the gain region immediately behind a stalled accretion shock, after core bounce. This model demonstrates that a negative entropy gradient is not sufficient to warrant linear instability. The stability criterion is governed by the ratio "chi" of the advection time through the gain region divided by the local timescale of buoyancy. The gain region is linearly stable if chi>3. For chi>3, perturbations are unstable in a limited range of horizontal wavelengths centered around twice the vertical size H of the gain region. The threshold horizontal wavenumbers k_{min} and k_{max} follow simple scaling laws such that Hk_{min}\propto 1/chi and Hk_{max}\propto chi. These scaling laws are understood as the consequence of a vortical-acoustic cycle within the gain region, fed by the Rayleigh-Taylor growth of vorticity perturbations during advection. The stability of short wavelength perturbations is compared to the "ablative stabilization" of accelerated ablation fronts. The convective stability of the l=1 mode in spherical accretion is discussed, in relation with the asymmetric explosion of core collapse supernovae. The advective stabilization of long wavelength perturbations weakens the possible influence of convection alone on a global l=1 mode. Convection may however cooperate efficiently with a global vortical-acoustic cycle extending below the gain radius