We analyze the linear stability of a stalled accretion shock in a perfect gas with a parametrized cooling function L ~ rho^{beta-alpha} P^alpha. The instability is dominated by the l=1 mode if the shock radius exceeds 2-3 times the accretor radius, depending on the parameters of the cooling function. The growth rate and oscillation period are comparable to those observed in the numerical simulations of Blondin & Mezzacappa (2006). The instability mechanism is analyzed by separately measuring the efficiencies of the purely acoustic cycle and the advective-acoustic cycle. These efficiencies are estimated directly from the eigenspectrum, and also through a WKB analysis in the high frequency limit. Both methods prove that the advective-acoustic cycle is unstable, and that the purely acoustic cycle is stable. Extrapolating these results to low frequency leads us to interpret the dominant mode as an advective-acoustic instability, different from the purely acoustic interpretation of Blondin & Mezzacappa (2006). A simplified characterization of the instability is proposed, based on an advective-acoustic cycle between the shock and the radius r_nabla where the velocity gradients of the stationary flow are strongest. The importance of the coupling region in this mechanism calls for a better understanding of the conditions for an efficient advective-acoustic coupling in a decelerated, nonadiabatic flow, in order to extend these results to core-collapse supernovae.

The puzzle of birth velocities of pulsars (pulsar kicks) could be solved by an asymmetric explosion of type II Supernovae. We propose a simple hydrodynamical mechanism in order to explain this asymmetry, through the advective-acoustic cycle (Foglizzo 2002) : during the phase of stalled shock, an instability based on the cycle between advected perturbations (entropy / vorticity) and acoustic perturbations can develop between the shock and the surface of the nascent neutron star. Eigenfrequencies are computed numerically, improving the calculation of Houck & Chevalier (1992). The linear instability is dominated by a mode l=1, as observed in the numerical simulations of Blondin et al. (2003) and Scheck et al. (2004). The frequency dependence of the growth rate reveals the presence of the advective-acoustic cycle.

The instability of Bondi-Hoyle-Lyttleton accretion, observed in numerical simulations, is analyzed through known physical mechanisms and possible numerical artefacts. The mechanisms of the longitudinal and transverse instabilities, established within the accretion line model, are clarified. They cannot account for the instability of BHL accretion at moderate Mach number when the pressure forces within the shock cone are taken into account. The advective-acoustic instability is considered in the context of BHL accretion when the shock is detached from the accretor. This mechanism naturally explains the stability of the flow when the shock is weak, and the instability when the accretor is small. In particular, it is a robust proof of the instability of 3D accretion when gamma=5/3 if the accretor is small enough, even for moderate shock strength (M sim 3). The numerical artefacts that may be present in existing numerical simulations are reviewed, with particular attention paid to the advection of entropy/vorticity perturbations and the artificial acoustic feedback from the accretor boundary condition. Several numerical tests are proposed to test these mechanisms.

The linear stability of stalled accretion shocks is investigated in the context of core collapse of type II supernovae. We focus on a particular instability mechanism based on the coupling of acoustic perturbations with advected ones (vorticity, entropy). This advective-acoustic cycle takes place between the shock and the nascent neutron star. Both adiabatic and non-adiabatic processes may contribute to this coupling, but only adiabatic ones are considered in this first approach. The growth time of the adiabatic instability scales like the advection time, and is dominated by low degree modes l=0,1,2. Non radial modes (l=1,2) found unstable by Blondin et al. (2003) can be related to this mechanism.