Type Ia supernovae are an important tool to determine the expansion history of our Universe. Thus, considerable attention has been given to both observations and models of these events. The most popular explosion model is the central ignition of a deflagration in the dense C+O interior of a Chandrasekhar mass white dwarf, followed by a transition to a detonation (TDD). We study in this thesis a new mechanism for this transition.The most robust and studied progenitor model and the postulated mechanism for the TDD, the so called 'Zel'dovich gradient mechanism', are presented. State of the art 3D simulations of such a delayed detonation, at the price of some adjustments, can indeed reproduce observables. But due to largely unresolved physical scales, such simulations cannot explain the TDD by themselves, and especially, the physical mechanism which triggers this transition - which is not yet understood, even on Earth, for unconfined media. It is then discussed why the current Zel'dovich mechanism might be too constraining for a SN Ia model, pointing to a new approach, which is the core result of this thesis.In the final part, our alternative model for DDT in supernovae, the acoustic heating of the pre-supernova envelope, is presented. A planar model first proves that small amplitude acoustic perturbations (generated by a turbulent flame) are actually amplified in a steep density gradient, up to a point where they turn into shocks able to trigger a detonation. Then, this mechanism is applied to more realistic models, taking into account, in spherical geometry, the expanding envelope. A parametric study demonstrates the validity of the model for a reasonable range of acoustic wave amplitudes and frequencies.To conclude, some exploratory 2D and 3D MHD simulations, seeking for realistic acoustic source compatible with our mechanism, are presented.
