The deacylation step of serine protease catalysis is studied using DFT and ab initio QM/MM calculations combined with MD/umbrella sampling calculations. Free energies of the entire reaction are calculated in the gas phase, in a continuum solvent, and in the enzyme elastase. The calculations show that a concerted mechanism in the gas phase is replaced by a stepwise mechanism when solvent effects or an acetate ion are added to the reference system, with the tetrahedral intermediate being a shallow minimum on the free energy surface. In the enzyme, the tetrahedral intermediate is a relatively stable species ( approximately 7 kcal/mol lower in energy than the transition state), mainly due to the electrostatic effects of the oxyanion hole and Asp102. It is formed in the first step of the reaction, as a result of a proton transfer from the nucleophilic water to His57 and of an attack of the remaining hydroxyl on the ester carbonyl. This is the rate-determining step of the reaction, which requires approximately 22 kcal/mol for activation, approximately 5 kcal/mol less than the reference reaction in water. In the second stage of the reaction, only small energy barriers are detected to facilitate the proton transfer from His57 to Ser195 and the breakdown of the tetrahedral intermediate. Those are attributed mainly to a movement of Ser195 and to a rotation of the His57 side chain. During the rotation, the imidazolium ion is stabilized by a strong H-bond with Asp102, and the C(epsilon)(1)-H...O H-bond with Ser214 is replaced by one with Thr213, suggesting that a "ring-flip mechanism" is not necessary as a driving force for the reaction. The movements of His57 and Ser195 are highly correlated with rearrangements of the binding site, suggesting that product release may be implicated in the deacylation process.