60-62, Boulevard Saint Michel 75272 PARIS cedex 06 FRANCEiii posons une séquence de pulses qui permettrait de rétablir cette information dans l'état de la cavité. Ceci pourrait prolonger la durée de vie d'un qubit de presque un ordre de grandeur. Le grand intérêt de cette proposition est double. Premièrement, elle nécessite un nombre de composants physiques minimal: seulement un qubit fortement couplé (dans le régime dispersif) à un mode de la cavité. Deuxièmement, nous avons des séquences de pulses pour l'encodage d'un qubit dans la cavité, la correction de l'état de la cavité suite à un saut quantique, et enfin le décodage de l'état de la cavité vers l'état du qubit.iv Abstract This thesis tackles the problem of preparing and stabilizing highly non classical states of quantum systems, as well as identifying their Hamiltonian. Most of the methods we propose were developed in close collaboration with experimentalists. Some of these ideas have already been implemented in the laboratory.First, we derived an observability result for Hamiltonian identification. We consider a typical setting in ultra-fast quantum control experiments where an observable is measured after a system (e.g a cloud of atoms) has interacted with a controlled electric field. We thus pose the problem of whether we can find a set of controls which would provide sufficient information to reconstruct the dipole moment matrix. We prove that under certain conditions, we can find controls (in analogy with Ramsey interferometry) which make the dipole moment matrix locally observable. The rest of the thesis is devoted to quantum state engineering and stabilization.The second contribution is on ensemble control. We propose a robust control which performs any state permutation on an ensemble of quantum ladder systems. The strength of this result which is based on adiabatic theory, is that this single control performs the same permutation on an ensemble of different systems. The results are based on a detailed asymptotic mathematical analysis giving error bounds.Third, we address the problem of quantum state stabilization by reservoir engineering in the context of cavity quantum electrodynamics (QED). We describe an experiment where flying atoms successively interact with one or two cavity modes. We find a control to apply during each atom-cavity interaction which would drive and stabilize the cavity modes close to a predefined non-classical target state. In particular, we are able to stabilize an entangled state which would lead to a violation of Bell's inequality for arbitrarily large times. The proposal is supported by numerical simulations and a first mathematical convergence analysis, where the Hilbert space is infinite dimensional (no truncation to a finite number of photons).Fourth, we introduce a control which performs quantum state preparation in the context of Josephson circuits (also called circuit QED). We consider one superconducting qubit coupled to a microwave resonator and describe a sequence of pulses which would generate any superpositio...