[en] This thesis investigates innovative fast reactors (FRs) and the possibility of using different liquid-metal coolants (Pb, Pb-⁷Li, Pb-Mg). The comparison is made for the requirements of 'fourth generation' reactors, as defined in the framework of the Generation IV international forum, and in particular reactor safety. More specifically, two important safety criteria are studied in detail: the issue of the void reactivity effect, which is an inherent drawback of FRs; and the behaviour of the reactor during unprotected transients. The potential of these innovative coolants is investigated through simulations of their use in a concept similar to the Russian project BREST-300. Two options were considered for the fuel: an oxide fuel, which is a mature technology, and an innovative, denser, nitride fuel. A study at low power was complemented by investigations of industrial size cores. A concept similar to the SFR reactor (Sodium Fast Reactor) is taken as the state-of-the-art reference point. These concepts were initially dimensioned using parametric studies covering the main core geometry options, taking into account both neutronics and thermal-hydraulics aspects in steady-state conditions. Furthermore, we propose a systematic approach for evaluating the tolerance to the main accident types (loss of primary flow, transient over-power, loss of heat sink, and over-cooling) based on a quasi-static reactivity approach complemented by the use of a dynamics code. In addition to these studies based on the U-Pu fuel cycle, we show the advantage of using the Th--U fuel cycle as an effective way of reducing the void effect for these innovative FR concepts. (author)