Abstract : The new systems developed by automotive sector require more and more electronics and will require more power than what can supply 12 and 24 Volts batteries of today. In this context new 42 Volts batteries are going to be introduced. In a first time, we showed that classical solutions were not suitable anymore to decrease the specific on-resistance of conventional unipolar devices to break the conventional "silicon limit". It is for this reason that our research works turned to new power MOS transistors (FLYMOS") adapted to 42 Volts. The work presented in this thesis is based on the concept of the Floating Islands (FLIDiode) and on its application to MOS devices (FLYMOS") in order to improve the "voltage handling capability/specific on-resistance" trade-off. The concept consists in the introduction of a P-buried floating layer in the N- drift region, in order to divide, under reverse bias condition, the maximal electric field in two parts and, consequently, to improve the breakdown voltage without degrading the specific on-resistance. In a second time, the technological parameters were optimized by "two dimensional physical simulations" to design the FLYMOS" transistor (edge and central cells). This FLYMOS" presents an optimal "voltage handling capability/specific on-resistance" trade-off. Then the floating island concept has been validated by the technical realization of low tension of FLIDiode and FLYMOS" (BVdss<100V). So the best realized FLYMOS" transistors present a breakdown voltage higher than the breakdown voltage of "plane junction" and a specific on-resistance strongly improved since it is close to the silicon conventional limit.