Among non-volatile memory technologies, NAND Flash represents a significant portion in the IC market and has benefitted from the traditional scaling of semi-conductor industry allowing its high density integration. However, this scaling seems to be problematic beyond the 22 nm node. In an effort to go beyond this scaling limitation, alternative memory solutions are proposed among which Resistive RAM (RRAM) stands out as a serious candidate for NAND Flash replacement. Hence, in this PhD thesis we try to respond to many open questions about RRAM devices based on hafnium oxide (HfO2), in particular, by addressing the lack of detailed physical comprehension about their operation and reliability. The impact of scaling, the role of electrodes, the process of defects formation and diffusion are investigated. The impact of alloying/doping HfO2 with other materials for improved RRAM performance is also studied. Finally, our study attempts to provide some answers on the conductive filament formation, its stability and possible composition.