Contents I. Introduction
Ferroelectrics are materials that show a remanent polarization that can be switched between two directions using an electrical field. The stable nonvolatile polarization paired with the purely field driven switching, make this class of materials a natural choice for nonvolatile memories overcoming the write inefficiency of most other emerging nonvolatile memory concepts [1]. However, typically the materials that show ferroelectricity have a complex structure with three or more compounds and in the case of oxides normally weakly bound oxygen (see fig. 1 b). These properties make them very hard to integrate into a state of the art CMOS process [2]. As a result, the established ferroelectric memory technologies are limited in scaling and therefore are only used in niche applications. Thus, after an intense phase of research and development between the late 1990s and the mid 2000s, the interest in ferroelectrics for memory applications diminished. The discovery of ferroelectricity in doped hafnium oxide (see fig. 1a) has changed that view since hafnium oxide is used as the gate dielectric in modern transistors since 2007. Traditionally, a 1 transistor – 1 capacitor cell (FeRAM see fig. 2a) inspired by DRAM is used in established ferroelectric memory devices. An alternative way is to integrate the ferroelectric into the gate stack of an MOS transistor (FeFET see fig. 2b). Finally, a resistive switching ferroelectric tunneling junction can be achieved when the ferroelectric is made very thin and allows for a polarization dependent tunneling current (FTJ see fig. 2c). All three memory concepts have seen increased activities by using hafnium oxide based ferroelectrics rather than traditional perovskites or layered perovskites.
