Single grain ferroelectrics: the holy grail of digital memory technology
With the evolution of data science, artificial intelligence and machine learning algorithms, it is necessary to store and retrieve a huge amount of data efficiently. This requires even larger device densities in memory technology in order to keep data center dimensions and power consumption within certain limits. Although the present high-density 3D NAND technology has been the main workhorse for nonvolatile solid-state memory, it is rapidly approaching its scaling limits. Therefore, the quest for a novel energy-efficient and scalable technology has started. One of the most promising candidates are ferroelectric memory devices, such as the ferroelectric field-effect transistor (Fe-FET), the ferroelectric capacitor (Fe-CAP) and the ferroelectric tunnel junction (FTJ). Using ferroelectric polarization as a means of storing information in a nonvolatile way, it is possible to achieve the lowest possible energy consumption per bit. However, the quest for a ferroelectric material that can be integrated in silicon is still ongoing. Perovskites were found to be less compatible with CMOS, restricting their usage in high density memories. On the other hand, the more recently discovered hafnia-based ferroelectrics are suffering from a highly defective structure containing different phases in a polycrystalline configuration, leading to reliability as well as variability issues. Some recent studies show that ferroelectric materials with a wurtzite-type crystal structure can be promising alternatives to the perovskite- and hafnia-based ferroelectrics. They owe this to their theoretically-predicted high remnant polarization and the possibility that the desired ferroelectric phase can be grown in a single grain. In such a single grain ferroelectric material, the more coherent ‘single dipole switching’ would enable a more rectangular hysteresis loop. As a consequence, this would allow an optimal separation between the read and write condition without compromising the low voltage and low power capability which is inherent to the ferroelectric switching mechanism. However, the wurtzite-based ferroelectrics require further optimization and tuning (such as doping, proper electrodes selection etc.) to achieve these desired properties. The purpose of this thesis is to study the concept of single grain ferroelectricity using different experimental characterization techniques on ferroelectric devices made up of new ferroelectric materials provided from the ferroelectric research program at imec. To support the arguments and analyze the results, further theoretical or simulation study may be necessary. Finally, the aim is to achieve a workable single grain ferroelectric memory: the holy grail for digital memory technology.