Title:Hafnia for analog memristor: Influence of stoichiometry and crystalline structure

Abstract:The highly non-linear switching behavior of hafnia memristor actually hinders its wide application in neuromorphic computing. Theoretical understanding into its switching mechanism has been focused on the processes of conductive filament generation and rupture, but possible phase transition and crystallization around the region of conductive filaments (CFs) due to the variation of O content have been paid less attention to. In this paper, HfO$\mathrm{_x}$ structural models covering the full stoichiometries from Hf to HfO$\mathrm{_2}$ were established, and the crystal structure evolution during the reduction process of hafnia was obtained through first-principles calculation. The electronic structures and O vacancy migration characteristics of these structures were analyzed. A criterion was prescribed to predict the mode of abrupt binary switching or gradual conductance modulation according to the structure evolution of the CFs. In particular, factors that influence the merging of tiny conductive channels into strong filaments are intensively discussed, including the anisotropy of O vacancy migration and the size effect. The feasibility of Mg doping to achieve robust gradual switching is discussed.