Experimental and Theoretical Study of Multifilamentary Resistive Switching in Nanoscale Transition Metal Oxide Films

The development of metal oxide based resistive random-access memory is of the greatest importance for creating a new class of memory in various applications, including large-scale memory arrays, analogue neuromorphic computing systems, and energy-efficient system-on-chips. The filamentary switching mechanism of resistive memory provides great scalability, while its nonvolatility, energy efficiency, high retention, and write/erase speeds make resistive memory a prospective universal memory. The ability of resistive memory to mimic the behavior of biological synapses coupled with its multilevel storage capabilities makes this novel type of memory a suitable hardware for analogue neuromorphic systems. However, practical application for large-volume memory arrays for neuromorphic systems is limited by unresolved fundamental issues including intrinsic cycle-to-cycle variability of a high resistance state and extrinsic cell-to-cell variability of switching from a low to a high resistive state. Mitigation of these issues is impossible without a clear understanding of physical processes occurring inside dielectric layers of resistive memory. In this paper, we address these issues via theoretical and experimental investigation and demonstrate that both issues are related to the formation of multiple conductive filaments.