Subfilamentary Networks and Their Dynamics As Source of Cycle-to-Cycle Variability in Reram Devices

Abstract

One major obstacle for the implementation of redox-based memristive devices in non-volatile memory or neuromorphic computing is the large cycle-to-cycle and device-to-device variability [1]. Here, we combine different complementary analysis methods such as atomic force microscopy (AFM), thermography and operando photoelectron emission spectroscopy (PEEM) [2] to monitor the filament shape and position in memristive SrTiO3 cells and correlate it to the observed variability of the resistance values. Our operando PEEM analysis revealed that some devices exhibit cycle-dependent variations in the shape of the conductive filament or in the oxygen vacancy distribution at and around the filament [3]. In other cases, even the location of the active filament changes from one cycle to the next. We propose that both effects originate from the coexistence of multiple (sub-)filaments observed by transmission electron microscopy [4] and that the active, current-carrying filament may change from cycle to cycle. Besides this, thermal imaging enabled us to monitor the position of the filament during repeated cycling. During cycling of some devices, we observed a movement of the filament towards the edge of the device that goes along with a drift of the low resistive state. As soon as the filament reaches the boarder, the resistance value stabilizes and remains constant during the following cycles. By modifying the interface at the top electrode, we could engineer the position of the filament and its stability against cycling. Our studies thereby represent the scientific basis, rather than prior purely empirical engineering approaches, for developing stable memristive devices.