Electrochemical processes and device improvement in conductive bridge RAM cells

Abstract

In this paper, we discuss the recent progress on the fundamental understandings of ECM/CBRAM cells but also on the improved device structures and reliability for high‐density applications. The influences of the local chemical environment and the material selection/combination are highlighted, and the filament dynamics is described in a general framework that relates all the reported switching modes. Furthermore we also detail some correlation evidences between the filament shape and device electrical characteristics. Finally, we discuss technological challenges related to current‐scaling and cell size‐scaling. Large switching variability associated to low current needs to be mitigated by appropriate Write‐verify methods, while cell scaling requires novel processing techniques such as Cu dry‐etch.

Microstructural transitions in resistive random access memory composed of molybdenum oxide with copper during switching cycles

“…These results for the Cu/MoO x reference devices reflect the typical switching characteristics of CBRAMs. 6,13 Therefore, the in situ TEM results that are described in the next section are expected to provide useful information to aid in understanding of the switching operation of CBRAMs.…”

“…In the second cycle, with its maximum voltage of +1.4 V, a weak hysteretic property was identified in the set process. According to the electrochemical model, 6,8,12,13 Cu is believed to be dissolved, and subsequently drifts in the MoO x layer to the BE, where it forms deposits. However, no change was seen in the TEM image.…”

“…The growth and shrinkage scheme observed here fits the behaviour of the filament models reported to date. 6,8,12,13 However, the CF in MoO x does not need to bridge the TE and the BE to realize a resistance reduction.…”

“…[4][5][6][7][8][9][10][11][12][13] This is occurring because the ReRAM shows strong potential for use as a nextgeneration memory; it offers a simple device structure that would be advantageous for integration, a high on/off ratio, and nonvolatility. In addition to binary operation, analogue memory operations that can be used for artificial neural network hardware have been investigated energetically in recent years.…”

“…The molybdenum oxide (MoO x )-based device with the Cu electrode investigated in this report is a type of CBRAM, 29 while MoO x shows ReRAM switching behaviour even without the Cu electrode. 30 Here, the ReRAM operation of the Cu/MoO x /TiN structure is briefly described based on the assumption of the conventional electrochemical model for the CBRAM, 6,8,12,13 where Cu is the active top electrode (TE) and TiN is the inactive bottom electrode (BE). Application of a positive voltage to the Cu TE induces oxidation of the Cu to produce cations.…”