Abstract
Repeated and fast switching between on/off state without read window degradation is key for the development of resistive random access memory (RRAM) to meet reliable nonvolatile storage demands. In cycling study of the cell, we first established an incremental step pulse programing (ISPP) algorithm for 1T-1R backfilled contact resistive random access memory (BCRRAM) array. Experimental data reveals a strong correlation between reset time and the types of conductive filament (CF) formed. A training methodology is proposed to reduce the chance of the stochastic CF growth inside BCRRAM to maintain efficient reset operations and extend the endurance capability.