Abstract
Variability has been one of the critical challenges in the implementation of large resistive random access memory (RRAM) arrays. Wide variations in set/reset, read and cycling characteristics can significantly reduce the design margin and feasibility of a memory array. Predicting the characteristics of RRAM cells is constructive to provide insights and to adjust the memory operations accordingly. In this study, a strong correlation between the cell performance and its initial state is found in contact RRAM (CRRAM) cells by 28nm CMOS logic technology. Furthermore, a verifyreset operation is proposed to identify the type of conductive filament (CF) in a cell. Distinctive CRRAM characteristics are found to be linked directly to initial CFs, enabling preliminary screening and adaptive resets to address the large variability problems in sizable CRRAM arrays.