Quantitative endurance failure model for filamentary RRAM

Abstract

Endurance in filamentary RRAM is modeled in the framework of the hourglass model. Two failure modes are distinguished: (i) stochastic set failure is caused by defect generation near the bottom electrode, and (ii) resistive window changes are controlled by T-activated changes of the number of filament vacancies. Bottom electrode/oxide interface optimization is the prime knob for endurance improvement. This model enables quantitative and predictive endurance simulations.