Identify the critical regions and switching/failure mechanisms in non-filamentary RRAM (a-VMCO) by RTN and CVS techniques for memory window improvement

Non-filamentary RRAM is a promising technology that features self-rectifying, forming/compliance-free, tight resistance distributions at both high and low resistance states (HRS/LRS). Direct experimental evidence for its physical switching & failure mechanisms, however, is still missing, due to the lack of suitable characterization techniques. In this work, a novel method combining the random-telegraph-noise (RTN), constant-voltage-stress (CVS) and time-to-failure Weibull plot is developed to investigate these mechanisms in the non-filamentary RRAM cell based on amorphous-Si/TiO2. For the first time, the following key advances have been achieved: i) Switching mechanism by defect profile modulation in a critical interfacial region has been identified from defect locations extracted by RTN; ii) Defect profile in this region plays a critical role in device failure, leading to different Weibull distributions during negative (LRS) and positive (HRS) CVS; iii) Progressive formation of a conductive percolation path during electrical stress is directly observed due to defect generation in addition to pre-existing defect movement; iv) Optimizing the critical interfacial region significantly improves memory window and failure margin. This provides a useful tool for advancing the non-filamentary RRAM technology.