Impact of the nucleation of charge clusters on the retention of memristors: a self-consistent phase field computational study

TL;DR

This study uses a phase-field computational model to show that charge-cluster nucleation during switching cycles causes retention loss in memristors, linking the phenomenon to cycle count and experimental observations.

Contribution

It introduces a phase-field model to explain retention loss in memristors as due to charge-cluster nucleation, a novel mechanistic insight.

Findings

Charge-cluster nucleation correlates with retention loss.

Retention loss probability increases with switching cycles.

Model aligns with experimental retention loss data.

Abstract

In recent years, resistive RAM often referred to as memristor is actively pursued as a replacement for nonvolatile-flash memory due to its superior characteristics such as high density, scalability, low power operation, high endurance, and fast operating speed. However, one of the challenges that need to be overcome is the loss of retention for both ON- and OFF-states; the retention loss. While various models are proposed to explain the retention loss in memristors consisting of a switching layer, in this paper, we propose that the nucleation of clusters made of electrical charges, charge-clusters, in the switching layer acts as a potential root cause for the retention loss. The nucleation results from localized electric-field produced intermittently during cyclic switching operations. We use the phase-field method to illustrate how the nucleation of charge-clusters gives rise to the retention loss. Our results suggest that the degree at which the retention loss arises is linked to the number of cyclic switching operations since the probability at which nucleation centers form increases with the number of cycle switching operations, which is consistent with a range of experimental findings previously reported.