Investigating Transient Characteristics of Volatile Hysteresis and Self-Heating of PrMnO$_3$ based RRAM

TL;DR

This paper investigates the transient and volatile hysteresis behavior of PrMnO$_3$-based RRAM, proposing a physical electrothermal model to understand its dynamics for neuromorphic device applications.

Contribution

It provides a comprehensive analysis of the hysteresis mechanisms and introduces a novel electrothermal simulation model based on SCLC and Fourier Heat equations.

Findings

Hysteresis window varies with voltage ramp rate

Self-heating significantly influences hysteresis behavior

The model accurately reproduces experimental hysteresis dependence

Abstract

PrMnO$_3$ (PMO) based RRAM shows selector-less behavior due to high-non-linearity. Recently, the non-linearity, along with volatile hysteresis, is demonstrated and utilized as a compact oscillator to enable highly scaled oscillatory neurons, which enable oscillatory neuromorphic systems found in the human cortex. Hence, it is vital to understand the physical mechanisms behind such a volatile hysteretic behavior to provide useful insights in developing a device for various neuromorphic applications. In this paper, we present a comprehensive investigation of the transient characteristics and propose a physical mechanism to replicate the observations by simulations. First, we investigate the complex dynamics of the hysteresis with the voltage ramp rate. We observe that the voltage window initially increases and later decreases as the ramp rate is increased - while the current window reduces monotonically. Second, an analytical electrothermal model based on space charge limited current (SCLC) and Fourier Heat equation is proposed to model the co-dependent heat and current flow. Finally, we show that the interplay between the self-heating due to the current and the current dependence on the temperature is accurately modeled to reproduce the hysteresis dependence on the various voltage ramp rate. Such a detailed understanding of the device PMO RRAM volatile hysteresis may enable efficient device design required in neuromorphic computing applications.