State Characterisation of Self-Directed Channel Memristive Devices

TL;DR

The paper tackles the challenge of reliably identifying and tracking the resistive state of self-directed channel memristors, proposing a physics-grounded state model and a noise-aware state estimation scheme. It introduces a parallel diode+Ohmic conduction model with state-dependent diode terms to capture nonlinear and asymmetric VI characteristics, and performs rigorous parameter fitting against GMSS baselines with a novel fitting-error metric. A minimum-variance estimator is derived to infer the state from noisy voltage–current measurements, and experimental data validate improved state identifiability and tracking over time. This approach enhances the reliability of SDC memristor state characterization for storage, neuromorphic computing, and related emerging applications.

Abstract

Knowing how to reliably use memristors as information storage devices is crucial not only to their role as emerging memories, but also for their application in neural network acceleration and as components of novel neuromorphic systems. In order to better understand the dynamics of information storage on memristors, it is essential to be able to characterise and measure their state. To this end, in this paper we propose a general, physics-inspired modelling approach for characterising the state of self-directed channel (SDC) memristors. Additionally, to enable the identification of the proposed state from device data, we introduce a noise-aware approach to the minimum-variance estimation of the state from voltage and current pairs.

Figures (7)

• Figure 1: The layer structure of the SDC (Knowm) memristor used in our experiments.
• Figure 2: Each state of the memristor parameterises a VI characteristic, where each voltage input corresponds to a unique current output. A change of state is associated with a shift to a different set of voltage-current pairs.
• Figure 3: Proposed physical model explaining the SDC memristor's switching mechanism, with the gaps between silver ion agglomeration sites resulting in an electron tunnelling barrier conduction characteristic yangMemristiveSwitchingMechanism2008. Increased density of hopping sites accelerates conduction.
• Figure 4: Illustrative figures showing the READ, SET, and RESET waveforms used to measure and set/reset the memristors as part of our experiments. $T_{read}$, $T_{set}$, and $T_{reset}$ are variables denoting the period of the waveforms, respectively. While, $A_{read}$, $A_{set}$, and $A_{reset}$ denote the given amplitudes.
• Figure 5: The measurement circuit for the memristor. $V_{\text{tot}}$ is the voltage measured across the entire circuit, while $V_{\text{series}}$ is the voltage across the series resistor. From these measurements, we can determine the voltage across and current through the memristor.