Memristor Model Optimization Based on Parameter Extraction From Device Characterization Data

This paper presents a memristive device model capable of accurately matching a wide range of characterization data collected from a tantalum oxide memristor. Memristor models commonly use a set of equations and fitting parameters to match the complex dynamic conductivity pattern observed in these devices. Along with the proposed model, a procedure is also described that can be used to optimize each fitting parameter in the model relative to an I-V curve. Therefore, model parameters are self-updated based on this procedure when a new cyclic I-V sweep is provided for model optimization. This model will automatically provide the best possible match to the characterization data without any additional optimization from the user. In this paper, multiple cyclic I-V characterizations are modeled from ten different tantalum oxide devices (on the same wafer). Additionally, studies were completed to demonstrate the amount of variation present between devices on a wafer, as well as the amount of variation present within a single device. Methods for modeling this variation are then proposed, resulting in an accurate and complete, automated, memristor modeling approach.

I. Introduction

The memristor was theorized in 1971 [1], and was first discovered in physical form in 2008 [2], [3]. The memristor devices we consider in this paper are nonvolatile nanoscale two-terminal passive circuit elements that have dynamic resistance dependent on the total charge applied between the positive and negative terminals. Given its unique device properties, the memristor has been proposed for use in many novel memory [4], logic [5], [6], and neuromorphic systems. Memristor-based neuromorphic systems are especially interesting, as they involve using memristors to mimic the functionality of a synapse in brain tissue [7], [8]. Just as electrochemical pulses are applied to a synapse to change connection strength, voltage pulses can be applied to a memristor to change the resistance. Memristors can be arranged in a crossbar [9] to provide high density and high connectivity, leading to massively parallel analog computation. Using this approach, several memristor-based neural systems have been proposed, including on-line learning [10] and deep network implementation [11]. Neuromorphic systems based on memristor crossbars have potential to perform at a power efficiency of 6 to 8 orders of magnitude greater than that of traditional RISC processors [12].