Applicability of Well-Established Memristive Models for Simulations of Resistive Switching Devices

Abstract

Abstract-Highly accurate and predictive models of resistive switching devices are needed to enable future memory and logic design. Widely used is the memristive modeling approach considering resistive switches as dynamical systems. Here we introduce three evaluation criteria for memristor models, checking for plausibility of the I-V characteristics, the presence of a sufficiently non-linearity of the switching kinetics, and the feasibility of predicting the behavior of two anti-serially connected devices correctly. We analyzed two classes of models: the first class comprises common linear memristor models and the second class widely used non-linear memristive models. The linear memristor models are based on Strukov's initial memristor model extended by different window functions, while the non-linear models include Pickett's physics-based memristor model and models derived thereof. This study reveals lacking predictivity of the first class of models, independent of the applied window function. Only the physics-based model is able to fulfill most of the basic evaluation criteria.

Improved Model of TiO2 Memristor

“…The transition between (10) and (11) is discontinuous. Therefore the direct use of those formulae in simulation models may bring serious problems both for model parameter identification and for time-domain integration.…”

“…To keep the state variable within physical bounds, the model was later completed with the so-called window functions by Joglekar [7], Biolek [8], Prodromakis [9] and others. These models are popular for their simplicity, easy implementation in SPICE and the possibility of obtaining an analytical solution but, as shown in a comprehensive study [10], they are not able to capture the complex nonlinear behavior of real devices.…”

“…The model represents very well both the static characteristics and the switching dynamics but its simulation is sensitive to changes of the excitation signal [10], [13]. Several authors have proposed simpler generic behavioral models that do not exhibit simulation problems and can also be used for other devices.…”

“…This class includes, for example, models proposed by Laiho [14], Chang [15], and Yakopcic [16]. However, the study [10] points out a possible danger of these models predicting incorrectly the memristor behavior.…”

Reliable Modeling of Ideal Generic Memristors via State-Space Transformation

“…Today frequently utilized "Hyperbolic Sine Model" [28] was generalized in [29], and another modification of this model was published in [30]. An excellent analysis of the hyperbolic models is given in [31].…”