A Physics-Based Circuit Model for Magnetic Tunnel Junctions

Steven Louis \affilmark 1*, Hannah Bradley \affilmark 2, Artem Litvinenko \affilmark 3, Vasyl Tyberkevych \affilmark 2 · 2025

Read A Physics-Based Circuit Model for Magnetic Tunnel Junctions by Steven Louis \affilmark 1*, Hannah Bradley \affilmark 2, Artem Litvinenko \affilmark 3, Vasyl Tyberkevych \affilmark 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-20813.txt.

Introduction

Steven Louis and co-authors wrote this engineering paper in 2025 on magnetic tunnel junctions, tiny stacks whose resistance flips with magnetism. Such parts anchor spintronic memory and brain-like chips. Designers need a circuit model that acts like the real part, not just a rough sketch. The team built one from the physics of the stack and ran it in LTspice, a common circuit simulator.

They checked the model against direct math solutions of the governing equation. The two agreed closely across key acts: resonance, switching by field or by spin push, and steady oscillation. The model takes live fields and voltages as inputs, which suits rich designs. The authors offer it as a fast, true stand-in that lets whole circuits be tested with such junctions inside.

Catalog record

CREATORSteven Louis \affilmark 1*, Hannah Bradley \affilmark 2, Artem Litvinenko \affilmark 3, Vasyl Tyberkevych \affilmark 2
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-20813

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A Physics-Based Circuit Model for Magnetic Tunnel Junctions Steven Louis \affilmark 1*, Hannah Bradley \affilmark 2, Artem Litvinenko \affilmark 3, Vasyl Tyberkevych \affilmark 2 Abstract This work presents an equivalent circuit model for magnetic tunnel junctions (MTJs) that accurately reproduces their magnetization dynamics and electrical behavior within the macrospin approximation. The model is validated through direct numerical simulations of the Landau–Lifshitz–Gilbert–Slonczewski (LLGS) equation, encompassing ferromagnetic resonance, field- and spin-torque-induced switching, and spin-torque-induced oscillations. Simulation results exhibit excellent agreement between the equivalent circuit model and the LLGS-based simulations, confirming the model accuracy and utility for efficient circuit-level analysis of MTJs. The capability of handling time-dependent magnetic fields and voltage-driven excitations renders the proposed model applicable to diverse areas, including neuromorphic computing, microwave signal processing, and spintronic memory technologies. By providing a computationally efficient yet physically rigorous circuit representation, this work facilitates seamless integration of MTJs into complex electronic systems, thereby accelerating the advancement of novel spintronic circuit architectures. \sptitle Magnetic tunnel junction equivalent circuit modeling \corresp Corresponding author: Steven Louis ([email protected]). {IEEEkeywords} Magnetic tunnel junction (MTJ), equivalent circuit modeling, spin-transfer torque, spintronics, magnetoresistive random-access memory (MRAM), LTspice, circuit simulation, ferromagnetic resonance (FMR). 1 INTRODUCTION In the past two decades, magnetic tunnel junctions (MTJs) have revolutionized data storage by enabling high-density recording in disk-based drives [Chappert, 2007]. MTJs have also been integrated with CMOS technology and commercialized in magnetoresistive random-access memory (MRAM) [Alzate 2015, Edelstein 2020] which leverages the non-volatility and fast switching properties of MTJs for energy-efficient, high-speed memory [Ikegawa 2020]. Beyond digital applications, MTJs may also be used as microwave sources [Fuchs 2004, Villard 2009], microwave detectors [Tulapurkar 2005], energy harvesters [Fang 2019], advanced microwave signal processing devices [Litvinenko 2022], and even in neuromorphic computing [Romera 2018, Louis 2024, Rodrigues 2023]. An MTJ consists of two ferromagnetic layers separated by a thin insulating barrier [Singh 2019, Yuasa 2007]. The operational principle of an MTJ is based on the tunneling magnetoresistance effect, which is sensitive to the relative orientation of magnetic moments in the ferromagnetic layers [Yuasa 2007], and on the spin-transfer-torque (STT) effect [Slonczewski 1989, Ralph 2008] that describes influence of electric current on magnetization dynamics. Advanced nanoscale fabrication enabled integration of MTJs into complex electronic systems, expanding their use in memory and logic devices [Alzate 2015, Edelstein 2020]. These distinctive properties position MTJs as important components in next-generation analog and neuromorphic computing systems [Incorvia 2024]. Figure 1: (a) Diagram of the MTJ, illustrating the free and fixed layers. (b) Equivalent circuit of the MTJ, as an intrinsic variable resistor. (c) Equivalent circuit representation of the MTJ free layer magnetization. (d) The complete MTJ circuit implemented in LTspice. Circuit simulation is essential in the design and analysis of electronic systems [Najm 2010]. An accurate circuit-equivalent model of an MTJ facilitates its integration into larger circuits and allows comprehensive performance evaluation, while preserving insight into magnetization dynamics. Prior works [Xu 2012, Panagopoulos 2012, Kim 2015] have used LTspice and other circuit simulators to numerically solve the LLGS equation, for example by mapping angular variables onto circuit elements. However, these approaches do not establish a physical correspondence between the circuit elements and the underlying energy storage, dissipation, or torque mechanisms present in the MTJ. There are models [Bendjeddou 2023, Abbasi 2017] that can emulate the MTJ impedance at high frequency, but do not take into account MTJ behavior under the influence of static inputs. This letter introduces an equivalent circuit model for MTJs that accurately reflects their underlying physical characteristics. Within the proposed model, conservative magnetic interactions (e.g., effect of a bias magnetic field or magnetic anisotropy) are modeled by equivalent nonlinear capacitance and inductance. Dissipative magnetic torques are represented by equivalent resistance (Gilbert damping) and current and voltage sources (STT). The model is realized as a standalone element in LTspice circuit simulation software and its validity is demonstrated by comparing LTspice simulation results with direct numerical simulations of the Landau-Lifshits-Gilbert-Slonczewski (LLGS) equation. 2 Physics-Based Equivalent MTJ Circuit An MTJ comprises a ferromagnetic fixed layer, a non-magnetic spacer, and a ferromagnetic free layer, as shown in Fig. 1 (a). The magnetization of the fixed layer is represented by \mathbf{p} , while the free layer magnetization is denoted by \mathbf{m} . The term “free layer” indicates that the magnetization direction can be altered by external stimuli, such as an applied magnetic field \mathbf{B}_{\mathrm{ext}} or electric current I . In contrast, the fixed layer has a stable magnetization direction that is largely unaffected by external perturbations. The dynamic behavior of \mathbf{m} arises from the interplay between magnetic fields, damping effects, and spin-transfer torque, producing complex magnetization dynamics. These dynamics determine the electrical behavior and switching characteristics of MTJ-based devices [Slavin 2009]. When an electric current I passes through the MTJ, it encounters a resistance denoted R_{\mathrm{mtj}} . The resistance R_{\mathrm{mtj}} reaches a minimum value, R_{\mathrm{P}} , when the magnetization vectors \mathbf{p} and \mathbf{m} are parallel. Conversely, the resistance is maximized, taking the value R_{\mathrm{AP}} , when \mathbf{p} and \mathbf{m} are antiparallel. If the vectors \mathbf{p} and \mathbf{m} are orthogonal, the resistance becomes R_{\perp} , defined by R_{\perp}=2R_{\mathrm{AP}}R_{\mathrm{P}}/(R_{\mathrm{AP}}+R_{\mathrm{P}}) . A general expression for the resistance of an MTJ is [Slonczewski 1989, Wang 2009], R_{\mathrm{mtj}}(\mathbf{m})=\frac{R_{\perp}}{1+\eta^{2}(\mathbf{p}\cdot\mathbf{m})}, (1) where \eta is the dimensionless spin polarization efficiency, given by \eta=\sqrt{(R_{\mathrm{AP}}-R_{\mathrm{P}})/(R_{\mathrm{AP}}+R_{\mathrm{P}})} . The MTJ shown in Fig. 1 (a) can be represented by the circuit in Fig. 1 (b), which incorporates an intrinsic magnetoresistor symbol as defined in [IEEE Standard] 2.1.7. The resistance R_{\mathrm{mtj}}(\mathbf{m}) is intrinsically nonlinear, depending on the free layer magnetization \mathbf{m} . The magnetization dynamics of the free layer can be effectively described using the macrospin approximation, which can be modeled by the Landau-Lifshitz-Gilbert-Slonczewski (LLGS) equation [Slavin 2009], given by \derivative{\mathbf{m}}{t}-|\gamma|\mathbf{B}_{\mathrm{eff}}\times\mathbf{m}=\alpha_{G}\mathbf{m}\times\derivative{\mathbf{m}}{t}-\frac{\gamma}{M_{s}V_{o}}\mathbf{m}\times\mathbf{J}_{s}\times\mathbf{m}, (2) where \gamma is the gyromagnetic ratio, and \alpha_{G} is the Gilbert damping parameter. The spin current \mathbf{J}_{s} can be written as \mathbf{J}_{s}=(\gamma\hbar\eta/2eM_{s}V_{o})(V_{\mathrm{mtj}}/R_{\perp})\mathbf{p} , where \hbar is the reduced Planck constant, \mu_{0} is the vacuum magnetic permeability, e is the elementary charge, V_{o} represents the volume of the free layer, and V_{\mathrm{mtj}} is the voltage across the MTJ. The effective magnetic field \mathbf{B}_{\mathrm{eff}} is given by \mathbf{B}_{\mathrm{eff}}=-(1/M_{s}V_{o})\partial E(\mathbf{m})/\partial\mathbf{m} , where E(\mathbf{m}) is the macrospin energy which is given by E(\mathbf{m})=M_{s}V_{o}\Big[-\mathbf{B}_{e}\cdot\mathbf{m}+\tfrac{1}{2}B_{d}(\mathbf{m}\cdot\hat{\mathbf{z}})^{2}-\tfrac{1}{2}B_{a}(\mathbf{m}\cdot\hat{\mathbf{x}})^{2}\Big], (3) where \mathbf{B}_{e} is the externally applied field, B_{a} is the easy-axis anisotropy field in the MTJ plane, and B_{d} is the effective demagnetizing field having contributions from dipolar field and perpendicular magnetic anisotropy (PMA). Please note that left side of ( 2 ) represents conservative precession, while the right side includes Gilbert damping torque and Slonczewski spin-transfer torque (STT). Equation ( 2 ) models the magnetization dynamics of the free layer and can be used to determine the macrospin vector \mathbf{m}(\theta,\phi) as a function of time, where \theta and \phi are the polar and azimuthal angles of \mathbf{m} (see Fig. 1 (a)). This dynamic behavior can be equivalently modeled by a nonlinear LC circuit by defining an effective capacitor charge Q and inductor flux \Phi Q=\frac{e}{\hbar}\frac{M_{s}V_{o}}{\gamma}\cos\theta,\hskip 11.38109pt\Phi=\frac{\hbar}{e}\phi. (4) The voltage V_{C} across a capacitor and the current I_{L} through an inductor can be related to the energy E of an LC circuit by V_{C}=\partialderivative{E}{Q},\hskip 11.38109ptI_{L}=\partialderivative{E}{\Phi}. (5) For an MTJ with an energy function ( 3 ), these definitions give V_{C}=\frac{\gamma\hbar}{e}\Big[B_{e,z}-\frac{(B_{e,x}\cos\phi+B_{e,y}\sin\phi)}{\tan\theta}-(B_{d}+B_{a}\cos^{2}\phi)\cos\theta\Big], (6) I_{L}=\frac{\gamma\hbar}{eZ_{s}}\Big[(B_{e,x}\sin\phi-B_{e,y}\cos\phi)\sin\theta+\frac{1}{2}B_{a}\sin(2\phi)\sin^{2}\theta\Big]. (7) Here, Z_{s}=\hbar^{2}\gamma/(eM_{s}V_{o})

Read the complete text