---
title: Magneto-Electric Circuit Approach
url: https://www.emergentmind.com/topics/magneto-electric-circuit-approach
type: topic
---

# Magneto-Electric Circuit Approach

Searching arXiv for recent and foundational papers on magneto-electric circuit approaches.
The magneto-electric circuit approach denotes a family of circuit-theoretic and device-level methodologies in which electric and magnetic variables are treated as coequal state variables, and in which magnetoelectric transduction is represented explicitly in the governing equations, equivalent circuits, or both. In contemporary arXiv literature, the expression covers several distinct but related programs: magnetic-oriented field–circuit coupling for magnetoquasistatic power devices, equivalent circuits for direct and inverse magnetoelectric read/write operations, physics-based spintronic logic models, probability-native mixed-signal nanocircuits, electromagnetic-field-based charge–flux circuit theory, and magnetic-circuit extensions for topological magnetoelectric media [2404.15438], [1710.10700], [2110.10890], [1504.04056], [2403.16025], [2304.11032]. A common feature is the elevation of charge, flux, polarization, magnetization, or magnetoelectric source terms to circuit-level primitives rather than treating them as secondary corrections.

## 1. Terminological scope and conceptual core

Across the cited literature, the term does not identify a single canonical formalism. Instead, it names a cluster of approaches that couple electrical and magnetic descriptions more tightly than conventional modified nodal analysis, standard equivalent-circuit reductions, or purely phenomenological compact models. Some works are formulated at the level of differential-algebraic equations for coupled field and network models; others operate at the device level through equivalent circuits and SPICE-compatible compact models; still others propose new circuit languages in which charge flow and flux flow are dual graphical primitives.

| Lineage | Core variables or mechanism | Representative papers |
|---|---|---|
| Magnetic-oriented field–circuit coupling | $\psi$, $\phi$, $q$, $\mathbf{a}$; exact power balance | [2404.15438] |
| Direct/inverse magnetoelectric equivalent circuits | $Q$, $m_x^2-m_y^2$, back-voltage $\partial E_m/\partial Q$ | [1710.10700] |
| Spintronic logic circuits | FE/AFM/FM dynamics, spin–orbit conversion, MTJ readout | [2110.10890], [1611.07738] |
| Probability-native nanocircuits | S-MTJ resistance states, mixed-signal probability arithmetic | [1504.04056] |
| Electromagnetic-field circuit theory | ECF/MFF diagrams, charge and flux conservation | [2403.16025] |
| Passive magnetic devices with ME media | extended Hopkinson’s law, Hall-current MMF terms | [2304.11032] |

The conceptual core is therefore not a unique device technology, but a style of modeling in which electromagnetic storage, dissipation, and interconversion are made explicit at circuit level. This suggests that the phrase is best understood as an umbrella category spanning energy-based coupling formalisms, equivalent-circuit abstractions, and magnetoelectric device architectures rather than a single standardized method.

## 2. Magnetic-oriented field–circuit coupling in the low-frequency regime

In low-frequency power-device modeling, the magneto-electric circuit approach appears as a magnetic-oriented alternative to electrically oriented MNA. Under magnetoquasistatic assumptions, displacement currents are neglected and the retained Maxwell relations are $\nabla\times\mathbf{H}=\mathbf{J}$ and $\nabla\cdot\mathbf{B}=0$, with linear constitutive laws $\mathbf{B}=\mu\mathbf{H}$ and $\mathbf{J}=\sigma\mathbf{E}+\mathbf{J}_s$. The magnetic-oriented formulation adopts an $A^\*$-gauge in which $\mathbf{e}=-\partial_t\mathbf{a}$ and $\mathbf{b}=\operatorname{curl}\mathbf{a}$, so that the eddy-current model becomes
$$
\sigma\,\partial_t \mathbf{a}+\operatorname{curl}\big(\nu\,\operatorname{curl}\mathbf{a}\big)=\mathbf{j}.
$$
After finite-element discretization, the device equations read
$$
M_\sigma\,\partial_t a+K_\nu\,a=X\,i_M,\qquad v_M=X^\top\,\partial_t a.
$$
The circuit side is parameterized not by node voltages and branch currents, but by magnetic node potentials $\psi$, branch fluxes $\phi$, and charges $q$, with $u=\partial_t\psi$, $v_\*=\partial_t\phi_\*$, $i_\*=\partial_t q_\*$, and $\phi_\*=A_\*^\top\psi$ [2404.15438].

The coupled field–circuit system is assembled by introducing terminal currents $i_M=\partial_t q_M$, terminal voltages $v_M=X^\top\partial_t a$, and an incidence matrix $A_M$ for the device ports. In compact notation, with $y:=(\psi,q_C,q_V,q_M,a)$, the model takes the geometric form
$$
C(\partial_t y)+\partial_y H(y)=f(y),
$$
where the stored energy for the linear case is
$$
H(\psi,q_C,a)=\tfrac{1}{2}\|A_L^\top\psi\|_{L^{-1}}^2+\tfrac{1}{2}\|q_C\|_{C^{-1}}^2+\tfrac{1}{2}\|a\|_{K_\nu}^2.
$$
The resulting continuous power identity is
$$
\frac{d}{dt}H(\psi,q_C,a)= -\|A_R^\top\partial_t\psi\|_G^2-\|\partial_t a\|_{M_\sigma}^2-\langle v_I,i_{src}\rangle-\langle v_{src},i_V\rangle.
$$
A central consequence is that resistive dissipation in the circuit, eddy-current losses in the device, and power exchange with sources are separated explicitly, while the coupling terms cancel exactly due to the interconnection structure.

This structure is preserved under implicit midpoint time discretization. For the rectifier benchmark with a transformer as the only energy storage element, the discrete identity
$$
d_\tau H^{n-1/2}= -\|d_\tau a^{n-1/2}\|_{M_\sigma}^2-\|A_R^\top d_\tau\psi^{n-1/2}\|_G^2-\langle v_{src}(t^{n-1/2}),i_V^{n-1/2}\rangle
$$
holds exactly for linear storage. The reported convergence data for a magnetic potential component $\psi_4$ are approximately second order: $\tau=0.0050\rightarrow\epsilon_\tau=0.0266$, $\tau=0.0025\rightarrow\epsilon_\tau=0.0078$ with e.o.c. $1.77$, $\tau=0.00125\rightarrow\epsilon_\tau=0.0021$ with e.o.c. $1.89$, and $\tau=0.000625\rightarrow\epsilon_\tau=0.0006$ with e.o.c. $1.79$. The discrete power-balance discrepancy $\epsilon_H$ is reported as $O(10^{-13})$ per step, consistent with the Newton tolerance $10^{-12}$. A recurring point in this line of work is that magnetic orientation is not merely a change of variables: it yields an explicit unified energy identity and typically reduced DAE index, often $\le 1$.

## 3. Equivalent circuits for direct and inverse magnetoelectric phenomena

A second major usage of the magneto-electric circuit approach is the construction of self-consistent equivalent circuits for write and read operations in magnetoelectric structures. In this setting, the magnetic state is not necessarily encoded by a net magnetization component. One formulation represents the bit by the pseudo-magnetization, or easy-axis indicator,
$$
q\equiv \mu \equiv (m_x^2-m_y^2),
$$
and couples it to circuit charge through a magnetic energy of the form
$$
E_m=(E_A+v_M Q)(m_x^2-m_y^2),
$$
or, for the MELRAM benchmark structure,
$$
E_m=-E_A m_xm_y+\frac{E_H}{\sqrt{2}}(m_x-m_y)+v_MQ(m_x^2-m_y^2).
$$
The inverse effect enters as a back-voltage
$$
V_{back}=\frac{\partial E_m}{\partial Q}=v_M(m_x^2-m_y^2)=v_M\mu,
$$
leading, for a capacitive load, to the circuit equation
$$
V_{IN}=\frac{Q}{C_L}+\frac{Q}{C}+v_M\mu,
$$
and, for a resistive load, to
$$
V_{IN}=R\frac{dQ}{dt}+\frac{Q}{C}+v_M\mu.
$$
The write threshold in the field-free easy-axis encoding mode is
$$
V_{th}=-\frac{E_A}{v_MC_{eff}},
$$
with $C_{eff}=(CC_L)/(C+C_L)$ [1710.10700].

Because the circuit and the stochastic LLG dynamics are solved self-consistently, the same framework supports deterministic write, tunable randomness, and non-volatile memory. In the randomness regime, with $E_A\to 0$ and modest ME coupling, the paper gives the Boltzmann average
$$
\langle \mu\rangle \approx -\frac{I_1(x)}{I_0(x)},\qquad x=\frac{Qv_M}{k_BT},
$$
while hysteresis and non-volatility appear when
$$
C_{eff}v_M^2>k_BT.
$$
The readout signal for a capacitive load is
$$
\Delta V_L=-\frac{C_{eff}}{C_L}v_M\mu=-\frac{C}{C+C_L}v_M\mu.
$$
A concrete example uses $C=C_L=50\,\mathrm{aF}$ and $v_M=10\,\mathrm{mV}$, giving $\Delta V_L\approx -0.5\times 10\,\mathrm{mV}\times\mu$.

This literature is notable for rejecting the assumption that magnetoelectric circuit design must rely on external magnetic fields or MTJs. The field-free mode is explicitly described as a “radical departure” from representing a bit by $m_x$, $m_y$, or $m_z$. The paper further states that the approach can be used to build a device exhibiting tunable randomness and suggests possibilities for extending it to non-volatile memory with read and write capabilities, without the use of external magnetic fields or magnetic tunnel junctions. The broader implication is that magnetoelectric circuit abstractions can encode anisotropy selection, not only magnetization reversal.

## 4. Spintronic logic realizations: MESO and MESL

In spintronic logic, the magneto-electric circuit approach takes the form of device-to-circuit models that combine magnetic dynamics, ferroelectric dynamics, and circuit transport. A physics-based MESO model comprises a multiferroic input stack and a spin–orbit output stack. On the input side, the FE polarization $\mathbf{P}$ evolves according to the Landau–Khalatnikov equation
$$
\Gamma_{FE}\frac{d\mathbf{P}}{dt}= - \frac{\partial}{\partial \mathbf{P}}\big(F_{bulk}+F_{elas}+F_{elec}+F_{me}\big),
$$
while FM and AFM macrospins follow LLG dynamics and interfacial exchange. On the output side, spin–charge interconversion in the SOC layer is represented by a $4\times 4$ block matrix with $6\times 6$ conductance submatrices, yielding a full $24\times 24$ relation between charge and spin currents and terminal voltages. Implemented in SPICE through modular FM, FM–NM, NM, and SO subcircuits, this model reproduces buffers, oscillators, and majority gates at $V_{DD}=100$–$200\,\mathrm{mV}$. Reported single-slice results include total vertical current $\approx 2.2\,\mu\mathrm{A}$ for $100\,\mathrm{mV}$ drive, $V_{cap}\approx 31\,\mathrm{mV}$ in $\approx 4\,\mathrm{ns}$, and $I_{cap}$ peaking at $\approx 0.2\,\mu\mathrm{A}$; five-slice SO structures raise the total current to $\approx 9\,\mu\mathrm{A}$ and $V_{cap}$ to $\approx 40\,\mathrm{mV}$ with saturation in $\sim 2\,\mathrm{ns}$. The majority gate requires $V_{DD}=200\,\mathrm{mV}$ for all $2^5=32$ initial conditions to succeed, and the optimum $V_{cap,\max}$ versus $\theta_{SH}$ occurs around $\theta_{SH}\approx 4$ [2110.10890].

MESL represents a related but distinct device family. It uses a four-terminal topology in which two nanomagnets are stacked vertically, each interfaced to its own ME oxide, and separated by an MgO spacer to form an MTJ. The write path is purely voltage driven through the ME oxide, while the read path senses MTJ resistance through a voltage divider and CMOS inverter; the paper emphasizes this as a decoupled write/read architecture. The effective ME field is modeled phenomenologically as
$$
\vec{H}_{ME}=\big(\alpha_{ME}(V_{ME}/t_{ME})\widehat{x},\,0\widehat{y},\,0\widehat{z}\big),
$$
inserted into the stochastic LLG equation together with demagnetization, interfacial anisotropy, and thermal fields. Gate-level constructions include XNOR, NAND, and NOR; cascadability is achieved with global reset and domino-style evaluation pulses. For the XNOR gate, the reported write energy is $E_{Swi~Total}\approx 5.5\,\mathrm{fJ}$, the read-out energy is $\approx 30\,\mathrm{fJ}$ for a $500\,\mathrm{ps}$ evaluation window, and the switching delay is $\approx 500\,\mathrm{ps}$. The paper also reports that ME switching shows a tighter distribution of switching times than STT in stochastic LLG simulations, and describes the logic family as having better robustness against thermal noise than conventional current-driven switching [1611.07738].

Taken together, these two spintronic lines define a magneto-electric circuit approach in which the central design problem is not only logical functionality but also faithful closure between materials parameters, multiphysics device dynamics, and circuit-level waveforms.

## 5. Probability-native magnetoelectric nanocircuits and charge–flux circuit theory

A third cluster of works uses magnetoelectric devices as circuit primitives for non-Boolean computation. In the S-MTJ probabilistic framework, probabilities are stored as spatially redundant vectors of straintronic magneto-tunneling junction states,
$$
\mathbf{p}=[p_1,p_2,\dots,p_n],\qquad p_i\in\{0,1,\dots,k-1\},
$$
with encoded scalar value
$$
P=\frac{\sum_{i=1}^{n}p_i}{n(k-1)}.
$$
A Probability Composer converts this redundant magnetic-domain representation into an analog current or voltage. In the binary case, a correction circuit yields
$$
I_{out}=\frac{V_{REF}}{\beta}\sum_{i=1}^{n}p_i=\frac{nV_{REF}}{\beta}P,
$$
while Elementary Arithmetic Composers implement addition and multiplication directly in the electrical domain:
$$
I_{out}\approx \frac{V_{REF}}{\beta}(P_A+P_B),\qquad
I_{out}\approx gV_{REF}P_AP_B.
$$
Hierarchical composition of these blocks gives self-similar Probability Arithmetic Composers suitable for Bayesian likelihood estimation and belief propagation. For that workload, the reported improvements relative to synthesized $45\,\mathrm{nm}$ CMOS baselines reach $79\times$ lower area than $4$-bit and $127\times$ lower area than $5$-bit multipliers, $142\times$ and $214\times$ lower active power, and, when external flash-memory reads are included, $\approx 70\times$ lower end-to-end latency, despite the Composer circuits being slower in pure computation by $\approx 288\times$ versus $4$-bit CMOS and $\approx 221.5\times$ versus $5$-bit CMOS [1504.04056].

At a more abstract level, electromagnetic-field-based circuit theory proposes dual electric-charge-flow and magnetic-flux-flow diagrammatics as a unified language for phase-independent and phase-dependent circuits. In the charge-based ECF representation, nodes are electric-charge containers with
$$
\mathbf{Q}_n=\mathbf{C}_n\mathbf{V}_n-\mathbf{Q}_e,
$$
pumps are electric-charge pumps with
$$
\mathbf{V}_{cp}=\mathbf{O}_n^{T}\mathbf{V}_n,
$$
and conservation is stated as
$$
\mathbf{Q}_n=\mathbf{Q}_{const}-\mathbf{O}_n\mathbf{Q}_{cp}.
$$
In the dual flux-based MFF representation,
$$
\boldsymbol{\Phi}=\mathbf{L}_m\mathbf{i}_m-\boldsymbol{\Delta}_e,\qquad
\mathbf{i}_{fp}=\mathbf{O}_m^T\mathbf{i}_m,\qquad
\boldsymbol{\Phi}=\boldsymbol{\Phi}_{const}-\mathbf{O}_m\boldsymbol{\Phi}_{fp}.
$$
The general device equations are written as $\mathbf{F}_{EIS}(\mathbf{V}_{EIS},\mathbf{Q}_{cp},t)=0$ for ECF and $\mathbf{F}_{EIP}(\mathbf{i}_{EIP},\boldsymbol{\Phi}_{fp},t)=0$ for MFF. The authors explicitly present CCL and FCL as replacements for KCL and KVL/FQL, and show how Josephson junctions and QPS junctions fit naturally into the pump formalism. This line of work broadens the meaning of magneto-electric circuit approach beyond material coupling alone, recasting circuit theory itself around the duality of charge transport and flux transport [2403.16025].

## 6. Magnetic circuits, extended Hopkinson’s law, and passive magnetoelectric devices

Passive magnetic devices provide another setting in which circuit reasoning is extended by magnetoelectric constitutive laws. In $\theta$-electrodynamics, the constitutive relations are
$$
\mathbf{D}=\varepsilon_r(\omega)\varepsilon_0\mathbf{E}+\chi\mathbf{B},\qquad
\mathbf{H}=\frac{1}{\mu_r\mu_0}\mathbf{B}-\chi\mathbf{E},
$$
with $\chi=(\theta/\pi)\alpha_0$. When $\theta$ is uniform in the bulk and varies only at interfaces, the surface Hall current density is
$$
\mathbf{J}_{Hall}=\frac{\alpha_0}{\pi}\nabla\theta\times\mathbf{E}.
$$
Under magnetoquasistatic conditions,
$$
\frac{r_2\omega_0}{c_i}\ll 1,\qquad r_2\mu_0\omega_0\chi\ll 1,
$$
this yields an extended Hopkinson’s law for a single-coil solenoid,
$$
\Phi\,\mathcal{R}=nI_f+I_{Hall,\varphi},
$$
and for a two-coil ideal transformer,
$$
\Phi\,\mathcal{R}=n_1I_f^{(1)}-n_2I_f^{(2)}+I_{Hall,\varphi}^{(1)}-I_{Hall,\varphi}^{(2)}.
$$
The induced voltages then acquire ME self and mutual inductance corrections, while the functionally passive part of the transformer, solenoid inductor, and solenoid actuator is stated to be indistinguishable from the conventional situation up to second order in the magnetoelectric susceptibility under low-power conditions. For bilayer NI–TI or TI–TI inductors with $\chi\approx 1000\,\alpha_0$, the reported inductance tunability exceeds $200\%$ up to $100\,\mathrm{GHz}$ at millimeter scale, whereas significantly strong ME response can restrict the operating frequency range below the ultra low frequency. The same paper argues that the principal benefit of topological insulator cores in the weak-response regime lies in lower power consumption because the bulk is insulating and eddy currents are strongly suppressed [2304.11032].

A related, non-topological magnetic-circuit methodology derives conventional transformer equivalent circuits entirely from reluctance, permeance, and complex permeability. In that framework,
$$
\mathcal{F}=NI,\qquad
\mathcal{R}=\frac{\ell}{\mu_0\mu_rA},\qquad
\Phi=\frac{\mathcal{F}}{\mathcal{R}}=P\mathcal{F},
$$
and the magnetizing inductance is
$$
L_m=N^2P=N^2\frac{\mu_0\mu_rA}{\ell}.
$$
Leakage inductances arise from air-path permeance, $L_\sigma=N^2P_\sigma$, while hysteresis and eddy-current losses enter through a complex reluctance $\mathcal{R}_c=a+j\beta$, giving
$$
R_c=\frac{N_p^2\omega\beta}{a^2+\beta^2},\qquad
X_m=\frac{N_p^2\omega a}{a^2+\beta^2}.
$$
This transformer literature does not treat explicit magnetoelectric coupling, but it supplies the magnetic-circuit baseline from which extended laws such as $\Phi\mathcal{R}=nI_f+I_{Hall,\varphi}$ can be viewed as generalizations rather than replacements [2103.17257].

## 7. Unifying principles, misconceptions, and open problems

Taken together, the cited works indicate that magneto-electric circuit approaches are united less by a single material platform than by a recurrent structural choice: the governing equations are organized around energy storage, dissipation, and interconversion between electrical and magnetic degrees of freedom. In some cases this appears as a Hamiltonian-plus-gradient DAE with exact continuous and discrete power balance; in others as a back-voltage source $\partial E_m/\partial Q$ in an equivalent circuit; in still others as coupled LK/LLG/SPICE modules, self-similar probability arithmetic, or dual ECF/MFF conservation laws. A plausible implication is that the most durable contribution of this literature is methodological: it relocates electromagnetic coupling from an implementation detail to the core circuit abstraction.

Several misconceptions are corrected by the primary sources. First, the approach is not synonymous with standard MNA plus a magnetic device block. The magnetic-oriented MONA formulation explicitly introduces $\psi$, $\phi$, and $q$, and its stated advantages include exact continuous power balance, preservation of energy structure under the midpoint rule, and typically reduced DAE index, often $\le 1$ [2404.15438]. Second, magnetoelectric circuit design is not restricted to encoding information in net magnetization or to reading it with MTJs; the easy-axis encoding based on $m_x^2-m_y^2$ was proposed precisely to avoid that restriction [1710.10700]. Third, stronger magnetoelectric response does not automatically imply stronger useful circuit functionality. In passive topological devices, the conventional behavior survives up to second order in susceptibility in the low-power regime, and the main benefit may instead be lower power consumption; conversely, in large-$\chi$ devices, tunability can increase while usable bandwidth shrinks [2304.11032].

The open problems are correspondingly heterogeneous. Power-device coupling still requires careful handling of gauging and boundary conditions, even though nonlinear energy-based models and more general coupling mechanisms are stated to be possible [2404.15438]. Mixed-signal S-MTJ architectures face analog precision, finite TMR, parasitic-capacitance, spacing, and fabrication-complexity constraints [1504.04056]. Physics-based MESO models explicitly omit thermal noise in the reported simulations, and their macrospin and quasi-static transport assumptions delimit predictive scope [2110.10890]. MESL relies on predictive ME parameters and does not analyze explicit ferroelectric dynamics or endurance [1611.07738]. Equivalent-circuit models based on $v_MQ(m_x^2-m_y^2)$ compress detailed electromechanical coupling into a single parameter that must be calibrated [1710.10700]. These limitations do not negate the approach; they delimit the domains in which a given magneto-electric circuit model should be interpreted as a faithful reduction rather than a convenient surrogate.

Source: https://www.emergentmind.com/topics/magneto-electric-circuit-approach