Mean-Field Dynamo: Theory & Mechanisms

• Mean-field dynamo is a framework that describes the generation of large-scale magnetic fields using averaged induction equations and turbulent transport coefficients.
• It employs methods like the test-field technique to derive scale- and frequency-dependent response functions, revealing nonlocality and memory effects.
• The framework encompasses classical α²/αΩ models alongside mechanisms such as negative eddy diffusivity, delayed transport, and fluctuating helicity effects.

Searching arXiv for recent and foundational papers on mean-field dynamos, including nonlocality, test-field methods, and representative mechanisms. Mean-field dynamo denotes the generation or maintenance of magnetic fields on scales larger than the underlying flow or turbulence by means of a closure for the mean electromotive force, $$\boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}$$, in the averaged induction equation. In its standard formulation, the magnetic and velocity fields are decomposed into mean and fluctuating parts, and the mean induction equation contains transport coefficients—most prominently $\alpha$ and turbulent magnetic diffusivity—that encode the effect of fluctuations on the mean field (Rädler, 2014). Modern work has made two points especially clear. First, mean-field dynamos are not restricted to the classical $\alpha^2$ and $\alpha\Omega$ paradigms; negative eddy diffusivity, delayed transport, incoherent helicity effects, and tensorial off-diagonal transport can also drive large-scale field growth (Devlen et al., 2012, Rheinhardt et al., 2014, Jingade et al., 2021). Second, the relationship between $\boldsymbol{\mathcal{E}}$ and the mean field is in general neither local nor instantaneous, so transport coefficients must often be treated as scale- and frequency-dependent response functions rather than constants (Rheinhardt et al., 2011, Brandenburg et al., 2018).

1. Mean-field formulation and closure problem

Mean-field electrodynamics begins with the decomposition

$$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$

followed by averaging of the induction equation to obtain

$$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$

This structure is common across kinematic DNS, spherical shell dynamos, shearing boxes, and nonlinear solar-cycle models (Devlen et al., 2012, Schrinner, 2011, Gressel et al., 2015, Safiullin et al., 2017).

Under ideal scale separation, $\boldsymbol{\mathcal{E}}$ is expanded in the mean field and its first derivatives,

$$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$

or, in isotropic notation,

$$\boldsymbol{\mathcal{E}}=\alpha\,\overline{\boldsymbol{B}}-\eta_t\,\mu_0\overline{\boldsymbol{J}}.$$

In this representation, $\alpha$0 is the inductive part associated in the simplest closures with helicity, while $\alpha$1 is turbulent magnetic diffusivity, usually positive (Devlen et al., 2012, Rädler, 2014). The effective diffusivity is then $\alpha$2 (Devlen et al., 2012).

A central modern refinement is that this closure is only approximate. In general, $\alpha$3 is a convolution in space and time,

$\alpha$4

or, in Fourier space,

$\alpha$5

so the relevant transport objects are kernels or response tensors depending on wavenumber and frequency (Devlen et al., 2012, Rheinhardt et al., 2011, Brandenburg et al., 2018). This shifts the closure problem from determining a few constants to determining $\alpha$6- and $\alpha$7-dependent transport functions.

2. Canonical dynamo classes and their generalizations

The classical mean-field taxonomy distinguishes $\alpha$8 and $\alpha$9 dynamos. In homogeneous helical turbulence with no strong shear, the $\alpha^2$0 mechanism can sustain a large-scale field through mutual regeneration of field components by the $\alpha^2$1-effect (Rädler, 2014). In differentially rotating systems, the $\alpha^2$2 dynamo combines shear amplification of toroidal field with $\alpha^2$3-driven regeneration of poloidal field; this remains the leading interpretation of large-scale magnetic cycles in stratified MRI turbulence and in many solar and stellar models (Gressel et al., 2015, Safiullin et al., 2017).

The canonical $\alpha^2$4 picture is quantitatively supported in vertically stratified, isothermal MRI shearing boxes. There, the relevant $\alpha^2$5-component scales linearly with shear, the turbulent diffusivity shows analogous scaling, and the measured coefficients reproduce the butterfly-cycle period through the standard $\alpha^2$6 dispersion relation (Gressel et al., 2015). In the solar context, nonlinear $\alpha^2$7 models augmented by dynamic helicity evolution and algebraic quenching generate cyclic solutions, grand-minimum-like episodes, and latitude–time activity patterns (Safiullin et al., 2017).

However, the modern literature emphasizes that mean-field dynamos are not exhausted by helicity-driven $\alpha^2$8-effects. Several additional mechanisms have been identified.

Negative eddy diffusivity: Roberts flow IV realizes a large-scale dynamo with $\alpha^2$9 and $\alpha\Omega$0, so that $\alpha\Omega$1 above a critical $\alpha\Omega$2 (Devlen et al., 2012).

Delayed transport: Roberts flows II and III produce large-scale dynamos even though turbulent diffusivity remains positive; the instability arises because off-diagonal transport coefficients are frequency dependent, so the EMF depends on the mean field at earlier times (Rheinhardt et al., 2014).

Fluctuating-helicity dynamos: In shearing flows, temporal fluctuations of kinetic helicity with zero mean can drive a large-scale dynamo when the helicity correlation time exceeds the velocity correlation time, even when turbulent diffusion remains positive (Jingade et al., 2021).

Diffusivity-fluctuation dynamos: Spatiotemporal fluctuations of turbulent kinetic energy, equivalently of turbulent diffusivity, can produce growth of a large-scale field even when kinetic helicity vanishes pointwise, especially when turbulent diamagnetism is included (Gopalakrishnan et al., 2023).

Rotation–shear–current and rotation–shear–vorticity effects: In MRI-driven turbulence treated via direct statistical simulation, the EMF contains explicit tensorial contributions associated with off-diagonal turbulent resistivity and large-scale vorticity, providing non-$\alpha\Omega$3 routes to large-scale field generation (Mondal et al., 2023).

This suggests that “mean-field dynamo” is best understood as a framework for reduced induction physics rather than as a synonym for the $\alpha\Omega$4-effect alone.

3. Determination of transport coefficients: the test-field method and DNS calibration

A decisive methodological development is the test-field method, which computes transport coefficients directly from DNS rather than imposing them a priori (Schrinner, 2011, Rheinhardt et al., 2011, Devlen et al., 2012). The method prescribes mean test fields, solves auxiliary induction equations for the associated fluctuations, computes the induced EMFs, and inverts the resulting linear system for $\alpha\Omega$5 and $\alpha\Omega$6 or their spectral counterparts (Devlen et al., 2012, Gressel et al., 2015).

In Roberts-IV flow, the test-field method shows that all $\alpha\Omega$7 components vanish within numerical accuracy, while the turbulent diffusivity tensor is isotropic in the horizontal plane: $\alpha\Omega$8 The mean-field growth observed in DNS is then quantitatively explained by the sign change of $\alpha\Omega$9 with $\boldsymbol{\mathcal{E}}$0 and by its frequency dependence (Devlen et al., 2012).

In fast-rotating global spherical-shell dynamos, the test-field method has been used to compute the tensors $\boldsymbol{\mathcal{E}}$1 and $\boldsymbol{\mathcal{E}}$2 entering

$\boldsymbol{\mathcal{E}}$3

from fully nonlinear DNS. For low Rossby number, chaotically time-dependent dynamos, the time- and azimuthally averaged EMF is sufficiently well parameterized by these coefficients, and the corresponding mean-field eigenmodes match the DNS mean field substantially better than in quasi-stationary cases (Schrinner, 2011). A notable conclusion is that chaotic time dependence can improve scale separation and thereby improve the mean-field description (Schrinner, 2011).

In stratified MRI turbulence, the quasi-kinematic test-field method yields full $\boldsymbol{\mathcal{E}}$4-dependent and $\boldsymbol{\mathcal{E}}$5-dependent $\boldsymbol{\mathcal{E}}$6 and $\boldsymbol{\mathcal{E}}$7, including off-diagonal components. The resulting closure coefficients capture vertical structure, anisotropy, and nonlocality, and provide a scale-separation ratio of about ten for the large-scale dynamo (Gressel et al., 2015).

These results support a broader implication: transport coefficients should be inferred from the underlying dynamics wherever possible. A plausible implication is that mean-field models relying on assumed positivity, isotropy, or locality of $\boldsymbol{\mathcal{E}}$8 risk missing operative dynamo branches (Devlen et al., 2012, Gressel et al., 2015, Mondal et al., 2023).

4. Nonlocality, memory, and delayed transport

Spatio-temporal nonlocality has become a central theme in mean-field dynamo theory. When scale separation in space and time is poor, the EMF kernel is better approximated by a response function than by local coefficients (Rheinhardt et al., 2011, Brandenburg et al., 2018). A widely used approximation is

$\boldsymbol{\mathcal{E}}$9

which in real space corresponds to an evolution equation for the EMF,

$$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$0

This converts the nonlocal closure into a PDE system tractable in spherical geometry and nonlinear settings (Rheinhardt et al., 2011, Brandenburg et al., 2018).

In oscillatory dynamos, nonlocality alters excitation conditions. For oscillatory $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$1-shear dynamos and equatorially antisymmetric $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$2 dynamos, the critical dynamo number is lowered by spatio-temporal nonlocality (Rheinhardt et al., 2011). In spherical mean-field models with strong radial variation of $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$3 and $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$4, nonlocality smooths sharp structures near the base of the convection zone, lowers the critical dynamo number, and lengthens the cycle period, while leaving surface patterns comparatively unchanged (Brandenburg et al., 2018).

The “delayed transport” dynamo is a more specific manifestation of memory. For Roberts flows II and III, DNS and test-field analysis show that the turbulent magnetic diffusivity remains positive, but large-scale growth still occurs because off-diagonal transport coefficients depend on frequency (Rheinhardt et al., 2014). In flow II,

$$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$5

with vanishing diagonal $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$6-components, whereas in flow III the relevant off-diagonal term behaves like a pumping coefficient $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$7 (Rheinhardt et al., 2014). In the instantaneous limit these terms would yield decay or pure oscillation, but once their $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$8-dependence is inserted self-consistently into the dispersion relation, the zero mean field becomes unstable (Rheinhardt et al., 2014). This establishes delayed transport as a distinct mean-field dynamo mechanism.

A related but separate memory effect occurs in negative-diffusivity dynamos. In Roberts-IV, comparison between DNS growth rates and test-field predictions requires evaluation of $$\boldsymbol{B}=\overline{\boldsymbol{B}}+\boldsymbol{b},\qquad \boldsymbol{U}=\overline{\boldsymbol{U}}+\boldsymbol{u},$$9, not merely $$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$0, and this frequency dependence explains oscillatory decay and growth near the stability boundary (Devlen et al., 2012).

5. Representative mechanisms beyond the classical $$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$1-effect

Several recent or revived mechanisms broaden the operational landscape of mean-field dynamos.

For Roberts-IV, the dynamo is a clean example of a pure negative-diffusivity mechanism. The mean field has the form

$$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$2

with $$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$3 and $$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$4 components growing independently, unlike in conventional $$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$5 dynamos (Devlen et al., 2012). The instability exists only at sufficiently large scales because $$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$6 becomes positive at larger $$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$7, stabilizing short-scale mean fields (Devlen et al., 2012).

For MRI turbulence, the unified second-order cumulant treatment identifies two explicit non-$$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$8 effects. The rotation–shear–current effect yields an off-diagonal turbulent resistivity

$$\frac{\partial \overline{\boldsymbol{B}}}{\partial t} = \nabla\times \left( \overline{\boldsymbol{U}}\times\overline{\boldsymbol{B}} + \boldsymbol{\mathcal{E}} - \eta\,\overline{\boldsymbol{J}} \right), \qquad \boldsymbol{\mathcal{E}}=\overline{\boldsymbol{u}\times\boldsymbol{b}}.$$9

which is negative for Keplerian rotation and therefore favorable for large-scale dynamo action (Mondal et al., 2023). The rotation–shear–vorticity effect arises from coupling between off-diagonal Faraday tensors and large-scale vorticity gradients, providing a distinct route to vertical-field generation (Mondal et al., 2023).

For helicity fluctuations in shearing flows, growth of the mean field occurs when $\boldsymbol{\mathcal{E}}$0, with $\boldsymbol{\mathcal{E}}$1, even if turbulent diffusion remains positive (Jingade et al., 2021). In this regime, the fastest-growing modes are axisymmetric and occur at wavenumbers smaller than the eddy wavenumber, and both the growth rate and the optimal wavenumber show a non-monotonic dependence on $\boldsymbol{\mathcal{E}}$2 (Jingade et al., 2021). This differs from Kraichnan’s negative-diffusion picture: shear plus finite helicity-memory is sufficient.

For turbulent-kinetic-energy fluctuations, a double-averaging procedure shows that $\boldsymbol{\mathcal{E}}$3-fluctuations and turbulent diamagnetism can generate mean-field growth even when kinetic helicity is zero pointwise (Gopalakrishnan et al., 2023). In the white-noise limit, these fluctuations reduce effective turbulent diffusion while generating a drift term that does not affect growth; with finite correlation time, growing solutions exist even with statistically isotropic $\boldsymbol{\mathcal{E}}$4-fluctuations (Gopalakrishnan et al., 2023).

Taken together, these cases undermine the misconception that large-scale mean-field dynamos require either nonzero mean helicity or negative total diffusivity.

6. Nonlinearity, saturation, and applications

Nonlinearity enters mean-field dynamos through field-dependent transport coefficients, magnetic-helicity constraints, and, in some settings, coupling to mean flows. A standard decomposition writes

$\boldsymbol{\mathcal{E}}$5

with $\boldsymbol{\mathcal{E}}$6 tied to kinetic helicity and $\boldsymbol{\mathcal{E}}$7 to current helicity or small-scale magnetic helicity (Rädler, 2014, Safiullin et al., 2017). In solar-cycle models, the magnetic contribution is evolved with a helicity-balance equation, while algebraic quenching suppresses both kinetic and magnetic parts of $\boldsymbol{\mathcal{E}}$8 at large field strength (Safiullin et al., 2017). This produces a nonlinear $\boldsymbol{\mathcal{E}}$9 system capable of cycle modulation and grand-minimum-like behavior (Safiullin et al., 2017).

In a different direction, semi-analytic work has shown that helical MHD turbulence can generate a mean flow $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$0 through small-scale Lorentz-force correlations, and that this self-generated flow feeds back on the EMF through terms proportional to small-scale cross-helicity and large-scale vorticity (0704.2217). In that framework, the canonical $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$1 and $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$2 terms remain formally unmodified, while the mean flow contributes additional EMF terms. The effect is stronger for large relaxation times in the minimal $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$3 approximation (0704.2217).

Applications of mean-field dynamos span rotating convection, accretion disks, stellar cycles, and laboratory-inspired flows. In rigidly rotating stratified convection without mean shear, DNS–mean-field comparison shows that a nonuniform $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$4-effect is sufficient to drive an oscillatory $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$5 dynamo wave with cyclic polarity reversals and spatiotemporal migration (Masada et al., 2014). In accretion disks, measured $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$6 and $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$7 support an $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$8 interpretation of the butterfly diagram and provide practical closure data for global disk models (Gressel et al., 2015). In solar applications, a nonlinear mean-field dynamo has been coupled to a Wolf-number budget based on negative effective magnetic pressure instability, linking large-scale dynamo fields to sunspot production (Safiullin et al., 2017).

There are also important negative results. Earlier claims that a modified Taylor–Green flow realizes a negative-eddy-diffusivity dynamo were not confirmed: DNS and test-field calculations show no growing large-scale field and $$\mathcal{E}_i=\alpha_{ij}\,\overline{B}_j-\eta_{ij}\,\overline{J}_j+\dots,$$9 in the modified Taylor–Green setups tested (Devlen et al., 2012). This suggests that identifying a mean-field mechanism requires simultaneous evidence from DNS morphology and transport-coefficient diagnostics, not growth alone.

7. Conceptual status and current outlook

Mean-field dynamo theory has evolved from a local $$\boldsymbol{\mathcal{E}}=\alpha\,\overline{\boldsymbol{B}}-\eta_t\,\mu_0\overline{\boldsymbol{J}}.$$0–$$\boldsymbol{\mathcal{E}}=\alpha\,\overline{\boldsymbol{B}}-\eta_t\,\mu_0\overline{\boldsymbol{J}}.$$1 closure into a framework centered on transport kernels, tensor anisotropy, and multi-scale response (Rädler, 2014, Rheinhardt et al., 2011). One persistent misconception is that failures of simplistic local closures imply failure of the mean-field concept itself. The literature surveyed here points in the opposite direction: when nonlocality, memory, tensor structure, and measured coefficients are included, mean-field descriptions often recover DNS behavior with considerable fidelity (Schrinner, 2011, Rheinhardt et al., 2011, Gressel et al., 2015).

Several themes now define the field. One is the move from prescribed coefficients to coefficient inference via the test-field method and related statistical closures (Devlen et al., 2012, Gressel et al., 2015, Mondal et al., 2023). Another is the recognition that non-$$\boldsymbol{\mathcal{E}}=\alpha\,\overline{\boldsymbol{B}}-\eta_t\,\mu_0\overline{\boldsymbol{J}}.$$2 mechanisms can dominate in settings as different as laminar Roberts flows, MRI turbulence, and models with helicity or diffusivity fluctuations (Rheinhardt et al., 2014, Mondal et al., 2023, Jingade et al., 2021, Gopalakrishnan et al., 2023). A third is that nonlinear saturation remains incompletely understood. Reviews emphasize that there is still no convincing comprehensive theory of $$\boldsymbol{\mathcal{E}}=\alpha\,\overline{\boldsymbol{B}}-\eta_t\,\mu_0\overline{\boldsymbol{J}}.$$3-quenching and mean-field dynamo saturation, despite major progress from magnetic-helicity-based approaches (Rädler, 2014).

A plausible synthesis is that “mean-field dynamo” now denotes a hierarchy of reduced models in which the large-scale induction problem is closed by transport coefficients or kernels extracted from the underlying flow. Within that hierarchy, classical $$\boldsymbol{\mathcal{E}}=\alpha\,\overline{\boldsymbol{B}}-\eta_t\,\mu_0\overline{\boldsymbol{J}}.$$4 and $$\boldsymbol{\mathcal{E}}=\alpha\,\overline{\boldsymbol{B}}-\eta_t\,\mu_0\overline{\boldsymbol{J}}.$$5 dynamos remain fundamental, but they are part of a broader family that includes negative eddy diffusivity, delayed transport, incoherent fluctuation dynamos, and tensorial shear- and vorticity-mediated effects (Devlen et al., 2012, Rheinhardt et al., 2014, Mondal et al., 2023).