Micro-tearing Modes in Fusion Plasmas

• Micro-tearing Modes are small-scale electromagnetic instabilities marked by tearing parity, localized current layers, and a transport signature dominated by electron heat flux.
• They are driven by electron temperature gradients with growth rates sensitive to collisionality and magnetic drifts, affecting confinement in a variety of magnetic geometries.
• Gyrokinetic theory and simulations identify MTMs through parity diagnostics, electron-diamagnetic propagation, and the observation of magnetic island overlap leading to stochastic transport.

Micro-tearing modes (MTMs) are small-scale electromagnetic tearing or drift-tearing microinstabilities with perpendicular scales of order a few ion gyroradii, localized current layers near mode-rational surfaces, and a transport signature dominated by electron heat flux. Across the literature considered here, they are consistently identified as modes driven primarily by the electron temperature gradient, with magnetic reconnection, magnetic flutter, and—when islands overlap—field-line stochasticity providing the operative transport pathway. Although MTMs were long associated especially with spherical tokamaks and pedestal physics, the available studies place them in a much broader class of regimes, including reversed-field-pinch internal transport barriers, conventional tokamak cores, pedestal-top plasmas, negative-triangularity scenarios, and a high-density-gradient Wendelstein 7-X discharge in which MTM turbulence dominates microtransport (Predebon et al., 2010, Applegate et al., 2011, Dickinson et al., 2012, Cu-Castillo et al., 22 Oct 2025).

1. Physical character and defining signatures

MTMs are electromagnetic instabilities with tearing parity in the fluctuating fields. The recurring eigenfunction signature is an even $A_\parallel$ and an odd $\phi$ across the resonant layer or ballooning coordinate, together with a narrow parallel-current sheet and reconnecting magnetic perturbations. In multiple studies, this structure is directly connected to the formation of microscopic magnetic islands at rational surfaces; when those islands overlap, the perturbed field becomes stochastic and electrons stream along perturbed field lines, enhancing radial heat transport (Predebon et al., 2010, Applegate et al., 2011, Chandran et al., 2022).

Their spatial scale is “micro” in the sense of ion-scale binormal wavelength rather than electron-scale turbulence. In the reversed-field-pinch and tokamak studies, the unstable spectrum lies at $k_y \rho_i$ or $k_\theta \rho_s$ of order unity or below, so MTMs are distinct from ETG turbulence even when both coexist in the same discharge (Predebon et al., 2010, Moradi et al., 2013, Giacomin et al., 2023). In the 2022 collisionless toroidal treatment, $\delta \hat A_\parallel$ is localized near $\theta \sim O(1)$, whereas $\delta \hat \Phi$ extends out to $|\theta| \sim (k_\wedge \rho_e)^{-1}$, which formalizes the usual picture of a tearing-like current layer embedded in a broader drift-wave-like electrostatic response (Chandran et al., 2022).

The propagation signature is electron-diamagnetic. In GENE and GS2 conventions used in several studies, this appears as a negative real frequency, whereas the gyro study of NSTX, ASDEX-UG, and JET uses the opposite sign convention and states that positive $\omega_r$ corresponds to the electron diamagnetic direction (Cu-Castillo et al., 22 Oct 2025, Balestri et al., 27 Jul 2025, Moradi et al., 2013). This sign-convention dependence is important: the physically relevant marker is electron-diamagnetic propagation, not the algebraic sign by itself.

2. Drive mechanisms and stability controls

The common free-energy source is the electron temperature gradient $\nabla T_e$. In RFX-mod, the growth rate rises strongly with the normalized temperature gradient and exhibits a stability threshold around $\phi$0; in NSTX it increases monotonically with $\phi$1; and in the W7-X discharge the MTM growth rate is strongly enhanced by larger $\phi$2 (Predebon et al., 2010, Moradi et al., 2013, Cu-Castillo et al., 22 Oct 2025). At the same time, the dependence on $\phi$3 is not universal: ASDEX-UG and JET show weaker and non-monotonic behavior, and MAST exhibits a branch change around $\phi$4 (Applegate et al., 2011, Moradi et al., 2013).

Collisionality is a second major control parameter, but its role depends on regime. Several core studies recover the familiar MTM result that finite collisionality is needed or strongly favorable: RFX-mod finds growth maximal when $\phi$5 is of order tens; ASDEX-UG and JET show non-monotonic dependence with no unstable MTM in the collisionless limit; and W7-X identifies a non-monotonic collisionality dependence as a classic MTM fingerprint (Predebon et al., 2010, Moradi et al., 2013, Cu-Castillo et al., 22 Oct 2025). By contrast, the pedestal-top study in MAST finds that the dominant edge MTM does not peak at finite collisionality and can remain strongly unstable at $\phi$6, which the authors interpret as a collisionless trapped-particle mechanism sensitive to magnetic drifts (Dickinson et al., 2012). The collisionless low-$\phi$7 toroidal dispersion-relation work reaches a related conclusion from a different direction: MTMs can be unstable in collisionless toroidal plasmas, and the electron-temperature-gradient drive is mediated by magnetic drifts (Chandran et al., 2022).

Magnetic drifts emerge repeatedly as a destabilizing ingredient. The detailed MAST analysis concludes that the instability is not substantially affected by the classic Hazeltine thermal-force or Catto–Rosenbluth trapped-particle boundary-layer mechanisms, but is strongly destabilized by magnetic drifts and by the electrostatic potential $\phi$8 (Applegate et al., 2011). The negative-triangularity study likewise identifies faster poloidally averaged magnetic drift velocity $\phi$9 as the core reason why MTMs are stronger in negative than in positive triangularity (Balestri et al., 27 Jul 2025). The pedestal-top work sharpens the point further by showing that the edge MTM is stable when magnetic drifts are turned off, and that trapped-particle drifts matter much more than passing-particle drifts in that regime (Dickinson et al., 2012).

Magnetic shear does not play a single universal role. In the W7-X study, MTMs are described as stabilized by magnetic shear $k_y \rho_i$0 through magnetic field-line bending, and very low magnetic shear helps facilitate onset (Cu-Castillo et al., 22 Oct 2025). In the negative-triangularity survey, larger $k_y \rho_i$1 helps push negative-triangularity cases into an MTM-dominated regime, summarized by the onset region $k_y \rho_i$2 (Balestri et al., 27 Jul 2025). In nonlinear MAST simulations, higher $k_y \rho_i$3 mainly amplifies transport by reducing rational-surface spacing and making island overlap easier, so the influence of shear there is strongly tied to saturation physics rather than merely linear onset (Giacomin et al., 2023). A plausible implication is that “magnetic shear” enters MTM physics through more than one channel: field-line bending, rational-surface packing, and drift geometry need not favor the same trend in every configuration.

The density gradient is similarly nuanced. In the W7-X case, the steep density gradient is important mainly because it suppresses competing ion-scale modes, especially ITG turbulence, while the MTM growth rate itself is only weakly sensitive to $k_y \rho_i$4 and even slightly reduced as $k_y \rho_i$5 increases (Cu-Castillo et al., 22 Oct 2025). The MAST study also describes the mode as only weakly dependent on density gradient (Applegate et al., 2011). The negative-triangularity work reports that doubling the logarithmic density gradient in DIII-D strongly reduces or removes MTMs in negative triangularity, leaving only higher-$k_y \rho_i$6 TEM-like destabilization (Balestri et al., 27 Jul 2025). These results collectively support a restricted statement: MTMs in these studies are not density-gradient-driven in the sense characteristic of $k_y \rho_i$7-driven TEM or universal-instability branches.

3. Identification in gyrokinetic theory and simulation

MTM identification is built from a recurring set of linear diagnostics. The most common combination is tearing parity, electron-diamagnetic propagation, and an electromagnetic transport signature dominated by electron heat flux. The triangularity study makes this explicit with a criterion table: MTMs have $k_y \rho_i$8, $k_y \rho_i$9 in the GENE convention, $k_\theta \rho_s$0, and $k_\theta \rho_s$1, whereas TEMs and ITG modes have electrostatic heat-flux dominance and $k_\theta \rho_s$2 (Balestri et al., 27 Jul 2025). In W7-X, the dominant branch over $k_\theta \rho_s$3 has negative real frequency, tearing parity, electron-dominated growth-rate contributions, and non-monotonic dependence on collisionality, while its $k_\theta \rho_s$4 lies close to the electron diamagnetic reference frequency used in the paper (Cu-Castillo et al., 22 Oct 2025).

Mode structure provides an independent confirmation. In MAST, Poincaré plots of perturbed field lines and current contours show magnetic islands at rational surfaces, with a current minimum at the island center, a current maximum at the X-point, and a narrow current layer around the resonant surface (Applegate et al., 2011). The 2022 low-$k_\theta \rho_s$5 asymptotic theory for axisymmetric toroidal geometry and the 2022 collisionless toroidal dispersion-relation study both formalize the same picture: an electron current layer localized to the rational surface is matched to a larger-scale electromagnetic region, and the passing-electron response acquires a discontinuity across the layer (Hardman et al., 2022, Chandran et al., 2022).

Several studies also stress that parity alone is not sufficient. The spherical-tokamak surrogate-model work therefore imposes physics-based filters: an accepted MTM must propagate in the electron diamagnetic direction, have tearing parity, and either lie within a prescribed frequency window around the analytic estimate or produce field-line deviation from the equilibrium flux surface, depending on the dataset construction (Hornsby et al., 2023, Hornsby et al., 2024). This is methodologically significant because active-learning or reduced-order models trained only on “odd/even parity” labels would mix MTMs with other branches.

4. Manifestations across confinement concepts

The available studies span a notably broad device set and show that MTMs are not confined to a single magnetic geometry or confinement region.

Configuration and regime	Reported MTM behavior	Paper
RFX-mod SHAx ITB	Dominant ion-scale turbulent mechanism in strong-$k_\theta \rho_s$6 barrier region	(Predebon et al., 2010)
MAST core	Linearly dominant ion-scale mode; prevalence linked to high $k_\theta \rho_s$7 and strong drifts	(Applegate et al., 2011)
MAST pedestal top	Distinct edge MTM with collisionless trapped-particle drift drive	(Dickinson et al., 2012)
NSTX, ASDEX-UG, JET cores	MTMs dominant linearly for experimentally relevant parameters, with machine-dependent $k_\theta \rho_s$8 response	(Moradi et al., 2013)
Negative/positive triangularity tokamaks and DEMO	NT more susceptible to MTMs; STs closer to MTM threshold than conventional tokamaks	(Balestri et al., 27 Jul 2025)
Wendelstein 7-X high-density-gradient core	First W7-X regime where MTM turbulence dominates microtransport	(Cu-Castillo et al., 22 Oct 2025)

In RFX-mod, the SHAx regime creates an internal transport barrier with strong electron temperature gradients, moderate $k_\theta \rho_s$9 of about $\delta \hat A_\parallel$0–$\delta \hat A_\parallel$1, and finite $\delta \hat A_\parallel$2 and collisionality, producing conditions favorable for MTMs inside a region where earlier discussions had focused more on large-scale MHD turbulence (Predebon et al., 2010). In the three-device tokamak comparison, MTMs are dominant over a broad $\delta \hat A_\parallel$3 range in NSTX and JET-like parameters, but over a narrower low-$\delta \hat A_\parallel$4 interval in ASDEX-UG, where ITG and ETG take over at higher $\delta \hat A_\parallel$5 (Moradi et al., 2013).

The pedestal-top regime is qualitatively different. Near $\delta \hat A_\parallel$6 in MAST, the dominant MTM peaks near $\delta \hat A_\parallel$7, high magnetic shear and large inverse aspect ratio amplify trapped-particle drift effects, and the instability remains strong as $\delta \hat A_\parallel$8, unlike familiar core MTM behavior (Dickinson et al., 2012). This directly motivates the later analytical effort to describe low-$\delta \hat A_\parallel$9, passing-electron-driven electromagnetic microinstabilities localized near mode-rational surfaces and to represent shaping through an effective $\theta \sim O(1)$0 (Hardman et al., 2022).

The most recent extension is the W7-X high-density-gradient case. At $\theta \sim O(1)$1, the representative parameters are $\theta \sim O(1)$2, $\theta \sim O(1)$3, $\theta \sim O(1)$4, $\theta \sim O(1)$5, $\theta \sim O(1)$6, and $\theta \sim O(1)$7, and MTMs remain unstable over $\theta \sim O(1)$8 (Cu-Castillo et al., 22 Oct 2025). The paper’s distinctive point is not merely that MTMs exist in a stellarator, but that the quasi-isodynamic, nearly max-$\theta \sim O(1)$9 geometry suppresses collisionless $\delta \hat \Phi$0-driven TEMs, so the steep density-gradient plasma does not transition into the TEM-dominated state often expected from tokamak intuition. This suggests that magnetic optimization can act as an instability-spectrum filter rather than as a universal turbulence suppressor.

5. Nonlinear transport, stochasticity, and experimental consistency

The transport consequence most consistently associated with MTMs is electromagnetic electron heat transport. In W7-X, the nonlinear flux spectrum is dominated by electromagnetic electron heat transport, the simulations recover the hallmark inequality $\delta \hat \Phi$1, and the calculated fluxes $\delta \hat \Phi$2 and $\delta \hat \Phi$3 are in good agreement with the measured values $\delta \hat \Phi$4 and $\delta \hat \Phi$5 (Cu-Castillo et al., 22 Oct 2025). The paper further notes that increasing $\delta \hat \Phi$6 by $\delta \hat \Phi$7 does not substantially change the nonlinear heat flux, consistent with the quasilinear scaling $\delta \hat \Phi$8 and with the interpretation that these MTMs are controlled mainly by $\delta \hat \Phi$9 and collisionality rather than by density gradient.

In RFX-mod, the nonlinear story is represented through quasi-linear stochastic transport. The Chirikov-overlap argument gives a threshold roughly $|\theta| \sim (k_\wedge \rho_e)^{-1}$0, whereas the dominant spectral peaks correspond to roughly $|\theta| \sim (k_\wedge \rho_e)^{-1}$1, so the mode spectrum satisfies the overlap condition over a substantial range of modes. With the Rechester–Rosenbluth model and a numerically estimated correlation length $|\theta| \sim (k_\wedge \rho_e)^{-1}$2, the paper obtains $|\theta| \sim (k_\wedge \rho_e)^{-1}$3, overlapping the experimental estimate $|\theta| \sim (k_\wedge \rho_e)^{-1}$4 (Predebon et al., 2010).

The dedicated nonlinear MAST study shows more sharply that linear instability is not enough. At $|\theta| \sim (k_\wedge \rho_e)^{-1}$5 with $|\theta| \sim (k_\wedge \rho_e)^{-1}$6, MTMs are linearly unstable but $|\theta| \sim (k_\wedge \rho_e)^{-1}$7 is negligible at the reference collisionality; ETG-driven heat flux is comparable to the experimental electron heat flux instead. At $|\theta| \sim (k_\wedge \rho_e)^{-1}$8 with $|\theta| \sim (k_\wedge \rho_e)^{-1}$9, $\omega_r$0 becomes significantly larger and comparable to $\omega_r$1 and $\omega_r$2 (Giacomin et al., 2023). The paper attributes this difference to island overlap and stochastic-layer formation: higher shear reduces resonant-surface spacing, making $\omega_r$3 much larger for similar fluctuation amplitude. It also identifies zonal $\omega_r$4 as a significant saturation mechanism, since removing zonal magnetic contributions leads to a substantial increase in heat flux.

The triangularity study reaches a related conclusion in a shaping context. In SMART, once $\omega_r$5 reaches about $\omega_r$6, negative-triangularity transport becomes much worse than positive-triangularity transport because MTMs dominate; the heat flux roughly doubles relative to positive triangularity, and the field-line diffusion coefficient rises by more than two orders of magnitude above the electrostatic regime (Balestri et al., 27 Jul 2025). In that sense, MTM turbulence is not merely a linear-stability label but a nonlinear state characterized by magnetic stochasticity, flutter-dominated electron heat transport, and weak particle-to-heat flux ratio.

6. Analytical theory, reduced models, and control-oriented interpretation

Two 2022 theoretical developments sharpen the analytical description of MTMs in toroidal geometry. The low-$\omega_r$7 asymptotic theory for axisymmetric toroidal plasmas models electron-driven electromagnetic microinstabilities with current layers localized near rational surfaces and binormal wavelengths $\omega_r$8, deriving orbit-averaged inner-layer equations together with a ballooning-space matching condition for passing electrons (Hardman et al., 2022). Its central construct is an effective $\omega_r$9 that incorporates shaping through a geometric factor $\nabla T_e$0, and the comparison with GS2 for MAST discharge #6252 shows that this effective $\nabla T_e$1 explains the dependence of local MTM growth rate on the ballooning parameter $\nabla T_e$2. The practical implication stated in the paper is that flux-surface shaping may be optimized to reduce MTM-driven transport.

The collisionless gyrokinetic dispersion-relation study solves the linearized gyrokinetic equation, quasineutrality condition, and Ampère’s law to produce a rapidly evaluated MTM dispersion relation in low-$\nabla T_e$3 toroidal plasmas (Chandran et al., 2022). Consistent with earlier simulations, it finds that MTMs are driven by the electron temperature gradient and that the drive is mediated by magnetic drifts; in the limit $\nabla T_e$4, the low-$\nabla T_e$5 mode is stable, confirming that density gradient alone is not the instability source in that regime. The paper also identifies a practical upper-wavenumber condition $\nabla T_e$6 for instability.

A separate line of work addresses MTMs as a reduced-order modeling problem. The spherical-tokamak surrogate papers build Gaussian-process classifiers and regressors for MTM growth rate $\nabla T_e$7, mode frequency $\nabla T_e$8, and quasi-linear electron heat flux across a 7D input space $\nabla T_e$9, using active learning to concentrate GS2 simulations near the unstable manifold (Hornsby et al., 2023, Hornsby et al., 2024). The reported workflow begins from a 300-point Latin hypercube sample, then adds batches with hit rates of $\phi$00 and $\phi$01, and the full model is built from about 5000 data points requiring roughly 1 million CPU-hours (Hornsby et al., 2024). The later extension argues that a single global GP plateaus in fidelity and is under-confident near marginal stability because multiple MTM sub-types may make the underlying mapping less smooth than expected; a clustering-based mixture-of-experts construction improves MSLL and uncertainty calibration (Hornsby et al., 2024). This suggests that MTM phenomenology is not always well represented by a single smooth branch, a point that resonates with the multi-branch behavior already seen in core-versus-edge and collisionless-versus-collisional studies.

Taken together, these theoretical and reduced-order results recast MTMs as a problem of mode selection and geometry-sensitive electromagnetic coupling rather than as a single textbook instability. The accumulated evidence supports a broad but specific picture: MTMs require electromagnetic coupling and electron-temperature-gradient free energy; their accessibility depends strongly on magnetic drifts, collisionality regime, competing instabilities, and geometric selectors such as shear, shaping, trapped-particle content, and max-$\phi$02 optimization; and their practical consequence is an electron-dominated transport channel whose nonlinear strength is set by magnetic-island overlap and stochastic-layer formation rather than by linear growth rate alone.