Quasi-Continuous Exhaust (QCE) in Tokamaks

• QCE is a distinct tokamak edge regime where high pedestal confinement is maintained without large Type-I ELMs, enabling quasi-steady heat and particle exhaust.
• It operates via localized ballooning modes and turbulence that generate small, high-frequency bursts, effectively regulating the pedestal transport.
• Simulations and experiments corroborate QCE’s reactor-relevant performance by showing favorable scaling with separatrix density, heat flux profiles, and normalized pedestal parameters.

Quasi-Continuous Exhaust (QCE) is a tokamak edge operational regime in which a standard H-mode-like pedestal is maintained without large Type-I edge-localized-mode (ELM) crashes, while heat and particle exhaust remain quasi-continuous because transport is concentrated near the separatrix and in the scrape-off layer (SOL) (Zhang et al., 3 Mar 2026, Dunne et al., 22 Apr 2026). In conventional H-modes, the edge pressure pedestal slowly grows until it crosses a peeling–ballooning boundary, triggering a Type-I ELM crash that dumps a large fraction of the pedestal energy in a short burst. In QCE, by contrast, high pedestal confinement, natural suppression of large Type-I ELMs, a continuous, broad SOL heat-flux decay length $\lambda_q$, and high separatrix density $n_{\rm sep}$ are realized simultaneously (Zhang et al., 3 Mar 2026). The literature describes this exhaust either as a spectrum of small, high-frequency edge bursts or, in more recent turbulence-resolved simulations, as self-limiting pedestal turbulence with quasi-coherent separatrix dynamics and blob-mediated SOL transport; both descriptions place ballooning-mode physics at the regime’s core (Harrer et al., 2021, Zhang et al., 3 Mar 2026).

1. Regime identity and phenomenology

QCE is not simply “mitigated” Type-I behavior but a distinct operational window. In advanced tokamak operation, it denotes a regime in which the usual large, low-frequency Type-I ELMs are replaced by a spectrum of small, high-frequency bursts that together carry particles and heat out of the plasma at an almost steady rate, while the pedestal at the edge of the plasma remains steep enough to sustain high confinement and high separatrix density (Harrer et al., 2021). In the more recent ASDEX Upgrade turbulence analysis, the same regime is characterized as one in which the pedestal limits its own growth via persistent turbulence, avoiding large intermittent bursts and producing a near-steady exhaust of heat and particles (Zhang et al., 3 Mar 2026).

Operationally, QCE combines high pedestal confinement, H-mode-like steep gradients, natural suppression of large Type-I ELMs, continuous broad SOL exhaust, and elevated $n_{\rm sep}$ leading toward detached divertor conditions (Zhang et al., 3 Mar 2026). In ASDEX Upgrade, access requires strong plasma shaping plus elevated $n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$ at the separatrix (Zhang et al., 3 Mar 2026). Earlier small-ELM descriptions report pedestal performance with $H_{98,y2}>1$ and $f_{GW,ped}\gtrsim0.87$, with small-ELM frequencies $f_{\rm ELM}\sim1$–$2\;{\rm kHz}$ rather than the $30$–$50\;{\rm Hz}$ typical of Type-I ELMs, peak ELM-induced divertor heat flux reduced by roughly a factor two over the QCE cycle, and a wetted divertor footprint broader by $n_{\rm sep}$0–$n_{\rm sep}$1 (Harrer et al., 2021).

A recurrent misconception is that QCE is merely a weaker form of the standard peeling–ballooning crash. The cited literature instead treats it as a regime in which the instability is narrowly localized near the last closed flux surface (LCFS) or separatrix, so that the edge transport barrier is “perforated” by small events or by persistent turbulence at the pedestal foot rather than catastrophically expelled across the full pedestal width (Harrer et al., 2021, Zhang et al., 3 Mar 2026).

2. Ideal-MHD access and separatrix stability

The basic access picture is an edge-localized ballooning instability. Near the LCFS, the competition between local pressure gradient and magnetic shear sets stability, with the high-$n_{\rm sep}$2 ideal-ballooning criterion written as

$n_{\rm sep}$3

and local magnetic shear

$n_{\rm sep}$4

In QCE plasmas, $n_{\rm sep}$5 just inside the separatrix locally exceeds $n_{\rm sep}$6 where $n_{\rm sep}$7 is small, triggering narrowly localized ballooning-mode bursts whose rapid repetition produces the small ELMs or equivalent quasi-continuous exhaust events (Harrer et al., 2021).

Plasma shaping enters through the field-line connection length between the good-curvature high-field side and the bad-curvature low-field side. A longer $n_{\rm sep}$8 reduces local shear stabilization at the low-field-side midplane and is accompanied experimentally by deeper $n_{\rm sep}$9 minima and stronger $n_{\rm sep}$0 shear,

$n_{\rm sep}$1

HELENA calculations organize this into a marginality factor,

$n_{\rm sep}$2

In QCE discharges, $n_{\rm sep}$3 at $n_{\rm sep}$4 just inside the separatrix, while $n_{\rm sep}$5 is recovered deeper in the steep gradient region around $n_{\rm sep}$6, where second ballooning stability correlates with improved pedestal confinement (Harrer et al., 2021).

A complementary cross-machine access model expresses QCE onset as a minimum separatrix density. Equating the edge drive to the ideal-MHD threshold yields

$n_{\rm sep}$7

For typical ASDEX Upgrade geometry and power-input scans,

$n_{\rm sep}$8

Measured normalized separatrix densities in JET and ASDEX Upgrade QCE discharges lie above the predicted threshold, around $n_{\rm sep}$9–$n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$0, whereas matched ELMy discharges lie below; in TCV the same scaling substantially over-predicts the threshold, indicating uncertainties in $n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$1 or $n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$2 extrapolations there (Dunne et al., 22 Apr 2026).

3. Quasi-coherent mode, electromagnetic self-organization, and radial coherence

Global two-fluid simulations with GRILLIX identify a quasi-coherent mode (QCM) straddling the separatrix, with narrow spectrum $n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$3 and propagation in the ion-diamagnetic direction (Zhang et al., 3 Mar 2026). Linear analysis at $n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$4 yields two local branches: a kinetic ballooning mode (KBM),

$n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$5

and a weaker resistive branch. The dominant QCM branch follows $n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$6 and is destabilized when the normalized ballooning parameter

$n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$7

approaches the ideal-ballooning threshold $n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$8 (Zhang et al., 3 Mar 2026).

The QCM is not described as a purely local mode. Its distinctive feature is an extended radial correlation length generated by electromagnetic self-organization of turbulence. Two mechanisms are singled out. First, Maxwell stress,

$n_{\rm sep}\approx4\times10^{19}\,\rm m^{-3}$9

transfers turbulent energy into zonal flows and shapes a broad $H_{98,y2}>1$0 well that pins the QCM between pedestal and SOL. Consistently, removing $H_{98,y2}>1$1 deepens the $H_{98,y2}>1$2 well, kills the QCM, and restores an ELMy-like pedestal. Second, finite-Larmor-radius (FLR) physics enters through the full polarization drift term

$H_{98,y2}>1$3

in the vorticity equation, shifting the KBM cross-phase toward $H_{98,y2}>1$4, i.e. anti-correlation of $H_{98,y2}>1$5 and $H_{98,y2}>1$6, and thereby allowing large-amplitude yet coherent transport events (Zhang et al., 3 Mar 2026).

The ballooning eigenvalue problem makes this FLR dependence explicit,

$H_{98,y2}>1$7

where the $H_{98,y2}>1$8 term comes solely from the FLR polarization drift. When that term is disabled, $H_{98,y2}>1$9 collapses from $f_{GW,ped}\gtrsim0.87$0 to $f_{GW,ped}\gtrsim0.87$1, $f_{GW,ped}\gtrsim0.87$2 transport spikes by an order of magnitude, and the pedestal is destroyed (Zhang et al., 3 Mar 2026).

A key mesoscopic signature is the radial correlation length $f_{GW,ped}\gtrsim0.87$3. In QCE, the QCM exhibits $f_{GW,ped}\gtrsim0.87$4 mm, comparable to or exceeding the local pressure-gradient length $f_{GW,ped}\gtrsim0.87$5, whereas inter-ELM KBM has $f_{GW,ped}\gtrsim0.87$6 mm. This longer $f_{GW,ped}\gtrsim0.87$7 enables mesoscopic “streamer” events that sweep the pedestal foot across the separatrix (Zhang et al., 3 Mar 2026).

4. Secondary X-point dynamics, ballistic blobs, and SOL profile decoupling

The recent turbulence-resolved account of QCE adds a second ingredient beyond the QCM itself: a resistive X-point mode (RXM). Near the X-point, magnetic diffusion grows as $f_{GW,ped}\gtrsim0.87$8, and a resistive branch with positive cross-correlation $f_{GW,ped}\gtrsim0.87$9 appears (Zhang et al., 3 Mar 2026). The point where $f_{\rm ELM}\sim1$0 marks an interchange layer. In the outer QCM region, $f_{\rm ELM}\sim1$1, this interchange injects positive density perturbations into the SOL via the $f_{\rm ELM}\sim1$2 flow, launching coherent “ballistic” blobs (Zhang et al., 3 Mar 2026).

The blob properties extracted from the simulations are explicitly quantified. The observed radial speed is

$f_{\rm ELM}\sim1$3

matching ASDEX Upgrade measurements reported by M. Griener et al. The perpendicular size is $f_{\rm ELM}\sim1$4 cm, the parallel length is $f_{\rm ELM}\sim1$5 m, and the ejection frequency is set by the QCM amplitude oscillation period $f_{\rm ELM}\sim1$6–$f_{\rm ELM}\sim1$7 ms (Zhang et al., 3 Mar 2026).

This mechanism explains how QCE can retain high pedestal confinement while broadening SOL exhaust. On the confined side, the QCM maintains a steep pedestal gradient with

$f_{\rm ELM}\sim1$8

near the separatrix. In the SOL, however, the electron-temperature fall-off is characterized by a distinct secondary decay length,

$f_{\rm ELM}\sim1$9

The time-averaged profile is therefore piecewise exponential,

$2\;{\rm kHz}$0

According to the simulation, this decoupling arises because QCM-driven pedestal reciprocation extends into the SOL only in the outermost layer, whereas blobs dominate heat transport further out (Zhang et al., 3 Mar 2026).

5. Numerical and experimental corroboration

The strongest direct validation currently reported is the GRILLIX simulation of ASDEX Upgrade discharge #36165 in QCE phase. It reproduces mean $2\;{\rm kHz}$1, $2\;{\rm kHz}$2, and $2\;{\rm kHz}$3 profiles within experimental uncertainties; pedestal-foot oscillation across the separatrix at the QCM frequency, matching coherent helium-beam and Thomson measurements; fluctuation amplitude and skewness with negative skew inside and positive skew in the SOL, including clear $2\;{\rm kHz}$4 blob thresholds; poloidal mode number $2\;{\rm kHz}$5 and real frequency $2\;{\rm kHz}$6; cross-phase $2\;{\rm kHz}$7 near the outer midplane, transitioning toward $2\;{\rm kHz}$8 at the blob birth radius; radial correlation lengths $2\;{\rm kHz}$9 mm for QCE KBM versus $30$0 mm for inter-ELM KBM; and blob radial velocity $30$1 with ejection statistics consistent with fast camera and probe data (Zhang et al., 3 Mar 2026).

Nonlinear JOREK simulations provide a complementary resistive-MHD description. With sufficient triangularity and an imposed $30$2 shear profile derived from experimental $30$3 measurements, the code develops broadband, medium-$30$4 ($30$5–$30$6) ballooning-like filaments that cycle quasi-periodically rather than producing a single giant crash. If $30$7 shear is artificially turned off, the code either fails to cycle or produces unrealistically violent turbulence after each ELM. Increasing the input heating power deepens the $30$8 well and restores a conventional Type-I ELM cycle, reproducing the experimental transition out of QCE (Harrer et al., 2021).

BOUT++ simulations of EAST show the same qualitative transition under a separatrix-density scan. Starting from a Type-I ELM baseline at $30$9, the simulated ELM size is $50\;{\rm Hz}$0; at $50\;{\rm Hz}$1, $50\;{\rm Hz}$2–$50\;{\rm Hz}$3; and at $50\;{\rm Hz}$4, $50\;{\rm Hz}$5, consistent with QCE-like small ELMs. The dominant toroidal mode number shifts from $50\;{\rm Hz}$6 at baseline to $50\;{\rm Hz}$7 and $50\;{\rm Hz}$8 at high separatrix density, and coherent density filaments detach and propagate into the SOL. The threshold for the Type-I to QCE-like transition is reported as $50\;{\rm Hz}$9–$n_{\rm sep}$00, with both linear and nonlinear outer modes localizing where $n_{\rm sep}$01 (Tang et al., 2022).

Taken together, these studies support a layered interpretation: ideal-MHD and reduced-MHD calculations capture the access boundary and narrow ballooning-unstable layer, while global fluid turbulence simulations resolve the self-organized QCM, the RXM-triggered ballistic blobs, and the resulting SOL profile decoupling. This suggests that the different modeling frameworks are emphasizing different scales of the same operational regime.

6. Cross-machine performance and reactor relevance

Across EUROfusion devices, QCE is treated as one member of a broader class of Type-I-ELM-free regimes, but its distinguishing feature is that the pedestal-top performance, when appropriately normalized, does not differ significantly from ELMy H-mode plasmas (Dunne et al., 22 Apr 2026). To compare pedestal performance across machines, the literature introduces a normalized poloidal-beta at the pedestal top,

$n_{\rm sep}$02

together with

$n_{\rm sep}$03

so that lines of constant $n_{\rm sep}$04 appear as straight isobars in the $n_{\rm sep}$05 plane. In JET, ASDEX Upgrade, and TCV, QCE points fall on the same $n_{\rm sep}$06 bands as ELMy data; in JET the overlap is near-perfect, while ASDEX Upgrade and TCV QCE operate at slightly higher $n_{\rm sep}$07 but in the same $n_{\rm sep}$08 range, $n_{\rm sep}$09–$n_{\rm sep}$10 (Dunne et al., 22 Apr 2026).

The regime is also presented as reactor-relevant on dimensional grounds. In the instability zone beneath the separatrix, ASDEX Upgrade QCE discharges exhibit

$n_{\rm sep}$11

closely matching the range expected at ITER- or DEMO-scale pedestals, $n_{\rm sep}$12–$n_{\rm sep}$13 and $n_{\rm sep}$14–$n_{\rm sep}$15 (Harrer et al., 2021). This similarity is used to argue that the same localized ballooning-unstable layer that yields QCE in present machines should persist in reactor-grade devices (Harrer et al., 2021).

For ITER 15 MA baseline conditions, the ideal-MHD threshold model predicts

$n_{\rm sep}$16

with ITER’s normalized pedestal-top point lying on the same $n_{\rm sep}$17 contours already accessed in QCE. Integrated modeling with IMEP further indicates that sustaining $n_{\rm sep}$18 at high shaping and $n_{\rm sep}$19 yields the desired pedestal without large ELMs (Dunne et al., 22 Apr 2026).

A more compact extrapolation framework is the separatrix operational space (SepOS), which maps $n_{\rm sep}$20 and uses a turbulence parameter

$n_{\rm sep}$21

together with a critical ballooning drive

$n_{\rm sep}$22

In this framework, QCE corresponds to $n_{\rm sep}$23. A predictive SPARC operating point is identified at

$n_{\rm sep}$24

which lies in the QCE region of the normalized SepOS (Eich et al., 2024).

The principal open issues are already delineated in the literature: scaling of filament heat loads to reactor-major radii, interplay with divertor detachment, and real-time control of the QCE foot mode (Dunne et al., 22 Apr 2026). Within those limits, QCE is presented as an operational paradigm in which separatrix-localized ballooning transport, rather than pedestal-wide crashes, regulates the edge and provides high confinement with continuous, tolerable exhaust.