Pedestal Relaxation Events in I-Mode Plasmas

• PREs are small, transient edge crashes in I-mode plasmas that induce a modest drop in stored energy and electron temperature at the pedestal.
• Experimental data across devices reveal PREs occur near the I–H transition with characteristic frequencies (50–600 Hz) and predominantly conductive energy losses.
• Recent simulations and measurements identify microtearing modes and density-gradient driven instabilities as triggers, linking PREs to broader pedestal-relaxation phenomena with reactor implications.

Pedestal Relaxation Events (PREs) are small, transient edge crashes observed primarily in I-mode plasmas close to the I–H transition. In Alcator C-Mod and ASDEX Upgrade, they relax the edge pedestal, transiently enhance the energy reaching divertor target plates, and produce ELM-like scrape-off-layer signatures while remaining distinct from type-I ELMs in amplitude, triggering, and stability properties (Silvagni et al., 2021, Silvagni et al., 2020). The contemporary literature places PREs within a broader class of pedestal-limited relaxation phenomena that includes the pedestal burst instabilities (PBIs) reported during the I-mode to H-mode transition in EAST and impurity-driven soft pedestal relaxations in DIII-D, while recent simulations of ASDEX Upgrade directly connect PRE triggering to microtearing modes (MTMs) (Zhong et al., 2021, Banerjee et al., 1 Jun 2026, Pitzal et al., 26 Sep 2025).

1. Definition, regime, and phenomenological scope

In the comparative C-Mod and ASDEX Upgrade study, PREs are described as small, transient edge crashes that occur only in a subset of I-mode discharges, mainly when the plasma is close to the H-mode transition (Silvagni et al., 2021). Their characteristic global signature is a modest decrease in stored energy, with $\Delta W/W_{\mathrm{MHD}} \sim 0.5\text{–}3\%$, accompanied by a sharp edge electron-temperature drop, a transient divertor heat pulse, and filamentary energy transport into the scrape-off layer (Silvagni et al., 2021). In ASDEX Upgrade, PREs were operationally identified as intermittent events in I-mode that cause macroscopic relaxation of edge electron temperature and density profiles, with ELM-like divertor signatures but with the pedestal far from the ideal peeling–ballooning boundary (Silvagni et al., 2020).

The temporal scales are also well constrained. In C-Mod, PRE frequency spans $\sim 50\text{–}600$ Hz, whereas in ASDEX Upgrade it is 100–400 Hz; the crash is fast, of order $O(100~\mu\mathrm{s})$, and recovery occurs on millisecond scales (Silvagni et al., 2021). These events are therefore neither steady background turbulence nor large type-I ELM crashes. They occupy an intermediate category: edge-localized, impulsive, but comparatively mild.

The literature also broadens the phenomenological scope of PREs beyond the original I-mode observations. In EAST, quasi-periodic pedestal burst instabilities during the I-mode to H-mode transition are explicitly interpreted as pedestal relaxation–type events: they are pedestal-localized, repetitive, and regulate the build-up of edge gradients before H-mode onset (Zhong et al., 2021). In DIII-D, a boron-injection study does not use the term PRE explicitly, but documents impurity-driven, turbulence-mediated soft pedestal relaxations that progressively reduce type-I ELM frequency and can sustain long ELM-free intervals (Banerjee et al., 1 Jun 2026). This suggests that “PRE” denotes both a specific experimental object in I-mode and a wider pedestal-relaxation behavior class.

2. Pedestal localization and profile evolution

The experimental picture is strongly localized to the edge pedestal. In ASDEX Upgrade, integrated data analysis based on ECE, TS, DCN, and Li-BES shows that the largest PRE-induced drops occur near the electron-temperature pedestal top at $\rho_{\rm pol}=0.94$: $\Delta n_e/n_e \approx 7\%$, $\Delta T_e/T_e \approx 12\%$, and $\Delta p_e/p_e \approx 18\%$ (Silvagni et al., 2020). Full radial profiles show a density pivot point at $\rho_{\rm pol}\approx 0.99$, with density inside the separatrix decreasing and SOL density increasing after the event, while the maximum pressure collapse is centered near $\rho_{\rm pol}\approx 0.95$ and the affected depth is about $\Delta\rho_{\rm pol}\sim 0.2$ (Silvagni et al., 2020). Recovery is asymmetric: $\sim 50\text{–}600$0 returns to its pre-PRE value within about 2.5 ms, whereas $\sim 50\text{–}600$1 recovers more slowly (Silvagni et al., 2020).

The cross-device study emphasizes that PREs primarily relax the temperature pedestal. In C-Mod and ASDEX Upgrade, ECE measurements at $\sim 50\text{–}600$2 show a sharp drop in $\sim 50\text{–}600$3, typically

$\sim 50\text{–}600$4

with the drop increasing at lower pedestal collisionality (Silvagni et al., 2021). Density changes are weaker in that formulation, so PREs are often treated as predominantly conductive edge relaxations.

In EAST, the analogous PBI cycle is resolved in greater temporal detail. Before each burst, the local density gradient near the pedestal top increases sharply; at burst onset, turbulence is excited, the density profile relaxes, and an outward particle flux is induced; after the burst, the pedestal gradient rebuilds until the next event (Zhong et al., 2021). Statistical analysis of 32 PBIs gives a relative density perturbation at the pedestal top of about $\sim 50\text{–}600$5, with a representative event showing $\sim 50\text{–}600$6 (Zhong et al., 2021). Doppler reflectometry, bolometry, SXR, ECE, Mirnov probes, D$\sim 50\text{–}600$7, and divertor Langmuir probes all place the event in the pedestal region, concentrated around the pedestal top (Zhong et al., 2021).

A simulation-based view from GRILLIX is complementary. In the ASDEX Upgrade I-mode discharge modeled in that work, simulated PREs appear as three intermittent bursts within a $\sim 50\text{–}600$8 ms window; each event has duration $\sim 50\text{–}600$9 ms, produces strong magnetic activity, and causes a marked relaxation of the electron-temperature gradient near the mode location while leaving density largely unchanged (Pitzal et al., 26 Sep 2025). That profile selectivity is consistent with the experimentally dominant role of the electron-temperature pedestal in PREs.

3. Operational space, thresholds, and collisionality dependence

PREs are not generic to all I-mode discharges. In the multi-device database, they appear in about 20% of I-mode discharges, specifically 79/375 in C-Mod and 26/109 in ASDEX Upgrade, and cluster just below an empirical H-mode access line (Silvagni et al., 2021). A proxy for the H-mode threshold is written as

$O(100~\mu\mathrm{s})$0

and I-mode discharges with PREs lie just below this line (Silvagni et al., 2021). In C-Mod, no PREs were observed at $O(100~\mu\mathrm{s})$1 T in unfavorable configuration, where no H-mode transitions occurred because external heating was below the I–H threshold (Silvagni et al., 2021). In ASDEX Upgrade, PREs occupy a narrow band near the I–H boundary, around the 4 kPa isobar in edge electron pressure and at $O(100~\mu\mathrm{s})$2, with no additional confinement improvement once PREs appear (Silvagni et al., 2020).

Pedestal collisionality organizes the PRE amplitude. The pedestal-top collisionality is defined as

$O(100~\mu\mathrm{s})$3

and the C-Mod/AUG database shows that both $O(100~\mu\mathrm{s})$4 and $O(100~\mu\mathrm{s})$5 increase as $O(100~\mu\mathrm{s})$6 decreases (Silvagni et al., 2021). At DEMO/ARC-like collisionalities, the relative pedestal-top temperature drop reaches $O(100~\mu\mathrm{s})$7 and the relative energy loss reaches $O(100~\mu\mathrm{s})$8 (Silvagni et al., 2021). The scaling is therefore qualitatively similar to type-I ELM behavior, though at smaller amplitude.

The EAST PBI study provides a complementary threshold description in terms of normalized pedestal density gradient. The first PBI appears at

$O(100~\mu\mathrm{s})$9

while the last PBI before H-mode occurs at

$\rho_{\rm pol}=0.94$0

and the value at actual H-mode onset is slightly higher than that triggering the last PBI (Zhong et al., 2021). PBI frequency is $\rho_{\rm pol}=0.94$1 kHz and is likely to be inversely proportional to the chord-averaged density and the loss power (Zhong et al., 2021). In that device, a prompt local increase of $\rho_{\rm pol}=0.94$2 at the pedestal top is therefore a direct precursor.

4. Triggering instabilities and microphysical interpretations

A central result of the PRE literature is that PREs are not type-I ELMs. In ASDEX Upgrade, ideal-MHD stability analysis with MISHKA places I-mode PRE operating points far below the ideal peeling–ballooning boundary, whereas H-mode type-I ELM points lie close to it (Silvagni et al., 2020). The C-Mod/AUG comparison reaches the same conclusion and instead identifies an electromagnetic precursor just inside the separatrix, propagating in the electron diamagnetic drift direction with high toroidal mode number, $\rho_{\rm pol}=0.94$3 (Silvagni et al., 2021). In AUG, the precursor frequency is $\rho_{\rm pol}=0.94$4 kHz, similar to the weakly coherent mode (WCM) frequency there; in C-Mod it is $\rho_{\rm pol}=0.94$5 kHz, with the same propagation direction and a similar mode number to the WCM (Silvagni et al., 2021). The proposed interpretation is that PREs grow out of an I-mode-specific edge fluctuation, rather than from classical peeling–ballooning growth (Silvagni et al., 2021).

The most explicit trigger identification comes from the GRILLIX simulations of ASDEX Upgrade I-mode. There, PREs are attributed to microtearing modes. The underlying mode is localized near $\rho_{\rm pol}=0.94$6, peaks near the outboard midplane, propagates in the electron diamagnetic direction, and has $\rho_{\rm pol}=0.94$7 (Pitzal et al., 26 Sep 2025). Its parity is canonical tearing parity, with even $\rho_{\rm pol}=0.94$8 and odd $\rho_{\rm pol}=0.94$9, and the observed frequency matches the analytical MTM dispersion relation

$\Delta n_e/n_e \approx 7\%$0

with

$\Delta n_e/n_e \approx 7\%$1

(Pitzal et al., 26 Sep 2025). The nonlinear simulation trajectory in $\Delta n_e/n_e \approx 7\%$2 space crosses from linearly stable to linearly unstable MTM regions before each burst and returns to stability after the burst, providing a threshold-relaxation cycle rather than an ideal-MHD crash (Pitzal et al., 26 Sep 2025).

A broader microstability framework emerges from DIII-D and MAST. In the DIII-D pedestal, MTMs and KBMs are the main ion-scale instabilities, but KBMs are often second-stable in the mid-pedestal while MTMs exhibit threshold behavior near pre-ELM pressure gradients and can constrain pressure rather than only electron temperature (Hatch et al., 24 Mar 2026). In MAST, local gyrokinetic analysis finds KBMs in the steep pedestal region and tearing-parity modes or MTMs just inside the pedestal top, with the unstable KBM/ballooning region broadening during the inter-ELM cycle (Dickinson et al., 2011, Dickinson et al., 2011, Dickinson et al., 2012). This suggests, as an inference, that PREs may arise from different pedestal-limiting branches in different regimes: WCM-linked electromagnetic edge modes near the separatrix, MTMs at the pedestal top or mid-pedestal, and density-gradient-driven edge modes in I–H transition phases.

5. Transport channels, SOL response, and divertor loading

PRE transport is mixed conductive and convective, but the conductive component is especially prominent. The energy decomposition used in the C-Mod/AUG study is

$\Delta n_e/n_e \approx 7\%$3

so the relative temperature drop at the pedestal top directly tracks the conductive contribution (Silvagni et al., 2021). In the ASDEX Upgrade PRE reconstruction, the profile-based estimate gives $\Delta n_e/n_e \approx 7\%$4, corresponding to $\Delta n_e/n_e \approx 7\%$5 and $\Delta n_e/n_e \approx 7\%$6, with conductive and convective contributions of about $\Delta n_e/n_e \approx 7\%$7 kJ and $\Delta n_e/n_e \approx 7\%$8 kJ, respectively (Silvagni et al., 2020). In EAST PBIs, by contrast, the dominant observable is a density relaxation with induced outward particle flux and only a modest relative energy loss of about 1% of stored energy (Zhong et al., 2021).

The scrape-off-layer response is distinctly ELM-like. In ASDEX Upgrade, PRE deposition time at the divertor is $\Delta n_e/n_e \approx 7\%$9 ms, helium-beam measurements show radial filament velocities of about $\Delta T_e/T_e \approx 12\%$0 km/s, and divertor heat-flux patterns imply a quasi mode number between 10 and 28 with mean $\Delta T_e/T_e \approx 12\%$1 (Silvagni et al., 2020). More energy reaches the outer divertor target, with

$\Delta T_e/T_e \approx 12\%$2

in upper single-null unfavorable drift configuration (Silvagni et al., 2020). In the cross-device study, the PRE parallel energy fluence is bounded from above by the Eich type-I ELM model and from below by roughly one third of that model (Silvagni et al., 2021).

That fluence comparison carries direct reactor implications. Using the type-I ELM model as an upper boundary and one third of it as a lower boundary, the projected peak parallel energy fluence is

$\Delta T_e/T_e \approx 12\%$3

for a DEMO I-mode scenario and

$\Delta T_e/T_e \approx 12\%$4

for ARC (Silvagni et al., 2021). With an ITER-like limit of $\Delta T_e/T_e \approx 12\%$5, unmitigated PREs would exceed nominal divertor material limits under those assumptions (Silvagni et al., 2021).

The DIII-D boron-injection work shows a different transport redistribution. There, impurity-driven turbulence increases inter-ELM transport power,

$\Delta T_e/T_e \approx 12\%$6

from about 0.18 MW in the reference case to about 0.43 MW in the high-B case, while ELM power decreases from about 0.81 MW to about 0.56 MW (Banerjee et al., 1 Jun 2026). The dominant transport channel is low-frequency IDD turbulence correlated with the divertor D$\Delta T_e/T_e \approx 12\%$7 baseline, indicating continuous convective particle exhaust that can replace part of the impulsive ELM loss channel (Banerjee et al., 1 Jun 2026). In PRE language, this is a turbulence-mediated soft-relaxation pathway.

6. Relation to PBIs, ELMs, and broader pedestal-relaxation regimes

PREs are most sharply distinguished from type-I ELMs by pedestal stability and loss amplitude. In C-Mod and ASDEX Upgrade, PREs occur while the I-mode pedestal remains ideal peeling–ballooning stable, and their relative energy loss is typically much smaller than the 3–10% associated with type-I ELMs (Silvagni et al., 2021, Silvagni et al., 2020). Yet the SOL and divertor response—filaments, asymmetric energy deposition, and ELM-like parallel transport waveforms—closely resembles ELM phenomenology once the energy has entered the SOL (Silvagni et al., 2020). PREs are therefore edge-relaxation events with ELM-like exhaust physics but a different pedestal trigger.

The EAST PBIs extend this category toward density-gradient-driven pedestal relaxation. Their first and last thresholds in $\Delta T_e/T_e \approx 12\%$8, the 6–8% density perturbation, the outward particle flux, and the localization at the pedestal top all support the description of PBIs as a particular, density-gradient-driven realization of PREs that regulates the edge during the I-mode to H-mode transition (Zhong et al., 2021). By contrast, the ASDEX Upgrade GRILLIX simulations emphasize an MTM-driven, electron-temperature-gradient-controlled PRE cycle (Pitzal et al., 26 Sep 2025). The DIII-D boron-injection study adds a third variant in which impurity-driven turbulence opens long ELM-free periods, reduces frequent hard crashes, and creates a feedback loop between turbulence, particle transport, and pedestal stability (Banerjee et al., 1 Jun 2026).

Inter-ELM studies in MAST and DIII-D broaden the conceptual frame further. In MAST, pedestal height increases mainly through broadening of a quasi-constant-gradient barrier, while KBM-unstable and ballooning-unstable regions broaden inward over the ELM cycle (Dickinson et al., 2011). In DIII-D, MTMs can act as the inter-ELM pressure limit in the mid-pedestal when KBMs are second-stable (Hatch et al., 24 Mar 2026). These results do not define PREs directly, but they furnish the general pedestal-relaxation background in which PREs sit: repeated, thresholded transport events can clip pedestal gradients well before a large macroscopic peeling–ballooning crash.

Several unresolved issues remain explicit in the literature. The C-Mod/AUG study identifies open questions on local beta thresholds near the separatrix, the conductive-versus-convective partition at very low $\Delta T_e/T_e \approx 12\%$9, and predictive control strategies based on precursor detection (Silvagni et al., 2021). The GRILLIX study emphasizes the sensitivity of MTM onset to Landau-fluid closure and resistivity modeling, noting that Spitzer resistivity likely underestimates damping in high-$\Delta p_e/p_e \approx 18\%$0 pedestals (Pitzal et al., 26 Sep 2025). Taken together, these results suggest that PREs are best understood not as a single instability with a universal signature, but as a family of pedestal-localized relaxation cycles whose specific trigger may depend on regime: WCM-linked electromagnetic edge modes, MTMs, or density-gradient-driven pedestal modes.